Discovering Factorization Surface of Quantum Spin Chains with Machine Learning
AI Breakdown
Get a structured breakdown of this paper — what it's about, the core idea, and key takeaways for the field.
Abstract
Entanglement in quantum many-body systems is required for a variety of quantum information tasks, making it crucial to identify the parameter space in which the ground state is fully separable, known as the factorization surface (FS). Nonetheless, the tuning parameters indicating FS for several quantum spin models remain unknown. We employ symbolic regression (SR), a supervised learning technique, to determine a closed-form expression in the parameter regime corresponding to FS of quantum many-body Hamiltonians. We verify the effectiveness of this method by examining the analytically tractable models, namely a nearest-neighbor (NN) quantum transverse XY model with additional Kaplan-Shekhtman-Entin-Aharony interactions, for which the FS is well-known. We construct an accurate expression for the FS of the XYZ model by providing the parameter set through the SR algorithm in which the ground state is derived by matrix product state formalism. With a satisfactory level of accuracy, we estimate the FS for the long-range XY model, and the NN XY model with Dzyaloshinskii-Moriya type asymmetric interaction for which the factorization surface is not known.