Sie sind auf Seite 1von 12

Herbert Ga jewski and Klaus Zacharias

W eierstra Institut fur Angewandte Analysis und Stochastik

Mohrenstr D

Berlin Germany

Key words and phrases Initial boundary value problem drift di usion pro

cesses a priori estimates Lyapunov function equilibria asymptotic behaviour

Mathematics Subject Classi cation F Q

Introduction

A D

We consider a mathematical model of polymerization which in the language of

chemistry

is called emulsion polymerization Roughly speaking polymerization is the formation

of huge molecules

the molecules of the polymer from smaller ones the molecules

of the monomer In the case of emulsion polymerization this process takes place in an

aqueous medium in the presence

of appropriate auxiliary substances As an example

we

mention that the well known polymer polyvinyl chloride PVC can be produced in this

way starting from the monomer vinyl chloride

There are di erent possibilities to operate a polymerization reactor One distinguishes

batch or discontinuous and continuous reactors The batch reactor is one where all

ingredients are charged at the

beginning of the polymerization and the reaction proceeds

over a certain interval of time Continuous reactors run with a continuous in ow and out

ow of material Mixed types of reactor operating are possible e g the semibatch mode

where part of the ingredients are added during the polymerization process A general

assumption is that the content of the reaction vessel is well stirred so that local inhomo

geneities can be neglected

The mathematical model presented here was proposed in the seventies by Min and Ray

It has been modi ed and extended in the research group of Dr Tauer

at the former Institute of Polymer

Chemistry Teltow Seehof At the

former Karl Weierstrass Institute of Mathematics Berlin the model has been investi

gated from the mathematical and numerical point of view during several years

The mathematical model

To describe the polymerization process we introduce a particle size distribution

function

f f t v depending on time t and a volume variable v The physical

interpretation of f may be given in a heuristic way as follows The di erential f t v dv

is the average number of particles or proportional to the average number of particles

whose volumes at time t belong to the in nitesimal volume interval v v dv That

is we make the usual assumption of statistical physics that the particle number has a

density which is justi ed by the large number of particles

The time evolution of f is is in uenced by chemical and physical processes like par

ticle growth particle coalescence particle nucleation the e ect of additional substances

e g catalyzers and emulsi ers and the reaction conditions For the more chemical and

physical side of the quite complicated interplay of all these factors we refer to the papers

mentioned in the introduction

Important quantities to characterize the particle size distribution function f are the

moments M of order de ned by

M f t Z v f t v dv

Especially we interpret

M f t Z f t v dv as particle number

M f t Z vf t v dv

as mean particle volume

M f t Z v f t v dv as mean particle surface

For the comparison with experiments the essential quantity is

Q t v

Z

v

wf t w dw

Z

wf t w dw

i e the relative contribution of particles of volume v to the mean particle volume at

time t

The model equation is formally a rst order partial di erential equation for f of the

form

f rf

t

v

a f g K f

which we consider on t v with the initial condition

f v f v v

and the boundary condition

f t t

The initial value f is a given nonnegative function satisfying the compatibility condition

f the boundary condition simply says that there are no particles of volume

zero By physical reasons one has to claim f t v for v t and the

asymptotic behaviour f t v for v

The

main constituents in the model equation are

r r v f the particle growth rate

K f the coalescence term

g g v the given seed particle number distribution function

The constant a is proportional to the reciprocal of the mean residence time

The particle growth rate r r v f is an expression of the form

r v f r M f M f b v b v

where b b are given nonnegative constants and r is a given nonnegative function

This is an empirical formula describing particle growth by

monomer and polymer particles

The coalescence term is de ned by

chemical reactions between

K f v k v w w f v w f w dw f v Z k v w f w dw

Z

v

with a given coalescence kernel k Following the proposals in

we take

i e a symmetric positive

constant C

k k v w C vw v w

and weakly singular integrable kernel with a given coalescence

All these terms show that

the model in its simplest form is a nonlocal rst order partial

integro di erential equation

Discussion of the model

To see that the model works properly it is instructive to consider several special cases

