<div> <div> <div> <p>The objective of this study is that combination of molecular dynamics (MD) simulations and machine learning to complement each other. In this study, four steps are conducted. </p> <p>First is based on the empirical potentials development in silicon nanowires for theory parts of molecular dynamics. Many-body empirical potentials have been developed for the last three decades, and with the advance of supercomputers, these potentials are expected to be even more useful for the next three decades. Atomistic calculations using empirical potentials can be particularly useful in understanding the structural aspects of Si or Si-H systems, however, existing empirical potentials have many errors of parameters. We propose a novel technique to understand and construct interatomic potentials with an emphasis on parameter fitting, in which the relationship between material properties and potential parameters is explained. The input database has been obtained from density functional theory (DFT) calculations with the Vienna ab initio simulation package (VASP) using the projector augmented-wave method within the generalized gradient approximation. The DFT data are used in the fitting process to guarantee the compatibility within the context of multiscale modeling. </p> <p>Second, application part of MD simulations, enhancement of mechanical properties was focused in this research by using MEAM potentials. For instance, Young’s modulus, ultimate tensile strength, true strain, true stress and stress-strain relationship were calculated for nanosized Cu-precipitates using quenching & partitioning (Q&P) processing and nanosized Fe3C strengthened ultrafine-grained (UFG) ferritic steel. In the stress-strain relationship, the structure of simulation is defined using the constant total number of particles, constant-energy, constant-volume ensemble (NVE) is pulled in the y-direction, or perpendicular to the boundary interface, to increase strain. The strain in increased for a specified number of times in a loop and the stress is calculated at each point before the simulation loops.</p></div></div> </div> <div> <div> <div> <p>Third, based on the MD simulations, machine learning and the peridynamics are applied to prediction of disk damage patterns. The peridynamics is the nonlocal extension of classical continuum mechanics and same as MD model. Especially, FEM is based on the partial differential equations, however, partial derivatives do not exist on crack and damage surfaces. To complement this problem, the peridynamics was used which is based on the integral equations and overcome deficiencies in the modeling of deformation discontinuities. In this study, the forward problem (i), if we have images of damage and crack, crack patterns are predicted by using trained data compared to true solutions which are hit by changing the x and y hitting coordinates on the disk. The inverse problem (ii), if we have images of damage and crack, the corresponding hitting location, indenter velocity and indenter size are predicted by using trained data. Furthermore, we did the regression analysis for the images of the crack patterns with Neural processes to predict the crack patterns. In the regression problem, by representing the results of the variance according to the epochs, it can be confirmed that the result of the variance is decreased by increasing the epoch through the neural processes. Therefore, the result of the training gradually improves, and the ranges of the variance are expressed as 0 to 0.035. The most critical point of this study is that the neural processes makes an accurate prediction even if the information of the training data is missing or not enough. The results show that if the context points are set to 10, 100, 300, and 784, the training information is deliberately omitted such as context points of 10, 100 and 300, and the predictions are different when context points are significantly lower. However, when comparing the results of context points 100 and 784, the predicted results appear to be very similar to each other because of the Gaussian processes in the neural processes. Therefore, if the training data is trained through the Neural processes, the missing information of training data can be supplemented to predict the results. </p> <p>Finally, we predicted the data by applying various data using deep learning as well as MD simulation data. This study applied the deep learning to Cryo-EM images and Line Trip (LT) data with power systems. In this study, deep learning method was applied to reduce the effort of selection of high-quality particles. This study proposes a learning frame structure using deep learning and aims at freeing passively selecting high quality particles as the ultimate goal. For predicting the line trip data and bad data detection, we choose to analyze the frequency signal because suddenly the frequency changes in the power system due to events such as generator trip, line trip or load shedding in large power systems. </p> </div> </div> </div>

