【Domestic Papers】Professor Cao Bingyang's team at Tsinghua University Adv. Mater.: Made new progress in the field of amorphous Gallium Oxide heat conduction

      Recently, Professor Cao Bingyang's research group at the School of Aeronautics and Astronautics of Tsinghua University successfully revealed the atomic structure characteristics, thermal transport properties and the internal influence mechanism and quantitative relationship of "structure-thermal transport properties" of amorphous Gallium Oxide system by combining machine learning, molecular dynamics simulation and experimental measurement. Because the current experimental technology is difficult to directly observe the three-dimensional atomic structure of amorphous materials, the research team used the machine learning potential function with quantum mechanical accuracy to simulate the melting and quenching process to accurately model the amorphous materials at the atomic scale, and used the non-equilibrium molecular dynamics simulation, The thermal conductivity of amorphous Gallium Oxide was studied by Harmonic Theory（Allen-Feldmen, AF） and Unified heat conduction theory (Unified Theory, UF). The experimental results show that the machine learning potential automatically generated based on random structure search and Maxwell-Boltzmann diagram sampling can accurately predict the structure and thermal conductivity of amorphous Gallium Oxide (FIG. 1), and reveal that the thermal conductivity of amorphous Gallium Oxide is dominated by modal coherence, and the contribution of phonon-like modal propagation to the thermal conductivity can be ignored.

FIG. 1. Comparison of experimental measurements of thermal conductivity of amorphous Gallium Oxide with theoretical predictions

      The absence of long-range disorder makes medium-short range order as the most important structural feature of amorphous materials, and the medium-short range order structure largely determines the physical and chemical properties of amorphous materials. To this end, the research team used machine learning-driven quenching simulations to reveal changes in the short-range ordered structure of amorphous Gallium Oxide from the high-density region to the low-density region (Figure 2). The results of the distribution function show that the average bond length of amorphous Gallium Oxide is about 1.9 A (Figure 2b). In addition, the average atomic coordination number of the amorphous Gallium Oxide atomic network increases with the increase of density, and the proportion of tetrahedral environment decreases, while the proportion of octahedral environment increases (Figure 2c-e). The statistical distribution of the shortest path ring decays rapidly with increasing density, indicating that the scale of the cell used in the simulation ensures the recurrence of medium-range ordered structures (FIG. 2f). At the same time, the results of the ring distribution show that the high-density system has a medium-range ordered structure that is more similar to the crystal.

FIG. 2. Characterization of short-range and medium-range ordered structures of amorphous Gallium Oxide

      In order to further analyze the mechanism of the effect of structural changes on the microscopic thermal conductivity of amorphous Gallium Oxide, the research team further calculated the reciprocal participation ratio and modal diffusivity of different amorphous Gallium Oxide systems (FIG. 3). The reciprocal participation ratio can measure the degree of localization of vibration modes, while the modal diffusivity can describe the rate at which vibration modes carry heat diffusion. The results show that the localization of vibration modes mainly occurs in the high frequency region, and the diffusivity of the local modes is generally low. With the increase of material density, the reciprocal participation ratio decreases, while the modal diffusivity increases, indicating that the proportion of modal localization decreases, while the spatial expansibility of the modes increases, which is ultimately reflected in the enhancement of material thermal conductivity. The effect of density on thermal conductivity can be more deeply attributed to the effect of atomic-scale structure on the heat transport process. As mentioned above, the increase of the density of the system leads to the increase of the average coordination number of atoms and the increase of the proportion of octahedral environment, which provides more action channels for heat transport, and thus leads to the enhancement of heat transport.

FIG. 3. Reciprocal participation ratio and vibrational diffusivity distribution of different amorphous Gallium Oxide systems

      From the perspective of material informatics, in order to establish the quantitative relationship between material structure and thermal conductivity, the Structural Similarity Factor (SSF) of amorphous materials with physical interpretability is proposed for the first time. From the perspective of atomic scale, amorphous materials and crystalline materials have similar polyhedral component units, and the differences in structure and properties of the two materials mainly come from the differences in connection number, orientation and distortion of the polyhedral units. SSF represents the structural characteristics of amorphous materials by measuring the similarity of the chemical environment between the amorphous materials and the crystalline materials. In essence, SSF has high sensitivity to the density and composition of the material system. The larger the SSF, the denser the atomic network, the higher the average coordination number, and the greater the thermal conductivity of the corresponding material. At the same time, the SSF cleverly quantifies the similarity of the middle program structure of the crystal and the amorphous material. The larger the SSF, the more similar the middle program structure of the amorphous material is to the crystal, which can be verified by the shortest path loop distribution shown in Figure 2f. The results show that there is a strong linear relationship between SSF and thermal conductivity, so a quantitative relationship between structure and thermal conductivity can be constructed using a small amount of data, which will help to predict thermal conductivity quickly and accurately directly from the structural information of amorphous systems, and accelerate the screening of amorphous materials with excellent thermal properties.

Figure 4. Relationship between density, component ratio and Structural Similarity Factor(SSF) vs thermal conductivity of amorphous Gallium Oxide

      The results have important implications for the development of thermal management techniques for amorphous Gallium Oxide electronic devices, and also demonstrate the ability of machine learning models to solve real-world physical problems. Given the complexity and importance of heat transport in the amorphous phase, this work provides a new starting point for accelerated exploration of heat transport properties and mechanisms of other important amorphous materials in the future.