(381f) Automatized Determination of Fundamental Eos Based on Molecular Simulations in the Cloud

Thermodynamic data are essential for the design and optimization of chemical engineering processes. Typically, the chemical industry is relying on experimental investigations to generate such data. Depending on the type of fluid, this undertaking is difficult to conduct or in the case of hazardous substances, e.g. being explosive, toxic or mutagenic, this might even be practically impossible. As an attractive alternative route, molecular modelling and simulation on the basis of classical force fields has been proven to be an adequate prediction method due to the steady development effort of the last decades. Along with its powerful predictive capabilities, large datasets of thermophysical data can be rapidly produced with considerably less effort than laboratory measurements.

It has recently been shown that molecular simulation data are useful for the construction of empirical fundamental equation of state (EOS) correlations that are explicit in terms of a thermodynamic potential, which is most often the Helmholtz energy [1]. Thermodynamic potentials are one of the most essential building blocks of thermodynamics, because every other time-independent thermodynamic property can be obtained as a combination of the partial derivatives of the thermodynamic potential with respect to its independent variables. Unfortunately, the exact mathematical expression of a thermodynamic potential, with the exception of very simple systems (e.g. ideal gas), is unknown. The construction of an EOS therefore necessarily means that the mathematical form of such a correlation has to be fitted to available thermodynamic data. Once the parameterized correlation is available, it provides thermodynamic data in a consistent manner. Naturally, the more thermodynamically independent properties are considered during the construction of the EOS, the better the representation of other properties is. Therefore, in order to create such a correlation, a large number of independent thermodynamic data points has to be at disposal. This scenario is in fact the application field for the statistical mechanical formalism proposed by Lustig [2,3]. Using his formalism, any derivative of the Helmholtz energy, which intentionally belongs to the same thermodynamic potential in which the EOS is explicit, can be obtained at a given state point by a single canonical ensemble molecular simulation. Since any thermodynamic property is just a combination of these derivatives, the lack of thermodynamically independent data is not a limiting factor anymore, and EOS can be created purely on the basis of molecular simulation data.

However, setting up molecular simulations and fitting an EOS using simulation data requires considerable expertise. The goal of this work is to automatize and optimize this workflow by combining the molecular simulation tool ms2 [4] with a simple EOS fitting algorithm, and providing an interface for the user that requires minimal knowledge about the background operations, making EOS development more appealing to the chemical industry. To be specific, this was done here by using a cloud-based approach employing a simple graphical user interface for creating the necessary input for simulations. In the background, a job manager efficiently allocates the molecular simulation runs to the available high performance computing (HPC) resources. Once the simulation results are ready, the EOS fitting software takes over control and provides the empirical EOS correlation along with the quality of representation of the simulation data considered. The capabilities of this approach are tested for ethylene oxide and phosgene [5], two particularly hazardous substances of industrial relevance with scarce experimental databases. References

[1] Rutkai, G.; Thol, M.; Lustig, R.; Span, R.; Vrabec, J. The Journal of Chemical Physics 2013, 139, 041102.

[2] Lustig, R. Molecular Simulation 2011, 37, 457â??465.

[3] Lustig, R. Molecular Physics 2012, 110, 3041â??3052.

[4] Glass, C. W.; Reiser, S.; Rutkai, G.; Deublein, S.; Köster, A.; Guevara-Carrion, G.;Wafai, A.; Horsch, M.; Bernreuther, M.; Windmann, T.; Hasse, H.; Vrabec, J. Computer Physics Communications 2014, 185, 3302â??3306.

[5] Rutkai, G.; Vrabec, J. Journal of Chemical & Engineering Data 2015, 60, 2895â??2905.