Water-mediated interactions between non-polar moieties play a crucial role in driving self-assembly processes such as surfactant micellization, protein folding, and many other diverse phenomena. Among a variety of forces contributing to the self assembly, hydrophobic interactions play a dominant role. Historically, thermodynamic models describing hydrophobic effects have invariably relied on macroscopic thermodynamic properties to infer this molecular behavior. Experimental studies help to probe the spatial correlations between model hydrophobic solutes and to measure their waters of hydration in order to examine structural perturbations in the surrounding water induced by the solute, or to measure directly the attractive forces between hydrophobic surfaces. Further, molecular simulations can be used to derive entropic and enthalpic contributions to the free energy of hydrophobic hydration in terms of water structure surrounding simple, model hydrophobic solutes, such as methane. Based on the results for simple solutes, these methods can now be extended to investigate the hydrophobic hydration of more complex molecular solutes of arbitrary size and shape such as micelles. Atomistic simulations of chemical systems provide a new perspective towards testing the theories behind the ubiquitous phenomenon of hydrophobic effect, and probe the underlying thermodynamic signatures. In this context, my research work delves into the water-mediated interactions leading to the hydrophobic hydration of short chain alkanes, volumetric properties of unfolded polypeptides and self-assembly mechanism in polymer-surfactant systems. The first part of my research involves re-optimization of existing force field interaction parameters for the CHn alkane sites (n=0 to 4) to accurately reproduce the experimental hydration free energies of linear and branched chain alkanes over a range of temperatures. This Hydrophobic Hydration-Alkane (HH-Alkane) model accounts for polarization effects in the alkane hydration and can be extended to polypeptides in water. Subsequent discussions will focus on the results from extensive molecular simulations of tri- and tetrapeptides to quantify the accuracy of the simulation model in capturing the volumetric properties of unfolded polypeptides. Group additivity correlation was used to calculate the partial molar volumes of the neutral sidechains of amino acids, glycine backbone unit and both zwitterionic and N-acetyl/amide terminal units. The simulation results will be compared to the experimental results to validate these observations. In addition, the research explores the self-assembly and aggregation mechanism in anionic sodium dodecyl sulfate (SDS) surfactant- non-ionic Polyethylene Oxide (PEO) and Poly vinyl pyrrolidone (PVP) polymer systems. Potential of mean force calculations at multiple temperatures show an increasing trend in hydrophobic attractions within the polymer-micelle system. Also, these simulations provide interesting insights into the experimentally observed phenomena between the polymers and the micelles starting from pre-formed structure as well as random configurations. / 1 / Lalitanand N. Surampudi

The work presented in my dissertation focuses on the conformational studies of bio-molecules including proteins and DNA using computational approaches. Conformational changes are important in numerous molecular bioprocesses such as recognition, transcription, replication and repair, etc. Proteins recognize specific DNA sequences and upon binding undergo partial or complete folding or partial unfolding in order to find the optimal conformational fit between molecules involved in the complex. In addition to sequence specific recognition, proteins are able to distinguish between subtle differences in local geometry and flexibility associated with DNA that may further affect their binding affinities. Experimental techniques provide high-resolution details to the static structures but the structural dynamics are often not accessible with these methods; but can be probed using computational tools. Various well-established molecular dynamics methods are used in this work to study differences in geometry and mechanical properties of specific systems under unmodified and modified conditions. Briefly, the studies of several protein and DNA systems investigated the importance of local interactions and modifications for the stability, geometry and mechanical properties using standard and enhanced molecular dynamics simulations. In addition to the conformational studies, the development of a new method for enhanced sampling of DNA step parameters and its application to DNA systems is discussed. Chapter 1 reviews the importance of the conformational changes in bioprocesses and the theory behind the computational methods used in this work. In the project presented in chapter 2 unbiased molecular dynamics and replica exchange molecular dynamics are employed to identify the specific local contacts within the inhibitory module of ETS-1. ETS-1 is a human transcription factor important for normal but also malignant cell growth. An increased concentration of this protein is related to a negative prognosis in many cancers. A part of the inhibitory module, inhibitory helix 1 (HI-1) is located on the site of the protein opposite to the DNA binding site and although loosely packed, stays folded in the apo state and unfolds upon ETS-1 binding to DNA. Our study investigated the character and importance of contacts between HI-1 and neighboring helices of the inhibitory module: HI-2 and H4. We also identified a mutant of HI-1, which possessed the higher helical propensity than the original construct. This study supported the experimental findings and enhanced the field by the identification of new potential target for experimental tests of the system, which plausibly inhibits binding to DNA. In the studies discussed in chapters 3-5 the conformational dynamics of DNA under normal conditions and upon specific epigenetic modifications are presented. Since DNA conformation can be accurately described by six base pair step parameters: twist, tilt, roll, shift, slide and rise, these were extensively analyzed and the results elucidated insights into the properties of the systems. In order to enhance unbiased simulations and allow for easier crossing of the energy barriers, we developed and implemented a novel method to control DNA base pair step parameters. With this approach we obtained the free energy estimates of e.g. DNA rearrangements in a more efficient manner. This advanced computational method, supported by standard and additional enhanced techniques, was then applied in the studies of DNA methylation on cytosine or adenine bases and oxidative damage of cytosine.

