From molecular simulation to thermophysical properties : can we bridge the gap between the nanoscale and process scale ?
Special Symposium - MultiScale Modelling
Invited Session on MultiScale Modelling (S-1)
Keywords: macroscopic properties, microscopic description, nanoscale, process scale
From a chemical engineering standpoint, predicting macroscopic properties of any system on the basis of its microscopic description is a Holy Grail. Many chemists have dreamt of models that could predict the properties of fancy mixtures before they are prepared, many polymer engineers would like to know precisely how mechanical properties are related to structure of a polymer material, many process engineers still expect a general method to model multiphase fluid flow. Is there a chance that some of these dreams turn to reality ? This talk aims at answering on the basis of recently published work.
In a first part, we consider "simple" cases, i.e. cases where the system is sufficiently homogeneous that a very small system sample (a few nanometers large), containing a few hundreds to a few thousands atoms, is statistically representative. Then the behaviour of the small size sample may be determined by molecular simulation, i.e. as the result of the microscopic interactions between its atomic components. Macroscopic properties can be obtained by suitable averages : time averages provide transport properties like diffusion coefficients or viscosity, statistical averages provide equilibrium properties like density, saturation pressure, etc. Three advances have helped to reach this stage : the continuous increase of computer capacity, the development of dedicated intermolecular energy models, and the availability of versatile models. Various examples are given for the cases that can be solved in this way, which are in fact not so uncommon : fluids (with the exception of the near-critical region), crystalline adsorbents (zeolites), and liquid-vapour interfacial tension.
In a second part, we consider cases where the quest is more difficult, i.e. the system is heterogeneous or the size of individual molecules is too large. Then alternative, system-specific strategies must be used to extrapolate from the microscopic to the macroscopic level. A first example is provided by near-critical behaviour of fluids, characterized by density fluctuations at a much larger scale than a few nanometers. Thanks to an appropriate scaling, it is possible to use small size molecular simulations to determine reliably the critical locus of pure components and binary mixtures. A second, more prospective example is provided by the solubility of gases in semi-crystalline polymers, which can be investigated by a detailed account of the amorphous fraction, coupled with an approximate description of the crystalline fraction of the material. When the mechanical behaviour of polymer materials is desired, it becomes necessary to lump several monomers in a blob, the properties of which can be obtained from atom-level simulations.
In a last part, we discuss the fields where heterogeneities are still preventing from extrapolating from microscopic to macroscopic behaviour. The computational fluid mechanics of dense suspensions of small solid particles is selected to illustrate which stages could be involved in such extrapolations. Other possible applications are outlined in conclusion.
Presented Monday 17, 16:10 to 16:40, in session Invited Session on MultiScale Modelling (S-1).
