Theor Chem Acc (2001) 105:328±337

DOI 10.1007/s002140000231

Regular article
The three-center-four-electron (3c-4e) bond nature revisited.
An atoms-in-molecules theory (AIM) and ELF study
Jose Molina Molina, Jose A. Dobado
Grupo de ModelizacioÂn y DisenÄo Molecular, Instituto de BiotecnologõÂ a, Campus Fuentenueva s/n, Universidad de Granada,
18071-Granada, Spain, e-mail: dobado@ugr.es; Tel.: +34-958-243186; Fax: +34-958-243186

Received: 1 June 2000 / Accepted: 4 October 2000 / Published online: 19 January 2001
Ó Springer-Verlag 2001

Abstract. Theoretical calculations (B3LYP/6±311+ clear from accurate ab initio calculations that d orbitals
+G**) were performed on a series of formally hyper- do not participate in bonding, but act solely as
valent compounds showing linear three-center geome- polarization functions [1, 5]. The bonding schemes
tries. The bonding nature was analyzed by the electron presented in the literature for hypervalent compounds
density, q(r), and electron-localization function (ELF) arise from di€erent analyses of the electronic wave
topologies, including calculations of the AIM charges function, indicating the importance of a rigorous and
and NMR chemical shifts (GIAO method). In addition, observation-based interpretation of the theoretical
a quantitative analysis was also performed of the results [6]. The bonding properties of di€erent linear,
localization and delocalization indexes, obtained from three-center hypervalent compounds have been charac-
the electron-pair density in conjunction with the de- terized by the Rundle [7] and Pimentel [8] three-center-
®nition of an atom in a molecule. Furthermore, the four-electron model (3c-4e). The application of this
populations and ¯uctuations in the ELF basins were also model to the [F3]) molecule is depicted in Fig. 1.
evaluated. The compounds studied presented linear (1± A valence bond description of the 3c-4e scheme was
5), T-shaped (6±9), and bipyramidal structures (10±15). performed by Linnett [9]. The 3c-4e model postulates
Our results support the 3c-4e model for the linear (1±5) three MOs, Y1 with bonding nature, Y3 with antibonding,
structures, but reveal for the T-shaped (6±9) structures and Y2 with no bonding character. These three MOs yield
only a small contribution from this model. In addition, an overall bond order of 0.5 with two electrons almost
there is no evidence to support the 3c-4e bond scheme localized on the external ¯uorines with atomic charges of
for the bipyramidal compounds (10±15). )0.5 e). This scheme has also been extended to other
hypervalent molecules such as F3Cl, [F3S]), F4S, F5P,
Key words: Ab initio calculations ± Atoms-in- [F5Si]) or F6S, with di€erent trigonal-bipyramid or oc-
molecules ± Bond theory ± Three-center-four-electron tahedral geometries. Furthermore, an explanation of the
bond ± Electron localization function dissimilar ligand equatorial and axial bond lengths to the
central atom is rationalized by these geometrical ar-
rangements. Another point of view in the description of
these hypervalent structures is the Generalized Valence
Bond approach summarized by the democracy principle
1 Introduction proposed by Cooper et al. [10], in which a particular atom
uses any of its valence electrons in the chemical bonding,
The bond nature in hypervalent compounds has long based on the principle of minimizing the total energy.
been controversial, and has been the object of several In recent years, an observation-based interpretative
recent reviews [1, 2]. Special attention has been focused tool has been extensively used, the atoms-in-molecules
on chalcogen oxides [3] and ylides [4]. Historically, the (AIM) theory [11]. In this context, AIM has been used in
bonding nature for these compounds has been based the bonding description of di€erent hypervalent com-
on an spxd y hybridization scheme. However, it is now pounds [3,4a,6,12]. A complementary method to study
the chemical bonding is the electron-localization func-
tion (ELF) [13] analysis of Becke and Edgecombe [14].
This work is dedicated to Professor SerafõÂ n Fraga
Moreover, the ELF topological analysis has been
Professor Jose Molina Molina passed away on the 13th June 2000 also used to de®ne chemical bonding including that in
Contribution to the Proceedings of the 2000 Symposium on hypervalent molecules [12].
Chemical Bonding: State of the Art in Conceptual Quantum An appropriate de®nition of bonding indexes is of
Chemistry. great importance for the bonding characterization of
 329

Overview of the q(r) and ELF topologies

The topology of the electronic charge density, q(r), as pointed out
by Bader [11], is an accurate mapping of the chemical concepts of
atoms, bonds, and structures. The principal topological properties
are summarized in terms of their critical points (CP) [11], and the
nuclear positions behave topologically as local maxima in q(r). A
bond critical point (BCP) is found between each pair of nuclei,
which are considered to be linked by a chemical bond, with two
negative curvatures, (k1 and k2) and one positive (k3) [denoted as
(3, )1) CP]. The ellipticity, e, of a bond is de®ned by means of the
two negative curvatures in a BCP as:

