Accelerating Quantum Simulations of Materials Through Parameter and Ansatz Transfer Strategies
Abstract
Quantum computing offers a promising route for electronic-structure calculations, but practical condensed-matter applications often require solving large families of related Hamiltonians arising from k-point sampling, compositional variation, defects, and changes in system size. Here, we develop strategies for transferring variational information between related Hamiltonians to accelerate quantum simulations. For Hamiltonians of the same embedded size, optimized variational parameters are transferred between related calculations. For different sizes, we introduce Augmented Ansatz Reuse for Target Initialization (AARTI), which transfers compatible Pauli-generator structure and optimized source parameters, then augments the ansatz with target-specific generators. Our workflow combines density functional theory, Wannier downfolding, embedded tight-binding Hamiltonians, variational quantum algorithms, and neural-network quantum states, with quantum-circuit simulations implemented using NVIDIA's CUDA-Q platform. Using LixCoO2 as a representative battery cathode material, we study Hamiltonian families spanning multiple delithiation levels, dense k-point meshes, and increasing sizes. Parameter and ansatz transfer substantially reduce the optimization steps required to solve these related Hamiltonians while maintaining the common target accuracy, with lower terminal errors in several cases. The calculated electronic-structure evolution with delithiation exhibits pronounced non-rigid-band behavior consistent with experiment. These results establish a framework for exploiting similarity among related Hamiltonians to improve the efficiency and scalability of quantum workflows for realistic condensed-matter systems.
Related
- Accelerating the discovery of nanoporous gas separation materials via Coulomb-enhanced classical density functional theory
- Accelerating Advanced Materials Innovation with New Approach Methodologies
- Bgolearn: a unified Bayesian optimization framework for accelerating materials discovery
- Realistic Simulations of Energy Materials Using Foundation Models and Electrode-Potential Learning
- Modelling Energy Transfer Efficiency in Collision Induced Dissociation and Unfolding Using Molecular Dynamics Simulations
BibTeX
@misc{shivpuje2026accelerating,
title = {{Accelerating Quantum Simulations of Materials Through Parameter and Ansatz Transfer Strategies}},
author = {Saurabh Shivpuje and Vinit Singh and Manas Sajjan and Sabre Kais},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.33617},
eprint = {2609.33617},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.33617},
}
RIS
TY - UNPB
TI - Accelerating Quantum Simulations of Materials Through Parameter and Ansatz Transfer Strategies
AU - Saurabh Shivpuje
AU - Vinit Singh
AU - Manas Sajjan
AU - Sabre Kais
PY - 2026
DA - 2026/09/27
PB - arXiv
DO - 10.48550/arxiv.2609.33617
UR - https://arxiv.org/abs/2609.33617
AB - Quantum computing offers a promising route for electronic-structure calculations, but practical condensed-matter applications often require solving large families of related Hamiltonians arising from k-point sampling, compositional variation, defects, and changes in system size. Here, we develop strategies for transferring variational information between related Hamiltonians to accelerate quantum simulations. For Hamiltonians of the same embedded size, optimized variational parameters are transferred between related calculations. For different sizes, we introduce Augmented Ansatz Reuse for Target Initialization (AARTI), which transfers compatible Pauli-generator structure and optimized source parameters, then augments the ansatz with target-specific generators. Our workflow combines density functional theory, Wannier downfolding, embedded tight-binding Hamiltonians, variational quantum algorithms, and neural-network quantum states, with quantum-circuit simulations implemented using NVIDIA's CUDA-Q platform. Using Li_xCoO_2 as a representative battery cathode material, we study Hamiltonian families spanning multiple delithiation levels, dense k-point meshes, and increasing sizes. Parameter and ansatz transfer substantially reduce the optimization steps required to solve these related Hamiltonians while maintaining the common target accuracy, with lower terminal errors in several cases. The calculated electronic-structure evolution with delithiation exhibits pronounced non-rigid-band behavior consistent with experiment. These results establish a framework for exploiting similarity among related Hamiltonians to improve the efficiency and scalability of quantum workflows for realistic condensed-matter systems.
ER -