Entropy-Stabilized Oxide Memristors

Memristor arrays have emerged as a promising hardware platform for efficient machine learning tasks. Traditional amorphous oxide-based memristor materials, however, suffer from device stochasticity and a lack of tunability which hinder applications requiring adaptive networks. Here I will present a study on tunable carrier transport and dynamics in single crystalline (MgCoNiCuZn)O entropy-stabilized oxide (ESO) thin films. We find that the ESO undergoes composition tunable hopping conduction in agreement with the composition dependent point defect formation and electronic structure from first principles calculations. Notably, the transport is bulk and non-filamentary. Pulsed measurements reveal a low resistance state with a short, composition-tunable retention time that can be harnessed for memristor function via temporal data processing. We interpret the carrier dynamics in terms of voltage modulated filling of deep level defects states that are controlled by composition as predicted by theory. The precise tunability of carrier transport in this ESO make it an excellent candidate for “task specific” neural network systems with record energy efficiency for temporal data processing.