Pulsed gradient spin-echo (PGSE) nuclear magnetic resonance (NMR) has become a method of choice for the determination of random motion (i.e., translational diffusion or self-diffusion) of molecules and small particles. This thesis focuses on the development of advanced PGSE NMR methods for conducting self-diffusion measurements on magnetically inhomogeneous (i.e., containing materials with different magnetic susceptibilities), aqueous or slowly diffusing samples. PGSE NMR diffusion measurements rely on the accurate encoding of the position of each diffusing particle by the use of pulses of spatially well-defined magnetic gradients. However, for a magnetically inhomogeneous sample, local magnetic field gradients (i.e., background gradients) with unknown intensities and directions, which are normally generated by the differences in magnetic susceptibilities inside and/or around the sample, can also lead to the position encoding of the diffusing particles. This position encoding may couple with the position encoding caused by the applied pulsed gradients and thus cause errors in PGSE NMR diffusion measurements. Although in numerous cases background gradients are assumed to be spatially and temporally constant in order to simply the analysis of the background gradient based position encoding, in general the background gradient experienced by each diffusing particle is non-constant both spatially and temporally because of the complex sample structure and particles diffusing through different magnetic environments. To suppress the deleterious effects of the (non-constant) background gradients, a new stimulated-echo (STE)-based PGSE method with the intensities of pulsed gradients at a magic ratio (MAG), MAG-PGSTE, was developed for the determination of self-diffusion in magnetically inhomogeneous samples. The method was tested on two water saturated glass bead packs (i.e., 212-300 μm and less than 106 μm glass beads). The MAG-PGSTE method was compared to the magic asymmetric gradient STE-based PGSE (MAGSTE or MPFG) method, which is newly developed for the suppression of the deleterious effects of the (non-constant) background gradients in only one transient, and Cotts 13-interval method, one of the most successful suppression methods based on the assumption of constant background gradients, using both glass bead samples. The MAG-PGSTE and MAGSTE method outperformed the Cotts 13-interval method in the measurement of diffusion coefficients; more interestingly, for the less than 106 μm glass bead sample with smaller bead sizes and thus higher background gradients, the MAG-PGSTE method provided higher signal-to-noise ratios and thus better diffusion measurements than the MAGSTE and Cotts 13-interval methods. In addition, when using relatively long pulsed gradients (e.g., 3 ms), the MAG-PGSTE method provided good bead size characterizations. Due to solubility problems, limited sample availability and/or aggregation, solvent (e.g., water) signals in NMR are typically 4-5 orders of magnitude higher than the solute signals. The huge solvent signal not only overlaps with the signals of interest but also saturates the NMR receiver and thus prevents the detection of the weak signals from the molecules of interest. To achieve a high degree of water suppression in PGSE NMR diffusion measurements, a new STE-based PGSE method incorporating the WATERGATE method, which is based on the selective manipulation of the water signal by the use of a binomial-like selective inversion sequence (i.e., a group of radio frequency (RF) pulses with symmetric pulse duration arrangement separated by equal delays), PGSTE-WATERGATE, was developed. The method is simple to set up and particularly suited to measuring diffusion coefficients in aqueous solution, which is commonly required in pharmaceutical and combinatorial applications. The method was tested on a sample containing lysozyme in 10% 2H2O and 90% 1H2O and a sample containing sucrose in 10% 2H2O and 90% 1H2O and provided superb water suppression and accurate diffusion measurements. Importantly, the new method provided the high degree suppression of the phase distortions in NMR spectra caused by the use of selective inversion pulses or