Toru Ujihara
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Toru Ujihara has a total of 418 co-authors. Here are the top co-authors by number of shared publications:
- Gen Sazaki
Institute of Low Temperature Science, Hokkaido University, N19−W8, Kita-ku, Sapporo 060-0819, Japan
47 shared publications - Noritaka Usami
Graduate School of Engineering
43 shared publications - Kozo Fujiwara
Institute for Material Science, Tohoku University, Sendai, Japan
40 shared publications Kazuo Nakajima
Tohoku University
38 shared publicationsShunta Harada
29 shared publicationsYoshikazu Takeda
Aichi Synchrotron Radiation Center 4 , Seto 489-0965, Japan
27 shared publications
Latest Publications
- 15 June 2023
- Published by Wiley in Advanced Theory and Simulations
- Vol. 6 (9)
Machine Learning for Semiconductor Process Simulation Described by Coupled Partial Differential Equations
Technology computer‐aided design (TCAD) simulation is an important tool for the development of semiconductor devices. Based on coupled partial differential equations (PDEs) for behaviors, TCAD can calculate objects such as impurities, point defects, and electronic carriers in semiconductors. However, over recent years semiconductor devices have become increasingly miniaturized and complicated, resulting in much longer calculation times for TCAD. Machine learning is one technology that may be used to overcome this simulation cost problem. In this study, a neural network architecture is proposed that considers the structure of the coupled PDEs. Features representing each concentration distribution of the calculation objects are extracted by convolution operations and their reaction is modeled by channel attention. The performance of the proposed architecture and of conventional neural network models is evaluated using a simulation dataset generated by 1D coupled PDEs that models the diffusion and reaction of vacancies and interstitial atoms. In addition, the advantage of the method is discussed through the analyses of error correlations of the two predictions and attention coefficients. The machine learning method developed in this study will be applicable to other physics described by coupled PDEs and is expected to speed up the computation of simulations in various fields.
- 15 June 2023
- Published by Wiley in Advanced Theory and Simulations
- Vol. 6 (9)
Machine Learning for Semiconductor Process Simulation Described by Coupled Partial Differential Equations
Technology computer‐aided design (TCAD) simulation is an important tool for the development of semiconductor devices. Based on coupled partial differential equations (PDEs) for behaviors, TCAD can calculate objects such as impurities, point defects, and electronic carriers in semiconductors. However, over recent years semiconductor devices have become increasingly miniaturized and complicated, resulting in much longer calculation times for TCAD. Machine learning is one technology that may be used to overcome this simulation cost problem. In this study, a neural network architecture is proposed that considers the structure of the coupled PDEs. Features representing each concentration distribution of the calculation objects are extracted by convolution operations and their reaction is modeled by channel attention. The performance of the proposed architecture and of conventional neural network models is evaluated using a simulation dataset generated by 1D coupled PDEs that models the diffusion and reaction of vacancies and interstitial atoms. In addition, the advantage of the method is discussed through the analyses of error correlations of the two predictions and attention coefficients. The machine learning method developed in this study will be applicable to other physics described by coupled PDEs and is expected to speed up the computation of simulations in various fields.
- 3 July 2022
- Published by Wiley in Advanced Theory and Simulations
- Vol. 5 (9)
A Transfer Learning‐Based Method for Facilitating the Prediction of Unsteady Crystal Growth
Real‐time prediction and dynamic control systems that can adapt to an unsteady environment are necessary for material fabrication processes, especially crystal growth. Recent studies have demonstrated the effectiveness of machine learning in predicting an unsteady crystal growth process, but its wider application is hindered by the large amount of training data required for sufficient accuracy. To address this problem, this study investigates the capability of transfer learning to predict geometric evolution in an unsteady silicon carbide (SiC) solution growth system based on a small amount of data. The performance of transferred models is discussed regarding the effect of the transfer learning method, training data amount, and time step length. The transfer learning strategy yields the same accuracy as that of training from scratch but requires only 20% of the training data. The accuracy is stably inherited through successive time steps, which demonstrates the effectiveness of transfer learning in reducing the required amount of training data for predicting evolution in an unsteady crystal growth process. Moreover, the transferred models trained with relatively more data (no more than 100%) further improve the accuracy inherited from the source model through multiple time steps, which broadens the application scope of transfer learning.
