Cooperative Research and Development Agreement (CRADA) Final Report Report Date: January 22, 2026 In accordance with Requirements set forth in the terms of the CRADA, this document is the CRADA Final Report, including a list of Subject Inventions. It is to be forwarded to the DOE Office of Scientific and Technical Information upon completion or termination of the CRADA, as part of the commitment to the public to demonstrate results of federally funded research. Parties to the Agreement: Lawrence Berkeley National Laboratory and Capra Biosciences, Inc. CRADA number: FP00017018 CRADA Title: Bioreactor Optimization through Multi-Phase Flow Models Responsible Technical Contact at Berkeley Lab: Dr. Ishan Srivastava Name and Email Address of POC at Partner Company(ies): Dr. Andrew Magyar, CTO, Capra Biosciences Inc., email: andrew@caprabiosciences.com Sponsoring DOE Program Office(s): EERE-AMMTO and EERE-ITO LBNL Report Number: LBNL-2001737 OSTI Number: [SPO to complete] Joint Work Statement Funding Table showing DOE funding commitment: DOE Funding to LBNL $300,000 Participant Funding to LBNL $0 Participant In-Kind Contribution Value $75,000 Total of all Contributions $75,000 Provide a list of publications, conference papers, or other public releases of results, developed under this CRADA: N/A Provide a detailed list of all subject inventions, to include patent applications, copyrights, and trademarks: N/A Executive Summary of CRADA Work: Chemical manufacturing uses 29% of energy in the United States and produces 925 million metric tons of CO2 annually. Biomanufacturing offers the potential to leverage America’s rich agricultural resources to produce critical chemicals such as lubricants, pharmaceutical precursors, and components of energetic materials that today are sourced extensively from overseas. The bioreactors used in biomanufacturing applications, such as one developed by Capra Biosciences, involve multiphase flow of biofilm-coated solid support particles that are continuously circulated in a fluidized state within the reactor along with a constant supply of oxygen via an aeration mechanism. In this project, Capra Biosciences and LBNL developed a multiscale modeling framework to simulate the multiphase flows of solid particles in a liquid-gas bubble mixture that occurs in the bioreactor using the current MFIX-Exa software, an open- source multiphase flow solver developed and maintained at LBNL and NETL. By leveraging HPC capabilities, this high-fidelity multiscale model was used to inform design decisions for bioreactor architecture to make them operationally efficient. We developed and incorporated several new numerical methods for multiphase flow modeling into MFIX-Exa, enabling more effective simulations of fluid dynamics in Capra’s industrial- scale bioreactors. These methods were essential for accurately representing the multiphase flows in a bioreactor, as MFIX-Exa was initially designed for gas-solid interactions. We incorporated new physical models (such as virtual mass force) to capture the dynamics of gas bubbles in a high-density, high-viscosity fluid. Another computational development was the implementation of the ability to read STL files, which allowed for the creation of embedded boundary surfaces to accurately depict the complex geometries of Capra Biosciences’ bioreactors. Using our developed computational model, we conducted numerous HPC simulations of multiphase flows in Capra’s bioreactors, utilizing ~15K GPU node-hours on OLCF Summit. We explored various gas flow rates, feedstock flow rates, and bioreactor geometries to optimize feedstock and gas mixing for aerobic bioreactions. A key insight from these simulations was that feedstock particles are fluidized and mixed within the bioreactor due to entrainment in the ascending plume of gas bubbles. We identified regions in the bioreactor that exhibited poor mixing, leading to increased production costs and waste, which encouraged us to explore alternative designs. These simulations along with experimental insights have revealed that aeration is a significant cost barrier in aerobic biomanufacturing and can account for up to 50% of operational expenses, thus prompting us to redirect our focus on understanding and optimizing for aeration within the bioreactors. Future work based on this project can build towards generalizing the computational model by integrating detailed gas bubble behavior, gas-liquid mass transfer, and oxygen uptake mechanisms. These advancements can help design industrial bioreactors with substantially improved oxygen efficiency and reduced energy demands. Furthermore, these models could serve as an important component of digital twins, which in combination with AI tools enable new modes for adaptive process control in bioreactors. Consequently, manufacturing costs for certain commodity chemicals could decrease significantly, bolstering U.S. market competitiveness against foreign imports. Summary of Research Results: Capra Biosciences Inc. has developed a new kind of bioreactor that uses biofilms rather than the traditional suspended cell culture that dominates industry today. Biofilm reactors offer several key