Giner 2018

Subscriber access provided by NORTH CAROLINA A&T UNIV
Quantum Electronic Structure

Interplay between electronic correlation and metal-ligand
delocalization in the spectroscopy of transition metal
compounds: case study on a series of planar Cu2+complexes.
Emmanuel Giner, David Tew, Yann Garniron, and Ali Alavi
J. Chem. Theory Comput., Just Accepted Manuscript • DOI: 10.1021/acs.jctc.8b00591 • Publication Date (Web): 22 Oct 2018
Downloaded from http://pubs.acs.org on October 23, 2018
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a service to the research community to expedite the dissemination
of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in
full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully
peer reviewed, but should not be considered the official version of record. They are citable by the
Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore,
the “Just Accepted” Web site may not include all articles that will be published in the journal. After
a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web
site and published as an ASAP article. Note that technical editing may introduce minor changes
to the manuscript text and/or graphics which could affect content, and all legal disclaimers and
ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or
consequences arising from the use of information contained in these “Just Accepted” manuscripts.
is published by the American Chemical Society. 1155 Sixteenth Street N.W.,

Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.
Page 1 of 42 Journal of Chemical Theory and Computation
1
2
3
4
5
6
7
8
9
Interplay between electronic correlation and
10
11
12 metal-ligand delocalization in the spectroscopy of
13
14
15
16
transition metal compounds: case study on a
17
18
19 series of planar Cu2+ complexes.
20
21
22
23
Emmanuel Giner,∗,† David P. Tew,‡ Yann Garniron,¶ and Ali Alavi‡
24
25
26
†Laboratoire de Chimie thorique, Sorbonne Universit, UMR 7616, 4 place Jussieu, 75252
27
Paris, France
28
29
‡Max Planck Institute for Solid State Research, Heisenbergstraβe 1, 70569 Stuttgart
30
31
¶Laboratoire de Chimie et Physique Quantique, UMR 5626, Universit Paul Sabatier, 118
32
33
route de Narbonne 31062 Toulouse, France
34
35
36
E-mail: emmanuel.giner@lct.jussieu.fr
37
38
39
40
Abstract
41
42
43 We present a comprehensive theoretical study of the physical phenomena that deter-
44
45 mine the relative energies of three of the lowest electronic states of each of the square-
46
47 planar copper complexes [CuCl4 ]2− , [Cu(NH3 )4 ]2+ and [Cu(H2 O)4 ]2+ , and present a
48
49 detailed analysis of the extent to which truncated configuration interaction (CI) and
50
51 coupled cluster (CC) theories succeed in predicing the excitation energies. We find
52
53 that ligand-metal charge transfer (CT) single excitations play a crucial role in the cor-
54
55
rect determination of the properties of these systems, even though the first impact of
56
these CT on the energetics of these systems appears at fourth order in perturbation
57
58
59 1
60 ACS Paragon Plus Environment
Journal of Chemical Theory and Computation Page 2 of 42
1
2
3
theory. We provide a minimal selected CI space for describing these systems with
4
5 multi-reference theories and use a high-order perturbation theory analysis within this
6
7 space to derive a simple and general physical picture for the LMCT process. We find
8
9 that CCSD energy differences agree very well with near full CI values even though the
10
11 D1 diagnostics are large, which casts doubt on the usefulness of singles-amplitude based
12
13 multi-reference diagnostics. CISD severely underestimates the excitation energies and
14
15 the failure is a direct consequence of the size-inconsisency errors in CISD. Finally, we
16
present reference values for the energy differences computed using explicitly correlated
17
18 CCSD(T) and BCCD(T) theory.
19
20
21
22
23
24
1 Introduction
25
26
27 Open-shell transition metal complexes, which are ubiquitous in biological and industrial
28
29 chemistry, represent one of the main challenges for present-day quantum chemistry, where
30
31 theory seeks to provide prediction and interpretation of key properties such as electronic
32
33 transition energies, spin-density maps and magnetic anisotropy. Complexes containing Cu2+
34
35 have been studied extensively, using Density Functional and wavefunction theories, 1–19 and
36
37 have been found to pose a tough test for electronic structure methods. Popular functionals
38
39 such as B3LYP and BP86 systematically underestimate the spin-density at the Cu atom,
40
41 provide poor d−d and ligand-to-metal exctiation energies 8,10,11 and misleading predictions of
42
43 magnetic anisotropy tensors. 8,9,13,20 Although it is possible to design tailored functionals for
44
45 these systems, with higher percentages of Hartree–Fock exchange, this pragmatic approach
46
47 has limited transferability and limited predictive power.
48
49
50 On the other hand, studies using wavefunction methods have also only been partially
51
52 successful. Transtition metal complexes are considered to be strongly correlated systems and
53
54 Complete Active Space Self Consistent Field (CASSCF) theory is usually applied, with multi-
55
56
reference perturbation or truncated CI corrections for dynamic correlation. The computed
57
58
59 2
1
2
3
energies are found to be highly sensitive to the choice of active space and the level of coupling
4
5
between the treatment of static and dynamic correlation, but the number of orbitals involved
6
7
in the coordination at the transition metal centre prohibits brute force convergence with
8
9 respect to the size of the active space. Although the relatively high density of low lying
10
11 electronic states and the large values of T1 diagnostics observed for transition metal complexes
12
13 discourages the use of single reference methods, the accuracy of single reference coupled-
14
15 cluster methods for these sytems remains an open question. This paper reports the results
16
17 of a series of careful benchmark calculations and detailed theoretical analysis, performed
18
19 on three Cu2+ complexes [CuCl4 ]2− , [Cu(NH3 )4 ]2+ and [Cu(H2 O)4 ]2+ , which enables us to
20
21 extend and generalise findings reported previously by one of us on the role of LMCT in metal-
22
23 ligand covalency and spin-densities of open-shell systems 18,21 and to provide a simple physical
24
25 picture of LMCT in the framework of perturbation theory. We address the question of what
26
27 characteristics an electronic structure method should have in order to correctly describe the
28
29 lower lying electronic states and the spin-densities at the Cu atom and analyse the sucesses
30
31 and failures of commonly applied single-reference and multi-reference wavefunction methods
32
33 for this class of open-shell transition metal complexes.
34
35
36 The three complexes [CuCl4 ]2− , [Cu(NH3 )4 ]2+ and [Cu(H2 O)4 ]2+ are all square-planar
37
38 coordinated and have a doublet ground state with a 3dx2 −y2 singly-occupied molecular orbital
39
40 (SOMO), which has the largest repulsion with the ligand lone-pairs that point at the Cu atom
41
42 along the x and y axes. Two of the three complexes, [CuCl4 ]2− and [Cu(NH3 )4 ]2+ , have been
43
44 studied extensively and the EPR spectra, spin density, g-tensor and electronic excitation
45
46 energies are well characterised experimentally. 22–25 The schematic ligand-field diagram 26 is
47
48 displayed in Fig. 1. The low-lying excited states all correspond to doublets where one of the
49
50 more low lying d orbitals becomes the SOMO.
51
52
In multi-reference computational studies of two of these systems 9,14 Neese et al. and
53
54
Pierloot et al. observed that in order to correctly describe the electronic spectrum and mag-
55
56
netic properties it is necessary to include the ligand donor orbital in the active space, even
57
58
59 3
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28 (a)
29
30 Figure 1: Crystal-field theoretical ordering of the orbitals and the orbital occupation char-
31
32
acteristics of the ground state and first (red arrow) and third (blue arrow) lowest d − d
33 electronic transitions.
34
35
36 though this orbital is doubly occupied and has a relatively low orbital eigenvalue. They also
37
38 found that CASPT2 performs poorly and sophisticated methods such as SORCI 9 or MS-
39
40 CASPT2 14 are required, which couple the dynamic correlation into the multi-reference treat-
41
42 ment. The general importance of ligand donor orbitals was highlighted by Nieuwpoort, Broer
43
44 and coworkers in their pioneering work on cluster models of transition metal oxides, 1,2,7,27
45
46 where they showed that ligand-to-metal charge transfer (LMCT), and associated orbital re-
47
48 laxation, forms a significant component of the wavefunction. Many subsequent studies have
49
50 confirmed the importance of LMCT in a range of transition metal systems, 3,6,12,28,29 and
51
52 recent work by one of us 17–19,21 found analogous correlation mechanisms in several open-
53
54 shell systems, both inorganic and organic. A common observation in all of these studies
55
56 is that the extent of metal-ligand delocalization can increase considerably as higher-order
57
58
59 4
1
2
3
correlation effects are taken into account, and the question of what level of theory is required
4
5
remains open.
6
7
8 In this work we provide a detailed wavefunction analysis of the role of LMCT in d − d
9
10 excitation energies and spin-densities of the [CuCl4 ]2− , [Cu(NH3 )4 ]2+ and [Cu(H2 O)4 ]2+ com-
11
12 plexes, and the extent to which these processes are captured in commonly applied methods.
13
14 The paper is organized as follows. Section 2 reports benchmark near FCI calculations on the
15
16 three lowest energy states of [CuCl4 ]2− , [Cu(H2 O)4 ]2+ and [Cu(NH3 )4 ]2+ and an analysis of
17
