Você está na página 1de 10

<

<

Molecular Geometry and Symmetry from


a Differential Geometry Viewpoint
ZBIGNIEW ZIMPEL, PAUL G. MEZEY
Mathematical Chemistry Unit, Department of Chemistry, University of Saskatchewan, Saskatoon,
Saskatchewan, Canada S7N 5C9
Received 1 July 1996; revised 30 January 1997; accepted 25 February 1997

ABSTRACT: Relations between an earlier generalization of molecular symmetry


called symmorphy and a molecular equivalence based on diffeomorphisms of electron
density functional graphs the so-called DFG equivalence introduced in our previous
work. are analyzed. Any two DFG-equivalent electron density functions can be derived
from one another by a suitable transformation of the spatial coordinates and the
electronic charge density scale; the classes of DFG equivalence are the orbits of a group of
linear operators operating in the space of electron density functions. Within the
symmorphy framework, the symmetry group is derived from the symmorphy group by
taking an intersection of a subgroup of the symmorphy group and the group of
isometries for a natural choice of the Riemannian metric tensor. The Riemannian metric
properties provide a choice for a suitable reference electron density function for each
class of equivalent densities. Such reference densities serve as tools for a systematic
classification of the infinite family of electron densities of molecular conformations.
Q 1997 John Wiley & Sons, Inc. Int J Quant Chem 64: 669]678, 1997

Key words: Molecular geometry; symmetry; symmorphy; topology; electron density

Introduction

his article describes a natural continuation of


our earlier study of a special four-dimensional 4D. representation of molecular electron
densities w 1x , motivated by the recent interest in
molecular similarity involving a variety of models

Correspondence to: P. G. Mezey.


Grant sponsor: Natural Sciences and Engineering Research
Council of Canada.
International Journal of Quantum Chemistry, Vol. 64, 669]678 (1997)
Q 1997 John Wiley & Sons, Inc.

and a rather wide range of approaches w 2]69x . In


the introduction, an informal description of the
background as well as a brief review of the concepts and the terminology are given. This is followed by the main topic of the work: classifying
molecular electron densities into equivalence classes based on transformations of the 4D functions
describing these electron densities.
In our previous study w 1x a representation of
molecular electron densities has been described
that combines the advantages of two earlier models: the density domain approach to chemical
bonding w 27, 41, 46, 49, 59, 68x and a 4D model of
CCC 0020-7608 / 97 / 060669-10

ZIMPEL AND MEZEY


molecular electron densities w 25, 26x where augmenting the three spatial coordinates, the density
value corresponds to the fourth dimension. As it
has been pointed out earlier w 25x , the 4D model is
not always advantageous when compared to the
conventional shape analysis technique of using
two-dimensional 2D. contour surfaces of electron
densities embedded in the ordinary three-dimensional 3D. space, nevertheless, such a 4D model
gives a concise curvature characterization for electron densities w 26x and a 4D extension of the shape
group methodology w 46x . A combination of the
density domain model and the 4D shape analysis
has been used in our previous study w 1x to explore
the properties of an equivalence relation between
molecules, based on a transformation in 4D.
In the present article, we shall use this equivalence relation as a tool for investigating the connections between the conventional concept of spatial symmetry and a generalization of symmetry,
called symmorphy w 69x . At first sight, this connection is rather trivial: symmorphy has been introduced originally as a straightforward generalization of symmetry w 69x , by replacing a restricted
family of linear operators symmetry operators. by
the family of all homeomorphisms of the space
which transform the molecular electron density r
into a density r X that is indistinguishable from the
original density r . These homeomorphisms preserve the morphology of the electron density and
are the symmorphy transformations of the molecule w 69x . Clearly, the symmorphy transformations
of any given molecule form a group, called the
symmorphy group of the molecule w 69x . There exist,
however, several, more subtle relations between
symmorphy and symmetry. For example, various
restrictions of symmorphy groups are possible using fixed point theory w 26x which imply special
connections to symmetry, and symmorphy groups
also provide tools for more general transformations involving seemingly intractable additivity
problems w 70x .
The actual representation used in this study
may be regarded as a tool for molecular similarity
analysis, based on diffeomorphisms of electron
density functional graphs introduced in our previous work, leading to the so-called DFG equivalence of electron densities. In the next section of
the present study some properties of the DFG
transformations of electron densities are described.
The third section is concerned with the proof of the
existence and a discussion of the uniqueness of the
representation of a DFG transformation between

670

two sufficiently similar electron densities, involving a transformation of spatial coordinates


combined with an appropriate rescaling of the
values of the electron density function. In the
fourth section of this report the properties of the
steps in the transformation are discussed and analyzed from a geometrical perspective.

