Yuhang Zhang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Yuhang Zhang has a total of 103 co-authors. Here are the top co-authors by number of shared publications:
- Yongfu Li
Shanghai Jiao Tong University
22 shared publications - Yabin Sun
East China Normal University
11 shared publications - Bingyi Ye
School of Integrated Circuit, East China Normal University, Shanghai, China
10 shared publications - Xiaojin Li
School of Integrated Circuit, East China Normal University, Shanghai, China
10 shared publications - Yang Shen
East China Normal University
10 shared publications - Jian Zhao
Shanghai Jiao Tong University
8 shared publications
Latest Publications
- 5 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Electron Devices
- Vol. 73 (6) , 3888-3895
SMS-EMOA-Driven Multiobjective Design-Technology Co-Optimization for 5-nm Gate-All-Around Transistors
Due to the complexity of physical effects and second-order effects, optimizing the electrical properties of sub-5-nm devices is exceptionally challenging. This article presents an S-Metric Selection Evolutionary Multiobjective Optimization Algorithm (SMS-EMOA)-driven Design-Technology Co-Optimization (DTCO) framework. By uniquely integrating a lightweight performance prediction surrogate model, it efficiently identifies Pareto fronts under strict manufacturing constraints. This achieves a $62\times $ speedup and reduces peak memory usage from 12GB to ~200 MB compared to direct technology computer-aided design (TCAD). Furthermore, artificial neural network (ANN)-based compact models can be automatically generated from $I$ – $V$ and $C$ – $V$ data for fast circuit-level verification. Considering back-end interconnect parasitics, DTCO is performed on a layout-aware 97-stage NAND2 ring oscillator. Compared with random baselines, the optimized device achieves up to 34% lower power and 65% higher performance, while delivering a > 25% improvement in SRAM Static Noise Margin (SNM) for robust stability. This framework establishes a tight, scalable connection across process, device, layout, and circuit co-optimization.
- 5 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Electron Devices
- Vol. 73 (6) , 3888-3895
SMS-EMOA-Driven Multiobjective Design-Technology Co-Optimization for 5-nm Gate-All-Around Transistors
Due to the complexity of physical effects and second-order effects, optimizing the electrical properties of sub-5-nm devices is exceptionally challenging. This article presents an S-Metric Selection Evolutionary Multiobjective Optimization Algorithm (SMS-EMOA)-driven Design-Technology Co-Optimization (DTCO) framework. By uniquely integrating a lightweight performance prediction surrogate model, it efficiently identifies Pareto fronts under strict manufacturing constraints. This achieves a $62\times $ speedup and reduces peak memory usage from 12GB to ~200 MB compared to direct technology computer-aided design (TCAD). Furthermore, artificial neural network (ANN)-based compact models can be automatically generated from $I$ – $V$ and $C$ – $V$ data for fast circuit-level verification. Considering back-end interconnect parasitics, DTCO is performed on a layout-aware 97-stage NAND2 ring oscillator. Compared with random baselines, the optimized device achieves up to 34% lower power and 65% higher performance, while delivering a > 25% improvement in SRAM Static Noise Margin (SNM) for robust stability. This framework establishes a tight, scalable connection across process, device, layout, and circuit co-optimization.
- Open Access
- 10 April 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Design Automation of Electronic Systems
- Vol. 31 (5) , 1-24
ATMAD: Agile Transistor Compact Modeling with Parameter Extraction Based on Automatic Differentiation
Compact models of transistors are essential for simulating and optimizing circuits with the use of SPICE simulation tool. Parameter extraction, which is calibrating these models, is essential to ensure their alignment with measured or simulated data. However, conventional parameter extraction methods are generally iterative and experience-dependent, requiring significant time and effort from modeling engineers. Moreover, as semiconductor devices and compact models become increasingly advanced, the need for a tailored extraction process for each model has become increasingly inefficient. To address the above challenges, this work proposes an agile transistor compact modeling framework, ATMAD. The proposed framework takes a compact model file and a set of electrical characteristic data as inputs, producing a calibrated model with minimal human intervention. ATMAD automatically retrieves the equations in the compact model and converts them into computational flow graphs, thus supporting different compact models with a generalized process. A graph unlooping technique is proposed to support automatic differentiation for compact models with implicit functions (e.g., series resistance and surface potential solving). Based on the computational flow graph, ATMAD adopts automatic differentiation technique to achieve automatic and parallel optimization of model parameters. The proposed ATMAD framework is validated on commonly-used compact models in academia and industry, showing its effectiveness for compact modeling for both I-V and C-V characteristics.
