Dipartimento di Chimica, Politecnico di Milano, Piazza Leonardo da Vinci 32, 1-20133 Milano, Italy
1. I N T R O D U C T I O N
After dissolution in appropriate solvents, polymer chains may attain an enormous number of conformations. As a simple example, a polyethylene chain with N chain bonds develops roughly 3Nconformations, since the rotation angle around each bond may be trans (T), gauche-plus (G÷), or gauche-minus (G-), corresponding to the minima of the torsional potential regarded as rotational states.' Of course, any existing side groups would add to the total number of conformations through the statistical combination of their own rotational states. Such a large number of accessible conformations naturally favours the use of statistical-mechanical methods where single polymer chains, if sufficiently long, may even take the role of canonical systems. In this paper we shall address some aspects of the conformational statistics of polymer chains in solution. Rather than in the classical thermodynamic properties, such as the free energy of mixing or the osmotic pressure, we shall be particularly interested in such conformational properties as the interatomic distances and the overall chain size, typically probed by elastic scattering techniques. Before examining the basic conformational aspects in solution, it is convenient to consider briefly a polymer chain in a typical melt. Specifically, we analyze the statistical behaviour of its chemical bonds, regarded as a sequence. We see that the relative spatial orientation of any two consecutive chain bonds is dictated by Boltzmann's statistical law applied to its conformational states. The energy of these states is basically determined by the local conformation of the chain; packing with the surrounding chains provides an energy contribution that is usually assumed as conformation-invariant, so that it may be effectively *In memory of Prof. Piero Pino. tAuthor to whom correspondence should be addressed. Editors' note: It was our editorial policy from the beginning, in putting together this issue, to refer to the metal complexes used in Ziegler-Natta polymerization uniformly as "initiators" or "initiator systems" rather than as "catalysts", and so the former term has been used in this paper and in the others in this issue. 463
464
G. A L L E G R A and F. G A N A Z Z O L I
neglected. In other words, the chain conformation is statistically influenced by the action of local forces, arising from bond-length, bond-angle and rotationangle potentials, in addition to interactions between atoms separated by no more than ,-, 10-15 bonds. If this behaviour dictated by local forces is taken to fully determine the overall conformation of the chain, no matter how long, we have the so-called phantom chain. 2 For this model, interactions between topologically distant atoms are completely ignored. However, after following its natural trajectory for many bonds, the chain will have wandered enough to come back to some region of space already visited. Unlike the phantom chain, the real chain will avoid the space conflict. In principle, we have a statistical dilemma: (i) either the chain makes what can be regarded as a local detour, without substantially changing its rotational-state statistics overall, or (ii) the chain coil undergoes a substantial expansion to avoid most space overlaps, the remainder being taken care of through local detours. We see at once that alternative (ii) strongly reduces the chain entropy because it implies discarding most conformations. Furthermore, although the chain avoids some space conflict with itself, it generates plenty of new conflict with the surrounding chains, with a further entropy reduction if these conflicts are to be avoided in turn. 3 In conclusion, alternative (i) is much to be preferred, and we have what shall be denoted as the unperturbed chain. It is important to maintain the distinction between this physical state and the phantom chain, wherein admissible forces between chain atoms are limited to short-range interactions, confined to a topological neighbourhood of ,~ 10-15 bonds; we stress that, in spite of the inherent approximation, the phantom chain is a useful and frequently adopted model.~ Although the unperturbed chain accounts for some volume exclusion, it does so through the action of forces falling off rather quickly with increasing interatomic topological separation (or distance along the chain contour). These forces will be denoted as medium-range interactions; 2"4 as we shall see in the solution state longer-range interatomic forces may also appear. Both the phantom and the unperturbed chain obey the important rule that their mean-square size is proportional to the chain length, if this is large enough. Indicating with r(N) and S respectively the end-to-end distance and the radius of gyration of the N-bond chain, we have ($2)o = (r2(N))0/6 oc N,
N >> l,
(1)
where the angular brackets stand for "conformational average" and the subscript zero refers to the unperturbed state of the chain. The same equation holds for the phantom chain, in which case the subscript ph will be adopted. Actually, scattering experiments may yield ($2)0, whereas the mean-square end-to-end distance (r2(N))0 is not usually an observable quantity, though it is easily obtained from computer simulations. Moreover, the relatively simple theoretical evaluation of (r2(N))0 directly gives ($2)0 through eq. (1).
LINEAR AND STAR POLYMERS 1N SOLUTION
465
w(r)
1 0
w.-
/
r
r,~
i
/
',
~e(r)
I i Ii I I
I
i! tl
F[6. I. Schematic drawing of the interatomic potential w(r) (solid line) in arbitrary units and o f the corresponding effective potential e(r) = 1 - e x p ( - w ( r ) / k B T ) (dashed line) as a function of the separation r. The distance r at the minimum of w(r). rvdw, and the separation r* at which it changes sign are also indicated.
