Domain-Wall-Mediated Ultralow-Barrier Sliding and Pinning in Ferroelectric Moiré Superlattices Revealed by Machine Learning

Jia-Wen Li, Sheng Meng, Xinghua Shi, Jin Zhang, Wei-Hai Fang · 2026

Read Domain-Wall-Mediated Ultralow-Barrier Sliding and Pinning in Ferroelectric Moiré Superlattices Revealed by Machine Learning by Jia-Wen Li, Sheng Meng, Xinghua Shi, Jin Zhang, Wei-Hai Fang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-20277.txt.

Introduction

Li and four colleagues use machine-learning molecular dynamics to watch sliding in ferroelectric MoS2 moiré superlattices. Heat alone drives layers to slide near 1 m/s at 300 K, as a broad drift of the pattern rather than stiff motion of the whole bilayer. With relaxed constraints the slide follows a near barrier-free path matching the drift, defying the small rigid-slide barrier. Sulfur gaps flip sliding to pinning, with just 0.1 percent sulfur gaps enough to turn long-range sliding into local wobbles, holding beyond small twist angles.

Catalog record

CREATORJia-Wen Li, Sheng Meng, Xinghua Shi, Jin Zhang, Wei-Hai Fang
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE28315
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.20277

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Domain-Wall-Mediated Ultralow-Barrier Sliding and Pinning in Ferroelectric Moiré Superlattices Revealed by Machine Learning Jia-Wen Li, Sheng Meng, Xinghua Shi, Jin Zhang, Wei-Hai Fang Abstract Sliding ferroelectrics built from stacked nonpolar monolayers enable out-of-plane polarization and unconventional switching via interlayer sliding, yet the microscopic sliding dynamics remain unclear. Using machine-learning molecular dynamics, we reveal spontaneous thermally driven interlayer sliding in ferroelectric MoS 2 moiré superlattices, with relative velocities on the order of 1 1 m/s at 300 K. Instead of rigid translation of the entire bilayer, the motion appears as a global drift of the moiré pattern. Such thermally driven sliding is inconsistent with the meV/atom-scale rigid-sliding barrier. In contrast, when constrained relaxation is allowed, the sliding proceeds along an almost barrierless pathway that directly reproduces the global drift of the moiré pattern. Furthermore, sulfur vacancies trigger a sliding-to-pinning transition, with ∼ \sim 0.1% S vacancies already sufficient to convert the long-range sliding into localized oscillations. Notably, these phenomena are not restricted to small twist angles, but arise generically in twisting-induced multidomain structures. These results reveal that the sliding process is governed by a domain-wall-mediated collective reconstruction pathway with an ultralow barrier, rather than rigid layer translation, deepening the understanding of microscopic dynamics in moiré superlattices and sliding ferroelectrics. Introduction Driven by the demand for next-generation storage devices, sliding ferroelectrics with switchable spontaneous polarization have emerged as exceptionally promising candidates [ 1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 ] . Many sliding ferroelectrics have been reported, such as 3R-MoS 2 [ 1 , 9 ] , 1T ′ -WTe 2 [ 10 ] , 1T ′ -ReS 2 [ 11 ] , \beta -InSe [ 12 ] , \gamma -InSe [ 13 ] , T d -MoTe 2 [ 14 ] , h-BN [ 15 ] , MoS 2 /WS 2 [ 16 ] and other transition metal dichalcogenides [ 17 ] . These materials show many advantages, including fatigue-free switching, high endurance, and compatibility with flexible and scalable nanodevices [ 18 , 19 , 20 ] . Their spontaneous polarization arises from the redistribution of interlayer charges driven by symmetry breaking at the van der Waals interface [ 8 ] . Reversible polarization switching can be realized by manipulating the in-plane sliding of these atomic layers [ 21 ] . Despite extensive research, the microscopic dynamics of interlayer sliding during polarization reversal remains unclear. One widely known picture views polarization switching as a collective rigid-like synchronized sliding of the two layers, offering an intuitive view for tracking local stacking evolution [ 8 , 21 ] . However, experiments have shown that switching is more naturally described in terms of domain-wall (DW) motion [ 1 , 5 , 17 ] . Recent theoretical works support this view by showing that polarization reversal proceeds through the formation and propagation of DWs, which show superlubric and soliton-like dynamics [ 22 , 23 , 24 , 25 ] . These results suggest that DWs are not merely static boundaries between domains, but highly active structural units that govern the sliding dynamics. Interestingly, a