advantages: (1) continuous synthesis, (2) more biomass per volume of water, (3) superior tolerance to chemical products, and (4) lower carbon-intensity compared to traditional technologies. Traditionally, biofilms have been limited by mass transport of feedstock and nutrients as well as challenges with product extraction. The Capra reactor design overcomes these limitations by using thin biofilms and small solid support beads (~100 microns) that are continuously circulated in an up-flow fluidized bed reactor. In these cylindrical bioreactors, fluid and air are injected from nozzles at the bottom of the cylinder at controlled flow rates, which fluidizes the solid particles within the reactor. When the project was first conceived, the hypothesis was that the pumping required to maintain fluidization of the solid support particles is the most energy-intensive step in the manufacturing process. Furthermore, the biofilm on particles introduced complex interparticle adhesion forces that complicated fluidization dynamics. In preliminary small-scale experiments, Capra demonstrated that introducing channeled structures to create angular flow within the bioreactor yielded ~10x reduction in the required pumping force to maintain fluidization. Although Capra has extensive expertise in bioreactor design and the bioengineering within the reactor, there was a lack of fundamental understanding of the fluidization dynamics in multiphase bioreactor flows, along with a lack of high-fidelity, computationally tractable models with predictive accuracy required for reactor design optimization. Furthermore, currently available commercial tools could not efficiently model such computationally intensive multiphase flows. As the project initiated, it was very soon realized that resolving the dynamics of aeration, i.e., flow of gas bubbles and their interactions with the solid support particles is the most critical step in the manufacturing process. The low solubility of oxygen in water was a key challenge for efficiently aerating bioreactors. The oxygen requirement for Capra’s microorganism is equivalent to only about 0.05 volumes of air per reactor volume per minute, and the key challenge was in getting as much oxygen dissolved into the liquid medium from the small bubbles that are introduced into the bioreactor as possible. As such, a significant cost barrier in their aerobic biomanufacturing was deemed to be reactor aeration, which could account for up to 50% of operational expenses. The overarching goal of this project remained the same, however, which was to develop a high- fidelity multiphase flow modeling framework to understand the bioreactor fluidization and aeration dynamics, and which can be used to optimize bioreactor design with the goal of enhanced energy/pumping efficiency while still maintaining good fluidization, well-mixed aeration and mass transport for manufacturability. This project required extensive HPC resources because of the complexity of this problem - in Capra’s small-scale reactor there are more than 11 million interacting particles, and the production-scale bioreactor contains >1000x more particles. The scale of the modeling required not only HPC resources but also expertise in multiscale modeling of multiphase flows to scale-up particle-scale physics towards reactor-scale fluid dynamics. Through concerted efforts by the national lab and industrial team, we incorporated several new physics models in the existing multiphase flow code, MFIX-Exa, to produce high- fidelity simulations of bioreactor dynamics that were conducted on OLCF and NREL HPC systems. The following summarizes the new numerical methods that were developed, the results of simulations, new insights that were gained, and important directions for future work. ● New numerical methods for multiphase flow modeling: We developed and incorporated numerical algorithms for new physics models into MFIX-Exa to effectively simulate multiphase fluid dynamics in Capra’s industrial-scale bioreactors. These numerical methods were needed to accurately simulate such flows in MFIX-Exa that was primarily written to simulate gas-solid flows. a. Ability to read STL files into MFIX-Exa towards creating embedded boundary surfaces for representing the complex geometry of Capra Biosciences’ bioreactors. b. Developed and incorporated a virtual force model for accurate representation of the dynamics of lighter Lagrangian particles such as air bubbles suspended in denser fluids. c. Developed numerical methods to interpolate fluid shear stresses as unresolved forces onto the solid particles and air bubbles in the simulation. While these effects are unimportant for high-speed solid particles in low-viscosity gas flows, bioreactors consist of high-viscosity fermentations (O(10x) viscosity of water), where fluid shear stresses have a pronounced effect on particle dynamics. d. Developed and implemented fluid-mediated lift forces on air bubbles in MFIX-Exa; these models were not present in the code previously that focused on high density particles. However, rising low-density air bubbles in a dense fluid experience lift forces in the direction of fluid vorticity, and