This document describes the implementation of molecular dynamics simulation as a standard engineering method for physical property estimation in ChemStations' chemical process simulator software. The success of this molecular model hinges on three fundamental premises. (1) The influence of repulsive forces dominates the physical properties. For example, the intermolecular distributions and their fluctuations are primarily influenced by how closely the atoms can approach each other. As another example, the entanglements that strongly influence transport properties occur because molecules cannot pass through each other, but must find a viable path for wriggling past each other. (2) Repulsive effects tend to be specific to the 3D structure of a molecule, necessitating molecular simulation of that specific molecule. In other words, accurately predicting the intermolecular distributions and their fluctuations from a generalized equation (e.g. Peng-Robinson or SAFT) or from integral equation theory (e.g. PRISM) is not reliable for molecules that may be composed of rings and branches. (3) Thermodynamic Perturbation Theory (TPT) is sufficiently accurate that a quantitative treatment of the attractive details of the potential can be derived from theory. Since the TPT contributions are directly related to the intermolecular distributions and their fluctuations, and these are accurately determined by the repulsive forces, this means there is no need to repeat the simulation for every possible specification of the attractive part of the potential. The parameterization of the attractive part of the potential can therefore be pursued in the manner of an engineering equation of state. The development of a prototype is progressing nicely. A preliminary demonstration package is available by clicking the link below.

We are referring to this version as "SPEAD" for Step Potential Equilibria And Dynamics. The developments to date can be best understood by executing a brief demonstration (~10 minutes). The demonstration is divided into two parts: the graphical user interface for defining the ".m3d" file, and the bat file execution for performing the molecular simulations. Instructions for conducting the demonstration follow below. This material is based upon work supported by the National Science Foundation under Grant No. 0226532. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.

The demonstration shows how a single brief simulation is conducted. We have performed many of these simulations to characterize the molecular interactions implicit in Table 1. The columns in Table 1 designated by TraPPE refer to the results of Siepmann and coworkers based on the Lennard-Jones potential model. Obviously, our results for vapor pressure are quite superior. Perhaps more importantly, however, is the temperature range over which our results are applicable. In general, DMD/TPT is accurate to reduced temperatures of 0.45 whereas all other united atom models are inaccurate below reduced temperatures of 0.6. Typical engineering equations of state like the Peng-Robinson equation are accurate to reduced temperatures of 0.45, so it is important to extend to lower reduced temperatures.

We have also completed the characterization of many other group types. Most of these are listed in Table 2. We achieve accuracy similar to the accuracy for the n-alkanes even for branched alkanes, alkenes, alkynes, and alcohols. These results are extremely encouraging. They validate our expectation that we can achieve engineering accuracy from molecular simulations within a fairly short time frame. To emphasize the significance of this accomplishment, we may compare to the status of the MSI software marketed by Accelrys. This is widely regarded as the state of the art software in molecular simulation. At the recent AIChE conference in Reno, I asked a presenter from MSI about their accuracy for vapor pressure. It was simply stated that they were not prepared to discuss vapor pressure at present. The MSI software and its inherent parameterization of the force fields have been under development for at least 15 years. Yet we have surpassed their accuracy and covered many of their molecular types in less than one year. Historically, the parameterization of MSI and its predecessors has focused on distribution functions, density, and solubility parameter. While these are important properties, they (1) tend to be less sensitive to details of the potential function than vapor pressure, and (2) play a less significant role in engineering calculations than vapor pressure.

Note that a substantial portion of the proposed work was directed at transport properties as well as thermodynamic properties. We have begun our analysis of transport properties by DMD/TPT and we again find validation for the perspective that we initially offered. Figure 1a below illustrates the trends in diffusivity for n-alkanes from methane to n-octane and compares to experimental data. Note that the diffusivity values reported here are based on molecular simulations of only the reference fluid, yet we are comparing to actual experimental data at 299K. The close agreement between DMD/TPT and experiment indicates that the disperse attractions really do play a minor role in determining transport properties. We are optimistic that these results will be true for viscosity and thermal conductivity as well.