18 the wavefunctions and LMCT is presented in section 3. In section 4 we discuss the perfor-
19
20 mance of multi-reference methods for these systems. We address the appropriate minimal
21
22 selected CI space required to capture the dominant physical processes at work, and pro-
23
24 vide a simplified general perturbational analysis that reveals the key features of LMCT in
25
26 this class of open-shell transition metal complexes. Single reference methods are discussed
27
28 in section 5, where we demonstrate that coupled-cluster methods perform very well, but
29
30 that non-size-extensivity errors severvely degrade theoretical predictions using truncated CI
31
32 methods, even for these small molecules near their equilibrium geometries. In section 6, near
33
34 basis set limit reference transition energies are reported. Our conclusions are summarised in
35
36 section 7.
37
38
39
40
41 2 Benchmark near FCI energies and wave functions
42
43
44
45 Near FCI wavefunctions in the 6-31G basis set for the ground, first and third electronic states
46
47 of each of the three complexes were computed using the CIPSI algorithm of the Quantum
48
49 Package software. 30 D2h symmetry (D2 for [Cu(NH3 )4 ]2+ ) was used and each state is the
50
51 lowest energy state in the symmetry block to which it belongs. The electronic state where
52
53 the dz2 is singly occupied is in the same symmetry block as the ground state and therefore
54
55
could not be studied with the available software. He, Ne and Ar cores were frozen in the
56
57
58
59 5
1
2
3
nitrogen, chlorine and copper atoms, respectively, resulting in 41 electrons in 50, 66 and 74
4
5
orbitals for the [CuCl4 ]2− , [Cu(H2 O)4 ]2+ and [Cu(NH3 )4 ]2+ molecules, respectively. ROHF
6
7
orbitals were used to ease comparison of the wavefunction parameters with those of CASSCF,
8
9 targeted CI and CC-based wavefunctions. The geometries of [CuCl4 ]2− and [Cu(NH3 )4 ]2+
10
11 were taken from Ref. 14. The geometry of [Cu(H2 O)4 ]2+ was optimized with D2h symmetry
12
13 at the unrestricted PBE 31 level of theory using a 6-31G* basis set. Cartesian structures of
14
15 all three complexes are listed in the supporting information.
16
17
18 The CIPSI approach approximates the FCI energy through an adaptively refined selected
19
20 CI procedure, corrected for discarded determinants through second-order multi-reference
21
22 perturbation theory. The CIPSI class of methods build upon selected CI ideas 32–38 and have
23
24 been successfully used to converge to FCI correlation energies, one-body properties and nodal
25
26 surfaces. 36,39–46 The CIPSI algorithm used in this work uses iteratively enlarged selected CI
27
28 and Epstein–Nesbet 47,48 multi-reference perturbation theory. The CIPSI energy is
29
30
31
32 ECIPSI = Ev + E (2) (1)
33 hΨ(0) |H|Ψ(0) i
34 Ev = min (2)
35 {cI } hΨ(0) |Ψ(0) i
36 X hΨ |H|µi2
(0)
X
37 E (2) = = e(2)
µ (3)
38 µ
Ev − hµ|H|µi µ
39 X
40 |Ψ(0) i = cI |Ii (4)
41 I∈R
42
43
44 where I denotes determinants within the CI reference space R and µ a determinant outside
45
46 it. To reduce the cost of evaluating the second-order energy correction, the semi-stochastic
47
48 multi-reference approach of Garniron et al 49 was used, adopting the technical specifications
49
50 recommended in that work. The CIPSI energy is systematically refined by doubling the size
51
52 of the CI reference space at each iteration, selecting the determinants µ with the largest
53 (2)
54 |eµ |, and the energy monitored as a function of the size of the reference space.
55
56
57
58
59 6
1
2
3
4
2.1 Reference near FCI energies
5
6
7 Fig. 2 plots convergence with respect to size of the CIPSI reference wave function for the elec-
8
9 tronic transitions of the [CuCl4 ]2− [Cu(H2 O)4 ]2+ and [Cu(NH3 )4 ]2+ complexes, up to 32×106
10
11 Slater determinants for the [CuCl4 ]2− , [Cu(H2 O)4 ]2+ and [Cu(NH3 )4 ]2+ molecule. Both Ev
12
13 and ECIPSI are displayed, the total energies can be found in the supporting information file.
14
15 Here and throughout D4h and D2d symmetry labels are used for the electronic states. The
16
17 CIPSI electronic transition energies for the [CuCl4 ]2− and [Cu(H2 O)4 ]2+ molecules are con-
18
19 verged with a sub-mH precision within 8×106 Slater determinants and the variational CI
20
21 transition energies agree with the CIPCI values to within 1 mH. Due to the larger Hilbert
22
23 space, the convergence for the [Cu(NH3 )4 ]2+ molecule is significantly slower. The variational
24
25 CI energy difference is not converged even using 32×106 Slater determinants, but the CIPSI
26
27 values do appear converged to within 1 mH with 32×106 Slater determinants, underlining
28
29 the importance of the second order correction to the energy. We note that in all cases there
30
31
is a clear trend: increasing the CI reference space increases the energy differences, which in-
32
33
dicates that the ground state has a larger correlation energy than that of the excited states
34
35
for each of the molecules. The transition energies are listed in Table 1, together with values
36
37
from several single and multi-reference methods, which are discussed in sections 4 and 5.
38
39 Table 1: Computed excitation energies (mH) for the [CuCl4 ]2− , [Cu(NH3 )4 ]2+ and
40 [Cu(H2 O)4 ]2+ molecules in the 6-31G basis and the frozen core approximation (see text
41 for details).
42
43 Transition ROHF UHF CISD CISD(SC)2 CCSD BCCD CCSD(T) BCCD(T) CAS(9,10) CAS(11,11) FOBOCI CIPSI
44 [CuCl4 ]2−
2
45 B1g − 2 B2g 30.2 30.8 38.3 43.6 43.2 42.6 43.9 43.8 31.1 39.7 43.2 42.0(1)
2
Eg − 2 B2g 38.2 39.0 46.7 52.2 51.9 51.3 52.6 52.6 39.9 48.8 51.4 52.1(2)
46
[Cu(H2 O)4 ]2+
47 2
B1g − 2 B2g 42.4 42.9 48.1 51.6 51.6 51.3 52.1 52.1 44.2 48.9 49.9 51.5(1)
48 2
Eg − 2 B2g 44.8 45.3 48.5 50.5 50.7 50.6 50.9 50.9 46.7 50.8 50.7 50.5(1)
49 [Cu(NH3 )4 ]2+
50 2
B1 − 2 B2 50.7 51.6 60.2 68.9 68.5 67.8 69.7 69.8 53.0 62.6 66.0 68.0(1)
2
51 E − 2 B2 62.7 63.7 72.1 80.5 80.5 79.8 81.6 81.7 65.6 75.3 78.3 79.9(1)
52
53
54
55
56
57
58
59 7
1
2
3
4
2.2 Composition of the ground state wave functions
5
6
7 The composition of the near FCI ground state wave functions on the [CuCl4 ]2− , [Cu(H2 O)4 ]2+
8
9 and [Cu(NH3 )4 ]2+ complexes present strong similarities in their dominant components: in all
10
11 cases there are clearly two leading Slater determinants, the ROHF determinant and a single
12
13 excitation where an electron has been excited from a doubly occupied ligand-based MO to the
14
15 3dx2 −y2 SOMO on the copper centre. Table 2 lists the amplitudes of these single excitations,
16
17 extracted from the largest CIPSI wave function. These singly excited determinants are
18
19 identified as a LMCT component of the ground state wavefunction and the orbitals involved
20
21 are plotted in Figs 3, 4 and 5. Since the ROHF orbitals are reasonably well localized on the
22
23 Cu atom or on the ligands, one can analyze the physical content of the CIPSI wave functions
24
25 in terms of valence bond (VB) structures. In all three ground states the ROHF determinant
26
27 corresponds to a VB form of the type Cu2+ X4 and the LMCT components correspond to a
28
29 set of four equivalent VB structures of the type Cu+ X+ X3 , where X denotes the ligand. In
30
31
the ROHF wavefunction, the spin density is concentrated at the copper atom, whereas in
32
33
the FCI wavefunction, the LMCT excitations delocalise the spin-density onto the ligands.
34
35
The spin-densities are listed in Table 3.
36
37 Table 2: Coefficient of the largest single excitations at various levels of theory in the 6-31G
38 basis set.
39
40 Electronic State CISD CISD(SC)2 CCSD FOBOCI CIPSI
41 [CuCl4 ]2−
42 2
B2g 0.071 0.175 0.165 0.149 0.156
43 2
44
B1g 0.032 0.083 0.078 0.064 0.069
2
45 Eg 0.031 0.089 0.085 0.078 0.074
46 [Cu(H2 O)4 ]2+
47 2
B2g 0.028 0.076 0.075 0.065 0.060
48 2
B1g 0.001 0.028 0.028 0.012 0.006
49 2
50 Eg 0.005 0.058 0.056 0.050 0.035
51 [Cu(NH3 )4 ]2+
52 2
B2 0.043 0.141 0.137 0.117 0.113
53 2
B1 0.001 0.015 0.033 0.016 0.009
54 2
55
E 0.001 0.014 0.028 0.014 0.006
56
57
58
59 8
1
2
3 Table 3: Spin density on the copper atom at various levels of theory using Mulliken popula-
4
tion analysis in the 6-31G basis set.
5
6 State ROHF UHF CISD(SC)2 CISD BCCD CAS(9,10) CAS(11,11) FOBOCI CIPSI BCCDa
7 [CuCl4 ]2−
8 2
B2g 0.93 0.91 0.80 0.89 0.80 0.93 0.86 0.81 0.81 0.62
9 2
B1g 0.99 1.02 0.99 1.00 1.00 0.99 0.98 0.99 0.99 0.93
10 2
Eg 0.99 1.01 0.99 1.00 0.99 0.99 0.98 0.99 0.99 0.92
11 [Cu(H2 O)4 ]2+
12 2
B2g 0.96 0.94 0.91 0.94 0.91 0.95 0.85 0.91 0.92 0.96
13 2
B1g 0.99 1.01 1.00 1.00 1.00 0.99 0.98 1.00 1.00 1.00
14 2
Eg 0.99 1.00 0.99 1.00 0.99 0.99 0.96 0.99 0.99 0.97
15 [Cu(NH3 )4 ]2+
16 2
B2 0.92 0.90 0.83 0.90 0.81 0.93 0.87 0.82 0.84 0.73
17 2
B1 0.99 1.02 1.02 1.00 1.02 0.99 0.99 1.01 1.01 1.02
18 2
E 0.99 1.02 1.02 1.00 1.02 0.99 0.99 1.01 1.01 1.03
19
20
21 a
From BCCD(T)-F12/aug-cc-pwCVTZ-DK calculations in section 6
22
23
24
25
26 2.3 Composition of the excited state wave functions
27
28
29 The composition of the CIPSI wave function for the various excited states presents strong
30
31 similarities with the ground state wave functions. In all cases, LMCT single excitations
32
33 appear where an electron is transfered from a doubly occupied ligand orbital with the same
34
35 symmetry as the SOMO, to the SOMO at the copper centre. The orbitals involved in the
36
37 LMCT processes are displayed in Figs. 9–14 and the amplitudes are reported in Table
38
39 2, as extracted from the largest CIPSI wavefunctions. The LMCT excitations are evidently
40
41 important in the 2 B1g and 2 Eg excited states of [CuCl4 ]2− and in the 2 Eg state [Cu(H2 O)4 ]2+ ,
42
43 but with an amplitude half the magnitude of that of the ground state, due to the weaker
44
45 overlap of the ligand and metal orbitals. The amplitudes of the LMCT excitations for the
46
2
47 B1g state of [Cu(H2 O)4 ]2+ and both states of [Cu(NH3 )4 ]2+ are between one and two orders
