Sustainable Synaptic Device with Two-Dimensional Ferroelectric Materials for Neuromorphic Computing.

Neuromorphic computing systems are based on new architecture inspired by biological learning behavior and biological structure to overcome the shortcomings of the von Neumann system, such as high energy consumption and generation of excessive redundant data. However, common synaptic‐device designs are still need to carry an external power source, which is detrimental to both the miniaturization of smart devices and the further reduction of energy consumption. Therefore, self‐powered optoelectronic synapses hold great promise for achieving energy‐efficient artificial intelligence applications. Given the advantages of heterogeneous integration and ferroelectric performance, the 2D ferroelectric In2Se3 has become a typical candidate material for biological synaptic devices and neuromorphic computing. Although, various structures of ferroelectric synaptic devices are developed, most of which are controlled by optical or electrical pulses, hardly any reports discuss the operation in self‐powered mode. In this work, an In2Se3/MoS2 ferroelectric lateral heterojunction‐type optoelectronic synaptic device is fabricated to mimic biological synaptic functions in voltage mode or self‐driven mode. In addition, two simplified image pre‐processing methods based on device arrays are shown to improve the recognition accuracy of a back‐end neural network. The proposed approach suggests a new route for designing and applying high‐performance synaptic devices.

In this work a discussion on the estimated volume trap densities (NT) compared to surface traps densities (Neff) related to identified traps in the Si fin of Si/SiGe superlattice I/O n-channel FinFETs using low frequency noise spectroscopy is made. The investigated devices present a fin width of 10 nm, a fin height of 10 nm and four fins in parallel, leading to an equivalent channel width of 120 nm. The equivalent oxide thickness (EOT) is 5.6 nm. More details on the device fabrication and experimental setup may be found in [1,2]. The noise spectra and the estimated surface traps densities are provided in [3]. In this work, focus is only on the identified T4 trap, for which using the linear dependence that should exist between A0i and t0i related to the same trap, without any other assumption, a surface trap density of 2.8∙1012cm-2 is obtained [3]. It should be noted that from the 1/f flat-band noise level an interface trap density value of about 1.9∙1018 eV-1cm-3 is obtained at 300 K.However, as the traps in the Si film are related to a volume phenomenon, two methodologies to estimate the volume trap densities are employed: one using the relationship between the surface trap density and volume trap densitiy [4], where B is a coefficient estimated to be 1/3 [4,5]; and a second one from the temperature (T) evolution at fixed frequency of the Lorentzian plateau level associated to the same trap (from equation 34 in [4]). Using the first method leads to a volume trap density of T4 of about 1.7∙1019 cm-3.The second method consists to use the maximum of the measured Svg_Lor(f 0,T) dependence with temperature. Indeed, the Svg_Lor(f 0,T) of Lorentzians associated to the same trap are proportional with ti(T)/{1 + [2pf 0ti(T)]2}. For a given frequency f 0, if 2pf 0ti(T) ≫ 1, SVg_Lor(f 0,T) ∝ ti(T)]-1, and SVg_Lor(f 0,T) increases with increasing temperature because ti decreases. If 2pf 0ti(T) ≪ 1, then SVg_Lor(f 0,T) ∝ ti(T) and SVg_Lor(f 0,T) decreases with increasing temperature, as explained in detail in [4]. The evolution of the Svg_Lor(f 0,T)∙f 0 in a temperature range where T4 traps are active is illustrated in Figure 1 and presents a bell-shaped behavior, as expected. Using this method, volume trap densities of about 1.25∙1018 cm-3 for f 0 = 10 kHz and of about 1.1∙1018 cm-3 for f 0 = 14 kHz are obtained. It may be observed that the estimated volume trap densities of the T4 defect are about one decade lower than when using the first method. This overestimation by the first method is related to the fact that the theoretical B coefficient was determined for conventional planar devices with one gate [4,5]. Moreover, it may be noticed that the second method is dependent on the fixed f 0 selected.The paper will present a discussion with more details, considering all identified traps in [1] and considering additional new low frequency noise spectroscopy results.References :[1] Hellings, H. Mertens, A. Subirats, E. Simoen, T. Schram, L.-A. Ragnarsson et al., “Si/SiGe superlattice I/O finFETs in a vertically-stacked Gate-All-Around horizontal Nanowire Technology”, in Tech. Dig. Symp. on VLSI Technology, The IEEE New York, 2018, p.p. 85-86, DOI: 10.1109/VLSIT.2018.8510654.[2] Boudier, B. Cretu, E. Simoen, R. Carin, A. Veloso, N. Collaert, and A. Thean, “Low frequency noise assessment in n- and p-channel sub-10 nm triple-gate FinFETs: Part I: Theory and methodology,” Solid State Electron., vol. 128, pp. 102-108, 2017, DOI: 10.1016/j.sse.2016.10.012.[3] Boudier, B. Cretu, E. Simoen, G. Hellings, T. Schram, H. Mertens and D. Linten, “Low frequency noise analysis on Si/SiGe superlattice I/O n-channel FinFETs”, In Proceedings of EUROSOI-ULIS’2019.[4] Lukyanchikova, “Noise and Fluctuations Control in Electronic Device”, edited by A. Balandin, American Scientific, Riverside, CA, 2002, pp. 201-233.[5] Yau and C-T. Sah, “Theory and experiments of low-frequency generation-recombination noise in MOS transistors”, IEEE Trans. Electron Dev., 1969, vol.16, pp. 170-177.Figure 1 : SVg_Lor(f 0,T)·f 0 versus temperature for the T4 trap identified in [3]; on the secondary Oy axis the characteristic frequency f 0i of the Lorentzians is displayed in function of temperature. Figure 1

