You are on page 1of 24

Learning Computational Chemistry

Kyogo Nagashima
Oberlin College, Chemistry
November 15th, 2023

Kyogo Nagashima, Nov 15th, 2023. Do not distribute without permission of the author.
Table of Contents
1. Quantum Chemistry & Hartree Fock ............................................................................. 3
1.1 The Hamiltonian Operator returns Energy as an Observable ..................................... 3
1.2 The Born-Oppenheimer Approximation ............................................................................ 4
1.3 The Hartree Fock Method ...................................................................................................... 5
1.4 The Fock Operator .................................................................................................................. 6
1.5 Hartree Fock Effective Potential Term ............................................................................... 7
1.6 Flaws of the Hartree Product ............................................................................................... 7
1.7 Slater Determinants ................................................................................................................ 8
1.8 The Hartree Fock Algorithm ................................................................................................. 9
References ........................................................................................................................................... 10
2. Limitation of Hartree Fock & Basis Sets.................................................................... 11
2.1 Limitations of Hartree Fock ................................................................................................ 11
2.2 Infinite Basis Sets? ............................................................................................................... 11
2.3 Slater type vs Gaussian type Orbitals ............................................................................. 12
2.4 Contracted Gaussian Type Orbitals ................................................................................. 13
2.5 Multiple - 𝝃 and Split Valence Basis Sets........................................................................ 14
2.6 Polarization Functions ......................................................................................................... 14
2.7 Diffuse Functions .................................................................................................................. 15
References ........................................................................................................................................... 15
3. Density Functional Theory ............................................................................................ 16
3.1 Electron-electron Repulsion............................................................................................... 16
3.2 Electron Density .................................................................................................................... 16
3.3 Away from Wavefunctions .................................................................................................. 17
3.4 Kohn-Hohenberg Existence Theorem ............................................................................. 18
3.5 Hohenberg-Kohn Variational theorem............................................................................. 19
3.6 Kohn-Sham SCF Method ..................................................................................................... 19
References ........................................................................................................................................... 21
4. Modeling Excited States................................................................................................. 21
4.1 Configurational Interaction and beyond ......................................................................... 21
4.2 Time-Dependent Perturbation/Density Functional Theory ........................................ 23
4.3 Solvation in Excited States ................................................................................................. 23

Kyogo Nagashima, Nov 15th, 2023. Do not distribute without permission of the author.
References ........................................................................................................................................... 24

1. Quantum Chemistry & Hartree Fock


In discussing computational chemistry, a brief discussion on Quantum Chemistry
cannot be skipped.

In fact, computational chemistry is all about “solving” the Schrödinger Equation,


especially for non-dynamical methods.

Simply put, the Schrödinger Equation describes the quantum mechanical nature
of systems.

The most common and simplest form of the Schrödinger Equation is the following.

̂ 𝜓 = 𝐸𝜓
𝐻
For any given chemical system, there exists a wavefunction (𝝍) that describes the
quantum state of the system.

When an operator (𝑯 ̂ ) acts on the wavefunction, the Schrödinger equation gives back
the observable (𝑬).

In other words, this equation can be described like a fortune teller.

If you throw a question at a fortune teller, they will provide you back with an answer.

Using this equation, we can solve for values of interest, like energies of a molecule.

However, there is an inherent problem to this.

The Schrödinger equation cannot be solved exactly for


anything above a hydrogen!!
That is why we turn to computational power to solve these problems.

1.1 The Hamiltonian Operator returns Energy as an Observable

As previously state, the Hamiltonian is like a fortune teller.

Given a wavefunction of a chemical system, it will return the energy of the chemical
system.

Kyogo Nagashima, Nov 15th, 2023. Do not distribute without permission of the author.
The Hamiltonian takes various forms of energies in a chemical system into account.

𝑁 𝑀 𝑁 𝑀 𝑁 𝑁 𝑀 𝑀
1 1 2 𝑍𝑘 1 𝑍𝑘 𝑍𝑙
̂ = − ∑ 𝛻𝑖2 − ∑
𝐻 𝛻𝑘 − ∑ ∑ + ∑∑ + ∑∑
2 2𝑀𝑘 𝑟𝑖,𝑘 𝑟𝑗 𝑟𝑘𝑙
𝑖=1 𝑘=1 𝑖=1 𝑘=1 𝑖=1 𝑗>𝑖 𝑘=1 𝑙>𝑘

The Hamiltonian written above account for the following:

1. Kinetic energy of electrons


2. Kinetic energy of the nucleus
3. Nuclei-electron attraction
4. Electron-electron repulsion
5. Nucleus-nucleus repulsion.

1.2 The Born-Oppenheimer Approximation

The nucleus is ~2000 times more massive than an electron.

If the nucleus is ~2000 times more massive, we can safely separate the movement of
the electrons with the nucleus.

This is called the Born-Oppenheimer approximation.

Therefore, we approximate that the nucleus is stationary.

If so, the previous Hamiltonian can be simplified.

This Hamiltonian that we described in the previous section can now be simplified.