48
49 of magnitude smaller.
50
51
52
53
54
55
56
57
58
59 9
1
2
3
4
3 Perturbation theory analysis of LMCT
5
6
7 Our benchmark near FCI wavefunctions and energies in the 6-31G basis set reveal that the
8
9 ground state is more correlated than the excited states in all three complexes, and that the
10
11 LMCT processes are stronger in the ground state than in the excited states. In this section
12
13 we analyse in greater depth the role of electron correlation and LMCT in the transition
14
15 energies from the perspective of single reference perturbation theory. Here and throughout,
16
17 all the orbitals doubly occupied in the ROHF Slater determinant are referred as i, the ligand
18
19 donor orbital for each state is called L, the SOMO as S and the virtual orbitals as a. Also,
20
1
21 the Sz component is assumed to be 2
so the unpaired electron has α spin. The LMCT
22
23 determinant is
24
25 |LMCTi = a†Sβ aLβ |ROHFi (5)
26
27
28 The prevalence of the LMCT determinants in all ground and some excited states wave-
29
30 functions can be considered as quite unusual at least for two reasons. First, all LMCT
31
32 determinants are found in our calculations to be more than 12 eV higher in energy than the
33
34 ROHF determinant, which is clearly not a near degeneracy situation. Second, the coefficient
35
36 of the LMCT determinant at first-order in Møller–Plesset perturbation (MP) theory 50 is
37
38
39 (1) hLMCT|H|ROHFi
cLMCT = =0 (6)
40 (0)
E0 − ELMCT
(0)
41
42
43 which vanishes because of the Brillouin Theorem. The large coefficients of the LMCT de-
44
45 terminant in the near FCI wave function come necessarily from their interactions with de-
46
47 terminants of higher excitation rank. The first non-vanishing contribution to the LMCT
48
49 coefficient appears at second order in the MP expansion:
50
51
52 (2)
X hLMCT|H|Di (1)
X hLMCT|H|Di hD|H|ROHFi X
53 cLMCT = (0) (0)
cD = (0) (0) (0) (0)
≡ δcD (7)
54 D E0 − ELMCT D E0 − ELMCT E0 − ED D
55
56
57
58
59 10
1
2
3
The contribution δcD of each double excitation |Di to the second-order coefficient can be
4
5
used to identify the most important double excitations for the LMCT. The largest values
6
7
of |δcD | correspond in all three states of all three molecules to the specific class of double
8
9 excitations that are single excitations from the LMCT determinant
10
11
12
†
13 |SL a,σ † †
i,σ i ≡ aaσ aiσ aSβ aLβ |ROHFi = aaσ aiσ |LMCTi (8)
14
15
16
17 Among these, the largest |δcD | occur when i is a 3d and a a 4d orbital, where the the
18
19 interaction elements hLMCT|H|SL ai i are found to be around 7 eV, which is very large for
20
21 off-diagonal Hamiltonian matrix elements. The large magnitude can be easily understood:
22
23 applying the Brillouin-Theorem and neglecting minor exchange contributions, the pertinent
24
25 matrix elements are
26
27 hLMCT|H|SL a,σ
i,σ i ≈ (ia|SS) − (ia|LL) (9)
28
29
30
where the standard chemical notation is used for the two-electron repulstion integrals. The
31
32
integrals (ia|SS) are very large (typically between 7 and 8 eV in our calculations) since
33
34
all orbitals are located at the copper atom, and the integrals (ia|LL) are small (typically
35
36
between 0.1 and 0.5 eV) since the distributions ia and LL are centered on different atomic
37
38
sites.
39
40 The fact that the single excitations from |LMCTi are important can be interpreted
41
42 physically as the need to relax the orbitals of the |LMCTi determinant, which has been
43
44 discussed extensively in a variety of contexts, most notably in the theoretical determina-
45
46 tion of magnetic coupling constants. 2,5,7,12,19,51 The ROHF orbitals are not optimal for the
47
48 |LMCTi determinant since the former represents the copper atom in its Cu2+ , wheras the
49
50 latter represents the copper atom in its Cu+ state where the orbitals are more diffuse. This is
51
52 nothing other than the breathing-orbital effect, well known in the VB framework. 52 An other
53
54 evidence of this physical effect is given by equation (9) which can be written as the coulomb
55
56 interaction between the distribution i(r1 )a(r1 ) with the dipole distribution S 2 (r2 ) − L2 (r2 ),
57
58
59 11
1
2
3
the latter corresponding to the leading term in the change of the electrostatic field between
4
5
the LMCT and ROHF determinants. These considerations all point to a subtle interplay
6
7
between electronic correlation and metal-ligand delocalization in the spectroscopy of these
8
9 transition metal complexes.
10
11
12
13
14
15
4 Multi-reference methods
16
17
18 When wavefunction approaches are applied to transition metal systems, multi-reference
19
20 (MR) methods are usually selected. The results obtained often depend critically on the
21
22 choice of active space and in this section we examine the influence of the active space on
23
24 the transition energies and spin densities. All CASSCF calculations were performed using
25
26 the Gamess(US) sofware 53 and all other CI calculations were performed using the Quantum
27
28 Package sofware. 30
29
30
31
32
33
4.1 CASSCF
34
35
36 A common choice of active space in transition metals is the so-called ’double d-shell’, which
37
38 involves all valence 3d electrons in the 3d and 4d orbital sets. For 3d9 copper complexes, this
39
40 results in a CAS(9,10), nine electrons in ten orbitals. The CASSCF transition energies and
41
42 spin densities are reported in Tables 1 and 3, respectively, and compared to the corresponding
43
44 near FCI values. Although this active space captures the dominant dynamical correlation of
45
46 the the 3d electrons, it is often insufficient for accurate results and this is also the case here.
47
48 The computed electronic transition at the CAS(9,10) level are 8 to 15 mH too low and the
49
50 spin density on the copper atom is overestimated, with almost no improvement over ROHF
51
52 for both quantities.
53
54 The analysis in the previous section highlights the importance of LMCT single excitations
55
56 3d → L, which are missing from the CAS(9,10) active space that contains only 3d → 4d
57
58
59 12
1
2
3
excitations. Adding the ligand donor orbital L to the active space results in a CAS(11,11)
4
5
and the CASSCF transition energies and spin densities are also reported in Tables 1 and 3.
6
7
The results are substantially improved due to the presence of LMCT single excitations in
8
9 the active space, but significant deviations from the reference CIPSI values remain.
10
11
12
13
14 4.2 A minimal CI space: FOBOCI
15
16
17 The perturbation analysis of section 3 suggests that it is possible to define a minimal selected
18
19 CI, refered to as FOBOCI 18,21,54 (first order breathing orbital CI) that contains the dominant
20
21 physical effects related to the LMCT determinant. The minimal CI should contain the ROHF
22
23 and LMCT determinants, together with all single excitations from these two configurations
24
25 to introduce the necessary orbital relaxation. The FOBOCI therefore contains all single
26
27 excitations and all double excitations of type |SL ai i.
28
29
30 The results obtained at the FOBOCI level for the electronic transitions are reported
31
32 in Table 1, the amplitude of the LMCT determinants in the FOBOCI wave function are
33
34 reported in Table 2 and the spin density on the copper atom is reported in Table 3. The
35
36 FOBOCI electronic transition energies are remarkably close to the near FCI values, with a
37
38 mean error of 1.2 mH and a maximum error of 2 mH. The amplitudes of the LMCT deter-
39
40 minants in the FOBOCI wave functions are also close to those of the CIPSI wave function,
41
42 as are the spin densities at the copper atom. The FOBOCI wavefunction clearly contains
43
44 the dominant differential physical effects involved in the spectroscopy of these complexes,
45
46 correctly balancing electron correlation and spin-delocalisation.
47
48 The success of the FOBOCI wavefunctions is even more remarkable considering that
49
50 they contain least 4 orders of magnitude fewer determinants than the CIPSI wave functions:
51
52 the largest FOBOCI wave function contains 1072 Slater determinants in the case of the
53
54 [Cu(NH3 )4 ]2+ molecule, compared to 32 ×106 Slater determinants of the CIPSI wave func-
55
56 tions. Although the FOBOCI wavefunction only recovers about 3% of the total correlation
57
58
59 13
1
2
3
energy of each state, this small fraction of the correlation contribution has a large differential
4
5
effect on the energies of the ground and excited states.
6
7
8 It is also interesting to note that CAS(11,11) performs systematically worse than FOBOCI,
9
10 even though the CAS(11,11) total energies are ∼0.1 Hartree below the FOBOCI values. The
11
12 main difference between the FOBOCI and the CAS(11,11) wavefunctions is that the former
13
14 contains |Sa
Li i determinants with i and a located on the ligands, which are missing from the
15
16 CAS(11,11) active space. The |Sa
Li i determinants with i and a are 3d and 4d orbitals allow
17
18 for the dilatation of the copper orbitals due to the transfer of charge from the ligand to the
19
20 copper, and the |Sa
Li i determinant with i and a located on the ligands allows for the cor-
21
22 responding relaxation of the ligand orbitals. Both are required for quantitative agreement
23
24 with the near FCI results.
25
26
27
28
29
4.3 Perturbation analysis of FOBOCI
30
31
32 Having established the accuracy and reliability of FOBOCI, this greatly simplified wavefunc-
33
34 tion can be analysed in detail to gain further insight into the relative levels of correlation-
35
36 induced spin-delocalisation among the low-lying states. We use the Møller-Plesset perturba-
37
38 tion series for this purpose, which corresponds to a Taylor expansion of the CI equations in
39
40 this subspace. At second order, neglecting exchange integrals, and the singles contribution,
