Skip to main content

Bridging Soil Scales

Full Campaign Name:

AI-Enabled Pore-Scale Modeling for Critical Mineral and Metal Applications

Heat map of pore networks displayed in orange, red, and black

Bridging molecular-scale reactivity to pore structural information is crucial for predicting the movement of critical minerals and materials (CMMs) in soils. 

EMSL is leading the development of agentic artificial intelligence (AI)-based approaches to streamline the integration of predictive flow and reactive transport modeling in soils with pore structural and molecular chemical data. This is accomplished through integrated pore-scale flow and transport modeling capabilities, such as PFLOTRAN, multimodal data fusion, and AI-enhanced workflows, including Pore2Chip and Chip2Flow. 

Through this Computing, Analytics, and Modeling Community Science Campaign, EMSL will partner with invited research community members to test agentic AI approaches that will integrate EMSL’s cutting-edge capabilities in pore-scale data transformations and modeling to streamline simulations (e.g., accelerate model setup and execution), enhance predictive accuracy, and optimize modeling efforts for CMM applications. 

By participating in this campaign, you will:

  • Help solve a big science problem 
  • Drive important outcomes 
  • Advance your own research

Participation

How researchers have been invited to participate 
  • A panel of researchers were invited to participate in the initial community science meeting based on their domain expertise, experience, and overlapping interests with the campaign topic. After this meeting to identify high-priority campaign objectives, a call for proposals was opened and shared with the community science meeting attendees. 
Required participant background 
  • Possess pore structural and chemical data that are approved/authorized for sharing 
  • Have an interest in using Molecular Observation Network (MONet) structural and chemical data in addition to their own data for scientific discovery through AI-enabled workflows 
  • Have expertise and an interest in pore-scale modeling (i.e., predicting mineral reactions, transport processes, and pore-structure dynamics in CMM applications) 
How will you contribute? 
  • Insight Into Current Methodologies and Techniques
    • Offer insights into transport processes, chemical reactions, coupled flow and reactive transport, relationships between pore structure and process behavior, and the evaluation of new pore-scale modeling approaches such as process-based simulations, AI/machine learning (ML) training, and agentic AI workflows. 
  • Supply data
    • Contribute use cases and pore structural and chemical data that can be used to advance campaign objectives. 
  • Feedback Through Community Science Meetings
    • Work with EMSL to identify high-priority science questions within Department of Energy, Office of Science, Biological and Environmental Research program areas where improved prediction or mechanistic understanding is needed. Help define the most important research problems where EMSL’s modeling, AI, and pore-scale analysis capabilities can make a useful contribution. 
  • Expert Input
    • Participants will work as a group to test agentic approaches for integrating cutting-edge capabilities in pore-scale data transformations and modeling to streamline simulation (e.g., accelerate PFLOTRAN simulation setup), enhance predictive accuracy, and optimize modeling efforts for CMM applications.

About the Campaign

EMSL has a long history of stewarding advanced imaging techniques that enable EMSL users to characterize structural and chemical information in soil pores. To interpret pore-scale processes observed in imaging and experimental data, EMSL has developed advanced flow and reactive transport modeling capabilities, multimodal data-fusion approaches, and AI-enhanced workflows, including PFLOTRAN, Pore2Chip, and Chip2Flow. These advanced tools enable detailed process-based modeling and imaging/experimentation (ModEx), bridging molecular-scale reactivity to pore-scale mineral dissolution and precipitation for EMSL users. 

EMSL resources include advanced imaging, spectroscopy, modeling, and AI-enabled data-integration capabilities that can support the pore-scale analysis of soils and porous materials. Data from techniques such as X-ray computed tomography (XCT), scanning electron microscopy–energy dispersive X-ray spectroscopy (SEM-EDX), optical coherence tomography (OCT), hyperspectral imaging, X-ray fluorescence (XRF), and Fourier transform ion cyclotron resonance (FTICR) mass spectrometry (MS) provide complementary structural and chemical information. However, integrating these diverse datasets into model-ready 3D representations and using them to simulate coupled flow, transport, and biogeochemical processes remains technically challenging, labor-intensive, and highly case specific. 

This campaign is needed to test and advance AI-driven and agentic workflows that can streamline these data-to-model transformations, improve reproducibility, and broaden access to pore-scale modeling capabilities for CMM-relevant science questions. AI-powered computer vision, multimodal data fusion, and agentic workflow approaches will be evaluated as tools for accelerating the integration of pore-structural and molecular chemical information into predictive flow and reactive transport simulations. 

By unifying diverse data modalities (e.g., time series, images), campaign participants will come together to apply AI-enhanced workflows that offer precision and clarity in understanding the multiphysics phenomena that underpin CMM extraction, addressing the complexity of multimodal data fusion and analysis. 

This campaign supports the recent presidential memorandum on fiscal year 2027 national research and development priorities and demonstrates how AI accelerates scientific discovery within Earth sciences.

Campaign Timeline

OCTOBER 2025 – CAMPAIGN TOPICS AND DESCRIPTIONS DRAFTED 

  • Identify community science campaign topics aimed at solving a significant scientific challenge or filling current gaps in knowledge. 

NOVEMBER 2025 – POTENTIAL CAMPAIGN PARTICIPANTS IDENTIFIED 

  • Strategically identify researchers with ideal domain expertise and experience to invite participation in the upcoming community science meeting. 

