Two-dimensional ferroelectrics hold great promise for ultrathin, energy-efficient electronic and memory devices. Unlike homobilayers, where the number of candidate bilayers scales linearly with the number of monolayers, the vast design space of heterobilayers, combined with their intrinsically sparse ferroelectric occurrence, renders brute-force screening highly inefficient. Here, we address this challenge by identifying key physical descriptors through high-throughput first-principles calculations on 33 102 hexagonal stacking configurations and by developing a monolayer-informed predictive model that enables rapid identification of sliding ferroelectricity in heterobilayers without explicitly constructing bilayer structures. Relying on monolayer-level descriptors, the model achieves high-fidelity predictions across a broad materials space. Applying this framework, we uncover more than 1000 sliding ferroelectric heterobilayers with out-of-plane polarization (OOP) exceeding that of experimentally reported ${\mathrm{MoS}}_2/{\mathrm{WS}}_2$. Notably, the top-performing candidate exhibits an OOP response approximately 130 times larger than ${\mathrm{MoS}}_2/{\mathrm{WS}}_2$. These results establish a large curated dataset of sliding ferroelectrics and demonstrate a scalable, structure-free discovery paradigm for ferroelectric heterobilayers, paving the way for accelerated design of next-generation two-dimensional ferroelectric devices.

Close

PrevNext

View figure in article

• Electric polarization
• Ferroelectricity
• 2-dimensional systems
• Ferroelectrics
• Density functional theory
• First-principles calculations
• High-throughput calculations
• Machine learning