The purpose of this investigating R&D committee on magnetic sensor with machine learning is to grasp the research and development trends in the basic technology of magnetic sensor with machine learning. Since 2020 we held 12 regular meeting and co-sponsored 5 joint technical meeting on "Magnetics". With 30 committee members and invited 10 guest speakers, the committee actively discussed and investigated 33 items. Active exchanges were promoted with 85 presentations, such as jointly holding with the IEEJ D and E departments at each of the 5 joint technical meeting.

We have developed a highly sensitive tunnel magneto-resistive (TMR) sensor. A sensitivity has been dramatically improved by development of soft-magnetic materials and optimization of structure of TMR devices. In addition, 1/f noise was also decreased by improvement of integration techniques and reduction in defects of tunneling barrier layer. Because of dramatic improvement of the sensitivity and 1/f noise, we have successfully achieved the magnetic field resolution of 0.4 pT/Hz^0.5 at 1 Hz. We have also demonstrated bio-magnetic field measurement at RT using the improved TMR sensors.

The purpose of this study is to increase the output voltage for magnetic bistable device. First, this paper surveys several magnetic bistable devices, and mentions the residual stress in the magnetic wires. Two experimental results present the investigation of enhancing output voltage for magnetic bistable device using amorphous magnetic wires. One is to use the bundle of magnetic amorphous wires, and the other is to apply twisting for the magnetic amorphous wires. From the results, this paper points out the importance of proper residual stress in the magnetic wire for enhancing out put voltage for magnetic bistable device.

Thin-film magnetoimpedance element utilizes impedance change based on ferromagnetic resonance and skin effect via magnetic permeability change when a magnetic field is applied. The impedance change ratio about 400% has been achieved with a film thickness of a few µm around 1 GHz. In this paper, we systematically investigated the impedance change in the GHz range and clarified that the impedance change ratio depends on the element dimensions and that it is due to the contribution of reactance portion. In addition, the dependence of the impedance change ratio on element dimensions can be explained qualitatively based on magnetic field simulations, which also indicates a possibility of quantitative estimation.

We have developed a novel nondestructive inspection technique for infrastructures using spintronics technologies. Since the tunnel magneto-resistance (TMR) effect has become dramatically more sensitive in recent years, TMR sensors can be applied to the highly sensitive nondestructive inspection named magnetic hammering test (MHT). The proposed MHT technique is based on the principle of detecting slight fluctuations in the spatial magnetic field caused by the vibration of steel materials. Since the fluctuations in the magnetic field occur with the natural vibration frequency of the steel materials, their condition can be detected from the change of natural frequency using highly sensitive TMR sensors. In this work, we have demonstrated that the size of steel plates was determined with high accuracy using the MHT technique.

Recently, optically pumped magnetometers (OPMs), which use the spin polarization of alkali metal atoms to measure magnetic fields, have attracted much attention. However, in order to realize high-sensitivity biomagnetic-field measurements using OPMs, it is necessary to reduce environmental magnetic noise and system noise. In this study, we investigate effectiveness of the noise reduction of OPM signals when principal component analysis, independent component analysis, and empirical mode decomposition are used for signal processing.

Recently, core loss evaluation and magnetization process analysis at high-frequency have become more important. Fiorillo et al. proposed the method to analyze the frequency characteristic in magnetization process by separating into reversible and irreversible processes. However, it is necessary that the complex convergence calculation and the core loss measurements to low-frequency for analysis. Therefore, the procedure is complicated and it is difficult to apply the method to the materials which are difficult for measuring the core loss to low-frequency. In order to solve the issue, we have examined the simpler method which is separating the magnetization process into reversible and irreversible.