(363j) Rethinking the Fate of Carbon: Leveraging Artificial Intelligence and Reaction Engineering to Enable New Hydrothermal Technologies

Motivation: Globally, humanity is producing 52 billion tons of carbon dioxide from anthropogenic emissions each year. Current efforts for sequestering carbon, such as direct air capture (DAC) and point source capture (PSC), are fundamentally limited by low concentrations of CO₂ (440 ppm – 0.1 atm). Additionally, the goal of these technologies is to utilize CO₂ as a feed which requires unique chemistries to reduce the energy needed for desorbing CO₂ after it has been captured. Often, these chemistries are sensitive to moisture and oxygen which require additional expensive separation steps. In this work, I propose that we let nature concentrate carbon in organic matter which can allow us to start with high-energy, cheap, and distributed feeds with biomass and organic municipal solids wastes.

Research Interests

Hydrothermal processes are a range of thermal methods that utilize the physical properties of water at different temperatures and pressures to convert wet organic feeds into biochar, biocrude, and biogas products. Due to the tunable nature, hydrothermal processes offer a versatile alternative to traditional management of organic wastes. My work presented here focuses on integrating machine learning and chemical reaction engineering to hydrothermal technologies to understand inherent multiphasic nature as well as optimize carbon recovery and utilization.

Defeatured Machine Learning Models: Hydrothermal liquefaction (HTL) operates near the critical point of water (250-350C) to create an environment that breaks down waste macromolecules to a biocrude that is in phase equilibria with water. With such complex phases, it is impossible to characterize every compound in the oil and water phases. Analytical methods like GC and GCxGC are limited in the size of compounds that can be analyzed and often produce noisy spectra due to coelution of compounds. With FT-ICR MS, unique molecular formulas of compounds in both aqueous and oil media can be determined with high resolution. While structural information is not provided with FT-ICR MS, I was able to develop a non-linear machine learning algorithm to predict a compounds partition coefficient based on the elemental composition of a molecule. This model, while dimensionally reduced, has an RMSE of 0.77 which was shown to be better than previously published models that require structural information. With this model, partitioning trends across various complex oil-water mixtures are now a possibility.

Developing Advanced Carbon Capture Processes: The fate of biocrudes produced in HTL is to be combusted as a fuel source. With a uniquely concentrated source of carbon and energy like waste, the utilization must take advantage of both. In this work, I present my patent-pending process hydrothermal mineralization (HTM) which sequesters carbon as an inert mineral through energy-positive processes. HTM converts organic waste into calcite (CaCO₃) in two thermodynamically favorable steps: (1) waste undergoes supercritical water oxidation (SCWO), a low-temperature combustion that drive carbon to a high-pressure CO₂, and (2) the high-pressure CO₂ drives mineralization in under 20 minutes. HTM is a thermal process that overcomes the shortfalls of current DAC technologies. Additionally, in this project, I show that HTM can be integrated directly into the production of alternative cements. The high-pressure CO₂ allows for directed precipitation along structure directing polymers to produce a strength of 9.23 MPa, double the expectation for non-load bearing cements.

Recovery and Extraction of Critical Resources: A critical assumption in the HTM process is the source of calcium that is used to form calcite (CaCO₃). Using calcium hydroxide (Ca(OH)₂) provided both stoichiometric amounts of Ca²⁺ and OH^- ions such that CO₂ could be converted into CaCO₃. Traditionally, Ca(OH)₂ is made by first sourcing CaCO₃, but if it is produced through the chloroalkyl process using calcium-rich brines the lifecycle impacts are reduced by 0.2 kg CO₂/kg Ca(OH)₂. This process is not widely used due to the greater abundance of CaCO₃ than calcium brines. With nearly 6 million tons produced each year, sewage sludge presents a unique opportunity as it is rich in minerals (i.e., calcium) and metals. In this work, I have designed and constructed a continuous supercritical salt precipitation (SCSP) reactor. This system will take the liquified product of HTL and separate the organic products from the inorganic with the physical properties of supercritical water.

Throughout my PhD, I have been able to apply new machine learning techniques and traditional chemical reaction engineering to develop systems that allow us all to see a world without waste. I am presently looking for a position that will allow me to leverage my skills to help develop the future of sustainable technologies and processes.

Biography

• Education: A. Chemistry, Assumption College; B.S. Chemical Engineering, Washington University in St. Louis; M.S. Chemical Engineering, Worcester Polytechnic Institute; Ph.D. Chemical Engineering, Worcester Polytechnic Institute
• Awards: ACS GCI Heh-Won Chang Ph.D. Fellowship in Green Chemistry (2024), American Institute of Chemists Student Award (2023)
• Certifications: Data Science, WPI 2022, Associate Systems Engineering Professional, INCOSE 2021, Fundamentals of Engineering, NCEES 2020

Publications

1. Kenney, D. H., Paffenroth, R. C., Timko, M. T., Teixeira, A. R. (2023), Dimensionally reduced machine learning model for predicting single component octanol–water partition coefficients, Journal of Cheminformatics, 15 (1), 9
2. Kenney, D. H., Charlebois, A., Wang. S., Rahbar, N., Timko, M. T., Teixeira, A. R., Negative Emission Waste-to-Cement by CO2 Oxidation via Novel Hydrothermal Mineralization Pathway, In Preparation