We have studied the accuracy of 16 different density functional theory methods to reproduce experimental data for bond lengths, angles, vibrational frequencies, as well as enthalpies and entropies for the binding of N₂, H₂, CO and hydride ions to various transition-metal complexes (with Fe, Ni, Cr, Mo and W) with relation to nitrogenase. We show that generalized gradient approximation functionals give better structure-related parameters, whereas hybrid functionals often give better energies. However, the BLYP and B97D functionals seem to give reasonably accurate results for both types of properties. Geometries and entropies are converged with split-valence basis sets, but energies and vibrational frequencies (and therefore also thermal corrections) in general require a basis set of triple-zeta quality. Dispersion corrections are important to obtain accurate energies (contributing by up to 57 kJ mol⁻¹), as well as structures.

我们研究了 16 种不同密度泛函理论方法的准确性，以重现键长、角度、振动频率以及 N ₂、H ₂结合的焓和熵的实验数据、CO 和氢化物离子与固氮酶相关的各种过渡金属络合物（含 Fe、Ni、Cr、Mo 和 W）。我们表明，广义梯度近似泛函提供更好的结构相关参数，而混合泛函通常提供更好的能量。然而，BLYP 和 B97D 泛函似乎对这两种类型的属性都给出了相当准确的结果。几何和熵与分裂价基组收敛，但能量和振动频率（因此还有热校正）通常需要三 zeta 质量的基组。色散校正对于获得准确的能量（贡献高达 57 kJ mol ^-1）以及结构非常重要。