New research shows that combining machine learning and advanced materials engineering can significantly improve the performance of hydrocarbon char, a waste-derived carbon-rich material, offering a promising path for sustainable agriculture and climate mitigation.
“By integrating machine learning and experimental design, we can predict and optimize hydrocarbon properties much more efficiently than traditional trial-and-error approaches,” said the study’s corresponding author. “This opens up new opportunities to transform agricultural waste into high-value environmental materials.”
Hydrocarbon carbon is produced by hydrothermal carbonization, a process that converts wet biomass, such as livestock manure, into a stable carbon material under moderate temperatures and pressures. Compared to traditional biochar production, this method requires less energy and eliminates the costly drying step. However, two key challenges limit its widespread use: the relatively low stability of carbon and the limited availability of phosphorus.
In the new study, researchers focused on pig manure, a major waste stream in the world that is rich in both carbon and phosphorus. They engineered the hydrocarbon by introducing ferric chloride during production and systematically changing reaction conditions such as temperature, acidity, and treatment time.
The results showed that iron modification significantly increases both carbon stability and phosphorus availability. Under optimal conditions, specifically acidic pH, 220 degrees Celsius, and 2 hours of reaction time, the engineered hydrocarbon char had improved resistance to decomposition, while at the same time releasing more plant-available phosphorus.
Carbon stability is critical for long-term carbon sequestration. This is because the more stable substances remain in the soil for longer periods of time, helping to reduce carbon dioxide in the atmosphere. At the same time, increased phosphorus availability promotes plant growth and reduces the need for synthetic fertilizers.
The researchers used multiple analytical techniques to understand these improvements. The modified hydrocarbon coal formed a more stable carbon structure and iron-related mineral phase, as shown in the thermogravimetric analysis and X-ray diffraction results in the paper. X-ray photoelectron spectroscopy further revealed changes in surface chemistry that contribute to enhanced carbon bonding and enhanced nutrient interactions.
To further optimize the experiment, the team applied five different machine learning models to predict hydrocarbon properties based on feedstock composition and processing conditions. Among these models, the generalized additive model performed best, achieving strong predictive accuracy with a correlation coefficient of 0.86.
Machine learning analysis has also provided new insights into what controls hydrocarbon performance. Carbon stability was mainly influenced by the hydrogen and oxygen content of the original biomass, whereas phosphorus availability was more dependent on the carbon, nitrogen, and oxygen compositions.
“These discoveries allow us to move from experimental testing to data-driven design of engineered hydrocarbons,” the authors said. “This means materials can be tailored to specific environmental applications more quickly and at lower cost.”
The impact extends beyond waste management. By converting livestock manure into engineered hydrocarbons, this approach helps reduce the risk of contamination associated with nutrient runoff while creating a valuable soil amendment. It also supports circular economy strategies by turning waste into resources.
As global demand for sustainable solutions in agriculture and climate mitigation increases, this research highlights how combining machine learning and materials science can accelerate innovation.
The researchers conclude that their framework can be applied to a wide range of biomass types and pave the way for the scalable production of next-generation hydrocarbons with optimized environmental performance.
===
Reference magazines: Xie, S., Zhang, T., You, S. Others. We applied machine learning to predict the properties and carbon and phosphorus fate of raw and engineered hydrocarbon coals. biochar 719 (2025).
https://doi.org/10.1007/s42773-024-00404-4
===
About biochar
biochar (e-ISSN: 2524-7867) is the first journal dedicated to biochar research across agriculture, environmental science, and materials science. We publish original research on biochar production, processing, and applications such as bioenergy, environmental remediation, soil improvement, climate mitigation, water treatment, and sustainability analysis. The journal serves as an innovative and professional platform for researchers around the world to share advances in this rapidly expanding field.
follow us above facebook, ×and blue sky.
Disclaimer: AAAS and EurekAlert! We are not responsible for the accuracy of news releases posted on EurekAlert! Use of Information by Contributing Institutions or via the EurekAlert System.
