CSF Flow in Complex Geometry: Validation of a High-Order CFD Transient Solver

The optic nerve and brain are enveloped by a biologically active layer of lepto-meningothelial cells (MECs), crucial for phagocytosis and cytokine secretion, including IL-6 and IL-8. MECs form the inner lining of the arachnoid and pial membranes, directly interfacing with cerebrospinal fluid (CSF). The visual system’s homeostasis depends strongly on the available MEC surface area for mass transfer, emphasizing the importance of precise morphometric characterization. Using synchrotron radiation-based micro-computed tomography on human optic nerve sections, we recently have obtained the first quantitative measurements of the complex MEC geometry. Our results revealed that trabeculae and septae within the subarachnoid space significantly amplify the interactive MEC surface, enhancing mass transfer rates by up to 17-fold, as confirmed by detailed large-scale analyses of laminar CSF flow dynamics.

Our computational approach combines a pseudospectral numerical scheme with a Brinkman penalty method, implicitly capturing the geometry through signed-distance transformations, on supercomputers. Given previous concerns regarding smeared or inaccurate predictions near phase interfaces even on simple geometries, we validated our method against the challenging FDA nozzle benchmark. This benchmark, supported by experimental data, includes complex phenomena such as laminar-to-turbulent transitions, turbulent boundary-layer development, and turbulent jet breakup, features historically challenging for grid-convergence studies. Our computational results across Reynolds numbers ranging from 500 to 6500, spanning laminar, transitional, and fully turbulent flow regimes, outperformed reference solvers CharLES and OASIS, based on unstructured meshes, providing superior accuracy overall and resolving prior skepticism regarding near-wall prediction fidelity.