Professional Documents
Culture Documents
PAVLO O. DRAL
Professor, State Key Laboratory of Physical Chemistry of Solid Surfaces, Fujian Provincial Key
Laboratory of Theoretical and Computational Chemistry, Department of Chemistry, College of Chemistry and
Chemical Engineering, Xiamen University, China
Elsevier
Radarweg 29, PO Box 211, 1000 AE Amsterdam, Netherlands
The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, United Kingdom
50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States
Copyright © 2023 Elsevier Inc. All rights reserved.
No part of this publication may be reproduced or transmitted in any form or by any means, electronic or
mechanical, including photocopying, recording, or any information storage and retrieval system, without
permission in writing from the publisher. Details on how to seek permission, further information about the
Publisher’s permissions policies and our arrangements with organizations such as the Copyright Clearance
Center and the Copyright Licensing Agency, can be found at our website: www.elsevier.com/permissions.
This book and the individual contributions contained in it are protected under copyright by the Publisher (other
than as may be noted herein).
Notices
Knowledge and best practice in this field are constantly changing. As new research and experience broaden our understanding,
changes in research methods, professional practices, or medical treatment may become necessary.
Practitioners and researchers must always rely on their own experience and knowledge in evaluating and using any information,
methods, compounds, or experiments described herein. In using such information or methods they should be mindful of
their own safety and the safety of others, including parties for whom they have a professional responsibility.
To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any liability for any
injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or
operation of any methods, products, instructions, or ideas contained in the material herein.
ISBN: 978-0-323-90049-2
v
vi Contents
7. Unsupervised learning 2
Rose K. Cersonsky and Sandip De
Machine learning potentials
Introduction 153
Notation guide 154 11. Potentials based on linear models
Methods 155 Gauthier Tallec, Gaetan Laurens, Owen Fresse-Colson, and
Case studies 171 Julien Lam
Conclusions 178
References 178 Introduction 253
Methods 255
Case studies 272
8. Neural networks Conclusion and outlook 276
Pavlo O. Dral, Alexei A. Kananenka, Fuchun Ge, Acknowledgments 276
and Bao-Xin Xue
References 277
Introduction 183
Methods 184 12. Neural network potentials
Case study 200 Jinzhe Zeng, Liqun Cao, and Tong Zhu
Conclusions and outlook 201
Acknowledgments 202 Introduction 279
References 203 Methods 281
Case studies 287
Conclusions and outlook 290
9. Kernel methods Acknowledgment 290
Max Pinheiro Jr and Pavlo O. Dral References 290
Introduction 391
Methods 392
4
Case studies 411 Machine learning-improved
Conclusions and outlook 423 quantum chemical methods
Acknowledgments 423
References 424 21. Learning from multiple
quantum chemical methods:
3 Δ-learning, transfer learning, co-kriging,
and beyond
Machine learning of quantum Pavlo O. Dral, Tetiana Zubatiuk, and Bao-Xin Xue
chemical properties
Introduction 491
18. Learning electron densities Methods 493
Bruno Cuevas-Zuvirı́a Case studies 503
Conclusions and outlook 505
Introduction 431 Acknowledgments 506
Methods 435 Author contributions 506
Case studies 445 References 506
viii Contents
Introduction 510
5
Methods 511 Analysis of big data
Case studies 523
Conclusions and outlook 527 26. Analysis of nonadiabatic molecular
Acknowledgment 527
dynamics trajectories
References 528
Yifei Zhu, Jiawei Peng, Hong Liu, and Zhenggang Lan
Please visit the below companion site to view repositories with programs, data, instructions,
sample input and output files required for case studies as well as any post-publication updates.
