Physics-Informed Neural Networks for Device and Circuit Modeling: A Case Study of NeuroSPICE

TL;DR

NeuroSPICE introduces a physics-informed neural network framework for device and circuit simulation, providing a flexible alternative to traditional methods with advantages in design optimization and modeling nonlinear systems.

Contribution

The paper presents NeuroSPICE, a novel PINN-based framework for circuit simulation that models device behaviors using analytical equations and solves DAEs via neural networks.

Findings

NeuroSPICE can simulate highly nonlinear devices like ferroelectric memories.

PINNs offer advantages in surrogate modeling and inverse design problems.

Speed and accuracy during training are comparable to traditional methods, with unique flexibility.

Abstract

We present NeuroSPICE, a physics-informed neural network (PINN) framework for device and circuit simulation. Unlike conventional SPICE, which relies on time-discretized numerical solvers, NeuroSPICE leverages PINNs to solve circuit differential-algebraic equations (DAEs) by minimizing the residual of the equations through backpropagation. It models device and circuit waveforms using analytical equations in time domain with exact temporal derivatives. While PINNs do not outperform SPICE in speed or accuracy during training, they offer unique advantages such as surrogate models for design optimization and inverse problems. NeuroSPICE's flexibility enables the simulation of emerging devices, including highly nonlinear systems such as ferroelectric memories.