𝑁 𝑀 𝑁 𝑀 𝑁 𝑁 𝑀 𝑀
1 1 2 𝑍𝑘 1 𝑍𝑘 𝑍𝑙
̂ = − ∑ 𝛻𝑖2 − ∑
𝐻 𝛻𝑘 − ∑ ∑ + ∑∑ + ∑∑
2 2𝑀𝑘 𝑟𝑖,𝑘 𝑟𝑗 𝑟𝑘𝑙
𝑖=1 𝑘=1 𝑖=1 𝑘=1 𝑖=1 𝑗>𝑖 𝑘=1 𝑙>𝑘

This is the Hamiltonian that has been simplified applying the Born-Oppenheimer
approximation.

Kyogo Nagashima, Nov 15th, 2023. Do not distribute without permission of the author.
𝑁 𝑁 𝑀 𝑁 𝑁
1 𝑍 1
̂ = − ∑ 𝛻𝑖2 − ∑ ∑ 𝑘 + ∑ ∑
𝐻
2 𝑟𝑖,𝑘 𝑟𝑗
𝑖=1 𝑖=1 𝑘=1 𝑖=1 𝑗>𝑖

Specifically, the term that describes the kinetic energy of the nucleus and the
term that describes the nucleus-to-nucleus repulsion has been removed.

If we are assuming that the nucleus is stationary, there would be no kinetic


energy.

If the nuclei are stationary, the nucleus-to-nucleus repulsion would be


constant; if the distance between two nuclei is the same, the strength of
repulsion shouldn’t change. If a term is a constant in an operator, that term
can be removed.

1.3 The Hartree Fock Method

The Hamiltonian that we are currently considering “bundles” multiple electrons into the
operator.
𝑁 𝑁 𝑀 𝑁 𝑁
1 2 𝑍𝑘 1
̂ = − ∑ 𝛻𝑖 − ∑ ∑
𝐻 + ∑∑
2 𝑟𝑖,𝑘 𝑟𝑗
𝑖=1 𝑖=1 𝑘=1 𝑖=1 𝑗>𝑖

More specifically, in the equation above, there are N-electrons.

The first term that describes the kinetic energy of electrons is a summation of kinetic
energies for N-electrons.

The second term that describes the electron-nucleus interactions is a summation of


these attraction between all combinations of N-electrons and M-nuclei.

Finally, the third term that describes electron-electron interactions is a summation of all
the possible combination of repulsive force between N-nuclei.

However, a Hamiltonian with multiple electrons is difficult to solve.

Instead, if there are N-electrons in the system, we want:

N number of 1 electron Hamiltonian.

Kyogo Nagashima, Nov 15th, 2023. Do not distribute without permission of the author.
This is straight forward for the first two terms since these terms can be broken down to
1 electron terms.

However, this is not necessary the case for the third term.
The third term can never be broken down to 1 electron terms because
accounting for electron-electron repulsion inherently requires us to consider two
electrons simultaneously.

Therefore, in the Hartree Fock method, the electron-electron repulsion term is forcefully
made into N number of 1 electron equations by considering electron-electron repulsions
as the repulsion of 1 electron against a field of all other electrons .

**insert image of 1 electron repulsion against a “field of electrons”.

This is called the Hartree Fock Approximation.

1.4 The Fock Operator

Each of these 1 electron Hamiltonians is now called a Fock operator.

Therefore, the Hamiltonian of the system will be expressed as the sum of N Fock
operators.

𝐻 = ∑ 𝑓(𝑥𝑖 )
𝑖=1

The 𝑓(𝑥𝑖 ) is the individual Fock operator.

Each of the Fock operators is now a modified version of the Hamiltonian with the Born-
Oppenheimer approximation.

The Hamiltonian changes from the following,

𝑁 𝑁 𝑀 𝑁 𝑁
1 𝑍 1
̂ = − ∑ 𝛻𝑖2 − ∑ ∑ 𝑘 + ∑ ∑
𝐻
2 𝑟𝑖,𝑘 𝑟𝑗
𝑖=1 𝑖=1 𝑘=1 𝑖=1 𝑗>𝑖

to the summation of Fock operators described below.

1 𝑍𝑘
𝑓(𝑥𝑖 ) = − ∇2𝑖 − + 𝑽𝑯𝑭
2 𝑟𝑖,𝑘

Kyogo Nagashima, Nov 15th, 2023. Do not distribute without permission of the author.
Notice that the final term that represents electron-electron repulsion is now replaced by
the Hartree Fock approximation.

Subsequently, the Schrödinger Equation for 1 electron becomes the following when the
Hartree Fock approximation is applied.

𝑓(𝑥𝑖 )𝑋(𝑥𝑖 ) = 𝜀𝑋(𝑥𝑖 )

𝑓(𝑥𝑖 ) is the Fock operator, 𝑋(𝑥𝑖 ) is the wavefunction, and 𝜀 is the energy for one
electron.

1.5 Hartree Fock Effective Potential Term

We will now look into the final Hartree Fock electron-electron repulsion term.
1 𝑍
In the Fock operator, 𝑓(𝑥𝑖 ) = − 2 ∇2𝑖 − 𝑟 𝑘 + 𝑽𝑯𝑭 , 𝑉𝐻𝐹 is characterized by the following
𝑖,𝑘
equation.
𝑁 𝑁
𝜌𝑖 ∗ 1
𝑉𝐻𝐹 = ∑∫ 𝑑𝑟 = ∑ ∫ 𝜓𝑖,𝑗 𝜓 𝑑𝑟
𝑟𝑖,𝑗 𝑟𝑖,𝑗 𝑖,𝑗
𝑖=1 𝑖=2

𝜌𝑖 is the electron density where 𝜌 = 𝜓 ∗ 𝜓.

