Theory of Molecular Size Distribution and

Gel Formation in Branched Polymers II.

General Cross Linking
Cite as: J. Chem. Phys. 12, 125 (1944); https://doi.org/10.1063/1.1723922
Submitted: 14 January 1944 . Published Online: 22 December 2004

Walter H. Stockmayer

ARTICLES YOU MAY BE INTERESTED IN

Statistical Mechanics of Cross-Linked Polymer Networks I. Rubberlike Elasticity

The Journal of Chemical Physics 11, 512 (1943); https://doi.org/10.1063/1.1723791

Analysis of Linear Viscoelasticity of a Crosslinking Polymer at the Gel Point

Journal of Rheology 30, 367 (1986); https://doi.org/10.1122/1.549853

Thermodynamics of High Polymer Solutions

The Journal of Chemical Physics 10, 51 (1942); https://doi.org/10.1063/1.1723621

J. Chem. Phys. 12, 125 (1944); https://doi.org/10.1063/1.1723922 12, 125

© 1944 American Institute of Physics.

THE JOURNAL OF CHEMICAL PHYSICS VOLUME 12, NUMBER 4 APRIL, 1944

Theory of Molecular Size Distribution and Gel Formation in Branched Polymers

II. General Cross Linking *
WALTER H. STOCKMAYERt
Department of Chemistry, Columbia University, New York, New York
(Received January 14, 1944)

A statistical calculation of the gel point and of the molecular size distribution is made for
cross-linked high polymers of arbitrary initial size distribution. The weight-average polymeriza-
tion degree of the cross-linked polymer depends only on the weight-average polymerization
degree of the initial polymer and the degree of cross linking. The results obtained also apply to
certain vinyl-divinyl copolymers. Polymerization degrees can be estimated from observations
of the gel point.

1. INTRODUCTION 2. THE GEL POINT

I Twhen
is well known and practically important that,
sufficiently frequent branching or cross
The polymeric molecules of various sizes which
initially comprise the system will be called
linking of chains occurs during the course of a chains, although they need not necessarily be
polymerization or polycondensation reaction, the linear. In the simplest case, each monomer unit
enhanced capacity for further growth of the in a chain bears one functional group which can
larger molecules (due to their larger number of be linked through chemical reaction to others of
unreacted end groups) leads ultimately to the its kind, either with or without the aid of an
formation of a few network structures so huge added cross-linking agent.6 All functional groups
that the material changes rather abruptly from are assumed to be similar and of the same reac-
a fluid to a gel. The statistical theory of such tivity. The term molecule will be reserved for the
reactions, developed by Floryl-4 and embellished aggregates which exist after some cross linking
by the writer,S has had conspicuous success in has taken place; a molecule therefore consists of
predicting the gel pointl for the several polycon- one or more chains. Let the system contain Nl
densations to which it has been applied. In this monomeric chains, N2 dimeric chains, ,
paper the theory is extended to a more general N. s-meric chains, - etc. The number-average
case than those previously considered; namely, size (polymerization degree) of the chains is thus
the cross linking of a system of chains with (1)
arbitrary initial size distribution. The rather for-
midable size distribution equations which are and the weight-average size is
obtained lead to relatively simple expressions
Aw = LS2 N.ILSN. = LSW., (2)
for the gel point and the several average poly-
merization degrees. It transpires that these equa- where w., the weight fraction of s-meric chains, is
tions should also apply to some vinyl-divinyl given by
copolymers, if certain plausible assumptions are W. = sNsILSNs. (3)
made.
Since each monomeric unit bears one functional
group, ~8 is also the fraction of groups residing
* Publication assisted by the Ernest Kempton AdamS
Fund for Physical Research at Columbia University. on s-meric chains.
t Present address: Research Laboratory of Physical I t will be assumed (as in the previous papers l- 5)
Chemistry, Massachusetts Institute of Technology, Cam-
bridge, Massachusetts. that when cross linking occurs no rings can form;
1 P. J. Flory, J. Am. Chern. Soc. 63, 3083 (1941). that is, any two chains can be cross linked to
2 P. J. Flory, J. Am. Chern. Soc. 63, 3091 (1941).
3 P. J. Flory, J. Am. Chern. Soc. 63, 3096 (1941). 6 Vulcanization processes are obviously of this type.
4 P. J. Flory, J. Phys. Chern. 46, 132 (1942). For several other examples, see B. S. Garvey, Jr., c. H.
·W. H. Stockmayer, J. Chern. Phys. 11, 45 (1943), Alexander, F. E. Kung, and D. E. Henderson, Ind. Eng.
hereinaf ter called 1. Chern. 33, 1060 (1941).
125
126 WALTER H. STOCKMAYER

