Ge2Sb2Te5 (GST) are phase-change materials that switch structure from amorphous to crystalline phase very fast upon heating. As such, they are considered the most promising candidate for non-volatile memory device. They also exhibit very rich topological properties in electronic structures. We expect many interesting physics playing behind this scene. Currently we are focusing on dynamics of their atomic structures.

 By performing DFT calculations and establishing simplified physical models, we pursue a deep understanding of the fundamental relation between structural variety and bonding character, and ultimately atomic dynamics of crystalline GST.

ATOMIC LAYER REARRANGEMENT

 We simulate atomic dynamics of layer switching in interfacial phase-change memory in terms of individual atomic energy. We develop a novel interpretation of energy barrier to analyze the nature of phase-change process.

RICH TOPOLOGICAL NATURE

 Variations of GST compounds exhibit interesing topological phases such as topological insulator and Weyl semimetal. We explore the relation between exotic topological nature and atomic structures.

BONDING AND PHONON PROPERTIES

 We study long-ranged bonding characters facilitated by so called “p-orbital network” in GST. We relate this unique bonding property with the phonon softening, which directly leads to dynamic instability and phase transitions.

EFFECT OF VACANCY

 Intrinsic or defective vacancies at cation sites in cubic GST affect material properties. We focus on the effect of vacancy distribution especially on the geometry of p-network, and its consequences on the structural and dynamic properties. 

 Migdal-Eliashberg theory gives exact descriptions for conventional superconductors, but it requires expensive calculations. Approximated formula, McMillan’s equation, is routinely used to reduce the computational load. Yet empirical parameters should be introduced. 

 A new numerical approach based on density functional theory (DFT), known as density functional theory for superconductors (SCDFT), is developed. SCDFT is the DFT version of Migdal-Eliashberg theory. The SCDFT is effective in reducing calculation cost without introducing any empirical parameters by solving the gap equation directly:

HYDROGEN-BASED SUPERCONDUCTORS

 In 2015, sulfur hydride set a new record for superconducting critical temperature. However, the required pressure is almost one million bar, very impractical for any purpose. Our goal is to find new hydride superconductors that have similar critical temperatures at low pressure.    

EXPANSION OF SCDFT FORMALISM TO UNCONVENTIONAL CASE

 Utilizing big data and neural network, or known as machine or deep learning is being a culture not only in computational community but also in materials research. Predicting materials properties by bypassing heavy demanding calculations at quantum mechanical levels but with similar accuracy would expedite the process of materials design. We adopt this methodology in our research of materials of current interest.

PREDICTING ENERGY AND FORCE VIA NEURAL NETWORKS

 The reliability of the MD simulation depends entirely on the accuracy of the energy and force calculations of the stationary system, and so is the speed of MD simulation. Neural-network-based PES (Potential energy surface) can be used to predict energies and forces nearly as accurate as but much faster than DFT calculation.

STUDY ON PHASE TRANSITION USING NEURAL NETWORK

 TMDCs are layered materials and can easily be synthesized in a-few-layered film, even in monolayer. Upon change in constituent elements, TMDCs exhibit many interesting physical properties unique in 2D such as charge density wave, long-lived excitons, quasi-1D transport, and moreover. 2D structures guarantee easy control of these properties by external field or strains. Direct-indirect gap transition and 2D quantum spin Hall phase have been reported in TMDCs. We explore the physics in TMDCs and its application using state-of-the-art DFT methods combined with phenomenological models.

ORIGIN OF STRUCTURAL VARIATION IN TMDCs

 Many interesting features of TMDCs are related to their atomic structure, so investigating the origin of the structure is important to understanding and applying their physics. For example, in group-VIB TMDCs, two very different phases, H (semiconductor with gap ~ 1 eV) and T (2D topological insulator) are observed. It is relevant to the orbital energy shift due to the relativistic effect in transition metal atom (Kim and Jhi, APL, 2017). On the other hand, group-VIIB TMDCs are in distorted T phase, which is Peierls-like distorted structure due to the Fermi nesting (Choi and Jhi, JPCM, 2018).

FERROIC (FERRO-ELECTRICITY, FERRO-MAGNETISM, FERRO-ELASTICITY) PHASES IN TMDCs

 Finding atomically thin ferroic materials is valuable challenge in physics and has potential to serve as new microscopic devices. TMDCs with proper composition of transition metal atoms have a possibility to become ferroelectric in monolayer, overcoming strong depolarization field by their 2D nature. Ferroelectricity can be coupled with other ferroic properties such as ferromagnetism and ferro-elasticity.