Orientador: Anna Diva Plasencia Lotufo / Co-orientador: Mara Lúcia Martins Lopes / Banca: Maria do Carmo Gomes da Silveira / Banca: Elizete de Andrade Amorim / Resumo: Com o aumento na demanda por energia elétrica o planejamento de geração, transmissão e distribuição bem como a operação são importantes para uma prestação de serviços de forma eficiente, econômica e confiável. Uma das ferramentas para gestão desses recursos são os modelos de previsão de séries temporais. Há diversos modelos na literatura, como os modelos de regressão, modelos estatísticos, dentre outros. Outro modelo que vem se destacando na literatura é a previsão utilizando as redes neurais artificiais, pela sua capacidade de aprendizado. As redes neurais possuem várias arquiteturas, e uma em particular, que é considerada padrão na literatura, é a rede perceptron multicamadas com o algoritmo backpropagation. O presente trabalho propõe uma rede neural híbrida composta pelo método de regressão linear com seleção de variáveis stepwise juntamente com a rede neural artificial perceptron multicamadas com o algoritmo backpropagation. O objetivo é obter um método simples e eficaz para redução de variáveis sem perda de qualidade de previsão. O modelo de regressão linear com o método de seleção de variáveis stepwise tem a função de selecionar as variáveis mais relevantes para compor o conjunto de dados de entrada para treinamento/diagnóstico da rede neural perceptron multicamadas com o algoritmo backpropagation que, consequentemente, é a responsável em realizar a previsão de carga elétrica. Com esta proposta busca-se uma metodologia que seja capaz de reduzir a quantidade de variáveis de entrada da rede neural e obter resultados satisfatórios, ou seja, boas previsões. Para corroborar a metodologia proposta são apresentados os resultados da previsão de carga elétrica a curto prazo em um período de 24 e 48 horas a frente, considerando-se os dados históricos de uma companhia do setor elétrico / Abstract: With the increase in electric energy demand the planning of generation, transmission and distribution as well as the operation are important to provide services efficiently, economically and reliably. One of the tools to manage those resources are time series model forecasting. There are several models in the literature, as the regression models, statistical models, among others. Other model that has been highlighted in the literature is the forecasting using artificial neural network, due to the capacity of learning. Neural networks have several architectures, and one in particular, that is considered standard in the literature is the multilayer perceptron network with the backpropagation algorithm. The present work proposes a hybrid neural network composed by the linear regression method with stepwise variable selection with the multilayer perceptron artificial neural network with the backpropagation algorithm. The aim is to provide a simple and effective method to reduce the variables without losing the forecasting quality. The function of the linear regression model with stepwise variable selection is to select the more relevant variables to compose the input data set to training/ diagnostic of the multilayer perceptron neural network with the backpropagation algorithm that, consequently, is the responsible to realize the electric load forecasting. The aim of this proposal is to find a methodology that reduces the amount of input variables of the neural network and obtain satisfactory results. To verify the proposed methodology results are presented for electric short-term load forecasting in a period of 24 and 48 hours ahead, considering the historical data obtained from a company pertaining to the electrical sector / Mestre

