Y Yin

First Contact: A Physical Language for Material Interfaces

First Contact: A Physical Language for Material Interfaces

Speaking script for 15 main slides. Target duration: approximately 12 minutes, including brief pauses on the scientific figures. The proposal comes first; the personnel section begins at slide 10.

1. First Contact: A Physical Language for Material Interfaces — 35 seconds

My Future Fellowship asks a simple question: which interactions must an interface preserve to retain its useful function?

When we bring materials together, the contact can create behaviour that neither material has alone. It can also remove the very property we wanted to use. My project will establish how local atomic environments and their surroundings work together, and which changes they can tolerate. The aim is to turn detailed calculations into physical principles for integrating quantum materials and developing more energy-efficient electronics.

2. Interfaces for quantum computing and advanced electronics — 45 seconds

Interfaces can create properties that neither material possesses alone, opening new possibilities for quantum computing and advanced, energy-efficient electronics. Advances in assembling atomically thin materials make interface design an increasingly promising route towards these technologies. Understanding which interface features enable useful behaviour, and how to preserve them during fabrication, is therefore both a fundamental scientific challenge and essential for developing reliable devices.

This matters for Australia because our ambitions in quantum technologies and advanced electronics depend on bringing materials into practical structures. My fellowship will establish which interactions must be retained, whether they can reproduce function in other interfaces, and where their influence reaches its limits. These are the physical principles that future fabrication research needs.

3. Which features must an interface preserve? — 45 seconds

There are two connected challenges. First, how do we know that our model contains the interactions responsible for the behaviour we want to understand? Second, how do we construct another interface that retains those interactions?

A list of atoms and bonds does not by itself tell us which features matter together. Nor does finding a stable structure guarantee a useful electronic response. I will test whether a limited group of interacting environments can explain that response, while making its dependence on the surrounding material explicit. The following preliminary observations show why this question is important.

4. Complex interfaces can retain strong electrical control — 55 seconds

This example, labelled MX₂–oxides–MX₂, contains chemically and geometrically different contacts. It builds on my broader experience of surfaces, interfaces and disorder, extended through the TQ Transistors screening programme.

Our completed calculations show that the assembled structure has a zero-field gap of about 0.4 electronvolts, substantially smaller than the isolated monolayer. Yet it responds much more strongly to an applied field than the bare bilayer. The important consequence is that substantial electronic reconstruction need not destroy electrical control.

Our preliminary interpretation includes interfacial doping. The fellowship will separate charge redistribution, orbital mixing and screening to establish which contacts reshape the electronic states, and which interactions preserve the strong response. That is the mechanism we need to understand before trying to reproduce it elsewhere.

5. An ordered layer may survive in less ordered surroundings — 50 seconds

The second example asks a complementary question about structural protection. This illustrative reconstruction represents an ordered Xene layer between a functionalised substrate and a lower-symmetry overlayer.

The interesting possibility is that an active layer can remain ordered even when its surroundings are not. But which part of the environment is responsible? Is it the substrate termination, confinement by the cap, or their joint action?

These alternatives can be tested by changing the lower and upper environments separately and together. I will assess structural persistence and electronic behaviour independently. Preserving an ordered layer is a useful starting point; whether it also preserves a useful quantum state is a further scientific question.

6. Feature Zones test which interactions preserve function — 55 seconds

A Feature Zone is a proposed description of interacting atomic environments and the surroundings needed to explain a response. It might link a bonded termination, a weak contact and an active layer. Its boundary must come from physical tests.

I will remove or change interactions separately and together to establish necessity. I will then recreate the required conditions in different, relaxed structures to test sufficiency. Finally, I will predict responses in withheld materials and at larger scales to establish transfer.

This makes the proposed physical unit testable. Segmentation suggests groups, while controlled calculations determine whether they explain the response. If a compact group fails because screening or collective coupling extends further, identifying that missing information is itself a scientific result.

7. Three work packages develop one physical explanation — 55 seconds

Each work package tests a different part of the same physical explanation. WP1 establishes which interactions are necessary and how far they act. The existing MX₂ and Xene observations provide starting evidence for new controlled perturbations.

WP2 recreates the proposed conditions in alternative relaxed structures. Its key outcome is whether those conditions are sufficient to reproduce the same mechanism and function. Failed reconstructions reveal missing physical constraints.

