[161], which are too reactive to be characterized by X-ray crystallography, but can
be characterized by (UV-Vis, EPR, (r)Raman, X-ray absorption) spectroscopy, and
sets challenges to be solved by a combination of theory and experiment. Being able
to know which theoretical method gives reliable results for which property is
therefore of the utmost importance.
Radon investigated recently [162] a number of iron complexes, with different
coordinating ligands: [Fe(H 2 O) 6 ]
3+ , [Fe(en) 3 ]
3+ , [Fe(tacn) 2 ]
2+ , [Fe(acac 2 trien)]
+
(en ¼ ethylenediamine, tacn ¼ 1,4,7-triazacyclononane, H 2 acac 2 trien ¼ Schiff
base obtained from the 1:2 condensation of triethylenetetramine with acetylacetone);
he used different QC methods, CCSD(T), CASPT2, NEVPT2, MRCISD+Q, and a
number of DFAs, to compare with experimental data (spin-forbidden transition
energies, spin-crossover enthalpies) that were corrected for environment effects for
a fair comparison. The main conclusions were [162] (1) the confirmation of the
validity of canonical CCSD(T), in particular when used with Kohn-Sham orbitals;
(2) systematic overstabilization of high-spin states by CASPT2, which can be
corrected partially by Pierloot/Harvey’s CASPT2/CC approach [133];
(3) NEVPT2 is performing worse than CASPT2; (4) the MRCISD+Q results depend
strongly on the size-consistency correction; and (5) few DFAs were able to give a
balanced description of all spin-state energetics (among which OPBE). In fact,
among the top performing DFAs are three developed in my group (OPBE, S12g,
SSB-D) and (surprisingly) B2PLYP-D 3 which was not reliable in other benchmark
studies; Kaupp’s local hybrid LH14t-calPBE [163] was also found among the best.
DFAs that performed well for the litmus test for spin states ([Fe
II (amp) 2 (Cl) 2 ]
0 and
[Fe
II (dpa) 2 ]
2+ ) [74] such as LC-wPBE or B97-D 3 (used by Radon with the D 2 form
for dispersion) are also present at the top. Surprisingly, neither the widely used
TPSSh [66, 152] nor Truhlar newest family [164] (MN15, MN15L) were found to
perform well. Moreover, strangely enough, while Perdew’s MVS functional [165]
performed excellently [85] for the litmus test, here it apparently gave disappointing
results (results that are similar to those of the S12h hybrid functional, which was
known [74] to fail as expected for spin states because of the inclusion of 25% HF
exchange). Finally, Radon also commented on the finding by Song and co-workers
who claimed that DFAs and CCSD(T) failed dramatically for spin-state splittings for
a number of Fe(II) complexes, for which they used Diffusion Monte Carlo (DMC)
data as reference (no experimental data were available to compare with). One of
the complexes, [Fe
II (NCH) 6 ]
2+ , had been used by others, who found, e.g., with
CCSD(T) within the complete basis set (CBS) limit that the high-spin (S ¼ 2) was
favored over the low-spin (S ¼ 0) state (i.e., ΔE HL ¼ ΔE HS – ΔE LS ¼ À2 kcal mol
À1 )
[126]. However, the DMC data favored clearly the high-spin (À27.1 kcal mol
À1 ),
close to the value of À28.9 kcal mol
À1 observed with BHandH (50% HF exchange!);
based on these results, the authors claimed that the DFAs and CCSD(T) were wrong
and DMC right. Radon argued that most likely the DMC data should not be trusted,
based among others on the results for the [Fe(tacn) 2 ]
2+ complex with a similar Fe
II N 6
octahedral coordination. This is a spin-crossover compound, favoring the low spin
experimentally at low temperatures (+3.8 kcal mol
À1 ), which was very well
reproduced by KS-UCCSD(T) data (+0.6 kcal mol
À1 ), and through an estimate for
204
M. Swart
be characterized by (UV-Vis, EPR, (r)Raman, X-ray absorption) spectroscopy, and
sets challenges to be solved by a combination of theory and experiment. Being able
to know which theoretical method gives reliable results for which property is
therefore of the utmost importance.
Radon investigated recently [162] a number of iron complexes, with different
coordinating ligands: [Fe(H 2 O) 6 ]
3+ , [Fe(en) 3 ]
3+ , [Fe(tacn) 2 ]
2+ , [Fe(acac 2 trien)]
+
(en ¼ ethylenediamine, tacn ¼ 1,4,7-triazacyclononane, H 2 acac 2 trien ¼ Schiff
base obtained from the 1:2 condensation of triethylenetetramine with acetylacetone);
he used different QC methods, CCSD(T), CASPT2, NEVPT2, MRCISD+Q, and a
number of DFAs, to compare with experimental data (spin-forbidden transition
energies, spin-crossover enthalpies) that were corrected for environment effects for
a fair comparison. The main conclusions were [162] (1) the confirmation of the
validity of canonical CCSD(T), in particular when used with Kohn-Sham orbitals;
(2) systematic overstabilization of high-spin states by CASPT2, which can be
corrected partially by Pierloot/Harvey’s CASPT2/CC approach [133];
(3) NEVPT2 is performing worse than CASPT2; (4) the MRCISD+Q results depend
strongly on the size-consistency correction; and (5) few DFAs were able to give a
balanced description of all spin-state energetics (among which OPBE). In fact,
among the top performing DFAs are three developed in my group (OPBE, S12g,
SSB-D) and (surprisingly) B2PLYP-D 3 which was not reliable in other benchmark
studies; Kaupp’s local hybrid LH14t-calPBE [163] was also found among the best.
DFAs that performed well for the litmus test for spin states ([Fe
II (amp) 2 (Cl) 2 ]
0 and
[Fe
II (dpa) 2 ]
2+ ) [74] such as LC-wPBE or B97-D 3 (used by Radon with the D 2 form
for dispersion) are also present at the top. Surprisingly, neither the widely used
TPSSh [66, 152] nor Truhlar newest family [164] (MN15, MN15L) were found to
perform well. Moreover, strangely enough, while Perdew’s MVS functional [165]
performed excellently [85] for the litmus test, here it apparently gave disappointing
results (results that are similar to those of the S12h hybrid functional, which was
known [74] to fail as expected for spin states because of the inclusion of 25% HF
exchange). Finally, Radon also commented on the finding by Song and co-workers
who claimed that DFAs and CCSD(T) failed dramatically for spin-state splittings for
a number of Fe(II) complexes, for which they used Diffusion Monte Carlo (DMC)
data as reference (no experimental data were available to compare with). One of
the complexes, [Fe
II (NCH) 6 ]
2+ , had been used by others, who found, e.g., with
CCSD(T) within the complete basis set (CBS) limit that the high-spin (S ¼ 2) was
favored over the low-spin (S ¼ 0) state (i.e., ΔE HL ¼ ΔE HS – ΔE LS ¼ À2 kcal mol
À1 )
[126]. However, the DMC data favored clearly the high-spin (À27.1 kcal mol
À1 ),
close to the value of À28.9 kcal mol
À1 observed with BHandH (50% HF exchange!);
based on these results, the authors claimed that the DFAs and CCSD(T) were wrong
and DMC right. Radon argued that most likely the DMC data should not be trusted,
based among others on the results for the [Fe(tacn) 2 ]
2+ complex with a similar Fe
II N 6
octahedral coordination. This is a spin-crossover compound, favoring the low spin
experimentally at low temperatures (+3.8 kcal mol
À1 ), which was very well
reproduced by KS-UCCSD(T) data (+0.6 kcal mol
À1 ), and through an estimate for
204
M. Swart