10 1016@j CCR 2017 03 018

Coordination Chemistry Reviews 344 (2017) 2–25
Contents lists available at ScienceDirect
Coordination Chemistry Reviews

journal homepage: www.elsevier.com/locate/ccr
Review
Covalency and chemical bonding in transition metal complexes:

An ab initio based ligand field perspective
Saurabh Kumar Singh a, Julien Eng a,c, Mihail Atanasov a,b,⇑, Frank Neese a,⇑
a
Max-Planck Institute for Chemical Energy Conversion, Stiftstrabe 34-36, D-45470 Mülheim an der Ruhr, Germany
b
Institute of General and Inorganic Chemistry, Bulgarian Academy of Sciences, 1113 Sofia, Bulgaria
c
School of Chemistry, Newcastle University, Newcastle upon Tyne NE1 7RU, UK
a r t i c l e i n f o a b s t r a c t
Article history: In this work, a general, user-friendly method – ab initio ligand field theory (AILFT), is described and illus-
Received 3 February 2017 trated. AILFT allows one to unambiguously extract all ligand field parameters (the ligand field one-
Received in revised form 8 March 2017 electron matrix VLFT, the Racah parameters B and C, and the spin-orbit coupling parameter f) from rela-
Accepted 24 March 2017
tively straightforward multi-reference ab initio calculations. The method applies to mononuclear com-
Available online 31 March 2017
plexes in dn or fn configurations. The method is illustrated using complete active space self-consistent
Dedicated to A.B.P. Lever in recognition of field (CASSCF) and N-electron valence perturbation theory (NEVPT2) calculations on a series of well doc-
his outstanding contributions to the field of umented octahedral complexes of CrIII with simple ligands such as F, Cl, Br, I, NH3 and CN. It is
inorganic spectroscopy and his activity as shown that all well-known trends for the value of 10Dq (the spectrochemical series) are faithfully repro-
Editor-in-Chief of Coord. Chem. Rev. during duced by AILFT. By comparison of B and f for CrIII in these complexes with the parameters calculated for
a period of 50 years and Editor of
Comprehensive Coordination Chemistry II.
the free ion Cr3+, the covalency of the Cr-ligand bond can be assessed quantitatively (the non-relativistic
and relativistic nephelauxetic effects). The variation of ligand field parameters for complexes of 3d, 4d
and 5d elements is studied using MCl3 III III III
6 (M = Cr , Mo , W ) as model examples. As reflected in variations
Keywords:
Transition metal complexes of 10Dq, B and f across this series, metal-ligand covalency increases from CrCl3 3
6 to MoCl6 to WCl6 .
3
Ab initio quantum chemistry Using the angular overlap model, the one-electron parameters of the ligand field matrix are decomposed
Ligand field theory into increments for r- and p- metal-ligand interactions. This allows for the quantification of variations in
Metal-ligand bonding r- and p-ligand donor properties of these ligands. Using these results, the well documented two-
Absorption spectroscopy dimensional spectroscopic series for complexes of CrIII is quantitatively reproduced. Comparison of the
Crystal field theory results obtained using CASSCF and NEVPT2 reveals the importance of dynamic electron correlation.
Angular overlap model (AOM) Finally, the limitations of the AILFT method for complexes with increasing metal-ligand covalency are
Ab initio ligand field theory (AILFT)
analyzed and discussed.
Complete active space self-consistent field
Ó 2017 Elsevier B.V. All rights reserved.
(CASSCF)
N-electron valence perturbation theory to
second order (NEVPT2)
Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2. Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.1. Complete active space self-consistent field theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2. N-electron valence and complete active space second-order perturbation theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3. Quasi-degenerate perturbation theory for relativistic effects. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.4. Ab initio ligand field theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3. AILFT extracted ligand field parameters for octahedral M3+(d3) ions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.1. Electronic structure of the free 3d, 4d and 5d M(d3) metal ions, M=Cr3+,Mo3+,W3+ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.1.1. AILFT parameters for the free M3+(d3) ions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.1.2. Interpretation of the AILFT parameters for the free M3+(d3) ions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2. Periodic trend across the group 6B: studies on [MCl6]3 (where M = Cr(III), Mo(III) and W(III)) complexes . . . . . . . . . . . . . . . . . . . . . . . . 10
⇑ Corresponding authors at: Max-Planck Institute for Chemical Energy Conversion, Stiftstrabe 34-36, D-45470, Mülheim an der Ruhr, Germany.
E-mail addresses: mihail.atanasov@cec.mpg.de (M. Atanasov), frank.neese@cec.mpg.de (F. Neese).
http://dx.doi.org/10.1016/j.ccr.2017.03.018
0010-8545/Ó 2017 Elsevier B.V. All rights reserved.
S.K. Singh et al. / Coordination Chemistry Reviews 344 (2017) 2–25 3
3.2.1. AILFT parameters for the [MCl6]3 (where M = Cr(III), Mo(III) and W(III)) complex series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.2.2. Interpretation of the AILFT parameters for the [MCl6]3 (where M = Cr(III), Mo(III) and W(III)) complex series. . . . . . . . . . . . . . 11
3.3. Insight into the spectrochemical series from AILFT: studies on [CrX6]3+/ (X = F, Cl, Br, I, NH3 and CN) complexes . . . . . . . . . . . . . 12
3.3.1. AILFT parameters for the [CrX6]3+/ (X = F, Cl, Br, I, NH3 and CN) complexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.3.2. Interpretation of the AILFT parameters for the [CrX6]3+/ (X = F, Cl, Br, I, NH3 and CN) complexes . . . . . . . . . . . . . . . . . . . 13
3.3.3. AILFT and molecular magnetism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4. Comparison of AILFT with experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4.1. Transitions from experiment and from CASSCF and NEVPT2 ab initio calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.2. Ligand field parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
5. Conclusions and outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
6. Computational details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
6.1. Cluster embedding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
6.2. Geometry optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
6.3. Deducing first (electrostatic) and second (covalent) order contributions to the octahedral eg-t2g orbital splitting D . . . . . . . . . . . . . . . . . 22
6.4. Ligand field computations and extraction of ligand field parameters from a direct fit to energies of d-d transitions . . . . . . . . . . . . . . . . . 22
Acknowledgement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1. Introduction nd orbitals in the complex [21–23]. This led to a static picture of

the d-electronic states in complexes where overlap between wave
Ligand field theory [1–8] is a model designed for an entire class functions located at transition metal and the ligands is ignored,
of compounds – namely coordination complexes of d-block transi- also known as crystal field theory [24,25].
tion metals (TM) in oxidation states that are not too high or too It was soon realized that such a localized description leads to
low, typically II and III. The electronic structure of the free ions energy effects (e.g. term energy splitting in crystals) that are too
with charges +2 and +3 has been studied in great detail by atomic small compared to those seen experimentally [26]. Electron para-
spectroscopy, more specifically by discharge spectra, which display magnetic resonance spectroscopy, which probes transitions
sharp d-d transitions [9–11]. For example, for the Cr(III) ion, which between spin-levels in a homogeneous magnetic field, is another
we will focus on in the present review, all S = 3/2 (4F and 4P) and method that proved unambiguously that metal-ligand covalency
S = 1/2 states (except of b2D) could be identified in the spectra, does exist [19,27]. Metal-ligand hyperfine interactions manifested
including their splitting into spin-orbit coupled multiplets. Such in the EPR spectra of Cu2+(d9) and Ir4+(low-spin d5) complexes with
19
spectroscopic studies served as a basis for the development of F and 35,37Cl (both possessing nuclear spins, I = 1/2 and 3/2,
the quantum theory of atomic structure [12]. respectively) provided information about the delocalized spin-
The electronic structures of the free ions change in a character- density, and correspondingly on the transferred hyperfine field
istic way upon embedding in a coordination complex with neutral that a ligand nuclei experiences from the spin of the transition
or anionic ligands. Such complexes often display geometries of metal bound to it [19]. This was the first direct manifestation of
high symmetry – octahedral (Oh) or tetrahedral (Td). This enables metal-ligand covalency. Note, that if the mechanism of spin-
the use group theory for the classification of localized ground- transfer to the ligand would be spin-polarization, one would
and excited electronic levels, commonly referred to as terms or expect negative spin density at the position of the ligands (as, for
multiplets [13]. For predominantly ionic (Werner type) complexes, example, in the methyl radical, where the hydrogen atoms reside
the ndq (n = 3, 4, 5, q = 1–9) electronic configurations were shown in a node of the singly occupied molecular orbital). However, pos-
by spectroscopy to be well defined [14]. The dn configurations then itive spin-density at the directly coordinating ligand atoms must
give rise to a multitude of electronic states and to electronic tran- come from symmetry-allowed covalent delocalization. In order to
sitions in the visible or near infra-red (IR) spectral region referred account for such effects, molecular orbital theory was developed
to as d-d or ligand field transitions [15]. Such transitions are man- to include these effects in a straightforward and transparent man-
ifested in absorption bands which gives rise to the beautiful and ner. When applied to transition metal complexes, the resulting
rich colors that are characteristic of coordination complexes with treatment was referred to as ‘‘ligand field theory” [1,3,25].
open d-shells. As they are partly filled, the ndq configuration also Ligand field spectra and ligand field parameters extracted from
leads to ground and excited electronic states with both spin and a best fit of these parameters to experimental energies of electronic
only partially quenched orbital angular momenta, and therefore transitions provided the first indirect information about metal-
exhibit a magnetic response to an applied external magnetic field ligand bond at a time where neither density functional theory
[16,17]. Numerous experimental techniques such as near IR, visi- (DFT) nor ab initio methods were broadly in use [14,15]. Ligand
ble, UV (UV–VIS) absorption and emission spectroscopies, mag- field parameters consist of 10Dq (parametrizing the splitting of
netic circular dichroism (MCD) [18], and electron paramagnetic the 3d-orbitals in octahedral and tetrahedral ligand fields) and
resonance (EPR) [19,20] have been employed widely to probe the the Racah parameters B and C, which quantify interelectronic
electronic structures of transition metal complexes. repulsion. Systematic studies of these parameters allowed the clas-
The early view of theory designed to interpret and classify the sification and ordering of TM complexes into series of increasing
spectroscopic and magnetic properties of transition metal com- 10Dq and B/Bo (Bo the free ion value)–the spectrochemical
plexes considered ‘‘d-electrons” as completely localized on the [14,15] and nephelauxetic [28], series.
metal [13]. As such, they were deemed to be solely affected by The reduction of the parameter B with respect to free ion values,
electrostatic interactions with surrounding ligands, where the lat- 1 B/Bo of the same element, otherwise known as the nephelaux-
ter have been modeled by point charges or point dipoles. In addi- etic effect, lead to the concept of ‘cloud expansion’ for the d-
tion to parameters accounting for the Coulomb repulsion electrons, which provided a first attempt at determining metal-
between the d-electrons, the potential due to ligand charges and ligand covalency on a rigorous experimental basis [14,15]. Yet,
dipoles has been used to explicitly calculate the splitting of the the information content of the ligand field parameters deduced
4 S.K. Singh et al. / Coordination Chemistry Reviews 344 (2017) 2–25
from experiments, more specifically the exploration of these This review will be organized as follows. In Sections 2.1–2.3, we
parameters from first principles, remained elusive for a long time. will provide a brief outline of the non-relativistic CASSCF and
In addition, the fitting of the ligand field parameters suffered from NEVPT2 methods and show how spin-orbit coupling between
a rather limited number of experimental d-d transitions, allowing non-relativistic states can be taken into account using a mean field
one to only get two or at most three ligand field parameters from approach. In Section 2.4, the derivation of the full set of one- and
reported spectra. In addition, spectra of different quality have led two-electron ligand field parameters for complexes of any dn con-
to ligand field parameters, which are difficult to compare for differ- figured TM ions using AILFT will be described. Expanding on this
ent complexes. Connecting ligand field theory to rigorous first- subject, a more elaborate analysis of the parameters of r- and p-
principles quantum chemistry has been proven elusive for a long antibonding derived from AILFT data will be explained employing
time. On one hand, this was due to the lack of powerful computers the angular overlap model (AOM). In Section 3, the electronic
and efficient implementations of post-Hartree-Fock methods, and structure of the Cr3+, Mo3+ and W3+ free ions and basic trends in
on the other due to the lack of a suitable conceptual framework the variation of 10Dq, B, the spin-orbit coupling parameters f,
[29]. and their dependence on the chosen ligands for Cr(III) and for
First attempts to extract ligand field parameters using average- MCl36 (M = Cr, Mo, W) down the 3rd group will be discussed.
of-configuration Density Functional Theory (DFT) calculations Ligand field splitting diagrams derived from AILFT will be
were pioneered by Daul [30]. By using the multiple scattering employed to show the impact of the method for the field of molec-
Xa-method along with average-of-configuration fractional 3d-MO ular magnetism (Section 3.3.3). Finally, in Section 4 we will com-
occupation calculations, ligand field parameters were derived pare AILFT results (which represent the best fit to the entire
which provided a satisfactory interpretation of the electronic spec- manifold of S = 3/2(d3) states) with fits to a limited subspace of this
trum of CrCl4 and other complexes [31]. Subsequently, Anthon and manifold. This will show that fairly different results for the ligand
Schäffer [36] combined DFT with multiplet structure theory and field parameters are obtained in these two cases, thus demonstrat-
therefore reformulated Slater-Condon theory for atoms using DFT ing that great care must be exercised when comparing AILFT calcu-
[32–35]. Near perfect agreement between DFT and experimental lated ligand field parameters to experimentally deduced ones.
electronic repulsion parameters has been achieved in this way. Conclusions and outlook will be the subject of Section 5, followed
The method has been extended and applied to mimic the two- by a closing Section 6 on Computational details.
dimensional spectrochemical series for tetrahedral VX 4 [37] and
CrX4 [38], (X = F, Cl, Br, I) complexes by the same group. This 2. Methods
greatly aided in justifying old concepts born out from the interpre-
tation of spectroscopic experiments, such as additive and non- The electronic structure methods used in the context of this
additive ligand fields, the two-dimensional spectrochemical series, review are well known, and hence we will only briefly describe
and the nephelauxetic effect. Following these developments, a gen- their features. A more thorough discussion can be found in a num-
eral and user oriented DFT based ligand field method (LFDFT) has ber of other reviews [60,57].
been proposed and applied to a wide range of hexa- and tetra-
coordinated 3d transition metal complexes [39,40]. Extensive 2.1. Complete active space self-consistent field theory
reviews on the topic have been published [41–43]. Recently, circu-
lar dichroism spectra of tris-chelate complexes with planar conju- In this work, we are using molecular orbital (MO) theory and
gate ligands were studied using a combination of time-dependent focusing on a series of electronic states that arise from different
density functional theory and the angular overlap model [44]. distributions of a fixed number of ‘d-electrons’ among a fixed num-
Over the past decade, powerful ab initio quantum chemical ber of orbitals (in this case the five d-orbitals). However, all elec-
methods became available that allowed for the treatment of spec- trons of the complex under investigation are treated. Therefore,
troscopic and magnetic phenomena of transition metal complexes the orbitals of the system can initially be conveniently divided into
of unprecedented size (for example the program packages MOLCAS three groups: (1) inactive orbitals that are always doubly occupied,
[45] and ORCA[46] feature extensive functionality in this area). (2) active orbitals with variable occupation, and (3) virtual orbitals
Using the well-established complete active space self-consistent that are always empty. A given distribution of electrons among
field method (CASSCF) [47] combined with a second-order pertur- spatial orbitals with occupation numbers 0, 1 or 2 is referred to a
bation treatment of the dynamic electron correlation (N-electron ‘configuration’ (CFG). For any given CFG, the unpaired electrons
valence perturbation theory to second order, NEVPT2[48–51]), an can be coupled in various ways to a given total spin S. Any given
ab initio based ligand field theory (AILFT) was developed [52] linearly independent spin-coupling inside a given configuration is
and implemented into the ORCA program [46]. This method has referred to as a ‘configuration state function’ (CSF). In general, for
been extensively used in the field of magnetism and d-d absorption N unpaired electrons and spin S there are
spectroscopy. It not only allowed for the interpretation but also the ! !
prediction of exciting spectroscopic and magnetic properties of a S N N
fN ¼
wide variety of systems [53–59]. 1
2
N S 1
2
N S1
The aim of this review is not to provide a comprehensive

