Journal of Molecular Modeling (2019) 25: 38

https://doi.org/10.1007/s00894-019-3924-8

ORIGINAL PAPER

Segmented all-electron basis sets of triple zeta quality

for the lanthanides: application to structure calculations of lanthanide
monoxides
A. Z. de Oliveira 1 & I. B. Ferreira 1 & C. T. Campos 1,2 & F. E. Jorge 1,3 & P. A. Fantin 1,4

Received: 24 September 2018 / Accepted: 3 January 2019 / Published online: 15 January 2019
# Springer-Verlag GmbH Germany, part of Springer Nature 2019

Abstract
Nonrelativistic and relativistic (Douglas-Kroll-Hess, DKH) segmented all-electron Gaussian basis sets of valence triple zeta
quality plus polarization functions (TZP) for the lanthanides were developed. As some atomic and molecular properties depend
on a good description of the electrons far from the nuclei, these basis sets are augmented with diffuse functions, giving rise to the
augmented TZP (ATZP) and ATZP-DKH basis sets. At the DKH level of theory, the B3LYP hybrid functional in conjunction with
the TZP-DKH basis set were used to calculate the atomic charges and valence orbital populations of the lanthanide and oxygen
atoms, the bond lengths, and the equilibrium dissociation energies of lanthanide monoxides. The DKH-B3LYP/ATZP-DKH
polarizability of Yb and the DKH-M06/TZP-DKH first ionization energies of the lanthanides are also reported. Compared with
the values obtained with a larger all-electron basis set, and theoretical and experimental data found in the literature, data obtained
by our compact basis sets are verified to be accurate and reliable. Unlike effective core potential valence basis sets, our basis sets
can also be employed in molecular property calculations that involve the simultaneous treatment of core and valence electrons.

Keywords TZ basis sets . Lanthanides . DKH2-B3LYP calculations . Lanthanide monoxides

Introduction theoretical methods but also for experimental techniques.

This difficulty is attributed to the 4f, 5d, and 6s valence or-
It is well known that lanthanides have peculiar physical and bitals of the lanthanides, which are partially filled, generating
chemical properties. Consequently, there is a growing interest a large number of degenerate and/or nearly degenerate elec-
in studying lanthanide compounds and their applications in tronic states in a given system. This degeneracy complicates
many areas [1–7]. These applications include catalysis, mo- the use of spectroscopic techniques to carry out measure-
lecular magnetism, constructions of cathode ray and optical ments. Theoretically, the degeneracy increases the number of
fibers, and therapeutic use. The study of the electronic struc- orbitals and electrons to be included in the calculations. In
tures of lanthanide compounds is a challenge not only for addition to these difficulties, one must simultaneously consid-
er the electronic correlation and the relativistic effects in the
calculations [8]. To overcome these problems, some theoreti-
cal approaches have been proposed. One such approach is the
* F. E. Jorge
Jorge@cce.ufes.br
effective core potential (ECP), which takes into account only
valence electrons with the purpose of reducing computational
1
time. Relativistic ECP, along with valence basis sets, has pro-
Departamento de Física, Universidade Federal do Espírito Santo, duced reliable results such as geometries and relative energies
Vitória, ES 29060-900, Brazil
2
[9–12]. However, as it is not easy to define the valence elec-
Instituto Federal do Espírito Santo, Campus Aracruz, Avenida trons for the lanthanides, several ECPs are reported in the
Morobá, 248, Bairro Morobá, Aracruz, ES 29192-733, Brazil
3
literature. Other deficiencies of ECPs have also been pointed
Unidade Acadêmica de Física, Universidade Federal de Campina out in the literature [13–15]. Certainly, ECPs cannot be
Grande, Campina Grande, PB 58429-900, Brazil
4
employed in calculations of properties involving inner shell
Instituto Federal do Espírito Santo, Campus Guarapari, Alameda electrons, such as EPR, Mössbauer and X-ray spectra and
Francisco Vieria Simoes, 720, Aeroporto, Guarapari, ES 29216-795,
Brazil electron density topological analysis. In these cases, all-
38 Page 2 of 9 J Mol Model (2019) 25: 38

