J. Am. Chem. SOC.

1995,117, 8407-8415

8407

Molecular Mechanics for Coordination Complexes: The Impact

of Adding d-Electron Stabilization Energies
Veronica J. Burton, Robert J. Deeth,*J Christopher M. Kemp, and
Phillip J. Gilbert
Contribution from the Inorganic Computational Chemistry Group, School of Chemistry,
University of Bath, Claverton Down, Bath BA2 7AY, UK
Received January 30, 1995@

Abstract: An electronic energy term for transition metals has been constructed for extending Molecular Mechanics
(MM) to open-shell, Wemer-type coordination compounds. The d-orbital energies from a generalized ligand field
calculation are used to compute the Cellular Ligand Field Stabilization Energy (CLFSE). The CLFSE models the
geometrical effects of the stereochemical activity of d electrons and can be computed for any coordination number,
molecular symmetry, and ligand type. In conjunction with ligand-ligand nonbonding and metal-ligand bond stretch
terms, CLFSEs provide a general framework for incorporating transition metals into MM. An explicit angle-bend
term is not required. After describing the theoretical basis of CLFSEs, the method is illustrated using a range of
six-coordinate high-spin and four-coordinate low-spin Ni amine complexes plus four-, five- and six-coordinate Cutl
amine systems. For the nickel complexes, the spin-state change is modeled simply by changing the d-orbital
occupancies. A single set of force field and CLF parameters simultaneously reproduces the metal coordination for
all ten nickel complexes with overall root-mean-square errors of 0.010 8, in Ni-N bond lengths and 0.621 in
N-Ni-N angles. For the copper compounds, the (slightly modified) force field automatically models the JahnTeller distorted structures of six-coordinate species, the planar coordination in the four-coordinate compounds, and
the distorted geometries of five-coordinate systems. The root-mean-square errors in bond lengths and angles for all
15 Cu molecules are higher (0.024 8, and 0.897, respectively) due to the inherent variability of the structural data.
Copper complexes, especially pentacoordinate ones, are intrinsically flexible or plastic which, with the added
influence of the Jahn-Teller effect, can result in large geometrical changes from relatively, minor crystal packing
effects.

Introduction
Molecular Mechanics (MM) is a popular method for modeling
molecular structure and conformational energies with wellparametrized force fields available for treating many essentially
organic problems in chemistry and
However, the extension of MM to inorganic chemistry and especially
to Transition-Metal (TM) systems presents geater ~hallenges.7-I~
One of the main difficulties with conventional MM treatments
of metal systems is the treatment of high coordination numb e r ~ ~and
. ~the definition of valence angle-bend terms. For
example, cis and trans L-M-L combinations in octahedral
symmetry must be handled differently. Many workers have now
overcome this so-called unique labeling problem. A particularly elegant solution is based on spherical coordinate^.^

* Author to whom all correspondence should be addressed.

Present address:

Department of Chemistry, University of Wanvick,

Coventry CV4 7AL, UK.
@Abstractpublished in Advance ACS Abstracts, July 15, 1995.
(1) Burkert, U.; Allinger, N. L. Molecular Mechanics; ACS Monograph
177; American Chemical Society: Washington, DC, 1982.
(2) Allinger, N. L.; Yuh, Y. H.; Lii, J. H. J . Am. Chem. SOC. 1989,111,
855 1-8566.
(3) Brooks, B. R.; Bruccoleri, R. E.; Olafsen, B. D.; States, D. J.;
Swaminathan, S.; Karplus, M. J . Comput. Chem. 1983, 4, 187-217.
(4) Weiner, P. K.; Kollman, P. A. J . Compur. Chem. 1981, 2, 287-303.
( 5 ) Weiner, S . J.; Kollman, P. A.; Case, D. A,; Singh, U. C.; Ghio, C.;
Alagona, G.; Profeta, S . ; Weiner, P. J . Am. Chem. SOC.1984, 106, 765784.
(6) Mayo, S . L.; Olafson, B. D.; Goddard, W. A. J . Phys. Chem. 1990,
94, 8897-8909.
(7) Allured, V. S.; Kelly, C. M.; Landis, C. R. J . Am. Chem. SOC.1991,
113, 1-12.
(8) Hancock, R. D. Prog. Inorg. Chem. 1989,37, 187-291 and references
therein.

Various metal species have been tackled ranging from

classical Wemer coordination and macrocyclic complexes7through metalloporphyrins to organometallic systems. However, all these methods only tackle part of the problem. None
has yet provided a general method for treating the additional
effects arising from electronic terms. In Wemer complexes,
the electronic effects of an open d shell can lead to severe

(9) Drew, M. G. B.; Rice, D. A.; Silong, S . ; Yates, P. C. J . Chem. SOC.,

Dahon Trans. 1986, 1081-1086. Castonguay, L. A,; Rappe, A. K.;
Casewit, C. J. J.Am. Chem. SOC.1991, 113, 7177-7183. Rappe, A. K.;
Colwell, K. S.; Casewit, C. J. lnorg. Chem. 1993, 32, 3438-3450.
Hambley, T. W.; Hawkins, C. J.; Palmer, J. A,; Snow, M. R. Ausr. J . Chem.
1981, 34, 2525-2542. Yohikawa, Y. J . Comput. Chem. 1990, 11, 326325. Bemhardt, P. V.; Comba, P. lnorg. Chem. 1993, 32, 2798-2803.
Bemhardt, P. V.; Comba, P.; Hambley, T. W. Inorg. Chem. 1993,32,28042809. Drew, M. G. B.; Jutson, N. J.; Mitchell, P. C. H.; Potter, R. J.;
Thomsett, D. J . Chem. Soc., Faraday Trans. 1993,89,3963-3973. Beech,
J.; Cragg, P. J.; Drew, M. G. B. J . Chem. SOC.,Dalron Trans. 1994, 719729. McDougall, G. J.; Hancock, R. D.; Boeyens, J. C. A. J. Chem. SOC.,
Dalton Trans. 1987, 1438-1444.
(IO) Bemhardt, P. V.; Comba, P. lnorg. Chem. 1992, 31, 2638-2644.
( 1 1) Hancock, R. D.; Weaving, J. S.; Marques, H. M. J . Chem. SOC.,
Chem. Commun. 1989, 1176-1178. Muiro, 0. Q.;Bradley, J. C.; Hancock,
R. D.; Marques, H. M.; Marsican, F.; Wade, P. W. J . Am. Chem. SOC.1992,
114,7218-7230. Shelnutt, J. A.; Medforth, C. J.; Berber, M. D.; Barkigia,
K. M.; Smith, K. M. J . Am. Chem. SOC. 1991, 113, 4077-4087. Sparks,
L. D.; Medforth, C. J.; Park, M. S.; Chamberlain, J. R.; Ondrias, M. R.;
Senge, M. 0.;Smith, K. M.; Shelnutt, J. A. J . Am. Chem. SOC.1993, 115,
581-592.
(12) Sironi, A. Inorg. Chem. 1992, 31, 2467-2475. Lauher, J. W. J .
Am. Chem. SOC.1986, 108, 1521-1531. Domain, T. N.; Landis, C. R.;
Bosmich, J. J . Am. Chem. SOC.1992,114,7264-7272. Howeler, U.; Mohr,
R.; Knickmeir, M.; Erker, G. F. Organometallics 1994, 13, 2380-2390.
Rudzinski, J. M.; Osawa, E. J . Phys. Org. Chem. 1993, 107-112. Choi,
M. G.; Brown, T. L. lnorg. Chem. 1993, 32, 5603-5610.
(13) Wiesemann, F.; Teipel, S.; Krebs, B.; Hoeweler, U. Inorg. Chem.
1994, 33, 1891-1898.

