Piotr Radomski, Dominik M. Florjan, Maciej J. Szary
Accurate treatment of van der Waals (vdW) interactions is essential for exploring the potential of transition-metal dichalcogenides (TMDs) in sensing, energy storage, and catalysis. Here, we present a comprehensive benchmark of 12 representative 2H-phase TMDs: NbS 2, MoS 2, TaS 2, WS 2, NbSe 2, MoSe 2, TaSe 2, WSe 2, NbTe 2, MoTe 2, TaTe 2, and WTe 2, using a suite of density functional theory (DFT) approaches. These include the semilocal PBE functional (bare and with D2, D3, D3M, D3+BJD, and D3M+BJD corrections) and six nonlocal vdW density functionals (vdW-DF, vdW-DF2, vdW-DF-OB86, vdW-DF-OBK8, vdW-DF-CX, and vdW-DF2-B86R). Both the physical accuracy and performance of each method were investigated. The results reveal that while no single vdW scheme is universally superior, a consistent subset, PBE+D3, vdW-DF-CX, vdW-DF-OB86, and vdW-DF2-B86R, provides a balanced and reliable description at a computational cost comparable to that of uncorrected PBE. Among these, vdW-DF-CX and vdW-DF2-B86R provide the most accurate and transferable performance, closely followed by the widely used PBE+D3, which remains a practical choice for adsorption and intercalation studies. In contrast, PBE+D2 shows erratic behavior for W- and Ta-based TMDs, while D3M-based corrections systematically overbind and original vdW-DF and vdW-DF2 underbind, leading to inaccurate layer separations. These results establish a clear hierarchy of vdW treatments for TMDs and provide concise, transferable guidelines for future first-principles studies of layered and hybrid materials, where dispersive interactions are essential.