Developing metamaterials, or artificial materials with properties not encountered in nature, invisible to the human eye is an arduous and fairly inefficient process which requires much trial and error in designing atoms smaller than the wavelength of light and making sure the polarisation and spin of light is where it needs to be to achieve the desired effect.
Luckily, a group of scientists from the Department of Mechanical Engineering and Department of Chemical Engineering at POSTECH have developed a technique which allows researchers to use deep learning to speed up the process of picking the right materials and designing required photonic structures.
Once provided with sufficient data, the new AI system was capable of correlating different photonic structures with their optical properties, and developing a design method for making a photonic structure with desired optical properties.
Analyses showed metamaterials designed using this method to have the exact properties predicted by the artificial neural network.
The new design method constitutes a real breakthrough, as it not only reduces the time needed for designing photonic structures, but also allows researchers to engineer new metamaterials without performing empirical testing.
New metamaterials produced with the help of this technique could be used in display, security, and military technologies, in addition to advance the technology used in materials science.
According to the lead researcher Professor Junsuk Rho, the system still requires users to input certain problem settings at the outset, and sometimes leads to untenable designs which are not suitable for producing desired metamaterials.
Considering the above, Professor Rho would like to take these findings “a step further by developing a complete design method of metamaterials using AI”, as well as to “make innovative and practical metamaterials by training AI with reviews of the design constructed in consideration of final products”.
A paper detailing the findings was published in the journals Applied Materials and Interfaces, Nanophotonics, Microsystems & Nanoengineering, Optics Express, and Scientific Reports.