- 10 April 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Design Automation of Electronic Systems
- Vol. 31 (5) , 1-24
Open AccessATMAD: Agile Transistor Compact Modeling with Parameter Extraction Based on Automatic Differentiation
Compact models of transistors are essential for simulating and optimizing circuits with the use of SPICE simulation tool. Parameter extraction, which is calibrating these models, is essential to ensure their alignment with measured or simulated data. However, conventional parameter extraction methods are generally iterative and experience-dependent, requiring significant time and effort from modeling engineers. Moreover, as semiconductor devices and compact models become increasingly advanced, the need for a tailored extraction process for each model has become increasingly inefficient. To address the above challenges, this work proposes an agile transistor compact modeling framework, ATMAD. The proposed framework takes a compact model file and a set of electrical characteristic data as inputs, producing a calibrated model with minimal human intervention. ATMAD automatically retrieves the equations in the compact model and converts them into computational flow graphs, thus supporting different compact models with a generalized process. A graph unlooping technique is proposed to support automatic differentiation for compact models with implicit functions (e.g., series resistance and surface potential solving). Based on the computational flow graph, ATMAD adopts automatic differentiation technique to achieve automatic and parallel optimization of model parameters. The proposed ATMAD framework is validated on commonly-used compact models in academia and industry, showing its effectiveness for compact modeling for both I-V and C-V characteristics.
- 6 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Electron Devices
- Vol. 73 (4) , 2343-2351
Novel Reconfigurable Feedback Field-Effect Transistor With Single Gate for Unified RAM (URAM) and In-Memory-Computing Applications
In this study, we have reported a single-gated reconfigurable feedback field-effect transistor (SGR-FBFET), integrating nonvolatile memory (NVM) and 1T-dynamic random access memory (DRAM) into a single FBFET device. Unlike previously reported reconfigurable FBFET, the proposed device enables reconfiguration with a single gate, thereby reducing the complexity of metal interconnection. Compared to existing unified random access memory (URAM), it requires lower supply voltage through its positive feedback mechanism. For NVM mode, our FBFET exhibits a faster read speed of 0.4 ns compared to other NVM devices, with an outstanding read current ratio ( $\gt\!10^{{6}}\text {)}$ . As 1T-DRAM, the sensing margin and retention time (RT) are about 22.0 $\mu $ A/um and 100 s, respectively, which are higher than previously reported URAMs. It also exhibits nondestructive read characteristics. Moreover, as a core element for reconfigurable logic-in-memory (R-LIM) cell, the reconfigurable devices show an on-current ratio of 1.06 for the n- and p-channel modes. The R-LIM cell can perform 8 Boolean logic operations and store output logic for >1 s without supply voltages. Our device exhibits great potential for 3-D memory integrations and future R-LIM applications.
- 6 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Electron Devices
- Vol. 73 (4) , 2343-2351
Novel Reconfigurable Feedback Field-Effect Transistor With Single Gate for Unified RAM (URAM) and In-Memory-Computing Applications
In this study, we have reported a single-gated reconfigurable feedback field-effect transistor (SGR-FBFET), integrating nonvolatile memory (NVM) and 1T-dynamic random access memory (DRAM) into a single FBFET device. Unlike previously reported reconfigurable FBFET, the proposed device enables reconfiguration with a single gate, thereby reducing the complexity of metal interconnection. Compared to existing unified random access memory (URAM), it requires lower supply voltage through its positive feedback mechanism. For NVM mode, our FBFET exhibits a faster read speed of 0.4 ns compared to other NVM devices, with an outstanding read current ratio ( $\gt\!10^{{6}}\text {)}$ . As 1T-DRAM, the sensing margin and retention time (RT) are about 22.0 $\mu $ A/um and 100 s, respectively, which are higher than previously reported URAMs. It also exhibits nondestructive read characteristics. Moreover, as a core element for reconfigurable logic-in-memory (R-LIM) cell, the reconfigurable devices show an on-current ratio of 1.06 for the n- and p-channel modes. The R-LIM cell can perform 8 Boolean logic operations and store output logic for >1 s without supply voltages. Our device exhibits great potential for 3-D memory integrations and future R-LIM applications.
