TOOLS FOR COMPUTATIONAL PHYSICAL-CHEMISTRY
(Permanently under construction - these pages are reorganised.)
Our work in the area of molecular simulations led us to develop a number of applications for the generation of initial polymer structures. The problem of building an initial configuration for an amorphous system of small molecules is, usually, trivial. Even if we start with a perfect crystal, the first few simulation steps will destroy its order and lead to an amorphous state. This is not the case for polymers, though. Polymer systems are characterised by very long relaxation time scales so that the final configuration of even very long simulations may not differ substantially from the initial structure. ‘his is especially true for Molecular Dynamics simulations as well as Monte Carlo simulations without connectivity-altering moves.
To deal with this situation and ensure that polymer simulations are reliable, one should start with a realistic structure from the outset. Torsion and bond angle distributions should be carefully chosen by means of appropriate algorithms while preserving reasonable distances for non-bonded atom pairs. Two of the commonest approaches for building initial polymer structures are the connectivity method (start with a set of disconnected atoms and try to build bonds linking them while respecting a reasonable distribution of bond lengths) and the chain growth approach (start with a set of end atoms or groups and grow chains by successively adding new bonds).
In our implementation we chose the second method because we felt it would be more appropriate for preserving the correct torsion distributions. We have developed a program for the generation of united-atom polyethylene-like systems of arbitrary user-defined polydispersity and another application that builds monodisperse systems of more complex architecture and generates the appropriate DL_POLY input files (FIELD and CONFIG). The latter can treat polydisperse systems too, as mixtures of pure species. Currently, we are working in the direction of unifying the two programs in one application that accepts polydispersity index as input for all species in the system.
As an offshoot of the above project, we have developed a polymerisation emulation program that mimics the growth processes taking place on the catalyst surface in gas phase polymerisation reactors. The user can define the rate of monomer adsorption to control the degree of polymerisation and molecular weight distribution of the polymer material and track them with time. A numerical trick to avoid overlapping atoms allows for the acceleration of the calculations. An example of growing chains subjected to two-dimensional periodic boundary conditions can be seen in this animation
Molecular Dynamics/Monte Carlo
In order to accelerate simulations of complex systems, one can resort to one of the following: use of simplified models that retain the essential features of a given system, improved algorithms and faster hardware platforms combined with the appropriate software to exploit their capabilities. Some times it is sufficient and more convenient to use simple models to deal with special systems and situations. We have developed an application for Molecular Dynamics simulations of molecules in porous materials. The material is modelled as an ensemble of pores defined by an axisymmetric potential that tends to infinity with increasing radial distance and may have a minimum near the pore "walls" or at the pore centre, depending on the input parameters. Currently, the program is being generalised to treat arbitrary porous networks of various shapes.
A Monte Carlo code for generating random conformations of an ideal gas single polymer chain has been extended to emulate the solvent effect by tweaking either the number of bonds between two interacting atoms or the interaction radius. This way, the effect of a good solvent, theta conditions or coil-to-globule transition can be captured. The code is using "global" moves (bond rotations and bond angle alterations) but we are now considering the addition of local moves and more sophisticated connectivity altering schemes.
Recent advances in the area of general purpose GPU computing (exploiting the inherent parallelism in graphics cards together with CPU to accelerate computations) pave the way to "high performance computing for the masses" thanks to the accessible cost of graphics units as compared to previous HPC infrastructure. Currently, an MD program is being developed by us, based on CUDA, a C-like platform that can be used on nVidia graphics cards.
Molecular Dynamics or Monte Carlo trajectories are useless without the appropriate processing of the raw output data. Our post-processing toolbox includes, among others:
POLYANA: (Version published to
An application for the effortless computation of radial
distribution functions, g(r), of molecular centres of mass based on
DL_POLY Molecular Dynamics trajectories. The current version is
described in detail in the above publication. A new version with more
features, such as the generation of potentials of mean force, is
currently tested and will be soon available.
An application, under development, for the computation of self-diffusion coefficients based on the well-known Einstein relation, Stefan-Maxwell diffusivity and the analysis of diffusive jumps based on our method presented in Raptis et al, J. Phys. Chem. B, 2007, 111, p. 13683 and Raptis et al., Mol. Phys., 2012, 110, p. 1171.
Calculation of various thermodynamic properties of pure compounds (density, isothermal compressibility, thermal expansion coefficient, interfacial tension etc.), property profiles, and calculation of solubility of other molecules in a given matrix by means of the Widom insertion (ghost-particle) method.
A novel cluster analysis scheme under development (RDCI), currently tested in a variety of case studies. It is partly described in
the following preprint [PDF]
At the heart of every molecular simulation lies a model equipped with a set of classical potentials, the force-field, that describe the energetics of the varying degrees of freedom in the simulated system. The parameters entering the functional form of these potentials are usually fit against thermodynamic, crystallographic or spectroscopic data. Another approach utilises quantum mechanical methods to obtain a set of optimised geometries and their energies relative to a global minimum. These data are then used as a basis to develop a simplified classical model in which the electronic and often the fastest vibrational degrees of freedom are eliminated in order to speed up calculations. Then, the model has to be validated by comparing the geometries it produces with the ones of the quantum mechanical training set (see for instance Raptis and Melissas J. Phys. Chem. B 2006, 110, p. 14929). We have developed two applications to carry out such a process, namely a Parameter Estimator that fits the desired potentials to quantum mechanical data and a Single Molecule Geometry Optimiser, which carries out molecular mechanics calculations to validate the parameters estimated by the first program.
Study of Multicomponent Hydrocarbon Mixtures Based On DePriester/McWilliams Approximation
In multicomponent systems as those met in fractionation columns one often needs to evaluate certain semi-empirical quantities known as the Distribution Coefficients. One of the first works towards an analytical approximation for them was originally given by Mark L. Williams in his paper "An equation to relate K-factors to pressure and temperature", Chemical Engineering, October 29, 1973, pp.138-140. The Williams approximation was based on the well-known DePriester charts that all chemical engineering students have heard about, so it can be used in their place, albeit within a narrower range of temperatures and pressures.
A simple tcl/tk calculator is provided in this Demo for educational purposes. It was created as a starting point to develop numerical optimization techniques for the study of phase equilibrium in such mixtures.
This is an umbrella for all unsorted material and auxiliary utilities that carry out minor but, actually, essential tasks. Here are some of them:
A library, under development, to help parse "SHELX input files" used in the study of crystal structures.
- Utilities for Format Interconversions
(e.g. DL_POLY to LAMMPS and vice versa, DL_POLY to HOOMD-blue and more...)
- Various tools to handle trajectory files, extract info, plot data and generate slideshows