Related Experiment Video
Updated: May 23, 2025
05:37
Fabrication of an Optical Cell Dryer for the Spectroscopic Analysis Cells
Published on: January 8, 2019
5.1K
Rapid Discovery of Graphene Nanoflakes with Desired Absorption Spectra Using DFT and Bayesian Optimization with
Şener Özönder1, Hatice Kübra Küçükkartal2,3
1Institute for Data Science & Artificial Intelligence, Boğaziçi University, İstanbul 34342, Turkey.
The Journal of Physical Chemistry. A
|May 8, 2025
Summary
This study introduces a faster, cheaper way to find new materials using Bayesian optimization and neural networks. The method significantly reduces computational costs for discovering materials with specific properties.
Area of Science:
- Computational materials science
- Artificial intelligence in chemistry
- Materials discovery
Background:
- Discovering new materials with desired properties often requires computationally expensive grid searches using methods like density functional theory (DFT).
- High-dimensional chemical spaces make exhaustive searches computationally prohibitive.
Purpose of the Study:
- To develop an efficient and cost-effective guided search strategy for exploring large chemical spaces.
- To minimize the number of computationally intensive calculations needed for materials discovery.
Main Methods:
- Utilized Bayesian optimization (BO) integrated with an artificial neural network kernel.
- Trained the kernel neural network on a limited set of DFT results to guide subsequent exploration.
- Applied the method to identify doped graphene quantum dots (GQDs) with maximal light absorption.
Main Results:
- Achieved significant computational cost reduction (approx. 80%) compared to a full grid search.
- Successfully identified high-light-absorbing GQD structures using only 12 time-dependent DFT (TDDFT) calculations.
- Demonstrated the method's effectiveness in accelerating materials discovery for doped GQDs.
Conclusions:
- The proposed Bayesian optimization approach with a neural network kernel offers a scalable and efficient solution for materials discovery.
- This method can be broadly applied to accelerate the search for new drugs, chemicals, crystals, and alloys in complex chemical spaces.