- 1 March 2026
- Published by IOP Publishing in Semiconductor Science and Technology
- Vol. 41 (3) , 035005
Novel FOM-enhanced sided-shield gate trench MOSFET with super-junction structure
In this paper, a novel sided-shield gate trench MOSFET with super-junction structure (SJ-SSGT) is proposed for the first time. By introducing both a SJ drift region and a SSGT together, the specific on-state resistance R ds(on) is reduced by about 21%, while the breakdown voltage is increased by about 57% compared with a conventional SGT MOSFET. In that case, the coupling of the SJ and the SSGT will lead to an approximately twofold improvement in the figure of merit of the device. The breakdown and conduction mechanism is investigated in detail by 3D TCAD simulations. Furthermore, the influences of drift-region doping concentration, SJ doping concentration, sided-shield gate length, and SJ length are systematically studied to optimize the proposed SJ-SSGT.
- 1 March 2026
- Published by IOP Publishing in Semiconductor Science and Technology
- Vol. 41 (3) , 035005
Novel FOM-enhanced sided-shield gate trench MOSFET with super-junction structure
In this paper, a novel sided-shield gate trench MOSFET with super-junction structure (SJ-SSGT) is proposed for the first time. By introducing both a SJ drift region and a SSGT together, the specific on-state resistance R ds(on) is reduced by about 21%, while the breakdown voltage is increased by about 57% compared with a conventional SGT MOSFET. In that case, the coupling of the SJ and the SSGT will lead to an approximately twofold improvement in the figure of merit of the device. The breakdown and conduction mechanism is investigated in detail by 3D TCAD simulations. Furthermore, the influences of drift-region doping concentration, SJ doping concentration, sided-shield gate length, and SJ length are systematically studied to optimize the proposed SJ-SSGT.
- 13 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Electron Devices
- Vol. 73 (3) , 1372-1380
Assessment of Capacitorless DRAM Based on Stacked Nanosheet Feedback Field-Effect Transistors With Tree-Type Channel
In this study, a novel vertically stacked nanosheet (NS) feedback field-effect transistor with tree-type channel (Tree-FBFET) is proposed for one-transistor dynamic random access memory (1T-DRAM) application. Compared with the conventional NS channel structure, an additional SiGe interbridge (IB) channel is vertically integrated in Tree-FBFET. The enhanced performance of Tree-FBFET stems from additional SiGe storage region which is recognized as the SiGe channel. The Tree-FBFET based 1T-DRAM shows an improvement of 2.8 times in sensing margin (SM) (from 126 to $351~\boldsymbol {\mu }$ A/ $\boldsymbol {\mu }$ m) and extends retention time (RT) from 0.6 to 32 s compared to 1T-DRAM based on conventional vertically stacked NS-FBFET. Moreover, the energy consumption is 35 fJ/bit for the write “1” operation and 0 fJ/bit for the hold operation. Geometric parameters of IB and NS, along with Ge mole fraction, are evaluated with respect to SM and RT. Furthermore, reliable $3\times 3$ array operation of the proposed 1T-DRAM, free from disturbance of write operation, is also verified.
