Molecular Dynamics

Ruben Santamaria
Molecular
Dynamics
Molecular Dynamics
Ruben Santamaria
Molecular Dynamics
Ruben Santamaria
Department of Theoretical Physics
Physics Institute, University of Mexico
(UNAM)
Mexico City, Mexico
ISBN 978-3-031-37041-0 ISBN 978-3-031-37042-7 (eBook)

https://doi.org/10.1007/978-3-031-37042-7
© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland
AG 2023
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse
of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Paper in this product is recyclable.

With Love To
Abril
Andrea
Mariana
I would like to express my gratitude for the
motivation given by my parents, brother, and
sisters, as well as the guidance and
knowledge provided by my thesis supervisors,
teachers, and colleagues. Additionally, I
appreciate the experiences gained from
working with my thesis students and friends.
I sincerely thank each and every one of them
from the bottom of my heart.
Preface
Molecular dynamics is a broad scientific discipline that covers multiple fields, from
physics to biology. This is because molecular dynamics can effectively describe
the workings of nature from the atomic perspective. The book’s organized structure
provides consistent and intelligible explanations, making it easy for readers with
diverse backgrounds to understand and make sense of molecular dynamics. The
book is self-contained, covering the fundamental concepts of mechanics, while also
catering to experts by including the required mathematical formalism. Hence, the
book is suitable for both non-experts and experts alike.
The book covers a range of topics, from classical mechanics to quantum mechan-
ics, and from few-body systems to many-body systems. It is particularly useful for
scientists studying the connection between small and large particle systems, whether
in a state of equilibrium or not. The book offers discussions at various levels of
theory, allowing for the exploration of complex chemical reactions through first-
principles molecular dynamics, while relaxed levels of theory permit the simulation
of atomic interactions with model potentials. The book is valuable for academics
conducting research at the intersection of molecular dynamics, thermodynamics,
and fluid dynamics.
The book is designed to be educational for both teachers and students, with
figures, illustrations, problems, solutions, and suggested readings included. It covers
the latest developments in molecular dynamics and is a valuable reference for
researchers and experts in various scientific fields who share an interest in exploring
and characterizing phenomena at the atomic level. The book’s discussions make
it suitable for researchers and students in physics, chemistry, biology, and inter-
disciplinary subjects like nanoscience, materials engineering, molecular genetics,
and more. Additionally, the book provides the formal and necessary foundations to
encourage readers to expand their knowledge of the microscopic world. By learning
and utilizing the theory of molecular dynamics, dedicated researchers can make
exciting new discoveries in their fields.
Mexico City, Mexico Ruben Santamaria
vii
Prologue
Molecular dynamics embodies the laws of physics, and thus, allows the study
of atomic-level processes. By understanding how microscopic attributes of nature
lead to macroscopic properties of matter, we can design new compounds and
improve processes. Molecular dynamics enable us to achieve this understanding.
The ultimate goal of molecular dynamics is to use natural laws to generate new
knowledge, create innovative technologies, and enhance the quality of life.
This book covers the fundamentals of molecular dynamics, integrated by formal
concepts, diverse mechanical variables, and dynamic expressions. Due to the
intricate nature of atomic-level processes, various levels of theory are unfolded to
investigate and characterize the complex world of atoms. Molecular dynamics is a
rapidly expanding field, thanks to both the scientific advancements in atomic-level
research and the swift progress of computer technology.
The book is constituted by five main parts, which are integrated by one or more
chapters:
• Basics of Classical Mechanics

• Basics of Quantum Mechanics
• First-Principles Molecular Dynamics
• Classical Molecular Dynamics
• Time Evolution Operators
Basics of Classical Mechanics
This part of the book begins by discussing the basic principles of classical mechan-
ics, as outlined in Newton’s theory. This is because Newtonian mechanics provides
a foundation for understanding particle motion and the mechanical variables that
are used to describe particle systems. When combined with model interaction
potentials, this approach to molecular dynamics becomes very versatile. It can be
applied to systems with few or many particles, regardless of whether they are in
thermodynamic equilibrium or not. However, when dealing with large numbers of
particles, it is important to pay attention to the collective behavior of the system
and to use statistical methods to analyze it. By doing so, it is possible to establish a
connection between molecular dynamics and statistical mechanics, and to compare
ix
x Prologue
the ensemble averages of the latter with the time averages of the former. In this
context, it is important to understand how microscopic and macroscopic properties
of matter are related. One way to do this is by using different types of mechanical
descriptors, which help to classify and distinguish the state of a system from a
dynamic point of view. The tools to describe macroscopic bodies are conveyed in the
rigid body approximation. The last chapter introduces the Lagrangian, Hamiltonian,
and Hamilton-Jacobi theories to not only provide the elegant formalisms and the
different paths in the formulation of the classical equations of motion, but also reach
a point where it is possible to create novel approaches in molecular dynamics. The
connections among all those classical theories are briefly analyzed.
Basics of Quantum Mechanics
This part of the book is an introduction to quantum theory. It explains the

fundamental concepts and equations of quantum mechanics that are required to
understand how atoms behave and are described at the quantum level. It covers
topics such as the dual nature of matter, the way energy is quantized, and the
behavior of microscopic particles with spin. The main equation used in quantum
molecular dynamics is the time-dependent wave equation, which is eventually
simplified into the time-independent wave equation.
First-Principles Molecular Dynamics
In this part of the book, we explore the time-dependent wave equations of electrons
and nuclei, which are essential for first-principles molecular dynamics simulations.
We use a self-consistent field approach and later achieve a stationary wave equation
for the electrons. To bridge the gap between quantum and classical molecular
dynamics, we discuss the mixed quantum-classical molecular dynamics scheme.
In this respect, we investigate the classical limit of nuclear motion and derive
Newtonian equations of motion for the nuclei. Through these chapters, we connect
quantum molecular dynamics to classical molecular dynamics, thus exploring
flexible approaches from rigorous approaches.
Classical Molecular Dynamics
In this part of the book, we focus on classical equations of motion for atoms and use
model functions for atomic interactions, avoiding the wave equation of electrons.
We discuss a variety of model interaction potentials, like those for biological and
non-biological molecules. These potentials simplify simulations, making it possible
to work with many-body systems since electron effects are parameterized. We
also dedicate a chapter to approximations that simulate the behavior of extended
systems like gases, fluids, and solids. These approximations are important for
Prologue xi
determining the dynamical and thermodynamic properties of compounds immersed

in an extended sample.
Time Evolution Operators
In this part of the book, we focus on how particle systems evolve over time. We
use the Poisson-bracket formalism to obtain classical time evolution operators,
which can be applied in the mixed quantum-classical molecular dynamics approach.
By calculating atomic forces on the fly, time integrators can be used to evolve
the positions and momenta of particles over time. These integrators have many
benefits: they are general, time-reversible, computationally efficient, and simplify
simulations by avoiding mathematical complications.
Contents
Part I Basics of Classical Mechanics

1 Principles of Classical Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1 Newtonian Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Space and Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.5 Electric Charge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.6 Reference System of Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.7 Newtonian Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.8 Linear Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.9 Angular Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.10 Descriptions Between Inertial Reference Frames . . . . . . . . . . . . . . . . . 19
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2 Foundations of Newtonian Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.1 First Newton’s Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.2 Second Newton’s Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.3 Third Newton’s Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.4 Reduced Mass of a Two-Particle System . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.5 Time Reversibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.6 Angular Momentum and Torque . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.7 Impulse, Work, and Power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.8 Kinetic and Potential Energies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.9 Energy Conservation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3 Many-Particle Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.1 Reference Frame of a Many-Particle System . . . . . . . . . . . . . . . . . . . . . . 41
3.2 Angular Momentum and Torque of a Many-Particle System . . . . . 43
3.3 Mechanical Energies of a Many-Particle System . . . . . . . . . . . . . . . . . 44
3.4 Transformation of the Energy Components . . . . . . . . . . . . . . . . . . . . . . . 47
3.5 Energy Balance Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.6 Statistical and Time Averages of Physical Observables . . . . . . . . . . . 50
3.7 Ergodic Hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
xiii
xiv Contents
3.8 Breaking the Ergodic Hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

3.9 Velocity Distribution Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.10 Temperature of a System of Particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.11 Temperature Scaling as a Thermostat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.12 Temperature Fluctuations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.13 Pressure and Volume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.14 The Virial and the Equation of State . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4 Mechanical Descriptors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.1 Caloric Curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.2 Interatomic Distance Fluctuations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.3 Root Mean Square Deviation of Positions . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.4 Orientational Order Parameter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.5 Probability Distribution Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.6 Correlation Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.7 Properties of Correlation Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.8 Vibrational Spectra from Autocorrelation Functions. . . . . . . . . . . . . . 81
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
5 Rigid Body . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.1 Angular Momentum of a Rotating System of Particles . . . . . . . . . . . 85
5.2 External Torques Acting on a Rotating Body . . . . . . . . . . . . . . . . . . . . . 89
5.3 Total Energy of a Rotating Rigid Body . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
A.1 Appendix: Matrix Diagonalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6 Analytical Mechanics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6.1 Action Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6.2 Principle of Stationary Action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
6.3 Classifying Molecular Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
6.4 Lagrange’s Equations of Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
6.5 Newtonian Equations of Motion from Lagrange Theory . . . . . . . . . 107
6.6 Non-uniqueness of the Lagrangian. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.7 Invariance of the Lagrange Equations of Motion. . . . . . . . . . . . . . . . . . 111
6.8 Motion with Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
6.9 Hamilton’s Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
6.10 Preservation of the Hamiltonian in Time . . . . . . . . . . . . . . . . . . . . . . . . . . 119
6.11 Conserved Observables and Symmetries . . . . . . . . . . . . . . . . . . . . . . . . . . 121
6.12 Space Homogeneity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
6.13 Space Isotropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
6.14 Uniform Passage of Time. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
6.15 Hamilton’s Equations of Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
6.16 Invariance Under Canonical Transformations . . . . . . . . . . . . . . . . . . . . . 129
6.17 Time Reversibility in Hamiltonian Theory . . . . . . . . . . . . . . . . . . . . . . . . 132
6.18 Hamilton-Jacobi Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
6.19 Illustrating with the Harmonic Oscillator . . . . . . . . . . . . . . . . . . . . . . . . . . 137
Contents xv
6.20 Contact Between Quantum and Classical Mechanics . . . . . . . . . . . . . 138

6.21 Poisson’s Brackets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
6.22 Classical Time Propagator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
Part II Basics of Quantum Mechanics

7 Wave-Particle Duality of Matter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
7.1 Young’s Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
7.2 Interference of Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
7.3 Photo-Electron Experiment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
7.4 Compton’s Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
7.5 Davisson-Germer’s Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
7.6 de Broglie’s Hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
7.7 Bohr’s Complementary Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
8 Quantization of the Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
8.1 Planck’s Energy Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
8.2 Blackbody Radiation Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
8.3 Rayleigh-Jeans Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
8.4 Wien’s Displacement Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
8.5 Ultraviolet Catastrophe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
8.6 Planck’s Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
8.7 Franck-Hertz Experiment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
8.8 Heisenberg’s Uncertainty Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
A.1 Appendix: Planck’s Radiation Intensity Law . . . . . . . . . . . . . . . . . . . . . . 170
A.1.1 Average Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
A.1.2 Density of States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
A.1.3 Energy per Volume per Frequency . . . . . . . . . . . . . . . . . . . . . . . 173
A.1.4 Irradiance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
A.1.5 Maximum Intensity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
A.1.6 Time Interval of Temperature Cooling . . . . . . . . . . . . . . . . . . . 177
Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
9 Quantization of the Angular Momentum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
9.1 Orbital Angular Momentum and Spin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
9.2 Characterizing a Particle with Spin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
9.3 Stern-Gerlach Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
9.4 Wave-Particle Duality and Spin of a Particle . . . . . . . . . . . . . . . . . . . . . . 185
9.5 Fermions and Bosons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
9.6 Pauli’s Exclusion Principle and Hund’s Rule . . . . . . . . . . . . . . . . . . . . . . 186
A.1 Appendix: Magnetic Moment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
A.1.1 Electric Current in a Circular Loop . . . . . . . . . . . . . . . . . . . . . . 188
A.1.2 Magnetic g Factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
A.1.3 Magnetic Energy and Magnetic Work. . . . . . . . . . . . . . . . . . . . 191
xvi Contents
A.1.4 Zeeman Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193

A.1.5 Electron Spin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
A.1.6 Paschen-Back Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
A.1.7 Applications of the Spin Resonance Technique . . . . . . . . . 199
B.1 Appendix: Spin Functions and Operators. . . . . . . . . . . . . . . . . . . . . . . . . . 200
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
10 Postulates of Quantum Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
10.1 Reformulating the Conceptual World . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
10.2 Postulates of Quantum Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
10.2.1 First Postulate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
10.2.2 Second Postulate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
10.2.3 Third Postulate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
10.2.4 Fourth Postulate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
10.2.5 Fifth Postulate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
10.2.6 Sixth Postulate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
10.3 Time Reversibility in Quantum Mechanics . . . . . . . . . . . . . . . . . . . . . . . . 215
10.4 Stationary States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
10.5 Superposition Principle of Quantum States. . . . . . . . . . . . . . . . . . . . . . . . 221
10.6 Bohr’s Correspondence Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
10.7 Selection Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
10.8 Pauli’s Principle in the Electronic Wave Function . . . . . . . . . . . . . . . . 228
10.9 Wave Function of the Electrons in a Molecule . . . . . . . . . . . . . . . . . . . . 229
10.10 Variational Principle of the Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
A.1 Appendix: Proposing the Wave Equation for Matter Waves . . . . . . 235
B.1 Appendix: Expansion of a Determinantal Wave Function . . . . . . . . 237
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
Part III First-Principles Molecular Dynamics

11 Dynamics of Electrons and Nuclei . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
11.1 The Electronic and Nuclear Dynamics Are Coupled. . . . . . . . . . . . . . 244
11.2 Molecular Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
11.3 Approximating the Total Wave Function . . . . . . . . . . . . . . . . . . . . . . . . . . 246
11.4 Time-Dependent Self-Consistent Field Equations . . . . . . . . . . . . . . . . 250
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
12 Classical Limit of the Nuclear Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
12.1 Polar Form of the Nuclear Wave Equation . . . . . . . . . . . . . . . . . . . . . . . . 256
12.2 Continuity and Hamilton-Jacobi Equations . . . . . . . . . . . . . . . . . . . . . . . 259
12.3 Conditions to Describe the Nuclear Particles Classically . . . . . . . . . 263
12.4 Simplification of the Nuclear Potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
12.5 Parameterizing the Potential Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
12.6 Total Energy of the Molecular System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
12.7 Establishing the Accuracy of Atomic Forces . . . . . . . . . . . . . . . . . . . . . . 277
12.8 Diffusion from the Continuity Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
Contents xvii
12.9 Diffusion Equation and Particle Flux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279

12.10 Expansion of the Electronic Wave Equation. . . . . . . . . . . . . . . . . . . . . . . 283
12.11 Expansion of the Newtonian Equation of the Nuclei. . . . . . . . . . . . . . 286
A.1 Appendix: Bohm’s Quantum Potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290
Part IV Classical Molecular Dynamics

13 Classical Molecular Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
13.1 Model Interaction Potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296
13.2 Forcefields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298
13.3 Atom Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
13.4 United Atom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 300
13.5 Bond Elongation and Compression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 300
13.6 Combination Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
13.7 Bond Angle Vibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
13.8 Plane Bending . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304
13.9 Angle Inversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
13.10 Torsional Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 306
13.11 Electrostatic Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
13.12 van der Waals Forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308
13.13 Interaction Potential Functions of Water. . . . . . . . . . . . . . . . . . . . . . . . . . . 311
13.14 Atom Polarization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312
13.15 External Fields and Potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314
13.16 Parameterization of Forcefields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315
13.17 Model Potentials of Non-biological Systems . . . . . . . . . . . . . . . . . . . . . . 317
13.18 Sutton-Chen Potential Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 318
13.19 Gupta Potential Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
13.20 Tersoff Potential Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 320
A.1 Appendix: Harmonic Model of the Dispersion Energy . . . . . . . . . . . 322
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325
14 Extended Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327
14.1 Fixed and Flexible Boundaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328
14.2 Periodic Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 330
14.3 The P BC System Is an Open System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
14.4 Electrostatics in the P BC Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 336
14.5 Ewald Sum Approach. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
14.6 Using the Poisson Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
14.7 Short-Range Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343
14.8 Electrostatic Self-Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344
14.9 Long-Range Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344
14.10 Ewald Electrostatic Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 348
14.11 Smooth Particle Mesh Ewald Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . 350
14.12 Shifted Potentials and Forces. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 356
xviii Contents
Part V Time Evolution Operators

15 Integrating the Equations of Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359
15.1 Liouville Operator as a Time Propagator . . . . . . . . . . . . . . . . . . . . . . . . . . 359
15.2 Discretizing the Time Propagator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361
15.3 Evolving Positions and Momenta . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363
15.4 Simplified Time Integrators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367
15.5 Leapfrog Algorithm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 368
15.6 Verlet Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 370
15.7 Bond Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 376
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379
Part I
Basics of Classical Mechanics
Principles of Classical Dynamics
1
Classical Mechanics is the theoretical framework of physics which describes the

mechanical and dynamical aspects of the classical particles and objects. This
chapter and the subsequent chapters present the fundamental concepts of classical
mechanics such as space, time, mass, and more and introduce the basic variables of
linear momentum, angular momentum, and torque from the Newtonian perspective.
The notion of time and time reversibility are established. The importance of a
reference system of coordinates and the transformation of the physical variables
between different inertial reference systems are shown.1
1.1 Newtonian Dynamics
Classical mechanics establishes the underlying concepts, mechanical elements,

and basic equations on the dynamics of particles. It is also given the name of
Newtonian dynamics in honor to Sir Isaac Newton, who formulated the laws of
classical mechanics. There are other formulations in classical mechanics, like those
proposed by Lagrange, Hamilton, Jacobi, Poisson, and others. Those treatments
were proposed after the conception of Newtonian dynamics and provide universal
concepts on the mechanical aspects of the particles, in addition to their own
particular equations of motion. The different formulations of classical mechanics
are in one way or another important tools in the analysis of particle systems, and
therefore, they all are of interest in molecular simulations.2
One of the main purposes of molecular dynamics is to compute the positions,
speeds, and forces at every instant of time of the particles. The Newtonian
theory applies to particles traveling with velocities sufficiently smaller than the
1 The discussion follows similar topics to those of: Alonso and Finn [1] and Halliday et al. [2].
2 A law in physics is a scientific proposal which tells the regular patterns in the behavior of physical
bodies and matter under precise conditions, by establishing the relations between different physical
variables. Once the law is formulated, it helps us predict the behavior of matter.
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 3

R. Santamaria, Molecular Dynamics, https://doi.org/10.1007/978-3-031-37042-7_1
4 1 Principles of Classical Dynamics
speed of light because, according to the theory of relativity, the bodies suffer
length contraction and time dilation at high speeds, which are features beyond
the consideration of Newtonian mechanics. The “exotic states” of matter and the
particles with very low or very high energies are excluded in Newtonian theory. In
similar form, the gravitational effects are out of the description because they are
negligible at the atomic level: the electric force, .fe , and the gravitational force, .fg ,
of a proton interacting with an electron separated by a distance of 1 Bohr give the
force ratio .fe ∼ 1039 fg . Another example illustrating the large differences in the
forces is the magnetic force. The magnetic force of a magnet is capable of attracting
a small metallic sphere, overcoming the gravitational force exerted by the entire
planet. In our classical approach, it is assumed that the particles and physical events
have character of “ordinary particles” and “ordinary events,” which are properly
described by Newtonian mechanics.
Finally, there are entities that are not defined in terms of more basic concepts,
variables, or objects in Newtonian theory. They are considered as fundamental
entities. The ones of interest in molecular dynamics are space, time, mass, energy,
and electric charge. It has been postulated that the fundamental entities emerged at
the first instances on the creation of the Universe, with relation to each other. Such
fundamental entities play important roles in molecular dynamics, and thus, they are
considered in our discussion.
1.2 Space and Time
Space and time are formally intertwined together into a single framework referred to
as spacetime. The spacetime concept is mathematically well accepted since it leads
to invariant metrics [3]. The spacetime is conceived as a set of four independent
dimensions. Three of them form the ordinary 3D space, where we are immersed
and can freely move, and the other dimension is time. We cannot freely move in
this last dimension. Time permeates the 3D space and never stands still. Time is
not necessarily regular and may be measured with time ticks at different rates.
According to the theory of relativity, time is not absolute and depends on the relative
motions of the observers. In consequence, simultaneous events are no longer so for
reference frames mainly moving with high relative speeds.
The four dimensions are neither perceptible nor distinguishable. The spacetime
may be considered either an intellectual artifact, with the purpose to construct
a conceptual structure of the Universe, or a real object, like a 4D limitless
fabric with indistinguishable regions, no discernible geometrical shape, and with
existence with no reference to any other object. The lack of perception of spacetime
demands measurements of relative type among the bodies. The above two points
(on conceiving the spacetime as a conceptual structure or a real structure) are in
contradiction. In spite of that, they have been used to establish the concepts and
formulations that explain natural phenomena, simultaneously allowing to reach a
deep understanding of the Universe.
1.3 Mass 5
1.3 Mass
In principle, the entities of the Universe may be classified into two classes: particles
of matter and particles of light, or photons. The particles of matter have mass.
Mass is not a fundamental property but an emergent feature of a subatomic (Higgs
quantum field) process due to the change in the energy landscape of vacuum. Yet
mass is of central importance. From a classical perspective, it is understood as the
property of a body to resist changes in its state of motion. Mass is a physically
measurable quantity, and therefore, is a physical observable.
According to the theory of general relativity, mass and light warp the 4D space-
time, Fig. 1.1. Mass and energy are equivalent, with the possibility of transforming
one onto another due to the high-energy physics expression:
. E = mc2 (1.1)
The sub-atomic experiments on particle production exemplify the transformation

of mass into energy and vice versa. A matter particle is considered an energy-
condensed structure. The mass is subjected to the passage of time, but the light
is not. The matter with mass travels slower than light, which has no rest mass. In
a closed system, the total energy of the particles (which also includes the energy
equivalent of the particle masses) is preserved with time. The mass immersed in a
gravitational field acquires weight, which is a type of force acting on the mass.
. w = mg (1.2)
The factor .g is the gravitational acceleration. In the absence of gravity, the object
has mass but not weight. A body isolated from any interaction maintains the initial
state of motion, but the state of motion changes under the action of a force, with an
acceleration proportional to the mass. A heavy mass is accelerated less rapidly than
a light mass due to the different inertias that the bodies with distinct masses present.
An inertial mass represents the resistance of the body to change its state of motion,
while the gravitational mass is the capacity of the body to curve the spacetime due
Fig. 1.1 According to the

theory of general relativity,
any mass bends the 4D space
to its gravitational field. In spite of the different conceptions, the inertial mass and
gravitational mass are equivalent.
The inertial mass is determined in relative terms, by making a couple of bodies
interact, assigning to one of them the unit mass, measuring the appropriate dynamic
variables of the experiment, and determining from the mathematical expressions
the unknown mass of the second body. On the other hand, the gravitational mass is
measured with a balance in the presence of a gravitational field. Let us consider a
falling body. If the body weight is .mg g with .mg the gravitational mass and .a the
acceleration of the inertial mass .mi , then we have:
. a = (mg /mi ) g (1.3)
When dealing with a different body whose gravitational mass is .m'g and its inertial
mass is .m'i , we also have .a = (m'g /m'i ) g. Under equal experimental conditions of
falling objects, the accelerations .a imprinted to the primed and unprimed objects are
observed to be the same. This leads to the free fall principle.
Statement 1.1 Free fall principle: all the bodies fall at the same rate.
Problem 1.1 If the force of gravity depends on the object mass, m, in the
form .f = GMplanet m/r 2 , show that two bodies with different masses fall by
the same rate.
Solution The gravitational force of .Mplanet exerted on body 1 is .f1 =