DFG Transformations and


DFG Equivalence
As in Ref. w 1x , the 4D representations of electron
density functions are called functional graphs not
to be confused with the customary molecular
graphs of chemical graph theory.. These functional
graphs are assumed to be smooth 3D manifolds
hypersurfaces. embedded in a 4D Euclidean space.
Using the acronym DFG constructed from the initials of diffeomorphism and functional graph, we
shall call a mapping from one electron density
functional graph into another a DFG transformation if and only if this mapping is a differentiable
and bijective transformation a diffeomorphism of
smooth manifolds. and if it maps section sets of
one functional graph onto section sets of the other
functional graph. Following the definition in Ref.
w 1x , we shall regard two electron density functions
DFG equivalent if there exits a DFG transformation of one electron density functional graph onto
the other. This will be shown to be equivalent to
the common sense idea of deformability using the
analogy of rubber geometry, in which two
molecules built from layers of electron density
level sets MIDCOs. w 27, 46x are considered homotopically equivalent if one molecule can be
smoothly reshaped into the other. Obviously, such
a reshaping transformation does not allow the isodensity surfaces to be cut or glued, which guarantees that the topological hierarchy generated by
the density thresholds i.e., by the heights of level
sets. is preserved. In this work we present an
elementary proof and study the consequences of
the fact that any two DFG-equivalent electron densities can be obtained from one another by a transformation of spatial coordinates and a transformation of the electron density scale. Thus, the DFG
transformations of functional graphs can be naturally extended to the whole space R 4 embedding
the functional graphs. These extensions of DFG
transformations form a group. The classes of
DFG-equivalent electron density functional graphs

VOL. 64, NO. 6

MOLECULAR GEOMETRY AND SYMMETRY


are orbits of this group. The DFG transformations
generate linear operators acting in the space of
electron density functions. These operators also
form a group homomorphic to the group of DFG
transformations of functional graphs and the DFGequivalent density functions form orbits of this
group of operators.
The DFG equivalence identifies a particular aspect of the general problem of molecular similarity. The DFG approach can be viewed as a method
for the study of molecular similarity based on a 4D
version of density domain analysis w 27, 41, 46, 49,
59, 61x , and focusing on a combination of continuum and discrete features of electronic densities.
If a functional graph of an electron density is
invariant under a transformation, this transformation can always be replaced by a transformation of
the spatial coordinates of points of equal electron
densities without any modification of the scale of
the electron density values. Such a transformation
of space was defined as a symmorphy transformation of the given electronic density w 69x , obtained
as a generalization of spatial symmetry. Symmorphy transformations have been studied as tools for
shape analysis w 26x and as a possible basis of more
general transformations involving additivity problems w 70x . The symmorphy transformations form a
groupthe symmorphy group of the molecule.
Symmorphy was introduced within a metric space
setting as generalizations of symmetry transformations; the molecular electron density is invariant
with respect to these transformations, and all symmetry operations of a molecule are also symmorphy operations of the same molecule. As the very
word implies, the original meaning of symmetry
involves a metric; without involving metric properties of the space, the identification of symmetry
transformations among the symmorphy transformations is not a simple task. However, a general
treatment can be formulated by introducing a Riemannian metric tensor w 71x and restricting the
symmorphy group to the isometries with respect
to the distance function generated by this metric
tensor. This leads to a general method for the
identification of the symmetry transformations, reducing these special symmorphies to the conventional representation of molecular symmetry. In
this study, the symmetry groups introduced in this
way are classified for all flat Riemannian geometries, which are compatible with the standard Euclidean metric structure of a differential 3D manifold with a single global system of coordinates.
The symmetry groups are shown to be isomorphic

with the subgroups of the orthogonal group.


Within a unifying scheme based on the geometrical viewpoint, the differences between any two
DFG-equivalent electron densities can be interpreted just as two different geometries of the same
generalized electron density.
These results provide a general framework for
molecular similarity analysis in terms of electron
densities, following some of the methods which
have been used earlier for symmetry analysis, but
now the scope can be broadened to more general
shape features. Based on the results, both the characterization and the classification of electron densities involve natural similarity concepts which can
be identified using the symmorphy symmetry formalism. The functions retain all information represented by standard molecular Hamiltonians, including electronic kinetic energy terms and electrostatic interactions between electrons and nuclei.
In general, small changes of the nuclear configuration are expected to lead to small changes in the
electron density distribution. In earlier studies, this
property of electron densities was the basis of
several approaches proposed for the generation of
approximate electron densities w 68, 72, 73x . These
approaches include the DER dimension expansion-reduction. method w 68x , the WAT weighted
affine transformations. method w 68x , and a stepwise technique w 72x based on a combination of the
Lowdin
transform and inverse Lowdin
transform

of density matrices generated using the ADMA


adjustable density matrix assembler. method w 73x .
The primary goal of the DER, WAT, and stepwise
approaches was the estimation of electron densities for a slightly distorted nuclear geometry from
the electron density calculated for a reference nuclear geometry.
Within the formalism of this work, the difference between two electron densities is regarded
nonessential if both belong to the same class of
DFG equivalence. Using a relatively simple test
of the DFG criteria, one is able to identify all
nonessential changes of a given molecular electron
density. If small variations of nuclear geometries
are considered, then the DFG equivalence and any
one of the algorithms for appropriate modifications of electron densities can complement the
lengthy ab initio computations. This is especially
desirable when modeling electronic densities of
molecules at finite temperatures w 74x , when the
nuclei are not treated as fixed in a certain configuration.