each other at only one point. This assumption about 4/3 to 3/2 at the gel point, although some
introduces only a moderate error in undiluted extremely large molecules are present.
systems, at least up to the gel point. Then when We wish also to consider the case of a co-
a fraction a of the functional groups in the polymer, in which only a fraction p of the
system has become involved in cross links, the monomeric units contained in the chains bears
number of molecules M is equal to the number functional groups susceptible to the cross-linking
of chains less the number of linkages, or reaction. If Nsl represents the number of s-meric
chains carrying I reactive groups each, the frac-
M = 'LN. - (a/2) 'LsN.
(4) tion of the reactive groups borne by chains
= 'LNs (1-aA n /2). carrying I such groups each is

The quantity 'Y=aAn is the (number) average

number of cross-linked units per chain, and was
called by Flory3 the cross-linking index. Repeating the previous argument, we now find
the average expectance of additional chains to be
A simple procedure l suffices for the prediction
of the gel point. Consider a particular cross- v= 'Ldla(I-1)
linked group attached to some chain chosen at
random from the system. The probability that = a'L. 'Lzl(l-1)N8l/'L. 'LIIN. z• (10)
the reacted group at the other end of the cross The general evaluation of this expression de-
link is on an s-meric chain is w •. Since this mands an explicit formula for N. l • If the reactive
second chain carries s groups in all, and since the units are distributed at random throughout all
probability that a group has formed a cross link the chains ("true copolymer"),7 we may write
is a, the expected number of additional chains
to which this second chain is connected is a(s-1).
Therefore the average expectance of additional
chains v (averaged over all possible chain sizes) is
where N. is the total number of s-meric chains of
all compositions. In this case the sums are easily
If v exceeds unity, there is a finite probability found, with the result
that a chain chosen at random from the system
(12)
belongs to an "infinitely large" molecule'! Hence
the critical value of a for the appearance of such so that the gel point is now given by
large molecules, physically identified with the
gel point, is (13)
(6)
as reported in a recent note. s Since the cross-
If the initial material is a fairly high polymer, so linking index is now given by 'Y = apA n , its critical
that Aw»1, the critical value of the cross-linking value is still (for long chains) equal to An/Aw as
index is expressed in Eq. (7).
(7) For other than random copolymers, Eqs. (12)
If all chains are of equal length, An = Aw and 'Y c= 1, and (13) will in general not be correct. They may
as previously found by Flory.3 The great hetero- be formally retained by altering the significance
geneity of the molecular size distribution at the of p to that of an unknown parameter which
gel point is illustrated by the value of the num- approaches the actual average composition of the
ber-average number of chains per molecule, which chains in the case of a true copolymer. In general,
equals this redefined p may vary with the extent of
(8) polymerization as well as with the average com-
position. It appears, however, that in some cases
from Eq. (4). For most synthetic linear polymers,
Aw/An lies between 1.S and 2, so that when such 7F. T. Wall, J. Am. Chern. Soc. 63,1862 (1941).
8W. H. Stockrnayer and H. Jacobson, J. Chern. Phys.
substances are cross linked 'LN./ M is only 11,393 (1943).
 THEORY OF BRANCHED POLYMERS 127