GMplanet m1 /r 2 , and on body 2 is .f2 = GMplanet m2 /r 2 . The acceleration
of .m1 and .m2 are .a1 = f1 /m1 = GMplanet /r 2 and .a2 = f2 /m2 =
GMplanet /r 2 . Thus, f 1 and f 2 are different, but .a1 = a2 .
The ratio .mg /mi in Eq. 1.3 is a constant and may be given the value of 1 to assert
that the inertial mass and the gravitational mass own the same value. This in turn
reveals the equivalence principle (in the weak form) referent to the inertial and
gravitational masses under the presence of a gravitational field of local character.
The equivalence principle also points out the equivalence between a gravitational
field and an accelerated frame of reference, Fig. 1.2 [4].
1.3 Mass 7
Fig. 1.2 A compartment with an observer in the interior is accelerated in the upward direction.
The observer registers a laser light to curve downward. A similar phenomenon is watched when
a planet with strong gravity bends a light beam. Thus, the gravitational force exerted on the
light beam in the downward direction may be interpreted in terms of the force exerted on the
compartment in the upward direction. However, in contraposition with the particles with masses,
the photons describing the light beam have no rest mass, and the light bending is associated with
the curvature of spacetime created by the planet
Statement 1.2 Equivalence principle (in the weak form): the acceleration
of gravity acting on a body and registered in a small spacetime region is
equivalent to a uniform acceleration of the body produced by an external
force.
The mass is a property that makes the dynamics of atoms different: atoms with small
and large masses have different inertias in the dynamics of molecular systems. The
atomic mass, abbreviated amu, with value:
. amu = 1.660, 538, 92 × 10−27 Kg
is the reference unit mass at the atomic level. The mass may be also given in terms
of energy units due to the mass-energy equivalence. One gram of mass is close to
89.9 T J of energy (.1T = 1012 ). The large energy value of the mass is due to the
light-speed factor taking part in the expression .E = mc2 .3
3 A quantity of 46.6 mg of mass is required to approximately produce a kiloton of T N T . The
energy released by the first atom bomb of World War II was close to a dozen T N T kiloton, coming
from .∼ 64 Kg of enriched uranium.
1.4 Energy
From the mathematical perspective, the energy is a scalar quantity whose conjugate
variable is the time. From the physical perspective, the energy is the capacity of a
system to do work. The energy satisfies some important principles.
Statement 1.3 The energy is always preserved in isolated systems, and it is

neither created nor destroyed but only transformed onto other energy forms.
The conservation of a physical observable with time is associated with a symmetri-

cal aspect of nature. The symmetry corresponding to the conservation of the energy
is the translational property of time. An event that occurred yesterday remains
unchanged if we consider that it occurred any other day due to the translational
symmetry of time. Under a general appraisal, the events occurring in nature are
independent of the time coordinate, as the intervals of time are indistinguishable
from each other. The laws of physics do not change under space and time translation
(this is considered a symmetrical aspect of nature), leading to the conservation of
energy (this is the conservation of a physical observable). In a closed system, the
energy is preserved; nevertheless, when the system is open (this is with external
interactions), the energy changes.4 The energy-time relation is entangled and is
present in every aspect of nature. The Heisenberg principle of quantum theory
represents an example of the energy-time entanglement, where the limits imposed
by nature to the simultaneous measurements of the energy and time are faced: it is
not possible to measure the time without affecting the energy and vice versa.
The energy is conveniently classified according to the state of the object. The
object may be static or non-static, and the type of energy may be in that order
associated with potential energy or kinetic energy. The potential energy depends
on the position coordinates and may also depend on the velocities, while the kinetic
energy depends on the speeds. An example of the two basic energy forms is provided
by a non-static pendulum. At the turning points of the oscillation, the energy of the
body is potential energy, and at the middle point of the turning points, it is kinetic
energy. The potential energy is determined by the location of the object in space
under the presence of a forcefield, and the kinetic energy is defined by the motion of
the object. The energy is not absolute and is measured with respect to a given state
of the object. For the pendulum, the point where the body reaches the maximum
4 Interestingly, in going from a closed system to an open one, we may insist on considering the
system as a closed one; however, the external interactions modify the physical behavior of the
system, and the modified behavior of the “closed system” may be attributed to a change of the
physical forces with time, leading to an energy change. The symmetrical aspects on the nature of
the system were modified, with the consequent change in the energy.
1.5 Electric Charge 9
speed is frequently considered as the reference point of zero-potential energy. There

are no rules to establish a reference point of zero-potential energy, but an adequate
reference point facilitates the computation of the energy when the system evolves
with time.
In the dynamics of classical particles, we have continuous energy conversions of
the type potential-kinetic energy and vice versa and energy transfer from one body to
another as well. The conversions of energy commonly observed at the macroscopic
level are interpreted as friction and thermal energy. A system with energy may be
used as a source of work or heat energy for other systems to overcome barriers
imposed by the potential energy and induce, for example, the chemical transforma-
tion of compounds. The first law of thermodynamics establishes the energy balance
of the macroscopic systems, including the possibility of exchanging energy through
matter transfer. The first law is formulated in terms of the thermodynamic variables;
still, it is the result of the interactions taking place at the microscopic level. In this
respect, the theory of thermodynamics has molecular origins.
Many questions of philosophical character remain. It is a mystery the type of
energy that created the Universe in the initial Big Bang event. Still, we know that the
energy is preserved in the Universe. For confined systems of particles, the energy is
quantized; nonetheless, it acquires continuous values at macroscopic scales, where
Newtonian theory is acceptable. The form of energy available for work is getting
reduced, since most thermodynamic processes are irreversible and dissipate energy
in ways that cannot be recovered and reused.5 In spite of the knowledge gathered on
the energy, the question on the type of dominant energy that shall persist in the last
episode of the Universe remains unclear.
1.5 Electric Charge
In the standard model of subatomic particles, the electric charge represents the
coupling strength of the particle to the electromagnetic field and, in this way, to
space. Fundamental particles like quarks interact among each other to give integer
values to the charges of the nucleons. In the formulation of molecular dynamics, the
electric charge shall be considered a fundamental quantity, and for such a reason, it
is not defined in terms of other physical variables. The magnitude and sign of the
electric charge associated with a particle are determined experimentally. It is not
a general property because there are particles with no charge. The electric charge
is independent of the motion state of the particle. The conservation of the electric
charge of a molecular system is formulated by an equation of continuity in terms of
the charge density and the electric-current flux. At the quantum mechanical level,
the charge conservation is due to a phase invariance of the wave representing the
particle. There are particles with positive charge and negative charge, which may
have equal or different masses. The charge comes in multiples of the elementary
5A reversible process is a phenomenon where energy dissipation takes no place in the system.
electric charge, .e = 1.602, 176, 634 × 10−19 C in SI units, except for some
sub-atomic particles. The proton and electron have positive and negative charges,
respectively. The particles with no charge are the neutral particles, like the neutron.
A charged particle produces an electromagnetic field that exerts electromagnetic
forces on charged particles, but has no interaction with neutral particles. The
electrostatic interaction is ruled by the Coulomb’s law, which is similar to the
gravitational force .fg = −G m1 m2 n̂/|r1 − r2 |2 . The Coulomb’s law states that
the electric interaction decreases with the square of the interparticle distance.
q1 q2 r1 −r2
. f= n̂
4π ε |r1 −r2 |2 ; n̂ = |r1 −r2 | (1.4)
f is the mutal force between a pair of particles with charges .q1 and .q2 located at .r1
.
and .r2 , with interparticle distance .|r1 − r2 |. The permittivity of space is .ε, while
the Coulomb constant .kc = 1/(4π ε) has value .kc e2 = 2.30708 × 10−28 J · m.
The electromagnetic forces are capable of accelerating charged objects resembling
the acceleration of the gravitational field. At sufficiently small dimensions, we are
not able to tell if the acceleration of a particle is due to a gravitational force or an
electromagnetic force.
The nucleus of an atom is made of protons and neutrons. The protons and
neutrons are more spatially localized than the electrons. Such a feature imparts
to some extent a classical character to the nucleus, and under certain conditions,
it may be visualized as a point particle with a net mass and a net charge. On
the other side, the electron is depicted as a negatively charged cloud surrounding
the nucleus. In the dynamics of particles, the nuclei of the molecular system may
move from one place to another carrying their own positive charges and forcing
the redistribution of the electron cloud. It is common to find the electric charge
irregularly distributed in molecular systems. In spite of that, all the charges of
the sub-systems add up to preserve the original charge of the particle system. For
example, the amount of charge remains the same in chemical reactions. When
the charges of different signs are spatially separated in the molecule, the system
is in a polarized state. The macroscopic objects are usually neutral, but exhibit
internal polarization. The polarization may have positive or negative directions,
where the sign of the polarization is defined in arbitrary manner for a given
direction. In some materials, like the metals, the charge mobility is observed
under certain circumstances, leading to electrical conductivity. The exotic states
in quantum mechanics that involve electrical conductivity are, for example, the
superconductivity and the Meissner levitation of magnets, which are not explained
under the classical version of mechanics.
The fundamental variables of space, time, mass, energy, and electric charge
represent universal concepts in molecular dynamics. In principle, they are inter-
twined together, with physical meaning when the particles interact. The details of
the fundamental variables proliferate, yet the basic aspects deployed here are the
ones of main concern in molecular dynamics.
1.6 Reference System of Coordinates 11
1.6 Reference System of Coordinates
The description of a set of particles requires a system of coordinates. The purpose

of the system of coordinates is to assign a position to every particle in 3D space.
In classical mechanics, the atoms are considered tiny entities, represented by points
with no internal structure. The position of a particle is unambiguously defined by
a set of three real numbers in 3D space at a given instant of time. The position is
a vector with units of distance. The most common reference frame of coordinates
is the Euclidean one, using the orthogonal coordinates x, y, and z to specify the
position vector .r of the particle.
. r = (x, y, z) (1.5)
The description may be always migrated to the moving reference frame of the
particle to simplify the equations of motion. Different reference frames are expected
to move independently and in rectilinear form with respect to each other; otherwise,
external forces are present and should be considered in the systems. The rules
to change reference frames that move with a relative velocity, v, are defined
by the Galilean transformations. In the theory of special relativity, the proper
transformations are the Lorentz transformations as they consider the dilation and
contraction of spacetime. The Galilean and Lorentz transformations are equivalent
when the velocity v is smaller than the light speed, c, that is, .v << c.
The particle motion can be described by a sequence of positions of the particle.
To do this, we require the introduction of a parameter, t, to impose an order in
the positions of the particle. The parameter t increases monotonically, and it is the
time. It is with the time that we are able to define the time evolution of the particle
from its motion . By using the time, we keep different events apart, even though
they may occur at the same point in space. In this context, we have the opportunity
to unequivocally speak of the past, the present, and the future events. Once the
parameter t is introduced, the different positions acquire a sequential order like the
frames of a movie, defining the particle trajectory.
The Newtonian reference frame of coordinates assumes that the space and time
are dimensions independent of each other, treating them in homogeneous and
isotropic manner.6 A reference system of coordinates that satisfies these properties
ensures identical functional forms of the laws of physics, receiving the name of
inertial reference frame. The difference between an inertial reference frame and a
non-inertial one implies the presence of a force acting on the non-inertial reference
frame rather than on the particle. In consequence, the particle motions in the inertial
6 An isotropic medium has the same properties in all directions, and a homogeneous medium has
the same properties in all locations. In the uniform translational motion of objects, we require
an isotropic medium. A space with a symmetrical protuberance (where we may be positioned)
is homogeneous, but the space is not isotropic due to the protuberance. In the uniform rotation
of objects around the protuberance, we require a homogeneous medium. None of such concepts
imply the other.
and the non-inertial reference systems are different. As an example, and in the case
of a non-inertial reference frame, consider a particle moving on the planar surface
of a rotating disc, and in the case of an inertial reference frame, consider the same
particle moving on the planar surface of a static disc. From the perspective of the
observer in the inertial frame, the force that makes the particle trajectory different is
sensed as a fictitious force. In the absence of forces, a particle travels with constant
speed in straight line. Such an observation allows to alternatively define an inertial
reference frame as one that moves with the particle subjected to no forces. In this
situation, the particle is observed to be at rest, remaining at rest in the absence of
external forces. In the case of a particle moving with uniform motion with respect
to the inertial reference frame, it will continue doing so unless a force acts upon the
particle. The laws of physics are identical in the inertial reference frames, acquiring
their simplest functional forms. The conceptions may be equally inverted to indicate
that the laws of physics define the inertial reference frames.
Under particular conditions, the concept of an inertial reference frame may be
generalized. Consider two bodies equally accelerated in linear form. We situate our
reference frame on one of the bodies, observing that the second body stays at rest
with respect to our reference frame. In this context, the chosen reference frame may
be taken as an inertial one. An exception occurs when we deal with a beam of light,
which exhibits a fixed speed independently of the state of motion of the emitting
body. A case like this is considered in the theory of special relativity.
Our planet rotates, and thus, the reference frame of coordinates in the planet is
not an inertial reference frame of coordinates. However, we may in the laboratory
consider the distant starts as “fixed objects” to establish an inertial reference frame
of coordinates (strictly speaking, the distant starts are not fixed due to an expanding
Universe with the possible intervention of forces from dark matter and dark energy).
This reference frame of coordinates is of terrestrial use and is relatively adequate in
most experiments.
1.7 Newtonian Time
The time is a complex concept that remains under debate. It may be considered
as an essential component of the Universe which introduces a new dimension
and is independent of any event. Nonetheless, it may be conceived in a different
way, this is, as a human contrivance to describe the particle dynamics. No time
quantization that reveals the time as a physical variable in quantum mechanics has
been firmly established at this stage, although there are investigations in this respect
[5]. The laws of physics are time reversible, and there is no time direction favored
over another. Contrary to the mechanical formulation, there is apparently a time
arrow in thermodynamic theory. The second law of thermodynamics postulates the
increase of entropy in one time direction. Yet the time direction does not necessarily
correspond to the direction of entropy.
In Newtonian mechanics, the time t is conceived as the parameter that labels
the positions of the particle in an event. In general, any physical variable of the
1.8 Linear Motion 13
particle (that changes values due to the interaction of the particle with other entities)
may carry the label t. We consider t as the ordinary time, measuring it with a
clock that gives regular ticks. The quantity t considered in this way is referred to
as the Newtonian time. Both the time and the reference system of coordinates are
indispensable components to characterize the particle dynamics.
Before introducing the general principles of classical mechanics, it is convenient
to define the basic variables that describe the different types of motion of particles.
1.8 Linear Motion
The combination of the position and time defines the velocity vector, .v. When a
body moves, the velocity of the body corresponds to the displacement .Ar divided
by the interval of time .At: .Ar/At = [r(t + At) − r(t)]/At. It corresponds to the
expression of a line slope. In the limit of small time intervals, .At → 0, the time
derivative denotes the instantaneous velocity.
. v(t) = lim [r(t + At) − r(t)]/At = dr(t)/dt = ṙ(t) (1.6)

At→0
The velocity owns units of .distance/time. The magnitude of the velocity, .|v|, is
the particle speed. We are assuming no warped spatial or temporal dimensions in
the kinematic expressions when the particle moves with high speed; otherwise, the
theory of general relativity should be employed. Acceleration is the rate of change
of the velocity with respect to time.
. a(t) = dv(t)/dt = v̇(t) = r̈(t) (1.7)
Acceleration owns units of .distance/time2 , or .velocity/time. A change in the

magnitude or the direction of the velocity with time involves an acceleration, .a,
which is a vector proportional to the force vector, .f, acting on the particle. The mass
m is the factor of proportionality.
. f(t) = m dv(t)/dt (1.8)
Consider a set of small displacements .xi = vi δti along the x axis in the time interval
δti . The particle is initially located at .x0 . The total displacement at the instant of time
.
t is given by the sum of the individual displacements.

E
. x = x0 + i vi δti (1.9)
A similar expression is displayed Efor the continuous (or infinitesimallyf small)

displacements, but the sum symbol . is replaced by the integral symbol . in this
case.7
ft
. x = x0 + t0 v(t) dt (1.10)
If the speed is constant in the time interval .t −t0 , the total displacements 1.9 and 1.10
are .x(t) = x0 + v(t − t0 ). When the position is replaced by the velocity, and the
velocity by the acceleration in the last two expressions, we achieve the velocity in
terms of the acceleration.
E ft
. v = v0 + i ai δti ; v = v0 + t0 a(t) dt (1.11)
An alternative expression relating the variables x, v, and a is obtained from the

identities .dv = adt and .v = dx/dt. By multiplying both expressions, we have
.vdv = adx. The integration of this equation gives the velocity v in terms of the
acceleration a, which in turn depends on the position x.

fx
. v 2 /2 = v02 /2 + x0 a(x) dx (1.12)
There is accelerated motion when the velocity and acceleration exhibit the same
directions; otherwise, it is a retarded motion. There is uniform acceleration when the
motion is rectilinear with constant acceleration. This is the case where the velocity
is:
. v = v0 + a(t − t0 )
By inserting this identity in Eq. 1.10 the position of the particle is:
. x(t) = x0 + v0 (t − t0 ) + a(t − t0 )2 /2
When the time interval .t − t0 is small, we may consider two consecutive times in
the trajectory of the particle and obtain the discrete version of the positions.
. x(ti ) = x(ti−1 ) + v(ti−1 )[ti − ti−1 ] + 1

2 a[ti − ti−1 ]2 (1.13)
The acceleration produced by an external force acting on the particle is constant

in the above time interval .[ti−1 , ti ]. Thus, the classical particle responds to the
external (or environmental) force, and there is a cause-effect action on the particle.
In general, the force exerted on the particle produces translational motion. A force
may also produce the rotation of the particle. Before studying the particle in rotation,
7 The letter S is employed in different forms to denote a summation. We use the (broken) Greek
E f
letter . i in the case of a discrete variable and the (soft) Latin letter . in the case of a continuous
variable. There is a summation over the corresponding type of variable in every case.
1.8 Linear Motion 15
it is convenient to introduce the torque, defined as the cross product of the position
vector .r = (x, y, z) with the force vector .f = (fx , fy , fz ).
| |
| ux uy uz |
| |
.τ = r × f = | rx ry rz | (1.14)
| |
|f f f |
x y z
The unit vector .û = (ux , uy , uz ) is perpendicular to the plane formed by the vectors
r and .f, with direction defined by the right-hand rule. Expression 1.14 is a symbolic
.
form to describe the torque given by:
. τ = (ry fz − rz fy )ux + (rz fx − rx fz )uy + (rx fy − ry fx )uz (1.15)
The torque is important in investigating the rotation of a particle around an axis

when a force is applied on the particle. The lever-arm is computed from the rotation
axis. When the forces are concurrent (i.e., they all have an intersection in a single
point) [6], the total torque is the net force .fr in cross product with the particle
position. In this special case, the torque is equally obtained by adding the torques
produced by the individual forces.
E E E
. τ = r × fr = r × i fi = i r × fi = i τi (1.16)
The net force produces the rotation of the particle and is equal to the rotation
produced by the concurrent forces. This is not necessarily the case of non-
concurrent forces acting, for example, on a rigid body, which is considered as a set of
particles that keep their distances fixed, with no possible deformation, and showing
spatial extension. In general, the net effect produced by the individual forces .fi
acting on different points of the rigid body cannot be reproduced by the single force
.fr obtained from the summation of the individual forces. In the situation of the rigid
body, the net force .fr = f1 + f2 + · · · and the net torque .τ r = τ 1 + τ 2 + · · · are not
necessarily perpendicular to each other, and expressions 1.16 are not well defined.
Problem 1.2 Consider a number of individual forces .fi exerted on different

points of a rigid body. Show that the torque is preserved when the forces are
applied along the line of action, namely, the parallel line to the net force .fr .
Find the line of action when the position .r of a particle and the force .f over
that particle are coplanar.
E
E The net force and torque on the rigid body are .fr =
Solution i fi , and
.τ = (r
i i − r) × f i , with reference point .r. When the reference point is
changed by an amount .λfr , which is a parallel vector to force .fr , we have:
(continued)
E E E
.
i (ri − r + λfr ) × fi = i (ri − r) × fi + λfr × i fi
From the properties of the cross product, .A×B = AB sin(θ ) n̂, where .θ is the
angle between the vectors .A and .B and .n̂ is the unit vector perpendicular to
vectors
E.A and .B in the direction given by the right-hand rule.EThe last term is
.λfr × i fi = λfr × fr = 0, and the original expression .τ = i (ri − r) × fi is
recovered. Thus, the torque exhibits no change along the line of action .λfr . On
the other hand, when .r and .F are coplanar (.z = 0 and .fz = 0 by supposition),
the torque is .τ = xfy − yfx , where .fx and .fy are the rectangular components
of the force. This expression is the equation of the action line, with x and y
the variables describing it, .y = (fy /fx )x − τ/fx .
The equilibrium of a particle demands a zero net force, and the equilibrium of a
rigid body additionally requires a zero torque. When the rigid body is static, no
translational neither rotational motions are observed.
E E
.
i fi (t) =0 ; i τ i (t) = 0
1.9 Angular Motion
When a particle travels along a curvilinear trajectory, the kinematic equations of

the particle change with respect to those of the linear motion. Consider a particle
moving along a small arc .As. The initial and final points of the trajectory are .ri
and .rf . The vector joining the two points along a straight line is .Ar, Fig. 1.3. The
velocity of the particle is given in terms of the position difference .Ar = rf − ri .
. v= Ar
At = Ar As
As At
(1.17)
When the arc segment is sufficiently small, the vector .Ar and the arc segment .As
define the tangent unit vector .ut to the trajectory, while the fraction .As/At is the
particle speed along the arc segment.
. v = ut v ; ut = lim Ar
; v = lim As
(1.18)
As→0 As At→0 At
The vector .v is the particle velocity; it changes with time since both v and .ut change
with time. The acceleration demands differentiation of the velocity with respect to
time.
d(vut ) dut
. a= dt = dv
dt ut + v dt
(1.19)
1.9 Angular Motion 17
Fig. 1.3 A particle travels

along a circular path. The
initial and final vectors .ri and
.rf describe the position of the
particle at different instants of
time. When the arc segment
.As is sufficiently small, the
vector .Ar and arc segment
.As define a tangent vector,
.Ar/As, to the circular path
Fig. 1.4 A particle travels along a circular path. The quantities .ut and .un are the tangent and
normal unit vectors to the particle trajectory. They are expressed as .ut = ux cos φ + uy sin φ and
.un = −ux sin φ + uy cos φ. The angle .φ is formed by the vectors .ut and .ux . The tangent and
normal vectors are orthogonal to each other
The acceleration is a vector usually described in terms of the Cartesian components

ux and .uy . Nevertheless, it is convenient to use the tangent vector .ut and the normal
.
vector .un in the case of the angular motion. The tangent and normal vectors are
perpendicular to each other and form a basis set for the curvilinear coordinates,
Fig. 1.4. By imposing the perpendicular condition .ut · un = 0, the form of the
normal vector .un is deduced.
. ut = ux cos φ + uy sin φ ; ut · un = (ux cos φ + uy sin φ) · (ux a + uy b) = 0

(1.20)
The components of .un that satisfy the above expression are .a = − sin φ and .b =
cos φ. The possible signs .(−, +) or .(+, −) of .(a, b) define the inward or outward
directions of vector .un with respect to the arc segment. .un is chosen to point toward
the radial center of the arc .As, and therefore:
. un = −ux sin φ + uy cos φ
The following step is to compute .dut /dt to determine the acceleration 1.19. By
using the unit vectors .ut and .un , and taking consideration of the angle .dφ, whose
definition in radian units is .dφ = ds/ρ, with the angle .dφ displaying the arc portion
ds and .ρ the curvature radius of such an arc, the time derivative .dut /dt is:
dφ dφ ds ( )
.
dt = ds dt = 1
ρ v ; dut
dt = −ux sin φ + uy cos φ dφ
dt = un
dφ
dt = un ρv
(1.21)
The incorporation of the last result in Eq. 1.19 gives the acceleration in terms of the
unit vectors.
. a = at ut + an un ; at = dv/dt ; an = v 2 /ρ (1.22)
The tangent component of the acceleration is given by the change of the speed over
time, and the normal component is associated with the change in directions of the
particle velocity, as .un is related to .dut /dt in Eq. 1.21. The normal component is
called the centripetal acceleration. A similar procedure may be followed using the
unit radial vector .ur and the unit transversal vector .uθ .
Problem 1.3 Explore the situations when, on the one hand, the tangent
acceleration is zero (.at = 0) and, on the other, the normal acceleration is
zero (.an = 0), in the angular motion of a particle.
Solution When .at = 0, we have .dv/dt = 0, and v is constant. In this case,

the particle is in uniform circular motion because the speed is constant. In
contrast, when the normal acceleration is zero, namely, .an = 0, we may either
have .v = 0 or .ρ = ∞. By avoiding the trivial case (.v = 0), we find that the
particle travels along a linear path when .ρ = ∞.
The curvature radius .ρ is constant in a circular motion. The expression .dφ = ds/ρ
is used to obtain the angular velocity and the angular acceleration. By using
expression 1.18 of v, it is possible to write it in terms of the angular velocity, .ω.
. v = ds/dt = ρω ; ω = dφ/dt (1.23)
The angular acceleration is calculated from the angular velocity by differentiation

with respect to the time.
. α = dω/dt = d 2 φ/dt 2 (1.24)
On the other hand, the components of .a acquire different forms when .ρ is constant.
By using the expressions of .at and .an of Eq. 1.22 in the form .at = dv/dt =
ρ dω/dt = ρ d 2 φ/dt 2 = ρ α and .an = v 2 /ρ = ρω2 , the expression of the
acceleration is written in terms of .ρ and .φ.
1.10 Descriptions Between Inertial Reference Frames 19
( ) ( )2
d2φ dφ
. a = ρ α ut + ρ ω2 un = ρ dt 2
ut + ρ dt un (1.25)
When .ω is constant, the particle describes uniform circular motion. In this case, .ω
of Eq. 1.23 is integrated to show that .φ depends linearly on the time t.
. φ = φ0 + ω(t − t0 ) (1.26)
A system like this exhibits periodic motion and may be used as a clock to give clock
ticks and measure the Newtonian time. The period of the particle motion is defined
as the required time to achieve a complete cycle, and the frequency is the number of
cycles registered in a time unit. If the particle performs n cycles in the time interval
.At, the oscillation period is .P = At/n, and the frequency is .ν = n/At. Both
quantities are related to each other.
. P = 1/ν (1.27)
If we impose the initial conditions .φ0 = 0 and .t0 = 0, then Eq. 1.26 is reduced to
φ = ωt. A complete cycle of the circular motion occurs when .φ = 2π and .t = P .
.
. ω = 2π/P = 2π ν (1.28)
The uniform accelerated circular motion corresponds to the case where .α is

constant. In this case, .α of Eq. 1.24 is integrated to show that .ω depends linearly
on the time t.
. ω = ω0 + α(t − t0 ) (1.29)
If we consider that .ω = dφ/dt, and use the above equation of .ω, then it is possible
to integrate and obtain an expression referent to the evolution of the angle .φ.
. φ = φ0 + ω0 (t − t0 ) + α2 (t − t0 )2 (1.30)
A discrete version of this equation is achieved by following similar steps to those in

the deduction of Eq. 1.13.
1.10 Descriptions Between Inertial Reference Frames
In the following, we describe the motion of a particle according to two observers

positioned on different inertial reference frames. The inertial reference frames have
motion relative to each other with constant speed .Vx along the x axis. The positions
of the particle in the non-primed and primed reference frames are .(x, y, z), and
(x ' , y ' , z' , ). Such positions differ in the x components and are identical in the other
.
dimensions.
. x ' = x − Vx t ; y' = y ; z' = z ; t' = t (1.31)
The Galilean transformations rule this change of reference frames. The events in
both frames are simultaneous since the observers register the same time (a low
speed .Vx with respect to the speed of light c is assumed). The relative velocities and
accelerations are deduced from the above expressions by differentiation in time.
vx' = vx − Vx ; vy' = vy ; vz' = vz

. (1.32)
ax' = ax ; ay' = ay ; az' = az
If we consider that the acceleration multiplied by the particle mass is the force
exerted on the particle, then the last line of expressions indicates that .f' = f.
Thereby, the force is an invariant quantity under transformations involving a uniform
translational motion.
The second situation of interest considers two reference frames sharing a
common origin, but one of them is in uniform angular motion (.ω = constant) with
respect to the other. The rotating frame is a non-inertial reference frame. The particle
velocities and accelerations should be related in the two reference frames. Before,
it is convenient to write the central expression .v = ρω, Eq. 1.23, of the angular
motion in vector form to facilitate the algebra. This is done by considering .ρ as
the projection of vector .r onto the circular plane described by the particle, Fig. 1.5.
According to such a geometry .ρ = r sin γ .
Fig. 1.5 The tangent velocity .v, and the angular velocity .ω of a particle with angular motion
are shown in the lateral views. The variable .r is the position vector of the particle, and .ρ is the
projection of .r onto the plane described by the particle’s motion. The conical angle is .γ , and the
angle on the plane defined by the particle with respect to an initial position is .θ. Several velocity
vectors of the particle at different instants of time are also illustrated in the top view
1.10 Descriptions Between Inertial Reference Frames 21
. v = ρω n̂ = rω sin(γ ) n̂ = ω × r (1.33)
ω is the vector perpendicular to the circular plane passing through the center, .γ is
.
the angle between .r and .ω, and the unit vector .n̂ is perpendicular to the vectors .ω
and .r, with direction given by the right-hand rule. The acceleration is obtained by
differentiating .v with respect to time, keeping in mind that .ω is constant.
. a = ω × v = ω × (ω × r) (1.34)
The following step is to find the positions and velocities registered by the observers
in their own reference frames.
r = ux x + uy y + uz z = u'x x ' + u'y y ' + u'z z'

