Browsing by Subject "Molecular Dynamics"

Now showing 1 - 13 of 13

Crystallographic Information Files (CIF) with atomistic models of interacting double-walled carbon nanotubes.
(2021-07-19) Dumitrica, Traian; dtraian@umn.edu; Dumitrica, Traian; Computational Nanomechanics Laboratory
The dataset provides Crystallographic Information Files (CIF) atomistic models of interacting double-walled carbon nanotubes diameter monodisperse and different mean diameter and standard deviations. These models can be used to reproduce Figures 6-10 of the referenced paper. The files can be visualized with molecular visualizers like OVITO and JMOL.
Development of Multistate Density Functional Theory for Photochemistry and Vibrational Dynamics using Quantum Vibration Perturbation Theory
(2018-01) Grofe, Adam
This dissertation is contains two separates areas of research. The first three Chapters focus on the development and several proof-of-concepts for multistate density functional theory (MSDFT) regarding excited state chemistry. When performing configuration interaction (CI) on density functional theory (DFT), there is a danger of double-counting with regard to correlation. It is shown that it is possible to remove double counting in a systematic way by combining wave function theory with DFT, and that MSDFT has the correct topology with regard to conical intersection because it fundamentally includes coupling between the excited state densities and the ground state density, which is not true for time-dependent DFT. This was revealed for the dissociation of ammonia, and Jahn-Teller conical intersections. Block localized DFT was used to assemble diabatic states that can be optimized separately and permits the use of chemical intuition by defining the states in a relatively straightforward manner. Finally, the MSDFT formalism was used to optimize spin-multiplet states with the correct degeneracy, which can be difficult in DFT due to the nonlinearity of the current exchange-correlation functionals. The second section of this dissertation regard vibrational dynamics. There is a multitude of methods for computing the vibrational frequency, but very few that simultaneously model anharmonicity and nuclear quantum effects in a manner that is efficient enough for computing frequency trajectories. Quantum vibration perturbation (QVP) theory satisfies all three of these criteria. This is accomplished by utilizing two approximations: (1) A discrete variable representation of the nuclear wave function is used that only requires single point energy calculations to optimize the wave function, (2) Perturbation theory is used to update the wave function across a set of configurations, which circumvents the need to solve the Schrödinger equation. The first application of this model is on hydrochloric acid in minimal solvation shells (water), and acetone bulk solvation dynamics. An implementation strategy is presented that allows for a reference normal mode to be applied across a trajectory. Then QVP is used to probe the condensed phase solvation dynamics of a carbonyl stretch and two silane stretches.
The Effect of α-Synuclein on Lipid Membrane Properties Characterized by Molecular Dynamics and Atomic Force Microscopy
(2018-08) Brummel, Benjamin
The protein α-synuclein (αSyn), primarily recognized for its link to neurodegenerative disorders, has multiple reported functions. One well-established role of αSyn is its ability to bind and remodel lipid membranes. This ability has been characterized in synthetic lipid bilayers and has been observed both in cellular and in vivo models. The native environment of αSyn—the presynaptic terminal of neurons—contains mitochondria and synaptic vesicles, which have unique membranes that differ from previously studied models. The goal of this dissertation was to characterize how lipids enriched in synaptic vesicles and mitochondria affect how αSyn changes membrane properties. First, molecular dynamics (MD) simulations of synaptic vesicle-mimic bilayers showed how lipids with polyunsaturated fatty acids modify membrane properties and interact with αSyn. Next, tubulation experiments were combined with MD simulations to explore how αSyn remodels bilayers containing cardiolipin and phosphatidylethanolamine, two lipids enriched in mitochondria. Finally, methods were developed to characterize lipid vesicle mechanical properties using pulsed force mode (PFM) atomic force microscopy (AFM). This work provides insight into the specifics of how αSyn affects the properties of synaptic vesicle and mitochondrial membranes and demonstrates how PFM-AFM can identify the mechanical properties of lipid vesicles.
