Abstract
Crystalline silicate minerals and carbonates make up 90% to 95% of the Earth's lithosphere, the rigid outermost shell comprising the crust and uppermost solid mantle. Crystalline materials are continuously modified by fluid–rock interactions including crystal growth occurring along fractures and within pore spaces. Crystals form under conditions dictated by extrinsic factors, such as temperature, pressure, fluid composition, and pore confinement, as well as intrinsic factors such as crystal structure, chemical bonding, and anisotropy. While the thermodynamic stability and structure of crystals are well established, their rates of heterogeneous nucleation and growth remain poorly constrained. This is particularly true in porous media, where nucleation is frequently a distinct and rate-limiting step rather than a negligible precursor to growth, especially where non-classical pathways preclude a single, well-defined interfacial free energy.
In this commentary, we outline how crystal nucleation and growth in porous media control pore-scale structure, flow, and transport, and we evaluate how these processes can be represented in numerical models across scales. We examine the distinct roles of nucleation pathways, growth kinetics, interfacial and confinement effects, and crystallization pressure in governing transient porosity, stress localization, and self-organized crystalline networks. Furthermore, we discuss how fluid flow and mixing modulate where and when these processes occur. These principles are illustrated through examples spanning well-controlled synthetic crystal growth in materials science to geo-energy and geoscience applications, such as basaltic CO2 mineralization, where probabilistic nucleation, reactive transport, and evolving pore architecture jointly control macroscopic behavior.
Finally, we assess the potential of machine learning and surrogate modeling approaches to bridge the persistent gap between molecular-scale mechanisms and field-scale predictions. By emphasizing mechanistic understanding over system-specific details, our goal is to help porous media scientists judge when crystallization is likely to matter, and to guide the selection of modeling approaches that remain both physically grounded and computationally tractable.
WHY THIS PAPER MATTERS
Crystal nucleation and growth can progressively clog pore spaces, redirect fluid pathways, and transform the mechanical and transport behavior of porous rocks across scales. Is it possible to use information we have about nanoscale processes to predict reservoir-scale changes? This InterPore Commentary article synthesizes how confinement, interfacial free energies, and crystallization pressures control where crystals appear, how fast they grow, and when they damage or seal flow networks in natural and engineered systems. It then outlines how machine learning can be used to connect molecular, pore, and field scales for predictive modeling.