Biodegradable and Flexible Polymer‐Based Memristor Possessing Optimized Synaptic Plasticity for Eco‐Friendly Wearable Neural Networks with High Energy Efficiency

Abstract

The ORCID identification number(s) for the author(s) of this article can be found under https://doi.org/10.1002/aisy.202200272.

Definition of a Localized Conducting Path via Suppressed Charge Injection in Oxide Memristors for Stable Practical Hardware Neural Networks

“…Hence, it can be thought that the electroforming process with the lower charge injection played a critical role in suppressing CF overgrowth, thereby preventing the deterioration of the memristor performance. For practical smart applications of the memristors, the stable pulse operation of the devices is important. , All of the prepared devices (device 1 and device 2) demonstrated stable resistive switching behaviors under the pulse modes, regardless of the conditions for the electroforming process, and the devices showed similar switching times (about 80 and 40 ns for the set and reset processes, respectively), as shown in Figure S7. However, in the pulse cycle tests, only device 2 showed stable and reversible switching performances without any breakdown problems (see Figure S8) during 3000 cycles, which is consistent with the results from the measurements in the voltage sweeping mode (see Figure b).…”

“…An image recognition system is a representative neuromorphic application that can be used to estimate the performance of hardware neural networks. , To further verify the importance of the limited charge injection in the electroforming process for the oxide memristors, we carried out the SPICE simulations for the offline-learning based two distinct pattern recognition systems utilizing device 1 and device 2, respectively (see Figure e). Each system was constructed for classifying complex fashion images from a data set provided by the Fashion Modified National Institute of Standards and Technology .…”

“…For practical smart applications of the memristors, the stable pulse operation of the devices is important. 7,33 All of the prepared devices (device 1 and device…”

“…Therefore, it is essential to develop a promising memory device with high integration density to achieve practical neural networks. , The mechanisms and applications of a resistive switching device known as a memristor have been studied using physical models and simulations . Such a device has been demonstrated as an ideal memory component for hardware neural networks from the perspectives of integration density and electrical characteristics. − Various types of memristors, such as a photonic memristor based on the phosphorene nanoparticles, an organic memristor, − a 2D material-based memristor, and an oxide-based memristor, , have been utilized as memory components in the neuromorphic systems. In particular, the oxide memristor has been evaluated as a promising option for artificial synapses in hardware neural networks due to its large scalability and high compatibility with the complementary metal oxide semiconductor. − For the oxide-based memristor, through the application of electric stimuli, a conducting filament (CF) composed of oxygen vacancies forms within an oxide-insulating film, leading to its resistive switching characteristics.…”

Engineered current path of vertical organic phototransistors for smart optoelectronic applications

“…Numerical simulations were performed to confirm the system performance for recognizing handwritten digit images from the Modified National Institute of Standards and Technology (MNIST). 50–53 Each system was composed of three layers of neurons (784 neurons for the input images, 32 neurons for processing, and 10 neurons for the classes of numbers), as shown in Fig. 4c.…”