ScholarWorks@성균관대학교

초록

The development of energy-efficient neuromorphic devices requires synaptic transistors capable of precise conductance modulation and scalable fabrication. Carbon nanotubes (CNTs) are attractive candidates owing to their high carrier mobility, mechanical flexibility, and chemical stability; however, their structural heterogeneity has necessitated costly and time-consuming sorting processes to remove metallic CNTs. Here, unsorted CNT (UCNT)-based synaptic transistors are reported that circumvent the need for sorting by employing an ion gel gate containing a huge amount of mobile ionic charges. The UCNT transistors exhibit an ON/OFF current ratio of approximate to 25 through reversible anion-driven doping/de-doping and demonstrate essential synaptic functionalities, including short-term plasticity, linear potentiation/depression across 50 states, paired-pulse facilitation with a 149% index, spike-frequency-dependent plasticity, and endurance over 1000 cycles (100 000 pulses in total). Operando Raman measurements confirm that mobile anions are responsible for p-doping and conductance modulation in the UCNT channel. Neuromorphic simulations using measured device parameters achieve an MNIST recognition accuracy of 88.75%, comparable to the ideal case of 92.75%. This study establishes a scalable and cost-effective route for CNT-based synaptic electronics, paving the way toward practical neuromorphic computing applications.

키워드