Putting the constant a we have the case of a batch reactor described by

f rf

t

v

K f

A simple calculation shows that the coalescence term K f satis es

Z

K f v dv M f

Z

vK f v dv

Using this and assuming f nonnegative we obtain from by a simple calculation

taking into account the boundary condition and the asymptotics as v

M t

d

d

dt M t dt M t

M t

This is just the expected behaviour the particle number decreases the

average par

ticle volume increases with growing time Reducing the model further by omitting the

coalescence term we obtain with the simpli ed growth rate

r

r M f M f

the model equation

f

t

r M f M f f

v

In this case the particle number is conserved i e with

appropriate scaling we have

M f t M f

With the initial condition f v f v continued formally by f v for v

we can solve by the method of characteristics We get

f t v f v M t

where M is the solution of the initial value problem

M r M M

Another reduced version of the model equation we obtain by putting r i e the case

of pure coalescence Then becomes

f

t

K f f v f v

This is the well known integro di erential equation of coagulation which

of processes where particles are lumped together beyond colloid chemistry

meteorology Formal integration gives

M t M t

describes a lot

e g in biology

i e the number of particles decreases and the mean particle volume M is conserved

Equation has been considered e g in

This equation allows to clarify the role of the coalescence term

K f v Z k v w w f v w f w dw f v Z k v w f w dw

v

Speaking heuristically the left hand side of is the time rate

of

change of the number

of particles of volume v The right hand side says that particles of volume v are formed

from particles of volume w v and of volume v w in a symmetric manner which

explains the factor and they disappear if particles of volume v agglomerate with

particles of volume w where the intensity of agglomeration is characterized by the

coalescence kernel k k v w This interpretation of the balance equation for

pure coalescence goes back to Smoluchowski see e g

Model extensions

F or the actual calculations the simpli ed model described

discuss two variants without going too deep into the quite

above has to be extended We

involved chemical details The

main

additional ingredient is the so called nucleation term which takes into account the

more

physical process of nucleation By nucleation appear

particles of a de nite size by

transformation of water soluble oligomers into particles

V ariant The model equation has the form

f

t

r f f

v

a f g K f N f v

where N f v is denotes the nucleation term This term depends on the

initiator concentration I

emulsi er concentration E

water volume fraction V

which are functions of the time The nucleation term has the form

N

f v c

I t V t E t v v

c

where v v denotes Dirac s Delta distribution with support v

The initiator

is a substance which furnishes the radicals necessary to initiate the polymer

ization reaction the emulsi er is a substance necessary to stabilize the growing polymer

particles The initiator concentration is

assumed to

satisfy the simple di erential equation

dI

dt

c c I I I

which gives

I t c c I c c e c t

The total emulsi er concentration S S t is given by

S

t S e at

and splits into the free emulsi er concentration in water E E t and emulsi er covering

the polymer particles Introducing the quantity A A t the degree of coverage of

particle surface with emulsi er we have the balance equation

E t p A t M f S t

Here appears the moment of order which can be interpreted as mean particle surface

see Section The free emulsi er in

linked by an adsorption isotherme

water and the emulsi er on the particle surface are

A

p E

p E

of Langmuire type

The water volume fraction V V t satis es the balance equation

V t H t x q where H t B exp t T

Here denotes x x t R M f the conversion where R is the rational function

R z

a z

b

z c

and q d m d p is the ratio of monomer and polymer density The constant B and the

mean residence time T are given constants as well as the quantities p p p S a b

c c c c c used above

The part r in the factorized form of the particle growth rate see has the form

r

M f M f

A M f

q M f

where

 