0002-786319511517-8407$09.00/00 1995 American Chemical Society

8408 J. Am. Chem. SOC.,Vol. 117, No. 32, 1995

geometrical distortions which are well-known in coordination
chemistry.I4 Of special note are the distorted non-cubic
geometries displayed by the majority of d9 CU" complexes. Sixcoordinate Curl species, for example, are virtually always
tetragonally elongated. Other geometries like those for planar
low-spin d8 species can be seen as the consequence of a JahnTeller distortion of the excited 'E state of the parent octahedral
system.15 Current MM formalisms would be obliged for a cuL6
species, for example, to define different parameter sets for
different metal-ligand bond lengths even though L might be
identical in both cases. In more complicated mixed-ligand
systems, one would have to decide a priori which parameter
set to apply to which ligand and this may not be desirable or
obvious.
Many distortions of coordination complexes can be rationalized in terms of the Jahn-Teller effect or, equivalently, in terms
of the Ligand Field Stabilization Energy (LFSE).I5 The LFSE
is a more general concept and applies beyond orbitally degenerate Jahn-Teller active systems like octahedral d9 Cult complexes. For example, the detailed variation of M-L distances
in a series of first-row metal complexes such as [M(saccharide)2(H20)4I2+l 6 and [M(Mestren)Br]+ l 7 can be traced to the
underlying arrangement of the d electrons. It appears necessary,
therefore, that a general MM treatment of open-shell TM
systems should include explicitly in the strain energy expression
some kind of Ligand Field Stabilization Energy term to account
for this "stereochemical activity" l7 of d electrons.
This paper describes such an energy term and our first
applications. (We presented a preliminary account at the
International Conference on Bioinorganic Chemistry, Oxford,
UK, 1991.18) The method is based on a combination of
conventional Molecular Mechanics and Cellular Ligand Field
(CLF) model of Gerloch and Wo01ley.I~ The so-called Cellular
Ligand Field Stabilization Energy (CLFSE) has several important and novel features. Not only does it provide a metalcentered term which automatically models the geometrical
consequences of the electronic stabilization energy of an
incomplete d shell but the CLFSE is also independent of any
assumptions concerning coordination number, geometry, and
ligand type.
The CLFSENM method can do more than mimic existing
conventional force fields for open-shell Wemer-type coordination compounds. As illustrative examples, we consider the
problem of high- and low-spin Ni" amine complexes and the
treatment of Jahn-Teller effects and the geometries in Cull
amine systems. Independent conventional force fields exist for
each Ni case8 yet, from a CLFSE viewpoint, the only relevant
difference is the d-orbital occupations. Using the same FF and
CLF parameters but with d-orbital populations that reflect the
desired spin state, the CLFSENM method accurately reproduces the structures of a range of complexes. That is, essentially
the same FF can model quite different geometries and Ni-N
interactions. Conventional approaches have also been suggested
for Cu" specie^'^.'^.^^ but can only be applied to limited sets
mainly of four-coordinate molecules. Attempts to treat sixcoordinate species require external constraints to force the
(14) Cotton, F. A.: Wilkinson, G. Advanced Inorganic Chemistry, 5th
ed.; Wiley-Interscience: New York, 1988: pp 766-771.
(15) Deeth, R. J.; Hitchman, M. A. Inorg. Chem. 1986,25, 1225-1233.
(16) Haider, S. Z.; Malik, K. M. A.; Ahmed, K. J.; Hess, H.; Riftel, H.;
Hursthouse, M. B. Inorg. Chim. Acta 1983, 72, 21-27.
(17) Deeth, R. J.; Gerloch, M. Inorg. Chem. 1985, 24, 4490-4493.
(18) Deeth, R. J.; Kemp, C . M.; Gilbert, P. J. Inorg. Biochem. 1991,43,
223-223.
(19) Gerloch, M.; Harding, J. H.; Woolley, R. G. Srruct. Bonding (Berlin)
1981. 46. 1-46.
(20) Comba, P. Coord. Chem. Rev. 1993, 123, 1-48.

Burton et al.

molecule along a tetragonal di~tortion.'~

In contrast, the CLFSEI
MM approach uses a single FF to compute the structures of
four-, five-, and six-coordinate complexes to good accuracy.
Of particular importance is that the Jahn-Teller distortion of
six-coordinate Curr is automatically accounted for.

Theoretical Basis
The general form for the extended CLFSENM strain energy, &,,
is

The terms in eq 1 refer respectively to bond stretch, angle bend,

torsional, nonbonding, and CLFSE interactions, respectively. Each
complex can be divided into a metal-based part and a ligand-based
part. The former comprises the metal and the atoms bound to it, while
the latter comprises the ligand groups. The M-N-L angles span both
parts.
The CLFSE and its derivatives are only computed for displacements
of the ligand atoms. This is reasonable for simple 0-bonding amines
which behave as monatomic ligands. Modifications to include displacements of the second coordination shell (Le. the atoms connected
to the ligand donors) are being developed for unsaturated, n-bonding
ligands and will be reported in due course.
The coordination geometry is specified by the M-L distances and
the L-M-L angles. The bond lengths are treated using a Morse
function

since some bonds, especially for distorted octahedral Cu complexes,

vary dramatically rendering a harmonic expression a poor approximation. Simple repulsion models, such as VSEPR theory21and Kepert's
approach,22show that, in the absence of extra electronic effects, ligandligand interactions determine the angular geometry. The CLFSE
implicitly contains an L-M-L angle bending contribution. However,
the method described here should be general and there are certain d
configurations which have a zero CLFSE (e.g. high-spin d5). Such
cases require an additional term to model the bond angles. One could
use conventional MM methodology with explicit angle bend terms but,
as mentioned above, this approach has its difficulties. Instead, the
whole unique labeling problem' can be avoided by using 1,3 (i.e.
ligand-ligand) nonbonded interactions (eq 3). A similar approach has
been adopted by Comba.20

