Delft University of Technology engineers use deep learning algorithms to back-calculate microarchitectures from desired properties.
Metamaterials offer engineers the chance to design devices that seem to defy the laws of nature: superlenses with resolutions beyond the diffraction limit, acoustics that can steer sound waves around obstacles, even the elusive invisibility cloak.
All of this is due to the fact that metamaterials derive their properties from their geometric structure at the nanoscale. The trouble is that these geometric structures are notoriously difficult to design. However, engineers at Delft University of Technology have developed deep learning models that can solve the so-called inverse problems of finding the right geometries to give rise to the desired properties.
“Even when inverse problems were solved in the past, they have been limited by the simplifying assumption that the small-scale geometry can be made from an infinite number of building blocks,” explained Helda Pahlavani, first author of the published research in a press release. “The problem with that assumption is that metamaterials are usually made by 3D printing, and real 3D printers have a limited resolution, which limits the number of building blocks that fit within a given device.”
With these new deep learning models, engineers can start with the number of building blocks accommodated by their respective manufacturing technique and then autonomously find the geometry that will yield the desired metamaterial properties based on their actual manufacturing capabilities. The result is autonomous discovery of fabrication-ready metamaterials.
Metamaterial Durability
According to the Delft engineers, the problem of metamaterial durability has also been overlooked in previous research. For this reason, most existing designs break after only a few uses.
“So far, it has been only about what properties can be achieved,” said Amir Zadpoor, co-author and professor in the Department of Biomechanical Engineering at Delft. “Our study considers durability and selects the more durable designs from a large pool of design candidates. This makes our designs really practical and not just theoretical adventures.”
Despite their huge promise, the practical constraints of discovering metamaterials as well as implementing them has relegated them largely to academic research. Thanks to AI, that could be changing. “We think the step we have taken is revolutionary in the field of metamaterials,” said corresponding author Mohammad J. Mirzaali. “It could lead to all kinds of new applications.”
Beyond those listed above, examples of new metamaterial applications include orthopaedic implants, surgical instruments, soft robots, adaptive mirrors and exo-suits. If this tool lives up to its promise, AI could be the key to unlocking metamaterials in all these applications, and more.
The research is published in Advanced Materials.
