Prediction of Diffusion Constants

The following description again follows the already quoted papers of Gusev and Suter. Using the Eins values mentioned in the foregoing section, the whole packing model in 

r1 (t)is the position vector of penetrant i and is the average over all possible time origins t = 0 and all simulated trajectories of a penetrant of a given kind. Again, as with the solubilities the Gusev-Suter method can only handle small penetrants in this way, because the respective polymer matrix cannot conformationally adjust to larger penetrants. Table 1.3 contains a comparison between experimental and 

In comparing simulated and experimentally measured transport parameters one has to be aware that experimental data in the literature depending, for example, on sample preparation conditions and the chosen measurement methodology can show a considerable scatter, often reaching a factor of two or even more. It is, for example well-known that polyimides often contain residual solvent filling a part of the free volume and thus leading to systematically lower S and D values from experiments than from simulations [34]. 

---

Two atomistic approaches have been presented briefly above molecular dynamics and the transition-state approach. They are still not ideal tools for the prediction of diffusion constants because (i) in order to obtain a reliable chain packing with a MD simulation one still needs the experimental density of the polymer and (ii) though TSA does not require classical dynamics it involves a number of simplifying assumptions, i.e. duration of jump mechanism, elastic polymer matrix, size of smearing factor, that impair to a certain degree the ab initio character of the method. However MD and TSA are valuable achievements, they are complementary in several... 

---