The accumulation and extrusion of Ca 2+ in the pre-and postsynaptic compartments play a critical role in initiating plastic changes in biological synapses. To emulate this fundamental process in electronic devices, we developed diffusive Ag-in-oxide memristors with a temporal response during and after stimulation similar to that of the synaptic Ca 2+ dynamics. In situ high-resolution transmission electron microscopy and nanoparticle dynamics simulations both demonstrate that Ag atoms disperse under electrical bias and regroup spontaneously under zero bias because of interfacial energy 2 minimization, closely resembling synaptic influx and extrusion of Ca 2+ , respectively.The diffusive memristor and its dynamics enable a direct emulation of both short-and long-term plasticity of biological synapses and represent a major advancement in hardware implementation of neuromorphic functionalities.CMOS circuits have been employed to mimic synaptic Ca 2+ dynamics, but three-terminal devices bear limited resemblance to bio-counterparts at the mechanism level and require significant numbers and complex circuits to simulate synaptic behavior [1][2][3] . A substantial reduction in footprint, complexity and energy consumption can be achieved by building a two-terminal circuit element, such as a memristor directly incorporating Ca 2+ -like dynamics.Various types of memristors based on ionic drift (drift-type memristor) 4-8 have recently been utilized for this purpose in neuromorphic architectures [9][10][11][12][13][14][15] . Although qualitative synaptic functionality has been demonstrated, the fast switching and non-volatility of drift memristors optimized for memory applications do not faithfully replicate the nature of plasticity. Similar issues also exist in MOS-based memristor emulators [16][17][18] , although they are capable of simulating a variety of synaptic functions including spike-timing-dependent plasticity (STDP). Recently, Lu's group adopted second-order drift memristors to approximate the Ca 2+ dynamics of chemical synapses by utilizing thermal dissipation 19 or mobility decay 20 , which successfully demonstrated STDP with non-overlapping spikes and other synaptic functions, representing a significant step towards bio-realistic synaptic devices. This approach features repeatability and simplicity, but the significant differences of the dynamical response from actual synapses limit the fidelity and variety of desired synaptic functions. A device with similar physical behavior as the biological Ca 2+ dynamics would enable improved emulation of synaptic function and broad applications to neuromorphic computing. Here we report such an emulator, which is a memristor based on metal atom 3 diffusion and spontaneous nanoparticle formation, as determined by in situ high-resolution transmission electron microscopy (HRTEM) and nanoparticle dynamics simulations. The dynamical properties of the diffusive memristors were confirmed to be functionally equivalent to Ca 2+ in bio-synapses, and their operating characteri...

No abstract

No abstract

No abstract

Memristors with tunable resistance states are emerging building blocks of artificial neural networks. However, in situ learning on a large-scale multiple-layer memristor network has yet to be demonstrated because of challenges in device property engineering and circuit integration. Here we monolithically integrate hafnium oxide-based memristors with a foundry-made transistor array into a multiple-layer neural network. We experimentally demonstrate in situ learning capability and achieve competitive classification accuracy on a standard machine learning dataset, which further confirms that the training algorithm allows the network to adapt to hardware imperfections. Our simulation using the experimental parameters suggests that a larger network would further increase the classification accuracy. The memristor neural network is a promising hardware platform for artificial intelligence with high speed-energy efficiency.