Machine intelligence accelerated design of conductive MXene aerogels with programmable properties

Authors: SHRESTHA, SNEHI - University Of Maryland; BARVENIK, KIERAN - University Of Maryland; CHEN, TIANLE - University Of Maryland; YANG, HAOCHEN - University Of Maryland; KESAVAN, MEERA - University Of Maryland; LITTLE, JOSHUA - University Of Maryland; TENG, ZI - Orise Fellow; Luo, Yaguang - Sunny; TUBALDI, ELEONORA - University Of Maryland; CHEN, PO-YEN - University Of Maryland

Submitted to: Nature Communications
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 5/14/2024
Publication Date: 6/1/2024
Citation: Shrestha, S., Barvenik, K.J., Chen, T., Yang, H., Kesavan, M.M., Little, J.M., Teng, Z., Luo, Y., Tubaldi, E., Chen, P. 2024. Machine intelligence accelerated design of conductive MXene aerogels with programmable properties. Nature Communications. 15. Article e4685. https://doi.org/10.1038/s41467-024-49011-8.
DOI: https://doi.org/10.1038/s41467-024-49011-8

Interpretive Summary: In agriculture and food research, it usually takes extensive screening through numerous formulas, treatments, and conditions to achieve optimal performance in each application. As more ingredients or treatments are introduced to the trial, the time expense on product screening increases exponentially, thus, adding tremendous workload and greatly increasing the risk of human error. Herein, scientists from USDA and University of Maryland demonstrated how robotic- and machine learning-empowered strategies significantly increase the efficiency and precision of formula and process screening. As an example, hundreds of aerogel samples (used for insulation and plant cultivation) were prepared and tested continuously without human intervention. Thereafter, a machine learning technology was applied to accurately predict and validate the optimal formula even if it has not been previously tested. This innovative technology is envisioned to greatly accelerate formula discovery and process optimization for various applications in agriculture and food industry.

Technical Abstract: Designing ultralight conductive aerogels with tailored electrical and mechanical properties is critical for various applications. Conventional approaches rely on iterative optimization experiments, which are time-consuming when exploring a vast parameter space. Herein, an integrated workflow is developed to combine collaborative robotics with machine learning to accelerate the design of conductive aerogels with programmable electrical and mechanical properties. First, an automated pipetting robot is operated to prepare 264 mixtures using four building blocks at different ratios/loadings (including Ti3C2Tx MXene, cellulose nanofibers, gelatin, glutaraldehyde). After freeze-drying, the structural integrity of conductive aerogels is evaluated to train a support vector-machine classifier. Through 8 active learning cycles with data augmentation, 162 kinds of conductive aerogels are fabricated/characterized via robotics automated platforms, enabling the construction of an artificial neural network prediction model. The prediction model can conduct two-way design tasks: (1) predicting the physicochemical properties of conductive aerogels from fabrication parameters and (2) automating the inverse design of conductive aerogels for specific property requirements. The combined use of model interpretation and finite element simulations validates a pronounced correlation between aerogel density and its compressive strength. The model-suggested conductive aerogels with high electrical conductivity, customized compression resilience, and pressure insensitivity allow for compression-stable Joule heating for wearable thermal management. The fusion of robotics accelerated experimentation, machine intelligence, and simulation tools expedites the tailored design and scalable production of conductive aerogels.