by Fong Hon Hang = 超薄二氧化硅的固定電荷分佈和電擊穿特性 / 方漢鏗. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2000. / Includes bibliographical references. / Text in English; abstracts in English and Chinese. / by Fong Hon Hang = Chao bo er yang hua gui de gu ding dian he fen bu he dian ji chuan te xing / Fang Hankeng. / ABSTRACT --- p.i / ACKNOWLEDGEMENTS --- p.iii / TABLE OF CONTENT --- p.iv / LIST OF FIGURES --- p.ix / LIST OF TABLES --- p.xiv / LIST OF SYMBOLS --- p.xv / Chapter Chapter1 --- Background of the thesis work / Chapter 1.1 --- Introduction --- p.1 / Chapter 1.2 --- Stability of charge on oxide --- p.1 / Chapter 1.3 --- Defects in SiO2/Si --- p.2 / Chapter 1.4 --- Objectives of the thesis work --- p.4 / Chapter 1.5 --- Organization of the thesis --- p.5 / Bibliography for Chapter1 --- p.6 / Chapter Chapter2 --- Theory of X-ray Photoelectron Spectroscopy (XPS) and Surface Charge Spectroscopy (SCS) / Chapter 2.1 --- Introduction --- p.7 / Chapter 2.2 --- X-ray photoelectron spectrometry (XPS) --- p.8 / Chapter 2.2.1 --- Binding energy reference for semiconductors --- p.10 / Chapter 2.2.2 --- Measurement of surface Fermi level --- p.15 / Chapter 2.2.3 --- XPS quantitative analysis --- p.17 / Chapter 2.2.3.1 --- Electron Inelastic Mean free Path --- p.16 / Chapter 2.2.3.2 --- Atomic concentration of a homogeneous material --- p.17 / Chapter 2.2.3.3 --- Determination of overlayer thickness --- p.19 / Chapter 2.3 --- Surface charge Spectroscopy (SCS) --- p.21 / Chapter 2.3.1 --- Principle of the SCS technique --- p.21 / Chapter 2.3.2 --- Control of the dielectric surface potential --- p.21 / Chapter 2.3.3 --- Dielectric layer surface potential --- p.22 / Chapter 2.3.4 --- Surface band bending --- p.23 / Chapter 2.3.5 --- Limitation of the dielectric layer thickness --- p.24 / Chapter 2.4 --- Applications of SCS on Metal-Oxide Semiconductor (MOS) --- p.24 / Chapter 2.4.1 --- Measurements of interface state density (Dit) --- p.24 / Chapter 2.4.2 --- Determination of density of fixed-oxide charges --- p.27 / Bibliography for Chapter2 --- p.28 / Chapter Chapter3 --- Instrumentation & methodology / Chapter 3.1 --- X-ray Photoelectron Spectroscopy (XPS) --- p.30 / Chapter 3.1.1 --- General description of the Kratos AXIS - HS XPS system --- p.30 / Chapter 3.1.2 --- X-ray source --- p.32 / Chapter 3.1.3 --- AXIS - HS electron analyzer and transfer lens system --- p.35 / Chapter 3.1.4 --- Laser alignment facility --- p.38 / Chapter 3.1.5 --- In-lens (Micro XPS) aperture --- p.38 / Chapter 3.1.6 --- Iris (Lens input aperture) --- p.39 / Chapter 3.1.7 --- Magnetic immersion lenses --- p.39 / Chapter 3.1.8 --- Lateral resolutions --- p.41 / Chapter 3.1.9 --- Charge neutralizer --- p.53 / Chapter 3.1.10 --- XPS imaging capability --- p.58 / Chapter 3.1.11 --- Angle-resolved X-ray photoelectron spectroscopy (ARXPS) --- p.58 / Chapter 3.1.12 --- Ion sputtering system and depth profiling --- p.59 / Chapter 3.2 --- Methodology for surface charging --- p.59 / Chapter 3.3 --- Sample preparation --- p.61 / Bibliography for Chapter3 --- p.62 / Chapter Chapter4 --- Fixed-oxide charge Qf(z) of thermally-grown SiO2/Si(100) / Chapter 4.1 --- Introduction --- p.63 / Chapter 4.2 --- Experimental results on oxide surface potential as a function of oxide thickness --- p.64 / Chapter 4.3 --- Calculation of fixed-oxide charge distribution --- p.69 / Chapter 4.3.1 --- Gauss's law --- p.69 / Chapter 4.3.2 --- Density of fixed-oxide charge --- p.70 / Chapter 4.4 --- Applications --- p.78 / Bibliography for chapter4 --- p.80 / Chapter Chapter5 --- Observation of dielectric electrical breakdown phenomena of SiO2/Si structure by SCS / Chapter 5.1 --- Introduction to electrical breakdown analysis in device electronics --- p.81 / Chapter 5.2 --- Experimental --- p.82 / Chapter 5.3 --- Results --- p.82 / Chapter 5.3.1 --- Analysis on 1000A Sio2/Si --- p.82 / Chapter 5.3.1.1 --- Variation of C 1s under charging --- p.82 / Chapter 5.3.1.2 --- Stochastic breakdown of SiO2 --- p.84 / Chapter 5.3.2 --- Analysis on 19k SiO2/Si --- p.91 / Chapter 5.4 --- Discussion --- p.93 / Chapter 5.4.1 --- Model of stochastic breakdown of SiO2/Si --- p.93 / Chapter 5.4.2 --- Variation of Si 2p under charging --- p.95 / Chapter 5.5 --- Summary --- p.96 / Bibliography for Chapter5 --- p.99 / Chapter Chapter6 / Conclusion --- p.100

Thesis (Ph.D.)--Michigan State University. Dept. of Physics, 2008. / Title from PDF t.p. (viewed on Mar. 27, 2009) Includes bibliographical references (p. 216-218). Also issued in print.

Made available in DSpace on 2015-10-06T13:03:11Z (GMT). No. of bitstreams: 0 Previous issue date: 2015-02-20. Added 1 bitstream(s) on 2015-10-06T13:18:50Z : No. of bitstreams: 1 000849384.pdf: 1047257 bytes, checksum: 45134ee2d113dbbeddabf3b345ce6cf9 (MD5) / Neste trabalho são propostas duas novas técnicas de parametrizações geométricas que se baseiam na análise da curva trajetórias de soluções (curva P-V) do fluxo de carga continuado e que permitem tanto o traçado completo das curvas P-V quanto a obtenção do ponto de máximo carregamento de qualquer sistema elétrico de potência. Estas técnicas surgiram diante das limitações de algumas técnicas de parametrização geométrica existentes para determinação do ponto de máximo carregamento e traçado dos perfis de tensão de sistemas cuja tensão de uma pequena área, ou magnitude de tensão de uma quantidade pouco significativa de barras, não permanece dentro da faixa normal de operação. Na primeira a adição de uma equação de segundo grau ao sistema de equações básicas do fluxo de carga continuado, a qual passa por três pontos no plano formado pelas variáveis perdas de potência ativa total e o fator de carregamento, mostrou-se eficiente quando aplicado aos sistemas do IEEE, 300, 638 e 787 barras do sistema Sul-Sudeste brasileiro. Mas, falha para sistemas com instabilidade de tensão com características predominantemente local, como o sistema de 904 barras do Sudoeste americano. Diante desta limitação é proposta uma nova técnica que consiste no acréscimo de uma equação de reta que passa por um ponto no plano formado pelas variáveis perda total de potência ativa e o fator de carregamento. É uma técnica robusta o que favorece sua aplicação com êxito em quaisquer sistemas do IEEE e os reais de grande porte, em particular o 904 barras, o que pode ser comprovado pela análise dos resultados obtidos. Também para ambas as técnicas, propõem-se a normalização da variável perda total de potência ativa, para uniformizar seus valores e a escala dos eixos propiciando a vantagem da definição de um processo eficiente e único de controle de tamanho de passo para o traçado completo da curva P-V para qualquer... / This work proposes two new geometric parameterization techniques that based on analysis of solutions trajectory curve ( P-V curve) of the continuation power flow and allow both the complete tracing of P-V curves as obtaining the maximum loading point of any electric power system. These techniques were developed before the limitations on the geometric parameterization techniques exist for determining the maximum load point and layout of system voltage profiles whose voltage profile of a small area or voltage magnitude of a little bit amount of bus not remains within the normal operating range. At first the addition of a second degree equation of the basic equations of the continuation power flow which passes through three points in the plane formed by the total power loss variable active and load factor was shown to be effective when applied to IEEE systems 300, 638 and 787 bus of the Brazilian South-Southeast system but fails for systems with instability with predominantly local voltage characteristics such as the American Southwest 904 bus system. Given this limitation we propose a new technique consisting of the addition of a line equation passing through a point in the plane formed by the variables total real power losses and loading factor is a robust technique which favors their successful implementation in any IEEE systems and large real in particular the 904 bus which can be confirmed by analysis of the results. Also for both techniques propose to normalize the total real power losses variable to standardize its values and the axes scale providing the advantage of defining an efficient and unique process step size control for the complete tracing of P-V curve for any operating condition