A

M f

with

A

if x R M f R

A x if x R M f R

A x c

x

qx

The constant c is given and we have the continuity condition

A A R

V ariant Compared with variant we modify the nucleation term and the particle

growth rate Both become conversion dependent and also the coalescence constant is

substituted by a conversion dependent quantity This modi cation allows to take into

account the gel e ect and the glass e ect These e ects appear as the polymerization

progresses and are a consequence of the growing viscosity of the reactor content If the

conversion is su

ciently high both particle growth

The nucleation term gets the form

rate and coalescence term disappear

N

f v c P t V t E t v

c

v

where P P t denotes the concentration of water soluble oligomers which satis es the

initial value problem

dP

dt

c I fc A t M f c g P P

M

f

M

f

The term r in is substituted by

where

r M f M f C

C x

c

x q

x

x

x A M f

q M f

if x R

if R x R

if R x R

if R x

The coalescence constant C which appears in the coalescence kernel see is sub

stituted by a conversion dependent function C C x of the following form

C x c

if x R

x

q if R x Q

x

if Q x

Again are c c c c c R R Q given constants

Numerics

We sketch only the

main ideas For the approximate solution of

the model equation

we use a di erence method Instead of the volume variable v we take the diameter D

as one of the independent variables linked with v by v D We discretize the

time variable t and the diameter variable D by the not necessarily equidistant grid

t i D k where

t t t i t M D D D k D N

and denote the corresponding step size by i t i t i and h k D k D k

i

respectively We put f k for the value of a function de ned on the grid in the grid point

t i D k and take

i

f k f t i D k

as an approximation of e g the particle density in t i D k We split the coalescence

term omitting the

time variable

K f v H f v f v G f v where G f v f v Z k v w f w dw

T o formulate a di erence method we follow the philosophy of as implicit as possible as

explicit as necessary Denoting by

k r f t i D k N k H f t i D k G k i G f t i D k

R i

k N t i D k H i

i

approximations of the

particle growth rate the nucleation

term and the splits of the

coalescence term we discretize the model equation by

i

f

k

f i

k

i

h k

R i

k f k

i

R i

k f k

i

a f

k

i

g k N

k H k f G i

i

i

i

k

k

for k N the boundary condition gives f The computa

f

i

i

tional

scheme runs as follows Assume that all quantities are

calculated in the grid points

D k k N on the time level t i From the di erence approximation and the

boundary condition we nd rst approximations for f on the new time level t i

k

i

in a knitting like manner for k N Then we compute by numerical integration

the nonlocal quantities like moments and the components of the coalescence term An

implicit di erence scheme Euler backward is used to nd an approximate solution of

on the time level t i

i P i
P

i

where is an approximation

c I t i t i P i P

t i c A t i M f t i c

M

f

t i

M

f

t i

calculated from the values available at this stage of the calculation Using the P i in

we nd a second approximation for the f This procedure is iterated until the

k

i

di erences between corresponding quantities are su ciently small That is we solve

by a quite simple iteration method If certain convergence

criteria are satis ed we

update and go on to the next time level t i The stopping criteria for the iteration

process as well as the step size for the time and the diameter variable were found by

numerical experiments supported by heuristics from polymer chemistry

Remark The most complex term is the convolution part H of the coalescence

term

Z

v

H f v k v w w f v w f w dw

Transformed to the diameter scale its value at the grid point D k k is

H f k Z k y yf

D

D k y f y dy

where k D k and simple symmetry properties of the convolution are used This

expression is approximated using a quadrature formula

of

trapezoidal type we omit the

details

Another possibility to calculate H assumes an equidistant grid in the volume variable

v given by v k k v k N Then a natural approximation of H is given by

v

k

X

j

H f k F k j F j k H f

where F F k v f k for k The convolution sum on the right hand side

k

can be calculated using the fast Fourier transform FFT Unfortunately the necessary

switching from a non equidistant D grid to an equidistant v grid spoils somewhat the

advantages of FFT

Asymptotics

In the following considerations we deal exclusively with Variant In the case of a

continuous reactor one expects that the time evolution leads to an asymptotic state as

t i e all quantities involved in the process assume their corresponding asymptotic

values Especially we have S t S H t B I t c c as t Denote

byE V the asymptotic values of E t V t and by M M M the

asymptotic values of the corresponding moments In addition we assume that

b g v

i e the particle growth rate has a special structure and there is no seed This is the