Electromechanical characterization of quasi-one dimensional nanostructures of silicon, carbon, and molybdenum disulfide via symmetry-adapted tight-binding molecular dynamics.
(2010-11) Zhang, Dong-Bo
With a newly developed symmetry-adapted tight-binding molecular dynamics (MD) capability, we performed microscopic calculations on a variety of quasi-one dimensional silicon, carbon, and molybdenum disulfide nanostructures. In symmetry-adapted MD the helical symmetry instead of the standard translational symmetry is used. In the considered nanostructures, equivalent calculations can now be performed with a substantial smaller, in terms of the number of atoms, repeating domain. The symmetry-adapted method was utilized in the studied highlighted below. The stability of the most promising ground state candidate silicon nanowires with less than 10 nm in diameter was comparatively studied with with nonorthogonal tight-binding and classical potential models. The computationally expensive tight-binding treatment becomes tractable due to the substantial simplifications introduced by the presented symmetry-adapted scheme. It indicates that the achiral polycrystalline of fivefold symmetry and the hexagonal (wurtzite) wires of threefold symmetry are the most favorable quasi-one-dimensional silicon arrangements. Quantitative differences with the classical model description are noted over the whole diameter range. Using a Wulff energy decomposition approach it is revealed that these differences are caused by the inability of the classical potential to accurately describe the interaction of Si atoms on surfaces and strained morphologies. The elastic response for a large catalog of carbon nanotubes subjected to axial and torsional strain was next derived from tomistic calculations that rely on an accurate tight-binding description of the covalent binding. The application of the computationally expensive quantum treatment is possible due to the simplification in the number of atoms introduced by accounting for the helical and ngular symmetries exhibited by the elastically deformed nanotubes. The elasticity of nanotubes larger than 1.25 nm in diameter can be represented with an isotropic elastic continuum. The torsional plastic response of single-walled carbon nanotubes is studied with tight-binding objective molecular dynamics. In contrast with plasticity under elongation and bending, a torsionally deformed carbon nanotube can slip along a nearly axial helical path, which introduces a distinct (+1,−1) change in wrapping indexes. The low energy realization occurs without loss in mass via nucleation of a 5-7-7-5 dislocation dipole, followed by glide of 5-7 kinks. The possibility of nearly axial glide is supported by the obtained dependence of the plasticity onset on chirality and handedness and by the presented calculations showing the energetic advantage of the slip path and of the initial glide steps. Symmetry-adaptedMD combined with density-functional-based tight-binding made possible to compute chiral nanotubes as axial-screw dislocations. This enabled the surprising revelation of a large catalog of MoS2 nanotubes that lack the prescribed translational symmetry and exhibit chirality-dependent electronic band-gaps and elastic constants. Helical symmetry emerges as the natural property to rely on when studying quasi-one dimensional nanomaterials formally derived or grown via screw dislocations. The nonlinear elastic response of carbon nanotubes in torsion was derived with the symmetry-adapted MD and a density-functional-based tight-binding model. The critical strain beyond which tubes behave nonlinearly, the most favorable rippling morphology, and the twist- and morphology-related changes in fundamental band gap were identified from a rigorous atomistic description. There is a sharply contrasting behavior in the electronic response: while in single-walled tubes the band-gap variations are dominated by rippling, multiwalled tubes with small cores exhibit an unexpected insensitivity. Results are assistive for experiments performed on nanotubes-pedal devices. Despite its importance, little is known about how complex deformation modes alter the intrinsic electronic states of carbon nanotubes. We considered the rippling deformation mode characterized by helicoidal furrows and ridges and elucidate that a new intralayer strain effect rather than the known bilayer coupling and &sigma-&pi orbital mixing effects dominates its gapping. When an effective shear strain is used, it is possible to link both the electrical and the mechanical response of the complex rippled morphology to the known behavior of cylindrical tubes. Moving on to graphene, to describe the strain stored in helical nanoribbons, we supplement the standard elasticity concepts with an effective tensional strain. Using &pi -orbital tight binding and objective molecular dynamics coupled with density functional theory, we show that twisting couples the frontier conduction and valence bands, resulting in band-gap modulations. In spite of the edges and ridges of the helical nanoribbons, from the effective strain perspective these band-gap modulations appear strikingly similar with those exhibited by the seamless carbon nanotubes.