https://www.elsevier.com/books-and-journals/book-companion/9780323900492
ix
This page intentionally left blank
Contributors
xi
xii Contributors
Yi-Fan Hou State Key Laboratory of Physical Hong Liu South China Normal University,
Chemistry of Solid Surfaces, Fujian Provincial Guangzhou, China
Key Laboratory of Theoretical and Zhi-Pan Liu Collaborative Innovation Center of
Computational Chemistry, Department of Chemistry for Energy Material, Shanghai Key
Chemistry, and College of Chemistry and Laboratory of Molecular Catalysis and
Chemical Engineering, Xiamen University, Innovative Materials, Key Laboratory of
Xiamen, Fujian, China Computational Physical Science (Ministry of
Education), Department of Chemistry, Fudan
Eugen Hruska Department of Chemistry,
University; Shanghai Qi Zhi Institute,
Emory University, Atlanta, GA, United States
Shanghai, China
Manabu Ihara Department of Chemical Science
Sergei Manzhos Department of Chemical
and Engineering, Tokyo Institute of
Science and Engineering, Tokyo Institute of
Technology, Tokyo, Japan
Technology, Tokyo, Japan
Bin Jiang Hefei National Research Center for
Philipp Marquetand Institute of Theoretical
Physical Sciences at the Microscale,
Chemistry, Faculty of Chemistry, University
Department of Chemical Physics, University
of Vienna, Vienna, Austria
of Science and Technology of China, Hefei,
Anhui, China Jiawei Peng South China Normal University,
Guangzhou, China
Hong Jiang Beijing National Laboratory for
Molecular Sciences, Institute of Theoretical Max Pinheiro Jr Aix Marseille University,
and Computational Chemistry, College of CNRS, ICR, Marseille, France
Chemistry and Molecular Engineering, Peking Aatish Pradhan Department of Chemical and
University, Beijing, China Biological Engineering, University at Buffalo,
The State University of New York, Buffalo,
Jun Jiang Hefei National Research Center for
NY, United States
Physical Sciences at the Microscale,
Department of Chemical Physics, University Jan Řezác Institute of Organic Chemistry and
of Science and Technology of China, Hefei, Biochemistry of the Czech Academy of
Anhui, China Sciences, Prague, Czech Republic
Grier M. Jones Department of Chemistry, P.D.Varuna S. Pathirage Department of
University of Tennessee, Knoxville, TN, Chemistry, University of Tennessee,
United States Knoxville, TN, United States
Alexei A. Kananenka Department of Physics Cheng Shang Collaborative Innovation Center
and Astronomy, University of Delaware, of Chemistry for Energy Material, Shanghai
Newark, DE, United States Key Laboratory of Molecular Catalysis and
Innovative Materials, Key Laboratory of
Julien Lam CEMES, CNRS and Universite de
Computational Physical Science (Ministry of
Toulouse, Toulouse Cedex, France
Contributors xiii
Education), Department of Chemistry, Fudan University of Warwick, Coventry, United
University; Shanghai Qi Zhi Institute, Kingdom
Shanghai, China Jiang Wu Hong Kong Quantum AI Lab and
Aditya Sonpal Department of Chemical and Department of Chemistry, The University of
Biological Engineering, University at Buffalo, Hong Kong, Pokfulam, Hong Kong
The State University of New York, Buffalo, Xun Wu The State Key Laboratory of Physical
NY, United States Chemistry of Solid Surfaces, Fujian Provincial
Peifeng Su The State Key Laboratory of Key Laboratory of Theoretical and
Physical Chemistry of Solid Surfaces, Fujian Computational Chemistry, and College of
Provincial Key Laboratory of Theoretical and Chemistry and Chemical Engineering,
Computational Chemistry, and College of Xiamen University, Xiamen, Fujian, China
Chemistry and Chemical Engineering, Bao-Xin Xue State Key Laboratory of Physical
Xiamen University, Xiamen, Fujian, China Chemistry of Solid Surfaces, Fujian Provincial
Huai-Yang Sun Beijing National Laboratory for Key Laboratory of Theoretical and
Molecular Sciences, Institute of Theoretical and Computational Chemistry, Department of
Computational Chemistry, College of Chemistry, and College of Chemistry and
Chemistry and Molecular Engineering, Peking Chemical Engineering, Xiamen University,
University, Beijing, China Xiamen, Fujian, China
Gauthier Tallec Institut des systèmes Jinzhe Zeng Shanghai Engineering
intelligents et de robotique (ISIR), Sorbonne Research Center of Molecular Therapeutics &
Universite, Paris, France New Drug Development, School of
