Machine intelligence accelerated design of conductive MXene aerogels with programmable properties

Designing ultralight conductive aerogels with tailored electrical and mechanical properties is critical for various applications but typically relies on iterative, time-consuming optimization across a large parameter space. This work integrates collaborative robotics with machine learning to accelerate the design of MXene-based conductive aerogels with programmable properties. An automated pipetting robot prepared 264 mixtures of Ti3C2Tx MXene, cellulose, gelatin, and glutaraldehyde at different ratios and loadings. After freeze-drying, aerogel structural integrity was used to train a support vector machine (SVM) classifier that defined a feasible fabrication space. Through 8 active learning cycles with data augmentation, 162 unique conductive aerogels were fabricated and characterized using robot-assisted platforms, enabling construction of an artificial neural network (ANN) prediction model. The model supports two-way design: predicting physicochemical properties from fabrication parameters and performing inverse design to meet target properties. Model interpretation and finite element simulations validate a strong correlation between aerogel density (via mixture loading) and compressive strength. Model-suggested aerogels exhibit high conductivity, customized strength, and pressure insensitivity, enabling compression-stable Joule heating for wearable thermal management.

Snehi Shrestha, Kieran James Barvenik, Tianle Chen, Haochen Yang, Yang Li, Meera Muthachi Kesavan, Joshua M. Little, Hayden C. Whitley, Zi Teng, Yaguang Luo, Eleonora Tubaldi, Po-Yen Chen