. (1.35)
v = ux vx + uy vy + uz vz ; v' = u'x vx' + u'y vy' + u'z vz'
Both reference frames share a common origin of coordinates, but the particle
position .r is described by the different unit vectors of the reference frames. The
relative circular motion implies a relation between the particle velocities .v and
' '
.v and between the particle accelerations .a and .a . An observer in one of the
reference frame considers the other observer in a rotating reference frame with
angular velocity .ω. The particle velocity is .v:
E ( du'i dri'
)
. v= dr
dt = i dt ri' + u'i dt ; ri' = x ' , y ' , z' (1.36)
Each unit vector .u'i rotates with velocity .du'i /dt. By considering Eq. 1.33, the first
term is described by .(ω × u'i ) ri' , where the basis vector .u'i plays the role of the
rotating vector .r in Eq. 1.33. The second term of the sum is the velocity .v' , Eq. 1.35,
of the particle in the rotating reference frame.
dx ' dy ' dz'

v = ω × (u'x x ' + u'y y ' + u'z z' ) + (u'x dt + u'y dt + u'z dt )
. (1.37)
= v' + ω × r' = v' + ω × r
E
Expression 1.35 was used to replace . u'i ri' with .r. Thus, the velocity of a particle
described by an observer in the non-primed reference frame of coordinates is the
same as the velocity measured in the primed reference frame of coordinates plus a
contribution coming from the relative angular motion of the reference frames. The
procedure to obtain Eq. 1.37 is a general one and implicitly defines a mnemonic rule
on the derivation of the vectors.
d '
.
d
dt b = dt b +ω×b (1.38)
where .b and .b' are vectors in the non-primed and primed reference frames. When
.b = r, expression 1.37 is recovered. The following step is to compute the
acceleration of the particle by either differentiating the velocity 1.37 or using the
rule 1.38, under the consideration of constant .ω.
dv'
.
dv
dt = dt +ω× dr
dt
(1.39)
The last term .dr/dt is the velocity .v and is replaced by the expression .v' + ω × r.
On the other hand, the first term .dv' /dt may be calculated from Eq. 1.35, where the
velocity .v' is shown in terms of the basis vectors .(u'x , u'y , u'z ), which change with
time. By following similar steps referent to the derivation of Eq. 1.37, and gathering
all the contributions together, we have:
a = a' + ω × v' + ω × (v' + ω × r)

. (1.40)
= a' + 2ω × v' + ω × (ω × r)
The observers register different accelerations in their reference frames. The discrep-
ancies are due to the Coriolis acceleration, .2ω × v' , and the centripetal acceleration
.ω × (ω × r). The force exerted on the particle is not an invariant quantity in this
case since the accelerations are different in the reference frames. Therefore, it is not
possible to formulate a conservation principle of the force.
References
1. M. Alonso, E.J. Finn, Physics (Pearson Prentice Hall, Essex, England, 1992)
2. D. Halliday, R. Resnick, J. Walker, Fundamentals of Physics, 9th edn. (Wiley, New Jersey, 2003)
3. A. Einstein, The Meaning of Relativity, 2nd edn. (Princeton University Press, Princeton, 2005),
with a new introduction by Brian Green
4. R.C. Tolman, Relativity Thermodynamics and Cosmology (Oxford University Press, Oxford,
1934), Sect. 72–74
5. M. Bauer, On the problem of time in quantum mechanics. Eur. J. Phys. 38, 035402 (2017)
6. O.J. Gutíerrez-Varela, R. Santamaria, Molecular nature of the drag force. J. Mol. Liq. 338,
116466 (2021)
Foundations of Newtonian Dynamics
2
The general purpose of molecular dynamics is to characterize the motion of

each particle of a particle system in response to the interaction with the other
particles and fields. This chapter presents the fundamental laws on the dynamics
of particles according to Newtonian theory. In the Newtonian approach, the atoms
are represented by points with no internal structure, and the mass and charge are
the intrinsic properties that distinguish them. Such properties are assumed to be
independent of the motion state of the atoms here. In the Newtonian approach, the
atoms are also referred to as the particles of the molecular system.
This chapter introduces additional concepts and basic variables to those already
defined in the previous chapter. The new concepts and the basic variables are
important to establish Newton’s laws and derive the equations of motion of the
particles. The laws are formulated after observations and experimentation and by
considering some idealizations in the proposal of the theory. The Newtonian laws
are used to explain the behavior of the classical particles. The linear and angular
motions of a particle are discussed from the concept of force. The time reversibility
of the motion equations of small and large particle systems is analyzed. We make
reference to the concepts of impulse, work, and power and classify important types
of mechanical energies. The Newtonian theory constitutes one of the fundamental
approaches of molecular dynamics to derive the functional forms of the forces,
establish the equations of motion, and solve them to determine the dynamics of
the system.
2.1 First Newton’s Law
Our everyday experience indicates that a particle changes its state of motion when
there is an interaction affecting it. In this regard, it is important to establish the
cause-effect relationship of the particle motion in a formal way. To do this, we
introduce the concept of force, understood as the entity responsible for producing
the change in the state of motion of the particle. Force has a magnitude and direction,

24 2 Foundations of Newtonian Dynamics
and therefore, it is a vector. Force is due to the correlation of the particle with the
other particles and fields, like the gravitational field, electromagnetic field, etc. The
macroscopic systems are made of many particles, and the force on a single particle
that builds the body is, in fact, the result of the many forces exerted on that particle
by the other particles of the system.
A reference system of coordinates is required to evaluate an external action
on the particle state of motion. The simplest reference system of coordinates is
positioned on an ideally isolated body, that is, a body free of interaction with any
other entity in the Universe, and moving along a straight line with constant speed.
According to the Galilean way of thinking, we have the opportunity to choose an
inertial reference frame where the particle is observed at rest, for simplicity. The
equivalence among the different inertial reference frames makes the particle at rest
physically indistinguishable from any other state of motion registered on a different
inertial reference system of coordinates. The particle owns the same inertia in the
different inertial reference frames. In our particular inertial reference frame, the
particle should remain static as long as it is free of forces from other physical
entities. Such an assessment leads to the first law of Newtonian dynamics.
Statement 2.1 First Newton’s law: a body stays at rest or continues moving
uniformly in a straight line unless acted upon by an external force that changes
the body state of motion.
In order to provide a mathematical formulation of the first law and be able to

evaluate the motion changes of the particle, it is convenient to introduce the concept
of linear momentum. The linear momentum, .p, is a vector proportional to both the
scalar mass m and the vector velocity .v of the particle.
. p = mv (2.1)
The mass distinguishes the particle, and in our case, it is independent of the motion
state. In terms of the linear momentum, the first principle of dynamics reads as
follows: a free particle moves with constant momentum. This principle implicitly
assumes the homogeneity of space and time to ensure the dynamic equilibrium of
the particle, provided it has no interaction with any other physical entity. When the
particle exhibits a momentum change, we infer that an external force acted upon
the particle. The first law can be generalized to systems constituted by two or more
particles, in spite of the possibility that the particles integrating the system may
interact with themselves, because the total momentum of the system corresponds to
that of an isolated body, but written in terms of the individual particle contributions.
E E
. P= i pi = i mi vi (2.2)
2.1 First Newton’s Law 25
The first law is translated to a version which involves the conservation of linear
momentum.
Statement 2.2 Conservation principle of the linear momentum: the total

momentum of an isolated system of particles is preserved in time.
The linear momentum is the central variable in the conservation principle. The
conservation principles are important to identify constants of motion and point
out the symmetrical aspects of nature. In molecular dynamics, we are interested
in such constants, as they help to both monitor the way in which the molecular
simulation unfolds and characterize the particle system. For instance, consider an
isolated system of particles where the momentum of a subset of particles increases.
The momentum of the complementary subset of particles is expected to decrease to
keep the total momentum of the isolated system with no change. New particles have
been discovered in experiments of sub-atomic physics by applying the conservation
principle of linear momentum. Concurrently, such experiments represent an indirect
proof of the conservation principle of linear momentum. The conservation principle
of the linear momentum also leads to the redefinition of mass from a dynamic
perspective: if we have two isolated particles with masses m and .mref , this last one
is the reference mass, and the particles interact in the x axis; then, after measuring
their speeds, the mass m is determined from the expression .m = mref vref /v.
Newton’s first law allows to identify the isolation of a particle, and thereby, it
is useful to determine inertial reference frames: if a particle has no change of its
momentum, then from the first law, we conclude that it is an isolated object, and it
may be used to define an inertial reference frame of coordinates. Based on the close
link between an isolated body and an inertial reference frame, the first Newton’s
law is also called the law of inertia. From a Galilean perspective, our observations
and measurements performed in different inertial frames are equivalent, and the
laws of physics formulated in the inertial reference frames have no change of
their functional forms; otherwise, a particular inertial reference frame is favored
or distinguished over the others. Such an observation leads to the invariance of the
physical laws.
Statement 2.3 Invariance of the laws of physics: the laws of physics are
mathematically invariant in inertial reference frames.
2.2 Second Newton’s Law
The purpose of introducing the concept of force is to evaluate the environmental

influence and the action of an external factor on the particle state of motion, without
paying too much attention to the details that provoke the motion disruption of the
particle. The force exerted on a particle does not necessarily imply a physical contact
with the particle. The force exerted by the planet makes the bodies fall to the planet
surface due to the gravitational force without ever exercising a physical contact.
The force is a thrust or drag capable of changing the particle velocity in magnitude
and direction. The acceleration of the particle is due to a force, and the force and
acceleration have the same direction.
In a system of interacting particles, the trajectory of a particle is determined by
all the acting forces on such a particle. In the special case where the addition of
forces exerted on the particle is null, there is a dynamic balance, and the particle
is in static equilibrium. In order to determine the momentum change of the particle
under the action of a non-zero net force, we consider the linear momentum of the
ith particle in separate form from the momenta of the other particles.
E
. pi + j /=i pj = constant (2.3)
The total momentum of the isolated system of particles is preserved at a different

instant of time.
E
. p'i + j /=i p'j = constant (2.4)
The primed and non-primed variables of Eqs. 2.3 and 2.4 indicate the momenta at
different instants of time. We subtract the quantity .p'i from .pi of the ith particle.
[E ( )] E
. pi − p'i = − j /=i pj − p'
j ⇒ Api = − j /=i Apj (2.5)
The expression indicates that an interaction of the ith particle with the other
particles, which constitute the environment for the ith particle, is a process where
the exchange of momenta may occur. A time interval has been considered in Eq. 2.5,
and we may deal with the rate of change of the linear momentum with time.
Api E Apj
.
At =− j /=i At (2.6)
In addition to the average value of the momentum, the instantaneous variation of the
momentum is also of interest. The instantaneous variation is obtained by taking the
limit .At → 0.
dpi E dpj
.
dt =− j /=i dt (2.7)
2.2 Second Newton’s Law 27
The mathematical definition of the force is introduced to describe the changes of

momenta of the interacting particles at a given instant of time.
dpi E dpj
.
dt = fi ; fi = − j /=i dt (2.8)
The change of the momentum of the ith particle is attributed to the force .fi exerted
by its environment (in this case, integrated by the other particles). In a similar form,
the momentum change of the sub-system composed by all the particles, except the
ith one which in this last case plays the role of the environment, is attributed to the
force exerted by the ith particle on the sub-system. In either case, the force exerted
upon the particle or the sub-system of particles corresponds to the time derivative of
the momentum. The generalization of the results to include different types of forces
on classical objects establishes the second Newton’s law.1
Statement 2.4 Second Newton’s law: the force exerted upon a particle
results in the change of the particle’s momentum.
If a body composed by many particles changes the state of motion as a single object,
it is due to the influence of an external force. The second law is complementary to
the first law by indicating that the preservation of the state of motion (established
by the first law) is altered by the implication of an external force (established by
the second law). The second Newton’s law may be shown in terms of the particle
position when we take the relation .vi = dri /dt between the velocity .vi and position
.ri into consideration.
. fi = d 2 (mi ri )/dt 2 (2.9)
For interactions where the mass is constant, the equation of the force transforms
onto:
d 2 ri dvi
. fi = mi dt 2
= mi dt = mai (2.10)
The acceleration is proportional to the force and inversely proportional to the mass.
By applying a force, we can characterize the amount of mass of two particles: if two
particles are embedded in the same environment and are acted upon by equal forces
1 The quantum origins of the basic forces are not fully clear. The basic forces in nature are the
strong force, weak force, electromagnetic force, and gravitational force. The strong force keeps
the nucleons together, and the weak force is responsible for the radioactive decay. The two forces
act at short distances, typical of nuclear sizes; instead, the electromagnetic and the gravitational
forces are long-range forces acting, respectively, on charged particles and masses. In principle, the
classical forces are explained in terms of the fundamental interactions.
but their accelerations are different, we conclude that their masses are different. On
the other side, if .m1 + m2 = m3 , then the acceleration of the system .m1 + m2 is
the same as the acceleration of the particle with mass .m3 when identical forces are
acted upon the two systems. The force satisfies the superposition principle, which
indicates that the acceleration of a particle produced by several forces .fi is equivalent
to the acceleration obtained by a unique force, which corresponds to the net force
.fnet .
E E
.
i fi = ma = fnet ; fnet = i fi (2.11)
The unit of force is the Newton, corresponding to .Kg m/s 2 in the MKS system
of units. The second law is compatible with the first law when the particle is free
to move or it is not interacting with anything else. In this case, the net force is
zero, .fnet = 0; the acceleration is also zero, .a = 0; and the momentum .p is
constant. While the mass quantifies the inertia of a particle, the force quantifies
the environmental influence on the particle. Both entities are related in the second
Newton’s law, which, additionally, is mathematically invariant in any other inertial
reference frame.
The particles are correlated by their interactions, and such interactions are
considered by the interatomic potentials. There is a type of forces that are deduced
from interatomic potentials and another type of forces with no association to an
interatomic potential. In the case of a force derived from an interatomic potential,
we have the following prescription to get the force:
. fi = −∇i V (2.12)
The force .fi acts on particle i and is due to a change of the potential V with the
particle position. In order to illustrate the effects of a potential on the particle,
consider a ball located somewhere in a mountain. The ball displacement depends
on the local steepness where the ball is positioned. From the physical perspective,
a force is produced when the particle, which is immersed in the potential, changes
positions. In this regard, the motion of the ith particle may be explained in terms of
the interaction potential.
. − ∇i V = mi ai
Whether the Newton’s law is written in terms of a force or in terms of an

inter-particle potential, it is a central equation in molecular dynamics. When the
functional form of the force or the inter-particle potential is known, we refer to such
a central equation as the equation of motion of the particle.
2.3 Third Newton’s Law 29
2.3 Third Newton’s Law
When two particles interact, the force between them acts equally on both particles.
This is due to the fact that the force is not defined on a single particle, but it is
exerted on the two interacting particles with no directional preference along the
line connecting them. The directional symmetry of the force in the dynamics of the
particle is formulated in the third Newton’s law, known as the action-reaction law.
Statement 2.5 Third Newton’s law: when two particles interact, the force
exerted on one particle is equal in magnitude but opposite in direction to the
force exerted on the other particle.
By introducing the force of Eq. 2.8 exerted on the ith particle, .fi = dpi /dt, and
the force exerted on the object, which is constituted by all the other particles, .f =
E
j /=i dpj /dt, the symmetry in the forces is observed.
dpi E dpj
.
dt =− j /=i dt ⇒ fi = −f (2.13)
In the particular case of dealing with two particles, we have the interaction of the
ith particle with the j th one. By taking .fj = dpj /dt, we obtain the symmetric
condition on the forces.
dpi dpj
.
dt =− dt ⇒ fi = −fj (2.14)
The summation .fi + fj is zero as the internal forces are balanced in an isolated
system of particles; otherwise, a self-acceleration of the particle system occurs,
contradicting the first Newton’s law. The particles of the system interact by
exchanging momentum while simultaneously preserving the total momentum of
the system. The conservation of the total momentum is verified by integrating the
equations of the internal forces.
The electrostatic interactions among the particles are not shown instantaneously,
because the interactions propagate with a velocity equal to the speed of light. The
finite speed propagation of the interactions results on a time delay, invalidating the
third Newton’s law. Nevertheless, when the particles are not far away, the third law
may be considered as an acceptable law.
2.4 Reduced Mass of a Two-Particle System
When two particles interact, it is possible to have a simplified description of the two-
particle system by introducing the reduced mass. In order to illustrate it, consider
the equations of motion of the two interacting particles.
. dv1 /dt = f12 /m1 ; dv2 /dt = f21 /m2 (2.15)
.fij is the force exerted by the j th particle on the ith particle. By subtracting the
equation on the right from the equation on the left, and using the third Newton law
.f12 = −f21 , the relative acceleration between the two bodies is obtained.
( )
. a12 = + ; a12 = d(v1 − v2 )/dt
1 1
m1 m2 f12 (2.16)
The two-particle system is reduced to the description of a single particle by

introducing the reduced mass, .μ, in terms of the individual masses.
. 1/μ = 1/m1 + 1/m2 (2.17)
The reduced mass is an effective mass. By making use of .μ, the equation of the
effective (single) particle is derived from expression 2.16.
. f12 = μa12 (2.18)
The reduced mass is important to simplify the description of a particle pair. When
the mass of one of the particles, say .m1 , is sufficiently heavier than the mass of the
other particle, .m2 , the reduced mass may be approximated by the light mass .m2 .
.
1
μ = 1
m1 + 1
m2 = 1
m2 ; m2 << m1 (2.19)
Thereby, in the interaction of a heavy mass with a light mass, the light mass is the
one essentially accelerated when the two particles are observed from a reference
system of coordinates fixed at some point on the line joining the two particles.
2.5 Time Reversibility
The time reversibility of the laws of physics is intimately linked to the conservation
of the energy. The time reversible property indicates that the laws of physics
are immutable to the passage of time, favoring no time arrow and ensuring the
preservation of the energy when the system is isolated. In this connection, the
equations of motion are capable of describing the reverse course of events which
were unfolded by the particle dynamics [1]. In order to determine the required
conditions that maintain the equations of motion time symmetric, let us consider
the instantaneous position of a particle, .x = x0 + vt + at 2 /2. The velocity and
2.5 Time Reversibility 31
acceleration satisfy the expressions .dx/dt = v + at and .dv/dt = a (the mass is

assumed to be constant and is avoided). When the change of t by .−t is produced,
and the velocity is reversed, .v → −v, the particle trajectory is recovered in any of
the previous equations, for instance, .x = x0 + (−v)(−t) + a(−t)2 /2.
This property is confirmed by considering that in the forward-time direction, the
trajectory of a particle is described by the variable .r = r(t), and in the reverse-
time direction, the trajectory is described by the primed variable .r' = r' (tf − t).
The time variable t is used with opposite signs in the previous cases. In the forward
direction, the particle travels from time .ti = 0 to .tf = 1 and in the reverse direction
from time .tf = 1 to .ti = 0. The initial and final conditions on the trajectories are
' ' ' '
.r(ti ) = r (tf − ti = 1) = r (tf ) and .r(tf ) = r (tf − tf = 0) = r (ti ). By employing
the chain rule, the forward and reverse velocities are the same, except for the change
of signs (or directions).
dr(t) dr' (tf −t) d(tf −t) ' '
.
dt = d(tf −t) dt = − drdt(t' )
A similar situation is observed by directly dealing with the time .−t instead of
tf − t: again, if .r' is the reverse time trajectory and .t ' is the reverse time, that
.
is, .t ' = −t, then .dr(t)/dt = [dr' (−t)/d(−t)] [d(−t)/dt] = −dr' (t ' )/dt ' . The
acceleration that participates in the equation of motion remains invariant due to
repeated derivations, .d 2 r(t)/dt 2 = d 2 r' (t ' )/dt '2 . Thus, for a particle describing the
trajectory .r(t) with velocity .v(t), the time reversibility is produced by reversing both
the time and the velocity.
. t → −t ; v → −v
In principle, it is possible to stop a particle traveling in the forward direction and

reverse its velocity with the same magnitude that it owned in the forward direction.
To include the necessary parts of this experiment, we also produce a specular
reflection of the Newtonian clock with the purpose to reverse the ticks of that clock,
Fig. 2.1. Thus, the reverse trajectory may be visualized as a movie of the particle
running in the backward direction.
Fig. 2.1 Two clocks are shown. Each one is the specular image of the other. The hands of the
clock on the right run clockwise, with time 5:39 hours, and the hands of the clock on the left run
counter-clockwise, with time 6:21 hours
A similar situation may be done for two or more interacting particles, and again,
the behavior of the particles is shown to be time reversed for an observer. Formally, if
the variables .r(t) and .v(t) are solutions of the equations of motion, then the variables
' ' '
.r(t ) and .v (t ) are also solutions of such equations.
}
[r(t), v(t)] 2 2 ' ' ' '
f(t) = m dv(t) d r(t) d r (t ) dv (t ) '
.
dt = m dt 2 = m dt '2 = m dt ' = f(t )
[r(t ' ), v' (t ' )]
(2.20)
The existence of the solution .[r(t ' ), v' (t ' )] gives the possibility to observe and
produce the reverse order of events. The invariance of the Newtonian equations of
motion under the above transformations of the time and velocity allows to establish
an equivalence between the forces:
. f(r, v, t) = f(r, v' , t ' ) (2.21)
In order to observe the reverse trajectory of a macroscopic object formed by many

microscopic particles, we have to apply the transformations on the time and the
velocity of every microscopic particle. In this respect, the macroscopic objects favor
no time arrow.
Problem 2.1 The position and velocity of a particle are .r and .v. Suppose that
the force on the particle is .f = αr × v, where .α is a constant. Show that the
system is not time reversible.
Solution From the expression of the force, we have that .f(r, v, t) = αr × v

is different from .f(r, −v, −t) = −αr × v. Thus, when the time and velocity
are reversed, the forces are different, and the system is not time reversible.
Frequently, the characterization of a system integrated by a large number of particles

is done with statistical mechanics or thermodynamics instead of Newtonian theory.
We introduce simplifications in statistical mechanics and thermodynamics to reach
more amenable expressions and procedures to deal with. However, by doing so,
there is a loss of information since many degrees of freedom (usually those asso-
ciated with the environment) are simplified, acquiring a statistical nature. In such a
reductionism, a complete description of the system is abandoned. In particular, we
miss the capability to analyze the particle trajectories in the forward and backward
time directions, because there are not enough details of the forces acting on the
2.6 Angular Momentum and Torque 33
particles. In this process, the evaluation of the energy is strictly affected, resulting
in conflict with the energy conservation and the time symmetry of the Newtonian
approach.2 Yet in order to characterize the system at the macroscopic scale, new
concepts with macroscopic meaning like the temperature, friction, heat, entropy,
and more are introduced. They have statistical character, resulting from averaging
the behavior of the microscopic particles when the system is energy equilibrated,
because in the equilibrium state, the macroscopic variables are observed to be
stationary parameters, that is, time-indifferent parameters. The energy relaxed
states of the many-particle system classify as thermodynamically stationary states.
In short, we face practical problems in solving a myriad of coupled equations of
motion for the accurate description of a many-particle system in Newtonian theory;
nevertheless, statistical mechanics and thermodynamics allow the characterization
of the many-particle system from a statistical perspective, with the cost of missing
some information, like the time reversibility of the macroscopic particle system.
2.6 Angular Momentum and Torque
In this section, we discuss the conservation law of the angular momentum since it is
frequently applied in molecular dynamics. Before, it is convenient to introduce the
following concepts to classify a particle system.
• When the system is free of interactions with the rest of the Universe, it is an
isolated system.
• When the system is in thermal or mechanical contact with the environment and
there is no exchange of matter, we have a closed system, like a closed jar of coffee
in contact with the surroundings.
• When the system exchanges matter and energy, it is an open system, like a jar of
coffee without being closed.
The angular momentum .l and torque .τ are physical observables of interest

to characterize a particle system. The angular momentum and torque are both
proportional to the mass m of the particle and involve the cross product of the
position vector .r and the velocity .v for the angular momentum and the position
vector .r and the acceleration .a for the torque.
. l = r × mv = r × p ; τ = r × ma = r × f (2.22)
2 The lack of a strict energy conservation of a particle system is commonly due to the oversim-
plification of the degrees of freedom associated with the external factors that affect the particle
system. Still, it is possible to introduce modified forces among the particles to compensate for
the environmental effects and formulate the preservation of a modified energy: Santamaria and
Soullard [2].
The angular momentum and torque are vectors that depend on the reference system
of coordinates. The angular momentum .l is related to the torque .τ by a time
derivative due to the use of .v in the angular momentum and .a (.= dv/dt) in the
torque.
. τ = dl/dt (2.23)
For a many-particle system, the individual terms .li = ri × mi vi and .τ i = ri × mi ai

are added to obtain the total angular momentum, .L, and the total torque, .τ tot .
E E
. L= i ri × mi vi ; τ tot = i ri × mi ai (2.24)
Some constraints on the dynamics of particles may be introduced by establishing

initial values of .L and .τ tot . For a system of n particles, the angular momentum
components are:
En y
Lx (t) = i=1 mi [yi (t) viz (t) − zi (t) vi (t)]
En
. Ly (t) = i=1 mi [zi (t) vix (t) − xi (t) viz (t)]
En y
Lz (t) = i=1 mi [xi (t) vi (t) − yi (t) vix (t)]
y
where .mi is the mass of the ith particle and .(xi , yi , zi ) and .(vix , vi , viz ) are the
position and velocity vectors. A similar expression is obtained for the torque .τ tot .
The total angular momentum of an isolated system is preserved with time because
in this case, there are no external torques on the system, .τ tot = ∂L/∂t = 0, and
.L = constant.
Statement 2.6 Conservation principle of the angular momentum: the total

angular momentum of an isolated system of particles is preserved with time.
In the closed and open systems, the angular momentum of the particle system may
change with time due to the influence of external factors.
2.7 Impulse, Work, and Power
The interaction of a particle with the rest of the Universe is described by the force
f, satisfying the equation .f = dp/dt. The integration of the force with respect to the
.
time defines the impulse, .I.

