Developing and Validating a High-Throughput Robotic System for the Accelerated Development of Porous Membranes

TL;DR

The paper addresses the slow, variable development of porous polymeric membranes fabricated by nonsolvent-induced phase separation (NIPS) and proposes a fully automated, high-throughput self-driving laboratory (SDL) platform that integrates solution preparation, blade casting, controlled immersion, and compression-based mechanical testing. By extracting elastic modulus $E$ and pore fraction $\phi$ from automated stress-strain curves, the approach provides rapid structural and mechanical proxies for membrane performance, demonstrated on polysulfone with PolarClean and water. The results reproduce established relationships: higher polymer concentration yields stiffer, denser membranes, while ambient humidity alters pore morphology and mechanics; dry nitrogen pretreatment can produce macrovoids with unexpectedly high stiffness. The platform delivers throughput under one hour per sample and enables data-driven optimization and future active-learning loops for autonomous membrane discovery and design.

Abstract

The development of porous polymeric membranes remains a labor-intensive process, often requiring extensive trial and error to identify optimal fabrication parameters. In this study, we present a fully automated platform for membrane fabrication and characterization via nonsolvent-induced phase separation (NIPS). The system integrates automated solution preparation, blade casting, controlled immersion, and compression testing, allowing precise control over fabrication parameters such as polymer concentration and ambient humidity. The modular design allows parallel processing and reproducible handling of samples, reducing experimental time and increasing consistency. Compression testing is introduced as a sensitive mechanical characterization method for estimating membrane stiffness and as a proxy to infer porosity and intra-sample uniformity through automated analysis of stress-strain curves. As a proof of concept to demonstrate the effectiveness of the system, NIPS was carried out with polysulfone, the green solvent PolarClean, and water as the polymer, solvent, and nonsolvent, respectively. Experiments conducted with the automated system reproduced expected effects of polymer concentration and ambient humidity on membrane properties, namely increased stiffness and uniformity with increasing polymer concentration and humidity variations in pore morphology and mechanical response. The developed automated platform supports high-throughput experimentation and is well-suited for integration into self-driving laboratory workflows, offering a scalable and reproducible foundation for data-driven optimization of porous polymeric membranes through NIPS.

Figures (8)

• Figure 1: Planned workflow for a NIPS-based self-driving lab (SDL). This workflow demonstrates an iterative process between a liquid handling robot, a temperature-controlled coagulation bath, and a compression tester, interconnected by a robotic arm. The liquid handling robot transfers and mixes the precursor solutions. The robotic arm performs blade casting and sample transferring between the different modules. NIPS takes place in the coagulation bath. Once a membrane is fabricated, the compression tester performs a series of tests on the sample to characterize its intra-sample consistency and mechanical properties. Upon post-processing and analysis, the next set of experiments can be initiated and operated iteratively.
• Figure 2: Experimental setup used for automated membrane fabrication and mechanical testing. (A) Overall system in a ductless fume hood. (B) The liquid handling robot is configured for polymer solution preparation. It uses pipette tips to transfer and mix the precursor solutions into the vials. The well plate can be heated to achieve more homogeneous mixing. The nitrogen blower introduces a laminar dry nitrogen flow over the cast film to reduce the humidity. The blade cleaning stage holds and cleans the doctor blade after each blade casting. (C) The temperature-controlled coagulation bath uses a copper coil that is cooled by a circulating chiller to achieve temperature control during phase separation. (D) The compression tester uses a 5 mm diameter flat punch to characterize the mechanical properties of the fabricated membranes. It is equipped with a load cell for force measurement and an external linear variable differential transformer (LVDT) for precise small displacement measurement.
• Figure 3: Demonstration of the automated stress-strain curve analysis. (A) Automatic segmentation of a stress-strain curve (engineering stress and engineering strain) using a designed pipeline in Python to identify key mechanical regions: elastic (blue), plateau (orange), densification (green), and creep (red). This segmentation enables estimation of the elastic modulus and pore fraction of the membrane, which for this sample are 166.1 bar and 0.57, respectively. Inset SEM images show the microstructure of a membrane before and after compression. (B) Data-processing pipeline for mechanical property extraction, consisting of four key stages: (1) Data Preprocessing, including unit conversion and curve alignment; (2) Curve Segmentation, where breakpoints are computed using derivatives and piecewise regression; (3) Property Extraction, determining elastic modulus, yield strength, densification, and creep properties; and (4) Quality Assessment, evaluating the quality and intra-sample consistency through statistical metrics such as the coefficient of variance (CV) and fitting success criteria.
• Figure 4: Intra-sample consistency analysis using the coefficient of variance (CV). A) Box plot showing the CV of stress-strain curves across three fabrication methods: 1) Manual Premixed (manual solution preparation and fabrication), 2) Auto Premixed (manual solution preparation with automated fabrication), and 3) Auto Mixed (fully automated mixing and fabrication). The horizontal line inside each box marks the median CV; the box itself spans the inter-quartile range (IQR) (25 %--75 %); the “whiskers” extend to 1.5 × IQR; points beyond the whiskers are outliers. Each dot represents a sample, color-coded by polymer concentration (wt%). The number of samples (n) for each category is specified in parentheses. Examples of stress-strain curves with B) a low CV obtained from a high polymer concentration sample and C) a high CV obtained from a low polymer concentration sample.
• Figure 5: Properties extracted from the stress-strain curves using the automated curve segmentation algorithm. (A) Elastic Modulus vs. Polymer Concentration and (B) Pore Fraction vs. Polymer Concentration across different humidity conditions: 1) $RH \geq 49\%$ (blue), 2) $RH < 49\%$ (orange), and 3) with dry nitrogen exposure (green). RH refers to the ambient humidity for experiments that did not use dry nitrogen during NIPS. Lines of best fit are drawn as a guide to the eye.