Solvers compared for modeling fractures in rocks dissolving from CO2 exposure
Solution Strategies for Dissolution-Driven Phase-Field Fracture in Geomaterials
Computational Engineering, Finance, and Science
Summary
Fractures in rocks caused by dissolving chemicals affect important places like CO2 storage sites. The authors study different computer methods to simulate these fractures accurately and efficiently. They test five solvers that handle the complex math, comparing how well each works on accuracy, speed, and reliability. Their results show trade-offs: some solvers are cheap but less accurate, others are accurate but costly, and a hybrid method balances these well. This helps engineers pick the best method depending on their simulation needs.
What this means in practice
- •For geotechnical engineers: Simulate long-term fracture development in rocks for assessing CO2 storage safety with better solver options balancing speed and accuracy.
- •For reservoir simulation teams: Model chemo-mechanical effects including CO2-driven fractures in geothermal and nuclear waste sites using reliable coupled solvers to predict integrity.
Authors
Maithah M. Alaleeli, Mohamed Elkhodbiaa, Mostafa E. Mobasher
Abstract
Dissolution-driven fracture governs the long-term integrity of geological CO2 storage, geothermal reservoirs, and nuclear waste repositories. Modeling it requires solving strongly coupled mechanical, chemical, and phase-field equations. The solver choice controls their accuracy, cost, and convergence. This study investigates the role of solution strategies in dissolution-driven fracture in geomaterials. The coupled model is implemented in Abaqus user-element (UEL) subroutines, with all tangent blocks derived analytically. Five solvers are compared: (1) block-diagonal Newton monolithic, (2) BFGS quasi-Newton monolithic, (3) single-pass fully staggered, (4) single-pass partially staggered, and (5) hybrid multi-pass fully staggered. Performance is quantified by accuracy (errors in peak load, peak and failure displacements, and local CO2 concentration), cost (CPU and wall-clock times), and robustness, measured by new factors that assess convergence over the whole simulation and capture iteration spikes. The solvers are tested on mechanical and chemo-mechanical tension, shear, and multiple-crack benchmarks, including CO2 diffusivities raised up to a thousandfold. The results reveal a clear trade-off between accuracy, cost, and robustness. The quasi-Newton solver matches the Newton reference at a fraction of its cost but can fail under intense transport. The single-pass fully staggered solver is inexpensive and always completes, but its post-peak response depends strongly on the time step. The hybrid multi-pass solver also always completes, retains near-reference accuracy, and avoids iteration spikes at moderate cost. The partially staggered solver is the least robust. These findings guide practical selection: the single-pass staggered solver suits rapid screening, quasi-Newton offers accuracy when convergent, and hybrid multi-pass provides the best balance.