2.7 Impulse, Work, and Power 35
fp ft ft
. I= p0 dp = t0 f dt ⇒ I = p − p0 = t0 f dt (2.25)
The impulse is a vector (with direction and magnitude) and corresponds to the
change of momentum due to a force acting on the particle. A large impulse is
produced on the particle when either a strong force is applied during a short time
interval or a weak force is applied during a long time interval. By integrating once
more, the position of the particle is determined.
f
1 t
. r(t) = r0 + v0 t + m t0 I dt (2.26)
The time evolution of the particle is obtained when the impulse in terms of the
time is known. A force exerted on the particle produces a work along the particle
displacement.
. dW = f · dr = f dr cos(θ ) (2.27)
θ is the angle between the force .f and the displacement element .dr. When .θ = π/2,
.
there is no work done by the force, and when .θ = 0, the maximum work on the
particle is performed. The total work in moving the particle from point A to point B
is given by the sum of infinitesimal displacements.
E fB
. W = i fi · dri ⇒ W = A f · dr (2.28)
The integral that defines the work is a line integral along the trajectory of the
particle. Contrary to the expression of the impulse, where we require to know the
force in terms of the time, the expression of the work requires to know the force in
terms of the position coordinates .(x, y, z). Under the situation of a constant force,
we have:
. W = f · (rB − rA ) (2.29)
The work depends on the spatial separation of points A and B. When the force is the
addition of several forces, .f = f1 + f2 + · · · , the work on the particle is the addition
of the works produced by the individual forces.
. dW = (f1 + f2 + · · · ) · dr = dW1 + dW2 · · · ; dWi = fi · dr (2.30)
The work on a particle may be done slowly or fast. The work per unit of time defines
the instantaneous power.
. P = dW
dt =f· dr
dt =f·v (2.31)
In many physical processes, the power plays a more important role than the work
itself.
2.8 Kinetic and Potential Energies
The work performed on a particle is written in terms of the speed .v.
. dW = f · dr = m dv
dt · dr = mv · dv
(2.32)
The integration of the above expression gives the total work on moving the particle
from point A to point B in terms of the speed.
fB mv2 (B) mv2 (A)
. W = A mv · dv = 2 − 2
(2.33)
The work is given by the quantity .mv2 /2 evaluated at the points A and B, and
in this expression, the work is independent of the nature of the force. The term
.mv /2 is proportional to the square of the momentum, .mv /2 = p /(2m), and
2 2 2
receives the name of kinetic energy. Expression 2.33 written in the form .W =
[p2 (B)−p2 (A)]/(2m) shows a quadratic dependence on the momentum. In contrast,
expression 2.25 of the impulse, .I = p(B) − p(A), shows a linear dependence
on the momentum. The kinetic energy is an energy of motion represented by
.Ek = p /(2m). The work is related to the change of kinetic energy.
2
. W = Ek (B) − Ek (A) ; Ek = p2 /(2m) (2.34)
This expression is general with many applications in mechanics. In the following

lines, we classify the force in terms of the type of work performed on the particle.
When the work done on a particle is independent of the particle trajectory but it
depends on the initial and final positions of the particle, the force responsible for that
work is recognized as a conservative force. An example of a conservative force is the
gravitational force. In the case of a conservative force, the work may be expressed in
terms of a function of the particle coordinates, which receives the name of potential
energy and is designated by .Ep . Thus, .Ep also represents energy which is gathered
with the possibility to perform work. The existence of conservative forces allows
to write the work on moving the particle from point A to point B as the change of
potential energies.
. W = Ep (A) − Ep (B) (2.35)
When the initial and final points coincide, .A = B, the work is zero as a result of
the conservative force acting on the particle. The point of zero potential energy is
arbitrarily decided, as the energy difference .Ep (A) − Ep (B) is the relevant quantity
and constitutes a physical observable. According to Eqs. 2.28 and 2.35 and the
fundamental theorem of calculus, the function .Ep is a scalar function, continuous
in the interval .[A, B], and is the anti-derivative of .f.
fB
. W = A f · dr = Ep (A) − Ep (B) (2.36)
2.8 Kinetic and Potential Energies 37
Any other anti-derivative of .f, say .Ep' , is equal to .Ep except for a constant c, .Ep =
Ep' + c. The constant c allows to arbitrarily define the reference point of potential
energy as previously mentioned. The function .f is written in the anti-derivative form:
. f = −∇Ep (r) (2.37)
This expression is an alternative definition of the potential energy .Ep by emphasiz-

ing the fact that a conservative force .f depends on the positions solely. The potential
energy constitutes a force field where the interactions among the particles are of
non-contact type and receive the name of forcefield. The forcefield tends to take
the particle to the position of minimum potential energy. Any work performed on
the particle has to counteract the forcefield at every point in space, since the work
attempts to increase the potential energy. This is a reason to have a negative sign
in the relation of the force to potential energy. Such a fact also explains the reverse
order of the potential energies in Eq. 2.35 with respect to that of the kinetic energies
in Eq. 2.34. In addition to expressions 2.36–2.37, the curl operator may be used to
define a conservative force:
. ∇ × f(r) = 0 (2.38)
where it is understood that the curl of the gradient of a scalar function, .∇ × (∇Ep ),
is zero. There are several examples of forcefields such as the gravitational field,
electric field (the magnetic field acting on a charged particle is a non-defined case),
the force exerted on a body by a spring, etc. In all those cases, there is a potential
energy associated with the particle position. The potential energy acquired by the
particle in the forcefield is independent of the trajectory that we chose to move the
particle from point A to point B. More precisely, the amount of energy .Ep (A) −
Ep (B) is constant regardless of the trajectory .A ↔ B. It is possible to recover the
energy that was inverted in moving the particle from point A to point B because the
conservative force preserves the potential energy of the particle in the forcefield.
There are forces that satisfy none of expressions 2.37–2.38. They are classified
as non-conservative forces. Examples of such forces are those with dependence
on the particle velocity, the forces changing in time, the friction forces, and the
viscous forces. The non-conservative forces are usually determined from statistical
considerations on the particle interactions, and do not constitute a forcefield.
Consider a solid sphere immersed in a fluid. The viscosity of the fluid is defined
by the statistical average of the interactions on the sphere. The viscosity of the fluid
only acts on the external particles of the sphere because it is not a forcefield [3]. In
contrast, the gravitational force is a forcefield acting on all the atoms of the solid
sphere, including the internal atoms of the sphere.
2.9 Energy Conservation
The energy is a scalar quantity which establishes a relation of fundamental variables.

It may be transformed onto different types of energy and even disassembled onto
different energy terms.
We have the possibility of writing the work in terms of changes of either the
kinetic energy or the potential energy. Still, we may combine both expressions using
Eqs. 2.34 and 2.35.
. Ek (B) − Ek (A) = Ep (A) − Ep (B)
By reorganizing terms, we obtain the energies of the particle at point A separated

from the energies of the particle at point B, reaching the expression .Ek (B) +
Ep (B) = Ek (A) + Ep (A). Since the points A and B are arbitrary, we conclude
that the kinetic energy plus the potential energy is preserved at any instant of time
and at any point in space under the consideration of conservative forces. Still, the
kinetic energy may be converted onto potential energy, and vice versa, as a form of
reversibility.3
. Ek + Ep = constant (2.39)
Problem 2.2 A particle in a forcefield moves in the x direction. The total

energy is .E = m(dx/dt)2 /2 + Ep (x). Find the equation of dx in terms of dt,
and obtain the time evolution .x(t) of the particle by assuming .Ep = −f x,
where f is a force.
Solution The expression of the energy contains .dx/dt. In this regard, we

solve for the differential dx in terms of dt.
( m )1/2 f x ft
.
2
dx
0 [E−Ep (x)]1/2 = 0 dt = t
This is a general equation in terms of the total energy and potential energy
√ as .Ep =
of a particle moving in the x axis. By taking the potentialf energy
−f x,√the previous expression is integrated using the fact . dx/ ax + b =
(2/a) ax + b.
(continued)
3 The conservation of the mechanical energy, .Ek + Ep , is a strict expression. In contrast, the
conservation of the thermodynamic energy proposed by the first law of thermodynamics is not a
strict expression due to the use of average values. This makes the first law of thermodynamics a
law on the statistical balance of thermodynamic energies (exclusive for states in thermodynamic
equilibrium), rather than a law on energy conservation.
2.9 Energy Conservation 39
( m )1/2
.
2
2
f [(E + f x)1/2 − E 1/2 ] = t
By solving for the position x, the result is .x(t) = (2E/m)1/2 t + (f/m) t 2 /2.
We impose the initial conditions on the particle motion. At the initial time
.t = 0, the position is .x = 0, and the total energy is .E = mv /2. We also have
2
0
.f/m = a. In this connection, the final result is .x(t) = v0 t + at /2.
2
The energy conservation is employed to characterize physical systems.4 In addition

to the classification of energy terms, we have the classifications of forces. When the
torque .τ = r × f is zero, with .r /= 0 and .f /= 0, then the vectors .r and .f are parallel.
Under such conditions, the force is known as a central force. It may be also stated
that for central forces, the torque is null. Many interaction potentials depend on the
distance r solely, making the force to also depend on the distance. This is the case
of bodies in orbital motion. In this situation, the total energy of the particle may be
unfolded, for example, in polar coordinates. By using the expression of the speed in
polar coordinates .v 2 = vr2 + vθ2 , with .vθ = rdθ/dt, and the angular momentum L
given by:
L = r × mvθ = mrvθ (r̂ × θ̂)

. ⇒ mvθ2 /2 = L2 /(2mr 2 )
the total energy is:

( dr )2 L2
( dr )2
E= m
2 dt + 2mr 2
+ Ep (r) = m
2 dt + Eeff (r) ;
. (2.40)
L2
Eeff (r) = 2mr 2
+ Ep (r)
where the effective potential energy .Eeff in terms of the radius r was defined. The
angular term .L2 /(2mr 2 ) is positive and increases the magnitude of the potential
energy .Ep at every point r. By following similar steps to those performed in
problem 2.2, and replacing .Ep by .Eeff , the equation of dr in terms of t is achieved.
( m )1/2 f r
. 2
dr
r0 [E−Eeff (r)]1/2 =t (2.41)
After integration, it provides the variable r in terms of t. On the other hand, if we

consider the relation between .θ and the angular momentum, .dθ/dt = L/(mr 2 ),
then the angle .θ in terms of t is:
4 The energy is strictly not preserved due to the expansion of spacetime.

ft
. θ = θ0 + L
0 mr 2 dt (2.42)
In order to have an equation of r in terms of .θ , we employ .(dr/dt)(dθ/dt)−1 . By

using the first expression of Eqs. 2.40 for .dr/dt and .dθ/dt = L/(mr 2 ), we reach:
[ ]1/2 ( 2 ) fr
.
dr
= 2
(E − Eeff (r)) mr
⇒ r0 dr
= θ − θ0
dθ m L [2(E−Eeff (r))/m]1/2 (mr 2 /L)
(2.43)
The last expression is the equation of the orbit, .r = r(θ ). The total work exerted
on a body is different from zero when non-conservative forces act on the body in a
closed-loop trajectory due to an irreversible transfer of energy, which results on a
net energy loss. The non-conservative forces are usually of statistical nature, where
we pay no attention to the individual particle interactions but to the effective force
of the surrounding particles on the particle of interest. In spite of the possibility of
associating a non-conservative force to a statistical average force of the surroundings
(acting on the particle), a non-conservative force is also characterized as a force
which is not derivable from a potential energy function.
References
1. B.W. Roberts, When we do (and do not) have a classical arrow of time. Philos. Sci. 80, 1112–
1124 (2013)
2. R. Santamaria, J. Soullard, Revisiting particle dynamics in the NPT ensemble under the
extended Lagrangian approach, Molec. Simul. 49, 9 (2023), https://doi.org/10.1080/08927022.
2023.2195010
3. O.J. Gutíerrez-Varela, R. Santamaria, Molecular nature of the drag force. J. Mol. Liq. 338,
116466 (2021)
Many-Particle Systems
3
Most of the systems of our real world are constituted by many particles. The purpose
of the present chapter is to introduce the relevant variables that characterize a many-
body system. Many-body systems are important to link Newtonian mechanics to
statistical mechanics and thermodynamics.
In a first stage, the reference frame of a many-particle system is established,
to later define the angular momentum and torque of the system. The different
types of energies of a many-particle system are considered. Heat is introduced
as an external factor to formulate an energy balance expression of the system in
interaction with its environment. In a second stage, with the purpose to close gaps
between Newtonian mechanics and statistical mechanics, the equivalence between
the time averages and the ensemble averages of physical observables is discussed.
The mechanical variables of pressure, volume, temperature, and work, which are
common in molecular dynamics and statistical mechanics, are derived in the last
sections. The combination of the pressure, volume, and temperature into a single
expression leads to an equation of state of the particle system.1
3.1 Reference Frame of a Many-Particle System
A reference frame of coordinates is required to describe a system of many particles.

A general approach to define the reference frame is to consider all the particles
without granting particular privileges to any specific particle. In this respect, the
position of the center of mass (cm) is chosen as the position of the many-particle
system.
E E
. rcm = i mi ri /mtot ; mtot = i mi (3.1)
1 The discussion follows similar topics to those of: Alonso and Finn [1].

42 3 Many-Particle Systems
In this expression .mi and .ri are the mass and position of the ith particle. The velocity
of the cm is obtained by derivation of the position variables with respect to time.
E
. vcm = i mi vi /mtot (3.2)
The momentum of the particle system is defined by the product of the body mass
with the body velocity.
E
. ptot = mtot vcm = i mi vi . (3.3)
The entire system is visualized as a single particle with mass .mtot and speed .vcm .
There is a relative motion of the individual particles with respect to the cm position.
The internal forces produce no change of .ptot because the third Newton law makes
them zero by pairs, .fij + fj i = 0, Eq. 2.14. Thus, the total momentum .ptot is
maintained constant when the system is isolated. Under such circumstances, it is
convenient to locate the origin of coordinates at the position of the cm to have an
inertial reference frame where the total linear momentum of the isolated system of
particles is zero. However, if an atom or set of atoms changes its momentum by
virtue of an external force, the total momentum of the system given by Eq. 3.3 is not
conserved. This occurs, for instance, when a protein interacts with external agents,
like the atoms of the solvent. There is an exchange of momenta between the different
systems of particles (namely, the protein and the particles of the solvent). Still, it is
possible to consider a single system of particles formed by the two sub-systems.
Let us denote with the letter S the system of particles and with the letter .S ' the
particles of the external system. The composed system is .S + S ' , and it is isolated
and preserves the total momentum.
'
. ptot = pStot + pStot = constant (3.4)
'
A change of the momentum .pStot implies a change of the momentum .pStot .
E E '
.
iεS ApSi = − i ε S' ApSi (3.5)
By differentiating every term of Eq. 3.5 with respect to time, the mutual external
forces acting on each system are obtained.
'
d
dt pStot = fext ; fext = − dt
d S
ptot
. (3.6)
'
d
dt pStot = f'ext ; f'ext = − dt
d S
ptot
The composed system is isolated, the total force is zero, .fext + f'ext = 0, and the
third Newton law in the subsystems S and .S ' is satisfied, .fext = −f'ext . The external
force .fext acts on the cm of system S producing the acceleration .aScm .
3.2 Angular Momentum and Torque of a Many-Particle System 43
E dvScm
. fext = d
dt pStot = d
dt iεS mi vi = mStot dt = mStot aScm (3.7)
The external force .fext acting on the whole system S is equivalently considered as
the outcome of all the individual external forces .fi acting on the single particles.
E E
. fext = d
dt iεS mi vi = i ε S fi (3.8)
The internal forces among the particles of the S system have no effect on the total
momentum of such a system. However, the external forces to the S system are the
forces capable of producing a non-zero contribution to the total momentum of the S
system.
3.2 Angular Momentum and Torque of a Many-Particle System
The angular momentum .l of a particle is .l = r × p, where .r and .p are the position

and momentum of the particle with respect to a given reference frame. In a system of
particles, the total angular momentum includes the contribution .li of every particle.
E E
. L= i li = i ri × pi (3.9)
The change of angular momentum .L of a many-particle system with time is related

to the external torque .τ ext .
E E E
.
d
dt L= d
i dt li = i ri × fi = i τ i,ext = τ ext (3.10)
The many-particle system is described using a system of coordinates, and the

angular momentum .li and torque .τ i,ext are described with respect to such a system
of coordinates. The internal forces .fj i between the arbitrary ith and j th particles
generate no change of .L because the internal forces lie along the separation vector
.rj i = rj − ri , resulting on zero contribution, .rj i × fj i = 0. Thus, Eq. 3.10 implies
zero contribution of the internal forces. In addition, the angular momentum of an

isolated system is preserved over time because the external forces are zero in this
case.
. τ ext = 0 ⇒ L = constant (3.11)
A link between the angular momentum in the reference frame of the laboratory, .L,
and the angular momentum in the cm reference frame, .Lcm , is obtained when the
change of coordinates .ri = r'i + rcm and velocities .vi = v'i + vcm is employed in
Eq. 3.9. If we consider that the non-primed variables describe the particles in the
reference frame of the lab, and the primed variables describe the particles in the cm
reference frame, then we have:
E E '
. L = Lcm + mtot (rcm × vcm ) ; mtot = i × p'i ]
mi ; Lcm = i [ri
(3.12)
We employed the facts .r'i × pcm = 0 as .pcm is static for .r'i , and .rcm × p'i = 0 as .rcm
is the origin for .p'i , where the position and velocity of the center of mass are .rcm
and .vcm . In an isolated system, .L and .Lcm are time indifferent. Thus, the angular
momenta in the two inertial reference frames are conserved for the isolated system
of particles. By differentiating the above expression, the relation between torques is
found.
. τ ext = d
dt Lcm + rcm × fext (3.13)
Here, we also employed .(drcm /dt) × vcm = vcm × vcm = 0 and

the facts E
E
d(mtot vcm )/dt = i mi dvi /dt = i fi = fext . We require to associate a torque
.
to the term .dLcm /dt. To do this, we use Eq. 3.10 to establish the relation between
the primed and non-primed variables.
E '
E '
. τ ext = i (ri + rcm ) × fi = τ cm + rcm × fext ; τ cm = i (ri × fi )
(3.14)
The net torque exerted on the particles as observed from the reference frame of the
center of mass is .τ cm . From Eqs. 3.13 and 3.14, it is concluded that:
. dLcm /dt = τ cm (3.15)
The torque exerted on the cm is .rcm × fext .
3.3 Mechanical Energies of a Many-Particle System
The interest now comes to obtain an expression of the energy of a many-particle

system. The process starts from the Newtonian equations of motion which should
be transformed onto an expression of the energy. The equation of motion of the ith
particle is given by:
EN
. mi ai = fi + j =1,i/=j fij (3.16)
The force .fi represents the external force acting on the ith particle, and the
summation on the j th index of .fij gives the total internal force, .f int
i , exerted on
such a particle. The internal force may be obtained from a potential energy function
.Epot by differentiation, .f
int
i = −∇i Epot ({rj }). The function .Epot depends on all
the coordinates of the particles, .{rj }. In the process of transforming the equation of
motion onto and energy equation, it is necessary to multiply each term of Eq. 3.16
by the displacement vector .dri .
EN
. mi ai · dri = fi · dri + j =1 fij · dri ; i /= j
3.3 Mechanical Energies of a Many-Particle System 45
All the equations of motion of the particles should be added up. Before, we illustrate
the way in which the third Newton law is considered in the addition of equations.
Let us deal with the internal and external forces on the ith and j th particles as given
by Eq. 3.16.
. mi ai · dri + mj aj · drj = fi · dri + fj · drj + fij · (dri − drj ) + · · ·
The third Newton law was employed in the last term of the above expression. It is
convenient to use the identities .ai · dri = (dvi /dt) · dri = dvi · vi to introduce the
velocities of the particles and integrate to apply the concepts of work and kinetic
energy.
f vb f vbj fb fb fb
.
vai
i mi vi · dvi + vaj mj vj · dvj = a fi · dri + a fj · drj + a fij · drij + · · ·
The difference .dri −drj is .drij . By similarly dealing with all the particles, we reach
an expression referent to the balance of the kinetic energy .Ekin :
(0)
Ekin − Ekin = Wext + Wint
. E 1 E fb E fb
Ekin = i 2 mi vi2 ; Wext = i a fi · dri ; Wint = N i<j a fij · drij
(3.17)
The kinetic energy .Ekin of the first line is given with reference to the initial kinetic
(0)
energy, .Ekin . This last energy is due to integrating and evaluating the result at the
lower integration limit. The external forces exerted on the particles produce the
external work .Wext . The last expression is the internal work .Wint , due to internal
forces. According to the expression in the first line, any change of the total kinetic
energy comes from the total work .Wext + Wint exerted on the particles. When the
internal forces acting on the particles are conservative, the internal work may be
(0)
written as the difference of internal potential energies, .Wint = Epot −Epot , Eq. 2.35.
The potential energy depends on the interparticle distances and is independent of the
reference frame of coordinates. The proper energy of the system is defined by:
. U = Ekin + Epot proper energy
The work exerted by the external forces on the system results on the change of the
proper energy.
. Wext = U − U (0) ; U = Ekin + Epot (3.18)
The energy increases, .U > U (0) , when the work is exerted on the particle system,
.Wext > 0. If the particles perform work, .0 > Wext , the energy diminishes,
ext = 0, the system is isolated,

.U
(0) > U . When there are no external forces, .W
and the proper energy U of the system is preserved in time. In an isolated particle
system, the energy preservation demands that any change of the kinetic energy
concurrently occurs with a change of the potential energy of identical magnitude
and opposite sign. By choosing a description of the kinetic energy from the position
of the center of mass, and recalling that the internal potential energy is independent
of the reference frame, it is possible to define the internal energy of the system.
. Uint = Ekin,cm + Epot internal energy (3.19)
Problem 3.1 Find the expression relating the proper energy .U = Ekin +
Epot to the internal energy .Uint = Ekin,cm + Epot in a reference frame of
coordinates located at the center of mass.
SolutionEThe kinetic energy given in the reference frame of the laboratory is

Ekin = i mi vi2 /2. The velocity .vi can be written in the form .vi = v'i + vcm ,
.
where .vcm is the velocity of the cm and .v'i is the velocity of the ith particle
with respect to cm. The insertion of the expression of .vi in the kinetic energy
leads to:
E ( )2 E
Ekin = 1
i 2 mi vi' + vcm ; mtot = i mi
.
E ( )2
= Ekin,cm + 1 2
2 mtot vcm ; Ekin,cm = 1
i 2 mi vi'
E
The total primed momentum, . i mi v'i , is zero in the reference frame of the
cm. In this regard, the kinetic energy .Ekin in the reference frame of the
laboratory is written in terms of the internal kinetic energy .Ekin,cm in the
reference frame of the cm, plus a kinetic energy of translation of the cm,
2
.mtot vcm /2. On the other side, the potential energy, .Ep , depends on the particle
distances and exhibits no change with respect to the two different reference
frames. By adding the potential energy to the above expression, the proper
energy U is written in terms of the internal energy .Uint .
. U = Uint + 12 mtot vcm

2 ; Uint = Ekin,cm + Epot
In an isolated system, U is constant, the translation of the system is constant,

2 /2 = constant, and the internal energy .U
mtot vcm
. int is equally constant.
Thereby, the energy is conserved in any inertial reference frame used to
compute the energy.
In a diluted gas at high temperature, the particles have large speeds, and the potential
energy of the particles is small with respect to the kinetic energy. In this case, the
conservation of internal energy is essentially a conservation of kinetic energy. On
the other hand, when the external forces in Eq. 3.18 are conservative, such as a
3.4 Transformation of the Energy Components 47
molecule in a gravitational field, the external forces are obtained from a potential
(0)
energy, .Wext = Epot,ext − Epot,ext , and it is possible to introduce the total energy
E of the system.
. E = U + Epot,ext total energy (3.20)
The preservation of the total energy is formulated from Eqs. 3.18 and 3.20 by .E =
E (0) , or .E = constant. The presence of external forces makes the energy E different
from U .
. E = U + Epot,ext = Ekin + Epot + Epot,ext (3.21)
In the absence of external forces or in an isolated system of particles, the proper

energy U is preserved, namely, .U = U (0) from Eq. 3.18, and in the presence of
conservative external forces, the total energy E is conserved, namely, .E = E (0) .
⎧
⎨ proper energy: U = constant energy no external forces
. particle system internal energy: Uint = constant energy no external forces
⎩
total energy: E = constant energy conservative external forces
3.4 Transformation of the Energy Components
When an interaction between two particle systems occurs, a transformation of the

kinetic energy onto potential energy, and vice versa, may be observed. A way to
evaluate the energy transformation is by introducing the q balance factor. If the
a + E a = E b + E b , and rewritten in the form .E a −
energy is constant, .Ekin pot kin pot kin
b = E b − E a , then the q balance factor is defined as:
Ekin pot pot
. q = Ekin
a − Eb = Eb − Ea
kin pot pot (3.22)
The variables with index b represent the system before the reaction, and the variables
with index a represent the system after the reaction. When .q < 0, there is a lowering
of kinetic energy and an increase of potential energy.
. q<0 ⇒ a < Eb
Ekin kin ; b < Ea
Epot pot
The opposite situation, .0 < q, corresponds to a lowering of potential energy and an

increase of kinetic energy.
. 0<q ⇒ a > Eb
Ekin kin ; b > Ea
Epot pot
In both cases, the interaction is an inelastic collision. When .q = 0, there is no

transformation of the energy components, and the interaction between the two
systems is an elastic collision.
The conservation of physical variables such as the linear momentum, angular
momentum, and energy are formulated in classical mechanics in terms of positions
and velocities of the particles. In this regard, a systems which is suspected to satisfy
a conservation law may be examined from the perspective of positions and velocities
of the particles, since these are basic variables to describe particle systems.
3.5 Energy Balance Equation
There is an additional energy ingredient of macroscopic nature capable of changing

the energy of a particle system. It is called heat and plays an important role
for objects in contact with different temperatures. Heat is a primitive concept in
thermodynamics only determined through observation, arising from the energy
diffusion between bodies with a temperature gradient at their contact points. In
Newtonian mechanics, heat is the average exchange energy between the system
and the surrounding environment due to the individual exchange of energy of the
particles of the system with those of the environment [2]. In statistical mechanics,
heat is the energy flowing between the heat source and the system of interest due to
their different temperatures.
Heat is characterized as an energy in transition, flowing from a body with high
temperature, known as the hot body, to another body with lower temperature, known
as the cold body. Such a type of flow is stated by the zero law of thermodynamics.
The energy transfer occurs via the exchange of kinetic and potential energies,
collisions, radiation, and other energy-transfer channels between two or more
connected systems with a given temperature difference. It is not possible to quantify
the heat of a body because the heat is an energy in transition.
Frequently, the macroscopic environment plays the role of a heat source and is
referred to as the heat reservoir. Heat is associated with forces of non-conservative
nature, namely, those forces which are not deduced from a potential energy function
and act on the particle system. On the other hand, an external work .Wext exerted
on the system is supposed to be generated by conservative forces. Heat, represented
by the symbol Q (of quantity), must be considered in expression 3.18 to properly
establish an energy balance of the closed system.
. Wext + Q = U − U (0) (3.23)
The above expression indicates that a change of energy U is produced by: (i) the
heat flux between the heat source and the system, and (ii) the work performed on
the system or by the system (the system energy may be also changed by adding
or subtracting matter, but it is supposed to be constant in this discussion). Heat
is created to the detriment of other energy forms: it is positive when the particle
system increases its internal energy (heat is absorbed by the system) and negative
3.5 Energy Balance Equation 49
when the particle system reduces its internal energy (heat is given by the system).
The experiment where the mechanical equivalent of heat is established puts in
evidence the transformation of mechanical energy onto heat and vice versa. The
energy cannot be created neither destroyed but transformed onto other energy forms.
The many phenomenological evidences on the transformation of the energy allow
to postulate the energy balance principle, Eq. 3.23, establishing the first law of
thermodynamics. Yet it is possible to derive it from first principles [3].
In addition to the heat, the energy balance demands the consideration of the
external work, .Wext . A differential work, dW , is related to an external force .f
by .dW = f · dr. If the external forces increase the system internal energy, then
the work .Wext is considered to be positive, and it is negative when the system
performs work. The previous proposals on the signs of heat and work follow the
I U P AC conventions.2 The work may be re-defined by introducing .Wsys in the
form .Wsys = −Wext , with the purpose to have a positive sign when the particle
system performs a work on the environment.
. Q − Wsys = U − U (0) (3.24)
Problem 3.2 The motion of a Brownian particle in a fluid is given by the

