8.

Basis sets
As we already know, each MO is expressed as a linear combination of basis functions (eqn. 40),
the coefficients for which are determined from the iterative solution of the HF SCF equations (as
flow-charted above). The full HF wave function is expressed as a Slater determinant formed from
the individual occupied MOs.
It is important to remember that the basis set expansion is another approximation: we are building
the molecular orbitals, that in general we have no idea how they should look like, from some
collection of convenient functions, such as atomic orbitals (AOs). It is not very difficult to see that
the arrangement of electrons along bonds in molecules may not be very well approximated by their
arrangement in individual atoms. But, for simplicity and convenience, the atomic orbitals are
almost exclusively used as MO basis functions.
Any choice of basis set necessarily restricts the space that the electrons are allowed to occupy. An
optimal description of the electron probability density would be achieved by an infinitely large
basis set. That would impose no restriction on the electrons and would give us the best possible
solution to the Schrodinger equation, within the limit of the theoretical method. The HF calculation
with the infinite basis set would give us the Hartree-Fock limit. Remember that the HF method is
flawed because it omits the electron correlation. But with infinitely large basis set, it would be as
good as it gets with the HF.
Obviously, in practice, one cannot make use of an infinite basis set. Thus, much work has gone
into identifying mathematical functions that allow wave functions to approach the HF limit
arbitrarily closely in as efficient a manner as possible.
Efficiency in this case involves three considerations. As noted above, the number of two-electron
integrals increases as M 4 where M is the number of basis functions. It makes a lot of sense to keep
the number of the basis functions to a minimum. In addition, however, it can be useful to choose
basis set functional forms that permit the various integrals appearing in the HF equations to be
evaluated in a computationally efficient fashion. Thus, a larger basis set can still represent a
computational improvement over a smaller basis set if evaluation of the greater number of integrals
for the former can be carried out faster than for the latter.
Finally, the basis functions must be chosen to have a form that is useful in a chemical sense. That
is, the functions should have large amplitude in regions of space where the electron probability
density (the wave function) is also large, and small amplitudes where the probability density is
small. The simultaneous optimization of these three considerations is at the heart of basis set
development.

8.1. Slater type orbitals (STO) and Gaussian type orbitals (GTO)
Slater type orbitals (STO) are essentially Hydrogen-like atomic orbitals that you know from the
general chemistry. The radial part consists of an exponential and a polynomial, while the angular
part is a spherical harmonic:
 ,nlm r , ,   r n1 exp( r )Yl m  ,  (56)
where  ( Greek zeta) is an exponent.

To use STO’s as the AO basis would therefore make a lot of sense. However, it is not practical
except for atoms and diatomic molecule calculations. The reason is that There is no analytical

1
solution available for the general four-index integral (eqn. 52) when the basis functions are STOs.
The requirement that such integrals be solved by numerical methods, which takes a long time,
severely limits their utility in molecular systems of any significant size.

It turns out that analytical solutions can be derived for any four index integral if chosen to have
the form of a Gaussian function, i.e. instead of an exponential exp( r) it needs exponential in r2:
exp( r2).

The general functional form of a normalized Gaussian-type orbital (GTO) in atom-centered

Cartesian coordinates is:
 
g ijk  ; x, y, z   x i y j z k exp   x 2  y 2  z 2  (57)

where  is again an exponent controlling the width of the GTO, and i, j, and k are non-negative
integers that dictate the nature of the orbital in a Cartesian sense:

 when all three of these indices are zero, the GTO has spherical symmetry, and is called an
s-type GTO.
 when exactly one of the indices is one, the function has axial symmetry about a single
Cartesian axis and is called a p-type GTO. There are three possible choices for which
index is one, corresponding to the px, py, and pz orbitals.
 when the sum of the indices is equal to two, the orbital is called a d-type GTO.

Note that there are six possible combinations of index values (i, j, k) that can sum to two. In eqn.
(56), this leads to possible Cartesian prefactors of x2, y2, z2, xy, xz, and yz. These six functions are
called the Cartesian d functions. However, we know that there are only five d functions in the H
atom usually referred to as xy, xz, yz, x2 − y2, and 3z2 − r2. The first three of these canonical d
functions are common with the Cartesian d functions, while the latter two can be derived as linear
combinations of the Cartesian d functions. The last remaining linear combination that can be
formed from the Cartesian d functions is x2 + y2 + z2, which is actually an s-type GTO (spherically
symmetric).
Different Gaussian basis sets adopt different conventions with respect to their d functions: some
use all six Cartesian d functions, others prefer to reduce the total basis set size and use the five
linear combinations. Note that if the extra function is kept, the linear combination having s-like
symmetry still has the same exponent  governing its decay as the rest of the d set. As d orbitals
are more diffuse than s orbitals having the same principal quantum number (which is to say the
magnitude of α for the nd GTOs will be smaller than that for the α of the ns GTOs), the extra s
orbital does not really contribute at the same principal quantum level.
As one increases the indexing, the disparity between the number of Cartesian functions and the
number of canonical functions increases. Thus, with f-type GTOs (indices summing to 3) there
are 10 Cartesian functions and 7 canonical functions, with g-type 15 and 10, etc. GTOs can be
taken arbitrarily high in angular momentum.