such variation can be rather small. 9 It is probably and

safe to expect that p will be approximated fairly N t= L: ntm(n.}, for all t. (17)
allln.l
well by the average composition if the chains
have been formed by a single batch copoly- The sums go over all possible sets {n.}, and M
merization. If, however, the chains are mechan- is given by Eq. (4). In the simpler cases treated
ical mixtures of rather different copolymers, such in I, the prototypes of Eqs. (16) and (17) required
an approximation will be poor. summations which were relatively easy to per-
It has already been pointed outS that in cases form directly; here an heuristic attack is of
where Eq. (13) holds it is possible in principle to advantage. Experience with the simpler cases
determine Aw experimentally by observations of suggests the substitutions
the gel point. Thus the weight-average poly-
s~.=w.a(1-a)8-2, (18)
merization degree of polyvinyl acetate, for ex-
ample, could be obtained by measuring the A = (L:sN.)(1-a)2/a . (19)
extent of its ester interchange with diethyl
phthalate at gelation. 6 It is shown in the Appendix that this choice
does satisfy Eqs. (16) and (17). The distribution
3. MOLECULAR SIZE DISTRIBUTION law, obtained by substituting Eqs. (IS), (18), and
In this section the methods developed in 15 (19) into Eq. (14), is thus
are employed to derive the actual distribution of (Lsn.- Ln.)!
molecular sizes up to the gel point. To avoid m{n.} = (L:sN.) a2:n.-1
excessively cumbersome equations, the detailed (L:sn.-2Ln.+2) !
treatment is restricted to the case that all units X (1-a)2:·n.-22:n.+2II. (w.n'/n.!). (20)
in the chains carry a reactive group (p= 1).
The description of any single molecule requires This law may be employed to derive the
specification of the number of chains of each weight-average molecular size Pw of the cross-
size which it contains. We denote by nl, n2, na, linked polymer. A molecule contains L:sn. units
.. ·n., etc., these numbers of chains, and for in all, so that the weight fraction of molecules
brevity represent any particular set of these described by I n.} is simply
numbers by the symbol I n.}. The symbol mIn.}
will be used for the number of molecules de-
W(n.} = (L:sn.)m(n.}/(L:sN.). (21)
scribed by a particular set In.}. The most Then the weight-average molecular size is
probable molecular size distribution is then
written (d. I) Pw = L: (L:sn.)Wln.). (22)
all In.l
mIn.} =Awln.lII. (~.n8/n.!), (14)
As shown in the Appendix, carrying out the sum-
where A and all the ~. are Lagrangian multipliers mation yields the result
to be evaluated subsequently, and wIn.} is the
number of different ways in which a molecule
described by In.} can be constructed from its Since this equation requires Pw to become in-
constituent chains. Generalization of the argu- finite at a c =(X w -1)-I, indicating the presence
ment of I (Appendix A) gives the result of some very large molecules, Eq. (23) may be
(Lsn.- Ln.)! regarded as furnishing an independent derivation
wIn.} = II. sn,. (IS) of Eq. (6).
(L:sn.- 2L:n.+2)! It is apparent that in the special case that all
Now the Lagrangian multipliers must be so chains have the same size s, Eq. (20) reduces to
chosen that (sn-n) !
M= L m{n.}, (16) m n = sNa n- l (l- a)8n-2n+2 , (24)
all In.l n!(sn-2n+2)!
9 C. S. Marvel and G. L. Schertz, J. Am. Chern. Soc. 65,
2054 (1943). which was derived in I. If, in addition, s»l, use
 128 WALTER H. STOCKMAYER