Langevin expression .m d 2 r/dt 2 = f(t) − ξ mv(t) + g(t), where .f(t) is a
systematic (external) force exerted on the particle, .ξ is the medium viscous
coefficient, and .g(t) is the stochastic punch of the environment on the particle
at the instant of time t. Obtain the energy equation of the particle.
Solution The net force is multiplied by the particle displacement .dr, and
integrated from point A to point B, to evaluate the particle energy in such a
process:
fB fB fB
.
A mv · dv = A f(t) · dr + A [−ξ mv(t) + g(t)] · dr
where the identity .a · dr = v · dv was used. The term on the left is the kinetic
energy difference between points B and A. The first term on the right is the
external work, .Wext , and the second is the energy Q due to the interaction
with the environment, like the friction forces and the medium strikes.
(B) (A)
. Ekin − Ekin = Wext + Q
2 TheInternational Union of Pure and Applied Chemistry (I U P AC) establishes the standards of
names, symbols, units, and more.
There are different pathways to take a system from an initial state to a predetermined
final state, for instance, by changing either the pressure, temperature, or volume. The
thermodynamic pathways convey energy and define the amount of heat and work
given to or received from the particle system. In the process of changing states, the
system may visit intermediate states. The work .Wext and heat Q have no dependence
on the coordinates of the system because they are external factors to the system, and
therefore, they are not considered as state functions of the particle system (under
a mathematical conception, the infinitesimal changes of .Wext and Q are inexact
differentials denoted by .d¯Wext and .d¯Q) [4]. Thus, the particle system in equilibrium
is not characterized by the heat and work;3 however, the difference between the
amount of heat and the amount of work .Q−Wsys defines the internal energy change
.U − U
(0) of the system. The internal energy U depends on the particle coordinates,
and it classifies as a state function of the system (under a mathematical conception,

the infinitesimal changes of U are exact differentials denoted by dU ).
The first law of thermodynamics relates the internal energy changes of the system
to the statistical forces produced by the macroscopic environment. In this respect,
the work and heat represent two forms of exchanging energy between the system
and the environment. Once the system changes its energy and reaches a state of
equilibrium, it is not possible to tell whether such a state was achieved after doing
work, transferring heat, or using both because, in addition to the time irreversibility
of thermodynamic theory, the work and heat are not state functions.
3.6 Statistical and Time Averages of Physical Observables
When a system is constituted by a large number of particles, the physical variables

of the individual particles are not considered observables anymore due to the
complexity in measuring the position and velocity of every particle, except under
particular situations as in the case of a Brownian particle, which is notably different
in geometrical size or inertial mass from the other particles. In a many-particle
system, the averages of physical variables acquire significance, rather than the
individual physical variables of the particles. The averages of the mechanical
variables may be computed in different ways after using either:
• a large number of similar particle systems, called replicas, and performing

a statistical average, also known as an ensemble average, of the mechanical
variables over the replicas
• a single molecular system and performing a time average, also known as a
trajectory average, of the mechanical variables over the trajectory.
The field of physics that deals with many particles, their replicas, and the ensemble
averages is statistical mechanics. According to statistical mechanics, the macro-
3 In statistical mechanics, a system subjected to heat transfer achieves a new level of disorder.
3.6 Statistical and Time Averages of Physical Observables 51
scopic observables of a many-body system are defined by the ensemble averages

on the basic mechanical variables using a large number of replicas of the particle
system. In contrast, the macroscopic observables in molecular dynamics are defined
by time averages of the basic mechanical variables using a single molecular system
evolved during a long period of time. The statistical and time averages are important
because they represent two forms to establish contact between the microscopic and
macroscopic worlds.
Based on the principles of statistical mechanics, the many-particle system is
required to be in the state of thermodynamic equilibrium to apply statistical mechan-
ics. The thermodynamic equilibrium demands that the rate of energy dissipation and
the rate of energy input are balanced and the system-intensive variables, like the
pressure P , temperature T , and chemical potential .μ, are constants over time. The
equilibrium of P , T , and .μ is a way to demand mechanical, thermal, chemical, and
diffusive equilibrium. The system in contact with a reservoir may use a permeable
wall to any of the previous intensive variables to force its homogeneity in space
and time and keep constant values of P , T , and .μ. The reservoir is supposed
to have large dimensions to be little affected by energy changes associated with
the contact with the particle system, or in any case, it should be capable of
rapidly fading away any noticeable energy gradients (such processes have been
formulated in fluctuation-dissipation theorems). The particle system flows to the
state of thermodynamic equilibrium when it is in contact with the reservoir, unless
it is prevented from doing so by external forces. The equilibrium state may be
reached in different stages, with some macroscopic variables possibly achieving the
equilibrium before others, or they all reach the equilibrium almost simultaneously
and in their own ways.
Once the thermodynamic equilibrium is achieved by the particle system, its
thermal fluctuations are small, and then, the mean value of a physical variable may
be identified with the value obtained at the macroscopic scale, in spite of the fact
that at the microscopic level there are ceaseless local energy transformations and
local flows of matter. The induced forces produced in this way give rise to the small
fluctuations observed at the macroscopic scale, yet they are continuously balanced
and dissipated.
The concept of thermodynamic limit plays an important role in systems with a
large number of particles because such a concept defines the required features to deal
with a macroscopic system and eventually allows the application of thermodynamic
theory on the many-particle system. More precisely, in the thermodynamic limit,
the system is considered to be of macroscopic dimensions, .N → ∞, with
natural fluctuations that are small at the macroscopic scale. Still, the size of the
fluctuations is important to determine the stability of the system. In this connection,
the thermodynamic equilibrium corresponds to a macroscopic stationary state,
where the average values of the macroscopic variables show no change over time.
When the average values, for example of the pressure, temperature, and chemical
potential are the same to those of the surroundings, the system stays in the state of
thermodynamic equilibrium insofar as there are no external forces or fields driving
it away from the stationary state.
Summing up, the averages of physical variables are calculated when the system
is in the state of thermodynamic equilibrium, as the macroscopic variables are
stationary and the experiments can be reproduced any number of times. The
ensemble averages obtained in the state of thermodynamic equilibrium correspond
to the measurements performed in the laboratory. Such experimental measurements
are also reproduced by computing the trajectory averages in molecular dynamics.
The equivalence between ensemble averages and trajectory averages is an important
hypothesis, which is briefly analyzed next.
3.7 Ergodic Hypothesis
The equivalence between a time average and a statistical average is postulated and
is known as the ergodic hypothesis. In order to illustrate it, consider a system with
constant N V E values. The discussion follows similar lines to those of Frenkel and
Smit [5]. We depart from a trajectory average to reach an ensemble average. Let
us denote with .Ā the trajectory average of observable A. It is calculated from the
sequence of instantaneous values .A(t1 ), · · · , A(tn ) registered along the trajectory
of the system and when the thermodynamic equilibrium has been achieved in the
molecular dynamics simulation.
En
. Ā(t) = 1
n i=1 A(ti ) (3.25)
When the time of the simulation is sufficiently long and the time step is small
enough, the time variable may be transformed from a discrete to a continuous
variable, and the summation becomes a time integral.
fτ
. Ā(t) = lim 1
0 A(t) dt (3.26)
τ →∞ τ
Problem 3.3 The time step in a simulation is .At = (t2 −t1 ), and the particles
evolve according to the time grid .t1 , t2 , · · · , tn+1 . After assuming a uniform
grid, show that the numerical integration of f , which is a function of t, is:
f tn+1 [ En ]
f (t1 )+f (tn+1 )
.
t1 f (t)dt = At 2 + l=1 f (t1 + lAt)
Find the result of integrating .f (t) when the grid is non-uniform.
Solution It is possible to compute the integral by considering the area of

small rectangles, where a rectangle area is .At[f (ti )+f (ti+1 )]/2. It is a linear
approximation since the function f between .f (ti ) and .f (ti+1 ) is considered
to be a straight line. The sum over all the small areas approximates the total
(continued)
3.7 Ergodic Hypothesis 53
area under the function .f (t).

f tn+1
t1 f (t)dt = 2 [f (t1 ) + f (t2 )] + 2 [f (t2 ) + f (t3 )] + · · ·
At At
2 [f (tn ) + f (tn+1 )]
+ At
.
= 2 f (t1 ) + At
At
[f (t2 ) + f (t3 ) + · · · + f (tn )] + At
2 f (tn+1 ))
[ E ]
= At f (t1 )+f2 (tn+1 ) + nl=1 f (t1 + lAt)
When the grid is non-uniform, it is not possible to factorize the time step, and
we have:
f tn+1 En
.
t1 f (t)dt = 1
2 i=1 (ti+1 − ti ) [f (ti+1 ) + f (ti )]
This numerical integration scheme is known as the trapezoidal integration

method. It is a linear method with deviations proportional to the second
derivative of f . The accuracy is improved when the time step is sufficiently
small [6].
A simulation may be performed in a single large run or in several small runs to

visit all the permissible configurations of the system that are compatible with the
imposed macroscopic conditions NV E. Let us consider several simulation runs.
They are started from different initial conditions .{r(ti ), p(ti )}; nevertheless, they all
are capable of reaching the NV E state of thermodynamic equilibrium of the system.
It is possible to add all the results of the many molecular dynamics simulations in
the thermodynamic equilibrium and obtain the average value .Ā. The simulations
and the summation process are represented by:
{ 1st simulation, N1 steps

' '' '
Ā = 1
M initial condition (1) : A(1) (t = 1) + · · · + A(1) (t = N1 ) +
2nd simulation, N2 steps

' '' '
.
+ initial condition (2) : A(2) (t = 1) + · · · + A(t=2) (t = N2 ) + · · ·
nth simulation, Nn steps }

' '' '
+ initial condition (n) : A(n) (t = 1) + · · · + A(n) (t = Nn )
(3.27)
Fig. 3.1 The sketch on the top illustrates different trajectories in the multidimensional space of
positions and momenta of a system that contains many particles. The trajectories are depicted
with lines, and the entire system is depicted with a ball. The different balls along the trajectories
are snapshots of the system, supposed to be in thermodynamic equilibrium. The trajectories are
generated using molecular dynamics. A single large trajectory or many small trajectories (created
from different initial conditions) cover the multidimensional space of positions and momenta. The
white islands are incompatible regions with the macroscopic observables of the system.
The sketch on the bottom illustrates that, once the multidimensional space is covered, the time has
no meaning, the trajectories may be dismissed (faded away in the sketch), and the time averages
may be replaced by statistical averages, because we have knowledge of the different microscopic
states or replicas which are compatible with the macroscopic state of the system.
In simulations like these, molecular dynamics meets statistical mechanics, and the average process
of physical variables may follow the rules of statistical mechanics
The terms .A(i) (t = 1), A(i) (t = 2), · · · , A(i) (t = Ni ), with .i = 1, · · · , n,

represent the values of A obtained in the ith simulation at the instants of time .t =
1, · · · , Ni . In principle, we obtain all the M configurations, .M = N1 +N2 +· · ·+Nn ,
that produce the NV E macroscopic state. When such configurations are registered
and numbered, the time has no more significance, and expression 3.27 may be
considered to represent a statistical average over the replicas of the system, Fig. 3.1.
In this respect, the time is dismissed, and we only maintain the indexes of A to
identify the values obtained from the different replicas of the statistical ensemble.
3.7 Ergodic Hypothesis 55
. Ā = [A(1) + · · · + A(N1 ) + A(N1 +1) + · · · + A(N1 +N2 ) + · · · + A(M) ] / M

(3.28)
Now, the nomenclature .A(i) indicates the value of observable A in the ith configura-
tion determined from the set of variables .{r, p} of all the particles. Procedures 3.27
and 3.28 establish a point of contact between molecular dynamics and statistical
mechanics. The summation 3.28 over replicas may be avoided by introducing a
weight function that contains the relevant information of the replicas to compute the
statistical average. In statistical mechanics, the statistical average is more precisely
referred to as the ensemble average, denoted in the form .<A> to distinguish it from
the geometric mean .Ā in molecular dynamics, and the weight function is known
as the probability distribution function. This function should be normalized and is
used to compute the average of the observable. When .M → ∞, expression 3.28 is
written as an integral over both the coordinates and momenta of the particles.
f
<A({r, p})> = 1
Z ρN V E A({r, p}) dr
. f (3.29)
Z= ρN V E dr ; r = d 3 r 1 · · · d 3 r N d 3 p1 · · · d 3 pN
The function .ρN V E is the probability distribution function with dependence on

the set of variables .{r, p} of all the particles and is characteristic of the particle
system under the given macroscopic conditions, NV E in our case. The factor Z
is the normalization factor and contains information of the microscopic states, or
microstates, through .ρN V E . By recalling that the average value of the observable A
is expressed in the form .Ā in molecular dynamics and .<A> in statistical mechanics,
the ergodic hypothesis of the particle system is abbreviated by:
. Ā({r(t), p(t)}) = <A({r, p})> (3.30)
The averages of an observable A computed in the form .Ā or .<A> are equivalent.
When the system is in thermodynamic equilibrium, it remains indefinitely in that
stable state, and the ergodic hypothesis is satisfied. Under the action of small
and transient external forces, the system in thermodynamic equilibrium recovers
the stability. However, when the external forces are large, they produce large
fluctuations capable of driving the particle system out equilibrium, changing the
macroscopic parameters that characterize the system. Contrary to the system in the
thermodynamic-equilibrium state, a system in a metastable state may be driven out
of equilibrium by small forces. When the system is in a non-equilibrium state,
the concepts of thermodynamic equilibrium and ergodicity have no application,
making the thermodynamics of the system meaningless. Yet the system in the out-
of-equilibrium state may be investigated with molecular dynamics since it only has
dependence on the fundamental interactions of the particles.
3.8 Breaking the Ergodic Hypothesis
The ergodic hypothesis is characteristic of systems in thermodynamic equilibrium.

In this regard, it is not fully satisfied when the particle system is out of ther-
modynamic equilibrium, since the mean values of physical variables have large
fluctuations and, in consequence, an average value is not properly defined. The
thermodynamic equilibrium of a system is a requirement that in most instances
is difficult to obtain in the laboratory. Nevertheless, the equilibrium condition
may be relaxed by demanding relative temporal equilibrium in such a way that,
after assessing the temporal stability of the macroscopic parameters, any required
measurement of a microscopic variable may be performed without time restrictions
on the physical system (in principle, the time interval of stability implied by
thermodynamics is a relative interval of time). The system may not be required
to be isolated, and the measurements may be performed on an open system when it
is in thermodynamic equilibrium with the surroundings.
The systems with a large number of particles and in thermodynamic equilibrium
show statistical deviations of order .N −1/2 , with N the number of particles. The
systems with a small number of particles like the atomic clusters exhibit large
fluctuations of their properties with respect to the average values of those properties.
Thus, care should be exercised in the study of small systems.4 Some examples of
systems out of thermodynamic equilibrium are those in the instances of a chemical
reaction, a phase transition, and more. Molecular dynamics may be applied to
investigate the behavior of small and large particle systems regardless of those
circumstances.
3.9 Velocity Distribution Function
In a molecular dynamics simulation, the atoms move and each of them has a
velocity. When the system is in thermal equilibrium, the speeds of the atoms satisfy
the Maxwell-Boltzmann speed distribution function:
[ ]3 [ ]
2 −mv 2
. f (v) = √4
π
m
2kB T v 2 exp 2kB T
(3.31)
where m is the particle mass, .kB is the Boltzmann constant, v is the velocity, and T is
the temperature. The most probable speed, .vp ; mean speed, .v; and root mean square
speed, .vrms , deduced from the Maxwell-Boltzmann distribution are, respectively:
√ √ √
. vp = 2kB T /m ; v= 8kB T /(π m) ; v rms = 3kB T /m
4 The clusters are of particular interest because they introduce the molecular size as a new physical
observable. The clusters change their properties with size. Such a feature fades away toward the
macroscopic scale, and therefore, has no counterpart in macroscopic systems.
3.10 Temperature of a System of Particles 57
These speeds are temperature dependent. In a simulation, the initial speeds, usually
referred to as the offset speeds, are frequently assigned from a Maxwell–Boltzmann
distribution function. This is done by taking in random manner the speeds from a
histogram that resembles a Maxwell-Boltzmann speed distribution function, where
the abscissa of the distribution function corresponds to the speed intervals and the
ordinate to the number of speeds that classify in the intervals.
The temperature, pressure, and volume are important features employed to
characterize the states of thermodynamic equilibrium at the macroscopic scale.
The goal of the following sections is to discuss those variables from a mechanical
perspective, namely, in terms of the basic variables of position and velocity of
the particles. The temperature, pressure, and volume may be also evaluated in
statistical mechanics after using the appropriate distribution functions, giving results
consistent with their mechanical counterparts.
3.10 Temperature of a System of Particles
In this section, we deal with the temperature since it is one of the properties that
characterizes the macroscopic state of a many-particle system. Temperature, T , is a
scalar parameter which indicates the ability of a system to give and receive heat from
the surroundings: the higher the temperature of the system, the better the ability to
receive and give energy. Temperature is one of the factors that we use to determine
the thermal equilibrium of the bodies. The concept of temperature is different from
that of heat. The difference may be illustrated with an example: two metallic rods
with equal geometry and different composition may have equal temperatures but
transmitting different amounts of heat to other bodies.
Suppose that a closed system has a higher temperature than that of its environ-
ment. The energy which the system gives to the environment is transferred in the
form of heat. The energy transfer decreases the kinetic energy of the system, cooling
the system down. The opposite situation occurs when the system temperature is
lower than that of the environment, warming the system up. According to our
observations, when body C is in thermal equilibrium with bodies A and B in
separate form, then body A and body B are found in thermal equilibrium when
they are in contact with each other. The zero law of thermodynamics defines the
temperature as a system property by introducing the thermal equilibrium between
bodies. The zero law may be explained from a statistical perspective by using the
equipartition of the energy, which indicates that for a system in thermal equilibrium,
the average kinetic energy of each independent degree of freedom is .kT /2, and
an additional amount of .kT /2 energy for the average potential energy of each
independent degree of freedom, inasmuch as the kinetic and potential energies are
quadratic functions of the independent degrees of freedom. Thus, the temperature
of a system of particles may be obtained from either the average kinetic energy or
the average potential energy of the independent degrees of freedom. Yet in many
particle systems, the potential energy does not necessarily contain quadratic terms
of the independent degrees of freedom, and then, it is convenient to determine the
temperature from the average kinetic energy, which usually involves quadratic terms
of the independent degrees of freedom of motion.
From the phenomenological perspective, we observe that when the average
velocity of the particles increases (decreases), the temperature linearly increases
(decreases). In other words, the temperature and the average kinetic energy exhibit
a linear dependence on each other. Still, it takes different amounts of energy to reach
the same average kinetic energy in different particle systems: it takes some amount
of energy to increase the average kinetic energy of the particles of an ideal gas and
a different amount of energy to reach the same average kinetic energy of a system
of bonded particles, where the potential energy among the particles modifies in an
effective way the inertia of the particles. Thus, according to our observations and
the equipartition theorem, the kinetic expression of the temperature may be obtained
from the microscopic motion of the particles without regard to the potential energy.
/E \
(3N −Nc ) N 1
. <Ekin (t)> = 2 kB T ⇒ T = 2
(3N −Nc )kB i=1 2 mi v2i (t)
(3.32)
There is an equivalence between a trajectory average and an ensemble average
in the above expression (hence, we also use brackets to represent time averages).
The factor 3N is the number of degrees of freedom of the N particles, and .Nc
is the number of constrained degrees of freedom. The kinetic energy is not only
averaged over a sufficiently long period of time but also averaged with respect to
the non-constrained degrees of freedom, .3N − Nc , in Eq. 3.32. Except for the units
introduced by the constant .kB , the temperature may be considered as an energy per
degree of freedom. Such an energy represents the equipartition of the kinetic energy
due to the homogeneous environment in which the particles are found in space and
time for the system in thermodynamic equilibrium. This fact establishes the zero
law of thermodynamics with reference to the kinetic energy per degree of freedom.
The factor .kB is the Boltzmann constant.
. kB = 1.380, 649 × 10−23 J /K.
.kB introduces the proper units to link the energy to the temperature. Apart from
units, if the .kB factor is omitted, then the temperature is of similar order to that of
the kinetic energy per degree of freedom. In this respect, the .kB factor is sufficiently
small (.∼10−23 ) to connect the macroscopic temperature with the microscopic
kinetic energy per degree of freedom. On the other hand, the symbols .mi and
.vi of Eq. 3.32 represent the mass and velocity of the ith particle. Temperature is
measured using the reference frame of the center of mass to avoid contributions of
the translational kinetic energy of the whole system. The temperature difference
between a system and its environment vanishes in time, reaching a thermal
equilibrium and making the average kinetic energy of the system stationary. The
thermal equilibrium should not be confused with the thermodynamic equilibrium
3.10 Temperature of a System of Particles 59
Fig. 3.2 The figure illustrates several temperature plots of a many-particle system with time. The
equilibrium temperatures are 200, 300, and 400 K. In a first stage, the system is cold and eventually
starts heating up. After a period of time, which depends on the system under investigation, the
thermal equilibrium is reached. The thermal fluctuations are relatively small with respect to the
equilibrium temperature but increase with the reference temperature
as in this last case, all the mechanical observables are stationary, while in the
thermal equilibrium, the temperature is stationary. The state of thermal equilibrium
is characterized by a Maxwell-Boltzmann distribution function of particle speeds.
In short, expression 3.32 is compatible with our observations and consistent with
the zero law of thermodynamics.
In the theory of molecular dynamics, the equation of motion of each particle is
integrated at every time step to generate a particle trajectory. The time average of
the kinetic energy is calculated from the sequence of values .Ekin (t1 ), · · · , .Ekin (tn )
for the system in thermodynamic equilibrium.
En
. <Ekin (t)> = 1
n k=1 Ekin (tk ) (3.33)
From the instantaneous values of the kinetic energy, the temperature may be
computed instantaneously, Fig. 3.2.
En EN
. T = 1
n k=0 T (tk ) ; T (tk ) = α 2
i=1 mi vi (tk )/2 (3.34)
Where .α = 2/(3NkB ). The average kinetic energy of an isolated system is

stationary; however, the temperature of a system out of equilibrium is undefined.
Problem 3.4 Consider a gas of N particles with equal masses. Show that
<vrms
.
2 (t)> is proportional to the temperature, where .v
rms is the root mean
square (rms) speed of the particles.
(continued)
E
Solution The kinetic energy is .Ekin = 2
i mi vi /2. If the gas contains
particles with equal masses, then
E it is possible to take the mass out of
the summation, .Ekin = (m/2) i v2i . The summation corresponds to the
2
rms speed of the particles, .vrms = (v21 + v22 + · · · )/N . Thus, .Ekin =
2
Nmvrms /2. On the other hand, the temperature is proportional to the kinetic
2 , .T = α<E (t)> =
energy, Eq. 3.32. Therefore, T is proportional to .vrms kin
2 2
(N αm/2)<vrms (t)>. By solving for .vrms in terms of T , we have:
< 2 >
.
m
2 vrms (t) = 3
2 kB T
In this particular case, we may think on the system temperature as the

temperature of a single particle with internal structure, mass m, and speed
.vrms .
3.11 Temperature Scaling as a Thermostat
In many simulations, it is necessary to stabilize the temperature or change it to

a different value. For instance, it is necessary to preset the reference temperature
T in simulations with fixed NV T values, where N is number of particles and
V is the system volume. In the laboratory, the temperature is regulated using a
thermostat, which is a device that automatically tunes the temperature by opening
and closing switches to permit contact of the system with a heat reservoir. In
molecular dynamics, there is no such device, but it may be simulated to control
the heat flow and set the required temperature. The simplest implementation of
a thermostat in molecular dynamics is based on an approach where the particle
velocities are scaled [7]. The scaling of the velocities changes the kinetic energy,
which in turn determines the system temperature. In order to implement this
thermostat, the scale factor .λ is introduced using the fact that the kinetic energy
is a homogeneous function of second order in the velocities.5 It is also convenient
to define the old and the new temperatures.
/E \ /E \
N 1 N
. Tnew = 2
3N kB i=1 2 mi (λvi )2 ; Told = 2
3N kB
1
i=1 2 mi v2i (3.35)
The value of .λ is determined from the new and old temperatures of the system in the
form .Tnew = λ2 Told . The heat flow between the heat reservoir and the particles is
regulated by .λ. The quantity .AT = Tnew − Told is the temperature change produced
by the thermostat.
5f is a homogeneous function of order n when it satisfies the expression .f (αv) = α n f (v), where
n is an integral number and .v is the independent variable.
3.12 Temperature Fluctuations 61
[/E \ /E \]
N N
. AT = 2
3N kB
1
i=1 2 mi (λvi )2 − i=1
1
2 mi v2i = (λ2 − 1) Told (t)
The new temperature .Tnew may be achieved in different ways during the simulation.
For example, .Told may be changed suddenly by scaling the velocities in a single
time step, or it may be gently changed in several time steps of the dynamics by
fractions of the scale factor until reaching .Tnew . In the scaling process, it may be
also convenient to distinguish particles with high and low speeds, with the purpose
of selectively applying the scale factor (or several scale factors) to the fastest and
slowest particles of the system, and homogenize the particle speeds. The scale factor
may also increase or reduce the magnitudes of the velocities in a predetermined
direction solely. Yet the way of introducing the thermostat effects is not genuine
enough because there is no consideration of the temperature fluctuations affecting
the particle system. In this respect, the scaling of velocities is only used to relatively
rapidly achieve the new temperature of reference from an old temperature of the
particle system, without employing the Maxwell-Boltzmann distribution function
in the assignation of the new particle speeds.
3.12 Temperature Fluctuations
The thermal fluctuations are important to determine the stability of the system,
Fig. 3.3. They are evaluated by computing the root mean square deviation (rmsd)
of the temperature with respect to the reference temperature .Tref during a period of
m time steps of the simulation.
/E
m
. rmsd(m) = i=1 (Ti − Tref )2 /m (3.36)
The temperature at the ith time step is .Ti . The rmsd factor, which has units of
temperature and depends on the number m of time steps, shows the temperature
variations of the system. In a simulation with say .n = 16,000 steps, the rmsd factor
Fig. 3.3 The rmsd factor, Eq. 3.36, is shown over time. The simulation consists of 16,000 time
steps with a reference temperature of 200 K. The rmsd factor is calculated with values of m equal
to 100, 300, 500, and 1000. The rmsd plots with small values of m show more structure than the
rmsd plots with large values of m
Fig. 3.4 The time evolution of the temperature is illustrated. The reference temperature of the
simulation is 200 K. The thin line running along the temperature curve represents the average
temperature every 100 steps of the simulation. The color band in the lower part of the plot is
the temperature spectrum. It shows the intervals of time where the particle system is hot (.T >
Tref + AT , red color), cold (.T < Tref − AT , blue color), or in the acceptable temperature range
(.T e [Tref − AT , Tref + AT ], green color) for a user-defined temperature range .At
may be computed every .m = 100 steps (.n >> m) and plotted to observe the thermal
variations in time. When m is small, the rmsd factor shows structure referent to
the partial temperature deviations. By using a large number m, the rmsd factor
averages the partial temperature deviations and shows less structure, Fig. 3.3. In
the region of thermodynamic equilibrium, the rmsd factor shows relative stability.
A complementary way to visualize the temperature deviations consists on plotting
a temperature spectrum for better analysis of the simulation, Fig. 3.4, or creating
a histogram of the particle speeds, which is expected to resemble a Maxwell-
Boltzmann distribution function when the system reaches the thermodynamic
equilibrium.
Once the system of particles is thermally equilibrated, it stays in an isothermal
state, and the sampling of important physical variables may be performed. The
sampling may follow the Nyquist-Shannon theorem, which indicates that a sampling
frequency equal to twice the vibrating frequency of the fastest-moving particles
should be considered to obtain accurate information of the physical variables with
time. For instance, if we have the behavior of a variable like .y(t) = cos(2π t) +
cos(4π t) + cos(5π t), then the corresponding angular frequencies are .ω1 = 2π ,
.ω2 = 4π , .ω3 = 5π . By considering that .ω = 2πf , the maximum frequency is
.fmax = f3 = 5/2. No further information is acquired by increasing the sampling
rate beyond twice the system bandlimit. Based on the theorem, it is possible
to acquire important information from the simulations with minimal frequency
sampling, .fsamp = 2fmax .6
6 Harry Nyquist in 1928 and Claude Shannon in 1949 proposed the theorem after investigating
telegraph communications in a transmission channel. According to the theorem, a bandlimit