2
8.2. Contracted Gaussian functions
Despite the great computational savings they offer, there are several problems with GTOs:
 incorrect shape of the radial portion of the orbital both for very short and long distances:
o For s type functions, GTOs are smooth and differentiable at the nucleus (r = 0), but
real hydrogenic AOs have a cusp.
o hydrogenic AOs have a radial decay that is exponential in r while the decay of
GTOs is exponential in r2this results in too rapid a reduction in amplitude with
distance for the GTOs.

STO - solid
GTO - dashed

 incorrect radial nodal behavior: no choice of variables lets the GTO (eqn. 56) to mimic a
2s orbital, which is negative near the origin and positive beyond a certain radial distance.

To get around it, instead a single GTO a linear combination of GTOs is used to reproduce as
accurately as possible a STO, i.e.

K
ijk  ; x, y, z    d a g ijk  a ; x, y, z  (58)
a 1

where K is the number of Gaussians used in the linear combination, and the coefficients d are
chosen to optimize the shape of the basis function sum and ensure normalization.

When a basis function is defined as a linear combination of Gaussians, it is referred to as a

‘contracted’ basis function, and the individual Gaussians from which it is formed are called
‘primitive’ Gaussians.
The more primitives that are employed, the more accurately a contracted function can be made to
match a given STO. However, note that a four-index two-electron integral becomes increasingly
complicated to evaluate as each individual basis function is made up of increasingly many
primitive functions, according to:

3
 1
(  |  )    * (1) (1) * (2) (2)dr1dr2 
r12
K K K K
1
   d a g a (1)  d a g a (1)  d  g  (2)  d  g  (2)dr dr
a a a a 1 2 (59)
a 1 a 1 r12 a 1 a 1

K K K K
1
     d  d  d  d   g  (1) g  (1) r
a 1a 1 a 1a 1
a a a a a a g a (2) g a (2)dr1dr2
12

Use of a contraction scheme also fixes the problem with the nodes. Since the contraction
coefficients d in (58) can be chosen to have either negative or positive sign, by making some of
the Gaussians negative, they can create the proper number and positions of the nodes.
It was discovered that the optimum combination of speed and accuracy (when comparing to
calculations using STOs) was achieved for K = 3. This defines the classic STO-3G basis set, which
is defined for most of the atoms in the periodic table. The STO-3G basically means an STO
approximated by 3 contracted Gaussians.

The radial behavior of various basis functions in atom-centered coordinates. The bold solid
line at top is the STO (ζ = 1) for the hydrogen 1s function; for the one-electron H system,
it is also the exact solution of the Schrodinger equation. Nearest it is the contracted STO-
3G 1s function (dots) optimized to match the STO. It is the sum of a set of one each tight
(dash-dots ), medium (dashes), and loose (solid ) Gaussian functions shown below. From
0.5 to 4.0 a.u., the STO-3G orbital matches the correct orbital closely. However, near the
origin there is a notable difference and, were the plot to extend to very large r, it would be
apparent that the decay of the STO-3G orbital is more rapid than the correct orbital.

Substituting equation (58) to the LCAO expansion, eqn. (40) gives:

M M K
 i
MO
  Ci     Ci  d a g ijk  a ; x, y, z 
AO
(60)
 1  1 a 1

4
This is the written-out expansion of the MO wavefunction in terms of primitive Gaussians.
Remember that:

 The coefficients C - the LCAO coefficients Ci - are optimized by the SCF procedure.

 The contraction coefficients, da are fixed during the SCF calculation: they were determined
beforehand by approximating an STO.

8.3. Minimal basis sets

The STO-3G is an example of a minimal basis set. Minimal basis sets contain the minimum
number of basis functions needed for each atom, as in these examples:

H: 1s
C: 1s, 2s, 2px, 2py, 2pz

In the special case of STO-3G as we have seen before, the basis functions are contracted from
three Gaussians. Other minimal basis sets include the MINI sets, which are named MINI-1, MINI-
2, etc., and vary in the number of primitives used for different kinds of functions.

