AEA Ignite Progress Report

Quantum Materials

Interface-first design of low-energy quantum devices

Yuefeng Yin*, Co-Investigator A, Co-Investigator B, et al.

INTRODUCTION: Low-energy electronics — topological transistors, spin–orbit logic and negative-capacitance devices — are built by stacking dissimilar two-dimensional (2D) and thin-film materials. Whether a device concept survives depends less on the isolated material than on its interfaces: band alignment, orbital hybridisation, strain and disorder decide what actually reaches a working stack.

RATIONALE: Interface screening has so far been done one "hero" calculation at a time. We built InterfaceHandler (InHand), an open-source package that automates lattice-matched supercell construction for arbitrary pairs of crystals, and coupled it to a high-throughput density-functional theory (DFT) workflow. The program targets three material families with established bulk physics — Heusler ferromagnetic films, 2D bismuth allotropes and topological pyrite-type OsX2 — and asks which interfaces keep their useful properties once contacted.

RESULTS: At the mid-point, InHand has generated 4,812 candidate interfaces across the three families, of which 1,241 show type-I or type-II band alignment and 292 have been converged with plane-wave DFT. Five of eleven milestones are complete and two more are in progress. Heusler films retain their large Berry curvature in 71% of converged stacks, and a first labelled dataset has been curated for the machine-learning surrogate that will replace most brute-force calculations in the second half.

CONCLUSION: An interface-first workflow turns device-stack selection from a case-by-case effort into a searchable design space. The remaining twelve months focus on the pyrite family, a first surrogate model and handing the best stacks to our industry partner for fabrication.

Corresponding author: Yuefeng Yin (yuefeng.yin@monash.edu) Cite this report as Y. Yin et al., Mid-term report, Monash University (2026).
Structurelibrary(MP · C2DB) match & build InHandinterfacelibrary DFT screening Interfaces generated 4,812interfaces Heusler films 2,214 (46%) Bi allotropes 1,492 (31%) Pyrite OsX₂ 1,106 (23%) Milestones 5 / 11complete Complete (5) In progress (2) Planned (4) Classification 1 Material family 2 Band alignment 3 Disorder tolerance Co₂MnGa Bismuthene OsX₂ type I type II ordered disordered
Workflow of the interface screening program. Candidate crystals from public structure databases are paired and lattice-matched by InHand to build an interface library, which is screened with DFT and classified by material family, band alignment and tolerance to disorder. Donuts show progress at the mid-term point (illustrative values).

Quantum Materials

Interface-first design of low-energy quantum devices

Yuefeng Yin1,2*, Co-Investigator A1, Co-Investigator B2, PhD Candidate C1, Industry Partner D3

Heterostructure interfaces decide whether a promising quantum material becomes a working low-energy device, yet they are rarely screened systematically. This mid-term report describes an interface-first program built around InterfaceHandler (InHand), an open-source tool that constructs lattice-matched supercells and feeds them to high-throughput first-principles calculations. We summarise progress across three material families, report the screening statistics to date, and set out the plan, risks and deliverables for the second half of the project.

Two-dimensional materials and ultrathin films have opened a route to electronics that dissipates far less energy than silicon (1–3). Topological edge channels, large Berry curvature and spin–momentum locking all promise low-dissipation transport, but every one of these properties is defined for an idealised, isolated crystal (4, 5). A real device places that crystal against a substrate, a gate dielectric and metallic contacts.

The physics at those contacts — charge transfer, hybridisation, strain and interfacial disorder — can switch a topological state off, pin the Fermi level or wash out the very signal a device relies on. Predicting it requires calculations on the interface itself, and the number of plausible interfaces grows combinatorially with the number of candidate materials, orientations and terminations.

This project replaces the "one hero calculation" approach with a systematic search. Its three streams are (i) software to generate interfaces automatically, (ii) high-throughput screening of three material families with density-functional theory (DFT), and (iii) machine-learning surrogates trained on the resulting data to extend the search at a fraction of the cost.

Interface construction with InHand

Given two bulk or 2D crystals, InHand enumerates supercells of each whose in-plane lattices coincide within a strain tolerance. For lattice vectors 𝐚1,2 (substrate) and 𝐛1,2 (overlayer), a match exists when integer matrices M and N satisfy

M(𝐚1𝐚2)≈R(θ)N(𝐛1𝐛2),|ε|<εmax
(1)

where R(θ) is an in-plane rotation and ε the residual strain applied to the overlayer. Candidate cells are ranked by strain, area and number of atoms so that the smallest physically faithful supercell is passed to DFT (Fig. 1).

A₁ A₂ Substrate Overlayer Matched cell Strain ε0.8 % Twist θ26.6° Atoms46 Area√5 × √5 illustrative values
Fig. 1. Coincidence-lattice matching. A square substrate (blue) and a rotated hexagonal overlayer (orange) share a √5 × √5 supercell spanned by A1 and A2. InHand scores each candidate by residual strain, twist angle and cell size.

Version 2.1 of InHand completes Phase 1 of the software plan: structure import from public databases, the √n supercell search and automatic termination handling. Phase 2 adds a stable Python API for workflow engines and is on track for month 16.