Arif Ullah State Key Laboratory of Physical Chemistry and Molecular Engineering,
Chemistry of Solid Surfaces, Fujian Provincial East China Normal University; NYU-ECNU
Key Laboratory of Theoretical and Center for Computational Chemistry at NYU
Computational Chemistry, Department of Shanghai, Shanghai, China
Chemistry, and College of Chemistry and Lina Zhang State Key Laboratory of Physical
Chemical Engineering, Xiamen University, Chemistry of Solid Surfaces, Fujian Provincial
Xiamen, Fujian, China Key Laboratory of Theoretical and
Gaurav Vishwakarma Department of Chemical Computational Chemistry, Department of
and Biological Engineering, University at Chemistry, and College of Chemistry and
Buffalo, The State University of New York, Chemical Engineering, Xiamen University,
Buffalo, NY, United States Xiamen, Fujian, China
Konstantinos D. Vogiatzis Department of Yaolong Zhang Hefei National Research Center
Chemistry, University of Tennessee, for Physical Sciences at the Microscale,
Knoxville, TN, United States Department of Chemical Physics, University
of Science and Technology of China, Hefei,
Jingchun Wang Department of Chemical
Anhui, China
Physics, University of Science and Technology
of China, Hefei, China Yi Zhao State Key Laboratory of Physical
Chemistry of Solid Surfaces, Fujian Provincial
Shuai Wang State Key Laboratory for Physical
Key Laboratory of Theoretical and
Chemistry of Solid Surfaces, Fujian Provincial
Computational Chemistry, College of
Key Laboratory of Theoretical and
Chemistry and Chemical Engineering,
Computational Chemistry, and College of
Xiamen University, Xiamen, China
Chemistry and Chemical Engineering,
Xiamen University, Xiamen, China Xiao Zheng Department of Chemical Physics,
University of Science and Technology of
Julia Westermayr Theoretical Surface
China, Hefei, China
Chemistry Group, Department of Chemistry,
xiv Contributors
Xinxin Zhong Collaborative Innovation Center Molecular Engineering, East China Normal
for Advanced Organic Chemical Materials University; NYU-ECNU Center for
Co-constructed by the Province and Ministry, Computational Chemistry at NYU Shanghai,
Ministry of Education Key Laboratory for the Shanghai, China
Synthesis and Application of Organic Yifei Zhu South China Normal University,
Functional Molecules, Hubei University, Guangzhou, China
Wuhan, China
Tetiana Zubatiuk Department of Chemistry,
Tong Zhu Shanghai Engineering Research Carnegie Mellon University, Pittsburgh, PA,
Center of Molecular Therapeutics & New United States
Drug Development, School of Chemistry and
Preface
Quantum chemistry is a very mature field Part 1 of this book is dedicated to introduc-
of science—its method development started ing the minimum of fundamental topics in
as early as 1926, when the Schr€ odinger both quantum chemistry and machine learn-
equation was proposed and, a year later, ap- ing, which are indispensable to lay the foun-
plied to solving the first molecular system dation for the other parts; thus, Part 1 with 10
H2. Nevertheless, the quick and astonishing chapters is the largest in the book. Quantum
progress in another discipline—artificial in- chemistry topics start with a brief introduc-
telligence and machine learning—has had a tion to quantum chemistry in Chapter 1,
big impact on how we perform quantum followed by density functional theory in
chemical simulations. Machine learning in Chapter 2, semiempirical quantum mechani-
quantum chemistry has emerged as a vi- cal methods in Chapter 3, and a brief dis-
brant and ever-changing field rich in many course dealing with quantum chemical
methodological advantages and has rapidly calculations of systems from molecules to
become a commonly used technique in com- solid-state materials in Chapter 4. Chapter 5
putational chemistry. This calls for a book covers an important introduction to dynam-
that can be used as a textbook in advanced ics simulations. Machine learning is intro-
courses on theoretical and computational duced in Chapter 6 followed by chapters
chemistry as well as a reference material covering specific key topics such as
for specialists and nonspecialists alike. unsupervised learning in Chapter 7, neural
Teaching machine learning methods in networks in Chapter 8, kernel methods in
quantum chemistry has become a must in Chapter 9, and Bayesian inference in
the university-level courses on theoretical Chapter 10.
