While machine learning interatomic potentials (MLiPs) have matured to revolutionize material science, deep learning models for electronic structure are just beginning to emerge and restricted, almost exclusively, to non-orthogonal basis Hamiltonians. We introduce G(Wa)NN, the first deep-learning model capable of generating the electronic Hamiltonian of solid-state systems in an orthogonal Wannier basis. G(Wa)NN is trained on an unprecedented, diverse dataset of more than 111K Wannier Hamiltonians (150M+ hopping matrices) spanning 69 elements. The combination of optimized inference and linear-scaling methods for orthogonal Hamiltonians unlock transport simulations at massive scales (10K+ atoms). Crucially, the framework supports local finetuning, allowing users to adapt the base model to custom Wannier Hamiltonian datasets. To seamlessly translate these predictions into physical observables, we introduce Tailwater, a Python package providing an API interface to G(Wa)NN alongside a high performance post-processing library. Tailwater enables automated projection of the predicted Hamiltonian into an arbitrary low-energy subspace-directly mirroring familiar Wannier90 workflows-and includes a suite of Kernel Polynomial Method (KPM) functions that exploit the orthogonal basis to achieve strict linear scaling for spectral observables. The Tailwater ecosystem, with the G(Wa)NN model at its core, aims to help bridge the gap between deep learning and macro-scale quantum transport simulations.
Accurate prediction of electronic Hamiltonians would enable broad property inference while avoiding the high computational cost of Density Functional Theory (DFT). However, progress toward general-purpose materials foundation models is limited by a data bottleneck: existing Hamiltonian datasets are typically small, lack structural diversity, and often omit essential relativistic physics such as spin--orbit coupling (SOC). We therefore construct UniHam, a large-scale Hamiltonian dataset and benchmark suite comprising 100,000+ DFT-computed complex-valued Hermitian Hamiltonians with full SOC, covering 72 elements and a wide range of crystal geometries and symmetries (spanning diverse lattice types and space-group families). Building on UniHam, we benchmark two representative state-of-the-art models under a standardized protocol and introduce complementary evaluation metrics that jointly assess three dimensions: (i) Hamiltonian reconstruction accuracy, (ii) out-of-distribution (OOD) generalization across composition/symmetry shifts, and (iii) the ability to support downstream property prediction from the predicted Hamiltonians. Experiments on UniHam demonstrate that the proposed benchmark and metrics effectively differentiate model capabilities, revealing intrinsic SOC- and element-dependent failure modes, large variations in compositional OOD robustness, and the necessity of spectral-level evaluation to assess whether Hamiltonian predictions reliably support downstream electronic-structure properties. Overall, UniHam provides a reproducible, SOC-complete benchmark that can sharpen model comparisons and accelerate the development of next-generation foundation models for quantum materials.
Yuewen Huang, Pin Chen, Yutong Lu· Proceedings of the 32nd ACM...· 0 citations
On-the-fly N${^2}$AMD (Neural network NAMD), a machine learning framework that makes on-the-fly NAMD in solids a reality, by employing an equivariant neural network to predict the system Hamiltonian, delivers excited-state energies, forces, and non-adiabatic coupling vectors at a fraction of the cost of ab initio calculations.
Changwei Zhang, Yang Zhong, Zhi-Guo Tao et al.· Physical Review Letters· 1 citation
This work proposes an operator-centric framework in which the external (nuclear) potential, expressed in an AO basis, serves as the model input and builds hierarchical, body-ordered representations of atomic configurations that closely mirror the principles underlying several popular atom-centered descriptors.
Jigyasa Nigam, T. Smidt, G. Dusson· Journal of Chemical Physics· 2 citations
Two-dimensional (2D) materials exhibit a wide range of electronic properties that make them promising candidates for next-generation nanoelectronic devices. Accurate prediction of their quantum transport behavior is therefore of both fundamental and technological importance. While the Non-Equilibrium Green's Function (NEGF) formalism coupled with Density Functional Theory (DFT) provides reliable insights, its high computational cost limits applications to large-scale or high-throughput studies. Here we present DeePTB-NEGF, a framework that combines a deep learning-based tight-binding Hamiltonian derived directly from first-principles calculations (DeePTB) with efficient quantum transport simulations implemented in the DPNEGF package. We validate the method on five prototypical 2D materials (graphene, hexagonal boron nitride (h-BN), [Formula: see text], [Formula: see text], and black phosphorus) demonstrating excellent agreement with conventional DFT-NEGF for band structures and transmission spectra. Beyond single-material benchmarks, we showcase the framework's versatility by exploring strain engineering (uniaxial strain on graphene and biaxial strain on [Formula: see text]), substitution doping in [Formula: see text], and current-voltage characteristics of a graphene field-effect transistor (FET). A scaling analysis reveals that DeePTB-NEGF can simulate systems with hundreds of atoms in minutes, achieving speed-ups of over [Formula: see text] compared to DFT-NEGF for heterostructures such as graphene/h-BN/graphene. These results establish DeePTB-NEGF as a powerful tool for autonomous, high-throughput design of quantum transport in microscopic heterostructures, enabling rapid prototyping of next-generation 2D devices.
The seamless integration of Density Functional Theory (DFT) with quantum variational algorithms is essential for the predictive simulation of strongly correlated materials. In this work, we present an end-to-end computational pipeline - comprising DFT geometry relaxation, non-self-consistent field (NSCF) calculations, and Wannier-based orbital localization - to prepare active-space Hamiltonians for quantum embedding. We utilize the Adaptive Variational Quantum Eigensolver (ADAPT-VQE) framework, significantly enhanced by a Greedy-Operator Commutativity Partitioning (GOCP) approach and a Taylor-expanded O(5) operator evolution strategy to efficiently manage the exponential scaling of the Hilbert space. We demonstrate this framework through a systematic benchmark study of Li-hBN, mapping the system onto qubit registers and investigating the convergence behavior as the active space is expanded from 8 to 14 spatial orbitals. Our results quantify the relationship between active-space size and computational demand, identifying a critical"scaling wall"where classical simulation costs transition from manageable to intractable. This study provides a rigorous performance baseline for the DFT-to-ADAPT-VQE workflow and offers empirical insights into the memory and processing limits currently facing hybrid quantum-classical architectures using advanced co-processing strategies.
The results indicate that symmetry compilation concentrates the expressive power of NQS on states relevant to the target problem, thereby reducing model size and training cost without sacrificing accuracy.
Turbasu Chatterjee, M. Sajjan, Songbo Xie et al.· 0 citations