Screening progress

All generated interfaces pass through the same three-stage screen: a fast band-alignment estimate from vacuum-aligned band edges, a relaxed low-precision DFT calculation, and a converged calculation with spin–orbit coupling for the survivors. Figure 2 shows the size of each stage per family.

0 500 1,000 1,500 2,000 Number of interfaces Heusler films 2,214 612 143 Bi allotropes 1,492 388 97 Pyrite OsX₂ 1,106 241 52 Generated Aligned (type I/II) Converged DFT
Fig. 2. Screening funnel by material family. Interfaces generated by InHand, those with type-I or type-II band alignment, and those converged with full DFT at the mid-term point (illustrative values).

Heusler ferromagnetic films

Co2MnGa-type films were previously shown to retain a large Berry curvature even when fully disordered (6). Within the converged set, 71% of film/substrate stacks keep an anomalous Hall conductivity within 20% of the free-standing value; the failures cluster on substrates with strong Mn–O bonding.

2D bismuth allotropes

For bismuthene, the question is whether the topological edge state survives a realistic contact (7). Screening is 70% complete; early results favour van der Waals contacts with large work functions, which preserve the band inversion.

Key result so far

Disorder tolerance, not lattice mismatch, is the main filter: most interfaces that fail do so because interfacial disorder closes the gap or quenches the Berry curvature.

Timeline and milestones

The project runs for 24 months across three research streams and a translation activity. Figure 3 and Table 1 give the status of each milestone at month 12.

M0 M3 M6 M9 M12 M15 M18 M21 M24 Stream 1 · InHand software Phase-1 release √n supercell search Phase-2 API Stream 2 · DFT screening Heusler films Bi allotropes Pyrite OsX₂ Stream 3 · ML surrogates Dataset curation Surrogate v1 Active learning Translation Industry workshop Device-stack handoff mid-term
Fig. 3. Milestone timeline. Filled bars are complete (blue) or in progress (orange, filled to the fraction done); open bars are planned. The dashed line marks this report.

Table 1. Milestone status at the mid-term point.

StreamMilestoneDueStatus
SoftwareInHand Phase-1 releaseM6Complete
Software√n supercell searchM9Complete
SoftwarePhase-2 Python APIM16In progress (55%)
ScreeningHeusler filmsM10Complete
ScreeningBi allotropesM14In progress (70%)
ScreeningPyrite OsX2M19Planned
MLDataset curationM13Complete
MLSurrogate model v1M21Planned
MLActive-learning loopM24Planned
TranslationIndustry workshopM7Complete
TranslationDevice-stack handoffM24Planned

M = project month. Illustrative example.

Risks and plan for the second half

  1. Compute budget. Converged spin–orbit calculations dominate cost; the surrogate model is designed to cut the number needed by roughly an order of magnitude.
  2. Pyrite family. OsX2 surfaces have several competing terminations; InHand's termination handling will be benchmarked on a small set first.
  3. Validation. The top-ranked stacks will be shared with experimental collaborators and our industry partner for growth and transport tests.

The second half shifts effort from building tools to using them, ending in a ranked shortlist of device stacks.

Methods

Interfaces were generated with InHand v2.1 using a strain tolerance of 2% and a maximum of 120 atoms per cell. Electronic structures were computed with the Vienna ab initio simulation package (8) using the projector augmented-wave method (9) and the PBE functional; converged calculations include spin–orbit coupling and van der Waals corrections. Band alignments were estimated from vacuum-aligned band edges. Calculations ran on NCI Gadi.

References and Notes

  1. K. S. Novoselov et al., 2D materials and van der Waals heterostructures. Science 353, aac9439 (2016).
  2. A. K. Geim, I. V. Grigorieva, Van der Waals heterostructures. Nature 499, 419–425 (2013).
  3. Reference placeholder — low-energy electronics review.
  4. M. Z. Hasan, C. L. Kane, Colloquium: Topological insulators. Rev. Mod. Phys. 82, 3045 (2010).
  5. Reference placeholder — Berry curvature and anomalous transport.
  6. Reference placeholder — disorder-robust Berry curvature in Heusler films.
  7. Reference placeholder — edge states in 2D bismuth allotropes.
  8. G. Kresse, J. Furthmüller, Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. Phys. Rev. B 54, 11169 (1996).
  9. P. E. Blöchl, Projector augmented-wave method. Phys. Rev. B 50, 17953 (1994).

Acknowledgments

Funding: Replace with funding sources, e.g. grant schemes and project numbers. Computing: Computational resources were provided by the National Computational Infrastructure (NCI) under an NCMAS allocation. Author contributions: Y.Y. conceived the project, developed InHand and wrote the report. Competing interests: The authors declare no competing interests. Data and code availability: InHand is open source; screening data will be released with the final report.

Author Affiliations

1Department of Materials Science and Engineering, Monash University, Clayton, VIC 3800, Australia. 2School of Physics and Astronomy, Monash University, Clayton, VIC 3800, Australia. 3Industry partner, Melbourne, VIC, Australia. *Corresponding author. Email: yuefeng.yin@monash.edu

Supplementary Materials

Full interface library, screening tables and convergence tests are available on request.

Report period: Oct 2025 – Sep 2026yyfforce.github.io