e ˆ k1 =k2 1; where jk2 j < jk1 j 1†

The ring CPs are characterized by a single negative curvature. Each

Fig. 1. Schematic molecular orbitals for the 3c-4e bond model of (3, )1) CP generates a pair of gradient paths [11] which originate at
[F3]) molecule a CP and terminate at neighboring attractors. This gradient path
de®nes a line through the charge distribution linking the neigh-
hypervalent compounds [15]. Therefore, much e€ort has boring nuclei. Along this line, q(r) is a maximum with respect
to any neighboring line. Such a line is referred to as an atomic
been made in the past to de®ne such a non-observable interaction line [11]. The presence of an atomic-interaction line
concept of chemical bonding [15, 16]. The corresponding in such equilibrium geometry satis®es both the necessary and suf-
covalent bond orders were de®ned both in the HF context ®cient conditions that the atoms be bonded together.
or at the correlated level, while addressing the necessity of The Laplacian of the electronic charge density, Ñ2q(r), describes
using the AIM atomic basins in the population analysis two extreme situations. In the ®rst, q(r) is locally concentrated
[Ñ2q(r)<0] and in the second it is locally depleted [Ñ2q(r)>0].
[16]. Moreover, other multi-center bond indexes have Thus, a value of Ñ2q(r)<0 at a BCP is unambiguously related to a
been proposed to study the 3c-4e model in the context of covalent bond, showing that a sharing of charge has taken place. In
hypervalent molecules [17]. Considering that the bond is a a closed-shell interaction, a value of Ñ2q(r)>0 is expected, as found
non-observable concept in the Quantum Chemistry de®- in noble gas repulsive states, ionic bonds, hydrogen bonds, and van
nition, electron-delocalization indexes have also been der Waals molecules. Bader has also de®ned a local electronic en-
ergy density, Ed(r), as a function of the ®rst-order density matrix:
de®ned based on the pair density to link Quantum Me-
chanical and Classical approaches to chemical structures. Ed (r) = G(r)+V(r) 2†
This de®nition was made [18] using the Mulliken ap-
where the G(r) and V(r) correspond to a local kinetic and potential
proximation to de®ne charge density, bond order, etc., energy density, respectively [11]. The sign of the Ed(r) determines
yielding indexes equivalent to those of Wiberg [19]. whether a charge accumulation at a given point r is stabilizing
Another alternative is the delocalization indexes for- [Ed(r)<0] or destabilizing [Ed(r)>0]. Thus, a value of Ed(r)<0 at a
mulated by integration of the pair density between the BCP presents a signi®cant covalent contribution and, therefore, a
atomic basins de®ned by the AIM theory [20, 21]. In this lowering of the potential energy associated with the concentration
work, the 3c-4e bond model is revisited based on an of charge between the nuclei.
The quantum-mechanical pair density in conjunction with the
adequate topological analysis of q(r) and ELF, together quantum de®nition of an atom in a molecule provides a precise
with the electron delocalization indexes proposed by determination of the extent to which electrons are localized in a
Bader, and the theoretical NMR chemical shifts. The given atom and delocalized over any pair of atoms [21]. The elec-
study has been performed for the hypervalent com- tron pairing is a consequence of the Pauli exclusion principle, and
pounds depicted in Figs. 2 and 3. Geometrical and the extent of spatial localization of the pairing is determined by the
corresponding property of the Fermi hole density. These ideas are
electronic calculations for 1±15 have already been made quantitative through the appropriate integration of the pair
extensively reported in the literature [22], including density to determine the total Fermi correlation contained within a
their q(r) and ELF analyses [12], and the results are single atomic basin, the quantity F(A,A), or F(A,B), the correla-
compatible with the VSEPR model [23]. tion shared between two basins. The quantity F(A,B) is thus
a measure of the extent to which electrons of either spin referenced
to atom A are delocalized into atom B with a corresponding
de®nition of F(B,A). Thus, F(B,A) ˆ F(A,B) and their sum,
2 Computational details F(A,B) + F(B,A) ˆ d(A,B), termed the delocalization index, is a
measure of the total Fermi correlation shared between the atoms.
General methods This delocalization index is calculated taking into account that
Density functional theory (B3LYP) calculations were performed XX
with the Gaussian 98 package of programs [24]. All the minimum F A; B† ˆ F B; A† ˆ Sij A†
Sij B† 3†
structures were fully optimized and tested by frequency analysis i j
at the B3LYP/6-311++G** level, yielding the minima with where Sij (A) is the corresponding atomic overlap matrix given by
constrained symmetries (see Figs. 2 and 3) and non-imaginary the PROAIM program at the Hartree-Fock level [20].
frequencies. The AIM analysis [11] has been performed with the The ELF function [13, 14], which was ®rst introduced by Becke
AIMPAC series of programs [25], using the DFT densities as the and Edgecombe [14], can be viewed as a local measure of the Pauli
input. repulsion between electrons due to the exclusion principle, enabling
The NMR chemical shifts were calculated by the GIAO method us to de®ne regions of space that are associated with di€erent
[26] using the tetramethylsilane (TMS) shielding as references electron pairs. The ELF function is expressed by
for the 1H (dH) chemical shift (31.98 ppm).
The Ñ2q(r) contour-map representations have been produced 1
ELF ˆ  2 4†
using the MORPHY98 program [27]. The ELF analyses have been 1‡ D
made with the ToPMoD package of programs [28]. Dh
330

Fig. 2. Hypervalent structures (1±8)

with the geometrical parameters,
charges and NMR chemical shifts,
(in parentheses symmetry)

where which P(Wi,Wj) is the pair population, and the ¯uctuation contri-
B
1 X
N
1 jrqj2
3
2=3 5=3 X
N bution PijB .
2 2
Dˆ ruj ; Dh ˆ 3p2 q ; qˆ uj i6ˆj
ij