Therefore, for a system of two electrons, the 𝑉𝐻𝐹 term will become the following.

∗ 1
1. 𝑽𝒊 = ∫ 𝜓𝑖,𝑗 𝜓𝑖,𝑗 𝑑𝒓𝒋
𝑟 𝑖,𝑗
∗ 1
2. 𝑽𝒋 = ∫ 𝜓𝑖,𝑗 𝜓𝑖,𝑗 𝑑𝒓𝒊
𝑟 𝑖,𝑗

1
Even though this equation includes the 𝑟 term, because they are integrated with
𝑖,𝑗
respect to either one of the electrons, the function can become a 1 electron equation.

1.6 Flaws of the Hartree Product

A wavefunction can be constructed from N-number of 1 electron wavefunctions, by


simply taking the product of the 1 electron wavefunctions like the following.

𝜓𝐻𝑃 (𝑥1 , 𝑥2 , … 𝑥𝑁 ) = 𝑋𝑖 (𝑥1 )𝑋𝑗 (𝑥2 ) … 𝑋𝑘 (𝑥𝑁 )

This is called the Hartree Product.

Kyogo Nagashima, Nov 15th, 2023. Do not distribute without permission of the author.
However, the HP doesn’t take the Pauli exclusion principle and
indistinguishability of electrons into account.

- Pauli exclusion principle: 2 electrons cannot have the same spin.

Therefore, when you exchange the spin/spatial coordinates of two electrons, the sign of
the wavefunction must change to follow the Pauli exclusion principle.

In other words, ψHP (x1 , x2 ) = −ψHP (x2 , x1 ) is satisfied.

A wavefunction that clears this criterion is antisymmetric.

• Indistinguishability of electrons: every electron is the same/indistinguishable.

Therefore, by theory, even if we exchange the electrons in a Hartree product, we should


be arriving at identical results.

In other words, Xi (x1 )Xj (x2 ) = Xi (𝑥2)𝑋𝑗 (𝑥1 ) must be satisfied.

However, this is not always the case as 1 electron wavefunctions are not all the same
and so exchanging the coordinates and spin state of electrons would lead to deviations.

To overcome some of these issues, the Slater Determinant was developed.

1.7 Slater Determinants

In short, the slater determinant is a “linear combination” of the Hartree Product.

We start by addressing the indistinguishability of electron problem.

2 electrons should be exchangeable if they are indistinguishable.

1. 𝜓𝐻𝑃 (𝒙𝟏 , 𝒙𝟐 ) = 𝑋𝑖 (𝒙𝟏 )𝑋𝑗 (𝒙𝟐 )


2. 𝜓𝐻𝑃 (𝒙𝟐 , 𝒙𝟏 ) = 𝑋𝑖 (𝒙𝟐 )𝑋𝑗 (𝒙𝟏 )

Then we take the linear combination of equation 1 and 2.

Kyogo Nagashima, Nov 15th, 2023. Do not distribute without permission of the author.
1
𝜓𝑠𝑙𝑎𝑡𝑒𝑟 (𝑥2 , 𝑥1) = [𝑋𝑖 (𝑥1 )𝑋𝑗 (𝑥2 ) − 𝑋𝑖 (𝑥2)𝑋𝑗 (𝑥1 )]
√2

1 𝑋𝑖 (𝑥1 ) 𝑋𝑗 (𝑥1)
𝜓𝑠𝑙𝑎𝑡𝑒𝑟 (𝑥2 , 𝑥1 ) = | |
√2 𝑋𝑖 (𝑥2) 𝑋𝑗 (𝑥2 )

This is the slater determinant.


1
The is a normalization factor.
√2

What would slater determinants for triplet state He(1s,2s) look like?

Here, we have decomposed the previous equation into spatial components and spin
components.

ψHP (x1 , x2 ) = 𝛟𝟏𝐬 (𝐱 𝟏 )𝛟𝟏𝐬 (𝐱 𝟐 )𝛂(𝐱 𝟏 )𝛃(𝐱 𝟐 )


ψHP (x2 , x1 ) = 𝛟𝟏𝐬 (𝐱 𝟐 )𝛟𝟏𝐬 (𝐱 𝟏 )𝛂(𝐱 𝟐 )𝛃(𝐱 𝟏 )

Since both electrons are in the same orbital, the spatial part of the function is identical.
However, the spins are opposite.

The slater orbital then becomes the following.

1
ψslater (x2, x1 ) = [ϕ1s(x1 )ϕ1s (x2 )α(x1 )β(x2 ) + ϕ1s (x2 )ϕ1s(x1 )α(x2 )β(x1)]
√2

It can be simplified into the equation below.

1
ψslater (x2 , x1 ) = [𝛟𝟏𝐬 (𝐱 𝟏 )𝛟𝟏𝐬 (𝐱 𝟐 )][𝛂(𝐱 𝟏 )𝛃(𝐱 𝟐 ) + 𝛂(𝐱 𝟐 )𝛃(𝐱 𝟏 )]
√2

This can be extended for systems with more than 2 electrons.

