Artificial synapses and neurons are recognized as key elements in building bioinspired, neuromorphic computing systems. However, synaptic and neuronal elements that have compatible material systems with each other with high scalability and biorealistic dynamics are yet to be demonstrated. Here we report a two-terminal memristive synapse that can realize short-term and long-term plasticity in both potentiation and depression processes. The Ag nanoclusters introduced at the interface can move, connect and redistribute in response to applied pulses, where their electrochemical migration and thermodynamic relaxation in dielectrics compete with each other and faithfully emulate the synaptic and neuronal dynamics in biology, which in turn allows the same devices to exhibit various synaptic functions and neuronal spiking in a scalable manner. The evolution dynamics of Ag nanoclusters was verified using high resolution transmission electron microscopy and compositional analyses. Based on the inherent state modulator and timing mechanism offered by such dynamics, the devices were able to naturally implement complex functions including metaplasticity, asynchronous classical conditioning and spike-timing-dependent plasticity without needing intentionally designed overlapping pulses, thus paving the way for the construction of intelligent neuromorphic systems capable of encoding and processing spatiotemporal information.

中文翻译：

---

金属纳米团簇的电化学和热力学过程实现了逼真的突触和泄漏整合并发射的神经元†

人工突触和神经元被认为是构建受生物启发的神经形态计算系统的关键要素。然而，具有高可扩展性和生物真实性动力学的具有彼此兼容的材料系统的突触和神经元元件尚未被证实。在这里，我们报告了两个末端的忆阻性突触，可以在增强和抑制过程中实现短期和长期可塑性。在界面处引入的Ag纳米团簇可以响应施加的脉冲而移动，连接和重新分布，它们在电介质中的电化学迁移和热力学松弛相互竞争，并忠实地模拟生物学中的突触和神经元动力学，从而允许使用相同的设备以可扩展的方式展现各种突触功能和神经元突增。Ag纳米团簇的演化动力学通过高分辨率透射电子显微镜和成分分析得到验证。基于这种动力学提供的固有状态调制器和定时机制，这些设备能够自然实现复杂的功能，包括超塑性，异步经典调节和依赖于尖峰时序的可塑性，而无需故意设计重叠脉冲，从而为构造铺平了道路。能编码和处理时空信息的智能神经形态系统