these effects have been shown to have a significant impact on bubble aggregation in bubble rising columns. ● Simulation results: We conducted numerous simulations of multiphase flows in Capra’s bioreactors, utilizing ~15K GPU node-hours on OLCF Summit. The inline figure illustrates a multiphase flow simulation depicting spatio- temporal evolution of aeration (top; represented by discrete gas bubbles) and feedstock (bottom; represented by discrete solid particles) in the bioreactor. We explored various gas flow rates, feedstock flow rates, and bioreactor geometries to optimize feedstock and air mixing for aerobic bioreactions. Additionally, we examined the local shear stresses on feedstock particles, as excessive shear stresses on biofilm-coated particles can damage the biofilms, while adequate fluid motion is required for achieving well-mixed conditions such as oxygen distribution. We validated the dynamics predicted by our simulations by comparing them to videos of dye-injected flows in the actual bioreactor (see figure). ● New insights gained: A key insight from these simulations was that feedstock particles are fluidized and mixed within the bioreactor due to entrainment in the ascending plume of gas bubbles. We identified regions in the bioreactor (see red dots in the figure) that exhibited poor mixing, leading to increased production costs and waste, which encouraged us to explore alternative designs. Furthermore, we noted that the energy consumption associated with aeration was substantial in our bioreactors, making accurate modeling essential. Using results of the current project, Capra was able to successfully redesign how gas and media is introduced into the bioreactor, which led to improved circulation and aeration in the reactor and the elimination of stagnant zones that can be sources of contamination. ● Future work: While we substantially enhanced the numerical code to simulate the multiphase bioreactor flows, there are two important areas of future code development. Firstly, the code does not account for gas bubble coalescence and breakage, two phenomena that are observed in practice in experiments, and which are vital for accurately simulating and optimizing aeration in industrial bioreactors. Additionally, the current code lacks the capability to model oxygen mass transfer and uptake, which is critical for understanding and optimizing dissolved oxygen distribution within the bioreactor. Future work along these research and development directions will make the software much more robust in accurately modeling the complex reactive fluid dynamics in industrial biomanufacturing. Combination of these models with advanced distributed sensing networks and AI tools can provide further insights into both bioreactor design optimization and enable new ways to actively control bioreactors with greater precision. APPENDIX A (Reference Only) This appendix has been developed by DOE to assist DOE Labs in drafting the Executive Summary and Summary of Research Results sections of the CRADA Final Report. Executive Summary of CRADA Work: Include a discussion of 1) how the research adds to the understanding of the area investigated; 2) the technical effectiveness of the materials, methods or techniques investigated or demonstrated, and their economic feasibility, if known; and 3) how the project is otherwise of benefit to the public. The discussion should be a minimum of one paragraph and written in terms understandable by an educated layman. Summary of Research Results: ● INCLUDE, IF APPLICABLE: "This product contains Protected CRADA Information, which was produced on [DATE] under CRADA No. [##-####] and is not to be further disclosed for a period of [ up to and not to exceed] five (5) years from the date it was produced except as expressly provided for in the CRADA." ● Summarize project activities for the entire period of performance, including original hypotheses, approaches used, problems encountered, any departure from planned methodology, and an assessment of their impact on the project results. Incorporate technical data, e.g. facts, figures, analyses, and assumptions used during the life of the project to support the technical conclusions of the work. It is acceptable to incorporate the technical data by reference to other publicly available sources, such as a publications or other reports, but not websites. Provide a comparison of the actual accomplishments with the goals and objectives of the project. Where possible, the summary should cover each task listed in the Statement of Work (SOW) and should note any deviations from the project plan, or lack of technical data. ● Identify products, potential applications, and technology transfer activities developed under the CRADA, including those completed and anticipated at the time of the report. These include but are not limited to: 1) networks or collaborations fostered; 2) technologies/techniques/methodologies; 3) other products that reflect the results of the project, such as commercial products, internet sites, data or databases, physical collections, audio or video, software, models, educational aid or curricula, and instruments or equipment.