TY - JOUR
T1 - Electrically reconfigurable dual-mode synapses based on topological insulator nanowires for high-accuracy biometric recognition
AU - Shi, Yuanhong
AU - Li, Haotian
AU - Hua, Qilin
AU - Zhai, Junyi
AU - Hu, Weiguo
N1 - Publisher Copyright:
© 2026 Elsevier B.V.
PY - 2026/11/1
Y1 - 2026/11/1
N2 - The hardware realization of artificial synapses capable of faithfully emulating the brain's rich, multiscale computational dynamics remains a key challenge in developing reliable biometric recognition systems. Here, we report an electrically reconfigurable memristive synaptic device based on vertically aligned Bi₂Se₃ topological-insulator nanowires with an Ag/PMMA/Bi₂Se₃/Pt heterostructure. Within a compact architecture, the device enables dual regulation of volatile short-term plasticity and non-volatile long-term plasticity. This mode-reconfiguration capability remains stable during alternating cycles, demonstrating reproducible dual-mode operation. The dual-mode behavior originates from the synergistic interaction between Ag electrochemical metallization in the PMMA barrier and defect/trap-regulated carrier transport in Bi₂Se₃, while the nanowire-array architecture spatially localizes the conductive pathways and electrically isolates neighboring units. The device exhibits reproducible multilevel conductance modulation, with six distinguishable states retained for more than 10,000 s. It achieves a SET response time of ≤10 ns and a RESET response time of approximately 40 ns, with corresponding operation energies of 25.6 and 269.2 pJ, respectively. Moreover, the device emulates representative synaptic functions over timescales ranging from nanoseconds to milliseconds, including paired-pulse facilitation/depression (PPF/PPD) and long-term potentiation/depression (LTP/LTD). Notably, in a complex fingerprint-recognition task, an artificial neural network constructed using the physical conductance states of the device achieves a high classification accuracy of 94%, comparable to that obtained with idealized software-simulated synapses. This work establishes a new class of topological-insulator synapses and demonstrates a synergistic design strategy that integrates electrical reconfigurability with high-precision analog signal modulation, laying a foundation for robust and energy-efficient hardware-native biometric intelligence.
AB - The hardware realization of artificial synapses capable of faithfully emulating the brain's rich, multiscale computational dynamics remains a key challenge in developing reliable biometric recognition systems. Here, we report an electrically reconfigurable memristive synaptic device based on vertically aligned Bi₂Se₃ topological-insulator nanowires with an Ag/PMMA/Bi₂Se₃/Pt heterostructure. Within a compact architecture, the device enables dual regulation of volatile short-term plasticity and non-volatile long-term plasticity. This mode-reconfiguration capability remains stable during alternating cycles, demonstrating reproducible dual-mode operation. The dual-mode behavior originates from the synergistic interaction between Ag electrochemical metallization in the PMMA barrier and defect/trap-regulated carrier transport in Bi₂Se₃, while the nanowire-array architecture spatially localizes the conductive pathways and electrically isolates neighboring units. The device exhibits reproducible multilevel conductance modulation, with six distinguishable states retained for more than 10,000 s. It achieves a SET response time of ≤10 ns and a RESET response time of approximately 40 ns, with corresponding operation energies of 25.6 and 269.2 pJ, respectively. Moreover, the device emulates representative synaptic functions over timescales ranging from nanoseconds to milliseconds, including paired-pulse facilitation/depression (PPF/PPD) and long-term potentiation/depression (LTP/LTD). Notably, in a complex fingerprint-recognition task, an artificial neural network constructed using the physical conductance states of the device achieves a high classification accuracy of 94%, comparable to that obtained with idealized software-simulated synapses. This work establishes a new class of topological-insulator synapses and demonstrates a synergistic design strategy that integrates electrical reconfigurability with high-precision analog signal modulation, laying a foundation for robust and energy-efficient hardware-native biometric intelligence.
KW - Artificial synapse
KW - Biometric recognition
KW - Memristor
KW - Neuromorphic computing
KW - Topological insulator nanowires
UR - https://www.scopus.com/pages/publications/105049040602
U2 - 10.1016/j.cej.2026.180619
DO - 10.1016/j.cej.2026.180619
M3 - Article
AN - SCOPUS:105049040602
SN - 1385-8947
VL - 547
JO - Chemical Engineering Journal
JF - Chemical Engineering Journal
M1 - 180619
ER -