Hence, the CLFSE/MM model requires three terms for the metal
part of the force field: (i) the CLFSE, (ii) M-L Morse functions for
bond stretching, E,,, (eq l), and (iii) atomic nonbonding terms, ,Evdw,
for ligand-ligand interactions (eq 2). The remainder of the molecule
is treated in a conventional way. The bending and torsional terms
employ the normal harmonic and cosine functions respectively with
certain parameters involving the metal atom set to zero (see Table 4
for a full listing of the FF). Apart from the 1,3 ligand-ligand terms,
Van der Waals interactions are only considered for atom pairs separated
by more than three bonds. In common with other ~ o r k e r s , ~ , ~ J ~ - ~ ~ , * ~
we have not yet included electrostatic interactions in the model.
A hybrid CLFSEJMM program package DOMMINO (D-Orbitals
in Molecular Mechanics of INOrganics) has been developed by adding
routines for computing the CLFSE and its derivatives to in-house MM
software.23 The modifications to the total energy and derivative
calculations are straightforward since the CLFSE part is virtually selfcontained. A simple interface to the SYBYL molecular modeling
(21) Gillesuie. R. J. J . Chem. Educ. 1970. 47. 18-23
(22) Kepe;, D. L. Comprehensive Coordination Chemistry: Pergamon
Press: Oxford, 1987; Vol. 1.
(23) In-house MM software was supplied by Dr. D. J. Osguthorpe,
Molecular Graphics Unit, University of Bath.

J. Am. Chem. Soc., Vol. 117, No. 32, 1995 8409

Molecular Mechanics for Coordination Complexes

18

2.2

24

26

28

Mnd length. r (AI

Figure 1. Schematic representations of the symmetry interactions

between the d orbitals local to a given cell and appropriate bonding
functions. The resulting changer in the d-orbital energies are disdaved
as functions of the CLF parameters e,,. e7". and e,.
package?' has been constructed for inputting structures to DOMMINO
and analyzing the results.
The CLFSE Term. To implement CLFSEs one must compute
derivatives with respect to atomic displacements. For ('normall. MM
energy terms. a simple function is employed which has analytical
derivatives. An analytical derivative of the CLFSE term cannot be
constructed since the CLFSE is computed from the d-orbital energies
of the underlying CLF caIcuIationlq.?' and this requires a 5 x 5
diagonalization.
In the CLF model, the potential sumounding the metal ion is
notionally divided into spatially discrete regions or cells where each
cell normally contains a single metal-ligand bond. The ligand field
perturbation in each cell can then be divided into separate a and n
components which are expressed in terms of the energy parameters e-,,
e,. e, as shown schematically in Figure 1. These parameters are
defined relative to the local M-L coordinate frame. The global ligand
field potential, VW. is then constructed by summing all the local
contributions together, taking due note of the relationship between the
local and global axis frame definitions as described by Gerloch."
The advantages of the CLF model are that (i) the symmetry of the
ligand field is implicit in the ligand coordinates and (ii) the CLF e2
parameters (2 = 0,,m. and ny)refer to individual M-L interactions.
The CLFSEs are therefore general and can handle any ligand type and
coordination environment. This paper deals only with saturated amines
which are the simplest ligand type requiring a single CLF parameter.
e,,. The extension to n-bonding ligands is straightforward and will be
the subject of future publications.
The mathematics of CLF theory is conveniently handled via tensor
operator theory whereupon the d-orbitai matrix elements become:

VLF is the ligand field potential between d orbitals i and j and the

T,,are unitary matrices defining the relationships between local and

global axis frames. Diagonalization of the ensuing 5 x 5 matrix yields
the d-orbital energies, <(d,).from which the CLFSE is computed simply

Figure 2. Graph showing the relationship between e, values for NiN(amine) bonds calculated using eq 6 (solid line) and empirically
observed values from ref 21 (triangles).
is therefore a strictly one-electron term with d-orbital energies that are
automatically barycentered. Different d configurations and spin states
can then be simulated simply by changing the d-orbital occupancies.
Implicit in such a treatment is the assumption that the spin state and d
configuration are well-defined. These restrictions presumably preclude
application of CLFSEs to organometallic systems although. fonunately,
it appears that such species are amenable to more conventional MM
approaches?
The diagonalization of VW precludes any simple analytical expression
for the CLFSE and its derivative. This represents the major difference
between our approach and conventional MM. However. our experience
to date indicates that numerical finite-difference derivatives are
adequate. It is then straightforward to insert the CLFSE code into the
MM program at the points where derivatives and total strain energies
are computed. The CLFSEJMM method described here is thus a
complete generalization of the simple formulas derived by Deeth and
Hitchmanls for treating tetragonal geometries in dP and dn species.
For many systems, the diagonalization will not exact an undue
penalty in computation time. The CLFSE contribution is not evaluated
for a11 atomic displacements. only those relevant to the M-L caordination. In a large metallwnzyme with only a few metal centers. for
example, the calculation would only be slightly slower than for a
comparable "all-organic'' system.
The CLFSE requires an expression relating the M-L bond length
io the CLF parameter value. Both simple theory and experiment suggest
for octahedral complexes that the ligand field splitting parameter lODq
or Am?should have an approximately I/$
dependence on the metalligand bond distance, r.1s.26 For simple 0-bonding-only ligands like
amines, & = 3e, implying that the CLF parameter should also display
a l/r5dependence.
For the particular case of high-spin Ni" complexes, however, it has
been shown" that e, varies linearly with Ni-N distance at least over
the range of about 2.0 to 2.3
Accordingly, the expression for e,, as
a function of Ni-N bond length, r. is

A.

e, = 21629 - 8235r

(6)

as:

(5)
Here. e(d,) are the d-orbital occupations (normally 0. I , or 2). Note
that we adopt a one-electron Ilsmrm,) coupling scheme as employed
in Master Equation M.2 of ref 25 (p 371) rather than the more general
many-electron ILSJM,) scheme of Master Equation M.3. The CLFSE
(24) The SYBYL Molecular Modelling package. Version 6. I. is available
fmm TRlPOS Associates: 1699 South Hanley Road. Suite 303. SI. Louis.
Missouri 63144.
(25) Gerloch. M. Mognnirm ond Ligand Field A m l w i . ~ :Cambridge
Uniterrity Press: New York. 1983.

