AKAbhishek Kumar

Abhishek Kumar

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Abhishek Kumar has a total of 43 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Complete Charge-Based Artificial Neural Network Augmented Compact Model for Common Multi-Gate FETs

    by Anamika Singh, Mohammad Ehteshamuddin, Abhishek Kumar, Mohit Shukla, Dariskhem Pyngrope, Srishti Parandiyal, Kumar Sheelvardhan, Amit Kumar Behera, Samrat Ray, Abhishek Somani, et al.

    This work presents a novel charge-based artificial neural network (ANN) augmented compact model that bridges physics-based and data-driven modeling for advanced semiconductor devices. Unlike prior ANN or knowledge-based ANN (KBANN) approaches that predict only terminal characteristics, our model embeds analytical charge expressions to ensure physically consistent predictions of current, capacitance, and higher order derivatives, enabling accurate RF, transient, and large-signal analysis. Fully compatible with the industry-standard Berkeley short-channel IGFET model—common multi-gate (BSIM-CMG), the proposed model integrates seamlessly into conventional simulation workflows. It achieves up to 80% reduction in simulation time for a 101-stage ring oscillator (RO) while maintaining accuracy. Furthermore, its runtime scales very slowly with increasing circuit complexity, in contrast to the quadratic scaling of BSIM-CMG, making it a scalable and efficient solution for next-generation electronic design automation (EDA) tools.

  • Open Access

    Artificial synapse based on ULTRARAM memory device for neuromorphic applications

    by Abhishek Kumar, Peter D. Hodgson, Manus Hayne, Avirup Dasgupta

    The memory demands of large-scale deep neural networks (DNNs) require synaptic weight values to be stored and updated in off-chip memory, such as dynamic random-access memory, which reduces energy efficiency and increases training time. Monolithic crossbar or pseudo-crossbar arrays using analog non-volatile memories, which can store and update weights on-chip, present an opportunity to efficiently accelerate DNN training. In this article, we present on-chip training and inference of a neural network using an ULTRARAM memory device-based synaptic array and complementary metal–oxide–semiconductor (CMOS) peripheral circuits. ULTRARAM is a promising emerging memory exhibiting high endurance (>107 P/E cycles), ultrahigh retention (>1000 years), and ultralow switching energy per unit area. A physics-based compact model of ULTRARAM memory device has been proposed to capture the real-time trapping/de-trapping of charges in the floating gate and utilized for the synapse simulations. A circuit-level macro-model is employed to evaluate and benchmark the on-chip learning performance in terms of area, latency, energy, and accuracy of an ULTRARAM synaptic core. In comparison with CMOS-based design, it demonstrates an overall improvement in area and energy by 1.8× and 1.52×, respectively, with 91% of training accuracy.

  • Novel Charge Partitioning Compact Model Including Field-Dependent Mobility Degradation Effect for Nanoscale MOSFETs

    by Mohit Shukla, Abhishek Kumar, Avirup Dasgupta

    Charge partitioning (CP) determines the segmentation of total channel charge among drain and source regions. It is widely utilized in the development of transient and high-frequency advanced compact models for MOSFET analysis. In this work, we have developed a CP compact model that is applicable for short as well as long-channel-length devices. In addition, it also incorporates field-dependent mobility degradation and bias dependency accurately. We have demonstrated accuracy improvement with transient analysis over the industry standard BSIM-CMG compact model, which uses the traditional Ward-Dutton (WD) CP scheme. We have also illustrated the presented model’s utility at high-frequency operation through two-port parameter analysis. Results show that the proposed model shows good agreement with the TCAD data. The proposed model has been implemented in Verilog-A and is compatible with the existing industry standard BSIM-CMG framework. This work provides the first analytical CP compact model formulation taking a second-order effect into account.


  • Open Access

    Physics, modeling, and benchmarking of ULTRARAM: A compound semiconductor-based memory device

    by Abhishek Kumar, Musaibh Farooq Dar, Peter D. Hodgson, Dominic Lane, Peter J. Carrington, Evangelia Delli, Richard Beanland, Shruti Mehrotra, Manus Hayne, Avirup Dasgupta

    ULTRARAM is a promising emerging memory exhibiting high endurance, long retention, and ultra-low switching energy per unit area. This compound semiconductor-based non-volatile memory utilizes triple barrier resonant tunneling (TBRT) through InAs/AlSb heterostructures. In this work, we have proposed a physics-based compact model of ULTRARAM memory device that captures the real-time trapping/de-trapping of charges in the floating gate and used to calculate the device characteristics. In addition, we have performed numerical simulations of the TBRT stack using the Schrödinger equation inside the quantum well with mono-layer (∼0.6 nm) variations to check the impact on memory characteristics of the device. Array-level simulations and benchmarking highlight the promise of this technology.