INTERNATIONAL JOURNAL OF QUANTUM CHEMISTRY

671

ZIMPEL AND MEZEY

Characterization OF DFG-Equivalent
Density Functions
Let r be an arbitrary electron density function
with a set of rather general properties as defined in
Ref. w 1x . The functional graph of r is defined as
M s  r, r r .. r g R 3 4 .

1.

There is a natural differential structure on M defined by the single global map M, pr M . with
pr: R 4 R 3 ,

pr r, a . s r,

2.

being the projection from R 4 to R 3. The inverse of


pr M is id, r . with id being the identity in R 3.
The manifold M can be decomposed into a family
of section sets G a. =  a4 , where
G a . s  r g R 3 r, a . g M 4

3.

are ordinary level sets of r at heights a.


We say that two electron density functions r
and r X are DFG equivalent if there exists a diffeomorphism F of the functional graph M onto M X ,
which induces a one-to-one correspondence, a
mapping F#, between their level sets or, equivalently, between section sets of their functional
graphs.. This diffeomorphism will be referred to as
a DFG transformation. We shall see that this transformation preserves the relative heights of level
sets. Our objective in this section is to find a
characterization of this equivalence relation in
terms of some natural properties of the electron
density functions. This will also lead to a physical
interpretation of the DFG equivalence.
The following two statements, I. and II., form
the basis for such a characterization:
I. Let A denote the height of the highest critical level of r . For every diffeomorphism y:
R 3 R 3 and every diffeomorphism h:
w 0, A x w 0, AX x such that h0. s 0 and h A.
s AX , the function

r X s h( r ( yy1

4.

is a generalized electron density function


equivalent to r .
II. For every pair of DFG-equivalent electron
density functions r and r X , with A and AX
being the heights of the highest critical lev-

672

els of r and r X , respectively, there exist:


(i) a diffeomorphism y: R 3 R 3, and
(ii) a diffeomorphism h: w 0, A x w 0, AX x
with h0. s 0 and h A. s AX such that
r X can be derived from r using Eq. 4..
Although these statements may appear as a natural consequence of the definition of the transformations chosen, they are fundamental for our subsequent arguments, and we shall present formal
proofs for both.
I. Given two diffeomorphisms y: R 3 R 3 and
h: w 0, A x w 0, AX x it is easy to show that r X s
h( r ( yy1 satisfies all conditions a. ] g. of Ref. w 1x ,
except for the normalization condition b. see the
paragraph following this proof .. This is equivalent
to saying that r X is strictly positive w satisfying a.x ,
has the same critical points as r w satisfying c., f.,
and g.x , is smooth w satisfying e.x , and vanishes at
infinity w satisfying d.x . Each of these assertions
naturally follows from the properties of compositions of diffeomorphisms; for example, to see that
d. is true, one should notice that since r is vanishing at infinity, then for every ) 0 there is a
sphere B in R 3 such that for every r f B one has
r r . F hy1 .; hence, for every r f y B . one finds
r X r. F . The same inequality evidently holds for
every r taken from the outside of a sphere containing y B .. This proves that r X is a not necessarily
normalized. generalized electron density function.
Moreover, the mapping
F : r, r r .. y r . , h( r r ..

5.

is a diffeomorphism from the functional graph M


of r onto the functional graph M X of r X . w In the
global coordinates defined by the maps M, pr M .
and M X , pr M X ., F can be identified with y ..
Next, for every r g G a. we obtain F r, a. s
y r., h a.. which proves that F maps G a. =  a4
into y G a.. =  h a.4 , which is a section set of M X
at height h a., since h is a strictly increasing function. Therefore the restriction F G a. =  a4 defines the transformation F# of level sets G a..
Consequently, the two electron density functions r and r X are DFG equivalent.
II. Inversely, given two DFG-equivalent electron density functions r and r X and a diffeomorphism F of their functional graphs M and M X , let
us define the diffeomorphism y as
y s pr M X . ( F ( id, r . .

6.

VOL. 64, NO. 6

MOLECULAR GEOMETRY AND SYMMETRY

GX aX . s F# G a .. s pr M X . ( F G a . =  a4 .
s pr M X . ( F ( pr M .

y1

( pr M . G a . =  a4 .
s pr M X . ( F ( id, r .
( pr M . G a . =  a4 .

