Memristors with threshold switching behavior are increasingly used in the study of neuromorphic computing, which are frequently used to simulate synaptic functions due to their high integration and simple structure. However, building a neuron circuit to simulate the characteristics of biological neurons is still a challenge. In this work, we demonstrate a leaky integrate-and-fire model of neurons, which is presented by a memristor-CMOS hybrid circuit based on a threshold device of a TiN/HfO2/InGaZnO4/Si structure. Moreover, we achieve multiple neural functions based on the neuron model, including leaky integration, threshold-driven fire, and strength-modulated spike frequency characteristics. This work shows that HfO2-based threshold devices can realize the basic functions of spiking neurons and have great potential in artificial neural networks.

With the advancement of artificial intelligence technology, memristors have aroused the interest of researchers because they can realize a variety of biological functions, good scalability, and high running speed. In this work, the amorphous semiconductor material silicon carbide (SiC) was used as the dielectric to fabricate the memristor with the Ag/SiC/n-Si structure. The device has a power consumption as low as 3.4 pJ, a switching ratio of up to 105, and a lower set voltage of 1.26 V, indicating excellent performance. Importantly, by adjusting the current compliance, the strength of the formed filaments changes, and the threshold characteristic and bipolar resistance switching phenomenon could be simultaneously realized in one device. On this basis, the biological long- and short-term memory process was simulated. Importantly, we have implemented leakage integration and fire models constructed based on structured Ag/SiC/n-Si memristor circuits. This low-power reconfigurable device opens up the possibilities for memristor-based applications combining artificial neurons and synapses.

As Moore's Law approaches physical limits, traditional von Neumann buildings are facing challenges. The application of memristors in multilayer storage, neuromorphic systems and analog circuits has the potential to overcome the von Neumann architecture bottleneck. Here, we fabricated high-performance memristors based on the Pd/La: HfO 2 /La 2/3 Sr 1/3 MnO 3 device on silicon substrate, which facilitate the compatibility with complementary metal oxide semiconductor processes. The memristor devices exhibited good cycling stability and multilevel resistive state storage capabilities. And the synaptic properties of the device, such as long-term potentiation/depression, short-term memory to long-term memory, spike time-dependent plasticity, and double-pulse facilitation, were also shown. Based on the brainlike synaptic behavior of the device, a high recognition rate of 91.11% was achieved in recognizing face images in neuralinspired computing. Through theoretical calculation and hardware associative learning circuit test, the hafnium-based ferroelectric memristor was successfully applied to biological associative learning behavior for the first time.

Biologically inspired neuromorphic sensory memory systems based on memristor have received a lot of attention in the booming artificial intelligence industry due to significant potential to effectively process multi‐sensory signals from complex external environments. However, many memristors have significant switching parameters disperse, which is a great challenge for using memristors in bionic neuromorphic sensory memory systems. Herein, a stable ferroelectric memristor based on the Pd/BaTiO3:Eu2O3/La0.67Sr0.33MnO3 grown on Silicon structure with SrTiO3 as buffer layer is presented. The device possesses low coercive field voltage (−1.3–2.1 V) and robust endurance characteristic (~1010 cycles) through optimizing the growth temperature. More importantly, an ultra‐stable artificial multimodal sensory memory system with visual and tactile functions was reported for the first time by combining a pressure sensor, a photosensitive sensor, and a robotic arm. Utilizing the above system, the sensitivity value of the system is expressed by the conductance of the memristor to realize the gradual change of external stimulus, and multi signals inputs at the same time to this system have faithfully achieved sensory adaptation to multimodal sensors. This work paves the way for future development of memristor‐based perception systems in efficient multisensory neural robots.image

No abstract