Partial least squares regression as a powerful tool for investigating large combinatorial polymer libraries.

Area of Science:

• Surface Science
• Materials Chemistry
• Chemometrics

Background:

• Partial Least Squares (PLS) regression is a multivariate statistical method.
• It is commonly used to correlate complex datasets, such as Time-of-Flight Secondary Ion Mass Spectrometry (ToF-SIMS) data, with specific material properties.
• Quantitative prediction of surface energy using PLS in surface analysis remains underexplored.

Purpose of the Study:

• To construct and evaluate a PLS model for predicting polymer surface energy from ToF-SIMS data.
• To investigate the impact of training set size and composition on PLS model performance and influential ion identification.
• To assess the predictive capability of the PLS model for polymers both within and outside the initial training set.

Main Methods:

• Development of a PLS regression model using ToF-SIMS and surface energy data from a library of 496 micro-patterned copolymers.
• Analysis of regression vector changes with varying numbers of training samples.
• Validation of the PLS model's predictive accuracy on synthesized and commercial polymers with diverse chemical compositions.

Main Results:

• The identity of influential ions in the PLS regression vector is sensitive to the training set composition.
• The PLS model systematically underestimated surface energy for copolymers with common monomers but different compositions.
• Predictive performance was poor for copolymers synthesized from novel monomers not present in the training set.

Conclusions:

• PLS regression can potentially predict surface energy for polymers composed of monomers represented in the training set.
• The accuracy of PLS predictions is highly dependent on the chemical relevance of the training data.
• A representative training set is essential for reliable quantitative predictions of polymer surface energy using PLS.