2 jˆ1
8 q 10 jˆ1

5† 3 Results and discussion

This de®nition gives ELF values between 0 and 1, with large Theoretical calculations have been performed on the
values where two anti-parallel spin electrons are paired in space
and small ones in the regions between electron pairs. Considering structures depicted in Figs. 2 and 3 together with their
the scalar character of the ELF function, the analysis of their corresponding reference non-hypervalent compounds
gradient ®elds yields their attractors (local maxima) and their (Fig. 4) at the B3LYP/6-311‡‡G** theoretical level,
corresponding basins. There are two type of basins: the core which has proven its utility for describing accurate
basins labeled by C(atom symbol) and the valence basins V(list geometrical and electronic features in similar hypervalent
of atoms). The valence basins are characterized by their synaptic
order (the number of core basins with which they share a common
derivatives [3, 4]. These ®gures illustrate the geometrical
boundary). Accordingly, they can be classi®ed as mono-, di- and parameters, yielding geometries in agreement with the
polysynaptic, corresponding to the lone pair, bicentric and values reported in the literature. In general, the geomet-
polycentric bonding region, respectively. The quantitative ric and electronic behavior are compatible with the
population on the di€erent basins are calculated by integrating VSEPR molecular geometry model [23].
a given density of property over the volume of the basins. The topology of the electron density, q(r), has been
The following de®nitions
R are used through the text: RbasinR
population N  Xi † ˆ 2 analyzed following the AIM theory, yielding maxima of
X1 q r†dr and its variance r Xi † ˆ Xi Xi
0 0   2
p r; r †drdr ‡ N Xi † N Xi † % in which p(r,r')
is the
two electron concentration also compatible with the VSEPR
electron density, covariance Bij ˆ N  Xi †
N
 Xj P Xi ; Xj % in model. The numerical values at the di€erent BCPs are
 331

Fig. 3. Hypervalent structures (9±

15) with the geometrical parameters,
charges and NMR chemical shifts,
(in parentheses symmetry)

depicted in Table 1. Integration of the density in the F-Cl lengthening ca. 0.4 AÊ compared to 19. The
atomic basins gave the AIM atomic charges. Moreover, negative charge was located mainly on the two
the GIAO method was used to generate the corre- terminal atoms, giving a highly positive hydrogen
sponding isotropic chemical shifts for the di€erent atoms atom in 1, and an slightly positive chlorine atom in 3.
(see Figs. 2±4). By contrast, the central atoms for 2 and 5 remained
The following trends are extracted from the geomet- almost neutral. Moreover, for 4 the ¯uorine had
rical parameters and electronic properties. a negative value and was similar to the terminal
chlorines.
1) The linear anionic structures (1±5) lengthened their
2) The T-shaped structures (6±9) (with lone pairs at the
bonds by more than 0.2 AÊ compared to the corre-
central atom) had larger ¯uorine axial bonds than the
sponding reference compounds (16±19). Further-
corresponding equatorial ones (>0.1 AÊ). However,
more, the main di€erences appeared in 4 with a
332

Fig. 4. Reference molecules (16±23)

for comparison, with the geometrical
parameters, charges and NMR
chemical shifts, (in parentheses
symmetry)

the di€erences with the reference compounds were bonds, but with notable covalent character in accord
shorter than in the linear structures. with q(r)>0.1 e/ao3 positive Ñ2q(r)>0.35 e/ao5 and neg-
3) The ¯uorine axial/equatorial bond length di€erences ative Ed(r) values. Furthermore, the di€erences in the
for 10±15 (without lone pairs at the central atom) numerical values for the axial/equatorial bonds were
were very small (<0.08 AÊ). very small.
4) The atomic charges on the Fa and Fe atoms were very The 19F-NMR chemical shifts were calculated to
similar for 7±15; however, the largest di€erence was evaluate the di€erences in the chemical environment,
found for 7 (0.16 e)). giving the linear structures trends similar to the ref-
erence compounds. Moreover, the T-shaped structures
The nature of the F-X bond has been analyzed,
(6±9) presented sharply di€erent axial shifts (dFa)
considering the q(r) topology (see Table 1), yielding the
compared to the equatorial values, the largest di€er-
linear structures (1±5) a very polar unstabilized behavior
ences being found for structure 7 and 8 (>100 ppm).
with small q(r) values, positive Ñ2q(r), small k1/k3
The axial/equatorial di€erences were due mainly to the
and Ed(r) values close to zero at the BCPs. However,
chemical environment of these compounds, with
structure 1 deviated from these trends due to the strong
equatorial disposition for the central atom lone pairs.
electrostatic interaction between the hydrogen and the
The largest di€erences in 7 and 8 are explained by the
two ¯uorines (q(r) ˆ 0.170 e/ao3 and Ñ2q(r) ˆ )0.29
existence of two lone pairs, in contrast to the only one
e/ao5). The sulfur derivatives enhanced the S-F covalent
for 6 and 9. The axial/equatorial shifts values were
character, giving higher q(r) and negative Ñ2q(r) values.
very similar for the structures without lone pairs on the
In addition, the numerical values matched the geomet-
central atom (10±15), excluding 12 which has another
rical results. The remaining structures with Cl (7), P (11±
two equatorial hydrogens.
13) and Si (14) as the central atom displayed highly polar
 333

