Modelling of the processing of incompatible polymer blends
-
Upload
khangminh22 -
Category
Documents
-
view
2 -
download
0
Transcript of Modelling of the processing of incompatible polymer blends
Modelling of the processing of incompatible polymer blends
Citation for published version (APA):Elemans, P. H. M. (1989). Modelling of the processing of incompatible polymer blends. [Phd Thesis 1 (ResearchTU/e / Graduation TU/e), Chemical Engineering and Chemistry]. Technische Universiteit Eindhoven.https://doi.org/10.6100/IR316189
DOI:10.6100/IR316189
Document status and date:Published: 01/01/1989
Document Version:Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)
Please check the document version of this publication:
• A submitted manuscript is the version of the article upon submission and before peer-review. There can beimportant differences between the submitted version and the official published version of record. Peopleinterested in the research are advised to contact the author for the final version of the publication, or visit theDOI to the publisher's website.• The final author version and the galley proof are versions of the publication after peer review.• The final published version features the final layout of the paper including the volume, issue and pagenumbers.Link to publication
General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright ownersand it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.
• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal.
If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, pleasefollow below link for the End User Agreement:www.tue.nl/taverne
Take down policyIf you believe that this document breaches copyright please contact us at:[email protected] details and we will investigate your claim.
Download date: 06. Aug. 2022
MODELLING OF THE PROCESSING OF
INCOMPATIBLE POLYMER BLENDS
PROEFSCHRIFT
ter verkrijging van de graad van doctor aan de
Technische Universiteit Eindhoven, op gezag van
de Rector Magnificus, prof. ir. M. Tels, voor
een commissie aangewezen door het College van
Dekanen in het openbaar te verdedigen op
dinsdag 5 september 1989 te 14.00 uur
door
Petrus Henricus Maria Elemans
geboren te Oss
Dit proefschrift is goedgekeurd door
de promotoren
prof. dr. ir. H.E.H . Meijer
en
prof. dr. P.J. Lemstra
Dit onderzoek werd mogelijk gemaakt door financiële steun
van DSM .
CONTENTS
1. INTRODUCTION
1.1. Structured blends
1.2. Modelling of mixing equipment
1.3. Distributive mixing
1.4. Dispersive mixing
l.S. Survey of the thesis
1.6. References
1
1
6
8
10
13
14
2. APPROACHES TO THE MODELLING OF MIXING EQUIPMENT 17
2.1. Mixing equipment 17
2.2. Modelling of mixing equipment 22
2.2.1. Liquid-liquid mixing 23
2.2.2. Solid-liquid mixing 28
2.3. References 32
3. MODELLING OF COROTATING TWIN-SCREW EXTRUDERS 35
3.1. Introduetion 36
3.2. Screw geometry
3.3. Analysis of simplified geometry
3.3.1. Relative lengtbs
37
41
42
3.3.2. Specific energy 43
3.4. Improved analysis 47
3.4.1. Leakage flows 47
3 . 4.2. Power consumption over the flights 49
3.4.3. Mixing elements 51
3.4.4. Sequence of screw elements 55
3.4.5. Nonisothermal powerlaw calculations 57
3.5. Calculated results 61
3.5.1. Specific energy
3.5.2. Combination of parts band c
3.5.3. End temperature
61
63
64
3.6. Experimental verification of the newtonian,
isothermal analysis
3.6.1. Throughput versus screw speed
characteristic
3.6.2. Pressure gradients
3.6.3. Filled lengths
3.6.3.1. Experimental setup
3.6.3.2. Results
3.7. Residence time distribution
3.8. Discussion
3.9. References
4. MODELLING OF THE CO-KNEADER
4.1. Introduetion
4.2. Screw geometry and working principle
4.3. Summary of the Newtonian, isothermal
analysis
4.4. Mixing
4.5. Experimental
4.5.1. Throughput versus pressure
64
65
68
73
73
75
77
79
80
82
83
84
87
91
93
characteristic 95
4.5.2. Filled length 100
4.5.3. Pressure gradients 103
4.6. Nonisothermal, non-Newtonian analysis 106
4.7. Residence time distribution 108
4.8. Discussion 111
4.9. References 111
5. SCALING
5.1. Sealing laws
5.2. Geometrical sealing
5.3. Thermal sealing
5.3.1. Laminar flows
5.3. 2. Ideally mixed annular
L/D = a constant
5. 3. 3. Ideally mixed annular
H/D = a constant
flow,
flow,
113
113
114
1Hi
117
118
120
5.4. Sealing laws
5.5. Example: Glass-fibre reinforcement
5.6. Conclusion
5.7. References
6. TIME EFFECTS IN THE DISPERSIVE MIXING OF
INCOMPATIBLE LIQUIDS
6.1. Introduetion
6.2. Affine deformation of droplets in
simple shear flow
6.3. Breakup of threads
6.4. Breakup of droplets
6.5. Experimental
6.5.1. Experimental setup
6.5.2. Model fluids
6.5.3. Results
6.5.3.1. Affine deformation
6.5.3.2. Breakup of threads in
120
121
122
123
124
124
125
127
129
134
135
135
136
137
simple shear flow 138
6.5.3.3. Stable deformation and
relaxation of droplets 139
6.6. Conclusion 141
6.7. References 142
7. MORPHOLOGY OF THE MODEL SYSTEM
POLYSTYRENE/POLYETHYLENE 144
7.1. Introduetion 145
7.2. Phase inversion 147
7.2.1. Materials 147
7.2.2. Blend preparatien 148
7.2.3. Phase inversion diagram 148
7.2.4. Influence of compression moulding 153
7.3. Film thinning 153
7.4. Influence of block copolymers 156
7.5. References
8. STABILITY OF MORPHOLOGIES, OR THE EXPERIMENTAL
DETERMINATION OF INTERFACIAL TENSION 161
8.1. Measurement of interfacial tension via
breakup of threads 161
8.2. Interfacial tension between two
homopolymers
8.2.1. Materials
8.2.2. Experimental procedure
8.2.3. Results
8.3. Influence of block copolymers on
interfacial tension
8.3.1. Materials
8.3.2. Results
8.4. Contact angle measurements
8.4.1. Experimental procedure
8.4.2. Results
8.5. Breakup of molten polymerie layers
162
162
163
166
168
169
171
173
175
176
- an illustration 177
8.6. Conclusion 179
8.7. References 179
9. COUPLING OF DETAILED AND OVERALL MODELLING 181
9.1. Examples of calculations on
dispersive mixing 181
9.1.1. Combined affine deformation
and reorientation 182
9 . 1.2. Dispersion of Rn isolated droplet
in a screw extruder 183
9.1.3 . Tot al shear in a corotating
twin-screw extruder 186
9.2. Breakup of threads 188
9.3. Experimental 190
9.4. Results 190
9.5. Conclusions 193
9.6. Reierences 1 94
-1-
CHAPTER 1
INTRODUCTION
Various morphologies can be realized via processing of
incompatible polymer blends, for instanee droplets or fibers
in a matrix and stratified or cocontinuous structures. The
structures induced are usually intrinsically unstable.
Modelling of extrusion processes and continuous mixers
yields expressions not only for the shear rate and shear
stress, but also for the limited residence time and the
number of reorientations. These results can be combined with
detailed knowledge of respectively distributive and
dispersive mixing processes to predict the development of
specific morphologies, i.e. structured blends, as a tunetion
of time.
1.1. STRUCTURED BLENDS
Similar to rubbers and thermosets, thermoplastic polymers
are hardly used in their pure form. Additives are needed to
improve for example processability and lifetime (lubricants,
stabilizers), modulus and strength (mineral fillers like
glass beads, chalk, clay, mica or glass- fiber
reinforcement), appearance and colour (pigments),
conductivity (conductive fillers like steelwire, aluminium
Reprinted partly from: H.E.H. Meijer, P.J . Lemstra and P.H.M . Elemans,
Makromol. Chem., Macromol. Symp., ~' 113 (1988), by permission of
Hüthig & Wepf Verlag, Basel.
-2-
flakes or carbon) or flammability (flame retardants) .
Despite of the continuous development of new polymers, a
large number of properties can only be obtained when
different polymers are combined. Well known examples are the
impact modified, (rubber) toughened polymers, where polymers
with different glass transition temperatures are blended,
and the group of barrier polymers for packaging, where
specific polar and apolar polymers are combined in order to
increase the resistance against water- and gas- (oxygen,
carbondioxide) transport simultaneously.
Of course there are various routes to combine polymers in
order to achieve optimum properties. Polymer blends can be
made directly on a microscopie scale in the reactor. The
other extreme, on a macroscopie scale, is co-extrusion to
produce multi-layered structures via casting, blowing, blow
moulding and injection-moulding. Extrusion (melt) blending
is a route in between and in principle a rather flexible
one. The limited miscibility of polymers (1,2) complicates
this processing route however.
Unless specific interactions exist, phase separation usually
occurs (3,4). Of course, processing of miscible polymer
systems is of interest since tailor made properties can be
obtained by just changing the volume fractions. Although
over 300 pairs of miscible polymers are known (2) only a few
systems have been commercialized. Well known is the
successful blend PPE/PS. Other systems of commercial
interest are PC/PET and PC/PBT (5) .
In general, however, we have to deal with incompatible
polymers and depending on the processing conditions various
morphologies can be obtained. Figure 1.1 displays some
characteristic morphologies as obtained by extruding the
incompatible blend of Polystyrene/High Density Polyethylene
(PS/HDPE). Figures 1.1a, 1.1b, and 1.1c display extrudates
obtained from a corotating twin-screw extruder. Figure 1.1d
shows a PS/PE composition made via the Multiflux static
mixer (6,7).
-3-
All these morpbologies were realized by extruding the model
system Polystyrene/High Density Polyethylene (PS/HDPE), by
changing the volume fractions, viscosity ratio or processing
route (8).
These structures have been classified befare (9) and have
been found in practice, for example, with SBS block
copolymers with different percentages of polybutadiene
(10,11). As one can imagine, there are one ortwoorders of
magnitude difference in the length scale between the
incompatible system PS/PE and the blockcopolymers of SBS.
Figure 1.1a . Figure 1.1b.
Figure 1.1c. Figure 1.1d.
Figure 1.1a. Scanning electron micrograph of the microtomed surface of a
85/15 PS/HDPE blend (viscosity ratio 1).
Figure 1.1b . As Figure 1.1a, of the edge of a microtomed surface of a
75/25 PS/HDPE blend (viscosit y ratio 2) . HOPE (in black)
still forms the continuous phase .
Figure 1.1c. As Figure 1.1a, of a 55/45 PS/HDPE blend.
Figure 1.1d. As Figure 1.1a, of 50/50 PS/HDPE Multiflux blend.
-4-
A lot of attention has been paid to the morphology shown in
Figure 1.1a, and especially to routes to obtain a small
partiele size. Experimental results have been reported by
Borggreve (12) and Wu (13) for the system PA/EPDM, where it
is clearly demonstrated that the tough-brittle transition
temperature is not only depending on the amount of rubber
(Figure 1.2) but also, at the same volume fraction, on the
partiele size of the dispersed phase (Figure 1.3). To obtain
this small partiele size of the dispersed rubbery phase
maleic anhydride modified EPDM had to be used (14).
1100 100 ---, .>!:
L 80 80 g, L ..-> ., g'
'-1il 60 ~ 60 ..-> u t);;' ~ ~-€ -~ 40 E --, 40 "8 ·- "'
"8~ ~ ~ u 20
~ 20 .,
L L ~ 15 0 0 c c
20 -40 -20 0 20 40 60 80 -20 0 T('C) T('C)
Figure 1-2- Figure 1-3
Figure 1.2. Brittle-tough transition in Nylon/rubber blends_
Effect of rubber concentration_ Data from Ref. 12.
40 60
(• 0;\/ 2.6;• 6.4;0 10.5;0 13.0;(', 19-6;e 26.1 voL%)
Figure 1.3. As Figure 1.2. Effect of partiele size_ Volume fraction
of the rubber is 26%. Data from ReL 12_ (• PA-6;
(j 1.59~;· 1.2~;\/ 1.14~;0 0.94~;0 0.57~;· 0.48~)
80
Structures as displayed in Figure 1.1b, i.e. flbrils in a
matrix are aimed for as reinforcement and in the fabrication
of synthetic paper or artificial leather (15,16,17).
-5-
Cocontinuous structures, Figure 1.1c, are usually obtained
if a 50/50 blend is extruded or, if the viscosity ratios of
matrix and dispersion differ from one, at other mixing
ratios as well (18). The morphology of the cocontinuous
structures is tosome e xtent similar to IPN's basedon
direct chemistry, although the scale is two orders of
magnitude higher. (19).
Layered structures, see Figure 1.ld, can be made rather
easily with specially designed static mixers like the Ross
mixer or even better with the Multiflux mi xer (6,7). The use
of layered structures is important for instanee in the area
of pac kaging. Figure 1.4 shows the barrier properties,
expressed in the effective diffusion coefficient, of a cast
or blown film as a function of the composition, with the
morphology as parameter.
-en ~ E 0 -
<D > -0 <D --<D
c
10- 6 ~-----------------------------------------------,
10-7
10-8
10-9
parallel
·~ EVOH spheres
lamel la;---.:=·~ -- ~ --- · -...........:::--:--.._ EVOH cylinders
~---- ~"':"--- ·~ ------- ·~
PP cylinders
-----~
---···---... ..__ ',<~', , --... ....._ ',\ ···-- \
PP spheres
10-104---------~--------r-------~r--------,--------~ 0 0.2 0.4 0.6 0.8
vol. fractlon EVOH
Figure 1.4 . Effective diffusion coefficient as a function o f composition
for a polypropylene/ethylene-vinyl alcohol copolymer
(PP/EVOH) blend, with different morphologies. After Ref. 20.
The two limiting curves correspond with the two extremes:
-6-
layers of the barrier material oriented either parallel or
perpendicular to the plane of the film. Note that a
logarithmic scale is used on the vertical axis, indicating
that the upper curve obeys the additive rule of mixtures.
As can be seen from the examples mentioned, these so-called
structured blends (21) all exhibit a distinct morphology.
It is important to understand how different morphologies in
a blend of two incompatible polymers can be obtained, and
guaranteed during subsequent processing. Detailed knowledge
of the processing equipment is necessary as well as an
understanding of the mixing process itself. Therefore, the
rest of this chapter will consist of a brief review of these
topics.
1.2. MODELLING OF MIXING EQUIPMENT
From simplified flow analysis inside extruders, important
overall parameters for mixing, such as residence time t,
shear stress "' shear rate y, total shear y and the nuffiber of
reorientations nr can be deduced, at least locally.
Especially if only melt-fed equipment is considered, all
geometries such as extruder channels and clearances, and
also converging flows with one or two moving boundaries,
e.g. the two roll mill, have been analysed (22,23,24,25). As
a consequence the local conditions present for mixing are
known even in typical compounding equipment like
batchmixers, counterrotating twin-screw extruders, Farrel
Continuous Mixers, corotating twin-screw extruders and
reciprocating pin extruders like the Buss Co-kneader.
Of course more elaborate calculations can be performed,
yielding the complete two- or three dimensional flow field
in the complex geometries of the mixing sections in a
corotating twin-screw extruder (26) and in the Co-kneader
(27).
-7-
However, it has to be postulated a priori that these mixing
sections are completely filled with melt and all
calculations are still isothermal. Here a more overall
investigation of these continuous mixers is developed, part
of which has been published already (28,29). By simplifying
the geometry again, the lengths which are completely filled
with melt are determined depending on the screw geometry
used and on processing conditions like screw speed and
(independently) metered throughput. Moreover, via an
averaged local heat balance, the temperature rise during the
compounding process and the specific energy, depending on
processing conditions, can be calculated .
If combined with criteria originating from a more complete
model of the dispersion process itself, this would be
sufficient to predict the morphology of an, as-processed,
blend. However, ignoring complicating factors like
coalescence of droplets, even the time effects af the
dispersion process are not well understood. For the more
simple dispersive mixing of carbon black in rubber in an
extremely simplified geometry of a completely filled
batchmixer (no time effects involved in breaking, no
influence of partiele size), see Figure 5, an interesting
analysis exist& (30,31), which is later extended
V
Q
HIGH SHEAR ZONE
Figure 1.5. High shear section in series with an infinitely well mixed
section. The fluid is continuously pumped from one section
into the other. From Ref. 30.
-8-
to two roll mills (32) . Provided that dispersive mixing of
blends is better investigated, these examples may be
extended to the rnadelling of the blending process in
continuous mixers, since the mathematical tools, necessary
for this kind of calculations, already exist from continuurn
mechanics, see for example (33) and (34).
1.3. DISTRIBUTIVE MIXING
For distributive mixing total shear y and number of
reorientations during the shear history are the only
determining factors. This has been clearly illustrated by Ng
and Erwin (35) who performed a c lassical experiment by
placing coloured slices of a polymer between two concentric
cylinders and rotate one of them. Either the number of
layers formed (measured radially) or the total interface,
bath being a measure for distributive mixing, is directly
proportional to the total shear.
A (A interfacial area) (1.1)
Since y = yt, shear rate and -time are interchangeable. If
the already formed layers are reoriented relative to the
direction of flow, mixing becomes much more effective. This
is illustrated by stopping the rotation, freezing, cutting
slices which are turned over 90°, an ideal reorientation,
heating up again and further shearing. If this procedure is
repeated n-1 times, Eq. 1.1 reacts (see Figure 1.6),
A = Ao ( 1 In y) n (1. 2)
A much more effective way of distributive mixing, because
reorientation does not cast energy (shear rate or -time)
Static mixers are the prime exponents of mixing by
reorientation rather than by total s hear y, but also in
corotating twin-screw extruders material is continuously
reoriented relative to the shearing motion of
-9-
Figure 1.6. Shearing and reorientation during shear of black and white
segments. From Ref. 36.
Figure 1.7. Mixing and
reorientation
in corotating
twin-screw
extruders.
the surfaces, when one screw scrapes the fluid from the
other one (Figure 1.7).
The number of reorientations can be estimated and forms,
tagether with the expressions for shear rate, time and total
shear, the basis for sealing rules for distributive mixing
in corotating twin-screw extruders (29) . Although the pins
of a Buss Co-kneader reorient the flow as well, see Ref. 27
and 37, the distributive mixing is better understood by
consictering the local weaving action of the pins (the screw,
as usual, is thought to be stationary and the barrel and
pins rotating and reciprocating, yielding sinuscictal
trajectories through the screw channel) .
Combined with an overall model of the continuous mixer (29)
-10-
this analysis directly provides insight in the distributive
mixing of additives, pigments, fillers and already dispersed
masterbatches in a matrix.
1.4. DISPERSIVE MIXING
If two incompatible polymers have to be blended, the
interfacial tension, which is directly related to the mutual
miscibility, becomes during the mixing process of the same
order of magnitude as the shear stress applied and will
dominate the resulting morphology. An order of magnitude for
the interfacial tension o is typically 10-2 [N/m], while the
shear stress ~ for polymer roelts is of the order of 10 4
(N/m2]. Consequently, if local radii are in the order of
10-6 [m], yielding o/R = 10 4 (N/m2 ], both stresses are
equal. Starting at the first stages of mixing, the droplets
undergo affine deformation according to Eq. 1.2 describing
the distributive mixing. The resulting long slender bocties
become instable due to the interfacial tension-driven
Rayleigh disturbances (38,39), see Figure 1.8.
Figure 1.8. Sinusoidal distortions on a PA-6 thread (diameter 55 ~)
embedded in a PS matrix at 230 ° C.
-11-
The draplets formed are again subjected to shear stresses
counterbalanced by the interfacial tension resisting the
deformation process. This process has extensively been
studied in the literature (38-45). Especially the work of
Grace (43) is worthwhile reading because of the large number
of experiments performed in shear and elongational flow with
liquids with a large range of viscosity ratios. The
stability of draplets turns out to be strongly dependent on
this viscosity ratio:
(1.3)
and of the ratio of the applied shear stress ~ = DeY and the
pressure due to the interfacial tension a/R, usually
referred to as the capillary number Ca:
Ca ~R/a (1. 4)
Quite a large difference exists between the (efficiency of)
shear- and elongational flows, especially if p ~ 1. This
difference can only partly be explained by the difference in
shear- and elongational viscosity (43). It is mainly due to
the difference in type of flow: weak vs. strong respectively
(46). See Figure 1.9.
1(0)
100 T.R/a
10
1
0.1 10-7 10-6 10-5 10-4 10-3 10-2 10-1 10° 101 102 103
Viscosity Ratio, p
Figure 1.9. Comparison of effect of viscosity ratio on critical shear
in rotational and irrotational shear fields. From Ref. 43.
-12-
Although all of these studies are performed with individual
draplets of model liquids at room temperature, they
emphasize the non-equilibrium state of the morphologies
given in Figures 1.1a-1.1d. The fibrous structures found in
PE/PS mixed on a corotating twin-screw extruder are half-way
the dispersive mixing process and are typically formed in
the strong elongational flow field between screw tips and
die and in the filament between the die and water quench.
This is clearly illustrated in Figures 1.10 and 1.11 showing
two different spots in the same filament. In one case
(Figure 1.10) some fibrils (the smaller ones, of course)
start to break up exhibiting Rayleigh disturbances while in
the second case (Figure 1.11) more fibers have broken up.
These effects, including coalescence, which is also found in
Figure 1.9, always occur during the mixing process. The
morphology will continuously change and adapt itself to
local situations.
Figure 1.10. Figure 1.11.
Figure 1.10. Scanning Electron micrographof a fracture surface
parallel to the direction of extrusion of an extrudate of a
45/55 PS/HDPE blend (viscosity ratio 1) . Fibreus PS is shown
in different stages of breakup and coalescence.
Figure 1.11. As Figure 1.10, but with more fibers breken up.
-13-
1.5. SURVEY OF THE THESIS
The main field of interest, as outlined in the previous
sections, can be further explored now. The final objective is
to combine the knowledge of specific areas into a more
complete model that can be of use in the processing of
polymer blends. The content of each chapter wi11 be briefly
indicated below.
Chapter 2 gives a review of the different approaches to the
rnadelling of compounding equipment.
Chapter 3 deals with the modelling of melt-fed corotating
twin-screw extruders. It concentrates on the calculation of
locally fi1led lengths, power, specific energy and
temperature rise. With regard to mixing, the shear rate,
shear stress, residence time and the number of reorientations
can be determined. Experiments supporting the analysis are
presented.
In chapter 4, the computational model for the corotating
twin-screw extruder is applied to the Cokneader. The course
of this chapter is analogous to that of chapter 3.
In chapter 5, methods of sealing will be introduced.
Chapter 6 discusses the processes involved in the mixing , of
two (incompatible) liquids, with emphasis on the time effects
of deformation and breakup of dispersed particles in
well-defined fields of flow.
Chapter 7 gives, as an example of an incompatible blend, the
morphology of the model system Polystyrene-Polyethylene, as
processed on a corotating twin-screw extruder. Parameters
varied are blend composition and viscosity ratio.
Chapter 8 describes model experiments concerning the
stabilization of specific, non-equilibrium morphologies.
Chapter 9 attempts to couple the knowledge of the previous
chapters. The origin and the development of blend
morphologies, made on a corotating twin-screw extruder, is
followed.
-14-
1.6. REFERENCES
1. 0. Olabisi, L.M. Robeson, M.T. Shaw, Polymer/Polymer
Miscibility, Academie Press, New York (1979)
2. L.M. Robeson, Polymer Compatibility and Incompatibility,
NMI Press (1982)
3. J.M. Barlow, D.R. Paul "Polyblends '87 ", Boucherville
Canada, Polym. Eng. Sci. Zl_, 1482, (1987)
4. W.H. Stockmayer, R. Koningsveld, E. Nies in 'EQuilibrium
Thermodynamics of Polymer Systems' Vol . .l: Polymer Phase
Diagrams, Oxford Univ. (1988)
5. NN, Plastics Eng. 2, 24, (1986)
6. R. Sluijters" De Ingenieur TI, 15 ,33 (1965)
7. D. S'cbil'O, I<. Ostertag, Verfahrenstechnik .6., 2, 45 (1972)
8. P.H.M. Elemans, J.G.M. van Gisbergen, H.E.H. Meijer in
'Integration of Polymer Science and Technology', Elsevier
(1988)
9. M. Matsuo, S • .Sagaye in 'Colloidal and Morphological
Behaviour of Block anà Graft Polymers', G.E. Molau, Plenum
(1971)
10. M. Matsuo, Japan Plastics, .2., 6, (1968)
11. J.A. Manson, L.H. Sperling, Polymer Blends aod Composites,
Plenum (1976)
12. R.J.M. Borggreve, R.J. Gaymans, J. Schuijer, J.F. Ingen
Housz, Polymer ~' 1489-1496 (1987)
13. S. Wu, Polym. Eng. Sci. 27, 335-343 (1987)
14.Du Pont, US Patent 580513
15. G.V. Vinogradov, N.P. Krasnikova, V.E. Dreval, E.V.
Kotova, E.P. Plotnikova, Int. J. Polym. Mat. ~, 187 (1982)
16.M.P.Zabugina, E.P. Plotnikova, G.V. Vinogradov, V.E.
Dreval, Int. J. Polym. Mat. lQ, 1 (1983)
17.M.V. Tsebrenko, Int. J. Polym. Mat. lQ, 83 (1983)
18. G.N. Avgeropoulos, F.C. Weissert, G.G.A. Böhm, P.H.
Biddison, ACS Rubber Division Meeting, Paper 3, New
Orleans (1975)
19.K.C. Frisch, D. Klempner, H.L. Frisch, Polym. Eng. Sci.
22, 17 (1982)
-15-
20. J. Sax, J.M. Ottino, Polym. Eng. Sci. ~' 165 (1983)
21.H.E.H. Meijer, P.J. Lemstra, P.H.M. Elemans, Makromol.
Chem., Macromol. Symp. 1Q, 113 (1988)
22. J.M. McKelvey, Polymer Processing, J. Wiley (1962)
23. J.M. Funt, Mixing of Rubbers, RAPRA (1977)
24. z. Tadmor, I. Klein, Engineering Principles of
Plasticating Extrusion, Reinhold (1971)
25. Z. Tadmor, C.G. Gogos, Principles of Polymer Processing,
J. Wiley (1979)
26.W. Szydlowski, K. Brzoskowski, J.L. White, Int. Polym.
Proc. l_, 207 (1987)
27.M.L. Booy, Y.K. Kafka, Submitted to Polym. Eng. Sci.
28.H.E.H. Meijer, P.H.M. Elemans, H.H.M. Lardinoye, G. Kremer
in: 'Wärmeübertragung bei der Kunststoffverarbeitung', VDI
Düsseldorf (1986)
29.H.E.H. Meijer, P.H.M. Elemans, Polym. Eng. Sci. ~'
275-290 (1988)
30. I. Manas-Zloczower, A. Nir, Z. Tadmor, Rubber Chem. Techn .
.5..5_, 1250 (1982)
31. I. Manas-Zloczower, A. Nir, Z. Tadmor, Rubber Chem. Techn.
;u, 583 (1984)
32. I. Manas-Zloczower, A. Nir, Z. Tadmor, Dispersive mixing
in two roll mills (submitted)
33. J.M. Ottino; Polym. Eng. Sci. ~' 7 (1983)
34.D.V. Khakhar, J.M. Ottino, Int. J. Multiphase Flow~' 7
(1987)
35. K.Y. Ng, L. Erwin, Polym. Eng. Sci. 21, 4, (1981)
36. G.M. Gale, Rapra Merobers Report No. 46 (1980)
37. L.Erwin, F. Mokhtarian, Polym. Eng. Sci. ~' 49-60 (1983)
38. Lord Rayleigh, Proc. Roy. Soc. (London) 2..2_, 71 (1879)
39. S. Tomotika, Proc. Roy. Soc. Al.2Q, 322 (1935)
40. G.I. Taylor, Proc.Roy. Soc. (London) A146, 501 (1934)
41. R.G. Cox, J. Fluid Mech. n, 601 (1969)
42. E.J. Hinch, A. Acrivos, J. Fluid Mech. ~' 305 (1980)
43. H.P. Grace, Chem. Eng. Comm. ll, 225 (1983)
44. J.J. Elmendorp, Ph.D Thesis Delft University of Technology
(1986)
-16-
45. C.D. Han, Multiphase Flow in Polymer Processing, Academie
Press ( 1981)
46.W.L. Olbricht, J.M. Rallison, L.G. Leal, J. Non-Newt.F.M .
.l.Q., 291 (1982)
-17-
CHAPTER 2
APPROACHES TO THE MODELLING OF MIXING EQUIPMENT
Mixing equipment, used in the compounding and blending of
polymers, is briefly reviewed, as well as the different
approaches used in the rnadelling of mixers.
Early rnadelling started with those sections that were
accessible in terms of flow field (two-dimensional; one
direction (lubrication approximation)), fluid (Newtonian),
and geometry (the unrolled extruder channel (1,2), the nip
region of a roll mill (3)). Throughout the years, the
analyses have become more sophisticated. Non-Newtonian,
non-isethermal effects (4), a three-dimensional (two
directions) description of the flow field (5), the use of
finite-element techniques (6,7) and even chaotic motions (8)
have been incorporated. However, they only give solutions to
local problems; overall answers are still hard to find.
Also combinations of (simplified) flow analysis with local
dispersion and breakup processes of solicts (9) or liquids
(10) are scarce.
2.1. MIXING EQUIPMENT
Mixing operations of highly viscous polymers take place in
various types of machines. These can be divided into batch
mixers (a few only) and (many types of) continuous mixers.
Batch mixers
Batch mixers are usually found in the processing and
compounding of rubbers. Most common types are the internal
mixer and the roll mill, because of their dispersive and
distributive mixing qualities (11). During each revolution,
-18-
the material is forced to pass through a nip, where the
deformation and breakup of the dispersed (solid or liquid)
particles takes place. Between two passes, the material is
reoriented, either via manual cutting (roll mill), or via
the lateral motion induced by the particularly shaped rotor
wings (internal mixer). See Figure 2.1.
c
-E@3 ~ B --== D~ ~-. ~· -
Figure 2.1. Exarnples of rotor designs (arrows indicate pumping action of
rotor wings). A: Banbury two-wing; B: Banbury four-wing;
C: Shaw Intermix three-wing; D: Werner & Pf1eiderer four-wing.
From (12).
Single screw extruders
Single screw extruders consist of a conveying screw fitting
closely in a cylindrical barrel. One wall (the barrel) is
stationary, while the other wall (the screw) is moving, thus
dragging the material in the direction of the die at the
outlet. A pressure flow is generated in the reverse
direction, due to the resistance of the die. Mixing is
achieved by the motion caused by the combination of drag and
pressure flow inthescrew channel (13,14). The mixing
quality of singlescrew extruders is generally poor (15),
but can be improved when the screw is equipped with extra
mixing elements that provide for periadie reorientation of
-19-
the material, such as barrier sections (16), pineapple
heads, blisters, eccentric disks or even cavity transfer
mixers ( 17) .
Co-kneader
More flexible in screw design is the Co-kneader (18). This
is a single screw extruder with a simultanuously rotating
and oscillating screw having interrupted flights. Pins from
the barrel are inserted into the screw channel. The combined
weaving motion of pins and flights gives rise to good
dispersive as well as distributive mixing.
Twin-screw extruders
J~~~ J~ J~
J~ G--~ ~0--~
Figure 2.2a Figure 2.2b
Figure 2.2 . Fully, partly and nonintermeshing types of corotating (Fig.
2.2a) and counterrotating (Fig. 2 . 2b) twin-screw extruders.
From (21).
As Figure 2.2 shows, twin-screw extruders may be divided
into counter- and corotating types and into closely, partly
-20-
and nonintermeshing systems (19-21). Apart from the
direction of rotatien of the screws, they can be subdivided
according to their transport mechanism: positive
displacement or drag flow. This division can be made by
investigating whether the channel is closed in the axial
direction (by the flight of the opposite screw) or open
(22-24) .