The fit between empirical e,, values and this straight line is shown
in Figure 2.
Having chosen a functional form for e,. the remaining parameters
ofthe M-N bond stretch were determined by fitting the Morse function
plus CLFSE for a hypothetical Ni-N diatomic to the harmonic
functions reported for high- and low-spin Ni"-N? For a hypothetical
octahedral high-spin d" NiNh complex, the CLFSE is -6/5Am or - I S /
Se., while for a hypothetical low-spin d" NiN. system it is -22/5e,.
Dividing through by six and four respectively. the relevant CLFSEs
for high- and low-spin diatomic systems are - 0 . 6 ~and - 1 . 1 ~ . A
process of trial and error then established the remaining Morse function
(26) Dceth. R. 1.: Gerloch. M. Inorz. Chem. 1984. 23. 3846-3853.
(27) Deeth. R. J.: Gerlach. M. Inorg. Chem. 1987. 26. 2582-2585.

Burton et al.

8410 J. Am. Chem. Soc., Vol. 117, No. 32, 1995

20 1

function, which minimizes at ro, tends to be a computed bond distance

somewhat less than ro.

Results and Discussion

3 1
x

-1004
1.6

1.8

:
2

2.2
2.4
bond IengLh, r (A)

+ DOMMINO

(Sd)

- Morse function

+ DOMMlNO (%I)
-Morse function

:
2.6

:
2.8

I
3

Hancock (S=O)
CLFSE

Hancock (S=l)

CLFSE

Figure 3. Comparison of bond stretch strain energy curves for the

harmonic functions reported by Hancock8 (circles) and the CLFSE plus
Morse functions used in DOMMINO (triangles). The separate contributions of the CLFSE (squares) and the Morse function (solid curved
line) are also displayed. (a) High-spin Ni", S = 1; (b) low-spin Nil1,
s = 0.
parameters of ro = 2.393 A, DO = 80.0 kcal/mol, and a = 0.45 as
capable of approximately reproducing both high- and low-spin harmonic
potential curves in the vicinity of the relevant minima. The requirement
of a single FF capable of treating both spin states simultaneously results
in slight displacements of the CLFSEMM curves relative to the
harmonic functions. A comparison of these various curves is given in
Figure 3.
For Cu, the only changes are a modified ro value of 2.283 A and a
Cu-N-C angular potential with ks = 14.39 kcaymol and 60 = 124'
(see Table 4). We recognize that the use of a linear function for the e,
values of copper complexes is a relatively poor approximation to the
empirically observed variation.26 However, our aim here is to reproduce
structures so that the actual values of the CLF parameters are irrelevant.
Should we wish subsequently to compute d-d transition energies, we
can always adopt the procedures introduced by Comba,20who employs
the Angular Overlap Model (AOM) to calculate various ligand field
quantities (d-d spectra, electron spin resonance g values) after first
determining the structure using more or less conventional MM. Given
the same e i parameter sets and molecular geometries, the AOM and
the CLF model produce identical results.
For both metals, the ro values are larger than the normally anticipated
M-L distances. (See Figure 3.) This is a general feature of the
CLFSEMM method. The CLFSE becomes increasingly negative as
the bond lengths decreases. The balance between the CLFSE, which
gives a (discontinuous) minimum for a zero bond length, and the Morse

Twenty-three amine complexes with saturated hydrocarbon

backbones and two hexaammine species have been located
comprising six octahedral high-spin and four planar low-spin
NilI species plus five six-coordinate, seven four-coordinate, and
three five-coordinate Cu" compounds. All the structures (except
for the [M(NH3)&+ species) were obtained from the Cambridge
Structural Database**and are displayed schematically in Figure
4 and listed in Table 1.
Nickel Complexes. The average observed and calculated
bond lengths and angles for the ten Ni" complexes are given in
Table 2 and, where possible, compared with the results from
the conventional MM treatment of Hancock.8 Individual and
average root-mean-square (rms) errors for these parameters
calculated over the individual bond lengths and angles are given
in the supplementary infomation. (A complete listing of all
the FF parameter values is included in Table 4.)
Overall, the performance of the CLFSENM method with
respect to the metal coordination environment is at least as good
as conventional MM. The average rms errors in bond lengths
and bond angles are only 0.010 8, and 0.62", respectively. For
individual complexes, the computed Ni-N distances are generally within three times the average rms error. The worst
agreement is for [Ni(dptn)2I2+(V) where the average Ni-N
rms error is 0.021 and the computed bond lengths are systematically too short by up to 0.08 A.
(28) Cambridge Structural Database, Version 4. Cambridge Crystallographic Data Centre, University Chemical Laboratory: Cambridge, 1994.
(29) Hummel, H. U.; Beiler, F. Z. Anorg. Allg. Chem. 1988, 565, 147153.
(30) Cramer, R. E.; Van Doome, W.; Huneke, J. T. Inorg. Chem. 1976,
15, 529-535.
(31) Andreetti, G. D.; Cavalca, L.; Sgarabotto, P. Gazz. Chim. Ital. 1971,
101, 494-501.
(32) Biagini, S.; Cannas, M. J. Chem. SOC.A 1970, 2398-2408.
(33) Biagini, S.; Cannas, M. J. Chem. SOC.A 1970, 2398-2408.
(34) Zompa, L. J.; Margulis, T. N. Inorg. Chim. Acta 1978, 28, L157L159.
(35) Stomberg, R. Acta Chem. Scand. 1969, 23, 3498-3512.
(36) Hambley, T. W. J. Chem. SOC.,Dalton Trans. 1986, 565-569.
(37) Ito, T.; Toriumi, K. Acta Cvstallogr. 1981, B37, 88-92.
(38) Hancock, R. D.; Dobson, S. M.; Evers, A,; Ngwenya, M. P.; Wade,
P. W.; Boeyens, J. C. A.; Wainwright, K. P. J. Am. Chem. SOC.1988, 110,
2788-2794.
(39) Distler, T.; Vaughan, P. A. Inorg. Chem. 1967, 6, 126-129.
(40) Bertini, I.; Dapporto, P.; Gatteschi, D.; Scozzafava, A. J . Chem.
SOC.,Dalton Trans. 1979, 1409-1414.
(41) Chaudhuri, P.; Oder, K.; Weighardt, K.; Weiss, J.; Reedijk, J.;
Hinrichs, W.; Wood, J.; Ozarowski, A.; Stratemaier, H.; Reinen, D. Inorg.
Chem. 1986, 25, 2951-2958.
(42) Stephens, F. S. J. Chem. SOC.1969, 2233-2241.
(43) Ammeter, J. H.; Burgi, H. B.; G a p , E.; Meyer-Sandrin, V.; Jensen,
W. P. Inorg. Chem. 1979, 18, 733-750.
(44) Fortier, D. G.;McAuley, A. M. J . Chem. SOC.,Dalton Trans. 1991,
101- 109.
(45) Cannas, M.; Cristini, A.; Marongiu, G. Inorg. Chim. Acta 1976,
19, 241-244.
(46) Lu, T. H.; Tahirov, T. H.; Bin, C. K.; Hsaio, C. D.; Liu, Y. L.; Lee,
J. C.; Chung, C. S. Acta Crystallogr., Sect. C.: Cryst. Struct. Commun.
1993, 49, 1747-1753.
(47) Vrabel, V.; Lokaj, J.; Garaj, J. Collect. Czech. Chem. Commun. 1983,
48, 2893-2902.
(48) Pajunen, A. Suom. Kemistil. E 1969, 42, 15-18.
(49) Pajunen, A. Suom. Kemistil. B 1969, 42, 261-266.
(50) Hamalainen, R. Suom. Kemistil. B 1973, 46, 237-240.
(51) Grenthe, I.; Paoletti, P.; Sandstrom, M.; Glikberg, S. Inorg. Chem.
1979, 18, 2687-2692.
(52) Walsh, A.; Hathaway, B. J. J. Chem. Soc., Dalton Trans. 1984,
15-18.
(53) Kamisawa, K.; Matsumoto, K.; Ooi, S.; Saito, R.; Kidani, Y. Bull.
Chem. SOC.Jpn. 1981, 54, 1072-1076.

