Matta, Boyd. The quantum theory of atoms in molecules
.pdf222 8 Topology and Properties of the Electron Density in Solids
pressible which, together with the weaker electrostatics, results in less stress over the halide anions. Thus, partitioning of thermodynamic data into atomic components aids interpretation of their values and provides a powerful tool for qualitative predictions based on atomistic intuition.
8.7
Obtaining the Electron Density of Crystals
Most solid state electronic structure methods make use of pseudopotential techniques to avoid the explicit calculation of the atomic-like core regions. The valence electron density obtained from those calculations can, in principle, be completed by addition of the missing atomic core densities [54–56]. It must be noticed, however, that there are many di erent strategies for obtaining pseudopotentials, usually involving quite di erent partitions of the atomic density into core-like and pseudovalence contributions. An inappropriate match between the core and valence portions can result in spurious topological features. Whereas the field is very promising, some e ort is still needed, in our opinion, to compare the results from the main types of pseudopotential currently in use. There is, therefore, a limited number of methods that provide the all electron crystalline electron density in and analytical form suitable for high-quality QTAIM studies.
Under the promolecular approximation the crystalline electron density is obtained as a sum of atomic contributions:
rð~rÞ ¼ Xj |
rjð~r ~RjÞ |
ð6Þ |
|
~ |
~ |
where the sum runs over all atoms, and rjðr |
RjÞ is the electron density of the |
~
atom situated at Rj. This simple formula not only enables rapid approximate estimation but lies at the core of the experimental determination of the electron density from X-ray and neutron di raction experiments. As a further simplification, the atomic densities can be assumed to retain the spherical symmetry of the free atoms. Although the promolecular model lacks the internuclear electron density accumulation characteristic of true covalent bonding, it has been observed to frequently retain the topological features of the density in nonmolecular crystals. Even such fine details as the tiny nonnuclear maxima of some alkaline metals [18] can be retained. In contrast, the promolecular model has been shown to miss intermolecular bond CPs, i.e. give a di erent topology, in some molecular crystal studies [57].
The ab initio perturbed ion (aiPI) [58, 59] method has been used quite successfully for calculation of the electron density of many ionic crystals [33–35, 37, 44, 51, 60]. The method solves the Hartree–Fock equations of a solid in a localized Fock space by partitioning the crystal wavefunction into local, weakly overlapping, group functions, each containing a fixed number of electrons and a single nucleus. The aiPI crystal density can be described as a promolecular-like formula
8.7 Obtaining the Electron Density of Crystals 223
where the atomic contributions have been self consistently adapted to the crystal environment.
The crystal software of Dovesi et al. [61, 62] exploits the same MO-LCAO (crystal orbitals by linear combination of atomic Gaussian functions should be the motto here) approach that dominate quantum chemical studies in the molecular realm. Hartree–Fock (HF) and density functional theory (DFT) calculations have been used to examine the electron density topology of many ionic, covalent, molecular, and metallic crystals [26, 63–65]. It must be remarked, however, that use of the rich and flexible bases that are common within the molecular regime usually gives rise to linear dependencies and other technical problems that render the crystal calculation unsuitable. The molecular basis sets must then be truncated and Gaussian exponents less than 0.03–0.05 bohr 2, approximately, become forbidden. To compensate for this problem, the lowest retained Gaussian exponents are usually massaged or optimized within the crystal. When those di culties are taken into account, the method remains one of the two best techniques available to obtain the crystalline electron density. The next version of the software, release of which is planned for 2006, promises a definitive solution to the basis set problem and could produce a significant breakthrough in the field.
The other candidate for the best technique crown is the full potential linearized augmented plane wave (fpLAPW) formalism, implemented in the wien [30, 66] software by Blaha et al. The method uses di erent treatment for nonoverlapping spherical regions close to the nuclei, mu n tins, than for the internuclear regions. Atomic orbitals inside each mu n tin are described as the product of spherical harmonics and fully relativistic radial functions. Plane waves and a scalar relativistic treatment is applied to the internuclear region. The wave functions of both pieces are forced to match on value and slope at the mu n boundaries, but perfect agreement is not possible, because it would require inclusion of infinite spherical harmonics within each atomic sphere. Some care is then required to ensure that any remaining discontinuity in the density or its slope is not seen by the topological algorithms. Saturating the internuclear region with plane waves is, in contrast, both easy and inexpensive.