DECEMBER 2025 – COMMUNITY SCIENCE MEETING 

  • Host a community science meeting to identify common pore-scale modeling questions for CMM extraction and recovery. 

DECEMBER 2025 – CALL FOR PROPOSALS 

  • Invite community science meeting attendees to submit proposals for projects contributing to overall campaign goals. 

JANUARY 2026 – CAMPAIGN PROPOSAL DEADLINE 

  • Deadline for invited campaign participants to submit proposals for projects contributing to overall campaign goals. 

JANUARY 2026 – INITIATE WORK ON ACCEPTED PROPOSALS 

  • Work begins on accepted proposals for projects contributing to overall campaign goals. 

FEBRUARY–SEPTEMBER 2026 – APPLY PORE-SCALE MODELING WORKFLOWS 

  • Apply EMSL pore-scale modeling workflows for key science questions developed by campaign participants. 

SEPTEMBER 2026 – COMPLETE CAMPAIGN 

  • Complete the campaign and publish AI/ML workflows.

Campaign Methods

ADVANCED PORE-SCALE MODELING 

  • Utilize EMSL’s advanced pore-scale flow and reactive transport simulation capabilities, including PFLOTRAN modeling, to characterize subsurface mineralogy, fluid–mineral interactions, and complex transport processes across molecular, pore, and core scales. 

MAPPING AND SCALING MINERAL–FLUID DYNAMICS 

  • Enable the precise mapping of mineral–fluid interactions at microscopic scales using Pore2Chip workflows, while leveraging Chip2Flow tools to scale these insights to system-level predictions for transport and reaction modeling across soil and subsurface environments. Pore2Chip and Chip2Flow are companion Python tools that turn 3D XCT scans of soil into simple, testable designs, and then simulate how fluids move through them. Pore2Chip converts the scans into 2D “lab-on-a-chip” blueprints that can be 3D printed or fabricated. Chip2Flow takes those designs and runs physics-based simulations (with the open-source PFLOTRAN engine) to predict flow and train AI models.  

HIGH-RESOLUTION IMAGING AND MULTIMODAL DATA FUSION 

  • Utilize high-resolution data-fusion algorithms, including vision language models, vision transformers, Segment Anything, and SegFormer, to achieve precise segmentation and co-registration of OCT, XCT, SEM-EDX, and chemical imaging datasets. 

AI-ENABLED MODEX 

  • Leverage ModEx workflows integrated with TerraForms and MONet soil databases to refine pore-scale modeling strategies. Enable AI-driven calibration and real-time feedback from experimental datasets to optimize sampling, imaging methods, and model accuracy while seamlessly connecting insights from EMSL and partner resources.

Solve a Big Challenge

This campaign helps address the challenge of comprehensively characterizing and optimizing the extraction of CMMs across natural and engineered systems. Traditional methodologies struggle to integrate complex multimodal datasets and scale insights across pore-scale and flow-level interactions, limiting our understanding of the biogeochemical processes governing CMM dissolution and precipitation. EMSL is overcoming this barrier by unifying diverse data modalities—such as XCT, SEM-EDX, hyperspectral imaging, and FTICR MS—through advanced AI-enhanced workflows like Pore2Chip and Chip2Flow. These tools enable detailed pore-scale modeling that bridges molecular reactivity with larger-scale system dynamics, empowering researchers to simulate coupled flow and biogeochemical phenomena with precision.

Expected Campaign Outcomes

INTEGRATED AI-AGENTIC WORKFLOWS FOR ENHANCED MODEL–DATA INTEGRATION 

  • The campaign will deliver advanced AI-agentic workflows that integrate multimodal experimental data and simulations through iterative process-model calibration. These workflows will refine the predictions of subsurface structure, flow dynamics, and chemical properties by enabling PFLOTRAN simulations, adaptive input deck development, and real-time feedback mechanisms. 

BENCHMARK EXPERIMENTAL DATASETS FOR CMMS 

  • Researchers will gain access to curated, high-resolution benchmark datasets designed for PFLOTRAN simulations, co-registration, and AI/ML model training. These datasets, generated from EMSL tools like TerraForms workflows (Pore2Chip and Chip2Flow), will validate and enhance predictive modeling efforts for various subsurface environments and serve as a foundational resource for future research. 

OPEN ACCESS TO PRETRAINED AI/ML MODELS 

  • The campaign will provide openly accessible, modular AI/ML models on platforms such as GitHub and EMSL’s Science Central. These pretrained tools will streamline workflows for tasks like mineral phase segmentation, permeability and porosity predictions, and reactive-transport modeling, enabling users to boost computational efficiency and scale applications across CMM research domains.

Advance Your Research

  • Accelerate your science
  • Access cutting-edge technology
  • Gain experience and key knowledge
  • Co-author scientific publications

Why "Community" Science Campaigns?

Each community science campaign is intended to bring together researchers with a wide variety of expertise to tackle the same strategically identified challenges that are bigger than what an individual principal investigator or small team research effort can accomplish alone. 

With the help of the scientific community, EMSL can ensure that the pore-scale modeling capabilities developed are addressing the most critical challenges, utilizing the right datasets, and refining workflows to advance the community’s understanding of complex subsurface processes.

Contacts

Campaign leader (science domain expert): Satish Karra | Website bio 

EMSL user program contact (logistics): Rick Washburn | Proposal calls