TY - JOUR
T1 - Ferroelectrically modulated α-In2Se3/ReS2 heterostructure for sensing-memory-computing devices
AU - Ma, Haiya
AU - Zhang, Xutao
AU - Liu, Liang
AU - Yin, Yihan
AU - Du, Wanyu
AU - Zhao, Chenyang
AU - Gan, Xuetao
N1 - Publisher Copyright:
© 2026 Author(s).
PY - 2026/3/1
Y1 - 2026/3/1
N2 - To address the von Neumann bottleneck, multifunctional optoelectronic devices integrating sensing and computing capabilities have garnered increasing attention. In this work, we demonstrate a ferroelectric heterojunction phototransistor based on an α-In2Se3/ReS2 stack, which combines sensing, memory, and computation capabilities. By modulating the polarization state of α-In2Se3 via gate voltages, the device dynamically tunes the built-in electric field and thereby regulates its conductance state. The device achieves a high responsivity of 17.2 A/W and a specific detectivity of 3.6 × 1012 Jones, together with excellent nonvolatile retention. Furthermore, it effectively emulates key synaptic behaviors including short-term plasticity, long-term plasticity, paired-pulse facilitation, and experience-dependent learning. When incorporated into neural network models, the device achieves recognition accuracies of 93.7% for handwritten digit classification and 84.0% for clothing recognition. These results underscore the potential of α-In2Se3/ReS2 heterostructures as building blocks for next-generation neuromorphic optoelectronic systems with integrated sensing-memory-computing capabilities.
AB - To address the von Neumann bottleneck, multifunctional optoelectronic devices integrating sensing and computing capabilities have garnered increasing attention. In this work, we demonstrate a ferroelectric heterojunction phototransistor based on an α-In2Se3/ReS2 stack, which combines sensing, memory, and computation capabilities. By modulating the polarization state of α-In2Se3 via gate voltages, the device dynamically tunes the built-in electric field and thereby regulates its conductance state. The device achieves a high responsivity of 17.2 A/W and a specific detectivity of 3.6 × 1012 Jones, together with excellent nonvolatile retention. Furthermore, it effectively emulates key synaptic behaviors including short-term plasticity, long-term plasticity, paired-pulse facilitation, and experience-dependent learning. When incorporated into neural network models, the device achieves recognition accuracies of 93.7% for handwritten digit classification and 84.0% for clothing recognition. These results underscore the potential of α-In2Se3/ReS2 heterostructures as building blocks for next-generation neuromorphic optoelectronic systems with integrated sensing-memory-computing capabilities.
UR - https://www.scopus.com/pages/publications/105032648124
U2 - 10.1063/5.0318879
DO - 10.1063/5.0318879
M3 - 文章
AN - SCOPUS:105032648124
SN - 1931-9401
VL - 13
JO - Applied Physics Reviews
JF - Applied Physics Reviews
IS - 1
M1 - 011425
ER -