Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
@i; PROYECTOCONACYT
Contributors
A. Berge
D. C. Blackley
A. S. Dunn
T. Ellingsen
Carlton G. Force
Robert G. Gilbert
A. E. Hamielec
F. K. Hansen
A. A. Khan
Gottfried Lichti
J. F. MacGregor
P. C. M0rk
Donald H. Napper
Mamoru Nomura
R. H. Ottewill
Gary W. Poehlein
Vivian T. Stannett
J. Ugelst~d
V. 1. Yeliseyeva
'.
EMULSION
POLYMERIZATION
Editad by
IRJA PIIRMA
Institute of Polymer Science
The Unversity of Akron
Akron, Oha
ACADEMIC
PRESS
----
7DX
LTD.
in Publication
.
Data
Emulsion polymerization.
Includes bibliographies
and index.
l. Emulsion polymerization.
l. Piirma,
QD)B2.E48E48
660.2'8448
ISBN 0-12-556420-1
81-17626
AACR2
9 8 7 6 S 4 3 2 J
Irja,
Contents
Contributors
Preface
ix
xi
1
2
6
8
14
16
17
19
22
26
27
28
31
35
36
39
40
42
45
47
47
1.
Introduction
Theory
51
54
56
63
73
82
86'
91
Contents
vi
94
115
141
142
143
Free-Radical Polymerization
D. C. Blackley
1. Introduction
11. Reaction Model Assumed
111. The Time-Dependent Smith-Ewan Differential Difference Equations:
Methods Available for Their Solution
IV. Solution for the Steady State
V. Solutions for the Nonsteady State
VI. Predictions for Molecular Weight Distribution and Locus-Size
Distribution
VII. Theory for Generation of Radicals in Pairs within Loci
List of Symbols
References
146
149
156
164
167
183
185
187
189
1. Introduction
11. Polymerization Rate Equations Involving Free-Radical Desorption
111. Derivation of Rate Coefficient for Radical Desorption from Panicles
IV. Effect of Free-Radical
Desorption on the Kinetics of Emulsion
Polymerization
List of Symbols
References
191
192
199
210
217
219
221
224
229
230
236
237
243
Contents
vii
7 Polyrnerization
of Polar Monorners
V. l. Yeliseyeva
1. Introduction
11. Interface Characteristics of Polymeric Dispersions
111. Relationship between Emulsifier Adsorption and the Difference in the
Boundary-Phase Polarity
IV. Mechanism of Particle Generation
V. Colloidal Behavior of Polymerization Systems
VI. Kinetics of Emulsifier Adsorption
VII. Mechanism of Formation and Structure of Particles
VIII. Polymerization Kinetics
IX. Relationship
between
Polymerization
Kinetics and Adsorption
Characteristic of the Interface
Nomenclature
References
247
249
250
257
261
268
27q
278
283
286
287
289
291
300
312
313
314
316
Introduction
Factors in Adhesion
Bonding Applications
Construction Applications
Rubber Goods
Properties of Various Latexes
References
Design, Oneration,
and Control
A. E. Hamie/ec
and J. F. MacGregor
1. Introduction
11. Batch Reactors
111. Continuous Stirred-Tank Reactors: Steady-State Dperation
IV. ContinuQus Stirred-Tank Reactors: Dynamic Behavior
V. Dn-Une Control of Continuous Latex Reactors
VI. Summary
Nomenclature
References
10
319
320
333
339
345
351
351
353
1. Introduction
. 11.
357
361
367
375
378
381
381
Contents
viii
11
of Monomer
Emulsions
383
384
392
396
401
408
411
412
lndex
415
418
433
436
437
447
451
Contributors
Numbers in parentheses indicate the pages on which the authors' contributions begin.
Contributors
Contributors
1m
Preface
Emulsion polymerization has been a very successful industrial
process for four decades. In recent years it has undergone revitalization:
while some old factories are closing, new and much more sophisticated
ones using emulsion polymerization are sprouting.
Historically, in an attempt to produce synthetic rubber, the polymerizations involving the use of an aqueous emulsifier solution resulted
in a product that in physical appearance resembled nature's latex. It also
gave a greatly improved synthetic rubber over that produced previously by
other processes, e.g., the sodium-butadiene process. This industrial success
was subsequently followed by a more theoretical approach to the problem
during the 1940s that resulted in the publication of the first scientific papers
in this field. Even then, since emulsion polymerization had been established
first as an industrial process, the theories proposed by the scientists were
primarily concerned with trying to put the industrial observations into a
general and workable scientific framework. Unfortunately, the emulsion
polymerization of styrene and the copolymerization of styrene-butadiene
fit beautifully into a simple kinetic scheme proposed by Smith and Ewart,
and during the two decades that followe9 almost all research efforts in the
field attempted to make the other monomers fit into this framework.
We now know that emulsion polymerization is not just another polymer
synthesis method and that the complexity of the interactions, whether
chemical or physical, must be considered before any control is possible
over the outcome of the reaction. The creation and nucleation of particles,
for example, is not necessarily and simply explained by the presence or
or absence of micelles, but needs the understanding of interactions of all the
ingredients present. Variables such as hydrophilic and hydrophobic associations or repulsions, polarity of the monomers, chemical structure of the
surfactants, have to be taken into account.
Research in the field is flourishing all over the world, and although numerous papers have been published and collections of papers have appeared
recently, they have the disadvantage of presenting fragments of the subject
and never the total picture. This book presents a collection of chapters, each
written by scientists in their fields of expertise. These experts come from all
over the world and thus sometimes represent different viewpoints of the same
subject. It is the hope ofthe editor and the contributors that we have been successful in presentirig a total picture of the current understanding ofthe subject.
xi
'"
1
xii
Preface
1 would like to thank Silvia Dolson for providing her time and talents to
assist in the book's cover designo
1
!1
S
~
-
1
The Stability and Instabiliy
01 Polymer Latices
R. H. Ottewill
1. Introduction .
11. The Nature of Polymer Latex Particles
111.
IV.
V.
VI.
VII.
VIII.
IX.
X.
XI.
XII.
XIII.
XIV.
xv.
XVI.
XVII.
XVIII.
XIX.
XX.
l.
1
2
6
8
14
16
17
19
22
26
27
28
31
35
36
39
40
42
45
47
47
Introduction
Over the last two decades work on the formation and properties
of polymer latices has developed extensively, and a very substantial amount
of work has been devoted to the study of the processes of formation of
polymer particles in a latex and to the characterization of the particles once
formed (see for example, Fitch 1980). It is now generalIy recognized that in
1
EMULSION POLYMERIZATlON
Copyright @ 1982 by Academic Press. Inc.
AIl rights of reproduction in any form reserved.
ISBN 0-12.556420-1
R. H. Ottewill
the majority of latices, the particles are within the size range 1 nm to 1 pm,
which designates them as colloidal, and the name "polymer colloids" is
becoming commonly used to describe this type of system. Consequently, the
entire field of polymer latex manufacture and utilization is very dependent
on an understanding of the basic principIes of colloid science in the widest
sense whether it is to stabilize or coagulate the particles or to control the
rheological properties of latex formation.
In using the word stability in the colloidal sense we understand that the
particles in the dispersion will remain for long periods of time, often years,
dispersed as single entities in Brownian motion. The size range 1 nm to
1 pm is chosen to exclude at the lower end of the range small single ion s and
molecules and at the higher end to exclude particles that settle under the
influence of gravity and do not remain dispersed by Brownian motion.
These are definitions of convenience rather than rigor and in considering
polymer latices we shall frequently need to consider particles with a size.
greater than 1 pm; in this case the colloidal arguments can be maintained
but, in addition, the effects of gravity have to be included. When the la,tex
loses its stability because the particles aggregate under the influence of
chemical additives or mechanical action, the terms coagulation and flocculation are applied (see later). By the term rheology we describe the flow
and deformation properties of the system, Le., whether it is viscous (either
Newtonian or non-Newtonian) or elastic or whether it possesses both these
properties and is viscoelastic.
ll.
I
j
j
I
j
j!
In aqueous-based emulsion polymerizations using water-soluble initiators, the surface groupings formed are frequently determined by the
nature of the initiator used (Ottewill et al., 1967; van den HuI et al., 1970;
Goodwin et al., 1973) and the following have been reported:
from hydrogen peroxide, persulfate,
bisazocyanopentanoic
acid
Weak acid
Strong acid
-0-S03
or
from persulfate
CH3 NH2
Base
Nonionic
,f'
-C-C~,
I
NH2
CH2
}
-OH
from azobisisobutyramidine
of
In addition, latex particles with mixed anionic and cationic groups on the
surface can be prepared (Bolt, 1978).
In an ionizing medium of high relative permittivity (e.g., water) the acidic
and basic groupings exist in the ionized form, depending on their pKa and
pKb values and the pH, and consequently the surface of the particle
becomes electrically charged. In addition the adsorption of other ionic
species, such as surface-active ions, can also contribute to the surface
charge. In physical terms the water is a good solvent for the ions and poor
for the latex particle; that is, most of the polymers used for latex preparation are totally insoluble in water. A schematic illustration of this situation
is shown in Fig. 1 where the particle surface is shown to be that of a smooth
sphere with the charges evenly distributed over the spherical surface. The
condition of electroneutrality is maintained by balancing the charge on the
latex surface by the charges on small ions of opposite sign in the solution
phase (counterions). This forms the so-called electrical double layer in
which an equilibrium is set up between electrostatic forces and diffusion
forces. As a consequence of its surface charge, the lat.ex particle surface has
an electrostatic surface potential rjlswhich can be either positive or negative,
depending on the nature of the surface groupings, relative to 'earth. This
potential falls off exponentially with distance from the surface of the sphere
according to the equation
rjlr= rjls(a/r)exp[
K(r - a)]
(1)
where rjlris the surface potential at a distance (r - a) from the surface of the'
sphere where a is the radius of the sphere. K is defined by
(2)
K2 = (8ne2NAl1000ekT)I
R. H. Ottewill
- -+ - -
,,/
o
+
/
I
I
"-
+ "-
+ \
+
\
\
I
I
++
\
\
+
"-
"-
-- +
.,//
1.
"-
\
.\
\
I
,
/
Fig. 2.
grafted nonionic
represents
. . . . . .'".. .
+
'
.,
- .
. +
.+
+
+
. +
+
. . .. . . .
Fig. 3.
Schematic iIIustration of a polymer latex core with grafted polyelectrolyte
chains attached to the core surface.
represents the range of electrostatic forces, and --represents the range of steric forces.
R. H. Ottewill
m.
10- 5 mol dm- 3, which increases the range of the eIectrostatic repulsive
forces, or the concentration of the particIes is increased, a situation is
Electrolyte 10- M
5
Remove electrolyte
Imixed-bed ion exchang~
Dlsperslon
Electrolyte
- sol
10-
resin)
Disorder
Random arrangement
of particles
Brownian motion
Repulsive contacts
Add electrolyte
Stable
Unstable
Electrolyte 0.2 M
Schulze
Hardy
Flocculatlon
Disorder
Weak attraction
Fig. 4.
partides
Coagulation
Disorder
Strong attraction
in an aqueous
rnediurn. (Reproduced
with perrnission
reached where the partic1es must maintain repulsive contacts over a long
period of time. Con sequen tIy, an ordered arrangement.of the partic1es is set
up so that the partic1es sit in lattice positions but remain well separated, Le.,
a "liquid crystal" arrangement is formed (see Fig. 4). When the interpartic1e
spacing is of the order of the wavelength of light, Bragg diffraction effects
become superimposed on the scattering from the partic1es and bright
iridescent colors can be seen visually. This is well demonstrated with
mono disperse latices which, in the right size and concentration range, show
brilliant colors (Hiltner and Krieger, 1969; Hachisu, 1973; Goodwin et al.,
1980).
R. H. Ottewill
IV.
Kh)/V2
(3)
1.
Primary
Maximum
. ...
l....
o
Secondary
Minim~m
Pri';;a~y- ,
Mnimum
h
Fig. 5.
ration curve to illustrate the rnain features used in discussing colloid stability. ~ Vr= energy
barrier to coagulation;
~Vb
i'd = dispersion
polyrner-water interface; Vm= height of the prirnary rnaxirnurn; VSM= depth of the secondary
rninirnurn.
(4)
(5)
R. H. Ottewill
10
2l
X2
+ 2x
6)
where x = hl2a and A = the composite Hamaker constant for the particles
in the mediumas givenby
(7)
TABLEI
Values of Reported Hamaker Constants for Various Polymers
Material
Poly(vinyl acetate)
Poly(vinyl chloride)
Poly(methyl methacrylate)
Styrene-butadiene
Polytetrafluoroethylene
Polystyrene
AII/J
(x 10-2)
8.84
12.4
10.0
6.3
5.6
3.7
7.6
AjJ
(xlO-21)
5.4
5.5
7.2
5.5
3.0-3.8
4.0
2.9
3.6
3.5
9.0
3.2
6.5
7.0
Reference
Dunn and Chong (1970)
Evans and Napper (1973)
Visser (1972)
Visser (1972)
Friends and Hunter (1971)
Evans and Napper (1973)
Force and Matijevi (1968a)
Fowkes (1967)
Rance (1976)
Gingell and Parsegiana (1973)
Visser (1972)
Krupp et al. (1972)
Gingell and Parsegiana (1973)
Gingell and Parsegianb (1973)
Evans and Napper (1973)
Lichtenbelt et al. (1974)
11
with All
(8)
where Vj is the dispersion frequency of the material (c/A.o,with A.o= the
dispersion wavelength), Ctjis the static polarizability and qj is the number of
atoms/molecules per unit volume of the material. Some typical values of
VA = - Aa/12h
150
11
.,
100 t+--
\
\
\
\
\.,
:i.
11.
kT
50':'
'\
"-
\
\
...
.\
"-
"". -- .;-..,
.. .........
"........
,.,
- - =.= .'::..~,
10
...-.--.-20
h (nm)
Fig. 7. Potential energy diagrams for the interaction between two spherical latex
particles of radius 0.1 Jlm at a constant potential of 1/1,= 50 mV, with A = 7.0 X 10-21 J and
T=298.2K
in a 1:1 electrolyte. -,
1O-3moldm-3;
---,
1O-2moldm-3;
-'-',
0.075 mol dm-3;
0.5 mol dm-3.
R. H. Ottewill
12
100
VT
kT
,......,
I!
r:
.,
"
'.
.'.
e.-- ..
'.
......-.
- .-.
k~..".-
O . ..-"
-8-
::
V..a.~.- . _8-
--
..- -
'8.~ I...:e-=."\:
f
h (nm)
Fig. 8.
Potential energy diagrams for the interaction between two spberical latex
1:1 eIectrolyte
concentration
37.5mV;-u-o 25mV;
12.5mV;
5 mV.
-.
witb
50 mV'; -'-,
the so-called retardation effect, and this has to be allowed for in the
calculation of VT. Equation (9) can then be rewritten in order to allow for
retardation as (Kitchener and Schenkel, 1960)
VA = -(Aa/h)(2.45/p
- 2.17/180p2 + 0.59/420p3)
(10)
"=""
13
"
\
200H-'I
I
I
I
I
\
\
\
\
\
\
100
O""'O\
'o
r!
,1
~,f
o
!II-:! 8_'-..1 -=
~ -= .":--'::.
20
-- ~ _8-:--=30
(n m)
-.
-
R. H. Ottewill
14
= kN~
(12)
= ko = 8nDR
= koNUW
(13)
= 28 r'1) exp(VT/kT)
Jo
dh
(14)
(h + 2a)2
and medium potentials (ljIs'" 50 mV) W can attain values of the order of
107 so that coagulation is imperceptible on a reasonable time scale. The
approach emphasizes the kinetic nature of the stability of lyophobic
colloids.
15
::
C
o
...J
0.5
-2.5
-2.0
Log
--~
-1.5
ce
Fig. 10.
Log W against log C. eurve for a polystyrene latex (a = 0.21 Jlm) in barium
nitrate solutions (Ottewill and Shaw, 1966).
indieates results obtained from light-seattering
indieates
R. H. Ottewill
16
that the gradient of the curve just before the cee was given as a first
approximation, based on Eqs. (3) and (9), by
(15)
From this equation it is clear that the gradient should be directly proportional to the particle radius a and inversely proportional to the square
of the valence of the counterion used. A number of experimental studies
have been made tha~ do not seem to confirm these predictions (Ottewill and
Shaw, 1966) and even more refined treatments of the kinetic process have
not removed the discrepancy (Derjaguin and Muller, 1967; Honig et al.,
1971). There is little doubt that as a kinetic process coagulation is rather
complicated; further discussion on this point will be given later.
VI.
(16)
(17)
Distance
>~
IJ)
c:
W
...
...
...
...
Fig. 11.
Sehematie potential energy eurve to iIIustrate eonditions leading to a theoretical definition ofthe cee. -,
VT;---, VAonly.
17
= 81tv2e2NAC/ekTx103
(18)
3.86 x 1O-25y4
-3
mol dm
A 2v
~
(19)
where Ccritis the ccc and we note the inverse dependence on the sixth power
of the valence of the counterion.
For low surface po tentials, t/ls< 25 mV, a further simplification can be
made to give
mol dm-3
(20)
which gives an inverse dependence on the square of the valence, so that for
univalent, divalent, and trivalent counterions, we obtain
VD.
R. H. Ottewill
18
0.2
>-
VI
c:
CD
c.
O
0.1
2.4
2.0
1.6
pH
Fig. 12.
Optieal density against pH al various times after addition of hydroehlorie aeid.
-0-,1
hr; -8-,
2 hr; -/;:,.-,26 hr; ---, cee 1 hr; -'-', cee 2 hr;
, cee 26 hr.
1.
19
11
Poly(vinyl chloride)
Counterion
cccjmmol dm - 3
H+
Na+
BaH
La3+ (pH 4.6)
CIBr1Na+
Na+
K+
MgH
BaH
La3+ (pH 3)
Na+
MgH
1.3
160
14.3
0.3
150
90
43
160-560
200
320
6
6
0.5
50-200
2-10
Reference
Ottewill and Walker (1968)
Storer (1968)
Ottewill and Shaw (1966)
Ottewill and Walker (1968)
Pelton (1976)
The cee values for a number of polymer latiees have been determined
and some typical values are reported in Table II. The trends observed are
qualitatively in agreement with those expected fram the theoretieal approach for particles with smooth surfaces, with 1/1.everywhere the same,
using simple eleetrolytes, i.e., those which do not interact chemically with
water to form new ionie species. These yalues should only be used for
qualitative guidance since, in addition to the factors already mentioned,
there can be variations of the cec with particle size, type and density of
surface groupings, and the presence or absence of stabilizing materials such
as surfactants. In practice it is advisable to determine the actual value for a
particular latex system.
VIII.
So far, the assumption has been made that the ions used in the
coagulation experiments do not interact with water. In a number of cases,
however, the ions do react with water under certain pH conditions to form
hydrolyzed species. For example, in the case of aluminum, the AI3+ ion
exists at pH values below about 3.3 as the hexaaquo ion, with six water
molecules in the octahedral coordinate positions. As the pH is slowly
increased, reaction occurs with water to form a sequence of species. The
chemistry involved in these reactions is somewhat complex and has not
1-
l
20
R. H. Ottewill
been fulIy resolved but a plausible reaction scheme can be proposed, for the
present purpose, as
~20
20
H20
",
--.
'~'~I~+
""
H;O
H20:
.'.
Ji20
:
'. ..
H20:
".
--.
''':~I~+
'..
'......
..........
lI20
~20
OH
'. H20
H;O
OH
/i.....
HO
'H20
:
:
"AI+
H20
"H20
H20
/
1+
H'~ r~~",r~H
"~I
/'~H/'"
[ HO
..
~!
((
lI20
..
"'}{20
1120
1
Alx[OHy]"+
1.
21
'"
'";;
O
z
-2
c:
.2 -3
-;;
...
:
Q)
()
c:
o
RESTABILlZATlON
... -4
al
o
~
ti'
o
...J
STABlE
-5
REGION
..
pH
Fig.13.
Log[(AIN03h/mol dm-3J against pH showing the positions of the coagulation domains for a styrene-butadiene latex. Curves constructed from the data of Matijevi
and Force (1968); reproduced with permission of Ko/loid Z.u.Z.fur Polymere.
.. ..
R. H. Ottewill
22
3.0
- 3.4
"O
<5
E
"-
,..,
u
u
u
':::! 3.8
'"
4.2
pH
Fig. 14.
nitrate
as the
coagulating electrolyte.
IX.
Secondary MinimumEtIects
1.
23
salt concentration, for a constant particle radius, the depth of the secondary
minimum increases; and although the magnitud e of Vrnis reduced, it
nevertheless remains distinct and positive. These trends can be clearly seen
in Fig. 15. The form of the potential energy curves indicates the possibility
that over this range of electrolyte concentration, once a particle enters a
secondary minimum, it will have a long residence time there and remain
separated from the second spherical particle by distances of the order of 6
to 10 nm. However, there remains a substantial primary maximum in
po tential energy to be overcome before the particles can come into contact
or enter the primary minimum. Association in the latter state clearly
corresponds to a condition where the particles come into close contact,
providing the possibility with subsequent thermal diffusion that they will
20 o
,"'
, '.
100
l'
i/\
... .' \
\ \
:!:L
kT
'
-',-'.- \
\
'.
".
.
:,
-100
:.,I
'.
------
;',.y
-----
--_o
.....;;-;:.-:~._- - - - -
00'
,,"
:,
h (nm)
Fig.
15.
VT/kT against
h for spherical
partic1es
of radius
= 1.62
J.lm at various
con-
i;
R. H. Ottewill
24
fuse together. Under these conditions, therefore, one would expect the units
formed to be hard, compact, and essentially nonreversible, and there is
compelling logic to term this state coagulation. On the other hand, when
association occurs in a secondary minimum, the particles remain separated
by a liquid film, which renders thermal diffusion of polymer chains between
the solid particles unlikely, and leaves the possibility that by decreasing the
salt concentration the particles can redisperse. The logical term for this
state is flocculation and there are strong reasons for distinguishing between
the coagulated and flocculated states when simple electrolytes are used to
produce them. It is clear from Fig. 15 that with continued addition of
electrolyte, a transition from the flocculated to the coagulated state should
also occur.
The spherical nature of the particles and the high degree of monodispersity in polystyrene latices makes them ideal systems for testing such a
hypothesis and exploring its practical implications. In some preliminary
experiments by Mardle (1980) the ccc values were determined for a series of
latices of different particle sizes; and then over a range of salt concentrations at and above the ccc he examined the effect on the associated
state of using dialysis to remove the salto The results obtained are summarized in Table III.
TABLE
111
eee/mol dm - 3
value for NaCI
0.21
0.208
0.208
Large aggregates
Nonreversible
0.40
0.155
Some redispersion
Some redispersion
Nonreversible
0.58
0.150
1.34
0.108
3.24
0.158
0.155
0.25
0.40
0.150
0.60
0.70
1.00
0.108
0.400
1.500
0.158
0.400
1.0
Eleetrolyte
NaCI eone
(mol
dm - 3)
Behavior
on dialysis
Some redispersion
Some redispersion
Some redispersion
Nonreversible
Redispersion
Redispersion
Redispersion
Redispersion
Redispersion
Redispersion
1.
25
C/')
:: 60
UI
41
:u 40
Q.
~
20
200
100
Time
300
(min)
Fig. 16.
Experimental determination of percentage of particles remaining at various
times after addition of electrolyte using a polystyrene latex with a particle radius of 1 Jlm.
Sodium chloride concentrations. -0-,
10-5 mol dm-3; -e-,
3 x 10-3 mol dm-3;
-f:::,-, 10-2 mol dm-3; --,
6 x 10-2 mol dm-3.
R. H. Ottewill
26
aggregation process. Assuming that at this stage doublets form the predominant associated units the rate of disappearance of single particIes
becomes
(21)
where k1 is the rate constant for entry into the secondary minimum, k2 the
rate constant for exit from the secondary minimum, N1 the number of single
particIes, and N2 the number of doublets. For a steady-state condition we
have
.
-dN/dt
=O
(22)
= k1NI/k2
(23)
x.
27
ANIONIC
ENHANCEO
/:~
STAS ItlTY
COAGU LA TI ON
CATIONIC
STASltlTY
--
..
Fig. 17.
Schematic iIlustration of the adsorption of anionic, cationic, and nonionic
surfactants on anionic polystyrene latex particles.
XI.
Ionie Head Group with a Charge of the Same Sigo as the Particle
R. H. Ottewill
28
TABLEIV
cee Values for PTFE latiees
Eleetrolyte
NaCI
Ba(N03h
AI(N03h (pH 3.0)
eee/mol dm - 3
dialyzed latex
eee/mol dm 3
in the presenee of
2.5 x 1O-2moldm-3
ammonium perfluoroaetanoate
4.7 X 10-2
7.4 X 10-3
1.6 X 10-4
2.2 X 10-1
7.4 X 10-3
5.4 X 10-3
increased and that nearly an order of magnitude increase in salt concentration was needed to produce coagulation. On the other hand with
barium nitrate the ccc value did not change. The explanation was that
barium perfluorooctanoate is essentially insoluble in water and the addition
of Ba2+ ions to the solution phase stripped the adsorbed perfluorooctanoate ions from the surface.
A further important point arises in the context of adsorption. It was
found that despite the fact that both fluorocarbons and hydrocarbons are
hydrophobic dodecanoate ions are only very weakly adsorbed on a PTFE
latex surface and do not form a monolayer, whereas perfluorooctanoate
forms a monolayer at relatively low concentrations (Rance, 1976).
Compatibility of the polymer surface and the surfactant is therefore an
important factor to consider if enhanced stability is required from addition
of the surfactant.
On basically hydrocarbon-hydrophobic substrates such as polystyrene,
it is well established that even on the negatively charged partic1es there is
adsorption of surfactant anions via the hydrocarbon chains. This is demonstrated in the work of Kayes (1976), who found a substantial increase in
the electrophoretic mobility of polystyrene latices with increase in the
concentration of dodecyl sulfate in the system, and in the work of Cebula et
al. (1978) on the adsorption of dodecanoate ions on polystyrene latex
partic1es.
XII.
:
L
IJ!!II
1.
29
;: 1
c>
o
..J
-1
with permission
of American
Chemical
Society.
= 48 nm.
R. H. Ottewill
30
given in Fig. 18, and it has also been examined theoretically (Ottewill et al.,
1960). At much higher additions of surfactant the electrolyte concentration
is appreciably increased and compression of the electrical double layer
occu.rs, leading to a second coagulation region. Under these conditions, i.e.,
with the ionic groups of the surfactant exposed to the solution phase, the
particles are well-wetted and the coagula usually sink if the density of the
particles is greater than that of the media. The conditions of the first
coagulat"ion region, however (i.e., zero surface potentiaI with hydrocarbon
chains orientated' toward the solution phase) leads to particles that are
easily dewetted and flotation is frequently observed (Connor, 1968).
In Fig. 19 values of the logarithm of the ccc obtained with a series of
alkyl trimethylammonium bromides are plotted against the chain length of
the surfactant for the coagulation of polystyrene latices. The curve is
essentially linear for chain lengths between C4 and Cl2 but deviates from
linearity for the short- and long-chain materials. A similar trend was
reported by Tamaki (1960) for the coagulation of silver iodide soIs by
alkylamine hydrochlorides. It appears that for chain lengths shorter than
C4 the ions behave in a manner similar to hydrated inorganic ions and that
the short hydiocarbon chain is'hydrated by the solvation sheath of the head
-1
-2
r--r
'"
le -3
"C
'O
-4
L
'" -5
o
...J
-6
Carbon
Atoms
10
in
12
14
Chain
Fig. 19.
Log cee against ehain length for the eoagulation of a polystyrene latex
(radius = 48 nm) by alkyl trimethyl ammonium bromides.
31
Surfaetant
CaH17 NMe; BrC1oH21 NMe; BrC12H2SNMe; Br-
polystyrene
dm - 3
eee/mol dm -
latex
PTFE latex
1.59 X 10-4
1.99 X 10-5
2.93 X 10-6
4.5 X 10-5
3.1 X 10-5
2.9 X 10-5
i
g
I
i
I
XIII.
Nonionic Surfactants
= (41tC;kT/3V1P~)(t/Jl -
(24)
R. H. Ottewill
32
the solvent. It is immediately clear that if rfl is taken as 0.5, then for
Xl < 0.5, V. is negative and the term becomes attractive, whereas for
Xl < 0.5, V. becomes positive and the term becomes repulsive. Moreover,
for this type of interaction no effect would be anticipated until h = 2<5,when
the adsorbed layers touch. An extension of this model which allows for a
redistribution of the adsorbed material during interaction has been proposed by Dorozklowski and Lambourne (1971), and application of similar
models to adsorbed polymer layers has been extensively examined by
Napper (1977) and his collaborators. The various models may be too simple.
but it does appear that for well-solvated materials XI is less than 0.5 and in
general, the interaction energy rises fairly steeply once the adsorbed layers
touch. The net effect of v., when it can be used in these circumstances, is to
impart considerable stability to the dispersion since it effectively presents
the particles from entering the primary minimum.
The form of V. against h is 'shown schematically in Fig. 20. It will,
however, seldom if ever act on its own, and usually the van der Waals
'. \ 1
..
'.\ 1
.\
'
.\
>-
O)
..
Q)
c:
W
.~
-Co
Q)
o
Q.
'.\
..
'.\
'. \
..
..
..
.. ........
h
Fig. 20.
Vs only;
, vT = Vs+ vA; -,
33
v.,so that
VT = Vs+
(25)
VA
a situation that can give rise to a shallow energy minimum, as shown in Fig.
20, and to a flocculated state in a dispersion that is easily reversed by
mechanical energy (Long et al., 1973).
In addition in an aqueous dispersion there will be some charge on the
layer, and hence for this situation
VT = Vs
+ VR + VA
(26)
and a potential energy curve of the form shown in Fig. 20 is obtained. Such
a system will be very well stabilized at low electrolyte concentrations and
will remain stabilized even at very high electrolyte concentrations since Vs
still provides the repulsive interaction energy.
Data which illustrate the latter point were obtained by Ottewill and
Walker (1968, 1974) using polystyrene latices of various sizes and various
electrolytes in th~ presence of nonionic surfactants. Figure 21 shows the
2
C)
...J
-4
-3
Log [La (N03h/mol
-2
dr3J
Fig. 21.
Log W against log concentration of lanthanum nitrate at pH 4.6 for a
polystyrene latex (radius = 52 nm) in the presence of various concentrations of a nonionic
surface active agent, C12E6. -0-,
CI2E6 absent; -e-,
lO-s mol dm-3 C12E6; -D.-,
1.5 x lO-s mol dm-3 C12E6; -0-,
2 x lO-s moldm-3.
R. H. Ottewill
34
curves of log Wagainst log C. obtained using polystyrene particles of 103nm diameter at various concentrations of C12E6 using lanthanum nitrate as
the coagulating electrolyte. It is clear that the ccc moves to higher lanthanum nitrate values as the concentration of C12E6 is increased. At
10-5 mol dm-3 C12E6, well below the critical micelle concentration, the
increase in ccc is over an order of magnitude. Close to the critical micelle
concentration with the system containing 5 x 10- 5 mol dm - 3 C12E6 the
latex was stable even at 0.3 mol dm - 3 lanthanum nitrate. The effectiveness
of C12E6 as a stabiiizing agent to the addition of electrolyte is particlesize dependent its effect being greatest with the smallest particles. This is
demonstrated by the data given in Table VI.
It should be noted, however, in connection with the use of nonionic
surfactants that tlocculation is usually observed at a temperature just below
that of the cloud point of the surfactant. The exact temperature at which
this occurs, however, can depend on the type of salt used and the presence
of other surfactants.
An interesting feature occurs in the use of nonionic surfactants with
PTFE latices. Experiments indicate that PTFE particles adsorb ethylene
TABlE VI
Coagulation of Polystyrene latices by Lanthanium Nitrate at pH 4.6 in
the Presence of C12E6
Number average
particle diameter
(nm)
60:t 10
Total C12E6
eoneentration
Equilibrium C12E6
eoneentration
(mol dm-3)
zero
X 10-6
x 10- 5
x 10- 5
x 10- 5
zero
1 X 10-5
1.5 X 10-5
2 x 10- 5
5 x 10- 5
zero
1 x 10- 5
2 x 10- 5
5 x 10- 5
10-4
zero
2 X 10-6
3 X 10-6
7 X 10-6
-
2.8
6.9
1.7
4.2
zero
6 X 10-6
1 x 10- 5
1.4 X 10-5
-
5.6
3.3
3.2
5.2
x 10-4
X 10-4
X 10-3
X 10-3
sa
x 10-4
X 10-3
X 10-3
X 10-3
zero
9.5 X 10-6
1.9 X 10-5
5.0 X 10-5
10-4
5.2
2.0
2.0
1.0
sa
x 10-4
X 10-3
X 10-3
X 10-2
5
1
2
5
103 :t 8
368 :t 17
S indieates
b cee eould
eee/mol dm - 3
La(N03h
(mol dm - 3)
that eoagulation
not be determined
slow eoagulation
oeeurred.
_b
nitrate.
1.
35
XIV.
\
80
C\I
Cii
60
CI)
.. .
\ '.
\
\
\
\
'\.
40
"
"
20
100
Yo
-,
Salt
Fig. 22.
Various effects obtained using mixed electrolyte systems as coagulating agents.
antagonism; ---, superadditivity;
, additivity; -'-', syngergism.
R. H. Ottewill
36
,,
,,
20
,,
,,
,,
,,
,,
,,
,,
...o.100
.
Fig. 23.
Results obtained for the coagulation of a polystyrene latex (radius = 0.109 pm)
using mixtures of magnesium sulfate and sodium nitrate at pH 8.5 and 298.2K. Percentages
are expressed in terms of the molar concentrations of the salts; --- indicates additivity.
which case the plot joining the two values is a straight line. Superadditivity
gives a curved line which is convex to both the abscissa and the ordinate.
Antagonism gives an even more pronounced convexity and the gradient at
the point of intersection on the ordinate is distinctly positive. The fourth
case is that of synergism, Le., when the curve is concave toward the axes.
Some results obtained by Storer (1968) using well-dialyzed polystyrene
latices at pH 8.5 and mixtures of magnesium sulfate and sodium nitrate as
the coagulating electrolytes are shown in Fig. 23. Distinct synergism was
observed over the entire concentration range. The use of activities of the
ions in the mixed systems, rather than concentrations, gave a reasonable
explanation for the form of the data, but it is also of interest that the
discrete-ion treatment of Levine and Bell (1965) also predicts synergism for
certain cases.
XV. Heterocoagulation
In the previous section the strong affinity of cationic surfactants for
negatively charged polystyrene latices was noted. This concept of a single
po sitive ion interacting with a negative charge on a surface can be extended
37
Fig. 24 in the form of log Wagainst N-/(N + + N-) where N+ and N_are
the number concentrations of the cationic and anionic lattices respectively.
The anionic latex used was composed of polystyrene particIes (diameter
0.1 6
O>
o
...1
0.0 8
0.1
Fig. 24.
0.2
0.3
004
= the
chloride concentration
= 10-3
mol dm-3.
3~
R. H. Ottewill
52.7 nm) with sulfate. groups on the surface; the cationic latex was also
polystyrene (diameter 43.4 nm) with surface amidine groups. It can be seen
from Fig. 24 that the system becomes completely unstable w~en the ratio
N_/(N+ + N_) reaches -0.25.
When the particles are of different sizes complete coverage of the bigger
particles by the smalIer can occur (Goodwin et al., 1978). This is demonstrated by the scanning electro n micrograph shown in Fig. 25.
An "alternative method of studying heterocoagulation is to use the
spinning disk technique, a method that provides well-defined hydrodynamic
conditions for examining the deposition of spherical particles on planar
surfaces. The method was used by Clint et al. (1973) to examine the
heterocoagulation of polystyrene latex particles (diameter 418 nm) onto
planar polystyrene surfaces in the presence of barium nitrate. The highest
rates .of deposition of the latex particles began to occur at about
0.02 mol dm - 3 barium nitrate. This corresponded to IjIs values of about
10 mV on the particles and about 6 mV on the planar surface. The
experiments were satisfactorily explained using the Levich theory (1962) of
diffusion to a rotating disk, with modification to include a potential energy
of interaction between the plate and a sphere.
Fig. 25.
Scanning electron micrograph showing the heterocoagulation of cationic latex
particles (radius = 22 nm) onto a negativeIy charged particle (radius = 1.07 11m).
1.
39
XVI.
Suace Coagulation
There are a number of cases with polymer colloid systems where the
coagulation process can occur at the liquid-air interface under conditions of
electrolyte concentration that are far removed from those required to
produce coagulation in the bulk solution. The processes that occur in this
type of coagulation are illustrated schematically in Fig. 26. The effect was
originally called mechanical coagulation, by Freundlich, but waf! more
appropriately termed surface coagulation by Heller and Peters (1970) who,
in conjunction with several co-workers, have carried out an extensive
investigation of this phenomenon. A theory of surface coagulation was
developed by these authors based on the following assumptions:
1. That the coagulation process proceeded exclusively at the liquid-air
interface and was a biparticle association.
2. That the contribution of aggregates to the rate could be neglected.
Moreover, since the aggregates returned to the bulk phase they had an
insignificant effect on the interfacial area available for occupation by
unreacted primary particles.
3. That sufficient convection occurred to exclude the diffusion of particles to and from the interface as a rate-determining factor.
4. That a steady state occurred for the distribution of primary particles
between the bulk dispersion and the surface, and this distribution could be
described by a Langmuir adsorption isotherm.
5. That the rate of adsorption was la,rge enough relative to the rate of
formation of fresh surface for the adsorption equilibrium to be unaffected.
Thus, putting the Langmuir adsorption isotherm in the form
c/r
= K1
+ K2c
(27)
Water
Dispersion
Fig.26.
Schematic
(5
Adsorption
illustration
co
Surface
Coagulation
Redispersion
40
R. H. Ottewill
When S is the surface area at a constant rate of surface renewal and Vis the
volume of solution, the rate at which the bulk concentration was changed
was given by
-dcjdt = (KOSjV)C2j(K2
+ K2c)2
(30)
XVll.
Peptization
t
41
>-
Q)
c:
L-- ( a)
(b)
Distance
-------
Fig.27.
repeptization.
AVb-kT
AVb-kT
(e)
(d)
42
R. H. Ottewill
One possibility is to provide the particles with a thin steric barrier such as
the "micro hairs" which could occur on the surface of a latex by solvation
of the polymer chain beyond the ionic end group. Indeed some evidence for
this occurs with certain latices. For example, Smithan et al. (1973) have
reported evidence of steric stabilization with polystyrene latices with a high
content of carboxyl groups on the surface prepared by an essentially
conventional emulsion polymerization method. Microsteric stabilization
with latices could be an important factor and this is undoubtedly an area
that needs more extensive investigation.
XVffi.
t~
43
I
I
/
POLYION
/
LAYER
/
/'
Fig. 28. Schematic ilIustration of a latex partic1e with an adsorbed or grafted layer of
polyelectrolyte.
represents the extension in space of the adsorbed layer, and --- represents
the extension in space of the electrical double layer. 'Reproduced with permission. of American
Chemical Society.
R. H. Ottewill
44
20
Fig. 29.
Data iIIustrating the temperature behavior of a latex with a polystyrene core
stabilized by grafted poly(acrylic acid) in 1.10 mol dm - 3 sodium chloride solution; IX= degree
of neutralization of the latex.
45
nature of the disperse phase, the particle size, the surface coverage and the
molecular weight of the stabilizing polymer. It was found that to obtain
colloid stability, it was necessary to use a dispersion medium that was better
than a () solvent for the stabilizing chains.
The implication of this observation was that the second virial coefficient
of the stabilizer needed to be po sitive so that the segmental excluded
volume was also positive. Under these conditions once overlap of the
stabilizing polymer layers occurred, the configurational entropy of the
molecules in the overlap region would become less than that of the
molecules in the dispersion medium and an excess osmotic pressure would
occur in the overlap volume. As a consequence of this, molecules of the
dispersion medium would difIuse into the overlap region, forcing the
stabilizing layers apart. This is essentially the basis upon which Eq. (24) was
formulated. It was found that with this type of system that peptization
could be achieved spontaneously after centrifugation or flocculation,' as
would be anticipated from Fig. 27d.
Napper (1970b) also investigated the flocculation of poly(vinyl acetate)
particles stabilized by poly(ethylene glycol) chains with a series of electrolytes. The order for the cations was
R. H. Ottewill
46
= 4na3PLNAi3Me = MdMe
(31)
(32)
= 4naa./[B(1 +
Ka)]
(34)
47
10
:;:
O>
+
2
.'
5
10
Radius
15
20
(nm)
Summary
R. H. Ottewill
48
10,173.
Goodwin, J. W., Ottewill, R. H., and Parentich, A. (1980). J. Phys. Chem. 84, 1580.
Gregory, J. (1975). J. Co/loid Interface Sci. SI, 44.
Hachisu, S., Kobayashi, Y., and Kose, A. (1973). J. Co/loid Interface Sci. 42, 342.
Hamaker, H. C. (1937). Physica 4, 1058.
Hardy, W. B. (1900). Proc. R. Soco London Ser. A 66, 110; Z. Phys. Chemie. 33.385.
Heller, W., and Peters, J. (l970a). J. Co/loid Interface Sci. 32, 592.
Heller, W., and Peters, J. (1970b). J. Co/loid Interface Sci. 33, 578.
Heller, W. and de Lauder, W. B. (l97Ia). J. Co/loid Interface Sci. 35, 60.
Heller, W. and de Lauder, W. B. (197Ib). J. Co/loid Interface Sci. 35, 308.
Hiltner, P. A., and Krieger, 1. M. (1969). J. Phys. Chem. 73, 2386.
Honig, E. P., Roeberson, G. J., and Wiersema, P. H. (1971). J. Colloid Interface Sci. 36, 97.
Hoy, K. L. (1979). J. Coat. Technol. SI, 27.
Kayes, J. B. (1976). J. Co/loid Interface Sci. 56,426.
Kitchener, J. A., and Schenkel, J. H. (1960). Trans. Faraday Soco 56, 161.
Kratohvil, S., and Matijevi, E. (1976). J. Co/loid Interface Sci. 57, 104.
Krupp, H., Schnabel, W., and Walter, G. (1972). J. Co/loid Interface Sci. 39,421.
Levich, V. G. (1962). "Physico-chemical Hydrodynamics." Prentice Hall, Englewood ClilTs,
New Jersey.
,,
~
r
49
~
\
2
Particle Formation Mechanisms
F. K. Hansen and John Ugelstad
1.
11.
111.
IV.
V.
VI.
VII.
Introduction.
Micellar Nucleation: The<Smith-Ewart Theory .
Radical Absorption Mechanisms.
".
Micellar Nucleation: Newer Models .
Homogeneous Nucleation.
Partide Coagulation during the Formation Period.
Nucleation in Monomer Droplets
References .
51
54
56
63,
73
82
86
91
l. Introduction
The nuc1eation stage constitutes the so-called Interval I in an
emulsion polymerization, the initial period in which the partic1e number is
changing. In Intervals n and nI the paftic1e number is believed to be
essentially constant. Nuc1eation of new partic1es may in some cases also
take place during Intervals n and In. This phenomenon is often referred to
as secondary huc1eation and may be encountered in systems with poor
stability (coagulation) or with changing composition (continuous and semicontinuous polymerizations). The present chapter will attempt to treat all
mechanisms that may lead to formation of polymer partic1es, in whatever
stage of the polymerization they take place.
All discussions of partic1e nuc1eation start with the Smith-Ewart theory
in which Smith and Ewart (1948) in a quantitative tre!ltment of Harkins'
micellar theory (Harkins, 1947, 1950) managed to obtain an equation for
the partic1e number as a function of emulsifier concentration and initiation
and polymerization rates. This equation was developed mainly for systems
of monomers with low water solubility (e.g., styrene), partly solubilized in
micelles of an emulsifier with low critical micelle concentration (CMC) and
p'?"Ts{lIledto work well for such systems (Gerrens, 1963). Other authors have,
,
however, argued against the Smith-Ewart theory on the grounds that (i) .
partic1es are formed even if no micelles are present, (ii) the equation for the
51
EMULSION POLYMERIZATlON
52
53
Honamer
H
Hice//es
Drap/ets
nitiatar
Emu/sifier
,
Aqueaus
sa/utian
Fig. 1.
emulsion polymerization
among phases.
of components
11
54
A
B
e
D
E
F
G
Result
Micellar initiation
Droplet initiation.
Radical disappearance
(particle growth)
Higher oligomers
"Dead" oligomers (may or may not lead
to nucleation)
HoIt1ogeneous nucleation
Nucleation
(homogeneous
or micellar)
n.
Smith and Ewart (1948) proceed as follows: radicals are absorbed into
monomer-swollen emulsifier micelles which then are transformed into polymer partic1es; the rate of radical absorption Is equal to the rate of initiator
decomposition p, which means that
dN/dt
= p
(1)
J1.
(2)
where v is the volume of the partic1e. The number of micelles will decrease
as the partic1es grow, giving an increasing surface which will adsorb
2.
55
(5)
api'
where
area of one
J1.
56
57
\
\
\
\
\
\
1
I
Halar fraction,
Concentration
r.e
XAW'CW
---
------
R Distance
Fig. 2.
Geometric and concentration conditions around an absorbing spherical partic1e. e is concentration and X is molar fraction of dilfusing species.
58
= (r/R2)[(r + b)/b](l -
XA) In[(1
XAa)j(l - XAw)]
(9)
=r
(10)
Usually the molar fraction of the diffusing species (A) is much smaller
than that of the' stationary component (B) so that 1 - XA ~ 1 and
In(l - XA) ~ XA, which gives
(11)
where
(12a)
and
(12b)
Ir b ~ r, Eq. (11) simplifies to.
A = -Dw4nr(Cw- Ca)
(13)
W=r-
r+~
r+ b
b f.r
et/lodR
exp--
( kT ) R2
(15)
where R is the distance from the oligomer to the center of the particle, e is
the electronic charge, and t/lois the (effective)surface potential.
The value of W' has been calculated by numerical integr~tion for the
case where b ~ r and has been given as a function of et/lo/kT and Kr where K.
is the inverse double-Iayer thickness (Hansen and Ugelstad, 1978). The term
eZin Eq. (14) may be considered as an activation energy (Boltzmann factor)
and accounts for the fact that in order to become absorbed the charged
radical s have tq surmount an energy barrier at the surface. The'term W' is the'
integral of this activation energy factor giving the retardation of a unit charge
upon diffusion to the surface. In the case where Cw~ Caez Eq. (14) reduces
to'
(16)
\.
..
2.
59
= -4nrDpC*(X ooth X - 1)
(17)
= r(k/Dp)1/2
(18)
= kpMp + nktp/vp
(19)
with
= aCa
(20)
60
leading to the following expression for the rate of absorbtion per particle
nA= 4nrDwCwF
(21) ,
(22)
Note that in Eq. (21) nA has a positive sign since it stands here for rate of
absorption.
It will appear from Eqs. (21) and (22) that if
W' ~ (Dw/aDp)[1/(X coth X
1)]eZ
(23)
which corresponds to Eq. (16). Equation (23) describes the situation when
the transport of the oligomer radical to the outer surface is the ratedetermining step in the absorption process.
In the case where W' 1
::::::
(24)
Figure 3 gives 1/F as a function of r for some probable values of the
parameters involved and with W' obtained by numerical integration of Eq.
(15). The more important features of Eq. (22) are evident from the figure.
1. When particles are large (r > 1000 nm) and a h.sa high value (oil-
effect(Fitch and Shih, 1975)in which the uncharged end pulls the oligomer
through the surface or reacts while the charged end still is in the aqueous
phase. This may be important especialIy for higher oligomers and may be
calculated if the van der Waals attraction is included in the activation
energy termo
Equation (22) may in some limiting cases be simplified.
Condition A:
X ~ 1
i.e.
X coth X
r(k/Dp)1/2~ 1
(25)
1 '" X2/3
(26)
2.
61
10
10
r{nm}
Fig. 3.
Inverse absorption
distribution constants a = 1 and 104. Elfective surface potential i/to = O (1), 50 mV (11),and
100mV
(III).
Other
parameters:
Dw = 5 X 10-8 dm2jsec,
Dp = 1 X 10-8 dm2jsec,
kp = 300 dm3jmol sec, Mp = 5 moljdm3, T= 298 K.
which
gives
(27)
Case A usually implies that the term W' in Eg. (22) may be neglected. In this
case Eg. (27) describes the curves with a slope of - 2 iti Fig. (3). The rate of
absorption is then
(28)
In this case the concentration gradients both on the agueous and the
particle side may be neglected. The rate of absorption is proportional to the
~
rate of reaction of the radical inside the particle and thus to the particle
volume.
62
Condition B:
x ~ 1 i.e. r(k/Dp)1/2~ 1
X cothX - 1 "" X
(29)
(30)
which gives
(31)
If condition B is operating it is more like1ythat W' is the dominant term
in the equation for "l/F, leading to the conclusion that the irreversible
absorption Eq. (23) will be operating. With low values of a and Dp and/or
high values of Z one may also find that the term W' may be neglected in this
case. Equation (31) then describes curves with a slope of -1, as found in
part of the most righthand curve in Fig. 3. The absorption rate will if the
term W' is neglected be
lA= 41tr2a(kDp)1/2/e'= apa(kDp)1/2/e'
(32)
63
Mechanisms
this fact should also come into play in that it leads to an increase in the
value of a with the same factor.
Application of a value of a given by Eq. (34) transforms Eq. (22) to the
form
!
F
Dw
aoDp X coth X - 1
+ w'
(22a)
A = vpaokCw
A = apao(kDp)
1/2
(32a)
Equation (23) for irreversible absorption will remain the same, the only
necessary condition for Eq. (23) should be to replace eZlawith 1lao.
~I
64
F. K. Hansen
9 seed
J. Ugelstad
Nsrs.1(
ea
and
O
11..9
21..21.
M
'e
1016
'Z
/'
./
;'
.,/
C=O
CMC
1
101'
001
0.1
1.0
e (free 50S)
Fig. 4.
initiator
K2S20S
10.0
(g dm-3)
of new partic1esN as a function of
(0.6 gjdm3).
(Reproduced
by permission
80 nm, temperature
of J. Polym. Sci.)
et al., 1976, 1980; Ugelstad et al., 1969). The often undramatic change in N
at the CMC may well be explained by homogeneous nucleation. The
question still remains as to whether the micelles present above the CMC
play any important role in the nucleation process. One method for investigating the competition between homogeneous and micellar nucleation
uses seed experiments. Results from such runs are shown in Figs. 4 and ~
for styrene and methyl metacrylate, respectively. The log(new particle
number) is plotted as a function of log(free emulsifier concentration) both in .
seeded and unseeded runs. Free emulsifier denotes the concentration of
emulsifier not adsorbed on seed particles.
In the case of styrene the curve for the unseeded runs shows a sharp
increase at CMC. Above the CMC, the increase in particle number with
increasing emulsifier concentration becomes smaller but is still significant.
2.
65
MMA
u
~
.u
0.1
1.0
10
Free SDS(gdm-3
Fig. 5. Emulsion polymerizatjon of methyl metacrylate.Number of new particles, N as
a fimction of the freeconcentration of SDS witb and witbout seed"
66
of particles below the CMC when compared to the case without seed.
Slightly above the CMC, however, the curves with and without seed
coincide completely. This would seem to indicate that with methyl methacrylate also the particle formation takes place in micelles when operating
above the CMC. It seems as if the main difference between styrene and
methyl methacrylate is that the chance for self-nucleation is better for the
latter. In conclusion it may be said that the seed experiments pro vide strong
arguments for the importance of micellar nucleation even for more watersoluble monomers. .
Further derivations based on the micellar theory have been advanced to
create a more detailed picture than that of Smith-Ewart. With these
theories the equations become so complex that a simple analytical solution,
like that of Smith-Ewart, cannot be given. Instead, numerical methods
must be applied to the integrations. Apart froIl1the calculations of Parts et
al. (1965) and Gardon (1968) already mentioned, such work has been
performed by Harada et al. (1972) (styrene), Nomura et al. (1976) (viny)
acetate), and Hansen and Ugelstad (1979b, styrene and 1979d, different
monomers). For most of these calculations nonsteady-state equations are
used to calculate the rate of particle nucleation and rate of volume growth
given by the average number of radicals per particle (). Generally, the rate
of micellar nucleation may be expressed by
(35)
wherePAis the total rate of radical generationin the aqueous phase and Pm
is the probability of absorption of radical s (oligomers) in the micelles. Often
(as in the Smith-Ewart theory) PA is set equal t 'Pi, the rate of initiator
splitting. Generally, radicals desorbed from the particles should also be
included in PA (Nomura et al., 1976; Ugelstad and Hansen, 1976). The
expression for PAwill then be
PA
= Pi +
kdN
(36)
2.
67
where Po and Pt are the probabilities for radical absorption in particles with
O and 1 radical, respectively. The rate of particle growth is given by
(39)
where kp is the propagation constant, N A Avogadro's number, <Pmpthe volume
. fraction of monomer in particles, and dm and dp the densities of monomer
and polymer, respectively. The last term includes the contribution to
volume growth by the micellar volume and is usually small. The volume,
radius, and surface area of each particle may be calculated from Vp and N
(this procedure involves an error in the calculated surface area because the
size distribution of the particles is not included, but it has been shown that
this error is small, usually about 5%).
Ir the above equations are integrated numerically with kd = O, constant
<Pmp,Pm = 1, and Po and Pt = O, one arrives at the Smith-Ewart upper
68
absorption rate nto the particle in question to the sum of all absorption
rates. For instance, the probability for absorption into micelles is given by
(43)
where Nm is. the number of micelles presento The appropriate rates of
absorption to be inserted into Eq. (43) is obtained from Eq. (21) with the
value of F given by Eq. (22).
Because oligomers of different degrees of polymerization may be absorbed, the expression for P may become much more complex than Eq. (43).
AIso, because the efficiency constant F for initiator radicals may differ from
monomer radicals, one should in fact operate with different probabilities for
those two radicals, which leads to further complications. These problems
will be dealt with in the next section. Satisfactory results have been
obtained with the relatively simple model outlined above. Nomura et al.
(1978) expressed Pmby
(44)
where
(45)
Here Mm is the agregation number for the micelles. It is seen that Eq. (44)
is identical to Eq. (43) if the efficiency factor in particles containing O and 1
radicals is the same. By numerical integration, using constants for vinyl
acetate, Nomura et al. found
(46)
where So and lo are the initial emulsifier and initiator concentrations,
respectively. Equation (46) explained well their experimental results with
vinyl acetate when B was held constant (1.2 x 107). A very important result
of these calculations is that the exponents in Eq. (46) are higher and lower,
respectively, than the Smith-Ewart values 0.6 and 0.4, respectively. This has
been shown
to be a common
feature
when desorptionjreabsorption
be-
by
(47)
Here b is an efficiency factor for radical capture in a micelle relative to a
particle of the same size and x is the order of capture rate with respect to
radius.
2.
69
0.1
1
10
50/(9 dm-3)
100
70
ve
1019
.... ....
...
-",,'"
--~.,,-'"
.... ..........
1_"''''''
'upper
,....-
...
I
~Iower
5-E
1017
10'5
1017
p!(dm-3
Fig. 7.
Sol)
as in Fig. 6. (Reproduced
more so the higher the value of the e[ective desorption constant. It turns
out that the particlenumber may be expressedby
(49)
where 0.6 < Z < 1.0.
Equation (49) was first obtained by Nomura et al. from a nonsteadystate treatment. The value of Z increases as the desorption and reabsorption of radicals increases. This also leads to an increase in the
particle number because desorbed radicals may also produce new particles.
On the other hand the polymerization rate per particle diminishes as the
high degree of desorption and reabsorption of radicals leads to a value of.
~1.
The above calculations explain well the qualitative results with monomers like vinyl chloride, vinyl acetate, methyl methacrylate, and styrene.
Quantitatively, the correlation between theory and experiment will depend
on the magnitude of J and x, which may also change during the formation
periodo Generally, if Z > 0.6, the particle number, as well as the value of Z,
71
72
t51
-I
r7.5
tO
5.0;I
g,
"
IC
0.5
)(
2.5
--
t (min)
Fig. 8.
The number of particles N mean radical concentration, , and monomer volume
fraction in particles, rPmp,as a function of time from numerical. integration of Eqs. (38-39) with
P m from
Eq. (21-22).
Monomer,
styrene;
emulsifier
SDS with
So
7 g/dm3,
73
(54)
volumeis then
(55)
where v is the molar volume of the repeating unit in the polymer, ktw the
aqueous-phase termination constant (in molecular units) and cjJppthe volume fraction of polymer in partic1es. The rate of partic1e formation is then
from Eq. (51)
dN/dt
= p[1 -
(nN)1/3{[3vkpMw/(4ktwcjJppNA)]
x In[cosh(pktw)1/2t]}2/3L]
(56)
..
74
2.
75
lnitiator radicals
dR/dt
= p -
kpRMw
ktwRRw
Lp kaipNp
(57)
(58)
l-mers from initiator radicals
(59)
l-mers from monomer radicals
(60)
76
dRr
d/ = kpRIO-1)Mw- RIj kpMw+ ktwRw + ~kaIjpNp
(61)
j-mersfrom monomerradicals
(62)
Here R and RM are the concentrations of radicals produced from the
initiator and by chain transfer to monomer, respectively, RIj and RMjare the
concentration of j-mers originating from these radical s, and Rw is the total
concentration of alI radicals in the aqueous phase
jIcr
Rw = R
L:
jMcr
RIj
+ RM +
j=l
L:
(63)
RMj
j=l
kaMjpNp
== kaIjmNm
== kaMjmNm
+ kaIjONO + kaIjlNl
+ kaMjoNO +
kaMj1N1
+ kaIjdNd
+ kaMjdNd
(64)
(65)
2.