41
42 the FOBOCI energy is
43
44
45
X Sa (1)
X (LS|ia)2
e(2) = cLi hROHF|H|SL ai i ≈ (10)
46
i,a i,a
L − S + i − a
47
48
49 The diagramatic representation is displayed in Figure 6. As previously noted, the second-
50
51 order energy already shows a differential role between the ground state and the excited
52
53 states. The electrostatic interaction between the SOMO S and donor ligand L orbitals
54
55 dictates the crystal field splitting and is therefore larger in the ground state than in the
56
57
58
59 14
1
2
3
excited states. The integrals (LS|ia) are correpondingly larger in the ground state, resulting
4
5
in a larger correlation energy, which raises the electronic transition energy. Figures 3, 4,
6
7
5, 9, 10, 11, 12, 13 and 14 depict the SOMO and ligand donor orbitals. This perturbation
8
9 perspective can be connected to the VB picture through a decomposition in terms of strongly
10
11 localised orbitals. This analysis is somewhat involved and is summarised in the supporting
12
13 information. The conclusions are that two physical effects are at work already at second
14
15 order: a small dispersive interaction between the ligand lone pairs and the electron in the
16
17 SOMO; and a comparatively large breathing orbital relaxation.
18
19
20 The first contribution to the energy from the LMCT determinant is obtained at fourth
21
22 order: the second order LMCT coefficient modifies the coefficients of the double excitations
23
(2)
24 at third order. Neglecting minor exchange contributions, the second order coefficient cLMCT
25
26 is
27 X (LS|ia) (ia|SS) − (ia|LL)
(2)
28 cLMCT ≈ (11)
29 ia
L − S + i − a L − S
30
31 The diagramatic representation is displayed in Figure 7. Since the integrals (LS|ia) are
32 (2)
33 larger in the ground state than the excited state, cLMCT is therefore also larger for the ground
34
35 state, which explains the general structure of the correlated wave functions computed here
36
37 such as CIPSI, FOBOCI or CCSD. The full fourth-order energy expression is involved even
38
39 for the FOBOCI space. The part that can be directly compared to the second-order energy
40
41 is dominant and is given by
42
43 !
44
X (LS|ia)2 ((ia|SS) − (ia|LL))2
45
e(2) + e(4) ≈ 1+ (12)
L − S + i − a (L − S + i − a )(L − S )
46 i,a
47
48
The diagramatic representation of the approximation to e(4) is displayed in Figure 8. As the
49
50
energy denominators are always negative, and the numerators always positive, the higher-
51
52
order effects enhance the second order energy correction through an interaction between
53
54
LMCT and the double excitations |SL ai i. The differential effects at second order are magnified
55
56
57
58
59 15
1
2
3
by the LMCT, which explains the importance of both the LMCT and the double excitations
4
5
|SL ai i in the correct prediction of the electronic transition energies.
6
7
8 The above analysis provides a clear and simple picture for understanding when LMCT
9
10 will be significant in general metal complexes. LMCT will occur when i) there is significant
11
12 overlap between an occupied ligand orbital and the SOMO at the metal, resulting in large
13
14 charge fluctuation integrals (LS|ia), and ii) the orbitals at the metal can relax to accom-
15
16 modate the additional charge density, resulting in significant LMCT enhancement factors.
17
18 These conditions will be met in many transition metal complexes, highlighting the central
19
20 role of LMCT processes in electron correlation within this class of systems.
21
22
23
24
25 5 Single-reference methods
26
27
28
29 The success of the FOBOCI shows that very reasonable descriptions of the wavefunctions,
30
31 electronic transitions and spin densities can be obtained at reduced computational cost
32
33 through a careful selection of the CI space. The excitation manifold in FOBOCI is a subset
34
35 of CISD, since it contains all single excitations and a specific class of double excitations,
36
37 and CISD may be anticipated to be even more accurate. This section is dedicated to the
38
39 investigation of the performance of single-reference wave function based methods. All CISD
40
41 and CISD self-consistent and size-consistent (CISD(SC)2 ) 55–57 calculations were performed
42
43 using the Quantum Package software 30 and all coupled-cluster calculations were performed
44
45 using the Turbomole V7.3 58 software.
46
47
48
49 5.1 CISD and CCSD: the size extensivity error
50
51
52 The results of the CISD calculations are reported in Tables 1, 2 and 3 for the electronic
53
54 transitions, the amplitudes of the LMCT and the spin density, respectively. Contrary to
55
56 expectation, CISD performs systematically worse than FOBOCI, underestimating the elec-
57
58
59 16
1
2
3
tronic transitions by at least 5 mH, with a corresponding underestimation of the amplitudes
4
5
of the LMCT by at least a factor of two. The results of CCSD calculations are also reported.
6
7
CCSD is in much better agreement with the near FCI results, with a maximum error of
8
9 only 1.6 mH for the transition energies. This is despite the presence of large D1 diagnostics,
10
11 the values of which are essentially the same as the T1 amplitudes listed in Table 2. In the
12
13 discussion below we demonstrate that the failure of CISD is a direct consequence of the lack
14
15 of size extensivity of the CISD wavefunction and energy.
16
17
18 The CISD and CCSD equations can be directly compared when using the unlinked CCSD
19
20 formalism. 59 In both CISD and CCSD, discarding the spin polarisation energy from the
21
22 Brillouin terms, the correlation energy is
23
24
25 X
26 Ecorr = hROHF|H|bc bc
jk i cjk (13)
27 jkbc
28 L UL
29 = Ecorr (LiSa) + Ecorr (LiSa) (14)
30
31
32 In the second line we have introduced a decomposition into linked and unlinked contribu-
33
34 tions with respect to to a particular double excitation |Sa
Li i. The unlinked correlation energy
35
36 UL
Ecorr (LiSa) is the sum over all quadruplet of indices (j, k, b, c) in eq. 13 which do not
37
38 L
match any of the four indices (L, i, S, a), whereas the linked part Ecorr (LiSa) is the sum over
39
40 all quadruplet of indices (j, k, b, c) in eq. 13 which match at least one of the four indices
41
42 (L, i, S, a). The CISD equation for the coefficient for the double excitations into |Sa
Li i is
43
44
45 1 h Sa X X i
46 cSa
Li = h Li |H|ROHFi + c b Sa
h
j Li |H|b
j i + c bc Sa
h
jb Li |H|bc
jk i (15)
47 ∆Sa
Li jb jkbc
48
49 ∆Sa Sa Sa L UL
Li = EROHF − hLi |H|Li i + Ecorr (LiSa) + Ecorr (LiSa) (16)
50
51
52
53
54
55
56
57
58
59 17
1
2
3
whereas the corresponding CCSD equation is
4
5
6 1 h Sa X X
7 cSa
Li = Sa
hLi |H|ROHFi + cbj hSa b
Li |H|j i + cbc Sa bc
jb hLi |H|jk i
8 ∆Li jb jkbc
9 i
10
X
+ {hSa
Li |H|IihI|CCSDi}L (17)
11 I∈T,Q
12
13 ∆Sa Sa Sa L
Li = EROHF − hLi |H|Li i + Ecorr (LiSa) (18)
14
15
16
17 where T, Q are triple and quadruple excitations, {}L denotes the linked part of a contraction
18
19 between two operators, and the cSa UL
Li Ecorr (LiSa) term from the denominator exactly cancels
20
the unlinked parts of I∈T,Q hSa
P
21 Li |H|IihI|CCSDi. In the CCSD equations, only the linked
22
23 correlation energy survives in the demoninator, whereas the total correlation energy remains
24
25 in the denominator of the CISD equations. For the copper complexes, the total correlation
26
27 energy is in the order of 10 eV, and the unlinked component accounts for more than 95%. The
28
UL
29 presence of Ecorr (LiSa) in the CISD equations therefore introduces a spurious 10 eV shift in
30
31 all energy denominators, which dramatically reduces the coefficients of the double excitations
32
33 cSa
Li . Consequently, the correlation energy in general, and the differential correlation effects
34
35 arising from cSa
Li in particular, are systematically underestimated at the CISD level, resulting
36
37 in poor transition energies.
38
39
Tables 1, 2 and 3 also report the results from the CISD(SC)2 method where the CISD
40
41 UL
equations are modified by removing Ecorr (LiSa) from the denominator. CISD(SC)2 repairs
42
43
the errors of CISD and indeed performs comparably to CCSD, indicating that the higher-
44
45
order linked terms that are missing from CISD(SC)2 do not play a large role in the energy
46
47
differences between the excited states. To conclude this analysis, we note that FOBOCI
48
49
does not contain unlinked terms with respect to |Sa
Li i. The FOBOCI correlation energy can
50
51
be expressed as
52
53
X
FOBOCI
Ecorr = cSb Sb
Lj hROHF|H|Lj i (19)
54
jb
55
56
57
58
59 18
1
2
3
and is therefore always linked with respect to any double excitation |Sa
Li i present in the
4
5
FOBOCI. The FOBOCI also therefore does not suffer from size inconsistency errors for
6
7
the terms that dominante the differential correlation effects among the low-lying electronic
8
9 states.
10
11
12
13
14 5.2 CCSD(T) and BCCD(T)
15
16
17 Table 1 reports the transition energies computed at the CCSD, BCCD, CCSD(T) and
18
19 BCCD(T) levels using the 6-31G basis set. The Brueckner coupled cluster results are included
20
21 since we have shown that orbital relaxation effects are important and the spin densities on
22
23 the copper atom from the Brueckner orbitals are listed in Table 3. In the 6-31G basis set,
24
25 the CCSD, BCCD, CCSD(T) and BCCD(T) results are all close to the CIPSI values, with
26
27 nothing to significantly favour one method over the other. The 6-31G basis is too small to
28
29 reliably assess the importance of triple excitations, so we performed additional CIPSI and
30
31 coupled-cluster calculations where f polarization functions are added to the copper atom,
32
33 which we denote the 6-31G*(Cu) basis. This was only feasible for the [CuCl4 ]2− molecule and
34
35 the results are collected in Table 4. The necessity for three-body correlation in approaching
36
37 near FCI quality transition energies is clearly apparant. Unfortunately, it is not possible to