- 13 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Electron Devices
- Vol. 73 (3) , 1372-1380
Assessment of Capacitorless DRAM Based on Stacked Nanosheet Feedback Field-Effect Transistors With Tree-Type Channel
In this study, a novel vertically stacked nanosheet (NS) feedback field-effect transistor with tree-type channel (Tree-FBFET) is proposed for one-transistor dynamic random access memory (1T-DRAM) application. Compared with the conventional NS channel structure, an additional SiGe interbridge (IB) channel is vertically integrated in Tree-FBFET. The enhanced performance of Tree-FBFET stems from additional SiGe storage region which is recognized as the SiGe channel. The Tree-FBFET based 1T-DRAM shows an improvement of 2.8 times in sensing margin (SM) (from 126 to $351~\boldsymbol {\mu }$ A/ $\boldsymbol {\mu }$ m) and extends retention time (RT) from 0.6 to 32 s compared to 1T-DRAM based on conventional vertically stacked NS-FBFET. Moreover, the energy consumption is 35 fJ/bit for the write “1” operation and 0 fJ/bit for the hold operation. Geometric parameters of IB and NS, along with Ge mole fraction, are evaluated with respect to SM and RT. Furthermore, reliable $3\times 3$ array operation of the proposed 1T-DRAM, free from disturbance of write operation, is also verified.
- Open Access
- 12 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Data Descriptions
- Vol. 3, 132-139
Descriptor: Emerging Semiconductor Device Electrical Dataset (ESDED)
This paper presents the Emerging Semiconductor Device Electrical Dataset (ESDED), a comprehensive collection of I-V characteristics data extracted from advanced CMOS devices. The ESDED dataset covers a range of device types, bias voltages, physical dimensions, and multiple process technologies. In this work, we meticulously document the design methodology used to create the dataset and provide a comprehensive overview of its format and accessibility. ESDED is designed to function as a standardized benchmark for evaluating and comparing emerging semiconductor models. It provides semiconductor researchers and model developers with a rich and extensive dataset, thereby facilitating the acquisition of sufficient data for compact models.
- 12 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Data Descriptions
- Vol. 3, 132-139
Open AccessDescriptor: Emerging Semiconductor Device Electrical Dataset (ESDED)
This paper presents the Emerging Semiconductor Device Electrical Dataset (ESDED), a comprehensive collection of I-V characteristics data extracted from advanced CMOS devices. The ESDED dataset covers a range of device types, bias voltages, physical dimensions, and multiple process technologies. In this work, we meticulously document the design methodology used to create the dataset and provide a comprehensive overview of its format and accessibility. ESDED is designed to function as a standardized benchmark for evaluating and comparing emerging semiconductor models. It provides semiconductor researchers and model developers with a rich and extensive dataset, thereby facilitating the acquisition of sufficient data for compact models.
- 13 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Electron Devices
- Vol. 73 (3) , 1639-1646
Novel Triple-Gate Nanosheet RFET With Embedded SiGe Surrounding Channel for Enhanced On-State Current
In this study, a novel triple-gate nanosheet reconfigurable field-effect transistor (RFET) featuring an embedded SiGe layer conformally surrounding the silicon channel (ESCRFET) beneath the control gate (CG) was proposed. The bandgap offset between SiGe and Si is utilized to modulate the energy band profile and optimize carrier injection in both n- and p-type operation modes. Compared to the traditional triple-gate RFET (TG-RFET), the proposed ESCRFET achieves 50.2% and 82.2% enhancement in Ion for n- and p-type devices, respectively, while maintaining comparable subthreshold swing characteristics. A neural-network-based compact model is developed to accurately reproduce the electrical characteristics, with an average error of less than 3%, enabling efficient SPICE-level circuit simulation. Circuit-level simulations of nor, AOI, and 1-bit full-adder circuits demonstrate that ESCRFET significantly reduces propagation delay, achieving a 40%–60% faster response compared with the TG-RFET, while also lowering the total transistor count from 28 to 18, corresponding to a 36% reduction. These results confirm the superior electrical performance and scalability of the ESCRFET for future reconfigurable logic and arithmetic circuit applications.