WP3 makes prospective predictions for reserved material pairs and larger interfaces. Independent comparisons will establish accuracy, errors and failure conditions. The scientific success is a tested range of explanation, whether or not it produces a better-performing candidate.

Both PhDs carry this sequence through their material family. I lead the theory, decisive comparisons and synthesis.

8. Does grouping add physical information? — 50 seconds

The central comparison uses grouped and ungrouped atomic graphs with the same physical inputs, training information and computational budget. Removing group membership isolates its contribution. For withheld structures, predictions use only information available before their electronic response is calculated.

WP2 develops six additional material pairs. Six further pairs are reserved for WP3, outside descriptor selection, fitting and constraint tuning. Each PhD predicts three reserved pairs, then studies two representative interfaces at increasing area. The tests assess mechanism and state character, alongside accuracy, compression and transfer.

This makes success measurable. It also makes an unsuccessful local description informative: it will reveal the nonlocal interaction or environmental condition that must be retained. The research therefore produces knowledge about both the reach and the limits of local interface control.

9. Tested interface principles and their predictive limits — 45 seconds

Materials physics will gain tested conditions explaining why different interfaces can share a function, and when local descriptions fail. Computer science will gain independently tested evidence of whether physical grouping improves representation and transfer. Tolerance maps will identify which interfacial variations future fabrication research must control.

For quantum electronics, these principles can guide substrates and protective layers that retain useful states. For energy-efficient electronics, they can guide contacts combining electrical control with useful current flow. Independent calculations and available measurements form the core validation. The Fellowship can therefore deliver its scientific outcomes without a new fabrication or phase-transition campaign.

Publications will report both successful predictions and explained failures. InterfaceHandler, or InHand, will share the supporting models and reproducible tests so other researchers can examine the findings.

10. I connect atomic mechanisms with material function — 45 seconds

I bring a sustained research background in surfaces, interfaces and atomic disorder. My original approaches connect first-principles calculations with efficient electronic models, allowing the interactions identified at atomic scale to be tested in larger structures and transport calculations.

This gives the proposed work a continuous foundation. Detailed calculations support the mechanism tests. Efficient models support comparisons across disorder and scale. Experimental collaborations help determine whether the explanation addresses observable behaviour. The fellowship develops a new question from this established expertise: when can interacting local environments provide a transferable description of an interface?

11. Original modelling is tested through collaborative science — 45 seconds

My studies of bismuthene show how substrate coupling and encapsulation symmetry can change or preserve electronic states. They establish the modelling foundation for the quantum-material part of this proposal.

My collaborative work extends that foundation to observations in real materials. Contributions in ACS Nano, Matter and Nature Communications connect interface interactions, magnetic disorder and atomic dynamics with experimental behaviour.

Across these studies, my role is to move from a calculated or measured response to a physical explanation that can be tested. That combination of original methodology and collaborative interpretation is particularly important here, where similar-looking structures may operate through different mechanisms.

12. I lead a funded programme in interface research — 55 seconds

My work with TQ Transistors demonstrates that I can organise substantial research around an original interface idea. Our completed screening includes 674 geometry–field calculations across 36 distinct stacks, supported by agent-assisted coding and high-throughput workflows.

I am Lead Entrepreneur of the AEA Ignite project with a total value of 650,000 Australian dollars and lead inventor on a pending patent application. My modelling role in the awarded 1.7 million dollar Linkage consortium also connects theoretical work with transistor development.

These programmes provide experience and starting evidence. The Fellowship will undertake new perturbation tests, relaxed alternatives and independent comparisons across material families. Its contribution is to establish physical sufficiency and transfer limits, with success measured by the explanation and its predictive reach.

13. My research leadership and two complementary PhD projects — 50 seconds

I will remain at the centre of the research, developing the theory, designing the decisive calculations, validating methods and integrating the findings.

Each PhD will own an independent material family and progress through identification, construction and transfer. One ARC stipend supports PhD 1 for four years. Monash’s Faculty of Engineering provides one strategic stipend for PhD 2 for 3.5 years. The Department of Materials Science and Engineering covers salary and stipend gaps.

Both students start in Year 1. PhD 2 completes in the first half of Year 4. I will lead synthesis and independent comparisons through Year 4 as PhD 1 completes. My experience includes three co-supervised PhDs and six further doctoral researchers mentored, with five completions across those roles.

