Laura H. Lewis

College of Engineering
Northeastern University
Boston, Massachusetts USA

Magnetism.... more than magnets
A discussion of interdisciplinarity & opportunities

The ability of magnetism to permit wireless transfer of energy and/or force without wires has built our civilization. While perhaps the oldest of the physical sciences, magnetism is also one of least popularly understood, a conundrum that contributes to “stove piping”, or lack of interdisciplinarity, in new technologies. Beyond conventional applications of magnetism in transformers, motors, generators and data storage, magnetism can enhance sensing, regulate biological function and aid in understanding the conditions that led to the formation of our solar system. In this presentation, examples of how magnetism has recently impacted these areas will be discussed, and opportunities for computation to deliver deeper mechanistic understanding will be highlighted. Finally, the concept of the magnetic domain, which controls the response of a magnetism system, will be explained and illustrated with a hands-on demonstration.

Brief Biography: Laura H. Lewis is Distinguished University and Cabot Professor of Chemical Engineering, Professor of Mechanical and Industrial Engineering and Affiliated Professor of Physics at Northeastern University in Boston, MA. Prior to her Northeastern University position, she was a research group leader and Associate Department Chair in the Nanoscience Department of Brookhaven National Laboratory (BNL). Concurrently, she was the Deputy Director of the BNL Center for Functional Nanomaterials, a DOE national user facility to provide researchers with state-of-the-art capabilities to fabricate and study nanoscale materials.

Laura’s research focuses on investigating the materials factors at the atomic level that provide functionality to magnetic and electronic materials, with particular expertise in advanced permanent magnets. She has participated on a number of advisory panels, including as a member of the Scientific Advisory Board of the Critical Materials Institute (a DOE Energy Innovation Hub). She is a Delegate of the U.S. Technical Advisory Groups to develop supply chain and sustainability standards to ISO TC298 (Rare Earths) and ISO TC333 (Lithium), on behalf of the American National Standards Institute (ANSI).

Laura, a Fellow of the IEEE, was Conference Editor of the IEEE Transactions on Magnetics (2008 – 2018) and was Chair of the IEEE Magnetics Society Technical Committee (2017-2019). She is also a Fellow of the American Physical Society, a Fulbright Fellow, a member of the Materials Research Society, the American Chemical Society and the American Society for Engineering Education and is an elected member of JEMS-EMA (The European Magnetism Association).

Control of falling liquid films with restricted observations

We propose a method to stabilise an unstable solution to equations describing the interface of thin liquid films falling under gravity with a finite number of actuators and restricted observations. As for many complex systems, full observation of the system state is challenging in physical settings, so methods able to take this into account are important. The Navier-Stokes equations modelling this interfacial flow are a complex, highly nonlinear set of PDEs, so standard control theoretical results are not applicable. Instead, we chain together a hierarchy of increasingly idealised approximations, developing a control strategy for the simplified model which is shown to be successfully applicable to direct numerical simulations of the full system.

Investigation of laser plasma instabilities driven by 1314 nm laser pulses at shock ignition conditions using multi-dimensional kinetic simulations

Session 2

Using Orbital-Constrained DFT to Simulate Surface Spectroscopy with Relativistic Corrections

TBC

Understanding Domain Reconstruction of Twisted Bilayer and Heterobilayer Transition Metal Dichalcogenides through Machine Learned Interatomic Potentials

In the study of twisted bilayer 2D materials, a detailed picture of the relaxations and layer-corrugations that occur due to interlayer interaction is crucial to predicting how their electronic and optical properties depend on twist angle and the resulting large-scale Moiré pattern. As the relative twist angle between the layers approaches 0º, referred to as parallel (P) stacking, or 60º, referred to as antiparallel (AP) stacking, reconstructions occur to maximize the area of low-energy stacking domains, with a lattice of solitons of high-energy stacking connected by domain walls (DWs). We show that Machine Learned Interatomic Potentials (MLIPs) can provide the combination of accuracy and scaling required to obtain atomistic insight into this behaviour. In contrast to empirical potential methods, MLIPs based on higher-order equivariant message passing, as implemented in MACE, can provide very precise energetics of stacking, strain, shear, and varying interlayer distances to exactly reproduce vdW-corrected DFT for systems dramatically larger than can be treated with ab initio methods. We predict, explain, and quantify the domain reconstruction patterns for all like-chalcogen combinations of the Transition Metal Dichalcogenides MoS2, MoSe2, WS2, and WSe2 down to twist angles approaching 1º. We demonstrate effects including DW-bending in AP systems, and the “twirling” that occurs around the nodes in heterobilayers.

