ADF/DFT and 1st row d-d transitions



 Hi, Netters!
 May be somebody could shed some light in a simple test of DFT and
 transition metal complexes with rather poor results.
 I am puzzled by the results obtained using ADF v. 1.0.2 (Aug. 1993):
 I am attempting to reproduce calculations on the spectroscopy of
 M(H2O)6 (+n), where M is a first row transition metal, and n varies
 from 2 to 3.
 The geometry is Th (although, the highest compatible symmetry
 available in ADF is D2h). This leaves the M-O, O-H, and M-O-H
 as the only internal coordinates necessary to specify the whole system.
 Th symmetry is forced through proper construction of the Z-mat. The
 density is symmetry constrained up to D2h symmetry.The excited states are
 computed at the optimized ground state geometry.
 I am using the II basis set (double zeta Slater functions with triple
 zeta for 3d orbitals in the metal), and the default xc functional
 (implementaion of VWN parameterization of CA MC simulations of an
 electron gas). All computations are unrestricted, including those
 with singlet states for the purpose of proper comparison with the
 excited states. The frozen core approximation is used with the core
 of the Metal corresponding to [Ne], so the 3s,3p electrons are also
 part of the valence shell.
 To speeds things: integration = 2.5 (lowest adequate accuracy for integrals),
 scfconvergence 0.001, gconv = 0.01 with M-O and H2O geometries as initial
 input corresponding to standard H2O geometry and M-O 1.95 - 2.15 Angstroms.
 This is what I get:
 	M	Configuration (&) 	   State	Energy (x 1000 cm-1)
  ----------------------------------------------------------------------------
 	Cr+3      d3 t2g(3)                 4A1.g              0
                      t2g(2) eg              4Tx.g(?)     16.9  (17.4, 17.6)*
 	Mn+2      d5 t2g(3) eg(2)           6A1.g              0
 		  d5 t2g(4) eg              4Tx.g(?)     12.1  (18.9)
 	Fe+3      d5 t2g(3) eg(2)           6A1.g              0
                      t2g(4) eg              4Tx.g(?)      9.7  (12.6)
         Co+3      d6 t2g(6)                 1A1.g              0
                      t2g(5) eg              1Tx.g(?)      9.7  (16.6, 18.9)
 (&) In Th symmetry the nomenclature Eg/u Tg/u is retained, but in D2h
 symmetry is reduced and Eg -> A1.g, with T2g -> B1.g + B2.g + B3.g. The
 ground state orbital energies do show BX.g to have proper degeneracies.
 Similarly, the two HOMO's in A1.g symmetry are degenerate.
 (?) x=1 or 2 the exact definition of the state is unclear to me. See below
 for the orbital occupation numbers.
 (*) values in parenthesis are from Anderson et.al. Inorg. Chem. 1986, 25,
 2728-2732. These are experimental results that a new version of INDO/S
 is able to reproduce.
 I am surprised that not even a qualitative trend is reproduced. Extensions
 to include Stoll's correction or increase the basis set to add polarization
 functions in the light atoms (Basis Set III) do not appear to improve the
 situation.
 Could you comment on these results? Are DFT methods inadequate to address
 spectroscopy questions such as the levels of d-d transitions?
 Do any HF based methods be adequate (e.g. HF/MP2, MCSCF etc)? We are
 doing this work as a test of the reliability of computations we intend
 to do on the spin density of metalloporphyrins and some Lanthanide complexes
 that have not been parameterized within the INDO/S formalism.
 Below I have reproduced the occupation numbers used in my calculations as
 specified in ADF.  spin 1 // spin 2
      			Cr
 A1.G	7 // 7 			A1.G	8 // 7
 B1.G	3 // 2			B1.G	3 // 2
 B2.G	3 // 2			B2.G	2 // 2
 B3.G	3 // 2			B3.G	3 // 2
 A1.U	0 // 0			A1.U	0 // 0
 B1.U	5 // 5			B1.U	5 // 5
 B2.U	5 // 5			B2.U	5 // 5
 B3.U	5 // 5			B3.U	5 // 5
 			Mn
 A1.G	9 // 7 			A1.G	8 // 7
 B1.G	3 // 2			B1.G	3 // 2
 B2.G	3 // 2			B2.G	3 // 3
 B3.G	3 // 2			B3.G	3 // 2
 A1.U	0 // 0			A1.U	0 // 0
 B1.U	5 // 5			B1.U	5 // 5
 B2.U	5 // 5			B2.U	5 // 5
 B3.U	5 // 5			B3.U	5 // 5
 			Fe
 A1.G	9 // 7 			A1.G	8 // 7
 B1.G	3 // 2			B1.G	3 // 2
 B2.G	3 // 2			B2.G	3 // 3
 B3.G	3 // 2			B3.G	3 // 2
 A1.U	0 // 0			A1.U	0 // 0
 B1.U	5 // 5			B1.U	5 // 5
 B2.U	5 // 5			B2.U	5 // 5
 B3.U	5 // 5			B3.U	5 // 5
 			Co
 A1.G	7 // 7 			A1.G	8 // 7
 B1.G	3 // 3			B1.G	3 // 3
 B2.G	3 // 3			B2.G	2 // 3
 B3.G	3 // 3			B3.G	3 // 3
 A1.U	0 // 0			A1.U	0 // 0
 B1.U	5 // 5			B1.U	5 // 5
 B2.U	5 // 5			B2.U	5 // 5
 B3.U	5 // 5			B3.U	5 // 5
 Gus Mercier
 mercie "at@at" cumc.cornell.edu