account on the application of ligand field theory to the field of a
linearly independent spin couplings ( is a binomial coefficient).
spectroscopy and magnetism of TM complexes. Rather than, we b
intend to show how the connection between first-principles elec- For any given state with spin S, there are 2S + 1 magnetic sublevels
tronic structure theory and ligand field theory can be used in prac- M = S, S1, . . . , S which are all degenerate so long as the Hamilto-
tice. We will demonstrate what AILFT offers as a complement to nian of the system is spin-independent (e.g. one is pursuing a non-
other approaches when applied to TM complexes with d3 configu- relativistic treatment). In the case of a non-relativistic treatment,
rations. The chosen systems are octahedral complexes of Cr(III) one may hence choose M = S for each state. In addition to S and M
with simple ligands such as I, Br, Cl, F, NH3 and CN. This being good quantum numbers, the electronic states of the system
choice is well suited to illustrate trends in the spectrochemical can be classified according to the irreducible representations (irrep)
and nephelauxetic series for Cr(III) and the variation of ligand field of the point group. For degenerate states this requires the symmetry
parameters down the 3rd row, taking [CrCl6]3, [MoCl6]3 and quantum numbers Cc, where C denotes the given irrep and c refers
[WCl6]3 as examples. to a component of the irrep. In general, A and B symmetry states are
one-dimensional, E states are two-dimensional and T-states are packages have sophisticated algorithms that greatly aid in the con-
three-dimensional. vergence of CASSCF wavefunctions.
Constructing all possible CSFs for a given S and C constitute a In this review, we are focused on the very small CAS(3,5)
‘complete active space’ (CAS). Thus, the CAS consists of a number active spaces and the states that result from it. In order to
E
Cc allow for connection between quantum chemistry and ligand
of CSFs that can be written as USM;
I , where I ¼ 1; . . . ; N CSF (N CSF
field theory, it is important that the five orbitals in the active
being the number of CSFs in the CAS). Since we focus on N-
space are predominantly of metal d-character. For most sys-
electrons in five d-orbitals for the purposes of this review, the
tems, this can be ensured by selecting proper MOs as an initial
dimension of the CAS is always small and never exceeds 75 (see
guess. The details of how to do this vary greatly among elec-
Table 1).
tronic structure programs. In any case, it is important for the
The possible non-relativistic many particle states in this mani-
user to verify that the active orbitals are indeed dominated to
fold are written as linear combinations of the CSFs:
a reasonable extent (where ‘reasonable’ is subjective and might
E X E
SM;Cc Cc be taken to be >65% for example) by basis functions of metal d-
WI ¼ C KI USM;
K ð1Þ
K
character.
Since there is a whole host of states which may be formed from
where the expansion coefficients C KI are found by minimizing the a CAS(N,5), it is customary to not minimize the total energy of each
total energy, leading to a linear matrix-eigenvalue equation: and every state individually, but to minimize the average energy of
several states.
HC ¼ EC ð2Þ
X D SM;Cc E
where the matrix element of the non-relativistic Hamiltonian is Eav ¼ W I WI ^ WSM;Cc
jHj ð5Þ
I
simply: I
D 0 0 0 0
E
Cc ^
HKL ¼ dSS0 dMM0 dCC0 dcc0 USM;
K jHjUSL M ;C c ð3Þ where the weights W I are user chosen. Typically, each state is
weighted equally. Here the user has to be careful since the
which emphasizes the block diagonal nature of the Hamiltonian results will vary with the number and identity of the states
with respect to the spin- and spatial symmetry quantum numbers. averaged over. Since most programs can only handle Abelian
Each CSF can, in principle, be expanded as a linear combination symmetry, care needs to be taken to include all members of a
of antisymmetrized orbital products (Slater determinants) that are given T or E term in averaging, otherwise uncontrolled symme-
built from a set of MOs fwg (note that in the interest of clarity, try breaking artifacts will arise. In the context of AILFT, it is
uppercase letters are used for many-electron quantities while low- common practice to average over all ligand field states. Some
ercase letters for one-electron quantities). The construction of CSFs programs will only allow one-irrep and one multiplicity (=2S
is elementary but somewhat laborious and hence, we will not go + 1) at a given time while other programs (like ORCA), allow
into any detail here and refer to the literature [61]. Each MO is in the user to average over any number of states for any number
turn expanded in a finite linear combination of basis functions of irreps and multiplicities. This is significant since optimizing
fug as: each irrep or multiplicity separately will result in several sets
X of MOs which are orthonormal within their own set but not
wp ðrÞ ¼ clp ul ðrÞ ð4Þ orthonormal between different sets.
l E
Cc
The result of a CASSCF calculation is a set of states WSM;
I;CAS with
For later convenience, note that inactive orbitals are given Cc
labels i; j; k; l, active orbitals the labels t; u; v ; w and virtual orbi- energies ESM;
I;CAS . For convenience, we will from now on drop the
tals the labels a; b; c; d while the labels p; q; r; s may refer to spin- and spatial symmetry labels for clarity, but with the under-
any orbital. The expansion of MOs in terms of basis functions standing that these are all good quantum numbers. The energy dif-
is exact if the basis set is mathematically complete (this is the ferences DI ¼ EI;CAS E0;CAS may be compared with experimental

basis set limit). In practice this is never the case, causing devia- (vertical) ligand field-excitation energies since the states WI;CAS
tions from the exact results due to basis set incompleteness. are many-particle equivalents of the multiplets that can be formed
These errors should be monitored to ensure that the results of in ligand field theory. However, through the minimization process,
a given study are not compromised by basis set incompleteness the orbitals and their energies reflect the nature of the ligands
artifacts. explicitly.
A complete active space self-consistent field (CASSCF) wave- CASSCF wavefunctions, while already being fairly elaborate, are
function is obtained if the total energy is not only minimized with still self-consistent field wavefunctions. The metal d-electrons (in
respect to the expansion coefficients C but also simultaneously with our case the active electrons), experience the average electric field
respect to the MO coefficients c. For N-electrons in M orbitals, one exerted by the nuclei and the other electrons of the system (core-
may refer to a wavefunction of this kind as CAS(N,M). This is a non- electrons on the metal and ligand electrons). They are very conve-
trivial, non-linear optimization problem, and quite frequently solv- nient wavefunctions, since they are still rather compact but have
ing it requires some skill and patience by the user of a given elec- the correct symmetry properties of the Hamiltonian with respect
tronic structure program. However, modern electronic structure to space and spin-symmetry and their number exactly equals the
number of possible ligand field excited states. This is a unique
advantage of this methodology [60] over density functional theory
Table 1 [62], where symmetry adapted states may, in general, not be
Number of CSFs that can be formed for a given multiplicity in each dn configuration.
formed, and hence the entire manifold of ligand field states is inac-
Configuration Multiplicities Number of CSFs cessible to this theory.
d1,d9 2 5 We should note in passing that CASSCF calculations require sig-
d2,d8 3, 1 10, 15 nificantly more computational resources than DFT calculations.
d3,d7 4, 2 10, 40 However, modern programs can easily treat systems with more
d4,d6 5, 3, 1 5, 45, 50
than 100–200 atoms, making molecular size hardly an issue for
d5 6, 4, 2 1, 24, 75
such studies anymore.
2.2. N-electron valence and complete active space second-order ðCASÞ

is one particular excited CSF. Here, the CASSCF coefficients C KI are
perturbation theory kept fixed at their CASSCF values. Clearly, the number of excited
CSFs one can form in this way is (nearly) independent of the num-
Despite the great convenience of CASSCF wavefunctions, CASSCF ber of CSFs in the CAS, which is a very large benefit of this approach.
energies are not highly accurate since they miss the all-important Thus, the appropriate ansatz for the wavefunction in intermediate
effects of dynamic electron correlation. By dynamic electron corre- normalization is:
lation, one refers to instantaneous electron-electron interaction E
~ X
‘events’ that go beyond the bulk of the electron-electron interaction WI ¼ WI;CAS þ T KI EK WI;CAS ð8Þ
energy that has already been captured in the Hartree-Fock SCF part KRCAS
of the calculation. The dynamic correlation energy is a small part of
the total energy (about 0.2%). However, given the total energy for where EK is a suitable excitation operator. Obviously, this ansatz is
already fairly small transition metal complexes is on the order of far less flexible than the more general form (Eq. (6)). However, a
thousands of electron Volts (eV), dynamic electron correlation plays host of experience indicates that the internal contraction in con-
a fundamentally important role for all of chemistry and spec- junction with second-order perturbation theory does not introduce
troscopy. We will extensively comment on the role of dynamic elec- important errors. Clearly, the excitation operator EK is confined to at
tron correlation in the context of ligand field theory throughout this most double excitations, since the Hamiltonian operator only con-
review. Here, we will only offer a few qualitative comments. tains one- and two-body terms. In general, there are eight distinct
One can think of dynamic electron correlation as an event classes of excitations, which correspond to different ways in which
where two-electrons come extremely close and avoid each other electrons can be promoted. These correspond to the excitations
by ‘jumping’ out of their mean-field (the so called Hartree-Fock ij ? tu, ta, ab, it ? uv, ua, ab and tu ? va, ab (note that tu ? vw does
sea) orbital into higher-lying, unoccupied orbitals. These events not occur since this is already part of the CAS). Of these, the excita-
can pertain to two, three, four or more electrons at a time. Hence, tion class ij ? ab contains the largest number of CSFs that grow as
one can calculate the dynamic electron correlation energy by form- the fourth power of the molecular size since the number of occu-
ing excited CSFs in which electrons, up to a given excitation level, pied and virtual orbitals both increase linearly with molecular size.
are excited into unoccupied orbitals. These CSFs, which will refer to This leads to a formally fifth-power scaling of the computational
E effort with respect to molecular size. However, efficient algorithms
~
as U K (the tilde denotes a dynamically correlated wavefunction), have been developed to do these calculations and even linear scal-
are then used in a linear expansion of the correlated wavefunction. ing algorithms have recently emerged [68,69].
E X X E A perturbation theory can then be formulated by choosing an
~ ~K
WI ¼ C KI jUK i þ T KI U ð6Þ appropriate 0th-order Hamiltonian. Here, the methods of NEVPT2
K2CAS KRCAS and CASPT2 differ significantly. The NEVPT2 formalism uses the
so-called Dyall-Hamiltonian [70] that contains one- and two-
Thus, one expands the many particle wavefunction in terms of body terms, while the CASPT2 formalism uses a pure one-body
the CSFs that are both inside the CAS (jUK i) and outside the CAS
E Hamiltonian. We do not want to enter into more detail here, and
~
(U K ). There are many ways to determine the expansion instead refer to the literature [48–51] for a more thorough discus-
coefficients C and T. If one applies the variation principle, the sion (see also [52,57]).
multi-reference configuration interaction (MR-CI) arises. In the The result of a CASPT2 or NEVPT2 calculation is a second-order
framework of coupled cluster (CC) theory, one of the very elaborate energy correction DEPT2
I , such that the total energy for each state is
multi-reference CC variants (MR-CC) arises. These are computa- EI ¼ EI;CAS þ DEPT2
I . However, it is important to understand that the
tionally very demanding calculations. However, the most straight- wavefunction is not changed by the treatment and still remains a
forward option is to apply perturbation theory. CASSCF wavefunction. Hence, all calculations of expectation val-
The lowest-order of perturbation theory at which electron- ues, magnetic operators or similar quantities are based on CASSCF
correlation effects arise is second-order. Given a reference wavefunctions in conjunction with second-order corrected total
wavefunction of the CASSCF type, second-order multi-reference energies.
perturbation theory (MR-PT2) is already a fairly elaborate under- The effects of dynamic electron correlation appear in multiple
taking. Only recently have programs become available which allow ways in the resulting electronic structure. One particular feature
for the application of these methods to large molecules with 100 or is that the dynamic correlation contributions for pairs of electrons
more atoms. Of the many possible variants of MR-PT2, we will of the same spin are much smaller than those for a pair electrons of
briefly comment on the second-order N-electron valence perturba- opposite spin. The reason for this behavior is that the self-
tion theory (NEVPT2) of Angeli, Malrieu and co-workers [48–52], consistent field already contains the so-called Fermi-correlation
the Generalized Møller-Plesset perturbation theory by Wolinski between electrons of the same spin and hence cannot come close,
and Pulay [63], and the complete active space second-order pertur- while opposite spin electrons, on average, repel each other too
bation theory (the CASPT2 method) by Roos and co-workers strongly in SCF type treatments like CASSCF. Therefore, there is a
[64–67]. Both methods (NEVPT2 and CASPT2) are fairly similar gross imbalance in the treatment of high- and low-spin states,
and many of the comments that we make below about our method where SCF treatments are hugely biased in favor of high-spin
of choice, NEVPT2, will also apply to CASPT2. states. Dynamic correlation introduces the so-called ‘Coulomb cor-
Both NEVPT2 and CASPT2 make use of the concept of ‘internal relation’ between electrons of opposite spin which restores most of
contraction’. By this one means that upon forming excited CSFs this balance. However, one needs to recover almost all of the
one does not excite each and every CSF inside the CAS individually, dynamic correlation energy and closely approach the basis set limit
but rather applies the excitation collectively to the entire CASSCF before quantitative accuracy is reached. Another typical failure of
wavefunction. For example, let Eai be an excitation operator that SCF methods is that ionic and neutral components of the wave-
promotes an electron from the doubly occupied orbital i to the function are almost equally weighted. This greatly disturbs the bal-
empty orbital a, then: anced description of chemical bonds, and metal-ligand bonds are
E X ðCASÞ a typically predicted overly ionic (metal character too high).
~a
Ui ¼ Ei WI;CAS ¼
a
C KI Ei jUK i ð7Þ The dynamic correlation treatment at the level of internally
K2CAS contracted second-order multireference perturbation theory does
not correct this failure. Hence, such calculations are best suited to 2.4. Ab initio ligand field theory
treat coordination complexes with fairly ionic bonds. Fortunately,
this is the case for very many Werner-type complexes. A cure for We are now in a position to discuss the connection between
this problem would involve incorporation of dynamic correlation first-principles CASSCF/NEVPT2 calculations and ligand field the-
effects in the orbital optimization. Such methods are in their ory, and how to uniquely extract ligand field parameters from
infancy at best. Alternatively, higher levels of correlation theory, these calculations [52].
for example MR-CI, MR-CC or difference dedicated CI (DDCI) [71] We have already established an active space CAS(n,5) that has
are able to restore the proper wavefunction even with suboptimal the same dimension as the full ligand-field problem. We have fur-
CASSCF orbitals. A detailed discussion of these subtle aspects thermore assumed that the CASSCF problem has been solved such
would lead beyond the scope of this article. that the five orbitals in the active space are of predominantly metal
If executed correctly, NEVPT2 and CASPT2 yield fairly similar character. Finally, we have added a correction for dynamic electron
results. As we will argue below, they capture about 70–80% of correlation.
the dynamic electron correlation energy. Hence, there are still The final step we must perform before a clean connection may
errors and challenges that arise from the electronic structure treat- be established is to canonicalize the active orbitals. It is well known
ment. However, for the intents and purposes of this review and that the CASSCF wavefunction and energy is invariant with respect
many practical applications in this area, the accuracy delivered to unitary transformations among the active orbitals. Thus, there is
by these methods relative to experiment is plenty sufficient (see flexibility that one can use here in order to arrange these active
Section 4). orbitals in a convenient manner. For an ideal match with ligand
field theory, one would want to have d-orbitals that are as ‘pure’
as possible. Furthermore, it is most convenient to use real d-
2.3. Quasi-degenerate perturbation theory for relativistic effects
orbitals in this treatment. Hence, the easiest way to ensure ‘clean’
d-orbitals is to diagonalize the z-component of the orbital angular
How to treat magnetic properties in the framework of
momentum operator over the set of active orbitals. This affords
CASSCF/NEVPT2 theory has been carefully reviewed recently
molecular orbitals that are ordered according to an approximate
[57] and therefore, we will not repeat this discussion at any
angular momentum quantum number ml ¼ 2; 1; 0; 1; 2, which
level of detail here. Briefly, spin-orbit coupling (SOC), magnetic
are then back transformed to their real spherical harmonics repre-
field effects and other spin-dependent operators are included
sentation wxy ; wyz ; wz2 ; wxz ; wx2 y2 . Note that these active orbitals are,
in the treatment by diagonalizing these relativistic operators
over the manifold of all CAS states (possibly corrected by of course, not pure d-orbitals but contain variable amounts of
NEVPT2 or CASPT2): ligand character as determined by the variational principle in the
preceding CASSCF calculation. Thus, neglecting metal-ligand over-
D E qffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðrelÞ SM;Cc S0 M0 ;C0 c0
HIJ ¼ dIJ ðEI;CAS þ DEPT2
I Þ þ WI;CAS jH SOC þ . . . jWJ;CAS ð9Þ lap, one may write wd ad ud þ 1 a2d uL , where ud is a (linear-
combination of) metal d-orbital basis function(s) and uL is a sym-
Please note carefully that these operators are not diagonal in the
metry adapted linear combination of ligand orbitals, while ad is
spin- or spatial symmetry quantum numbers and hence all mag-
the variationally determined mixing coefficient that can be related
netic sublevels of all multiplets formed in all multiplicities must
to ‘symmetry restricted covalency’. The latter will be subjected to a
be included in this treatment simultaneously.
deep analysis elsewhere. What has been achieved by canonicaliza-
The form of the SOC operator, however, deserves special men-
tion is that, to the largest extent possible, these MOs only have one
tioning. In general, the SOC operator is a two-electron operator.
real d-orbital component each, e.g. wxy , etc. Note that the phases of
However, it can be well approximated by an effective one-
the MOs must be adjusted such that the coefficient with which that
particle operator, the spin-orbit meanfield (SOMF) operator
d-orbital component enters in the given MO is positive.
[72,73]. The latter operator is of the form:
In ligand field theory, one works with fictitious pure d-orbitals
ðSOMFÞ
X
HSOC ¼ zSOMF si ð10Þ (of only metal character) that we order in the same way
i E
Cc
i dxy ; dyz ; dz2 ; dxz ; dx2 y2 . We now can build ligand field CSFs USM;
K;LFT
where zSOMF is an appropriately defined effective one-electron by distributing the n-electrons over the five ligand-field d-
i
operator that includes a one-electron term, a two-electron spin- orbitals. In the exact same way and order, we build the actual
same-orbit, and a two-electron spin-other-orbit term. The sum is CAS CSFs, distributing the n active electrons among the
taken over all electrons and si is the spin-operator for the ith elec- wxy ; wyz ; wz2 ; wxz ; wx2 y2 active orbitals. By this construction there is
tron where both si and zSOMF have three spatial components. Note a 1:1 correspondence of each and every ligand field CSF with each
i
that the occurrence of the spin of each individual electron is and every CAS CSF.
important since the SOC operator couples states of different spin Having established this 1:1 correspondence at the level of many
which would not be the case for the operator of the total spin S. particle basis functions, the remaining treatment is straightfor-
Hence, as argued many times previously, (e.g. [57]), the SOC oper- ward. Using the ligand field CSFs, we evaluate the ligand field
ator kLS, in which the total-spin operator S occurs, is strongly dis- matrix:
D E
couraged despite the fact that it is very prevalent in the ligand SM;Cc ^ SM;Cc
HLFT
KL ¼ UK;LFT jHLFT jUL;LFT ð11Þ
field literature. The operator kLS is an effective Hamiltonian that D E D E
Cc ^
operates within one Russel-Saunders multiplet of free atoms and ¼ USM;K;LFT jhLFT jUL;LFT
SM;Cc
þ USM; Cc
^
K;LFT jg LFT jUL;LFT
SM;Cc
ð12Þ
ions. It is very valuable in this application. However, outside this |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
f ðV LFT Þ f ðB;CÞ
special application it is ill-defined and will lead to erroneous
results. Note also that zSOMF
i is quite different from the ‘effective where the second line emphasizes, that the ligand field matrix
potential’ SOC operators that are commonly used in the DFT element consists of a one-electron term that is a function of the
framework. The latter miss the spin-other-orbit contribution, one-electron ligand field matrix elements:
which is relatively large. The SOMF operator is typically 99% accu-
D E
rate as will be shown below in comparisons with data from atomic V LFT ^ LFT jd0
d;d0 ¼ djV ð13Þ
spectroscopy.
where V^ LFT is the effective ligand-field potential, e.g. the potential Its matrix elements over the five d-orbitals are
SOC;LFT 0
created by the ligands at the place of the metal. We will not need hdd0 ¼ fhdjlsjd i, which can be compared to the SOMF matrix ele-
to know analytical expressions for these matrix elements as we will SOMF
ments hdd0 ¼ hwd jzSOMF sjwd0 i, resulting in an equally unambiguous
treat the matrix elements V LFT
d;d0 themselves as parameters. The two- fit of the ligand-field spin-orbit coupling constant f.
electron part of the ligand-field matrix hUSM;Cc
^ SM;Cc
K;LFT jg LFT jUL;LFT i is simply
At this point in the development one has unambiguously deter-
a function of the two-Racah parameters B and C. Hence, given that mined VLFT , B and C (actually C can only be determined if there is at
V LFT
is Hermitian, there is a total of fifteen one-electron parameters least one spin-multiplicity block with 2S + 1 = 2(Smax 1) + 1 pre-
d;d0
and two two-electron parameters to be determined. We will collect sent in the CAS as energies of states with the highest spin-
these parameters in one vector p. multiplicity 2Smax + 1 do not depend on C) and f.
At the level of the ab initio treatment, we have the CASSCF However, in chemistry one wants to dig more deeply into the
D E nature of the ligand field. Hence, there is an additional step – the
matrix elements HCAS ¼ USM;Cc jHj ^ USM;Cc that are evaluated with-
KL K L interpretation of the ligand field matrix in terms of a convenient
out any further approximation. model. Such a model can, for example, be the electrostatic model
Obviously, one would like to evaluate the parameters p such or the angular overlap model (as will be demonstrated below).
that the differences between the ab initio Hamiltonian and the It is of utmost importance to appreciate the different nature of
LFT Hamiltonian are minimized, e.g. we define the AILFT this step relative to what has been outlined before: the ab initio
functional: treatment unambiguously determines VLFT , B, C and f. Once the
X LFT 2 CASSCF calculation has been converged and the user requests the
RðpÞ ¼ ðHKL HCAS
KL Þ ¼ min ð14Þ
AILFT specific canonicalization, these values are automatically cal-
KL
culated and printed. For example, in the ORCA program package
Importantly, in Ref. [52] we have pointed out that the ligand the user requests ‘actorbs = dorbs’ in the input file. If the CASSCF
field expressions are linear in all ligand field parameters. Hence, space consists of just N electrons in five orbitals and there is a single
one can write: metal center (which the program checks for) and the active orbitals
X @HLFT are dominated by metal d-character (the program checks for that as
HLFT LFT
KL ðpÞ ¼ H KL ð0Þ þ
KL
pk ð15Þ well), the AILFT procedure is automatically performed. Importantly,
@pk
k there is no room for subjectivity or human interference. As long as
Inserting that linear expression and minimizing the AILFT func- the CASSCF calculation converges to a proper solution in the sense
tional yields a linear equation system: explained above, this is an immediate and well-defined result. By
contrast, the second step – the interpretation – is not automatic. It
Ap ¼ b ð16Þ consists of the scientist creating a model that introduces a number
with of one-electron parameters that can then be determined from
knowing the numerical values of the matrix VLFT . This step is subjec-
X @HLFT @HLFT
Akl ¼ KL KL
ð17Þ tive and depends on the human creativity. It may not be unique.
KL
@pk @pl For example, in a strictly octahedral complex, with Cartesian
axes parallel to the three C4 axes, the matrix VLFT will be diagonal
X @HLFT and the diagonal matrix will consist of a threefold degenerate
bk ¼ KL
HCAS ð18Þ
KL
@pk KL
value and a twofold degenerate value. The difference between
these two values is quantified by 10Dq in the electrostatic model
@HLFT
Hence, knowing the analytical expressions for KL
@pk
(which are sim- or 3er 4ep in the angular overlap model (note that in this case,
er and ep cannot be separately determined). How many parameters
ple numerical coefficients) and the numerical values of HCAS
KL , we
can be unambiguously determined depends on the specific case
obtain the solution:
and chosen model, but certainly not more than fourteen, which
p ¼ A1 b ð19Þ is the number of independent elements of VLFT that enter energy
differences.
This equation either has no solution, in which case there is a lin-
In our opinion, AILFT is a remarkable construction that provides
ear dependency, or a unique solution. This shows that by using the
a very powerful link between modern first-principles electronic
AILFT construction outlined above the ab initio electronic structure
structure theory and chemical concepts that are deeply engraved
defines the ligand field parameters unambiguously. We note in
into the language and thinking of coordination chemists. We will
passing that in practice we are solving the linear equation system
move on in the following paragraphs to describe and illustrate
by making use of the Moore-Penrose inverse (pseudoinverse).
the properties of this simple, yet highly effective theory.
In the case of NEVPT2, one can pursue an approximate treat-
ment, where one defines a NEVPT2 Hamiltonian in the following
way (a procedure referred as spectral resolution of the 3. AILFT extracted ligand field parameters for octahedral
Hamiltonian): M3+(d3) ions
* +
X CAS D E