- 3 July 2022
- Published by Wiley in Advanced Theory and Simulations
- Vol. 5 (9)
A Transfer Learning‐Based Method for Facilitating the Prediction of Unsteady Crystal Growth
Real‐time prediction and dynamic control systems that can adapt to an unsteady environment are necessary for material fabrication processes, especially crystal growth. Recent studies have demonstrated the effectiveness of machine learning in predicting an unsteady crystal growth process, but its wider application is hindered by the large amount of training data required for sufficient accuracy. To address this problem, this study investigates the capability of transfer learning to predict geometric evolution in an unsteady silicon carbide (SiC) solution growth system based on a small amount of data. The performance of transferred models is discussed regarding the effect of the transfer learning method, training data amount, and time step length. The transfer learning strategy yields the same accuracy as that of training from scratch but requires only 20% of the training data. The accuracy is stably inherited through successive time steps, which demonstrates the effectiveness of transfer learning in reducing the required amount of training data for predicting evolution in an unsteady crystal growth process. Moreover, the transferred models trained with relatively more data (no more than 100%) further improve the accuracy inherited from the source model through multiple time steps, which broadens the application scope of transfer learning.
- Open Access
- 1 March 2022
- Published by AIP Publishing in AIP Advances
- Vol. 12 (3) , 035310
Nucleation sites of expanded stacking faults detected by in operando x-ray topography analysis to design epitaxial layers for bipolar-degradation-free SiC MOSFETs
We investigated the nucleation sites of expanded single Shockley-type stacking faults (1SSFs) in a silicon carbide (SiC) metal–oxide–semiconductor field effect transistor (MOSFET) and demonstrated epitaxial layers designed for bipolar-degradation-free SiC MOSFETs. Since the sufficient hole density just below the basal plane dislocation (BPD)-threading edge dislocation (TED) conversion points induces 1SSF expansion, we derived the dependence of the nucleation depth on the applied current condition from the BPD-TED conversion points of 1SSFs. We first simulated and determined the three-step current conditions applied to a body diode in a SiC MOSFET so that a sufficient amount of holes would be supplied to the drift layer, to the buffer layer, and inside the substrate in the SiC MOSFET. An in operando x-ray topography analysis was conducted with the determined conditions for dynamically visualizing 1SSF expansion motions, and 1SSFs expanded at different forward current densities were successfully extracted. The depths of the BPD-TED conversion points of the extracted 1SSFs were analyzed, and it was experimentally clarified that these depths, i.e., the nucleation sites of expanded 1SSFs, became deeper with forward current densities. The bipolar degradation characteristics of SiC MOSFETs were evaluated as a function of the forward current density, and the validity of the simulation model was verified by experimental results. We also confirmed that bipolar degradation can be suppressed to some extent by using a substrate with a low BPD density, and SiC MOSFETs with a high-nitrogen-concentration epitaxial layer showed high reliability under bipolar operation. Depending on the application of SiC MOSFETs, the epitaxial layers should be designed to prevent the hole density inside the substrate from exceeding the threshold for 1SSF expansion.
- 1 March 2022
- Published by AIP Publishing in AIP Advances
- Vol. 12 (3) , 035310
Open AccessNucleation sites of expanded stacking faults detected by in operando x-ray topography analysis to design epitaxial layers for bipolar-degradation-free SiC MOSFETs
We investigated the nucleation sites of expanded single Shockley-type stacking faults (1SSFs) in a silicon carbide (SiC) metal–oxide–semiconductor field effect transistor (MOSFET) and demonstrated epitaxial layers designed for bipolar-degradation-free SiC MOSFETs. Since the sufficient hole density just below the basal plane dislocation (BPD)-threading edge dislocation (TED) conversion points induces 1SSF expansion, we derived the dependence of the nucleation depth on the applied current condition from the BPD-TED conversion points of 1SSFs. We first simulated and determined the three-step current conditions applied to a body diode in a SiC MOSFET so that a sufficient amount of holes would be supplied to the drift layer, to the buffer layer, and inside the substrate in the SiC MOSFET. An in operando x-ray topography analysis was conducted with the determined conditions for dynamically visualizing 1SSF expansion motions, and 1SSFs expanded at different forward current densities were successfully extracted. The depths of the BPD-TED conversion points of the extracted 1SSFs were analyzed, and it was experimentally clarified that these depths, i.e., the nucleation sites of expanded 1SSFs, became deeper with forward current densities. The bipolar degradation characteristics of SiC MOSFETs were evaluated as a function of the forward current density, and the validity of the simulation model was verified by experimental results. We also confirmed that bipolar degradation can be suppressed to some extent by using a substrate with a low BPD density, and SiC MOSFETs with a high-nitrogen-concentration epitaxial layer showed high reliability under bipolar operation. Depending on the application of SiC MOSFETs, the epitaxial layers should be designed to prevent the hole density inside the substrate from exceeding the threshold for 1SSF expansion.