Finally, it is valuable to demonstrate that the accuracy to be expected of predictions for mixtures by DMD/TPT is comparable to that of correlations by engineering equations. Figure 1b shows that the accuracy for the strongly non-ideal methanol+benzene system is comparable to equations like SAFT or ESD. In fact, if one considers the united atom models of the interaction sites as group contribution parameters, DMD/TPT is much like a blend of SAFT and UNIFAC. Yet there are important distinctions. UNIFAC provides only activity coefficients while SPEAD also provides fugacity coefficients, density, vapor pressure, enthalpy, diffusivity, viscosity, and thermal conductivity. Furthermore, the detailed simulation of the reference fluid structure provides greater specificity and accuracy in characterizing the molecular structure than SAFT or ESD, which effectively assume a string of tangent spheres.

Altogether, our results for this first year could hardly have been better. We are ahead of schedule in developing both the user interface and the statistical mechanics. We look forward to bringing these results to fruition, and possibly testing the software on the Internet, within the coming year.

To view first part of the demonstration, follow these steps:

1. Open a window showing the contents of the directory where the downloaded "Spead.zip" file is located. Move the zip file to C:\. Right click the zip file and select "Extract to C:\SPEAD."
2. Open the directory C:\Spead
3. Double-click the Spead.exe icon. From this point forward, you should view the demonstration instructions from the Help, Help Topics menu within Spead itself. The most current details are maintained in these Help files. The remainder of the discussion below is provided in case you would rather have a better idea of what the demonstration does before downloading and installing it.
4. From the "Commands" menu, select "Display 3D molecule", then select (e.g.) 2,3DimethylButane.mol. You should see a stick model of the molecule showing the hydrogens in white and the carbons in gray. If you see oversized grey spheres, select "View" and clear the "UAM Perspective" check box. Rotate the molecule by pressing x, y, z or shift-x, shift-y, shift-z. Zoom out by right-clicking the molecule and selecting "Zoom Out."
5. From the "View" menu, select "UAM Perspective." Zoom out by right-clicking the screen and selecting "Zoom to Fit." This displays the molecule in the manner that the DMD/TPT program models it. "UAM" stands for united atom model. Note that the hydrogen atoms are no longer displayed and the molecule looks more like a blob than a branched chain. This is because the diameters of the methylene groups are so large that they actually overlap significantly. Believe it or not, this is a more accurate picture of the way this molecule interacts with other molecules. Rotate the molecule by pressing x, y, z or shift-x, shift-y, shift-z. Zoom out by right-clicking the molecule and selecting "Zoom Out."
6. From the "Commands" menu, select "Build Groups." Spead identifies the groups of UAM site types that occur in the molecule, e.g., >CH- and CH3b. Note that CH3b is a secondary methyl group, as opposed to a primary methyl group that would occur in n-butane, or a tertiary methyl group in neopentane. We find that the overlaps between methyl groups differ for primary, secondary, and tertiary methyl groups and this affects the van der Waals interactions between the molecules. Siepmann and coworkers have made similar findings using Lennard-Jones models of the molecular interactions. It may also be possible that the methyl group in, say, methanol is special and needs a dedicated subgroup. The program is organized to easily allow 100 subgroups for each main group, where CH3 and >CH- are examples of main groups. Spead also identifies the coordinates of the atoms in Angstrom units.
7. Close NotePad by clicking the X in the upper-right corner of the NotePad window. You will be prompted to create an ".m3d" file. Select "Yes." NotePad will display the diameters of each group, the coordinates in nanometers, and the main and sub group integer identifiers. The .m3d file is the main one referred to by DMD/TPT. Close NotePad by clicking the X in the upper-right corner of the NotePad window. From the "File" menu, select "close." This will bring you back to the blank "SPEAD Utility" screen.

Note that the interface executed up to this point has very little to do with molecular simulation. It is primarily a graphical user interface (GUI) for downloading and displaying molecular graphics.

The primary focus at The University of Akron has been the operations that occur after creating the .m3d file. To illustrate, continue from the last step above.