and computational chemistry as I have ar- Part 2 is dedicated to one of the most
gued in my Perspective article with the same abundant applications of machine learning
title as this book in early 2020 [1]. I have when it is used as a surrogate for quantum
started to teach the very basics of machine chemical potentials; by potential we mean
learning topics myself in the spring semester the atomistic system total energy as a func-
of 2021 for undergraduate chemistry stu- tion of nuclear coordinates. Chapter 11
dents at Xiamen University. My lecture notes covers potentials based on linear models
turned out to be very useful when contribut- followed by Chapters 12 and 13 introducing
ing to some of the chapters. The aforemen- neural network potentials and kernel
tioned Perspective has also pretty much method potentials, respectively. Chapter 14
formed the blueprint for this book, largely covers the important topic of creating train-
defining the scope of the topics covered, ing data for such potentials via active learn-
and the reader may refer to this Perspective ing. Chapter 15 focuses on the challenges in
for a very concise, bird’s-eye view of ma- creating machine learning models for
chine learning in quantum chemistry. performing excited-state dynamics. Machine
xv
xvi Preface
learning also helps in vibrational spectros- studies section. The case studies in many
copy as discussed in Chapter 16 and in accel- chapters revolve around one of the simplest
erating molecular structure optimizations chemical systems, the H2 molecule, which is
with quantum chemical methods as shown intentionally chosen to demonstrate the basic
in Chapter 17. ideas of various methods to ensure that they
Part 3 provides a taste of how machine are well understood before moving on to
learning can be used to learn other important more complex examples. Many chapters
quantum chemical properties such as elec- also contain examples from modern studies
tron densities (Chapter 18), dipole moments, so that the reader can better understand
and polarizabilities (Chapter 19), and more advanced concepts. The reader is al-
excited-state properties (Chapter 20). ways encouraged to go beyond the instruc-
Part 4 describes how machine learning be- tions in the case studies and experiment on
comes an integral part of quantum chemical their own.
simulations by improving quantum chemical This broadness of the topics could not
methods, e.g., via general approaches such be covered by a single person within a
as Δ-learning (Chapter 21, which also covers reasonably short amount of time adequate
related topics of learning from multiple quan- for such a fast-paced field, especially as
tum chemical methods such as transfer learn- many colleges have already introduced
ing, co-kriging, hierarchical machine learning, machine learning-related topics for chem-
and their combinations) or dedicated ap- ists and we have needed a textbook pref-
proaches for accelerating coupled-cluster erably yesterday. Thus, this book is a
and perturbation theory quantum chemical product of a massive effort by 65 authors
methods (Chapter 22), redesigning density from very different backgrounds ranging
functional theory methods (Chapter 23), and from mathematics to physics, materials
improving semiempirical quantum chemical science, and chemistry. Many of the au-
methods (Chapter 24). Part 4 concludes thors are among the best specialists in
by overviewing how machine learning can the topics they covered, which is evident
learn the essence of quantum chemistry— from the fact that many of them were
wavefunction—in Chapter 25. speakers at the first International Sympo-
Part 5 concludes the book with examples sium on Machine Learning in Quantum
of analysis of big data in a quantum chemical Chemistry that was held online from No-
context, showing specific examples for ana- vember 12 to 14, 2021.
lyzing nonadiabatic molecular dynamics This is only the tip of the iceberg, as many
trajectories (Chapter 26) and searching for more people were peer reviewing and all of
organic materials with tailored optical prop- my group members (Lina Zhang, Arif Ullah,
erties in the wide sea of possible compounds Bao-Xin Xue, Yi-Fan Hou, Fuchun Ge,
(Chapter 27). Peikun Zheng, Shuang Zhang, Wudi Yang,
One of the main motivations of this book Ran Wang, Linqiang Wei) were heavily in-
has been to offer purely theoretical knowl- volved in creating this book by becoming
edge alongside practical knowledge by pro- co-authors, additional peer reviewers, and
viding mathematical expositions in the testers for the case studies. The publisher’s
Methods section and incorporating computa- team (Anneka Hess, Devlin Person, Jai Marie
tional and, sometimes, pen-and-paper exer- Jose, Howell Angelo M. De Ramos, Paul
cises at the end of each chapter in the Case Prasad Chandramohan, Indhumathi Mani,
Preface xvii
and many more behind the scenes), were in- and publisher’s team for their efforts that
strumental to accomplish this project from made this book possible.
initiation to finalizing and supporting it all Happy reading and learning!
the way through. Just to give a short perspec-
tive on the amount of effort put into this Pavlo O. Dral
book: from the suggestion to write the book Xiamen, China
on May 13, 2020 (by Anneka Hess from
Elsevier) until writing this Preface, my email
folder for this project contains around 2000
emails; this is not including all the discus-
sions via messaging apps and meetings to co- Reference
ordinate our efforts. I cordially thank all the
[1] P.O. Dral, Quantum chemistry in the age of machine
authors, peer reviewers, my group members,
learning, J. Phys. Chem. Lett. 11 (2020) 2336–2347.
