Project Overview

Artificial Intelligence (AI) algorithms such as Machine Learning and Deep Learning are transforming areas previously considered to be innately human activities. One area where AI methods have potential to make huge inroads is materials science, where complex interatomic potentials have historically been developed using a mixture of chemical and mathematical intuition. Development of new interatomic potentials is a tedious and non-unique process, and computer simulations of materials using AI methods offer a path to massively speed up the process. 
 

Deep learning networks have been found to excel in superior compact representations while being able to surpass generalization performances compared to traditional machine learning methods (e.g. Support Vector Machines) in many disciplines such as computer vision, machine translation, natural language processing and many data analytics applications in science and engineering. This ability to generalize at the algorithm level is ideal for interatomic potentials, where traditionally the practitioner has to intuit a mathematical expression which is simultaneously computationally efficient and chemically accurate. Deep learning breaks this impasse, albeit at the cost of introducing a ‘black box’ whose inner workings are opaque to the user.

Aims

The aim of this project is to apply AI algorithms to the simulation of solids containing carbon, tungsten and cobalt. The ultimate goal is an Machine Learning interatomic potential which can be used to perform cutting edge molecular dynamics simulations of high- technology hardmetal cutting tools. The resultant deep learning tool should be able to describe large systems, of order hundreds of thousands of atoms, far beyond the capability of density-functional-theory (DFT) while being able to train efficiently on datasets generated at quantum-mechanical accuracy using DFT.

Objectives

The project will be an interdisciplinary collaboration between Curtin University, University of Bochum (Germany) and Sandvik (Sweden). The three groups have recently collaborated in a pilot study, and this project builds upon initial progress. The project sits at the nexus of materials science and computer science, and suits a student with strong abilities in one discipline, and interest/familiarity in the other. The student will be guided to become competent across both disciplines, as the fundamental task is to synthesize the training needs of an AI with the local invariance constraints of an interatomic potential.  

The program will be built upon the new field of Atomic Cluster Expansion (ACE) potentials, which are a breakthrough in Machine Learning methods. The ACE family has unified the field of ML, and has led to new directions (MACE=multi-ACE and GRACE=graph-ACE) which offer quantum-level accuracy at vastly reduced computational costs. This project will apply the ACE family of methods to a complex three-element system with industrial significance.

Significance 

The industrial context of the proposal is the desire to remove cobalt from hard metal cutting tools used in mining and manufacturing. For 100 years the hard metal industry has been dominated by the compound tungsten-carbide-cobalt, in which tungsten- carbide (WC) forms the ‘hard phase’, and cobalt serves the role of binder that holds the WC grains together. Although cobalt works brilliantly in hard metals, there are strong drivers to find alternatives due to chemical toxicity, high prices due to demand in batteries, and social questions associated with mining practices in Africa. 

Sandvik and its competitors have spent tens of millions of dollars exploring alternatives to cobalt, but this trail-and-error approach has not been successful. As an alternative, Sandvik is interesting in computational aided design as a platform for developing better materials and products. This PhD is part of a broader program at Sandvik to develop new computational methods. Determining a pathway to describing a complex three-element system like WC-Co would lay the groundwork for other materials systems in which computational methods can be used to design future products.

An internship may be available for this project. The student would have an opportunity to spend time in Stockholm working in the Materials Modelling and Simulation Group. This would expose them to the daily practices and industrial context of computer-guided development of industrial products.

• Future Students

• Faculty of Science & Engineering
  • Science courses
  • Engineering courses

• Higher Degree by Research

• Australian Citizen
• Australian Permanent Resident
• New Zealand Citizen
• Permanent Humanitarian Visa
• International Student

• Merit Based

The annual scholarship package, covering both stipend and tuition fees, amounts to approximately $70,000 per year.

In 2024, the RTP stipend scholarship offers $35,000 per annum for a duration of up to three years. Exceptional progress and adherence to timelines may qualify students for a six-month completion scholarship.

Selection for these scholarships involves a competitive process, with shortlisted applicants notified of outcomes by November 2024.

1

All applicable HDR courses.

The HDR applicant will have a strong interest in computation and a high level of proficiency in either Computing or the Physical Sciences (i.e. Physics/Chemistry). Undergraduate experience spanning both domains would be an advantage. The applicant will be motivated to learn new computational skills beyond their core knowledge, and will have the ability to integrate concepts across discipline boundaries. Essential skills include proficiency in at least two of the following: Python, UNIX, high-performance computing and AI training. Desirable skills include experience with atomistic simulation packages (LAMMPS, VASP) and C++.
 

Please send your CV, academic transcripts and brief rationale why you want to join this research project via the HDR Expression of Interest form to the project lead researcher, listed below. 

You must be enrolled in a Higher Degree by Research Course at Curtin University by March 2025.

Project Lead: Associate Professor Nigel Marks