HNEVPT2
KL ¼ UK j jWI ðECAS
I þ DENEVPT2
I Þ WCAS
I jUL ð20Þ AILFT is equally applicable to metal complexes and to free metal
I ions. To explore trends across the series Cr3+, Mo3+ and W3+ we first
consider the electronic structure of the free ions. We then proceed
which is then used in place of HCAS
KL . A more rigorous way of using an
to analyze the trends of the ligand field parameters in complexes of
intermediate effective Hamiltonian which has been outlined in Ref.
series 6B taking MCl3 III
6 (M : Cr, Mo and W) as prototypes. In the
[52], but not yet implemented. Numerical results with that
last section, we study the series of complexes CrIIIX6 (X = I, Br,
improved scheme will be presented elsewhere.
Cl, F, NH3 and CN) with an aim to understand the trend of
The ligand field spin-orbit operator is:
X ligand field parameters. Moreover, special attention has been
HSOC
LFT ¼ f li si ð21Þ devoted to probe the nature of the metal-ligand interactions using
i AILFT computed ligand field parameters.
3.1. Electronic structure of the free 3d, 4d and 5d M(d3) metal ions, When applied to free ions, the AILFT module of ORCA yields the
M=Cr3+,Mo3+,W3+ parameters B, C, and in addition (provided relativistic SOC module
is switched on) the effective spin-orbit coupling parameter f. For
3.1.1. AILFT parameters for the free M3+(d3) ions the free ions Cr3+, Mo3+, and W3+ these parameters are listed in
Energies of electronic transitions (Table 2) have been expressed Table 2. Their values will serve as a reference in the analysis of
in terms of the parameters of interelectronic repulsion A, B and C the effects of covalency on the reduction of B, C and f in the pres-
(Racah-parameters), or alternatively in terms of Slater-Condon ence of ligand environments (nephelauxetic and relativistic
parameters F0, F2 and F4. Their relationship is given in Eqs. (22)– nephelauxetic effects, respectively in Sections 3.2 and 3.3 below).
(25).
A ¼ F 0 49F 4 ð22Þ 3.1.2. Interpretation of the AILFT parameters for the free M3+(d3) ions
The d3 configuration of the ions under consideration gives rise
B ¼ F 2 5F 4 ð23Þ to the following manifold of non-relativistic 2S+1L Russel-
Saunders multiplets: 4F (the ground state) and 4P, 2G, 2P, 2D1, 2H,
C ¼ 35F 4 ð24Þ 2
F and 2D2 (excited states). Their energies from CASSCF and
NEVPT2 calculations are listed in Table 3. Spin-orbit coupling
F 2nd F4 induces the splitting of each 2S+1L term, resulting in sublevels char-
F 0 ¼ F 0nd ; F 2 ¼ ; F 4 ¼ nd ð25Þ
49 441 acterized by total angular momenta J. Terms with the same J
belonging to different 2S+1L multiplets can therefore mix. In Table 3
While A quantifies the average repulsion between two d-electrons
we also include the spin-orbit sublevels 4FJ of the ground state.
of the d3 configuration it merely leads to a uniform shift of all states
The trends of the parameters B, C and f across the series Cr3+,
by 3A. The parameters B and C describe the angular dependence of
Mo3+ and W3+ are visualized in Fig. 1. Values of the parameter C
the five d-orbitals on the electron-electron repulsion and therefore
extracted from experiment shows non-monotonic changes: it
enter into state energy differences. For a free ion with a d3 configu-
decreases from Cr3+ to Mo3+, followed by an increase toward
ration, the multiplet energies are given in Eqs. (26)–(31) [74].
W3+. This increase may be attributed to 5d3 -5d26s1 configuration
Eð4 FÞ ¼ 3A 15B ð26Þ interaction, which shifts the transitions 4F(5d3) ? 2G(5d3) and 4F
(5d3) ? 2P(5d3) up in energy (these energies are proportional to
Eð4 PÞ ¼ 3A ð27Þ C, see below). By contrast, the energy of the spin-allowed transition
4
F(d3) ? 4P(d3) depends solely on B and presumably remains
Eð2 HÞ ¼ Eð2 PÞ ¼ 3A 6B þ 3C ð28Þ weakly affected by the mixing between the 5d3 and 5d26s1
configurations.
Eð2 GÞ ¼ 3A 11B þ 3C ð29Þ In order to analyze the trends of the parameters of the free Cr3+,
Mo3+ and W3+ ions on a more quantitative basis, we use a model
Eð2 FÞ ¼ 3A þ 9B þ 3C ð30Þ that relates B, C, and f with the radial part of the 3d, 4d and 5d
wavefunctions. In doing so, we make the strongly simplifying
1=2 assumption that the radial part of the metal d-orbitals can be suf-
Eð2 DÞ ¼ 3A þ 5B þ 5C ð193B2 þ 8BC þ 4C 2 Þ ð31Þ
ficiently well described by a single effective Slater radial function
Table 2
M3+(d3) (M = Cr, Mo, W) free ion AILFT parameters (in cm1) from AILFT calculations.a
Parameter Method Cr3+ Mo3+ W3+

B CASSCF 1,157 818 768
NEVPT2 991 686 653
C CASSCF 4,327 3,293 3,132
NEVPT2 3,844 2,835 2,698
f CASSCFb 284 787 2,892
a 3 3+
CAS(3,5) CASSCF and NEVPT2 results employing all 10 S = 3/2 and 40 S = 1/2 multiplets (roots) split out from the d configuration of M .
b
Note that there are no separate NEVPT2 values, since f is directly calculated from the CASSCF orbitals.
Table 3
Non-relativistic (obtained following averaging over the spin-orbit split components) and relativistic (4FJ, J = 3/2,5/2,7/2, 9/2 ground state sublevels) energy levels (in cm1) of the
free d3-ions Cr3+, Mo3+ and W3+ from CASSCF and CASSCF/NEVPT2 calculations.
Terms Cr3+ Mo3+ W3+

CASSCF NEVPT2 CASSCF NEVPT2 CASSCF NEVPT2
4
F 0 0 0 0 0 0
4
P 17,359 14,444 12,268 10,002 11,524 9,730
2
G 17,609 15,420 13,150 11,023 12,468 10,611
2
P 23,395 20,925 17,240 14,444 16,309 12,602
2
D2 25,455 21,658 18,893 15,944 17,894 15,114
2
H 23,395 20,925 17,240 14,444 16,309 14,102
2
F 40,755 35,138 29,508 24,310 27,883 23,182
2
D1 64,103 55,761 46,751 39,988 44,153 38,136
4
F3/2 0 0 0 0 0 0
4
F5/2 244 247 731 746 3,481 3,681
4
F7/2 579 582 1,674 1,692 6,914 7,055
4
F9/2 997 1,000 2,771 2,772 9,869 9,735
Z
a2 1 dV Z
fnd ¼ R2nl ðrÞ dr; V ¼ ð34Þ
2 r dr r
Employing an effective exponent of n = 3.11 (pertaining to
Mo3+(4d)) we plot in Fig. 2 the radial probability distribution of a
4d electron R24d ðrÞr2 along with dependence of the quantities
F2(r)/F2, F4(r)/F4 and f(r)/f on r (this implies integration of Eqs.
(33) and (34) over a sphere of finite radius r (F2,4 – the actual values
of these parameters computed with r ? 1). It is remarkable to
learn from these plots that R24d ðrÞr2 peaks at a distance r = 0.8 Å
which is 2–3 times smaller than typical metal-ligand single bond
lengths (R = 1.9–2.5 Å). This is also the radius of a sphere inside
which F2,4 and f gain a large amount of their values (50% for F2,4
and 90% for f). It also emerges from Fig. 2 that contributions to f
Fig. 1. Trends in the parameters B, C and f across the isoelectronic series Cr3+, Mo3+ are largely biased toward the core region establishing spin-orbit
and W3+ from a best fit to energies of electronic transitions from experimental data
coupling as a local property. Finally, Fig. 2 lends support to ligand
[11] (see Section 6.4 detailing the procedure of extraction) and from NEVPT2/AILFT
calculations.
field theory for complexes of transition metals in normal oxidation
states, justifying use of atomic-like parameters to describe inter-
electronic repulsion and spin-orbit coupling.
Table 4 Now, we can turn the argument around and ask for the orbital
Expressions for the Slater-Condon parameters Fo,F2, F4, Racah B, C and C/B, and the exponent that is required to reproduce the values of F2, F4, and f
effective Spin-Orbit Coupling parameter f in terms of the effective exponent n of which were determined in the AILFT treatment (Table 2) These
Slater type orbital Rnd ¼ Nrn1 expðnrÞ (in cm1) for gaseous Cr3+, Mo3+ and W3+ ions.
are in included in Table 5.
Parameter Cr3+(d3) Mo3+(d3) W3+(d3) Based on these results, we can state the following:
Fo 56650.1n3d 44089.9n4d 36157.9n5d
F2 610.3n3d 504.1n4d 432.2n5d (i) Effective exponents derived from F2, F4, and f are different.
F4 44.2n3d 37.4n4d 32.7n5d (ii) Effective exponents extracted from spin-orbit coupling
B 389.3n3d 317.1n4d 268.7n5d parameters are larger than those deduced from F2 and F4.
C 1547.0n3d 1309.0n4d 1144.5n5d
C/B 3.97 4.13 4.26
This reflects the sensitivity of spin-orbit coupling to electron
f 9.35n3d3 5.84n4d3 4.80n5d3 distributions closer to the nuclei. By contrast, F2 and F4 are
very sensitive to dynamical correlation, a point which will
be the subject of a more detailed analysis reported
(Eq. (32)). This affords analytical expressions for F knd (and following elsewhere.
Eqs. (22)–(25)) of B and C as well and f (Table 4) (iii) It was stated in Ref. [75], which is dedicated to spectra of Cr
Rnd ðr; nÞ ¼ ð2nÞnþð1=2Þ ½ð2n!Þ1=2 rn1 enr ð32Þ (III) doped Al2O3 and other complexes of Cr(III) (see also the
discussion in a monograph on ligand field parameters [76]),
as defined by Eqs. (33) and (34) (Z-the atomic number, r-the elec- that F2 is more sensitive to the outer part of the 3d-electron
tron nuclear distance, a = 1/c is the fine structure constant where density distribution than F4 is. Based on the Slater exponents
c 137 is the speed of light in atomic units), respectively. for Cr3+, we find an opposite order where F2 (n = 2.38)
ZZ slightly exceeds F4 (n = 2.27). However, the situation may
r k< 2
F knd ¼ R ðr 1 ÞR2nd ðr2 Þr 21 r22 dr 1 d2 ð33Þ change when moving to complexes. Similar close values of
r >kþ1 nd
n(F2) and n(F4) are found for Mo3+, while for W3+ n(F2) and
n(F4) differ significantly. Presumably, configurational mixing
between the 5d3 and 5d26s1 is responsible for the more com-
pact nature and larger value of n(F4) (=C/35). This readily
explains the ratio of C/B deduced from experiment (=6.94),
which is larger than expected (4.26, cf. corresponding
entries in Tables 4 and 5).
3.2. Periodic trend across the group 6B: studies on [MCl6]3 (where
M = Cr(III), Mo(III) and W(III)) complexes
3.2.1. AILFT parameters for the [MCl6]3 (where M = Cr(III), Mo(III)

and W(III)) complex series
The AILFT parameters for the octahedral [MCl6]3 complexes
(M = CrIII, MoIII, WIII) are collected in Table 6. All calculations have
Table 5
Effective exponents n(F2), n(F4) and n(f) extracted from a best fit of F2, F4 and f to
energies of electronic transitions from experimentally reported atomic energy levels
of Cr3+, Mo3+ and W3+.
Fig. 2. Radial probability distribution R4d2(r)r2 for a Slater type 4d orbital R4d(r) Slater exponent Cr3+(d3) Mo3+(d3) W3+(d3)
= Nr3exp(nr) taken as an example (n = 3.11, pertinent to Mo3+) and the dependence n(F2) 2.38 2.02 2.28
of the F2(r)/F2, F4(r)/F4 (F2 = F2(r = 1), F4 = F4(r = 1)) Slater-Condon and the effective n(F4) 2.27 2.07 3.00
spin-orbit coupling f(r)/f(f = f(r = 1)) parameters on the electron-nuclear distance n(f) 3.08(Z = 24) 5.23(Z = 42) 8.44(Z = 74)
r.
Table 6
Ligand field parameters (in cm1) for MCl6 complexes (MIII = CrIII, MoIII, WIII) from AILFT calculations.a
Method 10Dq B f er ep
3
[CrCl6] CASSCF 10,927 1,022 252 4,833 808
NEVPT2 11,332 971 252 5,060 862
[MoCl6]3 CASSCF 18,033 665 680 8,516 1,766
NEVPT2 18,684 635 680 8,409 1,480
[WCl6]3 CASSCF 20,931 608 2,492 12,351 3,995
NEVPT2 21,475 555 2,492 12,362 3,838
a
CAS(3,5) CASSCF and NEVPT2 results employing all 10 S = 3/2 states that arise from the d3 configuration of M3+.
been performed on the DFT optimized geometries (see Section 6.2

for more details). In these calculations only the ten S = 3/2 elec-
tronic states have been taken into account; including the 40
S = 1/2 states would lead to results which, because of the missing
dynamical electron correlation, are biased in the wrong direction.
The parameters amenable to the AILFT procedure are therefore
10Dq and B. As was stated in Section 2.4, the AOM parameters er
and ep (see [77] and original references therein) are covariant
(i.e. related through 10Dq = 3er 4ep) and therefore not accessible
for perfect octahedral symmetry. To extract these AOM parameters
from AILFT results Eqs. (35) and (36) are used for the energies of
the 3d-MOs in a trigonally distorted (D3d) octahedral complex. In
this point group symmetry, the t2g orbitals split into a1g and eg
and the latter orbital mixes with the one of the eg parent
symmetry.
Fig. 3. Variation of the nephelauxetic reduction of B (in %) with the nature of the
a1 : central metal atom M = Cr(III), Mo(III), W(III) for the series of [MCl6]3 complexes