38
39 comment on the relative merits of BCCD(T) over CCSD(T) based on these numerical tests.
40
41 Table 4: Excitation energies (mH) at various levels of calculations for the [CuCl4 ]2− molecule
42
in the 6-31G*(Cu) basis set.
43
44 Electronic transition ROHF CCSD BCCD CCSD(T) BCCD(T) FOBOCI CIPSI
45 2
B1g − 2 B2g 30.2 45.7 45.1 46.7 46.7 44.0 46.8(1)
46 2
Eg − 2 B2g 38.6 54.9 54.3 56.0 56.0 52.5 55.6(1)
47
48
49
50
51
52
53
54
55
56
57
58
59 19
1
2
3
4
6 Basis set limit CCSD(T) and BCCD(T) transition
5
6
7
energies
8
9
10 The success of CCSD(T) and BCCD(T) in reproducing near FCI transition energies in small
11
12 basis sets, encourages us to use these methods to obtain high-quality reference values near the
13
14 basis set limit. Table 5 reports the results of ROHF-UCCSD(T)(F12*) and UBCCD(T)(F12*)
15
16 calculations 60–62 using the aug-cc-pwCVTZ-DK basis sets (aug-cc-pVTZ-DK for H). The aug-
17
18 cc-pwCVTZ-DK basis sets for C, N, O and Cl were constructed by adding the wC functions
19
20 from the aug-cc-pwCVTZ basis set to the aug-cc-pVTZ-DK basis sets. 63 The aug-cc-pVTZ
21
22 MP2 fitting basis 64 was used as the complementary auxiliary basis set (CABS) for Cu and the
23
24 specially optimised aug-cc-pVTZ CA basis sets were used for H, C, N, O and Cl. 65 The X2C
25
26 method 66–68 was used to account for scalar relativistic effects and an exponent of 1.2 a−1
0 was
27
28 used in the F12 correlation factor. 69 Table 5 lists the additive contributions from the CCSD,
29
30 (T), the CABS singles correction 62,70,71 and the F12 correction for frozen core calculations,
31
32 where an argon core was used for copper, and a helium core for oxygen and nitrogen. The
33
34 core-valence correlation correction is the difference between the full valence only ROHF-
35
36 UCCSD(T)(F12*) or UBCCD(T)(F12*) calculation and calculations correlating all electrons
37
38 except the neon core at the copper. All calculations were performed using Turbomole V7.3 58
39
40 and proceeded by first performing a ROHF calculation in D4h or D2d symmetry, followed by
41
42 a CC calculation in C1 symmetry. As experimental electronic transitions are available for the
43
2
44 B2g state of the [CuCl4 ]2− molecule and the corresponding 2 B2 state of the [Cu(NH3 )4 ]2+
45
46 molecule, we also report this quantity at the various CC levels of theory, made possible by
47
48 the high point group symmetry available in the SCF routines.
49
50
51
Concerning the dependence of the excitation energies on basis set, we find that while
52
53
the values differ substantially from those computed with a 6-31G basis, the CABS singles
54
55
correction is negligible and the ROHF energies are converged to within 0.2 mH at the aug-
56
57
58
59 20
1
2
3
cc-pwCVTZ-DK level. The F12 contribution is also less than a mH, suggesting that short-
4
5
range dynamical correlation is not decisive in the ordering of the excited states and that the
6
7
CCSD(T) and BCCD(T) values are well converged with respect to one-particle basis set size
8
9 at the aug-cc-pwCVTZ-DK level. We note that since the F12 correction is based on the cusp
10
11 condition for the first-order amplitudes, 61,72 it contains contributions of the type f12 |Lii, but
12
13 misses F12 contributions of the type f12 |Sai, and can therefore be expected to give slightly
14
15 too low excitation energies with small basis sets. The magnitude of this effect, however,
16
17 would appear to be very small, particularly in the Brueckner calculations where the orbital
18
19 optimisation reduces this bias considerably. We ascribe a basis set incompleteness error bar
20
21 of 0.5 mH for CCSD(T)(F12*) excitation energies, and 0.2 mH for BCCD(T)(F12*) excitation
22
23 energies.
24
25
26 Concerning the dependence of the excitation energies on the level of correlation treat-
27
28 ment, we find that BCCD systematically predicts lower transition energies than CCSD. This
29
30 pattern is reversed when comparing CCSD(T) and BCCD(T), where the differential effect of
31
32 the (T) triples correction on the excitation energies is larger for BCCD than for CCSD. While
33
34 the inclusion of high-order orbital relaxation effects in BCCD would favour this method, the
35
36 (T) correction is anticipated to be biased to the ground state in both cases. Without bench-
37
38 mark calculations in a larger basis set, it is difficult to be sure which method is superior.
39
40 We therefore quote the UBCCD(T)(F12*) as reference values for the transitions and assume
41
42 that the difference between the ROHF-UCCSD(T)(F12*) and UBCCD(T)(F12*) values is a
43
44 minimum estimate for the error bar. Mulliken population analysis of the Brueckner orbitals
45
46 resulting from the BCCD calculations, however, provide approximate spin-densities at the
47
48 Cu centre, which can be compared to experimental values where available. These are listed
49
50 in table 3, where the theoretical value of 0.62 for [CuCl4 ]2− in its ground state is in perfect
51
52 agreement with the value inferred from experiment, 24,25 and for [Cu(NH3 )4 ]2+ the value of
53
54 0.73 is in line with that obtained with the SORCI method by Neese 9 which is 0.71.
55
56
The ab initio reference values for the excitation energies are not expected to agree
57
58
59 21
1
2
3
perfectly with experimental values, since the former are gas phase data and the latter
4
5
are obtained from electronic absorption spectroscopy of single crystals containing the gas
6
7
phase chromophore, which are subject to crystal field effects and geometric relaxation. For
8
9 [CuCl4 ]2− Solomon and coworkers estimated the effect of crystal lattice on excitation en-
10
11 ergies from lattice model calculations at the level of DFT calculations, 8 reporting that it
12
13 is at most 5 mH. The agreement between theory and experiment is within this error bar.
14
15 The experimental values for [Cu(NH3 )4 ]2+ in our table differ from those in other theoretical
16
17 works, 9,14 where the values for the electronic transitions to the 2 B1 , 2 E and 2 A1 states are
18
19 63.8, 79.7 and 75.3 mH, respectively. With these values, however, the discrepancy between
20
21 theory and experiment is much larger than expected, as indeed previously noted by Neese. 9
22
23 The reason for this discrepancy is that the experimental values were measured by Hathaway
24
25 and co-workers in 1969 for single crystals of Na4 Cu(NH3 )4 [Cu(S2 O3 )2 ], which was assumed to
26
27 have a square planar Cu(NH3 )4 2+ environment. Prompted by Morosin’s more accurate X-ray
28
29 data, 22 a year later Hathaway published a revised crystal structure interpretation indicating
30
31 that the experiments were actually performed on a crystal with a weakly coordinating mono-
32
33 ammonia adduct Na4 Cu(NH3 )4 [Cu(S2 O3 )2 ],NH3 that has a time average stereochemistry at
34
35 the Cu atom of a tetragonal-octahedron. 23 In that same work, the electronic spectrum of
36
37 Na4 Cu(NH3 )4 [Cu(S2 O3 )2 ],H2 O was reported and analysed, and shown to have an effective
38
39 square-planar CuN4 stereochemistry with a freely rotating water molecule in the pocket at
40
41 [0,0, 21 ]. The electronic transitions measured were 83.8 and 87.5 mH to the 2 B1 and 2 A1
42
43 states, respetively. We therefore use these values in our table and indeed they agree with our
44
45 computed values to within 5 mH, which can be attributed to be largely from small structural
46
47 and environmental effects.
48
49
a
50 Single crystal electronic absorption spectroscopy of square planar cupric chloride 73
51
b
52 Single crystal electronic absorption spectroscopy of Na4 Cu(NH3 )4 [Cu(S2 O3 )2 ],H2 O 23
53
54
55
56
57
58
59 22
1
2
3 Table 5: CC excitation energies (mH) for the [CuCl4 ]2− , [Cu(NH3 )4 ]2+ and [Cu(H2 O)4 ]2+
4
molecules.
5
6 Electronic transition ROHF CCSD ∆HF ∆F12 ∆(T) ∆CV CCSD(T)-F12 BCCD ∆Ref ∆F12 ∆(T) ∆CV BCCD(T)-F12 Exp.a,b
7 [CuCl4 ]2−
8 2
B1g − 2 B2g 34.0 54.7 -0.1 -0.8 2.9 0.4 57.2 53.2 -0.1 -0.4 5.5 1.1 59.3 57.0
2
9 Eg − 2 B2g 46.0 64.7 -0.1 -0.6 2.7 0.0 66.7 63.5 -0.2 -0.4 4.8 0.6 68.4 64.7
2
A1g − 2 B2g 48.0 71.8 -0.2 -1.2 4.0 -0.9 74.4 69.4 -0.3 -0.7 7.8 1.4 77.9 77.5
10 [Cu(H2 O)4 ]2+
11 2
B1g − 2 B2g 44.4 56.0 0.0 0.2 1.8 0.7 58.2 55.1 0.0 -0.1 2.8 0.9 58.7 -
12 2
Eg − 2 B2g 47.8 55.3 0.1 0.0 0.9 0.8 57.1 54.9 0.0 0.0 1.4 0.9 57.3 -
2
A1g − 2 B2g
13 49.5 58.7 0.1 -0.3 1.5 1.1 61.1 57.9 0.1 -0.2 2.5 1.3 61.6 -
[Cu(NH3 )4 ]2+
14 2
B1 − 2 B2 53.8 76.9 0.0 -0.3 3.6 0.0 80.2 75.0 -0.1 0.0 6.3 0.0 81.2 83.8
15 2
E − 2 B2 68.6 89.5 0.0 -0.3 3.6 0.0 92.7 87.6 -0.1 0.0 6.3 0.0 93.7 -
16 2
A1 − 2 B2 66.4 87.6 -0.1 -0.5 3.6 1.2 91.9 85.8 -0.1 -0.2 6.1 1.7 93.4 87.5
17
18
19
20 7 Conclusion
21
22
23
Through careful benchmarking and theoretical analysis, this work highlights that the correct
24
25 theoretical determination of the electronic spectroscopy and the ground state spin density
26
27 of open-shell transition metal complexes requires methods that correctly couple a range of
28
29 correlation processes.
30
31
32 Definitive reference energies and wavefunctions for the three low-lying spin states of
33
34 [CuCl4 ]2− , [Cu(NH3 )4 ]2+ and [Cu(H2 O)4 ]2+ , in a modest 6-31G basis, were obtained from
35
36 near FCI calculations performed using the CIPSI selected CI method. Analysis of these
37
38 states revealed the prevalence of a specific excited configuration in all of the computed
39
40 wavefunctions, which plays a decisive role in the spin density and energies of the spin-states.
41
42 This configuration corresponds in all cases to a single excitation from the ROHF determinant
43
44 where an electron is excited from a ligand-like orbital to the SOMO which is mainly localised
45
46 on the central Cu2+ ion. A valence bond-like analysis shows that the these excitations can
47
48 be identified as LMCT components of the ground state wave functions, which can therefore
49
50 be thought of as a superposition of Cu2+ and Cu+ oxidation states.
