59th Annual Report on Research 2014

Asphaltenes form the heaviest and most aromatic fraction of crude oil.  The aggregation of asphaltenes presents significant challenges to the petrochemical industry by severely hindering the extraction, transport, and processing of crude oil.  Accordingly, this project pursues a multiscale simulation strategy for investigating the molecular mechanism of asphaltene self-assembly.  This approach develops and employs highly efficient coarse-grained (CG) models in order to effectively address the time and length scales that are relevant for asphaltene aggregation.   Since asphaltenes are operationally defined by their insolubility in n-heptane and solubility in toluene, our initial efforts have focused on developing transferable CG models for heptane-toluene (hep-tol) mixtures.  To date, we have successfully developed structure-based CG models for hep-tol mixtures.  Presently, we are developing thermodynamics-based models for hep-tol mixtures and also parameterizing CG models for model asphaltene compounds. 

There exist two dominant paradigms for developing CG models of liquids and other soft materials.  Our primary efforts have focused on developing bottom-up (structure-based) CG models, since we anticipate that they will provide a more accurate description of the interactions and self-assembly of complex asphaltene molecules.  However, bottom-up models typically provide a limited description of thermodynamic properties, such as bulk density and compressibility.  In the first funding period, we developed the computational tools for both parameterizing and simulating bottom-up CG models that accurately model these thermodynamic properties.  In particular, we implemented and validated the pressure-matching method that was earlier proposed by Das and Andersen (DA).  This method parameterizes a volume-dependent correction to the virial equation for CG models.  In addition, we implemented and validated a modified barostat for simulations of the resulting CG models in LAMMPS.

In the second funding period, we extended and applied this methodology to parameterize transferable bottom-up CG models that not only accurately model the equilibrium structure, but also the density and compressibility of hep-tol mixtures.   Our initial application of the DA algorithm for heptane provided a significant improvement in the description of volume fluctuations.  Nevertheless, the resulting model did not quantitatively match the volume fluctuations sampled by the atomically detailed OPLS model.  Accordingly, we implemented more sophisticated basis functions into the pressure-matching calculation and also more carefully investigated the convergence of the DA method.  These studies indicated that the discrepancies between the volume fluctuations of the atomic and CG models arose due to subtle differences in the intermolecular packing generated by the two models.   Consequently, we developed a self-consistent method for parameterizing the correction to the CG virial equation such that the resulting CG model reproduced the atomic volume fluctuations with quantitative accuracy.  We employed this method to parameterize CG models at several different resolutions for heptane (1-, 2-, and 3- site models) and for toluene (1-, and 3-site models).  These models accurately reproduce the bulk density and compressibility of the OPLS model at each resolution. 

In particular, the above figure compares the experimental (orange) density and compressibility with the volume distributions and compressibilities generated by the atomically detailed OPLS model (black) and by several 3-site CG models for heptane.  While the original structure-based model (blue) badly underestimates the density of heptane, the CG model with the DA correction (green) reproduces the OPLS volume fluctuations to within 5%.  The method developed by Dunn and Noid (DN,red) quantitatively reproduces the OPLS volume fluctuations and, even more impressively, matches the experimental density and compressibility with greater accuracy than a top-down model that was explicitly parameterized to reproduce these thermodynamic properties. 

In addition, we have applied this method to parameterize volume-dependent corrections to the virial for application with transferable potentials for hep-tol mixtures.  We applied the extended ensemble framework to parameterize a single set of transferable potentials for describing the intra- and inter-molecular interactions of heptane and toluene in a range of hep-tol mixtures.  Given these transferable potentials, we employed the DN method to parameterize composition-dependent virial corrections for each hep-tol mixture.  The following figure demonstrates that the resulting structure-based CG models (dotted) accurately model the bulk density and compressibility of the atomistic OPLS model (solid) at each composition.    The figure also demonstrates that the coefficients for the pressure corrections (black, DN) appear to follow simple regular solution (blue) considerations.  Accordingly, we can quite accurately predict the appropriate correction for any hep-tol mixture.  By combining the transferable potentials with the regular solution prediction for the pressure correction, we have now developed a highly efficient CG model that accurately describes both the structure and volume fluctuations for any hep-tol mixture. 

Furthermore, we have also worked on developing a top-down (thermodynamics-based) CG model for hep-tol mixtures.  We had previously parameterized 3-site top-down models for both heptane and toluene.  However, we have now discovered that the 3-site top-down toluene model demonstrates significant finite size effects and is not useful for simulating larger systems.  Thus, while the previously published top-down toluene model employs 4 sites, we have found it challenging to determine a more efficient 3-site toluene model.  Accordingly, we are still working to develop a top-down 3-site hep-tol model for comparison with the bottom-up models that we have already parameterized.

Finally, we have also begun work on parameterizing CG models for model asphaltene molecules.  These models will be compatible with the hep-tol models that we have already developed and, thus, will enable efficient simulations of asphaltene self-assembly in a wide range of environments.