This page intentionally left blank
P A R T 1
Introduction
This page intentionally left blank
C H A P T E R
1
Very brief introduction to quantum
chemistry
Xun Wu and Peifeng Su
The State Key Laboratory of Physical Chemistry of Solid Surfaces, Fujian Provincial Key
Laboratory of Theoretical and Computational Chemistry, and College of Chemistry and Chemical
Engineering, Xiamen University, Xiamen, Fujian, China
Abstract
This chapter provides a brief introduction to quantum chemistry including the basic concepts and approxima-
tions of quantum mechanics, and a series of quantum chemical methods.
Quantum chemistry (QC) refers to the application of quantum mechanics to study the
chemical and physical properties of atoms, molecules, and materials. It is generally accepted
that the first QC calculation was performed by the German physicists Walter Heitler and Fritz
London on the hydrogen (H2) molecule in 1927. Now, QC has become an irreplaceable tool for
chemists to explore fundamental scientific questions in chemistry and related fields, [1–4]
with numerous applications to atomistic systems such as atoms, molecules or ionic solids.
This chapter introduces basic knowledge of quantum mechanics and QC methods for solving
the Schr€odinger equation.
Quantum Chemistry in the Age of Machine Learning 3 Copyright # 2023 Elsevier Inc. All rights reserved.
https://doi.org/10.1016/B978-0-323-90049-2.00006-8
4 1. Very brief introduction to quantum chemistry
function Ψ(r1, r2…rn) is a function of the positions of all the particles in the system, describing
a time-independent quantum mechanical system, in which r1, r2…rn are the positions of N
particles. The square of the wave function, jΨ(r1, r2…rn)j2, represents the probability of find-
ing N particles in a given place.
Every observable physical quantity, such as position, energy, or momentum, is described
Ð Ð ∗
by an operator. A Hermitian operator A ^ satisfies f ∗ Agdτ
^ ¼ g Af ^ dτ, where f and g are
well-behaved functions. Measurement shows a prediction as a result of an operator acting
^ Hamiltonian.
on a wave function. The most important operator is the total energy operator Hor
The measurement of the Hamiltonian operator is described by an eigen-equation called the
time-independent Schr€ odinger equation. Solutions for the time-independent Schr€ odinger
equation only exist for certain values of energy (energy is quantized in bound states). The
resulting eigenvalue equation is:
^ ðr1 , r2 ⋯rn Þ ¼ EΨðr1 , r2 ⋯rn Þ
HΨ (1)
In Eq. (1), E is the eigenvalue, which is the energy of the system described by Ψ(r1, r2…rn).
Moreover, the wave function must generally depend on time to reproduce information about
the system’s evolution over time. The time-dependent wave function, expressed as Ψ(r1, r2…rn, t),
satisfying the time-dependent Schr€
odinger equation:
∂Ψðr1 , r2 ⋯rn , tÞ ^
iħ ¼ HΨðr1 , r2 ⋯rn , tÞ (2)
∂t
Born–Oppenheimer approximation
The solution of the Schr€
odinger equation for molecules systems is the primary task of QC.
However, an exact solution cannot be obtained for multi-electron systems. For a multi-
electron system (atom or molecule), the time-independent Hamiltonian in atomic units is:
X
N elec
1 2
NX
nuclei
1 2
NX X
nuclei N elec
1 X
N X
elec N elec
1
NX X
nuclei N nuclei
Za Zb
^ ¼
H ri ra + +
2 2 jRa ri j rj ri jRa Rb j
i a a i i j>i α b>a
(3)
The Hamiltonian in Eq. (3) contains the electron kinetic operators (the first term), nuclear
kinetic operators (the second term), nucleus-electron attractive potential operators (the third
term), electron–electron repulsion (the fourth term), and nucleus-nucleus repulsion (the final
term). The Born–Oppenheimer (BO) approximation is introduced to simplify the Hamilto-
nian. In the BO approximation, the coupling between the nuclear r2 and the electronic wave
function is neglected. In most cases, the coupling between two electronic states is not strong,
i.e., the energy difference between two electronic states is big enough so that a small displace-
ment of nuclei will not change the electronic state, making the BO approximation valid, which
allows the separation of electrons (light particles) from nuclei (heavy particles) on time scales,
resulting in a considerable simplification.
