(225e) Molecular Simulation Strategies for Large Scale Thermodynamic Data Generation

Today almost every effort that aims at technological process design and optimization requires a sufficient amount of reliable thermodynamic data for the involved chemical compounds. Traditionally, thermodynamic properties are obtained from experiments and are subsequently expressed in terms of equations of state (EOS).  The development of a highly accurate EOS requires a broad basis of reliable thermodynamic data containing as many different properties as possible. Despite the extensive effort that was invested in measurements over more than a century, the data availability today is still surprisingly low. Among the about 1000 chemical pure compounds that are in technological use [1], a complete thermodynamic knowledge is available for about 10, and an advanced or satisfactory knowledge is available for less than 100 [2]. Considering mixtures, where the number of relevant systems is orders of magnitude larger, the data availability is much worse.

Recent progress in molecular simulation has shown that molecular models (force fields) have powerful predictive capabilities with respect to thermodynamic data [3]. Such information can also be straightforwardly accessed for fluids and states which are experimentally difficult to investigate. Moreover, molecular simulation allows for the generation of large data sets containing consistent information on arbitrary thermodynamic properties at low cost. As there is a dire need for raw thermodynamic data for EOS optimization, the idea to develop a new generation of EOS which are partly based on molecular simulation data comes naturally. In practice, however, the generation of extensive data sets that contain as many different thermodynamic properties as possible may be cumbersome. Standard textbook approaches in the molecular simulation literature indicate specific statistical mechanical ensembles for particular thermodynamic properties [4]. It is true that certain properties have simpler statistical analogs in certain ensembles and may be difficult to derive in other ensembles. However, any thermodynamic property that can be measured in a given statistical mechanical ensemble can also be measured in any other statistical mechanical ensemble [5]. This is a direct consequence of the physical equivalence of various forms of the thermodynamic fundamental equation, e.g. entropy S = S(N, V, E), Helmholtz energy A = A(N, V, T) or Gibbs energy G = G(N, p, T). As a consequence, any arbitrary thermodynamic property can be expressed as a combination of higher-order partial or mixed derivatives of any of the chosen fundamental equation representation from: S(N, V, E), A(N, V, T) or G(N, p, T), etc. [6, 7]. Since the NVT ensemble is a convenient option with respect to parameter settings, the representation A = A(N, V, T) was chosen here, considering the appropriate derivatives of the Helmholtz energy. This formalism was implemented in the molecular simulation tool ms2 [8] for electro-neutral Lennard-Jones based force fields with superimposed electrostatic sites. It can be shown that basically every static thermodynamic property can be determined using only NVT ensemble molecular simulations offering a new route to develop empirical EOS on the basis of hybrid data sets, composed of experimental data and molecular simulation data.

References

[1] K.E. Gubbins, N. Quirke, “Molecular Simulation and Industrial Application”, Gordon and Breach, Amsterdam (1996)

[2] R. Span, “Multiparameter Equations of State”, Springer, Berlin (2000)

[3] http://www.fluidproperties.org, Industrial fluid properties simulation collective (2011)

[4] M. Lagache, P. Ungerer, Phys. Chem. Chem. Phys., 3, 4333-4339 (2001)

[5] H.W. Graben, J.R. Ray, Mol. Phys., 80, 1183-1193 (1993)

[6] R. Lustig, J. Chem. Phys., 100, 3048-3059 (1994)

[7] K. Meier, S. Kabelac, J. Chem. Phys., 124, 064104-064114 (2006)

[8] S. Deublein et al., Comp. Phys. Comm., in press (2011)