Counterrotating twin-screw extruders
Counterrotating twin-screw extruders can be constructed with
small clearances, and the closely intermeshing types are
therefore often associated with positive displacement. In
practice, this does not prove to be very realistic because
apart from the typical tetrahedral gap between the sides of
the adherent screw flights and the necessary clearance
between barrel and screws, the so-called calender gap
between screw root and tip of the flight of the opposite
screw is often rather large. This gap drags material (with
two rnaving walls!) backward into the previous C-shaped
chamber. The counterrotating extruder is treated in detail
in Refs. 25 and 26 with the final result that the pumping
characteristic, throughput versus pressure buildup, is
rather easily obtained as the number of C-shaped chambers
becoming free per unit of time multiplied by the volume of
one chamber minus the sum of all leakage flows. Even with
small clearances the backflow caused by leakage can be in
the order of half of the positive displacement, depending on
the pressure at the die. Counterrotating twin-screw
extruders are almast exclusively used in poly(vinyl
chloride) processing, because of their mild treatment of the
melt. As far as mixing is concerned, they can be treated as
a continuous two-roll mill process.
Farrel Çontinuous Mixer
The Farrel Continuous Mixer (FCM) is a combination of an
-21-
internal mixer with a nonintermeshing counterrotating
twin-screw extruder (Figure 2.3). The mean residence time of
FCMs is in the order of 20 seconds, which is short.
Therefore, they are mainly used for the fast melting and
pelletizing of premixed powdered polymers (PP, ABS, or HDPE)
Figure 2.3. Typical screw geornetry of a Farrel Continuous Mixer (FCM)
Corotating twin-screw extruders
Corotating twin-screw extruders are, much more than
counterrotating twin-screw extruders or FCMs, preferred in
the processing of polymer blends. They operate almost
completely under atmospheric pressure, since their main
pumping mechanism is drag flow. Via openings or vent ports
in the barrel, material can be added to the melt (fillers,
stabilizers, pigments) or volatiles can be removed. To do
so, these extruders normally have to be underfed, and
pressure is only generated in those parts of the screw that
are completely filled, e.g. countertransporting sections.
Mixing is achieved very effectively in the intermeshing
region between the screws. The material is passed from one
screw to the other, and is thus reoriented.
Corotating Disk Processor
Though still rarely encountered as a continuous mixer, the
Corotating Disk Processor can be mentioned (27). Its basic
-22-
geometry (processing chamber) consists of two parallel disks
mounted on a shaft, and fitting in a cylindrical barrel. The
pumping action is very efficient, since it is due to two
jointly rnaving walls. Apart from channel blocks, which
separate the processing chambers from each other, mixing
pins can be inserted through the barrel wall. They provide
for reorientation by splitting the streamlines as well as
for dispersive mixing because the material is forced through
narrow clearances, see Figure 2.4. Without these pins, the
mixing action of the Corotating Disk Processor is as poor as
that of other effective pumps like a gear pump (28).
MIXING
"~~== MIXING BLOCK CVTLET DISK 2
Figure 2.4. Processing chamber with mixing pins of a Corotating Disk
Processor. Frorn (29).
2.2 MODELLING OF MIXING EQUIPMENT
In rnadelling of mixing equipment, one aften encounters
complex geometries and fluids exhibiting strongly
non-Newtonian behaviour. Different approaches have been
developed to get a better understanding of the mixing
process in batch and continuous mixers.
Usually the material is considered to be completely molten,
-23-
having a well-defined (Newtonian) viscosity. Approaches or
roodels that focus on single- or twin-screw extruders can
aften be applied to batch mixers or vice versa.
It is interesting to review the most important approaches,
thereby camparing their practical value in the field of
distributive and dispersive mixing.
2.2.1. LIQUID-LIQUID MIXING
Liquid-liquid mixing is the approach that is found in many
textbooks on polymer processing (30-33). The geometry of
most mixing equipment (extruders, internal mixers) is
locally represented as two parallel plates. The relevant
parameters for the mixing process are firstly combined into
the shear rate y and the average residence time t.
circumferential speed 11' D N y
loc al channel depth H (2 .1)
volume 11' D L H t (pressure dependent) throughput Q(P) (2. 2)
The total strain y
y = y t, (2 .3)
and the temperature T define the basic parameters for mixing.
From this simple starting point, complications can be
incorporated, such as the number of reorientations nr and
the effect of initial orientation. Due to the complex
geometries of most mixers, it is inevitable to introduce the
use of average shear rates. In addition, residence time
distribution and the weighted average total strain are
necessary to characterize the mixer performance.
The mixing process is thought to be in its initial stage
(see Se ctiens 1.3 and 1.4), with large strains imposed by
the matrix on the suspended particles of the minor phase.
The analyses are usually based on the isothermal flat plate
-24-
model of the extruder screw (34). Newtonian flow is assumed.
From the velocity field in down-channel direction as wel! as
in cross-channel direction, the average residence time, the
respective average shear rates y2 and Yx and the total shear
strain y, can be calculated.
The problem of averaging is quite complicated for the
different types of mixers. Each fluid element, starting at a
given initia! position in an extruder channel, wil!
experience a different strain history. This was quantified
by Lictor and Tactmor (35), who introduced the Strain
Distribution Function (SDF) f(y)dy. It is defined as the
fraction of the fluid in the mixer which has experienced a
shear strain from y to y+dy. The mean strain of the fluid at
the exit of the mixer is:
00
y I y f(y)cty (2 .4)
Yo
with YQ the minimum strain.
Pinto and Tactmor (36) propose the Weighted Average Total
Strain (WATS) to calculate the amount of strain experienced
by a fluid element in a single screw extruder. It is defined
by:
00
WATS I y (tl f(tl ctt (2 .5)
0
with y(t) the strain undergone by a fluid element at a
time t
{(t) the Residence Time Distribution (37)
Unfortunately, the WATS does not constitute a quantity that
can be determined experimentally. Neither can it account for
the initia! orientation or perioctic reorientations of the
material,as brought about by mixing sections of an extruder.
Its main importance is, therefore, on the theoretica! level.
-25-
Extensions of this approach can be found in the work of Bigg
and Middleman (38) . They study the evolution of the
interfacial area, which is a measure of the degree of
mixing, between two immiscible fluids by following
tracer particles in two-dimensional flow fields that are
present in extruder channels. Mixing sections in single
screw extruders greatly enhance the formation of interfacial
area (39,40). Ottino (15) presents a complete
three-dimensional description ('framework') of the flow
field in e.g. single screw extruders, with essentially the
same conclusion as Ref. 40: The initia! orientat i on and,
most of all, the number o f reorientations are the
determining factors in liquid-liquid mixing.
Although a complete description of the velocity field is
useful to understand the working principle of a mixer, it is
not sufficient. Even seemingly simple velocity fields can
give rise to a quite complex flow or 'motion' (in terros of
continuurn mechanics) of fluid particles (41,4 2 ). As stated
before (see Section 1.3), distributive mixing is usually
analysed in terros of the deformation (to a large extent) of
'blobs' or granules, schematically represented by a material
line element. Under certain conditions, these material lines
undergo e xponential stretching. In two dimensions, this is
possible for instanee in a hyperbalie flow field, and in the
flow inside a cavity that has periodically moving walls
(43). In the latter case, the mixing has become 'chaotic'.
A chaotic flow is characterized by either of the two
following (equivalent) criteria: (i) The flow has a positive
Liapunov exponent or (ii) the flow forms so-called 'Smale
horseshoes' (44).
The Liapunov exponent can be explained using the length
stretch of an infi nitesimal materia l filame nt dx undergoing
the flow
-26-
(2 .5)
with x the position of a given material point at time t
xo the initial position
~ the deformation gradient
The length stretch À is defined by (15):
À lim I dx I I I dxo I .f(~ mo mo) ldx0 l+o
with ldxol
~
the initial length
the Cauchy-Green deformation tensor: ~
the initial orientation dxo/ldxol
(2. 6)
In most flows, it is normally observed that À oc t for long
times (e.g. in a shear flow). For that reason, these flows
are called 'weak flows' (43). Strong flows, on the other
hand, show À - eet, with o the Liapunov exponent:
lim ( 1/t) ln À (2. 7)
t+o:>
The effect of o > 0 is that fluid particles, no matter how
close tagether initially, become separated exponentially
(43). This is ref1ected in the behaviour of the
two-dimensional 'blinking vortex flow' (45).
The Smale harseshoe function (44) is shown in Figure 2.5. It
involves the stretching and folding of a square with itself.
This is the only possible mechanism of increasing length for
a two-dimensional surface in a bounded flow.
The presence of harseshoes is shown by Chien et al. (46).
They study the deformation of a blob of tracer liquid in a
cavity that has periodically rnaving upper and lower walls.
An optimal value of the dimensionless frequency f exists
which produces maximum (i.e. exponential) stretching of the
-27-
initia! blob in a given time. With regard to mixing, it is
desirable that the horseshoe functions are present over a
large portion of the mixing region. This is still difficult
to predict.
I I I
I Figure 2.5. Representation of the Smale horseshoe function. After (46).
It is clear that the concept of chaotic mixing is far from
complete and limited in practical application, but still it
may give some more insight into the process of the rapidly
decreasing length scale of two initially segregated fluids.
Also, the existence of 'demixed' regions can well be
demonstrated, see e.g. Ref 47.
-28-
2.2.2. SOLID-LIQUID MIXING
A much more simple approach can be found in the work of
Manas et al. (9,48), who develope an analysis of the
dispersive mixing process in Banbury-type of mixers, which
is later extended to rol! mills (49). Although originally
proposed for the dispersive mixing of carbon black in rubber
(no time effects involved), the analysis can in principle be
applied to e.g. the blending of incompatible polymers.
The approach successfully integrates a number öf aspects
that are relevant for the modelling of the mixing process:
(i) an extreme simplification of the mixer geometry, (ii) the
influence of (initia!) orientation and reorientation of the
dispersed particles, and (iii) a criterium for breakup of
carbon black agglomerates.
During one pass through the nip region of the batch mixer
(the high shear zone in Figure 1.5), the carbon black
agglomerates are subjected to hydrodynamic forces exerted on
them by the matrix fluid (which is considered to exhibit
Newtonian or power law behaviour) . The effectiveness of
these forces depends on the instantanuous orientation of the
agglomerates, and can be expressed as
with x s
shape factor
characteristic cross-sectional area of an
agglomerate
• shear stress
e,~ instantanuous orientation angles
(2. 8)
The agglomerates are thought to consist of clusters of
aggregates, which are held together by cohesive forces (9).
In their most elementary form, these forces can be expressed
as
-29-
1-c
c ~
) d s
with c the relative void space between the aggregates
Co a constant
d diameter of the aggregate
Agglomerate breakup is assumed to occur when the
hydrodynamic forces exceed the cohesive forces:
~> F = 1
c
With the Eqs. 2.8 and 2.9, this yields
~ z sin2e sin (f/ cos (f/ Fe
8 1-c d with z 9 x 1: ( co c
(2 .10)
(2 .11)
(2.12)
Following this criterium, agglomerate rupture is independent
of agglomerate size. With the criterium, and with some
statistics concerning the distribution of passes of a given
agglomerate through the nip, this part of the problem is
essentially solved. After leaving the nip, the agglomerates
are reoriented in the mixer chamber before entering the
high-shear zone again with a random orientation distribution.
By defining the volume fraction of undispersed agglomerates
(i.e. those above a certain critica! diameter), calculated
predictions can be compared with experimental data, see
Figure 2.6a. The agreement is fairly good, given the
relative simplicity of the model and the large number of
assumptions on which the analysis is based.
Figure 2.6b shows the calculated fraction undispersed carbon
black as a function of rotor speed.
Figure 2.6c shows an interesting Qifference in machine
performance. When styrene-butadiene rubber (SBR, with
-30-
~O = 105 Pa.s) is compounded, the fraction undispersed (I
carbon black decreases monotonically with increasing rotor
tip clearance, i.e. throughput. This means: the available
hydrodynamic farces are always much larger than the cohesive
farces (compare Eq. 2.10). Upon compounding low-density
polyethylene (LDPE, with ~O = 103 Pa.s), however, the
fraction undispersed carbon black exhibits a minimum, and
increases again with larger tip clearance. Since now the
hydrodynamic farces are of the same order of magnitude as
the cohesive farces, only the more favourably oriented
agglomerates will break up.
Ul
~ ~ Q)
E 0 .8 0 0, OI
"' '0 0.6 Q) Ul
Q; a. "' '6 0.4 c: :J
0 c: 0.2 0
·;::; u "' L.t 0
0 100 200 300 400 500 600
time (s)
Figure 2.6a. Fraction (f) of undispersed carbon black as a function of
time. Parameter: diameter D (in mm) of the internal mixer.
Symbols indicate experimental data. From (9) .
In this thesis, this line will be extended to the analysis
of the dispersive mixing of two incompatible fluids. The two
most important continuous mixers are analysed in a rather
simple but overall way (yielding data on y, t, y, nr,
depending on processing conditions) . Next we will study the
time effects of the local dispersive mixing and finally an
attempt will be made to combine these two approaches in one
model.
-31-
"' ! "' (ij E 0.8 0 0> 0)
"' " 0.6 Q)
"' (ij 90 sec. a. "' 0.4 '6 c: :::l
0 c: 0.2 0
'<3 "' Lt 0
0 40 80 120 160 200
rotor speed (rpm)
Figure 2.6b. Simulated fraction of undispersed agglomerates in a Banbury
B mixer (D ~ 97 mm) as a function of rotor speed. Parameter:
mixing time. From (9).
"' ! "' (ij E 0.8 0 0> 0)
"' " 0.6 Q)
"' Îi) a. "' 0.4 '6 c: :::l -0 c: 0.2 0
·.::; SBR 0
"' Lt 0 3 4 5
gap height (mm)
Figure 2.6c. As Fig. 2.6b, as a function of rotor tip clearance,
for a mixing time of 120 seconds. Power law model
parameters: low density polyethylene (LDPE) no = 103 Pa.s,
n 0.46; styrene-butadiene rubber (SBR) no = 105 Pa.s;
n 0.3. From (9).
-32-
2.3. REFERENCES
1. Anonymous, Engineering (London), ~' 606 (1922)
2. H.S. Rowell and D. Finlayson, Engineering (London), ZQ,
249 (1928)
3. G. Ardichvili, Kautschuk Gummi Kunststoffe, ~, 23 (1938)
4. R.E. Colwell and K.R. Nickolls, Ind. Eng. Chem., ~' 841
(1959)
5. R.M. Griffith, Ind. Eng. Chem. Fund., 1, 180 (1962)
6. W. Szydlowski and J.L. White, Intern. Polym. Proc. II,
3/4, 142 (1987)
7. R. Brzoskowski, J.L. White, W. Szydlowski, N. Nakajima
and K. Min, Intern. Polym. Proc. ~, 3, 134 (1988)
8. J.M. Ottino, C.W. Leong, H. Rising and P.D. Swanson,
Nature, ~, 419 (1988)
9. I. Manas-Zloczower, A. Nir and Z. Tadmor, Rubber Chem.
Techn., ~, 1250 (1982)
10.D.V. Khakhar and J,M, Ottino, Int. J. Multiphase Flow,
u, 71 (1987)
11. J.M. Funt, Mixing of Rubbers, Rapra (1977)
12.H.Palmgren, Rubber Chem. Techn., ~' 462 (1975)
13. J.F. Carley, R.S. Mallouk and J.M. McKelvey, Ind. Eng.
Chem., ~' 974 (1953)
14. R. Chella and J.M. Ottino, Ind. Eng. Chem. Fund., ZA, 170
(1985)
15. J.M. Ottino and R. Chella, Polym. Eng. Sci., ~, 7 (1983)
16. B.H. Maddock, SPE J., ~' 23 (1967)
17. G.M. Gale, 41st SPE ANTEC, 109 (1983)
18.P. Schnottale, Ka~tschuk Gummi Kunststofte ~, 2/85, 116
(1985)
19. C.J. Rauwendaal, Polym. Eng. Sci., 21, 1092 (1981)
20.H. Herrmann and U. Burckhardt, Kunststoffe, Qa, 753 (1978)
21.H. Werner, Dissertation, Munich University of Technology
(1976)
22.K. Eise, S. Jakopin, H. Hermann, U. Burckhardt and H.
Werner, Adv. Plast Techn., 18 (April 1978)
-33-
23.U. Burckhardt, H. Herrmann, and S. Jakopin, Plast.
Compounding, Nov./Dec. 73 (1978)
24.M.L. Booy, Polym. Eng. Sci., ~, 606 (1975)
25. C.J. Rauwendaal, Polymer Extrusion, Hanser Publishers,
Munich (1986).
26.L.P.B.M. Janssen, Twin-screw extrusion, Elsevier,
Amsterdam, (1978).
27. Z. Tactmor, P. Hold and L. Valsamis, Plast. Eng. J., ~,
20 (1979), ibid., ll, 34 (1979)
28.B. David and z. Tactmor, Intern. Polym. Proc.~, 1, 39
(1988)
29. L.N. Valsamis and G.S. Donoian, Adv. Polym. Techn., ~,
131 (1984)
30. J.M. McKelvey, Polymer Processin~, J. Wiley, New York
(1962)
31. Z. Tactmor and I. Klein, Engineering Principles of
Plasticating Extrusion, Van Nostrand Reinhold (1970)
32. z. Tactmor and e.c. Gogos, Principles of Polymer
Processing, Wiley-Interscience, New York (1979)
33. S. Middleman, Fundamentals of Polymer Processing,
McGraw-Hill, New York (1977)
34. w.o. Mohr, R.L. Saxton and C. H. Jepson, Ind. Eng.
ti, 1857 (1957)
35. G. Lidor and z. Tactmor, Polym. Eng. Sci., .l.Q, 450
36. G. Pinto and z. Tactmor, Polym. Eng. Sc i., 10, 279
Chem.,
(1976)
( 1970)
37.P.V. Danckwerts, Chem. Eng. Sci., ~, 1 (1953), ibid.,~,
93 (1958)
38. D. Bigg and S. Middleman, Ind. Eng. Chem. Fund., ~, 184
(1974)
39.L. Erwin, Polym. Eng. Sci., ~, 572 (1978)
40. L. Erwin and F. Mokhtarian, Polym. Eng. Sci., ~, 49
(1983)
41. H. Aref, J. Fluid Mech . , ~, 1 (1984)
42.D.V. Khakhar, H. Rising and J.M. Ottino, J. Fluid Mech.,
.1..12_, 419 (1986)
43.D.V. Khakhar and J.M. Ottino, Phys. Fluids, ~, 3503
(1986)
-34-
44. S. Smale, Bull. Amer. Math. Soc., 21, 747 (1967)
45.M.F. Doherty and J.M. Ottino, Chem. Eng. Sci., ~' 139
(1988)
46.W.L. Chien, H. Rising and J.M. Ottino, J. Fluid Mech.,
l..l..Q., 355 (1986)
47.D.V. Khakhar, J.G. Franjione and J.M. Ottino, Chem. Eng.
Sci., .1.2_, 2909 (1987)
48. I. Manas-Zloczower, A. Nir and Z. Tadmor, Rubber Chem.
'i'echn. ;u, 583 ( 1984)
49. I. Manas-Zloczower, A. Nir and Z. Tadmor, private
communication (1983)
-35-
CHAPTER 3
MODELLING OF COROTATING TWIN-SCREW EXTRUDERS
In many operations in polymer processing, such as polymer
blending, devolatilization or incorporation of fillers in a
polymerie matrix continuous mixers are used, e.g. corotating
twin-screw extruders (ZSK), Buss Cokneaders and Farrel
Continuous Mixers.
Theoretical analysis of these machines tencts to emphasize
the flow in complex geometries rather than generate results
which can be directly used (1-5).
In this chapter a simple model is developed for the hot melt
closely intermeshing corotating twin-screw extruder,
analogous to the analysis of the single screw extruder
carried out in 1922 and 1928 (6,7).
With this model, and more specifically with its extension to
the complete nonisotherrnal, non-Newtonian situation, it is
possible to understand the extrusion process and to
calculate the power consumption, specific energy consumption
and temperature rise during the process. These calculations
can be performed not only with respect to the viscosity of
the melt, but also dependent on the screw speed and the
screw geometry (location and number of transport elements,
kneading sections and blister3, pitch (positive or
negative), screw clearance and flight width).
To support the theoretical analysis, model experiments with
a Plexiglas walled twin-screw extruder are performed, in
actdition to practical experiments with melts on small and
large scale extruders, with very reasonable results.
Reprinted partly frorn: H.E.H. Meijer and P.H.M. Elemans, Polym. Eng.
Sci., ~, 275 (1988) by parmission of the Society of Plastics Engineers.
-36-
3.1 INTRODUCTION
Corotating twin-screw extruders are very often encountered
in the compounding and blending processes of polymers,
although some important drawbacks exist. The maximum torque
of corotating twin-screw extruders is still relatively low
in comparison with single-screw extruders and (conical)
counterrotating twin-screw extruders. This is inherent to
the flexible screw design and self-wiping action, necessary
to perform the primary task of corotating twin-screw
extruders, which is either to remove volatiles from the melt
(water, solvents, monomers) or to add fillers (glass, chalk,
talc, mica) via openings in the barrel. As a consequence,
all corotating twin-screw extruders must be underfed, which
means that the throughput locally, at least in the open
barrel sections is only a fraction of the maximum
theoretical throughput. Correct metering of the individual
components is often stated to be 80% of the compounding job.
In the modelling, attention will be focused on the hot melt
extruder, which in practice is used for devolatilization or
which is present after the melting section in each
compounding extruder. Solicts conveying-, transition- and
melting sections are difficult to analyse because no
distinguished melting mechanism can be recognized as in
single-screw extruders (8-11). Rather a mixture of solicts
and melt exists as in the dissipative mix melting mechanism
(12). Nevertheless, incorporation of the rnadelling of the
melting section will be important because during compounding
most of the limited torque is used in this stage of the
process. Sametimes even an important part of the dispersive
mixing is already achieved bere, because of the high
viscosity (low temperature) of the mixture (13).
In practice, corotating twin-screw extruders are usually
constructed with narrow flights. An exception to this is the
feed section, where broad flights are sametimes used to solve
-37-
problems with difficult-to-transport powders. However their
tetrahedral gap is from the geometrical point of view always
much larger than the one in counterrotating twin-screw
extruders. Moreover, they are constructed with closely
intermeshing screws in order to promate the self-wiping. As
a consequence the flights leave a completely open 8-shaped
chamber. Therefore, the transport mechanism is drag flow.
The analysis of the corotating twin-screw extruder can be
found in Ref. 4, but as in Refs. 14 and 15 too much effort
is paid to a detailed treatment of the complex geometry and
the reader easily gets lost. Rauwendaal (14) pays some
attention to the rnadelling of corotating twin-screw
extruders, but the analysis is incomplete and therefore of
little practical use. Of course, the more important extruder
manufacturers have developed their own computer programs to
predict the performance of their extruders depending on
screw design and to scale up the results from laboratory
measurements to production size (16), but for obvious
reasans they do not always present their know-how to the
competitors in the open literature.
Here, the corotating twin-screw extruder will be dealt with
as a single-screw extruder using the theories developed in
1922 and 1928. This is allowed with respect to the transport
characteristics of the melt-filled sections because of the
completely open channel.
3.2 SCREW GEOMETRY
Different screw elements exist: Single-, double- or
triple-flighted screws with different pitch, even with
negative pitch, mixing and kneading elements. Screw
contiguration is extremely flexible, and can be fitted to
the job, this being one of the major advantages of this kind
of extruders. At present mostly double- flighted screw
elements are used because of the larger useful volume.
Single-flighted screws are less popular because mixing
performance increases with the number of flights.
-38-
The most elementary screw geometry is given in Figure 3.1
and consists of a sequence of transport elements with
positive pitch combined with an element with negative pitch.
The principle of the analysis will be explained using this
geometry.
Figure 3.1. Elementary screw geometry.
In this elementary screwpart three functional zones can be
distinguished: part 'a', partially filled having a degree of
fill f (0 < f < 1); part 'b', completely filled, pressure
generating; part 'c', completely filled, pressure consuming.
In principle, every screw can be thought to consist of parts
'a','b' and 'c'. Some examples are given schematically in
Figure 3.2.
Local pressure gradients (following from metered throughput
compared with theoretically maximum throughput) and lengths
determine their relative dimensions.
For the moment, only one screw is considered, thereby
neglecting the presence of the intermeshing region. The
screw is thought to be stationary and the barrel rotating in
opposite direction as usual. Furthermore curvatures are
neglected by either looking very locally or by unrolling
-39-
the screw channel. In first approximation, the screw channel
is thought to have a rectangular cross-sectional shape, with
average height H and width W.
11 1 I I 11 11 11115\\M\\\\~1>
\\\\\\\\\\\\\\\\\V '• ~\\\\\\\\\\\\\\111~
\SS'\'\ '\'\S:SS'\S:S: '\'\SS'\\\\\\\\\\\\\\\\\\\\\\\\\\\//\\\\\)
\\\\\\ \\\\\\\\\\S\Jnt!BU!Vl\S\ \\\\\\ \\\\'1\\\\ll!f\\\\IJI[\\\\\\\\\\\\\\\\\\\\>
~\\\\\\IIIIIIIIIIIIIIIIIIIIIIK\91111111mmlliiiiiK\9111111111miiiiiiiiiK\~IIIIIIIIIIIIIIIIIIIIIKS9mllllllllmiiiiiiiK\9111111111111111111111K\~IIIIIIIIIIIIIIIIIIIIIK\\'llllllllllllllllllllll@~lllllllllllllllllllijK\JIIIIIIIIIIIIIIIIIIIII@ll>
Figure 3.2. Examples of screw design.
The barrelwall moves with a velocity, V, over the screw
channel under an angle ~, the pitch angle, and drags the
liquid into the direction of the die (cos~ - 0.95 neglected
here for the time being) .
V 11'DN
with
N revolutions per second
D diameter
-40-
The maximum drag flow capacity iu parts 'a' is (17):
(3 .1)
Qd = "2VHW ( 3 . 2 )
The real throughput Q is always smaller than Qd
(3. 3)
with f the (local) degree of fill, (see Figure 3.3a), Q is
approximated to be half of the real throughput because only
one of the screws is considered here.
a
Figure 3.3. Transport in parts a,b and c.
-41-
Parts 'b' are completely filled; therefore Qd = ~VHW. They
are generating pressure, consequently (see Figure 3.3b):
Because the metered throughput Qm = Q
everywhere, the pressure flow equals
Qp = -(1-f)Qd
(3 .4)
f.Qd is constant
(3 .5)
Parts 'c' are completely filled and consume pressure because
the drag flow is in the negative transport direction: Qd
-~VHW (see Figure 3.3c). This must be overcompensated by a
pressure flow to yield a net transport. Because of
continuity of throughput, it follows that:
(3. 6)
In rectangular ducts of width W and height H (the influence
of side walls is neglected here) the expression for pressure
flow of a Newtonian fluid reacts (7,17,18)
1 dp 3 Q -- H W p = - 12~ dz (3. 7)
By substituting in Eqs. 3.2 and 3.7, Eq. 3.4 can now be
written as
( 3. 8)
3.3 ANALYSIS OF SIMPLIFIED GEOMETRY
Provided that the leads of the forward and reverse zone are
numerically equal, that the number of flights in all zones
is equal, and that the pressure generated in zone 'b' is
dissipated in zone 'c', the simplified geometry now can be
analysed.
-42-
3.3.1 RELATIVE LENGTHS
First, attention will be paid to the calculation of the
lengthof parts 'a', 'b' and 'c' from the local pressure
gradients.
, a, Q =
dp
dz
, b' Q =
dp hence,
dz
'c' Q =
dp hence,
Fr om
dz
Eqs.
dp dz I/
b
f.l,2VHW
0
1 dp 3 l,2VHW -
12}1 dz H W f.l,2VHW
-l,2VHW -
= -(1 +
3.12 and
dp
dz c
1 dp
12}1 dz
6}-lV f)-
H2
3.14 it
1 - f
1 + f
3 H W f.l,2VHW
follows that:
(3. 9)
(3 .10)
(3 .11)
(3.12)
(3 .13)
(3 .14)
(3.15)
(The minus sign occurs because of different signs of the
pressure gradients in parts 'b' and 'c' .) In case of an
isothermal, Newtonian fluid the pressure gradient is
constant:
dp ~p
dz L (3 .16)
It is now possible to determine the length ratio with Eqs.
3.15 and 3.16:
1 + f
1 - f (3 .17)
-43-
(minus sign omitted) . Let the total length of the screw be
L; then the lengthof parts 'a', 'b' and 'c' are given by:
With Eqs. 3.17 and 3.18:
2 La = L - 1 - f Lc
If relative lengths are defined:
L ___a L
L .
~ and ll.c
and ll.c is given, this yields:
2 ll.a 1 1 f JLC -
1 + f ll.b 1 - f JLC
JLC = JLC
Example: if f 0.3 and JLC 0.2
0.37
3.3.2 SPECIFIC ENERGY
L __c_ L
then ll.a
The shear stress (N/m2 ) at the wall equals
du 'tw = ~ dy I
y=H
The shear ra te at the wall reacts (7,17,18)
du V 1 dp
dy I H + - - H
y=H 2~ dz
(3 .18)
(3 .19)
(3 .20)
(3. 21)
0.43 and ll.b
(3.22)
(3.23)
-44-
So, for the three parts 'a' ,'b', and 'c' this yields:
'a' Eqs. 3.10, 3.22 and 3.23,
y_ 'tw = }l H
'b' Eqs. 3.12, 3.22 and 3.23,
'tw = }l ~ (4 - 3f)
'c' Eqs. 3.14, 3.22 and 3.23,
'tw = }l ~ (-4 - 3f)
(3.24)
(3. 25)
(3. 26)
The force (N) acting on the wall equals the product of
shear stress and surface area where the stress is active.
'a' Eqs. 3.21, 3.24 and 3.27,
V 2 F = f }l trDL(1 - !!.cl H 1 - f
, b' Eqs. 3.21, 3.25 and 3.27,
V 1 + f F = }l trDL(4 - 3f) l!.c H 1 - f
, c, Eqs. 3. 21, 3.26 and 3.27,
V
F = }l H trDL(4 + 3f)l!.c
(minus sign in 1:w should be omitted here)
The total force is the sum of these three forces:
V }l H trDLf(1 +
2(3f + 4)
f
(3.27)
(3 .28)
(3. 29)
(3 .30)
(3. 31)
-45-
The torque (Nm) equals the force times the radius:
T (3.32)
The power consumption (Nm/s) is torque times screw speed
(rad. /s):
p T.N.211' (3.33)
From Eqs. 3.31, 3.32 and 3.33 it follows that
V 2(3f + 4) p = \A H
11' 2D2LNf(1 + f l!.c) (3.34)
Or (with COS<p - 1 still)
Q f l,(zVHW' V 'Tl'DN; w 'Tl'Dsin<p (3 .35)
21.l11'DLNQ 2(3f + 4) p
H2 sin<p (1 + l!.c)
f (3. 36)
Without screw elements 'c' with negative pitch l!.c = 0 so
the term between the brackets equals 1 and the usual
expression for the power consumption of a single screw
extruder is found (19). With the elements 'c', the term
between the brackets yields the relative importance of the
presence of elements with negative pitch, see Table 3.1.
The lower the value of f, the more important is the
contribution of the completely filled part to the total
power consumption. In absolute sense, its contribution is
less, because the filled length Lb decreases with
increasing pressure gradient in part 'b'.
Another way to illustrate the relative importance of the
channel part which is completely filled (total relative
length l!.b+l!.c) by the presence of 'c' elements is given by
the ratio
~ p
a
l+f (4-3f) 1-f + 4 + 3 f
2
f (1-f)
-46-
4-3f2
f (3.37)
Thus: f = (0.4; 0.3; 0.2) yields Pb+c/Pa (9; 13; 20).
In words, for f = 0.3 a part with negative pitch involves
the consumption of 13 times the power relative to a partly
filled channel 'a' of the same length. These simple
calculations illustrate the relative importance of negative
pitch parts, yet it must be kept in mind that until now the
power consumption in the clearance between flight and
cylinder has not been taken into account .
Table 3 . 1. The relative importance of elements with negative pitch:
power consumption in a completely filled channel (%) .
Q,c 0.01 0.05 0.1 0.2
f = 0.4 26l1,c/(1 + 26l1,c)•lOO% 21 56 72 84
f = 0.3 33l1,c/(1 + 33l1,c)•100% 25 62 77 87
f - 0.2 46l1,c/(1 + 46l1,c)•100% 32 70 82 90
The specific energy (kWh/kg) is power (kW) divided by
throughput (kg/h). From Eq. 3.36 and with Q = (m3/s):
Qm = Q.3600.p
This yields:
E sp.