According to this approximation, the total Hamiltonian operator can be expressed as the
sum of electronic Hamiltonian and nuclear Hamiltonian. The electronic Hamiltonian is writ-
ten as:
Introduction—The foundations of quantum chemistry 5
X
N elec NX X
nuclei N elec X
N X
elec N elec
^ el ¼ 1
H r2i
1
+
1
(4)
2 jRa ri j rj ri
i¼1 a i i j>i
Unless otherwise specified, the Hamiltonian used in this chapter refers to the electronic
Hamiltonian of Eq. (4).
According to the Pauli principle, the wave function of a multi-electron atom or molecule
should be antisymmetric with respect to the exchange of two electrons. Antisymmetrization
leads to Fermi holes, preventing the close approach of the electrons with the same spin and
reducing the repulsion energy between them. Thus, the many-electron function can be ap-
proximately written as the antisymmetrized product of N one-electron functions in the form
of a Slater determinant (SD):
ϕ1 ðr1 Þ ϕ2 ðr1 Þ ⋯ ϕN ðr1 Þ
1 ϕ1 ðr2 Þ ϕ2 ðr2 Þ ⋯ ϕN ðr2 Þ
ψ ðr1 , r2 ⋯, rN Þ ¼ pffiffiffiffiffiffi (5)
N! ⋮ ⋮ ⋯ ⋮
ϕ ðr Þ ϕ ðr Þ ⋯ ϕ ðr Þ
1 N 2 N N N
Variational principle
Given a system whose Hamiltonian is time-independent and whose lowest-energy eigen-
values is E1, if ϕ is any normalized, well-behaved function of coordinates of the system’s par-
ticles that satisfies the boundary conditions, then
ð
^
ϕ∗ Hϕdτ E1 (6)
If the trial function has a minimum of energy with respect to variational coefficient c, we
can find it by
∂ ∂ D ^ E
h Ei ¼ ϕHϕ ¼0 (7)
∂c ∂c
The lower energy, the better the trial wave function. And if the trial function is accurate
enough, the function we get will be close enough to the true ground-state wave function.
To improve the quality of trial wave function, the linear variational principle is applied,
by which the trial wave function is constructed by a linear combination of n linearly indepen-
dent functions (χ 1, χ 2, ⋯ χ n):
ϕ ¼ c1 χ 1 + c2 χ 2 +⋯+cn χ n (8)
6 1. Very brief introduction to quantum chemistry
Perturbation theory
Perturbation theory provides a procedure for finding approximate solutions to the
Schr€odinger equation for a system that differs only slightly from a system for which the
solutions are known. This method is frequently employed to estimate the functions or values
based on partial knowledge about the solutions of the investigated problem.
odinger equation for the Hamiltonian operator H
If the Schr€ ^
^ n ¼ En ψ n
Hψ (9)
ð0Þ
^ :
cannot be solved exactly, but we know the exact solutions of the Hamiltonian operator H
^ ð0Þ ψ ð0Þ ¼ Eð0Þ ψ ð0Þ
H (10)
n n n
where the superscript (0) denotes zero-order, which means that the wave function and energy
are already known, the subscript n denotes the nth electronic state.
In perturbation theory, we assume that H^ may be expanded as:
^ ð0Þ + H
^ ¼H
H ^0 (11)
0
The operator H^ is called the perturbation. Since the operator H
^ differs only slightly from
ð0Þ
^ , the eigenfunctions and eigenvalues of H
H ^ will not differ significantly from those of the
ð0Þ
unperturbed Hamiltonian operator H ^ .
If we suggest that the wave function and energy can be expressed as the linear expansion
according to λ, which can be written as
ψ n ¼ ψ ðn0Þ + λψ ðn1Þ + λ2 ψ ðn2Þ +⋯ + λk ψ ðnkÞ +⋯ (12)
The first-order wave function correction ψ (1) n is orthogonal to the unperturbed wave
function.
D E
ð0Þ ^ 0 ð0Þ
X ψ m H ψ n
ψ ðn1Þ ¼ ð0Þ ð0Þ
ψ ðm0Þ (16)
m6¼n E n E m
It can be found that ψ (1) n can be expressed as the linear combination of m (m 6¼ n) electronic
th
th th
states. The smaller
D the energy
E difference between m state and n state is, or the larger the
ð0Þ ^ 0 ð0Þ
coupling term ψ m H ψ n is, the larger the combination coefficient of the mth electronic
state wave function could be.
