Lecture: ab initio Methods Overview of the Density Functional Theory for Molecules


For many years wave function functional theory (Schrödinger theory) has been the method of choice when performing electronic structure calculations on chemical systems. However, within the last few years, density functional theory has experienced a rise in its popularity for calculating structures and properties of chemical systems. The need to include electron-correlation in calculations on large chemical systems and the discovery of new and more accurate approximations to the exchange-correlation energy density functional has contributed to the increased popularity of density functional theory. Not only does density functional theory (DFT) provide computational advantages over wave function functional theory (WFT), but quantities of interest to chemists (such as electronegativity, hardness, softness, reactivity indices, etc) are readily defined.

Density functional theory is equivalent to solving Schrödinger's equation and is, therefore, an exact theory for describing the electronic structure and properties of matter. An important difference between DFT and WFT is the principle quantity of interest. In DFT this quantity is the electron density which, unlike the wave function, is an observable. As a result, quantities not readily accessible through WFT are conveniently defined. The computational advantage of DFT originates with the fact that the electron density has three spatial coordinates, regardless of the number of electrons in the chemical system. Thus, DFT allows the calculation of structures and properties of molecules with a couple hundred atoms, a feat not generally possible with high-level WFT methods.

In 1964 the perception of density functional theory was forever changed. That year, Hohenberg and Kohn provided the long sought after proof that density functional theory is in fact an exact theory for describing the electronic behavior of matter. This was accomplished by proving the following remarkable theorem: There exits a variational principle in terms of the electron density which determines the ground state energy and electron density. Further, the ground state electron density determines the external potential, within an additive constant.

This states that the ground state electron density, as obtained from the Hohenberg-Kohn variational principle, uniquely determines the ground state properties of the system of interest. The electron density is obtained from the variational principle

equation1

Here tex2html_wrap_inline144 is the ground state energy, as a functional of both the electron density tex2html_wrap_inline146 and external potential tex2html_wrap_inline148 (the nuclear-electron attraction potential in many cases), tex2html_wrap_inline150 (the chemical potential) is a Lagrange multiplier introduced to preserve the number of particles and N is the number of electrons. For emphasis, we now explicitly denote the functional dependence of the energy on the external potential. The solution of the resulting Euler-Lagrange equation

equation2
determines the ground state density and, therefore, all the ground state properties of the system. The subscript tex2html_wrap_inline148 on the functional derivative implies that tex2html_wrap_inline148 is kept fixed when performing the functional derivative with respect to tex2html_wrap_inline146. Equation [2] is the DFT counterpart to the Schrödinger equation. The functional for the ground state energy can be partitioned
equation3
where tex2html_wrap_inline158 is a universal functional of tex2html_wrap_inline146. This universal functional contains kinetic and potential energy contributions,
eqnarray4
where tex2html_wrap_inline162 is a universal functional for the kinetic energy, tex2html_wrap_inline164 is a universal functional for the electron-electron repulsion energy and tex2html_wrap_inline166 is a universal functional of the exchange-correlation energy. Unfortunately, the exact forms of tex2html_wrap_inline162 and tex2html_wrap_inline166 are not known. For this reason Eqs. [1] and [2] are not used in practical applications of density functional theory. However this formal development of density functional theory does provide important relationships, which lead to useful expressions for properties of chemical interest.

Orbital-Density Formalism

A practical description of density functional theory was given by Kohn and Sham, just one year after the momentous Hohenberg-Kohn paper. The Kohn-Sham implementation is based upon an orbital-density description of DFT which removes the necessity of knowing the exact form of tex2html_wrap_inline162. They proposed focusing on the kinetic energy of a non-interacting system of electrons as a functional of a set of single-particle orbitals which give the exact density. Levy has since presented a constrained search formulation of Kohn-Sham theory in which the kinetic energy of a non-interacting system of electrons is minimized with respect to a set of single-particle orbitals, subject to the constraint that the orbitals are orthonormal and that the sum of the squares of the orbitals give the exact ground state density.

In the Kohn and Sham formulation of DFT, the kinetic energy of a non-interacting system of electrons is minimized with respect to a set of single-particle orbitals,{ tex2html_wrap_inline174}, subject to the constraint that the orbitals are orthonormal and that the sum of the squares of the orbitals gives the exact ground state density. The Kohn-Sham variational principle is

equation5
where
equation6
is the kinetic energy for a system of non-interacting electrons,
equation7
is the Coulomb energy,
equation8
is the nuclear-electron attraction potential energy, tex2html_wrap_inline176 is the exchange-correlation energy (containing both kinetic and potential contributions to the correlation energy), the tex2html_wrap_inline178 are Lagrange multipliers introduced to assure that tex2html_wrap_inline174 are orthonormal and
equation9
is the ground state electron density. Here ni is the occupation number of the ith Kohn-Sham orbital, tex2html_wrap_inline174.

The Kohn-Sham variational principle leads to a simple differential equation whose solutions are the Kohn-Sham orbitals,

equation10
where
equation11
is the Kohn-Sham effective potential. This differential equation is reminiscent of the Hartree and Hartree-Fock equations, but it is fundamentally different in that it yields, in principle, the exact ground state density.

The Kohn-Sham Lagrange multipliers tex2html_wrap_inline178 are not excitation energies as are the orbital energies of Hartree-Fock theory. Instead, they reflect the response of the total electronic energy to changes in occupation number, i.e.

equation12
This equation implies that a knowledge of the tex2html_wrap_inline178 provide information about the electronegativity of the system. Indeed, the electronegativity can be obtained from tex2html_wrap_inline192 and tex2html_wrap_inline194. Although no Koopmans'-like theorem exists in Kohn-Sham theory, excitation energies can be derived from the tex2html_wrap_inline178. For instance, the first ionization energy can be found from
equation13
Slater has derived a useful numerical approximation to this expression, namely
equation14
That is, the first ionization energy is found by removing 1/2 an electron from the highest occupied Kohn-Sham molecular orbital. Similarly, the electron affinity is given by
equation15
which may be approximated by
equation16
where 1/2 an electron has been added to the LUMO, the lowest unoccupied Kohn-Sham molecular orbital. Alternatively, the electron affinity can be obtained directly from a knowledge of the ionization energy and electronegativity,
equation17
Of course, both I and A can be calculated from ground state energy differences, as given by Eqs. (27) and (29) respectively. DFT has been quite successful in predicting I and A for atomic and molecular systems.

Since the Kohn-Sham orbitals give the total ground state electron density of the system, they also provide a practical method for calculating the local chemical reactivity properties discussed above. The Fukui reactivity indices may be obtained from the following equations:

equation18
and
equation19
In many instances, the relaxation term, tex2html_wrap_inline198, can be ignored and the frontier molecular orbital theory as first described by Fukui is recovered.

Although the Kohn-Sham formalism is, in principle, exact, its application is made inexact since the exchange-correlation energy density functional, tex2html_wrap_inline176, is not known. Fortunately, reasonable approximations to tex2html_wrap_inline176 are known. The simplest approximation, due to Dirac and later parameterized by Slater, works surprisingly well. More elaborate approximations to tex2html_wrap_inline176 have been proposed and have proven useful in the calculation of molecular properties. For many applications, local approximations to tex2html_wrap_inline176 suffice. But, for those situations in which elongated bonds (transition states, hydrogen bonds, etc) are involved, non-local (gradient) corrections must be utilized. The consequence of using an approximation to tex2html_wrap_inline176 makes Kohn-Sham theory non-variational with respect to the exact ground state energy. However, Kohn-Sham theory is variational with respect to the model system described by the approximate tex2html_wrap_inline176.


Author: Ken Flurchick