J.nr DLN (1 +
2(3f + 4)
f Q, )
c
0.4
91
93
95
(3.38)
(3 .39)
Again Eq. 3.39 yields, for Q,c = 0, the identical result as
derived for single-screw extruders .
-47-
3.4 IMPROVED ANALYSIS
Until now the analysis was relatively naive and only
qualitative. Nevertheless, some understanding of the basic
working principles has been raised. Next, some improvements
will now be incorporated in the model.
3.4.1 LEAKAGE FLOWS
In corotating twin-screw extruders leakage flows are more
important than in single-screw extruders, because they are
relatively large with respect to the throughput, which is
only a part 'f' of the maximum theoretical throughput by
drag flow. Moreover, the screw speed is higher than that of
single-screw extruders because of its desired significant
contribution to the mixing process.
So far, we have not considered the influence of the flights
since we have tried to simplify the treatment as much as
possible. Improvements however can, and will, be easily
incorporated.
To compensate for the area occupied by flights (double
flighted screws in most cases), the channel width should be
reduced and replaced in all equations by:
w ~Dsin~ (1-ne) (3.40)
where e is the relative width of the flights, given by
~ e = w ~Dsin~ (3.41)
with Wf the flight width measured (as W) perpendicular to
the flight, and n is the number of flights. A typical value
of e is 0.1.
-48-
For the present analysis of double- or triple-flighted
screws, all flights can be taken together; mixing, which
indeed is dependent on the absolute number of flights, is
not being analysed bere. The corrected value of the channel
width must be introduced into all the equations derived so
far; among other things, this correction influences
directly the value of f, the degree of fill, in Eqs. 3.2
and 3.3.
Figure 3.4. Unrolled screw channel.
The local throughput that must be transported to the die is
larger than the metered throughput Qm, because of backflow
through clearances. Simple drag flow is enough to estimate
the backflow because, in general, pressure flows are
negligible in narrow channels. Therefore Qm should be
replaced by (see Figure 3.4):
(3.42)
with
~v&~Dsin~cos~(l-ne) (3.43)
where & flight clearance
-49-
The degree of fill, f, is influenced once more by this
correction. Having corrected for f, the cos~ which was
neglected so far also can be brought in, starting with Eq.
3 .1.
3.4.2 POWER CONSUMPTION OVER THE FLIGHTS
Over the flights power is consumed. The shear stress equals
V
}. = Jlf ó (3. 44)
with Jlf the viscosity above the flight and & the flight
clearance.
The surface area A equals
Af = nelTOL (3.45)
Therefore, force, torque and power are respectively
V Ff Jlf ó nelTOL (3. 46)
V Tf ~Jlf ó
ne1T02L (3. 4 7)
V pf = Jlf ó ne1T2o 2LN ( 3. 48}
A comparison of the power consumed above the flights with
that consumed in the channels can be made now if Eq. 3 .34
is corrected according to Eqs. 3.40, 3.42 and 3.43 and cos~
is incorporated:
V 2(3f + 4} Pc Jlc H
1T202LNf(l + f !te) (1-ne} (3. 49} and
Qm Qm + ~VólTOsin~cos~(l-ne) f
~Vcos~HW ~VH1TOsin~cos~(l-ne} (3.50}
-50-
and, rewriting this expression,
&
f = + ~2D2HNcos~sin~(l-ne) H
(3.51)
yielding
~f H ne 1 (3 .52) Pc ~c & (1- ne) (1 + 2(3f + 4)2-c)
In Eqs. 3.44 throügh 3.49 and Eq. 3.52, ~f and ~c are
introduced to allow one to be able to choose a somewhat
lower Newtonian viscosity above the flights. This is
necessary because of pronounced shear thinning and
temperature rise due to local dissipation here.
With the typical data
H
& 40; n 2; e 0.1;
\lf
~c = 0.15
Table 3.2 follows. As is clearly illustrated in thi s Table,
power consumption between the screw flights and the barrel
is large compared to the power consumption in the channel.
This is typical for these (usually) low values of degree of
fill, f.
Table 3 . 2. Power consumption above the flights in comparison with
that in channels (Ef/Ecl. Parameter: relative length of
parts c (l/, c ).
ll,c 0.01 0.05 0.1 0.2 0.4
f - 0.4 3.0 1.6 1.0 0.6 0 . 33
f ~ 0. 3 3.8 1.9 1. 2 0.66 0.36
f - 0. 2 5.2 2 .3 1. 3 0.74 0 . 39
-51-
3.4.3 MIXING ELEMENTS
So far we have only dealt with the simplest combination of
screw elements with positive and negative pitch; moreover,
only one screw has been analysed. The geometry of two
combined screws is more complex and also many other
elements exist, as is illustrated in Figure 3.5.
fffiH1ft ma· .. -m~ ...
I
~~~ ~ ~ ~
Figure 3 . 5 . Various mixing and kneading elements.
Performing calculations on these elements is now rather
straightforward. First, we have to evaluate in more detail
the effective channel width to be used in the calculations.
Bath screws have to be visually investigated
simultaneously, preferably in the unrolled state (see Refs.
1, 4, 5 and 14). For the throughput, it follows:
2n-1
n (3.53)
the total metered throughput, corrected for leakage flow
over the flights. The factor (2n-1) represents the number
of independent channels in a multiple flighted twin-screw
-52-
extruder, with n the number of flights (4). E.g. n=2 yields
(2n-1) = 3 channels; n=3 yields (2n-1) = 5 channels. The
actual channel width through which the material is being
transported amounts to:
2n-1 w n nDsin~(1-ne) (3.54)
Second, for the mixing elements a main pitch angle ~,
positive (e.g.~= 45°), neutral (~ = 0°) or 'negative'
(e.g. ~ = 135°) must be recognized. Finally, we must sum up
the number of holes in the flight leaving room for positive
or negative drag- and pressure flows. If the lengthof the
sum of the remaining screw flight is defined as being a
factor r of the totalscrew flight length (Lf rnD/cos~),
the sum of the holes equals the factor (1-r).
A leakage by drag flow back into the previous channel
exists, caused by the presence of the holes. Therefore the
throughput brought about by the extruder should be
increased to compensate for this leakage flow:
2n-1 Q n
with
2n-1 where FN =--n--(1-ne)
(3.55)
(3.56)
F0 is a correction factor, like Fp, that compensates for
the influence of the side walls (the flights) on the drag
flow in the case of the ratio of channel depth to channel
width not being negligible anymore (see Refs. 6, 18 and
20) . Not only is the throughput influenced
-53-
by the presence of the holes, but also is the pressure
gradient in axial direction dominating the one in
down-channel direction. One has to consicter the parallel
combination of two flow resistances, represented by their
characteristic cross-sections: The rectangular screw
channel and the passage through the holes (For more
details: See Chapter 4). This is reflected in the Eqs. 3.57
and 3.58.
k
Qd - J.l Ap
The factor k can be calculated:
1 k = 12 !!. H
3'11'DFPC(sin
2q> Fifp + (1-r) Fp')
(3.57)
(3.58)
In a neutral kneading element, for instance, neither flight
nor pitch angle can be distinguished (r = 0, Cf>= 0), so the
leakage flow (Eq. 3.55) completely compensates the drag
flow (Eq. 3.2), withall the correctionsas indicated. In
this case axial pressure gradients are the only transport
mechanism, compare Eq. 3.7.
The flow in the intermeshing region results in an extra
contribution to the conveying capacity and can be treated
using the description given by Booy (5) . The average
contribution, Qa, per N revolutions to the drag flow is
given by:
(3. 59)
with the displacement factor ka a tabulated function (see
Ref. 5)
-54-
This term can be introduced in the equations used so far,
Q (3. 60)
The analysis is summarized in Table 3.3, yielding the
expressions for the pressure gradients and volume flows
with all corrections brought in until now. Figure 3.6 shows
the throughput versus pressure characteristic for the
various screw elements mentioned above.
Q (g /mln)
IV
2
111
100 2 0 ~p/di.10E·3 (Palm)
11
Figure 3 . 6. Throughput versus pressure characteristics for various screw
geometries. I transport e1ements, ~ = 17.66 • ; II transport
e1ements, ~ = 180 ° -17.66 • ; III kneading e1ements, ~ = 45 ° ;
IV kneading e1ements, ~ = 135°.
All kinds of screw geometries can be modelled now, also
screws where no negative pitch elements are present or
screws which consist of any combination of transport,
mixing or kneading elements. Before dealing with that, we
have to analyse in yet more detail what happens if several
different screw elements are combined sequentially.
-55-
Table 3.3. Expressions fordrag and pressure flow in screw elements,
based on the total metered throughput, Qm•
Definition: Q = Qd + Qa + Qp
Q Qm + QL + QLD
(1)+(2) -Qp = (Qd + Qa- QL- QLD) - Qm
with
pressure flow
drag flow
flow in the
intermeshing
region
leakage flow
over flights
leakage flow
k - t.p 1.1
(1)
(2)
(3)
(4)
(5)
(6)
(7)
through flights QLD 1/2 v cos~ H(1-r) nu sin~ FN F0 F0c (8)
V nDN; Lf r = _...",.,. __
nD/cos<p
2n-1
n (1-ne);
k 1
3 ( . 2 F .,. + (1 ) .,. ') 12 ~ H nD FPC s1n ~ N"P -r "p
ka, F0, Fp, F0 ç, Fpc (see Refs. 5, 14, 17);
transport element: r = 1; 0 < ~ < 180°
kneading element: 0 < r < 1 ; 0 < ~ < 180°
3.4.4 SEQUENCE OF SCREW ELEMENTS
(9)
(10)
In the Newtonian, isothermal case with constant viscosity, a
combination of different screw elements can be investigated
independently of the preceding and following elements. To be
more precise: all elements of the same kind cao eventually
be packed together in order to calculate energy and specific
energy.
-56-
In this case, a complete screw like the one in Figure 3.7
can be dealt with as a pure a, b, d, c screw with a sequence
of partly filled parts 'a' (with total length equal to the
sum of the individualparts 'a'), completely filled
pressure-generating parts 'b', completely filled pressure
generating or pressure-consuming parts 'd' (mixing
elements), and completely filled pressure-consuming parts
'c' (all with total length equal to the sum of their
individual components).
When, however, more detailed information is required than
only (specific) energy - for instance, if one wants to know
where the screw is partly filled or where the viscosity
changes during extrusion (the non-isothermal, non-Newtonian
case) - then only a repetition of the same calculations for
the individual sequential components and combinations can
give the answers. In all cases, the influence of one element
(pressure-consuming) can be noticeable in the preceding
sections (in a complex way, because many leakage flows are
introduced) .
If a combination (a,b,d,c,a,b,d,c) is used (see Fig. 3.7),
then the lengthof the second partly filled part 'a' might
become negative with increasing throughput, Q. This means
that the pressure buildup capacity of the second part 'b' is
not enough to compensate for the pressure consumption in
parts 'd' and 'c'. As a consequence, the absolute pressure
at the end of the first part 'c' has a positive value, the
first part 'b' increases in length to generate this extra
pressure and the first part 'a' decreases in length. If even
the lengthof the first part 'a' is negative, the transport
of this throughput is no longer possible with this screw
design: The critical screw speed has been reached.
-57-
p Q » a >>
Q »
a b de a b de a b
Figure 3.7. Pressure profiles and filled lengths. Parameter: the metered
throughput Om·
3.4.5 NONISOTHERMAL POWER LAW CALCULATIONS
The isothermal Newtonian treatment of the flow in corotating
twin-screw extruders given above is very successful in
providing qualitative information about working principles
and consequences of different screw design in the melt
section. The best understanding originates from the most
simple model. However, quantitative answers cannot be given
because the influence of temperature and shear rate on the
viscosity (and on local pressure generating capacity) must
be included.
It is beyond the scope of this Chapter to give a detailed
analysis of the nonisothermal power law rnadelling (see Ref.
21), butsome general remarkscan be made. In principle, the
treatment is completely analogous to the Newtonian
modelling, but incremental steps over the length of the
extruder must be incorporated in order to be able to define
a local viscosity depending on local
-58-
shear rates, local temperature and, in the case of solvent
containing melts, local concentrations.
The viscosity of a power law fluid depends in general on the
temperature and on the second invariant of the rate of
strain tensor. The components thereof are calculated from
the individual drag- and pressure flows. The influences of
the main drag component V/H = ~DN/H in the channel and
V/& = ~DN/& in the clearance between flight and barrel are
dominant at first sight.
With the local viscosity,it is possible to compute the local
velocity, the velocity gradient, and the pressure gradient.
Given the local throughput Q and the conservation laws of
momenturn and mass, it follows that
dp d -b(T - T0 J du n
(J.Io e (-) (3. 61) dz dy dy
H
Q w f udy (3.62) 0
These equation:s finally result .in a set of two implicit
equations for the two unknowns (the pressure gradient and an
integration constant) . Successful salution of this system
(e.g., with a Newton-Raphson procedure) strongly depends
upon the first estimate of the values of the two unknowns. A
much more convenient way to solve the problem is to write
Eq. 3.61 in a different way:
dp d -b(T - T0 J du n-1 du (-)
dz dy (J.to e dy dy) (3.63)
which means:
dp d du
dz dy (J.I (z, y) dy) (3.64)
-59-
The tridiagonal matrix resulting from this expression
combined with a last row completely filled with the digits
1, 4 and 2, resulting from the discretisation of mass
continuity (Eq. 3.62) according to a Simpson-like
integration rule results in a straightforward solution. Very
fast recursion formulas exist for the composition and
salution of tridiagonal matrices (22), while the last row of
the matrix can be brushed by Gaussian eliminatien already
during the formation of the tridiagonal matrix. Besides, an
extension to the threedimensional flow field now easily can
be incorporated :
ap a avz
az ay (}.I (z, y) ay-> (3. 65)
ap a avx
a x ay (}.I (z, y) ay-> (3. 66)
-b(T - T0 > av av 2 (n-1) /2
l.l (z, y) = llo e ( (___z) 2
+ (______A) ) (3. 67) ay ay
For this (coupled) system the same tridiagonal matrix must
be solved for bath x- and z- directions; the only difference
being the boundary condition at the barrel:
V2 = Vcos~; Vx = Vsin~ (3. 68)
and the throughput Q2 and Qx
Coup1ing of the equations via the viscosity is not only due
to the shear rate dependency but also via the temperature
(see Eq. 3. 67).
The next step, therefore, is ca1cu1ating the 1ocal
temperature. This can be performed re1atively simp1y because
the flow in the channel is considered to be perfectly mixed
(one screw taking over the material from the other while
completely reorientating it) . Therefore a 1ocal heat balance
(Fig. 3.8) dictates that the heat stored in the local volume
(resulting in a temperature rise) equa1s the dissipative
heat generated by the velocity gradients plus the heat
exchanged with the wal1s (positive or negative) .
-60-
The last term is calculated with an average heat transfer
coefficient <a> multiplied by the temperature difference.
This engineering <a> is not known a priori, and experiments
must yield its value. After the new temperature has been
calculated (Tout in Fig. 3.8) the average value of the
viscosity is adapted to the average temperature in the
volume and calculation of pressure gradients and average
temperature is repeated. If the results are not improved
anymore (only a few iterations are needed), the next
incremental step is taken. See Table 3.4.
m l'q, Tin D~ m PCp Tout
dx
Figure 3.8. Heat balance over an incremental step dx.
Table 3.4. Nonisothermal power law calculation. Numerical scheme.
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
Newtonian analysis
Estimation of lengths
Step dz
Determination of local pressure gradients, Eqs.
3.59, 3.60 and 3.61.
Determination of local mean temperature, Fig. 3.8
Analysis of leakage flow, average temperature and
power consumption
Corrections on throughput and mean temperature in
channel because of leakage flow(s).
Return to 4, until no changes in mean temperature
and pressure gradients occur anymore.
Return to 3, until checkpoint on absolute pressure
is reached.
If not correct: improved estimation of lengths,
return to 2.
If correct, next screw section or end.
-61-
It is recommended not to make the incremental steps too
small (a good cernpromise is one third of the diameter) as
the elements are coupled via leakage flows and the
calculation must be iterated because of the inflow of
material (with an other mean temperature) from leakage gaps
and following channels.
Moreover, when the calculation is started, only analytica!
estimations of the lengths of parts 'a' and 'b' are known.
Therefore, testing at parts where absolute (atmospheric)
pressures are known (degassing houses; feed ports; in the
worst case at the end of the die) is necessary to check
whether the calculated pressure equals the known pressure.
If not, a new guess of the lengthof element 'a' and 'b'
must be made and the calculations repeated.
3.5 CALCULATED RESULTS
As a demonstratien only a limited nuffiber of results will be
presented here. At first, a comparison of simple isothermal
Newtonian calculations with those for the nonisothermal
power law case will be made.
3.5.1 SPECIFIC ENERGY
In Figure 3.9 the specific energy for a hot melt extruder is
plotted as a function of throughput with as parameters the
screw speed and the degree of fill. Figure 3.10 shows the
same plot for the nonisothermal power law calculations. The
conclusion of this comparison is that the Newtonian model
provides qualitatively the right information: minimum
specific energy only at maximum degree of fill. Quantitative
predictions, however, can only be made with the more
complicated calculations shown in Figure 3.10. Especially
when the influence of a screw modific ation is t o be
predicted only the nonisothermal calculations are important
(as experienced in practice), because changes in viscosity
through local heat dissipation in one part of the screw have
great influence on the heat dissipation in the following
parts.
ëi -" ....
0.5
~ 0.4 =. >OI êi; c Ql 0.3 0 Ql Q.
f/)
' •· ' I I
0.2
0.1
:-62-
Parameter 1: screw speed [i'P~ Parameter 2: filling ratio [ -]
160
140
120
100
80
60
40
0~---r----.----.----.----.----.-----.----.----.----, 0 2000 4000 6000 8000 10000
--.- .,. Output(kg/hr]
Figure 3.9. Specific energy as a function of throughput.
OI .5
.:ti. ...... -'= 3: .4 ~ >-Cl Qj .3 c::: Ql
u Gl D. .2 (IJ
.1
0
Parameter: screw speed, degree of fill.
Isothermal, Newtonian model. From Ref. 23.
0
40 60 ' 80
' ~_oo 120
2000
f=0.1
4000
Parameter 1: screw speed [rpm] Parameter 2: filling ratio [-]
6000 8000 10000 --.;,. Output [kg/hr]
Figure 3.10. Specific energy as a function of throughput.
Parameter: screw speed, degree of fill.
Nonisothermal, power law model. From Ref. 23.
-63-
3.5.2 COMBINATION OF PARTS BAND C
As a second example, a (local) combination of transport- and
mixing elements has been investigated: a positive, neutral
and negative mixing element as is shown in Fig. 3.11. The
difference in pressure gradient in the mixing elements is
clearly demonstrated by drawing horizontal lines of 'zero
pressure' from the end of the mixing element towards the
left; the filled length can be determined graphically. The
difference in pressure gradient between positive and neutral
mixing elements is not very large. However, because of the
positive transport, screws with positive kneading elements
remain (almost) empty. This is why those kneading blocks are
in practice always followed by a screw element (or kneading
block) with a negative pitch.
50,--------------------r------------------~ ZSK 40
N = 150 rpm
40 Q = 40 kg/hr
"L' 30 0
_Q .....__"
Q) 20 L
:J (/) (/) Q) L 0... 10 -- ------------- - --- ---- ---------- --- 111 .
0~------------.-----~-------.------,-----_, 0 40 20 20 40
transport sectlon x ( m) kneading sectlon -10J-------------------~------------------~
Figure 3.11. Pressure gradient in a combination of a transport
section (~b = 27.4 ° ) and kneading section with different
pitch angle: I ~c = 36.5°; II ~c = 0 ° ; III ~c = 180° -36.5°.
From Ref. 23.
-64-
3.5.3 END TEMPERATURE
The third example refers again to devolatilization extruders
and reveals the end temperature and specific energy in
dependenee of processing conditions and materials, see Table
3.5.
Table 3.5. Comparison between final results from model calculations and
those from measurements for a melt extruder (ZSK300) under
several processing conditions.
measured values calculated values
viscosity powerlaw
at y=l s-1 index
(Pa. s)
12,500 0.43
11,300 0.43
10,500 0.43
9,000 0.43
6,250 0.43
485 0.82
296
292
284
294
286
228
0.17 297 0.18
0.15 294 0.156
0.137 280 0.148
0.126 293 0.13
0.123 287 0.125
0.052 224 0.05
From these and many other simulations it can be stated that
the more exact predictions about extrusion behaviour can
only be obtained when a temperature (great influence) and
screw speed (smaller influence) dependent heat transfer
coefficient <a> is used. It is only the heat transfer
coefficient on the polymer side that is responsible for this
dependency because a thicker isolating polymer layer is
formed with lower wall temperatures, screw speeds, and with
larger leakage gaps (25,26). A procedure for calculation of
<a> is presented in Ref. 21.
3.6 EXPERIMENTAL VERIFICATION OF THE NEWTONIAN,
ISOTHERMAL ANALYSIS
The correlation between theoretica! predictions and
practical values of Section 3.5.3 is fairly good. However,
-65-
one has to realize that an automatic feed-back exists:
Because of the temperature dependency of the viscosity a
more or less limiting value for the melt temperature and the
(specific) energy is present.
Therefore, to provide for a more direct check on the
validity of the Newtonian, isothermal analysis, experiments
were performed with model liquids (paraffinic oil, ~1 23 oc 0.2 Pa.s) on a Berstorff ZE25 laboratory corotating
twin-screw extruder.
It was possible to investigate the drag flow capacity of the
two combined screws, including the extra conveying action of
the intermeshing region between the screws.
To determine the pressure gradients and (locally) filled
lengths, the extruder could be fitted with a Plexiglas
cylinder.
3.6.1. THROUGHPUT VERSUS SCREW SPEED CHARACTERISTIC
The purpose of this experiment was to determine the maximum
drag flow capacity of different screw geometries. The
maximum theoretical drag flow can be obtained by letting
Op = 0 in Eq. 3 of Table 3.3:
This yields: Om (3. 70)
Two distinct screw geometries were investigated:
(i) positive transporting elements, having no leakage
through the flights; Eq. 3.70 reduces to
Om = Od + Oa - OL (3. 71)
and
(Ü) positive kneading elements
By substituting Eqs. 5 through 9 of Table 3.3, the
throughput 0 is obtained as a function of N, the screw speed.
-66-
(i) positive transporting elements (3.72a)
ó 2n-l ~ 3 Qm = ~VCOSfP H ~Dsinff' (1 - H) ( -n-(1-ne)) JD EDe+ 4 D Ntanff'ka
(li) positive kneading elements (3.72b)
ó 2n-1 ~ 3 Qm = ~VcosfP H ~Ds inff' r ( 1 - H ) ( -n-( 1-ne) ) JD EDe + 4 D Ntanff'ka
Table 3.6 summarizes the relevant geometrical parameters
used in the calculations. In this Table, H represents the
nominal channel depth. The real cross-channel profile can be
calculated from Eq. 10-7 of Ref.27, which gives the
cross-channel profile of the extruder screw as a function of
the circumferential angle e:
H(9) (3.73)
with Lc = the centerline distance between the screws
In the evaluation of the experiments, this real
cross-channel profile will be approximated by a rectangular
cross-section, with depth Heff· See also Figure 3.12.
E' s I
1.0
0.0
1.0
2.0
3.0
4.0
5.0 +---,----.-,----,----,.-.,--l--,----,--,--,----,--c-----1 -6.5-5.5-4.5-3.5-2.5-1.5-0.5 0.5 1.5 2.5 3.5 4.5 5.5 6.5
x (mm)
Figure 3.13. Cross-channel profile of the self-wiping screw
geornetry of a corotating twin-screw extruder
calculated frorn Eq. 3.73. The dashed line indicates Heff·
-67-
Table 3.6. Numerical values for the geometrical parameters plus
correction factors of the Berstorff ZE25 corotating
twin-screw extruder. (i) positive transporting elements;
(ii) positive kneading elements
ge ometry (i) (ii)
diameter D (mm) 25 25
channel depth H (mm) 4 4
flight clearance & (mm) 0.15 0.15
flight width Wf (mm) 1.4 6.6
Heffective Heff (mm) 3.1 3.1
pitch angle Cf> 17.66 ° 45 °
number of flights n 2 2
relative remaining
screw flight r 1 0.23
correction factors (14) H Heff H Heff
Fo 0.75 0.81 0.76 0.84
Foc 1.02 1.02 0.88 0.9
Fp 0.73 0.79 0.74 1.0
Fpc 1.15 1.11 0.98 1.0
Fp, 0.85 0 . 85
displacement factor ka 0.17 0.17
(Eq. A-10 in Ref. 5)
The maximum (theoretica!) drag flow capacity, obtained from
Eqs. 3.72a+b and from the data in Table 3.6, can be compared
with experiments. The extruder was completely filled (f = 1)
with paraffinic oil (SHELL Ondina 68; Pi 23 oç = 872 kg/m3,
Dl23oc = 0.2 Pa.s). No die was present, to prevent pressure
generation. Care was taken that the oil level in the feed
port remained constant. The throughput was measured at
several screw speeds and corrected for gravity.-induced
leakage flow.
Figure 3.13 shows the measured throughput versus screw speed
characteristics for different geometries, as compared with
those calculated from Eqs. 3.72 a+b.
-68-
50
....... 40 ..: .c: ....... 0) ~
conveylng screw - 30 -::J Q. .c: 0) ::J 20 0 ... .c: -
10 knaadlng sorew
0 0 50 100 150 200 250
screw speed (rpm)
Figure 3.13. Comparison of measured and calculated throughput versus
screw speed characteristics for transport elements
(~ = 17.66° ) and positive kneading elements (~ = 45 ° ).
Transporting elements show a large deviation (+25%) from the
predicted throughput, whereas in the case of positive
kneading elements, the agreement between experimental and
calc ulated results is relatively good (-10%), consictering
the assumptions in the analysis. Fitting the measured
throughput data of the transporting elements with a channel
depth H = 4 mm gives better agreement. Obviously, the
transporting screw elements are not completely self-wiping,
so l arger leakage gaps e xi s t. The main differenc e, howeve r,
betwee n transporting and kneading element s is t heir ability
to generate pressure, which will be discussed in the next
Section.
3 .6. 2 PRESSURE GRADIENTS
As stated in Section 2 .1, corotating twin-screw e xtruders
normally operate almast c ompletely under atmospheric
-69-
pressure. The screw is partially filled, having a degree of
fill f < 1 (parts 'a'). Pressure is generated only in those
parts of the screw where a combination of elements with
positive pitch (parts 'b') and elements with negative or
neutral pitch (parts 'c' and 'd') is present. Parts 'b', 'c'
and 'd' are completely filled.
It is possible to determine the axial pressure gradient in
parts 'b', 'c' and 'd' visually via the setup shown in
Figure 3.14. The Plexiglas cylinder is fitted with a series
of open tubes, which are directly connected with the
extruder channel. By measuring the oil level in the tubes,
the axial pressure gradient can be obtained via regression
analysis and compared with results from calculations.
Figure 3.14 Visualization of the axial pressure gradient in a
corotating twin-screw extruder.
(a) Combination of transport elements ha ving positive pitch
('{'b 17.66°) with transport elements ha ving negative pitch
('{'c 180°-qlb; Lc = 50 mm). Om= 36 g/min.; N = 130 rpm.
(b) Combination of transport elements having positive pitch
('{'b 17.66°) with kneading elements having negative pitch
(~c = 180°-45°; Lc = 150 mm). Om= 26 g/min.; N = 190 rpm.
-70-
Figures 3.15 and 3.16 compare the calculated and
experimental pressure gradients for different screw
geometries and processing conditions. The pressure gradients
in parts 'b' and 'c' are calculated from the Eqs. 3 through
10 of Table 3.3, plus the geometrical data in Table 3.6.
They can also be determined graphically from the pumping
characteristic (Figure 3.6), by drawing horizontal lines,
according to the present throughput followed by
interpolation towards the desired screw speed.
The correspondence between calculated and experimental
pressure .gradients in the kneading elements having negative
pitch (Figure 3.16) is quite good. For the pressure
gradients in transporting elements, however, errors as large
as -25% (Figure 3.15) or even -60% (Figure 3.16) are found.
This can be explained by consictering the (axial) pressure
gradient as a function of the channel depth, as shown in
Figure 3.17. Transporting elements exhibita strong decrease
in pressure gradient as the channel depth increases from
Heff = 3 mm to H = 4 mm. In kneading elements, however, the
pressure gradient proves to be less sensitive to changes in
channel depth.
{ !:;,
~
JOOOOO,.--,..------,---------------,
200000
100000
0
transport
knea~~ -------·-·-2
H (mm) 4
Figure 3.17. Axial pressure gradient as a function of channel depth as
calculated frorn the equations of Table 3.3 and the data of
Table 3.6. Qrn = 150 g/rnin; N = 200 rprn.
"' ,. 0 "l
P !Palm) P (Palm) PI f-'· ·o t-' · colculated vol'u~s H rt §. 'ê cclculofed valuts 0=15 N=188
~ ::r f-'· H
CD rt ::l CD 3000 rt H PI CD PI rt w H ::l t-'·
c.o 0 ..... '0 ::l (J1
-;; 0 2000 H 0
rt Hl () t-'· PI
c.o ::l rt ..... 0 IQ H 0 H PI .: CD CD ::l ..... ,. ..... c.o PI
CD '0 rt c.o ~ 0 CD '0 H p.
CD ::l rt CD rt f-'· PI p. c.o ::l ::l ' IQ p. 1 0 140 120 100 80 60 40 ::r
1 (mm) I ex. !mml H PI CD CD IX . '0 ~ i-' x negatlve tranaport (l0 - 26 mm) pos. trensport negative kneeding tLc- 150 mm) pos. tranaport ë f-'· CD '0
::l i3 CD IQ CD H
::l ~-::l rt 0" CD c.o CD
P (Palm) P (Palm) IQ ::l f?~Perlmentol vcftJes experfmentol valuts PI ::r rt
rt rt PI PI ( ::r t-'· ~ i-' .._] H ~ t-'· 3000 3000 I-' 0 CD ::l '0 ( .: IQ H
IQ '0 CD ::r f-'· '0 c.o '0 rt 0 c.o .: 0 c.o .:
2000 rt ::r f-'· H rt CD
iä .:ë" t-'· ~ IQ ...... CD H
i3 PI t-' · '0 p. 2. ..... f-'· f-'·
"' rt CD 0 0 ::l
0 ::r rt I c.o
..... .:ë" ..J f-'·
::l 0 140 120 100 80 60 40 CS\
1 ex. (mm) 1 ex . !mml" CS\ PI 0 ..... negative tranaport (lc- 25 mm) pos. transport negative kneadlng (lc - 160 mm) poa. transport ..J
CS\ m
0
(a) (b)
'0 ::r '"'l P (Po/m) P (Po/ml 1-'· ll> 1-'· colculoted volut;ts N=188 colculoled volues N=235 ("t ~ <Q 0 1-'· c ::r ::;, t1
<Q (1) 3000 3000
~ '0 w 0 co .... 1-'· "' .... ("t
2000 w 1-'· U1 ~ )>
(1) co '0 '"'l 1-'· 1-'·
"' ("t <Q ll> 0 c t1 ::r t1
~ (1)
(1) ~ w ("t (1) .... t1 (J1 1 0 140 120 100 BO 60 40 ....
I ax. lmml I ax. lmml _, ("t Hl nagative transport llc • 25 mml pos. transport negative k.neading llc- 150 mml pos. transport ::r "' 0 t1 "' t1 0 0
c ll> <Q ::r :( 0
P (Pa/m) P (Palm) '0 1-'· 0 experimentol ..-olues experlmentol voNes c ("t g. I ("t ::r
1-'· -..1
~ )I" ::;, 3000 3000 [\.) ::;, ll> I
'- (1) ("t
s ll> 1-'· 1-'· 0. 0 2. 1-'· ::;,
::;, <Q 0 2000
Hl (1) .... ("t (1) t1 s ll> (1) ::;, ::;, co ("t '0 co 0
t1 ::r ("t
ll> 1-'· ~ ::;, 1-'· <Q 0 140 120 100 BO 60 40 ::;,
Iu. lmm) lex. (mml <Q (1) .... negative transport tl.c- 26 mm) pos. transport negative k.neading (Lc • 160 mm) pos. trensport ::;, ~ (1)
<Q (1) ll> ::;, ("t ("t
1-'· co ~ (1)
(a) (b)
-73-
By combining Eqs. 3.15 through 3.17, it is possible to
calculate the filled length Lb from the measured pressure
gradients:
(dp/d!!.l I L - L • ( ) bax.- cax. (dp/d!!.llb (3. 74)
The filled length thus obtained is only a crude estimate,
because of the errors in the pressure gradient. In the next
Section, a more accurate method will be discussed.