Second, if the zero-order wave function is degenerate, any arbitrary combination of the de-
generate wave functions is the eigenfunction of H^ ð0Þ. However, introducing the perturbation
0
operator H ^ will completely or partially remove the degeneracy. Thus, the most important
issue is to derive the combination coefficients of the degenerate wave function. It is called
a degenerate perturbative theory, which can be seen from the textbook.
Hartree–Fock method
The Hartree–Fock (HF) method seeks to approximately solve the electronic Schr€ odinger
equation with the assumption that the wave function can be approximated by a single SD
made up of one set of spin orbitals. The HF method is a wave function-based approach based
on the independent-particle model, which reduces the many-electron problem to indepen-
dent single-electron problems. It means that the interactions between the particles are approx-
imated by taking all interactions into account in an average fashion.
The HF method is derived from the variational principle, which is applied to an energy
functional that includes Lagrange multipliers for orthonormal constraints and defines the
HF matrix for each orbital. First, by using the Condon–Slater rules, the energy of a single de-
terminant can be expressed as:
X n
n
1X n X
E¼ hii + J ij Kij (17)
i¼1
2 i¼1 j¼1
where Jij and Kij are Coulomb integral and exchange integral between spin orbitals i and j,
respectively:
ðð
1
J ij ¼ hijjiji ¼ dτ1 dτ2 ϕ∗i ð1Þϕ∗j ð2Þ ϕi ð1Þϕj ð2Þ (18)
r12
Methods of molecular electronic structure computations 9
ðð
1
Kij ¼ hijjjii ¼ dτ1 dτ2 ϕ∗i ð1Þϕ∗j ð2Þ ϕ ð1Þϕi ð2Þ (19)
r12 j
hii is the one-electron integral
D E ð
hii ¼ i^hð1Þi ¼ dτ1 ϕ∗i ð1Þ^hð1Þϕi ð1Þ (20)
where ^hð1Þ is the one-electron operator including electron kinetic operator and electron-
nucleus potential operator shown in Eq. (3).
The objective now is to determine a set of MOs that makes the energy a minimum or at least
stationary with respect to a change in the orbitals. The variation must be carried out to ensure
that MOs remain orthogonal and normalized. Since E varies subjected to a constraint that the
spin orbitals are orthonormal,
hijji δij ¼ 0 (21)
where hi j ji is the Dirac notation of the inner product of two spin orbitals,
ð
hijji ¼ dτϕ∗i ϕj (22)
X 1 X X XX
δhii + δJ ij δKij εij δhijji ¼ 0 (25)
i
2 i j i j
where εij refer to the Lagrange multiplier matrix elements, and the notation δ indicates the
variation of the functional.
The variation of the energy leads to the equation below:
2 3
Xð X X
δϕ∗i ð1Þ4^
hð1Þϕi ð1Þ + ^ j ð 1Þ ϕ i ð 1Þ
^J j ð1Þ K εij ϕj ð1Þ5dτ1 + c:c: ¼ 0 (26)
i j j
In Eq. (26), c.c. is the complex conjugate of the summation ahead. δϕi indicates the variation
of the spin orbital ϕi (ϕi ! ϕi + δϕi). ^J ð1Þ and K
^ ð1Þ operators are connected with the potential
energy of electron 1 with all other electrons in the system. They are defined via the action on
any function of the position of electron 1:
10 1. Very brief introduction to quantum chemistry
X
^J ð1Þ ¼ ^J j ð1Þ
j
ð (27)
^J j ð1Þuð1Þ ¼ dτ2 1 ϕ∗ ð2Þϕj ð2Þuð1Þ
r12 j
X
^ ð1Þ ¼
K ^ j ð 1Þ
K
j
ð (28)
^ j ð1Þuð1Þ ¼ dτ2 1 ϕ∗ ð2Þuð2Þϕj ð1Þ
K
r12 j
According to Eq. (26), we have
X
^
hð1Þ + ^J ð1Þ K
^ ð 1Þ ϕ i ð 1Þ ¼ εij ϕj ð1Þ (29)
j
^
hð1Þ + ^J ð1Þ K
^ ð1Þ is defined as the Fock operator, which is an effective one-electron energy
operator. It describes the kinetic energy of an electron and the attraction to all the nuclei
and the repulsion to all the other electrons.