of the Stirling formula leads to the following proceeds much beyond the gel point. For inter-
expression for the weight fractions: preting the physical properties (such as rubber-
Wn=nn-1(')'e-'Y)n/n!')', (25) like elasticity) of these materials the "number of
molecules" has little significance; what is re-
where ,)" equal to sa, is the cross-linking index. quired is the number and size distribution of the
The above equation was derived by Flory 3 by chain portions lying between cross links.l°. 11
methods appropriate to this special case.
The molecular size distribution for the case 4. VINYL-DIVINYL COPOLYMERS
p < 1 (d. Section 2) could in principle be found The copolymerization of a vinyl derivative
by straightforward extension of the methods with sufficient quantity of a divinyl derivative
described above. It is sufficient to remark here also leads to gelled three-dimensional poly-
that for a "true copolymer" the formulae for the mers, 6. 12. 13 but the mechanism differs from the
average molecular sizes carryover if a is replaced cases we have treated above, in that polymeriza-
everywhere by ap. If the chains are long, the tion and cross linking are concomitant rather
average sizes are given by the simple expressions than consecutive. It nevertheless appears that
Pn =A n /(1-,),/2), (26) under certain conditions, which we shall discuss,
the results just obtained may be applied to these
Pw=Aw/(l-,),/'Yc), (27)
systems. s
with ')'=apAn and ')'c = An/Aw. Formally these Some of the assumptions required are quite
equations may be retained for other than random generally obeyed. 14 One of these is that at any
copolymers by substitution of an appropriate instant the fraction of material involyed in
parameter in place of p, but as stated previously, active growing chains is negligible compared to
such maneuvers require considerable understand- the amounts of stable polymer and residual
ing of the copolymerization reaction. monomer.
The foregoing results are unfortunately of The stable polymer consists of molecules, which
little aid in characterizing materials for which are subdivided for our purpose into chains. These
the cross-linking reaction has progressed beyond chains may be most easily pictured as the aggre-
the gel point. Here the inclusion of ring formation gates which would remain if suddenly every
is essential. The treatment employed by Flory 3 doubly-reacted divinyl unit in the polymer were
(which is probably a fair approximation) shows to be cut in half. If CH 2=CHY and CH 2=CH
in any case that the residual fraction of rela- -X-CH=CH2 represent the two monomers,
tively small molecules (the extractable "sol the molecule pictured below would consist of
fraction") becomes negligible if the reaction three chains.

10 H.M.Jamesand E.Guth, J.Chem. Phys.11,455 (1943). 14 For recent concepts of the mechanism of addition
11 P. J. Flory and J. Rehner, Jr., J. Chern. Phys. 11,512 polymerization, consult: J. Abere, G. Goldfinger, H. Mark,
(1943). and H. Naidus, Ann. N. Y. Acad. Sci. 44, 267 (1943);
12 H. Staudinger and W. Heuer, Ber. 68, 1618 (1935). c. C. Price, ibid. 44, 351 (1943); H. M. Hulburt, R. A.
13 R. G. W. Norrish and E. F. Brookman, Proc. Roy. Soc. Harman, A. V. Tobolsky, and H. Eyring, ibid. 44, 371
A163, 205 (1937). (1943).
 THEORY OF BRANCHED POLYMERS 129