3.13 Pressure and Volume 63
3.13 Pressure and Volume
When a system with a large number of particles is confined in a sealed container,

the volume of the container determines the confining volume V of the particle
system. The volume is linked to the pressure. Inside the container, the particles
are constantly interacting and colliding with the walls of the container. The time
interval between successive collisions of the particles with the confining walls is
sufficiently small from the macroscopic perspective that every wall is essentially
subjected to a constant force. The constant force originates a pressure P on the wall
and is calculated by:
. P = f · n̂/A (3.37)
The surface area where the force is acting on is A, and the direction perpendicular
to the surface area is indicated by the unit vector .n̂. The definition of pressure 3.37
has statistical character as it depends on the net force and the total surface area. In
the description of a movable piston, the pressure is written in terms of the external
work by using the identity .dWext = −f · n̂ dl.
. P = − dWext
dV ; dV = Adl (3.38)
The quantity dl is the infinitesimal displacement of the wall produced by the force
f. By writing .Wext = U − U (0) , Eq. 3.18, the pressure is expressed in terms of the
.
energy U of the system.
. P = − ∂U
∂V
(3.39)
When the energy U in terms of the volume V is known, it is possible to compute

the pressure. In a compression process, the volume changes from an initial volume
.Vini to a final volume .Vf in . With these limits, the external work is calculated by
integrating the pressure, Eq. 3.38,

f Vf in
. Wext = − Vini P (V ) dV (3.40)
One of the simplest situations occurs when the pressure is constant and the change
in volume is ascribed to the temperature. In this case, the external work is .Wext =
P (Vini − Vf in ). In other occasions, under conditions of constant temperature and
constant amount of matter, the pressure is found to be inversely proportional to
the volume, and thus, the pressure increases when the volume decreases and vice
versa. In addition, if it happens that for any two points the identity .P1 V1 = P2 V2
function can be reconstructed when the sampling frequency .fsamp is equal to or greater than twice
the bandlimit .fmax of the signal, .fsamp ≥ 2fmax . The theorem is used to sample continuous
(analog) signals and produce reliable discrete (digital) signals.
is satisfied, which is one of the features of an ideal gas, then we have .P V = c

or equivalently .P = c/V , where c is a constant. This expression is acceptable
for systems under moderate pressures and temperatures. The Boyle-Mariotte law is
obtained by integrating the pressure, Eq. 3.40, with .P = c/V .
. Wext = c ln(Vini /Vf in ) (3.41)
The deviations from the Boyle-Mariotte law occur under extreme conditions. For
instance, at high temperature, the relativistic effects are important, and at high
pressure and low temperature, the quantum mechanical effects are important. At
moderate pressures and temperatures, the molecules behave in classical ways
because the interparticle distances are greater than the thermal de Broglie wave-
length, which is considered a reference parameter to distinguish quantum particles
from classical ones.
. λT = h/(2π mkB T )1/2 (3.42)
Problem 3.5 Consider a system of confined diatomic molecules. The follow-

ing two expressions are equations of state for the system of particles. The first
expression deals with molecules and the second one with atoms.
( ) ( )
1/3 1/3 1/3
Emol = a + b exp αVmol + c Vmol exp αVmol molecule
. ( ) ( )
Eatm = a ' + b' exp α ' Vatm + c' Vatm exp α ' Vatm
1/3 1/3 1/3
atom
Show their relation through the energy parameters, and demonstrate that the
pressures derived from them are the same.
Solution If the expressions are equivalent, then we should find a transforma-

tion between the set of parameters a, b, c, .α of the molecular case and the set
of parameters .a ' , .b' , .c' , .α ' of the atomic case. We depart from the fact that the
number of particles and volumes is related by:
. natm = 2 nmol ; Vatm = V

natm ; Vmol = V
nmol = 2V
natm = 2 Vatm
When .Econf is the energy of confinement, a relation between the energies of

the system is established.
Econf Econf
. Eatm = natm ; Emol = nmol ; Emol = 2 Eatm
(continued)
3.14 The Virial and the Equation of State 65
The identities .natm = 2 nmol , .Vmol = 2 Vatm , and .Emol = 2 Eatm are used
to determine a connection between the equations of state. We start by writing
the energy of the molecules .Emol in terms of the volume .Vatm .
( ) ( )
1/3 1/3 1/3
. Emol = a + b exp 21/3 αVatm + 21/3 c Vatm exp 21/3 αVatm
If we consider that .2 Eatm = Emol , then the coefficients .a ' = a/2, .b' = b/2,
c' = 21/3 (c/2) and .α ' = 21/3 α may be defined. In this way, we recover .Eatm
.
from .Emol .
( ) ( )
Eatm = a ' + b' exp α ' Vatm + c' Vatm exp α ' Vatm
1/3 1/3 1/3
.
The pressure is calculated as:
. Pmol = − ∂E ∂(2Eatm )
∂Vmol = − ∂(2Vatm ) = Patm
mol
The pressures to which the molecules and atoms are subjected to are the same.
3.14 The Virial and the Equation of State
The purpose of this section is to deduce a relation between the kinetic energy of a
system of particles and the forces acting on the particles. Such a relation is the virial
theorem. The virial theorem allows to formulate an equation of state of the system
in terms of the pressure, volume, and temperature. The departure point is from the
scalar quantity A.
E
. A= i mi vi · ri (3.43)
The virial applies to systems in equilibrium and out of thermodynamic equilibrium

since A depends on the basic variables of position and velocity. It is necessary to
consider the time derivative of A to obtain the forces and kinetic energy of the
particles.
E E
.
d
dt A = i mi ai · ri + i mi vi2
It is convenient to consider the forces over the ith particle. This particle is subjected
to the internal force .fij from the j th particle and the external force .fi as well. If we
consider that the j particle is in turn subjected to the force .fij of particle i, then we
have terms like .fij · ri + fj i · rj , which can be written
E as .fij · (ri − rj ) after using the
third Newton’s law. The second contribution, . mi vi2 , is proportional to the kinetic
energy .Ekin . The expression becomes:

E E
.
d
dt A = 2Ekin + i fi · ri + i,j fij · rij ; rij = ri − rj
We are interested in the time averages of every term. The average term of the left is
zero.
E E
. <Ekin > = − 12 − 21 < i,j fij · rij > (3.44)
Problem 3.6 The physical observable A is a bounded function with time.

Show that the time average of .dA/dt is zero.
Solution The observable A does not grow unlimited with time because it
is a bounded function. The time average of the derivative .dA/dt demands
integration over time.
fτ
. < dt
d
A> = lim 1 d
0 dt A dt = lim (A − A0 )/τ = 0
τ →∞ τ τ →∞
The result is zero because .τ grows without limits, .τ → ∞.
Expression 3.44 is one of the several versions of the virial theorem. It relates the
kinetic energy to the external and internal forces. The following goal is to use the
virial theorem to establish the equation of state of the particles. The equation of
state is important as it relates the pressure P , volume V , and temperature T of
the particle system. The parameters P , V , and T are described using the positions
and velocities of the particles, and therefore, are related to each other. In order to
formulate the equation of state, the particles are assumed to be confined in a cubic
box of length .l, with one of the vertex located at the origin and three edges resting
on the x, y, and z axes. The six faces forming the container are identified with the
planes, Fig. 3.5.
ABGH : (x, y, 0) ; BCF G : (l, y, z)

. ABCD : (x, 0, z) ; EF GH : (x, l, z) (3.45)
ADEH : (0, y, z) ; CDEF : (x, y, l)
When a particle collides with a wall, the wall exerts a force on the particle in the
perpendicular direction to the wall surface. The contribution to the virial of a particle
colliding with surface ABGH is null because .f · r = −(0, 0, f ) · (x, y, 0) = 0.
The negative sign comes from a force pointing in opposite direction to the particle
motion. Similar results are obtained for the surfaces ABCD and ADEH . The
3.14 The Virial and the Equation of State 67
Fig. 3.5 The figure shows a particle confined in a cubic box. The position vector of the particle
is .ri . When the particle reaches a wall, say the ABGH surface, the position of the particle is
.ri = (x, y, 0), and the force exerted by the wall on the particle is .fi = (0, 0, −f )
contributions to the virial of the particles colliding on the other surfaces are not
zero. For instance, .f · r = −(f, 0, 0) · (l, y, z) = −f l for the BCF G surface.
The summation of the forces exerted by the walls BCF G, EF GH and CDEF
is .−3f l = −3P V , where the identities .P = f/A, and .V = l3 = Al were used.
The result is inserted in the term of the external forces in the viral expression. By
solving for P V , we have:
E
. P V = 23 <Ekin > + 13 < i,j fij · rij > (3.46)
When the system is in thermodynamic equilibrium, the average kinetic energy is

proportional to the temperature T of the system, Eq. 3.32.
E
. P V = NkB T + 31 < i,j fij · rij > (3.47)
This is the equation of state of the system. It establishes a relation between P , V ,

and T , where the internal forces also play an important role. When the contribution
of the forces is negative (positive), the pressure is reduced (increased). When the
contribution of the forces is sufficiently small, the equation of state of an ideal gas
is obtained.
. P V = NkB T (3.48)
Several important results are deduced from the ideal gas law when either the
temperature, pressure, or volume is constant. For two points, .(P1 , V1 , T1 ) and
.(P2 , V2 , T2 ), of the particle system, under the consideration that .NkB is constant,
we establish different laws:
T1 = T2 ⇒ P1 V1 = P2 V2 Boyle’s law
. V1 = V2 ⇒ P1 /T1 = P2 /T2 Gay-Lussac’s law (3.49)
P1 = P2 ⇒ V1 /T1 = V2 /T2 Charles’ law
References
1. M. Alonso, E.J. Finn, Physics (Pearson Prentice Hall, Essex, England, 1992)
2. M. Alonso, E.J. Finn, Física, vol. 1: mecánica (Fondo Educativo Interamericano, Mexico 1970).
Sect. 9.10, pp. 272–274
3. E. Hernández-Huerta, R. Santamaria, T. Rocha-Rinza, Thermodynamics from Lagrangian
theory and its applications to nanosize sytems. Molec. Phys. 119, Issue 14 (2021), e1940333.
https://doi.org/10.1080/00268976.2021.1940333
4. J. Eloranta, Classical Thermodynamics (Department of Chemistry and Biochemistry, California
State University at Northridge, California, 2009). http://www.csun.edu/~jeloranta/, Sect. 2.3, pp.
47–55
5. D. Frenkel, B. Smit Understanding Molecular Simulation: From Algorithms to Applications
(Academic Press, San Diego, 1996)
6. K.E. Atkinson, An Introduction to Numerical Analysis (Wiley, New York, 1978), Chapt. 5
7. P.H. Hünenberger, Thermostat algorithms for molecular dynamics simulations, in Advanced
Computer Simulation: Advances in Polymer Science, ed. by C. Holm, K. Kremer, vol. 173
(Springer, Heidelberg, 2005)
Mechanical Descriptors
4
Many particle properties change their values in space and time in molecular
dynamics. In this regard, it is of interest to introduce additional descriptors for the
analysis and characterization of the particle system in the simulations. We discuss
the caloric curve from an energetic perspective. A way to evaluate the fluctuations
and deviations of the particle positions with respect to reference points is presented
with the purpose to determine the changes in the orientations of the molecules,
assess the structural changes, and help disclose phase transitions. The correlation
functions are discussed since they are tools to reveal the interconnection of physical
variables in space and time.
4.1 Caloric Curve
When the total energy of the particle system in terms of the temperature is
determined, we obtain the caloric curve. The total energy E is given in terms of
the kinetic energy and potential energy.
.<E(T )> = <Ek > + <Ep >
The brackets represent time averages of .Ek and .Ep over the simulation with
temperature T . The phase transitions may be monitored by means of the caloric
curve. A direct change of .<E(T )> implies a first-order transition, and a change in the
derivative of .<E(T )> corresponds to a second-order transition. At low temperatures,
the system is in the solid-like region, with a harmonic behavior of the particles.
When the temperature increases, the total energy increases linearly, and as the
temperature is further increased, the particle system reaches a transition region
where the slope of .<E(T )> changes. The plot of .<E(T )> vs T shows a leap because
the liquid phase grows at the expense of the solid one plus the actions of the
external factors. The leap corresponds to the melting point (a first-order transition),
where the solid transforms onto a liquid due to the fusion heat. There may be

70 4 Mechanical Descriptors
an overestimation of the melting temperature due to a response lag (or hysteresis

effect) of the interacting particles. In the liquid-like region, the caloric curve keeps
increasing by increasing the temperature.
4.2 Interatomic Distance Fluctuations
An additional descriptor of interest is .δL, which defines the interatomic distance

fluctuations.
EN /
. δL = 2
N (N −1) i>j <rij >−1 <rij2 > − <rij >2 (4.1)
The quantity .rij is the distance between the ith and j th particles, the number of
atoms in the molecule is N , and the brackets indicate time averages. By assuming n
steps in the molecular dynamics simulation, the averages are numerically computed
in the form:
/ \ [ ]
. r
k = r k (t ) + r k (t ) + · · · + r k (t ) /n ; k = 1, 2 (4.2)
ij ij 1 ij 2 ij n
The quantity .δL of Eq. 4.1 gives the fluctuations of the interatomic distances. It
is also recognized as the Lindemann index [1]. It is important to: (i) elucidate the
disorder provoked by the temperature effects in molecules, like the proteins, (ii)
determine the moments of melting of a cluster or solid, and (iii) establish the phase
transition of a system of particles. The parameter .δL grows quasi-linearly with an
increase in temperature in the solid-like region. However, in the moments of the
solid-liquid transition, the particles acquire high kinetic energy to overcome the
potential energy barriers, and the interatomic distances increase rapidly, producing
a sharp growth of .δL. The parameter .δL is stabilized after some time in the liquid
region.
Some specific values of the Lindemann index have been proposed as threshold
values to determine the solid-liquid phase transitions. The value .δL = 0.1 puts in
evidence the fusion of some macroscopic solids; nevertheless, that value does not
necessarily apply to finite size systems like the atomic clusters. A practical way to
establish the melting of a system of atoms is to consider values of .δL in the interval
.[0.20, 0.25]; this interval was determined experimentally in crystals. The melting is
/
supposed to occur at a temperature T where the average amplitude . <A2 > of the
particle vibrations is greater than the product .<rij >δL/2.
/
. if <A2 > > (<rij > δL)/2 ⇒ melting of the particle system (4.3)
The Lindemann criterion is equally applied to a double DN A helix when, for

instance, it is unfolded into separate strands at high temperature, resembling a type
of melting process.
4.4 Orientational Order Parameter 71
4.3 Root Mean Square Deviation of Positions
Another important parameter to describe the structural changes of the molecules

is the root mean square deviation (rmsd) of positions. The rmsd evaluates the
statistical deviation of a molecular conformation at the instant of time t from a
reference molecular conformation taken at the instant of time .t0 .
[ EN ]1/2
. rmsd(t) = 1
N i=1 [ri (t) − ri (t0 )]
2 (4.4)
The vector .ri (t) is the position of the ith particle at time t, and the reference
position of this particle is .ri (t0 ). The summation runs over the number of particles
in the system, N. The rmsd is equally applied to other quantities, such as the
atomic velocities, forces, etc. The rmsd is used to estimate the structural similarity
of different molecules, which should be translated and rotated to maximize the
position coincidence. In the case of proteins, the .α carbons may be considered in
the evaluation of the structural similarity, because they form the protein backbone.
The rmsd is small for atoms in the solid phase; it rises linearly for atoms in the
liquid phase with time and is stabilized after a long period of time.
4.4 Orientational Order Parameter
In molecular dynamics, the molecular structures continuously change. The struc-

tural changes may be evaluated using the orientational order parameter. It has the
form of a quadrupole expression and employs the vector direction between atom
pairs to define the orientational order. The unit direction vector .ûij between the ith
and j th particles is:
ri (t)−rj (t)
. ûij (t) = |ri (t)−rj (t)|
The vector .ûij (t) defines the axis of the atom pair at the instant of time t. The
reference vector .n̂ij (t) is required to gauge the orientation of .ûij . The reference
vector .n̂ij is defined in terms of the time average of .ûij .
. n̂ij = <ûij (t)> / |<ûij (t)>|
The orientational order parameter for the pair ij of atoms with the form of a
quadrupole expression is [2]:
/ \
. Oij = 3
2 [ûij (t) · n̂ij ]2 − 1
3 (4.5)
The parameter .Oij involves the scalar product of the vector of interest .ûij with
the reference vector .n̂ij . The coefficients in the definition of .Oij introduce the
normalization factor. In a molecule, there may be many atom pairs, and it is

convenient to consider them by summing over all atom pairs, .Np .
E
. O = <Oij > = 1
Np i/=j Oij (4.6)
When the structural changes of the molecule are small, the vectors .ûij remain
essentially the same, and .<(ûij · n̂ij )2 > = 1, resulting on a maximal orientational
order, .O = 1. When the pairs of atoms point with equal probability in every
direction, the result .<(ûij · n̂ij )2 > = 1/3 is obtained, producing a minimal value,
.O = 0, characteristic of a low orientational order of the particles. When all the
vectors .uij point in the same direction (.n̂ij = n̂), the nematic order parameter is
reproduced by Eq. 4.6 [3].
4.5 Probability Distribution Functions
Suppose a system of N particles in the fluid state and confined in a volume V . The
distribution function of the N particles is .ρN , which depends on the coordinates
.(r1 , r2 , · · · , rN ). The function .ρN is normalized to unity.
ff
. ρN (r1 , r2 , · · · , rN ) dr1 dr2 · · · drN = 1 (4.7)
The reduced distribution function of order k is obtained from .ρN by integration.

N!
ff
. ρk (r1 , r2 , · · · , rk ) = (N −k)! ρN (r1 , r2 , · · · , rN ) drk+1 drk+2 · · · drN (4.8)
In particular, the one-body distribution function is obtained after integration over all
variables, except one of them.
ff
. ρ1 (r) = N ρN (r, r2 , · · · , rN ) dr2 · · · drN (4.9)
The one-body distribution function is the fluid density, .ρ1 (r) = ρ, which is constant.
The two-body distribution function is similarly computed.
ff
. ρ2 (r1 , r2 ) = N(N − 1) ρN (r1 , · · · , rN ) dr3 · · · drN (4.10)
The two-body distribution function in terms of the one-body distribution function is

written as:
. ρ2 (r1 , r2 ) = ρ12 g(r) (4.11)
We have implicitly defined the correlation function, .g(r), in terms of the distance
vector .r = r2 − r1 . The correlation function gives the probability to find a
number of particles in volume .dr around position .r when a different particle, say
4.5 Probability Distribution Functions 73
Fig. 4.1 An arbitrary particle of the system is taken as a reference point to make interparticle-
distance measurements. The process is repeated on all the particles to have many reference points.
By plotting a histogram of the whole set of distances, the average radial packing of particles around
any particle is determined. The function describing the packing is the correlation distribution
function, .g(r). It is interpreted as the probability to find a number of particles in volume .dr
around position .r when a different particle is located at .r = 0. The correlation distribution function
frequently exhibits peaks, indicating the local particle density and neighbor abundance
a central reference particle, is located at .r = 0, Fig. 4.1. The correlation function

characterizes the structure of the system as either a solid, amorphous solid, liquid,
or gas, because g provides the spatial distribution of the particles, which in turn
depends through the internuclear distances .rij on the temperature. For a crystal,
the correlation function g exhibits separate peaks not connected to each other (like a
half-wave rectified electric signal) and is related to the structure factor of the crystal.
For a liquid, g reveals a curve with structure like that in Fig. 4.1. For a gas, g shows
a continuous curve with almost no structure (like a DC electric signal with small
ripples).
When the correlation function is E known, it helps describe the interactions of two-
body potentials. Suppose that .U = i,j u(ri , rj ), .i < j , describes the interactions
between particle pairs. The average interaction energy is computed by:
E ff
. = i,j u(ri , ri ) ρN (r1 · · · rN ) dr1 · · · drN (4.12)
i<j
The integration variables are dummy variables (they play the role of placeholders
solely), and the summations can be reduced to a single term multiplied by the
binomial coefficient .N (N − 1)/2, which accounts for the total number of pairs
without repeating any of them.
N (N −1) ff ff
 = 2 dr1 dr2 u(r1 , r2 ) ρN (r1 · · · rN ) dr3 · · · drN
.
ff
= 1
2 ρ2 (r1 , r2 ) u(r1 , r2 ) dr1 dr2
The two-body distribution function may be written in terms of the correlation

function as .ρ 2 g. If the interaction potential u depends on the particle distance
.r = |r2 − r1 |, then it is possible to perform an additional integration to finally
determine the average potential energy.