Let us now consider a polymer solution. Within the present paper we shall be concerned with dilute solutions, where the polymer coils behave as independent entities. At a quantitative level, if the distance r between any two chain atoms is smaller than their van der Waals' distance, the interaction energy is large and positive (i.e. repulsive), otherwise changing to negative, that is, attractive, for larger r's and going asymptotically to zero for still larger distances (see Fig. 1). The resulting energy plot resembles the well-known Lennard-Jones function, but two considerations must be borne in mind in this regard. First, for any given distance r the plot is assumed to represent the average local energy w(r) between two chain atoms, considering all possible conformational states of their side groups. Second, w(r) is actually to be regarded as a free energy difference between the state under consideration and a reference state wherein the chain atoms are surrounded by the solvent only. In conclusion, the energy reported in Fig. 1, (i) contains an entropic contribution, and (ii) is affected by the solvent quality. In a dilute solution we may assume that the energy contribution prevails over the entropic one. Since the two interacting atoms of Fig. 1 in principle may be separated by all possible
466
G. ALLEGRA and F. GANAZZOLI
distances, it can be shown that the cluster expansion coefficient effectively associated with their interaction 5 is
fl(T) = fv [1 - exp ( - w ( r ) / k , T ) ] d V ,
(2)
where the integral is extended to all space. Since w(r) is mainly energetic, the so-called binary cluster integral fl increases with increasing temperature, at least within the temperature range of interest to us. We shall consider in the following the alternative case where w(r) is mainly entropic. Although we are confining our attention to two-body forces, it must be pointed out that they are to be counted properly. For example, it is possible to see that, if three non-bonded chain atoms come close at the same time, their contact energy is not reducible to a mere sum of two-body contributions o f the type given in Fig. 1. Through the cluster-expansion method it can be shown that the coefficient of each two-body contact needs modification to account for the three-atom contact. We have
Belt(T)
=
(3)
f l ( T ) + a3,
where fleeris the effective factor and 0"3 ( > 0) is the three-body repulsive correction. Although a3, unlike fl, may sometimes be considered as effectively temperature
independent, it is generally dependent on the location of the two atoms along the chain (mainly on their topological separation). Of course, higher-order corrections are also needed, in principle, but their importance quickly decreases with increasing order. Hence we shall retain 0"3 as the only important correction to ft. It may be possible to select a temperature ® (depending on the polymersolvent pair) such that fl is slightly negative, 0"3 turns out to be the same for all the atom pairs and we hence have 2'6 flerr(O) =
fl(O) + 0"3 =
0.
(4)
Under these conditions we essentially reproduce the same unperturbed state existing in the melt. But why is it that in a melt the unperturbed state persists over a range of temperatures instead of appearing at a single temperature, as in solution? Let us first consider that in a melt the potential w(r) (see eq. (2)), cannot contain an important energy contribution. In fact, if the distance between any two atoms is changed, other identical atoms will intervene to keep the density locally uniform, the number of the interatomic contacts will stay the same and so will the energy. Being therefore (mostly) of an entropic nature, w(r) is proportional to T and the integral giving fl is temperature independent. The fact that its value is exactly the one satisfying eq. (4) is not obvious but is suggested by the initial considerations (in fact, fleff :/: 0 would lead to either expansion or contraction of the chain). Going back to the solution state, if T > ®, in the light of the preceding discussion we have fl(T) > fl(O), and consequently flefr(T) > 0, see eq. (4). As
LINEAR AND STAR POLYMERS IN SOLUTION
467
we shall see, there is a predominance of intramolecular repulsions and we are in a good solvent. Conversely, for T < ®, flefr(T) < 0, the intramolecular attractions predominate and we have a poor solvent. As chain length increases, the change of solvent behaviour is usually manifested within a progressively narrower temperature range. We shall be interested in chains long enough that this temperature range is much smaller than O. It is now possible to show that in the unperturbed, or theta, state, where fl¢~ = 0 (see eq. (4)) intramolecular medium-range interactions between different chain atoms still exist. Let a chain comprise N atoms indexed sequentially from 1 to N, and consider two of them with index numbers i a n d j and a mean-square distance . We show that if their van der Waals distance rvow(see Fig. 1) is not negligible with respect to m, the effective value of fl applying to them is more positive than the one given by eq. (2). In fact, the integral must be weighted by the a-priori probability Wi.j(r) of finding the two atoms at a distance r. We shall assume a Gaussian distribution for the probability, i.e.
[3/(2zt(r2(i, j)>)]3/2 exp [ - 3r2/2]
Wi.j(r) =
[3/(2n(r2(i,j)>]3n[l
- 3r2/2(r2(i,j)> + . . . ] .
(5)
Referring to temperature T for generality, the interaction free energy between the two atoms is the average weighted with Wij (r), or (in kB T units, see also eqs (2-4)) a(i,j)
Iv Wi.j(r)[1 - exp ( - w(r)/kB T) + f3(r)]dV
=
(6)
where f3(r) is the overall interaction among atoms i, j and a third atom, the distance r(i,j) being equal to r. Obviously, we have Iv t"3(r) d V =
0.3.
(6A)
Remembering eq. (5), it is convenient to split the integral in eq. (6) as follows a(i,
-3/(2(r2(i'J)>) Iv fl[1 - exp ( - w(r)/ks T) + f3(r)]dV t . (6B) The first integral within curled parentheses is merely/3o~(T), see eq. (4), and we shall write 2'4 a(i,j) ==_ a2(i,j) + ~ a3(i,j, k) + a2s(i,j)
(7)
k
where a2(i,j)
=
[3/(2g(r2(i,j)>)]3n~(T);
X a3(i' j' k) = k
[3/(2rr)]3/2 0.3;
(7A)
468
a2ss(i, j )
G. ALLEGRA and F. GANAZZOLI
=
- zr[3/(2n)]5/2 fv r2[1 - exp ( - w(r)/ka T) + fa(r)] d V.
(7B) At T = 19 the sum of the first two terms in eq. (7) is zero, since fl(19) + 0"3 = 0, see eq. (4) and (7A). The integral in eq. (7B) is negative - in fact, the integrand resembles the function reported in Fig. 1 - and approximately constant in the vicinity of T = 19. Defining ~v r2[l -- exp ( - w(r)/kB T) + f3(r)]dV =