Made available in DSpace on 2014-06-11T19:22:33Z (GMT). No. of bitstreams: 0 Previous issue date: 2012-10-19Bitstream added on 2014-06-13T18:08:31Z : No. of bitstreams: 1 oliveira_cm_me_ilha.pdf: 579995 bytes, checksum: 40131b559f97d367bfaaf567b4130b1d (MD5) / PROPG - Programa de Pós-Graduação / Nos dias atuais, a previsão de cargas elétricas tem se mostrado cada vez mais importante às empresas distribuidoras de tal serviço, especialmente para que sejam possibilitados o planejamento, análise e operação dos sistemas elétricos, restando clara a necessidade de se antever o comportamento da carga, tornando possível o fornecimento eficiente de energia aos consumidores, visando que isso ocorra de forma econômica e contínua, valendo ressaltar ainda que, a tais empresas resta o interesse na lucratividade do setor para que se mantenham sólidas no mercado. Em um primeiro momento, a solução para tais problemas foi a utilização de técnicas matemáticas e estatísticas, podendo citar como exemplo, as séries numéricas, com resultados satisfatórios, mas de difícil modelagem. A inteligência artificial tem se mostrado uma técnica que supera os resultados anteriormente obtidos e, como prova de tal afirmação, a presente dissertação apresenta uma metodologia baseada em redes neurais, possibilitando a obtenção de resultados bastante satisfatórios, demonstrando ser um modelo robusto, com baixo custo computacional, rápido e eficiente. O objetivo deste trabalho é a utilização do hibridismo de redes neurais, sendo a primeira delas, uma rede ART Fuzzy e a segunda, a Perceptron multicamadas, via algoritmo backpropagation, aproveitando as melhores características de cada uma delas para a obtenção de resultado viável quando de sua utilização. A metodologia utilizada apresenta níveis de erro aceitáveis comparado a outros métodos que se encontram na literatura ou, até mesmo, em uso pelas empresas distribuidoras do setor elétrico brasileiro. No intuito de se obter a previsão de cargas citada acima, foi utilizado um banco de dados histórico de uma empresa distribuidora nacional, valendo-se de técnicas como o janelamento, entre outras que serão devidamente descritas no decorrer do texto / Electrical load forecasting is very important to the electrical companies to advice in planning, operation and analysis of electrical power systems. Knowing the load in advance leads to provide energy efficiently, continuously and economically. Firstly the solution for such problems was executed with mathematical and statistical techniques, e.g. time series analysis with good results but with difficult modeling. The use of Artificial Intelligence overcomes these problems and the proposal of this work is to present a robust hybrid method using neural networks providing good results with low computational costs. The hybrid model proposed in this work uses a Fuzzy ART neural network and a MLP (multi layerPerceptron) by backpropagation training emphasizing the best characteristics of each one. The errors obtained are compatible with those found in the literature. Results are shown for data from a Brazilian electrical company

Orientador: Anna Diva Plasencia Lotufo / Coorientador: Maria do Carmo Gomes da Silveira / Banca: Lilian Yuli Isoda / Banca: Gelson da Cruz Junior / Resumo: Nos dias atuais, a previsão de cargas elétricas tem se mostrado cada vez mais importante às empresas distribuidoras de tal serviço, especialmente para que sejam possibilitados o planejamento, análise e operação dos sistemas elétricos, restando clara a necessidade de se antever o comportamento da carga, tornando possível o fornecimento eficiente de energia aos consumidores, visando que isso ocorra de forma econômica e contínua, valendo ressaltar ainda que, a tais empresas resta o interesse na lucratividade do setor para que se mantenham sólidas no mercado. Em um primeiro momento, a solução para tais problemas foi a utilização de técnicas matemáticas e estatísticas, podendo citar como exemplo, as séries numéricas, com resultados satisfatórios, mas de difícil modelagem. A inteligência artificial tem se mostrado uma técnica que supera os resultados anteriormente obtidos e, como prova de tal afirmação, a presente dissertação apresenta uma metodologia baseada em redes neurais, possibilitando a obtenção de resultados bastante satisfatórios, demonstrando ser um modelo robusto, com baixo custo computacional, rápido e eficiente. O objetivo deste trabalho é a utilização do hibridismo de redes neurais, sendo