Multiscale Simulation of 2D Electronic Devices—From Contact Design to Machine-Learning-Guided Device Optimization

In nanoscale logic and memory devices based on new materials, atomistic scale features of materials and interfaces can play an important role in determining device characteristics and performance. Atomistic simulations of a practical logic or memory device, however, are computationally expensive, which hinders efficient device design. A multiscale approach to device simulation can address the above challenge and achieve physical accuracy and computational efficiency at the same time in device simulations. We have developed a multiscale simulation approach and machine-learning-guided design optimization methods to investigate the metal contacts to 2D materials, 2D-material-based nanoscale transistors, and ferroelectric tunneling junction (FTJ) memory devices, as described below.<br/>To understand and explore low contact resistance to 2D materials, a multiscale simulation approach is developed to simulate the contact transport properties between a semimetal to a monolayer two-dimensional (2D) transition metal dichalcogenide (TMDC) semiconductor [1]. The results elucidate the mechanisms for low contact resistance between semimetal and TMDC semiconductor contacts from a quantum transport perspective. The simulation results compare favorably with recent experiments. Furthermore, the results show that the contact resistance of a Bismuth-MoS₂ contact can be further reduced by engineering the dielectric environment and doping the TMDC material to . The quantum transport simulation indicates the possibility to achieve an ultrashort contact transfer length of ~1nm, which can allow aggressive scaling of the contact size.<br/>At the device level, scaling of transistors near the physical limits imposes significant technological and design challenges. Identifying and understanding optimum designs and trade-offs between multiple design targets, including speed, power or energy, and variability, is necessary. By integrating machine-learning-based optimization methods into device simulation, we developed a multiobjective device design method that can automatically and efficiently identify the most promising device designs that can simultaneously satisfy multiple design objectives [2] for 2D-material-based field-effect transistors (FETs) near the scaling limit. The multiobjective design framework performs gradient-free efficient global optimization based on an active learning method. Optimum designs with the trade-off between transistor speed, power, and variability are identified automatically for 2D FETs by applying the multiobjective design framework. It is shown that the International Roadmap of Devices and Systems (IRDS) target of 2025 and 2028 technology nodes can be met by the identified designs of 2D FETs [2].<br/>The multiscale simulation approach is further applied to simulate 2D-material-based memory devices. Atomically thin van der Waals (vdW) heterojunctions are investigated for ferroelectric tunnel junction (FTJ) device application by combining multiscale simulations from atomistic <i>ab initio</i> to quantum transport device simulations with experimental studies [3]. The simulation reveals that low quantum capacitance of graphene, weak electronic hybridization of vdW bonds, and high interface quality free of dangling bonds can lead to extremely large vdW interface barrier height modulation at the graphene-2D ferroelectric (FE) interface [3].<br/><br/><b>References</b>:<br/>[1] T. Wu and J. Guo, “Multiscale simulation of semimetal contact to transition metal dichalcogenide semiconductor,” submitted, (2022).<br/>[2] T. Wu and J. Guo, “Multiobjective design of 2-D-material-based field-effect transistors with machine learning methods,” <i>IEEE Trans. Electron Dev</i>., (2021), doi: 10.1109/TED.2021.3085701<br/>[3] N. Yang, H. Chen, J. Wu, T. Wu, J. Cao, X. Ling, H. Wang, J. Guo, “Multiscale simulation of ferroelectric tunnel junction memory enabled by van der Waals heterojunction: comparison to experiment and performance projection,” Int. Electron Dev. Meeting (IEDM), (2020).