2 9 2 3 ð35Þ (CASSCF results).
jdz i : ep sin 2h þ er ð1 þ 3 cos 2hÞ2
2 8
e:
2 3
jdyzðxzÞ i jdx2 y2 ðxyÞ i
6 2 3 7
4 3ep ðcos2 h þ cos2 2hÞ þ ð9=4Þ sin 2her ð3=2Þð4ep 3er Þ cos h sin h 5
3 2 2
ð3=2Þð4ep 3er Þ cos h sin h ½3ep ð1 þ cos2 hÞ þ ð9=4Þ sin her sin h
ð36Þ
To deduce the parameters er and ep, we carried out CASSCF and
NEVPT2 calculations on MX3 6 model complexes with metal-ligand
bond distances determined by DFT geometry optimization (see
Section 2 for their values and comparison with experimental data)
and a value of 60° for the polar angle h (cf. with the value for the
octahedron of 54.735°). Values for er and ep were then obtained
by a least-squares fit to the AILFT data. Their values are included
in Table 6.
3.2.2. Interpretation of the AILFT parameters for the [MCl6]3 (where

M = Cr(III), Mo(III) and W(III)) complex series Fig. 4. Variation of the nephelauxetic reduction of the effective spin-orbit coupling
The trends in covalency across the series are already nicely parameter f (in %) across the complex series MIIICl3 III
6 (M = Cr(III), Mo(III),W(III))
reflected in the CASSCF results, which show a reduction of B [in (AILFT/CASSCF); the numbers at the bottom refer to the value of the reduction of the
free ion values, f(M3+) = 284(Cr3+), 787(Mo3+), 2892(W3+) cm1 in the given
Fig. 3 we have plotted the percentage reduction in the B parameter, complex, Df = f(M3+)-f(MIIIX6).
((1 B/B0) 100)] and the spin-orbit coupling parameter 1 [see
Fig. 4, ((1 1/10) 100)] as one moves down the series; larger val-
ues of (1 B/B0) and (1 1/10) indicate larger covalency. covalent radius of W changes only marginally relative to Mo, since
A closer look at Fig. 3 shows that in case of [MCl6]3 complexes, it is affected by the contraction of the 4f shell (Lanthanide contrac-
the B value drastically reduces as one moves from the Cr(III) to Mo tion). For the W(III) ion, the two effects counteract one another,
(III); however, there is a slight reduction of B as we move further leading to a very small decrease in the B parameter as one moves
down from Mo(III) to W(III). As we move down from the first to down from the Mo(III) to W(III). This is true for both the free ion
the second-row transition metals, the covalent radii (as a measure (Table 2) and in the complex (Table 6). Within the adopted model,
of atomic size) increase with the principal quantum number, which the reduction of B in a complex has two sources: (1) changes in the
naturally increases the size of the orbitals. This explains the drastic radial wavefunction of the d-orbitals, which are related to the
reduction in the B parameter from Cr(III) to Mo(III). On the other decrease of the effective nuclear charge Z⁄ (central field covalency
hand when moving from Mo to W the principal quantum number [14]), and (2) mixing between the metal and ligand orbitals deter-
increases, which also increases the orbital size. Despite this, the mined by the variational principle, within the restrictions of point
group symmetry (symmetry restricted covalency [14]). The latter

effect requires a deeper analysis that will be the subject of a sepa-
rate report.
The second important parameter that reflects metal-ligand
covalency is the spin-orbit coupling constant (f). Orbital dilution
has a crucial impact on the SOC constant and thus the associated
properties as metal hyperfine coupling and the appearance of
ligand superhyperfine structure in ESR measurements. The rela-
tivistic nephelauxetic reduction of f is computed here at the
CASSCF level. The computed f parameters for complexes are much
lower compared to that of free ions – a typical feature of transition
metal complexes. To analyze the relativistic nephelauxetic reduc-
tion in Fig. 4 we have plotted the percentage reduction ((1 f/
f0) 100) across the series. A first glance at Fig. 4 shows no trend
in this reduction. There is a large increase of the percentage reduc-
tion in the f as we move from [CrCl6]3 to [MoCl6]3, while no
change is witnessed when moving further down to the 3rd transi-
Fig. 6. The two-dimensional spectrochemical series reflecting the r and p-donor
tion row. This is essentially due to the very large f of the free WIII character of involved ligands quantified using the angular overlap model applied to
ion which is present in denominator, making almost no change in the MIIICl6 complex series (M = CrIII, MoIII, WIII) values of the parameters er and ep
the percentage reduction from Mo to W. On the other hand, the (expressed in cm1) have been obtained from AILFT/CASSCF and AILFT/NEVPT2
Df = f(M3+)-f(MIIIX6) term changes abruptly as we move towards calculations employing trigonally compressed (D3d) model complexes with a polar
angle h(C3-M-Cl) of 60°.
the 3rd transition row. In comparison to the [CrCl6]3 complex,
the Df parameter is approximately 3.5 times higher for [MoCl6]3
complex and 13 times higher for the [WCl6]3 complex.
The quantity 10Dq is the energy separation between the eg and Taken together, metal-ligand covalency reflected by the
t2g orbitals in a cubic/octahedral environment. In general, for d3 nephelauxetic effect in B and f increases as we move down the
ions in octahedral environment, the 10Dq parameter is given by group, which is further supported by the er and ep parameters
the energy of the 4T2(F) term i.e. it is determined by the first extracted from AOM. These relationships have been documented
spin-allowed transition. The CASSCF and NEVPT2 values of 10Dq in many review articles (see e.g. [14] and two classical monographs
are plotted in Fig. 5. As expected, 10Dq increases linearly as we [15,18]). It is gratifying to see that AILFT is able to reproduce all of
move down the group from 3d to 5d. From, Cr(III) to Mo(III), we these well-known trends without any recourse to experimental
have noticed a 58 percent increase of 10Dq, while a 77 percent data. To the best of our knowledge this has never been achieved
increase has been calculated for Cr(III) relative to W(III). This before.
increase is attributed to the more diffuse nature of d-orbitals in
heavier transition metals. This enhances the overlap between the
ligand and metal orbitals, leading to a larger 10Dq value. This is 3.3. Insight into the spectrochemical series from AILFT: studies on
consistent with classic extend Hückel arguments, where the [CrX6]3+/ (X = F, Cl, Br, I, NH3 and CN) complexes
energy destabilization of a metal d-based orbital is roughly hML2/
(e(metal)-e(ligand)), where e(metal/ligand) are orbital energies for 3.3.1. AILFT parameters for the [CrX6]3+/ (X = F, Cl, Br, I, NH3 and
the interacting orbitals prior to complex formation. The resonance CN) complexes
integral hML is typically taken to be proportional to SML, the metal The AILFT parameters for the octahedral [CrX6] complexes
ligand overlap integral. (X = F, Cl, Br, I, NH3 and CN) are collected in Table 7. As for
A two-dimensional plot for er and ep parameters confirms the MCl3 III III III
6 (M = Cr , Mo , W ) only 40 S = 3/2 states out of the entire
presence of the r- and p-donor character of the ligand. From manifold of 120 microstates have been taken into account. This
Fig. 6 (see also Table 6), it is evident that both the er and ep param- yields the AILFT values of 10Dq and B. The parameters er and ep
eters increase as one moves down the group (for either CASSCF or were again deduced by making use of D3d distorted model com-
NEVPT2). The increase in er and ep is essentially due to the largely plexes following the recipe detailed in Section 3.2.1. Their values
diffuse nature of the 4d and 5d magnetic orbitals of Mo and W, are included in Table 7 and compared with er and ep extracted
respectively, as already discussed for 10Dq. from spectra of tetragonal complexes of Cr(III) [78–80].
Fig. 5. Trends in 10Dq across the 6B group of the periodic table as quantified using AILFT/CASSCF and AILFT/NEVPT2 values of 10Dq for the of the complex series [MIIICl6] 3
(left); An example of ligand field orbitals splitting for the [MoCl6]3 (right).
Table 7
Ligand field parameters (in cm1) for MX6 complexes (MIII = CrIII, X = F,Cl,Br, I, NH3, CN) from AILFT calculations and extracted from a least squares fit to d-d absorption
spectra of trans-tetramine complexes with the corresponding donors as co-ligands.a
Method 10Dq B f er ep
[CrF6]3 CASSCF 11,299 1,091 271 6,929 2,345
NEVPT2 11,227 1,007 271 7,019 2,420
expb – – – 7,400 1,700
[CrCl6]3 CASSCF 10,927 1,022 252 4,833 808
NEVPT2 11,332 971 252 5,060 862
expb – – – 5,500 900
[CrBr6]3 CASSCF 9,459 1,013 216 3,833 424
NEVPT2 10,168 945 216 4,180 485
expb – – – 4,900 600
[CrI6]3 CASSCF 9,200 986 151 3,452 189
NEVPT2 10,375 944 151 3,926 223
expb – – – 4,300 600
[Cr(NH3)6]3+ CASSCF 16,898 1,037 265 5,780 –
NEVPT2 17,913 1,073 265 6,147 –
expb – – – 7,000 0
[Cr(CN)6]3 CASSCF 24,313 961 254 11,306 230
NEVPT2 27,458 1,007 254 12,185 210
a
CAS(3,5) CASSCF and NEVPT2 results employing all 10 S = 3/2 states that arise from the d3 configuration of M3+.
b
Adopted from Ref. [78–80] extracted from the spectra of hexa coordinate amine complexes with tetragonal symmetry and corresponding X = F,Cl,Br,I and NH3 as co-
ligands.
3.3.2. Interpretation of the AILFT parameters for the [CrX6]3+/ (X = F,

Cl, Br, I, NH3 and CN) complexes
We have computed d-d transitions for [CrL6]3+/ (where L = F,
Cl , Br, I, NH3 and CN) energies. A comparison of their values

with experimentally reported ones will be the subject of Section 4.

The AILFT ligand field parameters were extracted using both
CASSCF and NEVPT2. The gap between the ground state, 4A2g,
and first excited state, 4T2g, is often considered as the ‘‘spectro-
scopic” 10Dq. It is highly sensitive to the dynamic electron correla-
tion as witnessed by the difference between the CASSCF and
NEVPT2 values. Upon inclusion of dynamic correlation, the
4
A2g ? 4T2g transition undergoes positive corrections, as the
ground state is stabilized more strongly than the excited state.
The correction for the first spin-allowed 4A2g ? 4T2g transition is
smallest in the case of CrAF bonds (1690 cm1) and increases as Fig. 7. AILFT derived 10Dq parameters (both at CASSCF and NEVPT2) for the [CrIII-
X6]3+/ (where X = F, Cl, Br, I, NH3 and CN) complexes.
we move towards the heavier halides; for the CrAI bond, the cor-
rection is computed to be as large as 4164 cm1. The changes in
the energies of d-d transitions have been noticed in our previous
review [52], where another approach, the direct fit of the ligand
parameters to a selected manifold of d-d transitions, was used.
The comparison between the two approaches in deriving the
ligand field parameters will be the subject of Section 4. When going
from the rather ionic CrAF to the more covalent CrACl, CrABr and
CrAI bonds, corrections from dynamical correlation increase con-
siderably. Among the halides, iodine possesses the smallest 10Dq
value, which gradually increases as we move towards lighter
halide ions (largest for F). The spectroscopically reported trend
in 10Dq [14] across the series F, Cl, Br, I, NH3 and CN is nicely
reproduced (Fig. 7). Taken together, both the CASSCF and NEVPT2
calculations provide the correct trend of 10Dq parameters for the
set of chosen ligands.
On the other hand, the computed interelectronic repulsion
parameter B shows a consistent decrease as one moves from CrAF
Fig. 8. Variation of the nephelauxetic reduction of B (in %) with the nature of the
to CrAI bonds. Here we have plotted the percentage reduction in coordinating ligand X for the series of [CrX6]3 complexes (AILFT/CASSCF).
the B parameter for all the complexes taking their free ions as ref-
erence (Fig. 8). It follows from Fig. 8 that, among the halides, the
reduction of B is smallest for CrAF (6%) and largest for the CrAI in the spectrochemical series. The observed trend is in agreement
(14%). This is consistent with the expected increase in the metal- with the nature of the CrAX bonds, where hard-soft interactions
ligand covalency. However, the largest reduction in the B parame- (CrAI and CrACN bonds) show relatively large reductions when
ter has been observed for the [Cr(CN)6]3 complex, which is compared to hard-hard interactions (CrAF and CrANH3), which
expected since the CN ligand is known to be the strongest ligand are also the more ionic in nature. Incorporation of dynamical
electron correlation leads to a further reduction of B. The observed

trend in the B parameter follows the trend of covalency. However,
it is important to note here that the reduction in the B parameter
upon incorporation of dynamic correlations is strictly two-
electron in nature. Thus, it should not be considered as directly
related to metal-ligand covalency. A more thorough discussion of
this important subject will be given elsewhere.
The computed nephelauxetic reduction of the effective spin-
orbit coupling parameter f is plotted in Fig. 9. For halides, we have
observed a very similar trend as seen for the B parameter; com-
puted values of Df (fMIII fMX6) are 13, 32, 68 and 132 for F,
Cl, Br and I respectively. The trend shows that the relativistic
nephelauxetic reduction of the spin-orbit coupling parameter 1 is
monotonically increasing along the series F, NH3, Cl, CN, Br,
and I, and is therefore consistent with the hard-soft concept.
Interestingly, the CN ligand shows a relatively small reduction
in the relativistic nephelauxetic parameter. Fig. 10. The two-dimensional spectro-chemical series reflecting the r and p-donor
At this point, we applied the AOM parameterization scheme to character of involved ligands quantified using the angular overlap model applied to
extract more detailed information about the nature of metal-ligand the CrIIIX6 complex series (X = I, Br, Cl F, NH3 and CN); values of the
parameters er and ep have been obtained from AILFT/CASSCF and AILFT/NEVPT2
interactions. The resulting er and ep parameters are depicted in
calculations employing trigonally compressed (D3d) model complexes with a polar
Fig. 10 (see Table 7 for their numerical values). Not surprisingly, angle h(C3-Cr-X) of 60°.
the AOM computed trend of 10Dq (not included in Table 7) is very
similar to what we have extracted directly from the AILFT calcula-
tions. We have plotted er and ep in a two-dimensional plot p-bonds. All of these trends are nicely consistent with the general
(Fig. 10), which offers a clear picture of the strength of nature of expectations that originate from decades of experience in studying
metal-ligand donor character. The CN ligand possesses the largest transition metal complexes [52].
er parameter in the two-dimensional plot. On the other hand, F The second noticeable feature is the incorporation of the
shows the largest ep parameter while NH3 has a value of zero for dynamic correlation that leads to an increase of both er and ep
ep. The presence of a positive ep parameter for all the halides is a parameters in all cases except CN, where changes in ep are negli-
typical characteristic for p-donor ligands. However, instead of gible. The erCASSCF values for the halides are computed to be
the expected negative ep value for CN, we compute a very small 6929 cm1, 4833 cm1, 3833 cm1 and 3452 cm1 for F, Cl, Br
positive ep. As will be shown by a more detailed analysis (vide and I respectively, which decrease with increasing metal-ligand
infra), this small ep value results from the cancellation of two dif- distances. The effect of dynamical correlation on er is reflected in
ferent contributions (a positive contribution due to the bonding p- the difference (erNEVPT2 erCASSCF); this difference is computed to
MO of CN and a negative one due to the antibonding p⁄ MO). increase from 90 cm1 (1.3%) to 227 cm1 (4.5%), 347 cm1 (8.3%)
Among the halides the most remarkable trend is reflected in the and 474 cm1 (12.3%) for F, Cl, Br and I respectively. Interest-
er and ep parameters, which decrease sharply with an increase in ingly, these differences increase in the given order in spite of the
the metal-ligand bond distances. In spite of the ligand orbitals fact that the metal-ligand distances also increase. A similar trend
becoming more diffuse for heavier halides, the large metal-ligand has been observed for the ep parameter, where computed (epNEVPT2
distances dominate the trends, thus leading to a sharp decrease – epCASSCF) values are 75 cm1 (3%), 54 cm1 (6%), 61 cm1(12%),
in the er and ep parameters across the series F, Cl, Br, and I. and 34 cm1 (15%) for F, Cl, Br and I respectively. However,
The drop in the values of the parameters er and ep along this series as follows from the listed results, the impact of the dynamic corre-
is more acute for ep compared to er. This is a distinct feature of lation is more pronounced for p - compared to r-interactions.
In summary, the AILFT computed two-dimensional plot shows
the correct ordering of er and ep parameters and offers a correct
quantitative picture of the ligand field paradigm. Remarkably,
NEVPT2 computed values of er and ep compare well with data
extracted from the interpretation of solution spectra of tetragonal
complexes of Cr(III) [78–80] (cf. Table 10).
To interpret the LF parameters, we analyzed the first and
second-order contributions to the eg-t2g splitting D; the first order
one (Del) is purely electrostatic in nature and related to (but not
identical with) what crystal field theory pretends to do. Electro-
static perturbations are due to repulsive interactions between
one transition metal d-electron and the electronic clouds on the
ligands. They can be formally described by two-electron repulsion
integrals and are therefore destabilizing. Another smaller contribu-
tion to Del is due to d-electron-ligand nuclear attraction, a term
which is stabilizing but overcompensated by the repulsive one.
The second order contribution to the 3d-orbital energy splitting
is due to metal-ligand bonding. For ligands acting as r- and p-
donors this one-electron energy can be related with the antibond-
Fig. 9. Variation of the nephelauxetic reduction of the effective spin-orbit coupling ing destabilizations of the eg and t2g metal based 3d-orbitals (the
parameter f (in %) with the nature of the coordinating ligand X for the series of covalent shifts). It can be roughly represented by second order per-
CrIIIX36 complexes (AILFT/CASSCF); the numbers at the bottom refer to the value of
turbation by two terms, one for r and another for p of the form
the reduction of the free ion value, f(Cr3+) = 284 cm1 in the given complex, Df = f
(Cr )-f(CrIIIX6).
III
hdðMÞjhjpðLÞi2 =ðeðdðMÞÞ eðpðLÞÞ. This allows one to split D into
ues of 10Dq. However, they are useful in analyzing the electrostatic

and covalency contributions to D. These results nicely demonstrate
the decrease in the percentage of electrostatic contributions (Del)
to D from 63% in the case of F to 42, 40, and 34 for Cl, Br and
I, respectively. The results also show the decrease of Dp across
the series accompanied by a similar (except for Cl) decrease in
Dr. When applied to Cr(CN)3 6 , the same analysis shows that the
splitting Dr dominates D, where contributions from Dp and Dp
are much smaller. As they are of opposite sign, they almost cancel
each other. Therefore, at least at the CASSCF level of theory, the
large and negative Dp (implying p-back bonding) claimed for Cr
(CN)3
6 (see [18] and references therein) is not supported by our
calculations. In our opinion, this is sensible since the negatively
charge CN will not easily accept additional charge. Hence, the
notation of CN as a p- acceptor is not supported. Rather, is a very
strong r-donor. Interestingly, a similar result is obtained for Fe

