Hardware-Efficient Exchange-Only QML: Singlet-Triplet Spin Chains via Inter-pair Coupling without Magnetic Gradients
Abstract
Standard universal quantum computing using exchange-only qubits typically requires three physical spins per logical qubit, leading to significant hardware overhead. Conversely, two-spin units offer higher density but rely on local magnetic field gradients for control, increasing integration complexity. In this paper, we propose a resource-efficient quantum machine learning (QML) architecture that achieves high expressibility using minimal two-spin units and Heisenberg exchange interactions alone, without any magnetic gradients. We shift the paradigm from universal gate-based control to utilizing the intrinsic, time-domain dynamics of a spin chain as a learning resource. Numerical simulations on MNIST digit classification demonstrate that the symmetry-protected constraints of isolated spin pairs are bypassed by leveraging inter-pair exchange coupling. This interference-mediated state mixing significantly enhances the expressibility of the Hilbert space. The model reaches a test-set accuracy of 90.9% +/- 0.2% over five independent seeds on the full 10,000-image MNIST test set. Under an identical linear readout, the trained quantum feature map (88.1%) clearly outperforms a classical linear baseline on the same PCA inputs (83.2%) as well as an untrained (reservoir-style) version of the same dynamics (53.0%), demonstrating that the learned, input-dependent exchange pulses implement a genuinely non-linear and trainable feature map. The protocol is also robust to experimentally relevant imperfections: accuracy remains at 89.9% under 10% quasi-static pulse-area noise and at 89.8% when every observable is estimated from 10^3 measurement shots. These findings suggest that competitive QML can be executed on the simplest possible semiconductor spin-chain hardware, bypassing the need for leakage-prone encodings or complex micro-magnet integration.
Related
- Hardware-Efficient Error Mitigation and Shot-Efficient Sampling on IBM Quantum Hardware
- Hardware-Efficient Reinforcement-Learned Scheduling for Quantum LDPC Decoders
- Hardware-Efficient Ground-State Preparation using Variational Imaginary-Time Majorana Evolution
- Efficient quantum state preparation on Quantinuum hardware
- Effects of residual exchange coupling on simultaneously driven spin qubits
BibTeX
@misc{minato2026hardwareefficient,
title = {{Hardware-Efficient Exchange-Only QML: Singlet-Triplet Spin Chains via Inter-pair Coupling without Magnetic Gradients}},
author = {Yuichiro Minato},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2608.29017},
eprint = {2608.29017},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2608.29017},
}
RIS
TY - UNPB
TI - Hardware-Efficient Exchange-Only QML: Singlet-Triplet Spin Chains via Inter-pair Coupling without Magnetic Gradients
AU - Yuichiro Minato
PY - 2026
DA - 2026/09/27
PB - arXiv
DO - 10.48550/arxiv.2608.29017
UR - https://arxiv.org/abs/2608.29017
AB - Standard universal quantum computing using exchange-only qubits typically requires three physical spins per logical qubit, leading to significant hardware overhead. Conversely, two-spin units offer higher density but rely on local magnetic field gradients for control, increasing integration complexity. In this paper, we propose a resource-efficient quantum machine learning (QML) architecture that achieves high expressibility using minimal two-spin units and Heisenberg exchange interactions alone, without any magnetic gradients. We shift the paradigm from universal gate-based control to utilizing the intrinsic, time-domain dynamics of a spin chain as a learning resource. Numerical simulations on MNIST digit classification demonstrate that the symmetry-protected constraints of isolated spin pairs are bypassed by leveraging inter-pair exchange coupling. This interference-mediated state mixing significantly enhances the expressibility of the Hilbert space. The model reaches a test-set accuracy of 90.9% +/- 0.2% over five independent seeds on the full 10,000-image MNIST test set. Under an identical linear readout, the trained quantum feature map (88.1%) clearly outperforms a classical linear baseline on the same PCA inputs (83.2%) as well as an untrained (reservoir-style) version of the same dynamics (53.0%), demonstrating that the learned, input-dependent exchange pulses implement a genuinely non-linear and trainable feature map. The protocol is also robust to experimentally relevant imperfections: accuracy remains at 89.9% under 10% quasi-static pulse-area noise and at 89.8% when every observable is estimated from 10^3 measurement shots. These findings suggest that competitive QML can be executed on the simplest possible semiconductor spin-chain hardware, bypassing the need for leakage-prone encodings or complex micro-magnet integration.
ER -