Molecular Mechanics for Coordination Complexes

J. Am. Chem. Soc., Vol. 117, No. 32, 1995 8411

Table 1. Chemical Formulas, Full Ligand Names, Coordination Numbers, and Cambridge Structural Database Reference Codes for the
Molecules Shown in Figure 1
formula

no.

coord no.

I
I1
I11
IV
V
VI
VI1
VI11
IX

6
6
6
6
6
6
4
4
4

X
XI
XI1
XI11
XIV

4
6
6
6
6
6

XVI
XVII
XVIII
XIX

5
5
5

xv

XXI
XXII
XXIII
XXIV

xxv

CSD refcode

ethylenediamine
1,3-diaminopropane
diethylenetriamine
bis( 3-aminopropy1)amine
1,4,7-triazacyclononane
ethylenediamine
tetra-N-methylcyclam
1,4,8,11[7R(S),14S(R)]-5,5,7,12,12,13-hexamethyItetraazacyclotetradecane

ENIACH
DAMPNI
AEAMNI 10
AMPRNI 10
BAZNNI
EANBAGOl
DITMUO
MAZNIA

1,4,7,10-tetraazabicyc10[8.2.2]tetradecane

GALZUO

ethylenediamine
1,4,7-triazacycIononane
diethylenetriamine
triaminocyclohexane

CUENCL
DUSJACOl
ETACUB
TACCUP
LATSII
JIBZUP
TENCUB
CEDHAU
DAPRCU
DMEDCU
CMENOX
ETEACU
CEFBEU
CHXCUA

2,5,8,11,14-~entaazapentadecane
1,4-bis(3-aminopropyl)-1,4,7-triazacyclononane
N,N-bis(2-aminoethyl)diethylenetriamine
ethylenediamine
1.3-diaminopropane
(dimethy1amino)ethyl)amine
N-methylethylenediamine
N-ethylethylenediamine

4
4
4
4
4
4
4

xx

ligand name

N,N-diethylethylenediamine

1,3-diaminocyclohexane

VI

XI\'

XI1

xv

XVII

XVI

xx

x IX

El

El

XXIll

XVlIl

Me

tvk

ML
I

XXI

El

XXIV

XXV

Figure 4. Schematic diagrams of the amine complexes used in this

study. See Table 1 for more details.

ref
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
46
44
45
47
48
49

50
51
52

53

A significant feature of the CLFSENM approach is that the

coordination geometry is implicit in the CLFSE. Hence, for
high-spin dx NiN6, the CLFSE maximizes for octahedral
symmetry while for low-spin ds NiN4, the maximum CLFSE is
for a planar geometry. This contrasts with conventional MM
where some form of explicit angle bend or out-of-plane term is
required to enforce planar coordination.
An indication of the method's performance for the rest of
the molecule was obtained by computing the rms errors for NiN-C angles and the remaining "organic" bond lengths and
angles (excluding any involving hydrogen) as well as the
maximum deviation for the Ni-N-C bonds in each complex.
(A full listing is included in the supplementary information.)
The latter is generally less than about 7" except for [Ni(B12aneN4)I2+(X) where one Ni-N-C angle differs by 13.2"
with respect to experiment. As for [Ni(dptn)2I2+,six-membered
chelate rings are involved (vide infra). The average rms error
for the Ni-N-C angle is 0.87" while the average rms errors
for the remaining bond lengths and angles (excluding any
involving H) are 0.008 8, and 0.598", respectively.
Overall, therefore, the agreement with experimental structures
for both high- and low-spin Niu amine complexes is good. There
is a sharp division between high-spin octahedral and low-spin
planar geometries with around a 0.2 8, shortening of the Ni-N
distances for the latter. The achievement of the CLFSENM
data is that only a single set of FF parameters is required in
contrast to conventional MM which employs independent FFs
for high- and low-spin systems.
These results suggest a possible future application of the
CLFSEMM method for modeling spin crossovers. Providing
that the initial and final spin states are well-defined and can be
represented in terms of simple one-electron configurations, one
could imagine following a spin crossover by progressively
mixing the relevant two sets of d-orbital occupations. The
crossover region would then be associated with non-integral
d-orbital populations.
Another important observation concerns a hypothetical "bare
ligand" NiN6 molecule. It has already been suggestedi5 that
the planar geometries of low-spin dx complexes can notionally

8412 J. Am. Chem. Soc., Vol. 117, No. 32, 1995

Table 2. Observed and Calculated Bond Lengths
[Ni(NH3)d2+(1)
calc
obsZ9
Ni-N

(A)

average
N-Ni-N(')

average
Ni-Ni-N
a

(")

(A) and Angles (") Describing the Ni Coordination Environment"

[Ni(en)3J2+(11)
calc
obs30 Han

[Ni(tn)3J2+(111)
calc
obs3'
Han

2.12

2.13

2.12

2.13

2.12

2.14

2.15

90.0
180.0

90.0"
180.P

81.8
173.1

81.9
171.8

84.3

83.8
173.1

86.9
175.8

[Ni(tcn)?l2+(VI)
Ni-N(A)

Burton et al.