The fpLAPW, HF-LCAO, and aiPI topological properties of MgO and ZnO, two representative examples of the crystals that can be examined with the three techniques, are compared in Table 8.6. As usually occurs, the three techniques arrive at the same topology, which validates the very rapid and very stable aiPI procedure for analyzing trends and exploring the e ect of geometry changes. They also agree on the properties of the main bond critical points. The aiPI calculation, however, underestimates the electron density at the ring and cage CPs and, in general, in the low-density regions of the crystal. The e ect is more significant, because the average coordination index becomes smaller. The basin integrated charges provide clear evidence of this – the charges in the oxide basin agree for MgO (LCAO: 1.795, and aiPI: 1.852e) but di er substantially for ZnO (LCAO: 1.496, and aiPI: 1.800e). The di erence between the fpLAPW and HF-LCAO results is mainly because of the correlation e ects taken into account in the first calculation: DFT-LCAO results (not shown in the table) close the gap
2248 Topology and Properties of the Electron Density in Solids
Table 8.6 Comparison of the crystal topologies obtained by use of fpLAPW (PBE96 [31] GGA), HF-LCAO, and aiPI calculations on MgO (rock-salt structure, a ¼ 4:210 A˚ ) and ZnO (wurtzite structure,
a ¼ 3:250 and c ¼ 5:207 A˚ ). The (n,b,r,c) column provides the number of CPs of each type found in the unit cell. The rA/rB column gives the distance from the nuclei to a bond CP along the bond path. Atomic units are used throughout.
Crystal |
Method |
(n, b, r, c) |
CP |
r(~rc) |
‘2r |
rA/rB |
|
|
|
|
|
|
|
MgO |
fpLAPW |
ð8; 48; 48; 8Þ |
bðMg; OÞ |
0.03911 |
0.21775 |
1.716/2.262 |
|
|
|
bðO; OÞ |
0.01770 |
0.04553 |
2.813/2.813 |
|
|
|
r |
0.01713 |
0.05651 |
|
|
|
ð8; 48; 48; 8Þ |
c |
0.00998 |
0.02402 |
|
MgO |
HF-LCAO |
bðMg; OÞ |
0.03599 |
0.25421 |
1.694/2.284 |
|
|
|
|
bðO; OÞ |
0.01611 |
0.04900 |
2.813/2.813 |
|
|
|
r |
0.01507 |
0.06309 |
|
|
|
ð8; 48; 48; 8Þ |
c |
0.00723 |
0.03096 |
|
MgO |
aiPI |
bðMg; OÞ |
0.03266 |
0.28543 |
1.704/2.274 |
|
|
|
|
bðO; OÞ |
0.01183 |
0.05363 |
2.813/2.813 |
|
|
|
r |
0.01075 |
0.06737 |
|
|
|
ð4; 10; 8; 2Þ |
c |
0.00444 |
0.02460 |
|
ZnO |
HF-LCAO |
bðZn; OÞ |
0.07870 |
0.47364 |
1.789/1.941 |
|
|
|
|
bðZn; OÞ |
0.07554 |
0.44617 |
1.802/1.962 |
|
|
|
bðZn; OÞ |
0.00650 |
0.02995 |
2.903/3.173 |
|
|
|
r |
0.00649 |
0.02990 |
|
|
|
|
r |
0.00437 |
0.01926 |
|
|
|
ð4; 10; 8; 2Þ |
c |
0.00177 |
0.00850 |
|
ZnO |
aiPI |
bðZn; OÞ |
0.07961 |
0.51117 |
1.797/1.932 |
|
|
|
|
bðZn; OÞ |
0.07568 |
0.48603 |
1.811/1.953 |
|
|
|
bðZn; OÞ |
0.00363 |
0.02172 |
2.809/3.268 |
|
|
|
r |
0.00363 |
0.02171 |
|
|
|
|
r |
0.00261 |
0.01354 |
|
|
|
|
c |
0.00082 |
0.00466 |
|
|
|
|
|
|
|
|
between the LCAO and LAPW techniques. The correlation tends to increase the electron density population of the bond CPs at the expense of the population of the low-density regions.
The theoretical electron densities can be corrected by using the crystal form
ð~Þ
factors, Fobs h , derived from experimental di raction measurements [67, 68]. In principle, the electron density of the crystal can be obtained directly from the form factors:
|
|
Xh |
|
~ |
|
|
r |
~r |
F |
~h e i2ph~r |
7 |
Þ |
|
|
obsð Þ ¼ |
|
obsð Þ |
|
ð |
~
226 8 Topology and Properties of the Electron Density in Solids
Fig. 8.5 Electron density of the Si crystal along the xxx crystallographic direction. This is the line connecting two nearest neighbors situated on x ¼ 1/8 and x ¼ 1/8, respectively. Lines (a) through (c) are the raw promolecular, HF-LCAO, and DFT-LCAO electron densities, respectively. Lines (d) through (f ) correspond to the same model densities, now corrected with the experimental form factors [69, 70].
is an interesting target. The use of a small or unbalanced basis set, di erences in the approximated treatment of the correlation, and other small computational details can modify the theoretical equilibrium distance enough to make the NNM appear or disappear. The experimental form factors are corrected, on their own, to avoid some well known e ects, for example the directionally dependent X-ray absorption or the di erent time expended on measurement of the di erent
ð~Þ
Fobs h . The experimental form factors also unavoidably contain some thermal motion noise, even if they have been deconvoluted from thermal agitation. Neither quantum mechanical calculations nor experimental di raction data are thus free from trouble when examining problematic systems lying at the edge of topological regions. The interesting thing is that both approaches can be combined in several ways, giving the researcher new tools to examine the stability of di cult topological features.