- 14 October 2021
- Published by AIP Publishing in Journal of Applied Physics
- Vol. 130 (14) , 145703
In-operando x-ray topography analysis of SiC metal–oxide–semiconductor field-effect transistors to visualize stacking fault expansion motions dynamically during operations
We developed an in-operando x-ray topography method for dynamically visualizing single Shockley-type stacking fault (1SSF) expansion motions in silicon carbide (SiC) metal–oxide–semiconductor field-effect transistors (MOSFETs) during their operations and investigated the effect of the operating condition applied to the body diodes in SiC MOSFETs on dislocation glide velocity. In-operando x-ray topography observations were carried out in reflection geometry, and a high-resolution x-ray camera was used as a detector to record topographies dynamically. The sequence of 1SSF expansion motions in the SiC MOSFETs was observed at a high resolution of 1 s in x-ray topographies, which is sufficient to analyze the dislocation glide velocity of a 1SSF expansion. The observation results of changing the forward current density applied to the body diodes in SiC MOSFETs revealed that each triangular and bar-shaped 1SSF expands at different forward current densities. The 1SSF expansion timings also differed, even in the same chip under the same current density. The dislocation glide velocity of each expanded 1SSF in SiC MOSFETs was extracted, and it increased with the forward current density. Our method enables the dynamic visualization of bipolar degradation in SiC MOSFETs during their operations, and we can accurately obtain the information of when, where, and which 1SSF expands in a SiC MOSFET.
- 14 October 2021
- Published by AIP Publishing in Journal of Applied Physics
- Vol. 130 (14) , 145703
In-operando x-ray topography analysis of SiC metal–oxide–semiconductor field-effect transistors to visualize stacking fault expansion motions dynamically during operations
We developed an in-operando x-ray topography method for dynamically visualizing single Shockley-type stacking fault (1SSF) expansion motions in silicon carbide (SiC) metal–oxide–semiconductor field-effect transistors (MOSFETs) during their operations and investigated the effect of the operating condition applied to the body diodes in SiC MOSFETs on dislocation glide velocity. In-operando x-ray topography observations were carried out in reflection geometry, and a high-resolution x-ray camera was used as a detector to record topographies dynamically. The sequence of 1SSF expansion motions in the SiC MOSFETs was observed at a high resolution of 1 s in x-ray topographies, which is sufficient to analyze the dislocation glide velocity of a 1SSF expansion. The observation results of changing the forward current density applied to the body diodes in SiC MOSFETs revealed that each triangular and bar-shaped 1SSF expands at different forward current densities. The 1SSF expansion timings also differed, even in the same chip under the same current density. The dislocation glide velocity of each expanded 1SSF in SiC MOSFETs was extracted, and it increased with the forward current density. Our method enables the dynamic visualization of bipolar degradation in SiC MOSFETs during their operations, and we can accurately obtain the information of when, where, and which 1SSF expands in a SiC MOSFET.
- Open Access
- 3 July 2021
- Published by Taylor & Francis in Journal of Asian Ceramic Societies
- Vol. 9 (3) , 1248-1254
Control of microstructure and mechanical properties of sintered aluminum nitride through addition of aluminum nitride whiskers
Aluminum nitride (AlN) usually exhibits a microstructure consisting of equiaxed grains. However, these grains do not allow for the positive control of the microstructure and mechanical properties of AlN. In this work, we attempted to control the microstructure of AlN through the addition of whiskers and investigated the relationship between the microstructure and mechanical properties. The sintered AlN showed a microstructure consisting of large, rod-like grains, which originated from the whiskers. Furthermore, the microstructure varied with the size and amount of whiskers added, and so did the flexural strength and fracture toughness. The flexural strength and fracture toughness of the whisker-added AlN samples ranged from 197 to 267 MPa and 5.4 to 6.4 MPa m1/2, respectively. In contrast, those of AlN without whiskers were 336 MPa and 3.1 MPa m1/2, respectively.