ff
 = 1
2 dr1 dr2 ρ 2 g(|r2 − r1 |) u(|r2 − r1 |)
.
f∞ f∞ (4.13)
N2 2π N 2
= 2V 0 g(r) u(r) dr = V 0 g(r) u(r) r 2 dr
In order to investigate the asymptotic limit of the correlation function, let us again
consider N particles confined in volume V . The fraction of particles .<nr,r+dr > in
the volume element dV , with respect to a central reference particle positioned at a
distance r, is:
. <nr,r+dr > = N
V g(r)dV
The volume element dV , keeping linear terms in dr, is:
. dV = 3 [(r
4π
+ dr)3 − r 3 ] = 4π 2
3 (3r dr + 3rdr 2 + dr 3 ) ≈ 4π r 2 dr
By using the result .dV = 4π r 2 dr, we observe .g(r) = (V /N)<nr,r+dr >/dV . When
.<nr,r+dr > → N , we have .dV → V , and therefore .g(r) = 1. In extended systems
.g(r) tends to 1 as r increases [4]. In atomic clusters, .g(r) tends zero as r increases
because clusters have finite sizes.
E
Problem 4.1 Write the virial equation of state .P V = N kB T + < N i,j =1 rij ·
fij >/3,EEq. 3.47, in terms of the correlation distribution function .g(r), and
.U = i,j u(ri , rj ), .i < j .
E
Solution
E By considering conservative forces . i ri · ∇i U =
i<j rij ∂u(rij )/∂rij , and following similar steps in the deduction of
., we have:
E ff ∂u(rij )
P V = N kB T − 1
3 i,j rij ∂rij ρN (r1 · · · rN ) dr1 · · · · · · drN
i<j
( )
2 f ∂u(r)
. = NkB T − 1
3
N
2V g(r) r ∂r (4π r 2 ) dr
2π N 2
f∞ ∂u(r)
= N kB T − 3V 0 g(r) r 3 ∂r dr
4.6 Correlation Functions 75
4.6 Correlation Functions
The particles interact and move in a dynamic system, and the physical variables
continuously change their magnitudes in space and time. In many instances, two or
more physical variables may be correlated, and it is difficult to observe their relation
due to the large number of particles in an apparent chaotic motion. In this respect, it
is important to discuss the way to evaluate the correlation of the physical variables
and classify them according to their dependencies.
The correlation of the physical variables may be of spatial or temporal nature.
The time correlation function .CAB gives the interconnection of the physical
variables .A and .B by evaluating .A at the instant of time .t1 and .B at the instant
of time .t2 and computing the average of their scalar product.
. CAB (t1 , t2 ) = 1
3 <A(t1 ) · B(t2 )> (4.14)
The time may be replaced by a spatial coordinate to define a spatial correlation

function. The scalar product in the correlation function 4.14 is written as:
. A · B = |A(t1 )| |B(t2 )| cos θ
The angle between vectors is .θ . The scalar product is positive when both vectors
point in the same direction and negative when they point in opposite directions.
Thus, the correlation function considers the positive and negative values of .A(t1 ) ·
B(t2 ) at different instants of time in the simulation. To some extent, the correlation
function establishes the degree of similarity of the physical variable .A with the
physical variable .B, with some time delay between such variables included. The
variables .A and .B are dissimilar when there is no privileged orientation or all the
orientations between them are equally possible in an isotropic environment. In this
case, the value of variable .A implies no value of variable .B and vice versa. In
general, the correlations fade away with many collisions of the particles, and in
this connection, the correlation function gives the time persistence of the .A ↔ B
relation before it vanishes.
The brackets of Eq. 4.14 indicate either an ensemble average or time average. When
we deal with an ensemble average, Eq. 4.14 is [5]:
f
CAB (t1 , t2 ) = <A(t1 )B(t2 )> = r f (r) A(r; t1 ) B ∗ (r; t2 ) dr
.
r = q1 , · · · , q3N , p1 , · · · , p3N ; dr = dq1 · · · dq3N dp1 · · · dp3N
(4.15)
The joint probability distribution function is f , and the physical observables are
A and B. The asterisk on B signals the complex conjugate of that variable. The
distribution function defines the probability of finding the observables A and B with
values in .[A, A + dA] and .[B, B + dB]. The integral is performed over all the phase
space variables, when the system is in thermodynamic equilibrium. Equation 4.15
may be compared with an ensemble average over a single variable.
f
. <A(t)> = r A(P ) f (P ) dr (4.16)
The phase-space point .P = (q1 , · · · , qn , p1 , · · · , pn ), where .qi and .pi are

generalized coordinates of particle i, determines the microscopic state of the
classical system. Function f is the canonical distribution function for the system
with fixed number of particles, volume, and subjected to a given temperature.
The product .f (P ) dr = f (q1 , · · · , qn , p1 , · · · , pn ) dq1 · · · dqn dp1 · · · dpn gives
the probability to find particle i with coordinates between .qi and .qi + dqi and
momentum between .pi +dpi , where .i = 1, · · · , n [6]. The correlation function 4.15
may be considered a generalization of the ensemble average 4.16 as Eq. 4.15
contains a pair of observables with their respective times. When the system is
ergodic, the ensemble averages .<A> and .CAB (t1 , t2 ) may be replaced by time
averages obtained from a molecular dynamics simulation.
fT
Ā = lim 1
0 A(t) dt
T →∞ T
.
fT (4.17)
CAB (t1 , t2 ) = A(t1 )B(t2 ) = lim 1
0 A(t + t1 )B ∗ (t + t2 ) dt
T →∞ T
The function .CAB is the outcome of averaging the product AB over the time interval
of thermodynamic equilibrium. The averaging process considers possible retarded
effects between A and B by dealing with times .t1 and .t2 . The difference .t2 − t1 is
a time delay in the response of every particle to the forces produced by the other
particles, which is typical of crowded environments. The correlation function may
exhibit a cause-effect relationship between the variables A and B. Yet care should
be exercised because the correlation of variables does not necessarily imply that one
variable causes an effect on the other.
By using the correlation functions, it is possible to tell when the motion of a
particle has a random aspect. Thus, some of the important applications of correlation
functions are given on the characterization of particle dynamics in environments
with fluctuations in the density, temperature, pressure, etc. For instance, consider a
body in a fluid. By supposition, it has bigger size than the sizes exhibited by the
fluid particles. We barely have predictive capacity of the body’s trajectory because
the body is constantly kicked by a myriad of surrounding fluid particles, making the
body describe a random trajectory. Such a trajectory has little meaning to us. In this
regard, a statistical characterization of both the environmental forces acting on the
body and the body’s trajectories is convenient.
The correlation function .CAB is also distinguished with the name cross-
correlation function. When the variables A and B in the cross-correlation function
are the same, that function is called autocorrelation function and is denoted
with a single index, .CA . The autocorrelation function is the self-correlation of
A at different instants of time or different spatial positions. In general, transport
quantities such as the diffusion, viscosity, and thermal conductivity of a fluid have
4.6 Correlation Functions 77
their own representations in autocorrelation functions. For instance, the diffusion

of a particle is computed after evaluating the autocorrelation function of velocities.
Such autocorrelation functions are able to characterize stochastic processes in
stationary states. The general scheme that establishes the relations between the
autocorrelation functions and the transport quantities is the Green-Kubo approach.
The autocorrelation function of variable A in the quantum mechanical case is:
E
. CAB (t1 , t2 ) = n pn <n|A(t1 )A(t2 )|n> ; pn = Z −1 exp[−En /(kB T )]
(4.18)
En is the system energy in the nth state, T the equilibrium temperature,
.
B the Boltzmann constant, .pn the canonical probability function, and .Z =

kE
.
n exp[−En /(kB T )] the normalization factor.
Problem 4.2 The equation of motion of the harmonic oscillator is .mq̈ =

−kq, with solution .q(t) = q(0) cos(ωt), and .ω2 = k/m. The distribution
function of the harmonic oscillator is .f = exp[−Etot /(kB T )]/Z, where .Etot
is the total energy of the oscillator and Z is the normalization constant. Obtain
the autocorrelation function .<q(t)q(0)>.
Solution The insertion of the energy in f gives .f = Z −1 exp[−(p2 /m +

mω2 q 2 )/(2kB T )]. From the solution of the equation of motion, the average
is .<q(t)q(0)> = <q 2 (0)> cos(ωt). We compute the value of .<q 2 > using the
probability function f .
f∞ p2 f∞
<q 2 > = Z1 −∞ exp(− 2mkBT
)dp −∞ q 2 exp(−αq 2 )dq ; α = mω2 /(2kB T )
.
f∞ f∞
= −∞ q 2 exp(−αq 2 )dq / −∞ exp(−αq 2 )dq
f∞ √
We use the results . −∞ exp(−αq 2 )dq = π/α, and
f∞ √
−∞ q exp(−αq ) dq = 1/(2α) π/α to simplify .<q >.
.
2 2 2
. <q 2 > = kB T /(mω2 )
In this connection, the autocorrelation function of the position in time is:

kB T
. <q(t)q(0)> = <q 2 (0)> cos(ωt) = mω2
cos(ωt)
4.7 Properties of Correlation Functions
The correlation functions satisfy several properties that may be used for the better
understanding of particle systems. The properties are based on the fact that the
particle system is in an stationary state, and thus, we are able to apply probability
distribution functions in the evaluation of ensemble averages. Yet according to the
ergodic hypothesis, the correlation functions may be equivalently computed from
trajectory averages in molecular dynamics since the ensemble averages and the
trajectory averages are equivalent under conditions of thermodynamic equilibrium
[7].
One of the properties of the correlation functions involves conjugate symmetry.
If we consider the change of variables .x = t − τ in:
f∞ ∗ (t
. CAB (τ ) = −∞ A(t)B − τ )dt (4.19)
∗ (τ ) =
f∞ ∗
we obtain .CAB
f −∞ B(x)A (x + τ )dx. By replacing .τ by .−τ , we reach
∗ ∞ ∗
.C
AB (−τ ) = −∞ B(x) .A (x − τ )dx. The integral corresponds to the definition
.CBA (τ ). Hence, the conjugate symmetry indicates that:
∗ (−τ ) = C
CAB (4.20)
. BA (τ )
When .A = B, the autocorrelation function is even in time, .CA (τ ) = CA (−τ ).

Another property is the transposition of the independent variables in the expression
of the correlation function.
. CAB (t1 , t2 ) = CAB (At, 0) ; At = t1 − t2 (4.21)
f ∞0) = CAB (0, −t).

The correlation function is also an even function in time, .CAB (t,
The functions A and B are orthogonal when .CAB (τ = 0) = −∞ A(t)B ∗ (t)dt =
0. The correlation function should not be confused with the convolution function.
The convolution equally represents a mathematical operation on a pair of functions,
specifically, on the product AB. It is defined by (the discrete version is also shown):
f∞ EN
. A(t) ∗ B(t) = −∞ A(τ − t)B(t) dt ; F ∗ I (x) = i=−N F (i)I (x − i)
(4.22)
There is a sign flip in an integration variable. In addition to the properties of the
correlation function, the autocorrelation function also satisfies some properties. It is
always equal or greater than zero when it is evaluated at equal times, .t1 = t2 = t,
commutative in the times .t1 and .t2 , constant for absolute times, and invariant with
respect to time shifts.
<A(t)A(t)> = <A2 (t)> ≥ 0 ; <A(t1 )A(t2 )> = <A(t2 )A(t1 )>

. (4.23)
<A(0)A(0)> = constant ; <A(t1 )A(t2 )> = <A(t1 + At)A(t2 + At)>
4.7 Properties of Correlation Functions 79
A time shift, .At, may be moved from an argument to the other with a change of
signs.
CA (t, t + At) = <A(t)A(t + At)> = <A(t − At)A(t)> = CA (t − At, t)

.
CA (t1 , t2 ) = CA (At, 0) ; At = t1 − t2
(4.24)
The autocorrelation function is stationary due to the dependence on relative times.
After a long time, .At → ∞, the particles become uncorrelated due to the continuous
collisions with many other particles. The time to reach that asymptotic limit is
characteristic of the particle system and its environment. The asymptotic form of
the autocorrelation function is written as the product of the individual averages of
the physical variable.
. CA (t1 , t2 ) = <A(t1 )A(t2 )> = <A(t1 )> <A(t2 )> = <A>2 (4.25)
At→∞
When the fluctuations of observable A with respect to its average value .<A> are
required, the correlation function of the deviation .δA = A − <A> may be computed.
fT
<δA(t1 )δA(t2 )> = lim 1
0 [A(t1 + t) − <A>] [A(t2 + t) − <A>] dt
T →∞ T
. (4.26)
= <A(t1 )A(t2 )> − <A>2
From Eqs. 4.25 and 4.26, the fluctuations of A at different instants of time are
uncorrelated in the asymptotic time limit, .CδA (At → ∞) = 0. When the times
.t1 and .t2 are zero, the variance .σ is obtained as a special case of the autocorrelation
function of .δA.
. σ 2 = <δA(0) δA(0)> = <A2 (0)> − <A>2 (4.27)
In many instances, we may be interested in analyzing a specific direction of a vector

variable, for example, a particular direction of the dipole moment. Under such a
situation, the projection of variable .A onto the direction of interest, say .n̂, is:
. CA (t) = <(A(0) · n̂) (A(t) · n̂)> (4.28)
In particular, the correlations of the velocities usually fade away in time due
to the particle collisions. Such a decay permits the integration of the velocity
autocorrelation function in time because it is a bounded function from above by
a finite value. When it is integrated using a large time interval, .[t = 0, t → ∞], it
results in the diffusion constant.
For a vector with components .x1 , x2 , · · · , xn , the corresponding correlation
matrix with elements ij and dimension .n × n is:
. Cij (r, s) = <xi (r) xj (s)> (4.29)
The variables r and s belong to a mathematically well-behaved space. The cor-

relation matrix may inherit the symmetries of the distribution function employed
in the evaluation of ensemble averages. For the spatial coordinates r and s, the
translational and rotational symmetries are shown, in that order, in the correlation
function as:
. Cij (r, s) = Cij (r − s) ; Cij (r, s) = Cij (|r − s|) (4.30)
In the situation of 2D discrete data, the correlation function is written in the form:
EN EN
. F ◦ I (x, y) = i=−N j =−N F (i, j ) I (x + i, y + j ) (4.31)
The function F plays the role of a filter, and function I represents the pixels of
an image. For a given central pixel, the correlation function and the convolution
function combine the neighboring pixels to replace the central pixel.1 The corre-
lation and convolution functions are ubiquitous in molecular dynamics and image
processing and are employed in many codes of machine learning.
The autocorrelation function is related to the Fourier transform. In order to have
this relation, we recall the Fourier transform of the variable A and insert it into the
autocorrelation function.
f∞
CA (t) = −∞ A(t + τ )A∗ (τ ) dτ
.
f∞ ( 1 f∞ −iω2 (t+τ ) dω
)( f
∞ −iω1 τ dω
)∗
= −∞ 2π −∞ A(ω 2 ) e 2
1
2π −∞ A(ω 1 ) e 1 dτ
f
By using the expression of the Dirac function . exp[−i(ω2 − ω1 )τ ] dτ/(2π ) =
δ(ω2 − ω1 ):
f∞ 2 e−iωt
. CA (t) = 1
2π −∞ |A(ω)| dω = F[|A(ω)|2 ](t) (4.32)
The Fourier transform is .F. Therefore, the autocorrelation function has a spectral
decomposition and may be computed from .|A|2 in the frequency domain. By dealing
with:
f
. CAB (t) = A(t + τ )B ∗ (τ ) dτ
and following the procedure presented above, the cross-correlation theorem is

obtained.
1 D.Jacobs [8]. Many freely distributed programs for image manipulation allow the use of
convolution to modify the original images.
4.8 Vibrational Spectra from Autocorrelation Functions 81
f f∞
. A(t + τ )B ∗ (τ ) dτ = 1
2π −∞ A(ω)B
∗ (ω) e−iωt dω = F[A(ω)B ∗ (ω)](t)
(4.33)
The Wiener-Khinchin theorem is a particular result of the cross-correlation theorem.
4.8 Vibrational Spectra from Autocorrelation Functions
In many dynamic systems, the particles self-correlate their motions, and it is impor-
tant to determine the main vibrational frequencies to characterize the system (the
synchronization of many metronomes represents an example on the self-correlation
of dynamic bodies). The velocity autocorrelation function, .Cv (t) = <v(0) · v(t)>, is
appropriate to reveal the collective vibrations of the particles, Fig. 4.2. In particular,
the power spectrum is defined in terms of the velocity autocorrelation function. By
starting from .Cv (t), applying the Fourier transform to the time function, we obtain:
f∞
. F (ω) = √1
−∞ <v(0) · v(t)> e
iωt dt (4.34)
2π
The power spectrum provided by the Fourier transform, .F (ω), includes all the
frequencies of the particle system. The peaks of the spectrum are associated with
intense vibrational modes, which are frequently the ones of interest. Thus, the
velocity autocorrelation function embodies the vibrational characteristics of the
system, and in this way, it shows the spreading of the energy.
Fig. 4.2 The velocity autocorrelation functions of 13 hydrogen molecules in confinement and
under temperatures of 51K and 569K are depicted. At 51K, the molecules vibrate with large
amplitudes around their equilibrium positions, giving evidence of the atomic localization, which is
a characteristic behavior of a solid with strong interactions. At 569 K, the atoms collide, producing
frequent changes of the velocity directions and leading to atomic delocalization. The velocity
autocorrelation function is an indicator on the melting of a molecular system [9]
The absorption of infrared radiation produces atomic vibrations. Similarly, the

inelastic scattering of visible light produces atomic vibrations [10]. The autocorre-
lation function of the molecular dipole moment gives information of the infrared
spectrum, and the autocorrelation function of the molecular polarizability gives
information of the Raman spectrum. In this last case, the molecular simulation
demands an electric field to determine the polarizabilities. The anharmonic vibra-
tions are inherently considered in the vibrational spectra determined from the
correlation functions because the atoms are not constrained to move in a harmonic
potential [11].
Problem 4.3 Show that the velocity autocorrelation function is related to

the position autocorrelation function by using the Fourier and inverse-Fourier
transforms:
f ∞
.X(ω) = F[x(t)] = x(t) e−iωt dt;
0
f ∞
−1 1
x(t) = F [X(iω)] = X(iω) eiωt dω
2π −∞
Solution We know that .ẋ(t) = dx(t)/dt. By applying the Fourier transform

to the positions, and renaming the temporal variables t and .t + τ , we obtain:
f ∞ f ∞ ( f ∞ )
d 1
. <ẋ(τ ) ẋ(t + τ )> dt = < X(iω) eiωt1 dω
−∞ −∞ dt1 2π −∞
( f ∞ )
d 1
× X(iω) eiωt2 dω >t1 =t2 =t dt
dt2 2π −∞
f ∞
= (iω) 2
<x(τ ) x(t + τ )> dt
−∞
The autocorrelation function of the velocities is related to the autocorrelation

function of positions by integration in time.
References
1. F.A. Lindemann, The calculation of molecular vibration frequencies. Physik Z. 11, 609–612
(1910)
2. J. Soullard, R. Santamaria, D. Boyer, Thermodynamic states of nanoclusters at low pressure and
low temperature: the case of 13H2 . J. Phys. Chem. A115, 9790–9800 (2011), pp. 9796–9797
3. P.G. de Gennes, J. Prost, The Physics of Liquid Crystals (Clarendon Press, Oxford, 1993). M.
Doi, S.F. Edwards, The Theory of Polymer Dynamics (Clarendon Press, Oxford, 1986)
References 83
4. R.O. Watts, I.J. McGee, Liquid State Chemical Physics (Wiley, London, 1976), p. 54
5. V. Calandrini, E. Pellegrini, P. Calligari, K. Hinsen, G.R. Kneller, nMoldyn – interfacing
spectroscopic experiments, molecular dynamics simulations and models for time correlation
functions. Collection SFN 12, 201–232 (2011), pp. 202–203
6. R. Kubo, H. Ichimura, T. Usui, N. Hashitsume, Statistical Mechanics, 6th edn., (North-Holland,
Amsterdam, 1981), pp. 1–32, Chap. 1
7. A. Tokmakoff, MIT Open Course Ware, 5.74 Introductory quantum mechanics II. MIT
Department of Chemistry (2009)
8. D. Jacobs, Correlation and convolution, class notes for CMSC 426, Fall 2005
9. R. Santamaria, J. Soullard, J. Jellinek, Thermal behavior of a 13-molecule hydrogen cluster
under pressure. J. Chem. Phys. 132, 124505 (2010), Eq. 9 and Fig. 5
10. E. Hecht, Optics, Global edn., 5th edn. (Pearson Education Limited, Essex, 2017), Sect. 3.6.3
11. M. Thomas, M. Brehm, R. Fligg, P. Vohringer, B. Kirchner, Computing vibrational spectra
from ab-initio molecular dynamics. Phys. Chem. Chem. Phys. 15, 6608–6622 (2013)
Rigid Body
5
The purpose of this chapter is to determine the main equations of a body integrated
by a number of particles which maintain their distances fixed from each other
and give shape and volume to the body. A particle system like this is known as
a rigid body, formed by many particles with inter-particle distances fixed. The
motion equation and energy of the rigid body depend on the spatial arrangement
of the particles, and henceforth, a set of geometric parameters are introduced to
characterize the rigid body’s degrees of freedom. Two main degrees of freedom are
identified: the translational motion, where the particles of the rigid body describe
parallel trajectories, and the rotational motion, where the particles rotate all together
as a single entity. The translational motion is frequently omitted as it is possible to
work in an inertial reference frame undergoing translational motion with the rigid
body. In this respect, we analyze the rotating motion of the rigid body in terms of
the angular momentum and the torque, which is produced by external forces acting
on the body.
5.1 Angular Momentum of a Rotating System of Particles
The description of the circular motion requires an axis of reference, and therefore,
the z axis is taken as the rotation axis of the particles that form the body. The
reference frame of coordinates is depicted in Fig. 5.1 (similar to the reference frame
of Fig. 1.5). The ith particle has position .ri = (xi , yi , zi ) and velocity .vi = ω × ri .
The angular velocity .ω is constant in an isolated system and is the same for all
the particles of the body. The velocity magnitude is .vi = ωri sin(γi ), where .γi
is the angle between vectors .ω and .ri . On the other hand, the angular momentum
.li is perpendicular to both .ri and .vi , because it is the result of the cross product
.li = ri × mi vi , with magnitude .|li | = mi ri vi sin(π/2) = mi ri vi . The component
of .li in the z axis is:
. liz = |li | cos(π/2 − γi ) = mi ri vi sin(γi )

86 5 Rigid Body
Fig. 5.1 A particle P of a

rigid body rotates around the
z axis. It has an instantaneous
position .r, tangent velocity .v,
angular velocity .ω, and
angular momentum .l. The
quantity .γ is the conical
angle, and .π/2 is the angle
between .l and the plane
formed by .r and .v
The trigonometric addition .cos(a + b) = cos(a) cos(b) − sin(a) sin(b) was used
in the last identity. It is convenient to simplify the expression of the component
.liz in terms of .ρi , the distance from the z axis to the particle position. To do this,
we consider the projection of .ri onto the plane described by the particle motion,
.ri sin(γi ) = ρi . Using this identity, the component of the angular momentum in the
z axis transforms onto .liz = mi ri vi sin(γi ) = mi ri [ωri sin(γi )] sin(γi ) = mi ωρi2 .

The same procedure applies to all the particles of the body, and a summation over
the number of particles gives the total z component of the angular momentum.
E
. Lz = i mi ωρi2 (5.1)
E 2
. i mi ρi defines the moment of inertia of the body with respect to the axis of
rotation and is designated with the letter I (not to be confused with the symbol
of impulse on a particle, which is a vector). The component .Lz is now written in
terms of I and .ω.
E
. Lz = I ω ; I= i mi ρi2 (5.2)
The inertia moment I is a geometric factor of the system defined by the masses
and positions of the particles with respect to the rotation axis. Therefore, I depends
on the mass distribution. We may deal with the rotation of the body under more
general terms to introduce the additional components of the angular momentum.
Consider a particle with mass m rotating around an axis with angular velocity .ω.
The angular momentum of the particle is .L = r × p = mr × (ω × r). Contrary
to the linear momentum .p, the angular momentum .L is not an intrinsic property
of the particle, and for such a reason, the reference frame of coordinates should be
specified when the angular momentum and torque are discussed. The components of
the angular momentum .(Lx , Ly , Lz ) are determined from the vector triple product
.a × (b × c) = b(a · c) − c(a · b). In this regard, by identifying vectors .a = r, .b = ω,
.c = r and applying the vector triple product, we have .L = mω(r · r) − mr(r · ω).
The components of .L are .(Lx , Ly , Lz ). For instance, the x component is:

5.1 Angular Momentum of a Rotating System of Particles 87
Lx = mωx (r · r) − mx(r · ω) = mωx r 2 − mx(xωx ) − mx(yωy ) − mx(zωz )

. = m(r 2 − x 2 )ωx − m(xy)ωy − m(xz)ωz = Ixx ωx + Ixy ωy + Ixz ωz
E
= 3i=1 Ixi ωi
It was defined above .Ixx = m(r 2 − x 2 ), .Ixy = m(xy), .Ixz = m(xz). A similar
situation is obtained for the other components, in such a way that the angular
momentum components associated with a general rotation of the body may be
written in the form:
(E E3 E3 )
3
.L = i=1 Ixi ωi , i=1 Iyi ωi , i=1 Izi ωi
This expression may be written in matrix form.

⎛ ⎞ ⎛ ⎞⎛ ⎞
Lx Ix Ixy Ixz ωx
. ⎝ ⎠ ⎝
Ly = Iyx Iy Iyz ⎠ ⎝ ωy ⎠ (5.3)
Lz Izx Izy Iz ωz
Equation 5.3 indicates that the inertia moment is a 3D tensor, .I = Iij , with
components:
E E E
Ix = mi (yi2 + zi2 ) ; Iy = i mi (xi2 + zi2 ) ; Iz = i mi (xi2 + yi2 )
i
.
E E E (5.4)
Ixy = − i mi xi yi ; Iyz = − i mi yi zi ; Ixz = − i mi xi zi
The term .(yi2 + zi2 ) defining .Ix is the distance of the ith particle to the x axis. The
diagonal element .Ix is the inertia moment in the x axis, with similar results for .Iy
and .Iz . The off diagonal elements, .Ixy , .Iyz , .. . ., are the inertia moment products. Six
quantities define the body rotation since the tensor .Iij is a real-symmetric matrix.
Equation 5.3 is written in matrix notation, .L = I ω. According to matrix theory, by
carrying a transformation, it is possible to deal with the diagonal form of matrix
.Iij . The diagonalization of .Iij is determined by solving the eigenvalue problem
.I ω i = λi ω i , where .i = x, y, z, and the eigenvectors .ω i and eigenvalues .λi are real.
A diagonalization process is illustrated in Sect. A.1. The eigenvalue solutions may

be reached for bodies with simple and symmetric shapes or complicated geometries.
⎛ ⎞
Ix 0 0
. Iij = ⎝ 0 Iy 0 ⎠ (5.5)
0 0 Iz
In short, for any shape and rotation of the body, it is possible to find three rotation
axes where the inertia moment products are zero. The angular momentum in terms
of the inertia moments associated with the three preferent axes is:
88 5 Rigid Body
Fig. 5.2 Two symmetric bodies rotate around a vertical axis. The line joining the bodies makes an
angle .θ with the horizontal plane. The angular momentum .L is not parallel to the angular velocity
.ω, as this last vector lies along the vertical axis. However, when the line joining the particles is
perpendicular to the vertical axis of rotation, the angular momentum .L is parallel to vector .ω
. L = u'x Ix ωx + u'y Iy ωy + u'z Iz ωz (5.6)
The unit vectors .u'i , where .i = x, y, z, are in motion with respect to a fixed
reference frame of coordinates in the lab. The unit vectors are mutually orthogonal
and, in turn, fixed in the body, which is rotating, to describe the three directions of
independent rotations in 3D space. Any rotation of the body can be described in
terms of the three main axes, which receive the name of principal axes of rotation.
The inertia moments .Ix , .Iy , and .Iz with respect to such axes are the principal
moments of inertia. The components of the angular velocity .ω in the principal axes
of rotation are .ωx , .ωy , and .ωz . In this regard, the angular momentum of Eq. 5.6 has
general character for any rotating body and indicates that .L = (Lx , Ly , Lz ) and
.ω = (ωx , ωy , ωz ) are not in general parallel, Fig. 5.2. Under special circumstances,
the body may rotate around a principal axis. In this case, the vectors .L and .ω are
parallel, and expression 5.6 of the angular momentum becomes .L = I ω, where
the factor of proportionality is I , namely, the principal moment of inertia for the
respective principal axis of rotation. This last expression shows that .L points in
the direction of the rotation axis. For symmetric bodies, the axes of symmetry
correspond to the principal axes of rotation since a rotation of the body around
an axis of symmetry keeps the angular momentum parallel to the rotating axis, and
.L = I ω.
When the body is dense, we introduce the density .ρ to relate the infinitesimal
mass dm to the infinitesimal volume dV by .dm = ρdV and transform the
summation over the number of particles onto an integral over the body volume for
the inertia moment of expression 5.2.
f
. I= vol ρ(r) r
2 dV (5.7)
5.2 External Torques Acting on a Rotating Body 89
f
When the mass density .ρ is constant, the inertia moment is .I = ρ r 2 dV . The
body geometric features and spatial extension are considered by the integral over
the body’s volume.
5.2 External Torques Acting on a Rotating Body
The purpose of this section is to use the inertia moment I in the equation of motion
of a rotating rigid body to get it simplified as, otherwise, the equation of motion of
every particle that constitutes the body has to be specified. The equation of motion
of a rotating body is proposed from the expression .dL/dt = τ , relating the angular
momentum .L to the torque .τ . The angular momentum and torque are described from
a common inertial reference frame. When the body is formed by many particles, the
total angular momentum and total torque are:
E E
. L= i li ; τ= i τi (5.8)
li and .τ i are the angular momentum and torque on the ith particle. The individual
.
torque .τ i is due to the external forces acting on the ith particle. In a general situation,
the body rotates around an arbitrary axis with simultaneous translatory motion. An
example of such a situation is a rotating yoyo, where the yoyo falls due to gravity.
It is convenient to consider the body rotation axis as the z axis and work with the
expression .dLz /dt = τz to describe the body rotation. The component .Lz may be
replaced by the product .I ω, Eq. 5.2. Further simplification is achieved when the
inertia moment I is time indifferent.
. I dω/dt = τz (5.9)
The torque component .τz is zero for constant angular velocity. In order to keep the
direction of the total angular momentum in the z axis, the torque components .τx
and .τy need to be different from zero to counteract the centrifugal forces exerted
on the particles moving in a circular path. Therefore, in addition to .τz , the torque
components .τx and .τy participate in the stability of the body rotation.
The description of the particle system is substantially reduced when the rotation
axis corresponds to a principal axis of rotation as, in this case, the total angular
momentum is written in the form .L = I ω, being the vectors .L and .ω parallel. The
equation of rotation is:
. d(I ω)/dt = τ (5.10)
For zero torque, the product .I ω is preserved. The expression .I ω = constant

discriminates between a rigid body and a non-rigid one. A non-rigid body demands
an increase of the inertia moment I for a decreasing value of .ω, and vice versa, to
keep the product .I ω constant. On the other side, the inertia moment I of a rigid body
90 5 Rigid Body
is fixed, and the condition .τ = 0 implies a constant angular velocity, .ω = constant.