We have used low-frequency noise (LFN) spectroscopy to characterize generation-recombination (G-R) centers in silicon nanowires grown using chemical vapor deposition. The LFN spectra showed Lorentzian behavior with well-defined corner-frequency indicative of single G-R center in the bandgap. From the temperature-dependent LFN measurement a single deep level at 0.39 eV from the bandedge is identified, which matches closely with the Au donor level in Si. The trap concentration was estimated at 2.0 × 1012 cm−3 with electron and hole capture cross-sections of 9.5 × 10−17 cm2 and 1.4 × 10−16 cm2, respectively. This study demonstrates the potential of the LFN spectroscopy in characterization of deep-levels in nanowires.

Nonvolatile memory (NVM)‐based neuromorphic computing has been attracting considerable attention from academia and the industry. Although it is not completely successful yet, remarkable achievements have been reported pertaining to synaptic devices that can leverage NVM capable of storing multiple states. The analog synaptic devices performing computation similar to biological nerve systems are crucial in energy‐efficient analog neuromorphic computing systems. To use NVM as an analog synaptic device, researchers focus on improving device characteristics. Among various characteristics, the most challenging one is linearity and symmetry of synaptic weight update that is required for on‐chip training. In this regard, this review paper discusses recent synaptic device improvements focusing on novel schemes tailored for each NVM device to improve the linearity and symmetry. In addition to device‐level studies, recent research achievements are reviewed expanded up to chip‐level studies because in realizing neuromorphic hardware systems beyond a single synaptic device, several considerations and requirements are needed to confirm for high‐level design, and accordingly, cooptimize among synaptic devices, synapse arrays, electrical circuits, neural networks, algorithms, and implementation. Also, this review introduces various circuit and algorithmic approaches to compensate for the non‐ideality of the analog synaptic device.

