Bridging Simulation and Experiment in Nanoscience with AI
V. Fonseca Hernandes (TU Delft - QN/Greplová Lab)
R. Hanson – Promotor (TU Delft - QID/Hanson Lab, TU Delft - QN/Hanson Lab)
E. Greplová – Copromotor (TU Delft - QCD/Greplova Lab, TU Delft - QN/Greplová Lab)
More Info
expand_more
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Abstract
This thesis explores how artificial intelligence (AI) can be used to bridge the gap between simulation and experiment in nanoscience. As both theoretical modeling and experimental techniques in nanoscience become increasingly sophisticated, AI is emerging as a powerful tool to tackle challenges such as tuning experiments, accelerating simulations, generating synthetic data, and automating data analysis. This work presents applications of AI in three main domains: neuroscience, quantum computing, and condensed matter physics.
In neuroscience, we address the problem of efficiently simulating neuronal activity data by implementing a quantum machine learning model that uses a reduced number of trainable parameters. On the experimental side, we develop a computational package that automates the analysis of micro-electrode array data using a neural network trained to replicate human expert detection of burst patterns from spiking activity.
In spin-based quantum computing, we develop a computational package that simulates charge stability diagrams (CSDs) based on device characteristics, enabling the efficient creation of synthetic datasets. We also demonstrate how machine learning models can be used to filter high-quality CSDs for building experimental datasets, and we present the first implementation of diffusion models to complete partially measured CSDs, an approach that can be integrated into measurement routines to accelerate the process.
In condensed matter physics, we leverage neural quantum states to detect phase transitions by analyzing the evolution of neural network weights, without the need to calculate order parameters, and discuss potential directions for combining this technique with neural quantum states trained on experimental data.
These studies show that AI can accelerate simulations and data analysis, while also supporting the design and interpretation of experiments, highlighting its growing and essential role in the future of nanoscience.
help_outline