[Ni(en)2J2+(VII)

[Ni(dien)?Jzf(IV)
calc
obs32 Han

[Ni(dptn)zI*+(V)
calc
Han

2.15

2.05
2.15

2.06
2.16

2.05
2.16

2.15
2.11

2.23
2.15

2.20
2.11

85.1

81.7
168.0

81.6
167.3

82.9

91.6
177.4

90.4
176.4

91.2

[Ni(tmc)J*+(VIII)

[Ni(meso14ane)12+(IX)

calc

obs34 Han

calc

obs35

Han

calc

obs36

Han

2.11

2.11

2.09

1.94

1.92

1.92

1.99

1.98

1.97

80.7
177.8

82.6
177.1

82.6

87.6
179.7

86.4
180.0

88.5
180.0

90.3
167.4

90.6
168.6

86.4
90.0
166.0 180.0

calc
1.98

obs3'

Han

[Ni(B-l2-aneN4)l2+(X)

calc

obs38

1.96

1.85

1.87

90.0
180.0

89.7
169.8

89.7
169.1

Han

1.86

The experimental values were quoted as average values for the specific bond angles.

be derived from a Jahn-Teller distortion of the 'E excited state

of the octahedral complex. In support of this, a low-spin d8
configuration yields a tetragonally elongated system with NiNe, = 1.90 and Ni-N,, = 2.78 relative to a computed Ni-N
distance of 2.1 A for the parent octahedral case.
Copper Complexes. For d9 copper complexes, the CLFSE/
MM approach automatically generates the tetragonal geometry
for the first-order Jahn-Teller active 2E, state. For the "bare
ligand" system, the Cu-Ne, distance is 1.96 A and Cu-N,, is
2.45 A relative to a computed distance of 2.13 for a regular
octahedron. A contour plot of the total strain energy as a
function of equatorial and axial bond lengths is shown in Figure
5. Others have also reproduced such distortions but with
external constraints to force the molecule along a particular
~ o o r d i n a t e . ' ~In. ~contrast,
~
these distorted geometries arise from
the CLFSE and no constraints are required. This is a significant
advance in molecular modeling of coordination complexes,
especially those of d9 metals.
Copper(II) is probably the hardest metal ion for conventional
MM. Virtually all applications are restricted to four-coordinate
speciesI3 which are generally more straightforward (vide infra).
The CLFSEMM scheme therefore represents the first general
method for handling any coordination number and geometry.
However, before discussing the results, we must consider the
experimental structures to which the computed structures are
compared.
Copper(II) complexes are notoriously "plastic".55 This means
that a range of structures are possible which have similar
energies and the final "observed" solid-state structure may
depend critically on relatively subtle crystal packing forces. As
an example, consider trigonal-bipyramidal (TBP) versus squarepyramidal (SQP) coordination.
In TBP symmetry, there is a hole in the d,2 orbital. The
resulting "stereochemical activity" l 7 of the d-electrons facilitates
a closer approach of the axial ligands relative to, say, a
comparable dIo zinc compound. Thus, the zinc species has
longer axial than equatorial bonds while the relative bond lengths
in d9 copper systems are reversed and one observes two short
axial contacts and three medium equatorial distances. In
contrast, SQP symmetry has the hole in the equatorial dX2+
orbital where the stereochemical activity yields four short
equatorial bonds and a quite long apical contact (see Figure 6).
CLFSEMM calculations illustrate this more quantitatively.
For a model TBP "bare ligand" CuN5, the optimized axial and
equatorial distances are 1.88 and 2.28 A, respectively, compared
(54)Comba, P.; Zimmer, M. Inorg. Chem. 1994, 33, 5368-5369.
(55) Gazo, J.; Bersuker, I. B.; Garaj, J.; Kabesova, M.; Kohout, J.;
Langfelderova, H.; Melnik, M.; Serator, M.; Valach, F. Coord. Chem. Rev.
1916, 19, 253-297.

2.45

CuN(ax)

Figure 5. Contour plot of the total strain energy for a hypothetical

"bare ligand" CuN6 complex as a function of the equatorial and axial
bond lengths. Successive contours are separated by 1 kcal/mol. Solid
lines represent positive energies, long dashes correspond to zero energy,
and short dashes indicate negative energies.

Tr
N '

Figure 6. Schematic representation of the stereochemical activity of

the singly occupied d orbital in d9 CuN5 complexes. The arrows
indicate the motions of the ligands relative to a closed-shell dl* systems.

with the equatorial and apical distances of 2.01 and 2.48 8, for
SQP symmetry. The energies are very similar (the C4" system
is 2 kcaUmol lower in energy). Experimentally, five-coordinate
Cu" frequently display intermediate and variable structures as
shown by the [Cu(2,2'-bipyridyl)2XIn+ complexes where Hathaway has mapped out a series of structures from near trigonal
bipyramidal through to near square pyramidal.56
Thus, unless the calculations explicitly include the effects of
the surrounding lattice, one cannot expect to get exact agreement
with the solid state structure of five-coordinate CUI]complexes
unless the crystal does not alter the structure from what would
be observed in the vacuum phase. Unfortunately, experimental
gas phase structures of these cationic copper complexes are not
available. One could attempt to verify the CLFSE/MM results
( 5 6 ) Hamson, W. D.; Kennedy, D. M.;Power, M.; Sheahan, R.;
Hathaway, B. J. J . Chem. SOC., Dalton Trans. 1981, 1557-1564.

J. Am. Chem. SOC., Vol. 117, No. 32, 1995 8413

Molecular Mechanics for Coordination Complexes

Table 3. Observed and Calculated Bond Lengths (A) and Angles (") Describing the Cu Coordination Environment"

[Cu(NH&I2+(XI)
calc
ob99

[Cu(en)3l2+(XII)
calc
obs40

[Cu(tcn)21z+(XIII)
calc

obs41

Cu-N (A)

2.48
2.04

2.45'
2.15'

2.44
2.44
2.09
2.08
2.03
2.03

2.49
2.33
2.1 1
2.09
2.06
1.91

2.36
2.35
2.12
2.10
2.02
2.01

2.34
2.30
2.08
2.06
2.05
2.05

2.45
2.42
2.04
2.07
2.14
1.98

2.46
2.35
2.13
2.07
2.04
2.03

2.29
2.01

2.35
2.07

average

90.0
179.9

90.W
180.0d

79.1
169.7

80.7
168.9

79.2
177.7

81.3
178.2

80.0
165.3

80.6
166.1

85.2

87.0

N-CU-N (")

[Cu(papd)12+(XVU
CU-N

(A)

