07.04.2013 Views

Essentials of Computational Chemistry

Essentials of Computational Chemistry

Essentials of Computational Chemistry

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

9.1 PROPERTIES RELATED TO CHARGE DISTRIBUTION 311<br />

(cf. Eq. (9.7)), and f is a damping factor raised to the nth power. New electronic populations<br />

are then computed according to<br />

Q (n)<br />

k<br />

= Q(n−1)<br />

k − <br />

Q (n)<br />

<br />

k→k ′ + Q (n)<br />

k ′ →k<br />

k ′ bonded<br />

to k<br />

χ k ′ >χk<br />

k ′ bonded<br />

to k<br />

χk>χ k ′<br />

(9.9)<br />

In the original PEOE method, the damping function f was simply taken to be the constant 0.5,<br />

which led to practical convergence in the atomic partial charges within about five iterations.<br />

No et al. (1990a, 1990b) subsequently proposed a modification in which different damping<br />

factors were used for different bonds (MPEOE) and observed that this, together with some<br />

other minor changes, gave improved charge distributions when compared to known multipole<br />

moments. More recently, Cho et al. (2001) proposed computing the damping factor as<br />

<br />

<br />

fkk ′ = min<br />

0,<br />

1 −<br />

r vdw<br />

k<br />

rkk ′<br />

+ rvdw<br />

k ′<br />

(9.10)<br />

where rkk ′ is the distance between the two atoms and rvdw is a parametric van der Waals<br />

radius. The use <strong>of</strong> Eq. (9.10) delivers geometry-dependent atomic charge (GDAC) values,<br />

which were found to improve additionally on computed electrical moments.<br />

Another Class I charge model that is also sensitive to geometry is the QEq charge equilibration<br />

model <strong>of</strong> Rappé and Goddard (1991). From representing the energy u <strong>of</strong> an isolated<br />

atom k as a Taylor expansion in its charge truncated at second order, one can derive<br />

uk =ũk + χkqk + 1<br />

2 Jkkq 2 k<br />

(9.11)<br />

where ũ is the energy <strong>of</strong> the neutral isolated atom, χ is the electronegativity (experimentally<br />

the average <strong>of</strong> the atomic IP and EA), and J is the idempotential, which is formally equal<br />

to IP − EA. With this formula in hand, we may write the electrostatic energy <strong>of</strong> a collection<br />

<strong>of</strong> N atoms as<br />

N<br />

U = (ũk + χkqk) + 1<br />

N N<br />

2<br />

k=1<br />

k=1 k ′ Jkk ′qkqk ′ (9.12)<br />

=1<br />

where J is a matrix <strong>of</strong> Coulomb integrals for which we have already defined the diagonal<br />

elements as the idempotentials. The <strong>of</strong>f-diagonal elements are computed as (aa|bb)<br />

where a and b are STOs on the centers k and k ′ , respectively (thereby introducing geometry<br />

dependence). QEq charges q are then determined from minimization <strong>of</strong> U subject to<br />

the constraint that the total molecular charge remain constant. Note the close conceptual<br />

similarities between QEq and SCC-DFTB described in Section 8.4.4.<br />

Eq. (9.12) does not require any specification <strong>of</strong> bonding – all atoms electrically interact<br />

with all other atoms. Sefcik et al. (2002) have combined QEq electrostatics with Morse<br />

potentials for non-electrostatic non-bonded interactions between all atom pairs to create a

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!