The diffeomorphism y can be treated as a system of new coordinates, a new global map R 3, y .
on the Euclidean space R 3 replacing the standard
global map R 3 , id.. This defines a new global map
M, y (pr M .. on M. Similarly, the function h
can be treated as a new scale applied to the level
heights of r . From this point of view, the electron
density r X equivalent to r represents the same
density r in the new system of space coordinates
y and in a new scale h. This can be summarized by
the following diagram:
M

7.

where, usually, n is the number of electrons in the


molecule. Note, however, that for the purposes of
shape comparisons between different molecules or
different molecular fragments, it is sometimes advantageous to choose an n value that differs from
the actual number of electrons. Such artificial normalization is a useful tool in the comparison of
quantum chemical functional groups w 68x . The requirement of preserving the normalization condition can be included into the definition of DFG
equivalence: two DFG equivalent electron densities must integrate to the same number of electrons. In such a case, the electron charge conservation imposes an additional condition on the
transformations of the preceding statements, and a
multiplication by a normalization constant must
be included in the definition of h in order to
satisfy the normalization condition w Eq. 7.x .

pr M

pr M

Hr r. dr s n,

MX .

Moreover, if G a. is a level set of r at height a,


then y G a.. is also the level set of r Y at the same
height a. Consequently, r X and r Y have an isomorphy-preserving, one-to-one assignment of level
sets. Hence, there exists a bijection h: w 0, A x
w 0, AX x , h a. s aX , such that r X s h( r Y s h( r ( yy1.
Obviously, the function h must be monotonic, and
because every electron density decreases to zero at
infinity, it must be monotonically increasing. This
implies that h0. s 0 and h A. s AX .
Finally, if h were not differentiable at some
height a g w 0, A x , then r X could not be differentiable at any point r g G a.; hence h must be
differentiable. The same argument applied to hy1
leads to the conclusion that h is, indeed, a diffeomorphism.
The electron density functions are usually normalized see condition b. of Ref. w 1x. as

s y ( pr M . G a . =  a4 . s y G a .. .

8.
X

According to statement I, the function r Y s r ( yy1


is an electron density function DFG equivalent to
r . Let GX aX . be a level set of r X such that GX aX . s
F# G a... Then

In Figure 1 two series of level sets of the molecule


chlorobromomethane are shown to illustrate the
concept of equivalence of generalized. electron
densities. Both series of level sets were obtained
from electron densities computed using the ab
initio self-consistent field Hartree]Fock SCF HF.
method with a 3-21G** basis for the bromine atom
and the 6-31G** basis for the chlorine, carbon, and
hydrogen atoms, and formally normalized to 10
electrons. The first series column A. was computed for the standard Euclidean coordinates
x 1 , x 2 , x 3 .. The second series column B. was obtained in a new system of coordinates y 1 , y 2 , y 3 .,
which are first twisted along the x 3 axis parallel
to the hydrogen]hydrogen axis and passing
through the carbon atom. by an angle ax 3 with
a s pr30. proportional to the distance x 3 of the
point x 1 , x 2 , x 3 . from the x 1 , x 2 . plane and then
stretched along the y 1 axis:
y 1 s 32 .w x 1cos ax 3 . q x 2 sin ax 3 .x
y 2 s yx 1 sin ax 3 . q x 2 cos ax 3 .
y3 s x 3 .

9.

A second example of electron density that belongs to the same equivalence class is that of
iodobromomethane that has similar qualitative appearance. This second example can also be derived
from the same DFG reference density by some
DFG transformation.
To address the question of uniqueness of the
representation 4. of r X one needs to find out
which pairs of diffeomorphisms y and h satisfy
the equation

INTERNATIONAL JOURNAL OF QUANTUM CHEMISTRY

r s h( r ( yy1 ,

10.

673

ZIMPEL AND MEZEY


symmorphy group of r and map every level set
G a. onto a DFG-equivalent level set G h a.. as
discussed in Ref. w 1x. . In a local sense, that is, in a
neighborhood of G a., for noncritical level sets y
is a contraction when hr a ) 1, and an expansion when hr a - 1.
The diffeomorphism h: w 0, A x w 0, AX x can be
thought of as a restriction of the diffeomorphism
h: w 0, `. w 0, `.. In view of this interpretation,
every pair of diffeomorphisms y and h defines a
4
4
4
transformation F: Rq
Rq
, where Rq
s r, a. a
G 04 , such that for every electron density r , the
image F y, h. M . s M X of the functional graph
M of r is a functional graph of another electron
density r X s h( r ( yy1 . These transformations F
form a group such that every class of DFG-equivalent electron densities functional graphs. corresponds to an orbit of this group. Using an operator
formalism, every pair of diffeomorphisms y and h
defines a linear operator F y, h. in the space of
electron density functions:
F y, h . r s h( r ( yy1 .

FIGURE 1. Level sets (MIDCOs, molecular isodensity


contours) of chlorobromomethane for three values of the
density threshold (in a.u.): 0.1 (top row), 0.01 (middle
row), and 0.001 (bottom row). Series A (left column)
corresponds to the standard Cartesian coordinates of
the Euclidean space; series B (right column) corresponds
to the nonlinear system of coordinates defined by Eq.
(9). A second example of electron density that belongs
to the same equivalence class is that of
iodobromomethane that has a very similar qualitative
appearance and also can be derived from the same DFG
reference density by some DFG transformation.
that is, which pairs of diffeomorphisms y and h
leave r unchanged. If h is the identity function,
the set of transformations y satisfying 10. forms a
group the symmorphy group w 69, 26x. , which will
be discussed in the next section.
In general, let h be a function such that h a c . s
a c for every critical level a c of r . Obviously, every
function satisfying 10. for some y must have this
property. On the other hand, if h does satisfy the
equation h a c . s a c for every critical level a c of r ,
then there exists a transformation y such that 10.
is valid. These transformations form a coset of the

674

11.