electron basis sets with a relativistic Hamiltonian must be reported. In the past, basis sets of similar quality for the ele-
used. We recall that the Douglas-Kroll-Hess (DKH) method ments H-La and Hf-Ac were generated by our research group
[16–18] considers most scalar relativistic effects. Care must be [31–39], and have been used extensively in electronic structure
taken in calculations of properties like the electric field gradi- calculations of atoms, molecules, and clusters (see, e.g.,
ent at a nucleus with two-component relativistic methods such [40–45]). Using the DZH-B3LYP/TZP-DKH procedure, atom-
as DKH. Total energy derivatives do not lead to a picture- ic charges and valence orbital populations of the lanthanide and
change error [19, 20], whereas a simple calculation of the oxygen atoms, equilibrium bond lengths, and equilibrium dis-
expected value from the two-component wave functions or sociation energies of the lanthanide monoxides (LnO) were
from the two-component Kohn-Sham orbitals does. calculated and compared with theoretical and experimental da-
Different approaches to eliminate most picture-change errors ta. The static mean dipole polarizability of Yb and the first
have been proposed (see [21] and references therein). ionization energies of the lanthanides are computed as well.
For the lanthanides, some all-electron basis sets have been
published in the literature, for example, the Slater-type basis
sets to be used with the zero-order regular approximation Basis set determination
(ZORA) that were included in the ADF code [22], the atomic
natural orbital (ANO) basis sets reported by Roos et al. [23] to TZ basis set for Ce-Lu
be used with the second-order DKH Hamiltonian (DKH2), and
the sets of Nakajima et al. [24, 25] and of Sekiya et al. [26] To generate the TZ set for the lanthanides, the following steps
generated from the third-order DKH and the relativistic were carried out. (1) The valence space (4f5d6s6p) was cho-
Hamiltonians, respectively. The last three hierarchical se- sen. (2) For each element, the uncontracted basis set
quences of large Gaussian-type-function (GTF) basis sets have (21s15p10d6f) containing 158 functions was optimized for
been applied successfully in ab initio calculations of small the ground-state [Ce (4f15d16s2, 1G), Pr (4f36s2, 4I), Nd
molecules. The hierarchical sequences of basis sets presented (4f46s2, 5I), Pm (4f56s2, 6H), Sm (4f66s2, 7F), Eu (4f76s2, 8S),
by Lu and Peterson [27] and by Gomes et al. [28] using the Gd (4f75d16s2, 9D), Tb (4f96s2, 6H), Dy (4f106s2, 5I), Ho
third-order DKH and the Dirac-Coulomb Hamiltonians, re- (4f116s2, 4I), Er (4f126s2, 3H), Tm (4f136s2, 2F), Yb (4f146s2,
1
spectively, should also be mentioned. As the general contrac- S), and Lu (4f 145d16s2, 2D)] using the minimum total
tion scheme was employed in the construction of some of the Hartree-Fock (HF) energy criterion. In this process, two sub-
basis sets cited above, these basis sets are less efficient than the routines developed by our research group, and included in the
segmented basis sets in density functional theory (DFT) calcu- ATOM-SCF program [46], were used. One subroutine con-
lations of medium-sized and large molecules, where the CPU tains the improved generator coordinate HF method [47],
time is proportional to the two-electron integral number to be while the other generates fully optimized basis sets. Except
computed. Another point worth mentioning is that most quan- for Lu, one more Gaussian function of d symmetry is added to
tum chemistry programs were developed with the segmented the uncontracted basis set with the purpose of better describ-
and not generally contracted basis sets in mind; thus, they do ing the electrons in these orbitals. As the 5d orbitals of the
not efficiently handle the latter contraction scheme. Segmented elements Pr-Eu and Tb-Yb are not occupied in the ground
all-electron basis sets of valence triple zeta quality to be states, the 4fn-15d16s2 excited state electronic configuration
employed with the DFT-DKH2 or DFT-ZORA method and is employed. From Ce through Yb, the 4f n-1 6s 2 6p 1
with the DFT-DKH2 method were generated by Pantazis and (4f146s26p1 for Lu) excited state is used to optimize the
Neese [29] and by Dolg [30], respectively. The uncontracted smallest p-type exponent. (3) Finally, using a subroutine de-
(23s16p12d6f) and contracted [18s12p9d3f] basis sets devel- veloped by us that searches by the best segmented contraction
oped by these authors are of similar sizes and have 173 and 120 scheme obtained from the uncontracted basis set of each ele-
functions, respectively. To carry out calculations of medium- ment, one obtains the TZ set, namely: [9s6p6d3f] for Ce-Yb
sized and large molecules containing one or more lanthanide and [9s6p5d3f] for Lu.
atoms, it is of interest to have more compact segmented all-
electron basis sets, which are still accurate. Thus, the main TZP basis set for Ce-Lu
objective of this article is to generate basis sets that meet these
requirements at the same time. By adding 2g polarization functions to the TZ set of the lan-
In this work, nonrelativistic and DKH segmented all- thanides, one obtains the nonrelativistic TZP basis set. These
electron basis sets of valence triple zeta quality plus polarization functions were optimized using the second-order Møller-
functions (TZP) for the lanthanides are developed. To obtain Plesset perturbation (MP2) theory and the minimum energy
reliable results for electron affinity, polarizabilities, optical ro- criterion. The frozen core approximation available in the
tation, and electronic circular dichroism, augmented TZP and Gaussian 09 code [48] was employed. The number of
TZP-DKH (ATZP and ATZP-DKH) basis sets are also correlated electrons for each element is as follows: Ce (4), Pr
J Mol Model (2019) 25: 38 Page 3 of 9 38

