GENERAL DISCUSSION Prof. F. Franks (Unilever) said: Fig. 8 of the paper by Hayter et al. shows that the extrapolated values of Dd -t do not correspond to the self diffusion coefficient of bulk water. This result is at odds with studies on other lamellar systems, notably clays and phospholipids.' Two possible explanations can be advanced for this unacceptable result : (1) If d is determined by X-ray diffraction then the measured length corresponds to the sum of the water layer and the amphiphile and it is assumed that the addition of water to the anhydrous amphiphile system does not affect the mean chain length or area/molecule. (2) Finer and Darke have shown that by plotting D against d-l linear regions are obtained which reflect the different diffusion states of the inter lamellar water.Fig. 4 indicates that the anisotropy of D increases as d decreases, as would be expected. However, if the range of observation of the motion is only < lOA then the observed low anisotropy is not surprising. On the question of lack of anisotropy as shown by the n.m.r. results (does not eqn (7) assume isotropic diffusion?) this, as well as the irreproducibility, might be due to bilayer defects and end effects. Over a period of 1 ms the distance diffused is > 1 pm. It is very difficult to produce lamellar systems of such dimension without serious defects which would allow water to diffuse out of the interlamellar channel and apparently diffuse through the bilayer, as suggested by the authors. Prof. A. Silberberg (Israel) said: Dr. Windrich and I have found that water penetrates fatty acid multilayers built up on glass slides by the Blodgett technique.These effects were studied and evaluated by their effects on the spectra of dye molecules incorporated into a selected one of the transferred monolayers as a function of the distance of this dye containing layer from the surface. Dr. D. S. Reid (Unilever) said: In scanning calorimetric studies of agarose gelation,2 we found that it was very difficult to obtain complete dispersion of the agarose. We finally settled on solution at 120°C in an autoclave, followed by filtra- tion. There, therefore, may still be a significant concentration of nuclei present after heating to 80°C. Whilst these may not be effective in a fast quenching experiment, I wonder how Prins' results (fig.5) would change if solutions were heated to >80°. Also, I wonder how relevant results from such quenching experiments are to the under- standing of agarose gels formed under slower cooling conditions. We have found, for example, that though agarose normally gels at around 40"C, a 1 % sample held for 24 h at 50°C will gel. Clearly, kinetic factors are very important in this gelation. A quenching experiment will not give us information on the nature of these kinetic barriers, and the final gel obtained may be very different from that obtained by slower cooling. Prof. W. Prim (Syracuse University) said: (1) The important point of 5 fig. in our paper is the occurrence of a Bragg spacing upon quenching to sufficiently low temper- * D. S. Reid and D.J. Tibbs, Thermal Analysis, Proc. 3rd I.C.T.A. (Davos, 1971), ed. H. G. E. G. Finer and A. Darke, Chem. Phys. Lipids, 1974, 12,l. Wiedemann (Birkhauser Verlag, Basel), vol. 3, p. 423. 156GENERAL DISCUSSION 157 atures. We have proved (see " note added in proof " above) that this spacing is caused by a nucleation-free spinodal phase decomposition mechanism. In so far as autoclaving seems to eliminate residual agarose aggregation, one might expect the regularity caused by the spinodal mechanism to be possibly enhanced upon quenching auto- claved agarose solutions to below the spinodal. (2) Slow gelling at a temperature just below the phase boundary leads to a gel structure without a Bragg spacing because now the phase decomposition takes place through a normal nucleation and growth mechanism.The residual aggregates in our samples heated at 80°C are prime candidates for the initiation of the gelation process, giving rise to a very inhomogeneous structure with large domains of polymer-rich phase. If the residual aggregation is removed by autoclaving, the gelation will have to take place through homogeneous, rather than heterogeneous nucleation and it is difficult to predict whether the final structure will be very inhomogeneous with large domains-and thus a lot of light scattering-or less inhomogeneous with many small domains-and thus less light scattering. Prof. P. J. Flory (Stanford) said: Is it conceivable that the long-range correlations could be due merely to fluctuations of cross-linking density that are entirely of random statistical origin ? Such an explanation would avoid the implication of nonuniformity in the chemical processes involved in network formation.Secondly, can Prins clarify the term " . . . radius of gyration of all segments belonging to a given crosslink "? The basis for assigning segments to individual cross-linkages is not obvious. Moreover, irrespective of how the segments should be apportioned to the various cross-linkages, in most networks these sets of segments do not occupy discrete regions of space; their domains must overlap copiously. It is essential in this connection to distinguish spatial neighbours from topological ones, the latter being first neighbours in sequence along a given primary molecule. The topological domain may envelop 10 to 100 spatial neighbours.Prof. W. Prins (Syracuse University) said: (1) The observed light scattering of swollen networks formed by a vinyl/divinyl crosslinking copolymerization is typic- ally a hundred times larger than the scattering arising from the thermal fluctuations in segment density. The latter are calculated by assuming that all chains between crosslinks have the same contour length. It is hard to imagine that a distribution of contour lengths of any reasonable width will cause the thermal fluctuations in segment density to increase by a factor of 100. Similarly, it is hard to see why a distribution of contour lengths centring around chains with say 100 monomer units, would give 5000 to lOOOOA correlation distances in the light scattering as observed experi- mentally. For these two reasons we believe to be dealing with spatially non-random networks. So far we have not carried out a calculation quantitatively to corroborate the above statements. (2) In our treatment the radius of gyration refers to the topological domain of the p , segments around each crosslink, but the local volume fraction, q5(ro) of polymer segments takes into account contributions at ro from the crosslink concentrations v, (ro + r) at any location ro + r by integration over r : 40-0) = Vl {P(r)vc(ro + r> dr where p ( r ) is the topological segment density belonging to a given crosslink and u, is the volume of a segment. The spatial overlap of topological domains is thus fully accounted for. W. Prins et a/., J. Polytner Sci. (Phys. Ed.), 1974, 12, 533.