Work

An NMR-Informed Multimodal–Multitask Data Science Framework for Multilevel Decoding of Biopolymer Degradation

Year: 2026
Type: dissertation
Abstract: The transition toward a sustainable society has increased the demand for biodegradable polymers that can serve as environmentally friendly materials while maintaining sufficient performance during use.For these materials, degradation is not only an end-of-life property, but also a design factor that must be considered together with mechanical stability, durability, and environmental compatibility.Understanding their behavior under realistic aquatic conditions is therefore important for sustainable polymer development.Biodegradable polymers have often been evaluated using endpoint mass loss, chemical composition, or individual material properties.Although these indicators are useful, they do not fully describe degradation in natural environments, where molecular structure, higher-order structure, segmental mobility, thermal behavior, mechanical properties, water exposure, and microbial activity are closely related.A single-parameter evaluation is therefore insufficient to explain why different polymers show distinct degradation behaviors under similar environmental conditions.Nuclear magnetic resonance (NMR) is well suited to this problem because it provides information on both chemical structure and molecular dynamics.Time-domain NMR (TD-NMR) reflects segmental mobility, phase heterogeneity, and relaxation behavior, whereas solution-state NMR provides information on local chemical environments and polymer microstructures.In this thesis, NMR data were analyzed together with data science methods to describe material states and their changes during degradation.This thesis examined biodegradable polymer degradation from three perspectives: time-dependent changes, relationships among multiple material descriptors, and isotope-resolved molecular tracing.The first study introduced a time-series perspective to clarify how structure-property relationships change during degradation.The second study used multimodal-multitask learning to relate polymer structure, molecular dynamics, thermal properties, and mechanical performance to degradability.The third study applied 13 C labeling to examine polymer-microbe interactions and molecular changes in an aquatic microcosm.In the first study, biodegradable polyesters were exposed to aquatic environmental conditions and analyzed at multiple degradation stages.TD-NMR, solution-state NMR, mechanical testing, thermal analysis, and machine learning showed that the factors associated with degradation changed over time.Early-stage behavior was mainly related to initial mechanical properties and material-state descriptors, whereas later stages showed stronger contributions from thermal parameters and NMRderived structural and dynamic features.These results indicate that degradation
Institution: Nagoya University
Language: English
Subfield: Biomaterials
Open Access status: green
OpenAlex ID: W7221244178
Cites: 123
Cited by:
Related to: 10