You are on page 1of 15

This is a pre-peer reviewed version of the following article:

http://onlinelibrary.wiley.com/doi/10.1002/ejic.201201289/full
MICROREVIEW

DOI: 10.1002/ejic.200

Progress in Electronic Structure Calculations on Spin-Crossover Complexes

Hauke Paulsen,*[a] Volker Schünemann[b] and Juliusz A. Wolny[b]

Keywords: Ab initio calculations / Computational chemistry / Density functional calculations / Quantum chemistry / Spin crossover

Spin-crossover (SCO) complexes are an ongoing challenge to The results of these methods serve as reference for other
quantum chemistry due to their delicate balance of electronic and electronic structure calculations that may be applied to larger
entropic contributions to the adiabatic enthalpy difference between systems as well. The methods of choice for the calculation of
the high-spin and the low-spin state. This challenge has fuelled the geometries and molecular vibrations of isolated SCO complexes
improvement of existing quantum chemical methods and the and of crystalline compounds are based on density functional
development of new ones and will go on to do so. The progress in theory (DFT). Recent hybrid functionals are able to calculate the
electronic structure calculations on SCO complexes in the last years adiabatic energy difference with an accuracy that is in some
has made quantum chemical methods valuable tools that may aid cases close to those of ab initio calculations, although no unique
the design of new SCO compounds with desired features. Post- functional has been identified up to now that is superior to other
Hartree-Fock ab initio methods are able to calculate the adiabatic functionals in all cases. DFT methods are applied now to
energy difference between the high-spin and the low-spin state with crystalline systems, too, and allow to investigate inter-molecular
satisfactory accuracy but are currently limited to model systems or effects that are important for an understanding of the
smaller molecular SCO complexes. cooperativity of the spin transition.

____________
combined with configuration interaction were computationally too
[a] Institut für Physik, Universität zu Lübeck, Ratzeburger Allee 160, D-
23562 Lübeck, Germany expensive. In the time that has passed since then computers have
Fax: ++49 451 500 4214 become considerably faster and much progress has been made with
E-mail: paulsen@physik.uni-luebeck.de algorithms. As a result the number of publications containing
Homepage: www.physik.uni-luebeck.de electronic structure calculations for SCO complexes has become by
[b] Fachbereich Physik, Technische Universität Kaiserslautern, Erwin-
Schrödinger-Str. 46, D-67663 Kaiserslautern, Germany
far too large to allow a complete survey of all articles within this
review. Instead the purpose of this article is to present a limited
number of representative studies that illustrate what can be
achieved with quantum chemical methods in the field of SCO
research.
An important - and still ambitious - aim of electronic structure
Introduction calculations is to predict whether a given complex will or will not
exhibit spin crossover. If we focus on the simple and idealized
Spin-crossover (SCO) complexes, which undergo a transition example of an isolated complex that can exist either in a LS or in a
from a low-spin (LS) to a high-spin (HS) state upon change of HS state, we can express the temperature dependent fraction of HS
temperature or pressure or upon irradiation with light or pulsed molecules by
magnetic fields, are fascinating objects of an experimental and
theoretical field of research at the borderline of chemistry and χHS(T) = [1 + exp(ΔGHL/kBT)]-1 (1)
physics (an all-encompassing overview has been edited by Gütlich
and Goodwin [1]). Within this field of research quantum chemistry where ΔGHL=GHS-GLS is the free enthalpy difference between the
is employed in order to interpret experimental data, to test HS and the LS state, T is the absolute temperature and kB is
empirical models and to rationalize the design of new materials. Boltzmann's constant. Since the pressure dependent contribution to
Reversely, SCO complexes are used to test and to improve the difference of the free enthalpy, pΔV, can be neglected in case of
quantum chemical methods. The mere fact of containing one or SCO complexes (usually it contributes less than 1 J mol -1 [3]), we
more transition metal centres indicates that SCO complexes are a can write
challenge for any quantum chemical method [2]. This is even more
the case since a spin crossover can only occur if there is a delicate ΔGHL = ΔEHL - TΔSHL (2)
balance of different contributions to the electronic energy on one
hand and electronic and vibrational contributions to entropy on the where ΔEHL=EHS-ELS and ΔSHL=SHS-SLS denote the HS to LS
other hand. differences of total energy and entropy, respectively. EHS and ELS
When an early review article on quantum chemical calculations contain the potential and the kinetic energy of electrons and nuclei.
of the electronic structure of SCO complexes appeared almost a
decade ago [3], nearly all calculations on real SCO complexes had
been performed with methods employing density functional theory
(DFT), since calculations with the Hartree-Fock (HF) method

Submitted to the European Journal of Inorganic Chemistry 1


Hauke Paulsen was born 1964 in Flensburg, Germany. He graduated in physics at the University of Kiel in 1990 and went
then to Professor Alfred X. Trautwein at the University of Lübeck where he wrote a Ph.D. thesis on electronic structure
calculations. Since a habilitation treatise on optically switchable iron complexes in 2002 he is Privatdozent at the Institute of
Physics of the University of Lübeck. His current research interests are molecular dynamics simulations and electronic
structure calculations.

Volker Schünemann was born in 1963 in Lübeck, Germany. He graduated in physics at the University of Hamburg in 1989
and then joined the group of Professor Alfred X. Trautwein at the University of Lübeck. Working on the preparation and
characterization of nanometer-sized iron particles in zeolites, he completed his Ph.D. in 1993. After a post doc with
Professor W. M. H. Sachtler at Northwestern University, Evanston, USA, he returned to the Trautwein group and changed to
the field of Mössbauer and EPR spectroscopy in molecular biophysics and bioinorganic chemistry. In 2004 he became
professor for biophysics and medical physics at the University of Kaiserslautern. His current research interest is the role of
iron in inorganic chemistry and biology. With respect to spectroscopic methods, he is interested in the application and
further development of conventional and synchrotron-based Mössbauer spectroscopy.

Juliusz A. Wolny graduated in chemistry at the University of Wrocław and completed his Ph.D. Thesis on inorganic
chemistry there. He has worked in the area of synthetic chemistry and molecular spectroscopy with the groups of Prof.
Mikołaj F. Rudolf in Wrocław, Prof. Alex von Zelewsky at the University of Fribourg, Prof. Hans Toftlund at the University
of Odense, and in the area of applied quantum chemistry and synchrotron spectroscopy with the group of Prof. Alfred X.
Trautwein at the University of Lübeck. Since 2006, he is researcher in the group of Prof. Volker Schünemann at the
University of Kaiserslautern. His scientific interest involves the spin-crossover effect, vibrational spectroscopy, conventional
and synchrotron-based Mössbauer spectroscopy, and the application of quantum chemical methods to inorganic metal
complexes.

Following a widespread convention we denote ΔEHL as adiabatic Nevertheless, it must be kept in mind that the entropy difference
energy difference to distinguish it from the so-called vertical on the right side of Eq. (3) depends itself on the temperature, so
energy difference. The former is evaluated by calculating EHS and that an accurate solution can be obtained only iteratively.
ELS for the HS and LS geometry, respectively, while the latter As a typical example for the temperature dependent HS fraction
refers to EHS and ELS calculated both either for the HS or the LS of a SCO complex, data for [Fe(tpa)(NCS) 2] (tpa=tris(2-
geometry. pyridylmethyl)amine) that was obtained by magnetic susceptibility
The kinetic energy of the nuclei, which is present even at zero measurements [5] is presented in Figure 1. This complex exhibits a
temperature, has often been neglected although it might contribute temperature driven abrupt one-step spin transition at about 105 K
a few kJ mol -1 to ΔGHL [4]. Actually ΔEHL depends on the with a hysteresis of almost 10 K. The solid lines in Figure 1 are
temperature, which, however, can usually be neglected without loss simulated curves that have been created with Eqs. (1) and (2) using
of accuracy. The entropy difference ΔSHL is built up out of two two sets of (not too realistic) fit parameters, namely (i) ΔEHL=1.4
components, an electronic one, ΔSel, and a vibrational one, ΔSvib. kJ mol-1, ΔSHL=13.4 J K-1 mol-1 and (ii) ΔEHL=70.6 kJ mol-1,
The electronic component is due to the higher multiplicity of the ΔSHL=665 J K-1 mol-1 for the flat and the steep curves, respectively.
HS state (we do not take into account a possible occupation of In parameter set (i) ΔSHL equals kB ln 5 which is exactly the
excited electronic states) while the vibrational component stems contribution from the multiplicity of the HS state (5T2g in this case)
from the fact that the loosened metal ligand bonds in the HS state and an appropriate value for ΔEHL has been chosen in order to
lead to lower vibrational frequencies and - at elevated temperatures match the experimental transition temperature. Clearly both values
- to an increase of entropy. In general, ΔSHL depends on the are too low leading to a much too soft transition. Better agreement
temperature but fortunately in the region of the spin transition, is achieved with parameter set (ii) resulting in a steep HS fraction
T≈T1/2 and above, this dependency is only weak. The transition curve that closely resembles the experimental curves except for the
temperature T1/2 is defined by the equal occupation of HS and LS hysteresis. However, in set (ii) both ΔEHL and ΔSHL are by far too
states (χHS(T1/2)=1/2). According to Eq. (2) the transition large, as a result of the attempt to match the steepness of the
temperature can be approximated by experimental spin transition.

T1/2 = ΔEHL / ΔSHL (3)

Submitted to the European Journal of Inorganic Chemistry 2


as abruptness, hysteresis or steps can't be understood at all without
regarding inter-molecular effects. The well known model from
Slichter and Drickamer [16] gives a simple phenomenological
description of the spin transition by writing a χHS dependent free
enthalpy in the form

G(χHS,T) = G(0,T) + χHS (ΔEHL – TΔSHL) + Γ χHS(1-χHS) - TSmix (10)

where Smix is given by

Smix = - R [χHS lnχHS + (1-χHS) ln(1-χHS)] (11)

and Γ is a phenomenological interaction parameter.