77
Probably, jMcr will be less than jIcr because the monomer radical is uncharged. The rate of formation of living (N1) particles is
= dN /dt -
dN/dt
(70)
The total rate of particle growth is expressed by Eq. (39), only the micellar
volume Vrnshould be exchanged by the volume of the precipitated oligomers. We do not differentiate between the volumes of dead and living
particles, since the particles rapidly change from active to inactive and vice
versa.
Equations (57-70) may be integrated numerically, expressing F by Eq.
(22). A simplification may be obtained by means of the steady-state
approximation, in which all rates of radical change in the aqueous phase are
set equal to O,expressing Rj by means of Rj-1
RIj/RIO-1)
RMj/RMO-1)
= fJIj
= kpMw!(kpMw + ktwRw + Lp
kaMjpNp)= fJMj
,
(71)
(72)
(74)
where
(75)
78
and
PM = 1 + ktwMRw + Lp kaMpNp
kpMMw
kpMMw
(76)
(77)
= k p M w p T1
PII
1=1
and
j
kdN1
RMj
(78)
= k p M w PM T1
PMI
1=1
FinalIy we have
R.~ k:~A~+%Q]p,.)]
(79)
+ ::Z~PM[[+%(QPM.)]
In the first paper on this subject (Hansen and Ugelstad, 1978) desorption
of monomer radical s was neglected. Furthermore, in order to simplify Eq.
(74), Pi was set equal to 1 (Le., termination and absorption of initiator
radicals was neglected). AlI ka1jpvalues were set equal to some average
absorption constant Kc(when micelIes are absent). As an approximation, Rw
could then be expressed by
Rw
= {[(KcN)2 + 4pktw]I/2
KcN}/2ktw
(80)
and dN /dt by
dN/dt
(81)
Ir I<cand Rw are constant during the nucleation period, Eq. (81) may be
integrated numericalIy. Assuming that as an approximation radical absorption may be neglected in the calculation of Rw we ha ve
Rw
= (p/ktw)I/2
(82)
and
N(t)
..
= {[k1Pjcrt +
(k2 + l)jcr]IUcr - k2
1}/k1
(83)
..
2.
79
where
(84)
and
kz = (ktwpY/Z/kpMw
(85)
Because r will increase with time, ke will also increase, and dN/dt will fall off
more with time than is expressedby Eq. (81). By using a plausible average
value for r, Eq. (83) may give an idea of the number of particles that may be
expected to be formed homogeneously. Assuming jer = 5, as found by
Goodall et al. (1975) for styrene, and F = 1 (irreversible absorption) it turns
out that N will be of the same order of magnitude as found experimentally
in emulsifier-free systems. In systems with emulsifiers, below the CMC, the
experimental particle numbers are much higher, indicating that the value of
ke (and F) is in reality much lower. The low particle number in emulsifierfree systems must therefore be attributable to limited coagulation.
By numerical integration of Eq. (73) it was shown (Hansen and Ugelstad,
1979b) that after 1 hr the ca1culated particle number was close to the
experimental particle number for styrene with SOS as emulsifier, Le.,
immediately below the CMC. The following simplifications were made: (1)
no aqueous phase termination and (2) radical s with lower degree of
= PPIh + kdNPMh
(86)
in micelles. Here
(87)
and
PMh
= [1 +
(88)
(Lp4nDwrpNpFMp)/kpMwrl
pA
jMcr
(89)
where R1j and R..jare given by Eqs. (77) and (78). It is obviously very
difficult even to estimate PA from Eq. (89). One method of simplification
may be the second one given above. This leads to
Lp k.1pNp
PA =p.
1 kpMw
k N
jcr
kpMw
1 kpMw
kpMw
Lp k.MpNp
ktwRw + Lp k.MpNp kpMw + ktwRw
jMCr
(90)
80
= (1 +
ktwRw/kpMw)-her
(91)
ktwRw/kpMw)-jMer
(92)
= (1 +
= 4.9
x 1016 dm
- 3, diameter
93 nm). Hawkett's
values for
PAI at the above values of [I] are approximately 0.2, 0.09, and 0.03,
respectively. Hawkett points out that ktwmay have a higher value than that
given above based upon measurements in dilute polymer solution, since the
termination reactions in water involve the reaction between small molecules. It should be pointed out that the above treatment do es not involve
competition between termination and absorption as in Eq. (90). From this
equation we should expect an influence of partic1e number and micelles on
absorption efficiency.
The assumption of such a low efficiency as found by Hawkett, where the
order of PAlwith respect to [I] approaches 0.5 with increasing [I], seems to
be in contradiction to both the absolute value of partic1es formed (both
below and above the CMC) and even more so with the observed exponent
in Eqs. (5) or (7) (i.e., 0.4). AIso, it would seem to contradict the well-known
fact that at high values of [I] (i.e., high a' = 2ki[I]Vp/Nk~) the rate of
polymerization is proportional to [I] 1/2 even in Interval II (Smith-Ewart
Case 3).
The effect of termination in aqueous-phase on partic1e formation will
depend on whether the terminated and usually doubly charged molecules
stay in solution, are absorbed on existing partic1es, or form the nuc1eus for
a new partic1e. It seems unlikely that doubly charged styrene oligomers with
2.
81
L DwjRj
jcr
f1N1)
(93)
j= 1
jcr
L R,
Rjcr+l-j
j=1
1=1
jcr
+ 4n(f1N1 - foNo)
L DwjRj
(94)
j=1
where Dwj is the diffusion constant for j-mers in water and A is the ratio of
termination by disproportionation to the overall termination reaction. The
first term in Eq. (93) accounts for the partic1e formation by precipitation of
oligomers that have a total degree of polymerization higher than jcr. The
rate expressions for the soluble oligomers Rj, otherwise c10sely follow Eqs.
(59-62). Arai et al. assume that Dwjfollows Wilke's equation
DWJ. = Dw/':1.0.6
(95)
They do not take into account desorption of monomer radicals or the
possibility for decreased rates of desorptioll'.
The equations for dNldt and dR/dt, as well as for dVldt [Eq. (39)J are
solved by numerical integration for the polymerization system MMAK2S20S-water, with rate constants obtained from the literature. The
initiator efficiency was set-equal to unity. Partic1e numbers between 1014
and 1015were obtained for initiator concentrations of 10-3-10-2 mol/dm3.
The calculations showed that N should be almost independent of the
chosen value of jcr for values between 5 and 70 (in strong contrast to our
calculations). The reason for this is probably that aqueous-phase termination with subsequent precipitation is the dominant partic1e-formation
mechanism in Arai's model, even more so with increasing initiator concentration. The theoretical partic1e-formation time was on the order of 2
sec, a very low value compared to the experimental results of Fitch and
Tsai. Arai et al. found that their calculated partic1e numbers were approximately in accordance with the experimental results of Yamazaki et al.
(1968) for emulsifier-free polymerizations. Arai's model does not inc1ude any
coagulation mechanisms. It will therefore have the same shortcomings as
most other models, namely that the strongly increased partic1e number in
82
VI.
2.
83
and coagulation took place during a longer time interval when emulsifier
was used, and the total particle number both at the maximum and at the
steady-state level increased.
Fitch and Watson (1979) also used a light-scattering technique to
investigate the early stages of the reaction in a homogeneous solution of
MMA in water. They used a photoinitiation procedure by which free
radical s could be p~oduced by means of a light flash, after which Rayleigh
ratios (Rgo) were measured as a function of time. Within the experimental
time scale of '" 2 min, Rgo increased linearly with time, indicating particle
coagulation. The rate of coagulation was clearly dependent on SDS concentrations and was most pronounced in emulsifier-free runs.
The stability of a dispersion may in the most simple way be described by
Smoluchowski's coagulation equation
(96)
-dNddt = kllNf
and in the integrated form
1/N1
= l/No +
kllt
= l/No(1 +
t/1:)
(97)
Here k11 is the second-order coagulation constant and 1:the half-life time
l/Nokll' This treatment assumes that kll is constant during the coagulation, which is only approximately correct during the early stages. When no
force exist between the particles, the coagulation was regarded as a diffusion
process with a value of kll given by
kll
= 4nDllrll .
(98)
where Dll and rll are the total diffusion constant and the collision radius of
the particles, respectively. Expressing Dll by the Stokes-Einstein relationship leads to the familiar expression for kll
(99)
k11 = 4n(2D1,)(2r1) = 16n(k:T/61ttlr1)r1 = 8kT/3r
When repulsive forces exist between the particles, the rate constant is
decreased and the ratio kll(slow)/kll(fast) is usually expressed as
kll(slow)/kll(fast)
in which W(fast) 1S usually set
ratio, expressed by
= 1.
= W/W(fast)
The parameter
reo
W = 2rlo exp(J!;ot/kT)
dh
(100)
W is
Fuchs' stability
(101)
where J!;otis the total potential energy and h is the distance between the
particles. The parameters for this equation [Eq. (101)] have been extensively discussed in the literature (Derjaguin and Landau, 1941; Verwey and
Overbeek, 1948).
84
(102)
where S is the molar concentration of SDS and is less than 9 x 10-4 M
(0.26g/dm3).Becauseof the linear relationship of Rayleigh ratios with time,
W wasindependentof time for a specificexperiment,whichindicatesthat W
is independentof particlesize. This was also found by Ottewill and Shaw
(1966) to be the case for larger particle. When the total particle surface is
small, and also at low SDS concentrations, the surface charge density will
probably be independent of particle size and increase linearly with S. This
would seem to imply that W has a linear dependence on surface charge
density in the experiments described by Fitch and Watson.
Munro et al. (1979) and Goodall et al. (1980) investigated particle
nucleation in emulsifier-freepolymerizationsof styreneby means of electron
microscopy, light scattering, and photo correlation spectroscopy, the latter
in unstirred reactions in which 11 styrene layer was kept on top of the
saturated water solution. The initiator was potassium persulfate. Particle
numbers were found to decrease with time from an initial high value (Le.,a
maximum); the maximum was more pronounced the higher the polymerization temperature. The time scale for the decrease was 10-100 min and
involved particles up to a size of about 75nm. GPC analysis of both the
aqueous phase and the particles during these early stages showed that the
predominant molecular weight of the species present in the aqueous phase,
measured as polystyrene, was about 520 and that of the polystyrene in the
particles about 1000. In addition, the particles contained some higher
molecular weight polymer (up to 106), the fraction of which increased after
the formation period was finished. The conclusions drawn were that if these
low molecular weight species are produced by termination by combination
it would appear that the water-soluble oligomers are produced from 2-3
mers and the oligomers in the particles from 4-5 mers. This is the basis
for assuming that in the case of styrene jcr = 4-5.
In the work on emulsifier-free polymerization of styrene Hansen and
Ugelstad (1979a) also concluded that coagulation during the formation
period was of great importance. By seeded polymerizations, it was found
that by including coagulation between new particles and seed particles one
could describe qualitatively the number of new particles formed. For the
homocoagulation between new particles, Eq. (96) was applied with ku
given by Eqs. (97-100).Heterocoagulation between seed (Nz) and the new
particles (NI) is described by
(103)
-dN/dt = klZNINz
2.
85
where
k12
= 41tp12r12/W2
= 2kT/317[(1/r
(104)
+ r2)/WI2]
= 2.7
constants
kllNi
k12N1N2
(105)
O and N2
N = !(p317kpMwWu/4nDwkTrF)
(106)
86
VII.
= t[(K2 + 4CXS)1/2 -
K]
(107)
2.
87
..
..tt-
Fig.
9.
Electron micrographs
of poly-
.
88
where
K
[{ex
(108)
+ CMC)/CMC](Ad/as) - S + ex
ifC;?;CMC
C=S-
(109)
Ad/as
= 3VJrd
(110)
where J-d and rd are the total volume and surface average radius of the
monomer droplets, respectively.
The rate of radical absorption in monomer droplets is given by Eq. (89) if
index p is taken to represent monomer droplets, or simplified by an
equation similar to Eq. (90)
PAd
ka1dNd
= p.
I kpMw
+ Pd
kpMw
hcr
kaMdNd
kpMw
jMcr
(111)
4nDwrdNdFd
+ 4nDwrdNdFd + 4nDwrmNmFm
kpMw
(112)
kpMw
kpMw
+ 4nDwrdNdFd + 4nDwrmNmFm
(113)
= N~iC - CMC)
(114)
..
--
2.
89
A2
per molecule.
S
rd (pm)
0.05
0.10
0.15
0.25
0.50
1.00
2.00
5.00
10.00
Nd(dm-3)
5.73
7.16
2.12
4.58
5.73
x
x
x
x
x
1017
1016
1016
1015
1014
7.16
8.95
5.73
7.16
X 1013
X 1012
X 1011
X 1010
= 2 gJliter
e (gJliter)
Pd
0.04
-1
-1
-1
-1
0.09
0.14
0.26
0.65
1.18
1.57
1.83
1.91
-1
0.997
0.847
0.162
0.037
S
Pb
e (gJliter)
and Probabilities
= 5 gJliter
Pd
1.7 x 10-5
0.11
4.2x 10-5
0.30
8.0x 10-5
0.61
-1
1.9 x
6.8 x
2.7 x
9.0 x
1.1 x
1.0 x
1.56
3.28 .
4.14
4.57
4.83
4.91
0.997
0.781
0.379
0.116
0.019
0.005
10-4
10-4
10-3
10-3
10-2
10-2
-1
-1
Pb
1.7 x 10-5
4.2
8.0
1.9
5.3
1.0
1.2
1.3
1.3
x
x
x
x
x
x
x
x
10-5
10-5
10-5
10-4
10-3
10-3
10-3
10-3
Calculated from Eqs. (107-110) and (112-113) as a function of average droplet size rd'
Monomer
volume 300cm3/dm3
H20. Other parameters:
Dw = 5 X 10-8 dm3/sec;
Dp = 1 X 10-8 dm2/sec; kp = 300 dm3/mol sec; Mw= 4 X 10-3 mol/dm3; Mp = 5 mol/dm3;
a = 104; 1/10"; 100 mV, K = 5 X 108 m-I, CMC = 1.5 g/dm3.
Q
90
of particIes stemming from the micelles Nmmuch higher than Nd. It should
be pointed out, however, that this does not necessarily mean that there is no
initiation in droplets. They will, however, in most cases be difficult to detect
and even to separate from particIes initiated in the aqueous phase. The
higher total rate of consumption of monomer in the large number of
particIes initiated in the aqueous phase implies that even if monomer
droplets have become initiated they will during polymerization shrink
because of transfer of monomer to the growing particIes initiated in the
aqueous phase.
The experimental results shown in Fig. 10 illustrate the condition for
droplet initiation. In this figure the total number of particIes at the end of
polymerization is plotted as a function of the concentration of free emulsifier in the aqueous phase for three different levels of emulsifier concentrations. One moves from right to left in the figure as the degree of
../
'
0\"
~~.
8,<:,0
4,\
~o
~~
\KtA8~
....
I
E
~~~,yl
"
VP'
~L~
7. I
~
A
1015.
0.1
~~
~it.
2g
5..
10..
PPS
~
4
Ct
1.0
BP
<>
V
10.0
2.
91
Azad,A. R. M., Ugelstad,J., Fitch, R. M., and Hansen,F. K. (1976).Am. Chem.SocoSymp. Ser.
24, 1-23.
.
Barrett, K. E. J., (1975). "Dispersion Polymerization in Organic Media." Wiley, New York.
Byron Bird, R., Stewart, W. E., and Lightfoot, E. N. (1960). "Transport Phenomena." Wiley,
New York.
Chen, C-Y., and Piirma, I. (1980). J. Polym. Sci. Polym. Chem. Ed. 18, 1979-1993.
Danckwerts, P. V. (1951). Trans Faraday Soco 47. 1014.
Derjaguin, B. V., and Landau, L. (1941). Acta Physicochim. USSR 14, 633.
Dunn, A. S., and Al-Shabib, W. A. (1980). In "Polymer Colloids 11" (R. M. Fitch, ed.), pp.
619-628. Plenum Press, New York.
Fitch, R. M. (1973). Dr. Polym. J. 5, 467.
Fitch, R. M., and Shih, L-B. (1975). Prog. Colloid Polym. Sci. 56, 1.
Fitch, R. M., and Tsai, C. H. (1971). In "Polymer Colloids" (R. M. Fitch, ed.), p. 73.
Fitch, R. M., and Watson, R. C. (1979).J. ColloidInterfaceSci. 68, 14-20.
Fitch, R. M., Prenosil, M. B., and Sprick, K. J. (1969). J. Polym. Sei. Part C 27,95.
Friis, N., and Hamielec, A. E. (1976). Am. Chem. Soco Symp. Ser. 24, 82-91.
Gardon, J. L. (1968a). J. Polym. Sci. Part A-l 6, 623.
Gardon, J. L. (1968b). J. Polym. Sci. Part A-l 6, 643.
Gardon, J. L. (1968c). J. Polym. Sci. Part A-l 6, 665.
Gardon, J. L. (1968d). J. Polym. Sci. Part A-l 6, 687.
92
3
Theoretical Predictions
of the Particle Size and
Molecular Weight Distributions
in Emulsion Polymerizations
Gottfried Lichti, Robert G. Gilbert, and
Donald H. Napper
94
95
96
99
105
105
109
114
115
115
116
120
124
134
139
141
142
143
POLYMERIZATION
Copyright
<!:>1982 by Academic Press, Inc.
Al! rights of reproduction
in any form reserved.
ISBN 0-12-556420-1
94
l.
95
A. Simple
~D =
{[f
D2n(D, t) dD
I f n(D,
t) dD
]-
[f Dn(D,
t) dD
If n(D, t) dD J}
1/2
It may appear at first sight that 15 and ~D in fact are the only measures
of the PSD necessary (in a unimodal system), because of the following
argument. If one supposed that each latex particle in an emulsion polymerization were to grow at constant rate in terms of the volume added per unit
time, then the mean square volume deviation of the PSD would be constant
after Interval I (the nucleation period). The PSD expressed in terms of
volume would thus translate along the volume axis (i.e., the distributions
would be superimposable) as the experiment progressed in time. On the
other hand, if each particle were to grow ata constant rate in terms of
radius added per second, then the radius distribution would translate along
the radius axis after Interval 1. In fact, a real latex system cannot obey
either type of behavior exactly because not all particles actually contain a
growing free radical at any given instant. The foregoing argument involving
the translation of the PSD on the size axis remains valid provided only that
the fluctuations in the average polymerization times from particle to
particle be negligible. This permits the real particle tha~ polymerizes for half
the time to be replaced by an "equivalent" particle that grows continuously
at half the rate of a particle containing one free radical. Such considerations
are clearly of only limited value, and more complete measures of the PSD
are required.
Another measure of the PSD, one used by Gerrens (1959), is the mean
square volume deviation, which is directly proportional to the mean square
deviation of the diameter cubed. Gerrens was aware that such a description
96
is only meaningful when all the PSDs to be compared have the same
Gaussian or normal distributions. His experimental evaluation of several
PSDs of polystyrene latexes showed that the distributions were almost
Gaussian when expressed in terms of the particle volume.
An additional measure of the PSD is the skewness parameter s, defined
by Gardon (1968b) as the exponent of the radius such that the PSD
expressed in the size parameter (radius)' is most nearly Gaussian.
The preceding approaches to characterizing the PSD are adequate if the
PSDs always closely approximate a particular type of distribution. If,
however, the shape of the PSD changed markedly during polymerization,
or if the PSD do es not conform to a recognizable functional form, greater
mathematical sophistication is required.
B.
(2)
for all i ~ O. Here p is the first-order rate coefficient for the entry of free
radicals into latex particles, k the first-order rate coefficient for exit (desorption) of free radicals from the particles and 2e pseudo-first-order rate
coefficient for the annihilation of pairs of free radical s by bimolecular
termination. Note that in Eq. (2), the term with a negative state coefficient
(for i
= O) is
termination
only contributes
if
i ~ 2. Note, too, that p and k give, respectively, the average number of free
radical s that enter and exit from each particle in unit time and that the
average lifetime of an isolated pair of free radicals in a latex particle is
about !e.
,;
...
97
N(t)
(3)
100
n(1, t) d(1
L n(1, t)
(4)
The n(1,
t) term is a function of two independent variables, (1and t. The
equations describing them are a family of partial differential equations. The
basic forro of these equations, as derived from population balance considerations, has been recognized by many authors (O'Toole, 1969; Sunberg
and Eliassen, 1971; Pis'men and Kuchanov, 1971; Min and Ray, 1974;
Lichti et al., 1977) and may be written in the compact but general form
.
onlot
= en - o(Kn)lo(1+ e
(5)
Here n(1,t) i~ a vector whose (i + l)th component represents the distribution n(1,
t) and e is a square matrix formed from the kinetic parameters
p, k, and c. The nature of e may be recognized by expressing Eq. (2) in
matrix form and identifying the resulting square matrix with e. e is called
here the Smith-Ewart coupling matrix because its elements in Eq. (5)
describe how latex particIes change state as a consequence of the SmithEwart mechanisms (entry, exit, and bimolecular termination).
The matrix K in Eq. (5) is a diagonal matrix ~hose (i + 1, i + l)th
element specifies the rate of growth of a latex particIe in state i. The term
-o(Kn)lo(1 in Eq. (5) accounts for the particIe growth process whereby
particles of size (1 are lost to the population n(1,
t) when they grow to
another size (1+ d(1,whereas particles are gained in the population n(1,
t)
when particIes of size (1 d(1 in state i grow to size (1(Lichti et al., 1977).
One common assumption regarding K is that
K+l,+l = iK
(6)
98
99
Batch Polymerizations
1. lnterval II
Interval 11 of an emulsion polymerization is characterized by polymerization in a constant number of latex particles in the presence of monomer
droplets (Le., nucleation is absent). This situation usually exists in an ab
initio polymerization immediately on completion of Interval 1; however, for
a seeded system, it may exist from the commencement of polymerization.
Clearly, if no new particles are formed and coagulation does not occur, the
vector e in Eq. (5) is zero. Moreover, in the presence of monomer droplets,
the concentration of monomer at each polymerizing site in the particles is
approximately constant. The growth factor K is thus known. The remaining parameters p, k, and e in Eq. (5), or rather size-averaged values for them,
can be evaluated from kinetic studies (Hawkett et al., 1980, 1981;
Lansdowne et al., 1980). All the parameters in Eq. (5) are then specified and
it is possible to predict uniquely the time evolution of the PSD. Of course,
the initial distribution n~q, t = O)must also be specified (Lichti et al., 1977).
Provided that a relatively monodisperse seed latex is used, the size-averaged
kinetic parameters can be employed without introducing significant error.
In this way, the predictions of Eq. (5) can be tested against experiment and,
e.g., the accuracy of the kinetic parameters checked by PSD evolution data.
Additional experiments with polydisperse seed latexes may permit the size
dependence of the parameters to be specified.
a. The Solution of the Population Bala'lce Equations. The solution of
Eq. (5), given the initial conditions and all the requisite parameters, can be
achieved in at least three different ways: (i) by analytic solutions, if they
exist; (ii) by finite difference numerical solutions which may always be
generated (Carnahan et al., 1969), although they may require considerable
computing time and are prone to inaccuracy; or (iii) by the method of
moments (Bamford and Tompa, 1954) which provides an efficient numerical
procedure for certain systems.
Analytic solutions for Eq. (5) provide the most direct path of the
prediction of PSD evolution. For batch polymerizations in Interval 11,
however, analytic solutions have only been achieved forothe so-called zeroone system (Lichti et al., 1981). These are systems wherein negligibly few
particles contain two or more free radicals because of the rapidity of the
bimolecular termination reaction (e.g., in styrene emulsion polymerizations
with smalllatex particles). In this case, Eq. (5) may be written as follows:
ano/at = - pno + (p + k)n
(7)
andat = pno (p + k)nl - aKndaq
(8)
100
The appearance of the term pnl in Eq. (7) reflects the fact that entry into a .
state 1 latex particIe causes a state zero latex particIe to be formed as a
consequence of rapid bimolecular termination. Ir all the paramaters p, k,
and K in Eqs. (7) and (8) can be taken to be independent of time, while
being permitted to be arbitrary functions of size, analytic solutions may be
generated for arbitrary initial conditions (Lichti et al., 1981). These solutions encompass the more restricted cIass of analytic solutions provided
by O'Toole (1969) and Watterson and Parts (1971). The use of the term
analytic is retained Qere even though, for arbitrary initial conditions, one
numerical integration is required; such integrations are very rapidly performed numericalIy. The importance of this analytic solution is that it enables
one to make a thorough comparison between experimental PSDs and
theoretical predictions, with a wide range of assumptions as to the values
and functional forms of the various contributing rate coefficients.
The analytic solution derived by Lichti et al. is applicable to any zeroone system in lnterval n, with no restrictions on the rate coefficients except
that they be independent of time (but may depend on size). Although a
zero-one description is applicable to some important systems, it is not
universally
t, general
analytic solutions
have yet to be developed, and we now examine methods for solving Eq. (5)
for systems of arbitrary .
The second method for solving the PSD evolution equations is bruteforce numerical solution using first-order finite difference. Whereas a solution can always be obtained by this technique, it suffers from numerical
instability, from the lack of any automatic check on accuracy, and from
requiring large amounts of computer time.
The third method for solving Eq. (5), the method of moments, was
exploited by Katz et al. (1969) in connection with MWDs (see Section
n,E,2) and by Sundberg and Eliassen (1971) in connection with PSDs. This
method is numerically efficient. The kth moment mf of the population
n(a,t) is defined by
mNt) = 1<0akn(a,t) da
(9)
Equations involving only a finite number of mf may be derived from Eq. (5)
under appropriate cIosure approximations. As only one variable (t) is then
involved, the moment equations may be readily solved and the populations
n(a,t) generated. This method will often fail, however, when Eq. (5) is
nonlinear (e.g., if reentry is important) or if large numbers of moments are
equired (e.g., for a polymodal PSD).
b. Theoretical Results. To illustrate the use of the analytic solutions
mentioned above and to highlight the general behavior exhibited by the
101
0">0
O"~O
and
(10)
which corresponds approximately to a Gaussian distribution function. In
Fig. 1, we chose O"= V (volume) as the size variable.
Figure 1 depicts the evolving population no(O",
t) and nl(O"'t), as well as
the overall observable PSD n(O",t). Initially, alllatex particles are nongrowing. As polymerization proceeds, more particles enter the growing state.
After some time, the total number of latex particles in each state is equal
(Le., = 0.5), although the population nl(O"'t) is marginally further advanced in size than the nongrowing population no(O",
t). This must be so because
the population nl(O"'t) is growing. The overall PSD moves along the size
axis and broadens (as measured by the decreasein peak height of the
normalized curves) with increasing time. The rate of broadening is greatest
when the polydispersity is least (e.g., the PSD broadens significantly more
in the initial time interval 0-2 than in the next interval 2-4).
It is stressed that the PSD may be expressed as a function of any
arbitrary size variable. For example, distributions n(V) and n(r)expressed in
terms of the particle volumeand radius, respectively,are related by
n(V) dV = n(r) dr
(11)
n(V) = n(r)/41tr2
(12)
102
Gottfried
Lichti el al.
(o)
100 r
80
60
>
o
c: 40
20
(b)
100 r
80
60
>
c: 40
1=2
20f
.
1
(c)
100 r
80
60
>
40
20
Fig. 1.
103
dV/dt
=K
(13)
If the same PSD is now expressed in terms of the radius, the term K(r)
corresponds to the radius increment to the host particle in unit time as a
result of growth of a single free radical. Equation (13) yields
dr/dt
= K/4nr2 = r(r)
(14)
which is the requisite form of the same growth factor expressed in terms of
the new variable r.
2.
lntervall
in Interval I are in
doubt and the relative importance of homogeneous (Le., oligomeric prcipitation) versus heterogeneous (Le., micelIar) nucleation mechanisms are
not fully understood. For this reason, detailed solutions to Eq. (5) in this
Interval, when e is nonzero, appear to be premature. Moreover, in many
emulsion polymerizations, the precise detail~ of events occurring in Interval
I are masked by the subsequent particle growth in Intervals n and nI.
The global strategy endeavors to enCQIIlpass all possible nucleation and
coagulation m~chanislI\s. Min and Ray (1974) have identified four different
mechanisms that lead to a change in the number of particles, although this
list may not be exhaustive: (i) the coagulation of latex particles with
micelles, (ii) the mutual coagulation of latex particles, (iii) the entry of free
radical s into micelles, and (iv) the precipitation of an aqueous-phase
oligomeric radical that has exceeded its criticallength.
The four processes
all contribute to the vector e in Eq. (5). These terms are evaluated in a
nonrigorous fashion. For example, the rate of coagulation of particles of
volume V1 with those of volume V2 is taken by Min and Rayas proportional to the product of the respective numbers of particles, the constant
of proportionality being given by
kc
= C1 exp( -
(15)
104
With this type of approach, two comments must be borne in mind: first,
at this level of complexity, the proliferation of adjustable parameters almost
assures that any experimental data can be fitted (Le., the model is virtuaIly
unfalsifiable experimentaIly); second, only plausibility arguments were used
to establish Eq. (15). It might be argued, for instance, that E* should be a
function of particle size for charged species and that a formalism that treats
miceIles and smaIl latex particles in an identical fashion is unrealistic.
Furthermore, all particle sizes under consideration in Interval I are relatively smaIl 10 nm typicaIly); arguments leading to Eq. (15) have only
been validated for considerably larger particles. These criticisms are meant
only to stress that the coIloid science associated with Interval I is still
poorly understood. Progress in this are a is therefore likely to be slow.
Fortunately, as mentioned previously, particle formation by whatever
mechanisms are operative usuaIly produces particles whose size is smaIl
relative to that attained at the end of Interval In. It then becomes possible
to assume with little loss of accuracy that aIl particles are formed at the
same (smaIl) size.
3. IntervalIlI
.
dCM/dt
= ACM
(16)
= i~O
n(a,t) da
(17)
...
105
Sem;cont;nuous
Emuls;on Polymer;zat;ons
Semicontinuous emulsion polymerizations are characterized by the continued addition of monomer to the reaction vessel. This permits the
production of latexes with high weight percentage solids while allowing the
initial burst of nuc1eation to be achieved in substantially aqueous surroundings. The theory for semicontinuous systems is substantially that set forth
for Interval In of batch polymerizations, except that the materials balance
equations [Eq. (17)] must be modified to inc1ude the flow of new material
into the reactor. The effect of the monomer input is twofold: first, the mass
of material present in the system is increased and second, the concentration
of other reagents may be reduced.
From a practical viewpoint, a semicontinuous process may be used to
increase the duration of Intervals n or In. Thus, for example, by extending
the duration of an Interval In which displays a Trommsdorff gel effect with
its concomitant rate acceleration, a shorter reaction time may be achieved.
E.
A continuous emulsion polymerization is characterized by (i) the continual addition of monpmer, surfactant, and initiator to the reaction vessel
and (ii) the withdrawal of a steady stream of reactor fluido Several such
reaction vessels may be set up in series with one another so that the effiuent
from the first feeds the second, etc. Any species in such a system remains in
the reaction vessel for a finite time, termed the "residence time." Such
reactor vessels may be either completely or incompletely mixed and operate
under either isothermal or nonisothermal conditions. The incompletely
mixed, nonisothermal reactor is characterized by the existence of spatial
concentration and temperature gradients in the reactor.
The principal advantage of continuous reaction vessels is that they
operate (after an initial transient period) under steady-state conditions that
are conducive to the formation of a highly uniform and well-regulated
product. In this section, we shall confine the discussion to continuous
stirred-tank reactors (CSTRs). These reactors are characterized by isothermal, spatially uniform operation.
106
-NJr
(18)
where t is the mean residence time. This may be readily caIculated from the
volume of the reactor and the efflux rateo
The addition of a term such as that in Eq. (18) to Eq. (5) makes a
profound difference to the evolution of the PSD in time. For example, true
steady-state solutions to Eq. (5) may be generated by setting the lefthand side equal to zero. The PSD does not evolve in time in that case but
remains static under the competing influences of particle nucleations and
remo val. This type of behavior is unique to continuous systems. Uually,
steady-state conditions prevail in eSTRs only after a period equal to
several mean residence times has elapsed.
Thompson and Stevens (1977) have numerically solved the steady-state
version of Eq. (5) for a multistate emulsion eSTR, allowing for all the
usually considered free-radical events. The problem contains only one
independent variable, the particle size (1.It may be solved with considerable
precision using sophisticated finite difference methods. The actual shape of
the steady-state PSD is quite unlike that observed in the batch process (see
Fig. 2), being a monotonically decreasing function of particle volume rather
than a peaked function. This feature of the PSDs of eSTRs arises from the
greater probability of removal of particles in the effluent as their times of
residence (Le., growth times) increases. It is emphasized that the foregoing
argument is valid only when the PSD is expressed in terms of volume; the
PSD produced in a eSTR does show a maximum when expressed in terms
ofradius [see Eq. (12)].
The PSDs produced in emulsion eSTRs tend to extend over a wider
range of particle sizes than those generated in a batch process. This arises
from the constant creation of new particles in a eSTR compared with the
short burst of nucleation in a batch system. One consequence of this
broadness is that reasonably accurate predictions of the steady-state PSD
can be achieved in such reactors without considering "stochastic" broadening factors (Le., broadening caused by the distribution of free radicals
among the particles, as illustrated in Fig. 1 for a batch system). The
formalism of Eq. (5) fully encompasses the contribution from stochastic
broadening by considering the portion of the overall population in each
state i. eSTR PSDs often display insensitivity to stochastic broadening.
This insensitivity to stochastic broadening considerably simplifies the problem of predicting the PSD in a eSTR system. De Graaf and Poehlein
107
(a)
>-
t-
Z
t.s
e
lb)
a:
t.s
ID
~
=>
z
VOLUME
Fig. 2.
Schematic comparison of the PSDs produced by (a) batch and (b) CSTR
emulsion polymerizations.
(1971) showed that it was possible to calculate the PSD from an expression
for the distribution of residence times (and hence growth times), as well as a
growth rate parameter that depends upon the particle size [see Eq. (21)]. The
crux of their argument will now be presented,
The lifetime distribution of any species in the CSTR g(t) is given by
g(t)
Co exp(
tlr:)
(19)
if the effiux rate is uniformo Here Cois a normalizing factor. Suppose that it
is possible to assign a unique value of particle size O'(t)to a particle that
leaves the reactor after a residence time t. This assumption actually
disregards stochastic broadening by ignoring the fact that particles that
leave the reactor after residence time t will in reality possess a size
distribution. The lifetime distribution expressed by Eq. (19) can then be
transformedinto the steady-statePSD as follows[cf. Eq."(11)]:
g(O')
Co exp[
- t(O')Ir:](dtldO')
(20)
= K"
(21)
W8
where ",is the size-dependent average number of free radicals per particle
calculated from a knowledge of the size-dependent parameters p, k, and e
[see Eq. (2)].
The foregoing argument is equivalent to the approach of Stevens and
Funderburk (1972), who used a more abstract line of reasoning. Suppose
that a summation of Eq. (5) is performed over all states i (Le., all the rows
are summed). Then Eq. (5) becomes
.0
(22)
Equation (22) follqws from Eq. (5) if the vector e is given the form
appropriate to CST,s [see Eq. (18)] for removal of latex particles in the
effluent stream and if the particle creation terms are considered as boundary
conditions operating at a = o. The term involvingthe Smith-Ewart coupling matrix vanishes on summation because of the conservation of particle
numbers. The value of ",is given by
(23)
",= L ini/L ni
so that Eq. (22) may be rewritten,as
on/ot
(24)
n
where
= coex{
Co is the number
-(f:da/Kn",-r:)JIK
of newly formed
particles
(26)
present
at a
= O.
109
Batch Polymerizations
Min and Ray (1974) in their review article detailed many experimental
studies on the PSDs of emulsion polymers. In this section, some more
recent results that compare the predictions of the population balance
equations [Eq. (5)] with experimental data will be discussed.
The actual acquisition of reliable experimental PSD data is a task of
considerable complexity. Such techniques as soap titration, turbidity, ultracentrifugation, and light scattering (Morton et al., 1954; Kerker, 1969;
Shaw, 1970) usualIy pro vide only average values for the particle diameter,
although some of these methods can provide PSD data if some functional
form for the distribution function is assumed. The Coulter Counter can
usualIy be exploited only for particles oflarger size (> 0.5 ,um), although it is
not an absolute method (Eckhoff, 1967). Fractional creaming of latexes has
also been proposed as a method for determining PSD (Schmidt and Kelsey,
1951; Schmidt and Biddison, 1960). This method is based on a relationship
between the concentration of creaming agent and the size of the creamed
particles, which is claimed to be independent of the nature of the dispersed
particles. It has the advantage of simplicity, but again calibration is
mandatory. Hydrodynamic and liquid exclusion chromatography (Nagy et
al., 1980; Singh and Hamielec, 1978) have also been proposed as methods
for determining the PSD, although difficulties still exist in relating the
detector signal to the PSD.
By far the most direct method for the measurement of PSDs is the use of
transmission electro n microscopy. The measured PSD is a distribution
expressed in terms of the unswolIen radius (or diameter) of the particles. AlI
sizes are treated with the same statistical weight (provided they are able to
be resolved); this is not the case with some of the sizing methods listed
above (e.g., light scattering), which weight the larger particles more heavily.
One major difficulty with electro n microscopy is that so me latex particles
are adversely affected by the electro n beam. Shadowing techniques and
hardening procedures (Corio et al., 1979) can sometimes be used to
overcome this problem. A further difficulty inherent in any PSD measurement by electro n microscopy resides in the need to measure a large number
of images of latex particles on a photographic print. Gerrens (1959) has
stated that at least 3000 particles must be measured in order to obtain a
statisticalIy meaningful distribution.
This involves considerable
human
110
effort because current attempts to automate the procedure have thus far
foundered on the difficulty of resolving the extensively overlapping particles
tht inevitably occur in samples of sufficient number density to give
meaningful statistics. It seems likely that sophisticated image analysis
software andjor software that can resol ve overlapping particles will become
commercially available in the near future.
An additional complication arises with relatively mono disperse samples
in that slight changes in the focus of the electro n microscope introduce
random errors in the exact magnification of the final print. These problems
of focus adjustment can be corrected for by the use of an internal standard
(frequently another polymer latex with a significantly larger average particle
size). It is usually mandatory for the PSD of the internal standard to be
completely separate from the PSD of the sample.
Sundberg and Eliassen (1971) ha ve compared the predictions of their
population balance formalism with the experimental PSD data obtained by
Gerrens (1959) for polystyrene latexes. Their model is a zero-one system
with no exit (desorption) from the particles. Although the neglect of radical
desorption for particles of this size is questionable (Hawkett et al., 1980;
Lansdowne et al., 1980), the theoretical curves are in qualitative agreement
with the experimental PSDs (see Figs. 3 and 4). It must be stressed,
however, that the experimental PSDs were taken at the conclusion of the
6
32
1
2
3
4
5
6
28
">
o
H
"
24
20
16
12
8
4
O
500
1000
1500
2000
Fig. 3.
111
z 0.6
52
1
Z
~
I.J...
Z
O
: 0.4
~
m
a:
1IJ)
~
0.2
...J
O
>
o
Fig. 4.
0.5
1.0
1.5
VOLUME
2.0
(1971). Initiator concentration curves (mol dm-3): 1, 1.1 x 10-3; 2, 2.2 x 10-3; 3, 5.5 x 10-3;
4, 1.1 x 10-2. Volume is dimensionless. (Reproduced with permission of Plenum Publishing
Corp.)
112
N
I
o
~
0.2
>-
1Vi
Z
UJ
e
...J
~ 0.1
Ir
o
Z
80
PARTlCLE
Fig. 5.
DIAMETER
160
Inm)
113
7e
~
60
50
PSD KINETICS
2.2
2.3
2.3
2.4
m3 $-1
2:
e
I
o
~\
\
.\
,\
10
Fig. 6.
Comparison of theory and experiment for a seeded emulsion polymerization.
Continuous lines: smoothed partic1e size distribution obtained experimentally; broken line:
theoretical fit using values of p and K shown in inset (values of p and K from kinetic study also
shown) at t = 45 mino (After Lichti et al., 1981; reproduced with permission of Journal of
Polymer Science.)
114
1.0
z 0.8
O
i=
=>
ID
Q?
1- 0.6
VI
i5
ILI
>
~ 0.4
-1
=>
=>
00.2
O'
100.
200
300
PARTICLE DIAMETER(n m)
Fig. 7.
A comparison of theory with experiment for the normalized cumulative PSD,
proportional to Son(O',t) dO',where O'= diameter, obtained by De Graaf and Poehlein (1971) for
a polystyrene CSTR: --- = 0.5; > 0.5; O experimento (After De Graaf and Poehlein,
1971; adapted with permission of Journal of Polymer Science.)
are of dubious validity for the small particle sizes studied (Hawkett et al.,
1980).
G.
115
A. lntroduction
The molecular weight distribution (MWD) of a polymer generated by
emulsion polymerizations can be fundamentally different from that generated in solution or bulk. For example, in styrene emulsion polymerization
the MWD of formed polymer has a much higher average molecular weight
than may be obtained using other methods. The basic reason for this was
postulated by Smith and Ewart (1948) to be the compartmentalization of
the polymerization reaction inside the latex particles which leads to the
isolation of free radicals. This isolation reduces the probability of bimolecular terminations and hence increases the degree of polymerization.
The fundamental difficulty in constructing a theory for the MWD in
emulsion polymers is to account for the compartmentalized nature of the
system. In the commonly occurring situation where particles contain only a
few free radical s at any given time, it is obviously incorrect to consider that
each latex particle behaves like a "mini-bulk" reaction vessel, and so the
conventional methods used for bulk polymerizations are inapplicable.
Nevertheless, some assumptions which introduce only minor errors may
often be made. The most important such assumptions is that the evaluation
of the MWD may be separated from that of the PSD. In other words,
provided that the MWD being produced at any given moment is the same
as would be formed in an equivalent set of monodispersed latex particle
systems [as expressed in Eq. (27) below], then the MWD evolved in a
system that is polydispersed in size may be computed trivially. Formally,
this is expressed as follows. Let S(M, a, t) be the MWD formed in a
monodisperse' system 'Of size a at time t; here M is the molecular weight
variable. In a polydisperse system with PSD n(a, t) the overall MWD at any
experimental time te, defined as S(M, te),is postulated as
""
re
(27)
S(M, te)= O dt O da n(a, t)S(M, a, t)
ii
A detailed critique of the validity of Eq. (27) is given in Section III. In brief,
two criteria are required for this validity, both of which are well satisfied in
ordinary emulsion polymerization systems. These are (i) that the time
required for formation of a single polymer chain be much less than that
over which significant changes occur in the rate coefficients governing the
MWD and (ii) that the average number of free radicals per particle of size a
is close to its steady-state value.
116
Elementary Concepts
In this chapter, the term MWD refers to the number density MWD,
S(M), which gives the relative number of polymer molecules of molecular
weight M. Another commonly used measure is the weight density MWD,
W(M), giving the relative weight of polymer molecules. The two distributions are related by
.
W(M)
= MS(M)
(28)
Thus, the relative number of chains in the sample with molecular weights in
the range MI::;;;M ::;;;M2 is given by
N[MI,M2J = rM2S(M)dM
(29)
JMI
and the relative number in a small range M to M + 11M by S(M) 11M. The
normalization (and thus the dimensions) of S(M) will be specified later.
In the ensuing development, S(M) will refer to the "instantaneous"
MWD, Le., the contribution to the MWD created over a comparatively
short time periodo Ir all polymerization occurs under steady-state conditions, S(M) then gives the final MWD of formed polymer directly via Eq.
(27); otherwise, the final MWD is given by a simple time integration over
S(M), as given in Eq. (27) and amplified in Section II,C,2.
1. Review of Bulk MWD Theory
In bulk or solution polymerizations, the following free-radical mechanisms are usually considered: initiation, propagation, chain transfer (to
monomer, polymer or transfer agent), and bimolecular termination (by
117
combination or disproportionation). Except for initiation, all these reactions are bimolecular. Initiation and transfer are chain-starting reactions,
whereas transfer and bimolecular termination are chain-stopping.
A particularly simple theory for the MWD in bulk or solution polymeri. zations
(30)
(31)
(32)
where we have used Eq. (31) together with Eq. (30) to relate M and t'.
However, in the case of chain stoppage by bimolecular combination,
pairs of free radicals join together to form a chain whose "equivalent"
growth time t~q is the sum of the growth times of the two contributing
chains. Now, for bulk or solution polymerization, the product G(l')G(t)
specifies the distribution of pairs of growing chains of growth times t' and
t, since in a bulk system any growing chain is equally likely to undergo
118
bimolecular combination with any other chain (it will be seen that this does
not hold in a compartmentalized system). The distribution of dead chains
created by bimolecular combination with equivalent growth time t~q,
denoted Sbc(M) (where M = at~q) is thus proportional to the sum of alI
possible pairs of growth times giving this equivalent growth time:
t~q
Sbc(M
= at~q)oc. o
(33)
G(t' - t~)G(t~)dt~
(34)
At~q)
J/[LooMS(M)
dM
(35)
If alI chains have the same molecular weight, P = 1; otherwise P > 1. For
monomodal MWDs, the larger the value of P, the broader the distribution.
For the MWD of Eqs. (32) and (33), P = 2 and 1.5, respectively.
2. Emulsion Polymer MWDs: Concepts and Limiting Cases
The concepts introduced above require extension and revision when we
turn to an emulsion polymerization. In view of the above discussion, we
may take the system to be monodisperse in volume. First, we briefly
consider the kinetics of such a system. As described in Section I,B, this is
given by the distributions N(t), which are in turn given by the first-order
rate coefficients for entry (p), desorption (k) and bimolecular termination (c).
We note here that
c
= Cbc + Cd
(36)
= ktr.MCM + ktr.A[A]
(37)
where ktr.Mand ktr.Aare the rate coefficients for transfer to monomer and to
chain transfer agent A; we assume that if transfer of free radical activity to
A occurs, the species so produced polymerizes further. It is postulated that
119
f = ktr.MCM
+ ktr.A[A]- k
(38)
We will consider the MWD in two simple cases. The first is when chain
transfer is sufficiently rapid to ensure that all other chain-stopping events
can be ignored. In such a situation, whereas the compartmentalized nature
of the reaction may affect the rate of initiation of new chains, it will not
affect the lifetime distributions of the chains once they are formed. The
MWD may then be found from the bulk formulas, provided only that , the
average number of free radicals per particle, is known. Such an approach
has been used by Friis et al. (1974) to calculate the MWD evolved in a vinyl
acetate emulsion polymerization. These authors included in addition the
mechanisms of terminal bond polymerization and of transfer to polymer
(both of which cause broadening). The formulas required for the incorporation of these mechanisms could be taken from bulk theory.
It is important to note that, even in this present limiting case of a
transfer-dominated system, the chain-stoppage mechanism can be changed
by compartmentalization. Thus, the MWD formed in the polymerization of
styrene appears to be transfer-dominated in some emulsion systems (Piirma
et al., 1975) but to be combination dominated in bulk or solution (George,
1967). This difference occurs because, in ~tyrene emulsion systems, the rate
of radical entry into a particle is slow, and most particles usually contain
either zero or one free radical. In the "state one" particles (Section I,B), the
growing free,radical ras time to undergo several transfer reactions before a
further entry causes radical annihilation.
The second simple case we consider in this section arises in a zero-one
system (Section I,C,l,a), where the average time between successive entries
of free radicals into a latex particle is short compared with the time taken
for a growing chain to undergo a transfer reaction. In this case, most chains
cease growth by bimolecular termination. Ir this happens by combination
with a low molecular weight entrant species, the following approach is
possible. When entry occurs in a state zero particle, a free radical is
formed and grows uninterrupted until the next entry event. Since the rate
coefficient for entry is essentially constant, Eq. (31) gives the distribution of
growing chains. Since the terminating free radical is of negligible molecular
weight, the MWD will then be given by Eq. (32). The average chain growth
time A.-1 is the average time between successive entries into the same
particle (viz., p -1; i.e., p = },).
120
In concluding this section, we note that except for the limiting cases
given above, the phenomenon of long-chain branching has not thus far been
incorporated into the models for emulsion polymer MWDs. In the general
MWD formulation given below, it is assumed that these branching events
occur with negligible frequency.
C.
te = t + t'
(39)
from the above definition. It will be seen that this feature of the formalism
has the important consequence that the evolution equations governing
N;(t, t') are simply differential equations containing terms in only one of the
independent variables, viz., t'.
For N;, the minimum value of i is unity, not zero, since the definition
requires the presence of at least one growing chain. If N;(t, t') is integrated
over all values of the growth time t', the resulting integral counts each of the
free radicals inside the particle. Thus, from Eq. (36), we have the following
normalization:
rte
iNi(te)
= lo
(40)
L Ni = 1),N; has
3.
121
(41)
This has been derived in detail by Lichti et al. (1980). In brief, this equation
states that singly distinguished latex particles in state i form (i) by entry into
an N-l type particle, (ii) by desorption of nondistinguishing radicals from
an N+l particle, (iii) by termination among any of the (i + 1) nondistinguishing radicals in an N+2 particle. Similar statements hold for the
loss term in Eq. (41).
Since Eq. (41) involves variation in t' only, it may be solved given the
initial conditions, viz., N(t, t' = O). These last quantities are the distributions of ordinary latex particles in state i inside which a new chain begins
growth at time t:
N(t, t' = O)= pN-l(t) + ifN{t)
(42)
Equation (42) asserts that entry and transfer (involving any of the i chains
present) are the only chain starting reactions.
Note that since N(t, t') usually decay monotonically in t', no steady-state
approximati<:>ncan ~e made to simplify Eq. (41). Nevertheless, there are
only a set of differential equations in one variable (as noted above) which
can readily be solved (e.g., as in Section II.D.2).
Having specified N, we now show how to compute the required MWD,
or more specifically, the component of the MWD arising from the mechanisms considered in this section. This is the distribution of dead chains
formed as a result of transfer, exit (desorption), and termination by disproportionation. We denote the components ofthe instantaneous MWD arising
from each of these mechanisms by Str(t., M), S.it., M) and Sd(t., M), where
t. is the experimental time. We have
L fN(t., t')
S.x(t., M) = L kN;(t., t')
i~l
Stit., M) =
Sd(t., M) =
(43)
i;?;l
L 2(i -
i?t2
I)CdN(t., t')
(44)
(45)
122
Here t' and M are related by Eq. (30). The right-hand sides of Eqs. (43-45)
are the rates at which the distinguishing growing chain can cease growth by
each of the events considered.
It can be seen that Stn Sex, and Sd each has dimensions of(time)-z. The
meaning of these distribution functions is as follows: the number of chains
formed by transfer whose growth times are between t' and t' + dt', over the
range of experimental time from te to te + dte, is given by SIrdt' dte (a
dimensionless quantity), and similarly for Sex and Sd' Note that the
expression for Sd is simpler than that given by Lichti et al. (1980), although
the more complex expression for Sd given by these authors reduces to Eq.
(45) after suitable manipulation.
2. " Doubly Distinguished" Particles
The N; values are insufficient to account for the contribution to the
MWD from bimolecular termination by combination: since tfiis requires a
knowledge of the growth times of each of a pair of growing chains.
In the bulk system considered in Section II,A,l, the distribution of free-
radical pairs with growth times '1/1 and t~ was shownto be givenby the
product G(t~)G(t~).However, the equivalent relation in an emulsion system
does not hold. A simple counter example will illustrate this.
We consider a system containing only two particles in state 2. Let
particle 1 contain growing chains with growth times t~ and tB' and particle
2 contain growing chains with growth times t~ and tD' The fraction of
particles in state 2 which contain a chain of growth time t~, which we
denote N~(t~), is 0.5, and similarly for N~(tB)'N~(tC),and N~(tD)' Note that,
as in Eq. (40)
N~(t~) + N~(tB)+ N~(t~) + N~(tD) = 2
since each particle is counted twice in this expression. Now, the fraction of
particles that contain a pair of chains of growth times t~ and t~ is evidently
zero, 110t N2,(t~)N~(tC)= 0.25, as would be given by the product formula.
The reason for this difference between bulk and compartmentalized systems
is that whereas all free radicals in a bulk system are mutually accessible, in
an emulsion system only radicals inside the same latex particle are mutually
accessible for bimolecular termination.
We therefore define a new distribution N?(t, t', t") as being that of
"doubly distinguished" latex particles in state i, Le., particles in which one
free radical began growth at time t and continued growing for a time t', at
which time (t + t') another free radical began growth in the same particle,
and both are still growing at time t" later, the particle being in state i. From
this definition, the experimental time te is given by
te
= t + t' + t"
(46)
123
The two free radical s which started at t and t + t' are termed "distinguishing," and the latex particle ceases to be doubly distinguished when either or
both of these distinguishing free radicals cease growth. Obviously, the
minimum value of i for which N7 is nonzero is 2.
We define N;' to be normalized such that the integral of N7 over all
possible values of t' and t" counts all possible pairs of growing free radicals
in particles in state i. Since the number of pairs in a single particle in state i
is ti(i
1), we have
fte ftte
124
We now show how to use N' to obtain the contribution to the MWD
arising from bimolecular termination by combination. When two distinguishing chains, one of which has been growing for time t" (this being the
second chain formed)
+ t"
combination, the resulting single dead chain has an equivalent growth time
of t~q= t' + 2t". The rate coefficient for this event is 2c/(the factor of 2
arising from the usual definition of c as being per pair of free radicals).
Thus, the relative number of chains of equivalent growth time t~qpresent at
experimental time te formed by bimolecular combination is
t,~q
Sbc(te, M) =
dt" L: 2ccN'(te
i
t~q
+ t", t~q-
2t", t")
(50)
~ te
to an excellent approximation.
Equation (50) defines a distribution as do Stn Sex, and Sd, so that the
number of chains formed of growth time in the range t~qto t~q+ dt~q over
the experimental time te to te + dte, is Sbc dt~q dte. These four contributions
to the instantaneous
MWD may thus be directiy summed to give the
overall instantaneous MWD from all sources, S(te, M):
S(te, M) = SIr + Sex + Sd + Sbc
te,
(51)
Finally, we recall that the instantaneous MWD, S(te, M), may vary with
The accumulatedoverall MWD of formedpolymer at time t* is simply
Ct'
(52)
If S is independent of te, one simply has S = teS; such is the case, for
example, during the Interval n steady state. In other cases (e.g., Interval In)
the variation of quantities such as CM (and hence a) with time will be known
from the kinetics; the evaluation of Eq. (52) is then a simple quadrature.
D.
of equations
125
1. Zero-One-Two Systems
We first consider a system in which the maximum number of free
radicals per latex particle is 2. This is termed a "zero-one-two" system. An
analytic solution to the MWD equations for such a system in the Interval II
steady state has been developedby Lichti et al. (1980).
The zero-one-two model accurately describes systems wherein (the
average number of free radical s per particle) does not exceed 0.7; it is thus
applicable to small-particle styrene (Hawkett et al., 1980),vinyl acetate
(Ugelstad and Hansen, 1976), and vinyl chloride (Friis and Hamielec, 1975)
emulsion polymerizations. In the Interval II steady state, the analytic
solutions to the various functions involved in the MWD obtained by Lichti
et al. (1980)are as follows; in these expressions,the dependence on te is
suppressed since a steady state is assumed
(53)
N'l(t') = B1 exp( - A+t') + El exp( - Aj)
N~(t') = B2 exp( where
-t{all
},+ t')
E2 exp(
Aj)
(54)
-p-f - k
k + p/2
-p-2k-f-2c
We further have
N;(t',I t") = [pN'I(t') + fN~(t')] exp( - Qt")
(55)
where Q = p + 2(f + k + c). These results give the singly and doubly
distinguished particle contributions for i = 1 and 2 (no higher value of i
being required for our zero-one-two system) in terms of the various rate
coefficients for the microscopic processes involved. The various components
of the instantaneous MWD are then found from Eqs. (43-45) and (50).
Using the above results, Eq. (50) is found to reduce to
(56)
126
0.5
0.4
10-5M
6
( b)
(o)
0.4
N
I
U
Q)
~ 0.3
z
O
::
::>
~ 0.2
a::
ti>
5
0.3
0.2
0.1
0.1
O
O
O
2
341
GROWTH TIME (sec)
Fig. 8.
Plots of instantaneous molecular weight distribution S arising from instantaneous termination, stoppage, and transfer, as a function of growth time (Iower abcissa) and
molecular weight (for a typical styrene system; upper abcissa), for a steady-state transferdominated zero-one-two emulsion polymerization, with e = O, p = 0.1 sec-I, k = 0.3 sec-I,
f = 0.9 sec-I. (a) Curves I and 2 give conlribulions lo S from N'I and N2, respeclively. (b)
Overall MWD, S. (Mter Lichli et al., 1980; reproduced wilh permission of Journal of Polymer
Science.)
127
(o)
.-
...
I
<.>
CII
'"
5.... 0.002
::>
ID
ir
\
e
TT
o
o
1.0
2.0
1.0
2.0
EQUIVALENT
GROWTH
TIME (sec)
Fig. 9.
(a) Plots
of components
of instantaneous
molecular
weight
distribution
arising from bimolecular combination (Sbe' labeled BC) and instantaneous termination (Sil'
labeled TI) for a steady-state zero-one-two system with termination by combination only,
assuming Ce= I sec - 1, P = 0.1 sec - 1, k = f = Cd= O. (b) Plot of S, the overall instantaneous
MWD, for the same system. Note = 0.55. (After Lichti et al., 1980; reproduced with
permission of Journal of Polymer Science.)