51
52
53
A perturbation analysis of the coefficient of these Slater determinants in the ground
54
55
and excited state wave functions revealed that these determinants arise predominantly due
56
57
58
59 23
1
2
3
to the so called breathing orbital effect, an orbital relaxation induced by the change in
4
5
oxidation state at the Cu as a consequence of correlating the electrons. This effect is primarily
6
7
responsibe for the energy differences among the spin states and must be properly represented
8
9 in the wavefunction for a correct qualitative description of this class of systems. Using these
10
11 insights, we propose a minimal CI space, the FOBOCI, which captures the key physical
12
13 effects, and we demonstrate numerically that it is able to reproduce quantitatively the energy
14
15 differences and spin density of these three Cu2+ complexes, even though it recovers less
16
17 than 3% of the total correlation energy. The numerical evidence is further supported by a
18
19 perturbational analysis, up to fourth-order in the energy, which, together with some simple
20
21 physico-chemical considerations, explains the success the FOBOCI in accurately describing
22
23 the energy differences and provides a simple and general physical picture for the LMCT
24
25 processes. Based on our analysis, we are led to conclude that LMCT will be significant in
26
27 general transition metal complexes and predict that it will always occur when
28
29
30
31 • There is significant overlap between an occupied ligand orbital and the SOMO at the
32
33 metal
34
35 • The orbitals at the metal can relax to accommodate the additional charge density
36
37
38
39 Having obtained detailed physical and mathematical insight into the theoretical description
40
41 of these systems, we proceeded to investigate the performance of the commonly applied
42
43 wave function based methods, both single- and multi-reference. Regarding multi-reference
44
45 methods, the performance depends strongly on the choice of active space. The minimal active
46
47 space required for a qualitatively correct description is one containing the ligand orbital
48
49 involved in the SOMO LMCT together with the double-d shell, for the orbital relaxation.
50
51 Regarding single reference methods, we find that the correct description is obtained
52
53 provided that
54
55
56
• The wavefunction contains both the ROHF and SOMO LMCT configurations and all
57
58
59 24
1
2
3
single excitations from each
4
5
6 • The wavefunction coefficients are obtained to at least 2nd order in perturbation theory
7
8 (fourth order in the energy)
9
10
11 • The wavefunction coefficients are obtained in a size extensive manner
12
13
14
In this respect, our study reveals that CC-based methods are perfectly suited for the study
15
16
of these Cu2+ complexes, since the excitation manifold of singles and doubles contains all
17
18
important configurations, the iteratively optimised amplitudes correspond to high order in
19
20
perturabtion theory, and the method is size extensive. We find that CCSD(T) performs well
21
22
despite exhibiting large T1 amplitudes and large D1 diagnostics for all wavefunctions (see
23
24
supporting information for numerical evidence). Indeed, BCCD(T) and CCSD(T) return
25
26
very similar results. These diagnostics, based on the singles amplitudes, are large when
27
28
there are strong orbital relaxation effects and are an indirect indication of multi-reference
29
30
character at best. In this case the assumption that large T1 and D1 values predict the failure
31
32
of CCSD(T) is incorrect.
33
34
35 Our study also reveals that CISD performs poorly. Our analysis proves that the non-size
36
37 extensive nature of the CISD equations leads to erroneous supression of correlating excita-
38
39 tions, biasing spin states with smaller correlation energies. We expect that our observation
40
41 that size extensivity errors plague calculations of vertical spectrum of molecular complexes
42
43 at equilibrium geometry as well as dissociation energies will be generally applicable to all
44
45 systems, since the errors simply grow with the magnitude of the correlation energy.
46
47
48
Having established the reliability of the CC-based methods for the determination of the
49
50
energies of the spin states, we performed CCSD(T) and BCCD(T) calculations in a large
51
52
basis set using explicitly correlated corrections in order to establish reference values for the
53
54
energy differences of these three Cu2+ complexes (see Table 5). Our near basis set limit
55
56
core-valence correlated energies with scalar relativistic effects included agree with observed
57
58
59 25
1
2
3
energy differences from single crystal electronic absorption spectroscopy to within 5 mH,
4
5
which is the same magnitude as the change expected due to placing the gas phase ion in the
6
7
solid state crystal environment.
8
9
10 This study provides futher confirmation of the importance of LMCT in the determination
11
12 of the properties of many 3d transition metal containing molecular complexes and highlights
13
14 once more that metal-ligand delocalisation is very sensitive to the level to which electronic
15
16 correlation is treated.
17
18
19
20
21 Supporting information content
22
23
24
25 The supporting information text file contains all geometries used for the calculations, the
26
27 total energies of the CIPSI calculations in the 6-31G basis set and the total energies of the
28
29 CCSD(T) and BCC(T) in the aug-cc-pwCVTZ-DK basis set. The supporting information
30
31 pdf file contains a bried derivation of the physical content of the second order perturbative
32
33 correction to the energy.
34
35
36
37
38 References
39
40
41 (1) Janssen, G. J. M.; Nieuwpoort, W. C. Band gap in NiO: A cluster study. Phys. Rev. B
42
43 1988, 38, 3449–3458.
44
45
46
(2) van Oosten, A. B.; Broer, R.; Nieuwpoort, W. C. Heisenberg exchange enhancement
47
48
by orbital relaxation in cuprate compounds. Chem. Phys. Lett. 1996, 257, 207 – 212.
49
50 (3) Calzado, C. J.; Sanz, J. F.; Malrieu, J.-P. Accurate ab initio determination of magnetic
51
52 interactions and hopping integrals in La2−x Srx CuO4 systems. J. Chem. Phys. 2000,
53
54 112, 5158–5167.
55
56
57
58
59 26
1
2
3
(4) de Graaf, C.; Sousa, C.; de P. R. Moreira, I.; Illas, F. Multiconfigurational Perturbation
4
5
Theory: An Efficient Tool to Predict Magnetic Coupling Parameters in Biradicals,
6
7
Molecular Complexes, and Ionic Insulators. J. Phys. Chem. A 2001, 105, 11371–11378.
8
9
10 (5) Calzado, C. J.; Cabrero, J.; Malrieu, J.-P.; Caballol, R. Analysis of the magnetic cou-
11
12 pling in binuclear complexes. I. Physics of the coupling. J. Chem. Phys. 2002, 116,
13
14 2728–2747.
15
16
17 (6) Cabrero, J.; Calzado, C. J.; Maynau, D.; Caballol, R.; Malrieu, J.-P. Metal Ligand
18
19 Delocalization in Magnetic Orbitals of Binuclear Complexes. J. Phys. Chem. A 2002,
20
21 106, 8146–8155.
22
23
24 (7) Broer, R.; Hozoi, L.; Nieuwpoort, W. C. Non-orthogonal approaches to the study of
25
26 magnetic interactions. Mol. Phys. 2003, 101, 233–240.
27
28
29 (8) Szilagyi, R. K.; Metz, M.; Solomon, E. I. Spectroscopic Calibration of Modern Density
30
31 Functional Methods Using [CuCl4 ]2− . J. Phys. Chem. A 2002, 106, 2994–3007.
32
33
(9) Neese, F. Sum-over-states based multireference ab initio calculation of EPR spin Hamil-
34
35
tonian parameters for transition metal complexes. A case study. Mag. Res. Chem. 2004,
36
37
42, S187–S198.
38
39
40 (10) A. Ramı́rez-Solı́s, A.; Poteau, R.; Vela, A.; Daudey, J. P. Comparative studies of the
41
42 spectroscopy of CuCl2 : DFT versus standard ab initio approaches. J. Chem. Phys.
43
44 2005, 122, 164306.
45
46
47 (11) Atanasov, M.; Comba, P.; Martin, B.; Müller, V.; Rajaraman, G.; Rohwer, H.; Wun-
48
49 derlich, S. DFT models for copper(II) bispidine complexes: Structures, stabilities, iso-
50
51 merism, spin distribution, and spectroscopy. J. Comp. Chem. 2006, 27, 1263–1277.
52
53
54 (12) Calzado, C. J.; Angeli, C.; Taratiel, D.; Caballol, R.; Malrieu, J.-P. Analysis of the
55
56
57
58
59 27
1
2
3
magnetic coupling in binuclear systems. III. The role of the ligand to metal charge
4
5
transfer excitations revisited. J. Chem. Phys. 2009, 131, 044327.
6
7
8 (13) Kossmann, S.; Kirchner, B.; Neese, F. Performance of modern density functional theory
9
10 for the prediction of hyperfine structure: meta-GGA and double hybrid functionals.
11
12 Mol. Phys. 2007, 105, 2049–2071.
13
14
15 (14) Vancoillie, S.; Pierloot, K. Multiconfigurational g Tensor Calculations as a Probe
16
17 for the Covalency of the Copper-Ligand Bonds in Copper(II) Complexes: [CuCl4 ]2− ,
18
19 [Cu(NH3 )4 ]2+ , and Plastocyanin. J. Phys. Chem. A 2008, 112, 4011–4019.
20
21
22 (15) Boguslawski, K.; Marti, K. H.; Legeza, Ö.; Reiher, M. Accurate ab Initio Spin Densities.
23
24 J. Chem. Theor. Compu. 2012, 8, 1970–1982.
25
26
27 (16) Boguslawski, K.; Jacob, C. R.; Reiher, M. Can DFT Accurately Predict Spin Densities?
28
29 Analysis of Discrepancies in Iron Nitrosyl Complexes. J. Chem. Theor. Comp. 2011,
30
31 7, 2740–2752.
32
33
(17) Caffarel, M.; Giner, E.; Scemama, A.; Ramı́rez-Solı́s, A. Spin Density Distribution
34
35
in Open Shell Transition Metal Systems: A Comparative Post Hartree Fock, Density
36
37
Functional Theory, and Quantum Monte Carlo Study of the CuCl2 Molecule. J. Chem.
38
39
Theor. Comp. 2014, 10, 5286–5296.
40
41
42 (18) Giner, E.; Angeli, C. Metal-ligand delocalization and spin density in the CuCl2 and
43
44 [CuCl4 ]2− molecules: Some insights from wave function theory. J. Chem. Phys. 2015,
45
46 143, 124305.
47
48
49 (19) Tenti, L.; Maynau, D.; Angeli, C.; Calzado, C. J. Highly efficient perturbative + varia-
50
51 tional strategy based on orthogonal valence bond theory for the evaluation of magnetic
52
53 coupling constants. Application to the trinuclear Cu(ii) site of multicopper oxidases.
