There are a wide range of processes that take place in the subsurface that involve the evolution of fractures, including both opening and closing due to some combination of mechanical and chemical stresses. Understanding subsurface processes, especially fractures, is critical for the development of safe and reliable long-term CO2 storage, geothermal energy, nuclear waste isolation, and petroleum extraction.
Underground infrastructure—such as wellbores for natural resource extraction, storage sites for hazardous materials, and portions of geothermal powerplants—is critical for supplying the U.S. with electricity and resources like natural gas. According to the Energy Information Administration, natural gas made up nearly 40% of U.S. electricity generation in 2022, and the EPA estimates that at least 1% of natural gas produced each year—about 6.5 million metric tons, is lost to leakage. Subsurface leakage is also a key factor in the safety and reliability of nuclear waste containment, geothermal energy production, and beyond. Applying high performance computing to model subsurface fracture formation will allow engineers to test different materials in various subsurface conditions as they are subjected to chemical and mechanical stresses, greatly improving the durability of subsurface infrastructure and supporting the safety and efficiency of resource extraction, power generation, and hazardous waste containment.
The Exascale Computing Project’s Subsurface team has created a software suite for modeling subsurface fractures in extreme detail. The suite enables detailed analyses at scales which can accurately capture the chemical and mechanical stresses leading to the formation fractures and pillars—the microscale structures which keep fractures open. These simulations will allow researchers to compare the physical and chemical resilience of various materials, and ultimately to select the materials and locations most suited to the task at hand. The Subsurface code suite can also be used to determine the risk of leakage in preexisting subsurface infrastructure such as wellbores, which are built to variable safety standards.
Previous subsurface simulations were modeled on large scales from 100-1000 meters and operated on timescales up to 10 years. This approach is useful for modeling leakage in subsurface infrastructure like wellbores, but lacks the resolution and appropriate physical parameters to accurately determine the cause and location of subsurface fractures. Without more accurate models of fracture formation, and the features like pillars and asperities that keep the fractures open, researchers and engineers can only mitigate the effects of leakage rather than predicting and preempting it.
To address this issue, the Subsurface team used the power of exascale computing to integrate a single high resolution multiphysics simulation across scales ranging from kilometers to microns. The Subsurface code models fracture sites at micron scales while coupling information from a coarser simulation of the changes to materials surrounding the fracture, enabling detailed study of the causes and effects of fracture formation. The code suite also incorporates variable chemical compositions in subsurface infrastructure and surrounding host rock, allowing for comparative analysis between potential locations and materials.
Exascale-enabled models will improve our ability to select safe and reliable sites and materials to minimize fracture formation in subsurface infrastructure projects, and to understand the controls of what governs whether fractures open or close due to some combination of chemical and mechanical stresses. These improvements will enable several critical developments, including improved efficiency and energy generation from advanced geothermal power plants, more feasible and secure storage of nuclear waste and CO2, and safer and more efficient extraction of natural resources.
An urgent challenge in the field of subsurface wellbores involves understanding and predicting the behavior of hundreds of thousands of deep wells drilled to locate and extract natural resources. The performance of a wellbore hinges on the behavior of very thin interface features controlling the leakage of fluids along the well casing–cement boundary. Similarly, leakage of buoyant fluids (e.g., CO2) through fractured caprocks may be controlled by micron-scale asperities in fracture networks that are themselves subject to geomechanical and geochemical modification. At the reservoir or field scale (~1–10 km domain size), multiphase flow and reactions in fractured porous media are typically modeled using continuum models that use averaged quantities and bulk parameters that do not fully take into account thermal, hydrological, chemical, and mechanical-related heterogeneity at different spatial and temporal scales. A more rigorous treatment is to resolve the pore-scale (0.1–10 micron) physical and geochemical heterogeneities in wellbores and fractures so as to improve their ability to predict the evolution of these features when subjected to geomechanical and geochemical stressors. The Subsurface project is using exascale to integrate the complex multiphysics processes occurring at multiple scales, from the micro to the kilometer scale, in a high-resolution reservoir simulator.
A wide range of processes take place in the subsurface that involve the evolution of fractures, including both opening and closing due to some combination of mechanical and chemical stresses. This project focuses on the science challenge of overcoming the failure of a wellbore for CO2 sequestration in saline reservoirs, with consideration of a wellbore segment of up to 100 m and times up to 1 year. Wells are considered to be high-risk pathways for fluid leakage from geologic CO2 storage reservoirs because breaches in this engineered system have the potential to connect the reservoir to groundwater resources and the atmosphere. The geologic carbon storage community has raised further concerns about wellbore stability because of acidic fluids in the CO2 storage reservoir, alkaline cement meant to isolate the reservoir fluids from the overlying strata, and steel casings in wells that are inherently reactive systems. This is of particular concern for the storage of CO2 in depleted oil and gas reservoirs with numerous legacy wells engineered to variable standards. The problem of wellbore failure is one of a broader class of subsurface problems involving fractures that may become leakage pathways.
In contrast to the conventional treatment of wellbore failure currently modeled at large scales on the order of 100 m to 1 km and 10 years, accurate prediction of fracture evolution depends on microscale resolution of fracture asperities (i.e., pillars) controlling permeability and chemical reactivity. Microscale resolution is also needed to accurately predict fracture permeability because very rough fractures are typically held open by pillars of this scale. Chemical corrosion (i.e., dissolution) or mechanical corrosion (i.e., pressure solution) of these asperities occurs at the same micron scale. The localized subdomain needed to resolve reactive transport processes at microscale resolution during fracture propagation is a domain size up to 10 cm (in the length of the wellbore) × 1 cm (along an azimuth in the cement annulus) × 1mm (in the radial direction) with 1 micron grid resolution. This domain size is assumed to be the minimum domain needed to capture coupled reactive transport and mechanics effects in a fracture (e.g., pillar collapse).
The Subsurface project addresses this exascale computing challenge by coupling two mature code bases: (1) Chombo-Crunch, developed at LBNL, which currently handles Navier-Stokes and Darcy flow coupled to multicomponent geochemical reaction networks, and (2) the GEOSX code, developed at LLNL, which handles geomechanical deformation and fracture+Darcy flow at a variety of scales.
A science Challenge Problem has been developed that focuses on the evolution of a single fracture in wellbore cement, beginning at Stage 1 with diffusion-controlled reaction and a weakening of the cement that leads to fracturing. The propagation of the fracture as a result of further chemical reaction and fluid pressure–driven deformation is simulated with 1 micron resolution within the fracture and is coupled to a coarser resolution (10 micron) representation of the porous cement adjacent to the evolving fracture. The resulting Challenge Problem is estimated to require 1 trillion grid cells with 16 trillion degrees of freedom once the hydraulic, mechanical, and chemical variables are included. Based on prior experiments and modeling, the challenge problem is estimated to extend for 10 days of simulation to capture the evolving fracture and associated reaction fronts.