A Fungal secretome tailored to enable a radical (oxidative) wood decay mechanism in brown rot fungi
(2021-07) Castano Uruena, Jesus
Wood is one of the most important carbon sinks on earth, and it is composed mainly of cellulose, hemicellulose, and lignin, which together form what is known as the lignocellulosic complex. The lignin component in wood is highly recalcitrant and prevents wood from being degraded by most organisms. Wood degrading fungi have evolved mechanisms to handle the lignin barrier effectively and harness the carbon present in wood, which makes them also plausible biological templates for industry-related applications such as the production of biofuels. However, there are still several knowledge gaps about how fungi orchestrate the complex mechanisms behind wood decay, which prevents further usage of these fungi in biotechnological fields. Wood degrading fungi were initially classified by the physical properties of the rotted wood as white, brown, and soft rot fungi – with brown and white rot fungi being the most efficient. These properties are correlated with important genetic and regulatory differences that distinguish the different modes of decay. Recently, it was shown that brown rot fungi evolved from white rot fungi losing an important number of carbohydrate-degrading enzymes (CAZy) and oxidoreductases. This loss was accompanied by regulatory changes that included the overexpression of retained CAZys, and the development of a two-step mechanism that segregated a Fenton-based oxidative phase from a hydrolytic phase in early and late wood decay, respectively. However, although some main differences between white and brown rot fungi have been identified, several details remain obscure. For example, it is unknown how brown rot fungi manage to produce highly oxidative and unspecific hydroxyl radicals while producing some glycosyl hydrolases necessary for the depolymerization of pectin and hemicellulose at early decay stages. It seems reasonable, that brown rot fungi could protect their enzymes by making them more naturally tolerant of reactive oxygen species (ROS) radicals than other wood degraders associated with a different rot type. Also, although transcriptomics and proteomics data suggest brown rot fungi use some Fenton chemistry at early decay, there is little information about how these fungi regulate the concentration of the chemical species that enable this chemistry. This dissertation contributes to widen the knowledge about fungal wood decay mechanisms, especially those used by brown rot fungi. For starters, by studying the wood decay progression with the brown rot fungus Rhodonia placenta, we found that this fungus displays different mechanisms to harness the use of ROS during wood decay without inflicting damage on itself. First, R. placenta controls the extracellular production of ROS by regulating the concentrations of H2O2 and Fe2+ in the media (avoidance mechanism). Second, this fungus presents a high antioxidant capacity as decay progresses, potentially to quench any possible leaks of ROS from earlier decay stages (suppression mechanism). Thirdly, several R. placenta secreted CAZys, important for early wood decay, displayed tolerance of high concentrations of ROS compared to the soft rot fungus Trichoderma reesei (an industrially relevant cellulase producer), which enables these enzymes to work under harsh operating conditions (tolerance mechanism). Collectively, this indicates that R. placenta uses avoidance, suppression and tolerance mechanisms, extracellularly, to complement intracellular differential expression, enabling this brown rot fungus to use ROS to degrade wood. After finding these results, we decided to incorporate a white rot fungus (Trametes versicolor) in the tolerance comparison and observed similar results, with tolerance of ROS only present in the side-chain hemicellulases of R. placenta. Proteomics analysis, meant to examine the presence of oxidative modifications induced after an in-vitro oxidative treatment, revealed that not only side-chain hemicellulases but also several other enzymes such as laccases, glutathione-S-transferases, and proteases were differentially tolerant of ROS compared to white and soft rot fungi. This suggest that the fungal secretome in brown rot fungi has been tailored as a whole to better endure the presence of ROS. Also, we found that several of the