Minimization of G(χHS,T) with respect to χHS gives the HS-fraction
as a function of temperature.
The quantum chemistry methods used for calculating the
electronic structure of SCO complexes can be divided into two
groups: post-HF ab initio and DFT calculations. A general
introduction to electronic structure calculations is far beyond the
scope of this article and good introductions to ab initio methods
and DFT can be found in textbooks like those from Cramer [17]
Figure 1. Temperature dependent HS-fraction of [Fe(tpa)(NCS) 2] (taken and from Jensen [18]. Introductions to DFT calculations can be
from [5]). ▲ and ▼ stand for measurements with in- and decreasing found in the books from Parr, from Nalewajski and from Koch and
temperature. The dashed lines serve as guides to the eyes, the solid lines are Holthausen [19] and - with special emphasis to transition metals -
simulated curves (parameters see text).
in the book edited by Kaltsoyannis and McGrady [20] and in a
recent review article from Cramer and Truhlar [21]. Despite the
fact, that DFT is used now for many years to obtain spin state
The paradoxic conclusion one may draw from this example is, that energies, only in the last few years it became clear that such
completely wrong values for ΔEHL and ΔSHL might lead to better calculations have a sound theoretical foundations. A detailed
agreement with experiment than the true values. This apparent review on studies on this topic was recently presented by Jacob and
paradox is due to the fact that the steepness of the spin transition Reiher [22].
(and clearly also its hysteresis) shown in Figure 1 is a cooperative In quantum chemistry the terms "ab initio" and "from first
effect that cannot be explained by a simple model for isolated principles" (that will be used synonymously here) stand for
molecules as it is done by using Eqs. (1)-(3). This has always to be approximative solutions of the Schrödinger equation without use of
kept in mind when judging calculated energy differences for any empirical data (except for the fundamental physical constants
isolated molecules on the basis of experimental data. This example that are already included in the Schrödinger equation). One might
also illustrates that the entropy difference due to the spin argue that some DFT methods fit to this definition (see for instance
multiplicity amounts only for a small fraction of ΔSHL, which the reflections about the construction of ab initio DFT from Neese
underlines the importance of the vibrational contribution. [23]) but for practical reasons we follow the common practice to
Electronic structure calculations have been used in the past two reserve the term ab initio to wave function based methods, being
decades to determine many different properties of SCO complexes, aware of a different use in part of the literature reviewed here.
like molecular geometry, charge and spin densities, normal modes Another criterion to classify computational studies on SCO
of vibration, UV/VIS., IR and Raman spectra, nuclear resonant complexes is the inclusion of intermolecular interactions. In a
vibrational spectra, Mössbauer parameters [6], dielectric constants simplistic view one may distinguish between calculations for a free
[7] and many more. In this review the focus shall be on those molecule ("in vacuo" or "gas phase") and calculations with
properties that are of interest particularly for SCO complexes, like periodic boundary conditions as they are standard in solid-state
the transition temperature, T1/2, the width of the hysteresis, ΔT, the physics.
adiabatic energy and entropy differences, ΔEHL and ΔSHL, The main part of this review is divided into four parts. The first
respectively, and (with regard to light induced effects) the two parts are dedicated to ab initio and DFT calculations, the third
excitation energies. As has been demonstrated before, ΔEHL and part deals with calculations that include inter-molecular effects and
ΔSHL are the prerequisites for the prediction of T1/2 and ΔT. The in the fourth part the calculation of vibrational properties is
calculation of the entropy difference is a routine work since some discussed.
time [8] and can be performed with reasonable accuracy now
[9,10]. Much more difficult is the calculation of ΔEHL, since it is
the difference of two absolute energies of the order of 10 7 kJ mol-1 Ab initio calculations
differing by only a few 10 kJ mol -1 [11]. Since no quantum
chemistry method can calculate the total energy of a SCO complex Within the Born-Oppenheimer approximation the wave function
with an accuracy of 1 ppm any attempt to calculate ΔEHL must to a that describes a molecule can be separated into an electronic and a
large extent be based on a mutual cancellation of errors. nuclear part. For many purposes it is sufficient to treat only the
The fundamental origin of the spin transition is intra-molecular - electrons on the basis of quantum mechanics while the nuclei are
even in liquids it is possible to populate the HS state by irradiation regarded as classical particles. The description of the molecule is
with light [12], comparable to the light induced excited spin state then equivalent to the solution of the electronic time-independent
trapping (LIESST) observed in solids [13] - but inter-molecular Schrödinger equation
interactions are obviously quite important and contribute
significantly to ΔEHL. Bučko et. al. [14], for instance, calculated an H(Rα, ... ,Rω) ψ(ra, ... ,rz) = E ψ(ra, ... ,rz) (4)
inter-molecular contribution of about 11 kJ mol-1, and for the
important class of polymeric SCO complexes this contribution is
probably even larger [15]. Many features of the HS-fraction curve

Submitted to the European Journal of Inorganic Chemistry 3


Here Rα, ... ,Rω and ra, ... ,rz are the coordinates of nuclei and CI is called correlation energy (Scheme 1, again under the
electrons, respectively, and the Hamiltonian H depends on the assumption of an infinitely large basis set).
position of the nuclei. Since analytic solutions for the electronic The energy obtained by the HF method with CI converges
wave function ψ are available only for one electron systems, Eq. towards the true ground state energy of the molecule, and if certain
(4) must be solved numerically. The first steps towards an conditions are fulfilled this convergence is monotonous.
approximate solution of this Schrödinger equation is given by the Unfortunately, in full CI the number of determinants grows
Hartree product factorially with the number of orbitals making this method
infeasible for anything else than small molecules with small basis
ψ(ra, ... ,rz) = φ1(ra) ∙ φ2(rb) ∙ ... ∙ φn(rz) (5) sets. For this reason post-HF methods often divide the set of
orbitals into three groups, one group containing orbitals that are
that represents the many-electron wave function ψ as a product always occupied, a second group of orbitals that are always empty,
of one-electron wave functions (called orbitals) φ1, φ2, ... The and a third group, the so called complete active space (CAS), with
Hartree product (5) can be used to solve the Schrödinger equation orbitals that may be filled or not [24]. The number of Slater
(4) iteratively. From the perspective of physics one may interpret determinants that have to be considered is now determined by the
the iterative solution in such a way that each electron "sees" the size of the CAS (usually of the order of 10). The complete active
average distribution of all other electrons. The Hartree space self-consistent field (CASSCF) method [25] iteratively
approximation violates the Pauli principle and leads to a too high refines a starting wave function, for instance the
potential energy because electrons with equal spins do not avoid monoconfigurational HF solution, until it converges towards a self-
each other sufficiently. A better approximation is given by the consistent multiconfigurational wave function. This wave function
Hartree-Fock method where the wave function is represented by a may be further improved by a subsequent perturbational treatment
so called Slater determinant of orbitals, which by construction in order to account for the remaining electron correlation effects
fulfils the Pauli principle. For a two electron problem the Slater that are not included in the CASSCF wave function due to the
determinant is written as limitation of the complete active space. In case of CASPT2 [26],
the most common procedure of this type, perturbational corrections
ψ(ra,rb) = [φ1(ra) ∙ φ2(rb) - φ2(ra) ∙ φ1(rb) ]/A (6) up to the second order are evaluated. Like the other perturbational
methods, CASPT2 is not variational, so in rare cases the calculated
where A is a normalization constant. As a result the probability to energy correction may be too large instead of too small.
find two electrons of same spin is reduced with respect to the An alternative approach to approximate full CI is the so called
simple Hartree product (5), leading to an energy reduction which is coupled-cluster (CC) theory, where the exponential of an excitation
termed exchange energy (Scheme 1). It has been tacitly taken for operator (called cluster operator) is applied to the HF wave
granted here that the orbitals φ1 and φ2 have no limits of flexibility, function in order to get the multiconfigurational CC wave function.
which requires numerical orbitals or an infinitely large basis set. Depending on the size of the molecule and the available computer
resources the cluster operator is truncated after single, double or
triple excitations. For the calculation of SCO complexes a full
inclusion of triple excitations is currently prohibitive and the most
favourable coupled cluster method for this purpose seems to be the
so-called CCSD(T) method [27], where S and D denote the
inclusion of single and double excitations, respectively, and (T)
stands for a singles/triples correction term.
High level ab initio calculations often serve as a reference for the
judgement of the quality of less demanding calculations but they
are currently restricted to very small model systems. For
calculations on larger systems and on solid samples it is inevitable
to use experimental data as reference. This implies that not only the
adiabatic energy difference ΔEHL has to be calculated but also the
entropy difference ΔSHL. The determination of the latter requires
the calculation of molecular vibrations.
First ab initio calculations for very simple models of transition
Scheme 1. Exchange and correlation energy as defined by the HF method metal complexes were reported 1994 by Akesson [28] and 1998
and configuration interaction. Bolvin published CASPT2 results that allowed to extract Racah's
parameter B and C and the crystal field splitting Δ for subsequent
use in ligand field theory [29]. It took a couple of years before the
The correlation between electrons of same spin that is enforced first studies of CASPT2 and CCSD(T) calculations on more
by the antisymmetry of the HF wave function (6) is sometimes realistic models were published [30,31,32]. These early studies will
called Fermi correlation. In this case the term Coulomb correlation be discussed in the following section since an important part of
is used for the remaining correlation that is not due to the Pauli them are comparisons with density functionals.
principle. Also within the framework of the HF method the On the example of the model system [Fe(NCH)6]2+ Kepenekian
individual electrons do not "see" where exactly the other electrons et al. [33] enlightened the role of basis sets and active spaces for
are at the very same moment. As a result electrons are coming too CASSCF and CASPT2 calculations of the HS-LS energy
close to each other and “feel” in average a too strong Coulomb difference ΔEHL. Most important according to their results are the
repulsion. In order to cure this defect, in post-HF methods the wave basis functions that are centred on the central metal atom and its
function is written as a sum of different Slater determinants where nearest neighbour atoms whereas ΔEHL is far less sensitive on the
one or more orbitals of the ground state determinant is exchanged size of the basis for the second and third nearest neighbours. For
by orbitals with higher energy, a procedure that is called the model system [Fe(NCH)6]2+ a 7s6p5d3f2g1h/4s3p1d basis for
configuration interaction (CI) or full CI, if really all possible Fe and N atoms is proposed as a reasonable compromise between
determinants are regarded. The energy reduction resulting from full accuracy and computational cost. Comparison with reference
CCSD(T) calculations suggest a CAS[10,12] active space for the

