Dual-Ferroelectric-Coupling-Engineered Two-Dimensional Transistors for Multifunctional In-Memory Computing

The Integration of Two-Dimensional Materials and Ferroelectrics for Device Applications

In recent years, there has been growing interest in functional devices based on two-dimensional (2D) materials, which possess exotic physical properties. With an ultrathin thickness, the optoelectrical and electrical properties of 2D materials can be effectively tuned by an external field, which has stimulated considerable scientific activities. Ferroelectric fields with a nonvolatile and electrically switchable feature have exhibited enormous potential in controlling the electronic and optoelectronic properties of 2D materials, leading to an extremely fertile area of research. Here, we review the 2D materials and relevant devices integrated with ferroelectricity. This review starts to introduce the background about the concerned themes, namely 2D materials and ferroelectrics, and then presents the fundamental mechanisms, tuning strategies, as well as recent progress of the ferroelectric effect on the optical and electrical properties of 2D materials. Subsequently, the latest developments of 2D material-based electronic and optoelectronic devices integrated with ferroelectricity are summarized. Finally, the future outlook and challenges of this exciting field are suggested.

Monolithically Integrated Complementary Ferroelectric FET XNOR Synapse for the Binary Neural Network

Neuromorphic computing, which mimics the structure and principles of the human brain, has the potential to facilitate the hardware implementation of next-generation artificial intelligence systems and process large amounts of data with very low power consumption. Among them, the XNOR synapse-based Binary Neural Network (BNN) has been attracting attention due to its compact neural network parameter size and low hardware cost. The previous XNOR synapse has drawbacks, such as a trade-off between cell density and accuracy. In this work, we show nonvolatile XNOR synapses with high density and accuracy using a monolithically stacked complementary ferroelectric field-effect transistor (C-FeFET) composed of a p-type Si MFMIS-FeFET at the bottom and a 3D stackable n-type Al:IZTO MFS-FeTFT, achieving 60F2 per cell (2C-FeFET). For adjusting the threshold voltage and improving the switching speed (100 ns) of n-type ferroelectric TFT, we employed a dual-gate configuration and a unique operation scheme, making it comparable to those of Si-based FeFETs. We performed array-level simulation with a 512 × 512 subarray size and a 3-bit flash ADC, demonstrating that the image recognition accuracies using the MNIST and CIFAR-10 data sets were increased by 3.17 and 14.07%, respectively, in comparison to other nonvolatile XNOR synapses. In addition, we performed system-level analysis on a 512 × 512 XNOR C-FeFET, exhibiting an outstanding throughput of 717.37 GOPS and an energy efficiency of 196.7 TOPS/W. We expect that our approach would contribute to the high-density memory systems, logic-in-memory technology, and hardware implementation of neural networks.

2D Dual Gate Field-Effect Transistor Enabled Versatile Functions

Advanced computing technologies such as distributed computing and the Internet of Things require highly integrated and multifunctional electronic devices. Beyond the Si technology, 2D-materials-based dual-gate transistors are expected to meet these demands due to the ultra-thin body and the dangling-bond-free surface. In this work, a molybdenum disulfide (MoS2 ) asymmetric-dual-gate field-effect transistor (ADGFET) with an In2 Se3 top gate and a global bottom gate is designed. The independently controlled double gates enable the device to achieve an on/off ratio of 106 with a low subthreshold swing of 94.3 mV dec-1 while presenting a logic function. The coupling effect between the double gates allows the top gate to work as a charge-trapping layer, realizing nonvolatile memory (105 on/off ratio with retention time over 104 s) and six-level memory states. Additionally, ADGFET displays a tunable photodetection with the responsivity reaching the highest value of 857 A W-1 , benefiting from the interface coupling between the double gates. Meanwhile, the photo-memory property of ADGFET is also verified by using the varying exposure dosages-dependent illumination. The multifunctional applications demonstrate that the ADGFET provides an alternative way to integrate logic, memory, and sensing into one device architecture.

A review on advanced band-structure engineering with dynamic control for nonvolatile memory based 2D transistors

With advancements in information technology, an enormous amount of data is being generated that must be quickly accessible. However, conventional Si memory cells are approaching their physical limits and will be unable to meet the requirements of intense applications in the future. Notably, 2D atomically thin materials have demonstrated multiple novel physical and chemical properties that can be used to investigate next-generation electronic devices and breakthrough physical limits to continue Moore's law. Band structure is an important semiconductor parameter that determines their electrical and optical properties. In particular, 2D materials have highly tunable bandgaps and Fermi levels that can be achieved through band structure engineering methods such as heterostructure, substrate engineering, chemical doping, intercalation, and electrostatic doping. In particular, dynamic control of band structure engineering can be used in recent advancements in 2D devices to realize nonvolatile storage performance. This study examines recent advancements in 2D memory devices that utilize band structure engineering. The operational mechanisms and memory characteristics are described for each band structure engineering method. Band structure engineering provides a platform for developing new structures and realizing superior performance with respect to nonvolatile memory.