Synaptic devices based neuromorphic computing applications in artificial intelligence

Filamentary and interface switching of CMOS-compatible Ta2O5 memristor for non-volatile memory and synaptic devices

As Moore's law approaches its limits, it is gradually harder to achieve the more and more complex tasks with computers with conventional von Neumann architecture. The hardware circuits for neuromorphic computing, with low power and high computing density, have attracted a lot of attention. Based on these requirements, how to choose the appropriate circuit components for neuromorphic computing becomes very important. Ambipolar synaptic devices with tunable and versatile behaviours, as a new type of circuit component, can mimic both excitatory and inhibitory behaviours, which has great potential in neuromorphic computing. In this chapter, we will introduce the definitions of ambipolar materials and synaptic devices. Then, the characteristics and applications of the synaptic devices will be discussed. The differences between conventional and ambipolar synaptic devices will also be compared. Finally, we will introduce the possibility of applying ambipolar synaptic devices in array to achieve more complex work.

9 - Novel photoelectroactive memories and neuromorphic devices based on nanomaterials

In this paper, gradual and symmetrical long-term potentiation (LTP) and long-term depression (LTD) were achieved by applying the optimal electrical pulse condition of the interfacial phase-change memory (iPCM) based on a superlattice (SL) structure fabricated by stacking GeTe/Sb2Te3 alternately to implement an artificial synapse in neuromorphic computing. Furthermore, conventional phase-change random access memory (PCRAM) based on a Ge–Sb–Te (GST) alloy with an identical bottom electrode contact size was fabricated to compare the electrical characteristics. The results showed a reduction in the reset energy consumption of the GeTe/Sb2Te3 (GT/ST) iPCM by more than 69% of the GST alloy for each bottom electrode contact size. Additionally, the GT/ST iPCM achieved gradual conductance tuning and 90.6% symmetry between LTP and LTD with a relatively unsophisticated pulse scheme. Based on the above results, GT/ST iPCM is anticipated to be exploitable as a synaptic device used for brain-inspired computing and to be utilized for next-generation non-volatile memory.

The development of synaptic devices featuring metallic nanostructures with brain-analog hierarchical architecture, capable of mimicking cognitive functionalities, has emerged as a focal point in neuromorphic computing. However, existing challenges, such as inconsistent and unpredictable switching, high voltage requirements, unguided filament formation, and detailed fabrication processes, have impeded technological progress in the domain. The present study addresses some of these challenges by leveraging periodic nanostructures of Ag fabricated via plasma-assisted nanosphere lithography (NSL). The triangular nanostructures with a preferred orientation offer enhanced localized electric fields, facilitating low voltage electromigration at the sharp edges to guide predictive filament formation. A thorough investigation into gap control between the nanostructures through oxygen plasma treatment enables the attainment of an optimized low switching voltage of 0.86 V and retention at an ultra-low current compliance of 100 nA. The optimized device consumes low power, typically in the fJ range, akin to biological neurons. Furthermore, the device showcases intriguing synaptic characteristics, including controlled transition from short- to long-term potentiation, associative learning, etc., projecting its potential in perceptive learning, memory formation, and brain-inspired computing. COMSOL Multiphysics simulation, supported by ex situ electron microscopic imaging, confirms the controlled and predictable filament formation facilitated by electric field enhancement across the strategic nanostructures. Thus, the work highlights the potential of NSL-based cost-effective fabrication techniques for realizing efficient and biomimetic synaptic devices for neuromorphic computing applications.

Neuromorphic computing can efficiently handle data-intensive tasks and address the redundant interaction required by von Neumann architectures. Synaptic devices are essential components for neuromorphic computation. 2D phosphorene, such as violet phosphorene, show great potential in optoelectronics due to their strong light-matter interactions, while current research is mainly focused on synthesis and characterization, its application in photoelectric devices is vacant. Here, the authors combined violet phosphorene and molybdenum disulfide to demonstrate an optoelectronic synapse with a light-to-dark ratio of 106 , benefiting from a significant threshold shift due to charge transfer and trapping in the heterostructure. Remarkable synaptic properties are demonstrated, including a dynamic range (DR) of > 60dB, 128 (7-bit) distinguishable conductance states, electro-optical dependent plasticity, short-term paired-pulse facilitation, and long-term potentiation/depression. Thanks to the excellent DR and multi-states, high-precision image classification with accuracies of 95.23% and 79.65% is achieved for the MNIST and complex Fashion-MNIST datasets, which is close to the ideal device (95.47%, 79.95%). This work opens the way for the use of emerging phosphorene in optoelectronics and provides a new strategy for building synaptic devices for high-precision neuromorphic computing.