situation met in all examples Multiplying the model equation by v and integrating

over we get

aM

A M

q M

q b M Cv C c c V E

c

c

Here is C the asymptotic factor in the nucleation term and v the support of the

Dirac

distribution

If we

in addition assume that there is no

coalescence the asymptotic model equation

corresponding

This equation

to becomes

P

v f

v

af C v v P

has the solution

A

M

q M

q

b

f v C

P

v

exp a v

P

v H v v

where H denotes Heaviside s function

H y

if y

if y

These relations are useful tools to control the computation

Regularization of singular initial states

It is quite natural to start from the reactor lled with initiator water which corresponds

in the mathematical

model to the

assumption of a vanishing initial distribution

f v

In this case we have a singularity in the growth term resulting from M f

To overcome this di culty we try

an expansion of the form

f t v t b v o t

with an appropriate positive power The coalescence term will be neglected because

of K f O t which is in good agreement with the fact that in the early stages of

the

polymerization coalescence is unimportant We have in the nucleation term

V E I t B t

We take and enter with all assumptions into Comparison of equal powers

of t details are omitted gives the following equation for the coe cient b of the

lowest order term

A

v f

b v c B v v where M b c B

q

M b

v

This is formally the same equation as

So we can use the solution to overcome

the mathematical di culties in the early stage of the polymerization

References

Aizenman M Th A Bak Convergence to equilibrium in a system of reacting

polymers Comm Math Phys

Burobin A V V A Galkin Solutions of the coagulation equation Di Urav

Dubovskii P B Generalized solutions of coagulation equations Funktsional nyi

analiz i ego prilozheniya

Gajewski H On a rst order partial di erential equation with nonlocal nonlin

earity Math Nachr

Gajewski H K Zacharias On an initial value problem for a coagulation equa

tion with growth term Math Nachr

Gajewski H K Za charias On an initial value problem

in polymer chemistry Tagung Numerische Behandlung von

Martin Luther Univ

for a transport equation

Di erentialgleichungen

Gajewski H K Zacharias Uber Transportgleichungen mit nichtlinearem

Koagulationsoperator Tagung Numerische L osung von Di erentialgleichungen Mat

zlow Garwitz Report

R Math Inst f Math AdW

Galkin V A On existence

and uniqueness of a solution

of

the

coagulation

equa

tion Di Urav

 

Galkin V A On stability and stabilization of solutions

of the

coagulation

equa

tion Di Urav

Galkin V A

P B Dubovskii Solution of the coagulation equation with un

bounded kernels Di Urav

Melzak Z A A scalar transport equation Trans Am Math Soc

Min K W H I Gostin Simulation of semi batch polymerization reactors for

polyvinyl chloride PVC system Ind Eng Chem Prod Res Dev

Min K W W H Ray On the

reactors J Macromol Sci Rev

mathematical modeling of emulsion polymerization

Macromol Chem C

Min K W W H Ray The computer simulation of

reactors

through a detailed mathematical model

J Appl Polym Sci

batch emulsion polymerization

Morgenstern D Analytical studies related to the

J Rat Mech Analysis

Maxwell Boltzmann equation

Pruppacher H R J D Klett Microphysics of clouds and precipitation Reidel

Dordrecht

Stewart I W A global existence theorem for the general coagulation

fragmentation

equation Math Meth in the Appl Sci

Stewart I W On the coagulation fragmentation equation J of Appl Math

Phys ZAMP

Stewart I W A uniqueness theorem for the coagulation fragmentation equation

Math Proc Camb Phil Soc

Tauer K V I Anikeev V A Kirilov Modellentwicklung zur Beschreibung

der kontinuierlichen Emulsionspolymerisation von Vinylchlorid Acta Polymerica

T auer K G Reinisch H Gajewski I M uller Modeling of emulsion poly

merization

of vinyl chloride J Macromol Sci Chem A

Tauer K I M uller Modeling sustained oscillations in continuous emulsion

polymerization of vinyl chloride Preprint Max Planck Institute for Colloid and In

terface Research