54
55 Phys. Chem. Chem. Phys. 2016, 18, 18365–18380.
56
57
58
59 28
1
2
3
(20) Remenyi, C.; Reviakine, R.; Kaupp, M. Density Functional Study of EPR Parameters
4
5
and Spin-Density Distribution of Azurin and Other Blue Copper Proteins. J. Phys.
6
7
Chem. B 2007, 111, 8290–8304.
8
9
10 (21) Giner, E.; Angeli, C. Spin density and orbital optimization in open shell systems: A
11
12 rational and computationally efficient proposal. J. Chem. Phys. 2016, 144, 104104.
13
14
15 (22) Morosin, B. The crystal structures of copper tetrammine complexes. A.
16
17 Cu(NH3 )4 SO2 .H2 O and Cu(NH3 )4 SeO4 . 1969, 25, 19–30.
18
19
20 (23) Hathaway, B. J.; Stephens, F. The electronic properties and crystal structure of
21
22 Na4Cu(NH3)4[Cu(S2O3)2]2,L (L = H2O or NH3). J. Chem. Soc. A 1970, 884–888.
23
24
25
(24) Gewirth, A. A.; Cohen, S. L.; Schugar, H. J.; Solomon, E. I. Spectroscopic and the-
26
27
oretical studies of the unusual EPR parameters of distorted tetrahedral cupric sites:
28
29
correlations to x-ray spectral features of core levels. Inorg. Chem. 1987, 26, 1133–1146.
30
31 (25) Didziulis, S. V.; Cohen, S. L.; Gewirth, A. A.; Solomon, E. I. Variable photon energy
32
33 photoelectron spectroscopic studies of copper chlorides: an experimental probe of metal-
34
35 ligand bonding and changes in electronic structure on ionization. J. Am. Chem. Soc.
36
37 1988, 110, 250–268.
38
39
40 (26) Cotton, F. A. I - Ligand field theory. J. Chem. Educ. 1964, 41, 466.
41
42
43 (27) Broer, R.; Nieuwpoort, W. Broken orbital symmetry and the description of valence hole
44
45 states in the tetrahedral [CrO4 ]2− anion. Theor. Chim. Acta 1988, 73, 405–418.
46
47
48 (28) Miralles, J.; Castell, O.; Caballol, R.; Malrieu, J.-P. Specific CI calculation of energy
49
50 differences: Transition energies and bond energies. Chem. Phys. 1993, 172, 33 – 43.
51
52
(29) Angeli, C.; Calzado, C. J. The role of the magnetic orbitals in the calculation of the mag-
53
54
netic coupling constants from multireference perturbation theory methods. J. Chem.
55
56
Phys. 2012, 137, 034104.
57
58
59 29
1
2
3
(30) Scemama, A.; Applencourt, T.; Garniron, Y.; Giner, E.; David, G.; Caffarel, M. Quan-
4
5
tum package v1.0. 2016.
6
7
8 (31) Perdew, J. P.; Ernzerhof, M.; Burke, K. Rationale for mixing exact exchange with
9
10 density functional approximations. J. Chem. Phys. 1996, 105, 9982–9985.
11
12
13 (32) Bender, C. F.; Davidson, E. R. Phys. Rev. 1969, 183, 23.
14
15
16 (33) Huron, B.; Rancurel, P.; Malrieu, J. J. Chem. Phys. 1973, 58, 5745.
17
18
19 (34) Buenker, R. J.; Peyerimholf, S. D. Theor. Chim. Acta 1974, 35, 33.
20
21
22
(35) Buenker, R. J.; Peyerimholf, S. D.; Bruna, P. J. Comp. Theor. Org. Chem.; Reidel:
23
24
Dordrecht, 1981; p 55.
25
26 (36) Evangelisti, S.; Daudey, J.-P.; Malrieu, J.-P. Convergence of an improved CIPSI algo-
27
28 rithm. Chem. Phys. 1983, 75, 91 – 102.
29
30
31 (37) Harrison, R. J. J. Chem. Phys. 1991, 94, 5021.
32
33
34 (38) Holmes, A. A.; Tubman, N. M.; Umrigar, C. J. Heat-Bath Configuration Interaction:
35
36 An Efficient Selected Configuration Interaction Algorithm Inspired by Heat-Bath Sam-
37
38 pling. J. Chem. Theor. Comp. 2016, 12, 3674–3680.
39
40
41 (39) Rubio, J.; Novoa, J.; Illas, F. Convergence of a multireference second-order mbpt
42
43 method (CIPSI) using a zero-order wavefunction derived from an {MS} {SCF} cal-
44
45 culation. Chem. Phys. Lett. 1986, 126, 98 – 102.
46
47
48 (40) Cimiraglia, R.; Persico, M. Recent advances in multireference second order perturbation
49
50 CI: The CIPSI method revisited. J. Comp. Chem. 1987, 8, 39–47.
51
52 (41) Angeli, C.; Persico, M. Multireference perturbation CI II. Selection of the zero-order
53
54 space. Theor. Chem. Acc. 1997, 98, 117–128.
55
56
57
58
59 30
1
2
3
(42) Angeli, C.; Cimiraglia, R.; Malrieu, J.-P. On a mixed MllerPlesset EpsteinNesbet par-
4
5
tition of the Hamiltonian to be used in multireference perturbation configuration inter-
6
7
action. Chem. Phys. Lett. 2000, 317, 472 – 480.
8
9
10 (43) Giner, E.; Scemama, A.; Caffarel, M. Using perturbatively selected configuration inter-
11
12 action in quantum Monte Carlo calculations. Can. J. Chem. 2013, 91, 879–885.
13
14
15 (44) Scemama, A.; Applencourt, T.; Giner, E.; Caffarel, M. Accurate nonrelativistic ground-
16
17 state energies of 3d transition metal atoms. J. Chem. Phys. 2014, 141 .
18
19
20 (45) Giner, E.; Scemama, A.; Caffarel, M. Fixed-node diffusion Monte Carlo potential en-
21
22 ergy curve of the fluorine molecule F2 using selected configuration interaction trial
23
24 wavefunctions. J. Chem. Phys. 2015, 142 .
25
26
27
(46) Giner, E.; Assaraf, R.; Toulouse, J. Quantum Monte Carlo with reoptimised perturba-
28
29
tively selected configuration-interaction wave functions. Mol. Phys. 2016, 114, 910–920.
30
31 (47) Epstein, P. S. The Stark Effect from the Point of View of Schroedinger’s Quantum
32
33 Theory. Phys. Rev. 1926, 28, 695–710.
34
35
36 (48) Nesbet, R. K. Configuration Interaction in Orbital Theories. Proc. R. Soc. London,
37
38 Ser. A 1955, 230, 312–321.
39
40
41 (49) Garniron, Y.; Scemama, A.; Loos, P.-F.; Caffarel, M. Hybrid stochastic-deterministic
42
43 calculation of the second-order perturbative contribution of multireference perturbation
44
45 theory. J. Chem. Phys. 2017, 147, 034101.
46
47
48 (50) Møller, C.; Plesset, M. Note on an Approximation Treatment for Many-Electron Sys-
49
50 tems. Phys. Rev. 1934, 46, 618.
51
52
(51) Calzado, C. J.; Cabrero, J.; Malrieu, J.-P.; Caballol, R. Analysis of the magnetic cou-
53
54
pling in binuclear complexes. II. Derivation of valence effective Hamiltonians from ab
55
56
initio CI and DFT calculations. J. Chem. Phys. 2002, 116, 3985–4000.
57
58
59 31
1
2
3
(52) Hiberty, P. C.; Shaik, S. Breathing-orbital valence bond method , a modern valence
4
5
bond method that includes dynamic correlation. Theor. Chem. Acc. 2002, 108, 255–
6
7
272.
8
9
10 (53) Gordon, M. S.; Schmidt, M. W. In Theory and Applications of Computational Chem-
11
12 istry: the first forty years; Dykstra, C. E., Frenking, G., Kim, K. S., Scuseria, G. E.,
13
14 Eds.; Elsevier: Amsterdam, 2005; pp 1167–1189.
15
16
17 (54) Giner, E.; Tenti, L.; Angeli, C.; Ferr, N. Computation of the Isotropic Hyperfine Cou-
18
19 pling Constant: Efficiency and Insights from a New Approach Based on Wave Function
20
21 Theory. J. Chem. Theor. Comp. 2017, 13, 475–487.
22
23
24 (55) Meller, J.; Heully, J. L.; Malrieu, J. P. Size-consistent self-consistent combination of
25
26 selected CI and perturbation theory. Chem. Phys. Letters 1994, 218, 276–282.
27
28
29
(56) Malrieu, J.-P.; Daudey, J.-P.; Caballol, R. Multireference self-consistent size-consistent
30
31
singles and doubles configuration interaction for ground and excited states. J. Chem.
32
33
Phys. 1994, 101, 8908–8921.
34
35 (57) Daudey, J.-P.; Heully, J.-L.; Malrieu, J.-P. Size consistent self consistent truncated or
36
37 selected configuration interaction. J. Chem. Phys. 1993, 99, 1240–1254.
38
39
40 (58) TURBOMOLE Version 7.3. For further information see http://www.turbomole.com.
41
42
43 (59) Purvis, G. D.; Bartlett, R. J. A full coupledcluster singles and doubles model: The
44
45 inclusion of disconnected triples. J. Chem. Phys. 1982, 76, 1910–1918.
46
47
48 (60) Hättig, C.; Tew, D. P.; Köhn, A. Communications: Accurate and efficient approxima-
49
50 tions to explicitly correlated coupled-cluster singles and doubles, CCSD-F12. J. Chem.
51
52 Phys. 2010, 132, 231102.
53
54
(61) Tew, D. P.; Klopper, W. Open-shell explicitly correlated F12 methods. Mol. Phys.
55
56
2010, 108, 315 – 325.
57
58
59 32
1
2
3
(62) Tew, D. P. Explicitly correlated coupled-cluster theory with Brueckner orbitals. J.
4
5
Chem. Phys. 2016, 145, 074103.
6
7
8 (63) de Jong, W. A.; Harrison, R. J.; Dixon, D. A. Parallel DouglasKroll energy and gradi-
9
10 ents in NWChem: Estimating scalar relativistic effects using DouglasKroll contracted
11
12 basis sets. J. Chem. Phys. 2001, 114, 48–53.
13
14
15 (64) Hill, J. G.; Platts, J. A. Auxiliary basis sets for density fittingMP2 calculations: Non-
16
17 relativistic triple- all-electron correlation consistent basis sets for the 3d elements ScZn.
18
19 J. Chem. Phys. 2008, 128, 044104.
20
21
22 (65) Yousaf, K. E.; Peterson, K. A. Optimized complementary auxiliary basis sets for ex-
23
24 plicitly correlated methods: aug-cc-pVnZ orbital basis sets. Chem. Phys. Lett. 2009,
25
26 476, 303 – 307.
27
28
29
(66) Kutzelnigg, W.; Liu, W. Quasirelativistic theory equivalent to fully relativistic theory.
30
31
J. Chem. Phys. 2005, 123, 241102.
32
33 (67) Liu, W.; Peng, D. Exact two-component Hamiltonians revisited. J. Chem. Phys. 2009,
34
35 131, 031104.
36
37
38 (68) Peng, D.; Reiher, M. Exact decoupling of the relativistic Fock operator. Theor. Chem.
39
40 Acc. 2012, 131, 1–20.
41
42
43 (69) Tew, D. P.; Klopper, W. New correlation factors for explicitly correlated electronic
44
45 wave functions. J. Chem. Phys. 2005, 123, 074101.
46
47
48 (70) Adler, T. B.; Knizia, G.; Werner, H.-J. A simple and efficient CCSD(T)-F12 approxi-
49
50 mation. J. Chem. Phys. 2007, 127, 221106.
51
52
(71) Köhn, A.; Tew, D. P. Towards the Hartree–Fock and coupled-cluster singles and dou-
53
54
bles basis set limit: A study of various models that employ single excitations into a
55
56
complementary auxiliary basis set. J. Chem. Phys. 2010, 132, 024101.
57
58
59 33
1
2
3
(72) Bokhan, D.; Bernadotte, S.; Ten-no, S. Explicitly correlated second-order M[o-
4
5
slash]ller–Plesset perturbation theory for unrestricted Hartree–Fock reference functions
6
7
with exact satisfaction of cusp conditions. J. Chem. Phys. 2009, 131, 084105.
8
9
10 (73) Neese, F.; Solomon, E. I. Magnetism: Molecules to Materials IV ; Wiley-Blackwell,
11
12 2003; Chapter 9, pp 345–466.
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59 34
1
2
3
4
5 0.0421 0.0522
6 0.0420 0.0521
7 0.0419 0.0520
8
Energy diference (Ha)