(CN)3
6 , for which Dp dominates over Dp, a result that is consistent
with DFT based ligand field analysis of the same complex [81].
3.3.3. AILFT and molecular magnetism

The magnetic moment of octahedral d3 complexes is isotropic
Fig. 11. Orbital interactions and the splitting of D into an electrostatic Del term and with zero first-order orbital angular moment in the 4A2 ground
covalent contributions described by Dr, Dp and D*p for M(CN)3 6 . state. In Oh symmetry the orbital angular momentum operators
^Lx , ^
Ly , and ^Lz transform under the irreducible representation T1.
Del and Dr, Dp and D⁄p according to Eq. (37). This decomposition of Thus, only excited states of T2 symmetry can mix with the ground
D is visualized for Cr(CN)3 state (A2 T1 = T2). This mixing is due to spin-orbit coupling with
6 in Fig. 11 (for NH3 the Dp and Dp terms
the 4T2 and 2T2 excited states, which couple with the S = 3/2
vanish). These terms
ground state spin. The energy splitting is controlled by the effective
D ¼ Del þ Dr Dp Dp ð37Þ spin-orbit coupling parameter f. The spin-orbit coupling interac-
tion between the 4A2 ground and the 4T2 and 2T2 excited states
have been estimated using canonical MOs resulting from a con- induces non-vanishing orbital angular momenta in the 4A2 ground
verged CASSCF/AILFT calculation (see Section 6.3. for a detailed state. Uniaxial distortions with three- or fourfold axes lead to ani-
description of the procedure). A more detailed analysis of the
parameters D, Del , Dr , Dp , and Dp in terms of one- and two-
electron integrals between localized 3d and ligand orbitals will be
the subject of a separate report.
In Tables 8 and 9 we list D and its components for CrX3
6 (X = F,
Cl, Br and I) and Cr(CN)3
6 , respectively. We should note that values
of D resulting from analysis of the HF-matrix built on a converged
CASSCF wavefunction differ from those reflected by the AILFT val-
Table 8
Ligand-field splittings D of the 3d-MOs given by the CASSCF Fock-Matrix h/i jF^/j in
the basis of the symmetry adapted (using the D2h subgroup of Oh) metal 3d and ligand
orbitals i, j and its decomposition into an electrostatic splitting terms Del and r and p

covalent shift terms Dr and Dp for complexes of CrIIIX3 III
6 (Cr : X = F ,Cl , Br , I , all
energies in cm1).a
Complex D Del (Del =DÞ% Dr Dp

CrF3
6 24,754 15,600(63) 14,614 5,460
CrCl3
6 21,090 8,959(42) 16,213 4,082
CrBr3
6 18,726 7,525(40) 13,972 2,771
3
CrI6 16,742 5,657(34) 12,992 1,907 Fig. 12. Trigonally elongated octahedral complexes; the extend of distortion is
quantified by the polar angle h < 54.735°; trigonal compressions imply distortions
a
D ¼ Del þ Dr Dp . in directions opposite to the one shown in the figure (h > 54.735°); the point group
symmetry for both signs of distortions is D3d.
Table 9
Ligand-field splittings D of the 3d-MOs given by the CASSCF Fock-Matrix h/i jF^/j in the basis of the symmetry adapted (using the D2h subgroup of Oh) metal 3d and ligand
orbitals i,j and its decomposition into an electrostatic splitting terms Del and r and p covalent shift for r-donor, Dr , p-donor, Dp and p-acceptor, Dp contributions for Cr(CN)3
6 and
1 a
Fe(CN)3
6 (all energies in cm ).
Complex D Del (Del =DÞ% Dr Dp Dp

Cr(CN)3
6 34,920 9,448(27) 25,313 1,859 2,016
Fe(CN)3
6 27,166 11,392(42) 13,076 701 2,698
a
D ¼ Del þ Dr Dp Dp .
sotropic magnetic moments and to splitting of 4A2, given by DMs2 Table 10

(Ms = ±3/2, Ms = ±1/2). Due to this splitting, the two magnetic sub- Ab-initio ligand field orbitals energies, electronic transitions and spin-orbit coupling
parameters and magnetic parameters (in cm1) for D3d trigonally distorted CrCl3 6 and
levels (Kramers doublets) Ms = ±3/2 and Ms = ±1/2 become sepa- WCl3
6 model complexes from NEVPT2 calculations.
rated by a gap of 2D, with (for D < 0) the Ms = ±3/2 Kramers pair
below the Ms = ±1/2 one. This creates anisotropic Ms = ±3/2 mag- CrCl3
6 WCl3
6
netic moments, which can eventually lead to magnetic bistability. h(°) 49.74 59.74 49.74 59.74
The ultimate goal of the field of molecular magnetism is to maxi- D3d(Oh) D3d D3d D3d D3d
e(t2) 0 400 0 867
mize a negative D, ideally creating a gap 2D comparable or even a1(t2) 546 0 540 0
larger than room-temperature kT. In order to rationally design e(e) 11,029 11,108 21,058 21,420
complexes with such a large D, knowledge about the strength of 4
A2 ? 4A1(4T2) 14,208 13,058 22,704 22,120
4
the metal-ligand bond and the dependence of D on the complex A2 ? 4E(4T2) 13,916 14,058 22,348 22,606
2
A2 ? 2A1(2T2) 23,099 26,000 16,092 16,919
geometry is crucial. 2
A2 ? 2E(2T2) 24,554 23,533 16,702 16,270
In this area, AILFT has turned out to be an extremely useful tool f 247 247 2,439 2,439
[53–59]. To illustrate this point, we consider the trigonally elon- D[4A1(4T2)] 1.64 2.05 101.81 131.68
gated(compressed) octahedral CrCl3 6 and WCl3 6 complexes D[4E(4T2)] 1.96 1.75 124.20 109.22
(Fig. 12), taking deviations of the polar angle dh = ±5° from the D[2A1(2T2)] 2.06 0.70 206.19 97.10
D[2E(2T2)] 1.59 1.89 124.94 180.13
octahedral value h = 54.75°. Such distortions may be due to ligand
Dtotal 0.77 0.80 100.44 102.47
packing or to ligand bridging functions (Fig. 13). Energies of the 3d- Dgz = gz-ge 0.038 0.047 0.258 0.318
MOs deduced from AILFT show a nice correlation with the sign of Dgxy = gxy-ge 0.045 0.040 0.302 0.272
D: trigonal compression, implying a splitting of the octahedral t2
ge = 2.0023 – the free electron g-value.
orbital with e < a1, correlates with a negative D, and vice versa. Sec-
ond order perturbation theory allows this correlation to be trans-
lated into a relationship between D and the ligand field splitting
diagram, resulting in an expression which relates D with the order orders of magnitude larger (Table 10). This is a consequence of
of the 4A1(4T2), 4E(4T2), 2A1(2T2) and 2E(2T2) excited states from one the very large spin-orbit coupling parameter f(W) (see Table 10).
side and the spin-orbit coupling parameter f from the other: In contrast to CrCl3
6 , the smaller values of B(W) and C(W), make
the lowest 2T2 excited state dominate the sign and magnitude of
4 2 1 1 D for WCl3
6 .
Dð4 T 2 Þ ¼ f ð38Þ
9 Eð4 A2 ! 4 A1 Þ Eð4 A2 ! 4 EÞ
4. Comparison of AILFT with experiment

1 1 1
Dð2 T 2 Þ ¼ f2 2 ð39Þ
3 Eð A2 ! 2 A1 Þ Eð2 A2 ! 2 EÞ
In this section, we wish to point out a somewhat subtle but
Contributions to D from these excited states (Table 10) show important feature of AILFT if comparison to experiment is the pri-
that 4 T 2 and 2 T 2 are equally important in determining both its sign mary goal of the investigation. The point is the following: by con-
and magnitude; both states introduce contributions to D of the struction, the ligand field parameters obtained from AILFT are the
same sign. Due to the small value of f(Cr), D is rather small best possible compromise that guarantees the closest possible
(±0.8 cm1) for CrCl3 6 (Table 10). Contrary to this, equally large match between ligand field theory and ab initio theory for all
trigonal distortions for WCl3 6 lead to values of D that are two ligand field states. However, in experimental investigations, one
Fig. 13. Structural diversity of CrIIIX6 complexes in various solids: (a) Cr-X-Cr bridging CrBr3 [84], and CrCl3 [85], CrF3 [86] (amidst) and K3CrF6 [87] (right); (b) Cr-X terminal
from left to right K2NaCrF6 [88], [Cr(NH3)6 (ClO4)3 [89], K3Cr(CN)6 [90], K3MoCl6 [91].
rarely has access to all d-d transition energies. Hence, the empirical (i) Considering the series CrX3 6 (X = Br, Cl, F, NH3, CN),
fits that determine the values of the ligand field parameters are one observes a dramatic improvement of the energy of
biased towards the subset of states that are actually observed (or the first spin-allowed d-d transition, defined in ligand field
computed). If all states would be observed, rather than a subset theory as 10Dq, upon inclusion of dynamical correlation
of them, different values for the ligand field parameters would when going from CASSCF to NEVPT2. In the case of the
result. complexes with larger Cr-X covalency, CrBr3 6 and CrCl3
6 ,
Please note that this is not a critique of the experimental extrac- NEVPT2 corrects the energy of the second transition
4
tion procedure. It is simply a part of experimental reality that not A2 ? 4T1(1) in the correct direction, i.e. to an increase,
all states can always be observed. Hence, if one compares AILFT but corrections are overestimated; NEVPT2 computed tran-
computed ligand field parameters to experimentally derived ones, sition energies overestimate experimental values by about
care must be taken not to over interpret the results or be disap- 2000–2500 cm1. In the case of the more ionic CrF3 6 , the
pointed by a lack of apparent agreement. In other words: it is pos- value of 10Dq is recovered by including dynamical correla-
sible for the calculated transition energies to match the experiment tion. However, the energy of the second transition remains
well while at the same time the agreement of the calculated and underestimated by the same amount, with 2500 cm1 for
experimentally derived ligand field parameters is poor. It is not 4
A2 ? 4T1(1) and even more, 3600 cm1, for 4A2 ? 4T1(2).
implied that one or the other set of data is ‘better’ than the other. Transition energies for Cr(NH3)3+ 6 at the NEVPT2 level of
In such situations they are simply different since they have been theory are excellent, while, due to higher covalency in Cr
derived in different ways. If only a subset of states is observed, (CN)3
6 , d-d transitions given by NEVPT2 are overestimated.
the consistent point of comparison would be to fit the ligand field Surprisingly, due to error compensation, transition energies
parameters to only the same subset from the ab initio calculations. given by CASSCF for the latter complex compare better
However, this is not what AILFT is about. A fit to a subset would with experiment.
take away the uniqueness of the AILFT construction. We will illus-
trate this point with some examples below. Spectroscopic data on complexes for MoIIIX6 are rather scarce
-only MoCl3 6 has been reported to display two transitions
identified as d-d (Tables 11 and 12); NEVPT2 energies for the
two transitions 4A2 ? 4T2 (r1) and 4A2 ? 4T1(1)(r2) are in excel-
4.1. Transitions from experiment and from CASSCF and NEVPT2 lent agreement with reported band maxima at 19,000 and
ab initio calculations 24,000 cm1, respectively. To the best of our knowledge, spectro-
scopic data for WIIIX3
6 appear to be entirely absent from the
Ligand field spectra of Cr3+(d3) octahedral complexes have been literature.
studied in powdered solids, and only rarely in single crystals. These
solids can be subdivided into crystal lattices, in which CrIIIX6 com- 4.2. Ligand field parameters
plex units share one or two halide X = F, Cl, Br ligands (Fig. 13a) and
solids in which CrIIIX6 complex units are not directly connected to In the following, we will focus on the values of B and 10Dq for
each other, but are separated by alkaline ions (Fig. 13b). It is there- octahedral complexes. Ligand field parameters are usually
fore expected that d-d absorption spectra of a complex in these extracted from observed (and properly assigned) spectroscopic
two different types of solids will differ, reflecting a direct influence transitions. Clearly, there is more than one way to do this extrac-
on the donor properties of a given ligand connected to a Cr(III) cen- tion. Perhaps the most unbiased approach is a least-squares fit of
ter due to interactions with its other Cr(III) neighbor(s). In Table 11 the ligand field parameters to the observed transition energies.
we present literature values for spin-allowed d-d transitions, Such a fit is nonlinear given that the transition energies depend
where for CrCl3 6 for example, Cl appears as terminal in Cs2- on the ligand field parameters in a nonlinear fashion, at least in
NaCrCl6 (of the same structural type as K2NaCrF6 depicted in cases where there is more than one term of a given symmetry (in
Fig. 13b, left) but, say Cl in CrCl3 acts as a bridging ligand(Fig. 13a, the present example, the transitions to two excited states of the
left). In spite of these differences in the structure, the d-d transi- same symmetry 4T1(1) and 4T1(2)). The least- squares fit will then
tions appear at nearly the same energy in both cases. Presumably, depend on whether two or all three spin-allowed ligand field tran-
due to the very localized nature of the d-d excitations, neither next sitions have been observed in the experiment (see values of 10Dq
nearest neighbors, nor the lattice topology affects d-d transitions to and B from experimental d-d transitions reported in Table 11).
a large extent. This is in agreement with the largely ionic nature of Alternatively, analytical expressions can be used to extract the
the metal-ligand bonds in the considered solids, which makes the ligand field parameters. In the case of d3 complexes, one can use
ligand field less dependent on the second coordination sphere Eqs. (39) and (40) to obtain values for B and 10Dq. Again, it will
effects (but see exceptions discussed and analyzed in Ref. [82]). depend on the availability of data for two (r1 (4A2 ? 4T2) and r2
This independence on second nearest neighbors is also strongly (4A2 ? 4T1(1), Eqs. (39) and (40a) or all three (r1, r2 and
supported by our cluster calculations using both electrostatic r3(4A2 ? 4T1(2), Eqs. (39) and (40b) spin-allowed transitions
embedding and/or a polarizable continuum model (COSMO [83]). which values for 10 Dq and B are obtained.
Both sets of calculations show only a minor effect on the electronic
structure of CrX3
10Dq ¼ Eð4 A2 ! 4 T2 Þ ð40Þ
6 on the adopted charge compensating model.
Importantly, in the spectra of CrCl3 6 , out of the three spin-
ð2r1 r2 Þðr2 r1 Þ
allowed transitions for CrCl3 4 4
6 , only two, i.e. A2 ? T2 (r1) and B¼ ð40aÞ
4
A2 ? T1(1) (r2) could be observed, the third, A2 ? 4T1(2) (r3)
4 4 27r1 15r2
being masked by ligand-to-metal charge transfer transitions; only
1
in three cases for CrF36 (Table 11) could all three spin-allowed d-d B¼ ðr2 þ r3 3r1 Þ ð40bÞ
transitions be resolved. 15
The near invariance of the d-d transitions with changing lattices We will numerically illustrate this dependency for two mem-
allows one to compare the experimental and computed CASSCF bers of our test set, the fairly ionic CrF3
6 and the more covalent
and NEVPT2 d-d transitions in Table 12. CrI3
6 (Table 13). As input data, we will use the calculated CASSCF
From this comparison, we can draw the following conclusions: transition energies from Table 12. In the actual situation, one
Table 11
Experimental Structural (R in Å) and Spectroscopic Data (in cm1) on octahedral CrIII complexes in various solids.
Complex Compound MIII site symm. R(M-X) Energies of Electronic Transitions 10Dq B B/Bor
4 4 4 4 4 4
A 2 ? T2 A2 ? T1(1) A2 ? T1(2)
CrBr3
6 CrBr3 D3 2.524a 13,400i 17,400i – 13,400 373 0.39
CrBr3
6 Cr3+:Cs2NaYBr6 Oh – 12,400j 17,700j – 12,400 543 0.57
3
CrCl6 CrCl3 C3 2.347b 13,700i 19,200i – 13,700 551 0.58
3
CrCl6 Cr3+:Cs2NaYCl6 Oh – 12,800k 18,200k 28,400 shk 13,700 550 0.58
CrCl3
6 K2NaCrCl6 Oh – 13,600l 19,350l – 13,650 600 0.63
CrCl3
6 Tl2NaCrCl6 Oh – 13,000l 18,150l – 13,000 530 0.56
CrCl3
6 Rb2NaCrCl6 Oh – 13,100l 18,520l – 13,100 570 0.60
CrCl3
6 Cs2NaCrCl6 Oh – 12,650l 17,820l – 12,650 540 0.57
CrCl3
6 Cs2K1/2Na1/2CrCl6 Oh – 12,850l 18,170l – 12,850 560 0.59
CrCl3
6 Cs2KCrCl6 Oh – 13,050l 18,470l – 13,050 570 0.60
CrCl3
6 Cs2Rb1/2K1/2Cl6 Oh – 13,000l 18,280l – 13,000 550 0.58
CrCl3
6 Cs2TlCrCl6 Oh – 12,800l 18,000l – 12,800 540 0.57
CrF3
6 CrF3 C3 1.901c 14,600i 21,500i – 14,600 741 0.78
3
CrF6 K3CrF6 Oh 1.975d 15,200m 21,800shm 35,000m 15,200 795 0.83
CrF3
6 K2NaCrF6 Oh 1.933e 16,100i 23,350i – 16,100 760 0.80
CrF3
6 Cr3+:K2Na2GaF6 Oh – 16,200n – – 16,200 – –
CrF3
6 Cs2KCrF6 Oh – 16,000l 23,000l – 16,000 724 0.76
CrF3
6 K2LiCrF6 Oh – 15,700l 22,800l 34,800l 15,700 746 0.78
CrF3
6 Cs2LiCrF6 C3 – 16,100l 23,050l – 16,100 715 0.75
CrF3
6 Cs2NaCrF6 C3 1.913e 15,650l 22,700l 34,450 15,650 739 0.78
Cr(NH3)3+ 6 [Cr(NH3)6](ClO4)3/H2O Oh 2.074f 21,550o 28,490o – 21,550 657 0.69
Cr(CN)36 K3Cr(CN)6/solid Oh 2.078g 26,500p 32,400p – 26,500 530 0.56
3
Cr(CN)6 K3Cr(CN)6/H2O – – 26,600p 32,500p – 26,600 529 0.56
MoCl36 K3MoCl6 Ci 2.445h – – – – – –
MoCl36 K3MoCl6/3nHCl,H2O Solution – – 19,000q 24,000q – 19,000 458 0.72
a
[84].
b
Average CrACl distance, [85].
c
[86].
d
[87].
e
[88].
f
[89].
g
[90].
h
[91].
i
[92].
j
[93].
k
[94].
l
[95].
m
[96].
n
[97].
o
[98,99].
p
[100].
q
[101].
r
Free ion B values: Bo(Cr3+) = 953 cm1; Bo(Mo3+) = 633 cm1.
would need experimental data; however for the sake of argument results from AILFT and from the direct fits are almost identi-
it is immaterial what data set is used here. cal, while for CrI3
6 the complete Fit (Fit(3)) is considerably
There are a few interesting observations from Table 13: closer to the CASSCF energies.
(d) Fits or extractions that only make use of the lowest two
(a) The use of Eqs. (39) and (40a) and the direct fit to the lowest energies of d-d transitions achieve exact results for the first
two d-d transitions lead to identical results. two transitions at the expense of the value for the third tran-
(b) CASSCF is very compatible with ligand field theory. This is sition. For the very ionic CrF3
6 , which is very well described
shown by the very small RMS deviations for any of the fits. by ligand field theory, this is acceptable, while for the more
However, the agreement between ligand field theory and covalent CrI36 , the position of the third band is about
CASSCF description is much better for the more ionic fluo- 2000 cm1 higher in energy, hence the overall RMS value
ride compound compared to the more covalent iodide com- is the highest. The underlying reason is that interelectronic
plex. Clearly, ligand field theory cannot describe effects like repulsion in the third excited state is described by a smaller
the anisotropy in the electron-electron repulsion (anisotro- B – a direct consequence of metal-ligand covalency that is
pic or symmetry restricted covalence) that is present in larger for the eg(r) than the t2g(p) orbitals.
more elaborate ab initio treatments (including CASSCF) (e) There is a substantial difference of more than 150 cm1 in
without introducing more parameters. the extracted B values, depending on whether two or three
(c) In principle, AILFT is mathematically, by construction, the energies are used in the extraction. Translated to the situa-
best approximation to the entire ligand field matrix. This tion where the source of data is not CASSCF but experiment,
means that the ligand field matrix elements are determined this means that the ‘‘experimental” values for 10Dq and B
such that each and every single matrix element of the CI are inflicted with significant uncertainties and it is impor-
matrix is reproduced as closely as possible. Clearly, this does tant to be aware of how exactly these numbers were
not guarantee that the eigenvalues of this matrix also show obtained if one wishes to study trends among a series of
the closest match. For example, for the more ionic CrF3 6 complexes.
Table 12
Comparison between experimental and directly computed CASSCF and NEVPT2 energies (in cm1) of the spin-allowed d-d transitions of octahedral MIIIX6)d3 complexes.
Complex d-d Transition energies (cm1)

