Kanda Kensuke

Abstract In this study, a physical reservoir computing system, a hardware-implemented neural network, was demonstrated using a piezoelectric MEMS resonator. The transient response of the resonator was used to incorporate short-term memory characteristics into the system, eliminating commonly used time-delayed feedback. In addition, the short-term memory characteristics were improved by introducing a delayed signal using a capacitance-resistor series circuit. A Pb(Zr,Ti)O₃-based piezoelectric MEMS resonator with a resonance frequency of 193.2 Hz was employed as an actual node, and computational performance was evaluated using a virtual node method. Benchmark tests using random binary data indicated that the system exhibited short-term memory characteristics for two previous data and nonlinearity. To obtain this level of performance, the data bit period must be longer than the time constant of the transient response of the resonator. These outcomes suggest the feasibility of MEMS sensors with machine-learning capability.

Vibration energy harvesters that use resonance phenomena exhibit a high output power density for constant frequency vibrations, but they suffer from a significant drop in performance for non-steady-state vibrations, which are important for practical applications. In this work, we demonstrate that the output power under an impulsive force can be increased significantly by placing a U-shaped metal component, called a dynamic magnifier (DM), under an MEMS piezoelectric vibration energy harvester (MEMS-pVEH) with a 6 mm long cantilever using a 3  μm thick Pb(Zr,Ti)O₃ film. Based on the results of numerical calculations using a model of pVEH with a two-degree-of-freedom (2DOF) system, the DM was designed to have the same resonant frequency as the MEMS-pVEH and a high mechanical quality factor ([Formula: see text]). The waveforms of the output voltage of the fabricated 2DOF-pVEHs were measured for impulsive forces with various duration times, and the output power was calculated by integrating the waveforms over time. The output power of the MEMS-pVEH placed on the DM with a [Formula: see text] of 56 showed a gradual change according to the duration of applying an impulsive force and a maximum of 19 nJ/G² (G: gravitational acceleration) when the duration of the impulsive force was 3.8 ms. This result was about 90 times greater than the output power of the MEMS-pVEH without a DM. While it is not easy to fabricate pVEHs with a complex 2DOF structure using only the MEMS process, we have demonstrated that the output power can be significantly improved by adding a spring structure to a simple MEMS-pVEH.

Similarly to a harmonica reed, a piezoelectric MEMS cantilever is self-excited by an airflow. An airflow-induced self-excited vibration can be utilized as an energy source for energyharvesting devices. In this study, with the aim of reducing the cut-in flow velocity, which is the lowest flow velocity required for resonant vibration, a thin MEMS structure with an intentionally warped shape was exploited in an energy harvester based on the principle of harmonica reeds. By compensating for the residual stresses of PZT and Pt electrode films, the cantilever warpage of the harvester structure can be controlled. The thin-film nature and the warped PZT/Si laminated MEMS structure enabled energy harvesting from an airflow at low flow velocities. Moreover, the cut-in flow velocity of the airflow-induced MEMS harvesting device was very low (1.2 m/s, one-tenth of that of a conventional device), and an output power of 3.84 μW was obtained at a flow velocity of 3.7 m/s.