Acknowledgements
We thank the Spanish Ministerio de Ciencia y Tecnologı´a for financial support under project BQU2003-06553.
|
|
|
References |
227 |
|
References |
|
|
|
|
|
|
|
|
1 |
R. F. W. Bader, Atoms in Molecules. |
21 |
T. S. Koritsanszky, P. Coppens, |
|
|
A Quantum Theory, Oxford |
|
Chem. Rev. 101 (2001) 1583–1627. |
|
|
University Press, Oxford, 1990. |
22 |
R. Destro, R. Bianchi, C. Gatti, F. |
|
2 |
R. F. W. Bader, Int. J. Quantum |
|
Merati, Chem. Phys. Lett. 186 (1991) |
|
|
Chem. 56 (1995) 409–419. |
|
47–52. |
|
3 |
R. F. W. Bader, Phys. Rev. B 49 (1994) |
23 |
C. Gatti, R. Bianchi, R. Destro, F. |
|
|
13348–13356. |
|
Merati, J. Mol. Struct. (Theochem) 87 |
|
4 |
R. F. W. Bader, Int. J. Quantum |
|
(1992) 409–433. |
|
|
Chem. 49 (1994) 299–308. |
24 |
P. F. Zou, R. F. W. Bader, Acta Cryst. |
|
5 |
R. F. W. Bader, P. L. A. Popelier, Int. |
|
A 50 (1994) 714–725. |
|
|
J. Quantum Chem. 45 (1993) |
25 |
V. G. Tsirelson, P. F. Zou, T. H. Tang, |
|
|
189–207. |
|
R. F. W. Bader, Acta Cryst. A 51 |
|
6 |
P. F. Zou, R. F. W. Bader, Int. J. |
|
(1995) 143–153. |
|
|
Quantum Chem. 43 (1992) 677– |
26 |
C. Gatti, Z. Kristall, 220 (2005) |
|
|
699. |
|
399–457. |
|
7 |
R. F. W. Bader, Chem. Rev. 91 (1991) |
27 |
M. Morse, S. S. Cairns, Critical point |
|
|
893–928. |
|
Theory in Global Analysis and |
|
8 |
R. F. W. Bader, J. Chem. Phys. 91 |
|
Di erential Geometry, Academic |
|
|
(1989) 6989–7001. |
|
Press, New York, 1969. |
|
9 |
R. F. W. Bader, Prog. Appl. Chem. 60 |
28 |
T. Hahn (Ed.), International Tables |
|
|
(1988) 145–155. |
|
for X-Ray Crystallography. A. Space- |
|
10 |
R. F. W. Bader, P. J. Macdougall, |
|
group symmetry, Vol. A of |
|
|
J. Am. Chem. Soc. 107 (1985) 6788– |
|
International Tables for X-Ray |
|
|
6795. |
|
Crystallography, D. Reidel, Berlin, |
|
11 |
R. F. W. Bader, Acc. Chem. Res. 18 |
|
1983. |
|
|
(1985) 9–15. |
29 |
M. I. Aroyo, J. M. Pe´rez-Mato, C. |
|
12 |
R. F. W. Bader, H. Essen, J. Chem. |
|
Capillas, E. Kroumova, S. Ivantchev, |
|
|
Phys. 80 (1984) 1943–1960. |
|
G. Madariaga, A. Kirov, H. |
|
13 |
R. F. W. Bader, T. T. Nguyen-Dang, Y. |
|
Wondratschek. Zeitschrift fu¨r |
|
|
Tal, Rep. Prog. Phys. 44 (1981) 893– |
|
Kristallographie 221 (2006) 15–27, |
|
|
948. |
|
http://www.cryst.ehu.es/. |
|
14 |
R. F. W. Bader, T. T. Nguyen-Dang, |
30 |
P. Blaha, K. Schwarz, J. Luitz, |
|
|
Adv. Quantum Chem. 14 (1981) |
|
WIEN97, a Full Potential Linearized |
|
|
63–124. |
|
Augmented Plane Wave package for |
|
15 |
R. F. W. Bader, S. Srebrenik, T. T. |
|
calculating crystal properties, (K. |
|
|
Nguyen-Dang, J. Chem. Phys. 68 |
|
Schwartz, Techn. Univ. Wien, |
|
|
(1978) 3680–3691. |
|
Vienna). Updated version of P. Blaha, |
|
16 |
S. Srebrenik, R. F. W. Bader, T. T. |
|
K. Schwarz, P. Sorantin, and S. B. |
|
|
Nguyen-Dang, J. Chem. Phys. 68 |
|
Trickey, Comput. Phys. Commun. 59, |
|
|
(1978) 3667–3679. |
|
399, 1990 (1999). |
|
17 |
R. F. W. Bader, Acc. Chem. Res. 8 |
31 |
J. P. Perdew, S. Burke, M. Ernzerhof, |
|
|
(1975) 34–40. |
|
Phys. Rev. Lett. 77 (1996) 3865. |
|
18 |
V. Luan˜a, P. Mori-Sa´nchez, A. |
32 |
A. Martı´n Penda´s, M. A. Blanco, E. |
|
|