oxidative modifications that occurred in the glycosyl hydrolases of T. reesei and T. versicolor happened in amino acid residues in the vicinity of the active site, which can be linked to the loss of enzyme activity after the oxidative treatment. In a follow-up study addressing the effects of these modifications, we used molecular dynamics to understand the effect of some of the oxidative modifications in an α-L-arabinofuranosidase of T. reesei. Even though the number of oxidative modifications that we could include in the modeled protein was limited due to the lack of force field parameters, the simulations still showed some of the negative potential outcomes when a number of amino acid residues become oxidized. For instance, there were significant alterations of the conformational stability of the protein when oxidized, as evidenced by changes in root mean square deviation (RMSD) and principal component analyses (PCA) trajectories. Likewise, enzyme-ligand interactions such as hydrogen bonds were greatly reduced in quantity and quality in the oxidized protein. In addition, free energy landscape plots showed that there was a more rugged energy surface in the oxidized protein, implying a less favorable reaction pathway. Collectively, these results revealed the basis for the loss of function in the α-L-arabinofuranosidase of the commercially-relevant fungus T. reesei. Finally, metabolomics experiments were carried to find out whether the different modes of decay translated into signature metabolite profiles that could be assigned to either brown or white rot fungi. For this purpose, we cultured two brown rot fungi (R. placenta and Gloephylum trabeum) and two white rot fungi (Pleurotus ostreatus and T. versicolor). The results showed that brown rot fungi have a distinct metabolite pattern at late decay stages that clearly distinguish them from white rot fungi. Different metabolites such as organic acids, sugars, pyranones, and furanones contributed to this result. The finding of several pyranones and furanones being differentially more abundant in brown rot fungi was interesting since it agrees with the expansion of polyketide synthase genes in brown rot fungi. In contrast to brown rot fungi, we could not find a lot of similarities in white rot fungi as deduced by the PCA plots and heatmaps. However, some commonalities were evident such as the presence of galactitol as a potential biomarker, and the higher efficiency of these fungi at removing phenolic compounds originally found in undecayed wood. When focusing on both types of decay, we found that wood degrading fungi tend to accumulate sugars and carboxylic acids at late decay stages. Also, as fungal decay progresses, we observed an accumulation of different furans such as furfural or 5-methylfurfural in all fungi.
Mesoscopic Distinct Element Method for Multiscale Modeling of Carbon Nanotubes
(2019-07) xu, hao
Super strong lightweight material systems comprising carbon nanotubes (CNTs) are especially suitable for aerospace applications. Assembles of CNTs obtained by mechanically stretching the CNT sheets, represent a promising material platform for developing composite materials with mechanical attributes approaching those of individual CNTs. In this quest, the guidance power of computational materials modelling is critical. Ideally one would like to investigate CNT assembles with all atom simulation methods, but this approach is computationally prohibitive. Due to the inherent spatial and temporal limitations of atomistic modeling and the lack of mesoscale models, mesoscopic simulation methods for CNT systems are missing. My work focuses on deriving ultra-coarse-grained models based on mesoscopic dintinct element method (mDEM). Our mDEM model is informed by atomistic data obtained with molecular dynamics (MD) and density functional theory-based tight-binding (DFTB) objective molecular modeling. Our mDEM model is capable of reproducing the atomistic elastic and frictional properties of CNTs. With the mDEM model, tensile tests of mesoscale CNT network were carried out, showing results in good agreement with experiments. The tensile tests revealed nanofriction was a key factor deciding the load transfer of CNT network. Our mDEM model serves as a powerful tool to expand the understanding and guide the development of CNT materials.
Molecular dynamics simulations of membranes and membrane proteins.