(5), Nd (6), Pm (7), Sm (8), Eu (9), Gd (10), Tb (11), Dy (12), shown in [50]. Throughout the atomic and molecular calcula-
Ho (13), Er (14), Tm (15), Yb (16), and Lu (17). tions, spherical harmonic GTFs are considered.
At the end of this procedure, the TZP basis sets for Ce-Yb A procedure that is not very different from that presented in
[9s6p6d3f2g] and for Lu [9s6p5d3f2g] contain 96 and 91 this work was used to develop the segmented all-electron rel-
functions, respectively. They represent a reduction of one- ativistically contracted (SARC) basis sets for the lanthanides
third of the size of the all-electron contracted sets reported in [29]. Initially, ROHF calculations were carried out to deter-
[29, 30]. Thus, the first aim of this work was achieved. mine the innermost radial expectation values of each atom in
the ground state; they were then employed to obtain the max-
imum exponents of all atomic symmetries. Subsequently, the
ATZP basis set for Ce-Lu
series αl χ-i (i = 1, 2, 3,. . . and χ = 2.20, 2.40, 2.50, and 2.60
for l = s, p, d, and f) was used to generate the uncontracted set.
To describe appropriately electrons away from the nuclei, dif-
Next, the contraction coefficients were determined from the
fuse functions were added to the TZP set. Then, for each s, p, d,
DKH2 and ZORA Hamiltonians.
and f symmetry, one more function is added and optimized
considering the criterion of minimal total HF energy. Next, a
diffuse function of g symmetry was added to the latter basis set,
Results and discussion
before being optimized at the ROMP2 level of theory. It is
important to note that all diffuse functions of each element
To calculate any property of compounds containing lantha-
were optimized considering the ground state energy of the
nides, relativistic effects must be considered. For that reason,
negative ion. As mentioned above, this set is designated
only the TZP-DKH and ATZP-DKH basis sets presented in this
ATZP, and can be used in electron affinity, polarizability, opti-
work and the DKH2 Hamiltonian will be employed hereafter.
cal rotation, and electronic circular dichroism calculations of
The B3LYP functional [51, 52] has been applied successfully
atoms, molecules, and clusters. The maximum number of func-
in studies of electronic structures of atoms, molecules, and clus-
tions per symmetry of the ATZP basis set for the lanthanides is
ters. For example, it was recently used by our research group
[10s7p7d4f3g]. For some elements, it is not possible to gener-
[33, 35] in calculations of spectroscopic constants of molecules
ate diffuse functions for all symmetries, so the augmented sets
containing elements of the fourth and fifth rows of the periodic
are smaller. Details of this and all other sets constructed in this
table. Therefore, this hybrid functional seems a natural choice.
work can be found at the website http://qcgv.ufes.br/.
The equilibrium structures of the lanthanide monoxides were
determined with the DKH2-B3LYP/TZP-DKH model. From
TZP-DKH and ATZP-DKH basis sets for Ce-Lu these structures, equilibrium dissociation energies were calculat-
ed and a natural bond orbital analysis [53] was carried out. At
For heavy elements such as lanthanides, the relativistic effects the DKH2-B3LYP/ATZP-DKH level, the static mean dipole
are large. The DKH method [16–18] is known to consider most polarizability of Yb was also evaluated. For first ionization en-
of the scalar relativistic effects. De Jong et al. [49], carrying out ergy calculations of the lanthanides, the M06 hybrid functional
relativistic calculations of some compounds containing halo- [54] was used because of this large applicability, and because it
gens, verified that, in order to avoid undesirable results, it was is the most versatile of the 06 functionals.
essential to recontract the cc-pVXZ (X = D, T, Q, and 5) basis
sets for H, He, B–Ne, Al–Ar, and Ga–Br with the DKH First ionization energy
Hamiltonian. A similar procedure was employed by Jorge
et al. [33–36] to generate the TZP-DKH basis set from the The DKH2-M06/TZP-DKH, DKH2-B3LYP/UGBS [29], and
TZP basis set for the elements from H to La and from Hf to Ac. experimental [55] first ionization energies for the lanthanides
Using a subroutine coupled to the Gaussian 09 code, which are shown in Table 1.
optimizes contraction coefficients of a basis set, and consider- The agreement between the first ionization energies calcu-
ing the minimum DKH2 energy criterion, the TZP-DKH basis lated with the TZP-DKH and UGBS (34s24p20d14f) basis
set for the lanthanides was developed. It is worth mentioning sets is, in general, very good. The largest difference
that, in this process, the chosen Gaussian function exponents (0.34 eV) occurring for Tb. Overall, the mean absolute devi-
and the segmented contraction schemes of the TZP set are not ation of our results with respect to the benchmark theoretical
altered. Finally, we note that this procedure achieves appro- results [29] is equal to 0.14 eV.
priate relativistic radial changes of the wave function, i.e., the
core and valence orbitals are contracted and expanded, respec- Natural atomic analysis
tively. Adding the diffuse functions of ATZP to TZP-DKH,
the set designed ATZP-DKH is obtained. For any atom of the For the lanthanide monoxides, the atomic charges (qLn) and
periodic table, the DKH2 Hamiltonian variational stability is the valence orbital populations of the lanthanide (nLn) and
38 Page 4 of 9 J Mol Model (2019) 25: 38

Table 1 First ionization energies

(in eV) of the lanthanides DKH2- DKH2- Experimentalc
M06a B3LYPb

Ce: 4f15d16s2 (1G) → Ce+: 4f15d2 (4H) 5.20 5.46 5.54

Pr: f3s2 (4I) → Pr+: f3s1 (5I) 5.34 5.39 5.47
Nd: f4s2 (5I) → Nd+: f4s1 (6I) 5.30 5.45 5.52
Pm: f5s2 (6H) → Pm+: f5s1 (7H) 5.38 5.51 5.58
Sm: f6s2 (7F) → Sm+: f6s1 (8F) 5.34 5.57 5.64
Eu: f7s2 (8S) → Eu+: f7s1 (9S) 5.45 5.62 5.67
Gd: f7d1s2 (9D) → Gd+: f7d1s1 (10D) 6.04 6.07 6.15
Tb: f9s2 (6H) → Tb+: f9s1 (7H) 5.47 5.81 5.86
Dy: f10s2 (5I) → Dy+: f10s1 (6I) 6.04 5.92 5.94
Ho: f11s2 (4I) → Ho+: f11s1 (5I) 5.94 6.01 6.02
Er: f12s2 (3H) → Er+: f12s1 (4H) 6.04 6.11 6.11
Tm: f13s2 (2F) → Tm+: f13s1 (3F) 6.26 6.19 6.18
Yb: f14s2 (1S) → Yb+: f14s1 (2S) 6.10 6.28 6.25
Lu: f14d1s2 (2D) → Lu+: f14s2 (1S) 5.25 5.39 5.43
MADd 0.14 – –
a
This work. TZP-DKH all-electron basis set generated in this work
b
From [29] (the uncontracted universal Gaussain basis set (34s24p20d14f) was used)
c
From [55]
d
Mean absolute deviation with respect to UGBS ionization energies