Costales, M. A. Blanco, A. Martı´n |
|
Francisco, Chem. Phys. Letters 417 |
|
|
Penda´s, J. Chem. Phys. 119 (2003) |
|
(2005) 16–21. |
|
|
6341–6350. |
33 |
A. M. Penda´s, A. Costales, V. |
|
19 |
A. M. Penda´s, A. Costales, M. A. |
|
Luan˜a, Phys. Rev. B 55 (1997) 4275– |
|
|
Blanco, J. M. Recio, V. Luan˜a, Phys. |
|
4284. |
|
|
Rev. B. 62 (21) (2000) 13970–13978. |
34 |
M. A. Blanco, A. Costales, A. M. |
|
20 |
A. M. Penda´s, J. Chem. Phys. 117 |
|
Penda´s, V. Luan˜a, Phys. Rev. B. 62 |
|
|
(2002) 965–979. |
|
(2000) 12028–12039. |
|
|
|
References |
229 |
|
|
|
|
|
|
and Theoretical Chemistry (2001). |
69 |
S. Cumming, M. Hart, Aust. J. Phys. |
|
|
URL http://www.wien2k.at/ |
|
41 (1988) 423. |
|
67 |
G. H. Stout, L. H. Jensen, X-ray |
70 |
T. Saka, N. Kato, Acta Crystallogr. A |
|
|
structure determination: A Practical |
|
42 (1986) 469. |
|
|
Guide, 2nd Edition, John Wiley & |
71 |
R. Hosemann, S. N. Bagchi, Nature |
|
|
Sons, New York, 1989. |
|
171 (1953) 785–787. |
|
68 |
C. Giacovazzo, H. L. Monaco, D. |
72 |
N. K. Hansen, P. Coppens, Acta |
|
|
Viterbo, F. Scordari, G. Gilli, G. |
|
Cryst. A 34 (1978) 909. |
|
|
Zanotti, M. Catti, Fundamentals of |
73 |
J. M. Zuo, P. Blaha, K. Schwarz, |
|
|
Crystallography, Oxford UP, Oxford, |
|
J. Phys. Condens. Matter. 9 (1997) |
|
|
UK, 1992. |
|
7541–7561. |
231
9
Atoms in Molecules Theory for Exploring the Nature of the Active Sites on Surfaces
Yosslen Aray, Jesus Rodrı´guez, and David Vega
9.1 Introduction
The atoms in molecules theory provides a rigorous definition of chemical bonds for all types of molecules and solids [1–10] and has proven useful in analysis of the physical properties of insulators, pure metals, and alloys [4–8]. In particular, it has been observed that the strength of the bonding between a given pair of atoms in a molecule correlates with the values of the electron density at the bond critical point, rb [1]. A simple relationship between the binding energy and rb in periodic solids has also been reported [4, 6, 8]. The AIM theory also leads to unique partition of three-dimensional space into a collection of chemically identifiable regions called atomic basins (the atoms in a molecule or in a crystal). These are the most transferable pieces one can define in exhaustive partitioning of the real space [1]. In this chapter we describe the implementation of an algorithm that uses the r(r) topological information to determine the main elements of the AIM theory for periodic systems, and discuss an application to nanocatalysis.
9.2
Implementing the Determination of the Topological Properties of r(r) from a Three-dimensional Grid
The CPs are usually calculated using the Newton–Raphson (NR) technique [11]. The NR algorithm starts from a truncated Taylor expansion at a point r ¼ r0 þ h, about r0 of a multidimensional scalar function ð‘rðrÞÞ:
‘rðrÞ ¼ ‘rðr0Þ þ H0h þ higher order terms |
ð1Þ |
where H is the Hessian (the Jacobian of ‘rðrÞÞ at point r0. The best step, h, to move from the initial r0 to the CP is h ¼ H 1‘rðr0Þ. This correction is then used to obtain a vector rnew ¼ rold þ th (t is a small value) and the process is iter-