[𝑋𝑎 (𝑥1 ) ⋯
1 𝑋𝑧 (𝑥1 )
𝜓𝑠𝑙𝑎𝑡𝑒𝑟 (𝑥1 , 𝑥2 , 𝑥𝑁 ) = ( ⋮ ⋱ ⋮ )
√2 𝑋 (𝑥 ) ⋯ 𝑋𝑧 (𝑥𝑁 )
𝑎 1

1.8 The Hartree Fock Algorithm

An iterative algorithm can solve the Hartree Fock Schrödinger Equation.

The one electron equation is the following.

𝑓(𝑥𝑖 )𝑋(𝑥𝑖 ) = 𝜀𝑋(𝑥𝑖 )


Going back to the Fock Operator,

Kyogo Nagashima, Nov 15th, 2023. Do not distribute without permission of the author.
1 𝑍𝑘
𝑓(𝑥𝑖 ) = − 𝛻𝑖2 − + 𝑉𝐻𝐹
2 𝑟𝑖,𝑘
𝑁 𝑁
𝜌𝑖 ∗
1
𝑉𝐻𝐹 = ∑∫ 𝑑𝑟 = ∑ ∫ 𝜓𝑖,𝑗 𝜓 𝑑𝑟
𝑟𝑖,𝑗 𝑟𝑖,𝑗 𝑖,𝑗
𝑖=1 𝑖=2

Then, for 𝜓 we can use the Slater Determinant.

1 𝑋𝑖 (𝑥1 ) 𝑋𝑗 (𝑥1)
𝜓𝑠𝑙𝑎𝑡𝑒𝑟 (𝑥2 , 𝑥1 ) = | |
√2 𝑋𝑖 (𝑥2) 𝑋𝑗 (𝑥2 )

These are all the pieces necessary in describing the algorithm.

1. The algorithm begins with making a guess for 𝑋𝑖 and 𝑋𝑗 .


2. Using 𝑋𝑖 and 𝑋𝑗 , construct the slater determinants (𝜓).
3. Use the slater determinants (𝜓) to solve for 𝑉𝐻𝐹 .
4. Then use 𝑉𝐻𝐹 to evaluate the Fock operator (𝑓).
5. Then, use the Fock operator (𝑓) to calculate new 𝑋𝑖 and 𝑋𝑗 .
6. Repeat steps 2 ~ 5 until the self-consistent field (SCF) convergence criteria is
met.

Self consistent field (SCF) convergence criteria


Once the orbital energies are consistent among the set of 1 electron equations, then we
can say that the convergence criteria have been met.

References

• Antisymmetric Wavefunctions can be Represented by Slater Determinants,


LibreTexts Chemistry.
https://chem.libretexts.org/Courses/Pacific_Union_College/Quantum_Chemistry/
08%3A_Multielectron_Atoms/8.06%3A_Antisymmetric_Wavefunctions_can_be_
Represented_by_Slater_Determinants (accessed 2023-09-01).
• Cramer, C. J. Essentials of Computational Chemistry, Theories and Models, 2 nd
edition; John Wiley & Sons, 2004.
• Hartree Fock Method: A Simple Explanation. https://insilicosci.com/hartree-fock-
method-a-simple-explanation/ (accessed 2023-08-29).
• The Self-Consistent Field Approximation (Hartree-Fock Method), LibreTexts,
Chemistry.
https://chem.libretexts.org/Bookshelves/Physical_and_Theoretical_Chemistry_Te
xtbook_Maps/Book%3A_Quantum_States_of_Atoms_and_Molecules_(Zielinksi_
et_al)/09%3A_The_Electronic_States_of_the_Multielectron_Atoms/9.07%3A_Th
e_Self-Consistent_Field_Approximation_(Hartree-Fock_Method) (accessed
2023-08-28).

Kyogo Nagashima, Nov 15th, 2023. Do not distribute without permission of the author.
2. Limitation of Hartree Fock & Basis Sets
2.1 Limitations of Hartree Fock

The Hartree Fock method introduced in the previous section can give very accurate
computations.

However, just like in physical chemistry, we will be comparing energies rather than
judging its absolute magnitude.

In other words, we need an extremely high level of accuracy because the errors will be
magnified when we are comparing two energies that are very similar.

Therefore, slight inaccuracies will become magnified when differences are taken.

The error inherent to HF is called correlation energy (𝑬𝒄𝒐𝒓𝒓 ).

𝑬𝒄𝒐𝒓𝒓 = 𝐸 − 𝐸𝐻𝐹

𝐸 is the true energy of the system and 𝐸𝐻𝐹 is the energy calculated using the Hartree
Fock method.
This inaccuracy arises because HF doesn’t accurately model correlated effects of
multiple electrons.

In Hartree Fock, the electron-electron repulsion term is only an approximation, a


sacrifice that was necessary in implementing Fock operators which are 1 electron
operators.

Given this situation, we can say that modern computational chemistry is about
minimizing the correlation energy.