8.4. Double-, Triple- etc. basis sets

One way that a basis set can be made larger is to decontract it. This will effectively increase the
number of basis functions per atom. For example, instead of contracting all three Gaussians in
STO-3G, one could construct two basis functions for each AO:
 the first being a contraction of the first two primitive Gaussians
 the second would simply be the uncontracted, normalized third primitive.

This prescription would not double the size of our basis set, since we would have all the same
individual integrals to evaluate as previously, but the size of our secular equation would be
increased: we would independently optimize the LCAO coefficients for both basis functions.

Modern examples of such basis sets are the cc-pCVDZ, cc-pCVTZ where the acronym stands for
correlation-consistent polarized Core and Valence (Double/Triple/etc.) Zeta. The correlation
consistent we will discuss below, the important part is that it has both Core and Valence.

8.4. Split Valence or Valence Double- basis sets

Unlike the full double zeta basis sets, which include two, three etc. basis functions for all atoms,
the Valence double- (triple etc.) zeta basis sets keep a single basis function for the core electrons,
but have two (or more) sizes of basis function for each valence orbital. This is because the core
electrons are not very much affected by bonding.

5
For example, hydrogen and carbon are represented as:
H: 1s, 1s’
C: 1s, 2s, 2s’,2px, 2py, 2pz, 2px’, 2py’, 2pz’

where the primed and unprimed orbitals differ in size. Note that for C the 1s is there only once,
without a primed one, because it is a core, not valence orbital.

Amongst the most widely used split-valence basis sets are those developed by John Pople and
coworkers (also referred to as Pople basis sets). These basis sets include 3-21G, 6-21G, 4-31G,
6-31G, and 6-311G.
The nomenclature is a guide to the contraction scheme. The first number indicates the number of
primitives used in the contracted core functions. The numbers after the hyphen indicate the
numbers of primitives used in the valence functions – if there are two such numbers, it is a valence-
double- basis, if there are three, valence-triple-.
For example, the 6-31G is a valence double-with one basis function contracted from 6 Gaussians
for all core electrons, and two basis functions for valence electrons: the first contracted from 3
Gaussian, the second an uncontracted primitive.
The Pople basis sets have seen sufficient use in the literature that certain trends have clearly
emerged. While a more complete discussion of the utility of HF theory and its basis set dependence
appears at the end of this chapter, we note here that, in general, the 4-31G basis set is inferior to
the less expensive 3-21G, so there is little point in ever using it. The 6-21G basis set is obsolete.

Other widely used valence double- basis sets are:

 MIDI and MAXI named MIDI-1, MIDI-2, etc., MAXI-1, MAXI-2, etc. and vary in the
number of primitives used for different kinds of functions.
 correlation consistent basis sets cc-pVDZ, cc-pVTZ, cc-pVQZ, cc-pV5Z and cc-pV6Z.
 SV, and TZV basis sets, which stand for, respectively “split valence” and “triple zeta
valence”, with Def2 prefix for their new formulations (Def2SV, Def2TZP).

8.5. Polarized basis sets

As we already mentioned, the molecular orbitals, which are eigenfunctions of a Schrodinger
equation involving multiple nuclei at various positions in space, require more mathematical
flexibility than do the atoms. Split valence basis sets allow orbitals to change size, but not to change
shape. Polarized basis sets remove this limitation by adding orbitals with angular momentum
beyond what is required for the ground state to the description of each atom.

For example, polarized basis sets add d functions to carbon atoms and f functions to transition
metals, and some of them add p functions to hydrogen atoms.

6
An example of the polarized basis set is 6-31G(d). Its name indicates that it is the 6-31G basis set
with d functions added to heavy atoms in the first and second rows. It also adds an f function to
the transition metals, since they already have d functions. This basis set is also known as 6-31G*.

Another popular polarized basis set is 6-31G(d,p), also known as 6-31G**, which adds p functions
to hydrogen atoms in addition to the d functions on heavy atom (or f functions to transition metals).

8.6. Balanced basis sets

There is a rough correspondence between the value of adding polarization functions and the value
of decontracting the valence basis function(s). In particular, there is a rough equality between each
decontraction step and adding one new set of polarization functions, including a new set of higher
angular momentum. That means, for example:

 Balanced double-ζ basis sets should include d functions on heavy atoms and p functions
on H, such as 6-31G(d,p).
 triple-ζ basis sets should include 1 set of f and 2 sets of d functions on heavy atoms, and
1 set of d and 2 sets of p functions on H, e.g. 6-311G(2df,2pd).

