Pore-scale investigation of residual DNAPL dissolution in fractures: Role of aperture-field anisotropy and source-zone architecture

Microfluidic rough-fracture experiments track residual DNAPL morphology, interfacial area, and dissolution. • Interfacial-area scaling depends on residual architecture and dissolution stage. • Interface renewal and hydraulic accessibility jointly control DNAPL dissolution in rough fractures. Residual dense non-aqueous phase liquids (DNAPLs) trapped in fractured rock can sustain groundwater contamination through long-term dissolution. In rough fractures, aperture-field anisotropy controls preferential flow and residual source-zone architecture, causing interfaces of similar geometric area to differ in hydraulic accessibility. However, direct pore-scale evidence remains limited on how residual architecture and hydraulic accessibility jointly control dissolution pathways and interfacial-area-based mass-transfer scaling. Here, microfluidic experiments were used to quantify the co-evolution of residual DNAPL saturation, specific water–DNAPL interfacial area ( a s ), and the image-derived apparent dissolution coefficient ( K app ). Aperture-field anisotropy and displacement history generated pool-dominated (PP), ganglia-dominated (GP), and mixed (MP) residual architectures with distinct dissolution pathways. Across Re = 0.395–6.583, flow rate altered dissolution kinetics and stage duration, but the dominant architecture-dependent response remained identifiable. PP systems underwent fragmentation-driven interface generation followed by contraction-dominated depletion, GP systems were dominated by ganglia contraction, and MP systems exhibited spatially sequential dissolution. GP trajectories generally occupied higher K app ranges, whereas MP trajectories occupied lower ranges. Aperture-field anisotropy modified residual morphology, pool elongation, interfacial-area evolution, and fragmentation persistence, indicating a pathway-level control on dissolution behavior. The proportional K app − a s relationship did not hold universally: fragmentation-dominated stages produced scattered relationships, whereas contraction-dominated stages yielded more coherent but architecture-dependent responses. During GP Stage III, K app decreased more rapidly than total a s , indicating reduced apparent transfer efficiency per unit geometric interface. These findings show that predictive models of DNAPL dissolution in fractured media should account for residual source-zone architecture, aperture-field anisotropy, dissolution stage, and the evolving hydraulic effectiveness of the water–DNAPL interface.