2.2 Infinite Basis Sets?

The first step in the Hartree Fock algorithm is to make a guess of the wavefunction. This
guess is called a trial function. Using this trial function, we can calculate the
associated eigen-value (Energy/any other observable) and iteratively improve our
solution following the variational principle.

Therefore, it is important that our initial trial function is reasonable.

Because every system that is to be computed will most likely be extremely unique, the
trial function must be constructed in a way that can be applied for any molecule input.

Kyogo Nagashima, Nov 15th, 2023. Do not distribute without permission of the author.
We accomplish this by constructing a trial function that is a linear combination of a set of
functions like the following.

𝑎1 𝐴1 + 𝑎2 𝐴2 + 𝑎3 𝐴3 + 𝑎4 𝐴4 + ⋯

Molecular orbitals can be represented as a linear combination of atomic orbitals.


This approach is called the linear combination of atomic orbitals approach, often
abbreviated as LCAO.

Mathematically, what that means is, the wavefunction of the molecule is constructed
from a linear combination of functions.

𝝍𝑴𝑶 = 𝑎𝝓𝟏 + 𝑏𝝓𝟐 + ⋯ + 𝑧𝝓𝑵

The set of functions used to construct the wavefunction of the molecule is called a
basis set.

If, we had an infinite number of functions that comprise the basis set, we will be able
to reach the HF limit, which is the best calculation the HF method can give.

But this is computationally expensive and not feasible.

If so, can we reach the HF limit by smarter choice of basis sets?

The linear combination of atomic orbitals (LCAO) approach sounds very reasonable,
especially for those with some training in chemistry. However, we must not forget
that in computational chemistry, the trial functions can have chemical meaning,
but they are “just” functions regardless.

Therefore, the functions that make the most chemical sense isn’t always the most
efficient computationally.

What matters the most is whether the basis set is flexible enough to allow placement of
electron density adequately.

2.3 Slater type vs Gaussian type Orbitals

Slater type orbitals (STO), as previously introduced have been very successful in
modeling electron densities of hydrogen atom like orbitals.

However, calculations involving STOs have been shown to be computationally


expensive.

On the other hand, Gaussian type orbitals (GTO) are computationally inexpensive.

Kyogo Nagashima, Nov 15th, 2023. Do not distribute without permission of the author.
These gaussian orbitals have several types:

- 1 s-type orbital
- 3 p-type orbitals (px, py, pz),
- 6 d-type orbitals which can be decreased to 5 to span then entire
space (dxy, dxz, dyz, dx2-y2, dz2-r2).

However, Gaussian type orbitals also have inherent flaws in modelling the orbitals.

2.4 Contracted Gaussian Type Orbitals

As a solution, contracted Gaussian type orbitals (contracted GTO) are commonly


used.

Contracted GTO are linear combinations of GTO that intend to model STO.

𝜓𝑐𝑜𝑛𝑡𝑟𝑎𝑐𝑡𝑒𝑑 𝐺𝑇𝑂 = ∑ 𝑐𝜙𝐺𝑇𝑂

The individual GTO are called primitives. The number of primitives correlate with the
accuracy and cost of the calculation.

Contracted GTO with only one primitive function assigned to core and valence electrons
of the atom/molecule are called minimal basis sets (single-𝜉).

For instance, H and He will have 1 primitive (1s) and Li ~ Ne would have 5 primitives
(1s, 2s, 2px, 2py, 2pz).

Kyogo Nagashima, Nov 15th, 2023. Do not distribute without permission of the author.
2.5 Multiple - 𝝃 and Split Valence Basis Sets

While Minimal basis (single-𝜉) sets are computationally cheap, they are not necessarily
flexible enough at times.

As a solution, multiple-𝝃 basis sets can be used in which a linear combination of GTO
represents one orbital.

Since two or more GTO are used in multiple-𝝃 basis sets, different parts of the linear
combination can be used to account for various parts of the orbital (ex. near the nucleus
vs away from the nucleus).

However, multiple-𝝃 basis sets can be computationally expensive although they will
start approaching the Hartree Fock limit.

Split Valence basis sets use a minimal basis set for core orbitals and multiple- 𝝃 for
valence orbitals because valence orbitals require more flexibility.
- For instance, an atom bonded to an electronegative atom, it will have a partial
positive charge. Then, the remaining electrons will be distributed more
“compactly”.
- If an atom is bonded to an electropositive atom, it will now have more electron
density and the electrons will be “loosely” bounded to the nucleus.

Therefore, flexibility in the valence electrons are an important criteria for a basis set.
2.6 Polarization Functions

Normal basis functions composed of orbitals in which the electrons are occupied, isn’t
always enough.

A molecule of NH3 has a trigonal pyramidal geometry.

However, if you use a basis set composed of only s and p orbital type contracted GTO,
computational results will predict a trigonal planar geometry.

True Geometry Predicted Geometry


While s and p orbitals are only occupied in nitrogen and hydrogen, mathematically GT s
and p-orbitals are not flexible enough.

Kyogo Nagashima, Nov 15th, 2023. Do not distribute without permission of the author.
By adding GT d-orbitals, this problem can be fixed: this is a Polarization function.

Polarization functions are added to most atoms except for H and (excluding C-H
activation, H atom transfer).

2.7 Diffuse Functions

Polarization provides great flexibility.