Submitted to the European Journal of Inorganic Chemistry 4


model system. Later Domingo et. al. [34] performed CASPT2 and Another new development is the density matrix renormalization
CCSD(T) calculations for a series of iron(II) model complexes and group (DMRG) algorithm. Although this algorithm can be traced
report that the stability of the LS state is directly related to the back to a work of Fano et al. [42] more than fourteen years ago,
amount of ligand-to-metal-charge-transfer configurations that are only very recently Marti, Ondik, Moritz and Reiher [43]
mixed to the wave function, or in other words to the amount of demonstrated that energies of different spin states of mono- and
ligand-σ donation. For some of the models CASPT2 and CCSD(T) binuclear transition metal complexes can be calculated by the
gave significantly different values for ΔEHL, a finding that was DMRG algorithm with the same accuracy as with CASSCF
ascribed to the intrinsic multiconfigurational character of the LS calculations. The essential advantage of the DMRG algorithm is
state that is not well described by the CCSD(T) calculations which the feasibility of huge active spaces, which make this method
has a monoconfigurational reference wave function. For excited especially interesting for the calculation of polynuclear transition
states the multiconfigurational character of the wave function is metal complexes.
usually even more pronounced than for the ground state.
Multireference ab initio calculations are, therefore, the method of
choice for the description of optical excitations in not too large
complexes. In the field of spin crossover this was exemplified early Density Functional Calculations
by three studies of Ordejon, de Graaf and Sousa [35,36,37] who
applied the CASPT2 method including spin-orbit coupling effects Although the Xα method proposed by Slater [44] in 1951 as a
to [Fe(tz)6] (tz=1-H-tetrazole) and [Fe(bpy)3] (bpy = 2,2'-bipy- simplification of the HF method could be regarded as the first DFT
ridine), the former a model complex, the latter a stable LS complex method, usually the Hohenberg-Kohn theorems [45] published in
that can be transformed into a HS complex by irradiation with light 1964 by Hohenberg and Kohn (who was awarded the Nobel prize
due to the LIESST effect. They investigated the population of for chemistry for the foundation of DFT in 1998) are considered as
states in the metal-to-ligand-charge-transfer (MLCT) band during starting point of density functional theory. These theorems state
this process, starting with the initial activation to a 1MLCT state that all ground state properties of a molecule are determined by the
and most probably followed by transient population of 3MLCT and electronic charge density ρ(r) and that this charge density can be
5
MLCT states before the final deactivation into the HS state takes found by a variational principle. In other words there exists a
place. The Fe-N distance dependence of the MLCT energies and functional EHK[ρ] that gives the lowest energy for the true ground
the spin-orbit coupling between these states that were obtained in state charge density ρ0(r). In this way DFT replaces the
these studies give a plausible explanation for the efficient transfer complicated N-particle wave function Ψ by the one-particle
to the HS state. A similar approach was applied independently by function ρ that has an intuitive meaning. Unfortunately, no analytic
Suaud et al. [38] to the SCO complex [Fe(dipyrazolpyridine) 2]2+ form for EHK[ρ] is known, so approximations have to be used. The
and to the model system [Fe(NCH)6]2+. Also these calculations most successful approach up to now to find an approximation to
suggested a pathway that involves excited singlet, triplet and EHK[ρ] has been proposed in 1965 by Kohn and Sham [46].
quintet states coupled by spin-orbit interactions. It turned out that The Kohn-Sham equations are used to iteratively solve the
the nature of the main reference function does not very sensitively Schrödinger equation for a system of non-interacting electrons (the
depend on the size of the basis set, in sharp contrast to the electrons "see" each other only by meaning of an effective potential
calculated spin state splitting ΔEHL. that depends on the charge density) leading finally to a one-
determinantal "non-interacting" wave function Ψ ni (which is not the
Table 1. Short list of acronyms frequently used in quantum chemistry. true wave function except for trivial cases) and the true ground
state charge density ρ0(r). In this formalism the energy can be
Acronym meaning written as
CAS complete active space
CASPT2 CASSCF with second order perturbation theory EHK[ρ] = Eni[ρ] + Exc[ρ] (7)
CASSCF complete active space self consistent field
CCSD(T) coupled cluster approach with single and double The first term on the right hand side, Eni, is the sum of the kinetic
excitations and a singles/triples correction term energy, readily obtained from Ψni, and the classical potential
CI configuration interaction energy of a charge density ρ(r) in its own field and in the field of
DFT density functional theory the nuclei. Eni does not equal the electronic energy of the molecule
DMRG density matrix renormalization group because the kinetic energy calculated from Ψni is only a very good
GGA generalized gradient approximation
approximation to the true kinetic energy, and, much more
important, the potential energy obtained by classical electrostatics
HF Hartree-Fock
is not equal to the potential energy obtained form quantum
LDA local density approximation
mechanics. These differences are lumped together into a so called
QCISD quadratic configuration interaction with single and
exchange-correlation functional Exc[ρ] the construction of which is
double excitations
a central problem of density functional theory. The oldest DFT
method was born when Slater [44] proposed his Xα functional
Further developments
Exc[ρ] = (3/2) α (3/π)1/3 ∫ρ4/3dV (8)
Due to the tremendous computational resources needed by
where α is an empirical parameter that is usually chosen to be
accurate ab initio methods they are unlikely to become in the near
about 0.75 [47]. This simple example of an exchange-correlation
future a standard tool for calculations on SCO complexes. There
functional, which belongs to the class of local density
are, however, new developments that may increase the
approximations (LDA), works surprisingly well and much effort is
applicability of ab initio methods, like the integration of QCISD
needed for the construction of functionals that lead to better results.
and CCSD into the framework of the local pair natural orbital
There are two classes of such improved exchange-correlation
concept developed by Neese et al. [39,40,41], which promise a
functionals, the so called hybrid functionals and functionals
dramatic reduction of computational demands with practically no
employing the generalized gradient approximation (GGA), also
loss of accuracy.
called gradient-corrected functionals. The latter are often

Submitted to the European Journal of Inorganic Chemistry 5


constructed by forming the sum of a LDA functional and a horse in the field of SCO complexes too, especially if larger
correction term that depends on the gradient of the charge density. molecules are investigated and geometries have to be optimized. A
Hybrid functionals are called in this way because next to a pure typical example for such a study was published by Zein et al. [56].
density functional term they include a HF exchange correlation Despite of all efforts to improve density functional methods with
term, which makes them computationally more demanding. respect to spin state energies B3LYP may still happen to give the
Usually even more demanding is the application of the new best description of an SCO complex (see Bannwarth et al. [57] for
double-hybrid density functionals (for instance B2PLYP) which a recent example).
include corrections from second-order many-body perturbation The observation, that for calculations with hybrid functionals
theory [48]. However, Neese et al. demonstrated that the ΔEHL varies almost linearly with the amount of exact HF exchange
computational expenses could be reduced drastically without loss included, guided Reiher et al. [58] to the construction of the
of accuracy by some approximations like density fitting [49,50]. reparameterized hybrid functional B3LYP* that turned out to be
While for full CI ab initio calculations the total energy converges especially useful for the calculations of ΔEHL in SCO complexes
towards the exact value with increasing basis sets, this cannot be [9,10]. B3LYP* contains only 15 % exact exchange energy as
stated with the same rigour for DFT calculations and it may happen compared to 20 % in the B3LYP functional. The good performance
that an enlargement of the basis set increases the error of the of B3LYP* was confirmed by more recent studies. Shiota et al.
calculated energy. However, as will be discussed below experience [59] tested BLYP, B3LYP and B3LYP* for the SCO complex
teaches that small basis sets lead to inaccurate energy differences. [Fe(2-pic)3]2+ (pic = picolylamine) and got best results with
In case of crystals where DFT calculations with periodic B3LYP*. A similar conclusion was drawn by Qu [60] who
boundary conditions are used, a popular method for accounting for calculated the SCO complex [Fe(bt)2(NCS)2] (bt = 2,2'-
electron correlation is the Hubbard correction, originating from the bithiazoline) with different density functional methods. The
Hubbard model for conductors and insulators [51]. In simple words energies of different spin states including zero point corrections as
the Hubbard correction can be described as a penalty for the partial well as vibrational spectra and UV/VIS. absorption spectra were
occupation of orbitals (in case of iron usually only d-orbitals) that calculated. The reparameterized hybrid functional B3LYP* was
leads to a localization of electrons. The Hubbard correction is reported to obtain the best results for this complex. For other SCO
tuned by two empirical parameters, J and U, the latter being more complexes the amount of exact exchange in B3LYP* seems to be
important and giving the method its acronym DFT+U. still too large, for instance in the case of [Fe(phen) 2(NCS)2] Reiher
Although the introduction of hybrid functionals is a progress in found best agreement with an admixture of 12 % exact exchange
many respects there are still good reasons to develop gradient [9]. On the basis of a systematic study for iron(II) and iron(III)
corrected functionals that behave as well since these could benefit tripodal imidazole complexes Brewer et al. [61] reached at an
more from the resolution-of-the-identity techniques, in case of free optimal amount of HF exchange contribution of 13 %. More recent
molecules, or from the fast Fourier transformation for plane-wave studies by Jensen, Cirera and Paesani [62,63] suggested that an
basis functions, in case of crystals. even lower exchange admixture of 10 % (as it is the case in the
TPSSh functional) would perform better with respect to the HS-LS
energy difference. In this study several iron and cobalt SCO
Performance of functionals complexes were investigated and the TPSSh functional was the
only one that reached an accuracy of about 10 kJ mol -1. In line with
In the last decades a large variety of different density functionals this finding are calculations for the SCO complex [Fe(salen)(NO)]
has been developed, each of them having its own advantages and reported by Boguslawski et al. [64].
disadvantages (selected functionals are compiled in Table 2). A Another functional that was reported to perform quite well with
good impression of the multitude of density functionals is given by SCO complexes, is the GGA method OPBE. Swart [65] studied a
a study of Zein et al. [52], to our knowledge the most large number of small iron model systems as well as five larger
comprehensive study of the performance of different density iron complexes including [Fe(phen)2(NCS)2] with several density
functionals on SCO complexes, which includes results for 53 functional methods and found best agreement with experiment
different functionals that were employed together with a triple zeta when using the functional OPBE.
plus polarization basis set. An important result of this study is, that
the difference of ΔEHL between two similar SCO complexes is far Table 2. Selected density functionals that are used for SCO complexes.
less dependent on the choice of the functional than ΔEHL itself.
Another detailed comparison of 15 different density functionals Abbreviation type comment references
for a series of iron(II) model complexes was made by Rong et al. B3LYP hybrid most widely used functional [67]
[53]. A special feature of this study is the decomposition of the B3LYP* hybrid similar performance as [58]
energy difference ΔEHL into contributions from the exchange B3LYP, better results for ΔEHL
energy, correlation energy, kinetic energy, classical Coulomb B2PLYP double small bias to HS state, good [68]
energy between electrons and electrons, electrons and nuclei and hybrid results for ΔEHL also for non-
nuclei and nuclei. It turned out that none of these contributions SCO complexes
dominated ΔEHL. LSDA local local spin density
Since long it is known that LDA functionals strongly approximation
underestimate the stability of the HS state and are for this reason LYP GGA correlation functional from [68a]
seldom used for the calculations of isolated SCO complexes. Due
Lee, Yang and Parr
to their computational efficiency they are, however, of some value
OPBE GGA good results for ΔEHL [66,69]
for calculations with periodic boundary conditions [54]. To a lower
OLYP GGA good results for ΔEHL [66]
extent also many gradient corrected functionals have a bias to the
TPSSh hybrid quite good results for ΔEHL in [70]
LS state. Conversely the basic HF method (due to lack of
SCO complexes
correlation between electrons with different spin) overestimates the
stability of the HS state [8,55]. The hybrid functional B3LYP (with VWN GGA correlation funtional from [70a]
20 % HF exchange) inherits this bias for the HS state but to a much Vosko, Wilk and Nusair
lower extent than in the HF method. B3LYP is probably the most
widespread quantum chemical method and has become a working

