Total: 1
Resolving and controlling the local dynamical properties of covalent organic frameworks (COFs) remains a central challenge, particularly when assembled from large, flexible building units. Here, we combine synchrotron X-ray pair distribution function (PDF) analyses with machine learning-accelerated molecular dynamics (MD) simulations to resolve the local structure and dynamics of two three-dimensional imine-linked COFs, COF-682 {[(DHP)(TAM)]$_{imine}$}, assembled from 6,13-dihydropentacene (DHP) and tetrakis(4-aminophenyl)methane (TAM), and COF-612 {[(HBC-LA$_{12}$)(HAPT)$_2$]$_{imine}$}, assembled from nanographene dodecabenzaldehyde hexakis{[3,5-bis($p$-formylphenyl)-4,6-dimethoxyphenyl]}hexabenzocoronene (HBC-LA$_{12}$) and 2,3,6,7,14,15-hexa(4-aminophenyl)triptycene (HAPT). Validated against the experimental PDFs through ensemble-averaged calculations, the simulations show that the exposed $π$-surface and V-shaped geometry of the DHP linker endow COF-682 with enhanced local flexibility through face-to-face and offset $π$-stacking interactions differing in both their average interplanar separation and their ring-plane tilt angle. In contrast, the extended nanographene linker rigidifies COF-612 by maintaining the planarity of its fused cores, while the linker pendant aryl rings equip both COFs with enhanced librational ability. The simulations further provide quantitative measures of the translational and reorientational mobility of the linkers, revealing how local COF dynamics may be tuned by balancing non-covalent interactions and different degrees of aromatic rigidity. The PDF-MD experimental-computational approach holds promise as a general method beyond conventional crystallography to gain insights into the local properties of COFs with the aim of directing their dynamic function.