oxygen (nO) atoms evaluated with the DKH2-B3LYP/TZP- compounds can be thought of as a mixing of ionic and covalent
DKH procedure are displayed in Table 2. characters. Thus, in addition to the ionic character, the valence
Considering that, for the lanthanide monoxides, purely ionic orbitals of the lanthanide and oxygen atoms must form covalent
bonding would have the form Ln2+O2− (predicted by ligand bonds. The oxygen 2s orbital does not participate in these
field theory, see [56] and references therein), the atomic charges chemical bonds. Consequently, the bonding molecular orbitals
listed in Table 2 indicate that the chemical bonds of these are formed mainly from the 4f, 5d, and 6s orbitals of lanthanides
and the 2p orbitals of oxygen. The origin of the σ bonding,
Table 2 Natural bond orbital analysis of the lanthanide monoxides from nonbonding, and antibonding molecular orbitals is due to the
DKH2-B3LYP/TZP-DKH calculations interaction between the lanthanide hybrid orbitals (sdσ and sfσ)
and the oxygen pσ orbital. In addition, the dπ and pπ atomic
qLna nLnb nOc
orbitals interact with one another to form bonding and anti-
CeO 1.08 6s(0.91)4f(1.30)5d(0.82) 2s(1.93)2p(5.13) bonding molecular orbitals. Finally, the dδ and dϕ atomic or-
PrO 1.08 6s(0.29)4f(2.87)5d(0.85) 2s(1.93)2p(5.14) bitals give rise to nonbonding orbitals because there are no
NdO 1.23 6s(0.03)4f(4.11)5d(0.71) 2s(1.95)2p(5.27) valence orbitals in oxygen with these symmetries.
PmO 1.18 6s(0.02)4f(5.10)5d(0.77) 2s(1.94)2p(5.23) To obtain information about the contributions of the lantha-
SmO 1.24 6s(0.03)4f(6.02)5d(0.96) 2s(1.95)2p(5.28) nide valence orbitals to the chemical bonds, a population analysis
EuO 1.30 6s(0.02)4f(7.06)5d(0.67) 2s(1.96)2p(5.33) is necessary. A quick look at Table 2 shows that the 4f popula-
GdO 0.82 6s(0.98)4f(7.96)5d(0.35) 2s(1.96)2p(4.85) tions increase from CeO to LuO. Studies of other lanthanide
TbO 1.36 6s(1.00)4f(8.23)5d(0.28) 2s(1.95)2p(5.36) compounds presented similar trends [29, 56–58]. In general, the
DyO 1.12 6s(0.93)4f(9.08)5d(0.83) 2s(1.92)2p(5.17) 6s orbitals lose one or almost two electrons to the valence orbitals
HoO 0.93 6s(0.95)4f(10.99)5d(0.23) 2s(1.95)2p(4.91) of the lanthanide and oxygen atoms. Consequently, the numbers
ErO 1.20 6s(0.98)4f(11.04)5d(0.80) 2s(1.95)2p(5.25) of electrons in the 6s orbitals decrease to approximately 1 or 0,
TmO 0.77 6s(0.96)4f(12.98)5d(0.30) 2s(1.95)2p(4.74) which can be understood as lone electrons or empty orbitals,
YbO 1.34 6s(0.12)4f(13.98)5d(0.59) 2s(1.96)2p(5.37) respectively. Increases in the occupation numbers of the 4f or-
LuO 1.17 6s(0.95)4f(13.99)5d(0.83) 2s(1.93)2p(5.23) bitals occur for CeO, NdO, PmO, SmO, EuO, and GdO (see
Table 2). The largest value (0. 96 e) occurred for gadolinium
a
Charge (e) of the lanthanide atom monoxide. In contrast, for PrO, TbO, DyO, and ErO, the 4f
b
Populations (e) of the lanthanide valence orbitals orbitals lose electrons. At the DFT level of theory, and using
c
Populations (e) of the oxygen valence orbitals the Mulliken population analysis, Wang and Schwarz [57]
J Mol Model (2019) 25: 38 Page 5 of 9 38

verified that, for GdO, the 4f orbital occupation number increases theoretical approaches overestimate the experimental bond
0.76 e. This finding, along with the bond length and dissociation lengths for up to 0.06 Å (see Table 3), the TZP-DKH errors
energy obtained for this monoxide, led these authors to conclude seem to be normal for lanthanide monoxides.
that the 4f orbitals of Gd participate in the chemical bonds. With It is clear from Table 3 that the DKH2-B3LYP/TZP-DKH
this in mind, a deeper analysis of the DKH2-B3LYP/TZP-DKH bond distances of the lanthanide monoxides do not follow the
natural bond orbitals was carried out; we observed that, for EuO, same trends as the lanthanide and actinide ionic radii [65],
GdO, HoO, YbO, and LuO, the percentage of f character to form which were designated lanthanide and actinide contractions.
the hybrid orbital of Ln is less than 1.1%, whereas for the early This effect is a characteristic of bonds that have a high ionic
lanthanide (Ce-Pm) monoxides, and for TbO, DyO, and TmO, character, such as those of the lanthanide and actinide
the f characters are in the range between 17% and 38%. However, trichlorides [66, 67]. In these compounds, one can note a de-
even for these high values, care must be taken regarding whether crease in the bond lengths from CeCl3 to LuCl3 and from
or not the 4f orbitals participate in chemical bonding. As the DFT ThCl3 to LrCl3 due to the lanthanide and actinide contractions.
is not self-interaction free, and as the self-interaction can be large In contrast, for the lanthanide monoxides, we verify in the
for compact orbitals, the 4f orbital energies are artificially shifted section Natural atomic analysis that the covalent and ionic
up, and can mix strongly with the other valence orbitals. This characters of the chemical bond are comparable.
mixing increases the 4f population and decreases the total energy Consequently, the DKH2-B3LYP/TZP-DKH bond lengths
of the system. Ramakrishnan et al. [59] demonstrated that remov- of these compounds do not exhibit systematic behavior from
al of 4f (Lu) 2p (F) spurious orbital mixing in DFT calculations of CeO to LuO. In fact, two maxima, occurring in EuO and YbO,
LuF3 does not change the values of the molecular properties they can be noted in the bond lengths. To understand this latter
computed. This was done through the inclusion of a Hubbard finding, it is necessary to use the results of Table 2: (1) The
term in the DFT + U methodology, Here, it is important to men- 4f orbitals of the lanthanides are, respectively, half-filled and
tion that this kind of calculation goes beyond the scope of our fully filled, then they do not participate in the chemical bonds.
article. For most lanthanide monoxides, the electron numbers of (2) In fact, the only orbitals that participate are 5d and 6s, with
the 5d orbitals also increase due to relocations of 6s electrons. the largest contribution of 6s to the chemical bonds occurring
Reductions in the 5d occupation numbers occur only for CeO, exactly for EuO and YbO. (3) Consequently, these monoxides
GdO, and LuO. Our results also show that the largest contribution have the highest atomic charges. (4) The 6s orbital radius is
to form the hybrid orbitals of the lanthanide atoms comes from larger than the 4f and 5d radii, which is the main reason why
the 5d orbitals followed by 6s. For some monoxides, a small EuO and YbO have the largest bond lengths.
participation of 6p orbitals is also found. Another point in Table 3 that deserves attention is the reg-
ular increase in the DKH2-B3LYP/TZP-DKH bond distances
Bond distance from HoO to YbO. For these monoxides, the π* and σ* anti-
bonding orbitals are filled. The reduction of the nonbonding 4f
As the lanthanide monoxides include all 4f elements, they can orbital energy for approximately 0.2 Hartree can explain the
be used to test the performance of a proposed new methodol- abrupt decrease in the LuO bond length.
ogy or basis set. However, the amount of experimental In general, the agreement among the corresponding bond
[60–64] and theoretical [56–58, 65] data found in the literature lengths computed with the four different theoretical proce-
for these molecules is still limited due to the various factors dures shown in Table 3 is good. Except for YbO (0.04 Å),
cited in the Introduction. the difference between the TZP-DKH and SARC results does
Table 3 shows the bond distances of LnO calculated not exceed 0.02 Å. It appears that the YbO bond distance is
with the DKH2-B3LYP method along with the TZP- overestimated by SARC, because the experimental value is
DKH and SARC [29] basis sets. The results [56, 58] exactly equal to our calculated value (cf. Table 3). These find-
obtained with the B3LYP functional and the 6–311++ ings show that the accuracy of our basis set in bond length
G(df) basis set for oxygen and the Stuttgard/Dresden calculations is similar to those calculated with larger all-
ECPs in conjunction with the [6s6p5d4f3g] valence ba- electron and ECP valence basis sets.
sis set (104 functions) for the lanthanides, and with the
ZORA-PBE frozen core approximation along with the Dissociation energy
TZ2P basis set of Slater-type-functions, are also includ-
ed in this Table. Theoretical and experimental [60, 64] dissociation energies of
Initially, to assess the accuracy achieved by our results, the lanthanide monoxides are shown in Table 4. With the
comparison with a few experimental values [60, 61] found exception of ErO and TmO, the DKH2- B3LYP/TZP-DKH
in the literature was performed. The difference between the results agree well with the equilibrium dissociation energies
corresponding DKH2-B3LYP/TZP-DKH and experimental evaluated with the DKH2-B3LYP/SARC and ECP-B3LYP
results did not exceed 0.02 Å. Considering that the other [56] procedures. For ErO and TmO, our results are closer to
38 Page 6 of 9 J Mol Model (2019) 25: 38