4
A2 ? 4T2 4
A2 ? 4T1(1) 4
A2 ? 4T1(2)
CrBr3
6 Exp. 12400 a
17700 a
–
CASSCF 9,128 15,301 27,751
NEVPT2 12,341 19,504 28,748
CrCl3
6 Exp. 13,000b 18,280b –
CASSCF 10,724 17,594 30,193
NEVPT2 13,479 20,943 30,936
MoCl3
6 Exp. 19,200c 24,000c –
CASSCF 17,917 24,676 39,225
NEVPT2 19,712 24,930 40,858
CrF3
6 Exp. 16,000b 23,000b 34,600b
CASSCF 11,272 18,511 31,714
NEVPT2 12,962 20,457 31,086
Cr(NH3)3
6 Exp. 21,500d 28,500d –
CASSCF 16,753 25,632 40,370
NEVPT2 20,168 28,581 43,555
Cr(CN)3
6 Exp. 26,500e 32,400e –
CASSCF 24,222 33,835 53,670
NEVPT2 29,319 37,467 60,721
a
[93].
b
[95].
c
[101].
d
[98].
e
[100].
Table 13
Ligand field parameters extracted from ‘‘experimental” data in different ways (Transition energies from CASSCF and NEVPT2 from Table 12 were used). The results depend on
whether analytical expressions are used or least squares fits are performed and whether two (Fit(2)) or three transition (Fit(3)) energies are used in the extraction.a,b
CrF3
6 CASSCF (39) + (40a) Fit(2) (39) + (40b) Fit(3) AILFT
10Dq – 11,272 11,272 11,272 11,271 11,299
B – 1,094 1,094 1,093 1,094 1,091
r1 11,272 11,272 11,272 11,272 11,272 11,299
r2 18,511 18,511 18,511 18,511 18,511 18,543
r3 31,714 31,714 31,714 31,714 31,714 31,719
RMS – 0 0 0 0 25
CrI3
6 10Dq – 8,633 8,633 8,633 8,703 9,200
B – 1,168 1,168 1,094 1,026 986
r1 8,633 8,633 8,633 8,633 8,703 9,200
r2 14,639 14,639 14,639 14,562 14,588 15,277
r3 26,902 28,780 28,780 27,746 26,911 27,113
RMS – 1,150 1,150 519 54 507
CrF3
6 NEVPT2 (39) + (40a) Fit(2) (39) + (40b) Fit(3) AILFT
10Dq – 12,962 12,962 12,962 13,133 11,227
B – 950 950 844 795 1,007
r1 12,962 12,962 12,962 12,962 13,133 11,227
r2 20,457 20,456 20,456 20,042 20,033 18,255
r3 31,086 32,680 32,680 31,504 31,291 30,531
RMS – 976 976 361 307 1,651
CrI3
6 10Dq – 12,797 12,797 12,797 12,872 10,375
B – 928 928 726 685 944
r1 12,797 12,797 12,797 12,797 12,872 10,375
r2 20,161 20,161 20,161 19,276 19,141 16,901
r3 29,128 32,151 32,151 30,005 29,750 28,384
RMS – 1,851 1,851 763 733 2,383
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
hP i
a
Root mean square errors (RMS) in this work are defined as: RMS ¼ oT 2
i ðri ri Þ =N (N = 9).
b
Least square fits to two or three d-d transitions were done as detailed in Section 6.4.
It is instructive to perform the same comparison for the case of (a) Variations of ligand field parameters from NEVPT2 are more
NEVPT2 transition energies. The latter include more intricate pronounced than for the CASSCF case. This is closer to the
effects of dynamical correlation and hence include more effects situation one would face by using experimental data in the
beyond the ligand field model than CASSCF. NEVPT2 is also more extraction procedure.
realistic, since it yields results for transition energies that are closer (b) Obviously, the experimental data include all effects that go
to the experimental values than CASSCF (see Table 12). beyond ligand field theory. It is therefore not surprising that
There are interesting differences between the CASSCF and ligand field theory is a worse approximation per se for the
NEVPT2 results (Table 13): more complete NEVPT2 data compared to the CASSCF data.
Hence, none of the extraction procedures are in perfect In summary, AILFT can very well be used to study chemical
agreement with the input data. trends in ligand field parameters across series of related
(c) The results of the direct fits to the energies (Fit(3)) produce compounds and it provides (in conjunction with CASSCF/NEVPT2)
results closest to the input data. The same comment applies semi-quantitative to quantitative agreement with experimentally
as for Table 13(CASSCF) – the AILFT procedure provides the derived values. However, deviations between experimentally
closest fit to the entire ligand field matrix, not necessarily extracted 10Dq and B values and AILFT are inevitable. They do
to the eigenvalues. In addition, the way the AILFT procedure not reflect a shortcoming or failure of the method itself, but is sim-
is setup by using the spectral resolution (see Section 2.4, Eq. ply a feature of its basic construction. Experiment and theory are
(20)) yields results that do not guarantee the minimal error best compared at the level of the transition energies, e.g. the actual
mathematically. Improvements are currently under investi- experimental observables. The same comments apply, of course, to
gation, which are based on intermediate effective Hamiltoni- the ligand field parameters extracted from subsequent procedures
ans [102], very much in the spirit of the original AILFT like the angular overlap parameters.
formulation [52].
5. Conclusions and outlook

The bottom line of this comparison is the following: It is obvi-
ous from the results that 10Dq and B cannot be uniquely deter-
In this work, we have presented the ab initio ligand field theory
mined from the experimental data. In particular, the value of B
(AILFT) as a practical procedure that allows one to translate results
can vary by as much as almost 25% depending on which method
from correlated multireference ab initio electronic structure calcu-
is used for the extraction. Clearly, using Eq. (39) is designed to
lations to the chemically meaningful language of ligand field the-
obtain the position of the first band exactly. The fitted value for B
ory. Ligand field theory is justified if: (i) dn configurations are
must then compensate for any possible shortcoming of the ligand
spectroscopically well defined, implying metal-ligand bonds dom-
field approach that identifies the energy of the 4A2 ? 4T2 transition
inated by ionic interactions; (ii) ligand-to-metal charge transfer
with the eg-t2g energy difference. If only two energies are observed,
states do not strongly interfere with dn based multiplets, an
the two parameters 10Dq and B can obviously be adjusted to fit
assumption which is certainly justified for the ground and lowest
these two bands perfectly. However, being biased, the third transi-
ligand field excited states of a myriad of complexes. However, it
tion will be predicted significantly less accurately. In the case of
might be violated for higher lying dn states in the proximity of
two available transition energies, the least square fit and the ana-
ligand-to-metal charge transfer states with which they mix.
lytical expression lead to identical results.
We have shown by detailed test calculations on several series of
The situation is different if three transition energies are avail-
d3 complexes that the parameters obtained by AILFT follow the
able. Here, the two parameters must describe three transition
established chemical trends. We have stressed that the parameters
energies and the least square fit is not equivalent to the analytical
extracted from AILFT are unique. These are the ligand field one-
expression. By fixing the value of 10Dq to the first transition, one
electron parameters [entering into the symmetric (5 5 or 7 7
effectively introduces a constraint. This means that the use of the
analytical expression does not guarantee the overall best fit to for d- and f-complexes, respectively) matrix VLFT ], the Racah
the original data. Hence, an unconstrained least- square fit (Fit parameters B and C, as well as the spin-orbit coupling constant f.
(3)) in Table 13 is preferable. We regard it as an asset of AILFT in that it provides an unambigu-
This non-uniqueness of the values of 10Dq and B has obvious ous bridge between ab initio electronic structure theory and ligand
consequences for the comparison of AILFT extracted ligand field field theory. However, in order to extract chemically intuitive
parameters with experimentally extracted parameters. Strictly information, the matrix VLFT must be related to a model. This can
speaking, these values cannot be compared, since they are based be an electrostatic model, the angular overlap model, or even
on different premises. Ligand field parameters extracted from something different. It is obvious that the further decomposition
experiment are not uniquely defined, while AILFT derived parame- of VLFT is not unique, but depends on the chosen model. The latter
ters are. Again, this is not a critique of the experimental extraction. is manifestly subjective and requires the insight and creativity of
It is simply a factual statement of the differences between the the scientist that carries out these studies.
experimental and AILFT methodologies. Our computational model of choice has been the CASSCF
It should be carefully appreciated that there are two different method followed by a second order correction for the dynamical
sources of discrepancy between the AILFT ligand field parameters correlation energy (NEVPT2). It has been demonstrated that the
and the experimentally extracted one: (1) the non-uniqueness results of NEVPT2 calculations are noticeably better compared to
issues discussed at length above and, (2) the shortcomings in the experiment than the CASSCF calculations. This, of course, is the
theory itself. CASSCF/NEVPT2 is a reasonable and computationally expected result since dynamic correlation effects are important
convenient methodology, but it is not high-accuracy quantum throughout chemistry. However, dynamic correlation also brings
chemistry. Thus, shortcomings of the calculated wavefunctions in a number of effects that are outside the ligand field model, such
and energies will lead to deviations of the 10Dq and B values from as a pronounced anisotropy in the electron-electron repulsion. This
their ‘ideal’ theoretical values that one would obtain from exact means that quite typically AILFT extractions from CASSCF calcula-
solutions of the Schrödinger equation (full CI calculations in an tions are more successful than those based on NEVPT2. By ‘success-
infinite one-particle basis set). Such calculations are presently ful’ we mean that the root-mean-square deviation between
not feasible, but it is foreseeable that more accurate methods than ab initio calculated excitation energies and ligand field calculated
CASSCF/NEVPT2 will become available in the not too distant future. excitation energies (using the AILFT extracted parameters), are
They will lead to different 10Dq and B-values. It is safe to predict noticeably smaller for CASSCF than for NEVPT2 (where deviations
that these B-values will be smaller than the NEVPT2 ones since can be as large as 4000 cm1). The differences become particularly
NEVPT2 misses a fraction of dynamic correlation energy. Since pronounced if states of different nature are studied that contain
the B-values reflect electron-electron repulsion, they will be very different dynamic correlation contributions, for example
reduced by properly taking care of the electron-electron cusp. This high-spin versus low-spin states. This is not a limitation of NEVPT2
will improve the agreement with the experimentally derived val- or AILFT but reflects inherent limitations of the ligand field model.
ues, but source (1) of the disagreement will forever be present While CASSCF/NEVPT2 is a computationally convenient theory, it is
and prevent perfect agreement. by no means the only possible choice for the AILFT extraction. In
forthcoming papers, we will present an improved scheme based on correlation potential, and once obtained cannot be further sys-
the theory of intermediate effective Hamiltonians [102]. tematically improved. AILFT results in principle can by going
There are several possible ways to compare the results of beyond CASSCF/NEVPT2.
AILFT calculations to experiment: (a) one can compare calculated
and experimental transition energies, which is unambiguous (to 6. Computational details
the extent that the experimental data is), or (b) one can compare
AILFT extracted ligand field parameters to experimentally All calculations were carried out with the ORCA suite of elec-
extracted ligand field parameters. In Section 4, we have dis- tronic structure programs [46]. In the first step, we have performed
cussed at length the possible pitfalls of the second type of com- DFT geometry optimization in order to generate the closest struc-
parison, namely that the extracted ligand field parameters ture to experiment. Here we have employed pure BP86 functional
depend on the subset of states that were included in their deter- [106,107] for the structure optimization. We have opted for the
mination. In AILFT, we always fit the ligand field parameters to all electron def2-TZVP basis sets [108–110] for the non-metals
all states (by construction). In the experiment, all states are and for the all electron DKH3-TZP-Sapporo basis set [111–115]
rarely observed and hence one fits only to a subset of states. for the first/s/third row transition metal ions. Since most of the
The obtained ligand field parameters for fitting subsets of states model complexes are either tri- or di- anionic in nature, we have
are necessarily different from those obtained by fitting all states. employed the Conductor-like Screening Model (COSMO) to com-
It is simply outside the capability of ligand field theory to fit all pletely screen the charge density of the complexes [83]. Dispersion
d-d states with uniform accuracy. Hence, one should not be sur- corrections have been taken care of by incorporating Grimme’s
prised to find differences in AILFT computed and experimentally atom-pairwise dispersion correction approach [116,117] as imple-
extracted values for 10Dq and B. Alternatively, one can fit the mented in ORCA. Scalar relativistic effects are treated using
same subset of states that were used in the experiment from, second-order Douglas-Kroll-Hess (DKH) method [118]. The
say, the NEVPT2 calculations. However, this would no longer obtained structures are in good agreement with experiments. In
be AILFT and would not be unique. The strength of AILFT is its order to compute properties, we have relied on the state-average
uniqueness which allows one to study trends of ligand field complete active space self-consistent field (SA-CASSCF) method
parameters also in the absence of experiments to allow with all electron basis sets as mentioned above. For the d3 ions,
predictions. the active space is comprised of CAS(3,5) (three active electrons
In this work, we have focused on dn configurations. However, in the five active orbitals). AILFT parameters have been computed
we have recently demonstrated [103], that the same concepts also including only the ten non-relativistic S = 3/2 states(roots). How-
apply to fn configurations in lanthanides and actinides equally suc- ever, when discussing magnetic properties, the entire manifold of
cessfully. An alternative ab initio ligand field model on the basis of the 10 S = 3/2 and the 40 S = 1/2 have been taken into the consider-
Stevens operators has recently been proposed and applied with ation. The molecular orbitals have been optimized in an average
success to complexes of lanthanides [104,105]. The model relies way for all the states belonging to the given multiplicity. To capture
on the assumption that the relativistic ground state of the free the effect of the dynamic correlation, we employed second order N-
ion J is well separated from excited states with which it may inter- electron valence perturbation theory to second order (NEVPT2) on
act though the ligand field or spin-orbit coupling one-electron top of the converged CASSCF wavefunction. In order to speed up
operators. While this is justified for the complexes of most lan- the calculations, the resolution of identity (RI) approximation
thanides, it is not applicable for 3d, 4d, and 5d elements, as well [119] has been used with the corresponding auxiliary basis sets.
for complexes of the actinide series for which the ligand field is The spin-orbit effects along with the Zeeman interactions are incor-
comparable but can become even larger than the intrinsic spin- porated by quasi-degenerate perturbation theory (QDPT) approach,
orbit coupling constant. For such cases, the AILFT of the present where spin-orbit mean field (SOMF) operator [72,73] accounts for
review is equally applicable and yields, in addition to the parame- mixing of state with different multiplicities (DS = 0, ±1).
ters of the ligand field (the 5 5 or 7 7 matrices), parameters of
inter-electronic repulsion and spin orbit coupling. 6.1. Cluster embedding
Extensions of the scope of AILFT to complexes with larger
metal-ligand covalency includes consideration of inter electronic In order to treat the effect of the lattice and surroundings on the
repulsion in combination with point group symmetry. This leads computed properties, we have employed the embedded cluster
to the notion of symmetry-restricted covalency. In this considera- approach [120,121]. In this approach, we divide the molecules into
tion, more symmetry independent parameters, rather than one B three different zones as shown in Fig. 14. The first zone is the quan-
and one C need to be explicitly taken into account. Within the tum cluster (QC), which is surrounded by the point charges in the
realm of ab initio quantum chemistry this is a rewarding goal, still outer region, referred as to the point charge (PC) region. Between
to be accomplished. Finally, the extension of AILFT to more than the PC and QC regions, we introduce a boundary region to prevent
one shell (for example a d- and a s-shell or an f- and d-shell) or any charge flow. The boundary region is comprised of the repulsive
to dinuclear transition metal complexes is certainly rewarding capping effective core potentials (c-ECPs). This approach is very
and is under active investigation in our laboratory. popular and effective in mimicking solid-state effects; however
Finally, we would like to comment on the differences the molecules which we have subjected to study are discrete. Here,
between ligand field parameters resulting from AILFT and we have chosen the {ML6}n as the QC regions and the surrounding
LFDFT. Compared to experimentally deduced values of the counter ions in the first layer are as the boundary region, while rest
parameters of interelectronic repulsion B and C, these are over- has been considered by point charges. To fulfill the charges in the
estimated(underestimated) by AILFT(LFDFT). The origin of the PC and boundary regions, we have chosen charge neutrality condi-
overestimation of B and C resulting from AILFT has already been tions, where qQC = (qBR + qPC). To satisfy the principle of charge
discussed before (vide supra). The reason for the underestima- neutrality, several different combination of charges are possible,
tion of these parameters by LFDFT is possibly the self- which apparently offer different electrostatic potentials to the QC
interaction errors inherent in any standard Kohn-Sham DFT cal- region. Here, we have considered the charges on ions to be as close
culation. The last point of this comparison is that results from as possible to its charge population according to Breneman and
LFDFT are strongly depending on the adopted exchange- Wiberg [122].
Fig. 14. 3 3 3 super cell packing diagram for complex [MoCl6]3. Color Code. Dark green (Mo); light green (Cl); pink (P). Molecule shown in the inset with different color
(yellow (Mo) and light blue (Cl)) represent the QC (for visualization). Schematic representation of the PC, BR, and QC zones corresponding to the embedded cluster approach,
adapted from [120].
6.2. Geometry optimization r-MOs :