a primeira delas, uma rede ART Fuzzy e a segunda, a Perceptron multicamadas, via algoritmo backpropagation, aproveitando as melhores características de cada uma delas para a obtenção de resultado viável quando de sua utilização. A metodologia utilizada apresenta níveis de erro aceitáveis comparado a outros métodos que se encontram na literatura ou, até mesmo, em uso pelas empresas distribuidoras do setor elétrico brasileiro. No intuito de se obter a previsão de cargas citada acima, foi utilizado um banco de dados histórico de uma empresa distribuidora nacional, valendo-se de técnicas como o janelamento, entre outras que serão devidamente descritas no decorrer do texto / Abstract: Electrical load forecasting is very important to the electrical companies to advice in planning, operation and analysis of electrical power systems. Knowing the load in advance leads to provide energy efficiently, continuously and economically. Firstly the solution for such problems was executed with mathematical and statistical techniques, e.g. time series analysis with good results but with difficult modeling. The use of Artificial Intelligence overcomes these problems and the proposal of this work is to present a robust hybrid method using neural networks providing good results with low computational costs. The hybrid model proposed in this work uses a Fuzzy ART neural network and a MLP (multi layerPerceptron) by backpropagation training emphasizing the best characteristics of each one. The errors obtained are compatible with those found in the literature. Results are shown for data from a Brazilian electrical company / Mestre

Orientador: Dilson Amancio Alves / Banca: Sergio Azevedo de Oliveira / Banca: Francisco Carlos Vieira Malange / Banca: Marcos Amorielle Furini / Banca: Rogério Rocha Matarucco / Resumo: Neste trabalho são propostas duas novas técnicas de parametrizações geométricas que se baseiam na análise da curva trajetórias de soluções (curva P-V) do fluxo de carga continuado e que permitem tanto o traçado completo das curvas P-V quanto a obtenção do ponto de máximo carregamento de qualquer sistema elétrico de potência. Estas técnicas surgiram diante das limitações de algumas técnicas de parametrização geométrica existentes para determinação do ponto de máximo carregamento e traçado dos perfis de tensão de sistemas cuja tensão de uma pequena área, ou magnitude de tensão de uma quantidade pouco significativa de barras, não permanece dentro da faixa normal de operação. Na primeira a adição de uma equação de segundo grau ao sistema de equações básicas do fluxo de carga continuado, a qual passa por três pontos no plano formado pelas variáveis perdas de potência ativa total e o fator de carregamento, mostrou-se eficiente quando aplicado aos sistemas do IEEE, 300, 638 e 787 barras do sistema Sul-Sudeste brasileiro. Mas, falha para sistemas com instabilidade de tensão com características predominantemente local, como o sistema de 904 barras do Sudoeste americano. Diante desta limitação é proposta uma nova técnica que consiste no acréscimo de uma equação de reta que passa por um ponto no plano formado pelas variáveis perda total de potência ativa e o fator de carregamento. É uma técnica robusta o que favorece sua aplicação com êxito em quaisquer sistemas do IEEE e os reais de grande porte, em particular o 904 barras, o que pode ser comprovado pela análise dos resultados obtidos. Também para ambas as técnicas, propõem-se a normalização da variável perda total de potência ativa, para uniformizar seus valores e a escala dos eixos propiciando a vantagem da definição de um processo eficiente e único de controle de tamanho de passo para o traçado completo da curva P-V para qualquer... / Abstract: This work proposes two new geometric parameterization techniques that based on analysis of solutions trajectory curve ( P-V curve) of the continuation power flow and allow both the complete tracing of P-V curves as obtaining the maximum loading point of any electric power system. These techniques were developed before the limitations on the geometric parameterization techniques exist for determining the maximum load point and layout of system