128
0.03
....
I
(\)
'"
~
0.02
z
o
~
:J
CD
g:0.01
IJ)
10
GROWTHTIME (sec)
20
Fig. 10.
Plot of instantaneous MWD S as function of growtb time in a zero-one-two
Interval 11 system witb termination by disproportionation only, assuming p = 0.1 sec-1,
Cd = 0.3 sec-1, c<=
J = k = O. (After
oJPolymerScience.)
witb permission
of Journal
129
oN'jot' = O'N'
where N' is the vector whose elements are N'l' N~, oo.,and the elements of
the matrix n' are defined in Eq. (41); the solution to Eq. (58) is
N'(t') =
exp(Ajt')Lej[RejN'(t'
= O)]
(59)
where the Aj, Lej and Rej are the eigenvalues, left (column) and right (row)
eigenvectors of n'. The initial conditions N'(t' = O)are determined from Eq.
(42) using the values of the Ni obtained as described above. Solution of Eq.
(59) thus requires a single numerical eigenvalue determination for a comparatively small matrix (e.g., a 10 x 10 matrix for = 3). Sd' SIn and S.x are
then found from Eqs. (43-45). The next step is the evaluation of the N7.
These are obtained by writing Eq. (48) in matrix form, as in Eq. (58),
carrying out the eigenvalue evaluation, and then determining Ni from the
equivalent of Eq. (59). The initial conditions N7(t', t" = O) are found from
Eq. (49) using the results obtained in N;(t') computation. Specifically, we
expand Eq. (59) to give
N;(t')=
Ij a;j exp(Ajt')
where
Im Le;m(pNm-1
a;j = Re;j
+ mfNm)
Re;j and Le;m'being tbe ith components of the jth right and mth left
eigenvectors of n' (corresponding to Aj), and Ni being the above-mentioned
steady-state solutions of the Smith-Ewart equations. Writing Eq. (48) in
matrix notation, we have
oN"jot" = n"N"
whose solution is written using the eigenvalue expansion as
N;'(t', t") =
Ij a;j exp(Ait")
(60)
the Aj being the eigenvalues of n". The a;j depend on t', and one finds that
Eq. (60) may be rewritten as
N;'(t', t") =
Lj LI bijl exp().;t'
+ Ait")
130
where
bjl
= Rej L
Le;;'lpa.m-l
m
+ (m - l)fam]
L LL {bjl[exp(tAjt') -
;>2
exp(l.t')]/[Aj- 2A]}
Some typical resu"lts found using the above formulas are shown in Fig.
11, where the MDS from a combination-dominated high system are
shown (=1.72, obtained with p=0.5sec-1,
cc=O.lsec-l,
k=Cd=
f = O) and, for comparison, a combination-dominated low system
( = 0.51,using p = 0.5 sec-1, Cc= 50 sec-l, k = Cd= f = O).Note that the
MWDs have been normalized, and for cIarity of graphical presentation that
for = 0.51 has been re-scaled by a factor of t. The pronounced difference
between Sbc for
= 0.51
illustration
of the effects of
ratio P for:
= 0.51
= 1.72.
This
0.04
0.51 )
0.03
'"
I
(=1.72)
~ 0.02
<11
'"
o
~
~
0.01
o
O
5
GROWTHTIME (sec)
10
Fig. 11.
Comparison of instantaneous MWD for combination-dominated systems
(where S = Sbc) with = 0.51 and 1.72, assuming p = 0.5 sec-I, k = Cd= f = O, with
Cc= 50 sec -1 and 0.1 sec -1, respectively.
131
= 0.51.
= 1.72
compared
= 2.
. that
system is indeed
a "minibulk"
= 1 sec-1.
presented by Katz et al. (1969), whose work in this area will be reviewed in
Section n,E,2. We show P computed both from the zero-one-two analytic
132
"c:
~
v 1.5
'"'-
---
----
"~
~
v
1.0
0.5
1.0
1.5
2.0
Fig. 12.
Plot of polydispersity ratio P = (Mw>/(Mn>, as function of for system where
termination is by combination alone. + zero-one-two system; - complete analysis.
p = l sec - 1, k = 1 = Cd = O,CC varied.
4.0
3.0
2.0
1.0
0.5
--------------
1.0
1.5
2.0
Fig. 13.
Plot of polydispersity ratio, P = (Mw>/(Mn>, versus for system where
termination is by disproportionation alone. + zero-one-two system; - complete analysis.
p = I sec-I, k =1= '.c = O,Cdvaried.
133
A
;;;
3.0
~
v
""A
~
~V
-------
2.0'~-------
1.0
0.5
i'i
1.0
1.5
134
(61)
The first comprehensive theory of the MWD in an emulsion polymerization system was given by Katz et al. (1969). These authors considered a
monodisperse system in which the only mechanisms operating were freeradical entry, chain propagation, and bimolecular termination by combination. They defined a distribution function Pi(te,MI, M2, ..., Mi), as the
135
fJPifJt.= -a
j<'
+ 2c
fJP,lfJMj
- [p + i(i -1)c]P, + pP'-1t5M;.o
I I
'<j~'+2
ff
(62)
Here a is defined in Eq. (30). The Kronecker t5 term gives the source terms
for each P,. These coupled partial integrodifferential equations in the
variables t, MI, oo.,Mn were derived by population balance applied to each
microscopic process. Since bimolecular termination by combination is the
only chain-stopping mechanism considered, Sbc is the only contribution to
the instantaneous MWD; this is given by
f ~j~
Sbc(t., M) = 2ccP~
x t5(Mj + M,
(63)
M) dMI, oo.,dM,
F(M)
f: Sbc(M)dM
(64)
The development of Katz et al. properly encompasses all compartmeIltalization effects in a system where combination is the only chain-stopping
mechanism. It has been shown (Lichti et al., 1980) that the equations in Eq.
(62) are equivalent to the evolution equations for N; and N;' given in
Section n,e,
that k
= f = Cd = O. The
Mrla, Mrla) =
fI L
i~ 1
(65)
l<j~i
M.. Mj+l'
dMj+l'
oo.,M'-I,
oo.,dM'-1
M.. MI+I,
dM,+I'
.oo,dM,
i ~ 2
(66)
136
Plots of the type shown in Fig. 12 (P versus ) were first given by Katz et al.
Their treatment, however, suffers from a number of limitations. These are (i)
the numerical labor involved in generating the moment expansion, (ii) the
imperfect resolution involved in generating the cumulative distribution [Eq.
(64)] rather than the fuIl distribution, and (iii) the neglect of exit, transfer,
and bimolecular combination by disproportionation. Point (ii) may be
iIIustrated as foIlows. Consider a combination-dominated system in two
limits: the fuIly compartmentalized case of = 0.5, when one has
Sbcex exp(-ca'), and the bulk case, where = 00, when one has
Sbc ex t exp( - at'), where a is a parameter proportional to cc' Figure 15
e
~
~
z
o
::
=>
m
ir
1(J)
e
w
>
~
=>
=>
(,)
GROWTH TIME
Fig. 15.
Plot
combination-dominated
curve 2 = oo.
of instantaneous
MWD
system, as functions
(S) and
cumulative
distribution
(F) for
units). Curve 1
= 0.5;
137
shows plots of Sbcand F [from Eq. (64)] for these two limits, with a = 2. It
can be seen that plots of the MWD, Sbc' show a pronounced qualitative
ditTerence between the compartmentalized and bulk limits. However, the
equivalent plots of F, the cumulative distribution of the tail, are qualitatively similar, and only show smalI quantitative ditTerences.This is because
the greatest ditTerencesbetween the Sbcfor the bulk and compartmentalized
limits occurs at smalI values of M (or t'), to which F(M) is insensitive.
3.
Min and Ray (1974) have presented a theory in which the MWD and
PSD evolutions are considered together. The proposition that these quantities may be evaluated separately is considered in Section III of ~his
chapter.
The treatment of Min and Ray is applicable to a system which is
mohodisperse in volume if one omits both particle nucleation terms and
terms involving O/OV,and if the volume dependences of any rate coefficients
are ignored. For such a system, Min and Ray define a function fm(i, te), the
distribution of latex particles in state i at experimental time te that contain a
growing radical with m monomer units. This is directIy related to the singly
distinguished distribution function Ni(te, t') by
Ni(te
(67)
a convolution expression involving products of the fm and indeed acknowledge that this assumption is open to question. It was shown in Section
II,C,2 by use of a particular example that such a product approximation is
138
invalid. We now give a more specific illustration. Consider the correct form
for Sbcin a zero-one-two system:
it'
Sbe(te' t' ) =
"
N"2e'
(t t' - 2t" , t" ) dt
(68)
tO
pio
NI 2e'
(t
tI
t") dt"
~~
JSbcdt' = J
SbC
dt'.
1 sec-1, Cc = 10 sec-1, k = f = Cd = O
(giving = 0.55). It is clear that the product assumption is both quantitativeiy and qualitatively inaccurate. Indeed, the limiting
,
.value of the
NI
o
~
~
o
31
~
is instantaneous
MWD
139
than the
140
141
for obtaining the various rate coefficients klP p, k, kp, and e from kinetic
studies (e.g., Hawkett et al., 1980, 1981). The theoretical evaluation of the
MWD is then trivial. For systems with a significant polydispersity in
volume, the MWD can readily be calculated using Eq. (27), given the
volume dependence of the various rate coefficients; however, it was pointed
out above that uncertainties in these quantities make an unambiguous
relation between theory and experiment impossible in such an experiment.
Experiments on a volume-monodisperse Interval III system could again be
readily compared with theory, using known values of the time variation of
CM, etc. (obtained from the kinetics), and integrating the instantaneous
MWD over time.
The ideal seeded experiment would be where seed Iatex particles are
prepared in which the MWD of the seed is significantly different from that
formed during subsequent seeded polymerization. The instantaneous MWD
formed at early times could then be estimated directly. Seed latexes of
different sizes could be used to probe the effect of compartmentalization (we
stress again that if the system is transfer-dominated, the MWD of formed
polymer should be independent of particle size).
The experimental strategy outlined above would give a unified set of rate
parameters to describe the kinetics, MWD, and PSD. If no such consistent
set could be found, new mechanisms may pro ve to be important. For
example, aqueous-phase termination, which is not usually incorporated in
MWD theories, would give rise to extensive low molecular weight fragments, which would significantly influence the MWD if these fragments are
subsequently incorporated into the latex p~rticles.
142
scales is plainly invalid at very early reaction times, when most of the
volume of the partic1e may be that of a single chain. However, since latex
partic1es usually contain many individual chains, the time scale separability
assumption for volume versus chain growth [i.e., (i) above] is c1early
relevant to mO$tof the growth process.
We now consider the second time scale separability, (ii) above. Again,
this will obviously break down at early times, but again will be valid for
most of the polymeriz.ation. This occurs for three reasons.
1. As partic1es increase in volume, a fixed percentage increment of
growth takes longer to achieve.
2. With increasing time, the system will attain and maintain its steady
state in the distribution of free radical s in all partic1es, with the possible
exception of growth during a particularly rapid Trommsdorff gel effect.
3. The PSD becomes smoother with the passage of time (e.g., Fig. 1),so
that partic1e growth [which disturbs steady state through the o/ay term in
Eq. (8)] becomes less significant.
.
Since the separate time scale aondition is c1early valid for most of the
polymerization process, one may say that each polymer chain is formed
inside a partic1e of unchanging size, wherein all rate coefficients are constant
and the distribution of free radicals has its steady-state value, for each
volume V. Any residual effect of the PSD on the MWD would reside
presumably in the effects of the PSD on the kinetic parameters (e.g., p, e,
and to a lesser extent k). Conversely, the MWD would possibly influence
the PSD through its effects on the swelling of the partic1es by the monomer;
the effect, if it exists, is likely to be small.
IV. Conclusions
The theoretical tools required to describe the evolution of the MWD and
PSD of an emulsion polymerization, in terms of a consistent set of rate
coefficients for the various microscopic processes involved, are fairly well
developed. Whereas chain branching (transfer to polymer) has yet to be
inc1uded, it is likely that bulk theories for this process (Bamford and
Tompa, 1954) can be readily modified to the emulsion polymer case.
Although the general theory is well developed, a real understanding of
the details of the processes involved is limited by the paucity of unambiguous experimental results to enable theory to be tested, modified, and
applied.
143
The support of the Australian Research Grants Committee, and of an Australian Institute
for Nuclear Science and Engineering postdoctoral fellowship for GL, are gratefully
acknowledged.
References
Ballard, M. J., Gilbert, R. G., and Napper, D. H. (1981). J. Polymer Sci. PoI. Phys. Edn.,
(in press).
.
Bamford,C. H., and Tompa, H. (1954).Trans. Faraday Soco SO, 1097-1115.
Camahan, B., Luther, H. A., and Wilkes, J. O. (1969). "Applied Numerical Methods." Wiley,
New York.
Corio, l., Mara, L., and Salvatore, O. (1979). Makromol. Chem. 180,2251-2252.
De Graaf, A. W., and Poehlein, G. W. (1971). J. Polym. Sci. Parl A-2 9, 1955-1976.
Eckhoff,R. K. (1967).J. Appl. Polym. Sci. 11, 1855-1861.
Ewart, R. H., and Carr, C. 1. (1954). J. Phys. Chem.58, 640-644.
Flory, P. J. (1953). "Principies ofPolymer Chemistry." Comell Univ. Press, 1thaca, New York.
Friis, N., and Hamielec,A. E. (1975).J. Appl. Polym. Sci. 19,97-113.
Friis, N., Goosney, D., Wright, J. D., and Hamielec, A. E. (1974). J. Appl. Polym. Sci. 18,
. 1247-1259.
Gardon, J. L. (l968a). J. Polym. Sci. Parl A-I 6, 665-685.
Gardon, J. L. (1968b). J. Polym. Sci. Parl A-I 6, 687-710.
Gerrens, H. (1959).Forlschr. Hochpolym.Forsch. 1, 234-328.
George, M. H. (1967). /n "Vinyl Polymerizations" (G. E. Ham, ed.), Vol. 1, part 1, p. 165.
Dekker, New York.
Grancio, M. R., and Williams, D. J. (1970). J. Polym. Sci. Parl A-I 8,2733-2745.
Harkins, W. D. (1945).J. Chem. Phys. 13, 381-382.
Harkins, W. D. (1946). J. Chem. Phys. 14,47-48. ,
Harkins, W. D. (1947). J. Am.Chem.Soco69, 1428-1444.
Harkins, W. D. (1950). J. Polym. Sci. 5, 217-251.
Hawkett, B. S., Gilbert, R. G., and Napper, D. H. (1980).J. Chem.SocoFaraday Trans. /76,
1323-1343.
.
Hawkett, B. S., Napper, D. H., and Gilbert, R. G. (1981). J. Chem. Soco Faraday Trans. /
(in press).
James, H. L., and Piirma, l. (1976)./n
eds.), pp. 197-209. American
"Emulsion
Polymerization"
Chemical
Lansdowne, S. W., Gilbert, R. G., Napper, D. H., and Sangster, D. F. (1980). J. Chem. Soco
Faraday Trans./76,
1344-1355.
Lichti, G., Gilbert,
1957-1971.
D. H. (\980).
B. S., Gilbert,
R. G., Napper,
D. F. (1981). J. Polym.
144
Lin, C. c., and Chiu, W. y. (1979). J. Appl. Polym. Sci. 23, 2049-2063.
Min, K. W., and Ray, W. H. (1974). J. Macromol. Sci. Rev. Macromol. Chem. C 11, 177-255.
Min, K. W., and Ray, W. H. (1978). J. Appl. Polym. Sci. 22, 89-112.
Morton, M., Kaizerman, S., and Altier, M. (1954). J. Colloid Sci. 9, 300-312.
Nagy, D. J., Silebi, C. A., and McHugh, A. J. (1980). In "Polymer Colloids II" (R. M. Fitch, ed.),
pp. 121-137. Plenum Press, New York.
O'Toole, J. T. (1969). J. Polym. Sci. Part C 27, 171-182.
Piirma, J. Kamath, V. R., and Morton, M. (1975). J. Polym. Sci. Polym. Ed. 13,2087-2102.
Pis'men, L. M., and Kuclianov, S. I. (1971). Vysokomol. Soedin. A13, 1055-1065.
Schmidt, E., and Biddison, P. H. (1960). Rubber Age 88, 484-490.
Schmidt, E., and Kelsey, R. H. (1951). Ind. Eng. Chem. 43, 406-413.
Shaw, D. J. (1970). "Introduction to Colloid and Surface Chemistry." Butterworths, London.
Singh, S., and Hamielec, A. E. (1978). J. Appl. Polym. Sci. 22, 577-584.
Smith, W. V., and Ewart, R. H. (1948). J. Chem. Phys. 16,592-599.
Stevens, J. D., and Funderburk, J. O. (1972). Ind. Eng. Chem. Process Res. Dev. 11, 360-369.
Sundberg, D. C., and Eliassen, J. D. (1971). In "Polymer Colloids" (R. M. Fitch, ed.), pp.
153-161. Plenum Press, New York.
Thompson, R. W., and Stevens, J. D. (1977). Chem. Eng. Sci. 32: 311-322:
Ugelstad, J., and Hansen, F. J. (1976). Rubber Chem. Tech. 49, 53Cr609.
Watterson, J. G., and Parts, A. G. (1971). Die Makromol Chem. 146, 11-20.
Wood, D., Lichti, G., Napper, D. H., ~nd Gilbert, R. G. (1981), to be published.
4
Theory 01 Kinetics
01 Compartmentalized
Free-Radical Polymerization
Reactions
D. C. Blackley
1. Introduction.
A. Definitionsand Introductory Concepts
B. Practical Significance of the Theory
C. Scope of Chapter .
11. Reaction Model Assumed
A. Description.
B. Constancy and Uniformity of Monomer
111.
IV.
V.
VI.
VII.
VIII.
146
146
147
149
149
149
151
153
156
156
160
164
167
167
172
176
177
183
185
187
189
145
EMULSJON POL YMERIZA TION
Copyright ~ 1982 by Academic Press, Inc.
AII rights 01 reproduction in any lorm reserved.
ISBN 0-12-556420-1
146
D. C. Blackley
l. Introduction
A.
4.
Kinetics of Compartmentalized
147
B.
Apart from intrinsic interest, the theory of compartmentalized freeradical polymerization reactions is of importance primarily because it is
believed that ITIost of the polymer which is formed in the course of an
emulsion polymerization reaction is formed via reactions of this type. The
general shape of the conversion-time curve for many emulsion polymerization reactions suggests (see Fig. 1) that the reaction occurs in three more-orless distinct stages or "intervals." The first of these, the so-called Interval 1,
is interpreted as the stage of polymerization in which the discrete reaction
loci are formed. In the second and third stages - Intervals 11 and 111- the
polymerization is believed to occur essentially by compartmentalized freeradical polymerization within the loci which were formed during Interval l.
148
D. C. Blackley
c.
o
.~
Q)
>
e
o
<>
INTERVAL m
time
Fig.1.
Conversion-time
polymerization
reaction.
The feature that distinguishes IntervallI from Interval 111is that monomer
droplets are present as a separate phase during the former only.
The theory also has relevance to the so-called "seeded "emulsion polymerization reactions. In these reactions, polymerization is initiated in the
presence of a "seed" latex under conditions such that new particles are
unlikely to formo The loci for the compartmentalized free-radical polymerization that occurs are therefore provided principally by the particles of the
initial seed latex. Such reactions are of interest for the preparation of latices
whose particles have, for instance, a "core-shell" structure. They are also of
great interest for investigating the fundamentals of compartmentalized freeradical polymerization processes. In this latter connection it is important to
note that, in principIe, measurements of conversion as a function of time
during nonsteady-state polymerizations in seeded systems offer the possibility of access to certain fundamental properties of reaction systems not
otherwise available. As in the case of free-radical polymerization reactions
that occur in homogeneous media, investigation of the reaction during the
nonsteady state can provide information of a fundamental nature not
available through measurements made on the same reaction system in the
steady state.
Scope 01 Chapter
Description
150
D. C. Blackley
4.
nevertheless remains a general feature of the theory developed for nonsteady state reactions.
A further general assumption is that the physicochemical parameters
that characterize the rates at which the various processes occur remain
constant throughout the reaction. However, as will appe~r subsequently,
there is one particular instance where it has been possible to remove this
restriction. Important consequences of this restriction are that the polymerj
monomer ratio in the reaction loci should not vary significantly as the
reaction proceeds, and that the volume of the reaction loci should not
change greatIy as a consequence of polymerization. The latter requirement
will be most obviously realized in a seed emulsion polYp1erization in which
the size of the initial seed is large and the extent of polymerization is small.
B.
There has been much discussion in recent years concerning the question
of whether or not, in a reaction system of the type considered in this
chapter, the concentration of monomer in the reaction loci is constant as
the reaction proceeds and uniform across the reaction locus. Morton et al.
(1954) made the reasonable proposal that the equilibrium swelling of
polymer partic1es by excess monomer in a system such as is envisaged here
is determined by the balance of two factors: (i) the decrease of free energy
that accompanies mixing of polymer and monomer and (ii) the increase of
interfacial free energy that accompanies the concomitant swelling of the
polymer partic1es. Theoretical considerations predict that at equilibrium
large partic1es should imbibe more monomer than small partic1es, and that
reduction of the interfacial free energy between polymer partic1e and
external phase shoulcl lead to increased swelling. Polymer molecular weight
is not a significant factor in determining equilibrium imbibition, but the
polymer-monomer solution interaction parameter is. Morton et al. (1954)
and Allen (1958) have provided experimental evidence to support these
predictions, the polymers used being polystyrene and natural rubber, respectively.Further information has been provided by Vanzo et al. (1965)
concerning the solubilities of vinyl acetate, vinyl hexanoate, styrene, and
benzene in poly(vinyl ,acetate), poly(vinyl hexanoate), and polystyrene
latices. Morton et al. (1954) also provided evidencein support ofthe view
that, as long as excess monomer is present as a separate droplet phase, the
monomer-swollen polymer partic1es remain essentially in the same state of
equilibrium imbibition during polymerization as in the absence of polymerization. This implies that the process of diffusion from droplets to polymerization loci occurs rapidly relative to the rate of polymerization, so that the
152
D. C. Blackley
consumption of monomer by polymerization has little effect on the concentration of monomer at the polymerization loci. A further implication is
that the concentration of monomer is uniform throughout the polymerization locus.
The views summarized in the preceding paragraph were generally accepted without serious challenge for reaction systems for which monomer and
polymer are miscible, until Grancio and Williams (1970) proposed that the
monomer-swollen polymer particles in a styrene emulsion polymerization
are nonuniform, and that in fact they consist of an expanding polymer-rich
"core" surrounded by a monomer-rich "shell," the latter being the major
locus of polymerization. The "core-shell" model was suggested by kinetic
evidence and supported by evidence obtained by examining ultra-thin
sections of latex particles by electro n microscopy. These ideas have been
further elaborated by Williams (1971), by Keusch and Williams (1973), and
by Keusch et al. (1973).Friis and Hamielec(1973)have challengedGrancio
and William's interpretation of their kinetic data and have suggested that
the observations can be explained by means of a homogeneous particle
model in which allowance is made for decrease in the rate coefficient for
termination as polymerization proceeds. Napper (1971) has also criticized
the core-shell morphology proposed by Grancio and Williams, and has
argued in favor of a core-shell morphology of the reverse kind, Le., a
particle that comprises a monomer-rich core surrounded by a polymer-rich
shell. Williams (1973) has challenged this view. Keusch et al. (1974) have
discussed the nature of the distribution of polymer segments in a swollen
latex particle and find that the concentration of segments should be high in
the central region of the particle and fall to zero at the particle surface, thus
supporting a core-shell morphology of the Grancio-Williams type. The
evidence for and against concentration gradients in polymerizing monomerpolymer particles in a reaction system of the type envisaged here has been
reviewed by Gardon (1973). Theconclusion
reached is that such concentration gradients as may form within such a particle are unlikely to
be large enough to affect the rate of polymerization significantly. A further
review of this subject has been published recently by Vanderhoff (1976).
Blackley and Haynes (1977) have argued that the marked reduction in the
rate of the emulsion polymerization of styrene that occurs when certain
organic diluents (e.g., ethylbenzene) are added is consistent with the view
that the equilibrium concentration of monomer in the polymerization loci is
less when polymerization occurs than when it does noto This interpretation
has recently been challenged by Azad and Fitch (1980).
The conclusions to which the present evidence seems to lead can be
summarized as follows:
1. The monomer concentration is probably essentially uniform across
the polyermization locus if the monomer is a good solvent for the polymer.
C.
154
D. C. Blackley
Little needs to be said here except to note that (i) the rate of propagation
is unlikely to be appreciably dependent on the size of the reaction locus,
whereas the rate of termination is likely to be appreciably size-dependent
and (ii) the rate of both propagation and termination will be reduced if the
viscosity of the reaction medium rises, but the rate f termination will be
reduced more than that of propagation.
The physical reason why the rate of propagation is not appreciably size
dependent is simply that a propagating radical is always surrounded by the
same concentration of monomer molecules wherever the radical is in the
reaction locus and, to a first approximation, whatever the size of the locus.
However, to the extent that the monomer/polymer ratio in the reaction
locus depends on the size of the locus (see Section II,B), then some
dependence of rate of.pro paga tion on locus size will be observed.
Since most of the monomer in a compartmentalized free-radical polymerization reaction is consumed in the propagation reaction, it is customary
to write the overall rate of polymerization as
dM/dt
= 1Nkp[M]
(1)
156
D. C. Blackley
A. Derivation
The fundamental equations that govern the behavior of a compartmentalized free-radical polyrnerization reaction in which the radicals are
generated excIusively in the external phase are most readily derived by considering the rates of the various processes by which loci containing exactIy
i propagating radicals are formed and destroyed. These processesare illustrated in Fig. 2 as transitions between various "states of radical occupancy"
of the loci, each state of occupancy being defined as the number of
0+2)0 +l)ni+2X
(j + Oni+1 k
i+2
i+1
ink
i(j -I)nx
njO'
H
ni-lO'
j-2
F.ig. 2. Transitions betweenstatesof radical occupancyrequired for derivation of timedependentSmith-Ewart differential differenceequation for state i.
.
4.
1 loci by acquisition
loci by loss of a radical, and (iii) from type i + 2 loci by mutual termination
within the loci. Likewise there are three processes by which loci of type i
can be destroyed: (i) by acquisition of a further radical, thereby becoming a
locus of type i + 1; (ii) by loss of one radical by exit, thereby becoming a
locus of type i - 1; and (iii) by loss of two radicals by mutual termination,
thereby becoming a locus of type i - 2. The differencebetween the sum of
the rates of the first group of processes and the sum of the rates of the
second group of processes gives dn;/dt, the overall rate at which loci of type
i are formed. The result for dn/dt is readily found to be
dl1;/dt
= (11-1 -
11)O'+
{(i +
l)n+1
il1}k
(2)
158
D. C. Blackley
1+2
n..,.
I
---1---1_--1--
i+1
- -------------
i-I
1-2
.
barrier has been drawn between the states i and i + 1. The O'Toole
approach is simpler than the earlier approach of Smith and Ewart (1948)
(which considers the balance of transitions from a particular state of radical
occupancy, as illustrated in Fig. 2) to the extent that only four transitions
locus populations ni, Inspection of Fig. 3 shows that the overall rate of
= O. If
state, then Eq. (3) gives merely the overall rate of transition of loci upward
across the barrier. To obtain an expression for dn;/dt, it is necessary to
observe that this quantity is the rate of accumulation of loci in state i, and
that this in turn is equal to the difference between the rate of transition
upward across the barrier between states i - 1 and i and the rate of
transition upward across the barrier between states i and i + 1 (see Fig. 4).
Thus, we have
(4)
Equation (2) then follows immediately by noting that
R-l-+i
= n-l(1-
(5)
---
- - - --- - - - - -1-1
1-2
la)
lb)
Fig. 4. Transitions between states of radical occupancy required for derivation of dnJdt
by the O'Toole approach. (a) shows transitions across barrier betwe~n states i and i + 1; (b)
shows transitions between states i - l and i.
160
~o Co Blackley
lo
or [volumeJ[timer 1, according to
dn/dt
(6)
The infinite set of equations can therefore be written as the single matrix
equation
d[n]/dt = [nJ[n]
(7)
where [n]T is the matrix [nonl n2 . . .J and [nJ is a matrix of coefficientso
The structure of [nJ is such that the jth row contains the following elements
in the U - l)th,jth, U + l)th, and U + 2) columns respectively, and all the
other elements of the jth row are zeros:
a
(8) .
To the extent that the elements of the matrix [nJ are dependent neither on
time nor on the nnor on each other, it is possible immediately to write down
in matrix form the solution ofthe differential matrix Eq. (7). It may, however,
be preferable first to transform the differential matrix equation, eogo,by
using a suitable similarity transformationo Having solved the transformed
equation, the n can then be obtained from the matrix of transformed ni by
applying the inverse transformation.
It should be noted that, beca use the set of equations contained in Eq. (2)
is theoretically infinite in number, the matrices [n] and [nJ of Eq. (7) are
infinite with respect to the number of elements they contain. What is done
2. Use of a Locus-Population
Generating Function
'P(~, t) =
L v(t)~
=o
(9)
where ~ is an auxiliary variable. This function is described as a "locuspopulation generating function," because once it is known, all the fractional
locus populations at all times can in principIe be obtained from it by
making use of the fact that
1 0'P
(10)
v(t) ="1
::I}'i
lo
( )
u,>
~=o
0'1'
00
Y(t) = 1=1
.L iy(t)
()
~u,>
(11)
~=l
and the variance of the distribution of locus populations at any instant can
be found from
00
V(t)
L (i -
=o
~
{ o~
Y)2y(t)
~0'P
= L i2V(t) -
o~ )} ~= 1
0'1'
/2
(o~ )~=
(12)
1
162
D. C. Blackley
'P(I, t) = L: v(t)= 1
=o
(13)
for all values of t. Thus, the time dependency of 'P(~, t) should always
disappear when the substitution ~= 1 is made-this is a consequence of the
assumption that the number of reaction loci in the system remains constant.
The function 'P(~, t) has another interesting property which as yet hasnever
been exploited as fai:as is known. Ir the substitution ~ = - 1 is made; then
we obtain
'P(-I, t) = vo(t) - v(t) + V2(t)- v3(t) + ...
(14)
Thus, 'P(-I, t) gives at any time the difference between the fraction of the
total number of loci that contain an even number of propagating radicals
and those that contain an odd number, the fraction containing an even
number inc1uding within it those containing no propagating radicals at all.
Ir it is desired to know the fraction of the total number of loci that do not
contain any propagating radicals, then this can be obtained from the relationship
vo(t) = 'P(O,t)
-(15)
Thus, it is c1ear that the single function 'P(~, t) is capable of yielding a great
deal of information concerning the distribution of locus populations in the
reaction system.
There are at least four important advantages to using locus-population
generating functions as compared with the matrix approach:
1. It is quite unnecessary to truncate the set of equations contained in
Eq. (2); in fact, the logic of the method requires that the set be maintained
infinite in extent. There is therefore no possibility that the set of equations
actually being solved is a distorted representation of the behavior of the
reaction system being investigated. To the extent that the requisite mathematical operations ha ve been correctly performed, the resultant expressions for v(t),etc. are free from approximation errors.
2. It is possible in several cases to obtain expressions for v(t), etc. in
c10sed analytic form rather than as somewhat unwieldy algebraic
expressions.
3. The method using the locus-population generating function is more
readily applicable to systems in the steady state.
4. It is in principIe possible to make inferences concerning the nature of
the distribution of locus populations if the generating function can be
recognized as that of the frequencies of some known distribution.
It should perhaps be pointed out that in several previous publications
(e.g., Stockmayer, 1957; Birtwistle and Blackley, 1977, 1978: Birtwistle et al.,
(16)
where in each case the summations cover all possible values of i. The
resultant single differential equation then readily transforms to
.8'P
-
8t
t
= O\~
- 1)'P + k(l - ~)-8'P + X(l - ~2)-82'P
8~
8~2
(17)
This is the partial differential equation for 'P(~, t) which has to be solved for
each particular case. The solution has to be subject to the initial (boundary)
conditions appropriate to that particular case.
For the special case of reaction systems that are in a steady state, n(t)
are invariant with time, and therefore 'P(~, t) is not a function of t. It is therefore denoted by 'P(~) in this case. Thus, in this case 8'P/8t = O,and the righthand side of Eq. (17) must be zero. What were hitherto partial differential
coefficients of 'P(~, t) with respect to ~ become ordinary differential coefticients. Dividing through by the factor 1 - ~ then gives
d'P
d2'P
a'P - k- X(l + J:) -
d~
..
de
=O
(18)
164
D. C. Blackley
as the ordinary differential equation to be solved for 'P in this case. This is
identical to the differential equation first given by Stockmayer (1957) for the
locus-population generating function that characterizes the behavior of
reaction systems in the steady state. Equation (18) can also be obtained
from the infinite set of equations for the steady-state as given by O'Toole
(1965). It will be recalled that this is the set of which Eq. (3) with the lefthand side set equal to zero is the typical member. Again the procedure is to
multiply each eq\lation by ~i/N, add together all the equations so obtained,
and then note the relationships summarized in Eq. (16) together with the
additional relationship
.
2'P
(i + 1)ivi+l~1 = ~ ~2
(19)
Approximation Methods
m = kv/kt = k/X
(20)
(22)
Two new variables x and y are introduced. These are defined by the
equations
x
y = 'P(~)/xl-m
= 2)e(1 + ~),
(23)
{(1 - , m)2
X2}
=O
(24)
where Il-m(x) and Im-l(X) are modified Bessel functions of orders 1 - m and
m - 1 respectively, and A and B are constants. Physical considerations (see
O'Toole, 1965) indicate that the A must always be zero. The value of B is
determined by the requirement that 'P(1) = 1. The final result for 'P(~) is
(\/+fl+1)
2(m-1j/2
'P(~) = Im-l(h) (1 +
~)(1-m)/2Im-l
(26)
hi2m-1-3i
and
Vi
= i! Im-l(h)
Im-l+i
(J2)
(27)
(28)
166
D. C. Blackley
p = p' + k
i= 1
ini - 2ktE e2
(29)
where p is the total rate of entry of radicals into all the N loci in unit
volume of reaction system, p' is the rate of generation of new radicals within
unit volume of reaction system, and e is the stationary concentration
of radicals within the external phase. The form of this equation implies that
all the radical s that become present in the external phase (whether by
generation or by exit from the loci) are potentially available for reentry,
except insofar as they are destroyed by mutual termination. Dividing
through by N, setting a = p/N, a' = p'/N, and y = pie, and introducing
parameters
e = av/ktP'
e' = a'v/ktp,
m = kv/ktP,
Z = 2NktpktE/yZv
(30)
(31)
Ugelstad et al. have used Eq. (31) in conjunction with Eq. (28) to calculate y
as a function of e' for various values of m and Z. It is valid to regard the e of
4.
Kinetics of Compartmentalized
167
Eq. (31) as equivalent to the h of Eq. (28) because the reaction system is
presumed to have reached a steady state in which, inter alia, ,/,and 1 have
attained stationary values.
..
o\f/ot
= c(~-
(32)
= ... = O
(33)
;=0
v(t)
vo(O)
(34)
168
D. C. Blackley
T{t) is a function of t only. The details of the process whereby the solution
can be obtained have been given by Birtwistle and Blackley (1977). The
result for 'P(~, t) is
'P(~, t) = exp{(o-k)(~ - 1)(1 - e-k/)}
(35)
It follows immediately from this that at all times the distribution of locus
populations with respect to radical occupancy is Poissonian, and that the
parameter of the distribution at any instant is (O"/k)(1- e-k/). These conclusions follow because the expression for 'P(~, t) is recognizable as the
frequency-generating function for a Poisson distribution (see Kendal and
Stuart, 1965). It then follows that
- e-k/)
} {
eXP
-~(1
- e-k/)
(36)
and
(37)
Alternatively, the results embodied in Eqs. (36) and (37) can be derived from
Eq. (35)using Eqs. (10)and (11).The general result for v(t) then shows that
the distribution of locus populations must always be Poissonian with timedependent
parameter
= 00 in
following predictions for the reaction system when it has attained the steady
state:
'P(~,oo)
= e(O'/k)(~-l)
1 O" i
v(oo)=i! k e-O'/k
()
Y(oo) = O"/k
(38)
V2(t)
;,;-0.5
i=2
10
15
20
Fig. 5.
reaction system for whieh radiealloss from reaction loei is exclusively by first-order processes.
Values taken ror (] and k are 1 x lO-s sec- and 5 x 10-4 see- respeetively. The ordinates
for i = Oare vo(t); those for i = 1 are 40 v,(t); those for j = 2 are 2 x 103 v2(t).(Reproduced with
permission of J. Chem. SocoFaraday l.)
..
Q
,x
10
15
20
Fig. 6.
Average number of radieals per loeus I(t) as a funetion of time t for reaetion
system to which Fig. 5 refers. (Reprodueed with permission of J. Chem. SocoFaraday l.)
170
D. C. Blackley
10
10
15
20
steady state being reached after about 104 seco The average number of
radical s per locus in the steady state is small (0.02) for this case.
The amount of monomer M(t) converted in a unit volume of the reaction
system after the eIapse of time t can be obtained by substituting for T(t) in
Eq. (1) from Eq. (37), and then integrating
= O to t = t.
The
result is
(39)
The prediction for M(t) as a function of t for the reaction system to which
Figs. 5 and 6 refer is shown in Fig. 7. Marked deviation from linearity in
the predicted curve for conversion as a function of time occurs over the
range t < 5 X 103 seco
An interesting generalization to the case X = O has been given by
Birtwistle and Blackley (1979). In this generalization, the theory is extended
to include cases where a and k are time dependent. In order to emphasize
the time dependence of a and k they are written as a(t) and k(t). What
Birtwistle and Blackley (1979) have shown is that the solution to Eq. (32)
always has the form
'I'(~, t) = e<~-1)8(t)
(40)
(42)
and
-8(/)
v(t) = {O(tW
., e
lo
(43)
If a and k are not time dependent, then O(t)has the value (a/k)(1 - e-kl)
and the results are identical to those obtained previously. The foIlowing
particular cases of time-dependent a and k have been considered in detail
by BirtwistIe and BIackley (1979):
1. The case where a decays exponentiaIly with time, and k is constant.
2. The case where a decays linearIy with time, and k is constant.
3. The case where a decays as a consequence of second-order depletion
of initiator, and k is constant.
4. The case where a is constant, but k varies because the loci grow at a
constant rateo
5. The case where a varies because radicals lost to the external phase
are available for reinitiation, and k is constant.
Of these cases, the first is probably of greatest interest because the
decomposition of moSt dissociative initiators is kineticaIly of first order with
respect to initiator concentration, and therefore both the initiator concentration and the rate of decomposition of the initiator wilI decay exponentiaIly with time. Setting a(t) = aoe-al, the solution for O(t)is
(44)
except for the unlikely special case where o: = k, in whichcase the result for
O(t)is
O(t)= aote-kl
(45)
D. C. Blackley
172
2
a/k
=O
a/k =0.01
..
Q
X
,11
q;
15
20
Fig. 8. Loeus-populationdistribution parameter O(t)and the average number of radieals per loeus T(t) as funetions of time t for reaetion system for whieh radiealloss from reaetion
loei is exc1usively by first-order proeesses and rate of radical entry deeays exponentially with
time. The values taken for (10and k are 1 x 10- s see- and 5 x 10-4 see-, respeetively; eurves
are given for various values of c(, indieated by the C(/k ratios
(Reprodueed with permission of J. Chem. Soco Faraday l.)
appended
to the eurves.
cause 8(t) to rise to a maximum and then eventually to decay to zero. This
ultima te decay to zero is a consequence of depletion of the source of
radicals. It carries the implication that eventually all v(t) for i > O themselves decay to zero and vo((0) = 1.
B.
4.
Eq. (20)J.Thus, a nonzero value of m implies that radicals can be lost from
reaction loci by first-order processes,whereasa zero value of m impliesthat
either radicals cannot be lost from reaction loci by first-order processes or
k, is truly infinite. The latter possibility seems physically unlikely.
Whatever the value of m, the equation to be solved is Eq. (17) with (1set
equal to zero, Le.,
(46)
subject to an appropriate
boundary
condition.
In any particular
case, the
T(t) = Ae).x
(47)
where A.is the "separation constant" and A is a constant whose value may
depend upon that of A..It is necessary to give careful consideration to the
values of
A. that
+ 1) where b =
0,1,2,
....
(49)
174
D. C. Blackley
(50)
BpJr 1.m-l)(~)e-t/tp
'I'(~, t) =
p=o
+m-
1)]
(51)
p = O, 1,2, ...
the Jpare Jacobi polynomials, and the Bpare coefficients whose values are
determined by the requirement that 'I'(~, t) at zero time must have a
particular form 'I'(~,O). The particular form for 'I'(~,O) in any given instance
is that which characterizes
the distribution
of locus populations
which
prevailed at the instant when the generation of new radicals ceased. The
general results obtained for v(t), T(t), and M(t) are as follows:
1
v(t) =
2.,1.
00
.'Bip
p='
(52)
- l);1~:/,m+-l)(O)e-t/tp
00
T(t) =
t Bp(p + m -
(53)
l)e-t/tp
p=l
and
00
Bp7:p(p+ m p=l
1)(1-
(54)
e-t/tp)
= 2p +
2.
1 12p+m-(h)
p + m- 1
(55)
Im-(h)
Case in which m = O
Setting m
= O in
A.
(1 - ~2) d~-; - xE
(56)
=O
as the differential equation that must now be satisfied by the function E(~).
This equation is a special case of the Gegenbauer differential equation. The
result obtained for 'P(~, t) is
C()
(57)
L KpC~-/2)(~)e-t/tp
'P(~,t) =
p=o
..
Xp(p
(58)
p = O,1, 2, ...
1)
"
C()
.
( 1)(P-/)/2
r[(p + i - 1)/2]
e -t/tp
(59)
where the summation extends over all values of p ~ i such that p - i is even
C()
I(t)
=-
L Kpe-t/tp
(60)
p=!
and
(61)
176
D. C. Blackley
The separate term for p = 1 in Eq. (61) arises because '1 is infinite in this
case, as well as 'o.
There is a very simple physical consideration that underlies the need for
separate treatment in the case m = O,and which explains in particular the
fact that the summation for v(t) given in Eq. (59)does not contain terms
that correspond
to p
(62)
It should be noted that this result does not follow from Eq. (55) by setting
m = O. However, the 'p for the case m = O can be obtained from those for
the case m # O by putting m
C.
= O.
a(~
1)\f!
(63)
Le., in the third term of the right-hand side of Eq. (17), the factor ~ is
replaced by the factor 1 -~. Some mathematical justification for the
4.
approximation can be given if it is noted that (i) the use which is made of
the function 'P(~, t) is such that it is required at the extremities of the range
O::;;~ ::;;1 only, (ii) over this range 1 ~ is a reasonably good approximation for 1 - ~2, and (iii) 1 - ~ and 1 - ~2 have the same values at the
extremities of this range. Equation (63) can be obtained from a modified set
of time-dependent Smith-Ewart differential difference equations in which
the coefficient of kt/v in the final term on the right-hand side of Eq. (2) is
replaced by {(i + 2)(i+ 1)ni+2-(i+ l)ini+}' The modified set ofequations is thus expected to be a reasonable approximation, provided that the
rate of loss of radical s from reaction loci by second-order processes is not so
great as to be in effect the dominant radical-los s mechanism.
The solution of Eq. (63) subject to the boundary condition 'P(~, O)= 1 is
found to be
'P(~, t) = e(~-l)<I>(I)
(64)
'f'(t) = 2 a
tanh(at/2)
(66)
Other Approximations
Hawkett et al. (1975) ha ve used the matrix method to obtain approximate predictions of the nonsteady-state behavior of reaction systems for
178
D. C. Blackley
P + P2)
+ 4(1 +
P+
p2)e-2at -
p2e-(2x+a-a2/20X)t
(67)
(68)
40 Kinetics 01 Compartmentalized
179
100
-=
;;;- 005
i=2-
30
60
lime {mio}
Fig. 9.
Fractional locus populations, v~t), as functions of time t for i = 0,1,2 for
reaction system for which radical loss from reaction loci is exclusiveiy by second-order
processes. Values taken for (f and X are l x 10- 3 sec- and 1 x 103, respectively. The ordinates
for i = O and i = 1 are vo(t) and v(t), respectively; those for i = 2 are 2 x 106V2(t).
(Reproduced
with permission
systems using the matrix method. They agree that the locus-population
generating function method is more dir~ct than the matrix method, but
correctly point out that treatment of the general case by the locuspopulation geneniting function method has so far not been possible. For
the case where all the loci are initially devoid of propagating radical s and
(f ~ (k + X), the general solution is given by Hawkett et al. (1977) as
co
v(t)
= L l~r~eAp'
p=o
(69)
where l~ is the zerotb component of the pth left eigenvector of the matrix
[O], r~ is the ith component of the pth right eigenvector of [O], and Ap is
the pth eigenvalue of [O]. The expression given for M(t) is
(70)
where A is a constant. Thus, the problem of obtaining predictions for the
v(t)and for M(t) reduces to that of finding the eigenvalues and eigenvectors
180
D. C. Blackley
of En]. The method which Hawkett at al. (1977) use is to express the matrix
En] in tHeform
En]
= [no] + 0"[n1]
(71)
where
-1
O
-1
[O,]
O O oo.
O O oo.
1 -1
O ...
(72)
[
and [no] is the matrix whose elements are obtained by subtracting O"times
those of [na from the corresponding elements of En]. Since O"is assumed
to be small compared with k and X, the eigenvalues and eigenvectors of En]
can be obtained by using non-Hermitian perturbation theory and treating
0"[n1] as a perturbation on [no]. Hawkett et al. consider that the restriction O"~ (k + X) is fulfilled by most, though not all, compartmentalized
free-radical polymerization re'actions in which the external phase is
aqueous.
Of the somewhat grosser approximations, consideration will be given
first to the "two-state" model of Lichti et al. (1977). Reaction systems that
conform to this model are those in which very few loci ever contain more
than one radical, Le., the rate of entry of radical s into loci is much less than
the sum of the rates of radical loss by first- and second-order processes.
Thus, the only significant populations that have to be taken into account in
calculating l(t) are those of loci for which i = O and 1. The number of loci
that contain three or more radicals is assumed to be strictly zero. The
number of loci containing two radical s is assumed to be very small and
stationary. Although the number of such loci is assumed to be insignificant
with regard to contribution to 1(t), the assumption that dn2/dt is zero can be
used to eliminate n2 from the expressions for dnjdt and dn2/dt. The
following treatment of the model is essentially that of Birtwistle and
Blackley (1981a). Setting n3 = n4 = O in the equation for dn2/dt [Le.,
Eq. (2) with i = 2], and setting dn2/dt = O,gives
n2 = O"nj[O"
+ 2(k + X)]
(73)
from which it is evident that n2 ~ n1 implies that O"~ O"+ 2(k + X),and thus
that O"can be neglected relative to 2(k + X). Substituting for n2 from Eq. (73)
into the equations for dno/dt and dnjdt [Le., Eq. (2) with i = O and i = 1,
respectively] gives equations for dl1o/dt and dnjdt which can be expressed
in matrix form as
~
dt
no
[J
n1
O" -q
no
(74)
Polymerization
Reactions
181
[ ]
= c [ A ] e)."
nl(t)
+c
[A ]
2
(75)
e).21
where Al and A.2are the eigenvalues of the matrix of coefficients in Eq. (74),
[Al] and [A2] are the corresponding column eigenvectors, and CI and C2
are arbitrary constants. The following results for vo(t), VI(t), and f(t) are
obtained by this method:
vo(t)
f(t)
=~
O'+q
= VI(t)
1+
(76)
~e-(<1+q)/
q
O'
= -{1-
(77)
e-(dq)/}
O'+q
= k + O'X/(k+ X).
Solutions for a "three-state" model have been published by Brooks
where q
(1978,1979).In this model, it is assumed that the n{t)values are strictly zero
for i > 2, so that the only locus populations that have to be considered are
'1
:+
y.
kt
Fig. 10.
Predictions for average number of radicals per locus I(t) as a function of time t
obtained by numerical solution of time-dependent Smith-Ewart differential difference equations, showing effect of increasing /. keeping (1 and k constant. Values taken for (1and k are
both 1 x 10-3 sec-'. Values taken for /. are as follows: A, O; B, 1 X 10-5 sec-'; e, 1 x 10-4
sec-I; D, 1 X 10-3 sec-t; E, 1 X 10-2 sec-'; F, I X 10-' sec-'. (Reproduced with per'mission of J. Chem. Sal'. Faraday /.)
182
D. C. Blackley
0.4
,-
0.2
A
time (sec)
Fig. 11 . Predictions for average number of radicals per locus, (t), as a function of time
t obtained by numerical solution of time-dependent Smith-Ewart differential difference equations, showing effect of decreasing k and increasing X, keeping u constant. Value taken for u is
1 X 10-5 sec-I. The values taken for k and X are as follows: A: k = 1 X 10-3 sec-I,
X = 1 X 10-8 sec-I; B: k = 1 x 1O-4sec-1, X = 1 X 10-6 sec-I; C: k = 1 X 10-5 sec-I,
X=lxl0-4sec-l;
D: k=lxl0-6sec-l,
X=lxl0-2sec-l;
E: k=lxl0-7sec-.,
X = 1 x 10 sec-I. (Reproduced with permission of J. Chem. SocoFaraday l.)
4.
Kinetics of Compartmentalized
183
that are initially devoid of radical s and in which, at a certain time (taken as
t = O),radicals suddenly begin to be generated in the external phase at a
constant rateo Figure 10 illustrates the effect of increasing Xon the variation
of l(t) with t, (1,and k being held constant. As expected, an increase in the
value of X leads to a reduction in the value of "t at any instant. Figure 11
shows the effect of increasing X and decreasing k on l(t) as a function of t,
with (1being held constant. As expected, for large values of X/k, the value of
i(t) at long times approaches the Smith-Ewart Case 2 value of 0.5.
Numerical solutions have also been obtained by Brooks (1980) using his
three-state model (see above). The relevant simultaneous differential equations were solved by Euler's method. Brooks has included in his examples
the case where (1 decays exponentially with time. In all the cases investigated, he finds that allowance must be made for bimolecular mutual
termination of radicals and for the re-entry of desorbed radicals into
reaction loci; he concludes that failure to take account of these possibilities
can lead to serious errors.
VI.
This chapter has been concerned exclusively with predictions for the
distribution of locus populations within a compartmentalized free-radical
polymerization reaction system. Other matters of considerable interest are
the distribution of polymer molecular weights which the polymerization
reaction produces, and the distribution of sizes of the reaction loci at the end
of the reaction. A significant literature concerning both these aspects is
beginning to develop, but beca use of the complexity of the subject and
limited space, oniy a brief summary of the various contributions can be
given here.
The distribution of molecular sizes produced by a given addition polymerization reaction is determined by the balance of processes such as
propagation, termination, transfer; branching, and cross-linking. In fact, the
final polymer molecule produced is the "corpse" of the kinetic chain by
which it was produced, as modified by transfer, branching, and crosslinking reactions. Early attempts to consider the distribution of molecular
weights produced by a compartmentalized free-radical polymerization reaction were made by Katz et al. (1969) and by Saidel and Katz (1969).
These workers derived a set of coupled partial differential equations, from
the solution of which the molecular-weight distribution can, in principIe, be
predicted. However, except for a few simple cases, solutions were restricted
to predicting the lower moments of the molecular-weight distribution, and
184
D. C. Blackley
even these predictions required moderately extensive numerical calculations. Furthermore, the treatment did not take into account several
important molecular events, such as the possibility of mutual termination
with disproportionation, and the various transfer reactions. An alternative
treatment, given by Min and Ray (1968), is also restricted to the prediction
of the lower moments of the molecular-weight distribution.
More recently, Sundberg and Eliassen (1971) have attempted a prediction of molecular-weight distribution for a reaction taking place isothermally in a well-mixed batch reactor. Micelles are assumed to be the only
source of reaction loci. Both micelles and loci receive radicals at arate
proportional to surface area. All the radicals produced in the external phase
are absorbed by micelles and loci. Termination takes place immediately
when two radical s become present in a locus and combination is the
outcome. Allowance is made for the reaction of transfer to monomer. The
predicted molecular-weight distribution is of the Flory "most probable"
type with a polydispersity ratio of 2.0. As might be expected, the value of
the rate coefficient for transfer to monomer relative to that for propagation
has an important effect on both the average molecular weight and the distribution of molecular weights.
An extensive treatment of this subject has been given very recently by
Lichti et al. (1980),and a brief summary was given in an earlier paper
(Lichti et al., 1978).The model assumed for this treatment is a three-state
model in which i is O,1,or 2. An earlier paper (Lichti et al., 1977)applied a
similar treatment to a two-state model in which i is O or 1. The treatment
allows for the possibility that mutual termination may result in either
combination or disproportionation. It also allows for the possibility of
transfer to monomer. It has not, however, been possible to make allowance
for branching and cross-linking. Prediction of the full distribution of
molecular sizes, and not merely of particular moments of the distribution,
has been achieved. The conclusion has been reached that compartmentalization of the reaction leads to a broadening of the molecular-weight
distribution.
An early contribution to the cognate matter of the prediction of the
eventual distribution of locus sizes was made by Ewart and Carr (1954).
The two principal factors that affect the nature of this distribution were
considered to be (i) the distribution of locus sizes present when locus
nucleation ceases, and (ii) the manner in which the loci grow subsequently as
a consequence of polymerization. A third factor that may influence the
growth pattern is the tendency for large loci to imbibe more monomer than
do smallloci.
O'Toole (1969) has discussed the extent of stochastic (as contrasted with
deterministic) con tribu tion s to the polydispersity of locus sizes. He con-
'P(~, t, v) =
v(t, v)~
=o
(79)
where the function v(t,v) is such that v(t,V)DVis the fraction of the total
number of reaction loci which at time t contain i propagating radical s and
which also ha ve a volume that lies between v and v + DV.The relationship
between this locus-population generating function and the generating function 'P(~, t) which has been used previously is elearly
'P(~, t) =
Vll.
(80)
Loo
'P(~, t, v) dv
186
D. C. Blackley
termination. Compartmentalization therefore enhances the rate of termination relative to propagation, and therefore reduces the rate of polymerization; indeed, according to Haward, polymerization can be effectively
suppressed altogether if the degree of compartmentalization is sufficiently
great. The average degree of polymerization of the polymer produced is also
reduced by compartmentalization. The expression derived by Haward for
the overall rate of polymerization per unit volume of disperse phase in the
steady state is as foUows:
d[M]/dt
-kp[MJX/2k\)1/2
tanh{(X/2k\)1/2vNA}
(81)
(i+l)ni+lk
(i+ I)lnl+lx
(82)
1+2
i+1
(83)
(84)
The solution of this equation consistent with the physical constraints of the
reaction system is
l-m
'P(~) = A(l
(85)
where the combinatorial symbols have their usual meanings, and j is zero
for even values of i.and unity for odd values of i. The average number. of
propagating radical s per locus is found to be given by
(87)
ih tanh(h/4)
List of Symbols
A
[AJ,(AzJ
a
B
Bp
constant
column eigenvectors
(4<TX+ kZ)IIZ or surface area of spherical reaction locus
constant
constant associated with term containing e-ti..
(88)
188
b
q-I/21(~)
nCr
e
C,C2
h
l.
;
1
Jt..lm
j
Kp
k
ko
kp
kt
ktE
k,p
[o
p
M
[M]
m
N
NA
ni
D. C. Blackley
0,1,2, ...
of order p and parameter
stationary
concentration
mutual
mutual termination
mutual termination
concentration
kvlkt
of monomer
k+~
x
y
of matrix [O]
locus
T(t)
t
V
v
X
in external phase
in reaction loci
= klx
0,1,2, oo.
rip
at.reaction
loci
termination
locus
[n;]
R-i+l
-1-
Gegenbauer polynomial
combinatorial
symbol
k+X
rate of transition of loci across national boundary between states of radical
occupancy ; and ; + 1
radius of sphericaI reaction locus
ith component of pth right eigenvector of matrix [O]
maximum number of propagating radicals in single reaction locus
part of 'I'(~, 1) which depends only on 1
time
variance of distribution of locus populations
volume of reaction locus
rate of formation of new radicals containing one monomer unit, in unit volume of
monomer
2';6(1 + ~) o/' (1 + W2
'I'(~)
xl-m
Z
ex
2Nk,pk,Jy2v
see (10
p
r
(1/2X
pIe
gamma function;
(u). is a Pochhammer
e
e'
u'v/k,
O(t)
).
AI,A2
Ap
J1
Vi
E(~)
~
p
p'
u
u'
Uo and C(
,
'p
q,(t)
X
'P
[n]
= u'/x
time-dependent
parameter of Poisson distribution
separation constant
eigenvalues associated with [A tJ, [A2]
pth eigenvalue of matrix [n]
parameter of initial Poisson distribution of locus populations
fraction of loci which contain i propagating radicals
part of 'I'(~, t) which depends only on ~
auxiliary variable of locus-population
generating function
rate of entry of radicals into all reaction loci in unit volume of reaction system
rate of formation of new acquirable radicals in unit volume of reaction system
average rate of entry of radicals into a single locus reaction locus, i.e., p/N, or rate
of formation of radicals within a single reaction locus
p'/N
are such that u = u o r"
average number of propagating radicals in single reaction locus
I/XP(P + m - 1)
time-dependent
parameter of Poisson distribution
kJv
locus-population
generating function
matrix of coefficients
= [no]
+ [ntJ
References
Allen, P. W. (1958). J. Col/od Sci. 13,483-487.