Table 3 Equilibrium bond

lengths (Å) of the lanthanide Multiplicity DKH2- DKH2- ECP- ZORA- Experimentale
monoxides B3LYPa B3LYPb B3LYPc PBEd

CeO 3 1.814 1.813 1.806 1.814 1.82

PrO 4 1.801 1.805 1.797 1.810 –
NdO 5 1.802 – 1.798 1.823 1.7991f
PmO 6 1.815 1.800 1.804 1.816 –
SmO 7 1.849 1.888 1.851 1.834 –
EuO 8 1.872 1.870 1.884 1.867 1.89
GdO 9 1.825 1.812 1.823 – 1.81
TbO 8 1.800 1.799 1.772 – –
DyO 7 1.775 1.798 1.834 – –
HoO 6 1.785 1.794 1.792 – –
ErO 5 1.787 1.800 1.793 – –
TmO 4 1.796 1.803 1.800 – –
YbO 1 1.813 1.854 1.873 – 1.81
LuO 2 1.781 1.794 1.791 – 1.79
a
This work. TZP-DKH all-electron basis set generated in this work (for Ce-Lu) and in [34] (for O)
b
This work. SARC (for Ce-Lu) and TZP-DKH (for O) all-electron basis sets taken from [29, 34], respectively
c
Results from [56]. The 6–311++G(df) basis set for oxygen and the Stuttgard/Dresden ECPs in conjunction with
the [6s6p5d4f3g] valence basis set for the lanthanides were used
d
Results from [58]. For the lanthanides and oxygen atoms, the frozen core approximation along with the TZ2P
basis set of Slater-type-functions were employed
e
From [60]
f
From [61]

the experimental results than those obtained with the other PBE dissociation energies [58] overestimate the experimental
theoretical approaches. The TZP-DKH set seems to work well values, reaching 1.55 eV for CeO.
in predicting equilibrium dissociation energies of lanthanide From CeO to EuO, the DKH2-B3LYP/TZP-DKH equilibri-
monoxides. It is important to note that, in general, the ZORA- um dissociation energies diminish monotonically, whereas for

Table 4 Dissociation energies

(eV) of the lanthanide monoxides Multiplicity DKH2- DKH2- ECP- ZORA- Experimentale
B3LYPa B3LYPb B3LYPc PBEd

CeO 3 8.08 8.29 8.23 9.78 8.23

PrO 4 6.82 7.40 6.83 8.64 7.74
NdO 5 6.56 – 6.12 7.73 7.33
PmO 6 6.19 6.36 5.90 6.71 –
SmO 7 5.12 5.06 5.12 5.80 5.90
EuO 8 4.50 4.95 4.60 4.32 4.96f
GdO 9 7.74 7.29 6.55 – 7.44f
TbO 8 7.46 6.89 7.05 – 7.30
DyO 7 6.57 5.64 6.75 – 6.25
HoO 6 7.05 6.57 5.12 – 6.39
ErO 5 6.20 3.19 4.84 – 6.30
TmO 4 5.30 4.37 4.06 – 5.76
YbO 1 4.42 3.90 3.37 – 4.11
LuO 2 7.17 6.80 6.20 – 7.04f
a
This work. TZP-DKH all-electron basis set generated in this work (for Ce-Lu) and in [34] (for O)
b
This work. SARC (for Ce-Lu) and TZP-DKH (for O) all-electron basis sets taken from [29, 34], respectively
c
Results from [56]. The 6–311++G(df) basis set for oxygen and the Stuttgard/Dresden ECPs in conjunction with
the [6s6p5d4f3g] valence basis set for the lanthanides were used
d
Results from [58]. For the lanthanides and oxygen atoms, the frozen core approximation along with the TZ2P
basis set of Slater-type-functions were employed
e
From [60]
f
From [64]
J Mol Model (2019) 25: 38 Page 7 of 9 38