- 13 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Electron Devices
- Vol. 73 (3) , 1639-1646
Novel Triple-Gate Nanosheet RFET With Embedded SiGe Surrounding Channel for Enhanced On-State Current
In this study, a novel triple-gate nanosheet reconfigurable field-effect transistor (RFET) featuring an embedded SiGe layer conformally surrounding the silicon channel (ESCRFET) beneath the control gate (CG) was proposed. The bandgap offset between SiGe and Si is utilized to modulate the energy band profile and optimize carrier injection in both n- and p-type operation modes. Compared to the traditional triple-gate RFET (TG-RFET), the proposed ESCRFET achieves 50.2% and 82.2% enhancement in Ion for n- and p-type devices, respectively, while maintaining comparable subthreshold swing characteristics. A neural-network-based compact model is developed to accurately reproduce the electrical characteristics, with an average error of less than 3%, enabling efficient SPICE-level circuit simulation. Circuit-level simulations of nor, AOI, and 1-bit full-adder circuits demonstrate that ESCRFET significantly reduces propagation delay, achieving a 40%–60% faster response compared with the TG-RFET, while also lowering the total transistor count from 28 to 18, corresponding to a 36% reduction. These results confirm the superior electrical performance and scalability of the ESCRFET for future reconfigurable logic and arithmetic circuit applications.
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Nanotechnology
- Vol. 25, 409-416
Electrical Isolation Enabling Stateful Logic-In-Memory Operation in an Improved SC-RFET
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Nanotechnology
- Vol. 25, 409-416
Electrical Isolation Enabling Stateful Logic-In-Memory Operation in an Improved SC-RFET
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
- Vol. PP (02780070) , 1
A Physics-Constrained Neural Network Method for Compact Modeling of Semiconductor Devices
This work proposes a physics-constrained neural network (PCNN) method for automated and physically consistent compact modeling of semiconductor devices. Traditional ANN-based approaches often exhibit limited enforcement of physical monotonicity and symmetry, suffer from poor generalization, and require complex manual data preprocessing. The proposed method addresses these challenges through two core innovations: a self-adaptive data preprocessing strategy that eliminates manual intervention, and a weight-generation neural architecture that embeds physics-inspired constraints to ensure monotonicity and symmetry. Extensive device- and circuit-level evaluations demonstrate the effectiveness of the method. On average, a 3.2× reduction in mean relative error (MRE) is achieved in modeling I-V and C-V characteristics across the devices evaluated in this work compared with ANN-based models. In circuit-level simulations, the method reduces the power-delay product (PDP) error from 50% to 7% in a 17-stage ring oscillator and lowers the gain error from 5.7% to 0.4% in a common-source amplifier. Additionally, transient simulation time is reduced by up to 70% compared to ANN-based models. These results highlight that the proposed method not only improves physical consistency and modeling accuracy but also enables scalable and efficient SPICE-level simulation, paving the way for practical AI/ML-assisted compact model development.
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
- Vol. PP (02780070) , 1
A Physics-Constrained Neural Network Method for Compact Modeling of Semiconductor Devices
This work proposes a physics-constrained neural network (PCNN) method for automated and physically consistent compact modeling of semiconductor devices. Traditional ANN-based approaches often exhibit limited enforcement of physical monotonicity and symmetry, suffer from poor generalization, and require complex manual data preprocessing. The proposed method addresses these challenges through two core innovations: a self-adaptive data preprocessing strategy that eliminates manual intervention, and a weight-generation neural architecture that embeds physics-inspired constraints to ensure monotonicity and symmetry. Extensive device- and circuit-level evaluations demonstrate the effectiveness of the method. On average, a 3.2× reduction in mean relative error (MRE) is achieved in modeling I-V and C-V characteristics across the devices evaluated in this work compared with ANN-based models. In circuit-level simulations, the method reduces the power-delay product (PDP) error from 50% to 7% in a 17-stage ring oscillator and lowers the gain error from 5.7% to 0.4% in a common-source amplifier. Additionally, transient simulation time is reduced by up to 70% compared to ANN-based models. These results highlight that the proposed method not only improves physical consistency and modeling accuracy but also enables scalable and efficient SPICE-level simulation, paving the way for practical AI/ML-assisted compact model development.