Artificial synapse that can mimic physiological synaptic behaviors has attracted extensive attentions in intelligent robots. However, it is an extreme challenge for artificial synapses to achieve self-optimized feedback of mimicking biological behavior. Herein, a novel self-powered artificial neural pathway (SANP) is developed by coupling CsPbBr_xI_(3-x)-based artificial synaptic device and triboelectric nanogenerator (TENG) for self-optimized neuromorphic computing. The TENG can convert external mechanical stimulation into electricity that acts not only as a supply source to power the SANP but also as electrical stimulation to transmit to the synaptic device for neuromorphic computing. The synaptic device’s conductance can be well modulated by the electrical stimulation, which tunes the height of Schottky barrier between Ag and CsPbBr_xI_(3-x), to simulate the regulation of synaptic plasticity. Simultaneously, the synaptic device can implement synaptic functions of learning and memory. Furthermore, the SANP as self-powered mechano-nociceptor can successfully mimic the nociceptor features of “threshold”, “relaxation” and “allodynia”. More importantly, under repeated mechanical stimulation, the SANP with synaptic self-optimized feedback features enables the learning and memory training and the robotic arm’s grabbing and spreading simultaneously. Consequently, the SANP can effectively accomplish signal transmission, processing, and learning tasks without external power supply, which demonstrates potential application in neuromorphic computing and intelligent robots.

中文翻译：

---

用于自我优化神经形态计算的全无机钙钛矿基人工突触器件


可以模拟生理突触行为的人工突触在智能机器人中引起了广泛关注。然而，人工突触实现模拟生物行为的自我优化反馈是一个极端的挑战。在此，通过将基于 CsPbBr_xI_（3-x） 的人工突触装置和摩擦电纳米发电机 （TENG） 耦合，开发了一种新的自供电人工神经通路 （SANP），用于自优化的神经形态计算。TENG 可以将外部机械刺激转化为电能，电能不仅可以作为为 SANP 供电的电源，还可以作为电刺激传输到突触装置以进行神经形态计算。突触装置的电导可以通过电刺激很好地调节，电刺激调节 Ag 和 CsPbBr_xI_（3-x） 之间的肖特基屏障高度，以模拟突触可塑性的调节。同时，突触装置可以实现学习和记忆的突触功能。此外，SANP 作为自动力机械伤害感受器可以成功模拟“阈值”、“放松”和“异常性疼痛”的伤害感受器特征。更重要的是，在反复的机械刺激下，具有突触自优化反馈功能的 SANP 能够同时进行学习和记忆训练以及机械臂的抓取和展开。因此，SANP 可以在没有外部电源的情况下有效地完成信号传输、处理和学习任务，这在神经形态计算和智能机器人中显示出潜在的应用。