Prediction of Process Variation Effect for Ultrascaled GAA Vertical FET Devices Using a Machine Learning Approach
TL;DR: An accurate and efficient machine learning (ML) approach which predicts variations in key electrical parameters using process variations (PVs) from ultrascaled gate-all-around (GAA) vertical FET (VFET) devices with the same degree of accuracy, as well as improved efficiency compared to a 3-D stochastic TCAD simulation.
Abstract: In this brief, we present an accurate and efficient machine learning (ML) approach which predicts variations in key electrical parameters using process variations (PVs) from ultrascaled gate-all-around (GAA) vertical FET (VFET) devices. The 3-D stochastic TCAD simulation is the most powerful tool for analyzing PVs, but for ultrascaled devices, the computation cost is too high because this method requires simultaneous analysis of various factors. The proposed ML approach is a new method which predicts the effects of the variability sources of ultrascaled devices. It also shows the same degree of accuracy, as well as improved efficiency compared to a 3-D stochastic TCAD simulation. An artificial neural network (ANN)-based ML algorithm can make multi-input -multi-output (MIMO) predictions very effectively and uses an internal algorithm structure that is improved relative to existing techniques to capture the effects of PVs accurately. This algorithm incurs approximately 16% of the computation cost by predicting the effects of process variability sources with less than 1% error compared to a 3-D stochastic TCAD simulation.
Citations
More filters
TL;DR: An intelligent OP prediction algorithm based on the improved cuckoo search (ICS) is presented and the results show that it has a better OP prediction performance than the existing algorithms.
Abstract: In the field of transportation, the Internet of Vehicles (IoV) is an important component of the Internet of Things. The vehicle-to-vehicle communication is particularly challenging in mobile IoV networks because they are operated in complex and highly variable environments. The mobile IoV transmission interruption level can be evaluated by the outage probability (OP) performance. If the OP performance can be analyzed and predicted accurately, the Quality of Service (QoS) in the mobile IoV networks can be improved. However, the analysis and prediction of mobile IoV transmission channels is very challenging because they are highly dynamic. In this article, the analysis and prediction of the OP performance for mobile IoV networks are investigated. A hybrid decode-amplify-forward (HDAF) relaying scheme with transmit antenna selection (TAS) is considered. The exact OP expressions are derived in a closed form, and the analytical results are verified. To realize the real-time analysis of the OP performance, an intelligent OP prediction algorithm based on the improved cuckoo search (ICS) is presented. The proposed algorithm is compared with different methods and the results show that it has a better OP prediction performance. The prediction accuracy of ICS-BP can be increased by 51.8% compared with the existing algorithms.
38 citations
Additional excerpts
...Machine learning (ML) has recently been applied successfully in a variety of wireless communication applications [23], [24]....
[...]
TL;DR: A variability-aware machine learning (ML) approach that predicts variations in the key electrical parameters of 3-D NAND Flash memories caused by various sources of variability and verified the accuracy, efficiency, and generality of artificial neural network (ANN) algorithm-based ML systems.
Abstract: This article proposes a variability-aware machine learning (ML) approach that predicts variations in the key electrical parameters of 3-D NAND Flash memories. For the first time, we have verified the accuracy, efficiency, and generality of the predictive impact factor effects of artificial neural network (ANN) algorithm-based ML systems. ANN-based ML algorithms can be very effective in multiple-input and multiple-output (MIMO) predictions. Therefore, changes in the key electrical characteristics of the device caused by various sources of variability are simultaneously and integrally predicted. This algorithm benchmarks 3-D stochastic TCAD simulation, showing a prediction error rate of less than 1%, as well as a calculation cost reduction of over 80%. In addition, the generality of the algorithm is confirmed by predicting the operating characteristics of the 3-D NAND Flash memory with various structural conditions as the number of layers increases.
17 citations
Cites methods from "Prediction of Process Variation Eff..."
...5 GHz × 16 GHz) processor with 128 GB of RAM, which is same with our previous work [17]....
[...]
TL;DR: In this paper, an efficient and accurate DL approach with device simulation for gate-all-around silicon nanowire metal-oxide-semiconductor field effect transistors (MOSFETs) to predict electrical characteristics of device induced by work function fluctuation was presented.
Abstract: Device simulation has been explored and industrialized for over 40 years; however, it still requires huge computational cost. Therefore, it can be further advanced using deep learning (DL) algorithms. We for the first time report an efficient and accurate DL approach with device simulation for gate-all-around silicon nanowire metal-oxide-semiconductor field-effect transistors (MOSFETs) to predict electrical characteristics of device induced by work function fluctuation. By using three different DL models: artificial neural network (ANN), convolutional neural network (CNN), and long short term memory (LSTM), the variability of threshold voltage, on-current and off-current is predicted with respect to different metal-grain number and location of the low and high values of work function. The comparison is established among the ANN, CNN and the LSTM models and results depict that the CNN model outperforms in terms of the root mean squared error and the percentage error rate. The integration of device simulation with DL models exhibits the characteristic estimation of the explored device efficiently; and, the accurate prediction from the DL models can accelerate the process of device simulation. Notably, the DL approach is able to extract crucial electrical characteristics of a complicated device accurately with 2% error in a cost-effective manner computationally.
16 citations
19 May 2021
TL;DR: In this paper, the authors investigated the possibility of using Machine Learning as a replacement for numerical TCAD device simulation and proposed to utilize machine learning method to establish mapping between the performance parameters and structural parameters of the nanoscale MOSFETs.