Submitted to the European Journal of Inorganic Chemistry 6


The invention of B3LYP* as well as the construction of new ligand, which is a challenge on its own. This complex, together
GGA functionals improved the performance of DFT calculations with other nitrosyl complexes and the small model system
for SCO complexes and led to the necessity of more detailed [Fe(NO)]2+, was investigated thoroughly by Boguslawski, Jacob
performance tests in order to find the most suitable functional for and Reiher [64] employing several density functionals as well as
the calculation of spin state energies. The performance of density the CASSCF method. Testing several active spaces for [Fe(NO)] 2+,
functional in this respect can be evaluated by comparison with converged results were obtained for CAS[11,15], delivering in this
reference values that may be either obtained from experiment or way a reference spin density. Some of the DFT calculations gave
from high level ab initio calculations. Since experimental values satisfactory results for the adiabatic energy difference ∆EHL, but
usually are available only for the solid state, a proper comparison none of the functionals was able to correctly reproduce the
requires periodic DFT calculations for the crystal. Such reference spin density. In a very recent study Boguslawski et al.
calculations are quite demanding and it is difficult to find out [75a] demonstrated that the spin density of [Fe(NO)] 2+ can be
whether discrepancies between experimental and calculated results calculated accurately using the DMRG algorithm. Previously, also
are due to a deficient description of the spin state energetics or due Conradie and Ghosh [76] found that OLYP and B3LYP* gave
to an insufficient description of inter-molecular interactions. These similar adiabatic energies for [Fe(salen)(NO)] but quite dissimilar
difficulties make high level ab initio calculations for isolated SCO spin densities. One may conclude that this failure of treating the
complexes quite desirable. spin density correctly is closely connected to the difficulties of
Guillon et al. [71] investigated a two-step SCO complex with a density functionals to predict reliably ∆EHL.
variety of density functionals. While good results for the
geometries were obtained, the HS-LS energy difference was not
calculated accurately by any of the functionals, including B3LYP*.
Functionals that performed better for other complexes failed for
this large complex.
The groups of Casida, Hauser, Neese and Reiher published a Substituent effects
series of four articles [30,31,72,73] that are dedicated to a detailed
study of the performance of density functionals on the calculation None of the so far discussed methods for electronic structure
of ΔEHL for the iron complexes [Fe(H2O)6]2+, [Fe(NH3)6]2+, calculations is able to exactly determine the adiabatic energy
[Fe(bpy)3]2+ and [Fe(L)('NHS4')] ('NHS4' = 2,2'-bis(2-mercapto- difference between the HS and the LS state, ΔEHL, for a large SCO
phenylthio)diethylamine dianion and L = NH 3, N2H4, PMe3, CO or complex. However, even if the calculated absolute value of ΔEHL
NO+). In the first two of these studies by Fouqueau et al. [30,31] obtained from a given quantum chemical method is quite
the DFT results were compared with results from CASSCF and inaccurate, the calculated change of ΔEHL after modification of the
CASPT2 calculations and from calculations with the SCO complex might be more accurate. The assumption justifying
spectroscopy-oriented configuration interaction (SORCI) this idea is that the errors of the calculated ΔEHL(A) for complex A
procedure [74]. In the latter two studies by Lawson Daku et al. [72] and of ΔEHL(B) for complex B are quite similar. Therefore, when
and Ganzenmüller et al. [73] DFT calculations were compared with the difference
experimental data as far as available. For all complexes the
geometries calculated with hybrid and gradient corrected ΔA,B = ΔEHL(A) - ΔEHL(A) (9)
functionals were in good agreement with those obtained from
CASSCF or experiments, respectively, provided sufficiently large is formed, the errors of ΔEHL(A) and ΔEHL(B) should cancel each
basis sets (triple zeta) were used. In general the DFT methods other to some extent. The chances that this happens should be the
underestimated ΔEHL, but reasonably good performance was better the more similar the complexes A and B are. Ideal objects to
observed for the hybrid functionals B3LYP, B3LYP* and PBE0 test this hypothesis are SCO complexes with substituted ligands.
and also for the GGA functionals RPBE and OLYP and the Meta- Early studies [4,77] on some iron(II) complexes with tris(pyra-
GGA VSXC. That the GGA functionals were able to compete with zolyl) ligands gave qualitative agreement between experimental
hybrid functionals is promising in view of calculations for large and calculated transition temperatures. In order to calculate the
molecules and crystals, for reasons stated above. transition temperatures, ΔEHL and ΔSvib(T) were calculated with the
The four studies just discussed were extended by Pierloot and gradient-corrected density functional BLYP and a double zeta basis
Vancoillie [32,75] who employed larger basis sets and performed with effective core potentials. A follow-up study [11] on the same
CASPT2 calculations for [Fe(bpy) 3]2+ too. From the density SCO complexes employing several different functionals and basis
functionals OLYP, PBE0, B3LYP and B3LYP*, the OLYP sets demonstrated that these methods gave similar accuracies for
functional gave best agreement with CASPT2 results for ΔEHL but ΔA,B, even with small basis sets, although the methods differed
proved inferior to B3LYP with respect to bond lengths. The significantly concerning the accuracy of ΔEHL.
admixture of HF exchange in hybrid functionals B3LYP and PBE0 A more recent study from Güell et al. [78] on quite similar
that was optimal with respect to the calculated adiabatic energy complexes reported energy differences calculated with a large
difference, varied considerably for different complexes. series of different density functionals and a triple zeta basis. Best
Ye and Neese [50] tested a wide range of GGA functionals as agreement with experiment was obtained with the OPBE
well as hybrid functionals and the double-hybrid functional functional. These calculations allowed to rationalize some intra-
B2PLYP for calculations to a series of model complexes as well as molecular interactions that influence the spin transition. This study
SCO complexes like [Fe(phen)2(NCS)2]. Interestingly, they also emphasized the strong influence of counter ions, an
observed only a negligible influence of the basis set on results observation that has also been made by Lemercier et al. [54] for
obtained with hybrid functionals, in contrast to earlier findings. another type of SCO complexes.
However, very large basis sets turned out to be necessary for Deeth at al. [79] proposed empirical ligand-field molecular
calculations with the double hybrid functional B2PLYP. The latter mechanics (LFMM) as a tool for fast calculations of ΔEHL for a
functional was found to be very useful for non-SCO complexes but series of relative large molecules. The obtained energies agree with
showed a slight bias in favour of HS states for SCO complexes. those obtained by DFT methods. On the other hand Swart et al.
While SCO complexes in general are challenging objects for [80] proposed the multi-scale method which integrates improved
quantum chemistry, this is even more the case for the SCO density functionals, a polarizable force field and hybrid QM/MM
complex [Fe(salen)(NO)] due to its noninnocent nitric oxide calculations. This approach led to prediction of correct transition

