Machine intelligence accelerated design of conductive MXene aerogels with programmable properties

Area of Science:

• Materials Science
• Robotics
• Artificial Intelligence

Background:

• Designing ultralight conductive aerogels with specific electrical and mechanical properties is challenging due to the vast experimental parameter space.
• Conventional methods are iterative and time-consuming, hindering rapid material development.

Purpose of the Study:

• To develop an integrated workflow combining collaborative robotics and machine learning to accelerate the design of conductive aerogels.
• To create a predictive model for aerogel properties and enable inverse design for specific requirements.

Main Methods:

• An automated pipetting robot prepared 264 unique Ti3C2Tx MXene-based aerogel formulations.
• Support vector machine and artificial neural network models were trained using active learning and data augmentation.
• Robotics-automated platforms were used for fabrication and characterization of 162 unique aerogels.

Main Results:

• An artificial neural network model was constructed to predict aerogel properties and perform inverse design.
• Model interpretation and finite element simulations confirmed a strong correlation between aerogel density and compressive strength.
• The workflow successfully designed aerogels with high conductivity, tailored strength, and pressure insensitivity.

Conclusions:

• The integrated robotics and machine learning workflow significantly accelerates the design and discovery of conductive aerogels.
• The developed predictive model enables efficient inverse design for specific material property requirements.
• The optimized aerogels are suitable for applications like compression-stable Joule heating in wearable thermal management systems.