But how about for anions? excited states? Pi-pi interactions? They require an even
greater flexibility.

This is because the added electron density will be even further away from the nucleus,
getting distorted to the extent by which polarization functions cannot take care of.

In such cases, diffuse functions must be added to account for this increased distortion
of electron density.

References

• Cramer, C. J. Essentials of Computational Chemistry, Theories and


Models, 2nd edition; John Wiley & Sons, 2004.
• Correlation Energy and Configuration Interaction, LibreTexts, Chemistry.
https://chem.libretexts.org/Bookshelves/Physical_and_Theoretical_Chemistry_Te
xtbook_Maps/Book%3A_Quantum_States_of_Atoms_and_Molecules_(Zielinksi_
et_al)/10%3A_Theories_of_Electronic_Molecular_Structure/10.09%3A_Correlati
on_Energy_and_Configuration_Interaction (accessed 2023-09-10).
• Orbital Polariztion Terms in Basis Sets, LibreTexts, Chemistry.
https://chem.libretexts.org/Courses/Pacific_Union_College/Quantum_Chemistry/
11%3A_Computational_Quantum_Chemistry/11.04%3A_Orbital_Polarization_Te
rms_in_Basis_Sets (accessed 2023-09-13).
• Gaussian Basis Sets, LibreTexts, Chemistry.
https://chem.libretexts.org/Courses/Pacific_Union_College/Quantum_Chemistry/
11%3A_Computational_Quantum_Chemistry/11.02%3A_Gaussian_Basis_Sets
(accessed 2023-09-14).

Kyogo Nagashima, Nov 15th, 2023. Do not distribute without permission of the author.
3. Density Functional Theory
3.1 Electron-electron Repulsion

The biggest approximation made in the HF-SCF method is the treatment of


electron-electron repulsion.

The behavior of electrons is dependent on one another. Therefore, treating them as a


collection of one-electron equations in HF method leads to errors.

People have tried to solve this issue from different angles: DFT is a good example.

3.2 Electron Density

Electron density is the most important variable in density functional theory (DFT).

Methods that have been discussed so far have focused on the importance of a
wavefunction.

But what’s a wavefunction in the first place? While we can always state the formal
definition, it is not something that is relatable. Electron density on the other hand is
a much more tangible observable that we can possibly base this new method on.

To construct the Hamiltonian, we need to know the total number of electrons, the
position of the nucleus, and the charge of the nucleus.

Kyogo Nagashima, Nov 15th, 2023. Do not distribute without permission of the author.
The integral of the density along the entire space gives the total number of
electrons.
𝑁 = ∫ 𝜌(𝒓)𝑑𝒓
𝑁 is the number of electrons and 𝜌 is the electron density.

The position with the local maximum density in the form of a cusp is the position
of nucleus. In other words, it is the location with maximum electron density.

𝜕𝜌(𝑟)
=0
𝜕𝑟

The height of that cusp, or how much electron density is around tells us what the
charge of the nucleus is.

𝜕𝜌
̅ (𝑟)
= −2𝑍𝜌(𝑟)
𝜕𝑟
𝑤ℎ𝑒𝑛 𝑟 = 0

Overall, the electron densities can be defined as the following.

𝜌 = 2 ∑ 𝜓𝑖∗ 𝜓𝑖
𝑖=1

3.3 Away from Wavefunctions

The initial stages of DFT were far from useful. It used classical physics to approximate
energies using electron densities (𝝆).

Using electron densities, the electron-nuclear attraction was modeled as the


following.
𝒁𝒏𝒖𝒄
𝑽[𝝆(𝒓)] = ∑ 𝝆(𝒓)𝒅𝒓
𝒓𝒆𝒍𝒆𝒄 − 𝒓𝒏𝒖𝒄

The electron-electron repulsion was reported using classical physics with a


correction term.

𝟏 𝝆(𝒓𝟏 )𝝆(𝒓𝟐 ) 𝟏 𝝆(𝒓𝟏 )𝒉(𝒓𝟏 ; 𝒓𝟐 )


𝑉[𝜌(𝑟)] = ∫∫ 𝒅𝒓𝟏 𝒅𝒓𝟐 + ∫ ∫ 𝒅𝒓𝟏 𝒅𝒓𝟐
𝟐 |𝒓𝟏 − 𝒓𝟐 | 𝟐 |𝒓𝟏 − 𝒓𝟐 |

Kyogo Nagashima, Nov 15th, 2023. Do not distribute without permission of the author.
The hole function (𝒉(𝒓𝟏 ; 𝒓𝟐 )) accounts for the electron hole that electron 1 creates
(There would be a “hole” in the electron density around electron 1 due to repulsion).
This term is also used to incorporate to exchange correlation energies.

3.4 Kohn-Hohenberg Existence Theorem

Section 3.2 described how the ubiquity and convenience of replacing wavefunctions
with electron-densities would be of any benefit.

However, in using electron densities instead of wavefunctions in solving the


“Schrödinger Equation”, there were two important fundamental theorems that were
proven.

The first of the two is the Hohenberg-Kohn Existence Theorem that accelerated the
development of density functional theory (DFT).

The Hohenberg-Kohn Existence Theorem proved that the electron density is tied closely
to the energy of molecule.

More formally, this theorem states that the ground state energy is a unique functional of
the electron density.

In other words, if you supply a function that describes electron density to an energy
functional, that functional will return an energy value.

Most importantly, this proves that electron densities and energy have a one-to-one
relationship.

Kyogo Nagashima, Nov 15th, 2023. Do not distribute without permission of the author.
3.5 Hohenberg-Kohn Variational theorem.

The previous Hohenberg-Kohn Existence Theorem precisely states that the electron
densities give us a one-to-one mapping to a Hamiltonian, then wavefunction, then
energy.

Therefore, the Hohenberg-Kohn Existence theorem is not enough in


escaping wavefunctions.

The Hohenberg-Kohn Variational theorem, as its name suggests, has proved that the
variational principle can be applied to predicting energies using electron densities.

The variational principle states that the energy calculated from the variational method is
always equal to or larger than the true energy.

Therefore, this theorem states that the lower the predicted energy is that electron
density is getting closer and closer to the “true electron density”.

However, these two theorems aren’t enough in allowing us to escape wavefunctions.

3.6 Kohn-Sham SCF Method

With the Hohenberg-Kohn theorem setting the grounds for further developments of
density functional theory, it was the Kohn-Sham Self-consistent-field method that laid
the basic algorithm that allowed the initial implementation.

This was the algorithm that allowed a switch from trial functions to trial electron
densities.

The Kohn-Sham SCF Method uses the relationship between wavefunctions and
electron densities to the full extent.

𝜌 = 2 ∑ 𝜓𝑖∗ 𝜓𝑖
𝑖=1

Then, most importantly, the Kohn-Sham equation must be defined.


The Kohn-Sham equation is composed of an energy functional, meaning the energy is a
“functional” of the electron density function as shown in the equation below.

𝐸[𝜌(𝒓)] = 𝐾𝑛𝑜−𝑖𝑛𝑡𝑒𝑟𝑎𝑐. [𝜌(𝒓)] + 𝑉𝑛𝑢𝑐−𝑒𝑙𝑒𝑐 [𝜌(𝒓)] + 𝑉𝑒𝑙𝑒𝑐−𝑒𝑙𝑒𝑐 [𝜌(𝒓)] + 𝐸𝑋𝐶 [𝜌(𝒓)]

Here we have an energy functional, which means it accepts a function, in this case
electron density. The energy can be broken down into roughly 4 parts.

Kyogo Nagashima, Nov 15th, 2023. Do not distribute without permission of the author.
1. 𝐾𝑛𝑜−𝑖𝑛𝑡𝑒𝑟𝑎𝑐. [𝜌(𝒓)] is the kinetic energy of non-interacting electrons. Here it is
assumed that the electrons are not interacting with each other.

2. 𝑉𝑛𝑢𝑐−𝑒𝑙𝑒𝑐 [𝜌(𝒓)] is the potential energy from the interaction between the nucleus
and the electrons.

3. 𝑉𝑒𝑙𝑒𝑐−𝑒𝑙𝑒𝑐 [𝜌(𝒓)] is the potential energy from the interactions between.

4. Finally, 𝐸𝑋𝐶 [𝜌(𝒓)] is the exchange correlation functionals which account for all
other forms of error that is not captured by the previous terms. This exchange
correlation functional will be re-visited in further depth.

With this energy functional defined, we can now define the Kohn-Sham equations.

This may look like an ordinary Schrödinger equation but there are a few key differences.

ℎ𝑖𝐾𝑆 𝜒𝑖 = 𝜀𝑖 𝜒𝑖

𝑀
1 𝑍𝑘 𝜌(𝒓)
ℎ𝑖𝐾𝑆 = − 𝛻𝑖2 − ∑ +∫ 𝑑𝑟 + 𝑉𝑥𝑐
2 𝑟𝑖,𝑘 𝑟𝑖
𝑘=1

Now, we have all the parts necessary to describe the Kohn-Sham Scheme.

1. As previously stated, instead of a guess trial wavefunction, we employ a trial


electron density (𝜌(𝒓)).

2. We apply this trial electron density to the Kohn-Sham equations to solve for a
single electron wavefunction.

3. Then, using this single electron wavefunction, you can calculate the electron
density.

4. Finally, the newly calculated electron density is compared with the initial guess
trial electron density. If they are different, the algorithm is re-iterated from step 2.
If they are the “same”, this means that the convergence criteria have been met
and the algorithm can be terminated.

Kyogo Nagashima, Nov 15th, 2023. Do not distribute without permission of the author.
References

- Cramer, C. J. Essentials of Computational Chemistry, Theories and


Models, 2nd edition; John Wiley & Sons, 2004.
- Pribram-Jones, A; Gross, D. A.; Burke, K. DFT: A Theory Full of Holes? Annu.
Rev. Phys. Chem. 2015, 66, 283-304. DOI: 10.1146/annurev-physchem-040214-
121420.
- Introduction to Density Functional theory [Part One] Background. Matt
Timm, PhD. November 1st, 2021.
https://www.youtube.com/watch?v=Ez_Fm4iTUeo&t=663s

4. Modeling Excited States


While we have so far focused on molecules and chemical systems in the ground state,
that is the lowest possible energy configuration, this type of calculation is not sufficient
for all systems.

