Abstract:Moore’s scaling law has survived during more than 50 years because the transistor fabrication recipes have been continuously adapted and technology boosters have been gradually introduced, e.g. strain, high- dielectrics, or 3-D FinFETs. The driving force behind these innovations has always been the intuition of clever researchers who benefited from classical technology computer aided design (TCAD) tools. The latter have been used in the semiconductor industry since the end of the 1970’s, when the first 2-D simulations of CMOS devices became feasible on a supercomputer [1]. Over the last 40 years, transistors have undergone tremendous evolutions, their dimensions being reduced by several orders of magnitude, while the physical models at the core of commercially available device simulators have remained the same: electron transport is still described by classical drift-diffusion (DD) equations, which have been modified to capture parts of the quantum mechanical effects affecting nano-transistors, e.g. geometrical confinement tunnelling leakages, or energy quantization [2]. A new generation of advanced TCAD tools that go beyond the DD equations and rely on atomistic quantum mechanical concepts is needed to reproduce the characteristics of today’s nanostructures and to predict the performance of not-yet-fabricated components. Such tools should combine different modelling approaches to be able to treat various device types, from state-of-the-art nano-transistors to photo-detectors based on two-dimensional materials or resistive switching random access memories. For example, to shed light on the behaviour of valence change memory (VCM) cells, which consist of metal-insulator-metal stacks, molecular dynamics (MD), kinetic Monte Carlo (KMC), density functional theory (DFT), and quantum transport (QT) should be allied, as illustrated in the accompanying figure. By doing so, the growth of nano-filaments through realistic oxides embedded between two metallic electrodes can be accurately simulated and the electrical current flowing through computed with, e.g., the Non-equilibrium Green’s Function (NEGF) formalism [3]. In this presentation the multi-method simulation environment shown the accompanying Figure will be briefly reviewed. The main focus will be set on the discussion of few applications, among them transistors and memory cells. References: [1] S. Selberherr, W. Fichtner, and H.W. Potzl, “Minimos - A program package to facilitate MOS device design and analysis”, Proceedings of NASECODE I, 275 (1979). [2] A. Wettstein, A. Schenk, and W. Fichtner, “Quantum device-simulation with the density-gradient model on unstructured grids”, IEEE Trans. On Elec. Dev. 48, 279 (2001). [3] M. Kaniselvan, M. Luisier, and M. Mladenovic, “An Atomistic Modelling Framework for Valence Change Memory Cells”, Solid-State Electronics 199, 108506 (2023). Figure Caption: Multi-method simulation framework dedicated to the investigation of resistive switching devices, here valence change memory (VCM) cells. First, oxide samples with a low defect concentration are constructed withclassical molecular dynamics (MD) using a melt-and-quench procedure. They are thentransferredto akinetic Monte Carlo (KMC) solver that determine s the distribution of oxygen vacancies (V O ,green spheres) within the oxide layer. All input parameters to KMC (di ff usioncoe ffi cient s and generation/recombination rates) are computed with density functional theory(DFT) , which is also used to calculate the Hamiltonian (H) and Overlap (S) matrices of the created VCM structure. Finally, these quantities are passed to a quantum transport (QT) tool to perform ab initiodevice simulations . Figure 1

Efficient modelling of ionic and electronic interactions by resistive memory-based reservoir graph neural network

An Improved Algorithm for Drift Diffusion Transport and Its Application on Large Scale Parallel Simulation of Resistive Random Access Memory Arrays

A Physics-Oriented Memristor Model with the Coexistence of NDR Effect and RS Memory Behavior for Bio-Inspired Computing

(Invited) Advanced Modeling of Nanoscale Devices

Resistive Memory-based Neural Differential Equation Solver for Score-based Diffusion Model

Echo State Graph Neural Networks with Analogue Random Resistive Memory Arrays

Fully Hardware Memristive Neuromorphic Computing Enabled by the Integration of Trainable Dendritic Neurons and High‐Density RRAM Chip

Atomic-Device Hybrid Modeling of Relaxation Effect in Analog RRAM for Neuromorphic Computing

Digital Biologically Plausible Implementation of Binarized Neural Networks With Differential Hafnium Oxide Resistive Memory Arrays

In-memory and in-sensor reservoir computing with memristive devices

Nano Device Fabrication for In-memory and In-sensor Reservoir Computing

Modeling-Based Design of Memristive Devices for Brain-Inspired Computing

LiNbO$_3$ volatile memristors for reservoir computing

Task-adaptive physical reservoir computing

Functional Materials for Memristor‐Based Reservoir Computing: Dynamics and Applications

CMOS Back-End Compatible Memristors for in Situ Digital and Neuromorphic Computing Applications.

FangTianSim: High-Level Cycle-Accurate Resistive Random-Access Memory-Based Multi-Core Spiking Neural Network Processor Simulator

Modeling of Gate Tunable Synaptic Device for Neuromorphic Applications

Acceleration of Graph Neural Network-Based Prediction Models in Chemistry via Co-Design Optimization on Intelligence Processing Units

In-Memory Computing With Memristor Arrays

Physical reservoirs based on MoS 2 –HZO integrated ferroelectric field-effect transistors for reservoir computing systems