Spin-Selective Transport in Closed Molecular Backbones: A Geometry-Based Quantum Model and Physics-Informed Machine-Learning Study
Year: 2026
Type: preprint
Abstract: We study how molecular geometry, spin–orbit coupling, and electrode placement affect spin-dependent electron transport in closed molecular backbones. We compare planar loops, nonplanar unknots, and trefoil knots using a 60-site spinful tight-binding model. Local curvature and binormal direction enter an effective spin–orbit hopping term. Non-equilibrium Green’s functions provide drain-resolved transmission and spin polarization for coherent multi-terminal devices and for devices with charge-conserving Büttiker voltage probes. The simulation dataset contains 25,920 drain-resolved observations across 180 independent geometry groups. Among 25,712 usable observations with transmission \(\:T\ge\:0.01\), 202 reach polarization magnitude \(\:\left|\mathbf{P}\right|\ge\:0.90\). All 202 occur in coherent two- or three-drain configurations; the largest value is \(\:99.9708\text{\%}\) at \(\:T=0.1368\). Strong polarization also appears in planar loops, so it cannot by itself establish a chirality-specific effect. We further construct 185 physics-informed features and train a gradient-boosting model to rank high-polarization candidates. On 1,836 active candidates from held-out geometries, its top-10% shortlist contains 63 of 191 high-response cases, giving \(\:34.24\text{\%}\) precision and a \(\:3.29\times\:\) enrichment over random selection. These results support simulation-based screening of spin-selective molecular transport while identifying the need for stricter chirality and robustness tests.
Source: Research Square
Authors: Shreyash Krishna, Mayank Goswami
Institution: Indian Institute of Technology Roorkee
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7226029153
