Interpretable Processing–Structure Relationships in Spirulina Bioplastics through Automated Spectral and Microstructural Analysis
Abstract
Algal bioplastics are emerging as a promising class of sustainable materials, but their development is limited by an incomplete understanding of how processing governs structure. Establishing principles connecting processing to structure is hindered by manual analysis and the complexity of structural data. Here thermomechanical processing conditions (time, temperature, and pressing force) are mapped to structural descriptors of Limnospira platensis (spirulina) bioplastics. Quantifiable features are derived from Fourier transform infrared spectroscopy (FTIR) and scanning electron microscopy (SEM) to summarize the chemical and physical structure. Informative FTIR regions are selected by principal component regression (PCR), and interpretable Gaussian bands are automatically deconvolved from the spectra. SEM micrographs are encoded using a self-supervised vision transformer (DINOv3), projected into two dimensions, and clustered. Overall, 201 features are combined into a structural fingerprint and fed to a gradient boosted model that is inverted to recover processing conditions. The inverted model reveals that temperature strongly influences the structural features, while pressing time acts as a secondary logarithmic modulator. Conversely, pressing force is recoverable only as a regime, behaving as a consolidation threshold near 5 kN. Together, these results demonstrate that complex structural signatures can be transformed into interpretable processing–structure relationships, enabling rational design of algal bioplastics.
Related
- Processing–structure–property relationships of Y2O3- and ZnO-modified MgO–Al2O3–SiO2 glasses for high-modulus fiber drawing
- Unlocking Nanocomposites via Cast Film Extrusion: Processing–Structure–Property Relationships of PLA Sheets Containing Silver Nanoparticles
- Structure–Processing–Property Relationships in PVOH/PBAT Blends Plasticized with Dipentaerythritol, ε-Caprolactone, and Water
- Development of biodegradable bioplastics from pineapple bromelain processing waste: Utilization of decanter and fiber flours
- Enhanced thermoelectric performance of Bi 2 Te 2.7 Se 0.3 through microstructural engineering and powder processing design
BibTeX
@misc{liao2026interpretable,
title = {{Interpretable Processing–Structure Relationships in Spirulina Bioplastics through Automated Spectral and Microstructural Analysis}},
author = {Kuotian Liao and Yiyang Sun and Ian R. Campbell and Hareesh Iyer and Richard Sheridan and Arthur Gim and Etta Shack and Sebastian Sosa and Thea Zabala and Linda S. Schadler and L.C. Brinson and Cynthia Rudin and Eleftheria Roumeli},
howpublished = {ChemRxiv},
year = {2026},
month = sep,
doi = {10.26434/chemrxiv.15008361/v1},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15008361/v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Interpretable Processing–Structure Relationships in Spirulina Bioplastics through Automated Spectral and Microstructural Analysis
AU - Kuotian Liao
AU - Yiyang Sun
AU - Ian R. Campbell
AU - Hareesh Iyer
AU - Richard Sheridan
AU - Arthur Gim
AU - Etta Shack
AU - Sebastian Sosa
AU - Thea Zabala
AU - Linda S. Schadler
AU - L.C. Brinson
AU - Cynthia Rudin
AU - Eleftheria Roumeli
PY - 2026
DA - 2026/09/06
PB - ChemRxiv
DO - 10.26434/chemrxiv.15008361/v1
UR - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15008361/v1
AB - Algal bioplastics are emerging as a promising class of sustainable materials, but their development is limited by an incomplete understanding of how processing governs structure. Establishing principles connecting processing to structure is hindered by manual analysis and the complexity of structural data. Here thermomechanical processing conditions (time, temperature, and pressing force) are mapped to structural descriptors of Limnospira platensis (spirulina) bioplastics. Quantifiable features are derived from Fourier transform infrared spectroscopy (FTIR) and scanning electron microscopy (SEM) to summarize the chemical and physical structure. Informative FTIR regions are selected by principal component regression (PCR), and interpretable Gaussian bands are automatically deconvolved from the spectra. SEM micrographs are encoded using a self-supervised vision transformer (DINOv3), projected into two dimensions, and clustered. Overall, 201 features are combined into a structural fingerprint and fed to a gradient boosted model that is inverted to recover processing conditions. The inverted model reveals that temperature strongly influences the structural features, while pressing time acts as a secondary logarithmic modulator. Conversely, pressing force is recoverable only as a regime, behaving as a consolidation threshold near 5 kN. Together, these results demonstrate that complex structural signatures can be transformed into interpretable processing–structure relationships, enabling rational design of algal bioplastics.
ER -