(2011-07) Perlmutter, Jason David
Membranes composed of a lipid bilayer and embedded proteins are ubiquitous in nature. They form the barrier which demarcates every cell from its environment and separates the distinct organelles within eukaryotic cells, implicating membranes in a wide range of biological processes. The function of membranes and membrane proteins are determined by their structure, and the central focus of this thesis is the use of computational molecular dynamics simulations to study experimentally inaccessible details of membrane structure. Firstly, we have simulated ternary lipid bilayers containing steroids with a range of headgroup hydrophobicities, observing a correlation between the membrane lateral organization and the orientation of the steroid. Based on these results we suggest a general framework to distinguish previously identified steroid domain promoters and inhibitors. Secondly, we investigate the role of interleaflet coupling in membrane structure. This includes describing a compositional dependence to the interleaflet organization of phase separated membranes, as well as investigating structural perturbations due to interleaflet differences in composition. Thirdly, we demonstrate a strategy for obtaining experimental verification through low angle X-ray scattering and discuss its potential application to complex phase separated mixtures. The second focus of this thesis is considering how the structural features of membranes affect the behavior of membrane proteins. The membrane protein α-Synuclein is of wide interest due to its association with Parkinson's Disease, but its physiological function remains unknown. A third focus of this thesis is the structure of membrane-mimetics, such as detergent micelles and amphipathic polymers, which are commonly used for the stabilization of membrane proteins. Their potential distinct influence on protein behavior currently remains an unresolved hindrance to experimental characterization. The simulations presented herein demonstrate a distinct effect of membrane curvature on α-Synuclein behavior and suggest a potential role in regulating vesicle fusion. Collectively, these simulations of model systems offer insight into the fundamental features which determine the behavior of complex biological membranes.
Nanoscale mechanics of helical and angular structures: exploring and expanding the capabilities of objective molecular dynamics
(2014-06) Nikiforov, Ilia Andreyevich
Objective molecular dynamics (OMD) is a recently developed generalization of the traditionally employed periodic boundary conditions (PBC) used in atomistic simulations. OMD allows for helical and/or rotational symmetries to be exploited in addition to translational symmetry. These symmetries are especially prevalent in nanostructures, and OMD enables or facilitates many simulations that were previously dicult or impossible to carry out. This includes simulations of pristine structures that inherently possess helical and/or angular symmetries (such as nanotubes), structures that contain defects (such as screw disclocations) or stuctures that are subjected to deformations (such as bending or torsion). OMD is already a powerful method, having been coupled with the quantum-mechanical density functional-based tight-binding (DFTB) method, as well as with classical potentials. In this work, these capabilities are used to investigate electromechanical properties of silicon nanowires, treating the mechanical simulation results in the context of continuum mechanics. The bending of graphene is studied, and the underlying molecular orbital mechanisms are investigated. The implications of the results on other simulation methods used to study bending of graphene are discussed. OMD is used in an experimental-theoretical collaboration studying the kinking of graphene and boron nitride nanoribbons. The simulations elucidate and quantify the underlying mechanism behind the kinking seen in experiments.Although theoretically, as a generalization, OMD can match or exceed the capabilities of PBC in all cases, OMD is a new method. Thus, practical implementation must be tackled to expand the capabilities of OMD to new simulation methods and simulation types. In this work, OMD is expanded to allow coupling with self-consistent charge (SCC) DFTB, by developing and implementing the required summation formulas for electrostatic and dispersion interactions. SCC-DFTB is an improved form of the standard DFTB method which includes explicit consideration of charge transfer between atoms. This allows for improved description of heteronuclear materials. To demonstrate this capability, proof-of-concept calculations are carried out on a boron nitride nanotube, a screw-dislocated zinc oxide nanowire, and a single-helix DNA molecule.Finally, preliminary development of heat current calculations under OMD is presented. Heat current calculations are used for calculating thermal conductivity of materials from equilibrium molecular dynamics. So far, heat current calculations have been implemented for the pairwise Lennard-Jones potential. The next development (not yet implemented) is the extension of the heat current calculation under OMD to the Tersoff interatomic potential. The challenges and considerations involved are discussed.
