In this article, we present a fragment model potential approach for the description of the crystalline environment as an extension of the use of embedding ab initio model potentials (AIMPs). The biggest limitation of the embedding AIMP method is the spherical nature of its model potentials. This poses problems as soon as the method is applied to crystals containing strongly covalently bonded structures with highly nonspherical electron densities. The newly proposed method addresses this problem by keeping the full electron density as its model potential, thus allowing one to group sets of covalently bonded atoms into fragments. The implementation in the MOLCAS 7.0 quantum chemistry package of the new method, which we call the embedding fragment ab inito model potential method (embedding FAIMP), is reported here, together with results of CASSCF/CASPT2 calculations. The developed methodology is applied for two test problems: (i) the investigation of the lowest ligand field states (2)A1 and (2)B1 of the Cr(V) defect in the YVO4 crystal and (ii) the investigation of the lowest ligand field and ligand-metal charge transfer (LMCT) states at the Mn(II) substitutional impurity doped into CaCO3. Comparison with similar calculations involving AIMPs for all environmental atoms, including those from covalently bounded units, shows that the FAIMP treatment of the YVO4 units surrounding the CrO4(3-) cluster increases the excitation energy (2)B1 → (2)A1 by ca. 1000 cm(-1) at the CASSCF level of calculation. In the case of the Mn(CO3)6(10-) cluster, the FAIMP treatment of the CO3(2-) units of the environment give smaller corrections, of ca. 100 cm(-1), for the ligand-field excitation energies, which is explained by the larger ligands of this cluster. However, the correction for the energy of the lowest LMCT transition is found to be ca. 600 cm(-1) for the CASSCF and ca. 1300 cm(-1) for the CASPT2 calculation.
Download full-text PDF |
Source |
---|---|
http://dx.doi.org/10.1021/ct7003148 | DOI Listing |
Enter search terms and have AI summaries delivered each week - change queries or unsubscribe any time!