sequences. In the NMR spectrum of an aqueous (bio-molecular) sample, especially a protein sample, the water signal almost always overlaps with the signals of interest. Therefore, a good water suppression technique should target only the water signal. The selectivity of the PGSTE-WATERGATE method depends on the selectivity of the binomial-like selective inversion sequence. The traditional way to enhance the selectivity of a binomial-like sequence is to increase the number of RF pulses contained in the binomial-like sequence. This type of selectivity enhancement is very effective but leads to longer sequence durations and thus the signal loss due to NMR relaxation. In this research, two 6-pulse phase-modulated binomial-like inversion sequences were developed by simultaneously optimizing the RF pulse durations and phases instead of increasing the number of the RF pulses. In combination with the excitation sculpting method, which contains two WATERGATE unit and affords the suppression of phase distortions caused by the phase modulations on the binomial-like sequences, both of the new binomial-like sequences outperformed the well-known 3-9- 19 sequence, a 6-pulse binomial-like sequence with a pulse duration ratio of 3:9:19:19:9:3, in selectivity and inversion width. The new sequences provided the similar selectivity and inversion width to the W5 sequence, a 10-pulse binomial-like sequence, but with significantly shorter sequence durations. When used in PGSTEWATERGATE, they afforded highly selective water suppression in diffusion experiments. The hydrodynamics of slowly diffusing molecules (e.g., protein) and small particles (e.g., protein aggregates) has long been of scientific interest due to its importance in the understanding of many biological and biophysical processes such as protein crystallization. However, the diffusion measurements on these molecules and small particles confront the PGSE NMR methods with huge challenges because the diffusion coefficients of these molecules and small particles are normally at or under the lower diffusion determination limit of the PGSE NMR methods, which is mainly controlled by the maximum pulsed gradient strength of the NMR probe (i.e., high pulsed gradient strengths are preferred for measuring slow diffusion), the pulsed gradient durations (i.e., long gradient durations are preferred for measuring slow diffusion), and the observation time (i.e., long observation times are preferred for measuring slow diffusion) which is limited by the NMR relaxation times of the molecules and small particles. To overcome this problem, multi-quantum NMR coherences may be utilized in PGSE NMR experiments as these coherences are more sensitive to the encoding by the pulsed gradients than normal single quantum coherences (e.g., the encoding effect of a pulsed gradient on a single quantum coherence may be doubled by using a double-quantum coherence). In this research, five new multi-quantum coherence encoding STE-based PGSE methods were successfully developed. The new multi-quantum PGSE methods were tested on a sample containing L-[3-13C]-alanine in 2H2O. The quadruple quantum and triple quantum coherences from 13CH3, which can greatly enhance the encoding by pulsed gradients, were successfully selected using these new sequences. This thesis contains eight chapters. In Chapter 1, the focus of this thesis, translational diffusion, is introduced and the difficulties in the PGSE NMR diffusion measurements on magnetically inhomogeneous, aqueous, and slowly diffusing samples are also introduced. In Chapter 2, to facilitate the understanding of how PGSE NMR diffusion measurements work, both the basic principles and simulation methods of NMR and the fundamentals of PGSE NMR diffusion measurements are given. In Chapter 3, a brief review on the current advanced PGSE NMR methods for measuring diffusion in magnetically inhomogeneous, aqueous, and slowly diffusing samples is given. In Chapter 4, the experimental details of the research work in this thesis are given. From Chapter 5 to Chapter 7, newly developed PGSE NMR methods and their applications are discussed in detail. In Chapter 8, the general conclusions are given. / Doctor of Philosophy (PhD)

