A reconfigurable gated Schottky diode is proposed as new high-density and low-power synaptic device that has near-linear changes in conductance. The device has a reverse current of less than 12 nA/μm ...and an effective device area of 6F 2 . Since the Al/poly-Si Schottky junction is located on the bottom gate, which has a SiO 2 /Si 3 N 4 /SiO 2 charge trap layer, the effective Schottky barrier height is modulated by the bottom gate bias or by the amount of charge trapped in the Si 3 N 4 layer. The Schottky reverse current has an exponential relationship with the effective Schottky barrier height associated with the amount of stored charge, and the amount of stored charge is logarithmically proportional to the number of potentiation pulses. Because the exponential and logarithmic relationships cancel each other out, a near-linear conductance response to the number of potentiation pulses is obtained from the proposed device.
In this work, the low-frequency noise (LFN) characteristics of hafnium-zirconium oxide (HZO) ferroelectric field-effect transistors (FeFETs) with and without high-pressure forming gas annealing (HPA) ...treatment are investigated. The origin of <inline-formula> <tex-math notation="LaTeX">1/ {f} </tex-math></inline-formula> noise in the FeFET without HPA is changed from carrier number fluctuation to Hooge's mobility fluctuation after wake-up due to the remote phonon scattering from the polarized HZO. Also, Hooge's parameter is increased by the program/erase (P/E) cycling-induced stress. On the contrary, only the correlated mobility fluctuation is increased after the wake-up in the FeFET with HPA. Furthermore, the LFN of the FeFET with HPA shows robustness to P/E cycling-induced stress after the wake-up, showing superb endurance performance.
In this paper, we propose an adaptive quantization method that can easily transfer the weights, which are trained in software network with floating point operation, to the real synaptic devices in ...hardware-based neural networks and maintain high performance. An n-type gated Schottky diode is investigated as a synaptic device, and the conductance behavior of this device is modeled successfully. Max value normalization and <inline-formula> <tex-math notation="LaTeX">3\sigma </tex-math></inline-formula> normalization are applied to the weights trained with an accuracy of 98.29% on fully connected neural network (<inline-formula> <tex-math notation="LaTeX">784\times 256\times10 </tex-math></inline-formula>) using software network. Then, the weights are quantized using the adaptive quantization method and can be transferred by adjusting the number of identical pulses applied to the synaptic devices. After applying the adaptive quantization method, accuracy rates of 98.09% and 97.20% in MNIST classification are obtained for both max value normalization and <inline-formula> <tex-math notation="LaTeX">3\sigma </tex-math></inline-formula> normalization, respectively. The proposed quantization method works well even when there is nonideality of synaptic devices such as nonlinearity of conductance behavior, limited conductance levels, and variation of conductance.
We investigate the effects of length (<inline-formula> <tex-math notation="LaTeX">{L} </tex-math></inline-formula>) and width (<inline-formula> <tex-math notation="LaTeX">{W} ...</tex-math></inline-formula>) scaling on the low-frequency noise characteristics of the ferroelectric tunnel junction (FTJ). The FTJ is composed of metal/ferroelectric/dielectric/semicondu- ctor (TiN/HfZrO 2 /SiO 2 /<inline-formula> <tex-math notation="LaTeX">{n}^{+} </tex-math></inline-formula> Si). In the high-resistance state, 1/<inline-formula> <tex-math notation="LaTeX">{f} </tex-math></inline-formula> noise increases proportionally to 1/<inline-formula> <tex-math notation="LaTeX">{W}^{\alpha }{L}^{\beta } </tex-math></inline-formula> (<inline-formula> <tex-math notation="LaTeX">\alpha \cong ~1 </tex-math></inline-formula>, <inline-formula> <tex-math notation="LaTeX">\beta >1 </tex-math></inline-formula>), whereas the shot noise has no scaling dependence. In the low-resistance state, the 1/<inline-formula> <tex-math notation="LaTeX">{f} </tex-math></inline-formula> noise of the FTJ shows a more sensitive dependence on <inline-formula> <tex-math notation="LaTeX">{L} </tex-math></inline-formula> scaling than <inline-formula> <tex-math notation="LaTeX">{W} </tex-math></inline-formula> scaling since the switching and conduction mechanisms are more affected by the process-induced damaged edge regions.
The demand for gas sensing systems that enable fast and precise gas recognition is growing rapidly. However, substantial challenges arise from the complex fabrication process of sensor arrays, ...time‐consuming data transmission to an external processor, and high energy consumption in multi‐stage data processing. In this study, a gas sensing system using on‐chip annealing for fast and power‐efficient gas detection is proposed. By utilizing a micro‐heater embedded in the gas sensor, the sensing material of adjacent sensors in the same substrate can be easily varied without further fabrication steps. The response to oxidizing gas is constrained in metal oxide (MOX) sensing material with small grain sizes, as the depletion width of grain cannot extend beyond the grain size during the gas reaction. On the other hand, the response to reducing gases and humidity, which decrease the depletion width, is less affected by grain sizes. A readout circuit integrating a differential amplifier and dual FET‐type gas sensors effectively emphasizes the response to oxidizing gases by canceling the response to reducing gases and humidity. The selective on‐chip annealing method is applicable to various MOX sensing materials, demonstrating its potential for application in commercial fields due to its simplicity and expandability.