Integration of image preprocessing and recognition functions in an optoelectronic coupling organic ferroelectric retinomorphic neuristor

The human visual system (HVS) has the advantages of a low power consumption and high efficiency because of the synchronous perception and early preprocessing of external image information in the retina, as well as parallel in-memory computing within the visual cortex. Realizing the biofunction simulation of the retina and visual cortex in a single device structure provides opportunities for performance improvements and machine vision system (MVS) integration. Here, we fabricate organic ferroelectric retinomorphic neuristors that integrate the retina-like preprocessing function and recognition of the visual cortex in a single device architecture. Benefiting from the electrical/optical coupling modulation of ferroelectric polarization, our devices show a bidirectional photoresponse that acts as the basis for mimicking retinal preconditioning and multi-level memory capabilities for recognition. The MVS based on the proposed retinomorphic neuristors achieves a high recognition accuracy of ∼90%, which is 20% higher than that of the incomplete system without the preprocessing function. In addition, we successfully demonstrate image encryption and optical programming logic gate functions. Our work suggests that the proposed retinomorphic neuristors offer great potential for MVS monolithic integration and functional expansion.

Reconfigurable Logic-in-Memory Computing Based on a Polarity-Controllable Two-Dimensional Transistor

Logic-in-memory architecture holds great promise to meet the high-performance and energy-efficient requirements of data-intensive scenarios. Two-dimensional compacted transistors embedded with logic functions are expected to extend Moore's law toward advanced nodes. Here we demonstrate that a WSe₂/h-BN/graphene based middle-floating-gate field-effect transistor can perform under diverse current levels due to the controllable polarity by the control gate, floating gate, and drain voltages. Such electrical tunable characteristics are employed for logic-in-memory architectures and can behave as reconfigurable logic functions of AND/XNOR within a single device. Compared to the conventional devices like floating-gate field-effect transistors, our design can greatly decrease the consumption of transistors. For AND/NAND, it can save 75% transistors by reducing the transistor number from 4 to 1; for XNOR/XOR, it is even up to 87.5% with the number being reduced from 8 to 1.

Wurtzite and fluorite ferroelectric materials for electronic memory

Ferroelectric materials, the charge equivalent of magnets, have been the subject of continued research interest since their discovery more than 100 years ago. The spontaneous electric polarization in these crystals, which is non-volatile and programmable, is appealing for a range of information technologies. However, while magnets have found their way into various types of modern information technology hardware, applications of ferroelectric materials that use their ferroelectric properties are still limited. Recent advances in ferroelectric materials with wurtzite and fluorite structure have renewed enthusiasm and offered new opportunities for their deployment in commercial-scale devices in microelectronics hardware. This Review focuses on the most recent and emerging wurtzite-structured ferroelectric materials and emphasizes their applications in memory and storage-based microelectronic hardware. Relevant comparisons with existing fluorite-structured ferroelectric materials are made and a detailed outlook on ferroelectric materials and devices applications is provided.

Carrier Modulation in 2D Transistors by Inserting Interfacial Dielectric Layer for Area-Efficient Computation

2D materials with atomic thickness display strong gate controllability and emerge as promising materials to build area-efficient electronic circuits. However, achieving the effective and nondestructive modulation of carrier density/type in 2D materials is still challenging because the introduction of dopants will greatly degrade the carrier transport via Coulomb scattering. Here, a strategy to control the polarity of tungsten diselenide (WSe₂ ) field-effect transistors (FETs) via introducing hexagonal boron nitride (h-BN) as the interfacial dielectric layer is devised. By modulating the h-BN thickness, the carrier type of WSe₂ FETs has been switched from hole to electron. The ultrathin body of WSe₂ , combined with the effective polarity control, together contribute to the versatile single-transistor logic gates, including NOR, AND, and XNOR gates, and the operation of only two transistors as a half adder in logic circuits. Compared with the use of 12 transistors based on static Si CMOS technology, the transistor number of the half adder is reduced by 83.3%. The unique carrier modulation approach has general applicability toward 2D logic gates and circuits for the improvement of area efficiency in logic computation.

Recent Advances in Ferroelectric-Enhanced Low-Dimensional Optoelectronic Devices

Ferroelectric (FE) materials, including BiFeO₃ , P(VDF-TrFE), and CuInP₂ S₆ , are a type of dielectric material with a unique, spontaneous electric polarization that can be reversed by applying an external electric field. The combination of FE and low-dimensional materials produces synergies, sparking significant research interest in solar cells, photodetectors (PDs), nonvolatile memory, and so on. The fundamental aspects of FE materials, including the origin of FE polarization, extrinsic FE materials, and FE polarization quantification are first discussed. Next, the state-of-the-art of FE-based optoelectronic devices is focused. How FE materials affect the energy band of channel materials and how device structures influence PD performance are also summarized. Finally, the future directions of this rapidly growing field are discussed.