In this work the phase behaviour of noble gases is studied comprehensively by different molecular simulation methods using different intermolecular potentials. The aim is to investigate three-body effects on the phase behaviour of noble gases. A true two-body potential model (Barker-Fisher-Watts potential) and the three-body potential model (Axilrod-Teller term) have been used. The results obtained from the two-body BFW potential with the three-body Axilrod-Teller potential included for the vapour-liquid and solid-liquid phase equilibrium properties of pure noble gases are compared with the calculations using the Lennard-Jones potential with different suggested parameter values. The results have been compared with experimental data and the best parameter values for simulating the thermodynamic properties of noble gases are found. Three-body effects on the phase behaviour of noble gases are reported for a large range of density, temperature and pressure. Simple relationships have been found between two-body and three-body potential energies for pure fluids and solids. Three-body effects on the vapour-liquid phase equilibrium properties of argon, krypton, xenon and argon-krypton systems are studied by the Gibbs-Duhem integration Monte Carlo method. Three-body effects on the solid-liquid phase equilibrium properties of argon, krypton and xenon are investigated by non-equilibrium and equilibrium molecular dynamics techniques. All the calculations have been compared with experimental data, which show that three-body interactions play an important role in the overall interatomic interactions of noble gases.

Bibliography: leaves [71]-[72]

Thesis (Ph. D.)--University of Missouri-Columbia, 2007. / The entire dissertation/thesis text is included in the research.pdf file; the official abstract appears in the short.pdf file (which also appears in the research.pdf); a non-technical general description, or public abstract, appears in the public.pdf file. Title from title screen of research.pdf file (viewed on December 6, 2007) Vita. Includes bibliographical references.

Thesis (Ph. D. in Chemical Engineering)--Vanderbilt University, Dec. 2007. / Title from title screen. Includes bibliographical references.

This thesis presents a study of applications and techniques for molecular dynamics simulations. Three studies are presented that are intended to improve our ability to simulate larger systems more realistically. A comparison study of two- and three-body potential models for liquid and amorphous Si0��� is presented. The structural, vibrational, and dynamic properties of the substance are compared using two- and three-body potential energy models against experimental results. The three-body interaction does poorly at reproducing the experimental phonon density of states, but better at reproducing the Si-O-Si bond angle distribution. The three-body interaction also produces much higher diffusivities than the two-body interactions. A study of tabulated functions in molecular dynamics is presented. Results show that the use of tabulated functions as a method for accelerating the force and potential energy calculation can be advantageous for interactions above a certain complexity level. The decrease in precision due to the use of tabulated functions is negligible when the tables are sufficiently large. Finally, an investigation into the benefits of multi-threaded programming for molecular dynamics is presented. / Graduation date: 1999

We present a statistical mechanical approach for quantifying thermodynamic properties of proteins in mixed solvents. This approach, based on molecular dynamics simulations which incorporate all atom models and the theory of preferential binding, allows us to compute transfer free energies with experimental accuracy and does not incorporate any adjustable parameters. Specifically, we applied our approach to the model proteins RNase A and T1, and the solvent components water, glycerol, and urea. We found that the observed differences in the binding of glycerol and urea to RNase T1 and A are predominantly a consequence of density differences in the first coordination shell of the protein with the cosolvents, but the second solvation shell also contributes to the overall binding coefficients. The success of this approach in modeling preferential binding indicates that it incorporates the important underlying physics of proteins in mixed solvent systems and that the difficulty in quantitative prediction to date can be surmounted by explicitly incorporating the complex protein-solvent and solvent-solvent interactions. / Singapore-MIT Alliance (SMA)