Khan Muhammad Umair, Kim Jungmin, Chougale Mahesh Y, Furqan Chaudhry Muhammad, Saqib Qazi Muhammad, Shaukat Rayyan Ali, Kobayashi Nobuhiko P, Mohammad Baker, Bae Jinho, Kwok Hoi-Sing
Department of Ocean System Engineering, Jeju National University, 102 Jejudaehakro, Jeju, 63243 Republic of Korea.
Department of Electrical Engineering and Computer Science, Khalifa University, Abu Dhabi, 127788 UAE.
Microsyst Nanoeng. 2022 May 26;8:56. doi: 10.1038/s41378-022-00390-2. eCollection 2022.
By exploiting ion transport phenomena in a soft and flexible discrete channel, liquid material conductance can be controlled by using an electrical input signal, which results in analog neuromorphic behavior. This paper proposes an ionic liquid (IL) multistate resistive switching device capable of mimicking synapse analog behavior by using IL BMIM FeCL and HO into the two ends of a discrete polydimethylsiloxane (PDMS) channel. The spike rate-dependent plasticity (SRDP) and spike-timing-dependent plasticity (STDP) behavior are highly stable by modulating the input signal. Furthermore, the discrete channel device presents highly durable performance under mechanical bending and stretching. Using the obtained parameters from the proposed ionic liquid-based synaptic device, convolutional neural network simulation runs to an image recognition task, reaching an accuracy of 84%. The bending test of a device opens a new gateway for the future of soft and flexible brain-inspired neuromorphic computing systems for various shaped artificial intelligence applications.
通过利用柔软且灵活的离散通道中的离子传输现象，可以使用电输入信号来控制液体材料的电导，这会导致模拟神经形态行为。本文提出了一种离子液体（IL）多态电阻开关器件，该器件能够通过将离子液体BMIM FeCL和水注入离散聚二甲基硅氧烷（PDMS）通道的两端来模拟突触模拟行为。通过调制输入信号，脉冲率依赖可塑性（SRDP）和脉冲时间依赖可塑性（STDP）行为高度稳定。此外，离散通道器件在机械弯曲和拉伸下表现出高度耐用的性能。利用从所提出的基于离子液体的突触器件获得的参数，对卷积神经网络进行图像识别任务模拟，准确率达到84%。器件的弯曲测试为未来用于各种形状人工智能应用的柔软灵活的受大脑启发的神经形态计算系统开辟了新的途径。