related picture has emerged in moiré superlattices, where atomic reconstruction typically produces distinct stacking domains separated by DWs, often referred to as solitons [ 26 , 27 ] . These DW networks host ultrasoft shear and phason-like collective modes [ 28 ] , which were recently observed experimentally in twisted bilayer WSe 2 [ 29 ] . These ultrasoft collective modes imply a weak restoring force along DW-associated sliding coordinates, suggesting a low-energy pathway for interlayer motion. Consistently, thermally driven moiré-pattern distortions and twist-angle relaxation were reported in experiments [ 30 , 31 , 32 ] . Meanwhile, classical molecular dynamics simulations revealed thermally driven drift of the moiré texture [ 33 ] , while other calculations suggested that twisting can lower sliding barriers [ 34 , 35 , 36 ] . These DW-associated ultrasoft modes echo the superlubric DW dynamics recently proposed in sliding ferroelectrics [ 22 , 23 , 24 , 25 ] , and naturally draw attention to ferroelectric moiré superlattices as a promising platform for exploring DW-mediated interlayer motion. Locally, ferroelectric moiré superlattices resemble conventional sliding ferroelectrics in that both consist of ferroelectric domains separated by domain walls [ 37 , 38 ] . This twist-induced multidomain texture with an extended DW network makes them an ideal platform for amplifying and probing DW-mediated sliding dynamics. In this work, we explore the molecular dynamics of ferroelectric MoS 2 moiré superlattices with a high-accuracy machine-learning potential (MLP). The simulations reveal spontaneous thermally driven long-range interlayer sliding, with velocities on the order of 1 m/s. Rather than a rigid interlayer translation, the observed motion manifests as a phason-like global drift of the moiré texture. Such thermally driven long-range sliding is inconsistent with the meV/atom-scale rigid-sliding barrier, which would also destroy the moiré pattern. We therefore calculated the relaxed sliding barrier by allowing full atomic relaxation while constraining the interlayer shift. This pathway reduces the barrier by nearly two orders of magnitude to a nearly barrierless level and directly reproduces the global drift of the moiré pattern. Together, these results reveal that the observed sliding arises from a DW-mediated reconstruction pathway with an ultralow effective barrier. In addition, this nearly barrierless state is highly susceptible to defects, as nearly 0.1% S vacancies could act as local pinning centers to convert this long-range sliding into bounded oscillations, thereby accounting for the absence of freely drifting moiré patterns in experiments. Furthermore, both the ultralow-barrier sliding and defect-induced pinning are not restricted to small twist angles, but persist across a broad range of twist structures. This shows that the essential ingredient is not a small twist angle with a large moiré period, but the twisting-induced multidomain structure and its DW networks. By investigating ferroelectric moiré superlattices, we show that their sliding process is governed by a DW-mediated collective reconstruction pathway with an ultralow energy barrier, rather than rigid layer translation, thereby deepening the understanding of moiré superlattices and sliding ferroelectrics. Results Performance of the MLP for bilayer MoS 2 The structure of bilayer 3R-MoS 2 is shown in Figure 1 (a), whose rhombohedral stack breaks interlayer inversion symmetry and gives rise to an out-of-plane ferroelectric polarization. Density functional theory (DFT) calculations yield an in-plane lattice constant of 3.16 Å and an interlayer distance of 3.05 Å, consistent with previous reports [ 1 ] . Interlayer sliding modifies the stacking registry and leads to several high-symmetry configurations, including MX, XM, SP, and SP ′ , which correspond to upward polarization, downward polarization, and nonpolar states, respectively, as illustrated in Fig. 1 (b). The associated sliding energy profile, shown in Fig. 1 (g), features nearly degenerate MX and XM minima separated by higher-energy nonpolar SP and SP ′ configurations, with a sliding barrier of \sim 3 meV/atom. Figure 1: Performance of the machine learning model on bilayer 3R-MoS 2 . (a) Side view of the bilayer 3R-MoS 2 . (b) Top views of representative sliding configurations, including MX (up-polarized), SP (nonpolar), XM (down-polarized), and SP ′ (nonpolar). (c, d) Comparison of the predicted and true energies (c) and forces (d), showing mean absolute errors (MAE) of 0.11 meV/atom

Read the complete text