Quantum chemical studies of actinides and lanthanides: from small molecules to nanoclusters
(2013-06) Vlaisavljevich, Bess
Research into actinides is of high interest because of their potential applications as an energy source and for the environmental implications therein. Global concern has arisen since the development of the actinide concept in the 1940s led to the industrial scale use of the commercial nuclear energy cycle and nuclear weapons production. Large quantities of waste have been generated from these processes inspiring efforts to address fundamental questions in actinide science. In this regard, the objective of this work is to use theory to provide insight and predictions into actinide chemistry, where experimental work is extremely challenging because of the intrinsic difficulties of the experiments themselves and the safety issues associated with this type of chemistry. This thesis is a collection of theoretical studies of actinide chemistry falling into three categories: quantum chemical and matrix isolation studies of small molecules, the electronic structure of organoactinide systems, and uranyl peroxide nanoclusters and other solid state actinide compounds. The work herein not only spans a wide range of systems size but also investigates a range of chemical problems. Various quantum chemical approaches have been employed. Wave function-based methods have been used to study the electronic structure of actinide containing molecules of small to middle-size. Among these methods, the complete active space self consistent field (CASSCF) approach with corrections from second-order perturbation theory (CASPT2), the generalized active space SCF (GASSCF) approach, and Møller-Plesset second-order perturbation theory (MP2) have been employed. Likewise, density functional theory (DFT) has been used along with analysis tools like bond energy decomposition, bond orders, and Bader's Atoms in Molecules. From these quantum chemical results, comparison with experimentally obtained structures and spectra are made.
Simulation data for "Adsorption of Charge Sequence-Specific Polydisperse Polyelectrolytes"
(2021-05-28) Sethuraman, Vaidyanathan; Zheng, David; Morse, David C; Dorfman, Kevin D; vsethura@umn.edu; Sethuraman, Vaidyanathan; Dorfman Lab
Simulation data and codes for reproducing the data in the paper titled "Adsorption of Charge Sequence-Specific Polydisperse Polyelectrolytes".
Thermal heat transport characterization for macroscale, microscale, and nanoscale heat conduction
(2008-12) Anderson, Christianne Vanessa Duim Riberiro
Several theoretical and experimental methods for predicting the thermal conductivity of thin dielectric ¯lms and carbon nanotubes are presented based on two schools of thought: (1) the physics of the Boltzmann Transport Equation (BTE), and (2) Molec- ular Dynamics (MD) simulations. First, in relation to models based on the BTE, this thesis highlights temporal and spatial scale issues by looking at a uni¯ed theory that bridges physical aspects presented in the Fourier and Cattaneo models. This newly developed uni¯ed model is the so called C- and F-Processes heat conduction model. The model introduces the dimensionless heat conduction model number which is the ratio of thermal conductivity of the fast heat carrier F-Processes to the total thermal conductivity comprised of both fast F-processes and slow heat carrier C-processes. Prior work has claimed that macroscopic heat transfer models cannot explain mi- croscale heat transfer. First, this dissertation provides arguments by showing how the C-F model is able to extend the use of \macroscopic" constitutive relations for the prediction of thermal conductivity at the \microscopic" level for thin ¯lms, mul- tilayer structures, and includes both dielectrics and metals. Second, the in°uence of external mechanical strain on the thermal conductivity of single-wall carbon nan- otubes is studied using direct molecular dynamics simulations with Terso®-Brenner potential for C-C interactions. Three types of external mechanical strain, namely, axial compression, tension, and torsion are studied. In all three cases, the thermal conductivity does not degrade much, i.e., it remains within 10% of the pristine nan- otube values for lower applied strain below the values required for structural collapse. At higher applied strain, structural collapse occurs, and major reductions in the ob- served thermal conductivity for axially compressed and torsionally twisted tubes are observed.
Thermobiomechanics of arteries
(2008-11) Venkatasubramanian, Ramji T.