By analogy with the group of equivalence classes


described in the original symmorphy approach w 69,
26x , the set of these operators forms a group. In the
present work this group will be referred to as the
group T of operators of DFG equivalence. The
multiplication in this group is defined by
F yX , hX . ( F y, h . s F yX ( y, hX ( h . ;

12.

the identity transformation is


I s F id, id R . ;

13.

and the inverse transformations can be obtained as


F y, h .

y1

s F yy1 , hy1 . .

14.

The group T generates complete classes of DFGequivalent electron density functions. In other
words, if L denotes the space of all electron density functions, then each class of DFG-equivalent
electron density functions can be identified with
an orbit of the group T, that is, with an element of
the quotient space LrT.
The structure of the smooth manifold we have
explored so far required only that the projection
pr M be a global map on M. This condition makes
the differential structure of M isomorphic with the
standard differential structure of R 3, which is the
best known, the most widely used, and, apparently, the most natural one. However, one must be

VOL. 64, NO. 6

MOLECULAR GEOMETRY AND SYMMETRY


aware of the fact that other differential structures
can also be used, leading to alternative choices
representing the equivalence of electron densities.

Symmorphy and Symmetry


The diffeomorphisms g: R 3 R 3 satisfying a
direct or either of two equivalent indirect conditions

r ( gy1 s r or

F g , h . M . s M or

F g , id R . r s r

which for the standard global map R 3 , id. is given


by
2

16.

with id s x 1, x 2 , x 3 . being the Cartesian coordinates in R 3. In this case the group of isometries is
generated by rotations, reflections, and translations.
One may select a special subfamily SX of symmorphy transformations S, where the center of the
electron density, defined as the point r 0 satisfying
the equation

15.

form a group S, the symmorphy group of r w 69,


26x . Unlike the symmetry group, the symmorphy
group is not required to preserve metric properties
of space in which it operates, when such metric
properties are defined. The level sets of r are
invariant sets of a symmorphy transformation g
and nondegenerate critical points of r are fixed
points of g. The latter condition means that if r is a
nondegenerate critical point of r , then g r. s r.
This fact limits the size of the symmorphy groups
significantly, although a typical symmorphy group
is still very large. The discrete subgroup H of S,
isomorphic to the quotient group SrSid. with
Sid. denoting the largest connected subgroup of
the group S defines the combinatorial permutational. symmetry of the moleculethe symmetry
of the molecular topological tree introduced in Ref.
w 1x . This symmetry group allows one to define
equivalent atoms and atomic clusters within a
molecule. Since the elements of the symmorphy
group S generate smooth deformations of molecular electron densities, evidently, the same applies
to any of the conventional or fuzzy subsets of
these densities. For example, as conventional subsets, the formal atomic basins w 75x and their
boundaries undergo such smooth deformations,
and so do the additive fuzzy density fragments
obtained using the MEDLA w 76x and ADMA w 73x
methods. The effects of local transformations can
eventually be combined with transformations of
H.
One of the most interesting subgroups of S is
the intersection of S with the group of isometries
of R 3. However, the group of isometries depends
on the metric properties of the space R 3. The most
widely used metric is inherited from Euclidean
geometry w 71x and is defined by the metric tensor

ds 2 s dx 1 . q dx 2 . q dx 3 . ,

r 0 s r r r . dr

Hr r. dr,

17.

is a fixed point of every symmorphy transformation of subfamily SX of symmorphy transformations, with respect to electron density r . As shown
earlier, such families of symmorphies with a set of
specified fixed points also form a group w 26, 69x .
Therefore, one can replace the group of isometries
of R 3 by the isotropy group of R 3 at r 0 i.e., a
subgroup of the group of isometries containing all
transformations for which r 0 is a fixed point.. In
the Euclidean geometry w defined by the simple
Riemannian metric 16.x the isotropy group is the
orthogonal group O 3. w 71x . Thus, the symmetry
group G of the electron density function r is the
intersection of SX and O 3.. Obviously, only a few
molecules have a nontrivial symmetry in the Euclidean geometry, since almost all symmorphy
transformations are not isometries in the Euclidean
geometry. Within the framework of Euclidean geometry, a vast majority of molecules is characterized by the trivial symmetry group containing
only one transformationthe identity.
Natural generalizations of the Euclidean space
are the flat spaces w 71x , in which the Riemannian
metric tensor has the form
2

ds 2 s dy 1 . q dy 2 . q dy 3 . ,

18.

for a global map R 3, y .. In this geometry the


isotropy group of R 3 at r 0 is
Oy 3 . s  y ( g ( yy1 g g O 3 .4 .