Azad, A. R. M., and Fitch, R. M. (1980). Paper presented
Polym.
173, 1998-2009.
175, 2051-2064.
177, 413-426.
177,1351-1358.
Birtwistle, D. T., Blackley, D. c., and JefTers, E. F. (1979). J. Chem. Soco Faraday / 75,
2332-2358.
Blackley, D. C. (1975). "Emulsion
Polymerisation:
Theory and Practice,"
Science, London.
Blackley, D. C., and Haynes, A. C. (1977). Br. Polym. J. 9, 312-321.
Brooks,
p. 95. Applied
3022-3056.
"Polymer
Colloids"
"Polyrner
Colloids"
Friis, N., and Hamielec, A. E. (1973). J. Polym. Sci. Polym. Chem. Ed.2,3321-3325.
190
D. C. Blackley.
Gardon,
Gardon,
Gilbert,
Gilbert, R. G., Napper, D. H., Lichti, G., Ballard, M., and Sangster, D. F. (1980). Paper
presented at Am. Chem. Soco Emu/sion Po/ym. Symp., Las Vegas Preprints, pp. 527-530.
Grancio, M. R., and Williams, D. J. (1970). J. Po/ym. Sci. Part A-1S, 2617-2629.
Hansen, F. K., and Ugelstad, J. (1978). J. Po/ym. Sei. Po/ym. Chem. Ed. 16, 1953-1979.
Hanseii, F. K., and Ugelstad, J. (1979a). J. Po/ym. Sei. Po/ym. Chem. Ed. 17,3033-3046.
Hansen, F. K., and Ugelstad, J. (1979b). J. Po/ym. Sci. Po/ym. Chem. Ed. 17, 3047-3068.
Haward,
R. N. (1949).
Hawkett,
B. S., Napper,
Po/ym.
Sci.
4, 273-287.
Hawkett, B. S., Napper, D. H., and Gilbert, R. G. (1977). J. Chem. Soco Faraday 173,690-698.
Hawkett, B. S., Napper, D. H., and Gilbert, R. G. (1980). J. Chem. Soco Faraday 176,1323-1343.
Katz, S., Shinnar, R., and Saidel, G. M. (1969). Adv. Chem. Ser. 91, A.C.S., p. 145-157.
Kendal, M. G., and Stuart, A. (1963). In "The Advanced Theory ofStatistics,"
2nd ed., Vol. 1,
p. 30. Griffin, London.
Keusch,
Keusch, P.,
Keusch, P.,
Lansdowne,
Faraday
1971.
Lichti, G., Gilbert, R. G., and Napper,
Latex Conf, London Paper No. 1.
Lichti, G., Gilbert,
1323.
Min, K.
Morton,
Napper,
O'Toole,
O'Toole,
Saidel, G. M., and Katz, S. (1969). J. Po/ym. Sci. Part C. 27, 149-169.
Smith, W. V., and Ewart, R. H. (1948). J. Chem. Phys. 16,592-599.
Stockmayer, W. H. (1957). J. Po/ym. Sci. 24, 314-317.
Sundberg, D. C., and Eliassen, J. D. (1971). In "Polyrner Colloids" (R. M. Fitch, ed.), pp.
153-161. Plenum Press, New York.
Ugelstad, J., and Hansen, F. K. (1976). Rubber Chem. Techno/. 49, 53/Hi09.
Ugelstad, J., Merk, P. c., and Aasen, J. O. (1967). J. Po/ym. Sci. Part A-I 5, 2281-2288.
van der Hoff, B. M. E. (1958). J. Po/ym. Sci. 33, 487-490.
van der Hoff, B. M. E. (1962). In "Polyrnerization
and Polycondensation
Processes," Adv. in
Chem. Ser. 34, pp. 6-31. American Chemical Society.
Vanderhoff, J. W. (1976). Proc. Water Borne Higher Solids Coating Symp. Paper No. 6.
Vanzo, E., Marchessault,
R. H., and Stannett, V. (1965). J. Col/oid Sci. 20, 62-71.
Watterson, J. G., and Parts, A. G. (1971). Makromo/. Chem. 146, 11-20.
Weiss, G. H., and Dishon, M. (1976). J. Chem. Soco Faraday 172,1342-1344.
Williams, D. J. (1971). J. E/astop/ast. 3, 187-200.
Williams,
5
Mamoru Nomura
1. Introduction .
11. Polymerization Rate Equations Involving FreeRadical Desorption .
A. Reaction Locus .
B. Polymerization Rate Equations
111. Derivation of Rate Coefficient for Radical Desorption
from Particles.
A. Definition of Radical Desorption
and Reabsorption
B. Derivation of Mass- Transfer Coefficient
for Radical Desorption and Reabsorption .
C. Derivation of Rate Coefficient for Radical
Desorption
IV. Effect. of Free-Radical Desorption on the Kinetics
of Emulsion Polymerization
A. Effect on the Rate of Emulsion Polymerization.
B. Effect on Micellar Particle Formation
V. List of Symbols .
References
l.
192
192
194
199
199
202
204
210
210
214
217
219
Introduction
POLYMERIZATION
Copyright
It> 1982 by Aeademie Press. 1ne.
AIl rights of reproduetion
in any form reserved.
ISBN 0-12-556420.1
192
Mamoru Nomura
radical s from particles. Although the importance of this physical phenomenon was pointed out by Smith and Ewart in the 1940s, the quantitative
understanding was insufficient to explain the kinetic deviation of vinyl
acetate and vinyl chloride emulsion polymerizations from the Smith-Ewart
Case 2 kinetic theory.
Recently, Ugelstad et al. (1969) proposed a semiempirical rate coefficient
for radical desorption in vinyl chloride emulsion polymerization. On the
other hand, Nomur.a et al. (1971, 1976) have derived arate coefficient for
radical desorption theoretically with both stochastic and deterministic
approaches and have successfully applied it to vinyl acetate emulsion
polymerization. They also pointed out that radical desorption from the
particles and micelles played an important role in micellar particle formation. Friis et al. (1973) also derived the rate coeffieient for radical
desorption in a different way. Litt et al. (1981) discussed in more detail the
chemical reactions incorporated in the physical process of radical desorption in the emulsion polymerization of vinyl acetate..
In this chapter, the polymerization rate equations for emulsion polymerization will be reviewed briefly: Then, the rate coefficient for radical desorption from the particles will be derived theoretically, and the effect of
radical desorption on the rate of emulsion polymerization and the micellar
particle formation will be discussed.
Reaction Locus
Harkins (1947) and Smith and Ewart (1948) assumed that in emulsion
polymerization the polymerization loei were inside the particles because the
rate of styrene emulsion polymerization, for example, was proportional to
the number of polymer particles presento However, the polymerization loci
in the emulsion polymerization of water-soluble monomers such as vinyl
acetate and vinyl chloride have long been discussed because the rate of
emulsion polymerization of these monomers was not proportional to the
number of polymer particIes presento
Patsiga et al. (1960) and Giskehaug (1965) carried out the seeded emulsion polymerizations of vinyl acetate and of vinyl chloride, respectively,
and found that in both systems the rate of polymerization was proportional
only to the 0.15-0.20 power of the number of polymer particles. From this
finding they concIuded that the main locus of the emulsion polymerization
of water-soluble monomers must be in the water phase. This conclusion
seems reasonable because the solubilities of these monomers are about 100
Free Radicals
193
times greater than that of styrene. But this inference may be excluded from
the following simple discussion.
Let us consider a seeded emulsion polymerization where no particle
formation occurs and hence the number of polymer particles is constant. At
a steady state, the overall rate of radical entry into the particles PA is
expressed by
=
PA
[
-
J+
(1)
Since the value of PA is not so different from the value of Pw, the rate of
radical production in the water phase, we get
(2)
It must be noted here that the quantities PA and Pware different from one
another.
Ir the diffusion law can be applied to the radical entry into the particles,
PA is given by
PA = ka[R~JNT~ 2ndpDw[R~JNT
(3)
(4)
tadd= l/kp[MwJ
(5)
(6)
where kp is the propagation rate constant, [MwJ is the monomer concentration in the water phase, and ktwis the termination rate constant in the
water phase. The values of tred, tadd' and ttermare calculated to be -10-5,
-4 x 10-4, and - 103 sec, respectively, when the following numerical constants are used: ktw= 6 X 108 (dm3/mol sec), Pw= 1016(molecules/dm3 sec),
194
Mamoru Nomura
Rp
= -dM/dt = kp[Mp][R;]Mw .
(7)
NT=No+Ni+N2+...+Nn+"'=
Nn
n=O
n=i
nNn/NA
(8)
(9)
= (~i nNn)j NT
(10)
Thus, the rate of emulsion polymerization given by Eq. (7) is also rewritten
as
Rp = kp[Mp]NTMw/NA
(11)
195
from the particles, and the rate of radical termination within the particles.
For the nonsteady state
dNn dt
PA
()
-( )
N T Nn-l
+ k eap
PA
n+ I
(~ )
()
N n+l
+ k tp
NT Nn - keap vp Nn - ktp [
(n + 2)(n + I) N
n+2
n(n - I)
vp
J Nn
(12)
where ke is the rate constant for radical desorption from the particles
(m/sec), ap is the surface area of a particle (m2), vp is the volume of a particle
(m3), and ktp is the rate constant for termination (m3/molecule sec). Since
the rate constant for radical desorption should depend on the radical chain
length, the rate constant ke represents an average value.
The rate coefficient for radical desorption from the particles kr is defined
as
(13)
wh~e
(14)
where kj is the rate constant
j-
mer radical s, 'j is the ratio of the number of j-mer radicals to the total
number of radicals nNn contained in Nn particle, and j = O indicates the
initiator radical.
By solving Eq. (12) for Nn and introducing it into Eq. (10), one can
calculate the value of , and accordingly, 'the rate of emulsion polymerization by Eq. (11).. Although many investigations have been reported
concerning the solution of Eq. (12), only the principal work will be reviewed
briefly here. For more details, the readers should refer to the review article
by Ugelstad et al. (1976).
Smith and Ewart (1948) did not obtain a general solution for Eq. (12) but
rather solved it for three limiting cases, applying a steady-state hypothesis,
i.e., dNn/dt = O.
Case 1 PA/NT ~ kr
In this case, the following relation holds:
(15)
Further,
(a) When termination in the water phase is dominant (ktp/vp ~ kr,
Pw~ 2ktw[R~]2)
(16)
Mamoru Nomura
196
(17)
kr ~ pAiNT ~ ktP/vp
= 0.5
(18)
= (PAvp/2ktpNT)1/2 ~
0.5
(19)
(20)
= -!a[Im(a)/Im-l(a)]
(21)
where
a2 = 80(
O(= PAvp/ktpNT
(22)
and
m = krvJktp
(23)
Gardon (1968) has solved Eq. (12) numerically for the case of negligible
desorption of radical s from the particles without assuming the steady state,
stating that the Stockmayer solution for is incorrect because there is no
steady state in principie and because Eq. (12) includes the time-dependent
parameter vp' However, the results of numerical calculation by Gardon
coincide almost completely with those predicted by the Stockmayer solution for no radical desorption from the particles. This also supports the
validity of applying the steady-state hypothesis to the solution for Eq. (12)
under normal conditions for emulsion polymerization.
Noting that in the treatments by Smith and Ewart, by Stockmayer, and
by O'Toole the "fa te " of the desorbed radicals was not necessarily specified,
5.
197
Ugelstad et al. (1967) claimed that Eq. (12) should be solved simultaneously
with the following balance equation on the radicals in the water phase.
(24)
Since the steady-state hypothesis is applicable to Eq. (24), it is rewritten in
the same form as Eq. (1):
PA = ka[R:JNT = Pw
+ kfnNn - 2ktw[R:J2
(25)
(27)
m = kfVp/ktp
(23)
= 2ktwktp/k;NTvp
(28)
and
y
Ugelstad et al. (1967) solved the simultaneous equations Eqs. (21) and (26)
for and plotted the calculated value of against the value of a' at fixed
value of Y, varying the value of m as a parameter. Since the radical
termination in the water phase is negligible under normal emulsion polymerization conditions as mentioned in Section ~I,A,the condition Y = Ois most
important for usual emulsion polymerizations. Figure 1 shows a plot of
10'
y=o
IC
164
10-2
104
a'
Fig. 1.
Relationship between and IX'when termination in the water phase is neglected.
IX'= Pwvp/NTk,pand m = kfVp/k,p. (Reprinted with permission of Journal of PolyrnerScience.)
198
Mamoru Nomura
log versus logO(', varying the value of m for the case of Y = O. The
treatment by Ugelstad et al. (1967) is most general and correct, and hence,
Eqs. (21) and (26) are important in predicting the value of , that is, the rate
of emulsion polymerization.
Since it is inconvenient to use Eqs. (21) and (26) directly, however, several
empirical or approximate equations for are derived for the case where
Y=O.
1. When m = O
1/2
0('
( )
n= -+4 2
(29)
= !(- e + J c2 + 2e)
e = O('/m= Pw/kfNT
(30)
= 0.5
(31)
(32)
0('
[(
n= -
4.
1+-
1 1/2
m)J
(33)
When m -4 00 or ~ 0.5
= (0('/2)1/2
(34)
199
= kv-2/3
= k'd-2
p
p
(35)
Nomura et al. (1971) also discussed this problem almost at the same time
and derived the theoretical expression for kf, which was inversely proportional to the square of particle diameter, as shown by Eq. (35). In this
section, the derivation of kf by the Nomura and Harada method is mainly
explained (Harada et al., 1971; Nomura et al., 1976; Nomura and Harada,
1981).
A.
200
Mamoru
INTERFACE
I
WATER PHASE
PARTlCLE
-DISTANCE
Fig. 2.
(Reprinted
Nomura
Cp
= n/vp.
interface between two phases, the concentration gradients near the phase
boundary can be assumed to' be as shown in Fig. 2. Applying Fick's
diffusion law, the rate of radical desorption from a single particle that
contains, for example, n radicals with same chain length, is generally
expressed by
JRo= -d(vpCp)/dt = -dn/dt = ksap(Cp- Cpi)= kwap(Cwi- Cw) (37)
where ks and kw, respectively, denote the film mass-transfer coefficients for
the inner and outer diffusion films adjacent to the interface between the
particle and water phases, Cp is the concentration of radical s in the particle,
Cw is the concentration of radical s in the water phase and i denotes the
interface. The concentration of radicals at the interface, which is at equilibrium and usually very low, can be expressed by the linear relationship ,
Cpi
= rnd Cwi
(38)
dn =
Cp
dt (l/ksap)
- rndCw
+ (rnd/kwap) -
Ksap(Cp
rndCw)
(39)
(40)
here
l/ks + rnd/kw
(41)
(42)
l/Ks
201
= Ksjap[(l/vp)
-dNlfdt
- mdJR:J]N/ + Ksjap[(O/vp)
- mdJR:J]
(44)
= -Ks/ap/vp)NJ
(46)
(47)
The second term on the right-hand side of Eq. (47) usually represents the
absorption rate of j-mer radical from the water phase into the active
particles that already contain a radical, and hence represents the rate of
increase in the number of particles containing two radicals, one of which is
a j-mer radical. However, this term must be changed from positive to
negative because it expresses the rate of decrease in the number of active
particles when instantaneous termination in the particles is assumed. Thus,
using Eq. (43), Eq. (47) can be rewritten as
dN* /dt = - ('1KsjN/)(ap/vp)- ( KWjmdJR:j])apN*
(48)
202
Mamoru
Nomura
kf
= LKsiNNN?i')(ap/vp)
= LKOJ{NNN*)
ka = (LKw~[R:j]/[R:])ap
KOj =
= Kai[R:j]/[R:])
Ksiap/vp) (49)
where
[R:] = [R:.J + [R:] + [R:2] + o.. + [R:j_] + [R:j] + 0.0 (51)
In this case, ke and (j defined in Eqso (13) and (14) are, respectively,
represented by
ke =
B.
KSJ{Nf/ N*)
Derivation 01 Mass-Transler
and Reabsorption
(j
= (NNN*)
In order for the rate coefficients kf and kaodefined by Eqso (49) and (50),
to be predicted it is necessary to know how KSj and Kwj are expressed in
terms of the chemical and physical properties involved in the desorption
and reabsorption processeso First of all, let us consider the mass-transfer
coefficient for a j-mer radical in the individual diffusion filmso There is a
large number of published studies concerning the mass-transfer coefficient
in an external diffusion film around a spherical particleo One of these is the
following semitheoretical equation proposed by Ranz and Marshall (1952)
Sh
=2 +
006R~/2S~/3
(52)
(53)
On the other hand, mass-transfer inside the viscous polymer particles will
occur by molecular diffusiono According to an analytical solution to the
diffusion equation inside the particles, average mass-transfer coefficient
between time O", t,ks is given as follows (Newman, 1931):
_k 0=
SJ
- dp 1n
6
t
[~ ~ 2- (
2 L-
1tn=n
2 exp
4n21t2Dpl
dp2
)]
(54)
203
Since this coefficient is too complicated to employ for our present purpose,
we use an approximate mass-transfer coefficient derived by the following
simple treatment. The average time t spent by a radical inside a particle
before it escapes from the particle can be ca1culated using the Einstein
diffusion equation. Nomura and Harada (1981) used the following value
for t:
(55a)
Recently, Chang et a/. (1981) considered two cases: (1) when a radical enters
at the edge of the particle t is given by
(55b)
and (2) when a radical is generated anywhere in the particle t is given by
(55c)
We assume that the rate of radical desorption from the particle can be
ca1culated approximately using the following equation obtained from
Eq. (37) by neglecting the term Cpji because the value of Cpji is usually very
low compared to that of Cpj.
(56)
Using Eq. (56), the average time t spent by a radical inside the particle
before it escapes fram the particle will be given by
t =. (CpjVp)/lR*= (Cpjvp)/ksjap Cpj = dp/6ksj
By comparing
= /3(Dp/dp)
/3 = i
-i
(57)
(58)
mdjdp
2Dwj
/3mdjDpj
-1
= 2Dwj
bj
(59)
mdjdp
and
KWj
= m.K
dJ
sj
- 2Dwjb.
l
(60)
dp
where
(61)
Mamoru Nomura
204
kf
[R~I
[R~l]
[R~2]
[R~s] "-
(62)
(63)
where KQIand Kal are the coefficients for initiator radical s and Ko and Ka
are the coefficients for radicals other than initiators. Discrimination between the initiator and other radicals is made here because their chemical
and physical properties are very different. Thus, the values of KQIand Ko
are different.
In order to get the relationship between Nj and N*, we have the balance
equations with respect to N1*and Nf which contain an initiator radical and
* See Nomura
and Harda,
1981.
5.
205
= KaI[R:]No -
=0
~~
dNNdt
= Ka[R:j]No
= O
+ kp[Mp]Nt-
~~
~~
+ kp[Mp]Ns*-1
+ kmrEMp] + kTf[Tp] + kp[Mp]
+ Ko)Ns*
~~
where k is the initiation rate constant, kTfis the chain transfer rate constant
to chain transfer agent, [Tp] is the concentration of chain transfer agent in
the polymer particles and kmf is the chain transfer rate constant to
monomer. Furthermore, by taking a balance on the water-phase radicals
and applying a steady-state hypothesis we have
(68)
d[R:]/dt
= KoNi -
Ka[R:]NT
=O
(69)
d[R;j]/dt
= KoNt -
Ka[R:j]NT
=O
(70)
(71)
(72)
Pw(1- ) -
n~
KOI + k[Mp]
n~
206
Mamoru Nomura
and
j-
fVt
kp[AIp]
Kon + kp[AIp]
)
)[( )+ ( )fV* + (~ )
fVt
=(
kp[AIp]
kmr
kTf[Tp]
kp
kp[ AIp]
fVt
kp
(75)
kmr
fVt
kr = KOI
kTr[Tp]
kfVt
kp[AIp]
(76)
Introduction of Ego (73) and fV* = fVTinto Ego (76) leads to
k - K
r-
Pw(1- )
O( (KO( + k[AIp])fVT
I(
kp[AIp]
]
) [k; +
j
kmr
+ Ko j= Ko + kp[AIp]
kTf[Tp]
PW(1- )k
kp[AI!,] + (KO(
+ k[AIp])kpfVT]
(77)
p=l-q
(78)
207
Fig. 3. Scheme for radical escape from polymer partic1e. (Reprinted with permission of
Journal of Chemical Engineering, Japan.)
transfer to monomer or transfer agent to desorb from the particles before
growing to (5 + 1)-mer radical is given by
<I\t = q(1 + P + p2 + p3 + ... + p' - 2 + p' - 1) = 1 - p'
(79)
(80)
Part of these radicals will escape from the particles before growing (5 + 1)-
208
Mamoru Nomura
radical entry into the partic1es should be the same as that of radical
desorption from the partic1es. Therefore, the decrease, due to radical
desorption, in the number of partic1es containing escaping radicals is partly
recovered by radical entry into the partic1es containing no radicals. Let N
be the number of polymer partic1es containing an escaping radical. Then,
KoN represents the rate of radical desorption, and KoN(N o/NT)is the rate
of recovery in the number of partic1es containing an escaping radical by
radical entry into. the partic1es containing no radicals. Therefore, the
apparent (or net) rate of radical desorption per partic1e containing an
escapingradical is given by
(84)
1 -'q
(85)
kp[Mp]
Ko + kp[Mp]
~ = Ko f.
kp[Mp]
)
j
= 1-
kp[Mp]j= 1 Ko + kp[Mp]
kp[Mp]
(Ko+ kp[Mp])
(86)
(87)
KOJ
Ko) + k[Mp]
It must be noted here that the rate coefficient for radical desorption kr is
related not to the apparent (or net) rate but to the true rate of radical
desorption by molecular diffusion. Using Eqs. (81) and (83), therefore, the
apparent (or net) rate of radical desorption from the partic1es can be
expressed by the use of the coefficient kr as
krN* -
krN*(No/NT)
= <1>M(kmr[Mp]
+ kTf[Tp])N*
+ Pw(No/NT)[<1 + ~(1
- <1)]
(88)
Introducing Eqs. (86) and (87) into Eq. (88) and rearranging leads to
Eq. (77).
kr
= KOJ
Pw(1
- )
(KOJn+ k[Mp])nNT
Pw(1 -
)k
kmr
] [
]i (
(KOJ+ k[Mp])kpNT
+Ko
kp
+ -kTf[1;']
kp[Mp]
kp[M p]
j= 1 Ko
+ kp[M p]
(77)
209
kmr + kTf[Tp]
kp
I(
kp[Mp]
kp[Mp]
(89)
(90)
j= 1 Ko + kp[Mp]
The value of Z depends upon the values of Ko and kp[Mp] and satisfies the
following inequality:
o< Z < s
(91)
If it is supposed further that only monomer radicals can escape from a
(Ko + kp[Mp])(
kr = KO
kmr + kTf[Tp]
kp
kp[Mp]
(92)
(93)
r-
kmr[Mp]
o Ko + kp[Mp]
) (
+K
kTf[1;']
OT KOT
+ kiT[Mp]
(92')
where KOTand kiT are the desorption rate constant and the reinitiation rate
constant for chain transfer agent radical s, respectively.
210
Mamoru
Nomura
In the absence of transfer agent, when the rate of radical desorption from
the partic1es and the reactivity of monomer are very high, as with vinyl
acetate and vinyl chloride, the condition kp[Mp] ~ Ko is fulfilled. For such
systems Eg. (92) is simplified further to
kf
= Ko(kmelkp) = (12DwDlrndd~)(kmelkp)
(95)
6pDp
2
dp
kmf ~ 6Dp
kp
kmf
(96)
2
dp kp
Using Eg. (55a), Friis and Nyhagen (1973) derived the following expression for kf:
kf
= (2Dpld~Df~:[Mp])
kmf[Mp]
(97a)
= 2Dp kmf
d2p
kp
(97b )
This is the same as Eg. (96) except for the value of the numerical constant.
Ugelstad and Hansen (1976) also derived arate coefficient for radical
desorption similar to those given by Egs. (92) and (95).
IV.
A.
1. Verification of Applicability of kf
211
= ka[R:]No -
kfN*
ka[R:]N*
=O
(98)
(99)
Substituting Eq. (99) into Eq. (98) and solving for , we reproduce Eq. (30):
= 1(- e + ~ C2 + 2C)
e = pw/kfNT
(30)
(100)
monomor
10-1,
50.C
Yinyl
chlaride
SO.C
kfNy
literature
4
5,7
22
1
Ihalsgd
(1
.6. Gi.........
11
L:. Peggion.
24
key
()
o
resean:he1's
Frils "
Nomura
Zollars
Id
IcJ3
te.
vinyl
acotate
[-J
212
Mamoru Nomura
TABLEI
Numerical Values of the Constantsa
Constants
(units)
Vinyl acetate
kp
(Iiterjmolsec)
kmrjkp
md
Dw
(cm2jsec) .
kd
(Iiterjsec)
[Mp). (molj1iter)
z
b
3340
2 X 10-4
28
1.9 x lO-s b
1.5 X 10-6 e
8.9
1.0
1.0
Vinyl chloride
10,000
1.2x 10-3
35
2.5 X lO~sb
1.5 X 10-6,
6.2
1.0
1.0
of the Journal of
growing polymer chains. This was supposed to result from a chain transfer
to monomer and the rate of the interruption reaction rf was experimentally
analyzed according to the expression
rf = k[R .][M]
(101)
213
conc1uded that Eqs. (30) and (95) are applicable to vinyl acetate and vinyl
chloride emulsion polymerizations.
On the other hand, for such monomers as methyl methacrylate and
styrene which are less reactive and less soluble in water than vinyl acetate
or vinyl chloride, Eqs. (30) and (76) [or (77)] are essentially applicable,
although the value of the term (kJkp)(N*jN*) may not necessarily be
neglected comparing the value of the term kmr/kp.
The rate coefficient for radical desorption in emulsion copolymerization
was also derived in the same way as described in Section III, and it was
successfully applied to explaining the rate of emulsion copolymerization of
methyl methacrylate and styrene (Nomura et al., 1978, 1979).
2. Effect of Radical Desorption on Rate of Emulsion Polymerization
Let us consider the rate of emulsion polymerization Rp where Eqs. (30)
and (95) can be applied. From Eqs. (30-32), it is c1ear that Eq. (30) can be
written in the form
(102)
The values of a varies from ! to O with an increase in the value of C.
Combining Eqs. (95) and (102) we have
I ] av2a/3
R p oc N T oc N1-a
(103)
] ad2aoc N1-a
T [I Op
T [ Op
After the complete absorption of monomer existing as monomer droplets in
the water phase, the following relationship holds:
NTvp= Vp
(104)
where Vpis the total volume of the partic1e per unit water volume. Inserting
Eq. (104) into Eq. (103) leads to
R p oc N(3
(105)
T - saJ/3[ I ] aJl:2a/3
p
When radical desorption
Eq. (105) becomes
R p oc N1.0
T [I ] 0Jl:0
p
(Le., a = O)
(106)
(107)
This theoretical equation explains very well the rate of seeded emulsion
polymerization of vinyl acetate and vinyl chloride found by Patsiga et al.
214
Mamoru Nomura
(1960) and Giskehaug (1965). Recently, Ugelstad et al. (1969) reported that
when the number of polymer particles present was large, the rate of vinyl
chloride emulsion polymerization is proportional to the 0.15 power of the
number of polymer particles and to the 0.3 power of the total volume of
polymer
particles
~.
Nomura
that
Eq. (107)
explained the dependence of the rate of vinyl acetate emulsion polymerization on NT, [/0], and ~. Van der Hoff (1956) reported that even in
styrene emulsion polymerization, the rate of polymerization was proportional to the 0.17 power of the number of polymer particles when the
number of polymer particles was very large. These experimental findings
support the validity of the theory developed in this chapter.
B.
Smith and Ewart (1948) proposed two idealized situations for the
formation of polymer particles, assuming that (i) partile nucleation occurs
in monomer-swollen emulsifier micelles, (ii) the volumetric growth rate of a
particle is constant in the interval of particle formation, and (iii) radicals do
not desorb from a particle.
Case 1: Ir only micelles can absorb initiator radicals, then
dNT/dt = Pw
(108)
(109)
where /l is the volumetric growth rate of a particle, as is the surface area per
unit amount of emulsifier, and Sois the total amount of emulsifier per unit
volume of water.
Case 2: Ir both particles and micelles absorb radicals at arate
portional to the surface area, then
dNT/dt = pw[A.,J(Am + Ap)]
pro-
(110)
where Am and Ap are the total surface areas of micelles and polymer
particles, respectively. In this case
NT
0.37(pw//l)0.4(asSo)0.6
(111)
215
Noting this, Nomura et al. (1972, 1976; Nomura, 1975) studied the effect
of radical desorption on the formation of polymer particles from micelles
for the case Y = o. They used the following non-steady-state treatment
under the assumption that the polymer particles contain at most one
radical.
Particle formation:
dNT
k1ms[R:J
klms
+ kzNT[R:J = PA klms + kzNT
dt = PA
-
k1ms[R:J
) (
PA
1 + aNT/Sm
)
(112)
where
(113)
and
PA
(114)
216
Mamoru Nomura
formation when the conditions k1 = nd~, k2= nd~,and kf = O are employed (where dmis the diameter of a micelle).
Nomura et al. (1972) solved Eqs. (112) to (118) on a digital computer and
analyzed the particle formation in styrene emulsion polymerization where
kf = O approximately held under normal conditions and found that the
value of 8 was 1.3 x 105. This value is about 102 times greater than that
predicted by the diffusion theory which assumes that k1 = 2nDdm and
k2 = 2nDdp,and whereD is the diffusioncoefficientof a radical in the water
phase.
On the other hand, in vinyl acetate emulsion polymerization the value of
8 was 1.2 x 107 (Nomura et al., 1976). This value is also about 104 times
greater than that predicted by the diffusion theory. The reason for this may
be that radical s have greater difficulty in entering micelles than polymer
particles, or it may be that radicals, having entered a micelle, may escape
from the micelle too rapidly to cause initiation, because the micelle has too
small a volume. Both factors will decrease the apparj::nt value of k1 and
hence increase the value of 8. Therefore, 8 can be regarded as a factor that
represents "the radical capture efficiency of a micelle relative to a particle."
Furthermore, they studied the effect of radical desorption from the
polymer particles on the particle number formed in Internal 1 of emulsion
polymerization, using above equations, and showed that radical desorption
leads to an increase in the number ofparticles. The reason for this is that: (1)
The desorbed radicals reenter the micelles and take part in particle nucleation: (2) Radical desorption decreases the particle growth rate and hence,
results in a decrease in the rate of micelle consumption. This also increases
the chance of radical entry into the micelles. They also showed that radical
desorption brought about the change in the orders of particle number with
respect to emulsifier and initiator according to the following relationship:
NT oc S~p~-a
(119)
The value of a increases from 0.6 to 1.0 with increasing radical desorption.
For vinyl acetate and vinyl chloride emulsion polymerization, the calculated
217
Furthermore, the value of a in Eq. (119) increases from 0.6 to 1.0 with
increasing transfer agent (Nomura, 1975; Nomura et al., in press).
List of Symbols
a
ap
a,
Am
Ap
C
Cp
Cp;
Cw
Cw;
dm
dp
Dp
f
I
Im,/m-1
[lo]
j
Jpo
k.
kd
k
k
kj'
(8:x)1/2
surface area of a partic1e
partic1e surface area per unit amount
total surface area of micelles
radicals
radical
kTf
k.
kw
m
md
'n~
Mm
Mw
[Mp]
[Mw]
radical in a particle
with j number
of monomer
units, j = O
(j = 1,2,3,
) rate constant for radical desorption
j-mer radicals
rate constant for transfer to monomer
k,p
k,w
of emulsifier
interface
to the
to the
218
Mamoru Nomura
Mo
n
N
N*
NA
N.,
initial
monomer
number
concentration
of radicals in a particle
average number
of radicalsper particle
number
number
Avogadro's number
(n =
NT
P
0, 1,2, 3...)
number
of particles containing
number
radicals
Re
Rp
[R:]
[R:]
5
Sc
Sh
So
Sm
t
unit
Reynold's number
overallrate of polymerization
concentration of radicalsin the particlesper unit volume of water
totalconcentration of radicalsin the water phase
maximum
number of monomer
units in a desorbed radical
Schmidt number
Sherwood
number
totalamount
amount
particle
tadd
red
particle
fterro
termination in
[1;,]
Vp
Vp
Xm
concentration of transferagent
volume
of a particle
ex
ex'
numerical
constant,
!- i
ratio of the outer film mass-transfer resistance to the overall mass-
Cl
J1
ratio of number
radicals
of radicals
PA
Pw
<1>m
fraction
of
tbe
transfer agent
in the water
radicals
to desorb
generated
from
phase
by
chain
transfer
to
growing
monomer
to (5 +
or
1)-mer
radicals
<1>
t/lc
monomer
unit
weight fraction
in the particles
atsaturation
with monomer
219
References
Chang, K., Litt, M., and Nomura, M. (1981). Macromolecules (to be published).
Friis, N., and Nyhagen, M. (1973). J. Appl. Polym. Sci. 17, 2311.
Gardon, J. L. (1968). J. Polym. Sci. Part A-l, 6, 665.
Giskehaug, K. (1965). Symp. Chem. Polym. Process, London, April.
Hansen, F. K., and Ugelstad, J. (1979). Macromol. Chem. ISO, 2423.
Harada, M., Nomura, M., Eguchi, W., and Nagata, S. (1971). J. Chem. Eng. Jpn. 4(1), 54.
Harkins, W. D. (1947). J. Am. Chem. Soco 69, 1428.
Lewis, W. K., and Whitman, W. G. (1924). Ind. Eng. Chem. 16, 1215.
Litt, M., Patsiga, R., and Stannett, J. (1970). J. Polym, Sci. Part A-l S, 3607.
Morris, C. E. M., and Parts, A. G. (1968). Macromol. Chem. 119,212.
Newman, A. B. (1931). Trans. AIChE 27, 310.
Nomura, M. (1975). Ph.D. Thesis, Kyoto Uniy., Kyoto, Japan.
Nomura, M., and Harada, M. (1981). J. Appl. Polym. Sci. 26, 17.
Nomura, M., Harada, M., Nakagawara, K., Eguchi, W., and Nagata, S. (1971). J. Chem. Eng.
Jpn.4(2), 160.
Nomura, M., Harada, M., Eguchi, W., and Nagata, S. (1972). J. Appl. Polym. Sci. 16,811.
Nomura, M., Harada, M., Eguchi, W., and Nagata, S. (1976). In "Emulsion Polymerization"
(1. Piirma and J. L. Gardon, eds.), ACS Symposium Series No. 24, pp. 102-122.
Nomura, M., Harada, M., and Eguchi, W. (1978). J. Appl. Polym. Sci. 22, 1043.
.
Nomura, M., Yamamoto, K., Fujita, K., Harada, M., and Eguchi, W. (1978). Preprints, 12th
Fall Meeting ofthe Soco ofChem. Engrs., Okayama, Japan, GS-207, pp. 439-440.
Nomura, M., Yamamoto, K., Harada, M., and Eguchi, W. (1979). Preprints, Int. Conf Surface
Col/oid Sci., 3rd, Stockholm, Sweden, pp. 148-149.
Nomura, M., Minamino, Y., Fujita, K., and Harada, M. J. Polym. Sci., (in press).
Patsiga, R., Litt, M., and Stannett, V. (1960). J. Phys. Chem. 64, 801.
Ranz, W. D., and Marshall, W. R. (1952). J. Chem. Eng. Prog. 4S, 141.
Smith, W. V., and Ewart, R. H., (1948). J. Chem. Phys. 16, 592.
Stockmayer, W. H. (1957). J. Polym. Sci. 24, 313.
O'Too1e, J. T. (1965). J, Polym. Sci.9, 1291.
Ugelstad, J. (1981). Pure Appl. Chem. 53, 323-363.
Ugelstad, J., and M0rk, P. C. (1970). Dr. Polym. J. 2, 31.
Uge1stad, J., and Hansen, F. K. (1976). Rubber Chem. Technol.49, 536.
Uge1stad, J., M0rk, P. c., and Aasen, J. O. (1967). J. Polym, Sci. Part A-l 5, 2281.
Uge1stad, J., M0rk, P. C., Dahl, P., and Rangnes, P. (1969). J. Polym. Sci. Part C 27,49.
Van der HotT, B. M. E. (1956). J. Phys. Chem. 60, 1250.
Wang, B. C. Y., Lichti, G., Gilbert, R. G., and Napper, D. H. (1980). J. Polym. Sci. Po/ym.
Letter Edi. IS, 711.
Wilke, C. R., and Chang, P. C. (1955). AIChE J. 1,264.
Zollars, R. L. (1979). J. Polym. Sci. 24, 1353.
6
Effects of the Choice
of Emulsifier in
Emulsion Polymerization
A. S. Dunn
1. Introduction .
11. Monomer Emulsification
A. Hydrophile-LipophileBalance
B. Effects of Mixtures of Anionic
and Nonionic Emulsifiers .
C. Preparation of Monodisperse Latexes
111. Emulsion Polymerizationwith Nonionic Emulsifiers
IV. Emulsion Polymerizationwith lonic Emulsifiers.
A. Effectof Emulsifieron Numberof Latex
Particles Formed
B. The Effectof Micelle Size .
V. Latex Agglomeration
VI. Other Effects of Emulsifiers
A. Monomer Solubilization
B. Polymer Solubilization .
C. Initiator Decomposition Induced by Emulsifier.
D. Catalysis of the Initiation Reaction .
E. Transferto Emulsifiers .
References
221
224
225
227
228
229
230
233
234
236
237
237
238
239
241
242
243
l. Introduction
Despite the name, neither the emulsification of monomer nor the
stabilization of polymer particles formed is an essential characteristic of
emulsion polymerization. The essentlal feature of emulsion polymerization
is that polymerization occurs in a large number of isolated particles that
normally contain no more than a single polymerizing radical: this permits
high molecular weight polymer to be formed at a high rate of polymerization, by contrast with bulk or solution polymerization in which an
221
EMULSION POLYMERIZATION
Copyright It 1982 by Aeademie Pres,. Ine.
AJI righlS of reproduetion in any fonn reserved.
ISBN 0-12-556420-1
222
A. S. Dunn
6.
223
A. S. Dunn
224
4
2.5
5.0
102SJ/mol
7.5
10.0
dm-3
Fig. 1.
Dependence of the rate of polymerization
(in Interval 11) of styrene emulsions on
the concentration
of the emulsifier (sodium dodecyl sulfate) at 60C. Initiator: 0.2% K2S20S
on aqueous phase. (From AI-Shahib, 1977.)
6.
225
droplets can compete effectively for initiator radical s and become a significant or even dominant locus of polymerization. It may occasionally be
desirable to have a monomer emulsion sufficiently stable to permit stirring
to be dispensed with (e.g., this would facilitate the use of the dilatometric
method for determining the rate of an emulsion polymerization), but the
difficulty here is that an emulsion that is stable so long as it is not
polymerized will break as polymerization proceeds because of depletion of
emulsifier from the monomer droplets by adsorption on the latex paitic1es.
This might be prevented by the use of sufficiently large concentrations of
emulsifier .
A. Hydl'ophile-LipophileBalance
Hydrophile-lipophile
balance
(HLB)
to the
selection of a suitable emulsifier system for a given disperse phase particulady when nonionic emulsifiers are to be used. Several studies of the effect
of HLB in emulsion polymerization
HLB number
in water
Not dispersible
Poorly dispersible
Unstable milky dispersion
Stable milky dispersion
Translueent solution
{!6
Clear solution
{
a After Adamson,
Applieation
8
10
12
14
16
18
1976. Reprodueed
Detergents
Solubilizers
with permission
Emulsions
226
A. S. Dunn
TABLE
11
groups (constant)
-S04Na
(38.7)
-COOK
-COONa
-S03Na
-COOH
-OH
-0-{CH2CH20)-
(21.1)
(19.1)
(1l.0)
(2.1)
(1.9)
(1.3)
of HLBa Valuesb
Lipophilic
-CH2-CH=
-CH3
-{CH2CH2CH20)-
groups (constant)
- Dlipophilic
with permission
(0.475)
(0.15)
group
of J.. T. Davies.
6.
227
styrene, and ethyl acrylate and styrene. They found the HLB to affect
reaction rate, latex stability, viscosity, and partiele size. Optimum latex
stability was obtained in the range 12.1-13.7 for poly(methyl methacrylate),
11.8-12.4 for poly(ethyl acrylate), 13.3-13.7 for polyacrylonitrile, and 11.9513.05 for the copolymer obtained from a 50% (by weight) mixture of methyl
methacrylate and ethyl acrylate. Eliassaf (1973) found a correlation between
the HLB of the suspension stabilizer and the morphology of PVC
[poly(vinyl chloride)J partieles prepared by suspension polymerization,
which varied from glassy beads to porous partieles having a high plasticizer
absorption.
However, as pointed out by Rosen (1978) the HLB method is useful only
as a rough guide to emulsifier selection since it indicates neither the
efficiency of the emulsificr (i.e., the concentration required) or its effectiveness (i.e., the stability of the emulsion produced) and does not take into
account the effect of temperature in varying the extent of hydration of
hydrophilic groups such as polyoxyethylene, with consequent change in the
action of the surfactant.
% anionic
emulsifier, although
A. S. Dunn
228
the micellar weight of the anionic emulsifier alone was much larger (23,500):
thus, there is a much larger number of much smaller micelles in the
mixed system than is obtained with either component alone. This was
reflected in an increase in the number of latex particIes formed and in the
subsequent steady rate of polymerization and in a reduction of the breadth of
the particIe size distribution. The highest rate and narrowest particIe size
distribution was obtained with 17 mol % of the ionic emulsifier when the
average micellar weight was 6,000: it was found that the probability of a
micelle nucIeating a latex particIe was a maximum at this composition
aIthough the probability remained low as an absolute value (0.00038).
C.
In the absence of particIe coalescence (i.e., when relativeiy large concentrations of emulsifiers are present sufficient to stabilize all primary
particIe nucIei formed) the essential condition for the' formation of a latex
with a very narrow particIe size.distribution is that the nucIei of all particIes
should be formed at very neariy the same time and should subsequently
grow at equal rates. The use of mixtures of anionic and nonionic emulsifiers
pro vides the requisite conditions as first shown by Woods et al. (1968).
AIthough initiation of emulsion polymerization with a water-soluble initiator has hitherto been credited with a 100% efficiency, the efficiency of
initiation of styrene by persulfate in seeded polymerizations of styrene in
which sodium dodecyl sulfate was used as the emulsifier has recently been
shown by Hawkett et al. (1980) to be rather low. Using an -10% emulsion
with 0.42% sodium dodecyl sulfate on the water phase, they find at 50C an
initiator efficiency of 44% at low persulfate concentrations (1.6 x 10-5
mol dm - 3) falling to only 1% at high concentrations (8.33 x 10- 2
mol dm - 3). Consequently, any improvement in initiator efficiency
should increase the rate of particIe nucIeation. Evidently, the increased
surface area and reduced surface charge density of mixed micelles achieves
this. The seed latex used by Hawkett et al. was monodisperse: it was prepared at a high temperature (90C) using sodium dioctylsulfosuccinate
(Aerosol MA, Cyanamid) as emulsifier with persulfate initiation. The high
temperature would give a high rate of radical formation from the initiator
but this emulsifier, having two alkyl chains, presumably forms micelles
with a lower surface charge density than emulsifiers with single alkyl
chains.
It has often been pointed out that the approach of a charged radical to a
charged micelle requires a Coulombic repulsive energy barrier to be overcome even though counterion binding reduces the effective charge on the
6.
229
micelle below that which would be inferred from the aggregation number of
the micelle. Consequently, it has been surmised that a sulfate radical might
hav.e to add several monomers in an aqueous-phase polymerization before
the van der Waals attractive force between the growing radical and the
micelle suffices to overcome the Coulombic repulsion. Some evidence in
support of this is provided by the observation (Waite, 1977) that monomers
even less soluble in water than styrene (e.g., tert-butylstyrene, dodecyl
methacrylate) cannot be polymerized in an emulsion with a water-soluble
initiator unless the monomer solubility in the aqueous phase is increased by
addition of alcohol or acetone. On the other hand the only actual estimate
of the magnitude of the energy barrier (Fitch and Shih, 1975) shows it to be
negligible in the case of latex particles stabilized by ionic initiator end
groups only. Exchange of individual emulsifier molecules between micelle
and solution is rapid but a minimum Cs alkyl chain is needed for micelles
to be formed at all by n-alkyl sulfates or carboxylates, from which one
might infer that a minimum of four monomers must be added before a
sulfate radical can enter a micelle or a latex particle.
The first published formulation for producing a mono disperse polystyrene seed latex (Bobalek and Williams, 1966) involved the use of octylphenoxypolyethylene
ill.
A. S. Dunn
230
IV.
6.
231
emulsifiers are used alone or when mixtures of ionic and nonionic emulsifiers are used. The parameter characteristic of the emulsifier in SmithEwart theory is as, the area occupied by an emulsifier "in a saturated
monolayer at the polymer-water interface. The criterio n for the transition
from Interval 1, in which the latex particles are formed, to Interval 11, in
which the particles increase in size at a uniform rate while their number
remains constant, is that the total surface area of the latex particles should
ha ve increased to an extent sufficient to adsorb all the emulsifier originally
present in the aqueous phase as micelles. It is assumed that the area
occupied by an emulsifier molecule on the latex particle surface and in the
micelle is the same. Since the presence of micelles containing solubilized
monomer is postulated as essential to the nucleation of latex particles, it
follows that it should not be possible to nucleate additional latex particles
once all emulsifier has been adsorbed on the existing particles. The reason
that emulsifier should adsorb on the latex particles rather than aggregate in
micelles in solution is not explained, although it is an experimental fact that
the micelles do usually disappear from solution (as shown by an increase in
the surface tension) at approximately the same time as new latex particles
cease to formo Actually, it is probable that the van der Waals interaction
energy between an emulsifier molecule and a latex particle is greater than
that between an emulsifier molecule and a micelle simply because the latex
particle is larger, but a quantitative demonstration of this has not yet been
gIven.
Despite the fact that Smith-Ewart theory (or Gardon's development of
it) does apply well to the emulsion polymeri?:ation of styrene at least, none
of these postulates seems to be strictly true. Micelles are not essential for
the nucleation of la'tex particles since the possibility of emulsifier-free
emulsion polymerization has been amply demonstrated. When high concentrations of a weakly adsorbed emulsifier are used, micelles do not
disappear at the beginning of Interval 11 and even remain at the end of the
reaction (Bakker, 1952), it being impossible to carry out a soap titration on
the latex because the surface tension has not risen above the value
characteristic of the CMC of the emulsifier (AI-Shahib, 1977). The value of
as is not independent of the nature of the polymer as was initially assumed
[and apparently confirmed by early experiments (Sawyer and Rehfeld,
1963)] but increases with the polarity of the polymer and the temperature
(Piirma and Chen, 1980) (Table III). Indeed if as were independent of temperature the van't Hoff Isochore would imply a zero enthalpy of adsorption
in contradiction to the general observation that physical adsorption is an
exothermic process. The real criterion for the cessation of micellar particle
nucleation appears to be not the absence of micelles but the fact that the
number and size of the latex particles has increased sufficiently to ensure
c~pture of all radicals generated in the aqueous phase in competition with
any micelles that may remain. Hence, as is probably not the parameter
A. S. Dunn
232
TABlE
111
Area (a's) Occupied by a Surfactant Molecule at a PolystyreneWater Interface at the Critical Micelle Concentration: Effect of
Alkyl Chain length
(a/A2)
Potassium carboxylates
Chain length
C10
C12
C14
C16
C18
68
66
48
41
34
25
23
52
40
25
EffeCI of temperature
Sodium dodecyl,sulfate
TemperatureOC
a~A2
Q
22
47
37
49
47
52
On butadiene-styrene rubber.
6.
233
TABLEIV
Adsorption of Soaps on PVCa.b
Soap
Potassium
Potassium
Potassium
Potassium
dodecanoate
tetradecanoate
tetradecanoate
tetradecanoate
a;jA2
aslA2
2.7KCI
O
1.7 K2HP04
5.7 K2HP04
51
52.9
47
40
26.5
24.4
25
25
as by extrapolation
of a linear plot of
According to the theories of Smith and Ewart and of Gardon, the effect
of emulsifiers on the number of polymer latex particles formed N (and hence
on the rate of polym~rization during Interval 11)is determined by the area
as which the emulsifier molecule occupies in a saturated monolayer at the
polymer-water interface,
N oc (asS)3/5
A. S. Dunn
234
Considerable discrepancies exist between the results of different techniques for determining micelle size (Anacker el al., 1964); but it has been
shown (Kratohvil, 1980) that serious discrepancies can arise from oversimplification of the treatment of experimental results and that the results of
6.
235
ea
e 10
'
e 12
::./
O
e 16
/!
~ 40
e 14
LL l
600
600
Time(mio)
'
600
e la
,l
600
600
60
Fig. 2. Course of the emulsion polymerization of styrene at 60C using the same
concentrations (0.012 mol dm - 3)above their critical micelle concentrations of the homologous
series of sodium alkyl sulfates. lnitiator: 0.20% KzSzOs on aqueous phase. (From AI-Shahib,
1977.)
different techniques can be reconciled when proper allowance for the effects
of counterion binding is made, although unfortunately the only emulsifier
for which sufficient data exists to allow this to be done is sodium dodecyl
sulfate.
However, although the absolute values of the aggregation numbers may
be incorrect, there is no doubt (Table V) that micelle size increases with
alkyl chain length. Consequently, at constant micellar concentration the
number of micelles must decrease as their size increases. Clearly the number
of micelles initially present do es not determine the number of latex particles
formed. However, at a given micellar concent~ation, the total surface area of
the micelles is constant because the increasing size of the micelle compensates for their decreasing number as alkyl chain length is increased.
Thus, it seems that the total surface area of the micelles may be theofactor
determining the rate at which initiator radicals are captured by micelles and
hence the number of latex particles formed.
TABLEV
Total Surface
Aggregation numbera
10 - 3 Total surface area m z dm - 3 b
a
b
27
4.4
CIO
CIZ
CI4
CI6
41
4.5
64
4.3
80
4.4
100
4.5
CIS
A. S. Dunn
236
.
V. Latex Agglomeration
The size of the latex particles obtained by an emulsion polymerization
using an ionic emulsifier is in the 3O-100-nm range. Particles exceeding 100
nm in diameter can be produced by multistage seeded emulsion polymerizations: and although this technique has been used in the laboratory to
obtain model colloids it is not suitable for use on an industrial scale,
although latexes with an average diameter in excess of 100 nm are required
for some applications. The viscosity of dispersions increases with decreasing
average particle size although it also depends on particle size distribution.
The solids content of a latex produced by the mutual recipe using 100 parts
by weight of monomer to 180 parts water carried to a 70% conversion is
28% after the residual monomer has been removed. This is much too low for
many applications in which the product is used as a latex: solids contents in
excess of 60% are often required. However, if the latex is concentrated by
evaporation it sets to a stiff paste when the solids concentration rises to
about 40%. This can be avoided if the average particle size is increased by
agglomeration. The principIe of agglomeration is that the latex is destabilized slightly when particle coalescence occurs until the total surface
area of the particles has been reduced sufficiently for stability to be restored.
This may be brought about by the addition of a solvent that swells the
particles and which has to be removed after agglomeration, by partial
neutralization of carboxylate emulsifiers, by addition of electrolytes, or by
freezing which increases the effective electrolyte concentration when some of
the water separates as ice. Unfortunately, all these processes are more or
less difficult to control and generally re'Sult in partial coagulation of the
latex as well as agglomeration. The process of pressure agglomeration
6.
237
VI.
Monomer So/ubilization
A. S. Dunn
238
P,'I
~'" 1.2
I
(a)
""
..
~O.B
;Q
"
~
00.4
<
~
o
8
16
24
Ofopolymerizedin 3 hr
P,'I-;241(b)
""
32
'"
:o..
.1:1
~"
16
'O
..
~
8
..
..
>.
~
O
O
0.4
0.8
1.2
DMAB solubilized(gdm'3)
Po/ymer So/uhi/ization
6.
239
polymerizations this effect could mean that the lowest molecular weight
fractions of polymer could remain in the aqueous phase at high surfactant
concentrations.
C.
+ RH =
R.
+ S20~- = RSOi
+ HSOi
+ SOi
A. S. Dunn
240
additives as was implied by the results reported by Morris and Parts (1968):
their results refer to rates of persulfate decomposition but they seem to ha ve
been vitiated by interference by the additives in the titrimetric method used
for determining persulfate concentrations.
Yurzhenko and Brazhnikova (1956) reported that hexane or toluene,
dispersed in water without an emulsifier, increased the rate of decomposition of persulfate by a factor of two or three. van der Hoff (1967)
showed that the rate of decomposition of persulfate was much increased in
dispersions of octane or benzene. Vinogrado v et al. (1962) found that
whereas persulfate decomposed at a much enhanced rate in an emulsion of
xylene with potassium oleate as emulsifier compared with a similar emulsion using the same concentration by weight of sodium dibutylnaphthalene sulfonate, the rate of polymerization of styrene emulsions was
actuaIly greater with the latter emulsifier! Volkov and Kulyuda (1978)
found nOl1ionicsurfactants to enhance the rate of persulfate decomposition:
the effect of a polyethyleneoxyalkylphenol was grea~er than that of a
polyethyleneoxymonododecyl maleate, which exceeded that of poly(vinyl
alcohol), which in turn was greattr than that of polyethyleneoxymonotetradecyl maleate; but the rate of polymerization of styrene was greater with the
second of the above emulsifiersthan with the first. Ryabova et al. (1977)
found that persulfate decomposed 50 to 100% faster in potassium decano ate
solutions than in water, whether the concentration was above or below the
CMC of the emulsifier: emulsification of styrene resulted in a further 100%
increase in the decomposition rateo Liegeois (1971) found that the rate of
initiation of vinyl chloride in aqueous solution by persulfate was that
expected from the rate of decomposition in the absence of additives,
although the rate of decomposition was increased by a factor of three in the
presence of vinyl chloride. Similariy, AIIen (1958) found that the rate of
initiation of methyl methacrylate emulsions by persulfate (as inferred from
measurements of overaIl polymerization rate and the degree of polymerization of the polymer) did not vary as the concentration of sodium dodecyl
sulfate emulsifier was increased. The general conc1usion is that the presence
of emulsifiers and emulsified monomers does not increase the rate of
initiation by persulfate but that the induced chain decomposition may
increase the rate of consumption of the initiator by a factor of three or
more. Thus, the effect will usually be negligible in laboratory studies in
which conditions are usually chosen so that the decrease in the concentration of the initiator will be negligible in the duration of the polymerization reaction, but under industrial conditions wasteful consumption of
persulfate could lead to premature exhaustion of the initiator and incom-
pletepolymerization.
6.
D.
241
53 kJ mol- 1, much lower than any value determined in bulk. Hui and
Hamielec's value of 115 kJ mol-I (1972) seems to be representative of the
data available on th(;:rmalinitiation in bulk. The concIusions of Asahara et
al. are based on observations of the temperature dependence of the degree
of polymerization and are open to several objections.
1. The viscosity-average degree of polymerization is measured when the
number.oaverage degree of polymerization is required, although this will not
lead to any error in deriving an energy of activation provided the molecular
weight distribution does not change with temperature.
2. It is assumed that all transfer reactins can be neglected so that the
kinetic chain length can. be taken as half the degree of polymerization: this
seems dubious when it is generally agreed that the degree of polymerization
of thermally polymerized styrene in bulk is limited by transfer to monomer
and when the possibility of using oil-phase initiation for emulsion polymerization is dependent on the escape of one of a pair of.radicals generated in
the oil phase to the aqueous phase which seems likely to involve transfer
either to monomer or to emulsifier.
3. In their calculation, they seem to have confused the energy of
activation for propagation with the overall energy of ~ctivation for polymerization: when this is corrected the energy of activation for initiation
becomes 80 kJ mol- 1 which is still significantly lower than the bulk value of
115 kJ mol- 1 but which may stiII be incorrect if the second objection above
is justified.
An alternative method of calculating the energy of activation for initiation in emulsion polymeriza60n was introduced by Bartholom et al.
A. S. Dunn
242
emulsifier.
'
E. Transfer lo Emu/sifiers
The extent and importance of transfer reactions to emulsifier molecules
in emulsion polymerization has not been extensively investigated. Tr.ansfer
to an emulsifier molecule that is subsequentIy desorbed from the micelle or
latex partic1e would seem to be an attractive mechanism that would
facilitate the escape of radicals of the less water-soluble monomers to the
aqueous phase, a process of greater importance than had previously been
thought. However, most emulsifiers do not contain any particularly active
hydrogen or other substituent atoms, and the topology of the system
implies that transfer constants determined from experiments in homogeneous solution do not necessarily apply to latex systems. Thus, Okamura
et al. (1958) found from experiments in homogeneous solution that the rate
constant for the transfer of poly(vinyl acetate) radicals to poly(vinyl alcohol)
(35 x 10-4 dm3 mol-l sec-l) was much higher than that to poly(vinyl
acetate) (1.5 x 10-4 dm3 mol-1 sec-1) so that it might be expected that
when incompletely hydrolyzed grades of poly(vinyl alcohol-acetate) are
used as emulsifiers in the emulsion polymerization of vinyl acetate, the
extent of formation of graft copolymers of poly(vinyl acetate) with the
emulsifier sho~ld be expected to decrease with acetyl content. In fact, the
contrary trend was observed. The acetyl blocks form the hydrophobic
portion of poly(vinyl alcohol-acetate) emulsifiers and are the only parts of
the molecule readily accessible to poly(vinyl acetate) radicals in the latex
6.