This is the polarization prescription adopted by the cc-pVnZ basis sets of Dunning and co-workers
already mentioned above, where n ranges over D (double), T (triple), Q (quadruple), five, and six.
Thus, for cc-pV6Z, for example, each heavy atom has one i function, two h functions, three g
functions, four f functions, five d functions, and six valence s and p functions, in addition to core
functions (using the canonical numbers of these functions, we have 140 basis functions for a single
second-row atom, so this basis set presently finds use only for the smallest of systems). Note that
while it would be an unpleasant exercise to try to draw an i function, it is straightforwardly defined
by taking the sum of i, j, and k equal to 6 in eqn. (57).

The SV and TZV basis sets also have variants with polarization functions, namely SVP, TZVP,
where the “P” stands for polarized. The corresponding newer definitions are Def2SVP, Def2TZVP
etc. The polarization functions are included in definition again using the balanced basis set
concept.

A partially polarized basis set, MIDI! (where the ‘!’ is pronounced “bang”; in some electronic
structure programs, the abbreviation MIDIX is employed to avoid complications associated with
interpretation of the exclamation point) although much smaller than the 6-31G(d) in direct
comparisons it yields more accurate geometries and charges as judged by comparison to much
higher level calculations. For cases where p polarization functions on H might be expected to be
important, the MIDIY basis set includes these as an extension to MIDI!

8.7 Diffuse basis sets

Diffuse functions are large-size versions of s- and p-type functions (as opposed to the standard
valence-size functions). They allow orbitals to occupy a larger region of space. Basis sets with
diffuse functions are important for systems where electrons are relatively far from the nucleus:
molecules with lone pairs, anions and other systems with significant negative charge, systems in
their excited states, systems with low ionization potentials, descriptions of absolute acidities, and
so on.

7
In the Pople family of basis sets, the presence of diffuse functions is indicated by a ‘+’ in the basis
set name. For example, the 6-31+G(d) basis set is the 6-31G(d) basis set with diffuse functions
added to heavy atoms. The double plus version, 6-3l++G(d), adds diffuse functions to the hydrogen
atoms as well. Diffuse functions on hydrogen atoms seldom make a significant difference in
accuracy.

Diffuse sp sets have also been defined for use in conjunction with the MIDI! and MIDIY basis
sets, generating MIDIX+ and MIDIY+.

For the correlation consistent basis sets cc-pVnZ, diffuse functions on all atoms are indicated by
prefixing with ‘aug’, which stand for ‘augmented’ (= augmented by diffuse functions). Here
though a set of diffuse functions is added for each angular momentum already present. Thus, aug-
cc-pVTZ has diffuse f, d, p, and s functions on heavy atoms and diffuse d, p, and s functions on H
and He. This generates quite enormous basis sets really fast.

8.8 Effective core potentials

Basis sets for atoms beyond the third row of the periodic table are handled somewhat differently.
For these very large nuclei, electrons near the nucleus are treated in an approximate way, via
effective core potentials (ECPs).

In other words, in the HF Roothan-Hall equations for N electrons, the ones in the “core” instead
of being counted explicitly, are replace by an effective potential function.

ECPs must properly represent not only Coulomb repulsion effects, but also adherence to the Pauli
principle. Furthermore, the core electrons in very heavy elements reach velocities sufficiently near
the speed of light that they manifest relativistic effects. A non-relativistic Hamiltonian operator is
incapable of accounting for such effects, which can be significant for many chemical properties.
The ECP on the other hand can include the relativistic effects

A key issue in the construction of ECPs is just how many electrons to include in the core. So-
called ‘large-core’ ECPs include everything but the outermost (valence) shell, while ‘small-core’
ECPs scale back to the next lower shell. Because polarization of the sub-valence shell can be
chemically important in heavier metals, it is usually worth the extra cost to explicitly include that
shell in the calculations. Thus, the most robust ECPs for the elements Sc–Zn, Y–Cd, and La–Hg,
employ [Ne], [Ar], and [Kr] cores, respectively. There is less consensus on the small-core vs. large-
core question for the non-metals.

Popular pseudopotentials in modern use include the Los Alamos National Laboratory (or LANL)
ECPs, in particular the LANZ2DZ, which is a valence double-zeta basis set combined with ECPs
for heavy elements, and Stuttgart-Dresden ECPs.