We present a new stabilized 3-D control volume finite element method (FEM) for massively parallel simulation of drift-diffusion transport in semiconductor devices. This new solver employs unstructured hexahedral elements, thus leading to a very efficient scheme; both Poisson and current continuity equations are discretized with control volume FEM, and thus, the accuracy and stability are improved. Furthermore, a fully coupled Newton’s method is applied to solve these nonlinear equations, thus remarkably increasing the numerical stability. Then, we apply this new solver to simulate diode, MOSFET, and multifinger MOSFET devices. Numerical results indicate that the hexahedron-based method requires fewer iterative numbers, and its computing speed is 3.44–4.61 times faster than the tetrahedron-based counterpart. Moreover, our fully coupled Newton’s method permits constant iterative numbers, showing strong numerical stability even when the drain voltage is 160 V.
26 January 2023 in IEEE Transactions on Microwave Theory and Techniques
We present a mesh skeleton-enhanced discontinuous Galerkin (DG) method, i.e., hybridizable DG, to solve the 3-D highly nonlinear semiconductor drift-diffusion model. This skeleton-enhanced DG algorithm is a remedy but a significant extension of the classical DG method, where only the degrees of freedom on the skeleton are involved as a globally coupled problem, thus reducing the global dimension from 3-D to 2-D and 2-D to 1-D. Furthermore, high-order nodal basis functions over tetrahedra are easily obtained, and meticulous mesh designs are circumvented for complex semiconductor modeling. Rigorous analytical solutions demonstrate that the convergence rate achieves an optimal order of $\emph{p}+1$ in the $\bm{L}^{\bm{2}}$ -norm. Then, we also apply our algorithm to solve semiconductor devices, including a bipolar transistor and a trigate fin-shaped field-effect transistor. Compared with the conventional finite volume and finite element solvers, the proposed algorithm exhibits superior stability, convergence, and efficiency.
13 July 2021 in IEEE Transactions on Electron Devices
This article presents the construction and analysis of a hybrid numerical method for the discretization of the convection-dominated nonlinear carrier transport process in semiconductor devices. In the construction of the method, the edge streamline upwind (SU) current density model is employed to overcome the nonconvergence brought by the nonlinear property of the current continuity equation, and the mixed finite volume-finite element method (FVFEM) is utilized to effectively extend the edge SU current density model to multidimensional applications. The proposed method is more reliable and flexible than the finite box (FB) method and streamline upwind Petrov–Galerkin (SUPG) method. The performance of several popular SU current density models is compared to find the optimal one. The proposed method is validated by comparing the calculated results with those calculated using commercial software. Based on the streamline upwind-finite volume-finite element method (SU-FVFEM), the finite element method (FEM), and the domain decomposition method (DDM), an in-house parallel computing electrothermal simulator is developed for large-scale semiconductor devices. The performance of the in-house developed parallel simulator is evaluated through simulations of a p-n junction diode and a 3-D bipolar transistor. The results demonstrated that the developed parallel simulator possesses good accuracy, applicability, and scalability.
3 December 2018 in IEEE Transactions on Electron Devices
A comprehensive quantum transport study of ultrathin-body silicon double-gate pMOSFETs with gate length Lg= 14, 10, and 7 nm is performed by in-house developed nonequilibrium Green's function solver employing the six-band k · p Hamiltonian. The effects of channel length, body thickness, as well as confinement and transport crystal orientations are studied systematically. Both I-V and C-V characteristics are analyzed in this paper. Furthermore, the CMOS performance benchmarking results, such as switching delay, energy consumption, and their product, are calculated with the influence of parasitic resistances and capacitances treated in an appropriate way. It is found that the crystal orientation effect on drive current is significant at both long and short channel lengths, while its effect on effective capacitance is more observable at Lg = 7 nm. The drive current at Lg = 14 nm and Lg = 7 nm can be well explained by average ballistic hole velocity and source-to-drain tunneling ratio, respectively. Meanwhile, the effective capacitance is not only determined by the density of states but also affected by the distribution of holes in the confined direction. Both intrinsic and extrinsic performance assessments suggest that (001)/[100] and (110)/[001] are the optimal confinement/transport crystal orientation configurations at Lg= 7 nm, while (110)/[110] and (110)/[111] are the best choices at longer gate lengths. As Lg shrinks, both intrinsic and extrinsic performances will be better at a scaled body thickness.
30 January 2020 in IEEE Transactions on Electron Devices
The resistive random access memory with a graphene electrode (GE) is carefully modeled and simulated by considering its geometrical and physical properties. Since the graphene oxide is generated through the migration of oxygen ions from metal oxide to GE affecting current transport in the memory cell, the model of its electrical conductivity with different oxidation degrees is developed, in which the conductivity is exponentially dependent on the oxidation degree. The simulated I-V characteristics, which agree well with those of the experiment, are obtained by selfconsistently solving the current transport, heat conduction, and oxygen vacancy migration equations. Based on the simulation, oxygen vacancy density, temperature distribution, and resistances of high- and low-resistance states are extracted and compared. A low-temperature increase due to low power consumption indicates its great potential for high-density integration.
21 July 2023 in IEEE Transactions on Components, Packaging and Manufacturing Technology
Bonding wire arrays (BWA) play an important role in multi-chip systems. Especially the inductance, Q value, and loss of BWA are closely related to the performance of the chip systems. However, in industrial applications, full-wave numerical design approaches brings data burden and high computational costs, particularly for large-scale BWA. To alleviate the problem, we propose the multi-fidelity data enabled neural networks for inversely designing BWA, which consist of an inversion network and a residual network. Specifically, we firstly introduce an approximate analytical model of BWA, and perform low-fidelity data collection with the analytical model to train the inversion network. Subsequently, a small set of high-fidelity data is used to train the residual network, which is targeted to learn the residual between the analytical solutions and high-fidelity results. During the testing process, we cascade the two trained networks. Intensive numerical results are used to test the proposed method, where the average testing accuracy reaches to 94.30% with 5000 tests. To demonstrate the effectiveness of the proposed method, the inversely-designed BWA is fabricated and measured experimentally, where the measured results satisfy the performance demands. Furthermore, a power amplifier (PA) using bonding wires as matching inductors is designed, fabricated, and measured by using the proposed method.