" # " #
cbM;eg caM;eg ebeg 0 ð42Þ
For the purpose of our study, we have focused our attention on Cr ¼ ; Kp ¼
the hexa-coordinated mononuclear complexes of 3d, 4d and 5d cbL;eg caL;eg 0 eaeg
ions possessing three unpaired electrons within the d-manifold
In the orbital eigenvectors Cp and Cr the elements c have been
(S = 3/2). To shed light on the nature of ligand field parameters,
defined as positive definite and have been identified with the
we have limited ourselves to the general molecular formula of
square roots of the Löwdin populations of the symmetry adapted
[ML6]n+/ (where M = Cr(III), Mn(IV), Mo(III), W(III)); while L = F,
metal and ligand based linear combinations. To this end we have
Cl, Br, I, NH3 and CN (n = 2,3)). Crystal structures and molecular
used the D2 Abelian sub-group of the Oh group in which the t2g
geometries of different qualities for many of these complexes are
orbitals yield b1(dxy), b2(dxz) and b3(dyz) irreducible representa-
reported in the literature. The optimized geometries at the DFT
tions, while eg gives rise to a1 (dz2,dx2-y2) and corresponding linear
level of theory are in the good agreement with experiment. Due
combinations of metal (M) and ligand (L) based orbitals. The col-
to negatively charged species, we have carried out optimizations
umns in Cp and Cr are normalized but the two columns are not
using the COSMO approach [83], which completely shields the
orthogonal. We use Löwdin orthogonalization to get an orthonor-
charge density of the complexes. In comparison to the regular
mal set of orbitals Up and Ur:
gas-phase optimized structures, the charge compensated COSMO
optimized structural parameters are closer to experiment. The Sp ¼ C0p Cp ; Sr ¼ C0r Cr ð43Þ
metal-ligand bond lengths become shorter by 0.02–0.05 Å when
we employ the solvent corrections, values which are also much yielding
closer to experiment (similar trend observed for all complexes). Up ¼ Cp S1=2 ; Ur ¼ Cr S1=2 ð44Þ
p r
Thus, for all AILFT calculations we have opted for the solvent cor-
rected geometries. Apart from the bond length, we have not from which we get the Fock-matrices:
observed any remarkable changes in the other structural param-
Hp;M Hp;ML
eters such as \LAMAL bond angle. DFT(COSMO) optimized Hp ¼ Up Kp U0p ¼ ð45Þ
metal-ligand bond distances (in Å) for the octahedral complexes Hp;ML Hp;L
of this review are: CrIIIAF 1.997; CrIIIACl 2.385; CrIIIABr 2.589,
CrIIIAI 2.882, CrIIIANH3 2.138, CrIIIACN 2.061, MoIIIACl 2.476; Hr;M Hr;ML
Hr ¼ Ur Kr U0r ¼ ð46Þ
WIIIACl 2.471 (cf. with X-ray diffraction reported values: R(Cr- Hr;ML Hr;L
X) = 1.933–1.975, 2.33–2.347, 2.47–2.52, for X = F, Cl and Br,
Using Eqs. (45) and (46) we deduce Del from the difference
respectively and R(MoACl) = 2.445, all expressed in Å (see
Hr;M Hp;M and the covalent shifts Dr and Dp from eaeg Hr;M and
Table 11).
eat2g Hp;M , respectively.
6.3. Deducing first (electrostatic) and second (covalent) order
contributions to the octahedral eg-t2g orbital splitting D 6.4. Ligand field computations and extraction of ligand field
parameters from a direct fit to energies of d-d transitions
Molecular orbitals from a converged CASSCF calculation of p
and r type are given by bonding (b) and antibonding (a) linear Ligand field calculations and extraction of parameters from
combinations and corresponding energies as follows: reported or theoretically (CASSCF/NEVPT2) computed energies of
d-d transition have been done with the aid of the angular overlap
p-MOs :
" # " # program package AOMX [123–125]. In these calculations, correct
cbM;t2g caM;t2g ebt2g 0 ð41Þ assignment of electronic states to the irreducible representation
Cp ¼ ; Kp ¼
cbL;t2g caL;t2g 0 eat2g of octahedral point group was assisted by the implemented algo-
rithm for treating space group symmetries of both Abelian and
non-abelian point groups on equal footing. Parameter optimiza- [29] M. Gerloch, J.H. Harding, R. Guy Woolley, The context and application of
ligand field theory, Struct. Bond. 46 (1981) 1–46 (Berlin).
tions were carried out using the Powell parallel subspace algorithm
[30] C.A. Daul, Description par orbitals moléculaires des états électroniques dans
[126]. les complexes métalliques, Thèse d’agrégation présentée à la Faculté des
Sciences de l’université de Fribourg (Suisse) por l’obtention de la venia
legendi, Fribourg, 1981.
Acknowledgement [31] J. Weber, C.A. Daul, A new method for the description of ligand field states
based on the multiple scattering Xa results: application to CrCl4, Mol. Phys.
The authors owe tanks to Casey van Stappen, Max-Planck 39 (1980) 1001–1011.
[32] J.C. Slater, Statistical exchange-correlation in the self-consistent field, Adv.
Institute for Chemical Energy Conversion, Mülheim an der Ruhr,
Quant. Chem. 6 (1972) 1.
Germany, for critical reading of the manuscript and valuable [33] R.P. Messmer, D.R. Salahub, Molecular orbital study of the ground and excited
comments. states of ozone, J. Chem. Phys. 65 (1976) 779.
[34] T. Ziegler, A. Rauk, E.J. Baerends, On the calculation of multiplet energies by
the Hartree-Fock-Slater method, Theor. Chim. Acta 43 (1977) 261.
References [35] C.A. Daul, Density functional theory applied to the excited states of
coordination compounds, Int. J. Quant. Chem. 52 (1994) 867–877.
[1] C.J. Ballhausen, Introduction to Ligand Field Theory, McGraw-Hill Book [36] C. Anthon, C.E. Schäffer, Toward understanding nephelauxetism:
Company, Inc., New York, 1962. interelectronic repulsion in gaseous dq ions computed by Kohn-Sham DFT,
[2] B.N. Figgis, Introduction to Ligand Fields, Interscience Publishers, A Devision Coord. Chem. Rev. 226 (2002) 17–38.
of John Wiley & Sons, New York, 1967. [37] C. Anton, J. Bendix, C.E. Schäffer, An average-of-configuration method for
[3] H.L. Schläfer, G. Gliemann, Introduction in Ligand Field Theory, Akademische using Kohn-Sham density functional theory in modeling ligand-field theory,
Vrelagsgesellschaft, Frankfurt am Main, 1967 (in German language). Inorg. Chem. 42 (2003) 4088–4097.
[4] C.J. Ballhausen, Molecular Electronic Structures of Transition Metal [38] C. Anton, J. Bendix, C.E. Schäffer, Mimicking the two-dimensional
Complexes, McGraw-Hill International Book Company, New York, 1979. spectrochemical series using density functional computations, Inorg. Chem.
[5] B.N. Figgis, M.A. Hitchman, Ligand Field Theory and Its Applications, Wiley- 43 (2004) 7882–7886.
VCH, New York, 2000. [39] M. Atanasov, C.A. Daul, C. Rauzy, New insights into the effects of covalency on
[6] C.J. Ballhausen, Quantum mechanics and chemical bonding in inorganic the ligand field parameters: a DFT study, Chem. Phys. Lett. 367 (2003) 737–
complexes. I. Static concepts of bonding; dynamic concepts of valency, J. 746.
Chem. Ed. 56 (1979) 215–218. [40] M. Atanasov, C.A. Daul, C. Rauzy, A DFT based ligand field theory, Struct. Bond.
[7] C.J. Ballhausen, Quantum mechanics and chemical bonding in inorganic 106 (2004) 97–125.
complexes. II. Valency and inorganic metal complexes, J. Chem. Ed. 56 (1979) [41] C. Anthon, J. Bendix, C.E. Schäffer, Elucidation of ligand-field theory.
294–297. Reformulation and revival by density functional theory, Struct. Bond. 107
[8] C.J. Ballhausen, Quantum mechanics and chemical bonding in inorganic (2004) 207–301.
complexes. III. The spread of ideas, J. Chem. Ed. 56 (1979) 357–361. [42] C.E. Schäffer, C. Anthon, J. Bendix, Kohn-Sham DFT results projected on
[9] C.E. Moore, Atomic Energy Levels, vol. I (H to V), 1949, vol. II (Cr to Nb), 1952, ligand-field models: using DFT to supplement ligand-field descriptions and to
vol. III (Mo through La and Hf through Ac) 1958, Circulars of the National supply ligand-field parameters, Coord. Chem. Rev. 253 (2009) 575–593.
Bureau of Standards. 467, US Government Printing Office. [43] R.J. Deeth, Ligand field and density functional descriptions of the d-states
[10] NIST Atomic Spectra Database, <http://physics.nist.gov/PhysRefData/ASD/ and bonding in transition metal complexes, Faraday Discuss. 124 (2003)
levels_form.html>. 379–391.
[11] A. Kramida, Yu. Ralchenko, J. Reader, Atomic Spectral Transitions for the free [44] S.E. Harnung, E. Larsen, Circular dichroism of tris chelates with planar
Cr3+, Mo3+ and W3+ Ions: NIST ASD Team, NIST Atomic Spectra Database (ver. conjugate ligands described by angular overlap model and calculated by time
5.3.), 2015. <http://physics.nist.gov/asd>. dependent density functional theory, Coord. Chem. Rev. 307 (2016) 81–103.
[12] E.U. Condon, G.H. Shortley, The Theory of Atomic Spectra, Cambridge Univ. [45] F. Aquilante, J. Autschbach, R.K. Carlson, L.F. Chibotaru, M.G. Delcey, L. De
Press, 1935. Vico, I. Fdez. Galván, N. Ferré, L.M. Frutos, L. Gagliardi, M. Garavelli, A.
[13] H. Bethe, Splitting of terms in crystals, Ann. Phys. 3 (1929) 133–206. Giussani, C.E. Hoyer, G. Li Manni, H. Lischka, D. Ma, P.Å. Malmqvist, T. Müller,
[14] C.K. Jørgensen, Recent progress in ligand field theory, in: Struct. and Bond., A. Nenov, M. Olivucci, T.B. Pedersen, D. Peng, F. Plasser, B. Pritchard, M.
vol. 1, 1966, pp. 3–31 (and original references cited therein). Reiher, I. Rivalta, I. Schapiro, J. Segarra-Martí, M. Stenrup, D.G. Truhlar, L.
[15] C.K. Jørgensen, Absorption Spectra and Chemical Bonding in Complexes, Ungur, A. Valentini, S. Vancoillie, V. Veryazov, V.P. Vysotskiy, O. Weingart, F.
Oxford, 1962 (and original references cited therein). Zapata, R. Lindh, MOLCAS 8: new capabilities for multiconfigurational
[16] J.H. Van Vleck, Valence strength and the magnetism of complexes, J. Chem. quantum chemical calculations across the periodic table, J. Comput. Chem.
Phys. 3 (1935) 807–813. 37 (2016) 506–541.
[17] J.H. van Vleck, The theory of electric and magnetic susceptibilities, Oxford at [46] F. Neese, Comput. Mol. Sci., The ORCA program system 2 (2012) 73–78;
the Clarendon Press, 1932. ORCA- An ab initio, DFT and semiempirical SCF-MO package - Version 4.0,
[18] A.B.P. Lever, Inorganic Electronic Spectroscopy, Elsevier, Amsterdam, 1984. Design and Scientific Directorship: F. Neese, Technical Directorship: F.
[19] J.H.E. Griffiths, J. Owen, Complex hyperfine structures in microwave spectra Wennmohs, Max-Planck-Institute for Chemical Energy Conversion Stiftstr.
of covalent iridium compounds, Proc. Roy. Soc. Chem. Ser. A Math. Phys. Sci. 34–36, 45470 Mülheim a.d. Ruhr, Germany, tccec@mpi-mail.mpg.de, With
226 (1954) 96–111. contributions from: D. Aravena, M. Atanasov, U. Becker, D. Bykov, D. Datta, A.
[20] A. Abragam, B. Bleaney, Electron Paramagnetic Resonance of Transition Kumar Dutta, D. Ganyushin, Y. Guo, A. Hansen, L. Huntington, R. Izsak, C.
Metals, Clarendon Press, Oxford, 1970. Kollmar, S. Kossmann, M. Krupicka, D. Lenk, D. G. Liakos, D. Manganas, D. A.
[21] R. Finkelstein, J.H. Van Vleck, On the energy levels of chrome alum, J. Chem. Pantazis, T. Petrenko, P. Pinski, C. Reimann, M. Retegan, C. Riplinger, T.
Phys. 8 (1940) 790–797. Risthaus, M. Roemelt, M. Saitow, B. Sandhöfer, I. Schapiro, K. Sivalingam, G.
[22] F.E. Ilse, H. Hartmann, Term systems of electrostatic complexions of Stoychev, B. Wezisla; And contributions from our collaborators:M. Kallay, S.
transition metals with one d-electron, Z. Phys. Chem. 197 (1951) 239–246 Grimme, E. Valeev, G. Chan, J. Pittner; Additional contributions to the manual
(in German language). from: G. Bistoni, W. Schneider.
[23] F.E. Ilse, H. Hartmann, The term system of a Ion with two d-electrons in [47] P.-Å. Malmqvist, B.O. Roos, The CASSCF state interaction method, Chem. Phys.
octahedral field, Z. Naturforschg 6a (1951) 751–754 (in German language). Lett. 155 (1989) 189–194.
[24] W.G. Penney, R. Schlapp, The influence of crystalline fields on the [48] C. Angeli, R. Cimiraglia, S. Evangelisti, T. Leininger, J.-P. Malrieu, Introduction
susceptibilities of salts of paramagnetic ions. I. The rare earths, especially of n-electron valence states for multireference perturbation theory, J. Chem.
Pr and Nd, Phys. Rev. 41 (1932) 194; Phys. 114 (2001) 10252–10264.
W.G. Penney, R. Schlapp, Influence of crystalline fields on the susceptibilities [49] C. Angeli, R. Cimiraglia, J.-P. Malrieu, N-electron valence state perturbation
of salts of paramagnetic ions. II. The iron group, especially Ni, Cr and Co, Phys. theory: a fast implementation of the strongly contracted variant, Chem. Phys.
Rev. 42 (1932) 666. Lett. 350 (2001) 297–305.
[25] C.J. Ballhausen, Approximate methods for the electronic structure of [50] C. Angeli, R. Cimiraglia, J.-P. Malrieu, N-electron valence state perturbation
inorganic complexes, in: G.A. Segal (Ed.), Semiempirical Electronic Structure theory: a spinless formulation and an efficient implementation of the
Calculations, Part B: Applications, Plenum Press, New York, 1977, pp. 129– strongly contracted and of the partially contracted variants, J. Chem. Phys.
162. 117 (2002) 9138–9153.
[26] H. Hartmann, H.-J. Schmidt, On the absorption spectra of complexes of the [51] C. Angeli, B. Bories, A. Cavallini, R. Cimiraglia, Third-order multireference
three-valent molibdenium, Z. Phys. Chem. (NF) 11 (1957) 234–250 (in perturbation theory: the n-electron valence state perturbation-theory
German language), Table 1. approach, J. Chem. Phys. 124 (2006) 054108.
[27] L.E. Orgel, An Introduction to Transition-Metal Chemistry, Ligand Field [52] M. Atanasov, D. Ganyushin, K. Sivalingam, F. Neese, A modern first-principles
Theory, Methuen & Co Ltd.; John Wiley & Sons Inc, London; New York, 1966. view on ligand field theory through the eyes of correlated multireference
[28] C.E. Schäffer, C.K. Jørgensen, The nephelauxetic series of ligands wavefunctions, in: D.M.P. Mingos, P. Day, J.P. Dahl (Eds.), Molecular
corresponding to increasing tendency of partly covalent bonding, J. Inorg. Electronic Structures of Transition Metal Complexes II, Springer, Berlin,
Nucl. Chem. 8 (1958) 143–148. Heidelberg, 2012, pp. 149–220.
[53] M. Atanasov, D. Ganyushin, D.A. Pantazis, K. Sivalingam, F. Neese, A detailed [81] M. Atanasov, P. Comba, C.A. Daul, DFT-based studies on the Jahn-Teller effect
ab initio first-principles study of the magnetic anisotropy in a family of in 3d haxacyanometalates with orbitally degenerate ground states, J. Phys.
trigonal pyramidal iron(II) pyrrolide complexes, Inorg. Chem. 50 (2011) Chem. A 111 (2007) 9145–9163.
7460–7477. [82] D. Reinen, M. Atanasov, S.-L. Lee, Second-sphere ligand field effects on oxygen
[54] J.M. Zadrozny, M. Atanasov, A.M. Bryan, C. Lin, B.D. Rekken, P.P. Power, F. ligator atoms and experimental evidence – the transition metal-oxygen bond
Neese, J.R. Long, Slow magnetization dynamics in a series of two-coordinate in oxidic solids, Coord. Chem. Rev. 175 (1998) 91–158.
iron(II) complexes, Chem. Sci. 4 (2013) 125–138. [83] A. Klamt, G. Schüürmann, COSMO: a new approach to dielectric screening in
[55] M. Atanasov, J.M. Zadrozny, J.R. Long, F. Neese, A theoretical analysis of solvents with explicit expressions for the screening energy and its gradients,
chemical bonding, vibronic coupling, and magnetic anisotropy in linear iron J. Chem. Soc., Perkin Trans. 2 (1993) 799–805.
(II) complexes with single-molecule magnet behavior, Chem. Sci. 4 (2013) [84] H. Braekken, Kongelige Norske Videnskabers Selskab, Die Kristallstruktur von
139–154. Chromtribromid 5 (1932), 42–42.
[56] J.M. Zadrozny, D.J. Xiao, M. Atanasov, G.J. Long, F. Granjean, F. Neese, J.R. Long, [85] B. Morosin, A. Narath, X-ray diffraction and nuclear quadrupole resonance
Magnetic blocking in a linear iron(I) complex, Nat. Chem. 5 (2013) 577–581. studies of chromium trichloride, J. Chem. Phys. 40 (1964) 1958–1967.
[57] M. Atanasov, D. Aravena, E. Suturina, E. Bill, D. Maganas, F. Neese, First [86] J.E. Jørgensen, W.G. Marshall, R.I. Smith, The compression mechanism of CrF3,
principles approach to the electronic structure, magnetic anisotropy and spin Acta Cryst. B 60 (2004) 669–673.
relaxation in mononuclear 3d-transition metal single molecule magnets, [87] Y. Xiao, Y. Su, H.-F. Li, C.M.N. Kumar, R. Mittal, J. Persson, A. Senyshyn, K.
Coord. Chem. Rev. 289–290 (2015) 177–214. Gross, Th. Brueckel, Neutron diffraction investigation of the crystal and
[58] E.A. Suturina, D. Maganas, E. Bill, M. Atanasov, F. Neese, Magneto-structural magnetic structures in KCrF3 perovskite, Phys. Rev. B 82 (2010) 094437-1–
correlations in a series of pseudotetrahedral [CoII(XR)4]2 single molecule 094437-5.
magnets: an ab initio ligand field study, Inorg. Chem. 54 (2015) 9948–9961. [88] W. Massa, D. Babel, M. Epple, W. Rüdorff, Sind Elpasolithe fehlgeordnet? -
[59] Y. Rechkemmer, F.D. Breitgoff, M. van der Meer, M. Atanasov, M. Hakl, M. Strukturbestimmungen an Einkristallen K2NaCrF6, Rb2NaFeF6 und Rb2KFeF6,
Orlita, P. Neugebauer, F. Neese, B. Sarkar, J. van Slageren, A four-coordinate Rev. Chim. Miner. 23 (1986) 508–519.
cobalt(II) single-ion magnet with coercivity and a very high-energy barrier, [89] E. Mikuli, N. Górska, S. Wróbel, J. Sciesinski, E. Sciesinska, Phase
Nat. Commun. 7 (2016) 10467, http://dx.doi.org/10.1038/ncomms10467. polymorphism, molecular motions and structural changes in [Cr(NH3)6]
[60] F. Neese, T. Petrenko, D. Ganyushin, G. Olbrich, Advanced aspects of ab initio (ClO4)3, Z. Naturforsch. A 62 (2007) 179–186.
theoretical optical spectroscopy of transition metal complexes: multiplets, [90] S. Jagner, E. Ljungström, Nils-Gösta Vannerberg, The Crystal Structure of
spin-orbit coupling and resonance Raman intensities, Coord. Chem. Rev. 251 Potassium Hexacyanochromate(III), K3[Cr(CN)6], Acta Chem. Scand. A 28
(3–4) (2007) 288–327. (1974) 623–630.
[61] Spin Eigenfunctions. Construction and Use. Hrsg. von R. Pauncz, Plenum [91] Z. Amilius, B. van Laar, H.M. Rietveld, The crystal structure of K3MoCl6, Acta
Press, New York-London, 1979. Cryst. B 25 (1969) 400–402.
[62] F. Neese, Prediction of molecular properties and molecular spectroscopy with [92] D.L. Wood, J. Ferguson, K. Knox, J.F. Dillon Jr., Crystal-Field Spectra of d3,7 Ions,
density functional theory: from fundamental theory to exchange-coupling, III. Spectrum of Cr3+ in Various Octahedral Crystal Fields, J. Chem. Phys. 39
Coord. Chem. Rev. 253 (5–6) (2009) 526–563. (1963) 890–898.
[63] K. Wolinski, P. Pulay, Generalized Moeller-Plesset perturbation theory: [93] F. Gilardoni, J. Weber, K. Bellafrough, C.A. Daul, H.U. Guedel, Excited State
second order results for two-configuration, open-shell excited singlet, and Properties of Cr3+ in Cs2NaYCl6 and Cs2NaYBr6, J. Chem. Phys. 104 (1996)
doublet wave functions, J. Chem. Phys. 90 (1989) 3647–3659. 7624–7632.
[64] K. Andersson, P.A. Malmqvist, B.O. Roos, A.J. Sadlej, K. Wolinski, Second-order [94] R.W. Schwartz, Absorption and magnetic circular dichroism spectra of
perturbation theory with a CASSCF reference function, J. Phys. Chem. 94 chromium(III) in dicesium sodium yttrium hexachloride, Inorg. Chem. 15
(1990) 5483–5488. (1976) 2817–2822.
[65] K. Andersson, P.Å. Malmqvist, B.O. Roos, Second-order perturbation theory [95] D. Reinen, M. Atanasov, P. Köhler, D. Babel, Jahn-Teller coupling and the
with a complete active space self-consistent field reference function, J. Chem. influence of strain in Tg and Eg ground and excited states – a ligand field and
Phys. 96 (1992) 1218. DFT study on Halide MIIIX6 model complexes [M = TiIII-CuIII, X = F, Cl],
[66] B.O. Roos, P.-A. Malmqvist, Relativistic quantum chemistry: the Coord. Chem. Rev. 254 (2010) 2703–2754.
multiconfigurational approach, Phys. Chem. Chem. Phys. 6 (2004) 2919–2927. [96] G.C. Allen, G.A.M. El-Sharkawy, K.D. Warren, Electronic spectra of the
[67] B. Roos, M. Fülscher, P.-Å. Malmqvist, M. Merchán, L. Serrano-Andrés, hexafluorometalate (III) complexes of the first transition series, Inorg.
Theoretical studies of the electronic spectra of organic molecules, in: S. Chem. 10 (1971) 2538–2546.
Langhoff (Ed.), Quantum Mechanical Electronic Structure Calculations with [97] J. Ferguson, H.J. Guggenheim, D.L. Wood, Crystal field spectra of d3,7 Ions, VII
Chemical Accuracy, Springer, Netherlands, 1995, pp. 357–438. Cr3+ in K2NaGaF6, J. Chem. Phys. 54 (1971) 504–507.
[68] Y. Guo, K. Sivalingam, E.F. Valeev, F. Neese, Sparse Maps-a systematic [98] M. Linhard, On the light absorption and constitution of inorganic
infrastructure for reduced-scaling electronic structure methods. III. Linear- complexsalts, Z. Elektrochem. Angew. Phys. Chem. 50 (1944) 224–228 (in
scaling multireference domain-based pair natural orbital N-electron valence German language).
perturbation theory, J. Chem. Phys. 144 (2016) 094111. [99] H.L. Schläfer, On the Photochemistry of Complexes of the Transition Metals,
[69] F. Menezes, D. Kats, H.-J. Werner, Local complete active space second-order Zeitschrift für Physikalische Chemie (Neue Folge), 11 (1957) 65–77 (in
perturbation theory using pair natural orbitals (PNO-CASPT2), J. Chem. Phys. German language); the Spectrum of [Cr(NH3)6]Cl3 H2O in acidic water
145 (2016) 124115. solution was reported as re-measured in this reference with, compared to
[70] K.G. Dyall, The choice of a zeroth-order Hamiltonian for second-order Ref.[98] slightly different energies of the two spin allowed d-d transitions
perturbation theory with a complete active space self-consistent-field (21500 and 28500 cm1), including a shoulder at about 15300 cm1 which
reference function, J. Chem. Phys. 102 (1995) 4909. was assigned to an inter combination d-d transition.
[71] J. Miralles, O. Castell, R. Caballol, J.P. Malrieu, Specific CI calculation of energy [100] J.J. Alexander, H.B. Gray, Electronic structures of hexacyanometalate
differences: transition energies and bond energies, Chem. Phys. 172 (1993) complexes, J. Am. Chem. Soc. 90 (1968) 4260–4271.
33. [101] H. Hartmann, H.-J. Schmidt, On the absorption spectra of complexes of three-
[72] B.A. Hess, C.M. Marian, U. Wahlgren, O. Gropen, A mean-field spin-orbit valent molibdenium, Z. Phys. Chem. (Neue Folge) 11 (1957) 234–250 (in
method applicable to correlated wavefunctions, Chem. Phys. Lett. 251 (1996) German language).
365. [102] J.P. Malrieu, P.h. Durand, J.P. Daudey, Intermediate Hamiltonians as a new
[73] F. Neese, Efficient and accurate approximations to the molecular spin-orbit class of effective Hamiltonians, J. Phys. A: Math. Gen. 18 (1985) 809–826.
coupling operator and their use in molecular g-tensor calculations, J. Chem. [103] D. Aravena, M. Atanasov, F. Neese, Periodic trends in lanthanide compounds
Phys. 122 (2005) 034107. through the eyes of multireference ab initio theory, Inorg. Chem. 55 (2016)
[74] J.S. Griffith, The Theory of Transition-Metal Ions, Cambridge at the University 4457–4469.
Press, 1971. [104] L.F. Chibotaru, L. Ungur, Ab initio calculation of anisotropic magnetic
[75] J. Ferguson, D.L. Wood, Crystal field spectra of d3,7 ions. The weak field properties of complexes. I. Unique definition of pseudospin Hamiltonians
formalism and covalency, Austr. J. Chem. 23 (1970) 861–871. and their derivation, J. Chem. Phys. 137 (2012) 064112.
[76] M. Gerloch, R.C. Slade, Ligand-Field Parameters, Cambridge at the University [105] L.F. Chibotaru, Ab initio methodology for pseudospin hamiltonians of
Press, 1973. anisotropic magnetic complexes, Adv. Chem. Phys. 153 (2013) 397–519;
[77] T. Schönherr, M. Atanasov, H. Adamsky, Angular overlap model, Compr. L. Ungur, L.F. Chibotaru, Computational modelling of magnetic properties of
Coordin. Chem. II 2 (2004) 443–455 (A.B.P. Lever (Ed.)). lanthanide compounds, in: R.A. Layfield, M. Murugesu (Eds.), Lanthanides and
[78] T.J. Barton, R.C. Slade, Chemical significance of ligand field parameters in Actinides in Molecular Magnetism, vol. 6, Willey-VCH, 2015, pp. 153–184.
chromium (III) complexes of quadrate symmetry, J. Chem. Soc., Dalton Trans. [106] A.D. Becke, Density-functional exchange-energy approximation with correct
(1975) 650–657. asymptotic behavior, Phys. Rev. A 38 (1988) 3098–3100.
[79] J. Glerup, O. Mønsted, C.E. Schäffer, Non additive and additive ligand fields [107] J.P. Perdew, Density-functional approximation for the correlation energy of
and spectrochemical series arising from ligand field parametrization the inhomogeneous electron gas, Phys. Rev. B 33 (1986) 8822–8824;
schemes. Pyridine as a nonlinearly ligating p-back bonding ligand toward J.P. Perdew, Erratum, Phys. Rev. B 34 (1986) 7406.
chromium(III), Inorg. Chem. 15 (1976) 1399–1407. [108] A. Schäfer, C. Huber, R. Ahlrichs, Fully optimized contracted Gaussian basis
[80] D.W. Smith, Applications of the angular overlap model, Struct. Bond. 35 sets of triple zeta valence quality for atoms Li to Kr, J. Chem. Phys. 100 (1994)
(1978) 87–118. 5829–5835.
[109] F. Weigend, R. Ahlrichs, Balanced basis sets of split valence, triple zeta [120] M. Roemelt, D. Maganas, S. DeBeer, F. Neese, A combined DFT and restricted
valence and quadruple zeta valence quality for H to Rn: design and open-shell configuration interaction method including spin-orbit coupling:
assessment of accuracy, Phys. Chem. Chem. Phys. 7 (2005) 3297–3305. application to transition metal L-edge X-ray absorption spectroscopy, J.
[110] A. Schäfer, H. Horn, R. Ahlrichs, Fully optimized contracted Gaussian basis Chem. Phys. 138 (2013) 204101-1–204101-22.
sets for atoms Li to Kr, J. Chem. Phys. 97 (1992) 2571–2577. [121] D. Maganas, M. Roemelt, M. Havecker, A. Trunschke, A. Knop-Gericke, R.
[111] T. Noro, M. Sekiya, T. Koga, Sapporo-(DKH3)-nZP (n = D, T, Q) sets for the sixth Schlogl, F. Neese, First principles calculations of the structure and V L-edge X-
period s-, d-, and p-block atoms, Theo. Chem. Acc. 132 (2013) 1363–1367. ray absorption spectra of V2O5 using local pair natural orbital coupled cluster
[112] T. Noro, M. Sekiya, T. Koga, Segmented contracted basis sets for atoms H theory and spin–orbit coupled configuration interaction approaches, Phys.
through Xe: sapporo-(DK)-nZP sets (n = D, T, Q), Theor. Chem. Acc. 131 (2012) Chem. Chem. Phys. 15 (2013) 7260–7276.
1124–1131. [122] C.M. Breneman, K.B. Wiberg, Determining atom-centered monopoles from
[113] T. Noro, M. Sekiya, T. Koga, H. Matsuyama, Valence and correlated basis sets molecular electrostatic potentials. The need for high sampling density in
for the first-row transition atoms from Sc to Zn, Theo. Chem. Acc. 104 (2000) formamide conformational analysis, J. Comput. Chem. 11 (1990) 361.
146–152. [123] H. Adamsky, T. Schönherr, M. Atanasov, AOMX: angular overlap model
[114] T. Noro, M. Sekiya, T. Koga, S.L. Saito, Relativistic contracted Gaussian-type computation, Compr. Coordin. Chem. II 2 (2003) 661–664 (Elsevier, Edited by
basis functions for atoms K through Xe, Chem. Phys. Lett. 481 (2009) 229– A.B.P.Lever).
233. [124] P.E. Hoggard, Sharp line electronic transitions and metal-ligand angular
[115] Y. Osanai, M. Sekiya, T. Noro, T. Koga, Valence and correlating basis sets for the geometry, Coord. Chem. Rev. 70 (1986) 85–120.
second transition-metal atoms from Y to Cd, Mol. Phys. 101 (2003) 65–71. [125] AOMX, a Fortran Program That Calculates dn Electron Term Energies in the
[116] S. Grimme, J. Antony, S. Ehrlich, H. Krieg, A consistent and accurate ab initio Framework of the Angular Overlap Model Including Electron Interaction and
parametrization of density functional dispersion correction (DFT-D) for the Spin-Orbit Coupling by H. Adamksy based on the AOM1 Program by P.E.
94 elements H-Pu, J. Chem. Phys. 132 (2010) 154104–154123. Hoggard with Contributions by M.Atanasov and K. Eifert; http://www.aomx.
[117] S. Grimme, S. Ehrlich, L. Goerigk, Effect of the damping function in dispersion de (an executable and source code can be obtained by one of us (MA) on
corrected density functional theory, J. Comput. Chem. 32 (2011) 1456–1465. request).
[118] B.A. Hess, Relativistic electronic-structure calculations employing a two- [126] M.J.D. Powell, An efficient method for finding the minimum of a function of
component no-pair formalism with external-field projection operators, Phys. several variables without calculating derivatives, Comput. J. 7 (1964) 155–
Rev. A 33 (1986) 3742–3748. 162.
[119] F. Neese, An improvement of the resolution of the identity approximation for
the formation of the Coulomb matrix, J. Comput. Chem. 24 (2003) 1740–1747.

