All-in-one two-dimensional retinomorphic hardware device for motion detection and recognition

Zhang, Zhenhan; Wang, Shuiyuan; Liu, Chunsen; Xie, Ruizhi; Hu, Weida; Zhou, Peng

doi:10.1038/s41565-021-01003-1

Cited by 288 publications

(291 citation statements)

References 22 publications

Supporting

Mentioning

249

Contrasting

Order By: Relevance

“…Furthermore, since the 2D TMC layers can be used as LEGO-like building blocks to create a variety of heterostructures, new possibilities for research into various types of electronic/optoelectronic memristors and synaptic devices have been opened. By stacking different 2D materials, a variety of vdW heterostructures with richer optoelectronic properties can be created, allowing for the realization of retinomorphic sensors which is integrated optical perception, memory, and computation functions in crossbar array devices [ 28 ]. Under external optical/electrical stimuli, changes in conductivity of 2D materials and their heterostructures enable perception of optical information and conversion of that information into electrical signals for transmission and processing, similar to retina [ 29 , 30 ].…”

Section: Introductionmentioning

confidence: 99%

Memristive Devices Based on Two-Dimensional Transition Metal Chalcogenides for Neuromorphic Computing

et al. 2022

View full text Add to dashboard Cite

Two-dimensional (2D) transition metal chalcogenides (TMC) and their heterostructures are appealing as building blocks in a wide range of electronic and optoelectronic devices, particularly futuristic memristive and synaptic devices for brain-inspired neuromorphic computing systems. The distinct properties such as high durability, electrical and optical tunability, clean surface, flexibility, and LEGO-staking capability enable simple fabrication with high integration density, energy-efficient operation, and high scalability. This review provides a thorough examination of high-performance memristors based on 2D TMCs for neuromorphic computing applications, including the promise of 2D TMC materials and heterostructures, as well as the state-of-the-art demonstration of memristive devices. The challenges and future prospects for the development of these emerging materials and devices are also discussed. The purpose of this review is to provide an outlook on the fabrication and characterization of neuromorphic memristors based on 2D TMCs.

show abstract

Section: Introductionmentioning

confidence: 99%

Memristive Devices Based on Two-Dimensional Transition Metal Chalcogenides for Neuromorphic Computing

et al. 2022

View full text Add to dashboard Cite

show abstract

“…Both NPC and PPC are demonstrated to realize all‐in‐one perception, memory, and computing capabilities. [ 40 ] 2D‐RPP takes advantage of tunable bandgap, dielectric constant and quantum‐well structure, and its application in memory should lead to the development of devices with better performance and versatile functions.…”

Section: Resultsmentioning

confidence: 99%

Photoinduced Multi‐Bit Nonvolatile Memory Based on a van der Waals Heterostructure with a 2D‐Perovskite Floating Gate

Lai

Zhou

et al. 2022

Advanced Materials

View full text Add to dashboard Cite

inevitable, which reduces the reliability in high-density integration. Furthermore, uncontrollable interface is detrimental to the retention time and operation speed of the memory device. 2D van der Waals (vdW) crystals manifest ultra-thin layered structure, highly anisotropic in-and outof-plane conductivity and superior carrier migration, [6] which are promising to solve the above problems. Over the past decade, 2D vdW atomic crystals, such as graphene, MoS 2 , ReS 2 , etc., are widely used as the composed elements in FGNVM, [7][8][9][10][11] and extended device structures are proposed to improve the performance. [12,13] It has been demonstrated that the FGNVM based on vdW heterostructure could enable atomically sharp interface for ultrafast operation at tens of nanosecond. [14][15][16] Nonetheless, the investigation is still in the initial stage and much research work has to be carried out before the commercialization of the vdW-heterostructure-based nonvolatile memory. In conventional FGNVM, the program/erase operations are accomplished by electrical stimuli. The repetitive voltage drives not only increase the power consumption, but also deteriorate the reliability and stability of the device. [17] Light, as an efficient non-contact signal, has a small energy dissipation when traveling through the free space. Therefore, using light illumination as an auxiliary programming method is beneficial to realize long-distance quantum communication, while effectively minimizing the power consumption, so as to improve the reliability The development of floating-gate nonvolatile memory (FGNVM) is limited by the charge storage, retention and transfer ability of the charge-trapping layer.Here, it is demonstrated that due to the unique alternate inorganic/organic chain structure and superior optical sensitivity, an insulating 2D Ruddlesden-Popper perovskite (2D-RPP) layer can function both as an excellent chargestorage layer and a photosensitive layer. Optoelectronic memory composed of a MoS 2 /hBN/2D-RPP (MBR) van der Waals heterostructure is demonstrated. The MBR device exhibits unique light-controlled charge-storage characteristics, with maximum memory window up to 92 V, high on/off ratio of 10 4 , negligible degeneration over 10 3 s, >1000 program/erase cycles, and write speed of 500 µs. Dependent on the initial states, the MBR optoelectronic memory can be programmed in both positive photoconductivity (PPC) and negative photoconductivity (NPC) modes, with up to 11 and 22 distinct resistance states, respectively. The optical program power for each bit is as low as 36/10 pJ for PPC/NPC. The results not only reveal the potential of 2D-RPP as a superior charge-storage medium in floating-gate memory, but also provides an effective strategy toward fast, low-power and stable optical multi-bit storage and neuromorphic computing.

