Machine intelligence-accelerated discovery of all-natural plastic substitutes

Area of Science:

• Materials Science
• Polymer Science
• Sustainable Chemistry

Background:

• Petrochemical plastics accumulate in the environment, necessitating biodegradable alternatives.
• Developing all-natural substitutes with properties like transparency and fire retardancy is challenging.
• Current discovery methods rely on time-consuming iterative optimization.

Purpose of the Study:

• To accelerate the discovery of all-natural biodegradable plastic substitutes.
• To create a workflow integrating robotics and machine learning for material design.
• To achieve programmable optical, thermal, and mechanical properties in eco-friendly materials.

Main Methods:

• An automated pipetting robot prepared 286 nanocomposite films for training a support-vector machine classifier.
• 135 all-natural nanocomposites were fabricated through 14 active learning loops with data augmentation.
• An artificial neural network prediction model was established for property prediction and inverse design.

Main Results:

• The prediction model successfully predicted physicochemical properties from nanocomposite composition.
• The model automated the inverse design of biodegradable plastic substitutes meeting user-specific requirements.
• Several all-natural substitutes with properties analogous to non-biodegradable plastics were prepared.

Conclusions:

• An integrated workflow combining robotics, machine learning, and simulation accelerates the discovery of eco-friendly plastic substitutes.
• This methodology enables the design of all-natural materials with tunable properties.
• The approach utilizes generally-recognized-as-safe building blocks for sustainable material development.