GdO a sudden increase is observed. From GdO to DyO, another conjunction with the corresponding basis sets reported previ-
gradual decrease occurs, which is followed by another slight ously for the atoms from H-La and Hf-Lr [31–39, 69].
increase in HoO, and then, a decrease from HoO to YbO oc- Below, we list some conclusions associated with the
curs. Finally, in LuO, the third maximum occurs. The experi- DKH2-B3LYP/TZP-DKH results.
mental data displayed in Table 4 present similar behavior. This
adds credence to the fact that the dissociation energies calculat- – The atomic charges of the lanthanide monoxides indicate
ed with TZP-DKH are reliable. In summary, the dissociation that the chemical bonds of these compounds present cova-
energies of the lanthanide monoxides present two decreases: lent and ionic characteristics. Some general characteristics
one from CeO to EuO and another from GdO to YbO. of the orbital populations of the lanthanides are verified: (1)
To obtain a better understanding of the above findings, the From CeO to LuO, the 4f populations always increase; (2)
stabilities of the molecule and the product as well as the bond in general, the 6s populations are close to 1 or 0; and (3)
strength should be considered. We recall that bond strength, most 5d populations come from the 6s orbital.
bond length, and dissociation energy are interrelated. For ex- – Regarding the bond lengths of the lanthanide monoxides,
ample, a shorter bond length in general leads to a stronger we noted maximum and minimum values for EuO and
bond and vice versa. Among the lanthanide monoxides, YbO and for LuO, respectively. An increase from HoO to
LuO has the smallest DKH2-B3LYP/TZP-DKH bond dis- YbO was also observed. The atomic charge and orbital
tance. Certainly, this is one of the reasons for LuO’s large population results helped us understand the behavior of
dissociation energy. The open shell electronic configuration the lanthanide monoxide bond lengths along the line.
of Lu (4f145d16s2) also contributes to increasing the dissocia- – Considering the dissociation energies, maximum values
tion energy of lutetium monoxide. On the other hand, the very occur for CeO, GdO, and LuO. In the ranges between
stable electronic configuration of Eu (4f76s2), along with its CeO and EuO and between GdO and YbO, the dissocia-
large bond distance, can explain the low dissociation energy tion energies decrease.
of this monoxide. Similar reasoning could be applied to un- – At the DKH2-B3LYP level of theory, the bond lengths
derstanding the dissociation energy value of YbO. and dissociation energies obtained in this work are gen-
erally in good agreement with those evaluated with the
Polarizability larger all-electron basis set designated SARC [29]. In
addition, our results also agree well with computed
Thierfelder and Schwerdtfeger [68] verified that spin-orbit cou- B3LYP-ECP results [56]. It should be mentioned here
pling is fundamental in polarizability calculations for lantha- that the ECP basis set used in the latter calculations has
nides. Unfortunately, for the 4f elements, the only DKH result 104 functions, while the THZ-DKH set has 96 functions.
available in the literature is for Yb. Then, to assess the perfor-
mance of the basis set generated in this work, the static mean Finally, we verified that the ATZP-DKH basis set seems
dipole polarizability of this atom is calculated with the DKH2- promising for calculating polarizabilities.
B3LYP/ATZP-DKH procedure (137.81 a.u.). At the DKH2- The TZP, ATZP, TZP-DKH, and ATZP-DKH basis sets for
B3LYP level, a benchmark theoretical value of 134.67 a.u. the atoms from H to Lr can be found at the web site http://
was reported by Thierfelder and Schwerdtfeger [68] using the qcgv.ufes.br/.
uncontracted all-electron basis set (37s34p25d16f3g2h). There
is a good agreement between these two results. It seems that Acknowledgments We would like to acknowledge the financial support
of Conselho Nacional de Desenvolvimento Científico e Tecnológico,
ATZP-DKH can be used for the estimation of polarizabilities.
Coordenação de Aperfeiçoamento de Pessoal de Nível Superior, and
Fundação de Amparo à Pesquisa e Inovação do Espírito Santo
(Brazilian Agencies).
Conclusions
Publisher’s note Springer Nature remains neutral with regard to jurisdic-
tional claims in published maps and institutional affiliations.
In this work, nonrelativistic (TZP and ATZP) and relativistic
(TZP-DKH and ATZP-DKH) all-electron basis sets for the
lanthanides are developed. As these sets are generated from
the segmented contraction scheme, they are more efficient References
than generally contracted basis sets in DFT calculations in-
volving large compounds containing 4f elements and, as far 1. Sakamoto M, Manseki K, Okawa H (2001) d–f Heteronuclear com-
as we know, they are the smallest all-electron basis set of plexes: synthesis, structures and physicochemical aspects. Coord
Chem Rev 219–221:379–414
valence triple zeta quality published so far. Actually, they are 2. Bunzli JC, Piguet C (2002) Lanthanide-containing molecular and
compact enough to be competitive with ECP basis sets. It is supramolecular polymetallic functional assemblies. Chem Rev 102:
appropriate to use the basis sets presented in this work in 1897–1928
38 Page 8 of 9 J Mol Model (2019) 25: 38