By utilizing embedded micro‐heaters for post‐deposition annealing, metal oxide grain size and gas responses of adjacent sensors in the same substrate can be controlled without further fabrication steps. The proposed dual FET‐type gas sensors with on‐chip annealing can directly determine the gas concentration in a mixed gas environment, introducing a new approach for a real‐time energy‐efficient gas sensing system.
Retention characteristics of 3D NAND Flash cells are investigated at various temperatures (<inline-formula> <tex-math notation="LaTeX">{T} </tex-math></inline-formula>) depending on the degree of ...program and erase. The <inline-formula> <tex-math notation="LaTeX">\Delta {V}_{\text {th}} </tex-math></inline-formula> for each condition is compared to understand the degradation of the retention characteristics attributable to vertical loss and/or lateral diffusion. In addition, the relationship between Program/Erase (PE) window (PGM <inline-formula> <tex-math notation="LaTeX">{V}_{\text {th}} </tex-math></inline-formula> - Erase <inline-formula> <tex-math notation="LaTeX">{V}_{\text {th}} </tex-math></inline-formula>) and <inline-formula> <tex-math notation="LaTeX">\Delta {V}_{\text {th}} </tex-math></inline-formula> are analyzed. In the case when PGM <inline-formula> <tex-math notation="LaTeX">{V}_{\text {th}} </tex-math></inline-formula> is the same, the <inline-formula> <tex-math notation="LaTeX">\Delta {V}_{\text {th}} </tex-math></inline-formula> decreases as the PE window decreases. At temperatures below 150 °C, <inline-formula> <tex-math notation="LaTeX">\Delta {V}_{\text {th}} </tex-math></inline-formula> and PE window show linear relationship, and as PE window decreases, <inline-formula> <tex-math notation="LaTeX">\Delta {V}_{\text {th}} </tex-math></inline-formula> also decreases to 0. On the other hand, at 250 °C, <inline-formula> <tex-math notation="LaTeX">\Delta {V}_{\text {th}} </tex-math></inline-formula> has a non-zero value even if PE window decreases to 0, thus has a non-linear relationship. The measurement results show that the lateral diffusion has a great influence on the short-term retention of 3D NAND flash cells.
We demonstrate a novel ferroelectric-gate field effect transistor with recessed channel (R-FeFET) to improve memory window (MW), program/erase speed, long-time retention, and endurance ...simultaneously. Based on technology computer-aided design (TCAD) simulations including calibrated ferroelectric material (FE) parameters, it is revealed that the polarization is enhanced by the larger electric field (e-field) across the FE compared to a conventional planar FeFET, resulting in the wider MW and the faster program/erase speed. Moreover, the endurance/retention of the R-FeFET is expected to be improved as the e-field across the SiO 2 interlayer is significantly reduced.
A positive-feedback (PF) device and its operation scheme to implement pulse width modulation (PWM) function were demonstrated. First, the device operation mechanism for the PWM function was analyzed. ...By adjusting the amount of the charge stored in the n - floating body ( Q n ), the potential of the floating body changes linearly with time. When Q n reaches to a threshold value ( Q th ), the PF device turns on abruptly. From the linear time-varying property of Q n and the gate bias dependency of Q th , voltage-to-pulse width conversion for the PWM function was implemented. A rectified linear or hard-sigmoid activation function was successfully obtained from a single PF device, and it is beneficial to extremely reduce the area of a PWM neuron.
A reconfigurable field-effect transistor (RFET) with memory functionality is proposed as a new high-density synaptic device capable of exclusive NOR (XNOR) operation for a binary neural network ...(BNN). A RFET with three gates is used in this letter, consisting of two program gates (PGs) electrically connected to each other and a control gate (CG) between the PGs. By changing the polarity of the PG bias or the polarity of the charge trapped in Si 3 N 4 layer on the PGs, the RFET operates as an <inline-formula> <tex-math notation="LaTeX">{n} </tex-math></inline-formula>- or a <inline-formula> <tex-math notation="LaTeX">{p} </tex-math></inline-formula>-MOSFET having a CG as a switching gate. The XNOR operation is successfully demonstrated in a fabricated RFET. The on/off current ratio (the ratio of 1 to 0 in the result of the XNOR operation) is ~10 4 and the current is ~1 nA/<inline-formula> <tex-math notation="LaTeX">\mu \text{m} </tex-math></inline-formula>. Using a RFET as a synaptic device for a BNN is efficient in terms of the area (~8F 2 /synapse) because XNOR operation can be performed using only one RFET.