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

Abstract

Unfortunately, the substantial power consumption has appeared to be a non-trivial burden on the training protocol of ANNs, making their feasibility challenging, especially on edge devices. One primary reason is that ANN processes information continuously with no temporal resolution, resulting in redundant power usage. Indeed, excessive computational operations in the current ANNs can barely mimic the biological behavior.On the other hand, as inspired by biological systems, the spiking neural network (SNN) has attracted ever-growing interest as the network communicates and transmits information using discrete spikes. [7][8][9][10][11][12] In biological systems, the sensory periphery receives stimulation from the surroundings and converts the analog stimuli to spikes that are subsequently relayed to the brain through sophisticated neuron connections, as shown in Figure 1a. The biological neurons collect the input spikes from other neurons, and the output spikes are generated once the membrane potential exceeds the threshold. The SNNs are believed to best represent their biological counterparts where the leaky-integrate-fire (LiF) model has been frequently implemented. [13,14] Several demonstrations have confirmed its competitiveness in various machine learning applications while consuming significantly lower energy than conventional ANNs. [7][8][9][10][11][12] Nevertheless, a biomimetic encoder is needed for converting the external stimuli, a continuous variable, to a spiking format before relaying the information to the neural network. It is crucial that the encoder must not deteriorate the performance of SNN, in which the encoding resolution, conversion speed, power consumption, and noise resilience are the essential metrics for evaluation. Although hardware encoders, including complementary metal-oxide-semiconductor (CMOS), memristors, and transistors, have been reported, the challenges remain as the encoding procedure is either sophisticated (e.g., digital circuit implementation) or power inefficient (e.g., low throughput). [15][16][17][18][19][20] Thus, this imposes stringent criteria on the fundamental level where innovative material solutions are practically indispensable. The hafnium oxide-based ferroelectric material system offers exceptional opportunities from the aforementioned aspects. We propose that the randomly distributed 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 (MoS 2 ) 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...

Dynamic Ferroelectric Transistor‐Based Reservoir Computing for Spatiotemporal Information Processing

“…This feature of ferroelectric semiconductor als is distinguished from other polarization dynamics in dielectric material such as Hf 0.5 Zr 0.5 O (HZO) and PbZr 0.4 Ti 0.6 O 3 (PZT) where n is constant due to nonexistence of mobile charges. [21,24] The endurance of volatile memory has been examined under 10 4 programing (À5 V, 100 ms)/erasing (5 V, 100 ms) cycles, shown in Figure 2i. The nondegradation of HRS and LRS of three devices reliably demonstrates that the devices are effectively switched with negligible cycle-to-cycle and device-to-device variability.…”

“…FigureS3, Supporting Information shows that fading conductance effect in α-In 2 Se 3 Fe-FET is observed in both negative and positive V p , which facilitates the implementation of highly efficient RC. This feature of ferroelectric semiconductor als is distinguished from other polarization dynamics in dielectric material such as Hf 0.5 Zr 0.5 O (HZO) and PbZr 0.4 Ti 0.6 O 3 (PZT) where n is constant due to nonexistence of mobile charges [21,24]. The endurance of volatile memory has been examined under 10 4 programing (À5 V, 100 ms)/erasing (5 V, 100 ms) cycles, shown in Figure2i.…”

Attack Resilient True Random Number Generators Using Ferroelectric‐Enhanced Stochasticity in 2D Transistor

“…Device Fabrication: The ≈12 nm HZO film was prepared using the thermal ALD at 250 °C on a sputtered 30 nm tungsten (W) bottom electrode. [35] The growth rate of HZO was found to be ≈1 Å per cycle using ellipsometry. The Tetrakis(ethylmethylamido)hafnium, Hf[N-(C 2 H 5 )CH 3 ] 4 and Tetrakis(ethylmethylamino)zirconium, Zr[N-(C 2 H 5 )CH 3 ] 4 precursors were used as the Hf and Zr metals sources, respectively.…”

“…The presence of Al 2 O 3 IL in the practical implementation is to preserve the interface integrity of the MoS 2 channel that improves the operational lifetime. [35] The contour plan of ΔV T under various pulse biases and duration is statistically calculated in Figure 2e, where the transfer characteristics can be found in Figure S14 (Supporting Information). One identifies that the magnitude of ΔV T increases with V P (V E ) and t P (t E ), where the variability (e.g., V T and I D distribution) can be enhanced, as shown in Figure S5 (Supporting Information) and Figure 2f.…”

Efficient Spiking Neural Networks with Biologically Similar Lithium-Ion Memristor Neurons

“…Inspired by the parallel and efficient processing of information in the biological brain, artificial neural networks (ANNs) have received a lot of attention and research − and have already achieved tremendous results in fields such as autonomous vehicles, biomedicine, natural language processing, and intelligent terminals . Most ANNs encode information as real-valued vectors for computation rather than as electrical spikes like the human brain, which leads to energy inefficiencies in ANNs.…”