3.6.3 FILLED LENGTHS
Via the method discussed in the previous Section, the
axially filled lengths can be estimated. Comparison of these
lengths with calculated values inevitably leads to large
deviations. The calculations (using the equations of Table
3.3) yield results that are valid in down-channel direction.
Of course, these results can be converted into the
corresponding values for the axial direction, but the best
evaluation can only be expected with filled lengths that are
measured in down-channel direction.
3.6.3.1 EXPERIMENTAL SETUP
Experiments were performed with paraffinic oil (SHELL Ondina
68, ~1 23 oc = 0.2 Pa.s) in the Plexiglas-walled laboratory
corotating twin-screw extruder. With the setup shown in
Figure 3.18, the combination of a part c (transport elements
having negative pitch or kneading elements - neutral or with
negative pitch) with transport elements can be investigated
directly, campare Sectien 3.5.2.
-74-
Figure 3.18. Setup of Plexiglas-walled laboratory corotating twin-screw
extruder with High Speed camera.
To measure the length of the channel that is completely
filled preceding a part c at different screw speed and/or
throughput, a High Speed camera was used. Figure 3.19
represents one image of the film, showing the transition of
a partly filled channel into a completely filled channel, as
can be seen with a Motion Analyzer.
Figure 3.19. Combination of a transport element having a pitch angle
~b = 17.66 ° with a negative transport element
(~c = 180°-~b). The screws rotate clockwise. N
Qm = 100 g/min.
300 rpm,
-75-
3.6.3.2 RESULTS
A large number of screw geometries was investigated using
the High Speed Camera. Same representative results will be
discussed here. For details, the reader is referred to the
report by J.H. Truijen (28).
Figure 3.20 compares the predicted and experimental filled
lengtbs in screws with different combinations of parts 'b'
and 'c'. The sections 'c' may consist either of kneading
elements with neutral or negative pitch, or of transporting
elements having negative pitch. It is important to notice
that all sections 'c' are of the same axial length, and that
the filled lengtbs are measured in down-channel direction,
indicated by the coordinate z. The coordinates z and ~are
related:
z = ~/sin~ (3.75)
In Figure 3.20, three levels of filled lengtbs can be
distinguished, each of which corresponds to the type of
section 'c'. Within each level, filled lengtbs are measured
at a different throughput. The general agreement between the
predicted values and those from the experiments is fairly
good. If the results are examined in more detail, however, a
few remarks have to be made.
At a screw speed below N = 200 rpm, the predicted and
experimental results start to deviate. This is observed for
all combinations and throughputs. It is mainly due to the
fact that gravity causes the fluid to prolapse. To our
regret, no accurate measurements of the filled length can be
expected for these screw speeds, when using a fluid of low
viscosity (~ = 0.2 Pa.s)*.
* A fluid of higher viscosity causes nonisothermal effects
and might exhibit elasticity, due to the presence of trapped
tiny air bubbles.
-76-
It is quite an interesting region of screw speeds, however,
where a large differentiation in filled length is predicted
by the model.
The geometrical description of kneading elements having
negative pitch is rather poor. Still, the assumption of a
rectangular channel with effective channel depth Heff, and a
relatively remaining flight length of about 25% yields
acceptable results.
800,---------~-------------------------------,
700 lc • 112.5 mm
neg. tranaporllng .. otlon
600
0
500
400 -E E -.a
300 ...1 .s::. -CD c Q)
200
100 neg. kneadlng eecllon
150 200 250 300 350 400 450
screw speed (rpm)
-77-
previous page:
Figure 3.20 Filled channel length Lb as a function of screw
speed for combinations of transport elements with various
parts 'c'. The axial lengthof parts'c', Lc = 112.5 mm in all
cases. Parameter: Om (0 100; <) 140; 0 170; e 240 g/min).
Drawn lines are values calculated from the equations of
Table 3.3. The symbols represent experimental results.
3.7 RESIDENCE TIME DISTRIBUTION
Apart from their local pressure generating or pressure
consuming capacity, the different screw geometries of the
corotating twin-screw extruder can be characterized by their
residence time distribution RTD (29,30).
At a given moment during the compounding process (in the
experiments, LDPE Stamylan 2300G, DSM was used), titanium
dioxide (Ti02 ) tracer particles are added as a pulse at the
entrance of the e xtruder. At the exit, samples are taken
from the strand at regularly spaeed intervals in time (~t
7.5 s). From the amount of tracer material that can be
detected in the sample via X-ray fluorescence (with an
accuracy of 1 ppm), the response is determined.
The results are presented in dimensionless farm. The value
of the cumulative response F at a time t represents the
fraction of the material with a residence time less than t.
It is clear that F ~ 1 as t approaches infinity. It is
camman toplot F(t) versus the dimensionless time t/t, where
t is the average residence time.
Two extreme screw geometries (see Figure 3.21) are compared
here. The one shown in Figure 3.21a is a conventional screw
with a moderate amount of kneading sections. Figure 3 .21b
shows a s c rew geometry which consists of a series of
alternating Couette elements and kneading elements. In its
most simple farm, the average residence time can be
estimated for bath screws by calculating
-78-
V 'II'DLH t
Q Q (3. 75)
with V the volume that is completely filled with melt,
i.e. corresponding more or less to the length of
the kneading section.
Q the volumetrie throughput
(a)
(b)
Figure 3.21 Screw geometries of a Werner & Pfleiderer
ZSK30 laboratory corotating twin-screw extruder
(L/D = 42D) . (a) Standard kneading screw;
(b) Couette-kneading screw.
The experiments were now performed on a Werner & Pfleiderer
ZSK30 laboratory corotating twin-screw extruder. Typical
data are: The length of the kneading sections (L z 900 mm
and 140 mm) . The channel depth H = 5 mm, and the diameter of
the screw D = 30 mm. The density of the LDPE z 800 kg.m-3.
Substituting these values in Eq. 3.75 yields forthescrew
geometries of Figures 3.21 a and b: tr z 50 s and 300 s,
respectively. In spite of this simplistic way of
calculating, these results correspond quite well to the t
from the RTD measurements, as is shown in Figure 3.22. In
this Figure, the Couette-kneading screw geometry has the
stronger tendency towards plug flow.
-79-
0 .8 i(/ I
- 0.6
l -Ir
0 .4
screw type (t) ' 0.2 /; ZSK 02, (306 s)
ZSKOB_,_ (~ __ ___ .
~lug flow
0 0 2 3 4
t/t
Figure 3.22 Residence time distribution in a ZSK30 corotating
twin-screw extruder. Screw speed = 250 rpm;
metered thoughput Q = 6 kg . hr-1 ; barrel temperature
Tw = 200 •c. Screw geometries as in Figure 3.19.
3.8 DISCUSSION
It has been shown that a simplified model for a corotating
twin-screw extruder is able to predict the correct energy
consumption, specific energy, and temperature rise, not only
over the extruder as a whole, but also locally during the
processing (depending on local screw geometry, processing
conditions, and material properties) . This is of great
practical importance in polymer processing because an
understanding of the process is within reach and gives a
perspective for solving problems in s c ale up. Despite o f
this, the mo re dire ct measuring o f transport c apac ity, loc al
pressure gradients and filled lengths with model liquids in
Plexiglas-walled extruders yields larger deviations, as
expected more or less. However, some of the differences
between calculated and measured values can be attributed t o
-80-
the difficulties in the experiments caused by the low
viscosity of the fluid used. Gravity-induced leakage flow
apparently can not always be neglected. Other deviations
originate from the geometrical approximation of the channel
depth. Small improvements of the model, such as a local wall
and screw temperature dependent leakage gap (important in
large extruders) can be brought in. An independent
determination of the (scale-up rules for the) heat transfer
coefficient is necessary, but most useful will be an
incorporation in the modelling of the pressure-generating
capacity of the feed section combined with a model of the
melting section. Only if these eonditions are fulfilled may
a complete model exist (which in its turn could be combined
with the existing theory of the dispersion process (see for
example, Ref. 31) in order to predict the compounding
performance for the blending operation.
3.9 REFERENCES
1.W. Szydlowski, R. Brzoskowski and J.L. White, Int. Pol.
Proc., 1., 207 (1987)
2. J.L. White and K. Min, Paperpresentedat the European
Symposium on Polymer Blends, Strasbourg, France (May
1987) .
3.M.L. Booy, Pol. Eng. Sci., .la, 973 (1978).
4. H. Werner, Ph.D Thesis, Munich University of Technology,
West Germany (1976).
5.M.L. Booy, Pol. Eng. Sci., ZQ, 1220 (1980)
6.Anonymous, Engineering (London), 114, 606 (1922).
7. H.S. Rowell and D. Finlayson, Engineering (London), llQ,
249,385 (1928).
8.B.H. Maddock, SPE ANTEC (1959),383
9. Z. Tadmor, Pol. Eng. Sci., .Q., 185, (1966)
10. J.F. Ingen Housz, H.E.H. Meijer, Pol. Eng. Sci., 21
(1981), 352
-81-
11. J.F. Ingen Housz, H.E.H Meijer, Pol. Eng. Sci., ~
(1981)' 1156
12. Z. Tadmor, P. Hold, L. Valsamis, SPE ANTEC (1979),193.
13. K. Nordmeier, Ph.D. Thesis, Aachen University of
Technology, West Germany (1986)
14.C.J. Rauwendaal, Polymer Extrusion, Hanser Publishers,
Munich (1986).
15. L.P.B.M. Jansen, Twin-screw extrusion, Elsevier, '
Amsterdam, (1978).
16. H. Gosler, (Werner & Pfleiderer), private communications.
17. J.F. Carley, R.S. Malleuk and J.M. McKelvey, Ind. Eng.
Chem., ü, 974 (1953)
18. J.F. Carley and R.A. Strub, Ind. Eng. Chem., ~, 970
(1953)
19. R.S. Malleuk and J.M. McKelvey, Ind. Eng. Chem., ~, 987
(1953)
20. C.D. Denson and B.K. Hwang, Pol. Eng. Sci., ZQ, 965 (1980)
21. C.P.J.M. Verbraak, H.H.M. Lardino ye and H.E.H. Meijer,
Polym. Eng. Sci. (in progress)
22. J.R.A. Pearson and S.M. Richardson (ed.), Computational
Analysis of Polymer Processing, Appl. Sci. Londen (1983)
23. H.E.H. Meijer, P.H.M. Elemans, H.H.M. Lardinoye and G.
Kremer in: 'Warmeobertragung bei der
Kunststoffverarbeitung', VDI Dosseldorf (1986)
24. H.H.M. Lardinoye and G. Kremer, internal report DSM (1988)
25. C.H. Jepson, Ind. Eng. Chem.,~, 992 (1953)
26. H. Janeschitz-Kriegl and J. Schijf, Plastics & Polymers,
Dec., 523 (1969)
27. M. L. Booy, Polym . Eng. Sci.,.l.5_, 606, (1975)
28. J.H. Truijen, internal report Eindhoven University of
Technology (1989)
29.P.V. Danckwerts, Chem. Eng. Sci., ~' 1 (1953), ibid. a, 93 (1958)
30. G. Pinto and Z. Tadmor, Polym. Eng. Sci., ~, 279 (1970)
3i. J.J. Elmendorp, Ph.D. Thesis, Delft University of
Technology (1986)
-82-
CHAPTER 4
MODELLING OF THE CO-KNEADER
The Buss Co-kneader is a single screw extruder with
interrupted flights. Pins from the barrel are inserted into
the screw channel. The screw is both rotating and
oscillating. Due to this action screw flights are
continuously wiped by the pins. During one passage of the
pin, the material is not only subjected to high shear
stress, but it is reoriented as well, thus promoting the
distributive mixing process by the local weaving action of
the pins and screw flights. Attempts to model the Co-kneader
tend to focus on a single pin passing through the hole in a
screw flight (1,2). However, a more overall model can start
with the same equations that apply to the corotating
twin-screw extruder (3). Because now not only the effect of
leakage flows on the local pressure gradient has to be
considered but also that due to the local dragging action of
the pins (neglecting the oscillatory action), experiments
with model liquids have been performed to evaluate the
overall model. Additional experiments with a
Plexiglas-walled Co-kneader support the calculations
concerning filled lengths in various screw geometries. These
results, and those of model calculations, which are extended
to the nonisothermal, non-Newtonian situation, will be
presented.
Reprinted partly from: P.H.M. Elemans and H.E.H. Meijer, Polym. Eng. Sci.
(submitted) by permission of the Society of Plastics Engineers.
-83-
4.1. INTRODUCTION
The Buss Co-kneader is a continuous mixer which is widely
used in many operations in the processing of polymers and
foodstock. It offers possibilities for homogenization,
colouring, reinforcing, and (highly) filling of critical and
difficult-to-handle compounds of thermoplastics, rubbers or
thermosets (4-7). Provided the roetering of the individual
components is correct, torque and temperature rise can well
be kept under control, even for thermally instable materials
and/or critical additives partly because both screw and
barrel are oil thermostated.
Invented in 1945 (8), practical application of the
Co-kneader has been far ahead of theoretical understanding.
According to the design specifications, the Co-kneader is
said to operate fully adiabatic and scale-up should be no
problem since every next size in the series has ten fold the
throughput of the previous size (PR46, PR100, PR200).
Although the more modern machines have not altered their
main werking principle, their length has been increased from
7D to 11D (for special purposes even 15D can be chosen) and
screw- and barrel design is flexible now, also on the
laboratory kneader. This flexibility now requires for the
development of roodels that can give predictive answers with
regard to energy, specific energy, temperature rise and
residence time, dependent on metered throughput, chosen
screw design and eperating conditions.
Recently, one of the first more serieus studies concerning
the rnadelling of flow in pin-barrel (rubber) extruders
appeared (9). Like Ref. 1, it is confined to calculations of
one type of screw element of invariable geometry only (no
sequence of different screw elements as used in the
Co-kneader) . The flow is said to be Newtonian and
isothermal. However, it can be expected that processing of
highly viscous model fluids will give rise to a considerable
increase in melt temperature. Moreover, the mixing of pin
-84-
screw extruders used in rubber processing is much less
effective than that of the weaving action of the pins and
remaining screw flights in the Co-kneader due to its
rotating and oscillating movements (see Figure 4.3).
Besides, apart from improved mixing performance, see Ref.
10, one of the most important effects of the pins in rubber
extruders could be the impravement of the transport
characteristics by breaking through slip layers in the
beginning of the process where an extremely high viscous,
plastic material is present. Kneading is normally promoted
in these types of extruders by the choice of the screw:
multichannel, multicompression and decompression with many
local barrier flights.
Still, it would be interesting to cernpare the results of
Ref. 7 with experiments in those sections of rubber
extruders where only melt is present and the roodels apply.
Here it is tried to model the Buss Co-kneader in a way
analogous to that of the corotating twin-screw extruder (3).
First we will only deal with the simplest case of a
Newtonian liquid under isothermal conditions. This analysis
can be extended, as in Ref. 3, to the non-Newtonian,
nonisothermal situation of a melt-fed Co-kneader.
4.2. SCREW GEOMETRY AND WORKING PRINCIPLE
The screw of a Co-kneader consists of interchangeable
elements. Figure 4.1 shows three common screw elements:
conveying elements (single flighted), having two rows of two
pins per lead inserted in the barrel, closed channel
kneading elements, double flighted, with three rows of one
pin per lead, and kneading elements, also double flighted
and having three rows of two pins per lead.
In contrast with the corotating twin-screw extruder, no real
negative transporting elements exist, except for the element
shown in Figure 4.6e, which has a locally negative pitch
angle. Staggering elements, however, are used. The most
-85-
common ones are buffle rings, used preceding degassing
sections in the barrel, and the die that connects the
kneader to the discharge extruder if present. See Figures.
4.2a and 4.2b.
An alternative setup, which replaces the discharge extruder,
to assure regular material outflow, as proposed in Ref. 11,
is discussed in the Section 'Experimental'.
Figure 4.1. Common screw e1ements a) conveying (2 rows of 2 pins);
b) closed channel kneading (3 rows of 1 pin);
c) kneading (3 rows of 2 pins) .
4.2a 4.2b
Figure 4.2. Staggering e1ements (schematica11y). After Ref.3.
a) buff1e ring; b) die connecting kneader and discharge
extruder.
-86-
Figure 4.3 illustrates the combined (self-wiping) action of
the interrupted flights of a kneading element with the
corresponding kneading bolt in the barrel wall (the kneading
bolts are usually diamond-shaped, except in the PR46, which
has cylindrical bolts) . It is due to this weaving pattern
that the Co-kneader possesses an excellent distributive
mixing quality. The unrolled "standard screw" geometry
(Figure 4.4) clearly shows the variety in mixing intensity
along the screw.
!A A•B
n]c c~~ ~c c
~ r ~ ~ ~J 0 18 0 A
~u ~. ~~ .~ ~ A
JU ~. ~ ~ ~ c I ~ ~ ~
~ I A AB
Figure 4.3. Trajectories of (stationary) pins relative to the shaded
flight (Figs. 4.3(a) - 4.3(d)). Fig. 4.3(e) shows the weaving
pattern as a result from the motion of the pins.
0 0 0 L(l
~ ~ ~ ~ 0 0 0 L(l . 0 0
....: r.: ....: r--: • N N N N Ltl
N N N UJ UJ UJ
0 0 0
r-: ....: ....: N N N
0
<Ti IC
Figure 4.4 . Unrolled "standard screw" geometry (D = 46 mm, L/ D = 11),
showing the sinuscictal trajectories of the pins relative to
the screw. The circles indicate the feed ports.
Explanation of screw elements' abbreviations: EZ: conveying ;
KE: kneading; GS: closed channel kneading; OS: buffle ring.
-87-
4.3. SUMMARY OF THE NEWTONIAN, ISOTHERMAL ANALYSIS
Like in the case of a corotating twin-screw extruder, the
analysis of the Co-kneader will depart from the combination
of a screw element (conveying or kneading) with a staggering
element.
The screw elements of the Co-kneader all have interrupted
flights, leaving room for the pins to pass during each
revolution. Thus, two main influences on the usual
drag-and-pressure flow can be distinguished: (i) the extra
conveying action due to the pins. And (ii): The leakage flow
through the flights' holes, back into the previous channel.
The remaining screw flight is considered to be a factor 'r'
of the total length of the screw flight (Lf = r • ~D/cos ~).
The sum of the holes' length equals the factor (1-r). The
pins are thought to cause a plug flow right in front of
them. This is expressed by the factor a: The aspect ratio of
the pins' surface over the channel surface, which equals
(dpin)2/(L•W/n). We will investigate these two influences
separately in order to understand the transport
characteristics of the kneader.
Although the kneader is rotating as well as asciilating in
axial direction, we will neglect the influence of the latter
in the analysis of the transport characteristics. The
equations for Newtonian, isothermal pressure and drag flow
in an unrolled screw channel will be applied, with the usual
assumptions, including the negleetien of pressure flow over
the flight clearance (3). Table 4.1 gives the resulting
equations that apply to the Co-kneader. Because of the large
number of relatively big holes in the interrupted screw
flights of the Co-kneader, the flow rates due to local
pressure gradients are important nat only in down-channel
direction, but also in axial direction. In Figure 4.5, the
influence of a hole on these pressure gradients is
visualized in the unrolled screw. The pressure difference ~P
between the two points A and B in the screw channel can be
-88-
generated in two ways. Either by going a distance z in
positive z-direction: AP = (dp/dz)•z, or by going a distance
x in negative x direction: ~ = (dp/dx)•x. In each direction,
the pressure flow equals the pressure gradient times a
geometrical resistance factor k for that particular
direction, see Eq. 5 of Table 4.1. The coordinates x and z
are related to the axial coordinate, ~:
z = ~/sinq>
x= -~/cosq>
(4 .1)
(4 .2)
Figure 4.5. Pressure generation in down-channel and in cross-channel
direct ion.
-89-
The total pressure flow equals:
1 dp Qp =- 12}.1 H31fD FPC(sin<p Fr/p dz +
(1-r) ---F'
COS(j) P ( 4. 3)
Combining Eq. 4.3 with the Eqs. 4.1 and 4.2 , the pressure
flow is given by Eqs. 5 and 11 o f Table 4.1.
Table 4.1. Expressions for drag and pressure flow in screw elements,
based on the total metered throughput, Q.
Qc hannel
Qchannel
(~in> 2 I (L W/n)
with
pressure flow
drag flow
leakage flow
over flight
leakage flow
through flight
c = efficiency
pararoeter
extra drag flow
due to pins n V H ~in (if f = l)
V 'TfDN Lf
r = ---:-"'---'TfD / costf
z = _2._ sin<p
F0 , Fp, F0 c, Fpc (see Re fs. 12,13 , and 14)
1-ne
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(9)
(10)
(11)
-90-
In Table 4.1, Foc and Fpc are correction factors that
account for the influence of curvature on the flow in the
kneaders' screw channels. The kneader has an H/D of about
0.15, which is rather large compared to most corotating
twin-screw extruders, where the H/D < 0.1.
It is important to know what value of r, the relative
remaining flight length, should be used in the calculations.
In principle, there are two ways to determine r, as is
demonstrated in Figures 4.6a and 4.6b. The results are shown
in Table 4.2. The fact that two sets of 'r' values can be
obtained wil! turn out to be important in the later
evaluation of the experiments.
(a) (b) (c) (d)
Fig. 4.6. Detail of unrolled screw channel.
a) conveying element, Lh/Lf = (1-r)
b) conveying elemerit, Lh/Lf = (1-r)
c) closed channel kneading element
d) kneading element
e) as Fig. 4.6c, with locally negative pitch angle
(e)
-91-
Tab1e 4.2. Geometrica1 estimation of r, the re1ative
remaining f1ight 1ength and ~ (aspect ratio
(dpin)2/(L W/n)) for severa1 screw e1ements.
r ~
according to
Figure
4.6a 4.6b
conveying 0 . 92 0.8 0 . 005
e.c. kneading 0.78 0.55 0.011
kneading 0. 71 0 . 4 0.011
4.4. MIXING
The mixing qualities of the Co-kneader, which have been
somewhat in the background during the analysis, will now be
discussed. The number of reorientations can be calculated by
starting with the analysis by Booy et al. (1), who consicter
the elementary passage of a single pin through the opening
between two adjacent flights of a kneading element. Their
assumptions are: (i) the element is completely filled with
melt, and (ii) after the passage of the pin, the material is
- at least locally - reoriented, whi c h can be expected from
the simulations in Ref. 15.
The average residence time in a screw element with length L
approximately amounts to:
t
where
11'DLH(1 - ne)
Q
D diameter
H c hannel depth
n = the number o f f lights
e relative flight width
Q throughput
( 4. 4)
-92-
The average residence time required for the passage of a pin
between two flights (the length of one flight is about 1/3
times the circumference) is given by:
1/3'11'0 2
1/2'1l'DN 3N
h d (s -1) w ere N = screw spee
The number of reorientations, nr, is given by:
nr N
3/2'1l'DLH ( 1-ne) Q
(4. 5)
(4. 6)
This analysis is valid if only one pin is present in the
screw channel. However, in a real Co-kneader screw element,
a total amount of p pins can be recognized. For each
element, p can be obtained by multiplying the number of pins
per lead of the screw by the number of pin rows on the
barrel circumference. (See Sectien 4.2). Thus, for closed
channel kneading elements, p = 3, whereas for kneading
elements, p = 6.
For the number of reorientations in a completely filled
screw element, this yields the final expression (with L
'll'Dtan~, the screw lead) :
nr 2 2 p.3/2'1!' D Htan~ (1-ne) N
Q (4. 7)
In the case of partially filled screw elements, the number
of reorientations is evaluated with respect t o the filled
length Lf.
nr N
p.3/2'1l'D L~tan~ (1-ne) Q ( 4. 8)
-93-
Example: consicter the completely filled kneading element
(L = 27 mm) of a Buss PR46 laboratory Co-kneader, with
typical data:
~ 10.5°
H 7.10-3 m
D 46.10-3 m
n 2
e = 0.133
N 240 rpm = 4 s-1
Q 20 kg.hr- 1 z 5.5.10-6 m3.s-1
With these data, the number of reorientations, calculated
from Eq. 4.8, becomes nr = 128. The amount of shear between
two reorientations, Yn• is estimated to be y = y.t = 2/3 ~D/H ~ 14.
4.5. EXPERIMENTAL
The goal of the experiments was threefold: (i) to provide a
check for the geometrically estimated values of the relative
remaining flight length; (ii) to determine the influence of
the pins; (iii) to check the complete Newtonian, isothermal
analysis of the kneader. To achieve this, extrusion
experiments were performed using silicon oil (~23oc
1 Pa.s) and paraffinic oil (~23 oc = 0.2 Pa.s), under
isothermal conditions. Throughput versus screw speed,
throughput versus pressure generation as well as locally
filled lengtbs were measured, using severa1 screw types
(conveying, closed channel kneading and kneading), different
number of pins mounted in the barrel wall (from 0 to max.)
and processing conditions (screw speed, throughput) .
Finally, attempts were made to measure the pressure gradient
directly, i.e. from the oil level in a series of open tubes
along the cylinder. All experiments were performed on a
PR 46-llD laboratory kneader, which could be fitted with a
Plexiglas cylinder,
-94-
~igure 4.7. The Co-kneader fitted with a Plexiglas cylinder. Via a system
of small glass tubes that act as open manometers, the
pressure gradient along the cylinder can be visualized.
see Figure 4.7. A problem arises when a constant kneader
outflow is required, for instanee for measuring the pumping
characteristics (Q - ~p) of the Co-kneader. The oscillatory
motion of the kneader screw relatively to the barrel causes
severe pressure fluctuations. Usually a discharge extruder
is needed to guarantee constant pressure at the die. This
can be avoided by using a special die (11). The cylinder is
fitted with a flange having the same boring as the kneader.
A ring fitting closely inside this flange is mounted to the
kneader shaft. The die opening now yields a constant
pressure and consequently gives smooth filaments of constant
diameter which can be caoled and cut. For measuring the
pumping characteristic of the kneader the resistance of the
die is made adjustable and the pressure is measured using a
manometer mounted at the end of the screw. See Figure 4.8.
The pumping characteristics of the screws as well as the
filled lengths were measured using this special die.
-95-
conneet ion to manome ter
conneetion to adjustab~e 'die
Figure 4.8. Setup used to equalize pressure fluctuations due to the
oscillatory motion of the screw.
4.5.1. THROUGHPUT VERSUS PRESSURE CHARACTERISTIC
Figures 4.9a through 4.9c show the throughput versus
pressure characteristic for various screw elements. Note the
large difference between conveying and kneading elements.
The intersections with the vertical axis represent the net
maximum outflow of the (completely filled) kneader. In this
case, no die was present. This was done to prevent pressure
generation due to the oscillation of the screw. Because
relatively low viscous fluids were used, the experiments had
to be corrected for gravity-induced flow.
The intersections with the horizontal axis yield the maximum
pressure generating capacity of the screws. Here, the axial
oscillatory motion was compensated for with the special die
mentioned in the previous Section.
For conveying elements - single-flighted, i.e. having a much
wider channel compared with the other, double-flighted, ones
- no distinct influence of pins on either drag or pressure
flow was observed.
-96-
1500·-,---------------------,
N • 180 rpm ~
è e ..... 1000-~
'S a_
"§, ~conveylng ~ 500~ -~. :5 '' ' k d ' ""' """'- ~.c. neo 109 · ""'
"""-~ .
f~kneoding """- , ~ ' 0~~~-,_--~_,~---,----,---_,
è e
0.00 0 .05 0.10 0.15 0.20 0.25
Ap (bar)
1500,---------------------,
N • 200 rpm
..... 1000
""-~. ~
··~.
""".conveying
""" """ """ """ -~ ~(jl
0.00 0.05 0. 10 0.15 0.20 0.25
i\p (bar)
Figure 4.9. Throughput versus pressure characteristics for conveying,
closed channel kneading and kneading screws in a 46 rnm
Co-kneader. Model liquid s: silicon oil Rhodorsil V1000
1 Pa.s) and paraffinic oil SUNPAR 2280
1 Pa.s). Parameters: Screw speed N (rpm) and
number of pin rows.
Open resp. closed symbols: 0 • conveying; 0 e c. c. kneading ;
a•kneading screw with resp . without pins.
-97-
With the equations of Table 4.1 and the data presented in
Figure 4.9, the two important parameters r (relative
remaining flight length) and c (efficiency parameter) can be
evaluated.
Overall equation: Q + QL + QLD (1-a.) •Qct + c•Q · + Q p1n p ( 4. 9)
no pins: a. 0, c 0
( 4 .10)
The vertical intersection is calculated from the condition:
Q = 0 p
Q
Qd (1 - r•ó/H - (1-r))
The horizontal intersection is calculated from the
condition: Q = 0
-Q p Qd r• (1 - &/H)
r J.l Qd (1 - ó/H)
k
with k a function of r: k
see last Eq. of Table 4.1.
C •r -1 ) '
(4 .11)
(4 .12)
(4 .13)
(4 .14)
-98-
with pins: a. ~ 0, c ~ 0
The vertical intersectien (Qp 0) is given by
Q r•Qd(1 - ó/H) - a.•Qd + c•Qpin ----
(4 . 15)
(4 .16)
The horizontal intersectien (Q
4.14:
0) yields, analogous to Eq.
Ap (4 .17)
------------Apla.=O
Table 4.3 shows the values of rand c that follow frorn the
respective horizontal and vertical intersections.
Table 4.3. Values for r (relative remaining flight length) and c
(efficiency of pins) as determined from horizontal and
vertical intersections in Figures 4.9(a) - 4.9(c),
r c
intersectien (hor.) (vert.) (hor . ) (vert.)
conveying 0 . 93 0 . 75 0.01 0.01
closed channel
kneading 0.90 0 . 6 0.04 0.15
kneading 0.52 0.18 0.09 0.21
-99-
In the case of conveying and closed channel kneading
elements, the values of the experimentally determined
relative remaining flight length (using Eqs 4.9 till 4.17)
as shown in Table 4.3 correspond reasonably well to those
that can be determined from geometrical considerations. The
results from the horizontal intersections correspond to the
values in the first column of Table 4.2, whereas the results
from the vertical intersections correspond well to the
values in the second column of Table 4.2. This can be
explained by realizing that the pressure gradients in the
Co-kneader are almost exclusively in axial, rather than in
down-channel direction, because of the large number of
holes. The values of the relative remaining flight length,
calculated from the horizontal intersection will be close to
the ones determined according to Figure 4.6a. In the absence
of pressure flow, the value of the relative remaining flight
length is determined from drag flow, which is assumed to be
present along the length of the flight in down-channel
direction. Consequently, the values of r calculated from the
vertical intersectien correspond to the ones determined
according to Figure 4.6b. In the case of kneading elements,
however, the holes' length is so large in comparison with
the remaining flight length that the model equations hardly
apply.
With regard to the evaluation of the influence of pins as
expressed in the efficiency parameter c we can conclude, as
could be expected, that no impravement in transport
properties in single-flighted elements is found. There is,
however, an effect in the double-flighted elements
(efficiency in the order of 10-20%) . The experimental values
of c also differ when determined from either vertical or
horizontal intersectien of the throughput-pressure
characteristic. The naive statement of full effectivity of
the transport capacity of pins expressed in the
straightforward expression used: Qpin = n V H dpin is
apparently an overestimation.This would be expected not only
from reasans concerning the bow-wave type of flow in front
-100-
of a pin, but also because of the consideration that this
flow is mainly in tangential direction (straight through the
holes in the flights), thus promoting leakage as well. The
pins might seem less important, yet it must be kept in mind
that pins are important for solicts transport and that the
presence of the pins is essential for the mixing principle
of the kneader.
4.5.2. FILLED LENGTH
The second way to evaluate the parameters r and c can be
realized with the setup shown in Figure 4.7. By
photographing the length of the channel that is completely
filled, the combination of a basic screw element with a
staggering element can be closely examined, see Figure 4.10.
Figure 4.10. Photograph showing the combination of a closed channel
kneading screw element, having 3 rows of pins. with a
Couette element (gap height 0.8 mm).
