Skip to main content
2025 AOCS Annual Meeting & Expo Main banner
Amphistar Ad
Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.
Back

2025 AOCS Annual Meeting & Expo.

Surfactants and Detergents

Modeling, ML & AI in Surfactants

Predicting Structure-Property Relationships for Non-Ionic Surfactants using AI and Molecular Modeling Techniques

Wednesday, April 30, 2025
11:15 AM – 11:35 AM PST
Room: B110-112

    Presenting Author(s)

  • Sarvesh Agrawal, PhD

    Customer and Application Development Scientist
    ExxonMobil Technology and Engineering Company
    Spring, Texas, United States

  • BW

    Brian Willett

    Research Scientist
    NobleAI, United States

    • Co-Author(s)

  • FS

    Francois Simal

    CI and LAO Product Principal
    ExxonMobil, Belgium

  • AJ

    Arben Jusufi

    Principal Scientist
    ExxonMobil Technology and Engineering Company, New Jersey, United States

  • CL

    Chunzhao Li

    Sustainability Principal Scientist
    ExxonMobil, United States

  • Austin Dulaney

    Lead Research Scientist
    NobleAI, California, United States

  • Kevin Shen

    Research Scientist
    Noble AI Inc, California, United States

  • AG

    Alicja Gos

    Solutions Engineer
    NobleAI
    San Jose, California, United States

  • KF

    Kyle Fujdala, PhD (he/him/his)

    Sr. Dir. Chemical Sciences
    NobleAI, United States

Abstract:

This study explores the structure-property relationship for commercially produced non-ionic surfactants, over a wide range of conditions, relevant for cleaning end-uses. Major properties explored are critical micelle concentration (CMC), dynamic surface tension, and phase behavior for non-ionic surfactants. We investigate the impact of structural parameters, such as polydispersity, length of hydrophobe and hydrophile, hydrophobe branching, and free alcohol content, on surfactant properties. Utilizing a combination of artificial intelligence and physics-informed methods, including probabilistic modeling, graph neural networks, and transformer-based chemical embeddings, alongside molecular modeling approaches, we attempt to predict surfactant properties with high accuracy. Our analysis compares branched, semi-linear, and linear industrial non-ionic surfactants, as well as their mixtures. The model is initially trained using single-component model surfactants from existing literature and subsequently expanded to incorporate the behavior of commercially produced surfactants. This requires modeling the behavior of complex mixtures of multiple molecular structures relevant to commercially produced surfactants, including the interactions between these different structures, which has not been considered previously in the ML literature. This study’s AI and molecular modeling techniques provide a robust framework for optimizing commercially produced surfactants, paving the way for innovations in next gen products.