Structure Analysis of Polymer Pyrolyzate by Pyrolysis GC-TOFMS and AI -A Case Study of AI Library Containing in silico Pyrolyzate

Introduction

We have developed msFineAnalysis AI which has an automated structural analysis method ^[1] (hereafter referred to as AI structural analysis) that combines data from a time-of-flight mass spectrometer (TOFMS), a high mass resolution GC-MS, and mass spectrum prediction by deep learning, as a solution for unknown compound analysis. This software contains a database (DB) of structural formulas and predicted EI mass spectra of approximately 100 million compounds obtained from the PubChem DB (hereinafter referred to as the AI library). By utilizing the molecular formula information determined by integrated analysis ^[2], it is possible to quickly estimate the structural formula of an unknown compound.
The pyrolysis GC-MS method is a widely utilized technique for the analysis of solid polymer samples, such as resins. A significant number of the pyrolyzates observed in pyrolysis GC-MS have not been registered in commercially available electron ionization (EI) mass spectrum databases. We attempted an analysis using the AI library and confirmed that the structural analysis of unknown compounds in acrylic resin can be performed with high accuracy (MSTips No.389). However, some oligomer components, such as trimers, lacked a correct structural formula in the AI library. This may be attributed to the limited number of structural formulas registered in the PubChem DB for polymer pyrolyzates.
Consequently, a novel AI library was developed, encompassing predicted EI mass spectra of approximately 25 million pyrolyzates. This library was derived from the exhaustive generation of pyrolyzate structure formulas by in silico. In 2024, we have released msFineAnalysis AI ver. 2 with a new AI library for 120 million compounds, including polymer pyrolyzates.
In this MSTips, we report a case study of analysis using the AI library including in silico pyrolyzates equipped with msFineAnalysis AI ver. 2.

Experiment

Commercially available acrylic resin was used as the test sample. The sample pretreatment system was an EGA/PY-3030D pyrolyzer (Frontier Labs Inc.), GC-MS was performed using a JMS-T2000GC (JEOL Ltd.), and an EI/FI combination ion source was used. The obtained data were analyzed using msFineAnalysis AI ver. 2 (JEOL Ltd.). For details of the measurement conditions, refer to MSTips No.389.

in silico Pyrolyzate　

The structural formulas for the in silico pyrolyzates were prepared by the following procedure. The homopolymer and copolymer species prepared for the AI library are shown in Table 1.
（1）49 monomer compounds are selected ^[3].
（2）Generate hexamer linear chain structural formulas of 49 homopolymers and 18 copolymers.
（3）Generate the structural formula by randomly cleaving 1 to 5 single bonds in the structural formula of (2).
（4）Hydrogen addition or double bond substitution at the cleavage position in the structural formula of (3).
（5）Perform the processes described in (2) and (3) for all patterns to create the pyrolyzates structural formula.
After all the steps, we got structure formulas for about 25 million pyrolyzates. These were put into the developed deep learning model, and the EI mass spectra of about 25 million compounds were predicted.

Table 1　Polymer list of the predicted EI mass spectrum DB for 25 million in silico pyrolyzates

Result and Discussion

TIC chromatograms obtained from pyrolysis GC-TOFMS measurements are shown in Figure 1. In this study, 48 components of the observed acrylic resin pyrolyzates were analyzed, which were detected in the trimer with a retention time of 14-20 minutes.

Fig.1　Py-GC-EI and FI TIC chromatograms for an acrylic resin.

A total of 48 compounds were detected in the trimer region. However, only two of these were registered in the NIST23 DB. The remaining 46 compounds were analyzed in the AI library, of which 26 were from the PubChem DB and 20 corresponded to the in silico pyrolyzates created in this study. It was possible to deduce the structural formulae of approximately 40% of the trimer region compounds from the in silico pyrolyzates information.

The analysis result window in msFineAnalysis AI including the mass spectra and estimated structural formula of ID:045 observed at 18.63 min is shown in Figure 2. This component was hardly observed a molecular ion by the EI method, but by the FI method, m/z 300.15728, which is considered to be a molecular ion, was detected as the base peak. The molecular formula was estimated to be C₁₅H₂₄O₆ with high mass accuracy. The molecular formula information is used as a filter in AI structure analysis to narrow down the number of candidates. The number of candidate structural formulas was significantly reduced from the AI library of approximately 120 million compounds. The highest scoring structural formula among the 7529 candidates was the in silico pyrolyzates that can arise from methyl polyacrylate.

Fig.2 msFineAnalysis AI result window,
ID: 045 mass spectra and estimated structure by in silico pyrolyzates

The estimated structural formulas of the compounds that corresponded to in silico pyrolyzates are shown in Figure 3. We were able to confirm the trimer component reflecting the structure of the acrylic resin measured in this study.

Fig.3　Estimated structures for 20 compounds in the trimer region
using AI structure analysis and in silico pyrolyzates information

Conclusion

In this MSTips, we introduced a case study of structural analysis in the trimer region of acrylic resin, where most of the detected components are not registered in the NIST DB. msFineAnalysis AI ver.2 newly includes predicted EI mass spectra of 25 million pyrolyzates, enabling more accurate pyrolysis GC -MS qualitative analysis of solid polymer samples. It is expected to be used for various pyrolysis GC-MS qualitative analysis application in the future.

Reference

[1] A. Kubo et al, Mass Spectrometry, 2023, 12, A0120.
[2] M. Ubukata et al, Rapid Commun Mass Spectrom., 34 (2020 ). DOI: 10.1002/rcm.8820
[3] Shin Tsuge, Hajime Ohtani, Chuichi Watanabe (2011), Pyrolysis - GC/MS Data Book of Synthetic Polymers, Elsevier

Solutions by field

Plastics / Polymer

Related information

JMS-T2000GC AccuTOF™ GC-AlphaHigh Performance Gas Chromatograph- Time-of-Flight Mass Spectrometer

msFineAnalysis AI Ver.2Unknown CompoundsStructure Analysis Software