3. Kido J, Okamoto Y (2002) Organo lanthanide metal complexes for 25. Nakajima T, Hirao K (2002) Accurate relativistic Gaussian basis
electroluminescent materials. Chem Rev 102:2357–2368 sets determined by the third-order Douglas–Kroll approximation
4. Fricker SP (2006) The therapeutic application of lanthanides. Chem with a finite-nucleus model. J Chem Phys 116:8270–8275
Soc Rev 35:524–533 26. Sekiya M, Noro T, Miyoshi E, Osanai Y, Koga T (2006) Relativistic
5. Marjolin A, Gourlaouen C, Clavaguéra C et al (2014) Hydration gibbs correlating basis sets for lanthanide atoms from Ce to Lu. J Comput
free energies of open and closed shell trivalent lanthanide and actinide Chem 27:463–470
cations from polarizable molecular dynamics. J Mol Model 20:2471 27. Lu Q, Peterson KA (2016) Correlation consistent basis sets for
6. Freire RO, Rocha GB, Simas AM (2006) Lanthanide complex co- lanthanides: the atoms La-Lu. J Chem Phys 145:054111
ordination polyhedron geometry prediction accuracies of ab initio 28. Gomes ASP, Dyall KG, Visscher L (2010) Relativistic double-zeta,
effective core potential calculations. J Mol Model 12:373–389 triple-zeta, and quadruple-zeta basis sets for the lanthanides La–Lu.
7. Yang Y, Fang Y, Liu J et al (2015) Complexation behavior of Theor Chem Accounts 127:369–381
Eu(III), Tb(III), tm(III), and am(III) with three 1,10- 29. Pantazis DA, Neese F (2009) All-Electron scalar relativistic basis
phenanthroline-type ligands: insights from density functional theo- sets for the lanthanides. J Chem Theory Comput 5:2229–2238
ry. J Mol Model 21:185 30. Dolg M (2011) Segmented contracted Douglas–Kroll–Hess
8. Salahub DR, Zerner MC (eds) 1989 The Challenge of d and f adapted basis sets for lanthanides. J Chem Theory Comput 7:
Electrons. Theory and computation, ACS symposium series, vol 3131–3142
394. American Chemical Society, Washington 31. Barbieri PL, Fantin PA, Jorge FE (2006) Gaussian basis sets of
9. Cundari TR, Stevens WJ (1993) Effective core potential methods triple and quadruple zeta valence quality for correlated wave func-
for the lanthanides. J Chem Phys 98:5555–5565 tions. Mol Phys 104:2945–2954
10. Ross RB, Gayen S, Ermler WC (1994) Ab initio relativistic effec- 32. Machado SF, Camiletti GG, Canal Neto A, Jorge FE, Jorge RS
tive potentials with spin–orbit operators. V. Ce through Lu. J Chem (2009) Gaussian basis set of triple zeta valence quality for the atoms
Phys 100:8145–8155 from K to Kr: application in DFT and CCSD(T) calculations of
11. Cao X, Dolg M (2002) Segmented contraction scheme for small- molecular properties. Mol Phys 107:1713–1727
core lanthanide pseudopotential basis sets. J Mol Struct 33. Campos CT, Jorge FE (2013) Triple zeta quality basis sets for atoms
(THEOCHEM) 581:139–147 Rb through Xe: application in CCSD(T) atomic and molecular
12. Hülsen M, Weigand A, Dolg M (2009) Quasirelativistic energy- property calculations. Mol Phys 111:167–173
consistent 4f-in-core pseudopotentials for tetravalent lanthanide el-
34. Jorge FE, Canal Neto A, Camiletti GG, Machado SF (2009)
ements. Theor Chem Accounts 122:23–29
Contracted Gaussian basis sets for Douglas-Kroll-Hess calcula-
13. Vyboishchikov SF, Sierraalta A, Frenking G (1997) Topological tions: estimating scalar relativistic effects of some atomic and mo-
analysis of electron density distribution taken from a lecular properties. J Chem Phys 130:064108
pseudopotential calculation. J Comput Chem 18:416–429
35. Martins LSC, Jorge FE, Machado SF (2015) All-electron segment-
14. Güell V, Luis JM, Solà M, Swart M (2008) Importance of the basis
ed contraction basis sets of triple zeta valence quality for the fifth-
set for the spin-state energetics of Iron complexes. J Phys Chem A
row elements. Mol Phys 113:3578–3586
112:6384–6391
36. Campos CT, de Oliveira AZ, Ferreira IB, Jorge FE, Martins LSC
15. Cirera J, Ruiz E (2008) Exchange coupling in CuIIGdIII dinuclear
(2017) Segmented all-electron Gaussian basis sets of double and
complexes: a theoretical perspective. C R Chim 11:1227–1234
triple zeta qualities for Fr, Ra, and Ac. Chem Phys Lett 675:1–5
16. Douglas M, Kroll NM (1974) Quantum electrodynamical correc-
37. Fantin PA, Barbieri PL, Canal Neto A, Jorge FE (2007) Augmented
tions to the fine structure of helium. Ann Phys (NY) 82:89–155
Gaussian basis sets of triple and quadruple zeta valence quality for
17. Hess BA (1985) Applicability of the no-pair equation with free-
the atoms H and from li to Ar: applications in HF, MP2, and DFT
particle projection operators to atomic and molecular structure cal-
calculations of molecular dipole moment and dipole
culations. Phys Rev A 32:756–763
(hyper)polarizability. J Mol Struct (THEOCHEM) 810:103–111
18. Hess BA (1986) Relativistic electronic-structure calculations
38. Camiletti GG, Canal Neto A, Jorge FE, Machado SF (2009)
employing a two-component no-pair formalism with external-
Augmented Gaussian basis sets of double and triple zeta valence
field projection operators. Phys Rev A 33:3742–3748
qualities for the atoms K and Sc-Kr: applications in HF, MP2, and
19. Kellö V, Sadlej AJ (1998) Picture change and calculations of ex-
DFT calculations of molecular electric properties. J Mol Struct
pectation values in approximate relativistic theories. Int J Quantum
(THEOCHEM) 910:122–125
Chem 68:159–174
39. Martins LSC, de Souza FAL, Ceolin GA, Jorge FE, de Berrêdo RC,
20. Baerends EJ, Schwarz WHE, Schwerdtfeger P, Snijders JG (1990)
Campos CT (2013) Augmented Gaussian basis sets for the elements
Relativistic atomic orbital contractions and expansions: magnitudes
K, Sc-Kr, Rb, and Y-Xe: application in HF, MP2, and DFT calcu-
and explanations. J Phys B: At Mol Phys 23:3225–3240
lations of molecular electric properties. Compt Theor Chem 1013:
21. Autschbach J, Peng D, Reiher M (2012) Two-component relativis-
62–69
tic calculations of electric-field gradients using exact decoupling
methods: spin−orbit and picture-change effects. J Chem Theory 40. Campos CT, Jorge FE, Alves JMA (2012) XZP+1d and XZP+1d-
Comput 8:4239–4248 DKH basis sets for second-row elements: application to CCSD(T)
zero-point vibrational energy and atomization energy calculations. J
22. Te Velde G, Bickelhaupt FM, Baerends EJ, Guerra CF, Van
Mol Model 18:4081–4088
Gisbergen SJA, Snijders JG, Ziegler T (2001) Chemistry with
ADF. J Comput Chem 22:931–967 41. Antusek A, Sulka M (2016) Ab initio calculations of NMR
23. Roos BO, Lindh R, Malmqvist P-Å, Veryazov V, Widmark P-O, shielding of Sc3+, Y3+ and La3+ ions in the water solution and
45
Borin AC (2008) New relativistic atomic natural orbital basis sets Sc, 89Y, 138La and 139La nuclear magnetic dipole moments.
for lanthanide atoms with applications to the Ce diatom and LuF3. J Chem Phys Lett 660:127–131
Phys Chem A 112:11431–11435 42. Jorge FE, Venâncio JRC (2018) Structure, stability, catalytic activity,
24. Tsuchiya T, Abe M, Nakajima T, Hirao K (2001) Accurate relativ- and polarizabilities of small iridium clusters. Chin Phys B 27:063102
istic Gaussian basis sets for H through Lr determined by atomic 43. Novikov AS, Ivanov DM, Avdontceva MS, Kukushkin VY (2017)
self-consistent field calculations with the third-order Douglas– Diiodomethane as a halogen bond donor toward metal-bound ha-
Kroll approximation. J Chem Phys 115:4463–4472 lides. Cryst Eng Comm 19:2517–2525
J Mol Model (2019) 25: 38 Page 9 of 9 38