Session 3

MemPrO: Membrane Protein Orientation in Lipid Bilayers

Membrane proteins play an important role in many vital systems of a cell, such as transport of ions and raw materials, communication between adjacent cells, and antibiotic resistant behaviours. Correctly orienting membrane proteins is almost always the first step in the molecular simulation and analysis of membrane-protein systems. The method presented aims to orient a wide range of proteins that interact with the membrane. Many such programs already exist such as OPM and MemEmbed, however these do not work for some situations such as multi-bilayer systems or peripheral membrane proteins. MemPrO also contains tools for further analysis of membrane-protein systems without the need for simulations to be run. Currently membrane deformation and localisation of negatively charged lipids can be predicted. The core method consists of a minimisation in a mean field of potential constructed using Martini3 CG parameters.

Session 4

Dislocation dynamics in Ni-based superalloys from atomistic simulations

Ni-based superalloys exhibit extraordinary strength at high temperatures, which results primarily from the nanoscale precipitates in their microstructures hindering dislocation motion. In our work, we study precipitation strengthening in Ni-based superalloys using Molecular Dynamics (MD) simulations with classical effective potentials. The motion of edge dislocations in pure face-centred cubic (FCC) Ni was observed from MD simulations and Differential Evolution Monte Carlo (DE-MC) was used to fit the parameters of an equation of motion to the extracted dislocation trajectories. Using DE-MC as a sampling approach produces parameter distributions and successfully captures the correlations between them. The parameter distributions determined from DE-MC were then used to quantify the uncertainties in the model predictions, namely the dislocation positions and velocities. The equation of motion considered was also extended to account for the presence of multiple dislocations and their interactions, which in addition to interactions with precipitates, are key to obtaining a more realistic representation of Ni-based superalloys. This work serves as a first step towards developing a more comprehensive surrogate model to describe the deformation behaviour of Ni-based superalloys with a focus on propagating and quantifying uncertainties, in addition to exploring ways to apply atomistic-scale insights to inform larger length-scale simulations of dislocations.

Cohort 4

Session 1

Chantal Baer

TBC

Yu Lei

Towards Atomic Resolution of Cryogenic Ptychography Single-Particle Analysis (Cryo-EPty SPA)

Cryo-EM with single particle analysis (SPA) facilitates the visualization of 3D macromolecular structures at an atomic scale. However, the electron sensitivity inherent in biological samples leads to low contrast in EM images. While imaging at a large defocus can improve contrast, it also limits the information transfer at high spatial frequencies with the sample. Ptychography diffractive imaging, a technique capable of reconstructing phase information from diffraction patterns using an iterative algorithm known as ePIE, holds great promise for achieving super-resolution, high-contrast, low-dose, and 3D imaging of biological samples in vitreous ice at low doses. Moreover, ptychography utilizes the entire diffraction patterns, making it particularly efficient in dose usage, especially when using direct electron detector data with a high signal-to-noise ratio at a low electron dose. Using ptychography we have successfully reconstructed the 2D phase images of rotavirus at cryogenic temperatures with a dose of 5 e/Ã…2 and have further demonstrated the visualization of 3D structures at nm resolution by integrating SPA. To show the potential that cryogenic ptychography (cryo-EPty) and SPA to achieve atomic level resolution, here we will employ apoferritin as a benchmark sample, and implement different convergence semi-angles (CSA) to reconstruct its structure. Subsequent SPA 3D density maps have shown resolutions of 1.1 nm and 0.8 nm, respectively. Our findings suggest that the promising capabilities for cryo-EPty combination with SPA pave the way for high-resolution 3D reconstructions of biological samples, potentially reaching atomic resolution.

Hubert Naguszewski

Investigating the efficacy of CNNs and GNNs at predicting the committor for the 2D Ising model

The presentation shall explain how convolutional neural networks (CNNs) and graph neural networks (GNNs) perform at predicting the committor, the probability that a microstate will evolve to a state B before returning to a state A. Having a tool to rapidly predict the committor for a given system would allow for calculation of nucleation rates without high computation costs which would be useful for systems where generating data is expensive. The Ising model is being used because it is possible to quickly compute large quantities of accurate training data. Studying the neural network performance on such a simple system should allow for the lessons learned to be transferable to more complex systems. In particular the presentation shall go over the importance of constructing appropriate networks for the task at hand and the importance of the training data used.

Session 2

Mariia Radova

Fine tuning of MACE foundations models with transfer and delta learning

The mace-freeze tool allows to freeze layers or parameter tensors in the MACE-MP foundations models and bespoke MACE models to fine-tune them for a particular dataset of interest. This approach allows to retain the learned features from the large-scale dataset of the baseline models and adapt the later layers to the new task. Results from mace-freeze will be compared to delta-learned models to determine which technique can offer the same accuracy of predictions as the bespoke MACE model trained for a specific task, with as few data points as possible. The dataset of interest used for fine-tuning is reactive Hydrogen on Copper surfaces.