Conventional treatments for arterial diseases, such as balloon angioplasty, often result in restenosis or re-narrowing of the arteries. In the last few years, the clinical importance of thermal therapies for atherosclerosis involving both freezing (cryoplasty) and heating (in-stent heating) has increased significantly because of their potential to control or minimize restenosis. An alternative to these therapies includes replacing the diseased artery through preserved arterial grafts which brings with it the need to effectively preserve them. Cryopreservation, i.e. preservation of tissues by freezing to very low temperatures, has therefore become an important problem in medicine. As mechanical properties of arteries play a large role in blood flow, a complete understanding of the biomechanical changes following thermal treatments and the underlying mechanisms is essential for further optimization of these treatments through controlling biomechanical changes. The objective of this dissertation was to quantify the biomechanical changes and investigate the underlying mechanisms post freeze-thaw. In this dissertation, the following specific aims were pursued: 1. Quantification of freeze-thaw induced biomechanical changes in arteries 2. Investigation of underlying mechanisms of thermobiomechanics SA1 involved quantification of freeze-thaw induced mechanical property changes in arteries using both uniaxial tensile tests and indentation. While uniaxial tensile tests were chosen for relatively easy sample preparation and testing, indentation was performed in order to study a more localized biomechanical response while characterizing the diseased artery response. SA2 involved investigation of the mechanisms underlying the biomechanical changes. This primarily involved understanding the changes to the collagen matrix and SMCs following thermal treatments. Changes to collagen matrix stability were assessed by quantifying the changes to the amide-III band using the FTIR spectroscopy. Changes in SMC function were studied from the response of arteries to norepinephrine and acetylcholine. Finally, MD simulations were performed as a tool to further investigate dehydration induced increase in thermal stability of the collagen matrix due to freeze-thaw at the molecular level. The important conclusions of this dissertation research are: 1. Freeze-thaw causes significant stiffening of the arteries. While, significant increase in the physiological elastic modulus (and reduction in toe region) was observed in the uniaxial tensile response, the peak and equilibrium modulus measured from indentation increased significantly following freeze-thaw. 2. Freeze-thaw induces significant changes in the collagen matrix and smooth muscle cells (SMCs) that are arguably the most important components of an artery. While dehydration accompanied by increased thermal stability was observed following freeze-thaw in the collagen matrix, it caused complete destruction of SMCs measured through loss in function. 3. At the molecular level, dehydration due to freeze-thaw (or any osmotic treatments) results in formation of new sidechain-backbone hydrogen bonds that are typically absent under hydrated conditions. These newly formed intra-protein hydrogen bonds in the absence of water molecules increase the thermal stability of the tropocollagen molecule.
Transferability of Empirical Potentials and the Knowledgebase of Interatomic Models (KIM)
(2016-04) Karls, Daniel
Empirical potentials have proven to be an indispensable tool in understanding complex material behavior at the atomic scale due to their unrivaled computational efficiency. However, as they are currently used in the materials community, the realization of their full utility is stifled by a number of implementational difficulties. An emerging project specifically aimed to address these problems is the Knowledgebase of Interatomic Models (KIM). The primary purpose of KIM is to serve as an open-source, publically accessible repository of standardized implementations of empirical potentials (Models), simulation codes which use them to compute material properties (Tests), and first-principles/experimental data corresponding to these properties (Reference Data). Aside from eliminating the redundant expenditure of scientific resources and the irreproducibility of results computed using empirical potentials, a unique benefit offered by KIM is the ability to gain a further understanding of a Model's transferability, i.e. its ability to make accurate predictions for material properties which it was not fitted to reproduce. In the present work, we begin by surveying the various classes of mathematical representations of atomic environments which are used to define empirical potentials. We then proceed to offer a broad characterization of empirical potentials in the context of machine learning which reveals three distinct categories with which any potential may be associated. Combining one of the aforementioned representations of atomic environments with a suitable regression technique, we define the Regression Algorithm for Transferability Estimation (RATE), which permits a quantitative estimation of the transferability of an arbitrary potential. Finally, we demonstrate the application of RATE to a specific training set consisting of bulk structures, clusters, surfaces, and nanostructures of silicon. A specific analysis of the underlying quantities inferred by RATE which are used to characterize transferability is provided.

University Digital Conservancy

Browse by Subject

Browsing by Subject "Molecular Dynamics"