Eq. (29) may be simplified by choosing a unitary transformation that makes the matrix of
Lagrange multipliers diagonal, that is εij ¼ 0. This particular set of MOs is called canonical
MOs and thus transforms Eq. (29) into a set of pseudo-eigenvalue equations:
^ð1Þϕi ð1Þ ¼ εi ϕi ð1Þ
F (30)
where εi is the orbital energy of the Hartree–Fock orbital.
D E X
εi ¼ i^ hð1Þi + ½hijjiji hijjjii (31)
j
The HF equations form a set of pseudo-eigenvalue equations as the Fock operator depends
on all the occupied MOs (via the Coulomb and exchange operators). A specific Fock operator
can only be determined if all the other occupied orbitals are known, and therefore the self-
consistent-field (SCF) procedure must be employed to solve the problem.
The HF wave function is not unique. The total wave function is unchanged by a unitary
transformation of the occupied MOs among themselves (rows and columns in a determinant
can be added and subtracted without affecting the determinant itself ). The canonical MOs
obtained from the Fock equations can be transformed into localized orbitals with various lo-
calization schemes.
The total HF energy of a molecule is expressed as:
X
n
1Xn X
n
EHF ¼ hii + ½hijjiji hijjjii + V NN (32)
i
2 i j
where VNN is nuclear-nuclear repulsion energy and is not related to wave function. For an
atom, VNN is zero. Considering Eqs. (31) and (32), it is noticed that the total energy is not sim-
ply a sum of MO orbital energies.
An electron at r is viewed as interacting with an “averaged” charge density, “mean-field”,
arising from the N 1 remaining electrons. Originally, the HF calculations were performed
using numerical methods to solve the HF differential equations, which are complex for
Another random document with
no related content on Scribd:
Assistiu-lhe o piedoso bispo até o ultimo valle, e logo seu corpo foi
levado por homens principaes ao Hospicio de Nossa Senhora da
Penha dos Capuchinhos Francezes, o dia em que chegavam as
novas da restauração do famoso Palmar a Pernambuco, que havia
de ser o sexto da victoria, pois tanto gasta um caminheiro apressado
de um logar a outro. Mas é em vão busca-lo em Pitta, auctor
moderno que d’isto tracta como si não tractára. E mais me
escandalisa que passasse em sua mesma patria por um poeta de tal
nome seu contemporaneo, com quem devia gastar parte d’aquelles
elogios. Morreu finalmente no anno de 1696 com edade de setenta
e tres annos.
Este é o mais abreviado resumo que posso dar da vida do meu
suspirado, quão dilectissimo poeta lyrico; e oxalá podéra eu publicar
os prodigiosos fundamentos do meu amor, derramando entre as
gentes o manancial thesouro de suas graças! Singular foi a estrella
que dominou em seu engenho; porque a toda a circumferencia das
luzes apolineas brilhou com egualdade senhoril; e não menos
prodigioso aquelle não sei que de sua guarda, porque offendendo ás
claras muitas pessoas, de quem o menor movimento seria sem
duvida uma tyranna morte, sempre se atreveu, e nunca de seu motu
proprio cautelou perigos; morrendo intacto de tão prolongados
mezes.
Muitos eram os feridos do seu ferro que consultaram o remedio no
mesmo instrumento da chaga, beijando a Achilles a lança que os
traspassára. Raro testimunho d’esta fatalidade foi a resposta que
deu a um queixoso certo governador severamente resoluto: «Não
faça V. Mᶜᵉ. caso (disse), porque isso tambem passa por mim, sem
que por mim passe a minima tenção de o castigar.»
Testimunho d’esta fatalidade são as duas quartas de um soneto,
que se fez em sua morte; o qual não escrevo por inteiro em razão
de que si os seus principios professam a verdade pura, os fins
todavia contém temeraria petulancia:
o quente da cama
Com Gonçalo, e com sua ama,
Dizendo estava comei-me, &.
Por vida do meu Gonçalo,
Custodia formosa e linda, &.
Madrasta do Gonçalinho,
Que é lindo enteado a fé, &.
Sim, por vida de Gonçalo, &.
Mas por vida de Gonçalo, &.
GLOSA.
AOS VICIOS
TERCETOS