Eco-Friendly Charge-Trapping Memory Transistors with Balanced Rewritability for Neuromorphic Computing

Sustainable memory devices that combine reliable multilevel data storage, stable synaptic-weight modulation, and environmentally responsive degradability are increasingly needed for next-generation Internet of Things, wearable, and data-centric electronics, where rapid device replacement cycles continue to intensify electronic-waste concerns. Charge-trapping memory transistors offer an attractive platform for this purpose owing to their multilevel memory operation through modulation of trapped charge density and excellent retention characteristics, both of which are essential for preserving programmed conductance states in neuromorphic computing. However, previously reported biodegradable and bio-derived memory and synaptic transistors have mainly been limited in their practical neuromorphic applicability because of write-once-read-many behavior, insufficient retention, or unbalanced rewritable operation. Therefore, biodegradable and eco-friendly charge-trapping memory transistor platforms that integrate electrically rewritable non-volatile memory operation with balanced synaptic-weight modulation remain largely unexplored for their applicability to practical memory and neuromorphic computing systems. Here, we report a biodegradable charge-trapping memory transistor based on an optimized blended charge-trap layer composed of carboxymethyl cellulose sodium salt (CMC), a biodegradable polymer, and tannic acid (TA), a plant-derived polyphenolic molecule. CMC and TA exhibit charge-trapping behavior with different dominant tendencies toward electron and hole trapping, respectively, enabling their bias-dependent trapping contributions to be balanced by controlling the blending ratio. The optimized CMC (88%):TA (12%) charge-trap layer, combined with a tunneling layer, enables repetitive and well-balanced electrical program/erase operation, stable retention with a memory on/off current ratio exceeding 103 for 1000 s, and improved transistor switching characteristics. The resulting stable multilevel conductance modulation supports reversible potentiation and depression, which are implemented as synaptic weights in a hardware neural-network simulation for handwritten-digit recognition, achieving an accuracy comparable to the ideal software system. Moreover, the biodegradable charge-trap layer undergoes water-triggered dissolution in deionized water, leading to rapid loss of device integrity after immersion. These findings represent a significant step toward biodegradable charge-trapping memory transistors that simultaneously enable stable non-volatile retention, balanced synaptic-weight modulation, and environmentally responsive degradability for sustainable neuromorphic electronics.