Structure-property Relationship Of Transition Metal Compounds From Data-driven Study

Understanding the structure-property relationships of compounds at the atomic scale is a fundamental problem in the field of materials science.Nowadays,harnessing the recent advance in density functional theory in high-throughput mode,it is feasible to obtain the physical and chemical properties of a given compound quickly and accurately,making it easy for us to explore the underlay mechanisms of many physical phenomena.Meanwhile,the high-throughput approaches,both computational and experimental,can produce a massive amount of data,pushing forward the materials research into a data-driven paradigm.In this thesis,the structure-property relationships of different transition metal compounds are investigated with the help of high-throughput first-principles calculations and data-driven approaches as follows.（1）The statistical trends of the local structure of condensed matter systems are important features to rationalize physics in compounds,hence are crucial to physics,materials science,chemistry,and many other disciplines.The values of the ionic radius widely used in the field of material science were obtained statistically from the crystallographic data of the 1970s,these data are outdated compared to what we have today.Leveraging the massive data accumulated in materials science in recent years,we systematically analyze detailed information of nearly 40,000 transition metal compounds,and extract detailed information of coordination environment,valence state,ionic bond length,atomic magnetic moment,local deformation due to Jahn-Teller effect,and structural stability of transition metals.It is found that each transition metal ion has its own unique shape,size,and atomic magnetic moment in solids,thus each transition metal ion has its identical“persona”.Using unsupervised machine learning,we further obtain structural similarities of transition metal ions and use them to guide the design and the rapid assessment of the stability of new materials.Based on the knowledge of ion structure similarity,more than 60,000 new structures are generated employing the ion substitution method,and 5000 thermodynamically stable compounds are discovered through the high throughput first-principles calculations.（2）Single-atom catalyst made with nitrogen-coordinated transition metal in carbon is a highly promising system for oxygen reduction reactions（ORR）as the catalytic activity can be effectively tuned and improved via the engineering of the local coordination environment around the transition metal atom.In this thesis,we have investigated the reaction mechanism of the two-dimensional single-atom cobalt catalyst Co-N₃-C,in which Co is unconventionally coordinated by three nitrogen atoms.In this catalyst,the happening of two-sided adsorption in the two-dimensional system enables a large number of possible reaction pathways.Employing the high-throughput first-principles calculation,the free energies of dozens of reaction pathways for Co-N₃-C and conventional Co-N₄ catalysts are all evaluated,and we find that Co-N₃-C undergoes an activation process to induce the adsorption of OH intermediates and forms Co-N₃-C-OH as the true reaction site.The adsorption-desorption pathway of oxygen-containing intermediates at this site is greatly optimized,thus reducing the energy barrier of its ORR intermediates and improving the reactivity.（3）Voltage,an important property of battery materials,has a significant impact on the energy density of batteries.We analyze the factors affecting the cathode voltage of the battery using the battery material data obtained from the first-principles.It is found that the voltage drop during the discharge process of Li-ion batteries originated from the change in thermodynamic stability of the cathode materials during the electrochemical reaction.When the voltage corrections arising from the thermodynamic stability are applied,the voltage of a transition metal redox pair can be normalized to a single value,namely the intrinsic voltage.The intrinsic voltage is fundamentally determined by the formation energy convex hull of the transition mental chemical system and is sensitive only to transition metal element species and valence state during the redox reaction.With the concept of intrinsic voltage,it is possible for us to quickly justify the voltage and voltage drop of the electrodes.In this thesis,the structure-property relationships of different transition metal compounds are investigated in detail from a high-throughput first-principles calculation and data-driven perspective,providing guidance for the design of functional materials with thermodynamic stability,high-performance catalytic materials and high-energy density battery materials.