voltage profiles whose voltage profile of a small area or voltage magnitude of a little bit amount of bus not remains within the normal operating range. At first the addition of a second degree equation of the basic equations of the continuation power flow which passes through three points in the plane formed by the total power loss variable active and load factor was shown to be effective when applied to IEEE systems 300, 638 and 787 bus of the Brazilian South-Southeast system but fails for systems with instability with predominantly local voltage characteristics such as the American Southwest 904 bus system. Given this limitation we propose a new technique consisting of the addition of a line equation passing through a point in the plane formed by the variables total real power losses and loading factor is a robust technique which favors their successful implementation in any IEEE systems and large real in particular the 904 bus which can be confirmed by analysis of the results. Also for both techniques propose to normalize the total real power losses variable to standardize its values and the axes scale providing the advantage of defining an efficient and unique process step size control for the complete tracing of P-V curve for any operating condition / Doutor

Thesis. 1976. Ph.D.--Massachusetts Institute of Technology. Dept. of Meteorology. / Microfiche copy available in Archives and Science. / Vita. / Bibliography: leaves 240-244. / Ph.D.

Thesis (Ph.D. (Physics)) -- University of Limpopo, 2020 / This work presents a computational modelling workflow that uniquely combines several techniques, proposed as a means for studying and designing high-energy-density electrodes for the next-generation of rechargeable batteries within the era of the fourth industrial revolution (4IR). The Self-Consistent Charge Density Functional-based Tight Binding (SCC-DFTB) parameterisation scheme for the Li-Si and Na-Si systems is presented. By using the Li-Si system, a procedure for developing the Slater-Koster based potentials is shown. Using lessons learned from the Li-Si framework, the parameterisation of the Na-Si is reported. The Li-Si SCC-DFTB parameter set has been developed to handle environments that consist of Si-Si, Li-Si and Li-Li interactions; and the Na-Si SCC DFTB parameter set is developed for Na-Na, Na-Si, and Si-Si interactions. Validations and applications of the developed sets are illustrated and discussed. By calculating equilibrium lattice constants, the Li-Si set is shown to be compatible with various phases in the crystalline Li-Si system. The results were generally within a margin of less than 8% difference, with some values such as that of the cubic Li22Si5 being in agreement with experiments to within 1%. The volume expansion of Si as a function of Li insertion was successfully modelled via the Li-Si SCC-DFTB parameter set. It was shown that Si gradually expands in volume from 53.6% for the LiSi phase composed of 50 atm % Li, to 261.57% for Li15Si4 with 78.95 atm % Li, and eventually shoots over 300% for the Li22Si5 phase with the expansion at 316.45%, which agrees with experiments. Furthermore, the ability of the Li-Si SCC-DFTB parameter set to model the mechanical properties of Si is evaluated by calculating the mechanical properties of pristine cubic Si. The parameter set was able to produce the mechanical properties of Si, which agree with experiments to within 6%. The SCC-DFTB parameter set was then used to model the volume expansion of amorphous silicon (a-Si) as a result of lithiation within concentrations ranging from 33 – 50 atm % Li. Consistent with experiments, the a-Si was found to marginally expand in a linear form with increase in Li content. a-Si was observed to exhibit a lower expansion compared to c-Si. Additionally, the structural stability of the amorphous Li-Si alloys was examined, and observations agree with experiments.vi The Na-Si SCC-DFTB parameter set produced equilibrium lattice parameters that agree with experiments to within 4% for reference structures, and the transferability was tested on three Na-Si clathrate compounds (i.e. the Pm-3n Na8Si46, the Cmcm NaSi6 and Fd-3m