Methods To Construct The Crystal Structure Prototype Database And Applications

Numerical computation is used by more and more scientific researchers to simulate the properties of materials,and they use programs and computers to screen vast amounts of experimental or theoretical data to predict and design materials.As computer technology is developing very rapidly,the computer based crystal structure databases make it very convenient for us to search and visualize the crystal structures.So,these databases are becoming more and more important in the area of material research.The crystal structure of the material is the key information to predict and understand its properties.The databases to store crystal structure information have important applications in the areas of physics,chemistry,material science and crystallography.The experimental and theoretical methods to determine atomic structures have been well developed.Up to now,numerous structures were proposed,and then stored in databases.Many databases have been constructed and become famous,such as the Cambridge Structural Database,the Protein Data Bank,the Inorganic Crystal Structure Database,the American Mineralogist Crystal Structure Database,and the Crystallography Open Database.These databases are widely used in many different areas and help researchers to develop materials.However,there's lots of redundant entries stored in the database recording the same or similar structures,and these entries in some extent hampers materials analysis and discovery.In order to solve this problem,a robust and efficient structure descriptor was proposed.We came up with a set of simple and unambiguous definition of the crystal structure prototype.In this definition,our structure descriptor is used to assess the similarity of crystal structures.Based on our definition of crystal structure prototype,we classified the structures in database and removed the similar structures.The remaining structures were used to construct the Crystal Structure Prototype Database(CSPD).The CSPD was then used to study the high pressure phases of La C2.The main results of the thesis are as follows:1.In order to consider the coordination characterization of the atoms in the structure,the interatomic distances are smoothed by the Gaussian kernel to construct the Coordination Characterization Function(CCF).The Pearson correlation coefficient R was used to assess the similarity between CCFs from two structures,and 1-R gives the structural distance d.Many tests demonstrated in the thesis proved the effectiveness of CCF,and a threshold dt was determined to judge whether the two structures are similar or not.The chemical composition,space group,and structural distance are used to define crystal structure prototype.Our definition is simple and unambiguous,and this definition enables us to develop a program to automatically classify structures in database.Based on our definition of crystal structure prototype,we developed a Fortran program to construct the CSPD.The structures in the database were classified according to their chemical composition by our program,and then the similarity of the structures was assessed to remove similar structures.A series of statistics describing the distribution of crystal structure prototypes in the CSPD was compiled to provide an important insight for materials design.Three applications were demonstrated in our thesis.The first application shows how to apply the CSPD to structure prediction.The structures in the CSPD are distinct from each other.Almost all these structures are determined by experiment.So,they are physically more justified than the structures generated by random method.This method can extract the essence structural information from big data(crystal structure databases),so we name this method Big Data Method(BDM).The procedure of generating structures by BDM comprises four main steps:(i)selecting structure prototypes in the CSPD for a given targeted composition type and formula unit;(ii)substituting elements for the selected structure prototype;(iii)adjusting the lattice parameters for a given volume;(iv)checking the minimal interatomic distances.We compared the efficiency of BDM and random sampling method(with symmetry constraints)in Ca F2,KN3,and Cu In S2 system.Our test results show that BDM outperforms random method in all three systems.Meanwhile,BDM also found two meatastable phase P-42 c and P-4m2 in Cu In S2 system.These two structures are only 1 me V/atom above the experimental phase,so we believe that the proposed metastable structures of Cu In S2 are likely to be synthesized by experiment.The second application is searching similar structures in database.This function is implemented in the program the Structure Prototype Analysis Package(SPAP)developed by us.When a new structure is proposed through experiment or theory,we can use SPAP to determine how many compounds in the database are similar with the newly proposed structure.The procedure of searching similar structures comprises two main steps:(i)screening structures of the same composition type as the given structure;(ii)assessing the distances between the given structure and screened structures.Additional information such as space group can be defined during the search.We can also choose how to handle lattice parameters of all the structures: rescale to the same atom number density or adopt the same lattice parameters.The last application shows how to use SPAP to analyze structure prediction results.CALYPSO generates many distinct initial structures during structure search.After local structure optimization,some of the structures go to the same local minimum on the potential energy surface.So,there is a proportion of similar structures in the prediction results.SPAP uses CCF to assess the similarity between structures,and removes similar structures in the prediction results.This greatly simplified the manual job to analyze prediction results.Another function of SPAP is collecting the predicted low-energy distinct structures to construct a theoretical structure database.Local structure optimization makes these low-energy structures physically more justified.So,they can be used to generate structures for structure prediction.In this work,we came up with a definition for crystal structure prototype and then constructed the CSPD,and introduced three successful applications.We believe that after further development and optimization of the CSPD,it will become a generally used database in the future.2.By using SPAP,we collected lots of theoretically predicted low-energy crystal structure prototypes.These structures can be used to generate structures for structure prediction.We used structure generating method BDM based on database and CALYPSO evolutionary method jointly to study high pressure phases of La C2.All these two methods successfully predicted all the stable structures.This helps us to crosscheck the correctness and reliability of our calculation results and prove the effectiveness of BDM under high pressure condition.Carbon has the capability of forming various bonding states to affect the structures and properties of transition metal carbides.In this work,structural search was performed to explore the structural diversity of La C2 at pressures of 0.0-30.0 GPa.Five stable structures of La C2 reveal a variety of carbon structural units ranging from dimer to bent C3,zigzag C4 and armchair polymer chains.A series of pressure-induced structural transformations are predicted,as I4/mmm(i.e.experimental ? phase)? C2/c ? Pnma ? Pmma,which involves catenations of carbon from dimer to zigzag C4 units and further to armchair polymer chains.The bent C3 unit appears in a novel Immm structure.This structure is the theoretical ground state of La C2 at ambient conditions,but is kinetically inaccessible from experimental ? phase.La C2 becomes thermodynamically metastable relative to La2C3 + diamond above 17.1 GPa,and eventually decomposes into constituent elements above 35.6 GPa.The presented results indicate that a catenation of carbon can be realized even in simple inorganic compounds at nonambient conditions.