44. Galembeck SE, Caramori GF, Misturini A, Garcia LC, Orenha RP 56. Wu ZJ, Guan W, Meng J, Su ZM (2007) Density functional studies
(2017) Metal–ligand bonding situation in ruthenophanes containing of diatomic LaO to LuO. J Clust Sci 18:444–458
mMultibridged cyclophanes. Organometallics 36:3465–3470 57. Wang SG, Schwarz WHE (1995) Lanthanide Diatomics and lan-
45. Zhang Y, Miao H, Liu L, Zhang X, King RB (2017) Binuclear thanide contractions. J Phys Chem 99:11687–11695
chromium carbonyl complexes of the highly basic small bite 58. Xu W, Ji W-X, Wang S-G (2015) The electronic configurations of
bidentate diphosphine bis(dimethylphosphino)methane. LnX (Ln = La–Eu, X = O, S, Se, Te): A FON–DFT investigation.
Polyhedron 138:194–205 Compt Theor Chem 1068:81–87
46. Chakravorty SJ, Corongiu G, Flores JR, Sonnad V, Clementi E, 59. Ramakrishnan R, Matveev AV, Rosch N (2009) The DFT + U
Carravetta V, Cacelli I (1989) Modern techniques in computational method in the linear combination of Gaussian-type orbitals frame-
chemistry MOTECC-89. ESCOM, Leiden work: role of 4f orbitals in the bonding of LuF3. Chem Phys Lett
47. De Castro EVR, Jorge FE, Pinheiro JC (1999) Accurate Gaussian 468:158–161
basis sets for second-row atoms and ions generated with the improved 60. Huber KP, Herzberg G (1979) Molecular spectra and molecular
generator coordinate Hartree–Fock method. Chem Phys 243:1–7 structure, constants of diatomic molecules, vol IV. Van Nostrand
48. Frisch MJ et al (2009) GAUSSIAN 09, revision A.02. GAUSSIAN Reinhold, New York
Inc., Wallingford CT 61. Shenyavskaya EA, Bernard A, Vergès J (2003) High resolution
study of near-infrared emission spectra of 142 NdO. J Mol
49. De Jong WA, Harrison RJ, Dixon DA (2001) Parallel Douglas–
Spectrosc 222:240–247
Kroll energy and gradients in NWChem: estimating scalar relativ-
62. Chandrasekharaiah MS, Gingerich KA (1989) In: Gschniedner Jr
istic effects using Douglas–Kroll contracted basis sets. J Chem Phys
KA, Erying L (eds) Handbook on the Chemistry and Physics of
114:48–53
Rare Earths, vol 12, Chapter 86. Elsevier, Amsterdam
50. Brummelhuis R, Siedentop H Stockmeyer E 2002The ground-state
63. Haire RG (1994) High-temperature vaporization of transplutonium
energy of relativistic one-electron atoms according to Jansen and
oxides. J Alloys Compd 213/214:185–190
Hess. Doc Math 7:167–182
64. Dulick M, Murad E, Barrow RF (1986) Thermochemical properties
51. Becke AD (1993) Density functional thermochemistry. III. The role of the rare earth monoxides. J Chem Phys 85:385–390
of exact exchange. J Chem Phys 98:5648–5652 65. Konings RJM, Beneš O, Kovács A, Manara D, Sedmidubský D,
52. Lee C, Yang W, Parr RG (1988) Development of the Colle-Salvetti Gorokhov L, Iorish VS, Yungman V, Shenyavskaya E, Osina E
correlation-energy formula into a functional of the electron density. (2014) The Thermodynamic Properties of the f-Elements and their
Phys Rev B 37:785–789 Compounds. Part 2. The Lanthanide and Actinide Oxides. J Phys
53. Reed AE, Curtiss LA, Weinhold F (1988) Intermolecular interac- Chem Ref Data 43:013101
tions from a natural bond orbital, donor-acceptor viewpoint. Chem 66. Kovács A, Konings RJM, Szieberth D, Krámos B (2014) Study of
Rev 88:899–926 the an-cl bond contraction in actinide trichlorides. Struct Chem 25:
54. Zhao Y, Truhlar DG (2006) The M06 suite of density Functionals 991–996
for Main group thermochemistry, thermochemical kinetics, 67. Kovács A, Konings RJM (2004) Structure and vibrations of lantha-
noncovalent interactions, excited states, and transition elements: nide Trihalides: an assessment of experimental and theoretical data.
two new Functionals and systematic testing of four M06 J Phys Chem Ref Data 33:377–404
Functionals and twelve other Functionals. Theor Chem Accounts 68. Thierfelder C, Schwerdtfeger P (2009) Relativistic couple-cluster
120:215–241 calculations. Phys Rev A 79:032512
55. Linstrom PJ, Mallard WG (eds) (2005) NIST Chemistry WebBook, 69. De Oliveira AZ, Campos CT, Jorge FE, Ferreira IB, Fantin PA
NIST Standard Reference Database Number 69. National Institute (2018) All-electron triple zeta basis sets for the actinides. Compt
of Standards and Technology, Gaithersburg Theor Chem 1135:28–33