Some Problems Precautions of ab initio Quantum Chemistry

1
Some Problems (Precautions) of ab initio Quantum
Chemistry
Jim Jr-Min Lin
An actual situation
Most papers only show the good sides of the
calculation and hide the problematic sides. Such
that, experimentalists tend to think that theory
is nearly perfect.
2
ab initio calculations
e-
Orbitals 1s, 2s, 2p, ...
e-
Configurations (1s)2(2s)2(2p)1... (1s)2(2s)1(2p)2
Wave Function
Correlation Energy
3
Methods (Approximation, Algorithm) HF-SCF CAS-SCF
CAS-PT2 (Rayleigh-Schrodiger) MP2
(Moller-Plesset) CCSD(T)/QCISD(T) Density
Functional B3-LYP B3-Half and Half
4
Basis Set (BS) numerical functions to fit
orbitals Gaussian BS 6-31G(d,p)
6-311G(3df,3p) Dunnings correlation
consistent BS cc-pVTZ, cc-pVQZ, cc-pV5Z cc-pVNZ,
N3,4,5 aug-cc-pVNZ, N3,4,5
Number of Zeta, z
5
One electron Wave Function gt Approximate Orbital
more number of z, more diffuse
(s, p, d, f, ... ) l(1,2, 3, 4, ...)
Polarization Functions
6
JPC, A 101, 7923 (1997)
7
Even methods that are widely used and known of
high accuracy may have Significant Errors.
F2
Energy Hartree
R(F-F)
8
JCP 127, 164317 (2007)
9
Usually big basis set cant solve the Defect of a
method
F2
10
Basis set size has significant effect on accuracy
of relative energy and less effect on the
geometry.
F2
11
Another Example of Basis Set size
ClF
12
Another Example of Methods
ClF
13
The Shape of Potential Energy Curve depends on
the Methods
F2
14
Complete Basis Set Extrapolation
ClF
cc-pVNZ, N3,4,5 gt Ninfinity
De(Cl-F) Hartree
v3z
v8z
0.01H6.3 kcal/mol
v5z
v4z
(N0.5)-4
15
Take Home Messages

• Error of ab initio calculation depends on
  geometry Ri and bonding character. Err f
  (Ri)
• Form of f (Ri) very sensitive to Methods,
• not so sensitive to Basis Set
• Especially for Transition State, different
  methods often predict quite different geometry.
• Most methods behave well at equilibrium
  geometries (Here the bonding is well described by
  most MO methods)
• Testing and Calibration are necessary.
• Most papers only show the good sides.

16
Check your results with other methods. Check if a
Multi-Reference (MR) calculation is necessary.