Figures 4.lla through 4.1ld show the comparison of
experiments and calculations for the various processing
conditions. In Figure 4.lla, the filled length is calculated
for conveying screw elements without pins, using a value for
-101-
the relative remaining flight length r = 0.93, obtained from
the horizontal intersectien (see Table 4.3). The calculated
filled length is quite sensitive to the value of r. Figure
4.11b shows calculated and measured filled lengths for
conveying screw elements with pins. As could be expected
from Figure 4.9, no distinct influence of the pins on the
filled length was observed.
For closed channel kneading elements without any pins, the
value of r = 0.9 (see the first column of Tab1e 4.3) is
correct to predict the filled length, as shown in Figure
4.11c. If pins are mounted in closed channel kneading
elements, a significant decrease in filled length is
observed, see Figure 4.11d. This effect is, however, not
satisfactorily described by carrying out the calculations
with an efficiency parameter c = 0.15. Because the exact
value of the filled length in closed channel kneading
elements is difficult to discern from the photographs, the
value of the parameter c will not be discussed further.
400
Ê 350 .§. .a 300 ..J .s::. Öl 250 c:
~ ~ 200
~ x 150
··~ ~ '0 100 Q) u---:--------=: ;;::
50 0 0
0 :sa 200 250 300
screw speed (rpm)
Figure 4.11 a) Cornpariaon of calculated (drawn linea) and rneaaured
fi11ed length Lb for a combination of conveying acrew
elernenta (no pina) with a Couette element. (r = 0.93;
a,c = 0 uaed in the calculations). Model fluid: paraffinic
0.2 Pa.s). Q (g/min): D 75.0; • 100; 0 130;
400
Ê 350 .§. .0 300 ..J
.c Ö> 250 <:: ~
2CO (ij )( ~
150
"0 100
ê 50
150
450
Ê 400 .§. .0 350 ..J
.c 300 C) <:: 250 ~
(ij 200 )( ~ 150
"0 Cl) 100
- 50
0 ' 150
<50
Ê 4CO .§. .0 350 ..J
.c 300 C) c: 250 ~
(ij 200 )(
~ 150
"0 Cl) 100
50
0 . 150
-102-
-·.....__
200 250
screw speed (rpm)
0
0
•
~ 200 250
screw speed (rpm)
~~ • -------- 0 ~ . ~
• • :: 0 0 0
200 250
screw speed (rpm)
b) 300
c) 300
d ) 300
Figure 4.11 b) As Fig ure 4.11a, with pins (r = 0.93; ~ = 0 .005, & = 0.01) .
Q (g /mi n): 0 BO; • 100; 0 140.
c) As Figure 4.1la, for closed channel kneading element s (no
pins, r ~ 0 . 9 ; ~,& = 0). Q (g/ rnin): 0 75; • 100; 0 120.
-103-
(continued from previous page)
d) As Figure 4.11a, for closed channel kneading elements
(three rows of pins, r = 0.9; « = 0.011, e = 0.15).
4.5.3. PRESSURE GRADIENTS
The third and most accurate way to evaluate r is by visually
investigating the axia1 pressure gradients with the setup
shown in Figures 4.7 and 4.12. From the data thus obtained,
and with the equations of Table 4.1, the relative remaining
screw flight length r can be calculated. The results are
given in Table 4.4, for different screw geometries and
processing conditions. (No pressure gradients were
determined for kneading elements, because of the low
pressure generating capacity of these elements)
Table 4.4. Experimental pressure gradients and relative remaining screw
flight length r in conveying elements (2 rows of pins) and
closed channel kneading elements *) (3 rows of pins) .
Q N
(g/min) (rpm)
conveying 170 150
170 200
200 150
e.c. 170 150
kneading 170 200
170 250
270 150
270 200
270 250
~ L lexp.
(Pa/m)
5390
11590
5120
2680
4200
5340
1700
3020
4390
r
0.88
0.92
0.88
0.82
0.83
0.83
0.77
0.80
0.81
*) Efficiency parameter e 0.15 used in the calculations.
-104-
Figure 4 . 12 Visualization of the pressure gradient in a closed channel
kneading element. N = 250 rpm; Q = 270 g ./min.
To conclude the evaluation of rand c, Table 4.5 gives a
listing of the values to be used in future calculations.
Table 4.5. Concluding values for r (relative
remaining flight length) and c
(efficiency of pins) for several
screw elements.
conveying
e.c. kneading
kneading
r c
0.9
0.8
0.2
0.01
0.15
0.21
With the model derived so far, it is now possible to
calculate a sequence of screw elements with different
combinations of screw speeds and throughputs. Figure 4.13
gives some examples of calculated pressure profiles in
common screw configurations, for various processing
conditions.
I L{)
0 .-l I
.--------------------,~,_g ~
0 0 U")
0 0 '<t
Ê
I r~~ I I
0 0 N
0 0
f---r---..---.,---...---...---+o 0 ",
U") N
0 U")
N - 9 U")
(.mq) a;nssa;d <---
0
l!s::: 3> 4 "L "
5 ;sr- ~-~-
r--------------------,~~g ~
0 0 U")
0 0 '<t
.......-... :-. -----E -::::::;::: EZ 2~----
gE ~ EZ27 ",'-"'
0 0 N
0 0
_l ::::::::;::. EZ 27.1l
R~ -·--··--
l!c:~ dtL '----··d-
l!c:?> d"L "--·--
r-~--r---r-....--;------1- 0 0 U") l"l N
0 U") 0 N - -
U")
(;oq) aJnssa;d <---
0
-106-
30 'L'
0 25 ...0
'--"' (!)
20 L :J , VJ (/) 15 (!) L
A 0...
10 1\ I 5 I I
0 0 100 200 300 400 500 600
L (mm)
", ", ", ", ", ", "' "' ", ", N N N N N N N N N N
0> ,N IN N UI N N N N N N N N N N N !" I ~
,;-> 1:-> ~ .... ;:-J I:-' :-> :-> :-> :-> :-> .... ,..., ,..., !m
,. i;.. i ~ ,;.. 151 j!S 151 i "' • 151 I"' 151 , ... I"' i"' !"' I I I ! I I I I
Figure 4.13. a) Calculated pressure profile and filled lengths in a
Co-kneader (D = 46 mm), ha ving standard s c rew geometry.
Model fluid: Linear low density polyethylene.
Newtonian, isothermal model . Parameter : Screw speed
b) As Figure 4.13a, with kneading screw.
c) As Figure 4.13a, with conveying screw.
4.6. NONISOTHERMAL, NON-NEWTONIAN ANALYSIS
In analogy with the analysis in Chapter 3, the NeWtonian
isothermal analysis of the Co-kneader can be extended to the
nonisothermal, non-Newtonian case. By taking incremental
steps over the length of the kneader, more quantitative
answers concerning specific energy and temperature rise can
be given. Over each incremental step a heat balance is
solved: the heat stared in the local volume (resulting in a
temperature rise) equals the dissipative heat generated by
the velocity gradients plus the heat exchanged with the
walls (positive or negative). Figure 4.14a shows the
pressure profile s and fill e d lengths for constant thr oughput
with varying screw speed. Figure 4.14b shows the temperature
profile for the same throughput, with as parameter the
initia! melt temperature.
'L' 0
..0 ....._"
Q) I... ::::s (/) (/) Q) I... 0...
!\ I I I
,.-.... u
....._"
Q) I... ::::s
-+-0 I... Q)
0...
E Q)
+-
1\ I I I
15
12
9
6
3
0 0 100 200
300
250
200
150
-107-
Q;::; 25 kg/hr
Tin= 200 c
screw speed N = 200 rpm
;---. N = 220. a!!!!__ . 1/ _____ N =-~~.l!.!J!!!' ___
1/ I
.1? /
300 400 500
Q;::; 25 kg/hr
N;::; 240 rpm
initia! tempareture TIn= 200 C
T In = 2,:Z.Q_f_. __ ..
!...!.!)_;: 24Q__Ç ________ _
600
100;--------r-------.------~--------.-------.-------~ 0 100 200 300 400 500 600
L (mm)
Figure 4 . 14 a) Calculated pressure profile and filled lengths in a
Co-kneader (D = 46 mm). Model fluid: Linear low density
polyethylene , ~ol2oo oc = 5000 Pa.s; b = 0 . 016; n = 0.43.
Twall = 200 oe .Parameter: Screw speed.
b) Calculated end temperatures in a Co-kneader (D = 46 mm) .
Model fluid: as in Figure 4.14a, parameter: Initial melt
temperature .
-108-
4.7. RESIDENCE TIME DISTRIBUTION
Analogous to the procedure of Section 3.7, the Residence
Time Distribution of the Buss Co-kneader can be
determined*. Some representative results will be discussed
here. Details can be found in the work of A. Engbersen (14)
0 .8
0 .6 --i:L
0.4
throughput, (t) 0 .2 12.5 kgL'hr, (90.2)
25kgf~G (47.11 _____
0 0 2 3 4
tlt
Figure 4.15 Residence time distribution in a PR-46 Buss
Co-kneader. Screw speed: N = 200 rpm, barrel
temperature Tw ~ 150 °C . Kneading screw geometry
(see also Figure 4 . 13). Parameter: metered throughput Q.
* In the experiments, poly(vinylchloride) (PVC) was used.
0.8
0.6 --i:L
0.4
0.2
0 0
//
t/t
-109-
2
adhering PVC, Ï=82.7
non-qd_h_ering P_Y_CJ Ï=76.!!
3 4
Figure 4.16 As Figure 4.15. Processing conditions: Q = 12.5 kg/hr,
N = 200 rpm, Tw = 150 oe. Standard screw geometry.
Material : PVC with internal and external lubrificants,
referred to as 'adhering' and 'non-adhering' (DSM
VARLAN S 5020 and S 7120, respectively).
~--"'· · · · "'''''''' r~--
.. p·
0.8
0.6 --- (t) i:L screw type standerd (43.2)
0.4 ~n~yif}SI {2_2.§},
buffle ring_(~
0.2 ~<;,__~~~1!lLi~~-fJ(]r_all<!l_pl(]le_ITl(],j<!l ___
0 0 2 3 4
t/t
Figure 4.17 As Figure 4.15 . Processing conditions Q = 37.5 kg/hr,
N = 200 rpm, Tw = 150 oe. Parameter: screw geometry .
The dotted line represents F(t) fora single screw extruder,
as predicted by Pinto et al. (16) .
-110-
0.8
0.6 --Li:'
0.4 screw type (t) slandard (82.7)
kneadi'!9. (104.()). __
0.2 buffle ring_(~
s.c_,_~"._9~'!9.J~JL
0 0 2 3
tlt
Figure 4.18 As Figure 4.17. Processing conditions: Q
N = 100 rpm, Tw - 150 °C.
4
12.5 kg/hr,
In Figure 4.15, the influence of the throughput is
demonstrated. The shape of the F(t) curve remains almost the
same, while t is being reduced by a factor 2 *.
Figure 4.16 shows a tendency towards plug flow (i.e. a more
narrow distribution) for the non-adhering PVC grade.
Figure 4.17 shows an overall comparison between the most
common screw geometries.
* By assuming the kneading elements plus buffle ring elements
to be comp1etely fi11ed with melt, t can be estimated to be
(see Sectien 3.7): 80 s and 40 s when Q = 12.5 kg.hr-1 and
25 kg.hr-1 , respectively.
-111-
4.8. DISCUSSION
It is shown that an analysis similar to that for the
corotating twin-screw extruder can be used to model the flow
in a Buss Co-kneader. Measuring the throughput versus
pressure characteristic, the filled lengths and, most
accurately, directly the pressure gradients are effective
ways to check (and estimate) important geometrical
parameters such as the effective relative remaining flight
length and the pins' influence on drag and pressure flow.
Extension of the model to nonisothermal, power law
calculations is possible. It will be important to model in
yet more detail the dissipation of heat in the narrow gaps
between pins and screw flights as well as the increased heat
exchange between the polymer and the kneader wall. At a
local scale, a finite-element-method calculation might be a
helpful tool in understanding these phenomena. Further
experiments, such as determination of critical screw speed
and temperature profile in the kneader, are needed to
evaluate the extended model.
4.9. REFERENCES
1. M.L. Booy and F.Y. Kafka, Soc. Plast. Eng. (Techn.
Papers),~' 87 (1987)
2. J. Jeisy, Y. Trouilhet and P. Grassmann,
Verfahrenstechnik, lQLZ, 79 (1976)
3. H.E.H. Meijer and P.H.M. Elemans, Polym. Eng. Sci.,
~,275-289 (1988)
4. P. Schnottale, Kautschuk und Gummi Kunststoffe, ~' 2/85,
116-121 (1985)
5. S. Jakopin, Adv. Polym. Technology, vol.~ (4), 365-381
6. S. Jakopin and P. Franz, Paperpresentedat the AIChE
Diamond Jubilee, Washington O.C., November 3 (1983)
7. O.B. Todd and J.W. Hunt, Soc. Plast Eng. (Techn. Papers),
U, 577 (1973)
-112-
8. German patent 944727 (H. List, 1945)
9. R. Brzoskowski, J.L. White, W. Szydlowski, N. Nakajima
and K. Min, Int. Polym. Proc. III (1988) 3, 134-140.
10. E.G Harms, Dissertation RWTH Aachen, West Germany (1981)
11.US patent 3642406 (H.F. Irving, 1971)
12. C. Rauwendaal, "Polymer Extrusion", Hanser Publishers,
Munich (1986)
13. J.M. McKe1vey, "Polymer Processing", John Wiley & Sons,
New York (1962)
14. z. Tactmor and I. Klein, "Engineering Princip1es of
Plasticating Extrusion", Van Nostrand Reinhold, New York
(1971)
15.L. Erwin and F. Mokhtarian, Polym. Eng. Sci., ~, 49
(1983)
16.A. Engbersen (DSM), private communications (1986)
-113-
CHAPTER 5
SCALING
The procedure adopted in the Chapters 3 and 4 reveals that
it is well worth examining the sealing up from a small
laboratory extruder to a larger one. The scale-up strategies
that are usually developed involve all aspects of the
extrusion process: solicts conveying, melting and metering. A
review of existing scale-up theories can be found in Ref. 1.
The analysis in this chapter will mainly focus on partially
filled continuous mixers.
The parameters to be analyzed are: Torque, power
consumption, throughput, specific energy consumption (the
power consumption divided by the throughput), mean residence
time, shearrate, total shear and specific surface (the -
heat exchanging - barrel surface divided by the throughput) .
5.1. DEFINITION OF SCALING POWERS
In order to evaluate the different scale-up procedures, all
variables are expressed as a power of the diameter ratio.
This results in the definitions of Table 5.1.
Reprinted partly from: H.E.H. Meijer and P.H.M. Elemans,
Polym. Eng. Sci., 28, 275 (1988) by permission of the Society
of Plastics Engineers.
-114-
Table 5.1. Definition of sealing powers. The index , o'
refers to the reference machine.
Screw No D0 n Specific ~= D es =(-) (-)
speed No Do energy ESP·o Do
Channel Ho D0 n Residence t D t =(-) (-)
depth Ho Do time to Do
Screw Lo Do Q. Throughput Q D q
=(-) (-) length Lo Do Qo Do
Power Po Do p Shearrate y D g =(-) (-)
Po Do Yo Do
Torque To D0 m Tot al y D gs =(-) (-)
To Do shear Yo Do
Specific s D s = (-)
surface So Do
5.2. GEOMETRICAL SCALING
Usually corotating twin-screw extruders scale (60 < D < 200
mm) geometrically; this implies that
H
D c;
L
D c; N = c
where c = a constant.
Furthermore (locally)
<p = c Q.c = c ; f = c
(5.1)
(5. 2)
These practical rules are based on the principles which keep
the mixing constant.
Shear rate y:
V 'II'DN y ~
H H (5. 3)
-115-
Mean residence time:
volume t
throughput
~ f~ 'lf DLH
Q
fl!.'lf DLH
f l.2VHW
with f~ = local degree of fill*
And total shear y yt = f9_ 2L f Hsinqo
(5 .4) f 'lfDNsinqo
(5. 6)
As is well known (2) reorientation relative to the direction
of shear is extremely important for the mixing efficiency
M.E . - (___l y)nr nr ( 5. 7)
(The werking principle of static mixers is completely based
on nr rather than on the total shear y) • During each
revolution of the screw, material is passed from one screw
to the other and is assumed to be reoriented.
This may be written as
nr n.N.t
with
nr the number of reorientations
n the number of flights
N screw revolutions per second
t the mean residence time according to Eq. 5.4.
* f is the degree of fill e.g. underneath the hopper used only
(instead of Q) as a kind of dimensionless measure f o r the metered
throughput (Eq . 3.3).
(5 . 8)
f~ is the local degree of fill. It is equal to one in the case of
completely filled parts, but may differ from f, for example, by a change
of pitch.
hence
nr 2 f~~DLNH
Q
-116-
(5.9)
From Eqs. 5.1 and 5.2, it can be seen that the degree of
mixing (Eqs. 5.3, 5.5 and 5.6) remains constant: f, L/D and
~c = constant implies ~b = constant and ~a = constant (Eq.
3.21). If the samematerial is used (~, p = constant) the
specific energy, Eq. 3.39, is also constant. Furthermore, it
follows that throughput, energy and torque all scale with
the third power of the diameter in this case, see Table 5.2.
5.3. THERMAL SCALING
Geometrical sealing is, however, allowed only if both
extruders are operating under fully adiabatic conditions.
As soon as heat exchange with the barrel wall is more
important in the process, throughput can only scale with the
second power of the diameter because the barrel surface
(~DL) scales with o2 . Otherwise, the temperature development
in small test extruders will be different from that in
larger ones, with all consequences for viscosity differences
(mixing !) and thermal degradation with the more recently
developed high-melting-point polymers.
The problem that arises is that sealing according to the
mixing rules does not provide the same powers as sealing
according to the rules of equal thermal development. In
practice, this means that choices have to be made and that
experiments and eventually changes in screw design on the
larger scale extruder will always be necessary. This can be
illustrated as follows.
For laminar flows, the temperature development is given by:
-117-
pc u a x
+ \l (5.10)
whereas for ideally mixed annular flows (see Fig. 5.1):
dT Pc u dx
<a.> (T - T ) w +
H
<U> (T-Tw)
L
W=Y't
Figure 5.1. Heat balance for an ideally mixed annular flow.
5.3.1. LAMINAR FLOW
(5. 11)
The dimensionless form of these equations yields the
requirements for equal temperaturé development. For laminar
flows
Gz u*
where
Gz
Br
Graetz number
du* 2 + Br • (--)
dy*
pcVH H convective --
À L conductive
(5 .12)
heat transfer
heat transfer
(5 .13)
!lv2 viscous dissipation Brinkman number 'MT heat conduction
(5.14)
-118-
If Gz- and Br-numbers are constant, solution of Eq. 5.11
always gives the same result. Provided that the same
material is used (À,p,c and ~are constant), the same wall
temperature (Tw) is employed, and (as usual) L/D = a
constant, then these conditions yield (Eq. 5.14):
V = constant; consequently N ~ D-1 (5 .15)
And from Eqs. 5.13 and 5.15 it fellows that
H ~ D0.5 (5.16)
These very severe demands result in Q ~ D1.5.
The conditions ~ = 1, n = -1, h = 0.5 and q = 1.5 (see
Table 5.2) are met in practice only when heat exchange with
the walls is extremely important, for instanee in the
melting sectien of a single-screw extruder with a grooved
feed zone (3) .
5.3.2. IDEALLY MIXED ANNULAR FLOW; L/D A CONSTANT
For ideally mixed (annular) flows, which may be assumed for
the radially well mixed flow in corotating twin-screw
extruders, the heat balance for an infinitesimal element dx
can be written as:
dT 11'. Ddx + ~ [
dÜ ] ~ p c u dx dx 11' DH <a.> (T- - Tw) dy DH dx
(5 .17)
dT <a.> ('r - T ) du 2 p c u dx H + J.l ( dy) (5 .18)
In order to make these equations dimensionless, let
u T - T dx dy * ;;* :kl * * u V , fJ.T dx L ; dy H (5; 19)
-119-
ctöf* llT <a.> T* llT ~ v2 [ :y"-r. p c ~*v dx* + H2
DH dx L H
(5 .20)
-* ctf* ~ v2 H """* p c VHu
[ :;*] 2 dx* T* +
<a.> llTH2 <a.> c (5. 21)
ctöf* -* i*
du 2 p * + B • (-) u
dx* dy* (5.22)
with
P number pc VH
<a.> L
Péclet
Nusselt
convective heat transfer
total heat transfer
(5 .23)
B number
2 ~ V
<a.> HllT
Brinkman
Nusselt * As viseaus dissipation
total heat transfer
Nusselt
Péclet
As L
H
<a.>L
pcVH
À
total heat transfer
conductive heat transfer
convective heat transfer
conductive heat transfer
Constant temperature developments in this case require
constant P- and B-numbers; therefore, using the same
material (p,c,M,À constant), the sameheat transfer
coefficient <a.>, wall temperature (Tw) and when L/D is
constant,
hence
v "' ol/3 '
and VH
D constant
(5.24)
-120-
and therefore N ~ o-2/3 and H ~ o2/3 (5.25)
For the throughput this implies, as expected, Q ~ o 2 .
Consequently: 2. = 1, n = -2/3, h = 2/3 and q = 2, (see
Table 5.2). The sealing law for channel depth (in Eq. 5.25)
is often found in practice for many extruders (3), of
course, because channel depths are determined during the
construction of the extruders. Screw speeds and throughputs
can be changed easily at any later moment.
5.3.3. IDEALLY MIXED ANNULAR FLOW; H/D A CONSTANT
If the restrietion of L/D = c is dropped, and H - D is
introduced, (as is practice in corotating twin-screw
extruders) then the requirement of constant P- and B
numbers yields
VD V "' oO. 5 d an L a constant, so
N ~ o- 0 · 5 , L ~ o 1 · 5 and H ~ D , hence Q ~ o2 · 5
and 2. 1. 5, h 1, n -0. 5, q 2.5, see Table 5.2:
5.4. SCALING LAWS
To summarize: the (usual) sealing laws can be derived,
either from literature (4-6) or from the following
equations of this thesis:
Eqs. 3.2, 3.35, 3.36 p 3 + 2n + 2. - h
Eqs. 3.2, 3.35 q 2 + n + h
es p - q 1 + n + 2. - 2h
Eqs. 3.2, 3.31, 3.32 m 3 + n + l1. - h
Eqs. 3.1, 3.35, 3.44 t h + 1 + l1. - q l1. - 1 - n
Eq. 3.42 g = n - h + 1
gs g + t l1. - h
s = (1 + !1.) - q n + h + l1. - 1
(5. 26)
-121-
The sealing laws result in the following numerical values,
depending on the methad of sealing (see text, Eqs. 5.1
through 5. 26)
Tab1e 5.2 Sca1ing powers, fo11owing from different methods
of sca1ing.
geometrical 1aminar idea11y mixed
n 0 -1 -0.667 -0.5
h 1 0.5 0.667 1
l!, 1 1 1 1.5
e 3 1.5 2 2.5
es 0 0 0 0
m 3 2.5 2.667 3
q 3 1.5 2 2.5
t 0 1 0.667 1
g 0 -0.5 -0.333 -0.5
gs 0 0.5 0.333 0.5
s -1 0.5 0 0
5.5. EXAMPLE: GLASS-FIBRE REINFORCEMENT
The applicability of the sealing concepts derived so far
can be illustrated in the case of the glass-fibre
reinforcement of a degradation-sensitive polymer.
Processing should be performed as close as possible to the
(high) melting point (Tm) and, consequently, the scale up
rules for equal temperature development should be fulfilled
(see Table 5.2). In experiments on two corotating
twin-screw extruders with different sizes the sealing
exponents for screw speed and throughput were evaluated.
They proved to be dependent on the requirements which were
put on the average temperature of the outcoming melt (see
Table 5.3). Via simulations of the last part of the
extruder (melt!) assuming an averagemelt temperature at
the inlet (the end of the melting section) and aiming for
-122-
the same melt end-temperature on both extruders, Tm + 25 in
this case, a theoretical value for the sealing exponent 'q'
is found which is close to the experimental one (see Table
5. 3) .
Table 5.3. Experimental (rows a and b) and theoretica!
(row c) values for the sealing exponents for
screw speed and throughput. From (7).
Average Sealing exponents
Melt n q
Temperature
a Tm + 15 -1 2.3
b Tm + 35 -0.6 2.5
c Tm + 25 -0.4 2.4
5.6. CONCLUSION
The conclusions that can be drawn from this short
investigation of the sealing laws are that geometrical
sealing yields maximum throughput and meets the
requirements of constant mixing, but results in a different
temperature development as soon as the extruders do not
operate in a completely adiabatic mode. Consequently,
mixing is not the same anymore because viscosities are
temperature dependent.
In the case of equal temperature development, screw speed
(absolutely) and throughput (relatively) are lower, with as
a consequence decreased mixing, energy consumption, and
torque. Large extruders (D > 200 mm) do not allow
geometrical sealing any more. The circumferential speed and
the wear it causes become too high.
-123-
5.7. REFERENCES
1. C. Rauwendaal, Polym. Eng. Sci., ll, 1059, (1987)
2. K.Y. Ng and L. Erwin, Polym. Eng. Sci., 21, 4 (1981)
3. P. Fischer, Ph.D. Thesis, RWTH Aachen, West Germany
(1976)
4. J.F. Ingen Housz, Intern. Zeitschrift for Lebensmittel
Technologie, 1La2, 48 (1982)
5. J.F. Ingen Housz, Polymer Extrusion II, Paper 8, London
(1982)
6. J.R.A. Pearson, Plast. Rubber Processing, Sept., 119
(1976)
7. H.E.H. Meijer and P.H.M. Elemans, Polym. Eng. Sci., ~.
275 (1988)
8. H. Potente "Auslegen von Schneckenmaschinen Baureihen"
(1981)
-124-
CHAPTER 6
TIME EFFECTS IN THE DISPERSIVE MIXING OF
INCOMPATIBLE LIQUIDS
The deformation and breakup processes of single droplets in
well-defined fields of flow have extensively been studied in
the literature. In spite of the fact that in real mixers the
conditions are far from equilibrium, most studies are
confined to (Newtonian) systems undergoing almost stationary
deformation (1-4). The time effects that occur during the
different stages of the dispersive mixing process are less
well understood.
In a Plexiglas-walled Couette-apparatus, the time-dependent
deformation of Newtonian droplets into extended threads has
been studied. When the shear rate is very slowly increased,
allowing for almost equilibrium deformation, the results of
the critical capillary number Ca as a function of viscosity
ratio, as reported in the literature (3), are reproduced.
However, in transient flows at capillary numbers Ca >>
Cacrit' droplets are deformed into long slender bocties which
can remain extended until the shear has stopped.
They then desintegrate into lines of droplets because of the
interfacial tension-driven Rayleigh disturbances. The time
for breakup can well be calculated from the existing theory
on the stability of extended liquid cylinders (5).
6.1. INTRODUCTION
When a blend of two incompatible polymers is subjected to
shearing forces, e.g. in a corotating twin-screw extruder,
droplets of the dispersed phase will deform into long,
threadlike particles (7), which can break up into smaller
droplets. The ultimate morphology - the relative
-125-
distribution of the constituents - depends on volume or
weight fraction (8), viscosity ratio of both polymers and
the type of flow (4).
The word 'polymer', deliberately used in the former
paragraph, in contrast with 'liquids', as mentioned in the
title of this chapter, suggests a similarity in the
behaviour of systems used in emu1sion rheology as well as in
polymer rheology. In the next Sections, therefore, we will
review some of the literature on time-dependent deformation
and breakup of Newtonian droplets. The emphasis will be on
shear flows. Although less efficient than elongational
flows, they can be realized easier in practice. In Chapter
9, results from experiments with these systems will be
translated to the processing conditions of a polymer blend.
6.2. AFFINE DEFORMATION OF DROPLETS IN SIMPLE SHEAR FLOW.
For large capillary numbers Ca (Ca= ~cyR/o), and for
viscosity ratios p < 1, droplets will deform affinely with
the matrix. Their deformation can be expressed in terros of
the total shear y (y = yt), imposed by the flow field (9).
In a simple shear flow, a spherical droplet having an
initia! diameter 'a' will deform into an ellipsoid with
length L and width B:
L/a ( 1 + 2 1/2
y ) (6 .1)
B/a (1 + 2 -1/4
y ) (6 .2)
hence
L/B (1 + y2) 3/4
(6.3)
-126-
or, by defining the deformation D (L - B) I (L + B)'
2 3/4 (1 + y ) - 1
D (1 + y2)3/4 + 1
Figure 6.1 shows the properties L/a, B/a and D as a
function of y, calculated from Eqs. 6.1, 6.2 and 6.4.
3,-----------~----------------------------~
2.5
2
~ _J
r5 1.5
~ 0
1\ I I I 0.5
2
D
8/a
3 4 5 6 7 8 9 10
---> 'Y
Figure 6.1. Relative length L/a, relative width B/a and
deformation D of an affinely deformed droplet as a
function of y.
(6 .4)
As stated befare (10,11), periodical reorientation of the
already deformed droplet increases the efficiency of the
mixing process. This is illustrated in Figure 6.2, which
shows the influence of reorientations on the decrease in
width of a droplet in simple shear flow.
0.8
0.6
~ m 0.4
1\ I I I 0.2
0 0
-127-
·--10
---> 'Y
without reorientations
--·---·-with reorientations
20 30
Figure 6.2. Effect of reorientation on the decrease of B/a,
the relative width, of a droplet undergoing affine
deformation in simple shear flow.
6.3. BREAKUP OF THREADS
Until now, the shear rate y and the time t have been
interchangeable. However, once the droplet has become
highly extended, it will exhibit initially small sinusoidal
distortions, each possessing a wavelength À along the
thread. Depending on the viscosity ratio p = ~dl~c' one
wavelength, Àmr will turn out to be dominant and will cause
breakup (5,6,12). For Newtonian threads of infinite length
in a fluid mat~ix which is at rest, the time for breakup,
tb, can well be calculated using the theory of Tomotika
(5, 6) .
where
1 o.a1 R0 ll,n( ) q a.o
R0 the initial thread radius
a.o the initial distortion amplitude
q the growth rate of a distortion
(6 .5)
-128-
Predicting the stability of a thread in a matrix which is
deforrning is rnuch more difficult, because the wavelength À
is changing with time. For instance, in a shear flow the
distortions are swept away (see Figure 6.3), whereas in an
elongational flow a potentially dominant wavelength will be
extending so fast with the matrix (6,7,13) that its effect
will be darnped out before having been able to cause breakup.
------~~~.~-----~ i~,..--. --"7 e=:==--...J
------~~~>~'-------- ~ ~.-----------i
Figure 6.3. Distortiens being swept away in simple shear flow.
From (14).
Frorn the detailed studying of the dynamics of this breakup
process in sirnple shear flow and hyperbalie extensional
flow, by Mikarni et al. (15) and later by Khakhar et
al. (16), it can be said that the presence of flow has a
stabilizing effect on the thread. In hyperbalie extensional
flow, the time for breakup increases slowly with the
capillary nurnber (16):
tb - log (Ca), (6. 6)
while the size of the draplets is inversely proportional to
the strain rate.
In the case of simple shear flow, the results are presented
in terros of the initial orientation angle e0 of the thread
in the flow field. The time for breakup is given by (16):
(6. 7)
where
-129-
c = 2 tan eo eo initial orientation angle
v a constant > 0
The other initial condition, the minimum amplitude tt0 required for magnification of a distartion caused by the
interfacial tension to occur, is taken from the work of
Kuhn (13). Assuming the distortions to be due only to
thermal fluctuations, he proposed the following estimate
for tt0 :
21 kT ./(
8 3/2 Tl' (1
(6. 8)
with k Boltzmann constant
T absolute temperature
From the analysis presented in Ref. 16, it follows that
under similar conditions, the drops formed in simple shear
flow are larger than those in hyperbalie extensional flow.
6.4. BREAKUP OF DROPLETS
Numerous studies have been devoted to the deformation and
breakup of small draplets in a matrix undergoing simple
shear flow or hyperbalie extensional flow. Good reviews of
the relevant experimental and theoretical work reported in
the literature can be found in the workof Elmendorp (7),
Rallison (17) and Acrivos (18). To be recommended
separately is the almost 'classical' workon droplet
breakup by Grace (3,11), who used a Couette-apparatus and a
four-roll mill to generate simple shear flow and hyperbalie
extensional flow, respectively.
