MOLECULAR DYNAMICS STUDIES OF SIMPLE MODEL FLUIDS AND WATER CONFINED IN CARBON NANOTUBE

Molecular Dynamics (MD) simulation is one of the most important computational techniques with broad applications in physics, chemistry, chemical engineering, materials design and biological science. Traditional computational chemistry refers to quantum calculations based on solving Schrodinger equations. Later developed Density Functional Theory (DFT) based on solving Kohn-Sham equations became the more popular ab initio calculation technique which could deal with ~1000 atoms by explicitly considering electron interactions. In contrast, MD simulation based on solving classical mechanics equations of motion is a totally different technique in the field of computational chemistry. Electron interactions were implicitly included in the empirical atom-based potential functions and the system size to be investigated can be extended to ~106 atoms. The thermodynamic properties of model fluids are mainly determined by macroscopic quantities, like temperature, pressure, density. The quantum effects on thermodynamic properties like melting point, surface tension are not dominant. In this work, we mainly investigated the melting point, surface tension (liquid-vapor and liquid-solid) of model fluids including Lennard-Jones model, Stockmayer model and a couple of water models (TIP4P/Ew, TIP5P/Ew) by means of MD simulation. In addition, some new structures of water confined in carbon nanotube were discovered and transport behaviors of water and ions through nano-channels were also revealed.

An Extended Kernel for Generalized Multiple-Instance Learning

The multiple-instance learning (MIL) model has been successful in areas such as drug discovery and content-based image-retrieval. Recently, this model was generalized and a corresponding kernel was introduced to learn generalized MIL concepts with a support vector machine. While this kernel enjoyed empirical success, it has limitations in its representation. We extend this kernel by enriching its representation and empirically evaluate our new kernel on data from content-based image retrieval, biological sequence analysis, and drug discovery. We found that our new kernel generalized noticeably better than the old one in content-based image retrieval and biological sequence analysis and was slightly better or even with the old kernel in the other applications, showing that an SVM using this kernel does not overfit despite its richer representation.