19.

Consequently, the symmetry group G of r , the


intersection of group SX of symmorphy transformations, and Oy 3. is always isomorphic with a
subgroup of the orthogonal group O 3.. These
subgroups are: O 3. the full orthogonal group.,
SO 3. the group of all rotations., O 1. the group

INTERNATIONAL JOURNAL OF QUANTUM CHEMISTRY

675

ZIMPEL AND MEZEY


generated by rotations about a fixed axis and inversion about the isotropy center., SO 1. the group
generated by rotations about a fixed axis., and
finite subgroups.
The above relations can also be interpreted in
the following way: if y ( g ( yy1 is a symmetry
transformation of r X in the Euclidean geometry,
then g is a symmetry transformation of r in the
geometry defined by the Riemannian metric tensor
18.. This interpretation can be used to select a
standard representative for each equivalence class
of electron density functions: This representative is
any electron density of the class of equivalent
electron densities for which the symmetry group G
in the Euclidean geometry contains the largest
number of elements of the finite combinatorial
symmetry group H s SrSid..

Conclusions
In our previous study Ref. w 1x. we analyzed
some of the consequences of the fact that the space
of density functions can be decomposed into disjoint classes of DFG-equivalent densities. These
equivalent densities were identified with their
functional graphshypersurfaces having a natural
structure of differentiable manifolds, leading to a
similarity measure between electron densities. In
this study we further develop some of those ideas.
The equivalence relation that has been defined in
terms of electron density functional graphs can be
represented as an equivalence where each functional graph of an equivalence class corresponds to
a single generalized electron density function, that
can, however, be interpreted in terms of different
spatial coordinates global maps on the manifold.
and different scales of the electron density values.
Every DFG transformation can be decomposed into
two transformations: a diffeomorphism transforming the space and a diffeomorphism transforming
the scale of electron densities. The symmorphy
group of the density function or of its functional
graph. is a group of all those diffeomorphisms of
the space which leave the density function or its
functional graph. invariant. Defining a distance in
this space via a Riemannian metric tensor, a subgroup of the symmorphy group with a fixed point
at the center of charge, composed of all isometries
and, consequently, isotropies at the center of
charge. can be defined. This subgroup is the symmetry group. The finite symmetry groups of DFGequivalent electron densities are homomorphic to

676

SrSid. although they may represent different sets


of space transformations for different space geometries. The finite symmetry group is an invariant of smooth coordinate transformations, i.e., it is
independent of a particular system of coordinates.
Moreover, this finite symmetry group is always
isomorphic with a finite subgroup of the orthogonal group, and for every equivalence class there
always exists an electron density for which the
symmetry group is a finite subgroup of the orthogonal group in the Euclidean space. This density
can be taken as a representative for the whole
class.
By analogy with the earlier DER w 68x , WAT w 68x ,
and stepwise w 72x techniques, the equivalence
classes of symmorphy transformations described
here can be used for small electron density deformations, in order to generate new, approximate
electron densities from reference densities obtained for slightly different nuclear arrangements.
Aided by the DFG equivalence as a criterion for
identifying small deformations in a topological
sense, or in a metric sense with respect to various
choices of the Riemannian metric beyond the simple choice used in the symmetry analysis., the
symmorphy approach provides a general scheme
for small deformations of electron densities. In
combination with the DER w 68x , WAT w 68x , or stepwise w 72x techniques, the approach described in
this study can also be used to replace the transformations of electron density fragments used in the
MEDLA method of electron density calculation for
large molecules w 76]81x . This combined approach
appears advantageous if electron density fragments stored in the MEDLA database are to be
applied to build approximate electron densities for
target molecules where the nuclear arrangements
are slightly different from the actual local nuclear
arrangements corresponding to the density fragments in the database.
ACKNOWLEDGMENTS
The financial support of this work by the Natural Sciences and Engineering Research Council of
Canada is gratefully acknowledged.

References
1. Z. Zimpel and P. G. Mezey, Int. J. Quant. Chem. 59, 379
1996..
2. R. Carbo,
L. Leyda, and M. Arnau, Int. J. Quant. Chem. 17,
1185 1980..

VOL. 64, NO. 6

MOLECULAR GEOMETRY AND SYMMETRY


3. E. E. Hodgkin and W. G. Richards, J. Chem. Soc. Chem.
Commun. 1986, 1342 1986..

30. J. Cioslowski and S. T. Mixon, J. Am. Chem. Soc. 114, 4382


1992..