Submitted to the European Journal of Inorganic Chemistry 7


temperature in Fe(II) complexes of tris-pyrazolyloborate contribute to ΔSvib and about 75 % of ΔSvib is due to the 15 modes
complexes. of the central FeN6 octahedron. Another example for a complex,
for which the normal modes of intra-molecular vibrations were
calculated with high accuracy, is [Fe(pmea)(NCS) 2] (pmea =
TD-DFT bis[(2-pyridyl)methyl]-2-(2-pyridyl)ethylamin). Remarkably, for
this complex ΔSvib is about twice as large as for [Fe(phen) 2(NCS)2]
In its original formulation density functional theory is limited to (roughly 40 versus 20 J mol-1 K-1) [93b].
the ground state of a system. The Runge-Gross theorem [81] Garcia et al. [94] calculated ΔSvib for Mn(III) and a Cr(II) SCO
extends the theory to excited states by introducing a time- complexes and they found that that due to elongation of only one
dependent external potential, which gives this extension of DFT its bond upon triplet (LS) – quintet (HS) transition in the d 4 system
name, time-dependent density functional theory (TD-DFT). The ΔSvib is only a fraction (ca. 1/3) of the values predicted for spin
currently most common application of TD-DFT is the calculation transitions in Fe(II) analogues, in line with former results of
of electronic excitation energies. An early application of TD-DFT Nakano et al. [95].
to SCO complexes was presented by Respondek et al. [82] who Wolf et al. [96] reported the normal coordinate analysis of the
investigated excited states of some slightly simplified models of Fe(II) complex [Fe(btpa)]+2 and its benzyl analogue. The evolution
iron(II) complexes with respect of the suitability of these of spectral patterns in time resolved picosecond IR experiments
complexes for the LIESST effect. Ando et al. [83] used TD-DFT to and previous nanosecond resonance Raman studies [97] were
calculate optical excitations for two isomers of the SCO complex interpreted in terms of red shift of trigonal bending bipyridine
[FeII(2-pic)3]2+ and an analogous iron(III) complex. The obtained d- modes occurring in 1015-1040 cm-1 region upon spin transition.
d transition energies were in agreement with the experimental The first system reveals two different HS states (HS1 and HS2),
observation of the LIESST and, separately, the reverse-LIESST differing in the ligand conformation and symmetry. Both were later
effect after irradiation with light of different wavelengths for the modelled with DFT methods (B3LYP*/CEP-31G) [98] to yield
iron(II) complexes and the lack of such observations for the their relative energies and the transition states of the HS1 -> HS2
iron(III) complex. In recent years the transient X-ray absorption pathway, obtained by means of QST3 approach (B3LYP/CEP-
spectroscopy has established itself as a new powerful method to 31G). TD-DFT calculations gave the estimate of the energies of the
study the photo induced spin transitions [84]. The TD and TP vertical excited states of the LS isomer.
(transient potential) DFT calculation were applied to various DFT methods were extensively used for simulations of NIS
spectral regions of the Fe K-edge of SCO complexes [85]. In a very (NRVS) spectra of trimeric [99] and polymeric Fe(II) SCO systems
recent work Qu [60] used the TD-DFT method to calculate based on 4-substituted-1,2,4-triazole ligands [100,101]. The
electronic spectra of [Fe(bt)2(NCS)2]. polymeric systems were modelled by means of pentameric models
An interesting study for the rare case of a four-coordinate Co(II) that gave a good fit of the experimental spectra. It was shown that
system revealing a LS (planar) – HS (tetrahedral) interconversion in order to get reasonable Fe-N and Fe-Fe distances a charge
was performed by Marshak et al. [85a] who calculated the compensation of the +6 and +10 cationic unit had to be introduced
electronic energies with B3LYP/TZVP for different values of the by means of including a shell of the anions in the model. The
dihedral angle between two planar ligands (Θ,tetrahedral twist). assignment of the observed NIS (NRVS) and Raman bands to
They found that the crossover of the EelLS(Θ) and EelHS(Θ) curves molecular modes was given.
occurs at Θ=57°, yielding an activation barrier of 6 kcal mol -1. DFT Similarly, the application of B3LYP/CEP-31G to the normal
results suggest the SOMO orbital has b2 symmetry, corresponding coordinate analysis of the [Fe(ptz)6](BF4)2 complex [102,103]
to a (dyz)1 ground state. TD DFT calculations were performed as allowed the assignment of observed NIS (NRVS) bands to the
well. individual modes of HS and LS isomers. Particularly, the observed
changes in the observed disordered phase were related to the
movement of the n-propyl group that was coupled to the modes
Vibrational Spectra
being essentially Fe-N stretching. Also the NIS (NRVS) spectra of
the dimeric [{Fe(L)}2(bbim)](ClO4)2 complex were calculated on
The vibrational properties are of key importance for
the basis of normal coordinate analysis. With a similar approach
understanding the spin-crossover phenomenon [86]. Recent
Trautwein et al. [104] used the B3LYP/6-311G method to interpret
developments in vibrational spectroscopy for SCO complexes are
the single crystal and powder NIS (NRVS) spectra as a function of
reviewed in ref. [87,88]. Three types of vibrational data may be
pressure and temperature for Fe(II)(STP)(ClO 4)2 and Fe(II)(ETP)
derived from quantum chemical calculations: a) normal modes of
(ClO4)2 complexes (ETP=1,1,1-tris{[N-(2-pyridylmethyl)-N-
vibration and the corresponding infrared, Raman or nuclear
methylamino]ethyl}-propane, STP=1,1,1-tris{[N-(2-pyridyl-
inelastic scattering (NIS, also called nuclear resonance vibrational
methyl)-N-methylamino]ethyl}-ethane). Also in the case of the
spectroscopy, NRVS) spectra, b) the vibrational contribution to the
Fe(tetraphenylporphinato) SCO complex [Fe(II)(TPP)CN] - DFT
entropy difference upon spin transition, ΔSvib, c) the change of
calculations (B3LYP/VTZ+6-31G*) were performed for both spin
vibrational energy upon spin transition (for T=0 K: difference of
isomers in order to assign the vibrational modes obtained in
zero-point corrections). The calculation of these data is usually the
variable temperature single crystal und powder NVRS (NIS)
domain of DFT methods since with ab initio methods the geometry
measurements [104a].
optimization and subsequent calculation of normal modes of
Two recent articles represent, to the best of our knowledge, the
vibrations would be too time consuming for any but the smallest
first examples of the vibrational calculations for the solid state of
complexes.
SCO systems. Middlemiss et al. [105] reported solid-state lattice
Among SCO complexes [Fe(phen)2(NCS)2] [89] is one of those
that have been investigated most thoroughly by experiment and by dynamics calculations using the hybrid density functionals
electronic structure calculations. This applies especially to accurate (combinations of LSDA, B88, UHF, LYP and VWN correlation
calculations of the normal modes of vibrations, vibrational spectra, functionals) for the CsFe[Cr(CN)6] system. Normal coordinate
vibrational entropies and zero point corrections analysis was performed and the vibrational entropy was estimated.
[9,10,90,91,92,93,93a]. Brehm et al. [93a] discovered the Bučko et al. [14] used the periodic dispersion corrected DFT
importance of low-frequency mode contributions to the entropy. calculations for [Fe(phen)2(NCS)2] to study the spin transition in
Later Ronayne et al. [90] confirmed that only low energy solid and for the isolated molecule. The PAW (projector-
vibrational modes (20 % of the 147 mode of the free molecule) augmented-wave) basis set was used in combination with the PBE

Submitted to the European Journal of Inorganic Chemistry 8


functional. Strong correlation of d-electrons was treated by an represented by a Madelung field that was calculated using X-ray
additional Hubbard-like term [106] to the Kohn-Sham Hamiltonian crystal structures and ab initio charge distributions of the isolated
and the PBE-D2 correction of Grimme [48] was introduced in molecule. In this way the interaction parameter Γ of the Slichter
order to account for van der Waals interactions. Normal vibrations and Drickamer model [see Eq. (10)] could be predicted without
as well as transition enthalpies and entropies were calculated. using empirical data other than the X-ray structures. These two
A sometimes overlooked problem is the vibrational energy studies [112,113] are rare examples for ab initio calculations that
difference (nuclear contribution to enthalpy, difference of zero- take into account inter-molecular effects. A DFT study employing
point energy corrections) upon spin transition. The vibrational a Madelung field was published by Zein et al. [114].
enthalpy difference for a spin transition at given temperature is the An example for the investigation of polynuclear fragments is
given by the DFT study of a family of three 1D-polymers
sum of the change of zero-point corrections (vibrational energies at
exhibiting spin transitions, carried out by Matouzenko et al. [115].
T = 0 K, ΔZP = Evib,HS- Evib,LS, which, as a rule, is negative) and the
The calculations were performed for binuclear fragments using the
thermal contribution to enthalpy at given temperature. Reiher
hybrid functional B3LYP as well as the gradient-corrected
[9,10] gave values for [Fe(phen)2(NCS)2] of -5 to -8 kJ mol -1,
functional BP86. Only the latter was able to predict the correct LS
depending on the functional used, while Bučko et al. [14] reported ground state, but for both functionals the calculated electronic
values of -5.7 and -4.7 kJ mol -1 for the isolated molecule and the energy differences were in the same order as the observed
solid state (taken from phonons at the Γ-point), respectively. transition temperatures.
Jensen and Cirera [62] stated that for a large series of SCO An alternative approach was chosen by Vargas et al. [116] who
complexes the largest value was -7 kJ mol -1. These data point to the investigated the influence of the supercage zeolite Y on the SCO
importance of the above effect for the relative stabilization of the complex cation [Fe(bpy)3]2+ by DFT calculations with a variety of
LS and HS states. functionals. Although very different values for ΔEHL were obtained,
all functionals consistently predicted that the supercage
environment destabilizes the HS state with regard to the LS state.
Intermolecular effects Lord et al. [117] successfully applied the B3LYP functional
together with a continuum solvation model to predict the redox
Despite of the importance of inter-molecular effects electronic potentials of transition metal complexes (with chromium,
structure calculations were for many years restricted to isolated manganese, iron cobalt and nickel centres) with variable spin states
mononuclear complexes, due to the limits of computer resources. in solutions. The COSMO solvation approach in connection with
However, molecular calculations were used to rationalize the the RPBE functional was applied by Deeth and Fey [118] to a
structure of polymeric SCO compounds. Le Guennic et al. [107] series of simple iron(II) and iron(III) model complexes. According
applied the BP86 and the B3LYP functionals to fragments of 1D- to their results the RPBE/COSMO approach slightly overestimates
polymers in order to determine the energy of different coordination the stability of the LS state.
modes. Independent of the functional, the calculations predicted
that in case of the pyim ligand the cis coordination was clearly
favoured in comparison to the trans coordination in agreement with Periodic boundary conditions
the observed zig-zag form of the polymer. The favouring of the cis
coordination was much less pronounced in case of the aqin The straightforward way to handle crystalline SCO compounds
polymers and indeed here linear chains with trans coordination are electronic structure calculations with periodic boundary
were observed experimentally. conditions as they are for instance used in solid state physics to
The first step to take into account cooperativity has been the investigate metals or semiconductors. While the latter materials,
calculation of bi- and tetramolecular complexes. Zein et al. [108] however, often consist of unit cells with only a few atoms, SCO
presented DFT calculations for five binuclear complexes, which compounds typically have unit cells with a few hundred atoms. For
exhibit one- or two-step spin transitions or no transition at all. The this reason periodic DFT studies for SCO complexes are
calculated energies for the [LS-LS], [LS-HS] and [HS-HS] states comparatively recent and the first one, we are aware of, was
allowed a correct prediction of the observed nature of the published in 2005 by Lemercier et al. [54], who studied complexes
transition. This means, for instance, in the case of the two of the type [Fe(trim)2]X2 (X = F, Cl, Br, I, trim = 4-(4-imidazolyl-
complexes with two-step transition, the calculated energies were in methyl)-2-(2-imidazolylmethyl)imidazole) by calculations with
agreement with the relation 2ELS,HS < EHS,HS + ELS,LS. Recently Verat periodic boundary conditions using LDA and atom centred basis
et al. [109] confirmed the usefulness of this approach by DFT functions. Full geometry optimizations were performed starting
calculation for a newly synthesized binuclear SCO complex. from crystal structures. The transition temperature of the
Zueva et al. [110,111] studied ten different tetranuclear SCO [Fe(trim)2]X2 is increasing with the size of the out-of-sphere anion
complexes in the form of [2×2] molecular grids with cyanides and X, ranging from T1/2 < 0 K (HS ground state) to 380 K, a trend that
with extended organic ligands as bridges. According to their was correctly reproduced by the LDA calculations. The success of
calculations with the PBE functional, the nature of the spin the simple LDA functional indicates that also in crystals the choice
transition depends sensitively on the rigidity or flexibility of the of the functional is less critical if similar complexes are compared,
ligands bridging the four iron centres, whereas the exchange as has been discussed above for isolated molecules.
interaction between the centres is of minor importance for the spin A similar approach with full geometry optimization was applied
transition. by Chumakov et al. [119] to a SCO polymer (using the PBE
Computationally efficient approaches to treat SCO complexes in functional instead of LDA) and by Matar and Letard [92] to
crystals are calculations for polynuclear fragments or for isolated [Fe(phen)2(NCS)2] (using plane waves instead of atom centred
molecules in the Madelung field of the crystal. The latter approach basis functions). In the latter study the results for the solid state
was taken by Le Guennic, Kepenekian, Borshch and Robert were compared in detail with calculations for the isolated
[112,113] who performed CASPT2 and CASSCF calculations for a molecule.
prussian blue analogue and for a series of quite similar Fe(II)N 6 Together with periodic DFT calculations also the Hubbard
SCO compounds that exhibit spin transitions with largely varying correction found usage for electronic structure calculations on SCO
amount of cooperativity (abrupt transition with or without compounds. The disadvantage of the DFT+U method, the loss of
hysteresis or smooth transition). Intermolecular interactions were predictive power by the introduction of the empirical parameter U,

