Sputtering enables uniform and clean deposition over a large area, which is an issue with exfoliation and chemi-cal vapor deposition methods. On the other hand, the process of physical vapor deposition (PVD) film formationhas not yet been clarified. We prepared several samples from the sub-monolayer region, and performed Ra-man spectroscopy, X-ray photon spectroscopy and high-angle annular dark-field scanning transmission electronmicroscopy. From these results, the internal stresses inherent to PVD films, the bonding states specific to sub-monolayers, and the unique film structure and the grain formation process of PVD films were discussed fromthe perspective of sub-monolayers. As a conclusion, we found that it is important to suppress the formation ofsub-monolayers on the substrate to completely form the first layer.

Zirconium disulfide (ZrS2) -an attractive next-generation channel material because of its high mobility-is stabilized in air by an underlying zirconium dioxide (ZrO2) film which functions as a high-k film in MISFET. We fabricated high-k/PVD-ZrS2 stacks with a self-oxidized ZrO2 film as an interfacial layer; their chemical properties were analyzed to clarify how each fabrication process affects the ZrS2 under the oxide film. The results clarified that sulfur vapor annealing (SVA) is critical for fabricating high-quality physical vapor deposition (PVD) ZrS2 films and that the change in surface potential of the ZrS2 films due to interface dipoles between the high-k and Zr-compound films is suppressed with scaling of high-k thickness. The SVA with high-k films also prevents degradation of crystallinity and stoichiometry, enhancing the quality of the ZrS2 films without affecting their surface potential. These achievements enable us to control of the threshold voltage in ZrS2 MISFETs.

A method to predict lightning by machine learning analysis of atmospheric electric fields is proposed for the first time. In this study, we calculated an anomaly score with long short-term memory (LSTM), a recurrent neural network analysis method, using electric field data recorded every second on the ground. The threshold value of the anomaly score was defined, and a lightning alarm at the observation point was issued or canceled. Using this method, it was confirmed that 88.9% of lightning occurred while alarming. These results suggest that a lightning prediction system with an electric field sensor and machine learning can be developed in the future.

No abstract

No abstract