14. A focused network supports the research I lead — 50 seconds

The network supplies complementary perspectives around the work I lead. Simon Granville at Victoria University of Wellington contributes quantum materials and disorder; David Cortie at Wollongong contributes disorder and dynamics; Deep Jariwala at Penn State, my Linkage collaborator, contributes two-dimensional interfaces; and Bhaskaran Muralidharan at IIT Bombay contributes transport.

At Monash, established materials and physics collaborations, eResearch support and computing allocation pathways provide the research environment. My NCI and Pawsey experience supports the management of large computational campaigns. Initial subscription support of 12,000 US dollars from OpenAI will extend researcher-supervised coding and analysis.

I will seek specialist machine-learning co-mentoring for segmentation and independent benchmarking. New US directions through TQ and independently explored RIKEN connections remain prospective. Computing access depends on allocations; continued OpenAI support is not assumed. I retain responsibility for scientific validation and interpretation.

15. Understandable and reproducible interface function — 35 seconds

The fellowship will answer three linked questions: which interactions matter, which arrangements can realise them, and where their influence remains transferable.

My contribution is to bring original electronic models, careful physical interpretation and experience of delivering collaborative research to that problem. The result will be a clearer basis for integrating quantum materials, designing electronic contacts and training researchers across materials physics and computation.

I want to establish what an interface must preserve to work—and give future research a reliable basis for reproducing it.

Likely interview questions

How is a Feature Zone different from an ordinary atomic graph?

An atomic graph records atoms and their relationships. A Feature Zone adds a tested statement about which environments must be retained together for a specified response, including necessary boundary information. That statement is the research hypothesis. I will compare grouped and ungrouped representations with identical physical inputs, training data and computational budgets. If grouping loses essential physics or does not improve compression or transfer, I will reject it for that case. The scientific value is in establishing when a collective local description works and what limits it.

Is this a fundamental research project or a software-development project?

The objectives concern new knowledge: identifying the interactions responsible for interface behaviour, establishing whether they can be realised together, and determining their transfer limits. Each work package has physical hypotheses, controlled comparisons and outcomes that can support or challenge the proposed explanation. Software is needed to carry out and reproduce those tests. InHand will share supporting implementations alongside the publications, allowing others to inspect the assumptions and extend the findings. It is the delivery mechanism for the research outputs.

What makes the programme feasible, especially the machine-learning component?

The core route builds on my established electronic models, completed screening workflows and experience of large computational campaigns. Two PhDs study complementary families, while I lead the shared theory and validation. Transparent descriptors and physically controlled comparisons provide the starting point; more complex learning is introduced where reference data demonstrate a benefit. Monash provides computing and data expertise, and I will seek specialist machine-learning co-mentoring for segmentation and independent benchmarking. The research can establish useful physical principles without depending on an unvalidated learning architecture.

How does this Fellowship differ from the AEA and Linkage programmes?

The existing programmes develop transistor technologies. The Fellowship asks which interacting environments and surrounding conditions are necessary and sufficient for an interface response, and where that explanation transfers. The overlap in materials and modelling expertise provides a foundation. The new contribution lies in controlled removal and retention of interactions, reconstruction of the same mechanism in different relaxed structures, and prospective tests in withheld materials and larger interfaces. Existing screening results are preliminary evidence. A finding that compact Feature Zones fail can still establish the nonlocal physics required. In delivery, inherited data and new Fellowship calculations will need clear provenance so that the same activity is not counted twice.

Source and claim notes

Supporting budget

The revised budget requests A$1,022,634 from the ARC, including the Fellow salary and one four-year PhD stipend. Other research expenditure totals A$99,970: A$60,345 for travel and research exchanges, A$21,625 for equipment and the VASP licence, and A$18,000 for publication and dissemination. Equipment comprises the Mac Studio at A$11,445, NAS at A$2,180 and VASP licence at A$8,000. These resources support local model comparisons and shared analysis across WP1–WP3, with the NAS supplying active datasets and checkpoints. Monash cash support totals A$235,180. Annual project costs, including the ARC stipend, are A$59,963, A$58,863, A$59,408 and A$59,088. Prices exclude recoverable GST. The final Mac Studio configuration and VASP quote remain to be confirmed. Travel and publication amounts are estimates.