Fraser Birks

QM/MM Style Mixing of Machine-Learned Interatomic Potentials to Accelerate Simulations

Quantum Mechanical/ Molecular Modelling (QM/MM) is a method that has been used historically to great success in the field of atomistic simulation, with archetypical problems including fracture and dislocation motion. This technique relies on the idea that a system of interest can be partitioned into a local region which needs to modelled accurately (QM) and a remainder which can be modelled cheaply (MM), with a coupling scheme between the two [1]. This idea naturally extends to modern machine-learned interatomic potentials (MLIPs) – where expensive and cheap models replace the QM and MM regions respectively. An enormous benefit of using MLIPs is that they have well-defined local energies; this can be exploited to attain excellent error convergence with QM region size [2]. An archetypal problem for this method would be the modelling of complex damage processes such as irradiation assisted stress corrosion cracking, where simulation domains contain not only large numbers of atoms but also isolated regions of high chemical complexity. This talk will present preliminary work demonstrating this method with two Atomic Cluster Expansion (ACE) potentials [3], with some simple examples of dynamic quantities of interest in Silicon and Iron.

[1] Kermode, J., Albaret, T., Sherman, D., Bernstein, N., Gumbsch, P., Payne, M. C., Csányi, G. and De Vita, A. (2008), Low-speed fracture instabilities in a brittle crystal. Nature 455, 1224–1227.
[2] Chen, H.; Ortner, C. QM/MM Methods for Crystalline Defects. Part 2: Consistent Energy and Force-Mixing. arXiv September 22, 2015.
[3] Drautz, R. Atomic Cluster Expansion for Accurate and Transferable Interatomic Potentials. Phys. Rev. B 2019, 99 (1), 014104.

Jacob Eller

TBC

Arielle Fitkin

Investigating the role of metal-organofluorine interactions in selective metal deposition

Controlled deposition of metals onto a given surface is a slow and costly process, but is essential for electronics and photovoltaics. Recently a novel method for selective deposition has been discovered by our experimental collaborators, the Hatton group, which uses a thin layer of specific organofluorine compounds to prevent metal atoms from being adsorbed. We have levaraged DFT to investigate the nature and strength of the interaction between various metals and organofluorines to determine the extent to which the direct interactions between a metal atom and organofluorine molecule affect this process of selective deposition.

This is the first step in understanding the complex interplay between the metal-organofluorine interaction strength and the polymer-polymer intermolecular interactions which allow specific organofluorines to prevent metal condensation on surfaces.

Vincent Fletcher

Thermodynamically Informed Phase Space Exploration for Optimal Autonomous MLIP Dataset Building

We present an optimal method of database generation for the training of machine learned interatomic potentials (MLIPs). Nested sampling is an unbiased Potential Energy Surface (PES) sampling technique that produces samples across all phases given no prior information. Since the accuracy of any MLIP depends on the underlying data it is trained on, and the data is required to undergo high cost ab-initio evaluation, selecting the fewest and most important data-points is a critical component in developing MLIPs efficiently. Samples generated by nested sampling form a sparse mesh of thermodynamically relevant points of the PES which creates a potent, low cost database that can be iteratively expanded through successive sampling runs. Based on the Atomic Cluster Expansion (ACE) we suggest a highly automated framework and, with our method, we reproduce fundamental properties of pure magnesium (vibrational and elastic properties, phase diagram, 0~K enthalpy curves) with remarkably small databases.

UK Ministry of Defence © Crown owned copyright 2024/AWE

Session 3

Sebastian Dooley

Data-driven equation discovery for liquid film flows thick and thin

Partial differential equation (PDE) discovery is an exciting alternative to the standard first principles-based methodologies regularly used in mathematical modelling, particularly in regimes outside the reach of traditional approaches. This talk explores the application of PDE discovery methods to a variety of PDEs. These include introductory PDEs such as the advection, heat and advection-diffusion equations, which are complemented by looking further to the complex equation environment of liquid film flows, with the aid of direct numerical simulation data. To begin with, we focus our attention on established thin film equations, outlining important derivation aspects to build analytical understanding into the data-driven process and provide reasons for interest in data-driven methods from the thin film fluid dynamics community. Subsequently, we outline the SINDy (sparse identification of nonlinear dynamics) equation discovery method, sharing results from its application to introductory PDEs. We then gently steer the developed framework into new regimes of interest, such as thick liquid film flows, in which classical physical understanding is lacking.

Joseph Duque-Lopez

TBC

Anson Lee

Tracking ions in a travelling-wave based ion mobility mass spectrometer

Ion detection and resolution in mobility-based mass spectrometers rely mainly on innate differences in mobility of different ion species. There are however additional physics phenomena that complicate this, for example, diffusive broadening and field surfing reducing resolving power. It is therefore instructive to study the causes of deviation from ideal ion transport and thus improve upon the ion guide design. This is done both analytically, to simplify numeric models and reduce computational time as well as to capture the essential physics; and computationally for visualisation of the phenomena. In particular, the study is done on a travelling-wave based confining electric field, demonstrating unseen effects on ion trajectories from guide geometry and potential waveforms.