show abstract

“…[ 41 ] Due to the great reduction of massive information transmission and interaction, power consumption and latency of this architecture will be highly suppressed, and thus it will perform well in time‐critical application fields. [ 67 ] Here, we proposed a system with an in‐sensor computing architecture on the flexible devices to demonstrate motion detection with UV information extraction for the first time, as shown in Figure 5f and Table S1 (Supporting Information). In this real‐time application, the scene used for motion detection was a video of a polar bear rolling (Video S1, Supporting Information), in which two device arrays were, respectively, used to perceive images at time t and t +Δ t .…”

Section: Resultsmentioning

confidence: 99%

Flexible VO₂ Films for In‐Sensor Computing with Ultraviolet Light

Xie

Zhang

et al. 2022

Adv Funct Materials

View full text Add to dashboard Cite

With their unique advantages in portability, shape adaptability, and human friendly surfaces, flexible electronics pave the way for the implementation of wearable electronic textiles and human-machine interfaces. Although organic materials are promising for flexible devices because of the low-cost manufacturing and inherent flexibility, they meet challenges in harsh environments such as ultraviolet (UV) irradiation, which limits their applicability in UV sensors. Here, a flexible UV neuromorphic sensor is presented using inorganic vanadium dioxide (VO 2 ) films grown on mica substrates. The flexible device shows UV photoinduced nonvolatile phase transition, and can be reversibly modulated using electrolyte gating. The optical responses remain almost unchanged after 10 000 bending cycles or at small bending radius, exhibiting high tolerance to the bending deformation. Besides, the variations in image recognition accuracy under different bending conditions keep within 1.6%, indicating that the device can be adapted to various deformation conditions. By constructing near-/in-sensor computing architectures using the flexible VO 2 neuromorphic sensors with photoinduced nonvolatile phase transition, both static image processing and motion detection are realized without redundant and massive information transfer. This result lays the foundation for the development of flexible UV neuromorphic sensors.

show abstract

All-in-one two-dimensional retinomorphic hardware device for motion detection and recognition

Cited by 288 publications

References 22 publications

Memristive Devices Based on Two-Dimensional Transition Metal Chalcogenides for Neuromorphic Computing

Memristive Devices Based on Two-Dimensional Transition Metal Chalcogenides for Neuromorphic Computing

Photoinduced Multi‐Bit Nonvolatile Memory Based on a van der Waals Heterostructure with a 2D‐Perovskite Floating Gate

Flexible VO₂ Films for In‐Sensor Computing with Ultraviolet Light

Contact Info

Product

Resources

About

All-in-one two-dimensional retinomorphic hardware device for motion detection and recognition

Cited by 288 publications

References 22 publications

Memristive Devices Based on Two-Dimensional Transition Metal Chalcogenides for Neuromorphic Computing

Memristive Devices Based on Two-Dimensional Transition Metal Chalcogenides for Neuromorphic Computing

Photoinduced Multi‐Bit Nonvolatile Memory Based on a van der Waals Heterostructure with a 2D‐Perovskite Floating Gate

Flexible VO2 Films for In‐Sensor Computing with Ultraviolet Light

Contact Info

Product

Resources

About

Flexible VO₂ Films for In‐Sensor Computing with Ultraviolet Light