243
References
Adamson, A. W., (1976). "Physics and Chernistry ofSurfaces," 3rd ed. Wiley, New York.
Ahmed, S. M., EI-Aasser, M. S., Micale, F. J., Poehlein, G. W., and Vanderhoff, J. W. (1980). In
"Polyrner Colloids 11" (R. M. Fitch, ed.), p. 265. Plenum Press, New York.
Allen,P. W. (1958).J. Polym. Sci. 31, 206.
AI-Shahib, A. W. (1977). Ph.D. Thesis, Manchester.
Anacker, E. W., Rush, R. M., and Johnson, J. S. (1964).J. Phys. Chem.68, 81.
Aniansson, E. A. G. et al. (1976). J. Phys. Chem.SO,905.
Asahara, T., Seno, M., Shiraishi, S., and Arita, Y. (1970). Bull. Chem. Soco Jpn. 43, 3895.
A. S. Dunn
244
Asahara, T., Seno, M., Shiraishi, S., and Arita, Y. (1973). Bul/. Chem. Soco Jpn. 46, 249.
Bakker, J. (1952). Phi/ips Res. Rep. 7, 344.
Bartholom, E., Gerrens, H., Herbeck, R., and Weitz, H. M. (1956). Z. E/ektrochem. 60, 334.
Benetta, G., and Cinque, G. (1965). Chim. Ind. (Mi/an) 47, 500.
Blackley, D. e. (1975). "Emulsion
Polymerisation:
Theory and Practice," p. 308, Fig. 7.22.
Applied Science, London.
Bobalek, E. G., and WiIliams, D. A. (1966). J. Po/ym. Sci. Part A-/ 4, 3065.
Carr, C. W., KoltholT, J. M., Meehan, E. J., and WiIliams, D. E. (1950). J. Po/ym. Sci. 5, 20\.
Davies, J. T. (1957). Proc. InI. Congr. Surface Activity, 2ndVol. 1, p. 426. Butterworths,
London.
Doherty, J. V. (1978). Unpublished
work.
Dunn, A. S., and AI-Shahib, W. A. (1978). J. Po/ym. Sci. Po/ym. Chem. Ed. 16,677.
Dunn, A. S., and AI-Shahib, W. A. (1980). In "Polymer Colloids 11", (R. M. Fitch, ed.), p. 619.
Plenum Press, New York.
Dunn,
13, 126\.
Symp. Ser.ll,
Washington
Fiteh, R. M., and Shih, L.-B. (1975). Prog. Col/oid Po/ym. Sci. 56, \.
Fitch, R. M., and Tsai, C. H. (1971), In "Polymer
Press, New York.
Colloids"
6.
245
Onischenko, T. A., Rabinovich, M. A., and Gershenovich, A. 1. (1970). Kol/od Zh. 32, 886
[English transl.: Colloid J. USSR 32,744.
Ono, H., Jidai. E.. and Fujii, A. (1974). J. Col/oid Interface Sci. 49, 155.
Ono. H.. Jidai, E.. and Fujii. A. (1975). J. Phys. Chem. 79. 2020.
Ottewill, R. H., and Shaw, J. N. (1966). Discuss. Faraday Soco 42. 154.
Paxton. T. R. (1969). J. Col/oid Interface Sci. 31, 19.
Paxton. T. R. (1977) B. F. Goodrich Co.. Brecksville. private communication.
Piirma, l., and Chang, M. (1980). Org. Coatngs Plast. Preprints 43, 104.
Piirma, l., and Chen, S.-R. (1980). J. Col/od Interface Sci. 74, 90.
Piirma, l., and Wang, P.-e. (1976). In "Emulsion Polyrnerization" (1. Piirma and J. L. Gardon,
eds.), Am. Chem. Soe. Symp. Series 24, p. 34. American Chemical Society, Washington D.C.
Rosen, M. J. (1978). "Surfactants and Interfacial Phenomenena," Wiley, New Y.ork.
Ryabova, M. S., Sautin, S. N., Beresnev, V. N., and Smirnov, N. 1. (l975a). Zh. Prikl. Khim. 48,
1101 [Enf{li.fh transl.: J. Appl. Chem. USSR 48, 1148].
Ryabova. M. S.. Sautin, S. N.. and Smirnov. N. 1. (l975b). Zh. Prikl. Khim. 48, 1577 [English
transl.: J. Appl. Chem. USSR 48, 1632).
Ryabova. M. S.. Sautin. S. N.. and Smirnov. N. 1. (1977). Zh. Prikl. Khim. SO. 1719 [English
transl.: J. Appl. Chem. USSR SO, 1648].
Sata, N., and Saito, S. (1952). Kol/oid Z. 128, 151.
Sawyer, W. M., and Rehfeld, S. J. (1963). J. Phys. Chem. 67, 1973.
Testa. F., and Vianello, G. (1969). J. Polym. Sci. Part C 27,69.
Ugelstad, J., and Hansen, F. K. (1979). J. Polym. Sei. Polym. Chem. Ed. 17, 3069.
Ugelstad, J., EI-Aasser, M. S., and VanderhotT, J. W. (1973). J. Polym. Sci. Polym. Lett. Ed. U,
503.
122120t
(1979).
Woods, M. E., Dodge, J. S., Krieger, 1. M., and Pierce, P. E. (1968). J. Paint Teehnol. 40, 543.
Waite, F. A. (1977). 1.e.1. Ltd., Paints Division, Slough, private cornmunication.
Yurzhenko, A. l.. and Brazhnikova, O. P. (1956). Zh. Obs. Khim. 26,1311 [English transl.: J. Gen.
Chem. USSR 26, 1481].
7
Polymerization
of Polar Monomers
v. l. Yeliseyeva
1. Introduction.
11. Interface Characteristics of Polymeric Dispersions
111. Relationship between Emulsifier Adsorption and
the Difference in the Boundary Phase Polarity
A. Monomer-Aqueous Phase Interface
B. Aqueous Phase-Polymer and Aqueous PhaseLatex Particles Interfaces.
IV. Mechanism of Particle Generation .
V. Colloidal Behavior of Polymerization Systems.
VI. Kinetics of Emulsifier Adsorption
VII. Mechanism of Formation and Structure of Particles
VIII. Polymerization Kinetics
IX. Relationship between Polymerization Kinetics
and Adsorption Characteristic of Interface.
X. Nor,nenclature .
References .
l.
247
249
250
251
254
257
261
268
270
278
283
286
287
Introduction*
Experience has shown that each monomer has its own specific
features in emulsion polymerization and requires a specific technology. To a
certain extent this is also true of other components in the reaction system,
e.g., the emulsifier and the initiator. Attempts to fit various emulsion
reaction systems into one physicochemical model or theoretical scheme
have therefore failed.
Nevertheless, several features that are common to all polymerizations in
colloidal systems have been established. They were formulated in the
* There is a Nomenclature
chapter.
247
EMULSION POLYMERIZATION
Copyright <t> 1982 by Academic Press. Ioc.
Al! rigbts or reproductioo io aoy rorm reserved.
ISBN 0-12-556420-1
248
V. 1. Yeliseyeva
lassical theory of Smith and Ewart (1948) which was created in the late
1940s and which is weIl known to researchers working in the field of
emulsion polymerization. The most important concIusion that foIlows from
the theory is that the process takes place in an equilibrium of swoIlen
polymer-monomer particIes (PMP) each of which may be "living" or
"dead" at any given time depending on the presence of a free radical.
Furthermore, Case 2 of the theory was developed for systems in which
styrene was used as the weakly polar model monomer. An anion-active,
miceIle-forming surfactant with a low CMC (critical miceIle concentration)
value (such as sodium dodecylsulfate) was used as the model emulsifier and
persulfate as the model water-soluble initiator. The process is characterized
by a period of constant rate, the rate being proportional to the number of
particIes and the average number of radical s per particIe and depending on
monomer reactivity. According to the theory based on the Harkins' (1947)
model oi emulsion polymerization [a similar model was independentIy
developed by Yurzhenko and Mints (1945)], PM:P are formed from
monomer-swoIlen emulsifier miceIles into which free radicals enter from the
aqueous phase; the number bf particIes remains constant during the
constant rate periodo
The derived equation of the rate dependence on the emulsifier (Ce) and
initiator (C)concentrations during the constant rate period
(1)
249
polarity of the organic phase. Hence, the use of the emulsifier concentration
value in the rate equation without taking into account its activity .is
inadequate. Therefore, in order to extend the applicability of Eqs. (1) and (2)
to the polymerization of monomers of various polarity in the presence of
emulsifiers of various structure, a parameter characterizing the emulsifier
activity in the given system should be introduced. Whereas this parameter
may be neglected when considering polymerization of some hydrophobic
monomers (hydrocarbons) with close adsorption characteristics, it must be
taken into account in the case of polymerization of highly polar monomers
(or their copolymerization with hydrophobic monomers) where the colloidal behavior of the system, and consequently the polymerization kinetics,
depends on emulsifier activity. Since the practical importance of latexes
obtained by polymerization of polar monomers or monomeric systems is
rapidly increasing, a consideration of the regularities of their formation,
taking into account a11important factors, is of considerable interest.
D.
(3)
250
V. 1. Yeliseyeva
tensions Y12 under the assumption that only dispersion interactions are
operative at the interface. For all the compounds studied Y12was found to
be higher than Y12' The difference between them may serve as a measure of
the energy, which includes various polar interactions. Fowkes established
that for most of the cases examined, interaction attributable to dispersion
forces makes the highest contribution to interfacial tension, although for
substances capable of forming hydrogen bonds the attraction due to
dispersion forces may be approximately equal to attraction due to hydrogen
bonds.
Wu (1974) obtained an equation for boundary phases with various
properties which relates Y12not only to dispersion interaction but also to
phase polarity. In his opinion, for an interface between high and low energy
phases, e.g., mercury or water and organic polymers, this dependence may
be represented by the geometric-harmonic mean ratio
Y12 = Yl + Y2 -
(4)
251
In order to analyse the colloidal processes occurring in colloidal polymerization systems containing monomer, polymer monomer and polymer
particles it is important to know about adsorption characteristics of the
emulsifier on the monomer-aqueous phase and polymer-aqueous phase
interfaces.
A.
252
V. 1. Yeliseyeva
0.1
0.2
0.3
Concmtration of n-lJutyrcacid
(mol/tim!)
Fig.1. Adsorption isotherms of'n-butyric acid at the interfaces: (1) water-air; (2) waterhexane; (3) water-benzene; (4) water-olive oil. T, 318 K.
various polarity (~G). It was established that this value decreases with
increasing polarity of the organic liquido For nonpolar liquids such as
cycJohexane, hexane, and heptane, this value is 3.6 kJ/mol, whereas for
polar liquids (alcohol s, nitro-compounds, ethers) it decreases to 1.892.98 kJ/mol.
.
Using the stalagmometric method, we determined the adsorption energy
(E'ds) of sodium dodecylsulfate (SDDS) at the water boundary with the
homologous series of alkyl acrylates at 293 K (Yeliseyevaet al., 1978).
Calculations were performed from the initial section of the surface tension
isotherm using the Langmuir equation. Table 1 lists values of the initial
interfacial tension Y12;the overall free energy of adsorption, E'ds>and the
TABLEI
Energy Characteristics of SDDS Adsorption on WaterAlkyl Acrylate Interface Depending on Alkyl Acrylate
Polarity
YI2
E.d'
Acrylate
(mJ/m2)
(kJ/mol)
t:.G
(kJ/mol)
Methyl
Ethyl
Butyl
8
13.7
23.2
20.0
22.1
25.6
1.69
1.85
2.10
253
free energy of adsorption LlG per methyl group of the alkyl (assuming that
the entire alkyl portion of the emulsifier is immersed in the monomer
phase). Data presented in Table I indicate a decrease in the energy of
interaction between the hydrocarbon portion of the emulsifier and the
organic phase (alkyl acrylate) with an increase in the polarity of the latter.
Accordingly, the overall adsorption energy also decreases. Several important practical conc1usions may be drawn from these data. On the one hand,
dispersions of polar monomers are stabilized to a lesser extent than
nonpolar monomers by emulsifiers containing an aliphatic hydrocarbon in
the oleophilic part of the molecule. Although the overall adsorption energy
increases only slightly with increasing length of the hydrocarbon portion,
as follows from data presented in Table 11,the solubility of the emulsifier in
water is significantly reduced, which may hinder its practical use. Therefore,
on the other hand, in order to increase the energy of interaction of the
oleophilic part of the emulsifier molecule with the polar organic phase,
polar groups should be introduced into its structure. This has been confirmed in practice.
A consideration of the adsorption kinetics is very important in an
estimation of the effectiveness of surfactants under the dynamic conditions
of emulsion polymerization. In a stalagmometric study of dynamic and
static adsorption of emulsifiers of various structure at the air-water interface, it was established that adsorption values of micelle-forming surfactants differ significantly in the period of drop formation (Nikitina et al.,
1961). This was explained by the consider~ble period needed for establishment of adsorption equilibrium connected with the kinetics of adsorption
layer formation. The authors conc1uded that for usual concentrations of
surfactant solutions the period of establishment of adsorption equilibrium
can be taken as equal to 2 mino Figure 2 shows the adsorption isotMrms of
TABLE 11
Alkyl su,lfate
C12H2SS04Na
C14H29S04Na
CI6H31S04Na
a The
E.d.
(kJjmol)
22.1
23.3
24.6
of
I:1G
(kJjmol)
1.85
1.69
1.55
254
V. 1. Yeliseyeva
-10
rmxa =3.ifx!O
moZ/cm2
2;5 5.0
Concfntration
Z5
10.0
of lIerosol (tOsmoZ/ dm J)
Fig. 2.
Adsorption isotherms of Aerosol AT plotted according to data of (1) static
(2 min) and (2) dynamic (2 sec) values of surface tension.
The general ideas developed for adsorption of surfactants on liquidliquid and liquid-air interfaces obviously cannot be completely transferred
to adsorption on interfaces with a solid phase.
Wolfram (1966) caIculated the packing density of surfactant molecules on
the polymer surface from the adsorption value at the CMC and showed
that it varies with polymer polarity. Thus, the adsorption area of a SDDS
molecule on various surfaces takes the following values (nm2): paraffin 0.41,
255
polyethylene 1.21, poly(methyl methacrylate) 1.32, poly(ethylene terephthalate) 1.42, and polycarbonate 1.48. On the basis of the results obtained
the author introduced the concept of "required area" of a surfactant
molecule and conc1uded that this area increases with increasing polarity of
the polymer. The established relationship he explains by the existence of
other interactions such as orientation, induction, and hydrogen bonds
between contiguous phases, in addition to dispersion forces. Because of
these interactions the molecules adsorbed on the solid phase-liquid interface may acquire a nonvertical orientation: they may be arranged with
greater planarity the more intensively the field (depending on polymer
polarity) acts upon the polar part of the surfactant molecule.
Several investigations have determined the absorption behavior of surfactant adsorption on partic1es of aqueous polymer dispersions by adsorption titration. The results have been similar to those observed by Wolfram
for adsorption on aplanar polymer surface determined from the wetting
angle. Thus, Paxton (1969) established that the area occupied by a sodium
dodecylbenzylsulfonate molecule in a saturated adsorption layer (AS1im)on
the surface of PMMA latex partic1es is 1.31 nm2, whereas on the surface of
polystyrene latex partic1es it is only 0.53 nm2. The author considers that
previous studies of adsorption of this emulsifier, which gave adsorption area
(ASlim)of 0.50 nm2, were carried out on interfaces with similar adsorption
characteristics. The intermolecular forces involved in adsorption between
water and air or water and hydrocarbon solvent or nonpolar polymer are
nearly the same. In his opinion the differences observed in surfactant
adsorption on polymers of different nature allows one to use the molecular
adsorption area value as a polarity characteristic of the polymer phase.
Dunn and Chong (1970) determined the value of ASlim for sodium
dodecylsulfate on poly(vinyl acetate) latex partic1es; it was fountl to be
1.24 nm2, i.e., 2.4 times higher than the minimal adsorption area of this
emulsifier. Zuikov and Vasilenko (1975) showed that the ASlim value for
sodium dodecylsulfate increases with polarity of the polymer and is inversely related to the interfacial tension at the interface of the corresponding
monomer with water. A correlation was established (Yeliseyeva and
Zuikov, 1976) between ASlimof a given emulsifier and its adsorption energy
at the corresponding monomer-water interface. The same correlation is
observed between ASlimof emulsifier at the given polymer-water interface
and its adsorption energy at the interface between the monomer and water
for the homologous series of alkyl sulfates (Table III).
It follows from the results obtained that the factors determining the
adsorption value of a given surfactant at the water interface with the
monomer and with the corresponding polymer are the same. Since with the
increase in the adsorption energy on the emulsifier-monomer interface, the
256
V. 1. Yeliseyeva
TABLE
111
Monomer
Emulsifier
MA
EA
EA
EA
BA
HA
Sty
C12H2SS04Na
C12H2SS04Na
C14H29S04Na
C16H31S04Na
C12H2SS04Na
C12H2SS04Na
C12H2SS04Na
aAt polymer-water
1'12
monomerwater
(mJjm2)
-8
13.7
13.7
13.7
23.2
27.2
33.8
E.d'
(kJjmol)
ASHm
(nm2)
20.0
1.51
0.92
0.82
0.74
0.67
0.52
0.48
22.1
23.3
24.6
25.6
-
interface.
7.
IV.
257
Numerous experimental data are available which suggest that both the
solubility of the monomer in water and its polarity affect the mechanism of
particle generation. Whereas during a conventional emulsion polymerization of styrene (or other hydrophobic monomer) examined by Harkins
(1947) and Yurzhenko and Mints (1945) the most important sites of particle
generation are monomer-swollen emulsifier micelles; with the increase of
monomer solubility in water particle generation in similar systems may be
shifted to the aqueous phase. Thus, the possibility of preparing concentrated and stable latexes by polymerization of methyl acrylate and ethyl
acrylate in the absence of an emulsifier was demonstrated by Yeliseyeva and
Zaides (1965) and by Yeliseye~a and Petrova (1970). It is also well known
that styrene latexes may be obtained by emulsifier-free polymerization; such
latexes are characterized, however, by larger particles and much lower
concentrations.
In all such cases the process initiated by persulfate begins in the aqueous
phase with the formation of water-soluble, surface-active polymeric radicals
which, after growing to a certain critical size, precipitate to form particles;
subsequent polymerization proceeds mainly within these particles. The
higher the solubility of the monomer in water, the more surface-aciive
radical s and therefore primary particles are formed and the higher the
stability and the concentration of the latex. A kinetic curve of the
emulsifier-free polymerization of ethyl acrylate confirming this scheme is
shown by Curve 1 in Fig. 3 (Yeliseyeva anq Petrova, 1970). The process
15
t(mini
25
Fig. 3.
Conversion kinetics (g/100 g of aqueous phase) of ethyl acrylate (1) in the
absence of emulsifier and (2) in the presence of 0.5 mol % of emulsifier E-30. Ammonium
persulfate, 0.33%; tert-dodecyl mercaptan, 0.33% in aqueous phase; T, 60C; phase ratio, 1: 3.
258
V. 1. Yeliseyeva
259
...
'.
2
1.1"~
260
V. 1. Yeliseyeva
250
200
~t::
150f4
4'
1J
0.7
100
1
2
J
'+
5
MonomerSolu.bilitgin water (%)
Monodispersity
coefficient
(K) and diameter(D)of latexparticlesformedby
Fig. 5.
polymerization
of monomers differing in solubility in water: (1) styrene; (2) butyl methacrylate;
(3) methyl methacrylate; (4) methyl acrylate. For conditions see legend to Fig. 4.
""
0.8
"-
0.0'+
0.05
I
0.05
/3
0.7
I
0.01
0.02
0.03
[s1)1)8] <%)
Fig. 6.
The effect of SDDS concentration (% in aqueous phase) on the monodispersity
coefficient of latex particles obtained by polymerization of (1) styrene; (2) methyl
methyacrylate; and (3) methyl acrylate. Ammonium persulfate, 0.05% in aqueous phase; T,
80C; phase ratio, 1: 10.
261
V.
262
V. 1. Yeliseyeva
50
40
--
s:::
"Jt
....
'->
-b.
s:::
1::
20
/.
4.-
""
Q.. 10
2
,
Fig. 7.
[RSOaNa](%)
concentration
in polymerization
of (1) methyl acrylate and (2) butyl methacrylate.
persulfate, 0.01% (1) and 0.1% (2) in aqueous phase; T, 60C; phase ratio, 1: 1.5.
Ammonium
order of magnitude higher for BMA polymerization. Since in both cases the
polymerization begins with the formation of very small particles, it may be
assumed that the flocculation of particles occurs during MA polymerization, even with excess emulsifier (run 4), and does not occur (or occurs on
a much smaller scale) during BMA polymerization. Morover, as can be seen
from Figs. 4 and 5, during the polymerization of 'polar monomers (MMA
TABLE IV
Colloidal Characteristics of MA and BMA Latexes Depending on Initial
Concentration of Sodium Alkylsulfonate in Aqueous Phase
No.
Emulsifier
concentration
in aqueous
phase (%)
Monomer
1
2
3
4
5
0.5
1.0
1.0
3.0
3.0
MA
MA
BMA
MA
BMA
y
(mJ/m2)
41.6
38.1
"
31.0 (-CMC)
-"
Partic1e
diameter
(nm)
Number of
partic1es
in 1 dm3
217
. 178
86
148
60
7.5 x 1016
1.4 xl017
1.25 x 1018
2.45 x 1017
3.7 X 1018
"We are unable to determine y precisely because of rapid film formation on the surface of
the p-BMA latex.
263
TABlE V
Colloidal Characteristics of Acrylate and Styrene Latexes Synthesized with SDDS
Monomer
Initial
SDDS conc
in aqueous
phase
(mol/dm3 x 103)
Latex
conc
(%)
Coagulum:
monomer
(%)
MA
EA
BA
St
8.68
17.36
26.64
26.04
20
18.7
20
19.1
0.20
0.11
0.00
0.00
Particle
diameter
(nm)
)'
(mJjmZ)
As
(nmZ)
Additional
adsorbed
amount of
SDDS
(moljg of
polymer x 103)
49.5
47.5
53.6
63.9
1.92
1.24
1.06
0.95
0.347
2.01
5.78
13.23
125
96
88
71
TAlE
Degree of
adsorption
saturation
ASlim
(nmZ)
( )
1.75
0.86
0.61
0.49
90.6
78.2
63.8
51.7
A:m
VI
Monomer
Solubility
in water
(%)
Latex
conc
(%)
Coagulum:
monomer
(%)
Viscosity
(cp)
)'
(mJjmZ)
MA
EA
BA
2-EHA
BMA
St
5.2
2.6
0.34
0.01
0.08
0.03
34.4
34.3
34.3
34.0
33.8
33.7
0.17
0.06
0.00
0.68
0.36
0.18
6.0
12.3
29.2
13.9
18.5
41.7
49.4
59.5
67.0
62.0
-
Mechanical
on a Marone-Ulevich
instrumento
Particle
diameter
(nm)
150
150
MO
97
120
100
Number of
particles
. in 1 dm3
(x 10-17)
1.66
1.67
2.19
6.17 .
3.58
6.10
Overall
surface
in 1 cm3
(m-Z)
As
(nmZ)
mechanical
(% formed
coagulum)a
eIectrolyte
(mi 0.1 M
CaClz)
11.7
11.8
13.5
18.5
16.1
19.1
1.05
1.12
1.31
1.87
1.51
1.80
2.1
1.6
O
0.2
1.0
0.5
0.75
0.95
1.25
0.15
0.05
0.02
7.
265
Yeliseyeva et al. (1975) investigated the effectiveness in emulsion polymerization of the alkyl acrylates, butyl methacrylate and styrene, in the
presence of another type of emulsifier, alkylac polyoxyethylenesulfate
(AAOES), obtained by exhaustive sulfonation of Triton X-100 (Rohm and
Haas, Philadelphia, Pennsylvania) with sulfamic acid and subsequent neutralization with ammonia. Polymerizations were carried out under identical
conditions by the batch method. Emulsifier concentration
in aIl runs was
26.4 x 10- 3 mol/dm3 in aqueous phase, ammonium persulfate concentration
1.1 x 10- 3 mol/dm3
for MA and EA polymerization
and
2.2 x 10-3 mol/dm3 for polymerization of the other monomers. AIl in dices
characterizing the latex stability have optimal values for the monomer with
medium solubility in water, namely for butyl acrylate (Table VI). However,
latexes based on more polar monomer are more stable when using this
emulsifir than latexes based on less polar, hydrophobic monomers. Surface
tension in the latter case is considerably higher, indicating a higher capacity
of the surface to adsorb the given emulsifier when formed during polymerization of less polar monomers. As in the case of alkylsulfonate (Table IV),
at the same initial concentration of the emulsifier a smaIler number of larger
particIes are formed in the case of polar monomers, where the surface is
better protected by the emulsifier. Thus, the area As occupied by the
emulsifier molecule on the surface of styrene latex particIes is 1.7-fold larger
than for the methyl acrylate latex, the diameter of the polystyrene latex
particIes being smaIler.
The l0l" stability to electrolytes of latexes based on hydrophobic
monomers obtained in the presence of AAOES play therefore be explained by
the rarefied adsorption layer on the particIe surface. Conversely, the
enhanced stability to' electrolytes of latexes of polar monomers may be
attributed to the denser packing of emulsifier molecules in the adsorption
layer at a given emulsifier concentration.
The somewhat higher mechanical
stability of latexes based on hydrophobic monomers may be associated with
their higher stability to orthokinetic fIocculation which is attributable to the
smaIler size of particIes, which to a certain extent compensates for the fact
that their surface is les s protected by the emulsifier.
Similar data on the adsorption saturation of acrylate latexes (based on
EA, BA, and 2-EHA) synthesized with the same emulsifier were obtained
recentIy by Snuparek and Tutalkova (1978). A lesser degree of filling of the
adsorption layer, 46.5 and 49.6%, was observed for more hydrophobic
monomers (2-EHA and BA): with a greater degree being observed in the
polymerization of the more hydrophilic EA, 66.3%.
Summarizing the described results, it may be concIuded that when using
the emulsifiers described above, the coIloidal behavior of polymerization
systems is mainly determined by monomer polarity. Latexes of polar
266
V. 1. Yeliseyeva
kp
7.
267
..
V. 1. Yeliseyeva
268
20
un
I (8)
..
.
....
'1::-,0
"':
20
40
20
40
COT!1rersion(%)
Fig. 8.
100;
7.
269
30
(S)
90
30
90
t (min)
Fig. 9.
Conversion (%) kinetics during polyrnerization of (1) rnethyl acrylate; (2) ethyl
acrylate; and (3) butyl acrylate in the presence of SDDS (A) and AAOES (B). For recipe see
Fig.8.
270
VII.
V. 1. Yeliseyeva
J
12
o
4-
20
40
60
Con1rerson
(Dio)
.
-
80
Fig. 10. Variation of the number of particles during polymerization of ethyl acrylate in
the presence of emulsifiers: (1) sodium alkylsulfonate;
(2) oxyethylated (30) cetyl alcohol; and
(3) AAOES. Recipe m.p.: ethyl acrylate, 100; emulsifier, 4.0; ammonium persulfate, 0.5; water,
900. T, 60C.
271
assessed from electro n micrographs of ethyl acrylate latex at 38.0% conversion (Fig. 11A) (before preparation the latexes were irradiated by a 60Co
source, at 20 Mrad, in order to avoid particle deformation). A similar
pattern is observed during the polymerization of a mixture of MMA and
BA (55:45) at 60% conversion (Fig. llB). The complex structure ofparticles
of carboxyl-containing acrylate latex MMA-BA-MAA (55:40:5) may be
revealed by increasing the pH above 9 (Fig. llC). The desaggregation
which occurs is attributable to the electro static repulsion of dissociated
carboxyl groups present in the macromolecules of primary particles.
We were able to reveal the structure of latex particles and films of polar
polymers (acrylate latexes) using a special technique for preparing samples
Fig. 11.
Electron rnicrophotographs
of latex particles: (A) ethyl acrylate at 38 %
conversion; (B) rnethyl rnethacrylate-butyl
acrylate copolyrner (60:40) at 60% conversion; (C) rnethyl rnethacrylate-butyl acrylate-rnethacrylic acid terpolyrner (55:40:5) at
pH 9 and 100% conversion.
272
V. 1. Yeliseyeva
Fig. 12.
Electron micrographs of polystyrene latex particles (A) before etching and (B)
after oxygen etching.
for electron microscopy (Yeliseyeva et al., 1967, 1968). The method of lowtemperature etching with active oxygen formed in an electrodeless highfrequency discharge was used. Under such conditions the less ordered,
looser regions are" burned out" initially, revealing the denser morphological regions (Jacobie, 1960). The latexes were placed on quartz supports, .
subjected to oxygen etching and shaded with platinum at a 20 angle. The
structure of films was studied by the method of single-step carbon replicas
shaded with platinum. Optical periods of etching were determined empirically. The effectiveness of this procedure may be assessed by comparing
electro n micrographs of polystyrene latex particles which were subjected
(Fig. 12B) and not subjected (12A) to oxygen etching. As can be seen, these
particles, contrary to expectation, are not coils of tangled macromolecules
but an arrangement of rather clearly defined polymeric globules (this may
provide a new insight into the mechanism of particle formation during
styrene polymerization). On the other hand, examination of the structure of
poly(vinyl acetate) emulsion particles by the same method gave quite
different results (Fig. 13A and B).
273
274
V. 1. Yeliseyeva
(B)
lefA)
Fig. 13.
they are the last to disappear (after 20 min of etching, Fig. 14C). After
etching for an hour all trace of particles on the substrate disappears. In
order to estimate the size of primary globules and their content in latex
particles, etching was carried out of films obtained by drying of the same
latexes (Table VII).
Electron micrographs of replicas of films obtained by drying latexes
numbers 1, 2, and 3 (Table VII) at room temperature after 20 min oxygen
etching are shown in Figs. 15A, 15B, and 15C. In the comer of each
micrograph a medium-sized particle of the corresponding latex at the same
enlargement is shown. The main structural elements of films are primary
globules which are much smaller than latex particles. Taking the diameter
of primary globules as the distance between their centers, for latex number 3
(Table VIII) we conclude that these distances coincide in micrographs of
particles and films, i.e., the structural element of both particles and films is a
7.
275
MAA
Latex
conc
(%)
Molecular
mass of
polymer
(10- 3)a
O
2.6
4.3
37.95
39.34
39.6
214
203
227
Polymer composition
(%mol)
No.
MA
BA
1
2
3
65
63.3
62.3
35
34.1 .
33.4
Molecular
mass of copolymers
was calculated
Partic1e
diameter
(nm)
Latex
pH
118-210
120--220
118-200
2.9
2.8
2.8
in
dioxane using Staudinger's coefficient for poly(alkyl acrylates) Staudinger and Trommsdorff,
1933).
276
V. 1. Yeliseyeva
Fig. 14.
Electron micrographs
of polyacrylate latex particles (Table
VII. run 3) after oxygen etching for:
(A) 2 min; (8) 10 min; and (C) 20 mino
7.
277
No.
3
4
Polymer
composition
(%rnol)
BA-MA
(65:35)
BA-MA-MAA
(63.3:34.1:2.6)
BA-MA-MAA
(62.3:33.4:4.3)
BMA"
Partic1e
diameter
(nm)
latex
prirnary
Numbers
of primary
partic1es
in latex
partic1e
Number of
macromolecules
in
primary
partic1e
118-210
40
-120
-90-150
120-220
25-30
-530
- 25-40
118-200
15-20
-1000
15-40
no
no
-5-10
u In order to ensure necessary stability of the latex during polyrnerization of BMA the
initial ernulsifier concentration in aqueous phase was taken equal to 3.3%.
Fig. 15.
Electron rnicrographs of replicas frorn the surface of polyacrylate latex filrns:
(A) run 1, Table VII; (B) run 2, Table VII; (e) run 3, Table VII.
278
V. 1. Yeliseyeva
(5)
where K is the proportionality constant which takes into account the value
of equilibrium adsorption.
In reality however it may occur that
dr/dt < K dS/dt
(6)
VIn.
Polymerization Kinetics
279
water, their reactivity, and the fact that they do not display any particularly
specific features of polymerization (as do vinyl chloride and vinyl acetate
which have a high constant of transfer onto the monomer or as do vinyl
chloride and acrylonitrile where there is low solubility of polymer in
monomer).
The differences in polarity of alkyl acrylates (alkyl methacrylates) is
understood as the difference in polarity at the interface with water, since
taken separately lower alkyl acrylates have practically equal dipole moments (Granzhan, 1969). However, the structure of the alkyl acrylate-water
interface layer changes with increasing alkyllength: the polarization of the
carbonyl group, which determines polarity of the asymmetric alkyl acrylate
molecule at the water boundary, decreases as a result of an increase of the
induction effect of the alkyl group. The degree of orientation of monomer
molecules decreases and interfacial tension increases. Medvedev (1956)
proposed that molecules of some acrylic monomers have a negative charge
in aqueous phase.
Yeliseyeva et al. (1976) studied the polymerization kinetics of polar
acrylic monomers using ethyl acrylate. The dependence of the polymerization rate on the initiator concentration (ammonium persulfate) was
examined at equal ionic strength in the presence of two emulsifiers, sodium
dodecylsulfate and oxyethylated (30) cetyl alcohol. Polymerization was
carried out in a dilato meter at 45C, monomer-water phase ratio 1:9, and
ammonium persulfate concentration 0.44-4.4 x 10-3 mol/dm3. The obtained loagrithmic dependence is shown in Fig. 16. In each cases the
1.5
0.5
1.0
1.5
2.0
-4 +tg Ct(mol/dm')
Fig. 16.
Dependence of ethyl acrylate polymerization rate (R mol/dm3 sec) on
a~monium persulfate concentration (mol/drn3) in the presence of (1) 13.8 x 10-3 mol/drn3
SDDS; (2) 2.55 x 10- 3 rnol/drn3 oxyethylated (30) cetyl alcohol. Phase ratio, 1: 9; T, 45C.
280
V. 1. Yeliseyeva
dependence is represented by a straight line with the same slope, the tangent
being equal to 0.5. It should be noted that such a clearcut dependence is not
observed when the ionic strength is not controlled. This may be associated
.
.
o
Fig. 17.
Variation of polymerization rate (R mol/dm3 sec) (Curves I and 2) and of the
number of partic1es (N) (Curves l' and 2') with conversion (5) during polymerization of ethyl
acrylate in the presence of (1) 0.3% and (2) 0.8% SDDS with respect to the monomer. Phase
ratio, 1: O; T, 45C.
7.
281
(In
(8)
i:I:t 0.5
20
50
20
80
C07!Versio7! (%)
Fig. 18.
Rate (R %/min) versus conversion dependence during polymerization of (1)
methyl acrylate; (2) ethyl acrylate; and (3) butyl acrylate in the presence of SDDS (A) and
AAOES (B). The graphs are plotted from the kinetic curves in Fig. 9. Initiator benzoyl
peroxide.
282
V. 1. Yeliseyeva
adsorption of emulsifier and the rate of attainment of equilibrium adsorption value is sufficiently high for protection of partic1es from flocculation.
The protection is also enhanced by the absence in the surface layer of
negatively charged S04' groups. Therefore, the probable ratio between
living and dead partic1es is retained at 1: 1 as required by the Smith-Ewart
(1948) theory, and the duration of the constant rate period approximately
coincides with the period of the existence of the monomer phase.
Because of the considerable solubility of ethyl acrylate in water in the
case of persulfate initiation, the process begins in the aqueous phase with
the formation of charged polymer radicals which, after reaching a certain
critical size, precipitate forming primary PMP. Establishment of equilibrium adsorption of the emulsifier, which ensures stability of the partic1es, is
preceded by the flocculation of partic1es with other primary PMP which at
the beginning is accomplished by interpartic1e termination of radicals. Thus,
a new termination factor appears which is not considered in the above
theory and which leads to a higher relative content of d;:adpartic1es. This is
confirmed by data on the average number of radicals per partic1e which
drops to 0.07 for persulfate ibitiation of ethyl acrylate polymerization
(Mamadaliev, 1978). The large number of equilibrium monomer-swollen
partic1es creates, after the disappearance of the monomer phase, a reserve of
monomer in the system which, being redistributed via the aqueous phase
into the growing partic1es, facilitates maintenance of equilibrium monomer
concentration and thus continuation of the constant rate period after
disappearance of the monomer phase. This may also be facilitated by the
complex structure of partic1es and the existence of a monomer concentration gradient.
The logarithmic dependence of the ethyl acrylate polymerization rate on
concentration of emulsifiers of various chemical structure is shown in Fig.
19 (Yeliseyeva et al., 1976). Polymerization was carried out in a dilatometer
f.O
2.0
- i.+ Lg Cc (mol/dm3)
Fig. 19.
Dependence of ethyl acrylate polymerization rate (R moljdm3sec) on
concentration (moljdm3) of (1) SDDS; (2)sulfated polyoxyethylated (10) octadecyl alcohol; and
(3) oxyethylated (30) cetyl alcohol.
7.
283
IX.
In order to establish whether Eq. (7) may be extended to other polymerization systems (Yeliseyeva and Zuikov, 1977; Yeliseyeva et al., 1976, 1978) a
study was conducted on the dependence of.the polymerization rate on the
concentration of dodecylsulfate and its homologs in the polymerization of
other homologs of alkyl acrylates. The logarithmic dependences of polymerization rate of alkyl acrylates on sodium dodecylsulfate concentration is
presented in Fig. 20. With the shortening of the alkyl chain, the slope of the
straight line is regularIy decreasing, indicating a weakening of the rate
dependence on emulsifier concentration. A similar dependence for a wide
range of monomers differing in water solubility was earlier observed by
Gershberg (1965). Data presented in Fig. 20 indicates the general trend of
the established dependence. This is further confirmed by the graph for
styrene which also conforms to this regularity. The value of x determined
from the slopes of the straight lines regularIy increase from MA to hexyl
acrylate (HA) to styrene (from 0.13 to 0.46 to 0.5, respective1y) and from
dedecylsulfate to hexadecylsulfate (from 0.29 to 0.33 respectively) (Table
IX).
In order to establish the correlation between the value of x and emulsifier adsorption characteristics in the given reaction system in latexes
prepared according to the same recipes, as was used in determination of
~eaction order, the adsorption areas of emulsifier molecules in the saturated
284
V. 1. Yeliseyeva
2.0
1.8
1.6
~gCe'.l11oZ
/dmJ)
Fig. 20.
Dependence of the polymerization rate (R mol/dm3 sec) of (1) methyl acrylate;
(2) ethyl acrylate; (3) butyl acrylate; (4) hexyl acrylate; and (5) styrene on sodium
dodecylsulfate concentration. Ammonium persulfate 0.025% in aqueous phase; phase ratio
1:4; T, 60%.
=K
(8)
s = K 1 . ASlim
(9)
285
Emulsifier
Ead.
(kJ/mol)
Asum
(nm2)
K (nm2)"
20.0
22.1
23.3
24.6
25.6
-
1.51
0.92
0.82
0.74
0.67
0.52
0.48
0.197
0.198
0.238
0.244
0.222
0.240
0.240
0.13
0.21
0.29
0.33
0.33
0.46
0.50
C12H2SS04Na
C12H2SS04Na
C14H29S04Na
C16H31S04Na
C12H2SS04Na
C12H2SS04Na
C12H2SS04Na
" K = x . Asum.
it follows that x
= K2/S.
in polymerization
6
20
5/.
'+0
o
60
Converston (0/0>
Fig. 21.
Dependence of overall surface of particles (5) on conversion during
polymerization of ethyl acrylate. SDDS 1.04 x 10-3 mol/dm3; ammonium persulfate
0.44 x 10- 3 mol/dm3; phase ratio, 1: 9; T, 45C.
286
V. 1. Yeliseyeva
N = const
Am = ASlim
monomers)
N = ((Ead"t)
S = const
= const
R = KC~_6C?_4
ASlim= I(Ead,)
R = kC~ . C?-5[X = ",(Ead,) =
tr.con~1 =
K/Aslim = K'/AsdyoJ
tr'con~1> !m'ph
t'm'ph
Nomenclature
Am
As
ASlim
C.
C
D
Ead'
I:1G
= DJD =
w
kp
N
propagation
rate constant
number of particles
p
R
(ASlim!As}-degree
of adsorption
polymerization
rate
n-D~
1/3
"~ n-D~
-1/3
~ni ) (LniD~)
~
of particles
of resultant
latex
s
x
287
1'12
interfacial tension at the boundary of the first and second phases
I'
surface tension at the ith phase
1'1,1'~2 surface or interfacial tension due to dispersion interaction
I'f, I'Y2 surface or interfacial tension due to polar interaction
r
adsorption value
'm.ph duration of monomer phase existence
'r'consl duration of the constant rate period
References
Bagdasar'yan, Kh. S. (1966). "Theory of Radical Polymerization," p. 102. Nauka, Moscow.
Bakaeva, T. V., Yeliseyeva, V. 1., and Zubov, P. 1. (1966). Vysokomol Soedin. 8, 1073.
Breitenbach, J. W., Kuchner, K., Fritze, H., and Tarkowsky, H. (1970). Dr. Polym. J. 2, 13.
Davis, J. T., Higuchi, T., and Rytting, J. H. (1972). J. Pharm. Pharmacol. 24, 30.
Dunn, A. S., and Chong, Z. C-H. (1970). Dr. Polym. J. 2, 49.
Fitch, R. M., and Tsai, C. H. (1971). "PolymerColloids" (R. M. Fitch, ed.), pp. 73,103. Plenum
Press, New York.
Fowkes, F. M. (1964). Ind. Eng. Chem. 56, 40.
Gerrens, H. (1964). DECHEMA Monogr., Frankfurt a/M 49, (859), 346.
Gershberg, D. (1965). J. Chem. Eng. Symp. Ser. Inst. Chem. Eng. N3, 3.
Granzhan, V. A. (1969). Zh. Vses. Khim. Obshch. 14, 223.
Gritsenko, G. M., and Medvedev, S. S. (1956). Dokl. Akad. Nauk SSSR 11, 235.
Harkins, W. D. (1947). J. Am. Chem. Soco 69, 522.
Jacobie, E. (1960). Electron Microsc. 1.
Kaeble, D. H. (1971). "Physical Chemistry of Adhesion," pp. 153-170. Wiley (Interscience),
New York.
Kiselev, M. R., Evko, E. t.. and Luk'yanovich, V. M. (1966). Zavodskaya Lab. 32, 201.
288
V. 1. Yeliseyeva
ofSynthesis,
Study of
V.
V.
V.
V.
l.,
1.,
l.,
l.,
114l.
Yeliseyeva, V. 1., Zuikov, A. V., and Lavrov, S. V. (1976b). Vysokomo/. Soedin. Ser. A 18,648.
Yeliseyeva, V. 1., Zuikov, A. l., Vasilenko, A. l., and Shehepetil'nikov,
B. V. (1978). Int. Symp.
Macromol. Chem. Tashkent 89-90.
Yuzenko, V. l., and Mints, S. M. (1945). Dokl. Akad. Nauk SSSR 47(2),106.
Zuikov, A. V., and Soloviev, J. V. (1979). Acta Polym. 30, 503.
Zuikov, A. V., and Vasilenko, A. 1. (1975). Kolloid. Zh. 37, 640.
Zuikov, A. V., Vasi1enko, A. l., and Yeliseyeva, V. l. (1976). Vysokomol Soedin. Ser. A 18,707.
CON/\.CYT
<i> PROYECTO
8
Recent Developments
and Trends in the
Industrial Use of Latex
Carlton G. Force
1. Introduction .
11. Factors in Adhesion .
A. Physical Adsorption
B. Diffusion.
C. Electrical Charge Distribution
D. Chemisorption .
E. Tack .
111. Bonding Applications .
A. Adhesives
B. Fiber Bonding
C. Tufted Carpet Applications
D. Paints and Industrial Coatings
E. Paper Coatings .
.'
IV. Construction Applications .
V. Rubber Goods
VI. Properties of Various Latexes
References
l.
289
291
291
296
297
298
298
300
300
302
306
309
310
312
313
314
316
Introduction
290
Carlton G. Force
latexes for very long periods even when there is a considerable discrepancy
between the density of the polymer and that of the aqueous medium.
This small particle size is a characteristic necessity of the emulsion
polymerization reaction permitting good heat control through the aqueous
medium at rapid polymerization rates. However, technology is emerging
(Vanderhoff et al., 1979) which permits the preparation of emulsions in the
latex particle size range from polymers that have previously been unavailable as latexes primarily because of sensitivity to water in their synthesis.
This technology is expected to expand rather rapidly as the understanding
of methods for producing microemulsions in the latex particle size range is
transferred from monomeric systems, such as those being dealt with in
tertiary oil recovery research (Shah and Chan, 1980), to polymeric applications. This type of technology has already been used commercially for
more than a decade for the preparation of polyethylene emulsions by the
indirect pressure method (von Bramer and McGillen, 1966).
Availability of such a universally convenient me~ium of application
permits optimization of product properties through polymer designo The
most fundamental property of polymers deposited from latexes is their glass
transition temperature. This dictates the temperature conditions under
which they form continuous films and the degree of softness or flexibility
imparted to the item containing the latex polymer. The glass transition
temperature is characterized by the monomer making up the backbone of
the polymer molecule. However, it can be modified without overwhelming
other desirable characteristics by the copolymerization of such comonomers
as low glass transition temperature acrylic esters with high glass transition
temperature vinyl chloride to produce copolymers incorporating the advantages of poly(vinyl chloride) (PVC) but with glass transition temperatures
down to below room temperature. Comparable technology is used in the
production of many plastics and elastomers, for example, poly(vinyl acetate)
and styrene-butadiene (S-BR). Flexibility can also be imparted to films
developed through the fusion of high glass transition temperature polymers
when a phase inversion takes place in the polymer melt to incorporate
thoroughly the plasticizer in which the polymer had been previously
suspended. Again, this is common practice in the utilization of poly(vinyl
chloride) produced by the emulsion polymerization process (e.g., plastisols
and organosols).
In the utilization of emulsion polymers many specific characteristics are
imparted by the specific polymer to the product. Although cost is always a
factor in producing any commodity, achieving optimum characteristics for
end use at an acceptable price is of course the ultimate aim. Latexes have
found many of their opportunities through permitting achievement of this
goal by unique designs producing equal or better utility more cost ef-
291
ficiently than the products they replace. Thus, the versatility permitted with
latexes for the ingenious incorporation of the proper polymer into a
construction allows tremendous latitude, with the price of the polymer often
being a minor consideration compared to the properties imparted. This
chapter is organized around current applications of latexes. The specific
properties imparted by the various latexes utilized for each application will
be pointed out. New and emerging technology in each field will be
emphasized. However, the main objective will be to pro vide sufficient
background and stimulation to enable the reader to begin synthesizing
innovative emulsion polymers.
11. Factors in Adhesion
Regardless of the end use, factors that have been recognized as important
in adhesion are usually involved to some extent. Therefore, an outline of
these factors is desirable. An excellent comprehensive review of this subject
was recently published by Wake (1978). The factors responsible for adhesion can be subdivided into four categories: (i) physical adsorption, (ii)
diffusion, (iii) electrical charge distribution, and (iv) chemisorption, all of
which i~teract to some degree in every adhesive bond. A fifth characteristic
relating primarily to the adhesive polymer is tack.
A.
Physical Adsorption
292
Carlton G. Force
293
must be positive. In Eq. (2), Ysvrefers to the surface free energy of surface
tension of the solid, YLv is the surface tension of the liquid, and YSLis the
interfacial tension between the solid and liquido In other words, the surface
free energy of the solid must be decreased by the combination of the surface
free energy of the liquid plus the interfacial free energy between the solid
and the liquid for spontaneous spreading and wetting of the solid by the
liquid to take place. For solids, neither Ysvnor YSLis directIy measurable.
However, two methods have been developed which permit determination of
whether this condition of wetting of the substrate by a proposed adhesive is
likely to be met. A strictIy experimental method deveIoped by Zisman and
co-workers (1968) at the Naval Research Laboratory is based on a value Ye,
defined as the critical surface tension of the surface. Ye is the surface tension
of the liquid which will just, and only just, wet the surface of the solido
In order to determine this value experimentalIy, use is made of the concepts developed by Young. The Young equation is the mathematical expression of the vector diagram of a drop (shown in Fig. 1) which only partialIy
wets a solid surface:
Ysv
(3)
In Eq. (3), the angle (Jis the contact angle made by the liquid on the flat
solid surface. It is dependent on the surface tension of the liquido The lower
the surface tension, the greater the tendency to wet the solid, and the
smalIer the angle (J.Ye is the value obtained by extrapolating to zero contact
angle the results of contact angle measurements made with a series of
liquids having different surface tensions on tQe solid in question. The Naval
Research Laboratory group has actively pursued this approach for welI
over a decade and has recentIy published (Shafrin, 1977) Ye values covering
some 150 synthetic and naturalIy occurring polymers. The values for some
of the materials of interest to the emulsion polymers industry are shown in
Table 1.
A more theoretical approach to determining whether an adhesive can be
expected to wet the surface of a solid has been developed and promoted by
Girifalco and Good (1957) and Fowkes (1964). Utilizing the geometric mean
Fig.1.
Vector diagram
Carlton G. Force
294
TABLEI
Critical Surface Tensions of Polymer Materials
Critical surface
tension (y c)
Polymer
Phenol/resocrinol
adhesive
Resorcinol adhesive
Wool
Cellulose (regenerated)
Nylon 2
Polyacrylonitrile
Casein
Nylon 7,7
Poly(ethylene oxide)
Poly(ethylene terephthalate)
Nylon 6,6
Nylon 11
Nylon 6
,
Cellulose, from cotton linters
Nylon
Poly(methyl acrylate)
Polyoxyphenylene
Polysulfone
Poly(vinylidene chloride)
Cellulose acetate
Poly(methyl methacrylate)
Poly(vinyl chloride)
Starch
Chloroprene
Polyisoprene, chlorinated
Poly(vinyl acetate)
Poly(vinyl alcohol)
Nylon 9,9
Poly(vinyl ethyl ether)
Rubber (natural)/rosin
adhesive
Cellulose, from wood pulp
Polyacrylamide
Polyacrylate
Polyepichlorohydrin
Nylon 8,8
Polyisoprene, cyclized
Poly(ethyl methacrylate)
Polystyrene
Nylon 10,10
cis- Poly-1,3-butadiene
Polyoxypropylene
trans- Poly-1 ,3,-butadiene
(dynecm-I)
52.0
51.0
45.0
44.0
44.0
44.0
43.0
43.0
43.0
43.0
42.5
'42.0
42.0
41.5
41.0
41.0
41.0
41.0
40.0
39.0
39.0
39.0
39.0
38.0
37.0
37.0
37.0
36.0
36.0
(36.0
35.5
3540
35.0
35.0
34.0
34.0
33.0
32.8
32.0
32.0
32.0
31.0
(continued)
295
TABlE I (continuad)
Critical surface
tension (Yc)
(dyne cm -1)
Polymer
31.0
31.0
30.0
29.0
29.0
29.0
28.0
28.0
27.0
27.0
25.0
22.0
Polyethylene
cis- Polyisoprene
trans- Polyisoprene
Poly-4-tert- butylstyrene
Polyurethane
Poly(vinyl methyl ether)
Poly(butyl acrylate)
Poly(vinyl butyal)
Butyl rubber
Polyisobutene
Poly-l,2-butadiene
Silicone rubber
a
Van Nostrand
Reinhold
Co.
to take into account the London dispersion forces between molecules across
the interface, the interfacial tension between two immiscible liquids, such as
mercury or water and a hydrocarbon, can be expressed by
Y12= Yl + Y2
2{r1Y1)1/2
(4)
(5)
296
Carlton G. Force
the polymer can also be calculated, such as the heat and free energy of
adsorption. The van der Waals attractive constant A can also be calculated
from yd, as was done for S-BR latex mentioned earlier. These dispersion
force interactions are quite strong, being approximately equal in strength to
hydrogen bonds. In cases, where solid and liquid systems can interact only
by London dispersion forces, Fowkes has demonstrated that for wetting to
occur
yty~ > 2YL
(6)
From Eq. (6), it can be determined that y~ must exceed 243 dynejcm for
water to wet a solid surface.
From this, it can be seen why meticulously clean surfaces are usually
necessary for good adhesion. Although the contaminant may wet and thus
coat the substrate, the contaminant will often have sufficiently low surface
free energy so that the adhesive in turn cannot wet it. It should be noted,
however, that contaminants do not always lead to poor adhesion. In some
cases, a contaminant can improve adhesion, for example, by providing
improved wetting on one constituent and adhesion by another mechanism,
such as chemisorption or hydrogen bonding, to the other constituent.
B. Diffusion
This mechanism of adhesion advanced and originally promoted by the
Soviets (1963) is of the greatest importance to applications where a material
is being adhered to itself, such as in the self tack of rubber. Another familiar
example is the application of the same adhesive to two different surfaces
from a solvent system, allowing the solvent to dry partially or totally, and
then combining the treated surfaces, such as in the use of "contact cement"
or self-sellingenvelopes.
In the theory of diffusion it is postulated that high molecular weight
polymer molecules interdiffuse with each other across the interface. The
term "autohesion" is often applied to this process in the adhering of
portions of the same plastic material together. Since molecules of the same
material diffuse across the interface, making the two layers one, the original
joint disappears. Once the joint is completely healed, there is little chance
that adhesive failure will occur at the original interface.
For autohesion to occur, the polymer must be well above its glass
transition temperature. It is the primary mechanism operating in the heat
sealing of plastics and the coalescence of droplets from emulsions of the
film-forming plastics such as S-BR rubber during drying.
297
298
Carlton G. Force
D. Chemisorption
Whereas the electrical charge distribution mechanism described in the
preceding section postulates a relatively nonspecific electrostatic attraction
or donor-acceptor interaction across the interface, chemisorption is considered to involve the formation of covalent or donor-acceptor bonds
between specific molecules on opposite sides of the interface. One of the
more popular current applications of this theory is the siloxane coupling
agents used for bonding glass or hydrated metal oxides to polymeric
adhesives. These coupling agents are provided with appropriately reactive
terminal groups on one of the organic linkages to the silicon atom. The
reactive group could be a double bond, for reaction with unsaturated
polyester resins dissolved in styrene; an amine or carboxyl group, for
reaction with epoxy resins; or a mercaptan, to interact in the vu1canization
reaction with rubber. Some materials which are blended with elastomers for
the purpose of reacting with the substrate, the elastomer, or both, to
produce chemisorptive adhesion, include isocyanates and resorcinolformaldehyde resins in the familiar RFL, or resorcinol, formaldehyde, vinyl
pyridine latexes used for tire cord dips. Patterson (1969) has written a good
survey of the technical and patent literature of fiber adhesion.
Chemisorption is strongly emphasized in this survey, although not to the
exclusion of other mechanisms.
E. Tack
Closely associated with the factors responsible for adhesion is the
tackiness of the polymer. An adhesive is said to possess tack if, under the
conditions of application, only light pressure is required to produce a bond
sufficiently strong to require work to restore the interfjlce to its original
separated state. This differs from the preceeding four factors in that the
objective in those four cases was to produce a sufficiently strong bond so
that failure would be cohesive in the substrate or adhesive, but not at the
original interface. Tack is generally measured by the energy required to
separate the bond after contact has been made, under carefully controlled
conditions, which should include pressure, time, and temperature. Two
primary factors are involved in such a measurement. First, the adhesive'
surface must readily wet the surface of the material with which it is brought
into contact in order to bond quickly with minimum pressure. Second, the
bulk of the adhesive should generally be capable of absorbing, through
deformation, a portion of the energy applied to separate the bond. Thus,
the viscoelastic properties of the adhesive are extremely important in
8.
299
;;-
ID
o
a:
CL
30
10
1400
50 70
200
100
400
900
1200
1000
<{
i5
800
'",
S
600
a:
UJ
CL
'"
'-'
:!
400
200
20
40
60
'Q
80
I
100
Fig. 2.
Ine., Wilmington,
Delaware.
Reprinted
by permission
of Van
Carlton G. Force
300
III.
Bonding Applications
The polymer particle size in latexes is far too small to be effective for
adhesive applications. Therefore, many of these very tiny building blocks
must coalesce through the diffusion adhesion mechanism to provide an
optimum sized critical mass of polymer for the particular adhesive requirement. Methods of bringing the crucial number of particles together at the
proper point of intersection for optimum bonding represents much of the
ingenuity and skill involved in the utilization of emulsion polymers in
adhesive applications.
Equally important is having the correct emulsion polymer for the
physical conditions available to create the bond. This involves such factors
as glass transition temperature of the specific polymer and molecular weight
of the polymer molecules. Too low a molecular weight do es not provide
sufficient chain length for adequate entanglement and diffusion across
particle boundaries. Too high a molecular weight makes the polymer so
viscous it cannot flow readily enough to pro vide maximum intimate contact
with the optimum amount of substrate during the time provided for
bonding under application conditions (Mallon et al., 1980). In emulsion
polymerization,molecular weight is generallycontrolled by the quantity of
chain-transfer agent (often an alkyl mercaptan) used in the recipe.
A.
Adhesives
301
Carlton G. Force
302
B.
Fiher Bonding
303
incorporated into the fibrous stock before formation of the paper (beateradd) in this initial development. Arter World War n, synthetic rubber
latexes provided the smaIler partiele size, versatility, and broader range of
properties which permitted expansion of the industry into new methods of
rubber addition to pre-formed sheets (dry web and wet web saturation).