In this example, R represents fragments from a With this assumption of equal reactiVity, a
peroxide type catalyst,14 and the cutting process becomes the reacted fraction of"the double bonds
described above would occur at the places indi- attached to divinyl units, and could be found
cated by dashed lines. With the above definition experimentally from the composition of the co-
of a chain, it is not necessary to make special polymer and the over-all reacted fraction of
provision for any branchingl6 which might result double bonds. The problem of evaluating the
from chain-transfer processes, and which seems sums of Eq. (10) in the general case has already
to occur to a small but noticeable extent in the been discussed. In the special case of a "true
polymerization of monovinyl derivatives. If such copolymer," Eq. (13), and the distribution laws
branching takes place, the "chains" will not all of Section 3, may be employed, with p equal to
be topologically linear, but this fact in no way the composition of the original system (expressed
alters the statistical problem. as the over-all fraction of double bonds residing
It is clear that, if all the double bonds located on divinyl units) and a equal to the over-all
on divinyl units have the same probability of reacted fraction of double bonds. It should be
reaction a, then Eqs. (9) and (10) can be applied noted that a is not identical with the weight
to the case at hand. Here N.l is the number of fraction of polymer formed, since the polymer
stable polymer chains, consisting of s reacted contains some unreacted double bonds (on
double bonds in all, which contain l reacted divinyl units of which only one bond has reacted).
double bonds attached to divinyl units. In order The quantity Aw is equal to the weight average
that a be an independent probability, the rates number of (reacted) double bonds per chain. 21
of all steps in the polymerization mechanism The ideal case just discussed might be realiz-
must be independent of the size of the molecules able experimentally if the two monomers were
to which the reacting groups adhere. It has been judiciously chosen. An especially interesting situ-
demonstrated experimentallyl6 that this situation ation would occur if the rates were independent
obtains in polyesterification reactions. With of the composition p, for then Aw in the copolymer
regard to vinyl polymerization, Norrish and would equal the weight average polymerization
Smith l7 have suggested that the termination rate degree of a polymer prepared from the pure
(which may be diffusion-controlled) decreases monovinyl reactant (p = 0) under identical con-
with an increase in chain length, at least for ditions. s
methyl methacrylate. It is difficult, however, to Apparently no suitable experimental data on
understand such a variation physically, in view the gel point in vinyl-divinyl copolymerizations
of the well-known segment-flow behavior of have yet been published. 22 The well-known case
flexible long chains. 1s An alternative explanation of styrene and p-divinylbenzene 12 is not described
of the effects observed by N orr ish and Smith is by Eq. (13), since the latter substance enters the
not improbable, and is to be preferred in the polymer more readily than styrene. Qualitatively
absence of further evidence. Herington and the agreement is satisfactory. If temperature and
Robertson,19 for example, find that the molecular concentrations are such that Aw is 2000, then
size distribution of polystyrene 20 is in satisfactory 0.05 equivalent percent (about 0.03 weight per-
accord with the assumption that the variation cent) of divinyl-benzene is sufficient to produce
of reactivity with molecular size is insignificant. an infusible polymer according to Eq. (13); the
actual limit would be expected to be even lower.
16 G. V. Schulz, Zeits. f. physik. Chemie B44, 227 (1939). The writer is grateful to Homer Jacobson for
16 P. J. Flory, J. Am. Chern. Soc. 61, 3334 (1939). helpful discussion and criticism.
11 R. G. W. Norrish and R. R. Smith, Nature 150, 336
(1942) .
18 W. Kauzmann and H. Eyring, J. Am. Chern. Soc. 62 21 If t?e term.ination process is a disproportionation of
3113 (1940). ' two radicals whIch leaves a double bond in one chain or if
19 E. F. G. Herington and A. Robertson, Trans. Faraday a solven~ takes part i~ c~ain transfer, slight modific;;.tions
Soc. 38, 490 (1942). are reqUired. At all slgmficant degrees of polymerization
20 G. V. Schulz, A. Dinglinger, and E. Husemann, Zeits. however, ~hese effects are negligible. '
f. physik. Chemie B43, 385 (1939). 22 Expenments are now in progress in this laboratory.
130 WALTER H. STOCKMAYER

APPENDIX

First it is to be shown that the distribution law

CL;sn.- L:n.)! w. n •
min.} = (L:sN.)a!:n.-l(l-a)!:sn s -nn.+2 II- (AI)
(L:sn.-2L:n.+2)! 8 n.!
satisfies the conditions
M= L: m{n.) = (L:N.)(1-aA n /2), (A2)
all In,}
and
N t = L: ntmln.l (for all t). (A3)
all In.}

Expanding the power of (I-a) in Eq. (AI), we have

(L:sn.- L:n.)! w. ns
M = (L:sN.) L: L:k (_I)ka!:n,H I II - . (A4)
all tn.1 k!(L:sn.-2L:n.+2-k)! n ns!
By putting L:n.+k=q and eliminating k, this may be written

~ (-a)q (-1pn'q!(L:sn.-L:n.)!w.'"
M=(L:sN.) L : - - L: II , (AS)
q=l q! !:n,5:q 8 n.!(q- L:n.) !(L:sn.- L:n.+2-q)!