-130-
More recently, Bentley et al. (19,20) developed a
computer-controlled four-roll mill. The continuous
adjustment of the speed of each individual roller enabled
them to position a droplet almost anywhere in the apparatus
and to keep it there for a considerable period of time.
Moreover, they succeeded in (approximately) generating a
wide range of flows, characterized by a single parameter «,
which is correlated to the ratio of deformation rate to
vorticity.
The velocity field is then given by (21) :
-1 < « < 1 (6.9)
where x 1 and x 2 represent coordinates along the axes that
pass through the eentres of the rollers.
The limiting cases are:
« 1 hyperbolic extensional flow
« 0 simple shear flow
« -1: purely rotational flow
The conclusions of their experiments on steady droplet
deformation and breakup are in general agreement with those
of earlier authors.
Continuing Bentley's work, Stone et al. (22,23,24) study
transient effects in droplet breakup. By slowly increasing
the shear rate, droplets are brought to a certain
elongation ratio L/a. The flow is then stopped. Under the
influence of interfacial tension, the droplets may return
to their original shape or alternatively break up into
several fragments. For a broad range of viscosity ratios p
-131-
(0.01 < p < 10), the breakup mechanism via Rayleigh
disturbances is only observed for L/a values > 15. For
lower L/a values (6 < L/a < 15), the draplets exhibit
'end-pinching' (22). The ends of the draplets become
spherical, while the overall length decreases. The ends
pinch off, with the newly formed fluid thread relaxating
further towards a sphere. Obviously, the time scale for
'end-pinching' is much shorter than that for the growth of
Rayleigh disturbances.
Figures 6.4 and 6.5 show the influence of the viscosity
ratio and initial shape on relaxation and breakup of an
extended droplet. (Note that these theoretica! curves show
remarkably good correspondence with the experimental ones
in Ref. 23)
(a) C1C& #::JD
(b) cCC:w;:: >t:J)
(c) OCz >(:JJ=)
(d) c:= CIG>--g s(CJD =)
Figure 6.4. Relaxation and breakup of an initially extended
droplet in a fluid matrix which is at rest.
Parameter: viscosity ratio. From Ref. 23.
p = 0.05
p = 0.1
p = 1.0
p - 7.5
p 10.0
Reprinted by parmission of Cambridge Univarsity Press.
-132-
c ( (~) ) -::::> L/a 5.3
L/a - 6.4
L/a = 7.5
Figure 6.5. As Figure 6.4., for low viscosity ratio droplets.
Parameter: initial shape. From Ref. 23.
Reprinted by permission of Cambridge University Press.
Figure 6.6 shows the calculated shapes of an extended
droplet (L/a = 14), which is breaking up under the combined
action of 'end-pinching' and Rayleigh disturbances. In this
Figure, the time is made dimensionless with the factor
~cRo/cr, and can be compared with the dimensionless time tb*,
required for breakup solely by capillary waves (Eq. 6.5).
1 0.81 Ro tb q !1-n (
<XQ )
* tb ~eRa/a ~cRo/a
(6.10)
where q, the growth rate of a distortion, is given by ( 5' 6)
q (6.11)
-133-
with
Q(À,p) a tabulated function (4-6)
p the viscosity ratio ~dl~c
R0 the initial thread radius
2 hence (6.12)
with Àm the dominant wavelength at which breakup occurs
240
t* 100
t* = 0
Figure 6.6. Calculated evolution of capillary waves during
the cornbined relaxation and breakup of an
initially highly extended droplet in a fluid
matrix which is at rest (p = 1). Frorn Ref. 23.
Reprinted by perrnission of Cambridge University Press.
-134-
In the calculations performed in Ref. 23, the viscosity
ratio p = 1, and the relative distartion amplitude
tto/Ro 3.10-3. The value for Q(Àro,p) = 0.07 can be deduced
graphically from Figure 2.7 in Ref. 6, or from Figure 8.1
in Chapter 8 of this thesis. The dimensionless time for
breakup thus becomes:
t b
* 2
0.07
0.81 !!.n ( 3) ~ 160
3.10-(6.13)
From this calculation, and from Figure 6.6, it is clear
that the desintegration of the droplet by the mechanism of
'end-pinching' occurs on a shorter time scale than by
Rayleigh disturbances, and that the slowly but continually
decreasirig L/a increases the time for breakup (tb* = 240
VS. 160) .
6.5. EXPERIMENTAL
Breakup experiments with Newtonian systems were performed
in a Plexiglas-walled Couette-apparatus. The purpose of the
experiments was twofold:
(i) to investigate the behaviour of draplets during
deformation and breakup under non-equilibrium conditions
(ii) to check the time-dependent deformation of draplets
at low capillary numbers.
Although limited in direct practical applicability, we feel
that these experiments might give some more insight into
the time scales involved in the deformation of draplets in
real mixers, e.g. in the strongly varying flow field
between the kneading flight and the barrel of a corotating
twin-screw extruder.
-135-
6.5.1. EXPERIMENTAL SETUP
A simple shear flow is generated in a Couette-type
apparatus. It consists of two counterrotating concentric
cylinders. The speed of both the inner and outer cylinder
can be adjusted to keep a droplet at the same position
while being deformed at a shear rate y, which for a
Newtonian fluid is given by (25):
where
the radius of the inner, resp. outer
cylinder
(6.14)
the angular velocity of the inner, resp. outer
cylinder
In the present apparatus, R1 = 50 mm, R2 = 63.2 mm. The
height of the gap is 120 mm, which is enough to avoid
disturbance of the flow field due to the bottom. Shear
rates up to 20 s-1 can be obtained. The deformation of a
droplet can be studied from above by a microscope plus a
video system, connected to a high-resolution monitor.
6.5.2. MODEL FLUIDS
Tables 6.1 and 6.2 list the characteristic properties of
the Newtonian fluids used in the experiments. As continuous
phase, silicon oil (Rhodorsil, Rhone-Poulenc) was used. As
disperse phase, solutions of Corn Syrup (Globe 01138) in
water were used. Viscosities at 23° C were determined on a
Bohlin Vor dispersion rheometer. Newtonian behaviour was
observed up to approximately y = 10 s-1. The interfacial
tension between the Corn Syrup (C.S.) solutions and the
silicon oil was measured using a Du Nouy-ring tensiometer.
-136-
Table 6.1. Viscosities of the Newtonian model fluids at 23° C.
11J23 ° C (Pa.s)
continuous phase Rhod V 60000 63.3
Rhod V 30000 29.9
Rhod V 12500 12.5
Rhod V 5000 4.95
Rhod V 1000 1.04
disperse phase
(wt% C.S./wt% H20) 96/4 28.0
95/5 16.25
90/10 4.06
85/15 1.13
83/17 0.75
Table 6.2. Interfacial tension between Corn Syrup (C.S.)
solutions and silicon oil.
wt% C.S./wt% H20
96/4 40.0
95/5 40.0
90/10 41.4
85/15 46.7
83/17 47.0
6.5.3. RESULTS
We will now discuss experiments with model fluids in simple
shear flow, each representing a stage of the dispersion
process that occurs on a distinct time scale. Only the most
important results will be presented here. Details can be
found in reports by Janssen (26) and Bos (27).
-137-
6.5.3.1. AFFINE DEFORMATION
*) A supercritical simple shear flow, i.e. Ca > Cacrit is
stepwise applied to a spherical droplet (typical diameter
about 1 mm). The length L or, at deformation D > 0.8, the
width B, is measured from the monitor. The deformation, D,
can be determined and compared with the theoretical
prediction as given by Eq. 6.4. Affine deformation occurs
at Ca/Cacrit ~ 2. See Figure 6.7.
0 x 0 .9
0 .8 • 0.7 0 • • 0.6
• • p=0.135
0 0 .5
1\ I 0.4 I I 0.3
0 Co/Ca cri!= 7.1
.... Co/Co cri! = 3.2
x Co/Co cri! = 2.1
0.2 • Co/Co cri! = 1.2
0 .1
5 10 15 20
---> y
Figure 6.7 . Affine deformation at supercritical capillary number.
*)
Viscosity ratio p = 0.135. The drawn lines represent the
deformation calculated from Eq. 6.4.
Cacrit is the capillary number for breakup under steady conditions.
-138-
6.5.3.2. BREAKUP OF THREADS IN SIMPLE SHEAR FLOW
A supercritical simple shear flow, with Ca >> Cacrit' is
stepwise applied to a spherical droplet, which will deform
affinely into a thread. At a given moment, the thread will
exhibit sinusoidal distortions which will cause breakup.
The time for breakup tb is defined as the elapsed time
between the onset of the flow and the moment at which a
line of seperate draplets has been formed in the central
region of the thread. Figure 6.8 compares the results
concerning the dimensionless time for breakup with data
from similar experiments by Grace (3). Upon exceeding
Ca/Cacrit' the time for breakup does not decrease, as in
Grace's experiments. Grace possibly observed end-pinching
(see Section 6.4), which indeed yields a much lower value
for tb. The absence of a clear-cut definition of the time
for breakup in Ref. 3 might explain the large difference
shown in Figure 6.8.
1\ I I I
10
---> Co/Cocrit 100
Figure 6.8. Effect of exceeding Cacrit on dimensionless burst time.
Symbols indicate the value of p. Open symbols reprasent
data from Ref. 3. The line for p = 1 is obtained by plotting
tb* at Ca/Cacrit = 1 on the vertical axis, and assuming an
identical slope.
-139-
6.5.3.3. STABLE DEFORMATION AND RELAXATION OF DROPLETS
A spherical droplet is subjected stepwise to a subcritical
shear flow with constant shear rate. The deformation is
measured as a function of time, for viscosity ratios p =
0.09; 0.325; 0.933; 2.24. Figure 6.9 gives a typical plot
of the results (p = 0.09).
Figure 6.10 compares the final stable deformations of the
present systems with data from similar experiments by
Rumscheidt et al. (28).
The final deformation (usually reached at a dimensionless
time of about 25) is, at least for this range, almost
independent of the viscosity ratio p.
0.8
0.6
0 1\ I 0.4 I I
0.2
10 20 30
---> t*
p = 0 .09
0 0
40 50 60 70
• Ca= 0.527
• Ca= 0.463
o Ca= OA30
0 Ca= 0.393
6 Ca= 0.352
<> Ca= 0.295
'\1 Ca= 0.248
Figure 6.9. Deformation D as a function of dimensionless time
t* = t ~/(DeR), at subcritical capillary nurnber.
Upon cessation of the flow, the draplets return to their
original shape. The required dimensionless time is about
15. Figure 6.11 shows this relaxation process for a
viscosity ratio p = 0.933. The smal! deformation (D = 0.05)
that remains is due to optica! distartion by the monitor.
-140-
0 .8
• •• 0 .6
0 • D
~~· D
• o •• 0
0 0 .4 .. 0
1\ ~ • p = 0.0904
I ~ • p = 0.325
I 0 I 0.2 • p = 0.933
D p = 2.24
2 = 4.8 {R &: t.4)
0 0.0 0.1 0 .2 0.3 0 .4 0.5
---> Ca
Figure 6.10. Stable deformation as a function of the capillary nurnber.
The drawn line represents data from Ref . 27.
As stated before in Section 6.4.2, droplets that are
extended beyond a critical elangation ratio (L/alcrit will
not return to their initial shape, but will break up into
several fragments via the 'end-pinching' rnechanisrn after
the flow has stopped. Table 6.3 gives (L/alcrit for three
viscosity ratios.
0 1\ I I I
0.7-.---------------------,
0.6 p = 0 .933
0.5
0.4
0.3
0.2
0.1
0 .0 0 4 B 12
---> t*
• Co= 0.381
• Co= 0.359
0 Ca= 0.344
D Ca= 0.334
I:J. Ca= 0.309
0 Ca= 0.259
V Ca= 0.239
0 Co= 0.216
16 20 24
Figure 6.11. Relaxation of deformed droplets upon cessation of the flow.
Viscosity ratio p = 0.933.
-141-
Table 6.3. Critica! elongation ratio upon cessation of the flow.
disperse phase cont. phase p (L/alcrit
wt% C.S./wt% H20 Rhodorsil
96/4 V 30000 0.933 9.8
90/10 V 30000 0.325 6.3
83/17 V 30000 0.025 5.8
These observations are in fair agreement with those of
Stone et al. (22). Unfortunately, the critical elangation
ratio at higher viscosity ratios can not be investigated in
the present Couette device, because of the limitations
imposed by the type of flow (11).
6.6. CONCLUSION
In the case of Newtonian draplets in a Newtonian matrix,
affine deformation occurs in simple shear flow at Ca/Cacrit
~ 2. The time required for deformation can be calculated
from Eq. 6.4 (with y = yt). Upon cessation of the flow, the
time for breakup of highly extended draplets (L/a > 15) can
be calculated using Tomotikas theory (Eq. 6.5), see also
Chapter 8. For moderately extended droplets (6 < L/a < 15),
the numerical calculations of Stone et al. (23) apply, see
Figures 6.4-6.6.
If viscoelasticity is introduced, deformation and breakup
of draplets is more complicated. Some scouting experiments
with Boger fluids have been performed by Bos (27). Results
of a more systematic study can be found in Ref. 30 for
extensional flow, but especially in Ref. 32 for shear flow,
giving an increased Cacrit with increasing elasticity for
almest all values of the viscosity ratio p.
-142-
6.7. REFERENCES
1. G. I. Taylor, Proc. Roy. Soc. (London), hl.la, 41 (1932)
2. G.I. Taylor, Proc. Roy. Soc. (London), A146, 501 (1934)
3. H.P. Grace, Chem. Eng. Comm., u, 225 (1982)
4. F.D. Rumscheidt and S.G. Mason, J. Coll. and Int . Sc i.,
.1..6., 238 (1961)
5. s. Tomotika, Proc. Roy. Soc. (London), ~, 332 (1935)
6. J.J. Elmendorp, Ph.D. Thesis Delft University of
Technology (1986)
7. S. Tomotika, Proc. Roy. Soc. (London), ~, 302 (1936)
8. G.G.A. BÖhm, G.M. Avgeropoulos, C.J. Nelson and F.C.
Weissert, Rubber Chem. Tech., ~, 423 (1977)
9. J.M. Starita, Trans. Soc. Rheol., .1..6., 339 (1972)
10. K.Y. Ng and L. Erwin, Polym. Eng. Sci., ~, 4 (1981)
11. This thesis, Chapter 1, Section 1.4
12. Lord Rayleigh, Proc. Roy. Soc., (London), 2..2_, 71 (1879)
13.8. Kalb, R.G. Cox and R. St. John Manley, J. Coll. Int.
Sci., .8.2., 286 (1981)
14.W. Kuhn, Kolloid z. 132, 84 (1953)
15. T. Mikami, R.G. Cox and S.G. Mason, Int. J. Multipbase
Flow, 2, 113 (1975)
16. D.V. Khakhar and J.M. Ottino, Int. J. Multipbase Flow,
u, 71 (1987)
17. J.M. Rallison, Ann. Rev. Fluid Mech., .1..6., 45 (1984)
18.A. Acrivos, 4th Int. Conf. on Physico-chemical
Hydrodynamics, Ann. N.Y. Acad. Sci., ~, 1-11
19.B.J. Bentley and L.G. Leal, J. Fluid Mech., lQl, 219
( 1986)
20.B.J. Bentley and L.G. Leal, J. Fluid Mech., lQl, 241
(1986)
21. J.M. Ottino and R. Chella, Polym. Eng. Sci., ~, 357
(1983)
22.H.A. Stone, B.J. Bentley and L.G. Leal, J. Fluid Mech.,
m, u1 (1986)
23.H.A. Stone and L.G. Leal, J. Fluid Mech., ~, 399 (1988)
-143-
24. H.A. Stone and L.G. Lea1, submitted to J. Fluid Mech.
(1989)
25.B.J. Trevelyan and S.G. Mason, J. Colleid Sci., Q, 354
(1951)
26. J.M.H. Janssen, internal report Eindhoven University of
Technology (1989)
27.H.L. Bos, internal report DSM (1988)
28.F.D. Rumscheidtand S.G. Mason, J. Coll. Sci., ~' 238
(1961)
29. S. Wu, Polymer Interface and Adhesion, Marcel Dekker
Inc., New York and Basel (1982)
30.C. van der Reijden-Stolk, Ph.D. thesis Delft University
of Technology (1989)
31.W.J. Tjaberinga, internal report Delft University of
Technology (1988)
32.R. de Bruin, Ph.D. Thesis Eindhoven University of
Technology (1989)
-144-
CHAPTER 7
MORPHOLOGY OF THE MODEL SYSTEM
POLYSTYRENE/POLYETHYLENE
Blends of incompatible polymers often exbibit cbaracteristic
morpbologies (1). As an example, tbe morpbology of a blend
of tbe model system polystyrene/polyetbylene (PS/PE) is
sbown as a function of tbe viscosity ratio and composition.
Tbe blends are made on a corotating twin-screw extruder,
under identical processing conditions.
At low volume fractions, tbe dispersed pbase is present as
more or less spberical particles in a matrix. When tbe
volume fraction is increased, no stable dispersions are
found.
At a viscosity ratio of about unity, tbe dispersed pbase
consists of long, ultratbin fibers. Under the influence of
tbe interfacial tension, tbey eitber desintegrate into
droplets or coalesce. Co-continuous morpbologies are formed
upon blending or after compression moulding.
Wben block copolymers are introduced in tbe system, mucb
finer morpbologies are obtained. Tbe non-equilibrium
cbaracter, bowever, is still present.
Reprinted partly from: P.H.M. Elemans, J.G.M. van Gisbergen and H.E.H.
Meijer in Integration of Polymer Science and Technology. Part 2. Eds.
P.J. Lemstra and L.A. Kleintjens, Elseviers Applied Science Publishers,
Londen and New York (1966) by permission of Elseviers Applied Science
Publishers.
-145-
7.1. INTRODUCTION
As discussed in the farmer Chapter, the deformation and
breakup behaviour of isolated draplets can well be studied.
In a real blend, the final partiele size can not simp1y be
predicted from theory. As soon as the volume fraction of the
dispersed phase exceeds a critical concentration (~d z 1%),
the particles will start to collide and eventua1ly coalesce
(2,3). At volume fractions ~d > 15%, the system can no
langer be considered dilute (4). Co-continuous structures
will emerge until finally a composition is reached at which
phase inversion occurs (5) . The position of this inversion
region depends on the viscosity ratio of the components, as
is shown in Figure 7.1.
TORQUE RATIO
EPOM PBO
8 6 4
2
-~ ! ! "' ~~--• t CONT.
I ccoÖNT. ~\ I I --i- -1--i--
:~ EPDM ti .I=- .I .4 CONT ,
• 0 0.25 0.50 0.75 1.0
VOLUME FRACTION PBO
Figure 7.1. Morphological changes of blends of EPDM and polybutadiene
as a function of composition and viscosity ratio (represented
by the torque ratio). From Ref. 5.
This diagram shows that the component with the lower
viscosity or higher volume fraction will farm the continuous
phase. By assuming a similar influence of bath the viscosity
ratio and volume fraction, a semi-empirical equation (only
valid at low shear rates) expressing the location of the
mid-point of the inversion region has been proposed by
Jordhamo et al. (6):
-146-
lf'1 (7 .1)
The phenomenon of the coalescence of droplets has received a
lot of attention in the literature. With respect to polymer
blending, the theoretica! modelling is still in its initia!
stage. A few contributions are worth mentioning here.
With the aid of gravity-induced coalescence experiments,
Elmendorp (2) shows that the mobility of the interfaces in
molten polymer systems is quite high: droplets with radius R
exhibit coalescence times te ~ R, indicating nearly mobile
interfaces (whereas for droplets having fully immobile
interfaces, te ~ R5 ) . Therefore, coalescence will occur on
a rather short time scale (7). Recently, Chesters (8)
proposed a global model for partially mobile liquid-liquid
interfaces. The rate-determining step of the coalescence
process, the removal of the liquid film between the two
interfaces, is assumed to be due to a shear stress in a
boundary layer inside the droplet phase. Hence, not only the
matrix viscosity, but also the viscosity of the dispersed
phase and the viscosity ratio 'p' play an important role.
The validity of this model can be investigated by observing
the thinning rate of the film between two approaching
interfaces (9).
In the case of the gravity-induced coalescence of a droplet
approaching a flat liquid interface, the Chesters model
yields an expression for the film thickness (h) as a
function of the contact time (t), see also Ref. 8.
with
A l")d ,f (g t::.p/3)
3/2 (]
(7. 2)
(7. 3)
where
R
'ld g
!lp
(1
the
the
the
the
the
-147-
radius of the droplet
zero-shear viscosity of the droplet phase
constant of gravity
density difference between the two phases
interfacial tension
7.2. PHASE INVERSION
The schematic phase diagram of Figure 7.1 originally refers
to EPDM-butadiene rubber blends (5). However, it can also be
used to investigate the morphologies that arise upon
melt-blending two incompatible polymers. As representative
of this large group of polymer pairs, the model system
polystyrene/high-density polyethylene (PS/HDPE) is chosen.
7.2.1. MATERIALS
The properties of the materials used are listed in Table 7.1.
The flow curves (Figure 7.2) are more or less parallel,
which to some extent guarantees a constant viscosity ratio
over a wide range of shear rates.
Table 7.1. General properties of the materials used in the
extrusion experiments.
material code
HDPE (DSM)
Stamylan 8760
Stamylan 7058
Stamylan 7359
PS (DOW Chemica!)
634
638
meltindex I2
(dg/min)
0.06
5
35
3.5
25
density p123
oc (kg/m3)
956
952
954
1050
1050
-148-
7.2.2. BLEND PREPARATION
Blends of PS and HDPE were made on a Werner & Pfleiderer
laboratory corotating twin-screw extruder. The throughput
Q = 6 kg/hr, and the screw speed N = 270 rpm were kept
constant during the experiments. At the exit of the
extruder, the strand (with diameter about 3 mm) was
immediately quenched in water.
..,.. ö
.!::-
=-A I I I
1000
100
10 100
-~-> y (s-1) 1000 10000
Figure 7.2 Viscosity ~as a function of shear rate y of PS
and PE as determined on a Göttfert Rheograph
2000 capillary viscosimeter.
7.2.3. PHASE INVERSION DIAGRAM
HOPE 7359
t!.!1ITZ2~--· HOPE 8760 __
LOPE 1808 AN 99 ~H~L ____ _ PS 634
Figure 7.3 and 7.4 show the phase inversion diagram for the
system PS/HDPE. The micrographs represent the view of the
central regions of the strands. As can be inferred from
these micrographs, the sameareasas in Figure 7.1 exist.
However, with regard to the co-continuous area, there is no
symmetry around p = 1: The imaginary lines indicating the
co-continuous area are not parallel.
Some details of Figures 7.3 and 7.4, showing the different
stages of thread formation and breakup, are shown in Figures
-149-
1.1a, 1.1b, 1.10 and 1.11 of this thesis. Figure 7.5 shows
the presence of thin fibers at a viscosity ratio p z 1.
Campare the Chapters 1 and 6 for an explanation of the
formation of these extended morphologies.
Figure 7.5. Scanning electron rnicrograph of the edge of a
rnicrotorned surface of a 45/55 PS/HDPE blend (p z 1) .
Capricieus PS threads are visible in the direction of
extrusion.
next pages:
Figure 7.3. Scanning electron rnicrographs of the rnicrotorned
surface perpendicular to the direction of extrusion.
Figure 7.4. As Figure 7.3, parallel to the direction of extrusion.
Figure 7.6. Scanning electron rnicrographs of the rnicrotorned surfaces
of PS/LDPE blends (p z 1). Upper series: Befere cornpression
rnoulding, perpendicular to the direction of extrusion.
Lower series: After cornpression rnoulding.
-150'-
3dOHI..lfSdl..l
0
"' ..... 0 ClO
0 (") ..... 0 ,..._
0 ~ ..... 0 co
0 lil ..... 0 lil
0 ,..._ ..... 0 (")
0 ClO ..... 0
"'
2
UI 11. 0 :z:
s:::-..... "' 11. s:::-
0.5
0.2
10/90 20/80 30/70 40/60 50/50 60/40 70/30 80/20
%HOPE %PS
I .... (.Tl .... I
-152-
p9pJnOW UOJSS9JdWOO
0 N ..... 0 CD
0 M ..... 0 ,......
0 ~ ..... 0 CD
0 10 ..... 0 10
0 CD ..... 0 ~
0 ,...... ..... 0 M
0 CD ..... 0 N
w a.. oen J:C.. $$
-153-
7.2.4. INFLUENCE OF COMPRESSION MOULOING
Blends of the system polystyrene/low-density polyethylene
(PS 638, OOW/LOPE Stamylan 1808 AN 00, OSM) were processed
under identical conditions as listed inSection 7.2.2.
Subsequent to blending, the strands were pelletized and
compression moulded for 15 minutes at 200 °C (standard PS
cyc1e). Figure 7.6 shows the morphologies of the blends
before and after compression moulding. At higher volume
fractions ~ ~ 30%, a coarsening of the blend towards a
co-continuous structure is visible.
7.3. FILM THINNING
The experiments of Section 7.2 yield results at a rather
macroscopie scale. As a contrast, a scouting experiment will
be discussed which invo1ves the gravity-induced coalescence
of a polystyrene droplet with a horizontal interface of
molten polyethylene, at a temperature of 200 °C in an
atmosphere of nitrogen.
Typical data of the system polystyrene (PS 638, OOW
Chemical)/high-density polyethylene (HOPE Stamy1an 7058,
OSM) at 200 °C: Interfacial tension a= 5.10-3 N.m-1 (10);
density difference ap = 202 kg.m-3 (11); zero-shear viscosity
of the PS droplet ~d = 1400 Pa.s; droplet radius R = 1.5 mm.
A PS droplet is inserted in a HOPE matrix at an
approximately constant height above the interface. After a
contact time 't' has elapsed, the system is cooled down
immediately. The sample is then carefully microtomed, and
the film thickness at the droplet's plane of symmetry is
determined via scanning electron microscopy (see Figure
7.7). By varying the contact time, a first estimate for the
value of the proportionality constant 'A' in Eq. 7.3 can be
obtained.
For the present system, the Chesters model is expected to
apply at film thicknesses h ~ 40 ~ (12) . The theoretical
-154-
value of the constant A= 1.108 s.m-2. Figure 7.8 shows the
film thickness versus the reciprocal contact time. From the
slope of the line, a value A= 2.5.108 s.m-2 is found, which
is in the order of magnitude of the theoretical value.
This result can be compared, in terms of the thinning rate
(dh/dt), with systems having mobile or immobile interfaces,
respectively.
From Ref. 2 it fellows that
1 dh 8 1f (}2
h 3 dt 2 (immobile interfaces) 3 '1m Req F
1 dh 2 (}
h dt 3 '1m Req (mobile interfaces)
From Eq. 7.3, dh/dt can be expressed as:
dh
hence
where F
Req A
11c
1 dh
h 2 dt
t:.p
2R
.i R3 g 31f
1.0.108 s.m-2
1800 Pa.s
(7. 4)
(7. 5)
(7. 6)
( 7. 7)
For the present system, a global comparison between the
thinning rates predicted by the three models can be made by
substituting the parameters given in this Section. A film
thickness h = 10 pro is assumed. The calculated thinning
rates are presented below with respect to the thinning rate
predicted by the Chesters model.
-155-
dh (dh) (_)!mobile limmobile dt dt 1
:::: 20 :::: -- (7. 8) (dh) (dh) 600
dt I Chesters dt I Chesters
This means: The assumption of immobile as well as mobile
interfaces yields values for the thinning rates which differ
at least one order of magnitude from the experimental one.
Given the inaccuracy of the experiment (errors in droplet
radius are magnified to the power three), these results give
some support to the validity of the Chesters model.
I mm
Figure 7.7 Scanning electron micrographof a PS droplet in a
HDPE matrix. The film thickness (which is rather
large here) can readily be determined.
t U)
I 0
800.-----------------------~----------------,
600
• .:::::.- 400 ..c 1\ I I I 200
A=2.5•10+8 (s/m2)
)//~~~> R3/l (10-13 m3/•) Oi-----~----~.-----.------,------,-----~
0 5 10 15 20 25 30
Figure 7 . 8 Film thickness h versus reciprocal time R3/t,
of a PS droplet, approaching a horizontal PS/PE
interface.
-156-
7.4. INFLUENCE OF BLOCK COPOLYMERS
As was shown in the previous Sections, a blend of two
incompatible polymers is inherently unstable. Thread-like
particles will desintegrate into droplets, while coalescence
will destabilize the dispersion or will lead to
co-continuous structures. In principle, there are two
general routes to stabilize, at least locally, a desired
morphology (13).
(i) Functionalizing the two polymers with reactive groups.
As an example, related to the model system PS/LDPE, a
morphology of finely dispersed particles was reported for
blends of an oxazoline-modified PS (14) and a carboxylated
LDPE (15). With regard to the mechanica! properties,
however, no impravement was observed for the elongation at
break and the impact strength in PS rich blends (15,16).
(ii) Actding block copolymer to the blend.
It is well known from the literature (17-22) that block
copolymers of the type A-B may act as emulsifiers in a
system containing the immiscible polymers A and B. The block
copolymer tencts to concentrate at the interface, thereby
promoting interactions between the two phases. As a result,
the interfacial tension will be reduced. Therefore, with
regard to mixing, droplets will breakup at a smaller radius
R, yielding a finer dispersion in which coalescence is
decelerated. Finally, due to the better acthesion between the
phases, mechanica! properties will be improved * (23).
*) The only commercial use, to our knowledge, of a PS/PE blend,
is found in so-called 'easy seal/easy peel' packaging systems.
-157-
The model system PS/PE (containing up to 9% of added
(tapered) diblock copolymer) has been extensively studied by
Heikens et al. (24) and Fayt et al. (25-28). When blended on
a two roll mill, very fine dispersions were obtained.
Improvements in mechanica! properties, such as a higher
tensile strength and elangation at break were found in bath
PS rich and LDPE rich blends (25).
An additional third route may be found in the crosslinking
of the dispersed phase by gamma or electron-beam (EB)
irradiation. This can be particularly useful to fixate
fibrous or highly extended non-equilibrium morphologies (29).
Although the results are quite encouraging, different
morphologies may result for blends having identical
composition, when blended on different machinery. The time
scale of processes during the mixing of the polymers A and B
+ block copolymer (e.g. diffusion of the block copolymer to
the interface) is less well understood. The average
residence time on a two roll mill is in the order of 15
minutes, whereas on a corotating twin-screw extruder
residence times are as short as 2 minutes.
As an illustration, the morphology is shown of PS/HDPE
blends with 5% tapered diblock copolymer added to the HOPE,
prepared on a corotating twin-screw extruder, under the
conditions listed inSection 7.2.2. The Figures 7,9 and 7.10
show that the overall length scale of the blend has
decreased considerably, yet the same non-equilibrium
fibrillar morphology (viscosity ratio p ~ 1) as in Figures
7.4 and 7.5 is found.
Figure 7.11 shows the morphology of this blend after
compression moulding. The originally extensional character
of the morphology has disappeared. Threads have been braken
up into droplets. No extreme coarsening of the blend
morphology is observed, campare Figure 7.6.
For understanding ultimate morphologies of blends, it is
important to study the influence of the presence of
-158-
compatibilizer at interfaces with regard to the
agglomeration process in yet more detail.
20/75/5 45/50/5 80/15/5 20/75/5 45/50/5 80/15/5
Figure 7.9 Figure 7.10.
HDPE/PS (reference)
+ 5% PS/PE tapered
diblock copolymer
Figure 7.9. Scanning electron micrographs of microtomed extrudate
surfaces of the system HOPE/PS with and without
compatibilizer, parallel to the direction of extrusion.
Figure 7.10 As Figure 7.9, perpendicular to the direction of extrusion.
20/75/5 30/65/5 45/50/5 80/15/5
Figure 7.11. Morpho1ogy of the system HDPE/PS + 5% tapered diblock
copolymer after compression moulding (15 min. at 200 °C).