Submitted to the European Journal of Inorganic Chemistry 9


is at the same time an advantage, since this parameter allows to spin, an intermediate-spin and two different high-spin phases that
fine tune the electronic energy difference ΔEHL. The first are characterized by pressure and temperature. A crucial test for the
applications of DFT+U to SCO compounds were published by predictive power of this method will be the experimental search for
Kabalan et al. [120] and by Lebègue et al. [121]. In the latter study those transitions that have not already been observed.
calculations for [Fe(phen)2(NCS)2] and a similar SCO complex
were presented that reached good agreement between experimental
and theoretical results for U = 2.5 eV. The calculations for Conclusions
[Fe(phen)2(NCS)2] were performed for supercells based on X-ray
structures at 15 K. A recent DFT+U study for the same compound The past years have seen considerable progress in electronic
by Bučko et. al. [14] emphasized the importance of van der Waals structure calculations. This progress is due to both the considerable
interactions between SCO complexes. In order to obtain good increase of computer power and the improved and newly
agreement of calculated and experimental lattice parameters, developed methods. However, spin-crossover complexes are still a
dispersion and Hubbard correction turned out to be necessary. challenge for quantum chemistry and there is still no method
While most of the periodic DFT studies have been performed available that would be able to calculate all properties of interest. It
with pure density functionals, recently few studies [105,122,123] is safe to predict that also in the near future a bundle of quantum
were published that employ hybrid functionals. Middlemiss et al. chemical methods has to be applied to get a comprehensive
[105,122] investigated thermally driven spin crossover in prussian theoretical description of a given spin-crossover complex.
blue analogues with GGA functionals and varying admixture of HF Ab initio methods like CASPT2 and CCSD(T) are currently the
exchange. For CsFe[Cr(CN)6], which had been studied earlier by most accurate tools of quantum chemistry that may be applied to
ab initio calculations for the [Fe(NC)6] 4- core combined with a models or even to small SCO complexes. The adiabatic energy
Madelung field [112], best agreement between calculated and differences calculated with these methods serve as reference values
measured transition temperatures was obtained with an admixture for all other methods. Unfortunately, these ab initio methods are
of HF exchange of 14 %, which is quite close to the corresponding computationally so demanding that they are limited to small
value for the reparameterized B3LYP* hybrid functional [58]. The molecules and cannot be used for the optimization of geometries,
phonon spectrum and its contribution to the entropy difference was the calculation of vibrations or the treatment of cooperative effects
obtained from calculations for a supercell. A similar approach was in crystals. This situation may change somewhat due to two new
chosen by Wojdeł et al. [123] who used GGA+U method together promising developments, the embedding of CCSD into the local
with an admixture of HF exchange. pair natural orbital concept by Neese et al. [39,40,41] and the
Any type of electronic structure calculation needs an initial density matrix renormalization group (DMRG) algorithm as
geometry for the molecule or crystal under study. Although this demonstrated by Marti, Ondik, Moritz and Reiher [43].
geometry may be relaxed using molecular dynamics or energy Density functional methods are computationally less demanding
minimization, the quality of the initial geometry is not without and they will stay the methods of choice for all types of
importance. No problem arises, if X-ray structures are available, calculations that are too demanding for ab initio methods. Detailed
but for the LS state this is infrequently the case, and for polymeric investigations of the role of exact exchange for the calculation of
SCO complexes almost never. While for small complexes a the adiabatic energy difference ΔEHL have led to the development
reasonable guess for the geometry of the isolated molecule is of improved hybrid functionals like B3LYP* or TPSSh that give
usually easy to obtain, a starting structure for a crystalline structure fairly good results for ΔEHL. Whether the success of these
is much more difficult to get. If no X-ray structures are available functionals will in future be surpassed by double-hybrid
the danger is high not to find the true structure via geometry functionals is not already clear. The computational demands of the
optimization. For instance, Sarkar et al. [124] observed that the double-hybrid functionals can be kept on an acceptable level using
optimized geometry depends sensitively on the method. For a approximations like density fitting [49,50].
bimetallic polymeric SCO compound they obtained an Despite of the success of hybrid functionals there are still good
intermediate-spin geometry with a GGA functional, while the use reasons to search for improved pure density functionals, especially
of a GGA+U method led to the correct HS geometry. In two in view of solid state calculations. Meta-functionals, that depend
studies Jeschke et al. [125,126] demonstrated the success of a two- also on the kinetic energy density, are promising candidates. The
step strategy consisting of a preceding geometry optimization with failure of all current functionals to give a proper description of the
a classical force field and a subsequent structure relaxation using spin density leaves room for the speculation that functionals that
Car-Parrinello molecular dynamics with a GGA functional and a depend on the charge density as well as on the spin density could
PAW basis. In this studies on polymeric SCO compounds magnetic give better estimates for the spin state splitting.
exchange and elastic coupling were found to be equally important Experimental results and ab initio calculations for suitably
for the cooperativity of the spin transition. chosen fragments of polymeric SCO complexes can be chosen as
references, that are used to find a suitable Hubbard constant U that
allows to fit the reference results with DFT+U calculations. The
Pressure induced SCO same type of calculations for slightly modified SCO may then be
used to predict transition temperatures, hysteresis widths or the
Studies with electronic structure calculations that are able to occurrence of the LIESST and the reverse LIESST effect. The
reproduce pressure induced spin transitions are extremely rare and methods described here may become useful parts of a quantum
have been published only very recently [127,128]. Hsu et al. chemical toolbox that may once be used for an in silico design of
investigated (Mg,Fe)SiO3 perovskite (the most abundant mineral of photo switchable materials.
the lower mantle of the Earth's interior) with a DFT+U approach
together with plane waves and pseudopotentials. After full
geometry optimization they predict a transition pressure for site B Acknowledgements
in the range from 29 to 70 GPa - depending on the choice of
functional (LDA or GGA) and Hubbard constant (empirical value This work has been supported by the German Federal Ministry of
Education and Research under contract number 05 K10UKA.
4.5 eV or self-consistent Usc). Tarafder et al. [128] applied a similar
____________
method to the polymeric SCO compound Fe2[Nb(CN)8]·(4-
[1] P. Gütlich, H. A. Goodwin (Eds.) Spin crossover in transition metal
pyridinealdoxime) 8·2H2O and predicted transitions between a low-
compounds, Springer-Verlag, Berlin, 2004, vol. 1-3.

Submitted to the European Journal of Inorganic Chemistry 10


[2] a) J. N. Harvey Annu. Rep. Prog. Chem., Sect. C: Phys. Chem., [31] A. Fouqueau, M. E. Casida, L. M. Lawson, A. Hauser, F. Neese J.
2006, 102, 203-226; b) M. Reiher Chimia, 2009, 63, 140-145; c) M. Chem. Phys., 2005, 122, 044110/1-13.
Swart Int. J. Quantum Chem., 2013, 113, 2-7.
[32] K. Pierloot, S. Vancoillie J. Chem. Phys., 2006, 125, 124303.
[3] H. Paulsen, A. X. Trautwein Top. Curr. Chem. 2004, 235, 197-219.
[33] M. Kepenekian, V. Robert, B. Le Guennic, C. de Graaf J. Comput.
[4] H. Paulsen, L. Duelund, A. Zimmermann, F. Averseng, M. Gerdan, Chem., 2009, 30, 2327-2333.
H. Winkler, H. Toftlund, A. X. Trautwein Monatsh. Chem., 2003,
134, 295-306. [34] A. Domingo, M. Angels Carvajal, C. de Graaf Int. J. Quantum
Chem., 2010, 110, 331-337.
[5] H. Paulsen, H. Grünsteudel, W. Meyer-Klaucke, M. Gerdan, H.
Winkler, H. Toftlund, A. X. Trautwein Eur. Phys. J. B, 2001, 23, [35] B. Ordejón, C. de Graaf, C. Sousa J. Am. Chem. Soc., 2008, 130,
463-472. 13961-13968.