- 3 July 2021
- Published by Taylor & Francis in Journal of Asian Ceramic Societies
- Vol. 9 (3) , 1248-1254
Open AccessControl of microstructure and mechanical properties of sintered aluminum nitride through addition of aluminum nitride whiskers
Aluminum nitride (AlN) usually exhibits a microstructure consisting of equiaxed grains. However, these grains do not allow for the positive control of the microstructure and mechanical properties of AlN. In this work, we attempted to control the microstructure of AlN through the addition of whiskers and investigated the relationship between the microstructure and mechanical properties. The sintered AlN showed a microstructure consisting of large, rod-like grains, which originated from the whiskers. Furthermore, the microstructure varied with the size and amount of whiskers added, and so did the flexural strength and fracture toughness. The flexural strength and fracture toughness of the whisker-added AlN samples ranged from 197 to 267 MPa and 5.4 to 6.4 MPa m1/2, respectively. In contrast, those of AlN without whiskers were 336 MPa and 3.1 MPa m1/2, respectively.
- 1 April 2021
- Published by Elsevier in Acta Materialia
- Vol. 208, 116746
Synchrotron X-ray topographic image contrast variation of screw-type basal plane dislocations located at different depths below the crystal surface in 4H-SiC
- 1 April 2021
- Published by Elsevier in Acta Materialia
- Vol. 208, 116746
Synchrotron X-ray topographic image contrast variation of screw-type basal plane dislocations located at different depths below the crystal surface in 4H-SiC
- 27 November 2020
- Published by IOP Publishing in Japanese Journal of Applied Physics
- Vol. 60 (SA) , SAAD03
Analysis of dislocation line tilt in GaN single crystal by Raman spectroscopy
The strain field of the tilted threading dislocation (TD) in a hydride vapor-phase epitaxy-grown c -plane (0001) GaN single crystal is demonstrated using Raman spectroscopy. In the two-dimensional mapping image of the E 2 H peak shift around TDs, high-contrast areas corresponding to pairs of high and low wavenumber regions of the peak shift are observed, and the edge component of the Burgers vector can be specified. Deformed contrast areas extending in the specified direction are also observed. This deformed contrast region indicates tilted dislocations since 2D mapping also detects strain deep below the surface due to the transmission of the laser beam. Simulation of the E 2 H peak shift mapping around the TD, tilted at 0° and 60° in the 11 2 ¯ 0 direction and having a Burgers vector b = 1 / 3 11 2 ¯ 0 , is also performed. The simulation of the 60° tilted dislocation shows that the contrast extends in one direction. The line profiles of the E 2 H peak shift by simulation are in agreement with the experimental results. We propose that the dislocation line tilt can be analyzed from the 2D Raman mapping images.
- 27 November 2020
- Published by IOP Publishing in Japanese Journal of Applied Physics
- Vol. 60 (SA) , SAAD03
Analysis of dislocation line tilt in GaN single crystal by Raman spectroscopy
The strain field of the tilted threading dislocation (TD) in a hydride vapor-phase epitaxy-grown c -plane (0001) GaN single crystal is demonstrated using Raman spectroscopy. In the two-dimensional mapping image of the E 2 H peak shift around TDs, high-contrast areas corresponding to pairs of high and low wavenumber regions of the peak shift are observed, and the edge component of the Burgers vector can be specified. Deformed contrast areas extending in the specified direction are also observed. This deformed contrast region indicates tilted dislocations since 2D mapping also detects strain deep below the surface due to the transmission of the laser beam. Simulation of the E 2 H peak shift mapping around the TD, tilted at 0° and 60° in the 11 2 ¯ 0 direction and having a Burgers vector b = 1 / 3 11 2 ¯ 0 , is also performed. The simulation of the 60° tilted dislocation shows that the contrast extends in one direction. The line profiles of the E 2 H peak shift by simulation are in agreement with the experimental results. We propose that the dislocation line tilt can be analyzed from the 2D Raman mapping images.
