Achieving Reliable and Ultrafast Memristors via Artificial Filaments in Silk Fibroin.

The practical implementation of memristors in neuromorphic computing and bio-mimetic sensing suffers from unexpected temporal and spatial variations due to the stochastic formation and rupture of conductive filaments (CFs). Here, we patterned the bio-compatible silk fibroin (SF) with an on-demand nanocone array by using thermal scanning probe lithography (t-SPL) to guide and confine the growth of CFs in the Silver/Silk Fibroin/Gold (Ag/SF/Au) memristor. Benefiting from the high fabrication controllability, cycle-to-cycle (temporal) standard deviation of the set voltage for the structured memristor has been significantly reduced by ∼95.5% (from 1.535 V to 0.0686 V) and the device-to-device (spatial) standard deviation has also been reduced to 0.0648 V. Besides, the statistical relationship between the structural nanocone design and the resultant performance was confirmed, optimizing at the small operation voltage (∼0.5 V) and current (100 nA), ultrafast switching speed (sub-100 ns), large on/off ratio (10 4 ), and the smallest switching slope (SS < 0.01 mV/dec). Finally, the short-term plasticity and leaky integrated-and-fire behavior were emulated, and a reliable thermal nociceptor system was demonstrated for practical neuromorphic applications. This article is protected by copyright. All rights reserved.