where the second sum is taken over all sets In.l satisfying the condition L:n.::; q. Consider now the
function
(A6)

where q is an integer greater than two. Repeated differentiation gives

d)q-2 F(x) ] (-1P"sq!{L:sn.-L:n.)!w. n •

[( dx
-
",=1
= L: II
!:n,5:q s n.!(q- L:n.) !(L:sn.- L:n.-q+2)!
, (A7)

which is recognized as a sum required in Eq. (AS). But since this derivative necessarily contains the
factor (1- L:w.) , its value is zero, and the sum of Eq. (AS) reduces to the terms in q= 1 and q= 2.
Evaluating these, we obtain

M = (L:sN.) (L:w./s-a/2) = (L:N.)(1-aA n /2), (AS)

where we have used L:w./s= L:N./L:sN.= l/A n • This proves (A2).

We must now evaluate
(L:sn.- L:n,)! Wtnt
L: ntm{n.l = (L:sN.) L: a:!:n.-l(1_a):!:8nB-nnB+2 II.' w.n'/n.!, (A9)
allln,l all (n.1 (L:sn.-2L:n.+2)! (nt-1)!

where the prime on the product sign indicates that the term in t is omitted. If we now put vt=nt-l,
and v.=n., this becomes
(L:sv.- L:vs+t-l)! w."
L: ntmln.l=wt(L:sN,) L: a:!:'·(I-a):!:8V,-nv.+t II-. (A10)
all In,} all (v.) (L:sv.-2L:v.+t)! 8 V.!
As before, we expand the power of (1- a) and obtain
 THEORY OF BRANCHED POLYMERS 131

The (q-l)st derivative of the function

(A12)
vanishes for x= 1 and shows that only the terms in q=O and q= 1 of Eq. (All) contribute to the
sum. Evaluating these, we find
L.: ntm{n.l =wt(L.:sN.)/t=Nt • (A13)
all In. I
Thus the assumed distribution law has been shown to satisfy all necessary conditions.
We now turn to a calculation of the weight average molecular size, which from Eqs. (21) and (22)
of the text may be written as

all In.l
(A14)
=[L.:t (t 2 L.: nt 2m {n.})+2L.: L.: (rt L.: nrntm(n.})J/{L.:sN.).
allln.l t r>1 all In.l

The procedure resembles that used above. Putting vt=nt-2 and v.=n., we find
(L.:sv.- L.:v.+2t- 2)! w;·
L.: nt(nt-l)m{n.}/(L.:sN.) =Wt 2 L.: a:!:v a+l(l_a):!:8V.-nV.+2t-2 II-
all In.) alllv.1 (L.:sv.-2L.:v.+2t-2)! 8 Vs!
CA1S)
00 (-a)q (-l):!:V'(L.:sV.-L.:v.+2t-2)!q!w;·
=Wt 2a L.: - - L.: II .
'1=0 q! :!:v.";q 8 (L.:sv.-L.:v.+2t-2-q) !(q-L.:v.) !v.!
Now define
(Al6)

Expanding multinomially and differentiating q times, we have

(Al7)

A direct evaluation of this quantity from Eq. (A16) gives

(AlS)

Thus, substituting in Eq. (AlS) we find

00

L nt(nt- 1)m{n.1 /(LsN,) =Wt2a L a q (A w -l) q=wt2a/[1-a(Aw-l)J. (A19)

all In.) '1=0

Therefore
(A20)
all Inal
where we have made use of (A13).
The second sum in Eq. (Al4) may be found by a similar procedure, the appropriate auxiliary func-
tion being lex) =xr+t-2(l- L.:w x 1)q. The result is8
S
-

(A2l)
all In.l
Substituting Eqs. (A20) and (A2l) into Eg. (Al4), we get
Fw= [Lt t2w t 2 +2L L.: rtwrw tJa/[l-a(A w -l)J+ Lt tWt=aAu,2/[1-a(Aw-l)J+Aw
t r>t
=Aw(1+a)/[l-a(Aw-l)], (A22)
which is the result given in Eq. (23) of the text.