Multifunctional CMOS-integrable and reconfigurable 2D ambipolar tellurene transistors for neuromorphic and in-memory computing

Abstract

Despite significant efforts to eliminate the von Neumann bottleneck with new two-dimensional (2D) nanomaterial-based cutting-edge device structures, there remains room for exploring alternative computing architectures that leverage 2D nanomaterials. This study introduced a groundbreaking strategy featuring a complementary metal-oxide semiconductor (CMOS)-integrable and reconfigurable ambipolar 2D tellurene (Te) transistor toward non-von Neumann computing architecture. The innovative scenario integrated seamlessly with CMOS technology, utilizing the p/n-switchable ambipolar characteristics inherited from precise Fermi-level alignment via thermal atomic layer deposition. Further, the architecture exhibited remarkable synaptic behavior while maintaining the conventional inverter performance within a compact single 2D Te device architecture. Expanding these findings, we demonstrated a compact programmable CMOS inverter with reduced spatial complexity and also visualized the construction of diverse complementary logic-in-memory computing. The results of this study pave the way for revolutionary in-memory computing that transcends the boundaries of the von Neumann architecture based on 2D nanomaterials.