Rapid Copolymer Analysis of Unresolved Mass Spectra by Artificial Intelligence

Area of Science:

• Polymer Chemistry
• Analytical Chemistry
• Spectroscopy

Background:

• Low-resolution mass spectrometry, particularly time-of-flight (TOF), presents challenges in determining copolymer composition due to overlapping peaks at higher molecular weights.
• Accurate characterization of copolymer structure is crucial for understanding material properties and performance.

Purpose of the Study:

• To develop a novel, simplified data analysis method for determining copolymer composition from low-resolution mass spectra.
• To extend the accessible molecular weight range for reliable copolymer analysis.
• To extract hidden information from unresolved mass spectral peaks.

Main Methods:

• Development of regression relationships between spectral parameters and copolymer characteristics (mole fraction, repeat unit count, monomer distribution).
• Construction of two regression models: one using conventional statistics and another using an Artificial Neural Network (ANN).
• Demonstration and validation using poly(N-acryloylmorpholine)-block-poly(N-isopropylacrylamide) (PNAM-b-PNIPAM) diblock copolymers.

Main Results:

• The novel method successfully determines copolymer composition from low-resolution mass spectra, even in higher mass regions.
• The approach effectively extracts information from overlapping peaks, overcoming limitations of traditional analysis.
• Validated against experimental data and a computationally intensive method, showing comparable accuracy with reduced complexity.

Conclusions:

• The presented data analysis method offers a simplified and effective approach for copolymer composition determination using low-resolution mass spectrometry.
• This technique enhances the ability to analyze complex polymer architectures, particularly at higher molecular weights.
• The use of both statistical and machine learning models provides flexibility and robustness for diverse applications.