-159-
7.5. REFERENCES
1. M. Matsuo and S. Sagaye, in Colleidal and Morphological
Behavier of Block and Graft Copolyrners, Ed. G.E. Molau,
Plenum, New York (1971)
2. J.J. Elmendorp, Ph.D. Thesis, Delft University of
Technology (1986)
3. N. Tokita, Paper 26, ACS Rubber Division Meeting, San
Francisco (1976)
4. D. Heikens and W. Barentsen, Polymer, ~, 69 (1977)
5. G.G.A. Bohm, G.M. Avgeropoulos, C.J. Nelson and F.C.
Weissert, Rubber Chem. Tech., ~, 423 (1977)
6. G.M. Jordhamo, J.A. Mansen and L.H. Sperling, Polym. Eng.
Sci., ZQ, 517 (1986)
7. A.K. van der Vegt and J.J. Elmendorp, Paperpresentedat
the Workshop on New Polymerie Materials, Leuven, Belgium,
October (1986)
8. A.K. Chesters, Proc. Conf. Turbulent two phase flow
systems, Toulouse, France (1988)
9. w.w. Jongepier, internal report Delft University of
Technology (1987)
10. S. Wu, Polymer Interface and Adhesion, Marcel Dekker
Inc., New York and Basel (1982)
11.J.H. Truijen, internal report Eindhoven University of
Technology (1989)
12. J.M.H. Janssen, internal report Eindhoven University of
Technology (1989)
13.M. Xanthos, Polym. Eng. Sci., ~, 1392 (1988)
14. R.W Hohlfeld, Plastics World, August (1985)
15.W.E. Baker and M. Saleem, Polym. Eng. Sci., 22, 1634
(1987)
16. J.G.M. van Gisbergen, internal report Eindhoven
University of Technology (1986)
17.R. Fayt, R. Jéröme and Ph. Teyssié, Polym. Eng. Sci.,22,
328 (1987)
18. s. Bywater, Polym. Eng. Sci.,~, 104 (1984)
-160-
19. G. Maglio and R. Palumbo in: Preprints of the 2nd Joint
Polish-Italian Seminar on Multicomponent Polymerie
Systems, Lodz (1981)
20. J.W. Barlew and D.R. Paul, Polym. Eng. Sci., ~, 985
(1981)
21. G. Riess, G. Hurtrez and P. Bahadur in: Encyclopedia of
Polymer Science and Technology. Vol. 2, Ed. J.I.
Kroschwitz, Wiley-Interscience, New York (1985)
22.R. Jéröme, R. Fayt and T. Ouhadi, Prog. Polym. Sci., 1Q,
87 (1984)
23.D.R. Paul, Ch. 12 in Polymer Blends. Vol. 2, Eds. D.R.
Paul and S. Newman, Academie Press Inc., New York (1978)
24.D. Heikens, N. Hoen, W.M. Barentsen, P.Piet and H. Ladan,
J. Polym. Sci., Polym. Symposia, QZ, 309 (1978)
25.R. Fayt, R. Jéröme and Ph. Teyssié, J. Polym. Sci.,
Polym. Lett. Ed., ~, 79 (1981)
26. ibid., ~. 1269 (1981)
27.R. Fayt, R. Jéröme and Ph. Teyssié, J. Polym. Sci.,
Polym. Phys. Ed., ~, 2209 (1982)
28.R. Fayt, P. Hadjiandreou and Ph. Teyssié, J. Polym. Sci.,
Polym. Chem. Ed., ~, 337 (1985)
29. J.G.M. van Gisbergen, H.E.H. Meijer and P.J. Lemstra,
accepted for publication in Polymer (1989)
-161-
CHAPTER 8
STABILITY OF MORPHOLOGIES, OR THE EXPERIMENTAL
DETERMINATION OF INTERFACIAL TENSION
This Chapter describes interfacial tension-driven
morphological transitions of systems in which molten polymer
threads or layers are present.
Extended (Newtonian) liquid threads suspended in a
(Newtonian) liquid desintegrate into lines of droplets
because of the interfacial tension driven Rayleigh
disturbances (1,2). From the growth rate of these
disturbances, which can well be predicted from Tomotikas
theory (2), the interfacial tension between the thread phase
and the matrix phase can be calculated (3) . Apart from the
influence of compatibilizers on the interfacial tension,
also the stability and breakup of embedded films will be
discussed shortly.
8.1. MEASUREMENT OF INTERFACIAL TENSION VIA BREAKUP OF
THREADS
It is well known that droplets that are highly elongated in
a flow field, exhibit sinuscictal distortions with wavelength
À which inevitably cause breakup after the flow has stopped.
For two-phase systems in which both phases are Newtonian
fluids, the growth rate q of these distortions is directly
related to the interfacial tension cr, see Eq. 6.14 (2).
By simply observing the desintegration process of a liquid
thread in a liquid matrix, this relationship can be used to
determine the interfacial tension between these phases (3).
Chappelear (4) was the first who applied this procedure to
measure the interfacial tension between some polymer melts,
such as polyethylene, polystyrene, polyethyleneterephtalate
and polyamide-6,6. The reported interfacial tensions have
-162-
errors of about 30%. This is mainly due to the fact that
Chappelear extrapolated viscosity data, obtained from
capillary viscosimetry, to the zero-shear rate region. With
the advent of more actvaneed rheological equipment, the value
of the zero-shear viscosity should not cause such
significant errors. In fact, it is the slowness of the
process that explains the successful application of
Tomotikas theory in these systems, as was shown by Elmendorp
(5,6). He found good agreement between the predicted and
experimental growth rate of the distortion, but formulated
no explicit conclusions with regard to the interfacial
tension.
The procedure for measuring interfacial tension outlined
above is generally referred to as the Breaking Thread Method
(7) . In contrast with other (statie) methods for measuring
surface and interfacial tension,such as pendent or spinning
drop techniques (7-10), the Breaking Thread Method does not
require data on the density difference between the two
phases. The method is also suitable for systems consisting
of highly viscous polymers, in which equilibrium drop shapes
are attained only after several hours (9). Finally, it can
be used to obtain information concerning the emulsifying
effect of (newly synthesized) block copolymers (11). Some
typical results wil! be discussed in the next Sections.
8.2. INTERFACIAL TENSION BETWEEN TWO HOMOPOLYMERS
To establish the validity of the method, experiments are
performed with homopolymers of which the interfacial tension
is known from other methods.
8.2.1. MATERIALS
The properties of the materials used are measured on a
Rheometrics mechanica! spectrometer, as listed in Table 8.1.
For all materials in this Table, the loss angle & is close
-163-
to 90°, implicating Newtonian behaviour during the
experiments.
Table 8.1 Zero-shear viscosity (~0 ) and loss angle (6) of the
homopolymers used in the interfacial tension measurements.
Material code T (oC) ~0 (Pa. s) &
HDPE (DSM)
Stamylan 7058 200 1800±3% 90o
230 1115±2% 88°
Stamylan 7359 200 250 90°
LDPE (BP Chem.)
BP33 190 62000 76°*
PS (DOW Chem.)
638 200 1400 90°
230 1000 90o
634 200 11000 90o
PS (BP Chem.)
HH101 190 31200 aso*
PA-6 (BASF)
B3 230 425 90°
8.2.2. EXPERIMENTAL PROCEDURE
Threads of the desired polymer are spun, e.g. from a
Melt-Index apparatus, a single screw extruder or simply from
a molten granule on a hot plate. After having been annealed
during 24 hours at 80 °C, the threads are embedded in the
matrix, which consists of two films of the second polymer,
* From: Refs. 12 and 15
-164-
each with dimensions 10 x 10 x 0.6 mm. The system is placed
under an optical microscope (Zeiss) in a Mettler FP2 hot
stage, and brought to the desired temperature.
At a given position along the thread, the distortien is
photographed at regular intervals in time. As long as the
distortien remains sinusoidal, its amplitude a can be
obtained from the expression:
a = b - a
4
where b the largest thread diameter
a the smallest thread diameter
(8 .1)
By plotting log(2a/Do) against time, the growth rate q of
the disturbance is calculated from the slope of the line
(2,4,7,13):
q = d (!!.n (2a/Dol)
dt
with Do the initial thread diameter
Substituting this value for q in Eq. 6.14 yields an
expression for the interfacial tension o:
0 =
where 11m viscosity of the matrix
llt viscosity of the thread
p viscosity ratio lltll'lm 0('\n,p) a tabulated function (3-6)' see also
Figure 8.1.
(8. 2)
(8. 3)
'\n the dominant wavelength at which breakup
occurs
-165-
A typical plot is given in Figure 8.2. The experimental
wavenumber x = ~Do/À is compared with the theoretical one
(Xm = ~Do/\n), to give an extra control on the experiment.
0.8
E :-..:
• 0.6 E
Cl
'( 0.4 I I
0.2
01-TTnm~~~~~mm~~~~~~~~~~~~ 0 0.001 0.01 0.1
---> p
10 100 1000
Figure 8.1. Growth function Q(Àro,p) and wavenumber Xm = nTiol\n as a
0
~ N
1\ I I I
function of viscosity ratio p. From Ref. 5.
10 20 30 40 50 60 ---> t (s)
Figure 8.2. The relative amplitude (2~/Do) of the distortion on a
PA-6 filament (Do = 55 ~' no = 425 Pa.s) in a PS
matrix <no - 1000 Pa.s). The measurements were
performed at 230 oe. <Xexp = 0.6; Xm- 0.61)
-166-
8.2.3. RESULTS
The interfacial tension between some incompatible
homopolymers, measured via the Breaking Thread Method, are
listed in Table 8.2. Where possible, reference data, taken
from the literature are given.
Table 8.2 Interfacial tension between some incompatible homopolymers.
The ~~ta in the last column are taken from Ref. 7, unless
otherwise stated.
Thread Matrix T
phase phase oe Pa.s Pa.s
p Xth Xexp 0.103 0.103
N.m- 1 N.m- 1
(exp) (Ref. 7)
HDPE PS 200 250 1400 0.179 0.59 0.59 4.9 4.7
PS LDPE 190 31200 62000 0.503 0.60 0.55 3.5 4.9
PA-6 PS 230 425 1000 0.425 0.61 0.60 20.0
HDPE PS 200 1800 1400 1.290 0.56 0.55 4.1 4 . 7
PA-6 HDPE 230 425 1590 0.267 0.59 0.60 13.5 10.7*
On the results shown in Table 8.2, a few remarks can be made
concerning the applicability of the Breaking Thread Method.
To avoid end-pinching (14) or relaxation, see Figure 8.3,
the thread should be highly elongated, and have a constant
diameter . In the present experiments, the ratio of length to
diameter
L/Do > 60. For the same reason, systems with a viscosity
ratio ~tl~ << 1 can be expected to give better results.
* at T - 250 oe (5)
-167-
Figure 8.3. Relaxation and breakup of a HDPE thread (diameter
Do ~ 20 ~) in a PS matrix at 200 °C.
The time needed to perfarm one experiment is dependent on
the viscosities of the polymers, the filament thickness D0 and the interfacial tension, and may range from 50 seconds
(PA-6/PS) to several hours (PS/LDPE) . In systems where bath
polymers are highly viscous, it is therefore desirable to
decrease the initia! thread diameter as much as possible, to
avoid degradation. E.g. a value a= 4.4 ± 0.3 mN.m- 1 was
measured by Elmendorp (15) for the system PS/LDPE (nps =
62000 Pa.s; nLDPE = 31200 Pa.s; T = 190 °C), when starting
with a thread of diameter Do = 5 ~- The experiment took
about 1 hour.
-168-
The optical microscope had to be fitted with a long-distance
(LD-40) objective, in order to cbserve the breakup process.
A straightforward evaluation of the tabulated function Q(À,p)
is only possible if thread breakup is generated by a
distortien having the dominant wavelength Àm· It is
therefore essential to campare the experimental wavenumber
with the theoretical ene. The values have to be the same. If
initial distortions of different wavelengths are not all of
the same amplitude, breakup might occur at some wavelength À
other than the dominant wavelength Àm· In this case, the
complete graphical representation of Q(À,p), coveringa range
of 10-3 < p < 102 and 0.1 < x < 0.9, as given in Figures 1
and 2 of Ref. 4, might be of help.
8.3. INFLUENCE OF BLOCK COPOLYMERS ON INTERFACIAL TENSION
It has been shown by several authors that bleek copolymers
lower the interfacial tension in a blend of two incompatible
polymers (16-18). This has a large influence on the blend
morphology. Rayleigh disturbances will start to grow at a
smaller fiber diameter (o/R is a constant) . Breakup occurs
at the same critical capillary number, so smaller draplets
will result. Simultaneously, coalescence of draplets is
decelerated, due to the immobilization of the interfaces by
the presence of the bleek copolymer (2), and the lower
pressure in the film between the interfaces (19).
Rumscheidt et al. (4) were the first to report on the
influence of surfactants on the breakup of liquid
(Newtonian) threads suspended in a surrounding (Newtonian)
fluid. They observed a non-uniform wavelength À along the
thread, which gave rise to considerable scatter in the
resulting values for the interfacial tension. A
well-controlled experimental procedure, however, makes it
worth while to use the Breaking Thread Methad for measuring
the effect of actding bleek copolymer to the model system
PS/HDPE.
-169-
8.3.1. MATERIALS
Gel Permeation ehromatography (GPe) was performed on
solutions of polystyrene in THF at 40 oe using Waters
~-Styragel columns 105, 104 and 103 A , and on solutions of
high-density polyethylene in 1-2-4 trichlorobenzene at
135 oe using PL-gel columns. See Table 8.3.
Table 8.3 Molecular weights (weight average) Mw and molecular weight
(number average) ~' of the homopolymers used in the
interfacial tension measurements.
Material code
HDI?E (DSM)
Stamylan 7058
PS (DOW Chem.)
638
634
kg/mole kg/mole
100.0
162.1
254.1
30.0
51.2
81.0
3.33
3.17
3.14
Diblock copolymers polystyrene-polyethylene (PS-b-PE) were
obtained via anionic polymerization of
polystyrene-polybutadiene (PS-b-PB) diblock copolymers,
followed by hydragenation of the PB block. All syntheses
were performed by W.J.J. de Vet at the Université
Sart-Tilman, Liege (Belgium) . The main properties of the
diblock copolymers thus obtained are given in Table 8.4. For
details concerning synthesis and characterization, see Ref.
20.
The block copolymer was introduced in the thread phase
(usually polyethylene) via melt-blending, either on a two
roll mill or on the extruder from which the thread was spun.
Rheological characterization was performed on a Rheometrics
Mechanica! Spectrometer, using a cone-and-plate geometry.
Table 8.5 gives the rheological properties of the
homopolymers + diblock copolymers at 200 oe. Also listed are
-170-
data on hornopolymer + SHELL KRATON G 1652. This triblock
copolymer of styrene-ethylene-butadiene-styrene (SEBS) is
cornrnercially available as cornpatibilizing agent.
-Table 8.4 Molecular weight (number average) ~ and rest percentage
Table 8.5
of homopolymer PS after anionic polymerization of the
diblock copolymers PS-b-PB.
Material code
BC-1
BC-2
BC-3
BC-6
BC-10
BC-12
~(PS)
kg/mole
39.4
37.0
44.9
8.5
66.6
78.0
Zero-shear viscosity <l'lo>
~ (PB) % (PS)
kg/mole homopolymer
46.3
96.5
5.5
11.5
46.4
143.0
11
8
19
15
and loss angle (6) at
of the combination homopolymer + block copolymer
used in the interfacial tension measurements.
Material code l'lo (Pa.s) s
HDPE (DSM)
Stamylan 7058+0 . 25% BC-6 2110±2.2% 89 0
+0.5% BC-6 2140±2.5% 89 0
+1.0% BC-6 2270±2.0% 88 °
+5 . 0% BC-6 2540±1.8% 85 °
Stamylan 7058+1.0% BC-1 2350±3.2% 86 °
+1.0% BC-2 2400±4.0% 87 °
+1.0% BC-3 2050±3.0% 85 °
+1. 0% BC-10 2430±2.8% 90 o
+1. 0% BC-12 2590±3.3% 86 °
Stamylan 7058+0.25% KRATON 2275±2.5% 87 °
+0.5% KRATON 2400±3.0% 87 °
+1.0% KRATON 2470±3.0% 87 °
PS (DOW Chem.)
638 +0.2% BC-6 2310±5.5% 88 °
+1.0% BC-6 1840±4.4% 89 0
+0.2% KRATON 1425±5.6% 89 0
+1.0% KRATON 1710±4 . 1 % 89 0
200 oe
-171-
8.3.2. RESULTS
The Figures 8.4 and 8.5 show the dependency of the
interfacial tension on the amount of block copolymer added
to the thread phase.
The wavelengths À at which breakup occurs are indeed less
uniformly distributed along the thread, as reported by
Rumscheidt et al. (3). Therefore, only those measurements 5~------------------------------------~
Ê ...... z .§. 4
c: ..2 ~ 3 Gl -QJ
ö 2 QJ -... Gl -.E A
I I I
• • • • •• ••• .. .... ··
• 0+-------~-----,,------,-------.------~
0 1 2 3 4 6
---> wt. % dlblock-copolymer Figure 8.4 . Dependency of interfacial tension between PS (matrix)
and HOPE (thread) on wt . % diblock copolymer added to
Ê ...... z .§. c: 0 ii c: Gl -QJ ö QJ -... Gl -.E A
I I I
the HOPE. (T = 200 ° C). 6,---------------------------------------ï
4
• 3 '• I • • • .. · • • • 2
i • • •
.. .. ·· . • .... • •
····· ···· ··· ·· ·····
0~------.-------.-------.-------.-------~ 0 1 2 3 4
---> wt.% trlblock-copolymer 6
Figure 8.5. As Figure 8.4, with KRATON G 1652 added to the HOPE.
-172-
are plotted in which the value of the wavelength is closest
to the dominant wavelength. The corresponding wavenumbers
x= ~Do/À are listed in Table 8.6.
Table 8.6 Influence of the actdition of block copolymer on the
interfacial tension of the system PS 638/HDPE 7058
at 200 °C.
Thread
phase
Matrix p
phase Xth. Xexp.
HDPE+0.25% BC-6 PS 638 1.5 111 0.54 0.62 3.4
HDPE+0.50% BC-6 PS 638 1.5 129 0.54 0.56 1.7
HDPE+1.0 % BC-6 PS 638 1.5 71 0.54 1.1
HDPE+2.5 % BC-6 PS 638 1.8 129 0.53 0.51 0.8
1.8 89 0.53 0.52 1.5
1.8 111 0.53 0.49 1.1
HDPE+5.0 % BC-6 PS 638 1.8 169 0.53 0.57 1.2
HDPE+1.0 % BC-1 PS 638 1.7 196
HDPE+1.0% BC-10 PS 638 1.7 116
HDPE+1.0 % BC-12 PS 638 1.4 34
HDPE+.25% KRATON PS 638 1.6 110
HDPE+.5% KRATON PS 638 1.7 90
HDPE+1. % KRATON PS 638 1.8 160
0.54 0.53
0.54 0.54
0.56 0.55
0.54 0.54
0.54 0.55
0.53 0.56
2.2
2.8
2.3
1.0
3.0
2.6
From the results it is clear that the interfacial tension is
considerably reduced by small amounts of block copolymer (~
0.50 wt.%). Actdition of more than 5 wt.% does not cause
significant changes anymore.
At concentrations which are higher than 0.50 wt.%, the block
copolymer molecules form micelles. Consequently, the
interfacial tension reaches a plateau value (Figure 8.4) or
exhibits a local minimum (Figure 8.5).
-173-
An explanation for the latter phenomenon might be that in
this case, all the interfacial impurities are concentrated
in the micelles. Due to this decrease of the concentratien
of impurities at the interface, the interfacial tension will
increase. Block copolymer molecules will be transported from
the bulk phase to the interface, until a new equilibrium is
reached.
8.4. CONTACT ANGLE MEASUREMENTS
It is obvious from the previous Sectiens that the Breaking
Thread Method does not only serve as an illustration of one
characteristic stage of the mixing process. The method is
also quite useful for measuring the interfacial tension in
systems containing two incompatible polymers plus a possible
compatibilizer. However, its dynamic character calls for a
second - preferentially static - method, to check the
validity of the results. To our regret, complications in the
experimental procedure prohibited the direct application of
methods such as spinning drop or pendent drop. Closely
related to the latter is a method in which the interfacial
tension is calculated from the contact angle between a
molten polymer droplet and a molten polymer substrate
(7,21,22). See also Figure 8.6.
;;9Ljdz ds
Figure 8.6. Contact angle of an axisymmetric droplet resting on a liquid
substrate (Pct< p9). After Ref. 7.
-174-
Assume an axisymmetric droplet resting on a liquid
substrate, as shown in Figure 8.6. The density of the
droplet is smaller than that of the substrate. At any point
(x,z) on the curved interface, the pressure P(x,z) amounts
to (7,23)
2cr p b + ilpgz (8. 4)
where er the interfacial tension
b the radius of curvature at the crigin (x,z)=(O,O)
/lp the density difference between the phases
g the constant of gravity
This pressure must be balanced by the capillary pressure
across the interface (24):
p = 0'
1 + -)
R2 (8. 5)
where R1 , 2 the local radii of curvature at (x,z),
parallel and perpendicular to the plane of
the paper
From an infinitesimal element with dimension ds 2 (ds the are
length) (Figure 8.6) it can be seen that
x ds
sine ; R2 de (8. 6)
dx ds.case dz = ds.sine (8. 7)
Substitution of Eqs. 8.4 and 8.6 into Eq. 8.5 yields the
Bashforth-Adams equation (23) :
de
ds +
sine
x
ilpg ---;- ) •z (8. 8)
-175-
This equation can be made dimensionless with respect to the
length scale (~pg/a)1/2:
~pg
c = s s.fc ; B bic z z.fc x x.fc, a
hence
cte sine 2
cts + x B + z
Combining Eqs. 8.7 and 8.10 finally gives implicit
expressions for X,Z and e:
dX co se
cte 2 sine + z -
B x
dZ sine - 2 sine cte
+ z -B x
(8. 9)
(8.10)
(8 .11)
(8.12)
Tabulated solutions of these equations can be found in Ref.
(25) . By measuring x, z and e, it is possible to determine
X, Z and B. The interfacial tension is then calculated from
the relationship (24):
x 2 a = ~pg (X)
z 2 a = tlpg ( Z)
8.4.1. EXPERIMENTAL PROCEDURE
(8.13a)
(8 .13b)
A granule of the less dense polymer is placed on a
compression moulded plate of the denser polymer. The sample
is then placed in a small oven between two hot plates
(temperature T = 200 °C, in an atmosphere of nitrogen), and
is left to itself for about 30 minutes. Upon removal from
the oven, the sample is immediately quenched in water.
-176-
The shape of the droplet at its plane of symmetry is
observed by optical or scanning electron microscopy. The
densities of the polymers at 200 oe are determined with the
aid of a pycnometer (26) . For each phase, it is assumed that
the presence of less than 1 wt.% block copolymer does not
affect its density. The interfacial tension is calculated
from the dimensions x, zand e, as shown in Figure 8.7.
Figure 8.7. The shape of the plane of symmetry of a HOPE droplet.
8.4.2. RESULTS
The results of the contact angle measurements are listed in
Table 8.7. As can be concluded from this Table, the values
correspond quite well to the ones determined by the Breaking
Thread Method, compare Tables 8.2 and 8.6 and Figures 8.4
and 8.5. E.g. consicter the system HDPE 7058+0.5% BC-6 vs. PS
638. For the interfacial tension, a value cr = 1.7 mN.m-1 is
found when using the Breaking Thread Method, whereas from
contact angle measurements cr = 1.4 mN.m- 1 . Inaccuracies are
mainly due to deviations in the measured contact angle, or
to the fact that some droplets do not attain a completely
a xisymmetric shape during the e xperiment. Fina lly, the erro r
in the density of each polymer is duplicated in the value
for the interfacial tension.
-177-
Table 8.7 Contact angle, interfacial tension and related quantities *
for the system PS/HDPE 7058+block copolymer at 200 oe. The
columns on the extreme right show values for cr calculated
from the radius (Eq. 8.13a) and the height (Eq. 8.13b) of
the droplet, respectively.
droplet substrate e x z x z cr.103 cr.1o3
phase phase mm mm N.m-1 N.m- 1
8.13<! 8.131:!
HD 7058 PS 638 73° 2.94 1.52 1.89 0.98 4.8 4.8
HD 7058 PS 638 130° 3.33 1.56 3.88 1.89 1.5 1.4
+.2% BC-6
HD 7058 PS 638 140° 3.20 1.32 3.85 1. 98 1.4 0.9
+.4% BC-6
HD 7058 PS 638 90o 3.70 1.30 4.43 1.51 1.4 1.5
+.6% BC-6
HD 7058 PS 638 125° 3.33 1. 47 4.29 1. 92 1.2 1.2
+.8% BC-6
HD 7058 PS 638 115° 2.63 1.08 4.35 1.84 0.7 0.7
+1% BC-6
8.5. BREAKUP OF MOLTEN POLYMERIC LAYERS - AN ILLUSTRATION
Polymer blends containing layered structures can be obtained
by coextrusion techniques, eventually combined with a
Multiflux or Ross static mixer (27), or via biaxial
stretching of a dispersion of the two polymers (28) . Due to
their geometry, molten polymerie layers with large aspect
ratios (width W >> height h) are expected to be more
resistant to interfacial tension driven distortions than
molten polymer threads. Their surface can only decrease by
relaxation into one or more spheres or by breaking up into a
two-dimensional pattern of draplets (29) .
*Th~ densities at 200 oe: PHDPE 7058 954 kg.m-3
-178-
The stability of thin layers in the molten state can be
studied with the setup of Section 8.2. Thin slices (diameter
about 2 mm, thickness 3 ~) are microtomed from a granule
and embedded between two sheets of the opposite polymer. The
system is then heated up above the highest melting
temperature. Figure 8.8 shows a typical example of the
breakup sequence of such a layer.
c (360 s.) d (600 s.)
Figure 8.8. Sequence of the breakup of a thin (3 ~) PA-6 film
ernbedded in a polypropylene (PP) matrix at 240 °C.
Material is transported from the centre to the edges of the
layer, where small 'fingers' start to grow, which rapidly
desintegrate into droplets via the mechanism described in
this Chapter and in Chapter 6.
-179-
8.6. CONCLUSION
The Breaking Thread Method can be used successfully for
measuring the interfacial tension between incompatible
molten polymers which exhibit Newtonian behaviour as the
shear rate approaches zero. An advantage is the fact that
density data of the polymers need not be known.
It is a dynamic, straightforward simple method which does
not require special equipment. Besides, only small
quantities of the materials are needed. Also the
determination of the interfacial tension via the static
contact angle yields good results, although there is some
scatter. The results from both methods should be compared
with these from spinning drop. Pendent drop proves to
introduce many experimental problems if used for polymer
melts (30).
The stability of (extended) molten polymer threads is
considerably increased by the introduetion of block
copolymers.
8.7. REFERENCES
1. Lord Rayleigh, Proc. Roy. Soc. (Londen),~, 71 (1879)
2. S. Tomotika, Proc. Roy. Soc. (Londen),~, 322 (1935)
3. F.D. Rumscheidt and S.G. Mason, J. Coll. Sci., ~' 260
( 1961)
4. D.C. Chappelear, Polym. Preprints, ~, 363 (1964)
5. J.J. Elmendorp, Ph.D. Thesis Delft University of
Technology (1986)
6. J.J. Elmendorp, Polym. Eng. Sci., ZQ, 418 (1986)
7. S. Wu, Polymer Interface and Adhesion, Marcel Dekker
Inc., New York and Basel (1979)
8. S. Wu, J. Coll. Int. Sci., J..l, 153 (1969)
9. J.J. Elmendorp and G. de Vos, Polym. Eng. Sci., ZQ, 415
(1986)
10. J.J. Elmendorp, M. de Wit and B. Oudhaarlem, Mikroniek,
2...8..Ll, 9 (1988) (in Dutch)
-180-
11.H.E.H. Meijer, P.J. Lemstra and P.H.M. Elemans, Makromol.
Chem., Macromol. Symp., ~, 113 (1988)
12. L.A. Utracki and P. Sammut, Poly,m. Eng. Sci., ~, 1405
(1988)
13. S. Wu, J. Macromol. Sci., Revs. Macromol. Chem., Ç1Q (1),
1 (1974)
14.H.A. Stone and L.G. Leal, J.Fluid Mech., ~, 399 (1989)
15. J.J. Elmendorp, private communication (1988)
16.P. Gaillard, M. Ossenbach-Sauter and G. Riess, Makromol.
Chem., Rapid Comm., ~' 771 (1980)
17. S.H. Anastasiadis and J.T.Koberstein, Polym. Preprints,
28, 24 (1987)
18.H.T. Patterson, K.H. Hu and T.H. Grindstaff, J. Polym.
Sci., ~' 31 (1971)
19 A.K. Chesters, Proc. Conf. Turbulent Two Phase Flow
Systems, Toulouse, France (1988)
20.W.J.J. de Vet, internal report Eindhoven University of
Technology (1986)
21.F. Neumann, Vorlesungen uber die Theorie der
Capillaritat, B.G. Teubner, Leipzig (1894)
22.H.M. Princen in : Surface and Colleid Science. vol. 2,
Ed. E. Matijevic, Wiley-Interscience, New York (1969)
23. S.B. Bashforth and J.C. Adams, An Attempt totest the
Theory of Capillary Action, Cambridge University Press
and Deighton, Bell & Co., Londen (1892)
24. P.S. de Laplace, Mécanique celeste, Suppl. au Xième
Livre, Courier, Paris (1805)
25. S. Hartland and R. Hartley, Axisymmetric Fluid-Liquid
Interfaces, Elsevier, Amsterdam, (1976)
26. J.H. Truijen, internal report Eindhoven University of
Technology (1989)
27.H.H.T.M. Ketels, Ph.D. Thesis Eindhoven University of
Technology (1989)
28.P.J. Lemstra and R. Kirschbaum, Polymer, ZQ, 1372 (1985)
29.N. Dombrowski in : An Album of Fluid Motion, Ed. M. van
Dyke, Stanford Parabalie Press (1982)
30.F. Trompetter (DSM), private comm. (1988)
-181-
CHAPTER 9
COUPLING OF DETAILED AND OVERALL MODELLING
In this Chapter, an attempt is made to coup1e some of the
knowledge from detailed studies on droplet breakup with the
overall rnadelling of continuous mixers. The development
during processing on a corotating twin-screw extruder of the
morphology of a blend is examined. The number of
reorientations in the extruder is of importance. The
expressions describing the time for breakup of molten
polymer threads can be used to predict the resulting
morphology. Besides, the influence of the contraction flow
through the die and the cooling conditions on the final
blend morphology can not be neglected.
9.1. EXAMPLES OF CALCULATIONS ON DISPERSIVE MIXING
When incompatible polymers are blended on a corotating
twin-screw extruder, a fine dispersion can be generated
within a residence time which is in the order of magnitude
of one or two minutes. At first sight, it is surprising that
the enormous increase in the number of particles can be
realized within this short time scale. Consicter for instanee
the example of a 30/70 PS/HDPE blend, starting with one
large granule PS in every gram of HOPE. An average partiele
diameter of 1 ~ in the final blend implies that the number
of particles per cm3 has been increased to roughly 1011 .
These particles originate from the desintegration of highly
extended polymer threads.
The aim of this Section is to show some of the mechanisms
that produce this large reduction in length scale in such a
short time. Calculations are presented concerning (i) the
combined action of affine deformation plus reorientation,
(ii) the highly simplified dispersion process of an isolated
-182-
droplet in a screw extruder, and (ili) the total shear in the
kneading sectien of a corotating twin-screw extruder.
9.1.1. COMBINED AFFINE DEFORMATION AND REORIENTATION
Consicter a Couette flow between two concentric cylinders, in
which a droplet with an initia! diameter of 1 mm is affinely
deformed into a thread with diameter 1 ~· The concentric
cylinders can be part of a screw extruder, with typical
dimensions: diameter D = 25 mm, gap width H = 2.5 mm, and
screw speed N = 200 rpm.
The available shear rate is:
V 'II'DN/60
y z H H
With Eq. 6.2, the required amount of shear can be
calculated:
10-6 10-3 B/a
10-3
(1 + 2 -1/4
y ) 10-3 + y z 106
(9 .1)
(9 .2)
The required time to establish this total shear amounts to:
t = y/y = 104 s, which is almest three hours.
If the same droplet is perfectly reoriented once per
revolution (once per tn seconds), then the decreasein
droplet width (after n-1 reorientations) is given by:
B/a
where n
2 -1/4 n ( (1 + Yn l l
the nuffiber of reorientations
(9 .3)
the amount of shear between two reorientations
the time between two reorientations
-183-
60/N 0.3 s (9.4)
100•0.3 30 (9.5)
The required number of reorientations can be calculated by
combining Eqs. 9.1 - 9.5:
(9.6)
yielding n 5.