- 11 November 2019
- Published by Royal Society of Chemistry (RSC) in CrystEngComm
- Vol. 21 (47) , 7260-7265
Application of C-face dislocation conversion to 2 inch SiC crystal growth on an off-axis seed crystal
Combination of outward solution flow and off-axis 4H-SiC seed crystal resulted in inhomogeneities in the surface morphology and thus a distribution of TSD conversion.
- 11 November 2019
- Published by Royal Society of Chemistry (RSC) in CrystEngComm
- Vol. 21 (47) , 7260-7265
Application of C-face dislocation conversion to 2 inch SiC crystal growth on an off-axis seed crystal
Combination of outward solution flow and off-axis 4H-SiC seed crystal resulted in inhomogeneities in the surface morphology and thus a distribution of TSD conversion.
- 2 July 2018
- Published by AIP Publishing in Applied Physics Letters
- Vol. 113 (1) , 012101
Direct observation of stacking fault shrinkage in 4H-SiC at high temperatures by in-situ X-ray topography using monochromatic synchrotron radiation
An in-situ X-ray topography system using monochromatic synchrotron radiation for the observation of the stacking faults in 4H-SiC during a high-temperature annealing process was developed. We demonstrated that the stacking faults in nitrogen-doped 4H-SiC not only expand but also shrink at high temperatures. Furthermore, it was confirmed that the types of the core structure of partial dislocations enclosing the stacking fault can be determined at high temperatures.
- 2 July 2018
- Published by AIP Publishing in Applied Physics Letters
- Vol. 113 (1) , 012101
Direct observation of stacking fault shrinkage in 4H-SiC at high temperatures by in-situ X-ray topography using monochromatic synchrotron radiation
An in-situ X-ray topography system using monochromatic synchrotron radiation for the observation of the stacking faults in 4H-SiC during a high-temperature annealing process was developed. We demonstrated that the stacking faults in nitrogen-doped 4H-SiC not only expand but also shrink at high temperatures. Furthermore, it was confirmed that the types of the core structure of partial dislocations enclosing the stacking fault can be determined at high temperatures.
- Open Access
- 1 July 2018
- Published by AIP Publishing in Review of Scientific Instruments
- Vol. 89 (7) , 073103
Development of angle-resolved spectroscopy system of electrons emitted from a surface with negative electron affinity state
We developed an angle-resolved photoemission spectroscopy system for the analysis of conduction-band electrons. By forming a negative electron affinity surface on a semiconductor surface, electrons in conduction bands are emitted into a vacuum and measured by using an analyzer. This method enables us to determine the energy and momentum of the conduction electrons. Furthermore, it can be used to determine unoccupied conduction band structures. The main challenges of this method are that the energies of the emitted electrons are extremely low and the trajectories of the electrons change due to various influences. We overcame these problems by placing the shielding mesh close to the sample and parallel to the sample surface. The entire chambers, including the shielding mesh, were grounded, and a negative bias voltage was applied only to the sample. This configuration realizes the acceleration of electrons while preserving the momentum component parallel to the sample surface. Another problem is the establishment of a method for converting a detected angle into the corresponding wavevector. We focused on the emission angle of electrons emitted from a sample and their minimum energy and then established an analytical method for converting detected angles into corresponding wavevectors on the basis of the minimum energy.
- 1 July 2018
- Published by AIP Publishing in Review of Scientific Instruments
- Vol. 89 (7) , 073103
Open AccessDevelopment of angle-resolved spectroscopy system of electrons emitted from a surface with negative electron affinity state
We developed an angle-resolved photoemission spectroscopy system for the analysis of conduction-band electrons. By forming a negative electron affinity surface on a semiconductor surface, electrons in conduction bands are emitted into a vacuum and measured by using an analyzer. This method enables us to determine the energy and momentum of the conduction electrons. Furthermore, it can be used to determine unoccupied conduction band structures. The main challenges of this method are that the energies of the emitted electrons are extremely low and the trajectories of the electrons change due to various influences. We overcame these problems by placing the shielding mesh close to the sample and parallel to the sample surface. The entire chambers, including the shielding mesh, were grounded, and a negative bias voltage was applied only to the sample. This configuration realizes the acceleration of electrons while preserving the momentum component parallel to the sample surface. Another problem is the establishment of a method for converting a detected angle into the corresponding wavevector. We focused on the emission angle of electrons emitted from a sample and their minimum energy and then established an analytical method for converting detected angles into corresponding wavevectors on the basis of the minimum energy.