Applications also expanded to inelude replacement of eloth as weIl as
leather. The uniqueness of the products often developed their own special
applications, such as in asbestos and fiber reinforced gasketing. The area
has now expanded to inelude nonwoven synthetic fiber composites prepared from both water and air layed structures. Those bonded by latexes
are primarily dry web saturated in one manner or another.
There has been considerable development over the last few years in
methods for saturating these nonwoven structures. Additional new methods
would be expected to emerge periodicaIly to provide the physical and
aesthetic properties required to broaden the rapid expansion of this industry into different applications characteristic of its growth. Among the
saturating techniques presently used for synthetic fiber nonwovens are print
bonding (the desired amount of latex is transferred by a patterned gravure
roIl of a two roIl nip to the web), saturation bonding (the synthetic web,
often supported, is saturated in latex bath and excess squeezed out in
passing through a two-roIl nip), impregnation bonding (the desired amount
of latex is transferred to the unsupported web by the lower roIl of a two-roIl
nip the bottom of which, is rotating in a latex bath), foam bonding (the
foamed latex is metered onto the fibrous web with a knife coater and sucked
into the structure by passing the web ov.er a vacuum box), and spray
bonding (the latex is sprayed onto a high bulk to weight ratio web for end
uses such as fiber-fiIl'insulation while a vacuum box below the web control s
penetration) (Brooks, 1979).
The synthetic fiber nonwoven industry developed around air lay fiber
substrates. The hydrophobic character of synthetic fibers in water makes
them difficult to disperse satisfactorily to produce suitable substrate using
the more rapid water lay papermaking technology. However, this has been
overcome by developing synthetic fibers having tiny hairlike protrusions
(fibrils) extending, verticaIly from their surfaces in water. These fibrils keep
the fibers physicaIly separated from each other and in good dispersion for
substrate formation on papermaking equipment. In order to provide adequate strength in the substrate for drying on paper machine drying drums
or other unsupported drying systems, it is often desirable to incorporate
some latex prior to sheet formation to assist the fibrils in fiber bonding until
the sheet is finaIly saturated.
The primar"y advantage imparted to fibrous structures by latex bonding
is toughness in the form of very good hand tear strength along a smooth cut
304
Carlton G. Force
305
anionic fiber furnish. Dow points out that the many previous attempts to
utilize cationic latexes in fiber furnishes ha ve shown no pronounced advantage because the latexes were stabilized with migratable monomeric cationic
emulsifiers (Heiser et al., 1962; Wessling, 1979). In aqueous solution,
sufficient of these emulsifier molecules quickly migrate to the negative
charge sites on the fibers making them neutral before the bulkier rubber
particles can travel and attach to these sites. In this same vein, care must be
taken not to add materials such as alum to the fiber furnish which will
make it cationic (Goring and Mason, 1950; Goring et al., 1950) before
addition of the cationic latex.
In latex saturation processes of preformed fiber webs, the predominant
force responsible for drawing the rubber particles to fiber crossings and
coagulating these particles into a coherent mass is the surface tension of the
evaporating water medium. Current trends are to develop latexes whose
surface tension is close to water by utilizing oligomeric emulsifiers (White,
1976) or emulsifiers polymerized into the polymer chain (Sweeney, 1958) to
maximize the surface tension forces.
This industry has remained highly technological, producing on relatively
slow equipment and employing a disproportionately large technical staff
compared to other areas of papermaking primarily because of cost considerations involved in the relatively inefficient utilization of rubber. With
the possible exception of optimized systems in which the latex is of opposite
charge to the fiber at the time of addition, paper products with suitable tear
strength ordinarily contain from 20 to 40% rubber. Inefficiency in anionic
latex beater-add sheet s is due to the frac.tion of filtered rubber particles
attached to only one fiber. In saturated, preformed webs most fiber
junctions are encapsulated with rubber, whereas much less would be
required if the rubber were located only between the fiber crossings. Again
considerable rubber is deposited on only one fiber.
Fiber bonding is a market strongly dependent on latex properties. The
tear strength is dependent on the viscoelastic properties of the rubber.
Adhesive strength of the rubber to the fiber must only be greater than the
tensile strength of the rubber itself. Thus, the low viscosity rubber will string
out and break near the center between two fibers before the attachment to
either fiber separates. This may welI be why carboxylated latexes show little
or no advantage in properties over uncarboxylated versions in these
products. The additional bond strength provided by carboxylation in many
applications is simply not required in fiber bonding. Rubber cure can also
be undesirable in some latexes in that it shortens the stretchability of the
rubber. Any degree of sulfur cure of S-BR bonded fibers drasticalIy reduces
tear strength of the product. However, slight sulfur cure of a natural rubberbonded product is desirable for improving the rubber oxidative stability
Carlton G. Force
306
without harming tear strength. Cure of other polymers such as the metal
oxide-catalyzed cure of neoprene or the heat cure of acrylics is desirable to
improve properties such as solvent resistance and often develops the
optimum degree of stretchability for tear strength (Sweeney, 1958).
A myriad of opportunities exist for the latex producer to design specific
properties into new latexes that will allow the bonded fiber manufacturers
to develop new products. These can be based on polymer properties such as
solvent resistance,' flexibility characteristics, and stre~gth. New developments can also be based on the colloidal characteristics 'of the latex. The
cationic latexes of Dow and Rohm and Haas are examples. Another
opportunity is in the development of methods for reducing the tendency of
latex particles to migrate with the evaporating water from the center of
saturated products to the hotter surfaces during drying. Unique colloid
stability characteristics might also permit the development of new beater
addition and saturation techniques for improved latex efficiency and
properties.
C.
Another area of fiber bonding where latex is the key ingredient is in the
m~nufacture of tufted carpets. In this process, the yarn tufts making up the
surface of the carpet are needle punched through a fabric known as the
primary backing. This was originally jute fabric (burlap). However, it has
been replaced by nylon nonwoven fabric which in turn is currently being
displaced in many grades of carpeting by a coarse woven ribbon-like
polypropylene fabrico Another backing that will sit on the underlay pad or
floor is then attached to this basic carpet structure. This secondary backing
is often fabric (Klein, 1979). Jute, the secondary fabric workhorse until the
fall Qf 1979, is presently being replaced by woven polypropylene and spunbonded polypropylene because of economics and the uncertainty of jute
supply from the Far East. Roughly one-fifth of tufted carpeting in the
United States and almost all tufted carpeting in Europe has, as the
secondary backing, attached foam rubber made from latex.
A latex adhesive is necessary to bond the tufts into the primary backing,
adhere the individual fiber strands in the yarn tufts together so they don't
separate and "pill" at the carpet surface, and attach the secondary backing
to this primary structure. This is accomplished almost exclusively by
carboxylated styrene-butadiene latex containing 400 or more parts per
hundred parts of rubber of a mineral filler. In addition to economics, this
filler helps pro vide density and stiffness to the carpet structure for the
desired "hand." Feldspar used to be the filler of choice, but less expensive
calcium carbonate has replaced it in most instances.
307
308
Carlton G. Force
more rapid remo val of the water. Third, the thinner films of rubber within
the bubble walls assure thorough coverage of the two substrate surfaces
with a lower quantity of adhesive if other characteristics such as bond are
adequate. Adhesive frothing, which has now expanded to over 90% of
United States industry, has reduced green tack adhesion problems considerably allowing acceptable properties to be achieved without the excess
of adhesive often applied to insure good early bonding. Frothed latex
application has been important in the drapery and furniture fabric industries for at least a decade to provide strength, the adhesive for fIock
coating and desired "hand" properties to the fabrics utilized.
Westvaco has recently introduced a water-dispersed tackifier to the
carpet industry which improves both green tack and final adhesion. Since
these properties are currentIy adequate, it is possible to reduce the amount
of rubber in the adhesive compound by substituting this tackifier and still
maintain the required properties in the carpe1. Good success is achieved
replacing up to 15% of the rubber in fabric backed carpets and 20% of the
rubber in the precoat upon which latex foam backing is to be applied.
Because of the difficult wetting characteristics of polypropylene, opportunities exist to provide latexes and ingredients that stimulate better adherence between substrates made from this polymer.
The current no-gel process of applying foam rubber to carpet tuft bonded
with a precoat that might contain 700 or more parts of mineral filler per
hundred parts of rubber became predominant in 1975. Prior to that, and
still used for applications such as waffie-designed carpet backing and
underlay, the more technically demanding gel system was utilized. Gel foam
is compounded with ammonium acetate which is decomposed to acetic acid
and ammonia gas by infrared heat just beyond the foam coat metering bar.
This acidifies the system sufficientIy so that zinc ions present enter a pH
region where they are hydrolyzed to a + 3 species which then electrolytically coagulates the rubber (Force, 1971). In the no-gel system, a frothing
agent (tallow amine sulfosuccinamate) produces a foam of sufficient stability
that the water can be dried from it without foam collapse. This simpler
technique has permitted the carpet milis to reduce costs by compounding
their own products for foam backing instead of relying on technical experts
who supplied the gel formulation. However, they have accepted some
disadvantages in doing this. One is a reduction in coating speed to permit
sufficient drying before the first turn roll in multipass ovens so the coating
will not fall off the carpe1. The second is greater carpet drying difficulty after
water spills in use because of the more hydrophilic character of the
surfactant used in the no-gelsystem. Finally, tallow amine sulfosuccinamate
does not produce as firm a rubber foam structure as the gel system. This
softer resiliance provides a less cushiony feel under foo1. In spite of these
-309
deficiencies, economics dictate that no process much more costly than the
no-gel system can gain prominence in this industry. However, the deficiencies also provide opportunities for latex producers and chemical suppliers
to develop new products and new processes for this application.
The predominant latex used for backing foam carpet is S-BR made by
the cold redox-initiated system. Smaller amounts of specialty latexes such as
poly(vinyl chloride) and polychloroprene are sometimes used for meeting
specific flame spread requirements where chlorine-containing polymers
compounded with suitable catalysts such as antimony salts pro vide distinct
advantages. A fraction of hydrated alumina in the filler portion of both the
adhesive and foam coating allow S-BR to meet government regulations for
flame spread for most applications. Urethane foam is making small inroads
into this market. A number of suppliers are currently developing urethane
foam systems. The cost and availability of natural gas normally used for
carpet drying will probably be the factor governing how quickly this
development matures. Materials that permit compounding of higher solids
latex foam systems should help deter penetration of urethane into this
industry. Systems that permit application of foam backing of greater
thickness than the roughly i in. limit currently practical could also expand
the United States market beyond the one-fifth of carpeting which it now
enjoys and utilize more latex per carpet.
D.
310
Carlton G. Force
E.
Paper Coatings
Paper and paperboard for printing are coated to improve both printability and appearance. The coating consists primarily of mineral pigments
bound to the celIulose surface with synthetic and natural polymers. A
variety of additives may also be incorporated into the coating such as
lubricants, antifoams, pigment dispersants, flow modifiers, and dyes. The
binder level is generalIy in the range of lOto 20% of the quantity of pigment
and is composed of a combination of synthetic latex and a natural polymer
such as starch, protein, or casein. In addition to bonding the pigment to the
substrate, each binder pro vides other desirable properties to the coating.
311
For example, carboxylated S-BR helps produce gloss in the coating and seal
the surface for good ink hold out. In offset printing both water and inks are
applied to the paper. Therefore, the coating must be hydrophilic and porous
enough for the paper to absorb the water during drying. Starch for
hydrophilicity and poly(vinyl acetate) which pro vides strong pigment bonding but is nonfilm forming to maintain porosity is the binder combination
of preference for coating offset printing paper. For gravure printing, which
is currently gaining in popularity, the ink used is of sufficiently low viscosity
that surface bond strength of the coating is of little importance whereas
surface smoothness is of greater significance. Thus, film-forming synthetic
latex is often the exclusive binder (Heiser, 1979).
Pigment and binder help fill in voids between the surface fibers. The
pigments main purposes, however, are to reflect more light back to the eye
for improved brightness of the surface and to help scatter light passing
through the sheet for better opacity to prevent show-through of the
printing. Because of its high refractive index, titanium dioxide is an
extremely effective light scatterer for opacity improvement. Small particle
size clays, calcium carbonate, and hydrated alumina make good brightnessimproving pigments. Dow has introduced some synthetic polymer pigments
in the form of latex-like water dispersions. Just as polystyrene latex binder
pro vides high gloss in supercalendered paper, these 0.4 to 0.5 Jlm pigment
particles provide gloss, optimized scattering for opacity due to the specific
particle size, good brightness, and ink holdout. Economics will be important in determining how much penetration plastic pigments make into this
market.
Carboxylated latexes made with a small fraction of polymerizable fatty
acid in the monom'er mix pro vide many advantages as coating binders.
These improvements include better adhesion to both the mineral pigment
and cellulose through the bound carboxyls enhancement of the electrical
charge distribution mechanism and hydrogen bonding mechanism, respectively. This provides advantages in gluability and wet rub. Other
advantages include improved compatibility with starch and better coating
blade performance.
Among current trends is the compounding of paper coatings to higher
solids for faster drying at lower energy requirements. Future opportunities
in which the latex manufacturer can participate include providing better ink
holdout and more uniform ink adsorption, enhancing easier finishing (or
even elimination of the need for finishing), providing specific latexes for low,
medium, or high gloss coatings, enhancing faster ink drying, providing
improved basestock reinforcing and better coverage at faster coating speeds,
and improving coating blade runability (Heiser, 1979). These challenges will
be met through both polymer design and colloid chemistry of the latex and
coating systems.
Carlton G. Force
312
IV.
ConstructionApplications
313
cation in parking garages, recreational are as, and industrial areas where
the resiliency property is also of prime importance.
A large application of latex is in blends with asphalt to produce a
waterproof membrane on cinder block and concrete walls. Here, as well as
when incorporated into concrete mortars, the rubber fraction produces
better adhesion, durability, and temperature expansion tolerance than the
asphalt or mortar base alone.
With the decreasing availability and increasing price of petroleum,
methods are being implemented to convert more of it to higher value
distilled fractions. As a result, the residual asphalt fraction, is changing in
properties. There are also vast differences in asphalt properties depending
upon the section of the world from which the crude oil is obtained. An
active program is just beginning among petroleum refiners to investigate
the potential for achieving and retaining optimum properties in asphalt
through the incorporation of a fraction of latex in their products.
v.
RubberGoods
A large proportion of solid polymers are made by emulsion polymerization covering the entire range from tire rubber to high technology
thermoplastic resins such as ABS. However, these solid polymers and their
production is too vast a subject for a single chapter. Thus, this section will
be confined to a couple of product applications whose production technology is dependent on the colloid chemistry of latexes, in addition to being
a convenient medium for production of the .rubber.
Elastic thread used in many garments such as pajama waists and sock
tops is made either by cutting from solid rubber or extruding from latex.
Cut thread has the disadvantage in some applications of having square
edges. Extruded thread is produced by running a continuous stream of latex
into a coagulation bath through a circular hole that pro vides cylindrical
thread of the proper diameter. For natural rubber latex, the coagulant is
usually acetic acid whose density is adjusted to allow the thread to remain
suspended whilt: being drawn through the bath. The rubber thread thus
formed is then drawn through a leach bath to remo ve salts and passed
through an oyen for drying and curing of the rubber. This method of elastic
thread production is less costly than cut thread. However, the dry heat
resistance, particularly under tension, of this extruded thread is not as good
as cut thread. Development of better rubber cure technology in the
extruded thread process should allow latex to take over the majority of this
market. Synthetic latexes can provide special properties to the thread (e.g.,
polychloroprene can pro vide good solvent resistance under stress). In such
314
Carlton G. Force
VI.
TABLE
Application
Natural rubber
Styrene-butadiene
Nitrile
Polychloroprene
Poly(vinyl chloride)
Poly(vinyl acetate)
Acrylic resin
Adapted
Properties of Commonly
Used Latexes8,b
Glass
Typical
transition
Uncured
Organic
Price
cohesive
Polymer Oxidative Color
solvent
Water
temp
Compounding Specific Gravity
(0C)
requirements latex solids ranking
strength Resiliency tack
stability stability resistance resistance
Latex
11
E
F
S
E
G
G
G
E
S
E
E
G
F
F
E
F
P-G
P-G
P-G'
F
F
F
G
G-Ed
G-E
E
E
F
F
F
F
. S
E
E
P
P
E
S-Ed
G-S
F-G
F-G
E
E
E
E
E
F
G-S
Armstrong
-67
-48
-25
-40
78'
30
-55 to 451
Extensive
Extensive
Average
Extensive
Minimal
Minimal
Minimal
0.95
1.00
1.00
1.12
1.17
1.09
1.06
Pennsylvania.
.
d
Cured systems.
Copolymerized
with acrylate or vinylidene chloride
1 Depending on acrylate monomers utilized.
monomers;
glass transition
temperature
Major
markets
(by topic
outline)
0.91
High B-I,3,D
0.99 Moderate B-l,2,3,4,5,C
1.00
High B-l,2,3,5,D
1.23
High B-l,2,3,C,D
1.41
Low B-2,3,4,5,C,D
1.18
Low B-l,2,4,5,C
1.10 Moderate B-l,2,3,4,C
316
reasons. Many of
latex and natural
uncured cohesive
synthetic emulsion
Carlton G. Force
the industries were developed utilizing natural rubber
rubber possesses a few important properties such as
strength and tack that have yet to be matched by
polymers.
Acknowledgment
The author wishes to express his appreciation to the following persons who graciously
provided information on the subjects dealt with in this chapter: William de Vry and David
Weaver of B. F. Goodrich Chemical Company; B. R. Vijayendran of Celanese Polymer
Specialities Company; Evans Townsend of the Norton Company; and J. J. Gorman, C. H.
Gilbert, and J. C. Fitch of E. I. duPont de Nemours and Co. Appreciation is also expressed to
Westvaco Corporation for permission to publish this chapter and to numerous colleagues at
Westvaco for their suggestions and assistance.
References
Alinee, B., Robertson, A. A., and Inoue, M. (1978). J. Co/loid Interface Sci. 65, 98.
Bierwagen, G. P. (1979). Adv. Emul. Polym. Latex Technol., Ann. Short Course Papers, 10th,
Vol. n. Lehigh Univ., Bethlehem, Pennsylvania.
Briscoe, B. J., and Tabor, D. (1972). Faraday Special Discussion, "Solid/Solid Interfaces,"
Chem. Soco No. 2, p. 7. Chemical Society, London.
Brooks, B. A. (1979). Adv. Emul: Polym. Latex Technol., Ann. Short Course Papers. 10th, Vol. n.
Lehigh Univ., Bethlehem, Pennsylvania.
Force, C. G. (1971). Paper presented at American Chemical Society, Rubber Div., Southern
Rubber Group Meeting, Memphis, Tennessee, October.
Force, C. G., and Matijevi, E. (1968). Ko/loid Z. Z. Polym. 224, 51-62.
Fowkes, F. M. (1964). 1nd. Eng. Chem. 56, (12) 40-52.
Fowkes, F. M. (1967). "Surface and Interfaces" (J. J. Burke, ed.), Vol. 1,p. 199. Syracuse Univ.
Press, Syracuse, New York.
Fowkes, F. M. (1973). "Recent Advances in Adhesion" (L. H. Lee, ed.), pp. 39-43, Gordon and
Breach, New York.
Girifalco, L. A., and Good, R. J. (1957). J. Phys. Chem. 61, 904.
Goring, D. A. l., Biefer, G. J., and Mason, S. G. (1950). Can. J. Res. 288, 339-344.
Goring, D. A. l., and Mason, S. G. (l950a). Can. J. Res. 288, 307-322.
Goring, D. A. l., and Mason, S. G. (l950b). Can. J. Res. 288,.323-338.
Green, B. W., and Saunders, F. L (1970). J. Co/loid Inte~race Sci. 33, 393-404.
Grimley, H. (1973). "Aspects of Adhesion" (D. H. Alner and K. W. Allen, eds.), Vol. 7, p. 11.
Transcripta Books, London.
Hamaker, A. C. (1937). Physica 4, 1058.
Harkins, W. D. (1941). J. Chem. Phys. 9, 552.
Heiser, E. J. (1979). Adv. Emul. Polym. Latex Technol., Ann. Short Course Papers, 10th, Vol. n.
Lehigh Univ., Bethlehem, Pennsylvania.
Heiser, E. J., Morgan, R. W., and Reder, A. S. (1962). TAPPI 45,588.
Hildebrand, J. H., and Scott, R. L. (1950). "Solubility of Non-Electrolytes," 3rd ed. Van
Nostrand-Reinhold, Princeton, New Jersey.
317
Klein, A. (1979). Adv. Emul. Polym. Latex Technol., Ann. Short Course Papers, 10th, Vo!. 11.
Lehigh Univ., Bethlehem, Pennsylvania.
Lee, Y. S. (1980). Polym. Con! Polym. Improved Energy Efficiency, 11th, Akron, Ohio.
Mallon, C. B., Hagan, J. W., and Hamer, A. D. (1980). Polym. Con[ Polym. Improved Energy
Efficiency, 11th, Akron.
Ottewill, R. H., and Shaw, J. N. (1966). Discuss. Faraday Soco 42, 154.
Patterson, H. T. (1969). "Treatise on Adhesion and Adhesives," (R. L. Patrick, ed.), Vol. 2,
Chapter 6. Dekker, New York.
Shafrin, E. G. (1977). "Handbook of Adhesives" (l. Skeist, ed.), 2nd ed., pp. 67-71. Van
Nostrand-Reinhold, Princeton, New Jersey.
Shah, D. O., andChan, K. S. (1980). J. Dispersion Sci. Technol. 1, 55.
Skiest, 1. (ed.) (1977). "Handbook of Adhesives," 2nd ed. Van Nostrand-Reinhold (Litton
Educational Pub!.), Princeton, New Jersey.
Sweeney, E. J. (1958). TAPPI 41,304.
Vanderhoff, J. W., EI-Asser, M. S., and Ugelstad, J. (1979). U.S. Patent 4,177,177, December 4.
Visser, J. (1972). Adv. Colloid Interface Sci. 3, 331.
von Bramer, P. T., and McGillen, W. D. (1966). Soap Chem. Spec. 42(12), 123.
Voyutskii, S. S. (1963). "Autohesion and Adhesion of High Polymers" (S. Kaganoff,
Translator), (lnterscience), Wiley, New York.
Voyutskii, S. S. (1971). J. Adhes. 3, 69.
Wake, W. C. (1978). Polymer 19, 29.
Watillon, A., and Joseph-Petit, A. M. (1963). Am. Chem. Soco Div. Water and Waste Preprints
61-65.
Wessling, R. A. (1979). Adv. Emul. Polym. Latex Technol.. Ann. Short Course Papers, 10th, Vol.
11. Lehigh Univ., Bethlehem, Pennsylvania.
White, W. W. (1976). J. Elastomer. Plast. 8, 475.
Zisman, W. A. (1968). "Contact Angel, Wettability and Adhesion" (R. F. Gould, ed.), Adv.
Chem. Ser. No. 43. American Chemical Society, Washington, D.C.
9
Latex Reactor PrincipIes:
Design, Operation, and Control
A. E. Hamielec and J. F. MacGregor
1. Introduction .
11. Batch Reactors .
A. Introduction .
B. Monomer Consumption Rate .
C. Polymer Quality .
D. Initiation Systems .
E. Heat Removal
111. Continuous Stirred-Tank Reactors: Steady-State
Operation .
A. Introduction .
B. Polymer Quality .
C. Optimal Reactor Type and Operation
for Continuous Emulsion Polymerization
IV. Continuous Stirred-Tank Reactors: Dynamic
..
Behavior .
A. Introduction .
B. A Dynamic Model Based on Age Distribution
Functions.
C. Consideration of Dynamics in Reactor Design .
V. On-Une Control of Continuous Latex Reactors
A. Introduction .
B. On-line Instrumentation
C. On-line Estimation .
D. On-line Control.
VI. Summary.
Nomenclature
References
319
320
320
322
325
329
330
333
333
334
338
339
339
340
344
345
345
346
348
349
351
351
353
l. Introduction
The principIes of latex reactor design, operation, and control
will be illustrated by a consideration of the homopolymerization of styrene
and vinyl acetate. Emulsion polymerization of vinyl acetate follows Case 1
319
EMULSION
POLYMERIZATION
Copyright
~ 1982 by Aeademie Press, Ine.
AIl rights of reproduetion
in any form reserved.
ISBN 0-12-556420-1
320
Batch Reactors
A. lntroduction*
We consider a polymerizing system with Np polymer particles per unit
volume of latex. Np is usually of the order 1016-1018 particles per liter of
latex. Early in their life when polymer particles are relatively small (1001000 ).and when the polymer concentration is relatively low, termination
with an entering radical will be effectivelyinstantaneous. Thus, particles will
contain either zero or one radical. Later in the polymerization particles may
contain two or more radicals. The polymerization rate in a single particle
with n free radicals is given by
= kp[Mp]n
R
pp
(1)
NA
Rp
"
(2)
= n=1 n;np
(3)
= kp[Mp]Np
(4)
and finally
Rp
section.
321
= VRp
(6)
NMn
To get a feel for a typical polymer production rate, let us employ reasonable
numerical values for the parameters in Eqs. (4) and (6). These follow:
kp
[Mp]
Np
= 103 liter/Gmol
v=
sec
= 5 Gmol/liter
= 1017 partic1es/liter
latex
1 liter
NMn= 5 Gmol
= 0.5 (Case 2 kinetics)
from which Eq. (6) yields dX/dt '" 30%/hr. This is certainly a desirable rate
of polymerization for commercial production. The production rate in
practice would likely be limited by the heat removal capacity of the reactor.
The number of partic1es Np is a function of emulsifier type and concentration and initiator level, although for monomers that obey Case 1
kinetics such as vinyl chloride and vinyl acetate, Np is almost independent
of initiator level (Ugelstad et al., 1969; Friis and Nyhagen, 1973). The
calculation of Np for various reactor operations will be discussed later. The
monomer concentration in the polymer partic1e [Mp] can be obtained using
a simple mas s balance. Assuming the monmer and polymer volumes are
additive, one obtains. the following relationship for the conversion interval,
Xc :<:::
X :<:::
1.0 where Xc is the monomer conversion at the end of Stage 11.
(7)
To calculate
= Xc
mination of requires, in general, solution of the Smith-Ewart recurrence formulas (O'Toole, 1965; Ugelstad et al., 1967). The general
solution for the case of negligible water-phase termination (Y = O)is shown
graphically in Fig. f. The parameters are given by
a = p1 V/(Npk,P)
a' = R.V(Npk,P)
m = kdeV/k,P
(8)
Y = 2Npk,pk'w/(k; V)
For most emulsion polymerizations, water-phase termination can be neglected (Y = O) (Ugelstad et al., 1969). However, when an active chain-transfer
322
y=o
-3 L..L
-8
-6
-4
-2
log a'
Fig. 1.
1. Case 1 Kinetics
Consider a monodispersed latex, where water-phase termination is pegligible and termination is instantaneous when a radical enters a polymer
particle containing one radical. By definition, N2 = N3 = ... = O and the
total radical entry rate per liter of latex equals pl. Application of the
stationary-state hypothesis gives
R)
= 2pl(N/Np)
(9)
323
Ugelstad et al. (1967, 1969) have derived a more general expression which
accounts for the possibility that two radicals can coexist in a polymer
particle:
R
p
kp[Mp]
(N ) (
A
1/2
RI12 VpNi + Np
2k.p
2kde
(13)
(14a)
(14b)
(15)
= No = tNp
(16)
In other words
NI
and
-1.
-2
(17)
324
One normally assumes that systems such as styrene and methyl methacrylate, where transfer to monomer is not prominent, follow Case 2 kinetics
when latex particles are small and termination in polymer particles is
instantaneous. It has recently been shown that at low initiation rates radical
desorption can be significant relative to radical absorption, and as a
consequence values appreciably smaller than 0.5 were found (Gilbert et
al., 1980).At higher initiation rates = 0.5 was approached. The use of
chain-transfer agents would of course increase the desorption rate and
lower .
For Case 2 kinetics the rate of polymerization is given by
Rp
= kp[Mp]Np/2NA
(18)
with the rate independent of initiator concentration for fixed Np. Ir the
initiation rate is reduced to a great extent, will fall below 0.5 as radical
desorption becomessignificant.
3. Stage III PolymerizatiolJs
In Stage III, the polymer concentration in the polymer particles increases
with ever increasing chain entanglements, and if the polymerization tem-:
perature is below the glass-transition temperature of the polymer being
syntht;sized, a glass-state transition will occur with the diffusion coefficients
of small molecules falling by several orders of magnitude. The pro paga tion
reaction becomes diffusion-controlled and a limiting conversion of less than
100% is reached where the rate of polymerization is effectively zero (Friis
and Hamielec, 1976). Marten and Hamielec (1978) pro po sed the following
models for the effect of polymer molecular weight and free volume on the
termination and propagation constants.
k
~=
M:
2a
1
---
VF
VFcrl
( ) [ (
~ = exp[- B(~ - ~ )]
k,o
kpo
a exp(b/RT)
~C<I
exp -A
Mw
VF
)]
(20)
VFcr2
(21)
= M~c:' exp(A/VFcr.)
VF = [0.025 + Cip(T-
(19)
Tgp)]<I>p
(22)
9.
325
TABlEI
Diffusion-ControlledTermination and Propagation Model Parameters
Polyrner
PMMA
PS
PVC
Tgp(0C)
ex
110
95
85
1.05
0.48
0.25
1.0
0.7
0.2
1.75
1.75
1.75
VFcr1
0.066
0.043
0.085
0.563
0.926
8900
3830
= Xc in Eq. (23).
C. Polymer Quality
1. Molecular Weight
Polymer can be produced via the followjng reactions: transfer to monomer to chain transfer agent and to polymer, termination reactions, and
terminal double bona reactions. Transfer to polymer and terminal double
bond reactions produce long-chain branches. Transfer to small molecules
and termination reactions produce linear polymer chains. A major simplifiGation in the modeling of molecular weight and branching development
results when it can be assumed that a negligible amount of polymer is
produced via termination reactions. Fortunately, in most emulsion polymerizations, transfer reactions are relatively more important than termination
reactions in the production of polymer. This is a direct consequence of the
compartmentalization of radical s in polymer partic1es which permits commercial polymerization rates at relatively low radical initiation rates. When
the amount of polymer produced by termination reactions is negligible,
molecular weight and branching development is independent of initiator
and emulsifier levels (or number and size of the polymer partic1es). The
appropriate equations that describe molecular weight and branching development, which have been proved valid for the emulsion polymerization
326
,
A. E. Hamielec and J. F. MacGregor
of vinyl chloride and vinyl acetate in a batch reactor, follow (Friis and
Hamielec, 1975).
For X < Xc
MR
KXc
1+CM
1 - Xc
)
2MR 1 + KXc
(
)
CM+-1+1 - Xc (1 - Xc)(
1 - XC)
EN= ~
( - )( + K)
MN=-
(25)
M=
l-Xc'
Xc
CM = krm/kp,
KXc
2CpXc
CpXc
CpMn
MR
Cp = kr~/kp,
K = k;/kp
(26)
(27)
(28)
For X > Xc
dQo = CM- KQo
dX
1- X
(29)
dQl = 1
dX
(30)
1
dQ2 =2
dX
KX
+ 1- X
)(
= MRQdQo
Mw = MRQ2/Ql
CpQ2
KX
+1=X+1=X
C
CpX
M+I-X
(31)
)
(32)
(33)
(34)
(35)
..
327
The prediction of particle number and size distribution has been far less
successful than the prediction of conversion time histories and molecular
weight development, given these parameters as initial conditions. The
nucleation and early growth of polymer particles is even today, after the
comprehensive investigations of Ugelstad and Hansen (1976), not. well
understood. There are several reasons for this. First, the measurement of
number
particles
smaller than
100
A presents
rather
difficult experimental problems. Many complex processes occur simultaneously including radical capture by micelles and polymer particles,
precipitation of growing radicals from the aqueous phase, and finally
particle flocculation and coagulation. To illustrate how certain variables
such as initiation rate and emulsifier concentra tion might affect particle
number, we will refer to a highly idealized model for particle number after
Smith and Ewart (1948). To simplify the analysis, let us assume that the
area occupied by an emulsifier molecule is the same on the micelle as on the
polymer particle. In other words
(36)
For typical emulsifiers, as has values of 30-100 A2jmolecule or 1.8-6 x 107
dm2jGmol. The detailed derivation may be found in the original reference
and the final result has the form
(37)
where (X= 0.53 when radical capture by polymer particles is neglectedand
= 0.37 when micelles and polymer particles both capture radicals and
assuming surface area on both kinds of particles are equally effective in the
capture process. The diffusion theory of radical capture would predict that
the micelles, being smaller, would provide more effective surface area for
radical capture and thus give an exvalue greater than 0.37. According to
Ugelstad and Hansen (1976), the capture process is more complex with the
possibility of both collision and diffusion theories of radical capture applying under different limiting conditions. To develop effective control strategies for polymer particle concentration, it is of considerable importance to
(X
328
establish the time scale for micellar nucleation of polymer particles. The
effectiveness of any control scheme would of course depend on the
nucleation time. Np is given by
Np
= NAR.tr
(38)
=6
X 10-3
Gmol/liter
gives
v
cum F(v) =
.J.tr
v
= O.53.J.2/5(as[S]/R.NA)3/5
(39)
It is clear that the PSD at time tr narrows with increasing initiation rate
and broadens with increasing emulsifier concentration. Ir the original PSD
at tr is very narrow, then all particles will grow at the same rate in Stage 11
and the fully converted latex will have a narrow PSD. On the other hand, if
the original PSD is broad, the final PSD will be broad. Data on the final
PSD after Gerrens (1959) showing the effect of initiator concentration on
the breadth of the PSD are shown in Fig. 2. As anticipated by Eq. (39), the
distribution broadens with decrease in initiator concentration. To predict
the PSD to the end of Stage 11 for a broad distribution requires a
knowledge of the radical capture mechanism. According to the collision
theory, surface area on a small particle is equally effective in capturing
radicals to that on a large particle. On the other hand, according to the
diffusion theory surface area on a small particle is more effective.
9.
329
1.0
0.8
~ 0.6
IL
E
::J
()
0.4
0.2
1.0
0.5
1.5
2.0
V (LlTERS X 1016)
Fig. 2.
Cumulative partic1e size distributions for styrene emulsion polymerization at
different initiator concentrations [1]0 (in g/liter of water) (adapted from Gerrens, 1959):
A, 5.41; B; 2.88; C, 1.80; D, 1.08; E, 0.541; F, 0.361.
D. lnitiation Systems
Hydrophilic initiators such as the persulfates and hydrogen peroxide and
initiators with varying solubilities in the polymer particles such as the
hydroperoxides with 'or without reducing agents are employed in emulsion
polymerization. A recent review of redox systems in emulsion polymerization was published by Warson (1976). The most efficient manner of
adding the components of a redox system is gradually in stages or by
adding the peroxide at the start and then the reducing agent gradually
(Warson, 1976).
A typical redox system for cold SBR production employs sodium
formaldehyde sulfoxylate (reducing agent), a hydroperoxide (oxidant), and
ferrous sulfate, plus' a chelating agent or a chelated irol) salt. A simplified
kinetic mechanism for this redox couple follows (Wright and Tucker, 1977).
ROOH + FeH
Fe3+ + XH
--
RO'+M-
ROM'
330
The function of the chelator is to complexthe ferrousion and thus limit the
concentration of free iron. Redox systems appear very versatile, permitting
polymerization at ambient temperatures and the possibility of control of the
rate of radical initiation versus polymerization time. This would thus permit
control of heat generation and the minimization of reaction time. The use of
the redox system ammonium persulfate (2 mmol) together with sodium
pyrosulfite (Na2S20S 2.5 mmol) together with copper sulfate (0.002 mmol)
buffered with sodium bicarbonate in 1 liter of water form an effective redox
system for vinyl acetate emulsion polymerization. The reaction was started
at 25C and run nonisothermally to 70C. The time to almost complete
conversion was 30 min (Warson, 1976; and Edelhauser, 1975).
The rational for the use of finishing initiators that have appreciable
solubility in the polymer particJe follows. At high conversions, the concentration of monomer in the aqueous phase is very low and water-phase
termination of hydophilic radical s becomes excessive. The rate of radical
entry into polymer particJes is thus greatly reduced an,d the polymerization
rate falls to a very low level prematurely.
E. Heat Removal
331
batch. In other words, the polymerization rate and associated heat generation rate should equal the heat removal capacity at all times during the
batch. For an isothermal polymerization this can best be done by controlling the initiation rate using a redox system. Of course, the sinallest reaction
time occurs with adiabatic polymerization. The temperature rise can be
reduced by controlled heat removal to moderate the spread in molecular
weights. Typical conversion and heat generation rate profiles for an isothermal polymerization with constant radical initiation are shown in Fig. 3. The
heat generation rate over the batch is far from optimal as the maximum
heat remo val capacity of the reactor is about 220 arbitrary units. The
polymerization time could be significantly reduced by employing a higher
initiation rate early in the polymerization and by letting it fall with time so
as not to exceed a heat generation rate of 220 arbitrary units. A similar
reduction in polymerization time could be achieved with semibatch monomer feed using monomer starvation to give a polymerization rate equal
to Rpmax'A cold monomer feed would also assist in heat removal. The
shortest batch time would be obtained using semibatch monomer feed
giving Rpmaxand permitting the temperature to rise.
In semibatch emulsion polymerizations the polymer particles are kept
monomer-starved to obtain higher rates of polymerization and to permit
easier control of the rate and particle size distribution. There are two
aspects to the control of PSD. The controlled addition of emulsifier during
particle growth stabilizes the particles without further particle nucleation.
The second aspect is related to the particle sticky stage which often occurs
1.0
2.5
0.8
1.5 ~
0.6
"el.
:J:
<1
1.0 I
X
0.4
0.5
t (hrl
Fig. 3. Conversion and instantaneous heat generation rate for a typical isothermal
batch emulsion polymerization.
332
.O
>C
U
Q)
lit
I
..
Q)
::
......
o
E
Q.
Q:
1.0
. 2.0
3.0
4.0
50
Conversion
Fig. 4.
semibatch
J
o
(%)
emulsion polymerization.
at intermediate conversion levels and certain polymerjmonomer concentrations. During the sticky stage wall fouling and formation of coagulum
often occur. The use of semibatch operation under monomer-starved
conditions effectively eliminates the sticky stage. To illustrate that it is
easier to control a semibatch emulsion polymerization under monomerstarved conditions, consider the rate curve shown in Fig. 4. Point P1 at low
monomer concentration is a stable operating point for semibatch feed of
monomer, whereas P2 is unstable. A consideration of perturbations in
monomer feed rate clearly illustrates this fact.
Semibatch operations are commonly employed in copolymerizations to
maintain a more uniform copolymer composition. The monomers are fed at
rates equal to their consumption rates in the reactor.
333
A. lntroduction
In this section, emulsion polymerization in continous stirred-tank
reactor(s) (CSTRs) operating at steady state will be treated, with emphasis
on polymer quality, Le., molecular weight, long-chain branching, and
polymer particle number and size distribution. The commercially important
problems associated with oscillations in conversion (Le., particle number
and size, weight average molecular weight, and long-chain branching) and
their control will be considered in following sections.
A single CSTR or a train of CSTRs may offer several advantages over
batch reactors both with regard to production rate and polymer quality.
The copolymer composition distribution obtained in a CSTR is generally
narrower than that for a batch reactor. To obtain fairly narrow copolymer
composition distributions without appreciable loss in productivity, one has
the flexibility to increase the number of CSTRs in series to approach plug
flow.
Because of diffusion-controlled termination and propagation with concomitant Rpmaxat a particular conversion, it is possible to operate a CSTR
at considerably higher production rates. Because of the additional beneficial
effects of cold monomer and water feeds on heat removal, much higher
production rates are possible than with a batch reactor of the same volume.
It should be remembered that polymer production rates are usually limited
by heat removal capacity.
Quality variations in the polymer produced in batch reactors are often
caused by slight vaI:iations in the reactor start up procedure. Furthermore,
the polymerization rate may change considerably during the batch and this
may give temperature variations that are difficult to reproduce causing
batch-to-batch variations in quality. These problems would be minimized
with CSTRs if the continuous reactor system could be operated for at least
several weeks before wall fouling and coagulum build up become critical
and require reactor shutdown for cIeanup. Ir an effective start-up procedure
for a continuous reactor train is not available, the costs associated with
offspec material' could make continuous operation uneconomical. In addition, with a continuous reactor system one loses the flexibility of batch
reactors when a multiproduct operation, with its short productions runs, is
involved.
During start-up and operation of CSTRs there is a potential safety
hazard created by the possibility of more than one stable steady-state
334
operating point (Ley and Gerrens, 1974). The start-up procedure should
ensure that the desired steady state is reached.
B. Po/ymer Qua/ity
1. Particle Number and Size Distribution
The frequency distribution of partic1e radius F(R) is. related to the exit
age distribution E(t) with the identity
F(R) dR
= E(t) dt
(41a)
which states that the number fraction of polymer partic1es in the exit stream
of a CSTR train with radius in the range R to R + dR equals the fraction of
fluid in the exit stream with age in the range t to t + dt. For a single CSTR
F(R) dR
= rr1 exp( -
tlO) dt
(41b)
Equation (41b) is valid when there are no polymer partic1es flowing into the
reactor with aIl the partic1es nudeated within the reactor. lt is assumed thar
density changes can be neglected and that partic1es foIlow the streamlines.
These are reasonable assumptions in view of the smaIl size of partic1es and
the smaIl density difference between particIe and water. When two or more
CSTRs are employed in series, however, one must remember that the total
residence time of a polymer partic1e is made up of different times in each
reactor in the train. The relative amounts of time spent in each reactor wilI
not matter when the volumetric growth rate of a partic1e is the same in
each. This would require that the temperature, monomer concentration,
and average number of radicals per partic1e be the same for each reactor, an
unlikely possibility. This idealization is useful, however, when ilIustrating
the effect of increasing the number of CSTRs in series on the breadth of the
partic1e size distribution.
The exit age distribution function for N tanks in series is given by
E(t) =
(42)
= KlR2
(43a)
where
K = kp[Mp]MR/4nNAdp
(43b)
9.
335
= (2R2/K()
(44a)
exp[ -(2/3K()R3J
= 1-
(44b)
exp[ -(2/3K()R3J
In a similar manner the following PSD may be derived for Case 1 kinetics
for a single CSTR (Friis and Hamie1ec, 1976).
(45a)
F(R) = K~() exp[ -(2~1() )R2 J
cum F(R)
K
1
(45b)
= 1 - exp[ -C~I()
(4/3n)2/3 R,
kp[MpJMR
dp
)R2 J
1/2
(45c)
J[ 16n2NANpke J
0.8
~0.6
=' 0.4
u
R ( Al(
Fig.5.
Cumulative
16
12
20
24
10-3)
for a single CSTR. Case 2 kinetics.
336
PSD broadens with increasing K(). For situations where a separate monomer phase exists, [Mp] is a constant independent of (/ and perhaps
varying slightly with temperature. K will therefore depend on temperature
alone and increase with an increase in kp. Thus, increase in temperature and
() broaden the PSD. The PSD from a single CSTR is excessively broad for
most applications and it is thus desirable to add additional CSTRs in the
train to narrow the PSD. Eq. (42) shows the effect of increasing the number
of CSTRs on the .exit age distribution which also narrows and approaches
that of a plug flow reactor (PFR) (or a batch reactor). The narrowest PSD
would be obtained in a PFR. The use of 5-6 CSTRs in series should give a
PSD which is insignificantly narrower than that for a PFR.
The total number of polymer particles per liter of emulsion (Np) in a
single CSTR may be calculated by
[S]o
[S]
Np
= ()RNA
as[S]
Ap
dR
(46)
=O
[S]
(47)
The last term in Eq. (46) gives the amount of emulsifier covering the
polymer particles in the exit stream. Equation (47) is obtained by equating
the rate of particle outflow to the rate of particle nucleation within th
(48)
= 1 + NAR(4n/as[S]0)(3K/2)2/3()S/3r(5/3)
Q.
Z
10
81hrl
Fig. 6.
9.
337
= (MjM~)
exp(-MjMN)
(49)
-1
M
N
1.fJ:
2
=M
krm + krx[xJ
R kp.
kp[M]
It should be noted that at high conversions some branched PVC may form
and it may therefore be necessary to consider the transfer to polymer
reaction (Hamielec et al., 1980b). When transfer reactions control molecular
weight development, all polymer particles in the first CSTR in the train will
have the same molecular weight distribution (MWD) (neglecting inflow of
polymer particles) given by Eq. (49). It there is no CTA and each reactor in
the train is at the same temperature then the MWD will be the same in all
the reactors in the train. Ir a CT A is employed for an isothermal train, the
contribution from the CT A will increase in downstream reactors as the
monomer concentration falls, and therefore molecular weight averages will
fall. Of course, the MWD in each reactor will be of the same form and given
by Eq. (49), but the final whole polymer will be a composite of the polymer
produced in each reactor. This can be easily calculated if the conversion in
each reactor is known. It should be remembered that less than one CTA
molecule is consumed per polymer chain and therefore the CT A consumption by reaction can often be neglected.
338
M = MR
N.
Mw
CM
where
KWp
(50)
KW
2MR 1 + 1 -
(
= ~
(c
BN
( + 1- W )
(
~)
M+1-W
Wp
)(
1-W
CpMn
+K
MR
(51)
CpWp- 2CpWp 1
p
)(
K~
+l-W
(52)
al. (1974) have measured the following parameters for vinyl acetate using
batch emulsion polymerization.
kp = 1.89 x 107 exp(- 5650/R T)
krm= 3.55 x 106exp(-9950/RT)
krp = 1.430 x 106exp(-9020/RT)
(k:)o = 1.07 x 107 exp(-5650/RT)
liter/mol sec
liter/mol sec
liter/mol sec
liter/mol sec
(53)
For Wp> 0.2, a decrease in k: was observed and this decrease is given by
(54)
339
IV.
A. lntroduction
Continuous emulsion polymerization is one of the few chemical processes in which major design considerations require the use of dynamic or
unsteady-state models of the process. This need arises because of important
problems associated with sustained oscillations or limit cyc1es in conversion, partic1e'nlJmber and size, and molecular weight. These oscillations
can occur in almost a11commercial continuous emulsion polymerization
processes such as styrene (Brooks et al., 1978), styrene-butadiene and vinyl
acetate (Greene et al., 1976; Kiparissides et al., 1980a), methyl methacrylate,
and chloropene. In addition to the undesirable variations in the polymer
and partic1e properties that will occur, these oscillations can lead to
emulsifier concentrations too low to cover adequately the polymer partic1es,
with the result that excessive agglomeration and fouling can occur.
Furthermore, excursions to high conversions in polymer like vinyl acetate
340
B.
P(t)
()
1 op(t, -r)
) ( t) +
f( tpt
-nt
( -r) d-r
,
'
1o ot
(55)
where () is the reactor residence time and f(t) is the net rate of particle
generation in the reactor. Using this equation, property balances for the
total number of particles, the total diameter, the area and volume of
9.
341
(56)
dD (t)
--f-
(57)
D (t)
= ---t-
+ 2~(t)N(t) + dp(t,t)f(t)
dAp(t)
+ 4nW)Dp(t) + tIp(t,t)f(t)
dt = - A~(t)
(7
dVit)
dt
=-
dX(t)
--
= (1 - e-I/6){
V~t)
(7
(58)
(59)
+ ~(t)Ap(t) + vp(t,t)f(t)
- X(t)
()
+ Rp(t)
(60)
[MFJ }
The exponential term in the conversion equation arises from the assumed
start-up condition of a reactor filled only with degased water and then fed
at time zero with a stream having monomer concentration [MFJ.
Assuming both micellar and homogeneous nuc1eation the partic1e generation ratef(t) can be written as
(61)
and following Fitch and Tsai (1971), the homogeneous rate constant is
assumed to decrease with partic1e area as k.h= kho(1- ApL/4). Applying the
stationary-state hypothesis to the radical balance gives the radical concentration in the wter phase as
(62)
where the rate of appearance p(t) of free radicals in the water phase is given
by
(63)
Assuming that radical termination takes place exc1usively in the polymer
partic1e,the average number of radicals in a partic1eof volume vp(t,-r)can
be obtained as (Ugelstad and Hansen, 1976)
q(t, -r) =
1
2kde(t, -r)n(t,-r)d-rJ
1/2
p"
Ap(t)
1/2
(64)
342
q(t,,)
R,
1/2
2kodeJ
1/2
a;(t,')
A p(t)
(65)
(66)
The partic1e generation rate f(t) can now be evaluated using Eqs. (61), (62),
and (66),in conjunction with the property balance equations (56-60).
The rate of polymerization Rp [see Eq. (2)] can be expressed here as
Rp(t) = (kp[Mp]/NA)
LtJ(t,,)n(t,,) d,
= (kp[Mp]/NA)(RJ2kge)1/2
A~/2(
(67)
(68)
The time function ~(t) in the property balance equations (57-59) is given by
kpdm R,
~(t) = NAdp 2kge
1/2
( )
cI>m(t)
A~/2(t) 1 - cI>m(t)
(69)
where cI>m(t)
is the monomer volume fraction in the polymer partic1es given
earlier in Eq. (23).
The property balances must be coupled with the balances for the
initiator and emulsifier given, respectively, by
d[1]/dt
= e-1([1F]- [1]) - R,
(70)
(71)
The micellar area Am(t) needed in Eqs. (61) and (62) can then be obtained as
Am(t) = ([S]
[Scmc])as- Ap(t)
(72)
343
30
t520
;
Q:
ILI
~
15
o
<..>
10
1]= 0.005
6
8
10
12
14
DIMENSIONLESS TIME (t / e)
16
18
20
et al., 1980a,b).
344
2.0
re
TQ
..
.....
..
...
= 1.5
......
C/)
UI
.J
o
~
1.0
el.
la.
O
a:
UI
m
:!! 0.5
:)
z
.J
~
O
1-
O
O
6
8
10
12
14
DIMENSIONLESSTIME (tIa)
16
18
C.
9.
345
cheaper or otherwise more desirable polymer may be used for the seed
partic1es, it usually results in costlier operations resulting from the need for
batch reactors to produce the seed partic1es.
This oscillation phenomenon, however, can be avoided quite easily
without the need for seeding. By employing as the first reactor in the train a
small, optimally sized CSTR reactor and splitting the emulsifier, initiator,
water, and monomer feeds to enter at various points along the train it is
quite possible to ensure that all partic1e generation occurs in the first small
reactor under nonoscillatory
or steady-state conditions, followed only by
growth of these partic1es in the subsequent reactors (Pollock, 1981;
Nomura, 1980). As a simple illustration one could feed most of the
emulsifier to the small first CSTR "seeding" reactor along with part of the
initiator and part of the water and monomer, thereby ensuring continuous
partic1e generation in condition of excessemulsifier. By feeding much of the
remaining water and monomer to the second reactor, the emulsifier is
diluted to a value below the CMC at which partic1e generation is avoided
and partic1e growth only is promoted. A good dynamic model is the key to
sizing the reactors determining an optimal split of the feed streams. As will
be discussed later, the dynamic model is also the key to designing an on-line
monitoring and control system that will ensure that the desired conditions
of latex quality and quantity are maintained.
It is also possible, and in some cases more desirable,to use a continuous
A. Introduct;on
As indicated above a number of fundamental control problems with
cOQtinuous emulsion polymerization reactors are associated with the initial
.
346
B.
On-Line Instrumentation
The most useful on-line instruments would be those capable of measUfing the latex properties given in Eqs. (56-60), the free soap concentration
(Am), and some molecular weight characteristics. In this section we will not
attempt on exhaustive survey, but will only concentrate on a few of the
more rugged and more promising on-line measurements that have been
used.
9.
347
P:OO)
may
o_
Pe
{'
- 1.mPm + JwPw
100 _
Pe
- 1.mPp
{'
+ JwPw
Here fm and fw are the feed stream weight fractions of monomer and water,
and Pm, Pw, and Pp are the densities of, respectively, monomer, water, and
polymer. Using the new on-line densitometers which rely on vibrational
frequencies, a precision of approximately :t 0.5% on conversion can be
reasonably attained in the polymerization of methyl methacrylate (Schork
and Ray, 1980) or vinyl acetate (Pollock, 1981).
3. Turbidity Spectra
Light transmission has been a standard method for the measurement of
the size of spherical particles for many years. Mie and Rayleigh scattering
theories show how the measured turbidity can be related to particle
diameter and number. In the Mie regime (1 < nD/Am < 10) information on
the number of particles and the size distribution could be obtained,
particularly if used in conjunction with other instruments such as a
densitometer (Maxim et al., 1969). In the Rayleigh regime (nD/Am< 1) only
a turbidity-average particle diameter is easily obtained. From the point of
view of on-line process measurements, turbidity instruments are rugged and
have been demonstrated to follow easily the oscillations in average particle
diameter in continuous emulsion polymerizations (Kiparissides et al.,
1980a). For example, Fig. 9 shows a plot ofthe absorbance at 350 nm and the
0.4
o~
o
o
10
0.31<')
o;t 3000
a:
IIJ
1IIJ
~ 2000
~
iS
IIJ
..J
U
: 1000
a:
~
Q.
w
u
0.2~
al
a:
o
0.1
10
12
14
(f)
al
~
o
16
Fig. 9.
= 0.01
mol/liter;
= 0.01
moljliter.
348
Surface Tension
On-Line Estimation
349
On-Line Control
. dx(t)/dt
(73)
where the vector of states x = (N, Dp, Ap, Vp,X, 1, S)Tand the vector of
manipulatable variables u = (IF, SF)' The vector of measurements can be
expressed generaIly as
y(t) = f[x(t), u(t)] + v(t)
(74)
where the vector y(t) and the function f[x(t), u(t)] wilI depend upon which
on-line instruments are available. The vectors x(t) and w(t) represent,
respectively, measurement noise and process disturbances arising from
other variations in the process not accounted for in the modeI. The
variance-covariance structure of these noise sources can be identified from
process data taken from the system.
Using this model of the reactor system, usuaIly discretized for some
sampling interval, near-optimal control algorithms can be formulated which
350
will manipulate the input vector at time instant (UI)'in such a way as to try
to optimize some performance index involving the system states (XI)and the
manipulations (UI)over some interval of time. The most commonly chosen
performance index is the quadratic one
N
J(x,u)= E
L [(XI 1=1
Xdes)TQl(XI
xdes)
where E denotes the expectation, xdes and Udesare the desired operating
levels, and Ql and Q2 are positive semidefinite matrices. For example, this
would inelude the case where it is desired to minimize a weighted sum of the
mean squared errors of the states about their desired values subject to a
constraint on the magnitude mean squared errors of the manipulated
variables about their target values.
Kiparissides et al. (1980b) illustrated the use of this approach, using a
simulation of their single CSTR for the emulsion polymerization of VAc.
They looked at control during reactor start-up, during switching from one
operating condition to another, and at trying to reduce the magnitude of
the oscillations in partiGR: properties while operating in the oscillation
region (i.e., at low emulsifier concentrations). Note that in the latter
situation one cannot entirely eliminate the oscillations since partieles must
be generated sometime.
To implement such near optimal control algorithms would require
considerable space and time on a process minicomputer because of the
severe nonlinearities in the process, and therefore it is probably impractical
in most situations. However, off-line simulations of this control policy
shows some upper-limits as to what could possibly be accomplished and the
nature and magnitudes of the feedrate manipulations that are necessary.
This can then lead to the development of simpler heuristic-type control
algorithms which capture the important aspects of the optimal control but
are more easily implemented on-line.
Some simpler approaches to conversion control using material and
energy balances on a train of CSTRs are described by Amrehn (1977).
351
models for the disturbances in the system, optimal controllers can also be
developed (Box and Jenkins, 1970; Astrom, 1970).
MacGregor and Tidwell (1979) iIIustrate some of the steps involved in
running plant experimentation, building these process and disturbance
models, and implementing simple optimal controllers on some continuous
condensation polymerization processes. A number of similar applications to
continuous emulsion polymerization processes have also been made.
VI.
Summary
In this chapter an attempt has been made to review the state of the art of
the design, operation, and control of latex reactors and to introduce some
promising new approaches to the design of latex reactor trains using
dynamic models. Highly sophisticated dynamic models which can accurately predict particIe number and size distribution and also molecular
weight averages and long-chain branching frequency are being developed.
These dynamic models should permit the design of a new generation of
commerciallatex reactor systems.
Nomenclature
as
ap(t,T)
Ap, Ap(t)
Am
Aw
A,B
a,b
BN
CM,Cp, K
dM
dp
dp(t,T)
!?)w
!?)p
DpDp(t)
E
E(t)
F(R)
fm.!",
f(t)
[l]
[/]0
[/F]
352
K
KI
kp
k,p
k,w
krm
kfx
krp
k*p
k,o
k.
k~.
kd.
km
kv
kh
L
m
M
MR
MN
Mw
Mw.,
[MpJ
[MFJ
Np
NA
Nn
NMo
NM
N,N(t)
n
. n(t, t) dt
p{t, t)
P(t)
for radical
absorption
into
a particle
of radius
defined in text
kd.(t, t) is the.rate constant for radical desorption from a particle of volume v (or
born at time t and present time t)
rate constant for micellar particle generation
volume of particle phase over volume of water phase
rate constant for homogeneous particle generation
distance a radical will dilfuse before precipitating
partition coefficient in Eq. (14)
molecular weight
monomer molecular weight
number average molecular weight
weight average molecular weight
weight average molecular weight when termination
class of
particles
Qo, Ql' Q2
Ql'Q2
QE
ij(t, t)
Rpp
RI
monomer
Rp
Rpma.
