Autonomous phase diagram mapping of polymer blends through SAXS-guided active learning

Autonomous phase behavior mapping from multicomponent copolymer formulation requires feedback that connects scattering measurements to experimental decisions. We couple robotic polymer formulation, automated small-angle X-ray scattering (SAXS) analysis, and Gaussian-process active learning using structure factor to autonomously map complex phase behavior of multiblock copolymer/homopolymer blends. This dimensionless descriptor, based on secondary and primary scattering peak ratio q2/q1, captures relative reciprocal-space geometry while removing a common structural length scale when the same reflections are tracked. An ordering model identifies compositions with sharp scattering features, and a conditional regressor learns the ratio surface. An acquisition function weighted by ordering probability combines predicted ratio gradients with length-scale-adjusted uncertainty to direct measurements towards candidate structural transitions. Ternary blends of poly(1,4-isoprene-block-styrene-block-2-vinylpyridine), polystyrene and poly(2-vinylpyridine) provide the experimental test system. Starting from 15 initial formulations, the algorithm progressively refines an irregular phase map over 10 additional rounds with 5 experiments for each additional round. Across iterations 9-11, the mean absolute change in predicted q2/q1 between successive maps decreases from 0.044 to 0.012 over their common predicted ordered region, while mean predictive uncertainty remains approximately 0.08, indicating Gaussian-process active learning captured the unknown phase map of ternary blend system. The active learning uses only 1.26% of the formulations in a complete ternary grid at 1% composition increments. The resulting phase maps reveal regions consistent with BCC-packed spheres, hexagonally packed cylinders, and lamellae. Weakly ordered structures appear at high PS loading, while boundary samples show signatures of coexistence or network morphologies. The workflow is adaptable to other periodically ordered materials by using resolved scattering-peak ratios as feedback and training the model on measurements from each system. This will enable autonomous phase mapping for ordered material with different structure factors before every region has a definitive morphology label, while retaining the evidence needed for materials-specific interpretation.

Authors

Institutions

Publication Details

Journal
ChemRxiv
Published
2026-11-16
DOI
https://doi.org/10.26434/chemrxiv.15012310/v1
Primary Topic
Block Copolymer Self-Assembly
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Autonomous phase diagram mapping of polymer blends through SAXS-guided active learning

Marcus Michael Noack, Boran Ma, Xiaodan Gu, Saroj Upreti et al.
ChemRxiv
Block Copolymer Self-Assembly
preprint

Autonomous phase diagram mapping of polymer blends through SAXS-guided active learning

Marcus Michael Noack, Boran Ma, Xiaodan Gu, Saroj Upreti, Yunfei Wang, Sabin Baral, Kailash Adhikari, Chenhui Zhu
preprint en

Abstract

Autonomous phase behavior mapping from multicomponent copolymer formulation requires feedback that connects scattering measurements to experimental decisions. We couple robotic polymer formulation, automated small-angle X-ray scattering (SAXS) analysis, and Gaussian-process active learning using structure factor to autonomously map complex phase behavior of multiblock copolymer/homopolymer blends. This dimensionless descriptor, based on secondary and primary scattering peak ratio q2/q1, captures relative reciprocal-space geometry while removing a common structural length scale when the same reflections are tracked. An ordering model identifies compositions with sharp scattering features, and a conditional regressor learns the ratio surface. An acquisition function weighted by ordering probability combines predicted ratio gradients with length-scale-adjusted uncertainty to direct measurements towards candidate structural transitions. Ternary blends of poly(1,4-isoprene-block-styrene-block-2-vinylpyridine), polystyrene and poly(2-vinylpyridine) provide the experimental test system. Starting from 15 initial formulations, the algorithm progressively refines an irregular phase map over 10 additional rounds with 5 experiments for each additional round. Across iterations 9-11, the mean absolute change in predicted q2/q1 between successive maps decreases from 0.044 to 0.012 over their common predicted ordered region, while mean predictive uncertainty remains approximately 0.08, indicating Gaussian-process active learning captured the unknown phase map of ternary blend system. The active learning uses only 1.26% of the formulations in a complete ternary grid at 1% composition increments. The resulting phase maps reveal regions consistent with BCC-packed spheres, hexagonally packed cylinders, and lamellae. Weakly ordered structures appear at high PS loading, while boundary samples show signatures of coexistence or network morphologies. The workflow is adaptable to other periodically ordered materials by using resolved scattering-peak ratios as feedback and training the model on measurements from each system. This will enable autonomous phase mapping for ordered material with different structure factors before every region has a definitive morphology label, while retaining the evidence needed for materials-specific interpretation.

ChemRxiv
Lawrence Berkeley National Laboratory (US), Advanced Light Source (US), University of Southern Mississippi (US)
Block Copolymer Self-Assembly
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon