Sc(III) porphyrins. The molecular structure of two Sc(III) porphyrins and a re-evaluation of the parameters for the molecular mechanics modelling of Sc(III) porphyrins

Article abstract of DOI:10.1016/j.molstruc.2007.02.018

The crystal structures of two new Sc(III) porphyrins, [Sc(TPP)Cl]·2.5(1-chloronaphthalene), (5,10,15,20-tetraphenylporphyrin)-chloro-scandium(III)·2.5(1-chlo ronaphthalene) solvate, (Mo Kα, 0.71073 A?, triclinic system P over(1, ?)_, a = 9.9530(

Full text of DOI:10.1016/j.molstruc.2007.02.018

Available online at www.sciencedirect.com
Journal of Molecular Structure 872 (2008) 47–55
www.elsevier.com/locate/molstruc
Sc(III) porphyrins. The molecular structure of two Sc(III)
porphyrins and a re-evaluation of the parameters for the
molecular mechanics modelling of Sc(III) porphyrins
a,
a
a
a
Alvaro S. de Sousa *, Manuel A. Fernandes , Winston Nxumalo , James L. Balderson ,
a
b
a,
*
Tanya Jefticˇ , Ignacy Cukrowski , Helder M. Marques
a
Molecular Sciences Institute, School of Chemistry, University of the Witwatersrand, PO Wits, Johannesburg 2050, South Africa
b
Department of Chemistry, University of Pretoria, Pretoria 0002, South Africa
Received 18 January 2007; received in revised form 8 February 2007; accepted 10 February 2007
Available online 21 February 2007
Abstract
The crystal structures of two new Sc(III) porphyrins, [Sc(TPP)Cl]Æ2.5(1-chloronaphthalene), (5,10,15,20-tetraphenylporphyrin)-
˚
ꢀ
˚
˚
chloro-scandium(III)Æ2.5(1-chloronaphthalene) solvate, (Mo Ka, 0.71073 A, triclinic system P1;a = 9.9530(2) A, b = 15.4040(3) A,
˚
c = 17.7770(3) A, a = 86.5190(10)ꢀ, b = 89.7680(10)ꢀ, c = 86.9720(10)ꢀ, 13101 independent reflections, R1 = 0.0712) and the dimeric
[l2-(OH)2(Sc(TPP))2], bis-(l-hydroxo)-(5,10,15,20-tetraphenylporphyrin) scandium(III) (Mo Ka, 0.71073 A, monoclinic system C2,
˚
˚
˚
˚
a = 24.2555(16) A, b = 11.1598(7) A, c = 25.6468(17) A, b = 91.980(2)ꢀ, 13084 independent reflections, R1 = 0.0485) are reported. In
˚
[Sc(TPP)Cl] the metal is five-coordinate and the porphyrin is domed with the metal displaced by 0.63 A from the mean porphyrin
towards the axial ClÀ ligand. The average Sc–N bond length is 2.143(3) A, which is shorter than the average bond length of previously
˚
reported structures. Two of the phenyl rings are nearly orthogonal to the porphyrin core and the other two are significantly tilted because
of contacts with 1-chloronaphthalene solvent molecules, and the phenyl rings of neighbouring porphyrins. In [l2-(OH)2(Sc(TPP))2] both
porphyrins are domed, with the metal displaced from the mean porphyrin plane towards the bridging hydroxo ligands. The average Sc–N
˚
bond length is 2.197(12) A, which is in the upper range of Sc–N bond lengths in known Sc(III) porphyrins but not dissimilar to the aver-
age Sc–N bond lengths in another other bis-l2-hydroxo Sc(III) porphyrin, [l2-(OH)2(Sc(OEP))2]. One porphyrin is rotated relative to the
upper porphyrin by 25ꢀ due to steric contacts between the phenyl substituents. We have used these new structures to re-evaluated our
previously reported molecular mechanics force field parameters for modelling Sc(III) porphyrins using the MM2 force field; the training
set was augmented from two to seven structures by using all available Sc(III) porphyrin structures and the two new structures. The mod-
˚
elling reproduces the porphyrin core very accurately; bond lengths are reproduced to within 0.01 A, bond angles to within 0.5ꢀ and tor-
sional angles to within 2ꢀ. The optimum parameters for modelling the Sc(III)–N bond lengths, determined by finding the minimum
difference between the crystallographic and modelling mean bond lengths with the aid of artificial neural network architectures, were
found to be 0.90 0.03 mdyn AÀ1 for the bond force constant and 2.005 0.005 A for the strain-free bond length. Modelling the seven
˚
˚
˚
Sc(III) porphyrins with the new parameters gives an average Sc–N bond length of 2.182 0.018 A, indistinguishable from the
crystallographic mean of 2.181 0.024 A.
˚
ꢁ 2007 Elsevier B.V. All rights reserved.
Keywords: Metalloporphyrins; Scandium; Artificial neutral networks; Force field; Molecular mechanics
1. Introduction
Abbreviations: TPP, 5,10,15,20-tetraphenylporphyrin; TTP, 5,10,15,20-
tetra-p-tolylporphyrin; OEP, 2,3,7,8,12,13,17,18-octaethylporphyrin.
We have previously [1] reported parameters for the
modelling of the porphyrin core of iron porphyrins using
*
Corresponding authors. Fax: +27 11 717 6749.
E-mail addresses: Alvaro.Desousa@wits.ac.za (A.S. de Sousa),
molecular mechanics methods and the MM2 force field
Helder.Marques@wits.ac.za (H.M. Marques).
0022-2860/$ - see front matter ꢁ 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.molstruc.2007.02.018

48
A.S. de Sousa et al. / Journal of Molecular Structure 872 (2008) 47–55
[2]. The development of the parameters was based on the
results of a survey of all crystal structures of iron porphy-
rins available in the Cambridge Structural Database
(CSD) at the time and the experience we had acquired
in the modelling of the cobalt corrinoids [3]. The force
field we developed reproduced porphyrin core bond
We report here the structure of two new Sc(III) porphy-
rins, [Sc(TPP)Cl] and [l2-(OH)2(Sc(TPP))2]. These have
been included in the training set which has therefore been
augmented from 2 to 7 structures. We have re-determined
the optimum Sc(III)–Nporph bond stretching parameter,
and report on this in this paper.
˚
lengths to better than 0.01 A, bond angles to better than
2.5ꢀ and torsional angles to better than 4.5ꢀ of the mean
crystallographic values; these results are within one stan-
dard deviation of the mean of the experimental results.
With very minor modifications, we have found that these
parameters in fact adequately reproduce the structural
parameters of all metalloporphyrins of the first transition
series [4–7].
In an endeavour to obtain a set of reliable parameters
that can be used to explore metalloporphyrins structures
of all metalloporphyrins of the d block, we have focussed
on the development of metal–porphyrin (M–Nporph) bond
stretching parameters and have used artificial neural net-
works to assist in the rapid development of these parame-
ters. We have reported parameters for modelling four
coordinate Cu(II), Co(II), Ni(II), Pd(II) [4] and Zn(II) [5]
porphyrins; five-coordinate Zn(II) porphyrins [6]; and the
metalloporphyrins of Sc, Ti, V and Cr in all oxidation
states and with all coordination geometries of the metal
[6], and of five- and six-coordinate porphyrins of all oxida-
tion states of Mn, Co, Ni and Cu [7].
In some cases, very few structures were available so that
the parameters we have reported were necessarily prelimin-
ary parameters. One such case was for Sc(III) porphyrins
where the training set consisted of only two structures,
[Sc(OEP)(Me)] [8] and [Sc(OEP)(CH(SiMe3)2)] [8] as we
had deliberately excluded dimeric porphyrins (to preclude
porphyrin–porphyrin dimer interactions skewing the struc-
tural data) and porphyrins in which an axial ligand is coor-
dinated in an g5 fashion (we have never modelled such
complexes). We have now extended the training set used
previously for the modelling of Sc(III) porphyrins. An
examination of the structure of dimeric [l2-(OH)2(Sc
2. Methods
2.1. Crystallography
Intensity data were collected on a Bruker APEX II CCD
area detector diffractometer with graphite monochromated
Mo Ka radiation (50 kV, 30 mA) using the APEX 2 [11]
data collection software. The collection method involved
x-scans of width 0.5ꢀ and 512 · 512 bit data frames. Data
reduction was carried out using SAINT+ [12]. Diagrams
and publication material were generated using SHELXTL,
PLATON [13], and ORTEP-3 [14].
2.1.1. [Sc(TPP)Cl]
The crystal structure was solved by direct methods using
SHELXTL [15]. Non-hydrogen atoms were first refined
isotropically followed by anisotropic refinement by full
matrix least-squares calculations based on F2 using SHEL-
XTL. Where possible hydrogen atoms were first located in
the difference map then positioned geometrically and
allowed to ride on their respective parent atoms. One of
the phenyl groups on the main molecule was found to be
disordered and in the final refinements was refined over
two positions with the final occupancies being 0.36(3)
and 0.64(3) for each orientation, respectively. The structure
contains three disordered chloronaphthalene molecules in
three different sites in the structure. The first of these was
refined over two positions using SADI, FLAT, SIMU
and DELU restraints. The second molecule was found to
be disordered in at least three orientations and was refined
in three orientations as rigid bodies. The third molecule
was similarly disordered over at least two orientations over
a centre of inversion making the use of disorder models
extremely difficult. Consequently, the structure was pro-
cessed with SQUEEZE [16], which accounted for 80 elec-
trons, close to the 84 required for a chloronaphthalene
molecule. In the final refinements, the contribution of the
chloronaphthalene molecule to the structure was added
to F(000), crystal density and unit cell contents values.
(OEP))2]
(bis[(l2-hydroxo)-(octaethylporphyrin)-scan-
dium]) [8] showed that the porphyrins are in fact suffi-
˚
ciently far apart (5.05 A) so as to have no undue effect
on each other’s structure. This structure has therefore
now been included in the training set. Two organometallic
g5-coordinated structures, (g5-cyclopentadienyl)-octa-
ethylporphyrin-scandium, [Sc(OEP)(cp)] [9], and (g5-inde-
nyl)-(octaethylporphyrin)-scandium, [Sc(OEP)(ind)] [8],
have been reported. We have shown previously that param-
eters for the equatorial M–Nporph bond are rather insensi-
tive to the value of the parameters for the axial M–Naxial
bond. We have therefore developed preliminary parameters
for modelling these organometallic complexes and also
incorporated the two structures in the training set.
2.1.2. [l2-(OH)2(Sc(OEP))2]
The crystal structure was solved by direct methods using
SHELXTL [15]. Non-hydrogen atoms were first refined
isotropically followed by anisotropic refinement by full
matrix least-squares calculations based on F2 using SHEL-
XTL. Hydrogen atoms were first located in the difference
map then positioned geometrically and allowed to ride on
their respective parent atoms. All but one of the eight phe-
nyl groups in the structure were disordered. Each of these
was refined as rigid bodies over two positions with the
Two other Sc(III) porphyrin structures, [Sc(TTP)Cl] and
[l2-(O)(Sc(TTP))2] have been reported [10] but only incom-
plete coordinates are available in the CSD. We could not
include these in the training set.

A.S. de Sousa et al. / Journal of Molecular Structure 872 (2008) 47–55
49
sum of the final occupancies being constrained to one. In
addition, it was found that the hydroxo groups attached
to the scandium atoms were also disordered. These were
refined over three positions using SADI and FLAT
restraints with the sum of the final occupancies also being
constrained to one. The final occupancies were 0.526(16),
0.129(6) and 0.306(15) for the pairs of atoms O1 and O2,
O1A and O2A, and O1B and O2B, respectively.
nate structural decomposition (NSD) procedure developed
by Jentzen, Song and Shelnutt [18].
2.3. Artificial neural network architectures
To derive the M–Nporph bond parameters, the values
of the strain-free bond length, lo, and the bond stretching
force constant, ks, were varied in a grid-like pattern for
˚
values of (i) lo between 1.85 and 2.10 A and (ii) values
of ks between 0.1 and 1.5 mdyn AÀ1. We defined the error
˚
2.2. Molecular modelling
function as the mean difference between the modelled and
the crystallographically observed M–Nporph bond lengths,
and determined its dependence on lo and ks. This range of
lo and ks provided a preliminary grid upon which to base
the artificial neural networks (ANN) calculations. These
initial points were supplemented with additional points
as the shape of the error function emerged. These calcu-
lations were performed using WinNN32 [19], as previ-
ously described [4–7]. ANN architectures that gave
similar results were used to generate the error response
surfaces and compute the optimum values of lo and ks.
The predicted global minimum was computed by averag-
ing the lo and ks values from those trained ANN architec-
tures where the predicted error in the M–Nporph bond
length was the smallest. The standard deviations for the
lo and ks values and the expected error in the average
computed M–N bond length were calculated from the
ANN results.
Molecular mechanics (MM) modelling was performed
with the version of MM2 [2] called MM+ in the HYPER-
CHEM suite of programmes [17]. The potential energy
functions used have been listed previously [1] and the atom
type definitions are given in Table S1 of the Supplementary
Information. The porphyrin core was modelled with the
parameters that we have previously reported [4,5]. Stan-
dard MM2 parameters were used as required for modelling
organic substituents on the porphyrin ring and the axial
ligands, and no MM+ parameters were used in the
modelling.
Modelling of the bonding to coordinated cyclopentadi-
enyl in [Sc(OEP)(cp)] and to indenyl in [Sc(OEP)(ind)]
was performed as follows. The Sc–C distance observed
˚
crystallographically is 2.51 0.03 A. We set the strain-free
Sc–C bond length to the crystallographic mean and used a
bond stretch constant of 0.50 mdyn AÀ1. Since MM+ uses
˚
a ‘‘points-on-a-sphere’’ model, all bond angles of the type
X–M–Y, where M is greater than 4-coordinate are ignored.
All torsions involving the Sc–C bond were set to zero. This
approach adequately reproduced the Sc–C bond distances
(2.51 0.05 A) with the mean plane through the alkyl
ligand and that through the porphyrin core remaining par-
allel as observed in the solid state structures.
Because of the use of the ‘‘points-on-a-sphere’’ model
implemented in MM+, and which, as previously described
[3,6,7], we have found to be inadequate for modelling the
structures of both cobalamins [3] and metalloporphyrins
[6,7], we added parabolic restraints to the force field to
model bond angles of the type X–Sc–Y; these are listed
in Table S2 of the Supplementary Information. The values
of the bond angles and torsions used were the crystallo-
graphic mean of the seven structures used in the training
set; the force constants were those used previously [6,7].
All dipole moments involving the metal ion were set to
zero. Unless otherwise indicated, the crystal structure was
used as the starting point for the calculations. The strain
energy was minimised using firstly 1000 cycles of a steepest
descent algorithm followed by a Polak–Ribiere conjugate
2.4. Synthesis
The synthesis of (5,10,15,20-tetraphenylporphyrin)-
chloro-scandium(III), [Sc(TPP)Cl], was a modification of
the methods described by Arnold et al. [8] and by Sew-
chock et al. [10]. A mass of 1.1 g H2TPP (Strem, 1.9 mmol)
was added to 13 mL 1-chloronaphthalene (Aldrich). The
solution was degassed using three freeze-thaw cycles,
˚
gradient algorithm with
r.m.s.d. < 0.01 kcal molÀ1
a
convergence criterion of
in the gradient. Minimisa-
À1
˚
A
tion usually occurred within 800 cycles of the latter
algorithm.
The extent of the distortion of each porphyrin from pla-
narity, Doop, was quantified in terms of the normal-coordi-
Fig. 1. UV–visible spectrum of (a) [H2TPP], (b) [Sc(TPP)Cl] and (c)
[l2-(OH)2(Sc(TPP))2] in dichloromethane.

50
A.S. de Sousa et al. / Journal of Molecular Structure 872 (2008) 47–55
Fig. 2. (a) ORTEP diagram of [Sc(TPP)Cl]Æ2.5(1-chloronaphthalene), with the major conformation of the porphyrin and one of the orientations of
1-chloronaphthalene in one of the solvent sites (see Fig. S1 of the Supplementary Information) for a diagram showing all conformations of the porphyrin
and the disordered solvent molecules in two sites. (b) ORTEP diagram showing one conformation of the dimeric porphyrin [l2-(OH)2(Sc(TPP))2]. See
Fig. S1 of the Supplementary Information. Hydrogen atoms have been omitted for clarity.
purged with Ar, refluxed for 2 h under a CaCl2 drying tube,
and then further degassed twice with an Ar purge. ScCl3
(Aldrich), 1.1 g (7.3 mmol), was added to the solution.
The flask was evacuated and purged with Ar, and then
heated and refluxed for 2 h. During the reflux, the reaction
flask was regularly evacuated and purged with Ar to
Table 1
Experimental data for the crystal structure determination of [Sc(TPP)Cl] and [l2-(OH)2(Sc(TPP))2] reported in this work
[Sc(TPP)Cl]
[l2-(OH)2(Sc(TPP))2]
CSD deposit number
Moiety formula
633411
44H28ClN4Sc, 2.5(C10H7Cl)
1099.63
633412
C88H58N8O2Sc2
1349.34
123(2) K
0.71073 A
Monoclinic
C2
a = 24.2555(16) A
b = 11.1598(7) A
C
Formula weight
Temperature (K)
173(2) K
˚
˚
˚
Wavelength (A)
0.71073 A
Triclinic
Crystal system
Space group
Unit cell dimensions/A
ꢀ
P1
˚
˚
˚
a = 9.9530(2) A
˚
˚
b = 15.4040(3) A
˚
˚
c = 17.7770(3) A
c = 25.6468(17) A
a = 86.5190(10)ꢀ
b = 89.7680(10)ꢀ
c = 86.9720(10)ꢀ
2716.67(9)
b = 91.980(2)ꢀ
3
˚
Volume (A )
6938.1(8)
Z
2
4
Dcalc (mg mÀ3
)
1.344
0.356
1136
0.48 · 0.20 · 0.11
1.15 to 28.00
À13 6 h 6 13
À20 6 k 6 20
À23 6 l 6 23
51095
1.292
0.253
2800
0.50 · 0.21 · 0.03
0.79 to 26.00
À25 6 h 6 29
À13 6 k 6 13
À31 6 l 6 31
33248
l (mmÀ1
F(000)
Crystal dimensions (mm3)
h range for data collection/ꢀ
Index ranges
)
Reflections collected
Independent reflections
Refinement method
Data/restraints/parameters
Goodness-of-fit on F2
Final R indices [I > 2r(I)]
13101 [R(int) = 0.0276]
Full-matrix least-squares on F2
13101/1051/670
1.110
R1 = 0.0712,
wR2 = 0.2058
R1 = 0.0859,
wR2 = 0.2190
1.233/À0.864
13084 [I(int) = 0.0596]
Full-matrix least-squares on F2
13084/1403/1148
0.993
R1 = 0.0485
wR2 = 0.0930
R1 = 0.1095
wR2 = 0.1162
0.284 and À0.398
R indices (all data)
3
˚
Largest diff. peak and hole/e A

A.S. de Sousa et al. / Journal of Molecular Structure 872 (2008) 47–55
51
remove the HCl formed. The solution turned violet on
completion of the reaction.
418 nm for [H2TPP] to 422 nm. In the visible region, the
four bands at 515, 552, 592 and 640 nm in the spectrum
of [H2TPP] are replaced by bands at 517, 554 and
593 nm, with the central band being the most intense.
The FT-IR spectrum of [Sc(TPP)Cl] shows a strong band
at 422 cmÀ1, which is characteristic of a Sc–N stretch,
reported at 417 cmÀ1 for [Sc(TTP)Cl] [10]. FAB-MS(M+)
gave the most intense peak at m/z 657.2, which corresponds
to loss of the chloride from the complex.
The product crystallises as purple crystals if the solution
is allowed to cool. Just before that occurred, the solution
was filtered under reduced pressure to remove unreacted
ScCl3 whilst preventing co-precipitation of the complex.
The solution was then allowed to cool down to room tem-
perature and was refrigerated (at 5 ꢀC) for 12 h. The prod-
uct was filtered, washed with 1 mL of 1-chloronapthalene
and 4 · 10 mL aliquots of hexane, and air-dried to obtain
a purple solid (1.148 g, 92% yield). To obtain diffraction-
quality crystals, the product was dissolved in a minimum
volume of degassed (Ar) 1-chloronapthalene and slowly
cooled.
We synthesised [Sc(TPP)Et] using the procedure
described by Arnold et al. [8] for the synthesis of
[Sc(OEP)Me]. All glassware was heated under reduced
pressure to approximately 300 ꢀC with the aid of a heat
gun; on cooling to ambient temperature it was purged with
argon that had been passed through dry silica gel. This pro-
cedure was repeated twice before 5 mL of dry diethylether
was added by injection through a septum and degassed. To
this continuously stirred solution was added a solution of
ethylmagnesium chloride (0.8 mL of 2.0 M, 1.6 mmol) in
diethylether. Anhydrous dioxane (0.16 mL, 1.94 mmol)
was then added dropwise by means of a gas-tight syringe
to the reaction mixture. The solution was stirred for a fur-
ther 8 h to ensure complete formation of diethylmagne-
sium. The ether solvent, in which Et2Mg was dissolved,
was removed under reduced pressure and the reaction flask
purged with argon.
Anhydrous toluene (8 mL) was placed in a separately
dried Schlenk tube and degassed by twice by repeating
freeze-thaw cycles, of approximately 25 min each, under
reduced pressure and purging with argon each time. To
the dry, degassed toluene a non-stoichiometric amount of
[Sc(TPP)Cl] was added (250 mg, 0.104 mmol) followed by
a single degassing freeze-thaw cycle. The resulting violet
solution was cannula transferred to the reaction flask con-
taining the diethylmagnesium and was continuously stirred
for 14 h. Filtration of the reaction solution through a sin-
tered glass frit under inert conditions and removal of all
solvents under reduced pressure resulted in an air-stable
crimson powder of [Sc(TPP)Et] (232.8 mg, 94%).
Several attempts were made at recrystallising
[Sc(TPP)Et] from a variety of solvents. We only obtained
red crystals of bis-l-hydroxo dimeric porphyrin, [l2-
(OH)2(Sc(TPP))2], despite all efforts to exclude moisture.
Attempts to crystallise [Sc(TPP)Et] are continuing and, if
successful, the structure will be reported elsewhere.
3.2. [Sc(TPP)Et]
The UV–visible spectrum in hexane shows the Soret
band at 404 nm with four bands in the visible region
appearing at 513, 549, 589 and 622 nm (Fig. 1). As for
[Sc(TPP)Cl], the FAB-MS(M+) spectrum showed the most
Table 2
˚
Bond lengths (A) and bond angles (ꢀ) of the coordination sphere of Sc(III)
of the Sc(III) porphyrins reported in this work
[Sc(TPP)Cl]
[l2-(OH)2(Sc(TPP))2]
Sc–N(1)
Sc–N(2)
Sc–N(3)
Sc–N(4)
2.145(3)
2.139(3)
2.144(3)
2.145(3)
N(1)–Sc(1)
N(2)–Sc(1)
N(3)–Sc(1)
N(4)–Sc(1)
N(1A)–Sc(2)
N(2A)–Sc(2)
N(3A)–Sc(2)
N(4A)–Sc(2)
O(1)–Sc(1)
O(1)–Sc(2)
O(2)–Sc(2)
O(2)–Sc(1)
2.186(3)
2.210(3)
2.190(3)
2.199(3)
2.204(3)
2.180(3)
2.213(3)
2.192(2)
2.053(8)
2.077(7)
2.044(8)
2.077(7)
82.47(10)
80.64(11)
81.29(11)
82.98(12)
82.48(12)
81.43(11)
80.99(11)
82.44(10)
137.24(12)
134.20(12)
133.96(12)
137.42(11)
69.4(3)
Cl(1)–Sc
2.3482(12)
N(2)–Sc–N(3)
N(2)–Sc–N(1)
N(3)–Sc–N(4)
N(1)–Sc–N(4)
84.79(11)
85.48(11)
85.31(11)
84.79(11)
N(1)–Sc(1)–N(2)
N(1)–Sc(1)–N(4)
N(3)–Sc(1)–N(2)
N(3)–Sc(1)–N(4)
N(2A)–Sc(2)–N(1A)
N(4A)–Sc(2)–N(1A)
N(2A)–Sc(2)–N(3A)
N(4A)–Sc(2)–N(3A)
N(1)–Sc(1)–N(3)
N(4)–Sc(1)–N(2)
N(1A)–Sc(2)–N(3A)
N(2A)–Sc(2)–N(4A)
O(1)–Sc(1)–O(2)
O(2)–Sc(2)–O(1)
O(1)–Sc(1)–N(1)
O(2)–Sc(1)–N(1)
O(1)–Sc(1)–N(2)
O(2)–Sc(1)–N(2)
O(1)–Sc(1)–N(3)
O(2)–Sc(1)–N(3)
O(1)–Sc(1)–N(4)
O(2)–Sc(1)–N(4)
O(1)–Sc(2)–N(1A)
O(2)–Sc(2)–N(1A)
O(1)–Sc(2)–N(2A)
O(2)–Sc(2)–N(2A)
O(1)–Sc(2)–N(3A)
O(2)–Sc(2)–N(3A)
O(1)–Sc(2)–N(4A)
O(2)–Sc(2)–N(4A)
N(3)–Sc–N(1)
N(2)–Sc–N(4)
145.37(12)
146.60(12)
N(2)–Sc–Cl(1)
N(3)–Sc–Cl(1)
N(1)–Sc–Cl(1)
N(4)–Sc–Cl(1)
106.61(9)
107.63(9)
107.00(9)
106.79(9)
69.6(3)
87.1(4)
131.6(4)
88.6(3)
135.6(4)
131.5(4)
85.5(3)
132.4(4)
85.1(3)
80.2(3)
144.9(3)
99.1(4)
118.9(4)
144.8(3)
79.5(3)
3. Results and discussion
3.1. [Sc(TPP)Cl]
116.5(5)
96.0(4)
The UV–visible spectrum of [Sc(TPP)Cl] in dichloro-
methane (Fig. 1) shows a shift in the Soret band from

52
A.S. de Sousa et al. / Journal of Molecular Structure 872 (2008) 47–55
˚
intense peak at m/z 657.2, indicating loss of the axial
ligand. In solution, the complex is air and moisture-sensi-
tive; exposure to the atmosphere results in formation of a
new species, which we have subsequently found to be a
bis-l-hydroxo dimer (see below) as evidenced by the for-
mation of a shoulder on the Soret band at 419 nm when
the spectrum is recorded in dry hexane or dry diethylether
which is left exposed to the atmosphere (Fig. S1 of the Sup-
plementary Information). In the visible region, the band at
622 nm disappears, whilst the other three bands do not
change position, but their intensity drops. The conversion
of [Sc(TPP)Et] to the bis-l-hydroxo dimer is typically com-
plete within 3 h under these conditions.
domed with the metal displaced by 0.63 A from the mean
porphyrin towards the axial ligand. An analysis of the por-
phyrin deformation using the NSD procedure [18] shows a
total deformation of Doop = 1.25 A from planarity, consist-
ing mostly of doming, although there is also some wave
deformation.
The average Sc–N bond length is 2.143(3) A, which is
shorter than the average bond length of previously
reported structures (2.183 0.021 A); indeed, only one
other structure, [Sc(OEP)(CH(SiMe3)2)] [8], has Sc–N
bonds that are of comparable length (two Sc–N bond
˚
˚
˚
˚
˚
lengths are 2.142 A and the other two are 2.150 A).
3.4. Structure of [l2-(OH)2(Sc(TPP))2]
3.3. Structure of [Sc(TPP)Cl]
The dimeric porphyrin crystallises in the C2 space group
ꢀ
[Sc(TPP)Cl] crystallises in the P1 space group
(Table 1). Both porphyrins are significantly distorted from
˚
(R1 = 7.12%, Fig. 2). The experimental data are summa-
rised in Table 1. Some of the bond lengths and angles
involving the metal ion are summarised in Table 2.
Two of the phenyl rings of the porphyrin in [Sc(TPP)Cl]
are nearly orthogonal to the porphyrin core (the torsion
angles about the meso C–phenyl bond are 82ꢀ and 80ꢀ,
respectively) whilst the other two are significantly tilted
(torsion angles are 61ꢀ and 71ꢀ, respectively) because of
contacts between H atoms on the phenyl substituents
and those on the 1-chloronaphthalene solvent molecules,
and between the phenyl rings of two neighbouring
porphyrins.
planarity (Doop = 1.27 and 1.65 A, respectively). As with
[Sc(TPP)Cl], the major deformation is doming; there is
some wave deformation and also saddling of the two
porphyrins.
The average Sc–N bond length (Table 2) is
˚
2.197 0.012 A, which is in the upper range of Sc–N bond
lengths in Sc(III) porphyrins reported prior to this work
˚
(2.181 0.024 A) but not dissimilar to the average Sc–N
bond lengths in the other bis-l2-hydroxo Sc(III) porphyrin
reported previously, [l2-(OH)2(Sc(OEP))2], 2.203
˚
0.008 A [8]. The structures of the two dimeric porphyrins
are similar (Fig. 3) except that, in the case of
[l2-(OH)2(Sc(TPP))2], the lower porphyrin is rotated rela-
tive to the upper porphyrin by 25ꢀ in a clockwise direction
Like all other monomeric Sc(III) porphyrins reported to
date, the metal is five-coordinate and the porphyrin is
Fig. 3. Two views of the structures of (left) [l2-(OH)2(Sc(TPP))2 and (right) of [l2-(OH)2(Sc(OEP))2]. Hydrogen atoms are omitted for clarity.

A.S. de Sousa et al. / Journal of Molecular Structure 872 (2008) 47–55
53
whereas in [l2-(OH)2(Sc(OEP))2] the projection of the two
porphyrins are virtually coincidental (their relative rotation
is less than 2ꢀ). This is almost certainly due to steric con-
tacts between the phenyl substituents in the former, even
though they are significantly tilted from the vertical (the
mean torsion about the meso C–phenyl bond is 67 5ꢀ).
The tilting of the phenyl groups is also likely to be due to
contacts between hydrogen atoms of the phenyl groups
on neighbouring molecules in the structure, i.e., packing
effects.
the optimum values for the Sc–N bond parameters, ks
˚
and lo, they were varied between 1.75 and 2.15 A, and 0.1
and 1.5 mdyn AÀ1, respectively. In a plot of lo against ks
˚
(Fig. 4a), the error function has a valley of minimum values
that runs diagonally from (large ks, long lo) to (small ks,
short lo), showing that the two parameters are correlated.
À1
˚
Thus, for example, a value of ks = 1.35 mdyn A and
˚
˚
lo = 2.10 A produces a mean value of 2.172 0.007 A for
˚
the Sc–N bond length, or a difference of 0.009 A between
this mean value and the crystallographic mean, well within
˚
the experimental standard deviation of 0.024 A. A value of
À1
˚
˚
˚
3.5. Molecular modelling
ks = 0.30 mdyn A and lo = 1.80 A gives 2.187 0.008 A
˚
for the Sc–N bond length, which is merely 0.006 A different
The average Sc–N bond length in the seven porphyrin
structures on which the molecular modelling was based (5
monomeric porphyrins, 2 dimeric porphyrins, 36 Sc–N
from the crystallographic mean. This strong correlation
between ks and lo, and the fact that there is an appreciable
variation in the range of experimentally observed Sc–N
bond lengths, means it is difficult to pinpoint the best val-
˚
bond lengths reported) is 2.181 0.024 A. To determine
À1
˚
Fig. 4. ANN fit of the error function for the Sc–N bond length in the modelling of Sc(III) porphyrins. The minimum occurs at ks = 0.90 mdyn A and
˚
lo = 2.005 A. (a) A contour plot and (b) the three-dimensional plot.

54
A.S. de Sousa et al. / Journal of Molecular Structure 872 (2008) 47–55
Fig. 5. Overlay of the molecular mechanics (—) and crystal structure (...) of (a and c) [l2-(OH)2(Sc(OEP))2]; (b and d) [l2-(OH)2(Sc(TPP))2]; (e)
[Sc(OEP)(cp)]; and (f) [Sc(TPP)Cl]. Hydrogen atom are omitted for clarity.
ues of ks and lo. The trained ANN architectures predicted
The porphyrin core is very well reproduced by the force
field, providing further evidence of the adequacy of the
parameters we have developed for modelling metallopor-
phyrins. Bond lengths are reproduced to 0.01 A or better,
bond angles to within 0.5ꢀ and torsional angles to within
2ꢀ Table S3 of the Supplementary Information.
the minimum in the erroÀr1function (Fig. 4b) to occur with
˚
˚
and lo = 2.005 0.005 A.
ks = 0.90 0.03 mdyn A
˚
Modelling the seven Sc(III) porphyrins with these values
˚
gives an average Sc–N bond length of 2.182 0.018 A, in
excellent agreement with the crystallographic mean of
˚
2.181 0.024 A.
The crystal structures of the porphyrins are well repro-
duced; Fig. 5 shows four as an example. In the case of
[l2-(OH)2(Sc(TPP))2], one porphyrin of the dimer is twisted
more relative to the other than observed in the solid state.
In the case of all TPP structures, the phenyl rings minimise
into structures where they are orthogonal to the porphyrin
ring whereas in the crystal structures they are invariably
tilted away from the normal because of close contacts with
other molecules in the lattice. As we have found before [6],
if [l2-(OH)2(Sc(TPP))2] is energy minimised in a lattice
where all nearest neighbours are confined to their lattice
positions, then the structure observed in the solid state is
indeed reproduced closely.
4. Conclusion
We previously determined parameters for modelling
Sc(III)–N bonds in Sc(III) porphyrins, based on only two
À1
˚
˚
structures, to be ks = 0.90 mydn A and lo = 1.95 A. We
have now increased the training set to seven structures by
including all available Sc(III) porphyrin structures as well
as providing two novel structures. This significantly
increases the confidence in the parameters for modelling
Sc(III)–N bonds in Sc(III) porphyrins. We find that ka
and lo are heavily correlated and that a wide range of val-
ues of ks and lo will reproduce Sc(III) porphyrin structures
reasonably accurately. The optimum values found for the

A.S. de Sousa et al. / Journal of Molecular Structure 872 (2008) 47–55
55
two parameters in this work are 0.90 0.03 mdyn AÀ1 and
with this article can be found, in the online version, at
doi:10.1016/j.molstruc.2007.02.018.
˚
˚
2.005 0.005 A, respectively. The porphyrin core is accu-
rately reproduced with the parameters we have previously
reported for modelling metalloporphyrins. Modelling the
seven Sc(III) porphyrins with the new parameters gives
References
˚
an average Sc–N bond length of 2.182 0.018 A, indistin-
[1] H.M. Marques, O.Q. Munro, N.E. Grimmer, D.C. Levendis, F.
Marsicano, T. Markoulides, J. Chem. Soc. Faraday Trans. 91 (1995)
1741.
˚
guishable the crystallographic mean of 2.181 0.024 A.
[2] N.L. Allinger, J. Am. Chem. Soc. 99 (1977) 8127.
[3] H.M. Marques, K.L. Brown, Coord. Chem. Rev. 225 (2002) 123.
[4] H.M. Marques, I. Cukrowski, Phys. Chem. Chem. Phys. 4 (2002)
5878.
[5] H.M. Marques, I. Cukrowski, Phys. Chem. Chem. Phys. 5 (2003)
5499.
Acknowledgements
This work was made possible by grants from the Na-
tional Research Foundation, Pretoria, South Africa, and
the University Research Committee of the University of
the Witwatersrand.
[6] C.E. Skopec, J.M. Robinson, I. Cukrowski, H.M. Marques, J. Mol.
Struct. 738 (2005) 67.
[7] C.E. Skopec, I. Cukrowski, H.M. Marques, J. Mol. Struct. 783 (2006)
21.
Appendix A. Supplementary data
[8] J. Arnold, C.G. Hoffman, D.Y. Dawson, F.J. Hollander, Organo-
metallics 12 (1993) 3645.
[9] J. Arnold, C.G. Hoffman, J. Am. Chem. Soc. 112 (1990) 8620.
[10] M.G. Sewchock, R.C. Haushalter, J.S. Merola, Inorg. Chim. Acta
144 (1988) 47.
[11] Bruker, APEX2, V. 2.0-1, Bruker AXS Inc., Madison, Wisconsin,
USA, 2005.
[12] Bruker, SAINT+, V. 6.02 (includes XPREP and SADABS), Bruker
AXS Inc., Madison, Wisconsin, USA, 1999.
Supplementary Information available, giving definition
of atom types; parameters for the parabolic restraints
added to model the coordination sphere of the metallopor-
phyrins; comparison of mean structural metrics of crystal-
lographic and molecular mechanics modelled structures of
seven Sc(III) porphyrins. Figures shown ORTEP diagrams
of [Sc(TPP)Cl]Æ2.5(1-chloronaphthalene), in which there is
substantial disorder in the solvent molecules and in one
of the phenyl rings of the TPP core, and of [l2-
(OH)2(Sc(TPP))2], in which seven of the eight the phenyl
rings are distributed over two sites and the hydroxo ligands
have been refined over three positions. Another figure
shows UV–visible spectra of the conversion of [Sc(TPP)Et]
to [l-(OH)2(Sc(TPP))2] on exposure of a dry hexane solu-
tion to the atmosphere. Supplementary data associated
[13] A.L. Spek, J. Appl. Cryst. 36 (2003) 7.
[14] L.J. Farrugia, J. Appl. Cryst. 30 (1997) 565.
[15] Bruker, SHELXTL, V. 5.1 (includes XS, XL, XP, XSHELL), Bruker
AXS Inc., Madison, Wisconsin, USA, 1999.
[16] P. van der Sluis, A.L. Spek, Acta Cryst. A46 (1990) 194.
[17] HYPERCHEM, V. 7.03, Hypercube, Inc., Gainesville, FL, 2002.
[18] W. Jentzen, X.-Z. Song, J.A. Shelnutt, J. Phys. Chem. B 101 (1997)
1684.
[19] Y. Danon, WinNN32, V. 1.6, <http://www.geocities.com/sciware/>,
e-mail: <danony@rpi.edu/>.