Physics 212: Statistical mechanics II, Fall 2006 Lecture XIX

Now we begin Part III of the course, which will examine a few applications beyond phase transitions of the ideas and methods already developed. Lattice models of polymer physics This rst lecture will focus on some standard models of homopolymers (polymers in which every unit is the same), and on the classic Flory theory which is essentially a mean-eld theory. The second lecture will explain how all of this is connected with magnetic phase transitions. Most of the polymer material covered here can be found in Cardy, but a specialist reference is C. Vanderzande, Lattice models of polymers, Cambridge UP. Polymers are the rst example well see of nontrivial scaling that is purely a consequence of a geometrical constraint like self-avoidance, rather than tuning of parameters in a free energy. Real polymers usually live in continuous space, but we will deal with lattice versions such as lattice random walks and SAWs. Experimentally this seems to be adequate for some universal properties determined by long length scales, just as the Ising model and binary uid seem to have the same universal properties. For example, consider random walks on a square lattice. Let the walk has steps of length 1 and be at position x, y after step N . Then its mean squared displacement grows at the next step as R2
N +1

R2

= =

1 (x + 1)2 + (x 1)2 2x2 + (y + 1)2 + (y 1)2 2y 2 4 1 (4) = 1. 4

(1)

So R2

= N.

We can imagine each step as one unit of a polymer, although this isnt a very physical model as it ignores the contact or excluded-volume interaction of the polymer. It turns out that, for a polymer with a nite number of congurations per monomer, as occurs if for example there are only a few possible bond directions, the bad approximation (the one that changes the universality class) is not in going to a lattice model but in ignoring the contact interaction. There may also be long-range interactions between parts of the polymer, which we discuss briey below. Ignoring all interactions, we will say that a polymer is in the random-walk universality class if it has the same universal scaling laws R2 cN N 2 = N, N N 1 , 1 2 = 1. =

(2)

Here cN denotes the number of lattice polymers of length N ; for random walks = z, the coordination number of the lattice (4 for the square lattice example above). The idea is that a more realistic polymer, or a polymer on a dierent lattice, will still have the same scaling R N 1/2 , so = 1/2, even though the constant of proportionality will be dierent. Hence might be a universal 1

quantity (some simple calculations on dierent lattices, or on a polymer with certain possible bond angles in continuous space, will help you understand why this is so). An amazing fact is that even though is not universal, since for random walks on regular lattices = z and z is lattice-dependent, is universal (here we are obeying convention in dening the power-law factor as N 1 ). More complicated polymer ensembles have noninteger values of that are still universal! We will understand why next lecture. Finally, let us introduce one last polymer exponent. For many purposes, it is important to consider not just linear polymers but ring polymers (for example, some biologically important molecules are rings). How many random walks are there of length N that return to the origin at the last step? In one dimension, the answer is given by the number of arrangements of N/2 steps up and N/2 steps down, N 2N N 1/2 = N N 3 . (3) N/2 Here = 2, just as for the open walks, and = 5/2. It is left to the reader to show that for closed walks is the same as for open walks, and that = 3 d/2, where d is the dimensionality. 1 Thus we have = 2 , = 1, = 3 d/2 for random walks. An essential ingredient in obtaining a more realistic polymer model is to take into account the local repulsion of the monomers. This will lead to the study of self-avoiding walks, the main topic of the next lecture. For now we work through a famous mean-eld-like theory (Flory theory) that gives a surprisingly good approximation for the exponent for self-avoiding walks. To calculate the mean radius R2 , our approach will be to write an eective free energy F (R) and then minimize it with respect to R. We approximate the entropy by its value for random (not necessarily self-avoiding) walks: the number of walks of radius R is equal to the total number of walks, multiplied by the probability of having radius R: cN (R) = cN PN (R). Then, since PN (R) is Gaussian for random walks, we have SN (R) = log cN (R) = where B and C are some constants. We also need an approximation to the interaction energy of the polymer. For a contact interaction, the simplest approximation is that the energy density is just proportional to the number of contacts if the polymer segments are randomly distributed: U R 2 , where is the density of polymer segments. This gives N N2 UN (R) ARd ( d )2 = A d . (6) R R Then minimizing the free energy FN (R) = UN (R) T SN (R) gives Rd+2 N 3 R N d+2 = N ,
3

(4)

BR2 +C N

(5)

3 for d 4. d+2

(7)

One caveat to the above is that above four dimensions, the internal energy integral goes to zero at long distances if < 1/2, so the above result is unphysical in d > 4; neglecting the internal energy 2

gives just the random walk result = 1 for all d 4. This suggests, correctly, that the self-avoiding 2 walk problem has an upper critical dimension, as in the problems that we studied previously. Amazingly, Flory theory is exact for in 2 and 4 dimensions and just barely wrong (0.6 instead of 0.588 . . .) in 3 dimensions. The validity of Flory theory in dimensions greater than or equal to four is a result of the reduced importance of geometric uctuations of how many segments are in a volume. Note that these results have a dimensionality dependence and hence are not exactly like our earlier mean-eld theory.