Correlation of natural honey-based RRAM processing and switching properties by experimental study and machine learning

Abstract

Natural honey is a promising material for hardware components of nonvolatile memory and artificial synaptic devices in emerging renewable and biodegradable neuromorphic systems. The resistive switching properties of these devices are closely correlated with device process conditions. In this paper, honey based resistive random access memory (RRAM) devices were fabricated with different metal electrodes and drying temperature and duration. SET and RESET voltages were measured and used as dataset to train machine learning algorithms. Four machine learning models were applied to process data and demonstrated an average accuracy of 89.9 % to 91.6 % to predict the SET voltages in the range of [0 V, 6 V]. This study established a useful practice for fabrication of RRAM devices based on honey and can be extended to other natural organic materials.