Abstract: This review investigates the possibility of using Machine Learning as a replacement for numerical TCAD device simulation. As the chip design is getting complex to incorporate more and more functionality in the devices, many chipmakers started exploring advanced techniques of machine learning to get rid of some big challenges faced by IC industry. In Machine learning, advanced algorithms are utilized to identify patterns in data and to predict about the required information. Machine Learning finds its application in semiconductor fabrication as well as parameter extraction in device modeling. It is also used in prediction of device reliability and its analysis. This work proposes to utilize machine learning method to establish mapping between the performance parameters and structural parameters of the nanoscale MOSFETs. Methods using Machine Learning are fast, highly efficient and computing resource saving over traditional methods.
8 citations
References
More filters
DOI•
20 Sep 2004
1,387 citations
"Prediction of Process Variation Eff..." refers background or methods in this paper
...The algorithm presented in this brief refers to the learned values of the GAA VFET device in the 6/5 nm technology node [17]....
[...]
...The variation of EOT is assumed to be 4% at 3σ [17]....
[...]
05 Jun 2017
TL;DR: In this paper, the authors demonstrate that horizontally stacked gate-all-around (GAA) nanosheet structure is a good candidate for the replacement of FinFET at the 5nm technology node and beyond.
Abstract: In this paper, for the first time we demonstrate that horizontally stacked gate-all-around (GAA) Nanosheet structure is a good candidate for the replacement of FinFET at the 5nm technology node and beyond. It offers increased W eff per active footprint and better performance compared to FinFET, and with a less complex patterning strategy, leveraging EUV lithography. Good electrostatics are reported at L g =12nm and aggressive 44/48nm CPP (Contacted Poly Pitch) ground rules. We demonstrate work function metal (WFM) replacement and multiple threshold voltages, compatible with aggressive sheet to sheet spacing for wide stacked sheets. Stiction of sheets in long-channel devices is eliminated. Dielectric isolation is shown on standard bulk substrate for sub-sheet leakage control. Wrap-around contact (WAC) is evaluated for extrinsic resistance reduction.
547 citations
"Prediction of Process Variation Eff..." refers methods in this paper
...The 3-D TCAD simulation was performed using Synopsys Sentaurus, which was carefully calibrated with the experimental data of 5-nm three stacked nanoplate (NP) FET by referring to [12]....
[...]
01 Dec 2011
TL;DR: In this paper, a comprehensive 3D simulation study of statistical variability and reliability in emerging, scaled FinFETs on SOI substrate with gate-lengths of 20nm, 14nm and 10nm and low channel doping is presented.
Abstract: A comprehensive full-scale 3D simulation study of statistical variability and reliability in emerging, scaled FinFETs on SOI substrate with gate-lengths of 20nm, 14nm and 10nm and low channel doping is presented. Excellent electrostatic integrity and resulting tolerance to low channel doping are perceived as the main FinFET advantages, resulting in a dramatic reduction of statistical variability due to random discrete dopants (RDD). It is found that line edge roughness (LER), metal gate granularity (MGG) and interface trapped charges (ITC) dominate the parameter fluctuations with different distribution features, while RDD may result in relatively rare but significant changes in the device characteristics.
268 citations
"Prediction of Process Variation Eff..." refers background in this paper
...2937786 work function variation (WFV) effects in high-k/metal gate (HK/MG) devices, which is noted as the most important statistical variability sources, should be considered to cope with a critical risk of device performance due to randomly occupied grain-orientation of gate metal [5], [8], [10]....
[...]
TL;DR: In this article, a new source of random threshold-voltage (V_th) fluctuation in emerging metal-gate transistors and proposed a statistical framework to investigate its device and circuit-level implications.
Abstract: This paper highlights and experimentally verifies a new source of random threshold-voltage (V_th) fluctuation in emerging metal-gate transistors and proposes a statistical framework to investigate its device and circuit-level implications. The new source of variability, christened work-function (WF) variation (WFV), is caused by the dependence of metal WF on the orientation of its grains. The experimentally measured data reported in this paper confirm the existence of such variations in both planar and nonplanar high-k metal-gate transistors. As a result of WFV, the WFs of metal gates are statistical distributions instead of deterministic values. In this paper, the key parameters of such WF distributions are analytically modeled by identifying the physical dimensions of the devices and properties of materials used in the fabrication. It is shown that WFV can be modeled by a multinomial distribution where the key parameters of its probability distribution function can be calculated in terms of the aforementioned parameters. The analysis reveals that WFV will contribute a key source of V_th variability in emerging generations of metal-gate devices. Using the proposed framework, one can investigate the implications of WFV for process, device, and circuit design, which are discussed in Part II.
159 citations
"Prediction of Process Variation Eff..." refers background in this paper
...In recent studies, variability analysis methods with low computation overhead were proposed based on compact modeling for analysis of circuit level, as well as device level, but overall this approach still has a high error rate because it is based on many preconditions for simplification of the formulae used [8]–[11]....
[...]
...2937786 work function variation (WFV) effects in high-k/metal gate (HK/MG) devices, which is noted as the most important statistical variability sources, should be considered to cope with a critical risk of device performance due to randomly occupied grain-orientation of gate metal [5], [8], [10]....
[...]
06 Sep 2000
TL;DR: In this paper, the authors compared physically-based full band Monte-Carlo simulations with drift-diffusion simulations for channel lengths from 150 nm to 40 nm and found that the drift diffusion was not robust to surface scattering conditions.
Abstract: Physically-based full band Monte-Carlo simulations are compared with drift-diffusion simulations for channel lengths from 150 nm to 40 nm. Errors in the drift diffusion simulated I/sub ON/, g/sub m/ and channel velocities are quantified through comparison with Monte-Carlo simulations under realistic surface scattering conditions. Suggestions for improving the drift-diffusion results are also discussed.
123 citations
"Prediction of Process Variation Eff..." refers methods in this paper
...For capturing the transport characteristics of nanoscale devices, ballistic transport was considered, and the drift-diffusion (DD) approximation was used for carrier transport [13]....
[...]