  • A Compact Model for Perpendicular STT-MRAM Incorporating Free and Pinned Layer Thickness Dependence, Accurate Bias Dependence, and Real-Time 3-D Switching Dynamics

    by Musaibh Farooq Dar, Gagan, Kumar Sheelvardhan, Abhishek Kumar, Tanmoy Pramanik, Sayeef Salahuddin, Chenming Hu, Avirup Dasgupta

    In this article, we present a compact model of spin-transfer torque (STT)-based perpendicular magnetic tunnel junction (pMTJ). We propose a 3-D Landau–Lifshitz–Gilbert (LLG)-based compact model that solves the coupled magnetization dynamics using RC circuits to more precisely capture torques, magnetic energy distributions, and real-time switching dynamics. A robust, flexible, and asymmetric bias-dependent tunneling magnetoresistance (TMR) and resistance model has been developed. In addition, we introduce a new model that includes variations in the TMR ratio dependent on free layer thickness (TFL). Furthermore, we have developed a pinned layer thickness-dependent (TPL) TMR model thereby fully integrating the geometrical variations on TMR and resistive switching behavior for an MTJ device. We have validated our proposed models with multiple experimental data.


  • Metalearning Based Adaptive Compact Modeling Framework for Advanced Transistors Across Technology Nodes

    by Srishti Parandiyal, Abhishek Kumar, M. Ehteshamuddin, Anamika Singh, Kumar Sheelvardhan, Samrat Ray, Abhishek Somani, Sourajeet Roy, Avirup Dasgupta

    This paper presents an adaptive and automated device modeling framework valid across different technology nodes and device structures. A novel metalearning-based surrogate model using Prior Knowledge Input with Difference Artificial Neural Network (PKID ANN) combined with advanced transfer learning (TL) is developed. This approach is validated using various advanced FET devices. In addition to transferring weights and biases from pretrained model, a scaled low-fidelity model is developed for efficient training of different primary target models. Two TL techniques, full and partial knowledge transfer, are compared, with PKID ANN with partial transfer learning (PKID-PTL) showing significant speed-up in all phases of model development. The proposed PKID-PTL technique is a potential candidate for efficient device modeling allowing seamless model automation across technology nodes and devices with the least human intervention.


  • Open Access

    A method to generate additional volatile states using single-domain ferroelectric field-effect transistor

    by Abhishek Kumar, Shruti Mehrotra, Anand Bulusu, Avirup Dasgupta

    In this paper, we have proposed a novel technique to generate the additional volatile states using an ultra-scaled 12 nm technology node fully depleted silicon on insulator Ferroelectric field-effect transistor (FeFET). The proposed method is applicable to all the existing single-bit non-volatile ferroelectric technologies without having any impact on the reliability of the device. Additionally, this method works with single as well as multi-domain ferroelectric films. The approach shows stability in all the states with variations in channel length, non-ferroelectric phase, and ferroelectric parameters. In addition, the states are independent of the memory window of the device and can work with thin ferroelectric films with a hysteresis loop. The proposed approach is voltage-controlled and does not require any physical changes in the single-bit non-volatile FeFET to generate the additional volatile states.


  • Open Access

    Performance Projection of Gate-All-Around (GAA)-Based Negative Capacitance Complementary FET (NC-CFET) Relative to Standard CFET

    by Abhishek Kumar, Shruti Mehrotra, Anand Bulusu, Avirup Dasgupta

    In this article, we have compared the negative capacitance complementary FET (NC-CFET) from a device-circuit perspective with the experimentally demonstrated CFET. NC-CFET could potentially offer additional advantages over conventional CFET, like enhancement in the sub-threshold behavior, reduction in threshold voltage variability, improvement in noise margin, low supply voltage, and high-speed operation, etc. The comparative study shows the improvement in the logic devices, as well as SRAM memory cell with the use of negative capacitance concept. Detailed circuit-level simulations have been used to analyze the performance gains with NC-CFET. Further, we have demonstrated multi layer NC-CFET with vertically stacked N/P channels for ON-current matching, symmetric inverter switching, and improved static noise margin (SNM) in SRAM.


  • Open Access

    Study of Enhanced Ferromagnetic Behavior of Mn-Doped 2D Janus Cr₂S₂I₂ Monolayer

    by Neha Mishra, Abhishek Kumar, Avirup Dasgupta, Sourajeet Roy

    In this work, we report the first-principle based study of the electronic properties of Mn-doped Cr2S2I2 (CrMnS2I2) monolayer for possible applications in various spin devices. The electric field (0.5 eV/Å) and compressive strain (5%) are applied externally on the Mn-doped Cr2S2I2 monolayer to study the impact on its magnetic properties. The long-range ferromagnetism is established by the exchange coefficient (J1). The exchange coupling is enhanced from 0.131 to 7.81 meV with the applied external electric field. The ferromagnetic behavior is confirmed in the pristine CrMnS2I2 monolayer and by two external methods. The FM monolayer shows perpendicular magnetic anisotropy (PMA) of 115.6 meV and high Curie Temperature of 476K (Tc). Both PMA and Tc are increased to 792 meV and 497K with applied electric field and compressive strain. Since the studies material shows excellent potential for use in spintronic application, we have estimated the spin barrier height (3.97) and filtration efficiency (64%) for CrMnS2I2 monolayer as spin filter as well as the tunnelling magnetoresistance (TMR) effect as FM layer in magnetic tunnel junctions (MTJ) respectively. The TMR is observed to increase from 78% to 359.52% by varying thickness of non-magnetic MoSe2 spacer