Van der Waals Ferroelectric Semiconductor Field Effect Transistor for In-Memory Computing

In-memory computing is a highly efficient approach for breaking the bottleneck of von Neumann architectures, i.e., reducing redundant latency and energy consumption during the data transfer between the physically separated memory and processing units. Herein we have designed a in-memory computing device, a van der Waals ferroelectric semiconductor (InSe) based metal-oxide-ferroelectric semiconductor field-effect transistor (MOfeS-FET). This MOfeS-FET integrates memory and logic functions in the same material, in which the out-of-plane (OOP) ferroelectric polarization in InSe is used for data storage and the semiconducting property is used for the logic computation. The MOfeS-FET shows a long retention time with high on/off ratios (>10⁶), high program/erase (P/E) ratios (10³), and stable cyclic endurance. Moreover, inverter, programmable NAND, and NOR Boolean logic operations with nonvolatile storage of the results have all been demonstrated using our approach. These findings highlight the potential of van der Waals ferroelectric semiconductor-based MOfeS-FETs in the in-memory computing and their potential of achieving size scaling beyond Moore's law.

Analog Resistive Switching and Artificial Synaptic Behavior of ITO/WOX/TaN Memristors

In this work, we fabricated an ITO/WO_X/TaN memristor device by reactive sputtering to investigate resistive switching and conduct analog resistive switching to implement artificial synaptic devices. The device showed good pulse endurance (10⁴ cycles), a high on/off ratio (>10), and long retention (>10⁴ s) at room temperature. The conduction mechanism could be explained by Schottky emission conduction. Further, the resistive switching characteristics were performed by additional pulse-signal-based experiments for more practical operation. Lastly, the potentiation/depression characteristics were examined for 10 cycles. The results thus indicate that the WO_X-based devices are appropriate candidates for synaptic devices as well as next-generation nonvolatile memory.

Natural Organic Materials Based Memristors and Transistors for Artificial Synaptic Devices in Sustainable Neuromorphic Computing Systems

Natural organic materials such as protein and carbohydrates are abundant in nature, renewable, and biodegradable, desirable for the construction of artificial synaptic devices for emerging neuromorphic computing systems with energy efficient operation and environmentally friendly disposal. These artificial synaptic devices are based on memristors or transistors with the memristive layer or gate dielectric formed by natural organic materials. The fundamental requirement for these synaptic devices is the ability to mimic the memory and learning behaviors of biological synapses. This paper reviews the synaptic functions emulated by a variety of artificial synaptic devices based on natural organic materials and provides a useful guidance for testing and investigating more of such devices.

A MoS2 Hafnium Oxide Based Ferroelectric Encoder for Temporal-Efficient Spiking Neural Network

Spiking neural network (SNN), where the information is evaluated recurrently through spikes, has manifested significant promises to minimize the energy expenditure in data-intensive machine learning and artificial intelligence. Among these applications, the artificial neural encoders are essential to convert the external stimuli to a spiking format that can be subsequently fed to the neural network. Here, a molybdenum disulfide (MoS₂ ) hafnium oxide-based ferroelectric encoder is demonstrated for temporal-efficient information processing in SNN. The fast domain switching attribute associated with the polycrystalline nature of hafnium oxide-based ferroelectric material is exploited for spike encoding, rendering it suitable for realizing biomimetic encoders. Accordingly, a high-performance ferroelectric encoder is achieved, featuring a superior switching efficiency, negligible charge trapping effect, and robust ferroelectric response, which successfully enable a broad dynamic range. Furthermore, an SNN is simulated to verify the precision of the encoded information, in which an average inference accuracy of 95.14% can be achieved, using the Modified National Insitute of Standards and Technology (MNIST) dataset for digit classification. Moreover, this ferroelectric encoder manifests prominent resilience against noise injection with an overall prediction accuracy of 94.73% under various Gaussian noise levels, showing practical promises to reduce the computational load for the neural network.

Nonvolatile Memories in Spiking Neural Network Architectures: Current and Emerging Trends

A sustainable computing scenario demands more energy-efficient processors. Neuromorphic systems mimic biological functions by employing spiking neural networks for achieving brain-like efficiency, speed, adaptability, and intelligence. Current trends in neuromorphic technologies address the challenges of investigating novel materials, systems, and architectures for enabling high-integration and extreme low-power brain-inspired computing. This review collects the most recent trends in exploiting the physical properties of nonvolatile memory technologies for implementing efficient in-memory and in-device computing with spike-based neuromorphic architectures.