4. R. Carbo
and Ll. Domingo, Int. J. Quant. Chem. 32, 517
1987..

31. J. Cioslowski and S. T. Mixon, Canad. J. Chem. 70, 443


1992..
32. C. D. Zachmann, W. Heiden, M. Schlenkrich, and J. Brickmann, J. Comput. Chem. 13, 76 1992..
33. R. Ponec and M. Strnad, J. Chem. Inf. Comp. Sci. 32, 693
1992..
34. J. Gasteiger, W.-D. Ihlenfeldt, R. Fick, and J. R. Rose, J.
Chem. Inf. Comp. Sci. 32, 700 1992..
35. G. Sello, J. Chem. Inf. Comp. Sci. 32, 713 1992..
36. S. Hanessian, M. Botta, B. Larouche, and A. Boyaroglu, J.
Chem. Inf. Comp. Sci. 32, 718 1992..
37. U.-M. Weigel and R. Herges, J. Chem. Inf. Comp. Sci. 32,
723 1992..
38. G. M. Maggiora, D. W. Elrod, and R. G. Trenary, J. Chem.
Inf. Comp. Sci. 32, 732 1992..
39. V. Kvasnicka, S. Sklenak, and J. Pospichal, J. Chem. Inf.
Comp. Sci. 32, 742 1992..
40. P. G. Mezey, J. Math. Chem. 11, 27 1992..
41. P. G. Mezey, J. Chem. Inf. Comp. Sci. 32, 650 1992..
42. P. G. Mezey, In Theoretical and Computational Models for
Organic Chemistry, S. J. Formosinho, I. G. Csizmadia, and
L. G. Arnaut eds.. Kluwer Academic, Dordrecht, The
Netherlands, 1991..
43. J.-E. Dubois and P. G. Mezey, Int. J. Quant. Chem. 43, 647
1992..
44. X. Luo, G. A. Arteca, and P. G. Mezey, Int. J. Quant. Chem.
42, 459 1992..
45. P. G. Mezey, J. Math. Chem. 12, 365 1993..
46. P. G. Mezey, Shape in Chemistry: An Introduction to Molecular
Shape and Topology VCH Publishers, New York, 1993..
47. P. G. Mezey, J. Chem. Inf. Comp. Sci. 34, 244 1994..
48. P. G. Mezey, Int. J. Quant. Chem. 51, 255 1994..
49. P. G. Mezey, Canad. J. Chem. 72, 928 1994. special issue
dedicated to Prof. J. C. Polanyi..
50. O. Tapia, M. Paulino, and F. M. L .G. Stamato, Mol. Eng. 3,
377 1994..
51. B. B. Stefanov and J. Cioslowski, J. Comput. Chem. 16, 1394
1995..
52. J. A. Grant and B. T. Pickup, J. Phys. Chem. 99, 3503 1995..
53. G. Naray-Szabo

and G. G. Ferenczy, Chem. Rev. 95, 829


1995..
54. J. Sadowski, M. Wagener, and J. Gasteiger, Angew. Chem.
Int. Ed. Engl. 34, 2674 1995..
55. J. H. Schuur, P. Selzer, and J. Gasteiger, J. Chem. Inf. Comp.
Sci. 36, 334 1996..
56. D. J. Wild and P. Willett, J. Chem. Inf. Comp. Sci. 36, 159
1996..
57. D. J. Hankinson, J. Almlof,
and K. D. Leopold, J. Phys.
Chem. 100, 6904 1996..
58. P. G. Mezey, Structural Chem. 6, 261 1995..
59. P. G. Mezey, In Topics in Current Chemistry, Vol. 173, Molecular Similarity, K. Sen ed.. Springer-Verlag, Heidelberg,
1995..
60. P. G. Mezey, In Molecular Similarity and Reactivity: From
Quantum Chemical to Phenomenological Approaches, R. Carbo

ed.. Kluwer Academic, Dordrecht, The Netherlands, 1995..

5. E. E. Hodgkin and W. G. Richards, Int. J. Quant. Chem. 14,


105 1987..
6. R. Carbo
and B. Calabuig, Comput. Phys. Commun. 55, 117
1989..
7. R. Carbo
and B. Calabuig, Int. J. Quant. Chem. 42, 1681
1992..
8. R. Carbo
and B. Calabuig, Int. J. Quant. Chem. 42, 1695
1992..
9. R. Carbo,
B. Calabuig, L. Vera, and E. Besalu, In Advances
in Quantum Chemistry, P.-O. Lowdin,
J. R. Sabin, and M. C.

Zerner eds.., Vol. 25 Academic Press, New York, 1994..


10. P. G. Mezey, Int. J. Quant. Chem. Quant. Biol. Symp. 12, 113
1986..
11. P. G. Mezey, J. Comput. Chem. 8, 462 1987..
12. P. G. Mezey, Int. J. Quant. Chem. Quant. Biol. Symp. 14, 127
1987..
13. P. G. Mezey, J. Math. Chem. 2, 299 1988..
14. S. E. Leicester, J. L. Finney, and R. P. Bywater, J. Mol.
Graph. 6, 104 1988..
15. R. P. Bywater, In Molecular Similarity and Reactivity: From
Quantum Chemical to Phenomenological Approaches, R. Carbo

ed.. Kluwer Academic, Dordrecht, The Netherlands, 1995..


