Emerging 2D ferroelectric materials for neuromorphic computing: progress, challenges, and opportunities

Abstract To meet the rapid advancements in artificial intelligence and information technology, neuromorphic computing systems have attracted significant attention as promising alternatives to the conventional von Neumann paradigm. Two-dimensional (2D) ferroelectric materials feature atomic-scale thickness, moderate bandgaps, and compatibility with van der Waals heterostructures. These properties offer unique opportunities to overcome the scaling limitations and functional constraints of conventional ferroelectrics, enabling the development of next-generation low-power and highly integrated electronic and neuromorphic systems. In this review, we systematically summarize the fundamental mechanisms and synthesis strategies of 2D ferroelectricity, including bottom-up synthesis techniques and top-down exfoliation approaches, and discuss intrinsic and extrinsic methods for tailoring polarization to optimize device performance. We focus on representative 2D ferroelectric device platforms, such as ferroelectric field-effect transistors (FeFETs), ferroelectric semiconductor field-effect transistors (FeS-FETs), ferroelectric tunnel junctions (FTJs), negative capacitance field-effect transistors (NC-FETs), and ferroelectric photovoltaic (FPV) devices, and examine their potential for in-memory computing, in-sensor computing, and neuromorphic logic. Finally, we identify critical challenges that must be addressed to advance 2D ferroelectric technologies toward practical deployment. These include scalable and phase–pure material synthesis, artifact-aware polarization characterization, improved device uniformity and reliability, and stable integration with complementary metal–oxide–semiconductor (CMOS) platforms. Vectorial polarization in 2D ferroelectrics offers functional advantages but requires precise structural and interfacial control to mitigate cross-axis coupling and instability. At the system level, monolithic stacking with CMOS and van der Waals (vdW) integration strategies enables compact structures that unify sensing, memory, and processing. Opportunities such as hybrid light–matter computing, reservoir-based architectures, and multifunctional in-sensor operations represent promising directions. Continued progress will depend on co-optimization across materials, devices, and circuits, supported by standard benchmarks and scalable fabrication frameworks, to realize energy-efficient, neuromorphic computing systems based on 2D ferroelectrics.