Zero-power optoelectronic synaptic devices

In this study, we demonstrate the capability to detect microscopic stable displacement nature created by proton irradiations with a method based on the discrimination of the three major low frequency noise sources generally observed in semiconductor devices. Low frequency noise measurements are carried out in silicon in-situ phosphorus doped polycrystalline silicon serpentine resistance used as test vehicle. We observe that for the pristine sample, the voltage noise power density can be explained only considering 1/f and thermal noise contribution on the total noise, while for irradiated sample generation recombination noise prevails. Preliminary results on low frequency noise spectroscopy, used as a diagnostic tool in order to identify stable traps created by protons irradiations are presented.

A continuous exponential rise has been observed in the storage and processing of the data that may not curtail in the foreseeable future. The required data processing speed and power consumption are restricted by the buses between the logic and memory devices that are characteristic of the von Neumann computing architecture. Bio‐mimicking neuromorphic computing has garnered considerable academic and industrial interest to resolve these challenges. Additionally, devices based on emerging nonvolatile memories capable of mimicking the behaviors of synapses and neurons, which are the main elements in biological computing systems (brains), are attracting significant interest from the device community. With the discovery of ferroelectricity in fluorite‐structured oxides, such as HfO2and ZrO2, which are compatible with the state‐of‐the‐art complementary‐metal‐oxide‐semiconductor processes, ferroelectric devices have rapidly evolved as the main direction of these research and development activities. Fundamental science related to fluorite‐structured ferroelectrics has been intensively studied over the last decade. At present, the focus is gradually moving to practical applications, including neuromorphic computing and advanced classical processing or memory units in the conventional von Neumann architecture. However, despite its rapid development, the wealth of recent progress in neuromorphic computing devices based on fluorite‐structured ferroelectrics has not been reviewed and systemized. This progress report comprehensively reviews and systemizes the recent progress in artificial synaptic and spiking neuron devices for neuromorphic computing based on fluorite‐structured ferroelectrics.image

Optical artificial synapses possess several advantages, including high bandwidth, strong interference immunity, and ultra-fast signal transmission, overcoming the limitations of electrically stimulated synapses. Among various functional materials, 2D materials exhibit exceptional optical and electrical properties. By utilizing van der Waals heterostructures formed by these materials through rational design, synaptic devices can mimic the information perception ability of biological systems. This lays the foundation for low-energy artificial vision systems and neuromorphic computing. This study introduces an inhibitory artificial synapse based on photoelectric co-modulation of graphene/WSe2 van der Waals heterojunctions. By synergistically applying gate voltage and light pulses, we simulate memory and logic functions observed in the brain’s visual cortex. We achieve the construction of inhibitory synapses, enabling properties such as postsynaptic current response, short-term and long-term plasticity, and paired-pulse facilitation. Additionally, we accomplish the inverse recovery of device conductivity through separate gate voltage stimulation. Through bidirectional modulation of the artificial synaptic conductance, we construct an artificial hardware neural network that achieves 92.5% accuracy in recognizing handwritten digital images from the MNIST dataset. The network also has good recognition accuracy for handwritten digital images with different standard deviation Gaussian noise applied and other datasets. Furthermore, we successfully mimic the neural behavior of aversive learning for alcohol withdrawal in alcoholic patients using the device properties. The promising capabilities of artificial synapses constructed through electrical and optical synergistic modulation make them suitable for wearable electronics and artificial vision systems.