N-CU-N (")

average
CU-N

(A)

average
N-CU-N

(")

calc

2.41
2.16
1.97
2.01
2.02
78.6
112.7
109.1
101.7
85.7
170.5
98.9
86.1
145.5
85.3

[Cu(dien)2I2+(XIV)
o

calc

[Cu(apt)I2+(XVII)

[Cu(tach)2I2+(XV)b
s

~ calc
~

OW

[Cu(adt)12+(XVIII)

calc

obs"

calc

2.16
2.09
2.01
2.03
2.02

2.33
1.96
2.10
2.03
2.05

2.25
2.03
2.08
2.06
2.03

2.04
2.14
2.47
2.02
1.97

2.04
2.09
2.09
2.01
2.01

84.9
111.8
104.9
108.0
85.1
167.8
99.6
84.5
140.2
84.4

80.5
94.1
102.3
77.9
89.9
171.9
86.2
174.7
92.7
91.7

84.0
105.3
101.6
81.5
90.3
173.2
86.3
172.9
87.8
94.9

110.0
75.6
101.4
112.4
84.3
143.1
99.1
85.4
169.2
85.8

113.1
83.2
116.7
105.0
86.1
128.2
96.3
86.4
169.6
84.1

[Cu(dmed)?]?+
(XXI)

[Cu(med)2l2+
(XXII)

01~45

[Cu(nen)2I2+

[Cu(deen)212+

[Cu(chn)2I2+

(XXIII)

(XXIV)

(XXV)

[Cu(en)2I2+

[Cu(tn)2lZ+

(XIX)
calc
obs4'

(XX)
calc

cak

obs5I

2.02

2.02

2.00

2.04

2.04

2.06

2.01
2.04

2.01
2.06

2.02
2.03

2.01
2.03

2.02
2.10

2.01
2.08

2.00

2.02

85.4
180.0

84.1
180.0

84.8
180.0

86.7
180.0

$6.7
177.3

85.3
180.0

86.3
180.0

85.9
180.0

86.2
180.0

85.0
180.0

85.2
180.0

84.9
180.0

85.5
180.0

87.6
180.0

s calc
~ ~ o

s calc
~ ~

cak

s cak
~ ~o

For [Cu(tach)2I2+a full set of experimental data was not available for the purpose of comparison with calculated values. The experimental
data for this value was auoted with an error of h0.28. The exDerimenta1 data quoted was an average value only. e The experimental data for this
value was quoted with an error of 10.08.
further either by computing gas phase structures using ab initio
theory or by extending the model to include the crystal lattice
effects. Both of these avenues are being explored.
Apart from the "plasticity" of d9 Cu species, six-coordinate
copper complexes are also subject to the Jahn-Teller effect.
Hence, not only can quite different structures be observed for
essentially the same complex due to plasticity, apparent multiple
structures can also result from static or dynamic Jahn-Teller
effects where the molecule may occupy more than one minimum
on the warped Mexican Hat potential energy ~urface.~'Again,
these features are sensitive to crystal packing. [Cu(tach)2I2+
(XV) with C104- counteranions yields a typical elongated
geometry (Cu-N,, = 2.353 A, Cu-Ne, = 2.071 i442)while
the nitrate salt has virtually equal Cu-N contacts (Cu-N,, =
2.173 A, Cu-Ne, = 2.164 A42). However, the latter structure
is, of course, an artefact of the X-ray diffraction experiment
since the d-d spectra of both salts are essentially identical. The
"intrinsic" structure of CUN6 species is a tetragonally elongated
octahedron and therefore the solid state structures with the
largest tetragonal distortion are used for any comparisons with
CLFSEMM results (see Table 1).
Four-coordinate d9 species are at the limit of a tetragonal
elongation and are therefore less plastic. This makes them the
simplest Cu" systems to model as shown by the number of
previous MM applications. However, crystal packing can still
(57) Reinen, D.; Friebel, C. Srrucr. Bond. 1979, 37, 1-80.

exert an influence. For example, the conformations of the

cyclohexane ring in [Cu(chn)2I2+(XXV)induce a 5" variation
of the ligand bite angle from about 93" for the nitrate s a l P to
87.6" for the bromide compound.52 (Figure 7.) Overall,
therefore, care is required when making comparisons between
computed and "observed" structures.
Nevertheless, the CLFSElMM approach provides satisfactory
structures. The same CLFSE term as for Ni" was used but with
a d9 configuration. A few of the other parameters required
minor modification as described above (see Table 4). Table 3
lists a detailed comparison of Cu-N bond lengths and N-Cu-N
bond angles for all 15 complexes. A full table of rms errors is
available in the supplementary information.
As expected, the four-coordinate complexes, where axial
interactions are minimal, are well reproduced. As for the lowspin d8 systems, the CLFSE automatically favors a planar over
a tetrahedral structure. The worst case is found for the sixmembered chelate rings in [Cu(tn)2I2+(XX)where the Cu-N
bond lengths are systematically underestimated by 0.03 to 0.05
A. In contrast, the agreement for [Cu(deen)2I2+ (XXIV)is
excellent where the Cu-N distances for the substituted (observed 2.08 A, calculated 2.10 A) and unsubstituted (observed
2.01 A, calculated 2.02 A) nitrogens are rather different.
The five-coordinate complexes show a mixed performance
~

(58) Kamisawa, K , Matsumoto, K , 001,S ; Saito, R., IQdani, Y Bull.

Chem Soc Jpn 1981, 54, 1072-1076

Burron er a/.

a=876'

bi934-

Figure 7. Molecular SINCIUWS from X-ray structural analyses of two

fomie of [Cu(chn)d'+showing the variation of the N-CU-N bite angle
as a function of the ligand conformation. The CLFSWMM method
agrees with the structure of the bromide salt (left)rather than the nitrate
salt (right).

Tahle 4. Complete Listing of the Force Field Parameters and

Functional Forms Used in This Work"
Funcrional Fonns
E.,, = DO[
I - exp(-e""-"" 11'
angle bend: E- = (1/2)k,d%- Bo)?
torsion: E,,, = K( I + S(cosno)
van der Waals: E,, = A/#' - B/+
CLFSE e,,= 21629 - 8235r(cm-')

bond stretch:

Bond S I W I C ~

Ni-N

80.0
80.0
72.0

CU-N
N-C

2.393 0.45 N-H

2.283 0.45 C-C
1.49 2.20 C-H

93.0
88.0
108.6

0.91 2.50
1.50 1.92
0.97 1.60

Torsion
K(kcnl\
.
. ~

Ni-N-C-C
Ni-N-C-H

F
.

n
..