9 0.0418 0.0519
10 0.0417 0.0518 E v (2 E g)−E v (2 A g )

11 0.0416 0.0517 E CIPSI (2 E g)−E CIPSI (2 A g)
12
0.0415 0.0516
13
14 0.0414 0.0515
E v (2 B 1g)−E v (2 A g )
15 0.0413 0.0514
E CIPSI (2 B 1g)−E CIPSI (2 A g)
16 0.0412 6 0.0513 6
17
7 8 7 8
10 10 10 10 10 10
N det N det
18
19 (a) (b)
20
21
0.0520 0.0510
22
23 0.0515 0.0508
24
25
0.0510 0.0506
26 0.0505 0.0504
E v (2 B 1g)−E v (2 A g )
27
28 E CIPSI (2 B 1g)−E CIPSI (2 A g)
0.0500 0.0502
29
30 0.0495 0.0500
31 0.0490 0.0498
E v (2 B 2g)−E v (2 A g)
32 E CIPSI (2 B 2g)−E CIPSI (2 A g)
33 0.0485 5
10
6
10 10
7 8
10
0.0496 5
10
6
10 10
7 8
10
34 N det N det
35 (c) (d)
36
37
38 0.070 0.080
39
0.068
40 0.078
41 0.066
42 0.076
0.064
43
44 0.062 0.074
45 0.060
46 0.072
47 0.058
48 0.056
E v (2 A)−E v (2 b2) 0.070 E v (2 b1)−E v (2 b2 )
49
2 2
E CIPSI ( A)−E CIPSI ( b2) E CIPSI (2 b1)−E CIPSI (2 b2)
50 0.054 5
10
6
10 10
7 8
10
0.068 5
10
6
10 10
7 8
10
N det N det
51
52 (e) (f)
53
54 Figure 2: Convergence of the 2 B1g − 2 B2g and 2 Eg − 2 B2g electronic transitions at the vari-
55 ational and CIPSI level in the 6-31G basis set for the [CuCl4 ]2− (a and b) and [Cu(H2 O)4 ]2+
56 (c and d) complex and of the 2 B1 − 2 B2 and 2 E − 2 B2 electronic transition (e and f) for
57
58
the [Cu(NH3 )4 ]2+ complex as a function of the size of the reference CIPSI wave function.
59 35
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23 (a) SOMO (b) Donor orbital
24
25 Figure 3: SOMO S at the ROHF level (a) and ligand donor orbital L (b) in the 2 B2g ground
26 state of the [CuCl4 ]2− molecule. The isoamplitude is set to 0.05.
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
52
53 Figure 4: SOMO S at the ROHF level (a) and ligand donor orbital L (b) in the 2 B2g ground
54 state of the [Cu(H2 O)4 ]2+ molecule. The isoamplitude is set to 0.05.
55
56
57
58
59 36
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
23
24 Figure 5: SOMO S at the ROHF level (a) and ligand donor orbital L (b) in the 2 B2 ground
25 state of the [Cu(NH3 )4 ]2+ molecule. The isoamplitude is set to 0.05.
26
27
28
29
30
31 L S i a
32
33
34
35
36
37 Figure 6: Main diagrams involved in the calculation of e(2) within the FOBOCI space.
38
39
40
41
42
43 L S S L
44
45
46
47 i a i a
48
49
50
51
(a) (b)
52
53 (2)
54 Figure 7: Main diagrams involved in the calculation of cLMCT within the FOBOCI space.
(LS|ia) (ia|SS) (LS|ia) (ia|LL)
55 (a) corresponds to L − and (b) to − L −
S +i −a L −S S +i −a L −S
56
57
58
59 37
1
2
3
4
5
6
7
8
9
10
11
12 L S i a S L i a
13
14
15
16
17
18
19
20
21 i a i a
22
23
24
25
26 (a) (b)
27
28
29
30 S L i a L S i a
31
32
33
34
35
36
37
38
39
40 i a i a
41
42
43
44
(c) (d)
45
46
47
Figure 8: Main diagrams involved in the calculation of e(4) within the FOBOCI space. (a) cor-
(LS|ia)2 (ia|SS)2 (LS|ia)2 (ia|LL)2
48 responds to L − S +i −a (L −S +i −a )(L −S )
, (b) corresponds to L − S +i −a (L −S +i −a )(L −S )
,
49 (LS|ia) 2 (ia|SS)(ia|LL)
50
and (c) and (d) to − L − S +i −a (L −S +i −a )(L −S )
.
51
52
53
54
55
56
57
58
59 38
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
24
25 Figure 9: SOMO S at the ROHF level (a) and ligand donor orbital L (b) in the 2 B1g excited
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
52
53 Figure 10: SOMO S at the ROHF level (a) and ligand donor orbital L (b) in the 2 Eg excited
55
56
57
58
59 39
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
24
25 Figure 11: SOMO S at the ROHF level (a) and ligand donor orbital L (b) in the 2 B1g excited
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
52
53 Figure 12: SOMO S at the ROHF level (a) and ligand donor orbital L (b) in the 2 Eg excited
55
56
57
58
59 40
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
24
25 Figure 13: SOMO S at the ROHF level (a) and ligand donor orbital L (b) in the 2 B1 excited
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
52
53 Figure 14: SOMO S at the ROHF level (a) and ligand donor orbital L (b) in the 2 E excited
55
56
57
58
59 41
1
2
Ground state LF excited state
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
LMCT
20
21
22
23
24
25
26
27
28
29
30
31
32
33
35
36

Giner 2018

Uploaded by

Copyright:

Available Formats

You might also like

Giner 2018

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Giner 2018

Uploaded by

Copyright:

Available Formats

Subscriber access provided by NORTH CAROLINA A&T UNIV

Quantum Electronic Structure

is published by the American Chemical Society. 1155 Sixteenth Street N.W.,

Energy diference (Ha)

10 0.0417 0.0518 E v (2 E g)−E v (2 A g )

Energy diference (Ha)

Energy diference (Ha)

You might also like