R
[8]0
polymerization
rate in Gmoljsec liter latex
maximum polymerization
rate
polymer particle radius and gas constant
radical concentration
in water phase
concentration
of emulsifier
initial concentration
of emulsifier
[SmJ
[KJw
[8]
rate in Gmoljsec
particle
concentration
concentration
of emulsifier adsorbed
of emulsifier at CMC
353
on polymer partic1es
and polymer
vector of manipulatable
variables (Ud.. is desired operating level)
volume of polymer partic1es per liter of latex
free volume fraction
free volume fraction when termination becomes diffusion controlled
~
W(M)
W(t)
X
X.
[X]
x"x(t)
y(t)
c<,o:l,m,Y
Cl
1>m,1>m(t)
1>p, 1>p(t)
S(t)
f1
T
lJ
conversion
concentration
of chain-transfer agent in polymer partic1es
vector of states (xd.. is the desired operating level)
vector of measurements
kinetic parameters defined in Eq. (8)
differences between coefficients of volumetric expansion above
(polymer and monomer)
exponent in model for diffusion-controlled
termination [Eq. (19)]
volume fraction monomer in polymer partic1es
volume fraction polymer in polymer partic1es.
defined in text
and
below
1'.
References
Amrehn, H. (1977). Automatica 13, 533-545.
Astrom, K. J. (1970). "Introduction
to Stochastic Control Theory." Academic Press, New York.
Box, G. E. P., and Jenkins, G. M. (1970). "Time Series Analysis, Forecasting and Contro1."
Holden Day, San Francisco.
Brooks, B. W., Kropholler,
H. W., and Purt, S. N. (1978). Polymer 19, 193-196.
354
Colloids."
NATO Advanced
Colloids"
Study lnstitute,
Trondheim,
Gershberg,
AIChE
Gilbert,
R. G., Napper,
J. E. (1961). In Polymerization
Systems.
D. F. (1980). In Emulsion
A. H. (1970). "Stochastic
Processes
and Filtering
properties
Theory."
Academic
Press, New
Jo, J. H., and Bankoff, S. G. (1976). Am. Inst. Chem. Eng. J. 22, 361-369.
Kiparissides,
c., MacGregor,
Kiparissides, C., MacGregor, J. F., and Hamielec, A. E. H. (1980a). Can. J. Chem. Eng. 58,48-71.
Kiparissides,
(5).
c., MacGregor,
A. E. H. (l980b).
Kratky, O., Leopold, H., and Stabinger, H. (1973). Methdds Enzymol. 27, 98-110.
Ley, G., and Gerrens, H. (1974). Makromol. Chem. 175, 563-575.
Lichti, G., Gilbert, R. G., and Napper, D. H. (1977). J. Polym. Sci. 15, 1957-1969.
Lichti, G. (1980). PhD Thesis, Univ. of Sydney.
M. (1980). In "Emulsion
Polyrnerization,"
Las
Vegas.
Nomura, M., Kojima, H., Harada, M., Eguchi, W., and Nagata, S.(1971).J. Appl. Poly~. Sci.15,
675-695.
355
Nomura, M., Harada, M., Eguchi, W., and Nagats, S. (1976). In "Emulsion Polymerization" (1.
Piirma and J. L. Gardon, eds.), pp. 102-121. Am. Chem. Soco Symp. Ser. 24.
O'Toole, J. T. (1965). J. Appl. Polym. Sci. 9, 1291-1302.
Pollock, M. (1981). PhD Thesis, McMaster Univ.
Schork, F. J., and Ray, W. H. (1980). In "Emulsion Polymerization," pp. 823-828. Am. Chem.
Soco Preprints, Las Vegas.
.
Smith, W. V., and Ewart, R. H. (1948). J. Chem. Phys. 16, 592-606.
Sundberg, D. c., Soh, S. K., and Hsieh, J. Y. (1980). In" Emulsion Polymerization," pp. 520-525.
Am. Chem. Soco Preprints, Las Vegas.
Ueda, T., Omi, S., and Kubota, H. (1971). J. Chem. Eng. Jpn. 4, 50-63.
Ugelstad, J., and Hansen, F. K. (1976). Rubber Chem. Technol. 49, 536-561.
Ugelstad, J., M0rk, P. C., and Aasen, J. E. (1967). J. Polym. Sci. A-I 5, 2281-2296.
Ugelstad, J., M0rk, P. c., Dahl, P., and Rangnes, P. (1969). J. Polym. Sci. C 27,49-60.
Warson, H. (1976). In "Emulsion Polymerization" (1. Piirma and J. L. Gardon, eds.), pp.
228-235. Am. Chem. Soco Symp. Ser. 24.
Wright, D. E., and Tucker, H. (1977). In Polymer Reaction Engineering-Process TechnologySynthetic Elastomers. McMaster Univ. Professional Development Course.
10
Emulsion Polymerization
in Continuous Reactors
Gary W. Poehlein
1. Introduction .
A. Background Information .
B. Reactor Configurations and Flow Possibilities .
C. Differencesbetween CSTRsand BatchReactors
D. Scope of This Chapter .
11. Smith-Ewart Case 2 Model for a CSTR .
A. Particle Formation .
B. Particle Growth .
C. ParticleAge and Size Distributions .
D. Number of Particles and PolymerizationRate .
E. Comparison with a Batch Reactor
F. Effect of Inhibitors in Feed Streams .
G. Use of a Seed-Particle Feed Stream .
111. Deviations from Smith-Ewart Case 2.
A. Large Particles or Slow Termination Rates .
B. Radical Desorption .
.. .
C. Other Particle Formation Mechanisms .
D. ExperimentalResults
IV. Transient Behavior of CSTR Systems.
A. ExperimentalObservations.
B. Physicochemical Mechanisms
C. Control Methods
.
V. Strategies for Process and Product Development .
A. Reactor Design Considerations .
B. Product Development Considerations
VI. Summary.
References
357
357
358
359
361
361
361
362
363
363
366
367
367
367
368
369
370
374
375
375
375
378
378
378
380
381
381
l. Introduction
A.
Background lnformation
358
Gary W. Poehlein
359
360
Gary W. Poehlein
10.
361
D.
Particle Formation
Free radical s, initiated in the aqueous phase, can form polymer particles
by a number of mechanisms. Smith and Ewart (1948) considered two
quantitative models for predicting the number of particles formed. One of
these models is shown as Eq. (1) below.
(1)
362
Gary W. Poehlein
= RfJNA[Arf(asSNA)]
(2)
where () is the mean residence time of the reactor, the working volume
divided by the volumetric effluent flow rateo In order to calculate N one
needs to determine Af from equations for partic1e size distributions. These
equations in turn, come from partic1e growth and reactor residence time
relationships.
B.
Particle Growth
The basic tenet of the Smith-Ewart Case 2 partic1e growth model is that
each partic1e will contain an active free radical one half of the time. Thus,
the average number of free radicals per partic1e is 0.5. The rate of
polymerization Rp is given by
Rp
= kp[M][R.] = kp[M](N/NA)
(3)
10.
363
(5)
v=
Vo +
0.SK1[M]7:
(6a)
r3 = r~ + 3K[M]7:/Sn
(6b)
where ro and Voare the radius and volume respectively, of the particIe at
nucIeation.
= 0-1 exp(-7:/0)
(7)
The particIe size density function U(r) is obtained by combining the growth
equation with the age distribution.
U(r)
= f(1:)/(dr/dt)
U(r)
= [Snr2/(K
U(r)
(Sa)
(Sb)
(9)
The last equation has been derived by assuming that ro, the nucIeation
particIe radius, is equal to zero. Ir the conversion in the reactor is modest
and the particIe sizes are not too large, Eq. (9) can be used for accurate
predictions of particIe size distributions in styrene latexes. Figure 1 is a
typical comparison of Eq. (9) with experimental data.
D.
Equation (2) for the number of particIes per unit volume of aqueous
phase has one undetermined parameter Af which is the difference between
the surfactant charged and the surfactant adsorbed on the surface of the
polymer particIes.
(10)
364
Gary W. Poehlein
1.0
;:
=>
al
0.8
,~
~,,/
~ 0.6
(J)
15
eI '
ILI
~ 0.4
~
~
=>
:::!:
e'I
I
a 0.2
,,
e,'
,
,I
" ,/
.,,,,,;I
200
400
600
PARTlCLE
800
1000
DIAMETER
(Al
1200
1400
Fig. 1. Particle size distribution for a styrene latex (Poehlein and De Graff, 1971).
T = 70oC: 0= 15min; (S)= 2.79g/lOOg H20; (/) = 0.35g/IOOg H20. ---, Dn ($-E Case
2) = 690 ; -,
Dn(De Graaf) = 700: O, Dn(Exper.) = 720.
Combining Eqs. (10), (8b), and (2) yields the following relationship for N
(RfJNAfN) = 1 + (C<oRO/asS)[kp[M]O/(1
C(1[M])]2/3
(11)
or
1t1 = 1
+ 1t2
where the 1t1 and 1t2 are the groups indicated, c(o= 3.85(MoVp)2/3,
C(l= (MoVrnx 10-3), Mo is the monomer molecular weight, and Vp and Vrn
are specific vqlumes of polymer and monomer, respectively. Equation (11) is
written in dimensionless formo The second term on the right hand side 1t2
represents the ratio of surfactant adsorbed on the polymer particIes to the
free surfactant Ar. This ratio is normally much greater than 1 and Eq. (11)
reduces to
(12)
Combining Eqs. (3) and (11) permits the calculation of polymerization rateo
(13)
or
1t3
= 1+ 1t2
Figures 2 and 3 show a comparison of Eqs. (11) and (13) with data for
styrene emulsion polymerization over a wide range of experimental conditions. The rate data fit the theory quite well and the particIe number data
10.
Emulsion Polymerization
in Continuous
DEGRAFF
"
GERSHBERG a
GERRENSa KUCHNER
365
Reactors
LONGFIELD
CALCULATED FROM
R~ ASSUMING
1\
1/2
STYRENE
lO'
102
I
Fig.2.
1971).
103
+ lf2
'"
o
DEGRAFF
GERSHBERG a LONGFIELD
GERRENS a KUCHNER
STYRENE
101
Fig.3.
1971).
loS
102
+ 1T2
Gary W. Poehlein
366
also tit in a reasonable manner. The particle number data are usually
somewhat lower than model predictions. There are at least two reasons that
the pro po sed theoretical model would predict more particles than actually
measured. First, the procedures used for counting particles would normally
error on the small side. Second, the model equations are based on the
assumption that the particle nucleate size ro is zero. Thus, the model would
overestimate the number of particles.
E.
= kR?4(asSNA)0.6
(14)
(15)
O.577NBATCH
Thus, fewer particles will be produced from the same recipe in a eSTRo The
particle concentration in a eSTR can be increased by increasing the
emulsitier concentration or by using seed particles in the feed streams.
TABLEI
Model Equation Exponents for
Smith-Ewart Case 2 Models
Equation exponents
Parameter
R;
S
Batch reactor
0.4
0.6
NA
CSTR
O
1.0
-0.67
367
368
Gary W. Poehlein
= (a/4) (Io[a]/l
(18)
l[a])
1.0
0.5
IC:
O>
!2
-0.5
- 1.0
-4
-3
-2
-1
log 01
10.
369
08
z 0.6
o
::
:o
ID
:
....
'"
o 0.4
"'
>
::
<t
..J
:o
;;
0.2.
/
0/
/
/
/
/
500
1000
PARTICLE
1500
2000
2500
3000
3500
4000
OIAMETER (A)
Fig. 5. Particle size distribution for a styrene latex made at a mean residence time of
59min (Poehlein and De Graff, 1971). T=50C:
O=59min;
(S)=0.6g/100g
H20;
(1) = 1.5 g/100 g H20. ---, Dn (S-E Case 2) = 858 ; -,
Dn (De Graaf) = 1018 ; 0, Dn
(Exper.) = 1090 .
.
Radical Desorption
(19)
m = k~v/kt
(20)
y = 2Nktktw/k;v
(21)
where k~ is the rate constant for radical desorption from particles, ktw is a
termination rate constant for the water phase, and ka is arate constant for
radical absorption. Ugelstad's theory allowed for the reabsorption of
previously desorped radials; a factor not handled properly by Stockmayer.
370
Gary W. Poehlein
2
IC
'"
.2
-1
-2
-6
-8
-o
-2
-4
logo'-
Fig. 6.
Average number
of radicals
= 10-2
r>O
(22)
Smith and Ewart (1948) assumed that particles are formed when free
radical s diffuse into monomer-swollen micelles. Roe (1968) demonstrated
later that the S-E Model, in a mathematical sense, did not depend on the
concept of a micelle. Roe referred simply to the stabilizing capability of the
10.
371
HYDROPHILlC
-S04'
+M
---
-S 04M
SURFACE
ACTIVE
--
I
I +M
,-
-S04Mn
HYDROPHOBIC
I
HOMOGENOUS I+M
NUCLEATION
ADSORPTION
ONEXISTING
R\RTICLES OR
MONOMERDROPS
CAPTUREBY
MICELLES
so;
FLOCCULA
FLOCCULATION
OF PRIMARY
R\RTICLES
TION
ONTO
MATURE
R\RTICLES
+MaS
POLYMERIZATION
GROWTH
CONTINUED GROWTH BY
POLYMERIZATION
I FLOCCULATION,
OR
Fig.7.
BOTH
in the aqueous
phase.
372
Gary W. Poehlein
TABLE
11
Monomer
Water
Styren"e
Butadiene
Vinylidene chloride
Vinyl chloride
Methyl methacrylate
Vinyl acetate
Ethylene
Acrylonitrile
Acrolein
0.005
0.015
0.066
0.11
0.15
0.3
0.3-0.6
1.75
3.1
Polymer
partic1es
5.4
6.5 .
1.1
6.0
7.0
7.6
5.0
373
PRIMARY
PARTICLE
Fig. 8.
374
Gary W. Poehlein
.30
'"
"
.20
>1j
Z
LrJ
o
.J
~
~
a:
o
z
.10.
40
80
DIAMETER
120
(nm)
Fig. 9. Experimental and predictedparticle sizedistribution for a MMA latex (Ray and
Min, 1976).0,Gerren'sdata: -,
Model prediction (97.8%conversion).
D.
Experimental Results
The model based on S-E Case 2 kinetics has been quite successful in
handling steady-state data for styrene emulsion polymerizations in a CSTR.
One or more of the mechanisms described above, however, generally cause
other monomer systems to deviate from this simple model. The nature of
these deviations varies among the different monomers. Ir published literature data are fitted to equations of the type listed below one can obtain
values for the exponents a, b, and c.
Rp = kRRf[S]bOc
(23)
= kNRf[S]bOC
(24)
Table III gives the results of this fitting for styrene, methyl acrylate, methyl
methacrylate, vinyl acetate, vinyl chloride, and ethylene. Only the styrene
data agree with the simple theory.
The exponents on R are usually greater than 0.0 (the theory prediction).
This can be explained by radical desorption from the particles, by slow
t.erminatiop in the particles, or by a combination of these factors. Data on
the other variables ([S] and O)are incomplete but scattered results show
significant deviations from the S-E Case 2 Model. More complete models,
such as those presented in this book by Hamielec, have been successful with
nonstyrene monomers.
10.
375
Steady-State
111
Behavior of CSTRs
Approximate exponents
a
Theory (Rp and N)
Styrene (Rp and N)
MA(N)
MA(Rp)
MMA (Rp)
VA (Rp)
VC(Rp)
Ethylene (Rp)
IV.
A.
0.0
0.0
0.0
0.65
0.8
0.8
0.5 or less
0.5
1.0
1.0
0.85
0.0
0.9
SmaIl
?
0-0.3
-0.67
-0.67
-0.67
+0.43
?
0.0
SmaIl
Small
B.
Physicochemical Mechanisms
376
Gary W. Poehlein
70
60
STYRENE
40
z
O
C
a: 30
UJ
>
Z
O
u
-~
- -
-~
80
120
RUN
Fig. 10.
Conversion transients
160
TIME
at
start-up
200
7O'C
_STAGE2
- ~
7O'C
.: STAGEI
.100'C
20
10
O
StAGE 3
240
280
{min}
for
styrene
emulsion
polymerization
50
40
30
z
020
C
a:
UJ
>
10
METHYL
METHACRYLATE
O
O
Fig. 11.
TIME
10
12
14
(t le)
DIMENSIONLESS
(S) = 0.01molj\iter;
= O.OL (NaCI)
= 0.02.
= 0.43;
377
-,
',
,1 \
"'
"
"
-,
'1
\
\
\
\,
\,
"
i,
I
::>
i
I
\'
11
11
Q: 1.0
w
ID
I
I
I
"-
1Z
w
<.)
Q:
W
a.
/'.,
'
'Y..
'
I "
'-
'-
'
I
I
\",
\
\
0.3
/ /
/-..
0.1
O
500
1000
1500
2000
2500
3000
~RTICLE DIAMETER
(A)
Fig. 12. PVC particle size distribution transients (Berens,1974).-,
'-, t = 120mins;---, t = 240mino
t = 60 min; -
efficiency in particle formation rema ins high and the particle concentration
pass es its steady-state value. As these parti~les grow they genera te too much
surface to be saturated by the emulsifier and the particle formation rate
become much smaller, perhaps zero.
The effiuent from the reactor carries some of this unsaturated surface and
new emulsifier is being continuously added with the feed. Eventually, the
surface will be saturated and free emulsifier will be available to stabilize new
particles. If the particle formation and growth rates are appropriate,
continuous limit-cycle oscillations can be observed.
The particle formation rate depends on the rate of production of free
radicals and on the relative rates of the various mechanisms in which these
free radicals participate. A high rate of particle formation in the presence of
small amounts offree emulsifier will contribute to unstable cyclic behavior.
The rate of growth of particles depends on the concentration of the
reagents in the particles, radicals and monomer, and on the propagation
rate constant. The gel effect, which causes the termination rate constant to
be lower at higher conversions, can cause higher free radical concentrations
in the particles and thus higher particle growth rates. This effect also
contributes to conversion oscillations.
378
Gary W. Poehlein
C. Control Methods
A number of methods have been considered for controlling transients in
eSTRs. One simple method for trying to handle transients involves the use of
a smalI eSTR as a prereactor. The natural oscilIations and transients from
this reactor are damped by backmixing in the larger reactors that folIow.
Greene el al. (1976) demonstrated that strong initial overshoots and
conversion oscilIatiQns could, in some cases, be avoided by careful start-up.
The reactor was started either with a seed latex from a previous run or fuIl
of distilIed water. Both procedures were successful in achieving steady
operation in some cases. In other cases, however, significant transients were
obse!ved after a smooth start-up.
Hamielec el al. (1980) and Ray and Schork (1980) have worked on
potential feedback control schemes. Hamielec and co-workers have developed an on-line light scattering device that dilutes the latex and measures
scattering at several waveIengths. This data is then used to predict changes
in particIe concentrations and size characteristics.
.
Ray and Schork (1980) have developed an on-line device for measuring
surface tension. A bubble rise technique is used. Ir this method or the light
scattering technique of Hamielec were used in a control system one could
presumably alter the feed rate of a key ingredient, probably emulsifier, to
prevent undesired transients. More work is needed to demonstrate the
potential of these methods.
Berens (1974) and Gonzalez (1974) have shown that the use of a seed
latex in the feed stream can prevent oscilIatory behavior. Berens used a
latex seed generated in a separate reactor. He demonstrated that the
particIe size distribution reached a constant state in a pve polymerization.
Steady states were not observed in unseeded reactions.
Gonzalez placed a continuous tubular reactor in front (upstream) of the
eSTRo In this case the particIe seed was formed in the tube from a recipe
that did not contain seed. Gonzalez found the tube-CSTR system to be
quite stable so long as the conversion in the tu be was adequate to prevent
significant particIe formation in the eSTRo
The use of a particIe seed has the added advantage of aIlowing more
careful control of particIe concentration and reaction rate in the eSTRo
v.
A.
1
~
~
1
10.
379
understood. The factors listed below need to be considered to help minimize the wall polymer problem.
380
Gary W. Poehlein
;;
~
~
~
~
~
~
B.
381
VI. Summary
Some of the fundamentals of reactor systems have been presented and
discussed in this chapter. This material should assist the reactor engineer in
planning appropriate pilot studies for a new or revised process. Equally
important, the concepts reviewed should help the product development
personnel in devising reactor systems and operating procedures that will
produce satisfactory products.
More complete discussions of polymerization kinetics and reactor
modeling are contained in other chapters of this book. These factors will
also be of great value to the reactor engineer and the development chemist.
As the kinetic mechanisms and reaction models become better developed
we can look forward to more efficient research and development programs
for new products and processes.
Acknowledgment
This material is based, in part, upon work supported by the National Science Foundation
under Grant No. CPE-8011455.
References
Aho, C. E. (1958). U.S. Patent 2,831,842 (to E. 1. duPont de Nemours and Co., Wilmington,
Delaware), April 22.
Berens, A. (1974). J. Appl. Polym. Sci. 18,2379.
Calcott, W. S., and Starkweather, H. W. (1954). U. S. Patent 2,384,277 (to E. l. duPontde
Nemours and Co., Wilmington, Delaware), September 4.
382
Gary W. Poehlein
Durbin, D. P., EI-Aasser, M. S., Poehlein, G. W., and Vanderhoff, J. W. (1979). J. App/. Po/ym.
Sci. 24, 703.
Feldon, M., McCann, R. F., and Lundrie, R. W. (1953). Indian Rubber Wor/d 128(1), 51.
Fitch, R. M., and Tsai, C. H. (1971). In "Polymer Colloids" (R. M. Fitch, ed.), Chapters 5 and 6.
Plenum Press, New York.
Gerrens, H., and Ley, G. (1974). Private cornmunication.
Gershberg, D. B., and Longfield, J. E. (1961). Symp. Po/ym. Kinet. Cata/o Syst., 45th AIChE
Meeting, New York Preprint 10.
Ghosh, M., and Forsyth, T. H. (1976). In "Emulsion Polymerization" (1. Piirma and J. L.
Gardon, eds.), pp. 367-378. American Chemical Society Symp. Series 24, Washington, D.C.
Gonzalez, P. R. A. (1974). M.S. Thesis, Chem. Eng. Dept., Lehigh Univ., Bethlehem,
Pennsylvania.
Co.), January
4.
11
Effect of Additives
on theFormation
of M onomer Emulsions
and Polymer Dispersions
J. Ugelstad, P. C. M0rk, A. Berge. T. Ellingsen, and
A.A. Khan
1. Introduction .
11. Thermodynamic Treatment of Swelling
and Phase Distributions
111. Rate of Interphase Transport .
IV. Preparation of Polymer Dispersions .
A. Polymerization with Initiation in Monomer
Droplets Formed by the Diffusion Process .
B. Emulsificationof PolymerSolutions .
V. Effectof Addition of Water-Insoluble Compounds
to the Monomer Phase .
VI. Emulsificationwith Mixed EmulsifierSystems .
List of Symbols .
References
383
384
392
396
396
399
401
408
411
412
l. Introduction*
Quite recently, new methods for the preparation of emulsions by
diffusion processes have been described in a number of papers. The methods
in volved the diffusion of slightly soluble compounds' (ZI) through the
continuous phase to become absorbed into droplets of relatively low
molecular weight; highly insoluble compounds (Z2) or into particles consisting of polymer and highly insoluble compounds (Z2)' The swelling capacity
of such systems was in several cases shown to exceed that of pure polymer
particles by a factor of more than 1000.
* For an explanation
383
EMULSION POLYMERIZATION
Copyright @ 1982 by Academic Preso. loc.
All righto of reproductioo io aoy forro reserved.
ISBN 0-12-556420-1
384
J.Ugelstad et al.
A section of the present chapter is devoted to a rather detailed description of the basic thermodynamic and kinetic principIes involved in the
various swelling procedures. This seems justified in view of the present, and
potential, applications of these principIes to the preparation of emulsions
and polymer dispersions. The swelling procedures developed so far have led
to the first successful methods for preparation of large, monodisperse
partic1es.
Addition of small amounts of Z2 to solutions of polymers has been
shown to facilitate formation of aqueous emulsions of 'such solutions. A
very recent application involves the addition of Z2 to the monomer phase in
an ordinary emulsion polymerization. This resuIts in a decrease in the
concentration of monomer in the polymer partic1es and thereby in a change
in the kinetics of polymerization.
Another, more specific method for the preparation of emulsions of ZI
involves the addition of ZI to a preformed mixture of an ionic emulsifier, a
long-chain fatty alcohol, and water. In this way, the. rapid formation of a
stable emulsion may be obtained at ordinary stirring with relatively modest
amounts of emulsifier. The mechanism of this process is still not satisfactorily explained. AIso, subsequent polymerization (in the case where ZI is a
monomer) may lead to polymerization with initiation in monomer droplets.
The present chapter reviews the most recent work on the abovementioned topics and inc1udes some hitherto unpublished theoretical and
experimental material.
11. Thermodynamic Treatment of Swelling and Phase Distributions
The formation of emulsions and the swelling of polymer-oligomer
partic1es involve the mixing of amorphous substances of different molecular
weights. These processes will be discussed in terms of the original FloryHuggins theory for the free energy of mixing. For a mixture consisting of n
components, the free energy of mixing may be expressed as
n
I1G/RT
n- 1
L...
i=1
A..
'rt
i=lj=i+1
~.
J.NA..
I ro/) Xt,)
(1)
I1G/RTV =
i=1
(<Pi In <Pi/VM)
+ (1/Y.)(<pI
<P2X~,2
<PI<P3X~,3
<PI <P4X~,4
(2)
385
where V.Mis the partial molar volume of component i and V. is the volume
of one segment. According to Eq. (2) the terms representing the combinatorial entropy are inversely proportional to V.M' Therefore, with large
values of V.M,as in the case of the mixing of two polymer compounds, the
combinatorial entropy makes a very small contribution to the free energy of
mixing.
Expressing the interaction energy in terms of Hildebrand solubility
parameters (Hildebrand, 1964),one obtains
Xl.)~. = Vs (>.I
2
>.
J)
/RT
(3)
Accordingly, polymers will only be miscible when they have almost equal
values of '>,or if specific interactions between the two types of polymer
chains are present, in which case the values of xi.j may be negative. In
practice it turns out that the value of xi.j is not only a function of
temperature but also varies with the composition. More recent treatments
applying the lattice theory of Flory-Huggins have taken this into account.
AIso, the original Hildebrand approach has been refined to take into
account the contribution of polar groups and hydrogen bonds to the
solubility parameters. These modifications of the Flory-Huggins theory and
of the solubility parameter concept have made these methods an even more
useful tool in the description of solutions, especially of mixtures containing
polymer compounds. A comprehensive treatment of these extensions of
Flory-Huggins' and Hildebrand's theories, as well as the new "equation of
state" approach of Flory (1965), has recently been published (Shinoda,
1978; Olabisi et al., 1979).
.
The systems considered in the present context may consist of several
phases, some or all of which may be present as droplets. The differential
molar free energy f9r an n-component system is given by
n
dG
= V dP - S dT + L GndNn +
y dA
(4)
386
J. Ugelstad el al.
The condition for equilibrium in the system is that the activity of any
component i is the same in all phases. Choosing the pure component in
bulk as the reference state, the equilibrium conditions may be written
l:1G ia
= l:1G
ib
= l:1G ic = ...
(6)
where a, b, c oo.refer to the phases that make up the system. In practice, the
equilibrium state of one or several of the components may be so slowly
established that a semiequilibrium state is reached. In this state, the
compounds that are more easily transferred between the' phases will have
attained their semiequilibrium distribution. Throughout the following discussion, the low molecular weight compounds that are equilibrated between
the phases will be denoted by Zl, the relatively low molecular weight
compounds which are effectively hindered from being transferred between
the phases are denoted by Z2, and the polymer by Z3'
Emulsions of Zl that are slightly soluble in the continuous phase may be
stabilized toward degradation by diffusion by the presence in the droplets of
Zl of a small amount of a Z2 that is insoluble in the continuous phase. This
idea was first suggested by Higuchi and Misra (1962). Later, a more general
thermodynamic discussion was presented by Ugelstad (1978) and Ugelstad
et a/. (1977, 1978a, 1980a,b; Ugelstad and Hansen, 1976). At the instant the
droplets in such a system are formed, their composition, with respect to Zl
and Z2, is identical. Consider two droplet sizes raoand rg where rao> rg. The
initial driving force for the diffusion of Zt from b to a droplets
expressed as (l:1GlbO - l:1G1ao)is then given by
l:1G tbO
l:1G taO
= 2yV1M(l/rg-
l/raO)
(7)
+ 2YVMlra
= l:1Grb + 2YVMlrb= RTlnaiw
(8)
where the values of the interfacial tension and radii are those at equilibrium (see note on p. 413). The index w is used to denote the continuous
phase, most commonly water. It should however be noted that it mayas well
be a mixture of water and nonaqueous compounds or pure nonaqueous
solvents. The partial molar free energy differences may be obtained from
387
4>2b)
+ 2V1My(1/ra - 1/rb)/RT = O
+ (4)1a -
4>1b)x12
(9)
388
J. Ugelstad el al.
l/rb)/RT
=O
(10)
where raand rb are the radii of the swollen particles and the droplets of Zl
at equilibrium. In case no polymer is present, the terms '1Vithindex 3 should
be omitted.
Figure 1 shows the volume (V1)of Z1 which, according to Eq. (10), may
be absorbed per unit volume of polymer (V3)+ Z2 (V2)for different values of
J2fJ1 as a function of y/ro. Here, y is the interfacial tension at equilibrium
and ro is the radius of the polymer-oligomer particles prior to swelling with
Zl' The values of the interaction parameters are arbitrarily chosen to be
X12 = X13 = 0.5 and X23 = O. Furthermore,
l/rb is considered negligible
when compared to l/ra.
o
3
4
Lag I ~/ro)
(nm-2)
Fig. 1.
Swelling capacity of polyrner-oligomer partic1es as a function of 'Ilrofor various
values of J2fJ. as ca1culated fram Eq. (10): ro = radius of polymer-oligomer partic1e prior to
swelling with ZI, V2 = V3= 0.5, X12= XI3= 0.5, X23= O, V.M= 10-4 m3/mol, T = 323 K.
389
Figure 1 reveals that the swelling capacity of polymer particles containing 50% of a relatively low molecular weight, water-insoluble compound is
drastically increased compared to the case with pure polymer particles of
the same size (J2fJI = 00). It also appears that the swelling capacity of
polymer-Z2 particles is far more dependent on the value of y/re than is the
case for pure polymer particles. This has been verified experimentally by
Ugelstad et al. (1978b).Calculations of the swellingof pure droplets of Z2
with ZI give similar curves (Ugelstad et al., 1980a).
Figure 2 illustrates the swelling capacity dependence on V2 at a constant
value of J2fJI = 5. It appears that even low amounts of Z2 should give a
substantial increasein swellingcapacity at low values of y/re. This effecthas
been verified experimentally (Ugeistad et al., 1979a).
In the course of the swelling process, LlGlasteadily increases. This means
that
the driving
force, LlGlb
LlGla
--
00), steadily
V:1.0
0.67
0.50
0.20
f'r 3
0.10
0.05
0.02
o
..J
21
0.01
O
o
3
t.
Log(~/ro)
Fig. 2. Swellingcapacity of polymer-oligomerparticles versus y/ro for different values
of v2: ro = radius of polymer-oligomer particles prior to swelling with Zt> V2+ V3 = 1,
J2/J, = 5, X12= X13= 0.5, X23= O, V1M= 10-4 m3/mol, T = 323 K.
390
J. Ugelstad et al.
0.002
0.000
- 0.002
- O.OO~
1~:;:-0.006
-0.008
-0.010
-0.012
-0.014
10
1000
10,000
VI/IV2>V3)
Fig. 3. Partialmolar free energy of ZI in swelling partic1es versus swelling ratio, for
various values of "t/ro: XI2 = XI3 = 0.5, X23= O, J2/J1= 5, V2= 2/3, V3= 1/3,
V1M= 10-4 m3/mol,T= 323K.
391
<Pla/<Plb
X23JlfJ2)
<Plb)X13
1/rb)/RT= O
(11)
Eq. (11) with a material balance for Zl allows the caIculation of the radius
of the two types of particIes as a function of the total amount of Zl
absorbed. Excellent agreement between caIculated and experimental results
has been obtained for such systems (Ugelstad et al., 1979b, 1980a).
2. Every compound that may be transferred between the phases within
the time scale of the experiment will equilibrate between the various phases.
This means that if for instance one type of Zl has been incorporated in a
polymer particIe in a first step, and another type of Zl is added in a second
step, the equilibrium' swelling of the particIe will not deviate noticeably from
the one obtained in the presence of either Zl' The reason for this is that the
type of Zl first absorbed will diffuse out of the particIes and become mixed
with the second type of Zb with the concomitant gain in entropy that this
mixing process involves. Similarly, if the system consists of droplets of the
insoluble Z2, or of polymer particIes containing Z2, and if a mixture of two
types of Zl with similar J values is added, the resulting swelling will be
approximately the ~ame as that obtained with either of the two types of Zl'
3. Equation (10) may easily be extended to incIude the swelling of
particIes containing any number of different Z2' Again it should be noted
that for a given particIe size, the presence of a mixture of two or more types
of Z2 will not in crease the degree of swelling beyond that obtained with the
same amount of one type of Z2 as long as the various types have similar J2
values. Points 2 and 3 (above) emphasize the fact that the driving force for
the swelling process is mainly the gain in entropy obtained by mixing of the
various components of the system.
392
J. Ugelstad el al.
Transport
la
= 4nraNaDw(CI- Clas)
(12)
= Clooexp(AGla/RT)
(13)
= Rla = 4nDwraNa[CI -
Cloo exp(AGla/RT)]
(14)
where the concentrations are given in v/v units. In the case where the
solution is saturated with ZI (relative to pure ZI in bulk), Eq. (14) may be
written
dVla/dl
= 4nDwraNaCloo[1- exp(AGIJRT)]
(15)
393
fbNb
- exp(dGla/RT)]
(16)
The values of dGla and dGlb may be calculated using the appropriate forms
of Eq. (5). Two limiting cases of Eq. (16) will now be treated.
Case 1
Equation (16) is then simplified to
dVlJdt
= 4nfaNaClooDw[exp(dGlb/RT)
exp(dGla/RT)]
(17)
(faNa ~ fbNb)
fb = 10 f.1.m,
(18)
394
J. Ugelstad el al.
180
160
140
o.
>
....
120
100
->
80
60
40
20
10
20
30
40
50
80
10
Tim_1
10
10
100
110
120
130
minI
Fig. 4;
Volume of 21 absorbed per unit volume of polymer-oligomer (v.O) versus time,
calculated
from
Eq. (16): Dw = \O-10 m2jsec,
CI", = 6 X \0-4 m3jm3,
r~ = 2 x \0-7 m,
V2a = V3a = 0.5 V.o,
Ya= Yb= 5 mNjm,
X12 = X23 = 0.5,
X23 = O,
J2fJI = 3,
VIM= \0-4 m3jmol, T=
are assumed
to be present (r~
to
the case exemplified by Eq. (17). Curve B is caIculated for a case where
the concentration of 21 in the aqueous phase is constant, equal to the
saturation concentration relative to aplane phase (CI",).
In curve C, a large excess (Vb;v..= 105) of lO-jlm droplets
of 21 is
395
D
14
12
.=
10
...
..
..
8
E
111
e
6
Time (hrl
396
J. Ugelstad el al.
IV.
A.
397
1973, 1976, 1978). Low molecular weight, water-insoluble compounds (oligomers) may be incorporated into the partic1es during their preparation by
addition of a chain-transfer agent during polymerization (Ugelstad, 1978;
Ugelstad et al., 1978b), or may alternatively be introduced into the preformed polymer partic1es. One way of doing this is to swell the polymer
particles with monomer and initiator and carry out the subsequent polymerization under conditions that lead to the formation of oligomers.
Another procedure requires the seed in a first step to be swollen with a
low molecular weight, water-insoluble Z2' (Ugelstad et al., 1978a,b, 1979a,
Fig.6.
Scanning
electron micrograph
of monodisperse
5.3-Jlm polystyrene
particles.
398
J. Ugelstad el al.
Fig. 7.
Electron micrograph of particles resulting from polymerization of a polystyrene
seed (500 nm) swollen with a mixture of acrylic monomers (magnification 4080 x ).
399
The recent trend to avoid organic solvents has led to substantial elforts
to prepare aqueous dispersions of polymers that cannot be prepared
directly by emulsion polymerization. This includes such important polymers as epoxy resins, polyurethanes, and silicones. One method. to achieve
this is by emulsifying solutions of the polym~r by homogenization, followed
by evaporation of the solvent. Most often this method gives particles in the
range of 1 to 211m. 1'his may be a disadvantage in that the dispersion is
unstable toward settling of the particles on storage, and furthermore, large
particles tend to give inferior film properties. Vanderholf et al. (1979)
showed that addition of small amounts of Z2 to the solution of a polymer
before homogenization led to much more finely dispersed droplets of the
polymer solution. A number of dilferent polymer dispersions have since
been prepared (EI-Asser et al., 1977a,b; Vanderholf et al., 1978; Miscra et
al., 1978).
Ugelstad (1978) and Ugelstad et al. (1978a) have explained the elfect of
adding small amounts of Z2, along the same lines as the thermodynamic
treatment above. In the present discussion the previous explanation of the
effect of Z2 is somewhat extended.
The solvents applied in practice are all slightly water soluble, i.e., Zl
types. At a given composition of the solution of polymer (Z3) and solvent
(Zl) there is a theoretical minimum in the droplet size that may be obtained
before phase separation should occur. A droplet of this minimum size
400
J. Ugelstad el al.
contains a given amount of polymer. The minimum droplet size corresponds to the swelling capacity of a partic1e consisting of this amount of
polymer. Therefore, the minimum droplet size will in crease with increasing
amount of solvent (ZI)' If a small amount of Z2 is added before homogenization, the droplets formed contain a small amount of Z2 in addition to Z3'
The swelling capacity of a partic1e in which even a small amount of Z3 is
replaced by Z2 is considerably higher than that of pure Z3' Therefore, in the
presence of Z2 the. minimum diameter that may be obtained with a given
amount of solvent (ZI) may be drastically reduced. 'Calculation of this
minimum diameter may be carried out by applying Eq. (10). An example of
such a calculation is given in Table 1. It appears from the table -that the
minimum diameter in creases with increasing amounts of ZI' It also appears
that even with small amounts of Z2 it is possible to obtain small partic1es.
This is also the case where without Z2 one gets phase separation even at
infinite diameters of the partic1es.
During homogenization of a solution of polymer (Z3)' solvent (Z), and
water, droplets of different sizes are formed. Becase of the interfacial
energy difference, a transport of ZI from the smaller to the larger droplets
will take place. If droplets smaller than the critical size are formed, these
droplets, especially, will rapidly lose ZI to the surrounding larger droplets.
The viscosity within these small droplets will increase and a further
degradation may be mechanically hindered. In the presence of Z2, two
effects that will facilitate subdivision of the emulsion may be encountered.
1. The degree of diffusion between small and large droplets will be
reduced because the presence of Z2 means that a counteracting concentration potential is more rapidly established.
TABLEI
The Minimum Diameter That May Be Obtained by Homogenization of a Polymer
Dissolved in a Solvent (Z,) as a function of Volume V, of the Solvent
and Volume V2 of a Z28.b
X = 0.5
= 0.3
O
0.01
0.02
0.10
10
10
21
20
19
16
110
103
96
66
415
370
334
190
48
44
40
24
454
338
269
105
2980
1500
1005
280
147
108
85
32
From
Ugelstad
b XI2 =
O, XI3
T= 300 K.
X =0.6
(1978)
= X23
with
10
2545 3 x 105
146
356
permission.
401
V.
4>f3X13
= In 4>lb + (1 -
+ 2VMy/r3RT
JjJ2)4>2b
4>ibXl2
+ 2V1My/rbRT
(19)
given initial volume fraction of Z2, 4>Rb' in the monomer phase on 4>13/4>13
(V2= O)(Le.,the ratio of the volume fractions of monomer in the partic1es
with and without addition of Z2 to the monomer phase) for various values
of vNV30,the initial.ratio of monomer to polymer.
In order to illustrate the etfects one may expect and the problems in
interpretation of experimental results, some ca1culationsshowing the etfect
of addition of Z2 to the monomer phase have been carried out using
Eq. (19).
In Figures 8 and 9, 4>13/4>13
(V2= O)is given as a function of VIO/V30for
various values of 4>fb = V2/(V2+ VIO)and V2, respectively.In these ca1culations, the radius of the polymer partic1es is assumed to be r~ = 5 x 10- 8 m
400
J. Ugelstad el al.
contains a given amount of polymer. The minimum droplet size corresponds to the sweIling capacity of a particIe consisting of this amount of
polymer. Therefore, the minimum droplet size wiII increase with increasing
amount of solvent (21), If a smaIl amount of 22 is added before homogenization, the droplets formed contain a smaIl amount of 22 in addition to 23'
The sweIling capacity of a particIe in which even a smaIl amount of 23 is
replaced by 22 is considerably higher than that of pure 23' Therefore, in the
presence of 22 the minimum diameter that may be obtained with a given
amount of solvent (21) may be drasticaIly reduced. Calculation of this
minimum diameter may be carried out by applying Eg. (lO). An example of
such a caIculation is given in Table 1. It appears from the table 'that the
minimum diameter increases with increasing amounts of 21, It also appears
that even with smaIl amounts of 22 it is possible to obtain smaIl particIes.
This is also the case where without 22 one gets phase separation even at
infinite diameters of the particIes.
During homogenization of a solution of polymer (23), solvent (21), and
water, droplets of different sizes are formed. Because of the interfacial
energy difference, a transport of 21 from the smaIler to the larger droplets
wiII take place. If droplets smaIler than the critical size are formed, these
droplets, especiaIly, wiII rapidly lose 21 to the surrounding larger droplets.
The viscosity within these smaIl droplets wiII increase and a further
degradation may be mechanicaIly hindered. In the presence of 22, two
effects that wiIIfacilitate subdivision of the emulsion may be encountered.
1. The degree of diffusion between smaIl and large droplets wiII be
reduced because the presence of 22 means that a counteracting concentration potential is more rapidly established.
TABLEI
The Minimum Diameter That May Be Obtained by Homogenization of a Polymer
Dissolved in a Solvent (Z1) as a function of Volume V1 of the Solvent
and Volume V2 of a Z2a.b
X = 0.3
x = 0.5
x = 0.6
>\'
10
10
O
0.01
0.02
0.10
21
20
19
16
110
103
96
66
415
370
334
190
48
44
40
24
454
338
269
105
2980
1500
1005
280
147
108
85
32
From Ugelstad
b X12
10
2545 3 x 105
146
356
= O, XI3= X23= X, V2+ V3= 1, J2/11= 5, VIM= 10-4 m3/mol, y = SX10-3 N/m,
T= 300 K.
401
v.
+ 1/J3a +
I/Jfax13+ 2VIMy/raRT
(19)
402
J. Ugelstad et al.
1.0
0.9
0.8
-o
0.7
N
>
0.6
"
o
02b
"
61
.....
!!
61
0.2
o
02b =0.3
0.5
0.1,
0.3
0.2
0.1
1.0
2.0
5.0
3.0
VI/V3
Fig. 8.
V2 O versus initial ratio of ZI to Z3 for various values of the initial volume fraction of Z2 in
the b phase: r~ = 5 x 10-8 m, X12 = X13 = 0.5, V1M= 10-4 m3/mol, V2M= 3 X 10-4 m3jmol,
y = 5 mNjm,J2jJ = 3, T = 323K.
and the monomer droplet radius is set equal to infinity. The values of X12
and X13are arbitrarily chosen to be 0.5 and the molar volumes of 21 and 22
are set equal to 10-4 and 3 x 10-4 m3/mol, respectively. The interfacial
tension at equilibrium is assumed to be 5 mN/m and T = 323 K.
It would
appear
<PlaNla
(V2
= O)
passes
through
0.9
V2=0.S
V2=1.0
0.8
403
0.7
11
N
0.6
ni
SI
-ni
SI
o.S
0.4
03
0.2
01
2
IC
1,
14
16 18 20
v / V3
Fig. 9.
cP.Jc/>1a
(V2= 0.)versus vNvJoat various 'values of the volume (V2) of Z2 in the b
phase. V~ = 1. Parameter~ equal to those in Fig. 8.
+ (VIO -
ra = r~[(V3a
VI
= V1a+
Vlb)(djd3)
+ Vla)/V30]
1/3
V1b
404
J. Ugelstad el al.
8
7
6
-...
E S
't:I
...
o
E 4
..
u 3
0.1
0.2
0.3
0.4
0.5
Q6
0.7
0.8
0.9
10
Fig. 10.
Concentration of monomer in the polymer particJes during polymerization as
a function of conversion when Z2 is present in the monomer phase: V2= 0.1, V30= 1, VIOas
indicated, r~= 5 x 10-8 m, VIM= 10-4 m3/mol, J2/JI = 3, X12= X13= 0.5, y = 5 mN/m,
T= 323K.
as 1J= (V? -
0.16, respectively.
It would appear that the initial rate is lower the higher the value of 4>fb'
as would be expected under Smith- Ewart n conditions from the lowering of
the concentration of monomer in the particles. At increasing conversions it
405
- -
- -
150
o'"
J:
....
E
<f)
o..
I
100
DI
-~
50
.::.
H20
10"10Hexadecane
O
0.6 g K2S20a/dm3
20"10 Hexadecane
H20
10
~--3
0.6 g K2S20a/ dm
~-l
60
120
lao
21,0
300
360
TIME
420
I,ao
51,0
600
(min)
Fig. 11.
Conversion versus time curves for seeded emulsion polymerizations
with hexadecane present in the monomer phase. T = 333 K.
of sty~ene
<P'a/4J,a(V2
= O) is
is
406
J. Ugelstad et al.
300
oN
:x:
PI
e
~ 200
(.)
>
o.
el
100
100
Fg. 12.
chloride
200
TIME (min)
300
with hexadecane
present
in the monomer
phase.
[K2S20SJ
= 2.7
x 10-3 moljdm3
Vs = V30(3r;c5
- 3rac52 + c53)/(r~)3
= V1sIVs
<Pla
where VIS is the volume of monomer in the shell of thickness c5and total
volume Vs. The other parameters used in calculating the curves in Fig. 13
are the same as in Fig. 8.
Comparing Fig. 13 with Fig. 8 it appears that in this case one may
expect a considerably larger effect of the additive. Thus, for VIO
fV30= 0.05
and <Pfb= 0.5, the value of <Plal<Pla(V2
= O) is 0.75 when the swelling is
11.
1.0
407
..
0.9
0.8
a7
-o
"
N
>
..'"
ii
.....
02b
0.2
0.6
0.5
SI
0.4
0.3
02
0.1
05
1S
10
2.0
V~/V3
Fig. 13.
to polymer.
Other parameters
408
J. Ugelstad el al.
VI.
409
degradation by diffusion. The fact that with fatty alcohols below a certain
chain length one gets very unstable emulsions indicates that also in this case
the degradation by diffusion is a major cause of instability and that a fatty
alcohol may act as a "Z2" whose stabilizing effect is determined by its
solubility in the continuous phase. This does not exc1ude the possibility that
a condensed layer of emulsifier and fatty alcohol is formed at the interface.
Such a complex will in crease the stability toward coalescence and may also
decrease the rate of degradation by diffusion by forming a barrier to
transport of Z through the interface (HalIworth and Carless, 1974). Such a
hindered transport through the interface was also suggested as an explanation for the decrease in the rate observed in the seeded polymerization of
vinyl chloride when mixtures of ionic emulsifier and fatty alcohols were
applied (Ugelstad el al., 1973b).
The ease of formation of these emulsions is still not satisfactorily
explained. When an ionic emulsifier and a fatty alcohol are dissolved in
water, mixed micelIes, liquid crystals, and crystalline, rodlike partic1es (1-2
..mlength, 0.1-0.2 ..mdiameter) are formed (Chou el a/., 1980). Chou el a/.
carried out an extensive study of the formation of emulsions with mixed
emulsifiers. The emulsifier systems were prepared by heating water with
fatty alcohol and cationic emulsifier. To this mixture styrene was added
gradualIy, and the resulting emulsion was examined spectroscopicalIy and
by mesurement of the electrical conductivity. The conductivity of the mixed
emulsifier system was found to be considerably lower than that of an
equimolar solution of pure ionic emulsifier, and more so the higher the fatty
alcohol-emulsifier
ratio. This result is expected since the complexation of
the emulsifier with the fatty alcohol leads to a reduction in the amount of
free emulsifier in the' aqueous phase. When styrene was added, the conductivity at first decreased and then increased on further addition of styrene
before the curve leveled off or started to decrease slightly. With high fatty
alcohol-ionic
surfactant ratios, when the conductivity before addition of
styrene was at its lowest, the conductivity increased from the start when
styrene was added.
The most stable emulsions were formed with a molar ratio of fatty
alcohol to ionic emulsifier of about 3: 1, whereas higher ratios gave more
unstable emulsions. Similar results were obtained using systems with anionic emulsifier (Hansen et a/., 1973).
Chou el al. observed that addition of relatively smalI amounts of styrene
caused the rodlike crystals to disappear. They did not come to any
conc1usion about the mechanism of the emulsification process but suggested
that the droplets may have been formed by diffusion of styrene into the
crystalline, rodlike partic1es.
Chou el al. ascribed the stability of the emulsions to the formation of an
emulsifier-fatty alcohol complex at the surface of the droplets. As discussed
410
J. Ugelstad el al.
above, it seems plausible that even if such a complex may be formed and
even if it increases stability toward coalescence, the major stabilizing effect
is the prevention of degradation by diffusion, caused by the fact that a longchain fatty alcohol may act as a Z2'
Another explanation that has been advanced by Azad et al. (1975)
implies that when stirring the mixture, the mixed micelles collide with
droplets of Zt with the result that fatty alcohol is transferred to the
droplets. During this process, a transient high concentration of fatty alcohol
in the outer layer of the droplets of Zt may cause a spontaneous emulsification of parts of the droplet. It is well known that spontaneous
emulsification may be obtained with 20-30% fatty alcohol (Schulman et al.,
1940).
Another explanation that has more recently been advanced by Ugelstad
and Fitch (to be published) is based on the principIe of formation of
emulsions by diffusion. During stirring, droplets of Zt with fresh surface
and therefore a high interfacial tension are formed. Diffusion of Zt from the
surface of these droplets into neighboring mixed micelles takes place
rapidly. The swelling of the mixed micelle is determined by its fatty alcohol
content, acting as a Z2'
As the mixed micelles are very small (10-20 nm) the initial swelling must
be very limited. To be able to absorb more of Zt, the initial small droplets
must be furnished with more emulsifier, and even more importantly, with
more fatty alcohol (Z2)' This may be achieved by coalescence of initial
droplets or by absorbtion of mixed micelles from the surroundings. The
assumption that the emulsification takes place by a diffusion process seems
to be supported by experiments with mixed systems of ionic emulsifier and
fatty alcohol and various dispersed phases, showing that a necessary
condition for a rapid emulsification is that the compound to be emulsified
have slight water solubility. Furthermore, it has been observed that if even
small amounts of Z2 are added to Zt before addition to the water-mixed
emulslfier system, the extent of emulsification is reduced and the resulting
emulsion becomes less stable.
Another interesting phenomenon is that post-addition of small amounts
of a highly insoluble compound to an emulsion leads to a rapid degradation
of the emulsion (J. Ugelstad, unpublished results). It has also been observed
that if the molar ratio of fatty alcohol to emulsifier is increased above four,
at a constant concentration of emulsifier, the stability of the emulsions
decrease (Lange et al., 1973; Chou et al., 1980). In all these cases, a bulk
phase containing Z2 will be present, even if an emulsion of Zt is formed.
Thus, the emulsion may degrade by diffusion of Zt from the emulsion
droplets to the bulk phase containing Z2'
The increase in conductivity observed by Chou et al. at increasing
411
List o Symbols
ZI
Z2
aw
a and b
A
CI
Cta.
Clo:>
di
Dw
dG
t:.G
t:.Gi
density of component i
diffusion constant of Z. in the continuous phase
differential molar free energy
free energy of mixing
partial molar free energy of mixing for any component
surface) as reference state
partial molar
surface
from
with pure ZI in
has aplane
t:.Gix
t:.GxO
i
j
Ji
N.
Nb
Ni
P
type of
number
number
number
number
pressure
rO
x
R
Rix
(dV;Jdt)
S
t
T
V;
"Ix
"1M
V;~
t:.Gi in phase x (x = a or b)
initial t:.Gix before any mass transport
type of component
component
of segments
of partic1es
of partic1es
of moles of
rate
has occurred
between phases
in component i
with radius r.
with radius rb
component i
of transport
ZI from
the bulk
of the continuous
phase
temperature
volume of component i
volume of component i in phase x
partial molar volume of component
initial "Ix before any mass transport
i
has taken place between phases
to the
surface
412
v.
J. Ugelstad et al.
Vs
segment volume
total volume of outer shell (of thickness
vIS
i'
b
volume of compound
interfacial tension
shell thickness
bi
r
rPi
rPix
b) of a particle
rPi~
initial
Xi.;
xi.;
Xi)Ji
rPix
= interaction
parameter
i with component
References
Azad, A. R. M., Ugelstad, J., Fitch, R. M., and Hansen, F. K., (1975). Paper presented at Symp.
Emul. Polym. Philadelphia, Pennsylvania Vol. 1. American Chemical Society Ser. 24,
Washington, D.C.
Azad,A. R. M., Nomura, M, and Fitch, R. M. (1980). Org. Coat. Plast. Chem. Preprints43, 537.
Chou, Y. J., EI-Asser, M. S., and Vanderhoff, J. W. (1978). Paper presented at Symp. Phys. Chem.
Properties Colloidal Partides, Miami.
Chou, Y. J., EI-Asser, M. S., and Vanderhoff, J. W. (1980). "Polymer Colloids" (R. M. Fitch,
ed.), Vol. 11.Plenum Press, New York.
Davies, S. S., and Smith, A. (1974). Paper presented at Symp. Theory Practice Emul. Technol.
Brunel Univ. p. 325. Academic Press, New York.
EI-Asser, M. S., Vanderhoff, J. W., and Poehlein, G. W. (1977a). Preprints Am. Chem. Soc. Div.
Organ Coating Plast. Chem. 37, 92.
EI-Asser, M. S., Miscra, S. C., Vanderhoff, J. W., and Manson, T. A. (1977b). J. Coatings
Technol. 49, 71.
Flory, P. J. (1965). J. Am. Chem. Soco 86, 1833.
Ghanem, A. H., Higuchi, W. l., and Simonelli, A. P. (1969). J. Pharm. Sci. 58, 165.
Ghanem, A. H., Higuchi, W. l., and Simonelli, A. P. (1970a). J. Pharm. Sci. 59, 232.
Ghanem, A. H., Higuchi, W. l., and Simonelli, A. P. (1970b). J. Pharm. Sci. 59, 659.
Goodwin, J. W., Hearn, J., Ho, C. C., and Ottewill, R. H. (1975). Br. Po/ym. J. 5, 347.
Goodwin, J. W., Hearn, J., Ho, C. C., and Ottewill, R. H. (1976). Colloid Po/ym. Sci. 60,173.
Goldberg, A. H., and Higuchi, W. 1. (1967). J. Pharm. Sci. 56, 1432.
Goldberg, A. H., Higuchi, W.I., Ho, N. F. H., and Zografi, G. (1969). J. Pharm. Sei. 58,1341.
Hallworth, G. W., and Charles, J. E. (1974). Paper presented Symp. Theory Practice Emul.
Technol., Brunel Univ. p. 305. Academic Press, New York.
Hansen, F. K., Ugelstad, J., and Lange, S. (1973). Abstr. Seand. Symp. Surface Chem., 5th, Abo.
Fin/and.
Hansen, F. K., Bauman Ofstad, E., and Ugelstad, J. (1974). Paper presented at Symp. Theory
Practice Emul. Techno/., Brunel Univ. p. 1. Academic Press, New York.
Hansen, F. K., and Ugelstad, J. (1979). J. Po/ym. Sci. Po/ym. Chem. Ed. 17, 3069.
Higuchi, W. J., and Misra, J. (1962). J. Pharm. Sci. 51, 459.
Hildebrand, J. H. (1964). In "The Solubility of Non Electrolytes" Dover, New York.
Lange, S., Ugelstad, J., and Hansen, F. K. (1973). Abstr. Scand. Symp. Surface Chem., 5th, Abo,
Fin/and.
Mfutakamba, H., Mork, P. C., and Ugelstad, J. (1979). Unpublished results.
413
Miscra, S. c., Manson,J. A.,and Vanderhoff,J. W (1978). Preprints Am. Chem. SocoDiv. Organ.
Coatings Plast. Chem. 38(1), 213.
Napper, D. H., Netschey,
Netschey, A., Napper, D.
Olabisi, O., Robeson, L.
Academic Press, New
Schulman, J. A., and Cockbain, E. G. (1940). Trans. Faraday Soco 36, 651.
Shinoda, K. (1978). In "Principies of Solutions and Solubility." Decker, New York.
Smoluchowski,
M. V. (1918). Z. Phys. Chem. 92, 129.
Ugelstad, J., Merk, P. c., Dahl, P., and Rangnes, P. (1967). J. Polym. Sci. Part C 27,49.
Ugelstad, J., EI-Asser, M. S., and Vanderhoff, J. W. (1973a). J. Polym. Sei. Polym. Letl. Ed.
11, 505.
Ugelstad, J., Flegstad, H., Hansen, F. K., and Ellingsen, T. (1973b). J. Polym. Sei. Part C42, 473.
Ugelstad, J., Hansen, F. K., and Lange, S. (1974). Makromol. Chem. 175,507.
