The development of next‐generation in‐memory and neuromorphic computing can be realized with memory transistors based on 2D ferroelectric semiconductors. Among these, In2Se3 is the interesting since it possesses ferroelectricity in 2D quintuple layers. Synthesis of large amounts of In2Se3 crystals with the desired phase, however, has not been previously achieved. Here, the gram‐scale synthesis of α‐In2Se3 crystals using a flash‐within‐flash Joule heating method is demonstrated. This approach allows the synthesis of single‐phase α‐In2Se3 crystals regardless of the conductance of precursors in the inner tube and enables the synthesis of gram‐scale quantities of α‐In2Se3 crystals. Then, α‐In2Se3 flakes are fabricated and used as a 2D ferroelectric semiconductor FET artificial synaptic device platform. By modulating the degree of polarization in α‐In2Se3 flakes according to the gate electrical pulses, these devices exhibit distinct essential synaptic behaviors. Their synaptic performance shows excellent and robust reliability under repeated electrical pulses. Finally, it is demonstrated that the synaptic devices achieve an estimated learning accuracy of up to ≈87% for Modified National Institute of Standards and Technology patterns in a single‐layer neural network system.

中文翻译：

---

通过 FWF 方法合成的 In2Se3 用于神经形态计算


下一代内存和神经形态计算的开发可以通过基于 2D 铁电半导体的存储晶体管来实现。其中，In2Se3 很有趣，因为它在 2D 五元组层中具有铁电性。然而，以前从未实现过大量 In2Se3 晶体与所需固定相的合成。在这里，展示了使用闪中闪焦耳热方法合成 α-In2Se3 晶体。这种方法允许合成单相 α-In2Se3 晶体，而不管内管中前驱体的电导如何，并能够合成克级的 α-In2Se3 晶体。然后，制备 α-In2Se3 薄片并用作 2D 铁电半导体 FET 人工突触器件平台。通过根据门电脉冲调节 α-In2Se3 薄片的极化程度，这些器件表现出独特的基本突触行为。它们的突触性能在重复电脉冲下表现出出色而稳健的可靠性。最后，证明突触设备在单层神经网络系统中对修改后的美国国家标准与技术研究所模式的估计学习精度高达 ≈87%。