News Release

2D natural clay offers a platform for machine learning algorithm

Peer-Reviewed Publication

Science China Press

Figure 1

image: (a) Camera image of mineral mica and (b) AFM image of 2D mica flakes. (c) Schematic of a 2D mica-based RRAM, (d) schematic of a non-Markov chain algorithm. view more 

Credit: @Science China Press

Algorism plays a significant role in predicting future states of a system. Particularly, non-Markov chain algorithm has been widely applied in epidemic spreading processes, social and man-made memory networks, the environment-related quantum entangled states, and artificial algorisms such as face pose tracking. Traditionally, a large number of memories and computing cells are integrated to achieve these goals by software algorisms, showing high complexity. In the paper published in Science Bulletin, a group led by Bilu Liu and Hui-Ming Cheng from Tsinghua-Berkeley Shenzhen Institute (TBSI) of Tsinghua University has realized a non-Markov chain algorithm in a single resistive random access memory (RRAM) based on 2D mineral material for the first time and revealed the related mechanism.

The researchers found that 2D mica is an excellent ionic conductor, the internal potassium ions (K+) in which controllably migrates under the stimulation of a cyclic electric field to induce resistance switching (RS) phenomena. It is interesting that the related RRAM device exhibits both single-window and bipolar RS behaviors, which is modulated by the strength of the electric field. The migration of the intrinsic K+ contributes to the high on/off ratio of 103, long retention time of more than 108 s, high stability and reliability of the 2D mica-based RAAM, superior to the ones relied on the conduction of external ions.

Accordingly, the authors have successfully achieved the non-Markov chain algorithm in a 2D mica-based RRAM with three states (Figure 1). Different polarities of the input voltages were applied to stimulate the device and produce certain passing paths. By this way, the output signal of the device is not only related to the current input voltage but also the previous state, and a multi-path non-Markov chain is realized. This research reveals the controllable ion transport in 2D layer mineral materials and provides a guideline to design and engineer the related functional devices for realization of algorithms in future.


See the article: Rongjie Zhang, Wenjun Chen, Changjiu Teng, Wugang Liao, Bilu Liu, Hui-Ming Cheng. Realization of a non-markov chain in a single 2D mineral RRAM. Science Bulletin,66,doi: 10.1016/j.scib.2021.04.025

Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.