[6] J. A. Wolny, H. Paulsen, H. Winkler, A. X. Trautwein, J.-P. [36] C. de Graaf, C. Sousa Chem. Eur. J., 2010, 16, 4550-4556.
Tuchagues Hyperfine Interact., 2006, 169, 1389-1392. [37] C. de Graaf, C. Sousa Int. J. Quantum Chem., 2011, 111, 3385-3393.
[7] S. Bonhommeau, T. Guillon, L. M. Lawson Daku, P. Demont, J. [38] N. Suaud, M.-L. Bonnet, C. Boilleau, P. Labeguerie, N. Guihery J.
Sanchez Costa, J.-F. Letard, G. Molnar, A. Bousseksou Angew. Am. Chem. Soc., 2009, 131, 715–722.
Chem. Int. Ed., 2006, 45, 1625-1629.
[39] F. Neese, F. Wennmohs, A. Hansen J. Chem. Phys., 2009, 130,
[8] H. Paulsen, L. Duelund, H. Winkler, H. Toftlund, A. X. Trautwein 114108.
Inorg. Chem., 2001, 40, 2201-2203.
[40] F. Neese, A. Hansen, D. G. Liakos J. Chem. Phys., 2009, 131,
[9] M. Reiher Inorg. Chem., 2002, 41, 6928-6935. 064103.
[10] O. Salomon, M. Reiher, B. A. Hess J. Chem. Phys., 2002, 117, 4729- [41] A. Hansen, D. G. Liakos, F. Neese J. Chem. Phys., 2011, 135,
4737. 214102.
[11] H. Paulsen, A. X. Trautwein J. Phys. Chem. Solids, 2004, 65, 793- [42] G. Fano, F. Ortolani, L. Ziosi J. Chem. Phys., 1998, 108, 9246.
798.
[43] K. H. Marti, I. M. Ondik, G. Moritz, M. Reiher J. Chem. Phys.,
[12] J. J. McGarvey, I. Lawthers J. Chem. Soc., Chem. Comm., 1982, 906. 2008, 128, 014104.
[13] S. Decurtins, P. Gütlich, C. P. Köhler, H. Spiering, A. Hauser Chem. [44] J. C. Slater Phys. Rev., 1951, 81, 385-390.
Phys. Lett., 1984, 105, 1.
[45] P. Hohenberg, W. Kohn Phys. Rev., 1964, 136, B864.
[14] T. Bučko, J. Hafner, S. Lebegue, J. G. Angyan Phys. Chem. Chem.
Phys., 2012, 14, 5389-5396. [46] W. Kohn, L. J. Sham Phys. Rev., 1965, 140, A1133.
[15] a) H. Spiering Top. Curr. Chem., 2004, 235, 171; b) J. Kröber, J. P. [47] K. Schwarz Phys. Rev. B, 1972, 5, 2466-2468.
Audière, R. Claude, E. Codjovi, O. Kahn, J. G. Haasnoot, F.
Grolière, F. Jay, A. Bousseksou, J. Linares, F. Varret, A. Gonthier- [48] S. Grimme J. Comp. Chem., 2006, 27, 1787-1799.
Vassal Chem. Mater., 1994, 6, 1404; c) J. Linares, H. Spiering, F. [49] F. Neese, F. Wennmohs, A. Hansen, U. Becker Chem. Phys., 2009,
Varret Eur. J. Phys. B, 1999, 10, 271. 356, 98.
[16] C. P. Slichter, H. G. Drickamer J. Chem. Phys., 1972, 56, 2142. [50] S. Ye, F. Neese Inorg. Chem., 2010, 49, 772-774.
[17] C. Cramer Essentials of computational chemistry, John Wiley & [51] J. Hubbard Proc. R. Soc. Lond. A, 1963, 276, 238–257.
sons, Chichester, 2004.
[52] S. Zein, S. A. Borshch, P. Fleurat-Lessard, M. E. Casida, H.
[18] F. Jensen Introduction to computational chemistry, John Wiley & Chermette J. Chem. Phys., 2007, 126, 014105.
sons, Chichester, 2006.
[53] C. Rong, S. Lian, D. Yin, B. Shen, A. Zhong, L. Bartolotti, S. Liu J.
[19] a) R. G. Parr Density-functional theory of atoms and molecules, Chem. Phys., 2006, 125, 174102.
Oxford university press, Oxford, 1989; b) R. W. Nalewajski (Ed.)
Density Functional Theory, Springer, Berlin, 1996, vol. 1-5; c) W. [54] G. Lemercier, N. Brefuel, S. Shova, J. A. Wolny, F. Dahan, M.
Koch, M. C. Holthausen A Chemist's Guide to Density Functional Verelst, H. Paulsen, A. X. Trautwein, J.-P. Tuchagues Chem. Eur. J.,
Theory: An Introduction, Wiley-VCH, Weinheim, 2000. 2006, 12, 7421-7432.
[20] N. Kaltsoyannis, J. E. McGrady (Eds.) Principles and Applications [55] J. N. Harvey Struct. Bond., 2004, 112, 151-183.
of Density Functional Theory in Inorganic Chemistry,
Springer-Verlag, Berlin, 2004, vol. 1-2. [56] S. Zein, G. S. Matouzenko, S. A. Borshch Chem. Phys. Lett., 2004,
397, 475-478.
[21] C. J. Cramer, D. G. Truhlar Phys. Chem. Chem. Phys., The Royal
Society of Chemistry, 2009, 11, 10757-10816. [57] A. Bannwarth, S. O. Schmidt, G. Peters, F. D. Sönnichsen, W.
Thimm, R. Herges, F. Tuczek Eur. J. Inorg. Chem., 2012, 2776–
[22] C. R. Jacob, M. Reiher Int. J. Quantum Chem., 2012, 112, 3661- 2783.
3684.
[58] M. Reiher, O. Salomon, B. A. Hess Theor. Chem. Acc., 2001, 107,
[23] F. Neese Coord. Chem. Rev., 2009, 253, 526-563. 48-55.
[24] B. O. Roos, P. R. Taylor, P. E. M. Siegbahn Chem. Phys., 1980, 48, [59] Y. Shiota, D. Sato, G. Juhasz, K. Yoshizawa J. Phys. Chem. A, 2010,
157. 114, 5862-5869.
[25] B. O. Roos, P. R. Taylor, P. E. M. Siegbahn Chem. Phys., 1980, 48, [60] Y. Qu Int. J. Quantum Chem., 2012, DOI: 10.1002/qua.24040.
157.
[61] G. Brewer, M. J. Olida, A. M. Schmiedekamp, C. Viragh, P. Y.
[26] a) K. Andersson, P.-Å. Malmqvist, B. O. Roos, A. J. Sadlej, K. Zavalij Dalton Trans., 2006, 5617-5629.
Wolinski J. Phys. Chem., 1990, 94, 5483; b) K. Andersson, P.-Å.
Malmqvist, B. O. Roos J. Chem. Phys., 1992, 96, 1218. [62] K. P. Jensen, J. Cirera J. Phys. Chem. A, 2009, 113, 10033-10039.

[27] J. D. Watts, J. Gauss, R. J. Bartlett J. Chem. Phys., 1983, 98, 8718. [63] J. Cirera, F. Paesani Inorg. Chem., 2012, 51, 8194-8201.

[28] R. Akesson, L. G. M. Petterson, M. Sandström, U. Wahlgren J. Am. [64] K. Boguslawski, C. R. Jacob, M. Reiher J. Chem. Theory Comput.,
Chem. Soc., 1994, 116, 8691-8704. 2011, 7, 2740-2752.

[29] H. Bolvin J. Phys. Chem. A, 1998, 102, 7525. [65] M. Swart J. Chem. Theory Comput., 2008, 4, 2057-2066.

[30] A. Fouqueau, S. Mer, M. E. Casida, L. M. L. Daku, A. Hauser, T. [66] N. C. Handy, A. J. Cohen Mol. Phys., 2001, 99, 403.
Mineva, F. Neese J. Chem. Phys., 2004, 120, 9473-9486.

Submitted to the European Journal of Inorganic Chemistry 11


[67] a) A. D. Becke J. Chem. Phys., 1993, 98, 5648; b) A. D. Becke Phys. [93b] G. Brehm, M. Reiher, B. Le Guennic, M. Leibold, S. Schindler, F.
Rev. A, 1988, 38, 3098; c) C. Lee, W. Yang, R. G. Parr Phys. Rev. B, W. Heinemann, S. Schneider J. Raman Spectrosc., 2006, 37, 108-
1988, 37, 785. 122.
[68] S. Grimme J. Chem. Phys., 2006, 124, 34108. [94] Y. Garcia, H. Paulsen, V. Schünemann, A. X. Trautwein, J. A.
Wolny Phys. Chem. Chem. Phys., 2007, 9, 1194-1201.
[68a] C. Lee, W. Yang, R. G. Parr Phys. Rev B, 1988, 37, 785-789.
[95] M. Nakano, G. Matsubayashi, T. Matsuo Adv. Quant. Chem., 2003,
[69] a) J. P. Perdew, K. Burke, M. Ernzerhof Phys. Rev. Lett., 1996, 77, 44, 617-30.
3865; b) J. P. Perdew, K. Burke, M. Ernzerhof Phys. Rev. Lett.,
1997, 78, 1396E. [96] M. M. N. Wolf, R. Groß, C. Schumann, J. A. Wolny, V.
Schünemann, A. Døssing, H. Paulsen, J. J. McGarvey R. Diller
[70] a) J. Tao, J. P. Perdew, V. N. Staroverov, G. E. Scuseria Phys. Rev. Phys. Chem. Chem. Phys., 2008, 10, 4264-4273.
Lett., 2003, 91, 146401; b) J. P. Perdew, J. Tao, V. N. Staroverov, G.
E. Scuseria J. Chem. Phys., 2004, 120, 6898–6911. [97] C. Brady, P. L. Callaghan, Z. Ciunik, C. G. Coates, A. Døssing, A.
Hazell, J. J. McGarvey, S. Schenker, H. Toftlund, A. X. Trautwein,
[70a] S. H. Vosko, L. Wilk, M. Nusair Can. J. Phys., 1980, 58, 1200-1211. H. Winkler, J.A.Wolny Inorg. Chem., 2004, 43, 4289-4299.
[71] T. Guillon, L. Salmon, G. Molnar, S. Zein, S. Borshch, A. [98] J. A. Wolny, H. Paulsen, J. J. McGarvey, R. Diller, V. Schünemann,
Bousseksou J. Phys. Chem. A, 2007, 111, 8223-8228. H. Toftlund Phys. Chem. Chem. Phys., 2009, 11, 7562-7575.
[72] L. M. Lawson Daku, A. Vargas, A. Hauser, A. Fouqueau, M. E. [99] J. A. Wolny, S. Rackwitz, K. Achterhold, Y. Garcia, K. Muffler, A.
Casida ChemPhysChem, 2005, 6, 1393-1410. D. Naik, V. Schünemann Phys. Chem. Chem. Phys., 2010, 12,
[73] G. Ganzenmuller, N. Berkaine, A. Fouqueau, M. E. Casida, M. 14782.
Reiher J. Chem. Phys., 2005, 122, 234321. [100] J. A. Wolny, S. Rackwitz, K. Achterhold, K. Muffler, V.
[74] F. Neese J. Chem. Phys., 2003, 119, 9428. Schünemann Hyperfine Int., 2012, 204, 129.

