Research: Overview

Overview of Research Interests


Modern theoretical methods have advanced to an extent that the microscopic details of chemically important processes can be now be investigated using novel algorithms and high speed computer. My research aims to further develop and exploit these emerging capabilities by devising new molecular dynamics and electronic structure techniques and applying these in chemically and biologically important systems. Specific current projects going on in my group are:

1. Ab initio molecular dynamics investigations of the addition of organic molecules to semiconductor surfaces.

2. Ab initio molecular dynamics and path integral studies of the solvation and transport of charged defects in hydrogen-bonded liquids including water, ammonia, and methanol, and at the corresponding vapor/liquid interfaces.

3. Theoretical drug binding studies of novel fullerene based HIV protease inhibitors.

4. Theoretical studies of the structural and dynamical properties of silicate glasses and liquids.

5. Development of field theoretic approaches to orbital localization in real and reciprocal spaces.

6. Development of new, simple real space basis set approaches for density functional based ab initio molecular dynamics.

7. Development of novel adiabatic dynamics and variable transformation based methods for conformational sampling and protein folding.

8. Development of novel dual gridding approaches for application in mixed ab initio/force field based (QM/MM) calculations in chemically and biochemically active systems.

9. Development of theoretical statistical mechanical analysis techniques for non-Hamiltonian dynamical systems used in molecular dynamics calculations.

10. Development of a large-scale, parallel, object-oriented simulation package (PINY_MD) capable of performing force field based, ab initio and path integral molecular dynamics calculations in a variety of ensembles.

Results from these studies are summarized below:



1.The chemistry of hybrid structures composed of organic molecules and semiconductor or metal surfaces is opening up exciting new directions if molecular electronics and nanoscale devices. Theoretical studies of the addition of 1,3-butadiene, a conjugated diene, to the Si(100)-2x1 surface obtained a distribution of addition products. These include a [4+2] Diels-Alder adduct with a single surface dimer, a [4+2]-like adduct in which the butadiene bridges two dimers within a row, a [4+2]-like adduct, in which the butadiene bridges two dimers in neighboring rows, and a [2+2]-like adduct bridging two dimers within a row. The proportion of each adduct in the distribution is in agreement with STM measurements. Moreover, a common underlying mechanism capable of rationalizing the formation of each adduct was proposed. Finally, it was shown that modification of the butadiene by substitution of a fluorine for a hydrogen lowers the free energy barrier for removal of the molecule from the surface, a result that has important implications for surface lithography.









2.Proton transport processes are fundamental in numerous biologically and technologically important systems. Our studies of proton solvation and transport in hydrogen-bonded liquids such as water, methanol and ammonia have elucidated microscopic mechanisms, resolved a number of long-standing controversies, and have uncovered a number of general patterns and principles that appear to govern this process in different liquids.












3. Fullerene-based HIV protease inhibitors are of interest because their structural complementarity to the hydrophobic cavity region of the protease containing the active site. They also show good anti-viral activity and are of relatively low toxicity. Our studies have shown a quantitative connection between the opening and closing of the cavity (through the motion of the flexible flaps) and binding ability and have also characterized the structure of the relatively small amount of water in the cavity region in the presence of a fullerene-based inhibitor.








4. Liquid silicates and their glasses are of industrial and geophysical importance. Our studies have shown that many interesting experimentally known facts about calcium aluminosilicate (which contain both network modifiers and network formers) glasses are also true for the liquid. These include an unusually high fraction of non-bridging oxygens, violation of the aluminum avoidance principle, and high aluminum coordination.





5. Orbital localization in ab initio molecular dynamics has the advantage of yielding orbitals with a chemically intuitive interpretation. In addition, when these orbitals are expanded in a localized basis set, linear scaling methods can be developed. We have shown that the Car-Parrinello Lagrangian lends itself to dynamical orbital localization via techniques used in quantum field theory. specifically, the Car-Parrinello Lagrangian is reformulated so that it is invariant to time-dependent unitary transformations, also known as time-dependent gauge transformations. Then by applying gauge-fixing techniques, it is possible to derive equations of motion that naturally evolve the maximally localized orbitals.




6. Real space approaches in electronic structure have typically employed Gaussian basis sets, which, although very efficient, possess many technical complexities. Plane wave basis sets, on the other hand, although very simple and elegant, often suffer from poor convergence. By employing discrete variable representation (DVR) techniques with plane wave basis sets for long range interactions and our novel screening function methodology, we have developed a simple and rigorous real space based approach for density functional based electronic structure calculations which lead to an order of magnitude gain in efficiency over plane wave methods but retain all of the simplicity of the latter.




7. The conformational sampling problem is one of the computational grand challenges. If solved, problems such as protein and nucleic acid folding will be significantly impacted. By introducing novel variable transformations into the mathematical expression for the statistical mechanical partition function and introducing adiabatic dynamics schemes, we have developed novel molecular dynamics approaches that lead to very large gains in efficiency in the ability to sample statistically relevant conformations of large chain molecules and to compute free energies and other equilibrium properties.




8. Mixed ab initio/force field based (QM/MM) calculations are becoming an increasingly popular method for treating large systems for a small chemically active region (e.g. enzyme catalysis). By employing a novel cardinal B-spline technique similar to that used in particle-mesh Ewald methods, we have developed an efficient method for handling long range interactions in QM/MM calculations.

9. Non-Hamiltonian dynamical systems are now widely used in modern molecular dynamics calculations in order to generate phase distributions consistent with the canonical (NVT), isobaric (NPH) and isothermal-isobaric (NPT) ensembles as well as to model systems away from equilibrium. However, up to now, no rigorous statistical mechanical theory of such systems, allowing for an unambiguous analysis of their phase space distributions, had been developed. By employing geometric phase space techniques, we have developed such a theory and shown how it can be applied in the design of very general non-Hamiltonian based molecular dynamics algorithms.

10. The wide range of projects described above would not be possible without a multi-purpose, flexible, easily extensible software platform in which new developments can be easily implemented and tested and which is also capable of carrying out large-scale simulations on a wide variety of platforms. We have been heavily involved in the development of such a platform (PINY_MD), which is a parallel, object-oriented software package capable of carrying out force field based, ab initio, and QM/MM calculations (including nuclear quantum effects via path integrals) using state-of-the-art algorithms. We are currently extending the capabilities of this package and working to achieve good scaling on very large numbers of processors. The code will soon be made available to the community under a GPL via the Web.