Table 1. Electronic charge

density, q(r), its Laplacian, Structure Bond q(r) Ñ 2 q(r) e |k1/k3| Ed(r)
Ñ2q(r), ellipticity, e, curvatures, (e/a3o) (e/a5o ) (Hartree/a3o)
k1/k3, and electronic energy
density, Ed(r), at the B3LYP/ 1, FHF) F-H 0.174 )0.290 0.000 0.645 )0.186
6-311‡‡G** theoretical level 2, FFF) F-F 0.100 0.539 0.000 0.199 0.015
for the di€erent BCPs 3, FClF) F-Cl 0.099 0.257 0.000 0.254 )0.014
4, ClFCl) Cl-F 0.076 0.241 0.000 0.222 0.000
5, [Cl3]) Cl-Cl 0.063 0.117 0.000 0.259 )0.006
6, H2CSF2 C-H 0.287 )1.022 0.004 1.371 )0.285
C-S 0.262 )0.469 0.781 2.094 )0.354
S-F 0.132 0.091 0.060 0.411 )0.069
7, ClF3 P-Fe 0.201 0.033 0.149 0.511 )0.125
P-Fa 0.156 0.234 0.115 0.363 )0.061
8, [SF3]) S-Fe 0.160 )0.052 0.509 0.690 )0.137
S-Fa 0.110 0.148 0.270 0.366 )0.040
9, F4S S-Fa 0.170 )0.062 0.068 0.589 )0.133
S-Fe 0.205 )0.062 0.041 0.568 )0.226
10, F4SO S-O 0.312 1.140 0.139 0.250 )0.386
S-Fa 0.194 )0.204 0.027 0.794 )0.202
S-Fe 0.208 )0.102 0.050 0.612 )0.233
11, F2PH3 P-F 0.130 0.387 0.000 0.233 )0.085
P-H 0.184 )0.113 0.001 0.618 )0.191
12, F3PH2 P-Fe 0.164 0.759 0.018 0.218 )0.101
P-Fa 0.141 0.477 0.095 0.240 )0.095
P-H 0.190 )0.148 0.034 0.667 )0.202
13, PF5 P-Fe 0.173 0.812 0.053 0.221 )0.112
P-Fa 0.164 0.645 0.000 0.228 )0.115
14, [SiF5]) Si-Fe 0.116 0.808 0.034 0.174 )0.024
Si-Fa 0.110 0.682 0.000 0.177 )0.028
15, F6S F-S 0.210 )0.127 0.000 0.627 )0.242
16, FH F-H 0.369 )2.799 0.000 1.252 )0.784
17, F2 F-F 0.267 0.594 0.000 0.352 )0.119
18, Cl2 Cl-Cl 0.134 0.022 0.000 0.470 )0.052
19, FCl F-Cl 0.182 0.061 0.000 0.452 )0.108
20, FPH2 P-F 0.139 0.487 0.083 0.235 )0.088
P-H 0.167 )0.130 0.202 0.745 )0.168
21, F3P F-P 0.157 0.620 0.070 0.232 )0.102
22, F4Si F-Si 0.138 1.120 0.000 0.164 )0.026
23, F2S F-S 0.177 0.001 0.568 0.610 )0.173

According to the above-mentioned trends, the struc- nal nuclei should appear in this model. Compounds
tures can be classi®ed in three main groups. One has 10±15 (above-mentioned group 1) with a bipyramidal
structures 10±15 without lone pairs on the central atom, structure, displayed linear F-X-F moiety and a 3c-4e
in which the axial/equatorial ¯uorine bonds had similar bond has been assigned, particularly for 13 (PF5). Com-
behavior for the geometries and electronic properties. pounds 10±14 displayed two di€erent bonding schemes
The second group (6±9) has marked di€erences in the for the axial/equatorial bonds, with the 3c-4e for the axial
axial/equatorial geometries and electronic properties. ones and three 2c-2e for the equatorial ones. In contrast,
For these structures, the di€erences may be due to the the bonding scheme in 15 was previously described in the
axial/equatorial bonding nature or to the electronic en- literature as three 3c-4e bonds with an additional bonding
vironment. In the third group, linear structures 1±5 have MO [2b], resulting in six equivalent F-S bonds. Com-
markedly di€erent properties compared to the parent pounds 13 and 14 displayed D3h symmetry with three
compounds (16±19). Among the three groups, the latter equivalents equatorial bonds and two axial ones. The
displayed characteristics compatible with the 3c-4e bond theoretical axial/equatorial bond lengths are similar with
scheme. longer axial distances. Furthermore, as previously men-
For a qualitative as well as quantitative evaluation of tioned, all the geometrical and electronic properties were
the 3c-4e bond, the electron localization and delocal- similar. In addition, the localization and delocalization
ization indexes [21] were calculated, and the numerical indexes for both bonds were equivalent [F(Fa,Fa)=9.21
values are presented in Table 2. In addition, a topolog- and F(Fe,Fe)=9.20 e) for 13]. In general, the delocal-
ical ELF study was also made, including the numerical ization indexes were very similar with larger values for the
populations and ¯uctuations for the di€erent basins, to X-Fe bonds. On the other hand, the electron population
evaluate the electronic delocalization for these structures of the ELF valence basins had the same trends
(see Table 3). [V(Si,Fe)=1.24 and V(Si,Fa)=1.18 e) for 14] and the
The Rundle [7] and Pimentel [8] 3c-4e bond model has monosynaptic valence basins at the ¯uorine-atom popu-
a non-bonding Y2 MO with two electrons (see Fig. 1). lations (lone pairs) were also similar and independent of
Therefore, an electron delocalization between the termi- the axial/equatorial nature for 13 and 14. In addition, the
334

Table 2. Atomic localization index, F(A,A), and delocalization index d(A,B)

A F(A,A) (%)a A,B d(A,B) A F(A,A) (%)a A,B d(A,B)

)
1, FHF ; áá F 9.59 97.4 F,H 0.28 12, F3PH2; áá P 10.35 88.2 P,Fa 0.48
H 0.04 12.3 F,F 0.22 Fa 9.27 95.0 P,Fe 0.54
2, FFF); áá Ft 9.96 94.7 Ft,Fc 0.71 Fe 9.24 94.8 P,H 0.64
Fc 8.38 92.2 Ft,Ft 0.29 H 0.99 65.7 Fa,Fa 0.03
3, FClF); áá F 9.15 95.2 F,F 0.15 Fe,Fa 0.17
Cl 15.99 95.3 Cl,F 0.78 Fa,H 0.06
4, ClFCl); áá Cl 16.76 96.6 Cl,F 0.70 Fe,H 0.15
F 8.64 92.5 Cl,Cl 0.42 13, PF5; áá P 10.15 89.3 P,Fa 0.44
5, [Cl3]); áá Clt 16.94 97.0 Clt,Clc 0.79 Fa 9.21 94.6 P,Fe 0.51
Clc 16.27 95.4 Clt,Clt 0.27 Fe 9.20 94.7 Fa,Fa <0.01
6, H2CSF2; áá S 12.88 88.4 S,C 1.76 Fe,Fe 0.06
F 9.08 94.4 S,F 0.77 Fa,Fe 0.20
C 4.54 69.7 C,H 0.93 14, SiF5); áá Si 10.0 92.8 Si,Fa 0.28
H 0.32 38.1 F,F 0.09 Fa 9.45 95.9 Si,Fe 0.33
C,F 0.16 Fe 9.45 96.1 Fa,Fa <0.01
7, ClF3; áá Cl 14.41 91.0 Cl,Fa 0.91 Fe,Fe 0.05
Fa 8.84 93.7 Cl,Fe 1.05 Fe,Fa 0.17
Fe 8.62 92.8 Fa,Fa 0.15 15, F6S; áá S 10.68 84.4 S,F 0.66
Fe,Fa 0.14 F 8.93 93.4 Fa,Fa <0.01
8, [SF3]); áá S 13.86 91.9 S,Fe 0.93 Fe,Fa 0.15
Fe 9.00 94.0 S,Fa 0.76 16, FH; áá F 9.44 97.3 F,H 0.52
Fa 9.19 95.1 Fa,Fa 0.08 H 0.04 13.1
Fa,Fe 0.11 17, FF; áá F 8.37 93.1 F,F 1.25
9, F4S; áá S 12.14 87.9 S,Fe 0.86 18, ClCl; áá Cl 16.34 96.1 Cl,Cl 1.32
Fe 8.90 93.3 S,Fa 0.81 19, FCl; áá F 8.74 93.5 F,Cl 1.21
Fa 8.97 93.9 Fa,Fa 0.06 Cl 16.05 96.4
Fe,Fe 0.11 20, FPH2; áá P 12.03 90.9 P,F 0.73
Fa,Fe 0.15 F 9.27 95.1 P,H 0.84
10, F4SO; áá S 10.72 84.3 S,O 1.23 H 0.97 64.5 F,H 0.11
Fa 8.93 93.5 S,Fe 0.73 H,H 0.12
Fe 8.88 93.2 S,Fa 0.65 21, F3P; áá P 11.73 91.7 P,F 0.71
O 8.21 89.7 Fa,Fa <0.01 F 9.23 94.8 F,F 0.15
Fe,Fe 0.06 22, F4Si; áá Si 9.99 92.5 Si,F 0.40
Fa,Fe 0.19 F 9.43 96.2 F,F 0.12
O,Fa 0.21 23, F2S; áá S 13.87 93.0 F,S 1.04
O,Fe 0.11 F 8.96 93.9 F,F 0.13
11, F2PH3; áá P 10.45 87.6 P,F 0.48
F 9.31 95.3 P,H 0.66
H 0.99 65.5 F,F 0.04
H,H 0.05
F,H 0.13
a
Percentage of electron localization=(F(A,A)/NA) ´ 100%

overall electronic localization trends for the axial/equa- of this group, in agreement with HaÈser's results based on
torial bonds were highly similar to the parent compounds a one center-expansion technique for the PF5 molecule
(21 and 22). For compounds 13 and 14, non-electron [22d]. The small di€erences found between the axial/
delocalization was found between the axial ¯uorines, with equatorial bonds are explained by the di€erent chemical
only an axial/equatorial delocalization similar in magni- surroundings for these bonds.
tude to the d(F1,F2) in the parent compounds. The second group of compounds (6±9) displayed
Similar observations resulted analyzing the ¯uctua- di€erent T-shaped geometries with lone pairs on the
tion contributions of the ELF valence basins in the V(F) central atom. These structures presented marked dif-
basins (see Table 3), with an extremely small axial-axial ferences in the chemical axial/equatorial properties
contribution (<1%). (bond lengths, charges and chemical shifts), the di€er-
Compounds 10 and 15 had sulfur as the central ences in the localization and delocalization for these
atom and the localization characteristic for the ¯uo- structures between the axial/equatorial bonds were now
rine in 15 and both Fa and Fe in 10 had the same not negligible with an increase in the d(Fa,Fa) values
values (ca. 8.9 e)). Analogous behavior was found for from 0.06 in 9 to 0.15 in 7. The V(Fa) basin ¯uctuation
the remaining compounds in this group (11 and 12), contributions from the other axial basins also have
including no electron delocalization between axial larger values than in the former group (ca. 2%). These
atoms. results support a small contribution to these structures
From all the above electronic localization and delo- of the 3c-4e model, due mainly to a slight electron
calization results, we conclude that there is no evidence delocalization between the axial ¯uorine atoms. How-
to support the 3c-4e model in the Fa-X-Fa substructure ever, the di€erent behavior of the axial/equatorial atoms
 335

Table 3. Basin population, N, variance, r2 and contribution ¯uctuations

`N r2 Contribution ¯uctuations (%) `N r2 Contribution ¯uctuations (%)

)
1, FHF V(H) 0.30 0.26 29 V(F); 12, F3PH2 V(P,H1) 2.10 0.58 13.5 V(P,H2); 8 V(Fe); 6 V(P,Fe);
19 V(F!) 16 V(F2); 8 V(P,F2)
V(F1!) 0.94 0.66 8 V(H); V(P,Fe) 1.15 0.81 4 V(P,H); 86 V(Fe); 3 V(F2); 3 V(P,F2)
80 V(F1);
3.2 V(F2);
2.1 V(F2!)
V(F1) 6.76 1.02 8 V(H); V(P,F2) 0.95 0.70 7 V(P,H); 3 V(Fe); 3 V(P,Fe); 69 V(F2)
52 V(F1!);
5.3 V(F2);
2 V(F2!)
2, FFF) V(F1)c 6.93 1.03 32 V(F2) V(Fe) 6.68 1.37 2 V(P,H); 58 V(Fe); 20 V(P,Fe); 2 V(F2);
1 V(P,F2)
V(F2)t 7.33 0.85 39 V(F1); V(F2) 6.81 1.35 3 V(P,H); 1.5 V(Fe); 1 V(P,Fe); 62 V(F2);
15.6 V(F3) 17 V(P,F2)
3, FClF) V(F) 7.44 0.83 44.5 V(Cl); 13, PF5 V(P,F1)e 1.23 0.86 67 V(F1); 3.5 V(P,F2); 1.6 V(P,F1);
8.1 V(F2) 3.6 V(F2); 2 V(Fe)
V(Cl) 6.68 1.32 28 V(F) V(P,F2)a 1.14 0.81 3.5 V(F1); 3.6 V(P,F1); 66 V(F2)
)
4, ClFCl V(F) 7.15 1.03 32 V(Cl) V(F1) 6.58 1.36 57 V(F1); 21 V(P,F1); 3 V(F2)
V(Cl1) 7.22 1.12 30 V(F); V(F2) 6.75 1.20 2 V(F1); 64 V(F2)
17 V(Cl2)
5, Cl3) V(Cl1) 6.96 1.30 28 V(Cl2) 14, SiF5) V(Si,F1)e 1.24 0.85 74 V(F1); 2.2 V(Si,F2); 1 V(Si,Fe); 3 V(F2)
V(Cl2) 7.33 1.09 35 V(Cl1); V(Si,F2)a 1.18 0.81 2.5 V(F1); 2.3 V(Si,F1); 73 V(F2)
11 V(Cl3)
6, H2CSF2 V(S,C) 2.67 1.31 37 V(S); V(F1)e 6.60 1.36 58 V(F1); 22 V(Si,F1)
6 V(F)
V(S) 3.78 1.67 29 V(S,C); V(F2)a 6.65 1.17 19 V(Si,F2); 64 V(F2); 2 V(F1)
20 V(F)
V(F1) 7.45 1.43 3 V(S,C); 15, SF6 V(S,F1) 1.05 0.77 60 V(F1); 4 V(S,F2); 4 V(F2)
11 V(S);
69 V(F1);
1.4 V(F2)
7, ClF3 V(F1!) 0.15 0.14 62 V(F1); V(F1) 6.86 1.14 66 V(F1); 11 V(S,F1)
11 V(Cl);
8.5 V(Cl!)
V(F1) 6.93 1.30 60 V(F1); 16, FH V(FH) 1.45 0.77 93 V(F)
3.5 V(F1!);
6.5 V(Cl);
4 V(Cl!);
2 V(F2)
V(F2) 7.26 1.40 7 V(Cl); V(F) 6.41 1.06 68 V(F,H)
65 V(F2);
2.5 V(F3)
V(Cl!) 0.27 0.25 41 V(F1); 17, FF V(F1!) 0.08 0.08 54 V(F1); 38 V(F2); 3 V(F2!)
5 V(F1!);
19 V(Cl);
3 V(F2)
V(Cl) 2.82 1.26 14 V(F1); V(F1) 6.78 0.97 4.3 V(F1!); 54 V(F2); 3 V(F2!)
28 V(Cl);
4 V(Cl!);
16 V(F2)
8, SF3) V(S,F1) 0.40 0.35 59 V(F1); 18, ClCl V(Cl1!) 0.36 0.31 50 V(Cl1); 30 V(Cl2); 11 V(Cl2!)
13 V(S);
2.5 V(F2)
V(S) 2.45 1.50 15 V(F1); V(Cl1) 6.53 1.18 14 V(Cl1!); 32 V(Cl2); 8 V(Cl2!)
4 V(S,F1);
27 V(S2);
16 V(F2)
V(F1)e 7.17 1.40 68 V(F1); 19, FCl V(F,Cl) 0.38 0.33 56 V(F); 36 V(Cl)
6.5 V(S,F1);
5 V(S);
1.5 V(F2)
336

Table 3. Continued

`N r2 Contribution ¯uctuations (%) `N r2 Contribution ¯uctuations (%)

V(F2)a 7.49 1.41 1.7 V(F1); V(F) 6.93 1.00 19 V(F,Cl); 43 V(Cl)
6 V(S);
71 V(F2);
1 V(F3)a
9, F4S V(S,F3)e 0.89 0.68 9.5 V(S); V(Cl) 6.41 1.14 39 V(F); 11 V(F,Cl)
4 V(S,Fe);
3 V(S,Fa);
3.5 V(Fa);
58 V(F3);
3.6 V(Fe)
V(S,F1)a 0.52 0.44 14.5 V(S); 20, FPH2 V(P,H1) 2.00 0.58 26 V(P,H2); 37 V(P); 13 V(F); 6 V(P,F)
5 V(S,Fe);
56 V(F1)
V(S) 2.50 1.13 5.5 V(S,Fe); V(P,F) 0.79 0.61 6 V(P,H); 9 V(P); 67 V(F)
5.5 V(S,Fa);
17 V(Fa);
10 V(Fe)
V(F3)e 6.88 1.37 4 V(S); V(P) 2.09 0.81 27 V(P,H); 15 V(F); 7 V(P,F)
14.6 V(S,F3);
59.5 V(F3);
1.5 V(Fe);
1.5 V(Fa)
V(F1)a 7.18 1.40 7 V(S); V(F) 7.91 1.40 3 V(P,H); 60 V(F); 15 V(P,F)
8 V(S,F1);
65 V(F1);
2 V(Fe)
10, F4SO V(S,F3)e 0.97 0.73 5 V(S,O); 21, PF3 V(P,F1) 0.96 0.71 9 V(P); 67 V(F1); 3 V(P,F2); 3.5 V(F2)
2.6 V(S,F4)e;
9 V(Fe);
4 V(Fa);
58 V(F3);
5 V(S,Fa)
V(S,F1)a 0.93 0.71 8 V(S,O); V(F1) 6.80 1.40 6 V(P); 18 V(P,F1); 57 V(F1)
5 V(S,Fe);
57 V(F1);
5 V(Fe)
V(F3)e 6.84 1.37 15.6 V(S,F3); V(P) 2.23 0.85 7 V(P,F); 16 V(F)
1.5 V(Fa);
59 V(F3)
V(F1)a 6.93 1.23 11 V(S,F1); 22, SiF4 V(Si,F1) 1.24 0.83 76 V(F1); 1.4 V(Si,F2)
69 V(F1)
11, F2PH3 V(P,H1) 2.07 0.58 17 V(P,H2); V(F1) 6.33 1.09 63 V(F1); 20 V(Si,F1); 1 V(F2); 2.5 V(F2)
8 V(P,F);
15 V(F)
V(P,F1) 0.87 0.65 6.5 V(P,H); 23, SF2 V(S,F) 0.58 0.48 62 V(F); 12 V(S); 1.5 V(S,F2); 2.5 V(F2)
69 V(F1);
<1 V(P,F2)
V(F1) 6.85 1.10 2.5 V(P,H); V(F) 6.95 1.35 61 V(F); 10.5 V(S,F); 6 V(S); 1.5 V(F2)
13 V(P,F1);
35 V(F1);
<1 V(F2)
V(S) 2.29 0.94 17 V(F); 6 V(S,F); 30 V(S2)
a
Arrows indicate a nonsynaptic basin directed to the joined atom

arose mainly from their di€erent chemical surroundings The latter showed a noticeable electron delocalization.
(lone pairs). The delocalization observed in the linear compounds (1±
The linear structures presented the largest overall 5) is re¯ected also in the ELF ¯uctuation contributions
di€erences compared to 16±19. The localization indexes values between the terminal basins, e.g., a value of 17%
were highly di€erent, and the delocalization ones for 4. This delocalization observed for the linear com-
between the terminal atoms were larger with values in pounds strongly supports the 3c-4e model, in agreement
the 0.15±0.42 range between terminal atoms, for 3 and 4. with previous results [17a].
 337

4 Conclusions 7. (a) Hach RJ, Rundle RE (1951) J Am Chem Soc 73: 4321; (b)
Rundle RE (1963) J Am Chem Soc 85: 112
8. Pimentel GC (1951) J Chem Phys 19: 446
Theoretical calculations (B3LYP/6±311‡‡G**) were
9. Bilham J, Linnett JW (1964) Nature 201: 1323
performed on di€erent structures, showing three-center 10. Cooper DL, Cunningham TP, Gerratt J, Karadakov PB,
linear and pseudo-linear arrangements. The compounds Raimondi M (1994) J Am Chem Soc 116: 4414
studied represented linear, T-shaped, and bipyramidal 11. (a) Bader RFW (1990) Atoms in Molecules: a Quantum Theory.
structures, with the latter group lacking electron pairs on Clarendon Press: Oxford; (b) Bader RFW (1991) Chem Rev 91:
the central atom. The electronic behavior of the q(r) 893
topology yielded the ¯uorine bonds to the central atom 12. Bader RFW, Johnson S, Tang T-H, Popelier PLA (1996) J Phys
Chem 100: 15398
with a highly ionic character. The overall geometrical 13. (a) Silvi B, Savin A (1994) Nature 371: 683; (b) Savin A, Nesper
and electronic characteristics, including the electron R, Wengert S, FaÈssler TF (1997) Angew Chem Int Ed Engl 36:
delocalization indexes for the bipyramidal compounds 1809 (c) Marx D, Savin A (1997) Angew Chem Int Ed Engl 36:
(10±15), are similar for the axial and equatorial bonds. 2077
In addition, no electron delocalization was found 14. Becke AD, Edgecombe KE (1990) J Chem Phys 92: 5397
between the axial ¯uorines for these compounds. 15. Bachrach SM (1994) In: Lipkowitz KB, Boyd DB (eds),
Reviews in Computational Chemistry. VCH: New York, vol.
Consequently, no 3c-4e bond scheme exists for these 5, pp 171±227, and references cited therein
structures. However, the electron delocalization in- 16. AÂngyaÂn JG, Loos M, Mayer I (1994) J Phys Chem 98: 5244 and
creased for the T-shaped (5±9) and linear (1±5) struc- references cited therein; (b) AÂngyaÂn JG, Rosta E, SurjaÂn PR
tures, indicating a participation of the 3c-4e model. The (1999) Chem Phys Lett 299: 1 and references cited therein
T-shaped structures presented di€erent geometrical as 17. (a) Kar T, SaÂnchez Marcos E (1992) Chem Phys Lett 192: 14;
well as electronic behavior between axial/equatorial (b) Ponec R, Mayer I (1997) J Phys Chem A 101: 1738
18. Ponec R, Strand M (1994) Int J Quantum Chem 50: 43
bonds with a small amount of electron delocalization 19. Wiberg KB (1968) Tetrahedron 24: 1024
between axial atoms, and consequently a small contri- 20. Bader RFW, Stephens ME (1975) J Am Chem Soc 97: 7391
bution for the 3c-4e model. However, the linear 21. Fradera X, Austen MA, Bader RFW (1999) J Phys Chem A
structures showed larger di€erences in the geometry, 103: 304
compared with their parent compounds, and maximum 22. (a) Fujimoto H, Yabuki R, Tamao K, Fukui K (1992)
electron delocalization values (with a noticeable value of Theochem 260: 47; (b) Ziegler T, Gutsev GL (1992) J Chem
0.42 electron pairs for 4) compatible with the 3c-4e bond Phys 96: 7623; (c) Moc J, Morokuma K (1994) Inorg Chem 33:
551; (d) HaÈser M (1996) J Am Chem Soc 118: 7311; (e) King
model. RA, Galbraith JM, Schaefer III HF (1996) J Phys Chem 100:
6061; (f) Landrum GA, Goldberg N, Ho€mann R (1997)
Acknowledgments. Computing time was provided by the Univers- J Chem Soc Dalton Trans 3605; (g) Moc J, Morokuma K (1997)
idad de Granada (Spain). We are grateful to Professors R.F.W. J Mol Struct 436: 401; (h) Ponec R, Duben AJ (1999) J Comput
Bader and B. Silvi for a copy of the AIMPAC and TopMoD Chem 20: 760; (j) Fuster F, Silvi B (2000) Theor Chem Acc
package of programs. We thank Professor E. Sanchez Marcos 104: 13
for their fruitful discussion on the manuscript. We also thank 23. Gillespie RJ, Robinson EA (1996) Angew Chem Int Ed Engl 35:
D. Nesbitt for reviewing the language of the English manuscript. 495
24. Frisch MJ, Trucks GW, Schlegel HB, Scuseria GE, Robb MA,
Cheeseman JR, Zakrzewski VG, Montgomery Jr JA, Stratmann
References RE, Burant JC, Dapprich S, Millam JM, Daniels AD, Kudin
KN, Strain MC, Farkas O, Tomasi J, Barone V, Cossi
1. Gilheany DG (1994) Chem Rev 94: 1339 M, Cammi R, Mennucci B, Pomelli C, Adamo C, Cli€ord S,
2. (a) Akiba K (1999) In: Akiba K (ed), Chemistry of hypervalent Ochterski J, Petersson GA, Ayala PY, Cui Q, Morokuma K,
compounds. Wiley-VCH: New York; (b) Curnow OJ (1998) Malick DK, Rabuck AD, Raghavachari K, Foresman JB,
J Chem Educ 75: 910 Cioslowski J, Ortiz JV, Baboul AG, Stefanov BB, Liu G,
3. (a) Dobado JA, MartõÂ nez-GarcõÂ a H, Molina J, Sundberg MR Liashenko A, Piskorz P, Komaromi I, Gomperts R, Martin RL,
(1998) J Am Chem Soc 120: 8461; (b) Dobado JA, MartõÂ nez- Fox DJ, Keith T, Al-Laham MA, Peng CY, Nanayakkara A,
GarcõÂ a H, Molina J, Sundberg MR (1999) J Am Chem Soc 121: Gonzalez C, Challacombe M, Gill PMW, Johnson B, Chen W,
3151 Wong MW, Andres JL, Gonzalez C, Head-Gordon M, Repl-
4. (a) Dobado JA, MartõÂ nez-GarcõÂ a H, Molina J, Sundberg MR ogle ES, Pople JA (1998) Gaussian 98, Revision A.7. Gaussian,
(2000) J Am Chem Soc 122: 1144; (b) Dobado JA, MartõÂ nez- Inc., Pittsburgh PA
GarcõÂ a H, Molina J (1999) J Inorg Chem 38: 6257 25. Biegler-KoÈnig FW, Bader RFW, Tang T-H (1982) J Comput
5. (a) Reed AE, Weinhold F (1986) J Am Chem Soc 108: 3586; (b) Chem 3: 317
Magnusson E (1990) J Am Chem Soc 112: 7940; (c) Reed AE, 26. Wolinski K, Hilton JF, Pulay P (1990) J Am Chem Soc 112:
Schleyer PvR (1990) J Am Chem Soc 112: 1434; (d) Suidan L, 8251
Badenhoop JK, Glendening ED, Weinhold F (1995) J Chem 27. MORPHY98, a program written by Popelier PLA with a
Educ 72: 583; (e) Kutzelnigg W (1984) Angew Chem Int Ed contribution from Bone RGA, UMIST, Manchester England,
Engl 23: 272 EU (1998)
6. (a) Ciowsloski J, SurjaÂn PR (1992) Theochem 255: 9; (b) 28. Noury S, Krokidis X, Fuster F, Silvi B (1997) ToPMoD
Cioslowski J, Mixon ST (1993) Inorg Chem 32: 3209. package, Universite Pierre et Marie Curie