Abstract
<title>Abstract</title> <p>The accuracy of density functional theory (DFT) calculations for zeolite simulations critically depends on the choice of exchange-correlation functional, yet a systematic benchmark of modern functionals remains lacking. We present a comprehensive evaluation of 26 DFT functionals, spanning LDA, GGA, and meta-GGA families, against experimental data for four distinct zeolite frameworks (α-quartz, NAT, HEU, and OFF). Performance is assessed across ten criteria covering structural properties (bond lengths, angles, surface area, cell volume, density) and energetic properties (total, binding, kinetic, electrostatic, exchange-correlation, and Jellium energies). A key innovation is the normalization of volumetric and surface properties to the tetrahedral (TO₄) unit, enabling fair comparison across frameworks with different unit-cell dimensions. GGA-HCTH emerges as the optimal functional for general-purpose simulations, achieving the highest combined score (0.888), ranking 1st in Si-O-Si angle (2.348°) and cell density. m-GGA-M06-L is superior for electronic properties, ranking 1st in kinetic, electrostatic, and exchange-correlation energies. GGA-PBE-Grimme offers the highest thermodynamic accuracy (total energy RMSE = 0.0011 eV), but its success relies on error cancellation (the "PBE-Grimme paradox"), as evidenced by a high DFT-D correction error (17.56%) and poor structural performance. GGA-PW91-OBS provides the most physically sound van der Waals correction (DFT-D error = 5.42%). Three functionals (m-GGA-MS0, m-GGA-M11-L, GGA-BLYP-TS) are completely unsuitable. Application-specific guidelines for functional selection are provided, enabling informed decisions for zeolite simulations ranging from structural optimization to adsorption and catalysis.</p>