Ugelstad, J., and Hansen, F. K. (1976). Rubber Chem. Technol. 44(3), 536.
Ugelstad, J., Hansen, F. K., and Kaggerud, K. H. (1977). Faserforsch. Textiltech.-Z.
Polym.
Forsch. 28, 309.
Ugelstad, J. (1978). Makromol. Chem. 179, 815.
Ugelstad, J., ElIingsen, T., and Kaggerud, K. H. (1978a). Int. Conf Organ. Coatings Sci. Technol.
Preprillts, p. 425.
Ugelstad, J., Kaggerud, K. H., and Fitch, R. M. (1978b). Symp. Phys. Chem. Properties Colloidal
Particles. Miami.
Ugelstad, J., Kaggerud, K. H., Hansen, F. K., and Berge, A (1979a). Makromol. Chem. 180,737.
Ugelstad, J., Merk, P. C., and Kaggerud, K. H. (1979b). Paper presented at Int. Conf Surface
Co/loid Sci. 3rd. Stockholm Abstracts, p. 344.
Uge1stad, J., Merk, P. C., Kaggerud, K. H., ElIingsen, T., and Berge, A. (1980a). Adv. Co/loid
Interface Sci. 13, 101.
Ugelstad, J., Merk, P. C., Hansen, F. K., Kaggerud, K. H., and Ellingsen, T (1980b). J. Pure
Appl. Chem. (in press.)
Ugelstad, J., Merk, P. c., and Khan,
43,514.
Ugelstad, J., Ellingsen, T., and Kaggerud, K. H. (1980). Adv. Organ. Coatings 2, 1.
Ugelstad, J., Kaggerud, K. H., and Fitch, R. M. (1980). "Polymer Colloids" (R. M. Fitch, ed.),
Vol. n. Plenum Press, New York.
Vanderhoff, J. W., Vitkuske, J. F., Bradford, E. B., and Alfrey, T. (1956). J. Polym. Sei 20,225.
Vanderhoff, J. W., van den Hui, H. J., Tausk, R. J. M., and Overbeck, J. Th. G. (1970). In "Clean
Surfaces. Their Preparation
and Characterization
for Interfacial Studies" (G. Goldfinger,
ed.), Dekker, New York.
Vanderhoff,
J. W., EI-Asser,
Vanderhoff,
J. W., EI-Asser,
12
Radiation- lnduced
Emulsion Polymerization
Vivian T. Stannett
415
1. Introduction .
418
11. LaboratoryResults with Different Monomers
418
A. Styrene
424
B. Acrylonitrile .
425
C. Methyl Methacrylate
428
D. VinylAcetate.
429
E. VinylChloride
431
F. Miscellaneous Monomers .
433
111. Copolymerizations .
433
A. Random Copolymers .
434
B. Block and Graft Copolymers .
IV. Radiation-Induced Emulsion Polymerization Using
436
Electron Accelerators
437
V. Pilot Plant and Related Studies
437
A. Introduction .
437
B. VinylAcetate.
440
C. Ettylene .
442
D. Tetrafluorethylene .
E. Tetrafluorethylene-Propylene and Other
444
Copolymers .
447
References
l. Introduction
The fitst.reported work on the radiation-inducedpolymerization
of a vinyl monomer in emulsion appears to be that of Ballantine (1954) at
Brookhaven National Laboratory in 1953. This work was confined to
styrene and utilized a tantalum y source with a dose rate of only 60,000
radjhr. Smooth conversions at high rates and high molecular weights to more
than 90% were obtained at 25C with about 2 Mrad. Soap concentration,
monomer to water ratio, and temperature were varied. These and other
results obtained with styrene, the classical monomer for such studies, wil\ be
described later. The motivation for the work was the considerable industrial
415
EMULSION POLYMERIZATION
Copyright ~ 1982 by Aeademie Pros,. Ine.
An right, or reproduetion in any form reserved.
ISBN 0-12-556420-1
416
Vivian T. Stannett
interest in emulsion polymerization and the high yield of free radical s from
the radiolysis of water. This latter feature in addition to the high kinetic
chain lengths associated with emulsion po!ymerization kinetics has led to a
continued interest in radiation as the initiator for a number of monomers.
There are a number of other practica! advantages associated with the use of
radiation-induced emulsion polymerization. These will be described in the
next few paragraphs.
Radiation, particularly using isotopes such as 60Co, can give an essentially unlimited range of radical fluxes from zero to those equivalent to
many moles per liter of chemical initiators, which would c1early be impracti. cal. Associated with this range is the ease with which the fluxes can be
monitored during the course of the polymerization reaction as the kinetics
and other considerations demando The fluxes can be programmed and built
into modern process control techniques leading in principIe to the orderly
control of molecular weight and partical size distributions and to the
elimination of residual monomer. Alternatively, there need be no change in
the radical flux during the reaction, Le., no initiator exhaustion, and the
initiation process can be started and stopped at will by simply dropping
away or raising the radiation source. This could be useful for changes in the
feed in copolymerizations and, in principIe, for block and graft polymerization in emulsion. Initiation with radiation is essentially temperature
independent. This leads to comparatively low temperature dependencies for
the overall reaction, the activation energy dropping from about 20 for
chemical initiation to only about 7 kcaljmol. This difference, especially
when coupled with the ease of removing the initiation source, makes the
possibility of exothermic, runaway reactions extremely low indeed.
Furthermore, initiation is not a function of polymerization temperature and
hence is completely uncoupled, in direct contrast to chemical initiation.
The lack of any activation energy with the initiation reaction has already
been emphasized. This also means that polymerizations can be conducted
at will at any temperature and at any initiation rateo In principIe, this can
be, and is indeed, accomplished with chemical initiation inc1uding redox
systems. In practice, however, especially low-temperature initiation is not
easy to achieve and control, particularly with polar monomers. The lowtemperature polymerization of vinyl acetate is of particular interest. It is
well known that monomers such as vinyl chloride and vinyl acetate, whose
degree of polymerization is mainly governed by chain transfer to monomer,
have a negative dependence of molecular weight on temperature. To
achieve high molecular weights, therefore, low-temperature polymerizations
are necessary.
The radicals produced by the radiolysis of water are hydrogen atoms
and, mainly, hydroxyl radicals. These are neutral and highly reactive
417
radicals which lead to efficient initiation and there are no electrolytes such
as arise from most chemical systems; this leads to somewhat higher surface
tension latices with lower ionic strengths and higher pH values, often
eliminating the need for adding buffers as, for example, is the case with
potassium persulfate initiation. No contamination with residual initiator
frag~ents occurs. Another advantage that could conceivably become of
considerable importance is that radiation is ideal for initiating the polymerization of systems using cationic soaps. which often interact with chemical
systems. This feature was pointed out and investigated many years ago in
Japan by Inagaki et al. (1960a,b,c); such cationic lattices were found to be
quite stable. There are disadvantages to the use of radiation; for example,
there are no ionic end groups such as arise from persulfate initiation. These
could, in principie, lead to some stabilization of the resulting latex. In
addition, the radiation attacks all the components including the emulsifier
and the polymer as it is formed. The former could lead to a very small
amount of grafting but is probably a negligible factor. The latter, however,
leads to branching and a very small loss of acetate groups. It is doubtful,
however, under the conditions that would prevail in an industrial process
where low total doses would be used, that any of these problems could
cause difficulties. A more important problem, which could arise in flow
reactors or even kettle systems, is possible build-up of polymer on the walls
of the reaction vessel caused by diminishing flow rate approaching the
walls. This could also arise because the radiation is at its most intense near
the radiation elements. However, with suitable reactor designs this problem
could also be eliminated, as well as the general coating problems that can
arise with chemical initiation. In fact, with kettle reactors these effects have
not been observed, at least over the limited time periods involved in pilot
plant studies. With wide-tube flow reactors some coating of the walls has
been observed, but again it is probably not a major problem, at least over
reasonable time periods.
In general, for most investigations, 60Co }' radiation has been used
because of its high degree of penetration and the comparative ease of
estimating dose-depth characteristics and because radical fluxes comparable
to those used witr. chemical initiation can easily be achieved. There have
also been a few, comparatively brief, studies using electr.on accelerators to
initiate emulsion polymerization in emulsion. These have mainly been
conducted in Japanese laboratories.
From a fundamental point of view it is interesting to speculate on the
differences that could exist between the kinetics of emulsion polymerization
initiated by }' radiation and those of a conventional chemical initiator with,
for example, potassium persulfate. At dose rates giving a free radical flux
comparable to those achieved with chemical initiation any differences
418
Vivian T. Stannett
A. Styrene
In his early work, Ballantine (1954) used 1% of an amine long-chain
alcohol sulfate (Duponol G) as the emulsifier and monomer to water ratios
of 1:7 and 1:9 at 25, 35, and 45C. The overall activation energy for the rate
was found to be 3.7 kcaljmol. The molecular weights decreased slightly with
conversion but were very high, close to or more than one million. The rates
were quite high reaching, for example, 54% per hr at 45C with molecular
weights up to 2 million. Abkin et al. (1959) presented the next published
work, followed closely by Inagaki et al. (1960a,b,c) and Bradford et al.
(1961). Hummel et al. (1962) conducted a long series of studies with styrene
and other monomers beginning in 1962. There was an interesting paper by
Acres and Dalton in 1963. The present author and his -colleagues began
419
work on the radiation-induced polymerization of styrene and other monomers in 1964, with the first published report in 1967.
Styrene is one of the best-behaved of monomers in that it tends to follow
most closely classical Ewart and Smith (1948) kinetics. There are many
exceptions, however, even with potassium persulfate, the standard chemical
initiator. The best agreement was found with potassium persulfate as the
initiator and sodium lauryl sulfate as the emulsifier. Until recentIy (Garreau
et al., 1979), the various studies were mainly conducted with other systems
and for reasons other than to check whether there are, indeed, any real
differences between radiation and chemieal initiation in such an ideal
system.
Abkin et al. (1959) found the dependence of rate on the dose rate to be
close to 0.4, the Classical Smith-Ewart (1948) Case 2 value. The molecular
weights wc-reclose to those caIculated from the chain transfer to monomer
values. However, they found close to zero activation energy for the rates
and substantial post-irradiation polymerization. It was pointed out that the
radiation could also generate hydrogen peroxide as an additional initiation
source. One cationic and three anionic emulsifiers were studied, and the
cationic emulsifier gave a very low rateo Inagaki et al. (1960a,b,c) concentrated their studies on the elfect of different emulsifiers: anionic, nonionie, alld cationie. In all cases the rate increased with increasing emulsifier
concentration. The rate, in general, was greatest with anionic emulsifiers
and least with cationic. The rate and molecular weight dependencies (on the
soap concentrations) were in the range 0.14-0.40. Cationie lattices are not
easy to prepare with chemical initiators, and this earIy work pioneered the
use of y irradiation for their production. Stable lattiees with up to 50%
solids content were made with an overall activation energy of 6.2 kcaljmoI.
Bradford et al. (1961) conducted a detailed study using the dihexyl ester
of sodium sulfosuccinate as the emulsifier. Their main emphasis was on the
competitive growth of particles using seeded systems and was related to
their development of highly uniform particle size polystyrene lattices (e.g.,
Bradford and Vanderhoff, 1955). The paper also included, however, a rather
detailed kinetic study leading to values of the termination rate constant kl'
the average nUII).berof radieals per particle , and the activation energy
associated with die molecular weights produced (7.2 kcaljmol). Rather
similar results were obtained with 60Co y radiation as with potassium
persulfate initiation. However, the number of particles decreased with
temperature using radiation but increased with potassium persulfate. This
was attributed to the increase in the rate of initiation with temperature in
the case of persulfate initiation. In general, their values for the activation
energy, the efficiencies of initiation, and the rate constants for termination
werein good agreementwith those obtained later by Garreau et al. (1979).
420
Vivian T. Stannett
12.
Radiation-Induced
421
Emulsion Polymerization
100
75
c:
.o
Q)
>
c:
o
50
eQ)
(,)
...
,f
25
Fig. 1. Typical styrene conversion curve at 30C (dilatometer trace). Dose rate
0.02 Mradjhr, 25%solids, 6.7%sodium lauryl sulfateon total volume.
of Styrene in Emulsion
Preparation
y-Radiation
Property
pH
Surface tension (dyne/cm)
Particle size (Jm)
Molecular weight (viscosity)
K2S20S
at 600C.
3.8
61.1
0.13
2,700,000
9.2
68.9
0.10
1,664,000
7.9
69.0
0.07
413,000
.25% solids and 6.7% sodium lauryl sulfate based on water content
422
Vivian T. Stannett
B
24....
22
20
18L
16
14
e
:o
12
10
o
C 8
6,
:1
250
-Diameter
2000
(,.\.)
Fig. 2
Differential particle size distribution of polystyrene lattices initiated by: A,
radiation OC; B, radiation 600C; C, potassium persulfate 60C (Araki et al., 1967).
ratio and to the 0.4 power of the dose rate (Fig. 3). The activation energy
(Fig.4) associated both with the rate and the molecular weight was
7.9 :t 0.6, in good agreement with Van der Hoff et al. (1958) and with the
literature values for Ekp' This is reasonable since little or no activation
energy is involved in the initiation and termination steps. The dependence
of the rate and the molecular weight on the emulsifier concentration was
about 0.7, in reasonable agreement with the Smith-Ewart value of 0.6. The
number of particles, however, was of the correct order of 0.6 with respect to
the emulsifier concentration. The termination rate constants were caIculated
using the method of Van der Hoff (1958). The values found were in excellent
agreement with those of Van der Hoff for persulfate initiation and Bradford
et al. (1961) for y initiation. The propagation rate constant kp was caIculated from Case 2 Smith-Ewart kinetics and found to vary between 32 and
423
90
80
70
60
<D 50
& 40
30
.~
<D
20
E
>-
o
a. 10
.2
.3
.4 .5.67.8.9
4 5 6 789
424
VivanT. Stannett
':"
le
f" 20 'ol-e
'-..)
G
..
I
t.
lO
lO
9
8
7
6
o 9
-E 8
7
Z 6
O
e
.
...
N
Z
O
e
N
...
!
...
O
A.
...
O
...
...
111:
o
...
Q
...
O
A. 3
...
O
...
e
111:
3.1
3.2
3.3
3.4
3.5
3.6
Fig. 4. Temperature dependenceof the rates and degreesof polymerization for the
radiation induced polymerization of styrene. Dose rate 0.093MradJhr (Garreau et al., 1979;
reproducedwith permissionof Journal of Col/oid and Interface Science.)
B. Acrylonitrile
The emulsion polymerization of acrylonitrile is quite complex because of
the high solubility of the monomer in water (7.4% at 25C) and the
insolubility of the polymer in the monomer. Hummel et al. (1967), using a
sophisticated recording dilatometer, showed that the rate-conversion
curves
425
were very complex, with an initial rise followed by two periods of almost
constant, but different rates, and finally a first-order reaction rate at high
conversions. Partic1e formation in the aqueous phase was assumed. A few
studies reported by Glazkova et al. (1971) were conducted below the critical
micelle concentration of the emulsifier used. Karpov et al. (1974) reported a
rather thorough study using three different emulsifiers: potassium laurate,
sodium alkyl sulfonate, and a cationic emulsifier, cetyl pyridinium chloride.
Contrary to the general experience with styrene, vinyl acetate, and methyl
methacrylate, the rates were markedly higher with the cationic emulsifier.
True latices were formed when the monomer concentration was below the
solubility limit, Le., no monomer droplets were present. With potassium
laurate, however, only coagulum was formed in every case. The number of
partic1es formed in the latex systems were found to be independent of the
dose rate but proportional to the emulsifier concentration to about the 0.6
power. With high concentrations (0.5%) of cationic emulsifier the number of
partic1es increased linearly with conversion, but at lower concentrations and
always with the sulfonate the number remained constant. The molecular
weight did not vary more than about twofold over a wide range of
emulsifier, with monomer concentration and dose rate being of the order of
106 viscosity average. There is much useful practical data in the paper but
the results are confounded by the presence of heavy concentrations of
coagulum.
O'Neill and Stannett (1974) also found, with sodium lauryl sulfate as the
emulsifier, that when a separate monomer phase was present a milky slurry
that was 100% filterable was formed even ~t low conversions. However, in
the water-soluble regio n low solid latices were obtained. The initial rates
were much higher than in the absence of emulsifiers as was previously
reported by Karpov et al. A plot of the initial rates versus the monomer
concentration is shown in Fig. 5. A first-order dependence on monomer
concentration was found. Again, as with Karpov et al., considerable useful
data is presented for systems where a separate monomer phase is present.
Since these are not truly emulsion polymerizations, further details will not
be discussed in this chapter. The molecular weights were essentially independent of the emulsifier concentration and in the same range, about 106,as
found by Karpov et al. Izumi et al. (1967; Izumi, 1967) obtained similar
results with ammonium persulfate as the initiator.
C.
Methyl Methacrylate
426
Vivian T. Stannett
400
200
e
.2100
'"
Q;
O
u
80
60
'<
IX
40
.....
20
10
0.1
/8
0.2
0.4
2.0
[M]c!ml-')
Fig. 5.
Dependenee of the rate of the radiation-indueed polymerization of aerylonitrile
in aqueous solution, linear portions of rate eurves, on the monomer eoneentration at 25C 8
in the absenee of sodium lauryl sulfate, O with 0.21% (O'Neill and Stannett, 1974: reprodueed
with permission of Journa/ of Macromo/ecu/ar Science-Chemistry AS, p. 949, by eourtesy of
Mareel Dekker, Ine.)
12.
Radiation-Induced
Emulsion Polymerization
427
found to be 0.3, 0.4, and 0.2 for anionic, nonionic and cationic, respectively.
The dose rate dependence of the rate was 0.3 with the cationic emulsifier,
stearyl trimethyl ammonium chloride, and the activation energy was
6.7 kcal/mol. Allen et al. (1962) conducted a few experiments in connection
with their work on block and graft polymer synthesis in emulsion. An
estimate of 92 M-1 sec-1 at 25C for the rate constant for propagation was
obtained, in reasonable agreement with the published solution and bulk
values. As with Abkin et al. the possible contribution to the initiation step
from peroxides arising from the irradiation was stressed and some supporting evidence presented.
A more detailed fundamental study was reported by Hummel et al.
(1962; Hummel, 1963). Using a sensitive recording dilatometer of their own
design they could follow the variation of the rate of polymerization with
time and conversion. There was an abrupt rise in the rate for the first few
percent conversion attributed to particle formation. There followed a
constant rate period up to about 30% when the free monomer phase
disappeared. An increase in rate to a maximum was then observed, ascribed
to the gel effect, followed by a steady decrease in rate due to monomer
depletion and slower diffusion of the monomer to the active sites in the
highly viscous particles.
diameter, compared with those often obtained with chemical initiation (e.g.,
Zimmt, 1959). This presumably explains the comparatively close adherence
to the simple Smith-Ewart picture up to the appearance of the gel effect.
Acres and Dalton (1963a) also studied the emulsion polymerization of
methyl methacrylate initiated by 60Co y raqiation using a recording dilatometer. Only the conversion-time curves were measured with constant dose
rate, varying monomer concentration, and with constant monomer concentration at different dose rates. Except at the lowest monomer concentration a clear gel effect was observed, with linear rates up to that point.
The linear rates increased with increasing monomer concentration up to
.
about 0.4 mol/liter and then leveledoff. The dependenceof the rate, before
the gel effect, on the dose rate was 0.4 and, unlike their findings with
styrene, not dependent on the monomer concentration. Their results w~re
consistent with those of Hummel et al. that methyl methacrylate follows,
with y radiation, the generally accepted Smith-Ewart Case 2 kinetics
except for the marked gel effect.
Hoign and O'Neil (1972) studied several features of the y radiationinitiated polymerization in emulsion. Sodium lauryl sulfate, dioctyl and
dibutyl sodium sulfosuccinates, and two nonionic, polyoxyethylene-type
emulsifiers were used. Sodium lauryl sulfate gave, by far, the highest rates
and most stable lattices and was used for all of the results reported. The
rates were found to be 0.43 order on the dose rate and 0.53 on the
428
VivianT. Stannett
D.
Vinyl Acetate
429
or dose rate
0.5- 1.0
0.8
0.2
0.35
0.1-0.3
Radiation
Smith-Ewart 2
0.5-0.9
0.26
0.7
0.2
0.3-1.0
0.4
O
1.0
O
0.6
E.
Vinyl Chloride
Wang (1962) was the first to report studies on the y-radiation initiated
polymerization of vinyl chloride in emulsion at room temperature. Rapid
rates to high conversions were obtained after rather long induction periods
of 1 to 3 hr. The degrees of polymerization were constant within experimental error at ab0l!t 2000, in keeping with termination being dominated by
chain transfer to monomer. Little or no dependence .of the rate on the
emulsifier concentration or the monomer concentration was observed.
However, the rates were proportional to the 1.22 power of the dose rateo
Some preliminary results were presented by Barriac et al. (1969) and
followed by a more detailed report (Barriac et al., 1976) wherein two
emulsifiers were used. Again at 25C, for example, the degree of polymerization was essentially independent of the dose rate or the emulsifier
concentration and similar to that obtained by Wang. A strong post effect
430
VivianT. Stannett
was observed, and even at low conversions these effects were eliminated
experimentally by freezing in liquid nitrogen immediately after irradiation.
The conversion curves were linear up to about 80% conversion with a small
gel effect apparent at the lowest emulsifier, Le., largest size particles,
concentration. The rates of polymerization and the number of particles
were proportional to the 0.4 power of the dose rateo The rate was
dependent on the 0.3 power, and the number of particles on the 1.0-1.3
power of the emuls.ifier concentration and independent of the monomer to
water ratio. The activation energy for the rate for both emulsifiers was
4.0 kcal/mol and for the number of particles the unusualIy high value of
17.0 kcal/mol. A detailed kinetic analysis of these and the other results was
made. The mechanism and kinetics were in accord with those derived by
Gardinovacki et al. (1971) and Hansen and Ugelstad (1976), modified
somewhat due to the very smalI particles obtained with y initiation. The
smalI size and correspondingly larger number of particles is often found to
be a feature of radiation initiation. It has been explained by the lower ionic
strength of the aqueous phase leading to. the stabilization of more and
smalIer particles in addition to the faster diffusing and neutral hydroxyl
radicals and hydrogen atoms compared with the negatively charged, larger
sulfate-type radicals.
Karpov et al. (1977) studied the polymerization in the presence of a
number of anionic emulsifiers below the critical micelIe concentration. High
molecular weight polymers with low concentrations of impurities were
obtained at high rates. The overalI rate was found to be about 0.5 order
with resp~t to the dose rate with an activation energy of 5 kcal/mol, in
reasonableagreementwith those reported by Barriac et al. (1976).
A number of interesting and unusual phenomena were reported by the
same group of workers. With cationic emulsifiers suspensions rather than
latices were obtained although similar systems gave stable latices with
chemical or UV initiation (Karpov et al., 1971). It was hypothesized that
the radiation conferred some negative charges on the particles, reducing
their stability. Karpov et al. (1972)reported that higher rates were obtained
with cationic emulsifiers below their critical micelIe concentration but that
the reverse was true with hydrogen peroxide. With anionic sulfonate
emulsifiers, Karpov et al. (1973b) found that with monomer concentrations
below 35% stable lattices were obtained but that above 35% free flow
suspensions were obtained. Latex inversion was suggested as the explanation. Kinetics similar to those found with anionic emulsifiers was reported by Karpov et al. (1973c) when cationic emulsifiers below the critical
micelIe concentration were used.
431
Miscellaneous Monomers
1. Vinylidene Chloride
Panajkar and Rao (1979) have reported a rather extensive study of the y
radiation-initiated polymerization of vinylidene chloride in emulsion. With
sodium lauryl sulfate as the emulsifier smooth polymerization-time curves
at high rates were obtained, up to more than 90% conversion. Between 45
and 60% conversion, the linear region, the rate was 0.3 order with respect to
the emulsifier concentration. The molecular weights were found to increase
with conversion and values up to 79,000 were obtained. Some reasons for
the departure from Smith-Ewart behavior were suggested. Earlier, Hummel
et al. (1967) had presented some interesting data on a closely related system,
a similar rate-time behavior was observed and a tentative explanation proposed. Both discussions were based on the insolubility of the polymer in its
own monomer.
2.
Butadiene
in diameter, obtained with radiation. This was partly due, however, to the
rate of radical production at the dose rates used since styrene under similar
conditions had only about double the diameters.
3.
Methyl Acrylate
432
Vivian T. Stannett
Ethyl Acrylate
This monomer also appears to have been studied only by Hummel et al.
(1962, 1967; Hummel, 1963). The rate-time curve shows an initial sharp rise
followed by a slow decline, with no zero-order portion.. The curve can be
adequately explained by the assumption of a strong gel effect and the high
solubility of the polymer in the monomer; presumably the separate monomer phase disappears early in the reaction whereas the creation of new
particles continues longer than, for example, in the case of styrene because
of the comparatively high solubility of the monomer in water, 1.8 compared
with 0.012% with styrene.
5. n-Butyl Acrylate
Hummel et al. (1967) reported a brief study of this monomer with
sodium lauryl sulfate as the emulsifier. Behavior similar to ethyl acrylate
was observed. The rate increased rapidly during the first 10% conversion,
presumably because of particle formation. There followed a long first-order
decline in the rateo This behavior, like ethyl acrylate, was ascribed to the
high degree of swelling of the polymer by the monomer leading to the early
(-15% conversion) disappearance of the separate monomer phase. A more
detailed study with the same emulsifier has been described by Hoign and
O'Neill (1972). The overall activation energy of the rate was very low, only
0.5 kcaljmol; that of the molecular weight, however, was 3.9 kcaljmol. The
dependence of the rate was about 0.7 order with respect to dose rate and
0.48 with respect to the emulsifier concentration. The molecular weight was
essentially independent of the emulsifier concentration and the dose rateo
The values were high, 2.6-6.2 million, viscosity average, but they could be
systematically reduced by the addition of n-butyl mercaptan. The rate
dependencies are similar to those found with potassium persulfate
initiation.
433
6. Other Monomers
Hummel et al. (1967) also conducted brief studies of a number of other
monomers. Chloroprene behaved similarly to styrene. Methacrylonitrile
behaved somewhat like methyl methacrylate with a detinite gel effect. Butyl
methacrylate behaved somewhat like styrene but with two small maxima in
the rate-conversion curves; the reasons for this are unknown but the
second small peak could arise from the gel effect. Decyl methacrylate showed
only one maximum rate at about 50% conversion. Again the reasons for
this behavior are unclear. Isoprene did not polymerize in emulsion at either
low or high dose rates. Kalyadin et al. (1975) have presented a study of the
radiation-induced polymerization of vinyl fluoride in emulsion.
111. Copolymerizations
% styrene
with
87,000 for pure styrene under similar conditions. The reasons for the low
rates, as with butadiene itself, are not clear. Acrylonitrile, however, as the
comonomer with butadiene behaved normally with easy copolymerizations
at rapid rates. G(-M) values reached 100,000 with 90% acrylonitrile. In the
range used for nitrile rubbers G values were still around 25,000, a quite
practical yield from the industrial viewpoint. The copolymer reactivity
ratios were normal. Strong post effects were observed showing a clear tirstorder termination process. Kamiyama and Okamura (1969) copolymerized
VivianT. Stannett
434
B.
435
The grafting efficiencies and molecular weights of the side chains were both
much higher with radiation initiation. An interesting observation was that
the y-grafted latices formed strong continuous films on drying whereas the
redox grafts gave cracked or crazed films. This difference was attributed to
the chemical grafts being located mainly on or near the rubber particIe
surfaces, producing a heterogeneous polymer system. Evidence to support
this idea was provided by adding oil- and water-soluble retarders in the
radiation grafting experiments. The oil-soluble ones also gave cracked films
since they presumably reduced the amount of homogeneous grafting
whereas excellent films were obtained in the presence of water-soluble retarders. Cooper et al. (1959a-c; Cooper and Vaughan, 1959) also carried
out extensive studies of the direct radiation grafting of methyl methacrylate.
A general kinetic scheme was developed for the grafting system and the
activation energy determined as 3.1 kcal/mol for the rate, whereas the
molecular weight of the grafted side chains was little affected by temperature. A number of other interesting and useful observations were reported
in these papers.
More recentIy, Ishigure et al. (1973) and Garreau et al. (1980) reported
rather detailed investigations of the direct radiation grafting of styrene to
polybutadiene latices. The motivation for the work was for the possible
synthesis of such latices for use in the production of high-impact polystyrene. Both papers described direct comparisons between radiation- and
potassium persulfate-initiated systems. The earlier paper used gel-free smallparticle-size latices whereas the second used a larger particle-size crosslinked latex of the type used commercial~y for high-impact polystyrene
production. The former types of polybutadiene latex yielded propagation
rate constants and activation energies in good agreement with the literature
values for styrene. The grafting efficiencies decreased with conversion but
were similar in both radiation and chemical initiation. A simple model was
presented of the basis for the grafting efficiencies, and it gave a cIearer
picture of the various reactions involved.
The second paper (Garreau et al., 1980) with the larger cross-linked
polybutadiene latex gave a much more complicated pattern of behavior
with significant departures from the simple Smith-Ewart picture. It was
shown that the extracted homopolystyrene had molecular weights similar to
the grafted side chains after destruction of the polybutadiene backbone
polymer. Radiation was found to lead to more grafted side chains of lower
molecular weight than was observed for persulfate initiation. This was
explained by the higher rate of initiation in the former case and led to
somewhat lower grafting efficiencies. Nevertheless, under proper conditions
up to 80% efficiency was obtainable. The conversions also produced
436
Vivan T. Stannett
economically favorable yields. These results, when coupled with the ease of
control and other radiation features described in the introduction, make
radiation a viable and effective alterna tive to chemical initiation for grafting
processes on an industrial scale.
437
Finally, a very recent study (Hayashi and Okamura, 1980) was conducted in a flow system with electro n beam irradiation. Styrene gave very
low rates with broad distribution molecular weights (mn) averaging about
1000. Vinyl acetate, on the other hand, polymerized at much higher rates
reaching 60% conversion in 200 sec, in some cases at a dose rate of
0.1 Mradjsec at 40C. Trimodal molecular weight distributions were obtained at peak s of about 400, 40,000, and 400,000. The highest peak may
have been cross-linked or highly branched. Further studies of emulsion
polymerization at high dose rates could be quite rewarding and possibly of
industrial interest in certain cases.
V. Pilot Plant and Related Studies
A. lntroduction
The previous sections discussed the results of a large number of laboratory experiments concerning radiation-induced emulsion polymerization.
In this section the results obtained from a number of investigations using
small-scale pilot plant type equipment will be described. A full discussion of
these, more engineering oriented experiments, is beyond the scope of this
chapter. It was considered important, however, to summarize these studies
and to inelude those references which present the full details of the work.
Engineering flow systems and small scale autoelaves, designed largely for
batch experiments were developed for these investigations. Experiments of
this kind are necessary preludes to the eventuallarge-scale use of radiation
for industriallatex production.
.
B.
Vinyl Acetate
The basic laboratory studies described earlier were with batch systems;
in addition, one larger size and two micro pilot plants have been constructed and operated. AII were based on the general principIe of recirculation from a stirred vessel away from the radiation source. The
emulsions were ~ontinuously circulated through tubing to an in-so urce
"plug-flow" reactor and back to the kettle. The two micro pilot plants were
largely based on the originally designed and operated pilot plant described
by Stahel and Stannett (1969, 1971). The larger pilot plant, constructed by
Neutron Products, Inc., was designed and constructed simultaneously but
quite independently of the micro plants (Allen et al., 1969). The radiation
emulsion pilot plants described by Stahel et al. were investigated for
styrene, vinyl chloride-vinyl acetate copolymers (Stahel et al., 1979), and for
grafting to latex (Memetea et al., 1977).
438
VivianT. Stannett
The first micro pilot plant for vinyl acetate to be described was that of
Hoign el al. (1972), which was essentially an all glass apparatus. The
semicontinuous equipment consisted of a stirred, constant temperature
3000-ml glass reservo ir from which the emulsion was pumped through a
double helix 240 mI-reactor situated within the radiation cavity. On leaving
the reactor the emulsion flowed through a cooler and back into the
reservoir. The reservoir has inlets for monomer, soap solution, and the pure
nitrogen with which the apparatus was purged prior to and during the run.
The conversion curves were found to be linear between 1,5 and 60%. No
post effects were observed although this could have been caused by traces of
oxygen in the system beca use of the diffusion of air through the thin-walled
polyethylene transfer lines. Increasing the flow rate decreased the rate of
polymerization, and at very rapid rates of flow the polymerization ceased
entirely. This effect again was attributed to traces of oxygen in the equipment since the residence time in the oxygen-free reactor itself became too
short for the oxygen to be consumed. The dependence of the rates was
found to be about 0.3 and 0.5 order on the emulsifier' concentration and
dose rate, respectively.
An experimental flow equipment was also built by Friis (1973). There
were two important differences from the Hoign el al. system. First, after
initial purging of the system with nitrogen the polymerizations were conducted in the absence of any gas phase, and second the conversion-time
curves were measured by means of a dilatometer rather than sampling. In
the absence of any oxygen, increasing the flow rate increased the rate of
polymerization, confirming perhaps the explanation of Hoign el al. by the
opposite effect of the flow rateo A very small post effect was found.
A large-scale pilot plant constructed by Neutron Products, Inc., (Allen el
al., 1969) was built as a model for a commercial-scale facility, i.e., it had
features that would be important in an industrial process such as heat
exchangers and pumps capable of handling high-viscosity polymeric emulsions, multiple-stage processing, and the capability for continuous operation. Furthermore, all materials used in the latex formulations were of
commercial grade with no purification prior to reaction. The pilot plant
consisted of four complete reactor loops. The reactors varied in total
volume from 3 to 35 gal and were all interconnected to permit multistage
processing.
The reactants were pumped from the surge tank through a heat exchanger embedded in a 6 ft concrete floor that separated the radiation zone
from the rest of the pilot plant. The reactants were then circulated around
the source in a specified configuration, through another heat exchanger,
then back to the surge tank. All or part of the emulsion returning from the
radiation zone could be drawn off for continuous operation.
439
The majority of the efforts of AlIen et al. at the Neutron products facility
were directed toward producing a poly(vinyl acetate) latex suitable for
production of high-quality paints. They employed a practical commercial
formulation containing a comonomer, a complex emulsifier system, a
stabilizer, and a buffer. The highest quality latexes were produced with a
two-stage batch or semibatch process. The first stage, for initiation and
propagation, was operated on either a batch or semibatch basis with
continuous monomer addition. Generally, the first stage was operated to
about 12% residual monomer content at which time the product was
transferred to the second stage for finishing under different conditions. The
second stage was operated in either a batch or a continuous manner.
Experiments were also performed with the plant perating on an entirely
continuous basis and a good quality latex was obtained.
The properties of the latices and the paints produced from them were
evaluated by several commercial firms. The latex properties varied somewhat from sample to sample but, in general, the paints produced from the
latexes were comparable to those produced from high-grade commercial
latexes made by conventional means. One property of the paint produced
from the radiation-catalyzed latex, enamel holdout, which is one measure
of the capability of a paint to serve as a primer for enamel on unpainted
wood, was superior to that of paint made from conventionally produced
latex.
The results showed that production of latices by radiation catalysis is a
commercially feasible process. However, the authors did not appear to have
resolved two process-development problems: the reduction of residual
monomer to commercially acceptable levels and the eIimination of polymer
build-up inside the process lines. The authors noted, however, that these are
problems of formulation and operation rather than of the basic process
itself.
Using the data obtained in the pilot plant, a conceptual design and cost
analysis was made for a commercial-scale radiation catalysis plant (AlIen et
al., 1969, 1971). The conceptual design was based on a two-stage semicontinuous process. The first stage is operated on a semibatch basis with
continuous monomer addition and intermittent feed to the second stage
which is operated on a batch basis. Economic analysis indicated that
capital and operating costs for radiation catalysis and conventional catalysis plants would be about the same; but the authors felt that the greater
enamel holdout of the radiation-initiated product would justify the use of a
radiation catalysis plant. In addition, the early work of Araki et al. (1969)
showed that better wet-scrub resistance was obtained with paints formulated with high molecular weight radiation-produced homovinyl acetate
latices compared with their conventional commercial counterparts.
440
Vivan T. Stannett
The results of the laboratory and pilot plant studies show clearIy that a
radiation-induced emulsion polymerization process for vinyl acetate is of
considerable interest. In addition to the general advantages discussed in the
introduction, the smooth and easy polymerization at low temperatures is of
special interest for vinyl acetate and its copolymers since it leads to very
high molecular weight products. High molecular weight-polymerized acetate latices are known to lead to paints and coatings with better ink and
enamel holdout and superior wet-scrub resistance.
C.
Fig. 6.
Flow sheet of apparatus: (1) ethylene cylinder, (2) air-operated automatic
intensifier, (3) autoclave, (4) temperature-controlled oi! bath, (5) 60Co radiation source, (6)
buret for flushing of medium, (7) sampling line, (8) shielded room (Senrui et al., 1974a:
reproduced with permission of Journal of Polymer Science.)
12.
Radiation-Induced
Emulsion Polymerization
441
442
VivianT. Stannett
electrolytes. With seeded polymerization the rate was found to be proportional to the 1.8 power of the ethylene pressure and to the 2.5 power of
the fugacity. The temperature dependence again slowed a maximum in the
rate, but at 87 rather than at 800C. Mixed emulsifiers showed that the
overall rate was close to the sum of the separate rates. It was believed that
the two, very different emulsifiers, behaved independently due to the big
difference in the solubility parameters
of n-dodecane and nperfluoroheptane (7.8 and 5.6, respectively). Alcohols depressed the rate,
presumably because of chain transfer since they should increase the solubility of the ethylene in the aqueous phase. The addition of salts increased
the rate of polymerization, and this effect was shown to be attributable to
an increase in the number of particles.
Experiments using no emulsifier were conducted in the same stainless
steel autoclave equipment described above (Machi et al., 1975). Stable
latices were obtained, believed to be achieved by hydroxyl end groups and
adsorbed hydroxyl ions. As with a number of the experiments with emulsifiers the polyethylene had a considerable cross-linked gel contento
Finally, the same group of workers studied the radiation-induced emulsion
polymerization of ethylene in a ftow system (Kodama et al., 1974). Both
potassium myristate and ammonium perfluorooctanoate were used as emulsifiers. At longer residence times (above 0.2 hr) the rate of polymerization
was essentially constant. As with the batch system it was assumed that the
number of particles remained constant. In this region the rate was found to
be proportional to the 0.3 power of the potassium myristate concentration
and the 0.5 power of the dose rate, not too different from the batch systems.
The kinetics was developed and estimates of the propagation rate constants
obtained. Despite other similarities between the two systems, these were
quite different, however, from those extracted from the batch experimento
These extensive and important studies of ethylene polymerization have
only been summarized in this review. The original papers should be
consulted for the full details.
D. Tetrafluoroethylene
Machi et al. (1974) first reported an investigation of the radiationinduced emulsion polymerization of tetraftuoroethylene, with ammonium
perfluorooctanoate as the emulsifier. A 200-ml stainless steel autoclave,
equipped with a magnetically driven propeller-type stirrer, was used. The
standard recipe used was 28 gm of monomer in 150 mI of water with 1%
emulsifier (based on the water). n-Hexadecane (2.0 mI) was added to inhibit
any gas-phase polymerization. The polymerizations were conducted at
443
25C. The rate of stirring had a considerable effect on the reaction rateo The
dose rate exponent of the reaction rate was found to be 0.8, with little effect
on the molecular weights. The emulsifier concentration hardly affected the
rates but greatIy changed the molecular weights, which increased with
decreasing emulsifier concentration. The particIe shapes were also affected,
being spherical at less than 0.5% and rod-like and eventually fibrillar above
this concentration. There was considerable post-irradiation polymerization
with increasing molecular weight. Molecular weights were in the 105-106
range. In the absence of emulsifier much higher molecular weights were
obtained, presumably due to lower chain-transfer reactions.
There followed a very important series of papers concerning the
emulsifier-free system (Machi et al., 1978, 1979a-d). The equipment used
was a modification of that used earlier, but the tetrafluoroethylene pressure
was continuously recorded with the use of a strain gauge. In the first paper
of the series (Machi et al., 1978) the rate of polymerization was shown to be
proportional to the 1.0 and 1.3 powers of the dose rate and the initial
pressure, respectively. The activation energies were 0.8 above and - 5.2
kcal/mol below 70C. There was a maximum in the molecular weights at
about the same temperature. This behavior is reminiscent of the behavior of
ethylene and was again attributed to the increased mobility of the growing
chains above the maximum temperature. The very low mobility would also
account for the first-order dependence of the rate on the dose rate below
70C. As before, n-hexadecane pro ved to be an excellent inhibitor of
polymerization in the gas phase. ParticIe sizes in the range of 0.1-0.2
microns were obtained.
A detailed study of the size, distribution, and number of particIes was
presented in the secorid part (Machi et al., 1979a). It was found that the
particIes grow faster at higher dose rates in concert with the higher reaction
rates. At the higher dose rates the particIe size levels off with reaction time
but continues to grow at the lower dose rates. The size and number of
particIes are essentially independent of the temperature from 30-100oC. The
effect of pressure was complex but the particIe size tends to increase and the
number of partic1es to decrease with increasing pressure. The partic1e size
distribution also becomes broader, perhaps due to increased flocculation
effects at higher pressures. It was concIuded that after the 'generation of the
particIes, which tends to be in the first five minutes, they grow by propagation at the surfaces. The particIes are believed to be stabilized by
hydroxyl and carboxyl end groups and possibly by adsorption of ions. A
discussion of the mechanism of particIe formation and .the loci of the
polymerization was also presented.
Machi et al. (1979b) also investigated the concurrent formation of
hydrofluoric acid during the polymerization due to the radiolysis of the
444
Vivian T. Stannett
445
stainless steel autoclave with 60Co radiation. Ammonium perfluorooctanoate gave the highest rates and molecular weights and was therefore
chosen for the detailed investigation. Increasing the stirring speed first
increased the rates and stability but then essentially leveled off. The
emulsion system gave a higher rate and molecular weight than the corresponding polymerizations in bulk. Increasing the temperature slightly
decreased the rate and increased the molecular weight. The copolymers
were essentially alternating, as were the bulk copolymerizations previously
mentioned. Oxygen caused a small induction period which was followed by
a resumption of the normal rates. The effect of the emulsifier concentration
was to cause an increase in the rate and molecular weight above 1% and
then to have little further effect. The number average molecular weight
ranged from 3 to 8 X104. Matsuda et al. (1974b) continued their detailed
studies using ammonium perfluorooctanoate as the emulsifier. The rate of
polymerization was found to be 0.9 order and 0.26 order with respect to the
dose rate and the emulsifier concentration, respectively. Both the rate and
the molecular weight increased with increasing tetrafluoroethylene content
of the feed. The copolymer composition, however, remained alternating
over a very wide range of compositions. The dose rate dependence indicated
that most of the termination was by degradative chain transfer to the
propylene monomer.
Danno et al. (1974) presented additional work showing that above the
critical micelle concentration (CMC) of the emulsifier the dose rate dependence of the rate was actually 0.7 and below the CMC as 0.6. The molecular weight was higher above the critical micelle concentration and independent of the dose rate and lower below the CMC and decreasing with
increasing dose rateo These and other details of the polymerization reaction
were discussed further in terms of the particle sizes, which are larger and less
numerous at zero or low-emulsifier contents and at lower dose rates and
degradative chain transfer to propylene monomer. Okamoto and Suzuki
(1974) presented the results of a brief study of the effects of a number of
additives on the polymerization. Hexachlorethane and carbon tetrachloride
were found to decrease markedly the molecular weight; the former had no
effect whereas .the latter decreased the rateo 1,1,2-Trichloro-l,2,2trifluoroethane had no effect on either the rate or 'molecular weight.
Alcohols reduced the rate in the order isopropyl > ethyl > methyl, in the
same order as the ease of hydrogen abstraction but had no effect on the
molecular weight. This indicated that the primary radical s were reacting
with the alcohols in the aqueous phase but not in the particles themselves.
Diphenyl picrylhydrazyl and p-benzoquinone, as expected, completely inhibited the polymerizations above 3 x 10- 3 molar concentration. Okamoto
and Suzuki (1975) also investigated the hydrofluoric acid yields during the
polymerization. The yields were first order with the dose rateo The presence
446
Vivian T. Stannett
ve
mor.omer
VA
monomer
Metering
pump
Metering
pump
Soap
solution
tank
Solvent
tank
Liquid
level
gauge
EmuHifying
tank
Sample
Temperature
bath
Metedng
pump
Vent
Rotameters
Off produc.t
drum
Fig. 7.
Schematic diagram for radiation-induced copolymerization pilot plant (Stahel et
al., 1979; reproduced with permission of Journal 01 Applied Polymer Science.)
.
12. Radiation-Induced Emulsion Polymerization
447
of oxygen increased the yield, but was decreased greatly when more than
1% of emulsifier was used. G values were much lower than with pure
tetrafluoroethylene emulsion polymerization. Some tentative explanations
were presented.
Ito et al. (1974b) also studied the copolymerization of ethylene and
hexafluoropropylene with the same autoclave batch system used with
ethylene and with ammonium perfluorooctanoate as the emulsifier. The
rate of polymerization was found to be proportional to the 1.6 power of the
ethylene fugacity. The ratio of ethylene units in the copolymer increased
linearly with the reaction rate, extrapolating to zero at zero rateo This
agrees with the fact that hexafluoropropylene does not homopolymerize. A
number of thermal and other properties of the copolymers were determined.
Properties varied from semicrystalline polyethylene types to amorphous
rubbers. Memetea et al. (1977) attempted to graft styrene to a previously
radiation prepared, poly(vinyl chloride) latex. A recirculating flow system
was used, very similar to those used earlier by Stahel et al. (1969) and others
and shown schematically in Fig. 7. In contrast to earlier results published
by Wang (1963) very little grafting occurred if the correct extraction
procedures for the homopolystyrene were used. Interestingly, the polymerization of styrene on the poly(vinyl chloride) seeds ciosely approximated
Smith-Ewart Case 2 behavior. The less than 3 %grafting was attributed to
the low total dose needed to effect the essentially complete polymerization
of the styrene, only about 0.1 Mrad. Parallel experiments with styreneswollen films and the literature results did give reasonable grafting yields at
higher doses (e.g., 33% with 3.6 Mrad).
Finally, Stahel et al. (1979) studied the radiation-induced emulsion
copolymerization of vinyl chloride and vinyl acetate in an engineering flow
system with a l-gallon reactor, shown schematically in Fig. 7. The copolymerization proceeded smoothly to high conversions. The rate was found to
be proportional to the 0.17 power of the emulsifier concentration, within
the range reported for the individual monomers. The activation energies
were 5.5 and -2.0 kca1/mol for the rate and molecular weight, respectively,
in good agreement with literature results for similar, chemically initiated
systems. The res.ults showed that radiation does represent a viable and
practical method for producing stable lattices with high conversions and
solids content in engineering systems'.
References
Abkin, A. D., Mezhirova, L. P., Iakovleva, M. K., Matveeva, A. V., Khomikovskii, P. M., and
Medvedev, S. S. (1959). Vysokomo/. Soedin. 1, 68.
Acres, G. J. K., and DaIton, F. L. (I963a). J. Po/ym. Sci. A 1, 3009.
448
Vivian T. Stannett
449
Inagaki, H., Yagi, K., Saeki, S., and Okamura, S. (1960b). Chem. High Polym. (Jpn.) 17,37.
Inagaki, H., Yagi, K., Saeki, S., and Okamura, S. (l960c). Chem. High Polym. (Jpn.) 17, 135.
Ishigure, K., and Stannett, V. (1974). J. Macromol. Sci. Chem. AS, 337.
Ishigure, K., Tabata, Y., and Sobue, H. (1964). J. Polym. Sci. A2, 2235.
Ishigure, K., Yoshida. K., and Stannett, V. (1973). J. Macromol. Sci. A7, 813.
Ishigure, K., O'Neill, T., Stahel, E. P., and Stannett, V. (1974). J. Macromol. Sci.-Chem. AS, 353.
Ito, l., and Watanabe, T. (1965). J. Chem. Soc. Jpn.-/nd. Chem. Sect. 68, 552.
Ito, M., Matsuda, O., Okamoto, J., Suzuki, N., and Tabata, Y. (1974a). J. Macromol. Sci.-Chem.
AS, 775.
Ito, M., Senrui, S., and Takehisa, M. (1974b). J. Polym. Sci. 12, 627.
Izumi, Z. (1967). J. Polym. Sci. A-l 5, 469.
Izumi, Z., Kuichi, H., and Watanabe, M. (1967). J. Polym. Sci. A-l 5, 455.
Kalyadin, V. G., Sirlibaev, T. S., and Tirkashev, 1. (1975). Nauch Tr. Tashkent Un-T (462) 100;
(1974) CHAB. 83, 193858f.
Kamiyama, H. (1974). Ann. Rep. Osaka Lab. JAER15029, 33.
Kamiyama, H. (1975). Ann. Rep. Osaka Lab. JAER/ M-6260, 23.
Kamiyama, H., and Okamura, S. (1969). Ann. Rep. Osaka Lab. JAER/5022, 33.
Kamiyama, H., and Saito, K. (1975). Ann. Rep. Osaka Lab. JAERI 5030, 147.
Kamiyama, H., and Shimizu, Y. (1975). Ann. Rep. Osaka Lab. JAER/503O, 148.
Kamiyama, H., Kitayama, M., and Okamura, S. (1970). Ann. Rep. Osaka Lab. JAER/5026(3),
26.
Karpov, V. L., Lukhovitskii, V. l., Polikarpov, V. V., Lebedeva, A. M., and Lagucheva, R. M.
(1968). Vysokomol. Soedin. AI0, 835.
Karpov, V. L., Lukhovitskii, V. l., Polikarpov, V. V., Lebedeva, A. M., and Lagucheva, R. M.
(1969). Khim. Vys. Energ. 4, 173.
Karpov, V. L., Smirnov, A. M., and Lukhovitskii, V. 1. (1971). Khim. Vys. Energ. 5, 470.
Karpov, V. L., Smirnov, A. M., and Lukhovitskii, V. 1. (1972). Vysokomol. Soedin. B-l 4, 6.
Karpov, V. L., Lukhovitskii, V. l., and Lebedeva, A. M. (1973a). Vysokomol. Soedin. A15, 2465.
Karpov, V. L., Smirnov, A. M., and Lukhovitskii, V. 1. (1973b). Vysokomol. Soedin. B15, 726.
Karpov, V. L., Lukhovitskii, V. l., and Pozdeeva, R. M. (1973c). Vysokomol. Soedin. B15, 907.
Karpov, V. L., Polikarpov, V. V., Lukhovitskii, V. l., and Pozdeeva, R. M. (1974). Vysokomol.
Soedin. A16, 2207.
Karpov, V. L., Smirnov, A., Lukhovitskii, V. l., and Pozdeeva, R. M. (1977). Plast. Massy
(USSR) 3, 18.
Klein, A., Stannett, V., and Litt, M. (1975). Br. Polym. J. 7, 139.
Kodama, A., Senrui, S., and Takehisa, M. (1974). J. Polym. Sci. 12, 2403.
Konishi, K., Senrui, S., Suwa, T., and Takehisa, M. (1974). J. Polym. Sci. 12,83.
Litt, M., Patsiga, R., and Stannett, V. (1960). J. Phys. Chem. 64, 801.
Litt, M., Patsiga, R., and Stannett, V. (1970). J. Polym. Sci. A-l S, 3607.
Machi, S., Suwa, T., and Takehisa, M. (1974). J. Appl. Polym. Sci. lS, 2249.
Machi, S., Suwa, T., N'akajima, H., and Takehisa, M. (1975). J. Polym. Sci. Polym. Lett, Ed.
13, 369.
.
Machi, S., Suwa, T., Watanabe, T., and Okamoto, J. (1978). J. Polym. Sci. Chem. Ed..16, 2931.
Machi, S., Suwa, T., Watanabe, T., Seguchi, T., and Okamoto, J. (1979a). J. Polym. Sci. Chem.
Machi, S., Suwa, T., Watanabe, T., and Okamoto, J. (1979b). J. Polym. Sci. Chem. Ed. 17, 129.
Machi, S., Watanabe, T., Suwa, T., and Okamoto, J. (1979c). J. Appl. Polym. Sci. 23, 967.
Machi, S., Suwa, T., Watanabe, T., and Okamoto, J. (l979d). J. Polym. Sci. Chem. Ed. 17,503.
Matsuda, O., Suzuki, N., and Okamoto, J. (1974a). J. Macromol. Sci.-Chem. AS, 793.
Matsuda, O., Suzuki, N., and Okamoto, J. (l974b). J. Polym. Sci. Chem. Ed. 12, 2045.
450
Vivian T. Stannett
Memetea, T., Mitri, K., Stahel, E. P., and Stannett, V. (1977). J. Macromol. Sei.-Chem. All, 337.
Mukoyama, E., and Toriaka, S. (1962). Chem. High Polym. (Jpn.) 19, 337.
Ohdan, K., and Okamura, S. (1969). Ann. Rep. OsakaLab. JAERI 5022(2),37.
Ohdan, K., Kamiyama, H., and Okamura, S. (1970). Ann. Rep. Osaka Lab. JAERI5026(3), 15.
Okamoto, J., and Suzuki, N. (1974). J. Polym. Sci.-Chem. Ed. 12, 2693.
Okamoto, J., and Suzuki, N. (1975). J. Macromol. Sci.-Chem. A9, 285.
O'Neill, T., and Stannett, V. (1974). J. Macromol. Sci.-Chem. AS, 949.
Panajkar, M. S., and Rao, K. N. (1979). Radial. EjJecls 41,71.
Senrui, S., and Takehisa, M. (1974). J. Polym. Sci. 12, 535.
Senrui, S., Suwa, T., and Takehisa, M. (1974a). J. Polym. Sci. 12,93.
Senrui, S., Suwa, T., and Takehisa, M. (1974b). J. Polym. Sci. 12, 105. .
Smith, W. V., (1948). J. Am. Chem. Soco 70, 3695.
Stahel, E. P., and Stannett, V. (1969). "Large Radiation Sources for Industrial Processes," pp.
135-150. IAEA, Vienna.
Stahe1, E. P., and Stannett, V. (1971). USAEC Rep. ORO. 3687-1.
Stahe1, E. P., Tsai, J. T., and Stannett, V. (1979). J. Appl. Polym. Sei. 23, 2701.
Sunardi, F. (1979). J. Appl. Polym. Sci. 24, 1031.
Van der HofT, B. M. E. (1958). J. Polym. Sci. 33, 487.
Wang, U. P. (1962). J. Chinese Chem. Soco (Taiwan) Seco II 9, 195.
Wang, U. P. (1963). J. Chinese Chem. Soco (Taiwan) 4, 171.
.
Zimmt, W. S. (1959). J. Appl. Polym. Sci. 1,323.
lndex
D
A
Diffusion, 296
306
E
Electrostatic
Electrostatic
Electrostatic repulsion,
Emulsification
B
Butadiene,
effects, 6
forces, 3
399
372, 431
tran~fer, 242
Emulsion polymerization
choice of emulsifier, effect of, 221
distribution of components and phases, 53
emulsifier-free, 257
initiatibn mechanisms, 51
Chemisorption, 298
Coagulation, 2, 7, 24
with aluminum salts, 19-22
as kinetic process, 14
measurements of rate, 15
Colloidal behavior of polymerization
261
Colloidal characteristics
systems,
particle stability, 45
radiation-induced, 415
of latexes, 264
Ethylene, 440
Colloidal stability
definition, 2
theory, 8
Constant rate period, 286
Copolymerization
block and graft, 434
random, 433
Hamaker constant,
451
10
452
Index
1
Initiation
emulsifier-induced decomposition,
in monomer droplets, 396
radiation-induced, 415
systems, 329
Interface characteristics, 249
liquid-liquid,
251
liquid-polymer,
254
Interfacial tension, 249, 251
monomer-water,
256
Ions
115
239
372
organic, 26
31
homogeneous, 52, 73
micellar, 54, 63
in monomer droplets,
Latex application properties,
Latex particles
.
colloidal behavior, 2
86
315
M
Macromolecules, oligomeric, 53
adsorbed on particles, 5, 42, 43
grafted on particles, 5, 42, 43
Micellar nucleation
newer methods, 63
Smith'- Ewart theory, 54
Micellar size, 234
effect of polymerization, 235
Mixed electrolyte systems, 35
Molecular weight, 325
of styrene, 421
theoretical predictions, 93
o
Oligomeric radicals, 53
P
Paints and industrial coatings, 309
Paper coatings, 310
Particle generation
homogeneous nucleation, 73
mechanisms, 51, 257, 270, 370
micellar, 54, 63
in monomer droplets, 86
Particle number, 327, 334, 363
effect of emulsifier, 233
initial emulsifier concentration,
of,69
rate of initiator decomposition,
70
as function
as function of,
90
approaches,
95
453
Index
in batch polymerization, 99
in continuous polymerization,
cumulative, 335
experimental investigations,
functions, 102
mechanisms,
56
reabsorption, 199
Radical desorption, 191, 199, 204, 369
105
109-114
on emulsifier
experimental, 90
theoretical, 69
energy characteristics,
Solvation effects, 6
polymerization,
R
Radical absorption
inverse absorption efficiency,
energy, 252
kinetics, 268
of, 344
61
418
41
252, 253
effect
454
Index
T
Taek, 298
Tetrafluoroethylene,
442
v
Vinyl ehloride, 429
latex panicle size distribution,
Vinylidene ehloride, 431
377