Exfoliation via Liquid Phase (ELP) is one of the promising mass production methods of 2D materials with applications in water desalination, energy storage, optoelectronic device, DNA sequencing, and energy generation. Compared with other 2D material production methods such as mechanical exfoliation or chemical vapor deposition, ELP has a higher yield at a lower cost. In general, ELP is a simple method, which uses an external force (ultrasonic) on the bulk material immersed in a solvent. One of the most challenging issues hindering its industrialization and academic applications is the design of solvent which optimizes the yield, stability, and quality of produced 2D materials. 

Due to the microscopic nature of exfoliation, mere knowledge of macroscopic properties of solvent and bulk material is not enough to select the optimal solvent. Recent computational studies using molecular dynamics simulation have provided insight into the microscopic phenomena occurring during the exfoliation process. Even though these studies are crucial in understanding the key parameters of the exfoliation process, they do not contribute directly to the design and screening of optimal solvents. On the other hand, data-driven and machine learning methods are successful in finding optimal processes in various tasks, especially in physical problems. The synergy between machine learning and high-throughput simulation can solve the problem with the selection of optimal solvents for ELP.

In this tool, we plan to combine high-throughput computational simulations of various solvents with machine learning algorithms to relate interactions between 2D nano-sheets to the composition of the solvent. By applying the machine learning algorithm, the best solvent composition for the exfoliation process is predicted. The study presented here, therefore, can significantly improve 2D material production through advanced machine learning algorithms, which have been successfully applied to various problems. The machine learning algorithm in the ELP tool is designed to follow an active-learning protocol by letting the dataset grow through adding new materials into the tool dataset. The current material dataset includes room temperature ionic liquids with two cations and four anions, but, there is tremendous room to expand the dataset.

For further details about the tool, please visit the following video. 

At the moment, the dataset is not final, in the next update, we will also make them available to the public.

Researchers should cite this work as follows:

• Alireza Moradzadeh, Narayan Aluru, Darren K Adams (2020), "Machine-learning Assisted Virtual Exfoliation via Liquid Phase," https://nanohub.org/resources/mavelp. (DOI: 10.21981/HN7S-5326).

  BibTex | EndNote

1. nanoMFG Node