From dyes to photosensitizers to photocatalysts, there are many chemical systems


in which exploring the system in its excited state is of high interest.

4.1 Configurational Interaction and beyond

Current methods can be argued as a variation of the Hartree Fock method in which we
use a trial wave function or its equivalent to calculate the lowest possible energy.

While this is an extremely important method, it is not all encompassing.

We must not forget that the lowest eigen-value is “not the only solution”. While the
lowest energy eigen-value corresponds to the ground state, higher energy eigen values
will similarly correspond to excited states.

For some systems, computing/accounting for the possibility of access to higher energy
states can be helpful. The simplest way to account for this is the Configuration
Interaction method.

In the Hartree Fock method, a Slater determinant is often utilized to account for
electrons filling the various molecular orbitals.
𝑜𝑐𝑐𝑢𝑝 𝑣𝑖𝑟𝑡

𝜓 = 𝜓𝐻𝐹 + ∑ ∑ 𝑎𝑖𝑥 𝜓𝑖𝑥 + ⋯


𝑖 𝑥

Kyogo Nagashima, Nov 15th, 2023. Do not distribute without permission of the author.
For example, in a system with 4 electrons, 2 of which is occupied, the rest being virtual
orbitals, the 𝜓𝐻𝐹 accounts for the ground state, and the additional summation terms
account for all combination of singly excited states.

Because they are a linear combination of functions, we can further expand this basis set
to allow more flexibility for higher excited states.

In this current model of configurational interaction, literally all possible excitations are
being considered. This is also called the full configuration interaction method.

However, excitation to a LUMO or LUMO+1 is much more likely than an excitation to a


LUMO+5.

In methods such as CAS-SCF and RAS-SCF, there are stricter restrictions made on
which electrons can be excited.

Kyogo Nagashima, Nov 15th, 2023. Do not distribute without permission of the author.
In these methods, the user can specify the orbitals in which the electronic excitations
are mainly explored.

In the No transition orbitals, doubly occupation of each orbital is enforced. No


transitions from these orbitals are explored. In the restricted active space, there is a
restriction on the number of electrons that can be excited. Only in the
complete active space is all the combinations tested.

Common methods are CAS-SCF, RAS-SCF, and CASPT2.

4.2 Time-Dependent Perturbation/Density Functional Theory

While I will not go into the detailed theory on how time-dependent perturbation/density
functional theory works, I would like to go into general approaches that are being made.

As its name suggests, the time-dependent Schrödinger equation must be considered in


which a time-dependent is introduced.

ℏ 𝜕𝜓
− ̂𝜓
=𝐻
𝑖 𝜕𝑡
The wavefunction for this can be thought of as a correction of the time-independent
Schrödinger equation (Φ).

𝑖𝐸𝑡
−( )
𝜓=𝑒 ℏ Φ

A system in an excited state must be provided with an energy. In the case of UV/vis-
spectroscopy, the system is excited with UV radiation. Therefore, the Hamiltonian must
also take this into account.

The Hamiltonian that was previously discussed is now perturbed by a term that
simulates incident radiation on the system.

𝐻 = 𝐻 0 + 𝑒0 𝒓sin (2𝜋𝑣𝑡)

𝑒0 gives the charge, 𝒓 the position, and a frequency and time-dependence are also
introduced here.

4.3 Solvation in Excited States

Solvents is another consideration that must be made. Solvation effects are often
underestimated in computational calculations. For excited state calculations, serious
consideration of solvation is necessary.

Kyogo Nagashima, Nov 15th, 2023. Do not distribute without permission of the author.
The presence of solvents can significantly change the energies and optimal geometries
of structures.

There are two key types of solvation: equilibrium and nonequilibrium solvation.

In a ground state system, the electronic and nuclear components of the solvent is
oriented in a way that is most comfortable (lowest energy configuration). This is the
equilibrium solvation.

Therefore, in modelling this computationally, we must have a wavefunction that


minimizes the Hamiltonian in the gas phase + potential interactions between the
solvent + reorganization energy for the solvent/solute to reorient themselves.

However, when this system is excited, the electronic distribution of the molecule of
interest will change. With this change, the electronic distribution of the solvents will also
make subsequent changes. However, in the timescale of electronic excitation, the
nuclear arrangement of the solvents simply cannot change. This is a non-equilibrium
solvation.

References

- Configuration Interaction, LibreTexts, Chemistry.


https://chem.libretexts.org/Bookshelves/Physical_and_Theoretical_Chemistry_Te
xtbook_Maps/Book%3A_Quantum_States_of_Atoms_and_Molecules_(Zielinksi_
et_al)/09%3A_The_Electronic_States_of_the_Multielectron_Atoms/9.08%3A_Co
nfiguration_Interaction (accessed 2023-09-23).
- Cramer, C. J. Essentials of Computational Chemistry, Theories and
Models, 2nd edition; John Wiley & Sons, 2004.
- Foresman, J. B. Exploring Chemistry with Electronic Structure Methods, 3 rd
edition; Gaussian, Inc. 1993.

Kyogo Nagashima, Nov 15th, 2023. Do not distribute without permission of the author.

You might also like