0.ooO
O.Oo0

0.0
0.0
O.O(KI 0.0
0.ooO 0.0
0.474 3.0
1.423 3.0

CU-N-C-c

Cu-N-C-H
H-C-C-C
H-C-C-H

0.0
0.0
0.0
0.0

K(kcnl\
.
. ~

H-C-N-H

F
.

n
..

1.423
1.423
0.474
0.474
0.474
0.474

3.0 -1.0
3.0
1.0
3.0
1.0
3.0
1.0
3.0
1.0
3.0
1.0

B(")
64.786
64.786
64.786
51.815
47.497

109.492
110.492
109.492
109.378
108.977

H-C-C-N
C-C-C-N

C-N-C-C

1.0 C-N-C-H
1.0 N-C-C-N

Angle Bend

k. (kcal)
N-Ni-N

0 ("1

N-CU-N

0.ooO
47.497
64.786
28.786

H-N-H

N-C-H
C-N-C

0.ooO

O.Oo0
0.m
108.977
109.492
109.492

ko(kcal)
C-C-C
N-C-C
H-N-C
H-C-C
H-C-H

V"" der Wnnlr

Ni
CU
H

0.0
0.0
654.295

0.0
0.0
41.085

10155.9
14539.0

326.517
397.436

" All energies are calculated in kcal/mol. This requires convening

the CLFSE from wavenumbers

ranging from reasonable for [Cu(apt)]*+ (XVII) to apparently

quite poor for [Cu(adt)]*+ (XVIII) and [Cu(papd)]'+ (XVI).
The CLFSWMM scheme favors SQP structures (see above).
Experimentally, however, only [Cu(apt)]'+ (XVII) could be
even approximately described as square pyramidal?' The other
two complexesJs.* are closer to aigonal bipyramidal which leads
to apparently poor agreement between theory and experiment.

However, we attribute these discrepancies to the subtle effects

of crystal packing. If the metal coordination environment is
fixed at that observed crytallographically and the ligand
minimized, an estimate of the strain energy difference between
the "in crystal" and the "in vacuo" structures is obtained. The
"in crystal" energy is, as expected. always slightly higher but
by only about 1.6 kcallmol for [Cu(papd)J2+(XVI) and [Cu(apt)]'+ (XVII) and by 3.25 kcallmol for [Cu(adt)]'+ (XVIII).
The six-coordinate species are relatively well treated and are
all tetragonally elongated. Given the large experimental errors
associated with [Cu(NHh]*+ (XI) and the apparent uncertainty
regarding its actual exi~tence?~
the data given in Table 3 should
be considered as illustrative only. The computed results for
the other complexes are rather more symmetrical than the
experimental data, as shown, for example, by the Cu-N,,
distances in [Cu(en),]*+ (XII) which are calculated to be
identical (2.44 A) while the observed values differ by 0.16
A similar pattern emerges for [Cu(tcn)2]*+ (XIII) and [Cu(dien)#+ (XIV). Again, packing forces may be to blame.
Alternatively, the "observed solid state smcture may represent
contributions from more than one minimum on the Mexican
Hat potential energy surface.
The greater flexibility of the copper species relative to the
nickel complexes is reflected in somewhat larger average rms
errors of 0.0248, and 0.897" for Cu-N distances and N-Cu-N
angles, respectively. For the "organic" parts of these molecules
(excluding the hydrogen atoms), the agreement for the copper
species is marginally better than that for the Ni complexes. (A
full listing is included in the supplementary information.) The
maximum deviation of any Cu-N-C is about 3' less than the
worst case for a nickel complex. The similarities across all the
molecules studied here presumably reflect the common FF
parameters used for the "organic" parts.

Conclusions
The Cellular Ligand Field Stabilization Energy provides a
method for treating the important geometrical effects associated
with the electronic energies of an incomplete d shell. In
conjunction with bond-stretch and ligand-ligand nonbonding
terms, the CLFSE provides a general framework for explicitly
incorporating open-shell transition metals into a conventional
molecular mechanics treatment for Werner-type complexes. The
absence of an L-M-L angle-bend term facilitates a uniform
treatment of wider ranges of coordination numbers.
The CLFSE term can be designed to mimic existing molecular
mechanics force fields. However, the same CLFSWMM force
field is also able to reproduce simultaneously the quite different
structures and Ni-N distances of octahedral high-spin and
planar low-spin dXNi amine complexes simply by altering the
d-orbital occupations to reflect the desired spin state. The
uniform treatment of both high- and low-spin systems opens
the possibility for following certain types of spin crossover.
With only minor modifications, the CLFSWMM FF provides
a good treatment of four.. five-. and six-coordinate d9 Cu amine
complexes including planar CuN4 systems and the auromaric
generation of tetragonally elongated CuN6 centers. The JahnTeller effect is implicit in the CLFSE and requires no external
constraints. This represents a significant first in molecular
modeling of d9 complexes. For pentacoordinate systems, the
CLFSE favors a square-pyramidal geometry whereas two of the
three complexes available are better described as trigonal
bipyramidal. However. the apparently poorer treatment of fivecoordinate complexes reflects the high flexibility or "plasticity"
(59) See discussion in: Bunon. V.I.: Deeth.
Commun. 1995. 573-574.

R. 1. J. Chrm. SK., Clwn.

J. Am. Chem. SOC., Vol. 117, No. 32, 1995 8415

Molecular Mechanics for Coordination Complexes

of copper complexes which can result in large geometrical
changes for relatively subtle variations in crystal packing.
The CLFSEMM scheme appears to be the first general,
empirical method for modeling a variety of Cu" geometries and
represents a significant advance in molecular mechanics methodology for d9 copper in particular and open-shell coordination
complexes in general. The next phase of this work is to extend
the CLFSE term to n-bonding ligands and a wider variety of
metal centers.

Acknowledgment. We thank the Engineering and Physical

Sciences Research Council [formerly the Science and Engineering Council (SERC)] for financial support (P.J.G.) and for
studentships (C.M.K. and V.J.B.). The calculations were canied
out in part on a SPARC station 2GX donated by Sun Microsystems Computer Corp. and on other hardware provided by

the SERC Computational Science Initiative and the Science and

Materials Computing Committee. We also acknowledge the
support of T W O S Associates in the provision of computer
hardware and the SYBYL molecular modeling package.

Supporting Information Available: Tables of rms errors

in Ni-N and Cu-N bond lengths and N-Ni-N and N-Cu-N
bond angles and rms errors and maximum absolute error for
Ni-N-C and Cu-N-C bond angles (3 pages). This material
is contained in many libraries on microfiche, immediately
follows this article in the microfilm version of the journal, can
be ordered from the ACS, and can be downloaded from the
Internet; see any current masthead page for ordering information
and Internet access instructions.
JA9503138