16. N. L. Allan and D. L. Cooper, J. Chem. Inf. Comp. Sci. 32,
587 1992..
17. D. L. Cooper and N. L. Allan, In Molecular Similarity and
Reactivity: From Quantum Chemical to Phenomenological Approaches, R. Carbo
ed.. Kluwer Academic, Dordrecht, The
Netherlands, 1995..
18. G. A. Arteca, V. B. Jammal, and P. G. Mezey, J. Comput.
Chem. 9, 608 1988..
19. G. A. Arteca, V. B. Jammal, P. G. Mezey, J. S. Yadav, M. A.
Hermsmeier, and T. M. Gund, J. Molec. Graph. 6, 45 1988..
20. M. A. Johnson, J. Math. Chem. 3, 117 1989..
21. G. A. Arteca and P. G. Mezey, J. Phys. Chem. 93, 4746
1989..
22. G. A. Arteca and P. G. Mezey, IEEE Eng. Med. & Bio. Soc.
11th Annual Int. Conf. 11, 1907 1989..
23. M. A. Johnson and G. M. Maggiora eds.., Concepts and
Applications of Molecular Similarity Wiley, New York, 1990..
24. C. Burt, W. G. Richards, and P. Huxley, J. Comput. Chem.
11, 1139 1990..
25. P. G. Mezey, In Reports in Molecular Theory, H. Weinstein
and G. Naray-Szabo

eds.. CRC Press, Boca Raton, 1990.,


Vol. 1, pp. 165]183.
26. P. G. Mezey, In Concepts and Applications of Molecular Similarity, M. A. Johnson and G. M. Maggiora eds.. Wiley,
New York, 1990..
27. P. G. Mezey, In Reviews in Computational Chemistry, K. B.
Lipkowitz and D. B. Boyd eds.. VCH Publishers, New
York, 1990..
28. P. G. Mezey, J. Math. Chem. 7, 39 1991..
29. A. Good and W. G. Richards, J. Chem. Inf. Sci. 33, 112
1992..

INTERNATIONAL JOURNAL OF QUANTUM CHEMISTRY

677

ZIMPEL AND MEZEY


61. P. G. Mezey, In Molecular Similarity in Drug Design, P. M.
Dean ed.. Chapman & Hall]Blackie Publishers, Glasgow,
U.K., 1995..
62. P. D. Walker and P. G. Mezey, J. Comput. Chem. 16, 1238
1995..
63. P. D. Walker, G. M. Maggiora, M. A. Johnson, J. D. Petke,
and P. G. Mezey, J. Chem. Inf. Comp. Sci. 35, 568 1995..
64. P. G. Mezey, Theor. Chim. Acta 92, 333 1995..
65. P. G. Walker, P. G. Mezey, G. M. Maggiora, M. A. Johnson,
and J. D. Petke, J. Comput. Chem. 16, 1474 1995..
66. P. G. Mezey, In Computational Chemistry: Reviews and Current Trends, J. Leszczynski ed.. World Scientific, Singapore, 1996..
67. P. G. Mezey, In Fuzzy Logic in Chemistry, D. H. Rouvray
ed.. Academic Press, San Diego, 1997..
68. P. G. Mezey, In Advances in Quantum Chemistry, P.-O.
Lowdin,
J. R. Sabin, and M. C. Zerner eds.. Academic

Press, New York, 1996..


69. P. G. Mezey, Topology of Molecular Shape and Chirality, New
Theoretical Concepts for Understanding Organic Reactions, J.
Bertran ed.., NATO ASI Series Plenum, New York, 1989..

678

70. S. Arimoto and P. G. Mezey, J. Math. Chem. 16, 93 1994..


71. Y. Choquet-Bruhat, Geometrie
Differentielle
et Systemes

`
Dunod, Paris, 1968., Chap. 3.
Exterieurs

72. P. G. Mezey, Int. J. Quant. Chem. 63, 39 1997..


73. P. G. Mezey, J. Math. Chem. 18, 141 1995..
74. P. Constans, R. Carbo-Dorca,
and P. G. Mezey, to appear.

75. R. F. W. Bader, T. T. Nguyen-Dang, and Y. Tal, Rep. Prog.


Phys. 44, 893 1981..
76. P. D. Walker and P. G. Mezey, J. Am. Chem. Soc. 115, 12423
1993..
77. P. D. Walker and P. G. Mezey, J. Am. Chem. Soc. 116, 12022
1994..
78. P. D. Walker and P. G. Mezey, Canad. J. Chem. 72, 2531
1994..
79. P. D. Walker and P. G. Mezey, J. Math. Chem. 17, 203
1995..
80. P. D. Walker and P. G. Mezey, J. Comput. Chem. 16, 1238
1995..
81. P. G. Mezey, Z. Zimpel, P. Warburton, P. D. Walker, D. G.
Irvine, D. G. Dixon, and B. Greenberg, J. Chem. Inf. Comp.
Sci. 36, 602 1996..

VOL. 64, NO. 6

Você também pode gostar