Saroj UpretiSelf-driving labs for materials discovery
Control room

A robotic platform for mapping copolymer phase behavior

Hundreds of thin-film samples fabricated in hours, measured by synchrotron scattering, and turned into the first high-throughput map of the order–disorder transition landscape in triblock copolymer–homopolymer blends.

Headline result
100sof thin-film samples fabricated and characterized per day
Period
2023 – 2026
With
Lawrence Berkeley National Laboratory · Brookhaven National Laboratory
  • High-throughput experimentation
  • Robotics
  • SAXS / GISAXS
  • Phase behavior
  • Open data

Throughput first

A closed-loop optimizer is only as good as the throughput underneath it. I led development of the polymer processing robots that fabricate hundreds of thin-film samples within hours — and, just as important, make them the same way every time, so variation across a library reflects the composition rather than the hand.

The formulation and coating robot (NOVA) and its modules. One arm moves vials, pipettes and wafers between stations, so a formulation becomes a coated film automatically.Adapted from Upreti et al., ACS Nano 20, 17303 (2026), Fig. 2A, under CC BY 4.0.
A sample library being fabricated: dispense, mix, coat, rack, repeat. Each well becomes one point in the composition–processing space; a full library becomes one region of a phase map. Repetitive segments sped up.

Mapping the order–disorder transition

The target was the phase behavior of ABA triblock copolymer–homopolymer blends. The order–disorder boundary is easy to step over with a coarse grid; robotic fabrication feeding SAXS/WAXS at the ALS and NSLS-II, with high-throughput AFM for real-space confirmation, sampled it densely enough to resolve it. A reduction pipeline built on Bluesky and Tiled kept pace with fabrication.

(A) GISAXS patterns from ordered to disordered films. (B) The order–disorder boundary across homopolymer size and blend ratio, coloured by peak width. (C) The 1D profiles it was read from. The boundary region is where the physics is, and it is precisely what a low-resolution grid cannot see.Adapted from Upreti et al., ACS Nano 20, 17303 (2026), Fig. 3, under CC BY 4.0.
Real space agrees with reciprocal space: AFM phase images across the same grid, with the GISAXS order–disorder boundary drawn in red.Adapted from Upreti et al., ACS Nano 20, 17303 (2026), Fig. 5, under CC BY 4.0.

Six chemistries, one database

PS-P2VPPS-PEOPI-PS-P2VPPS-P4VPPB-PS-PBPI-PS-PI

The data is published openly at polymermorphology.org and became the training set for the machine learning framework and the groundwork for the closed-loop work.

Published in