[75] K. Pierloot, S. Vancoillie J. Chem. Phys., 2008, 128, 034104. [101] S. Rackwitz, J. A. Wolny, K. Muffler , K. Achterhold , R. Rüffer , Y.
Garcia, R. Diller, V. Schünemann Phys. Chem. Chem. Phys., 2012,
[75a] K. Boguslawski, K. H. Marti, Ö. Legeza, M. Reiher J. Chem. Theory 14, 14650.
Comp., 2012, 8, 1970-1982.
[102] J. A. Wolny, K. Muffler, H.-J. Krüger, S. Reh, H. Kelm, A. I.
[76] J. Conradie, A. Ghosh J. Phys. Chem. B, 2007, 111, 12621. Chumakov, V. Schünemann Hyperfine Int., 2012, 204, 63.
[77] H. Paulsen, J. A. Wolny, A. X. Trautwein Monatsh. Chem., 2005, [103] L. H. Böttger, A. I. Chumakov, M. Grunert, P. Gütlich, J. Kusz, H.
136, 1107-1118. Paulsen, U. Ponkratz, R. Rüffer, V. Rusanov, A. X. Trautwein, J. A.
Wolny, Chem. Phys. Lett., 2006, 429, 189-193.
[78] M. Güell, M. Sola, M. Swart Polyhedron, 2010, 29, 84-93.
[104] A. X. Trautwein, H. Paulsen, H. Winkler, H. Giefers, G. Wortmann,
[79] R. J. Deeth, A. E. Anastasi, M. J. Wilcockson J. Am. Chem. Soc., H. Toftlund, J. A. Wolny, A. I. Chumakov, O. Leupold J. Phys.:
2010, 132, 6876-6877. Conf. Ser., 2010, 217, 012125.
[80] M. Swart, M. Guella, M. Sola Phys. Chem. Chem. Phys., 2011, 13, [104a]J. Li, Q. Peng, A. Barabanschikov, J. W. Pavlik, E. E. Alp, W.
10449–10456. Sturhahn, J. Zhao, J. T. Sage, W. R. Scheidt Inorg. Chem., 2012, 51,
[81] E. Runge, E. K. U. Gross Phys. Rev. Lett., 1984, 52, 997–1000. 11769-11778.

[82] I. Respondek, L. Bressel, P. Saalfrank, H. Kämpf, A. Grohmann [105] D. S. Middlemiss, D. Portinari, C. P. Grey, C. A. Morrison, C. C.
Chem. Phys., 2008, 347, 514-522. Wilson Phys. Rev. B, 2010, 81, 184410.

[83] H. Ando, Y. Nakao, H. Sato, S. Sakaki Dalton Trans., 2010, 39, [106] a) V. Anisimov, I. Solovyev, M. Korotin, M. Czyzyk, G. Sawatzky
1836-1845. Phys. Rev. B, 1993, 48, 16929; b) A. Liechtenstein, V. Anisimov, J.
Zaanen, Phys. Rev. B, 1995, 52, R5467.
[84] C. Bressler, C. Milne, V. T. Pham, A. ElNahhas, R. M. van der
Veen, W. Gawelda, S. Johnson, P. Beaud, D. Grolimund, M. Kaiser, [107] B. Le Guennic, G. S. Matouzenko, S. A. Borshch Eur. J. Inorg.
C. N. Borca, G. Ingold, R. Abela, M. Chergui Science, 2009, 323, Chem., 2008, 3020-3023.
489. [108] S. Zein, S. A. Borshch J. Am. Chem. Soc., 2005, 127, 16197-16201.
[85] B. E. Van Kuiken, M. Khalil J. Phys. Chem. A, 2011, 115, 10749- [109] A. Y. Verat, N. Ould-Moussa, E. Jeanneau, B. Le Guennic, A.
10761. Bousseksou, S. A. Borshch, G. S. Matouzenko Chem. Eur. J., 2009,
[85a] M. P. Marshak, M. B. Chambers, D. G. Nocera Inorg. Chem., 2012, 15, 10070-10082.
51, 11190-11197. [110] E. M. Zueva, E. R. Ryabikh, A. M. Kuznetsov, S. A. Borshch Inorg.
[86] J.-P. Tuchagues, A. Bousseskou, G. Molnar, J. J. McGarvey, F. Chem., 2011, 50, 1905-1913.
Varret Top. Curr. Chem., 2004, 235, 85, and ref. therein. [111] E. M. Zueva, E. R. Ryabikh, S. A. Borshch Inorg. Chem., 2011, 50,
[87] J. A. Wolny, R. Diller, V. Schünemann Eur. J. Inorg. Chem., 2012, 11143-11151.
2635-2648. [112] B. Le Guennic, S. Borshch, V. Robert Inorg. Chem., 2007, 46,
[88] J. A. Wolny, H. Paulsen, A. X. Trautwein, V. Schünemann Coord. 11106–11111.
Chem. Rev., 2009, 253, 2423-2431. [113] M. Kepenekian, B. L. Guennic, V. Robert J. Am. Chem. Soc., 2009,
[89] F. König, K. Madeja Inorg. Chem., 1967, 6, 48. 131, 11498-11502.

[90] K. L. Ronayne, H. Paulsen, A. Höfer, A. C. Dennis, J. A. Wolny, A. [114] S. Zein, G. S. Matouzenko, S. A. Borshch J. Phys. Chem. A, 2005,
I. Chumakov, V. Schünemann, H. Winkler, H. Spiering, A. 109, 8568-8571.
Bousseksou, P. Gütlich, A. X. Trautwein, J. J. McGarvey Phys. [115] G. S. Matouzenko, M. Perrin, B. L. Guennic, C. Genre, G. Molnar,
Chem. Chem. Phys., 2006, 8, 4685-4693. A. Bousseksou, S. A. Borshch Dalton Trans., 2007, 934-942.
[91] G. Baranovic, D. Babic Spectrochimica Acta, Part A: Molecular [116] A. Vargas, A. Hauser, L. M. Lawson Daku J. Chem. Theory
Spectroscopy and Biomolecular Spectroscopy, 2004, 60, 1013-1025. Comput., 2009, 5, 97-115.
[92] S. F. Matar, J.-F. Letard Z. Naturforsch. B, 2010, 65b, 565-570. [117] R. L. Lord, F. A. Schultz, M.-H. Baik J. Am. Chem. Soc., 2009, 131,
[93] V. K. Palfi, T. Guillona, H. Paulsen, G. Molnar, A. Bousseksou C. 6189-6197.
R. Chimie, 2005, 8, 1317-1325. [118] R. J. Deeth, N. Fey J. Comp. Chem., 2004, 25, 1840-1848.
[93a] G. Brehm, M. Reiher, S. Schneider J. Phys. Chem. A, 2002, 106, [119] Y. Chumakov, G. S. Matouzenko, S. A. Borshch, A. Postnikov,
12024-12034. Polyhedron, 2009, 28, 1955-1957.

Submitted to the European Journal of Inorganic Chemistry 12


[120] L. Kabalan, S. F. Matar, M. Zakhour, J. F. Letard Z. Naturforsch., [128] K. Tarafder, S. Kanungo, P. M. Oppeneer, T. Saha-Dasgupta Phys.
2008, 63b, 154-160. Rev. Lett., 2012, 109, 077203-077207.
[121] S. Lebegue, S. Pillet, J. G. Ángyán Phys. Rev. B, 2008, 78, 024433.
[122] D. S. Middlemiss, C. C. Wilson Phys. Rev. B, 2008, 77, 155129.
[123] J. C. Wojdeł, I. de P. R. Moreira, F. Illas J. Chem. Phys., 2009, 130,
014702. Received:
Published online:
[124] S. Sarkar, K. Tarafder, P. M. Oppeneer, T. Saha-Dasgupta J. Mater.
Chem., 2011, 21, 13832-13840.
[125] H. O. Jeschke, L. A. Salguero, B. Rahaman, C. Buchsbaum, V.
Pashchenko, M. U. Schmidt, T. Saha-Dasgupta, R. Valentí New J.
Phys., 2007, 9, 448.
[126] H. O. Jeschke, L. A. Salguero, R. Valentí, C. Buchsbaum, M. U.
Schmidt, M. Wagner C. R. Chimie, 2007, 10, 82-88.
[127] H. Hsu, P. Blaha, M. Cococcioni, R. M. Wentzcovitch Phys. Rev.
Lett., 2011, 106, 118501.

Submitted to the European Journal of Inorganic Chemistry 13


Entry for the Table of Contents

Post-Hartree-Fock ab initio and density


functional methods have made Hauke Paulsen,* Volker Schünemann
considerable progress in the last years and Juliusz A. Wolny
and can be used to obtain a fairly
accurate description of the challenging Progress in Electronic Structure
electronic structure of spin-crossover Calculations on Spin-Crossover
complexes. A suitable combination of Complexes
these methods may aid the design of
new photo switchable materials with Keywords: Ab initio calculations /
favourable properties. Computational chemistry / Density
functional calculations / Quantum
chemistry / Spin crossover

Submitted to the European Journal of Inorganic Chemistry 14

You might also like