This means: the required reduction in droplet size can be
achieved within t = n.tn 1.5 seconds, provided
Ca >> Cacrit' and provided a mixing apparatus which
combines shearing and reorientation.
For equation 9.3 to be valid it is assumed that after each
reorientation, the extended droplet is cut in pieces which
have the size 'B' of the original droplet. In the flow
inside extruders, stretching and folding is the only
mechanism that produces an increase of the interfacial
area, see Section 2.2.1.
It is therefore more likely to expect the droplet to be
folded at each reorientation, followed by stretching, as is
shown in Figure 2.5. In this case, the value of Yn (which is
30 in the present example) only applies during the time
preceding the first reorientation. For long times, Yn = 2,
as in some static mixers.
Substituting Yn = 2 in Eq . 9.6 yields a number of
reorientations n = 18, and hence a mixing time of
t = n.tn z 6 seconds for the required reduction in droplet
size.
9.1.2. DI SPERSION OF AN ISOLATED DROPLET INA SCREW
EXTRUDER
Consicter an extruder, with dimensions:
-184-
diameter D 25 mm
flight clearance ó 0.25 mm
tip width b 2 mm
length L/D 5
channel depth H 2.5 mm
pitch angle Cj' 20°
relative flight width e = b/ (11'DsinC!') :::: 0.1
die diameter d 3mm
and further:
throughput Q 50 g/min (single screw extruder,
Section 9.1.2)
Q 100 g/min (corotating twin-screw extruder,
Section 9.1.3)
screw speed N 200 rpm
density p 800 kg/m3
It is assumed that the material makes one pass through the
nip between screw flight (or barrier flight, as present in
Maddock-type of mixers) and barrel wal!. The material is
then transported towards the die by the drag flow in the
screw channel.
The shear rate over the flight is:
y :::: 'II'DN/60
ó :::: 1000 s - 1 ( 9. 7)
Polystyrene (PS), which has a viscosity nd = 100 Pa.s at
this shear rate, is dispersed in high-density polyethylene
(HOPE) with corresponding viscosity nc = 150 Pa.s. Hence,
the viscosity ratio p = 0.7. The interfacial tension is
about 5 mN.m-1 (1), so for droplets having a radius
R = 1 ~ the local capillary number Ca can be calculated:
Ca n y R ~
(1 30 (9.8)
-185-
This capillary number exceeds Cacrit' required for breakup,
by a factor 40 (Cacrit z 0.7, campare Figure 1.8 in Sectien
1.4). However, the residence time in the clearance is very
short:
B t z 0.015 s.
V
This yields a dimensionless timet*:
t* 0
t Rz0.5, 11c
(9. 6)
(9.7)
which is too short for the droplet to break up. Campare
Figure 6.8, where a dimensionless time t* ~ 9 is required
(at p = 1) for droplet breakup at Ca/Cacrit z 40. Because of
the large capillary number, the droplet will deform
affinely into an extended thread. Its dimensions fellow
from the total shear over the flight: y = y.t z 15, hence
the elangation ratio L/2R becomes (with Eq. 6.1):
L/2R
After passing through the nip, the droplet will be
subjected to much lower shear stresses in the e xtruder
channel. The shear rate becomes (Eq. 9.1): y = 100 -1 s ,
( 9. 8)
and
the capillary number (with a matrix viscosity 11c = 700 Pa.s)
Ca z 14, which is high enough to keep the droplet extended.
In this case, the value Ca/ Cacrit z 14/0.7 = 20. The
dimensionless time for breakup is about 100, see Figure 6.8.
At a throughput Q = 10-6 m3/s, the average residence time
in the extruder channel amounts to
'II'DLH
t Q
z 24 s, hence t* z 170 (9. 9)
The available time is in the same order of magnitude as
the required time, allowing for interfacial tension-driven
disturbances to cause breakup.
-186-
9.1.3. TOTAL SHEAR IN A COROTATING TWIN-SCREW EXTRUDER
The total shear (with and without reorientations) in a
neutral kneading section of a corotating twin-screw
extruder can be calculated. Consicter a kneading section as
given in Figure 9.1.
The shear rate Yf between a kneading flight and the barrel
wall can be calculated to be: Yf z 1000 s-1 (Eq. 9.7).
Figure 9.1. Cross-sectien of a neutral kneading sectien of a
corotating twin-screw extruder.
The shear rate Yr in the region between two flight passages
(in tangential direction) amounts to: Yr = 100 s- 1
(Eq. 9.8) and the residence time tt is estimated to be
'II'D/2 tt = l,21!'DN/60
Assuming that all
reg ion, the total
60 = 0.3 s N
of the material passes
shear Yt in tangential
(9.11)
through the nip
direction becomes:
(9.11)
The average residence time ta in axial direction over the
kneading section* is
2 'II'DLH (1-2e)
Q z 15 s
* the relative flight width e 0.2 here.
(9.12)
-187-
Tagether with the number of (tangential) passes over the tip
of the kneading flight, v,
N v :::: 2 60 tv = 100 (9.13)
this fina1ly yields for the total shear y Yt·v:::: 4500
Therefore, if no reorientation is assumed, a droplet with
initia! diameter 'a' = 3 mm is deformed into a thread with
diameter B:
2 -1/4 -1/2 B = a. ( 1 + y ) :::: a. y :::: 45 J.1ffi (9.14)
Suppose that all of the material is perfectly reoriented
once per revolution of the screw, then the total number of
reorientations over the kneading section is:
N
n = 60 .tv 50 (9.15)
Hence, the amount of shear between two reorientations, Yn'
is given by
1 1 Yn = n ·Y = 50 . 4500 :::: 90 (9.16)
This leads to a drastic, although irrealistic reduction in
droplet diameter:
B a. 2 -n/4 (1 + y )
-25 a. (90) :::: 10-52 m. (9.17)
In reality, the mechanism of reorientation is far less
efficient. A material line passing through the intermeshing
region may be stretched and folded as indicated in Figure
9.2. Suppose that every line is subjected to Yn = 2 (see
Figure 9.2 b) then the calculation of the last paragraph of
Sectien 9.1.1 yields a droplet diameter B:
-188-
B 2 -n/4 a. (1 + 2 ) z 1o-11 m (9.18)
Even this is an overestimation, since the major part of the
material 'escapes' in axial direction by flowing between the
kneading flights. It can be expected that only a minor part
is reoriented. Moreover, from the experiments with the
Couette device there proves to be a limiting value for the
ratio B/a, especially for viscoelastic threads, and at smal!
local radii R (~ B), Rayleigh disturbances wil! start to
occur. Yet these calculations once more underline the
necessity of reorientation in reducing the length scale of a
blend. 2
[SJ[Z] CSJEJ CS]~ [SJ[!]
Figure 9.2. Stretching and folding of a material line upon passing
through the intermeshing region of a corotating twin-screw
extruder.
9.2. BREAKUP OF THREADS
Once a certain length scale is reached, the relatively
slower breakup processes govern the course of the mixing
process. This means: except for threads having a very smal!
diameter, the time scale at which breakup occurs is still
rather large, compared with the overall residence time. This
is shown in Table 9.1, in which initia! distortions of
relative amplitude a 0 /Ro = 10-3 are assumed to cause thread
-189-
breakup in the fastest process, i.e. in a stationary matrix.
With the typical data ~c = 1400 Pa.s and c = 5 mN.m- 1 , the
time for breakup in the absence of flow can be calculated
from Eq. 6.5.
Table 9.1. Time for breakup (tb) for PS filaments in a HDPE matrix
which is at rest. Parameters: filament diameter Do and
viscosity ratio p.
p Do (~) tb (s)
1 10 300
5 150
1 30
0.5 15
0.1 3
0.2 10 100
5 50
1 10
0.5 5
0.1 1
Apart from coalescence phenomena, the resulting draplets
wil! have a diameter which is approximately twice the
diameter of the thread (2).
This simple calculation is valid for polymers exhibiting
Newtonian behaviour as the shear rate approaches zero. E.g.
on a two rol! mil!, this is the case for the time interval
spent by the material on the rolls between two passages
through the nip. In a corotating twin-screw extruder, low
shear rates may be expected in partly filled channels
between two kneading sections (see Chapter 3). In this
respect, also the contraction flow into the die during
stranding and the subsequent cooling conditions may
introduce morphological transitions. See Chapter 7, where
long ultrathin fibers are formed in the strand at a
viscosity ratio p = 1 for a broad range of compositions.
Upon renewed heating, these fibers may either desintegrate
into draplets or coalesce to form a co-continuous
-190-
structure (3) . In the next Section, experiments will be
discussed which illustrate some of these time-dependent
transitions in the model system PS/HDPE.
9.3. EXPERIMENTAL
The polymers used in the experiments are HOPE Stamylan 7058
and 7359 (DSM), and PS 638 (DOW Chemical). The properties of
these materials are given in Sectiens 7.2.1 and 8.2.1.
Blends are made on a Berstorff ZE25 laboratory corotating
twin-screw extruder at a throughput Q = 100 g/min and screw
speed N = 200 rpm. The extruder geometry is shown in Figure
9.1. While eperating the machine, samples can betaken from
the vent port. After exiting the die, the strand is quenched
in water. It is possible to vary the position of the water
bath, while keeping the diameter of the strand constant. The
time between the die and the water level can thus be varied
from 1 second to approximately 50 seconds.
9.4. RESULTS
0 :..: 96g/min
90% HDPE 7058 63!3
200'C
... --~ ~'" ~ q, "'~y~
----.-- ..... ~ ~ -~ -
Figure 9 . 3. Extruder geometry, processing conditions and morphology
while processing HDPE/PS blends, with different cooling
conditions. The average residence times are indicated
where necessary.
-191-
Figu~e 9.3 shows the resulting morphology of a 90/10 HOPE
7058/PS 638 blend (viscosity ratio p = 1), depending on
cooling conditions . When the strand is immediately quenched
in water, an oriented morphology is formed. However, a
droplet-in-matrix structure is found if the strand touches
the water after 6 seconds.
In the kneading section, the material is molten and
subsequently deformed. The total shear generated in this
kneading sectien is calculated to be y = 4500 (see Sectien
9.1.3). This means (Eq. 6.2): a droplet with initia!
diameter 'a' = 3 mm is deformed into a thread with diameter
B = 45 ~, neglecting the effect of reorientations. The size
of the spherical particles that are found in the sample
taken from the vent port is in the order of 1.5 ~· This
clearly indicates the presence of reorientations in the
kneading section according to the mechanism proposed in
Sectien 9.1.3.
The draplets are extended again in the contraction flow
through the conical die. A theoretica! value for the maximum
elangation ratio À follows from the surface ratio of the
rectangular screw channel and the die. With typica1 data as
in Sectien 9.1.2, À becomes:
3 Y.~Dsin~(1-2e)
~/4 d2 z 25 (9.19)
For the blend morphology in the experiments, this means:
Droplets with a diameter 'a' = 1.5 ~are present in the
extruder channels preceding the die. Upon passing the die,
these draplets are extended into threads having a diameter B
2 -1/4 = a. (1 + À ) = 0.3 ~· Cooling conditions and coalescence
wil! determine the final features of the morphology.
The difference in resulting morphology can be explained by
using the breakup times of Table 9.1. When the strand is
immediately quenched upon leaving the die, threads are
-192-
visib1e (diameter ~ 0.3 ~) having a breakup time of about 9
seconds, see Table 9.1. When a time of about 6 seconds is
allowed between the die and the water bath, the same strand
exhibits a morphology which consists of finely dispersed
droplets (diameter ~ 0.6 ~). This is in fair agreement with
the predicted time for breakup.
If the volume fraction is increased, the morphology of the
blend contains threads with a larger diameter due to
coalescence (4). Consequently, larger breakup times {up to
30 seconds for threads of 1 ~) are required for the
transition of these threads into droplets. Figures 9.4 and
9.5 show the morpbologies of HDPE/PS blends with different
cooling conditions. The breakup times, as predicted in Table
9.1, are confirmed. The 50/50 HDPE/PS blend, however, still
forms a highly oriented morphology, which transforms into a
droplet-in-matrix structure after annealing for 10 minutes.
Apparently, the breakup of the threads is decelerated by the
presence of neighboring threads.
10/90 20/80 30/70 40/60 50/50
%PS/% HOPE
Figure 9.3. Scanning electron micrographs of the microtomed
surfaces of PS/HDPE blends, parallel to the direction
of extrusion. The time between the die and the water
bath is 7 seconds.
-193-
20/80 30/70 40/60 50/50 %PS/% HOPE
Figure 9.4. As Figure 9.3, but now with a time of 50 seconds
between die and water bath.
9.5. CONCLUSIONS
From the calculations and experiments, the following
conclusions concerning the coupling of detailed and overall
rnadelling can be stated.
The complicated flow field inside extruders generate
deformation processes having an inherently transient
character. (Results of studies concerning deformation under
equilibrium and steady conditions have only limited
practical value) . Upon passing through the high shear zone
between flight tip and barrel wall, the draplets are
stepwise subjected to a supercritical flow , i.e.
Ca>>Cacrit' and are deformed into threads with a finite
elangation ratio L/almax' limited by the interfacial tension
and the total deformation y. Upon leaving the nip, the
extended dropiets may either return to their original shape
or desintegrate into droplets.
Even in kneading elements in corotating twin-screw
extruders, only a minor part of the material passes through
the nip, which repesents the high-shear zone, and
-194-
reorientation is not perfect. This can be abducted from the
ca1culations, but can also be seen in flow visualization
experiments in Plexiglas-walled extruders. It is therefore
worthwhile to study the flow inside these geometries in yet
more detail (5) . The emphasis should be on the process of
reorientation upon passing the material from one screw to
the other, and on the the material flowing sideways in axial
direction, thus avoiding the high-shear zones. Perhaps this
can give an optimum kneading flight width and clearance &, as in the Tactmor model for internal mixers (6).
Experiments on deformation and breakup of droplets in
various combinations of shear flow and elongational flow (as
present in most extruder geometries) have been reported by
Bentley et al. (7,8). Systems with viscosity ratios p > 1,
often encountered in the rubber toughening of glassy
polymers, can thus be investigated.
Systems with viscoelastic fluids have been investigated
elsewhere (9-13). The modelling of coalescence in polymer
systems is still in its initial stage, although a few
successful attempts have been made (12,15,16).
9.6. REFERENCES
1. S. Wu, Polymer Interface and Adhesion, Marcel Dekker, New
York and Basel (1979)
2. s. Tomotika, Proc. Roy. Soc. (London), ~. 302 (1936)
3. J.J. Elmendorp, in Mixing in Polymer Processing, Ed. C.
Rauwendaal, Marcel Dekker, New York and Basel, to appear
4. W.M. Barentsen, Ph.D. Thesis, Eindhoven University of
Technology (1972)
5. W. Szydlowski and J.L. White, J. Non-Newtonian Fluid
~eh., ~. 29 (1988)
6. I. Manas-Zloczower, A. Nir and z. Tadmor, Rubber Chem.
Techn., ~. 1250 (1982)
-195-
7. B.J. Bentley and L.G. Leal, J. Fluid Mech., lQl, 241
(1986)
8. H.A. Stone, B.J. Bentley and L.G. Leal, ibid., ~, 131
(1986)
9. R.W. Flumerfelt, Ind. Eng. Chem. Fund., ~, 312 (1972)
10.W.K. Lee, K.L. Yu and R.W. Flumerfelt, Int. J. Mult.
Flow, ]_, 385 (1981)
11.S.L. Goren andM. Gottlieb, J. FluidMech., 12..Q., 245
(1989)
12. J.J. Elmendorp, Ph.D. Thesis Delft University of
Technology (1986)
13.C. van der Reijden-Stolk, Ph.D . Thesis Delft University
of Technology (1989)
14.R, de Bruin, Ph.D. Thesis Eindhoven University of
Technology (1989)
15.B.K. Chi and L.G. Leal, J. Fluid Mech., 2Ql, 123 (1989)
16.A.K. Chesters, Proc. Conf. Turbulent Two Phase Flow
Systems, Toulouse, France (1988)
-196-
SUMMARY
Mechanica! blending of two (or more) incompatible polymers
offers interesting routes towards new materials with
specific properties. Since desired properties of a blend
strongly depend on the resulting morphology, it is necessary
to understand the parameters governing the blending process.
Two main areas are of concern, namely the modelling of
compounding equipment and the studying of the dispersion
processes. This thesis addresses topics in both these areas.
A simple computational model is proposed for the two
important continuous mixers: The corotating twin-screw
extruder and the Co-kneader. By distinguishing between
extruder channels that have a degree of fill f < 1 and these
that are completely filled, it is possible to calculate
locally filled lengths, power, specific energy and
temperature rise. With regard to mixing, the shear rate,
shear stress, residence time and the number of
reorientations can be determined. Because of the
comparatively simple geometry in single screw extruders with
segmented screw design, including mixing sections, the model
can be used to analyse these extruders as well.
Detailed studies of deformation and breakup processes of
droplets in well-defined flow fields can be incorporated in
the practical model derived above if the time-effects during
the different stages are taken into account. The time
dependent deformation of Newtonian droplets into extended
threads is studied by means of a Plexiglas-walled
Couette-apparatus . When the shear rate is slowly increased,
allowing for a nearly equilibrium deformation, the
dependenee of the critica! capillary number Ca (ratio of
viscous and interfacial forces) on the viscosity ratio, as
reported in the literature, is reproduced.
-197-
However,in transient flows at Ca>>Cacrit' draplets are
deformed into long slender bocties which remain extended
until the shear has stopped.
Extended polymer threads suspended in another polymer matrix
desintegrate into arrays of draplets driven by Rayleigh
disturbances due to the interfacial tension. This process
can be studied using an optical microscope. From the growth
rate of the disturbances, the interfacial tension can be
determined. The time for breakup is greatly increased if a
suitable compatibilizer, acting as an interfacial agent, is
introduced in the system. Connected to this is the stability
of polymer layers in a matrix. Although seemingly stable,
these layers also break up, in a way similar to that of
polymer threads in a matrix.
Several non-equilibrium morphologies as obtained with blends
of the model system polyethylene/polystyrene are discussed.
These blends are made on a corotating twin-screw extruder.
By varying the viscosity ratio and composition, an
interesting spectrum of non-equilibrium morphologies is
obtained. In general, these morphologies are present in
blends of incompatible polymers (e.g polystyrene/polyamide-6
or polypropylene/ethylene-vinylalcohol copolymer) . Due to
coalescence, no stable dispersions are found when the volume
fraction is increased. Co-continuous morphologies are formed
upon blending or after compression moulding.
The rate-determining step of the coalescence process can be
investigated by observing the thinning behaviour of the
layer between two colliding spheres, or between a sphere
approaching a plane of the same polymer. A relatively simple
model can give insight into the dynamics of the thinning
process.
When block copolymers are introduced in the model system
polyethylene/polystyrene, more finely scaled morphologies
are obtained. The non-equilibrium character, however,
remains present.
To study the coupling of overall and detailed modelling,
-198-
the development of the morphologies of blends is followed
during the mixing process. It can be concluded that in the
nip of a kneading element in corotating twin-screw
extruders, which represents the high-shear zone, the
critical capillary number is exceeded by at least one order
of magnitude, resulting consequently in affine deformation.
However, the local residence time is far toa short for
breakup to occur. A second conclusion, drawn from the
diameter of the droplets, is that only a minor part of the
material passes through the nip, and reorientation is nat
complete.
-199-
SAMENVATTING
Het mechanisch mengen van twee (of meer) polymeren die in
thermodynamische zin niet mengbaar zijn biedt interessante
routes om te komen tot nieuwe materialen met specifieke
eigenschappen. Omdat de gewenste eigenschappen van een
mengsel sterk afhankelijk zijn van de resulterende
morfologie is het nodig de parameters die het mengproces
beheersen te begrijpen. Dit proefschrift beschrijft
onderzoek aan onderwerpen uit de twee belangrijkste
aandachtsgebieden: modelvorming van kompoundeerapparatuur en
bestudering van het dispergeerproces.
Een eenvoudig rekenmodel wordt gepresenteerd voor twee
belangrijke kontinue mengers: de meedraaiende
dubbelschroefsextruder en de Kc-kneder. Door onderscheid te
maken tussen extruderkanalen die een vulgraad 'f' hebben ten
opzichte van geheel gevulde, is het mogelijk om lokaal de
gevulde lengte, vermogen, specifieke energie en
temperatuurstijging te berekenen. Ten behoeve van de
mengwerking kunnen worden bepaald: de afschuifsnelheid,
afschuifspanning, verblijftijd en het aantal herorientaties.
Vanwege de minder ingewikkelde geometrie in
enkelschroefsextruders, die voorzien zijn van een
opbouwschroef (inklusief mengsekties), kan het model ook
gebruikt worden om deze machines te analyseren.
Gedetailleerde studies naar de deformatie- en
opbreekprocessen van druppels in goed gedefinieerde
stromingsvelden kunnen in het hierboven beschreven model
worden ingebracht mits de tijdseffekten gedurende de
verschillende stadia bekend zijn. De tijdsafhankelijke
deformatie van Newtonse druppels tot lange draden is
bestudeerd in een Couette-apparaat, voorzien van een
plexiglazen wand. Bij een zodanig langzaam opvoeren van de
afschuifsnelheid, dat de druppel steeds een
evenwichtsdeformatie kan aannemen, kan het uit de literatuur
-200-
bekende verloop van het kritische kapillairgetal Cacrit (de
verhouding van viskeuze krachten en grensvlakkrachten bij
breuk) als funktie van de viskositeitsverhouding worden
gereproduceerd. In tijdsafhankelijke stromingen bij Ca >>
Cacrit worden de druppels gedeformeerd tot lange draden die
uitgerekt blijven totdat de afschuiving gestopt is.
Gesmolten polymere draden, gesuspendeerd in een polymere
matrix veranderen in rijen druppels ten gevolge van Rayleigh
verstoringen onder invloed van de grensvlakspanning. Dit
proces kan worden bestudeerd onder een lichtmikroskoop. De
grensvlakspanning kan worden bepaald uit de aangroeisnelheid
van de verstoringen. De opbreektijd (tijd tot breuk) wordt
sterk vergroot door het introduceren van een geschikte
compatibilizer, die als grensvlakaktieve stof fungeert in
het systeem. Gerelateerd hieraan is de stabiliteit van
gesmolten polymere lagen in een matrix. Alhoewel
ogenschijnlijk stabiel, breken deze lagen op eenzelfde wijze
op als polymere draden in een matrix.
Verschillende niet-evenwichtsmorfologieën, verkregen met
mengsels van het modelsysteem polyetheen/polystyreen worden
besproken. Deze mengsels zijn gemaakt op een meedraaiende
dubbelschroefsextruder. Door de viskositeitsverhouding en
samenstelling van de mengsels te variëren ontstaat een
interessant spektrum van niet-evenwichtsmorfologieën. In het
algemeen zijn deze morfologieën aanwezig in mengsels van
niet mengbare polymeren (bijvoorbeeld polystyreen/nylon-6 of
etheen-vinylalkohol kopolymeer) . Bij grotere volumefrakties
zijn de gevormde dispersies niet stabiel als gevolg van
koalescentie. De snelheidsbepalende stap van het
koalescentieproces kan worden onderzocht door de afname van
de filmdikte tussen twee elkaar naderende grensvlakken te
volgen. Een relatief eenvoudig model kan inzicht geven in
de tijdschaal van dit proces.
Door de introduktie van blokkopolymeren die fungeren als
compatibilizer in het modelsysteem polyetheen/polystyreen
ontstaan veel fijnere morfologieën. Het
niet-evenwichtskarakter blijft echter gehandhaafd.
-201-
De koppeling tussen de globale en de gedetailleerde
modellering is bestudeerd door het ontstaan te volgen van
morfologieën van blends tijdens het mengproces. Een eerste
konklusie is dat in de spleet van een kneeddeel in
meedraaiende dubbelschroefsextruders (een gebied van hoge
afschuiving) het kritische kapillairgetal met tenminste een
grootte-orde wordt overschreden. Dit resulteert in affiene
deformatie. De lokale verblijftijd is echter veel te kort om
opbreken te laten plaatsvinden. Een tweede konklusie, die
volgt uit de uiteindelijke diameter van de druppels is dat
slechts een klein gedeelte van het materiaal de zone van
hoge afschuiving passeert. Volledige herorientatie vindt
niet plaats.
NOMENCLATURE
A constant in Eq. 7.2
A
B
(interfacial) surface
drop width
B coefficient in Eq. 5.22
Br Brinkman number
-202-
~ Cauchy-Green deformation tensor
ca capillary number
D deformation
D
r. F
diameter
deformation gradient
force
F (cumulative) residence time distribution
Fx correction factors (F0 , F0 c, Fp, Fpc, FN)
Gz Graetz number
H
L
Lc M
N
p
p
p
Q
(local) channel depth
(channel) length
centerline distance
molecular weight
screw speed
power
pressure
coefficient in Eq. 5.22
throughput
R radius of curvature
R
s T
T
V
w a
b
b
c
radius
area
torque
temperature
circumferential speed
channel width
droplet diameter
radius of curvature
temperature coefficient of viscosity
specific heat
c constant in Eq. 6.7
m
m
N
m
m
m
Pa
m
m
m
-203-
d diameter
e relative flight width
f degree of fill
constant of gravity
film thickness
gap height
Boltzmann constant
(flow) resistance factor
displacement factor
relative length
(axial) length
m orientation vector
n number of flights
n power law index
n number of reorientations
nr number of reorientations
p
p
p
q
r
t
t
u
V
x
z
pressure
viscosity ratio
number of pins in Co-kneader
growth rate of a distortien
relative remaining flight length
time
(average) residence time
velocity
velocity
position vector
coordinate in transverse direction
coordinate in down-channel direction
Greek symbols
ll = difference
Q(À,p) tabulated growth function
dimensionless parameter in Eq. 6.9
distortien amplitude
ex aspect ratio
<ex> = average heat transfer coefficient
m
m
m
m
Eq. 1.2
Pa
s
m
m
m
m
y y
s s
shear rate
total shear
(flight) clearance
loss angle
c efficiency parameter
-204-
c relative void space between aggregates
n viscosity
e orientation angle
e circumferential angle
heat conduction coefficient
À elongation ratio
wavelength
length stretch
viscosity
v constant in Eq. 6. 7
p
0
0
density
interfacial tension
Liapunov exponent
shear stress
~ orientation angle
~ pitch angle
x wavenumber
x shape factor
Subscripts
D drag
DC drag, (due to) curvature
L leakage (over flights)
LD leakage (through flights)
N number of flights p pressure
PC pressure, (due to) curvature
a average
a axial
a,b,c =part a, b, c
b = breakup
m
Pa.s
m
Eq. 2. 6
Pa.s
kg.m-3
N.m- 1
Eq . 2. 7
N.m- 2
Fo
Foc
QL
QLD
FN
Fp
Fpc
ta
Qa, ka
La,Q.a
tb
c continuous phase
c cohesive
c channel
d drag
d droplet
d disperse phase
f flight
f flight
h hydrodynamic
h hole
i intermeshing region
2. local
m dominant
m metered
m ma ss
m melt
n between two reorientations
0 reference
0 initial
p pressure
r residence
r region between two flights
s substrate
sp specific
t tot al
t tangential
t tot al
x,z= x, z direction
w wall
Superscripts
e energy
es specific energy
g shear ra te
gs tot al shear
h channel depth
-205-
'lcrlf'c Fe (Eq. 2.9)
Ec,J.lc
Qd
Pct
llctrlf'ct
Ef,Tf,Ff,J.lf,Lf
'Yt Fh (Eq. 2.8)
Lh
Fi
f2.
Àmr Xm
Qm
Qm ( Eq. 3 . 3 8)
Tm
Yn,tn
J.lo
ao, 90, Ao
Qp
tr
Yr Ps
Esp
kt
tt
Ft
-206-
~ screw length
m torque
n screw speed
q throughput
s specific surface
t mean residence time
T transponed f.T
* dimensionless u*
average T
-207-
CURRICULUM VITAE
De schrijver van dit proefschrift werd op 8 maart 1960 in
Oss geboren. Hij behaalde het diploma Gymnasium-~ in 1978
aan het Titus Brandsma Lyceum te Oss. In datzelfde jaar
begon hij met de studie Werktuigbouwkunde aan de Technische
Hogeschool Twente (die inmiddels Universiteit Twente heet)
in Enschede. Het kandidaatsexamen werd afgelegd in september
1982. Het doktoraalexamen werd behaald op 29 mei 1985 met
als afstudeerrichting Kunststoftechnologie (prof.ir. J.F.
Ingen Housz) .
Vanaf september 1985 tot september 1989 was hij werkzaam als
wetenschappelijk medewerker aan de Technische Universiteit
Eindhoven binnen de vakgroep Polymeerchemie en
Kunststoftechnologie van de fakulteit Scheikundige
Technologie.
Stellingen.
1. Het feit dat uit het model van Manas et al. blijkt dat de
fraktie ongedispergeerde roetagglomeraten afhankelijk is
van de diameter van de gebruikte kneders, duidt erop dat
deze slecht ontworpen zijn.
I. Manas-Zloczower, A. Nir and z. Tadrnor, Rubber Chemistry and
Technology, ~. 1250 (1982)
2. Het associëren van een negatieve spoed met een negatieve
spoedhoek in dubbelschroefsextruders is onjuist, daar dit
leidt tot negatieve waarden voor kanaalsnelheid en
-breedte.
Dit proefschrift, hoofdstuk 3
3. De nauwkeurigheid van de meeste statische methoden ter
bepaling van de grensvlakspanning tussen gesmolten
polymeren staat of valt met betrouwbare
dichtheidsgegevens. Hoewel principieel beter, is het
rechtstreeks meten van het dichtheidsverschil tussen twee
gesmolten polymeren praktisch moeilijk te realiseren.
S. Wu, Polymer Interface and Adbesion, Marcel Dekker Inc.,New York
and Basel (1979)
4. De bepaling van de grensvlakspanning uit de relatie van
Taylor tussen deformatie D en afschuifsnelheid y biedt
twee voordelen: (i) de dichtheden hoeven niet bekend te
zijn en (ii) het grensvlak wordt voortdurend ververst.
G.I. Taylor, Proc. Roy. Soc. (Londen), ~. 501 (1934)
H.P. Grace, Chem. Eng. Comm., 14, 225 (1983)
5. De semi-empirische relatie van Yinxi et al. om de ligging
van het ko-kontinue gebied in een tweefasen polymeer
mengsel te beschrijven heeft geen praktische waarde.
L. Yinxi, K. Kubota, S. Amano en Y. Oyanagi, paper 13-22 op de Se PPS
Meeting, Kyoto, 1989.
6. De door Ng en Erwin beschreven exponentiële afname van de
laagdikte in gekombineerde afschuiving en heroriëntatie
betreft het optimale geval, waarbij de lengte van een
deeltje na heroriëntatie gelijk is aan de beginlengte.
Zelfs in zeer goed ontworpen dynamische mengers is dit
moeilijk te realiseren en verloopt het proces van
laagdikteverkleining veel langzamer.
K.Y. Ng en L. Erwin, Polymer Engineering and Science, 21, 4 (1981)
Dit proefschrift, hoofdstuk 9
7. Gezien het feit dat de meeste Nederlanders pas om 'het
milieu' gaan geven wanneer de vervuiling ervan hun
persoonlijke levenssfeer raakt, verdient het aanbeveling
het drinkwater op ongeregelde tijdstippen bruin te
kleuren.
8. Een onderbelicht aspekt van de beoogde verschuiving in
het privé-vervoer van auto naar fiets is het feit dat
ruim 40% van alle Nederlandse autoritten korter is dan
vijf kilometer.
Tweede Struktuurschema Verkeer en Vervoer, deel a: Beleidsvoornemen,
(Handelingen der Staten Generaal: bijlagen Tweede Kamer nr. 20992),
p. 78, Staatsuitgeverij (1988)
9.Een betrouwbare grootheid voor de stand van de ekonomie
is de dagelijkse hoeveelheid ongevraagde reklame via
brievenbus of telefax.
lD.Wanneer het jaarlijkse aantal vliegtuigongelukken wordt
uitgezet tegen het totale aantal reizigervlieguren in
plaats van reizigerkilometers, blijkt dat vliegen niet
veel veiliger is dan autorijden.
H. Antonisen, Intermediair, ~, 25 (1989)