Na24Si136). By employing the approach used when lithiating Si, the sodiation of crystalline silicon (c-Si) was modelled. It was predicted that c-Si expands by over 400% at 77 atm% Na and shoots above 500% for concentrations exceeding 80 atm% of Na. By comparing how c-Si expands as a result of lithiation to the expansion consequent to sodiation for concentrations ranging from 66.6 – 81.4 atm%, c-Si is shown to be unsuitable for Na-ion batteries. As a test, the ability of the developed Na-Si SCC DFTB parameter set to handle large and complex geometries was shown by modelling the expansion of a-Si at 33 atm% Na. It was deduced that a-Si would be more preferable for Na-ion batteries since at 33 atm% Na, a-Si expanded a lot less than when c-Si was used. Using the Li-Si and the Na-Si SCC-DFTB parameter sets, it was noted that amorphisation appears to lower the magnitude by which Si expands, therefore agreeing with experiments in that amorphous structures are reported to exhibit a buffering effect towards volume expansion. The material space for the Li-Si alloy system is explored through crystal structure predictions conducted via a machine learning powered cluster expansion (CE). Using the FCC and BCC – based parent lattice in the grid search, 12 thermodynamically stable Li-Si alloys were predicted by the genetic algorithm. Viz. the trigonal Li4Si (R-3m), tetragonal Li4Si (I4/m), tetragonal Li3Si (I4/mmm), cubic Li3Si (Fm-3m), monoclinic Li2Si3 (C2/m), trigonal Li2Si (P-3m1), tetragonal LiSi (P4/mmm), trigonal LiSi2 (P-2m1), monoclinic LiSi3 (P2/m), cubic LiSi3 (Pm-3m), tetragonal LiSi4 (I4/m) and monoclinic LiSi4 (C2/m). The structural stabilities of the predicted Li-Si alloys are further studied. With focus on pressure, the thermodynamic conditions under which the Li-rich phase, Li4Si (R 3m), would be stable are tested. Li4Si (R-3m) was subjected to pressures during geometry optimization and found to globally maintain its structural stability within the range 0 – 25GPa. Hence, Li4Si was predicted to be a low pressure phase. In studying the PDOS, the Li4Si (I4/m) was noted to be more stable around 40GPa and vii 45GPa, which is consistent with the prediction made from other works, wherein intelligence-based techniques were used. A test for exploring the Na-Si material space was done using insights acquired from the Li-Si framework. Three thermodynamically stable Na-Si (i.e. the I4/mmm Na3Si, P4/nmm NaSi and Immm NaSi2) were predicted. Using the Na-Si SCC-DFTB parameter set, a correlation of the total DOS in the vicinity of the Fermi level (Ef) with the structural stability of the three Na-Si alloys is done. NaSi (P4/nmm) was shown to be unstable at 0GPa, NaSi2 (Immm) is found to be stable, and the Na-rich Na3Si exhibited metastability. The stability of Na3Si was seen to improve when external pressure ranging from 2.5 – 25GPa was applied; hence, suggesting Na3Si (I4/mmm) to be a high-pressure phase. Furthermore, expanding on the groundwork laid from the Li-Si and Na-Si CE, the Mg-Si system was tested to illustrate that the approach can be used to rapidly screen for new materials. The ground-state crystal structure search predicted 4 thermodynamically stable Mg-Si alloys. Viz. Mg3Si (Pm-3m), MgSi (P4/mmm), MgSi2 (Immm) and MgSi3 (Pmmm). Lastly, to highlight the power of combining various computational techniques to advance material discovery and design, a framework linking SCC-DFTB and CE is illustrated. Candidate electrode materials with nano-architectural features were simulated by designing nanospheres comprised of more than 500 atoms, using the predicted Li-Si and Na-Si crystal structures. The stability of the nanospheres was examined using SCC-DFTB parameters developed herein. The workflow presented in this work paves the way for rapid material discovery, which is sought for in the era of the fourth industrial revolution. / National Cyber Infrastructure System: Center for High-Performance Computing (NICIS-CHPC) for computing resources, the National Research Foundation (NRF) and the University of Limpopo

Mémoire (M.Sc.)--Université du Québec à Chicoutimi, 1978. / Document électronique également accessible en format PDF. CaQCU