10 1016@j CCR 2017 03 018

Загружено:

Сведения о документе

Исходное описание:

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

10 1016@j CCR 2017 03 018

Загружено:

Авторское право:

Доступные форматы

Coordination Chemistry Reviews 344 (2017) 2–25

Contents lists available at ScienceDirect

Coordination Chemistry Reviews

Covalency and chemical bonding in transition metal complexes:

1. Introduction nd orbitals in the complex [21–23]. This led to a static picture of

2.2. N-electron valence and complete active space second-order ðCASÞ

Parameter Method Cr3+ Mo3+ W3+

Terms Cr3+ Mo3+ W3+

3.2.1. AILFT parameters for the [MCl6]3 (where M = Cr(III), Mo(III)

been performed on the DFT optimized geometries (see Section 6.2

3.2.2. Interpretation of the AILFT parameters for the [MCl6]3 (where

group symmetry (symmetry restricted covalency [14]). The latter

3.3.2. Interpretation of the AILFT parameters for the [CrX6]3+/ (X = F,

with experimentally reported ones will be the subject of Section 4.

electron correlation leads to a further reduction of B. The observed

ues of 10Dq. However, they are useful in analyzing the electrostatic

3.3.3. AILFT and molecular magnetism

Complex D Del (Del =DÞ% Dr Dp

Complex D Del (Del =DÞ% Dr Dp Dp

sotropic magnetic moments and to splitting of 4A2, given by DMs2 Table 10

 4. Comparison of AILFT with experiment

Complex d-d Transition energies (cm1)

5. Conclusions and outlook

6.2. Geometry optimization r-MOs :

Вам также может понравиться

Complex D Del (Del =DÞ% Dr Dp Dp

4. Comparison of AILFT with experiment