We present a benchmarking study for the evaluation of electronic couplings in donor–bridge–acceptor systems with the Fock matrix reconstruction-bridge (FMR-B) method. We compile a data set for the benchmark that contains 29 molecules for which reliable experimental coupling values are available: DBA29. This data set is general and includes different types of donor, acceptor, and bridge units as well as different bridge lengths, and it spans a range of couplings from 0.1 to 0.8 eV. We use DBA29 to test FMR-B with 11 density functionals belonging to different classes (pure, global hybrid, and range-separated) and the Hartree–Fock (HF) method. We also test a subset of these methods with nine basis sets from the Pople and Dunning families, which include a varying number of polarization and diffuse functions. We find that the best accuracy and lowest computational cost is obtained with range-separated functionals and compact basis sets. Global hybrids with a large amount of HF exchange also work well because of error cancellation between the approximate exchange-correlation kernel and the HF part. Pure functionals, although less accurate, still provide reasonable results with a consistent underestimation of the experimental values, and they can be used for larger and more computationally demanding systems.

中文翻译：

---

FMR-B方法在供体-桥-受体系统中电子耦合的基准研究

我们用Fock矩阵重构桥（FMR-B）方法评估供体-桥-受体系统中电子耦合的基准研究。我们为基准数据汇编了一个数据集，其中包含29个分子，这些分子具有可靠的实验耦合值：DBA29。此数据集是通用数据集，包括不同类型的供体，受体和桥单元以及不同的桥长度，并且跨度范围从0.1到0.8 eV。我们使用DBA29来测试FMR-B，它具有11个属于不同类别（纯净，全局混合和距离分隔）的密度泛函和Hartree-Fock（HF）方法。我们还使用Pople和Dunning系列的9个基础集测试了这些方法的子集，其中包括变化数量的极化和漫射函数。我们发现，使用范围分隔的功能和紧凑的基集可以获得最佳的准确性和最低的计算成本。由于近似交换相关内核和HF部分之间的错误消除，具有大量HF交换的全局混合器也可以很好地工作。纯粹的功能虽然准确性较低，但仍可提供合理的结果，并且始终低估实验值，并且它们可用于更大和计算要求更高的系统。