Examples of these two cases correspond to the different spins that a cooked egg and
a raw egg exhibit. A cooked egg is solid and spins relatively easy, while the liquid
of the raw egg takes time to acquire the spin and also slows down the rotation by
friction.
In many situations it is convenient to take the body center of mass as the origin of
coordinates to describe the angular momentum. Once the reference frame is located
at that point, any external torque should be also described from the center of mass
to have consistent descriptions of the angular momentum and the external torques.
. dLcm /dt = τ cm (5.11)
Again, the equation is simplified when the rotation axis of the body is a principal
axis of rotation because .Lcm = Icm ω, and the equation of motion is transformed
onto .Icm dω/dt = τ cm . In the absence of external torques, the angular momentum
may be considered an intrinsic property of the body because .Lcm shows no change.
In such a case, the angular momentum is referred to as the body spin. A spinning
body may additionally have an orbital angular momentum, like that of a spinning
planet in orbital motion around the Sun.
Problem 5.1 Suppose that a body is translating and rotating. The axis of
rotation passes through the body’s center of mass (cm). Show that the angular
momentum, .L, may be written in terms of an orbital component plus a spin
component.
f
Solution The body’s angular momentum is .L = r × v dm. The factor dm
is a mass element of the body. The total body’s velocity is decomposed in
.v = vcm + vrot , with .vcm the cm translational velocity and .vrot the body’s
rotational velocity with reference to the axis passing through the cm. We have
' '
.vrot = ω × r , where .r is the position of the mass element dm with respect
to the cm reference frame of coordinates. The total velocity reads .v = vcm +

ω × r' , and the angular momentum becomes:
f
L = (rcm + r' ) × (vcm + ω × r' ) dm
.
f
= Lcm + r' × (ω × r' ) dm
Lcm = mtot (rcm × vcm ) and the body’s total mass is .mtot . In thisf regard, the
.
angular momentum has components .Lorb = Lcm and .Lspin = r' × (ω ×

r' ) dm.
5.3 Total Energy of a Rotating Rigid Body 91
In the following lines, we are interested in describing the torque from different
reference frames. To do this, consider a pair of reference frames with common
origin. One of them has angular motion with respect to the other. The torque in the
non-primed reference frame of coordinates, .τ = dL/dt, is related to the torque in
the rotating (or primed) reference frame of coordinates, .τ ' = dL' /dt. The relation
is established by using the mnemonic rule 1.38, .db/dt = db' /dt + ω × b. The
angular momentum .L replaces vector .b in Eq. 1.38. Since the reference frame co-
rotates with the body, and the axes of the rotating reference frame are assumed to
be the body principal axes of rotation, the quantity .ω is the angular velocity of the
primed reference frame. The relation between the two torques is:
. τ = τ' + ω × L (5.12)
This identity is known as the Euler equation for the torques computed in the
different reference frames. The Cartesian components of the equation are:
τx = Ix' + (Iz' − Iy' )ωy ωz τy = Iy' + (Ix' − Iz' )ωx ωz

dωx dωy
dt ; dt
. (5.13)
τz = Iz' dωz
dt + (Iy' − Ix' )ωx ωy
The Euler equations are simplified when there is no angular acceleration because
in this case .dω/dt = 0. Still, there are other terms of the torque components
.(τx , τy , τz ) that are not zero and act on the body. If the body’s rotating axis is the
'
.z principle axis of rotation, .ωz /= 0, then all the torque components .τi in Eqs. 5.13
are zero because the angular velocity components .ωx and .ωy are zero, and the body
rotates steadily due to the absence of torque components.
5.3 Total Energy of a Rotating Rigid Body
The following quantity of interest is the total energy of the rigid body. It involves
the kinetic energy and potential energy. The purpose is to separate the translational
kinetic energy from the rotational kinetic energy in the expression of the total kinetic
energy and separate the external potential energy from the internal potential energy
in the expression of the total potential energy. To do this, it is convenient to establish
a reference frame of coordinates in the body’s center of mass (cm). The rotational
kinetic energy is given in terms of the translational velocity, .vcm , and rotational
velocity, .ω × r' .
f
. Ekin = 1
2 dm (vcm + ω × r' )2 (5.14)
The translational kinetic energy is that of the center of mass, .mtot v2cm /2, with .mtot
the mass of the rigid body. The rotational velocity is .ω × r' , where .r' is the position
of the mass element dm with respect to the cm reference frame of coordinates and
92 5 Rigid Body
ω is the angular velocity. We use the angular momentum .L and the angular velocity
.
ω to rewrite the kinetic energy.

.
f
. Ekin = 1
2 mtot v2cm + 12 ω · dm r' × (ω × r' ) = 1
2 mtot v2cm + 1
2 ω·L (5.15)
We have translational and rotational components of the kinetic energy.
Problem 5.2 Obtain the kinetic energy 5.15 from expression 5.14 in the case
of many particles forming the body, namely, the particles discrete case.
Solution TheEdiscrete version of Eq. 5.14 (omitting primes) is .Ekin =

mtot v2cm /2 + i mi (ω × ri )2 /2. The product of vectors in this expression
has the general form .(a × b) · (c × d). We identify vectors: .a = c = ω and
.b = d = ri . The purpose is to demonstrate that .a · [b × (c × d)] is equivalent
to .(a × b) · (c × d). We start from the expression:
. b × (c × d) = (b · d)c − (b · c)d
By carrying the dot product of .a with every term of the equation, we have:
a · [b × (c × d)] = (b · d)(a · c) − (b · c)(a · d)

. = (a · c)(b · d) − (a · d)(b · c)
= (a × b) · (c × d)
E using this result, the kinetic energy is .Ekin = mtot vcm /2 + (ω/2) ·
By 2
m
i i ir × (ω × ri ). The factor .ω × ri represents the particle rotational
velocity. Thus, the summation gives the total angular momentum of the body,
.Ekin = mtot vcm /2 + ω · L/2.
2
We know that when the rotating axis is a principal axis of rotation passing through
the center of mass, we have .L = I ω. The equation of the kinetic energy is simplified
to .Ekin = mtot v2cm /2 + L2 /(2I ). The total kinetic energy may be alternatively
written as .Ekin = mtot v2cm /2 + I ω2 /2. If we use the angular momentum .L = I ω in
terms of the inertia moment tensor .I , Eq. 5.3, the kinetic energy in matrix form is:
. Ekin = ωt I ω/2 (5.16)
where the translational kinetic energy is neglected for the isolated rigid body and .ωt
is the transpose of the angular velocity. Equation 5.16 in terms of the components
of .ω and .I is:
A.1 Appendix: Matrix Diagonalization 93
⎛
⎞⎛ ⎞
Ix Ixy Ixz ωx
Ekin ⎝
= 2 (ωx , ωy , ωz ) Iyx Iy Iyz
1 ⎠ ⎝ ωy ⎠
.
Izx Izy Iz ωz (5.17)
= 12 (Ix ωx2 + 2Ixy ωx ωy + 2Ixz ωx ωz + Iy ωy2 + 2Iyz ωy ωz + Iz ωz2 )
By assuming a system where the inertia moment products have null contribution,
and using expression 5.6 of the angular momentum, the kinetic energy is written in
terms of the principal moments of inertia solely.
( )
. Ekin = L2x /Ix + L2y /Iy + L2z /Iz /2 (5.18)
The internal potential energy of the rigid body is constant. However, in order to
formulate the total energy of the system, we are required to also deal with any
external action on the rigid body capable of producing a change of its kinetic energy.
(0)
In this regard, we recall the energy balance .Wext + Q = Ekin − Ekin , Eq. 3.23,
where the external work .Wext is due to external conservative forces and Q is due to
external non-conservative forces. The external work is expressed as the difference
(0)
of external potential energies, .Wext = Epot,ext − Epot,ext .
(0) (0)
. Ekin + Epot,ext = Ekin + Epot,ext +Q (5.19)
When .Q = 0, the total energy .Etot at different instants of time is preserved.
. Etot = 1
2 mtot v2cm + 1
2 Icm ω
2 + Epot,ext = constant (5.20)
When the force of gravity is the external potential energy, we have .Epot,ext =
mtot ghcm , where .hcm is the position of the mass center with respect to ground level.
. Etot = 1
2 mtot v2cm + 1
2 Icm ω
2 + mtot ghcm = constant
A.1 Appendix: Matrix Diagonalization
Suppose that A is a matrix with dimension .n × n, and it can be diagonalized. The

diagonalization of matrix A demands the writing of A in terms of matrices .Ad and
T.
. A = T Ad T −1 (5.21)
Ad is the diagonal matrix whose components are the eigenvalues of A, while T

.
is the transformation matrix whose components are the eigenvectors of A. The

transformation matrix T acting on both sides of A gives .Ad .
94 5 Rigid Body
. T −1 AT = Ad (5.22)
In the first stage, the eigenvalues of A are obtained from the determinant:
. |A − λI | = 0 (5.23)
I is the unit matrix. The expansion and evaluation of determinants are given in
Sect. B.1 of Chap. 10. The determinant unfolds onto its characteristic polynomial of
order n in .λ. The polynomial equation may be solved for the roots or eigenvalues
.λ1 , · · · , λn . In the second stage, we observe that .AV = λV, and therefore, .(A −
λI )V = 0. The matrix of coefficients is .(A − λI ), and the vector of variables is .V,

supposed to be a non-trivial vector. We must work with each eigenvalue separately
to obtain the corresponding eigenvector. The first matrix equation .(A−λ1 I )V1 = 0,
where .V1 is not .0, is proposed.
⎛ ⎞⎛ ⎞
A11 − λ1 A12 · · · A1n v11
⎜ A21 A22 − λ1 · · · A2n ⎟ ⎜ v12 ⎟
.⎜ ⎟⎜ ⎟=0 (5.24)
⎝ ··· ··· ⎠⎝ ··· ⎠
An1 An2 · · · Ann − λ1 v1n
The components of the eigenvector .V1 = (v11 , · · · , v1n ) are obtained from the
system of n equations with n unknown variables implied by the first matrix
equation. Similarly, the components of the linear independent eigenvector .V2 =
(v21 , · · · , v2n ) are obtained from the second matrix equation .(A − λ2 I )V2 = 0
and so on. In principle, the number of eigenvectors is n, and they are linearly
independent from each other (when there are repeated eigenvalues, the linear
independence property may be additionally used to find the eigenvectors). After
finishing the whole process, we have the n eigenvalues and the corresponding
eigenvectors of A.
λ1 ; V1 = (v11 , · · · , v1n )
λ2 ; V2 = (v21 , · · · , v2n )
.
··· ···
λn ; Vn = (vn1 , · · · , vnn )
The matrices .Ad and T are consequently determined from the eigenvalues and
eigenvectors.
⎛ ⎞ ⎛ ⎞
λ1 0 0 · · · v11 v21 · · · vn1
⎜ 0 λ2 0 · · · ⎟ ⎜ ⎟
. Ad = ⎜ ⎟ ; T = ⎜ v12 v22 · · · vn2 ⎟ (5.25)
⎝··· ···⎠ ⎝ ··· ··· ⎠
0 0 0 λn v1n v2n · · · vnn
A.1 Appendix: Matrix Diagonalization 95
The order in writing the eigenvalues follows the order in writing the eigenvectors
and vice versa. The inverse matrix .T −1 is determined from the condition on the
matrix product .T T −1 = I . Once we determined .Ad , T , and .T −1 , the identity
.A = T Ad T
−1 may be confirmed. Finally, the diagonalization of matrices is a
process among many processes in linear algebra, but it acquires especial importance
in quantum physics, for instance, for the determination of state energies, the finding
of vibrational modes and frequencies, etc.
Analytical Mechanics
6
The formulations presented in this chapter establish the principles of analytical

mechanics. The chapter starts by introducing the action as the fundamental function
that contains the physical information of the particle system. By considering the
stationary principle on the action, the Lagrangian theory is derived, paying special
attention to the equations of motion. Based on the Lagrangian formalism, we show
that a symmetry in nature implies a conserved observable via Noether’s theorem.
Later, by using a Legendre transformation, the relation between the Lagrangian
function and the Hamiltonian function is achieved. Both functions codify the
physical information of the system but employ different conjugate variables. A
formulation of mechanics in terms of the Hamiltonian, which is equivalent to
Lagrangian mechanics, is now established with the respective equations of motion.
The Hamiltonian is demonstrated to be a constant of motion, corresponding to the
energy of the system when the Lagrangian is time independent. The possibility of
dealing with different Hamiltonians of a given system takes us to the discussion
of the canonical transformations with the purpose to link the different Hamiltonian
functions. The specific canonical transformation where the basic conjugate variables
of the system are constants introduces the Hamilton-Jacobi scheme as a branch of
Hamiltonian theory where the solutions of the equations of motion may be found
in a distinct manner. The Hamilton-Jacobi theory not only brings the connection
back to the action function but also provides the first relative contact of classical
mechanics with quantum mechanics.
The particles dynamics is analyzed according to the Poisson formulation, because
this offers the opportunity to derive a classical time propagator from the Liouville
operator. The functional form of the propagator resembles that of a quantum
mechanical propagator and allows to have the time evolution of physical observables
in terms of the positions and momenta. All the previous mechanical formulations
are linked together and are equivalent to Newtonian theory. Yet, every formulation
provides from different perspectives the fundamentals of analytical mechanics. The
analytical theories of mechanics shall be used in later chapters to establish contact

98 6 Analytical Mechanics
with quantum theory. The chapter is finished by discussing the general and modern
concepts of mechanics.1
6.1 Action Function
In daily life, an action is the accomplishment of a task using a force in an interval

of time. In mechanics, it has a more precise meaning. The action is a mathematical
functional that, by using the function L, contains the physical characteristics of
the particle system, such as the number of particles, their types, forms of kinetic
energy, and kinds of interactions, and from which the information can be extracted in
systematic manner to establish the equations of motion of the particles. The method
to obtain information from the action is general and plays a central role not only
in mechanics but also in other fields of physics like in electromagnetism, general
relativity, high-energy physics, etc. The action is represented by the letter S; it is
proposed as an integral of the function L over time and depends on the time limits
.ti and .tf .
f tf
. S= ti L({qi }, {q̇i }, t) dt (6.1)
The function L is the Lagrangian. The Lagrangian provides key information on the
motion and interaction aspects of the particles. It is given as the balance between
the kinetic energy, T , and potential energy, V . The units of S are those of angular
momentum .(energy × time).
. L = T ({q̇i }) − V ({qi }) (6.2)
The formulation in terms of the action demands a bounded trajectory in time and
space and is invariant to the coordinates representation. In this respect, the variables
.qi and .q̇i (the dot indicates differentiation in time) are the generalized coordinates
and generalized velocities in the action. They constitute the independent degrees of
freedom of the theory and may have units that are not necessarily those of position
and velocity. By dealing with the action, it is possible to determine the equations
of motion of the particles. When such equations are solved, we are able to describe
the variables .qi and .q̇i with time. The time evolution of the system corresponds to
the time evolution of the particles, supposed to evolve along the path that goes from
.{qj (ti ), q̇j (ti )} to .{qj (tf ), q̇j (tf )}, with .ti and .tf the initial and final times. The sets
of generalized coordinates and generalized velocities define the state of the system.
1 The reader may also follow the discussions of: Goldstein [1] and Arnold [2].
6.2 Principle of Stationary Action 99
6.2 Principle of Stationary Action
The path followed by the particles is observed to make the action an extremum. In
this regard, the action is postulated to be stationary for the real trajectory of the
particles. The origin of the postulate comes from experimentation and calculations
on both the paths of diffracted light at the interface of two mediums and the paths
described by the particles in time and space. In reference to the paths described in
time and space, consider a bead that slides without friction along a thin wire due
to the action of gravity. The bead departs with zero speed from point A at a fixed
height and reaches the lowest point B describing a particular trajectory, Fig. 6.1.
The experiments show that the sliding bead moves along the trajectory with the
shortest traveling time. In mathematical terms, that path is an extremum trajectory,
and in the case of the sliding bead, the trajectory corresponds to a brachistochrone.2
The brachistochrone trajectory of the sliding bead is intimately related to the
trajectory of a light beam diffracted at the interface of two mediums.3 These
observations are generalized by assuming similar behavior of the particles at the
microscopic level, leading to a more general scenario. Thereby, the principle of least
action, also known as the Hamilton’s principle, is a fundamental piece of mechanics,
allowing in a formal and systematic way to mathematically reach the equations of
motion of the particles in general.
Fig. 6.1 A bead with mass m slides along a wire without friction from point A to point B under the
action of gravity. The principle of stationary action indicates that among many possible trajectories,
the path with minimum time of descent is the favored trajectory by nature
2 In common jargon, nature is efficient as it considers the least-time path in the time evolution of a
particle.
3 The name brachistochrone is due to the Greek language (brachisto .∼ shortest, and chrone .∼
time). The trajectory that makes the action an extremum was found by Johann Bernoulli and
corresponds to a brachistochrone curve. On the other hand, when light crossing two mediums with
different indexes of refraction travels from point A to point B, it does so taking the fastest path.
The light path is determined using the Fermat’s principle and the Snell’s law. The brachistochrone
trajectory and the light path are closely related to each other: Levi [3].
Statement 6.1 Principle of stationary action: the motion of a particle from

time .t1 to time .t2 is described by the stationary solution of the action. The
action is S and is given as an integral of the Lagrangian over time.
f t2 f t2
. S= t1 L dt = t1 (T − V ) dt (6.3)
The Lagrangian L is .T − V , where T is the motion energy and V is the

interaction energy. The stationary condition is .δS = 0.
In contraposition to Newtonian mechanics, which is based on a cause-effect

paradigm for describing the motion of the particles under the action of vector forces,
analytical mechanics considers a common pattern in the motion of the particles,
because it deals with the whole path of a particle in an interval of time in the
definition of the action, postulating that the trajectory followed by the particle is a
path of least time and formally establishing a variational principle on the action. The
action has dependence on a Lagrangian function, which introduces the information
of the system through the energy difference .T − V , a scalar quantity which must
have small variations over time, thus avoiding to directly deal with vector forces.
The Lagrangian equations of motion are deduced from the condition:
. δS = 0 (6.4)
The principle of least action applies to the particle system in any time interval, and
therefore, the time evolution of the system satisfies the Lagrangian equations of
motion at every instant of time. The stationary condition indicates that the action
may represent a minimal, maximal, or saddle point of a multidimensional surface
generated by the generalized coordinates and generalized velocities. However, we
consider a minimal value of S for the trajectory with the shortest traveling time.
From the physical point of view, the stationary principle on the action replaces the
Newtonian principles and is of special interest when the forces have little sense,
such as in quantum mechanics. From a mathematical perspective, the stationary
principle constitutes a variational principle that introduces an alternative mechanical
formulation of Newtonian theory. The principle of stationary action demands the
functional derivative of S to be zero, .δS = 0, for small perturbations of .L =
T − V in terms of the generalized coordinates and generalized velocities. The
proper mathematical field to deal with the stationary condition on a function of
several variables is calculus of variations. In the process of connecting calculus
of variations to differential calculus, and imposing the stationary condition on the
action function, we shall reach differential equations of the generalized variables
.qi and .q̇i . Such differential expressions are the equations of motion of the particles.
This formulation makes use of the Lagrangian function and is known as Lagrangian
6.3 Classifying Molecular Systems 101
theory. After applying the initial conditions on the dynamics of the system, the
solutions of the differential equations are unique and provide the time evolution of
the particles, namely, the particle trajectories with time.
The Lagrangian theory has applications in different domains of physics, as in
electromagnetic theory, quantum mechanics, relativistic mechanics, quantum field
theory, etc. The quantum path integral method has been found to explain to some
extent the principle of stationary action. This method takes every path of a quantum
particle, like that of an electron or photon, into consideration to estimate the different
probabilities to reach a final quantum state from an initial one. Here, the Schrödinger
quantum equation plays the role of the diffusion equation. The wave phases far from
a minimum-action path are found to cancel each other, only surviving the wave
phases close to the minimum-action path. In the classical limit, when .S(x) >> h̄ and
.h̄ is the Planck constant, the path integral method leads to the principle of stationary
action and explains the Fermat’s principle of optics, which specifically states that a
light beam travelling from point A to point B takes the fastest trajectory.
6.3 Classifying Molecular Systems
There are general definitions and important classifications which are required for the
characterization of the particle systems. The definitions and the classifications shall
be used later in the formulations of analytical mechanics. The system of particles
may be able to preserve the energy when the forces exerted upon the particles are
conservative, because they are derived from a potential function V in the form
.fi = −∇i V ({rk }). When the forces are not associated with a potential function,
they are non-conservative forces, and the system may absorb or dissipate energy.4
On the other side, a system may be subjected to forces that impose constraints, for
example, those forces that keep selected atomic bonds rigid. There are different
types of constraints, and they have to be classified as well. The constraints are
holonomic when they are written in terms of the particle coordinates, including the
time, or when the constraints are written as differential equations. An example of
constraints relating the coordinates is:
. f (qi , t) = 0 (6.5)
An holonomic constraint satisfies the identity .df/dt = 0. The constraints may be

expressed in differential form.
E ∂f ∂f
. df = j ∂qj dqj + ∂t dt (6.6)
4 Suppose a great number of particles interacting with a relatively large body. The individual
particle forces are conservative and decelerate the body’s velocity. If the interactions with the
body are statistically averaged, we may write the average force as a frictional force, .−ηv, where
.η is the friction factor and v is the body’s velocity. The negative sign indicates the opposition
of the statistical force to the body’s motion. The frictional force is non-conservative, in spite of
representing a statistical average of the conservative forces. In summary, a simplified description
of many conservative forces may result on a single non-conservative force acting on the body.
When the constraints are written in terms of mathematical inequalities, the con-
straints are non-holonomic. An example of a non-holonomic constraint involves a
moving particle on the external surface of a sphere, in such a way that the particle
is in contact with the sphere surface (.x 2 + y 2 + z2 ≥ R, where the coordinates
of the particle are .(x, y, z) and the radius of the sphere is R). The particle moves
and eventually falls from the sphere at some stage due to the action of gravity. The
non-holonomic constraints may additionally include the velocities, .f (qi , q̇i , t) = 0.
In general, it is difficult to transform non-holonomic expressions onto holonomic
constraints, as the non-holonomic expressions are usually not integrable. There is
no general method to deal with non-holonomic constrains. Yet, in many instances, it
is possible to apply several methods at different stages of the problem for describing
the particle motion. The Lagrangian equations of motion consider holonomic and
non-holonomic constraints that are integrable.
The constraints may or may not show time dependence. In the case of an
explicit time dependence, the system classifies as rheonomous; otherwise, it is
scleronomous. The minimum number of independent variables required to specify
the state of a particle system is referred to as the degrees of freedom of the
system. In the general case, there are 3N spatial degrees of freedom of a system
composed by N particles. If there are m spatial constraints relating the variables,
then the number of independent degrees of freedom is reduced to .3N − m. The
spatial degrees of freedom are given by the independent variables .qi (t), which all
together built a mathematical space identified as the configuration space. When the
momenta .pi are included in the description, the space formed by all the sets of
variables .{qi (t), pi (t)} is the phase space. Without constraints on the coordinates
and momenta, the number of degrees of freedom is 6N . The set of 6N variables
is visualized all together as forming a single point in the multidimensional phase
space. In this case, the time evolution of the system is described by the trajectory of
that single point in phase space, Fig. 6.2.
Fig. 6.2 Inset A: the phase space is a mathematical space with dimension determined by the total
number of degrees of freedom of the particles. A state of the particle system is visualized as a
single point in the multidimensional phase space. Inset B: the time evolution of the particle system
defines the trajectory of the single point in phase space. In this regard, the phase space point is
associated with the dynamical state of the mechanical system
6.4 Lagrange’s Equations of Motion 103
6.4 Lagrange’s Equations of Motion
Consider a set of interacting particles. The kinetic energy of the system is T , and
the interaction among the particles is described by the potential energy V , a scalar
function that has dependence on the coordinates .qi , velocities .q̇i , and time t. The
purpose of this section is to deduce the Lagrange equations from the energy terms
T and V . The Lagrangian formulation is an alternative approach to the Newtonian
equations to describe a system of particles.
The Lagrange equations may be derived either from the differential principle
of d’Alembert or the integral principle of Hamilton. In the d’Alembert’s approach,
the virtual displacements .δri occur in an instant of time t, not in the time interval
dt when the real forces change. Yet, the virtual displacements are supposed to be
compatible with the constraints and the real forces acting on the system. An impor-
tant advantage of the d’Alembert’s approach is the deduction of the Lagrangian
equations of motion from the Newtonian equations of motion, giving evidence of the
connection between Newtonian mechanics and Lagrangian mechanics. However, in
the present section, we depart from the Hamilton’s integral principle because it is a
common way to deduce the equations of motion in many fields of physics.
According to the Hamilton’s principle, the stationary requirement on the action,
.δS = 0, represents a formal equation of motion. There are no perceptible forces
pushing the particles in the principle of least action, and in spite of that, the particles
follow an optimal path.5 For practical reasons, we favor differential equations of
motion over an integral equation of motion. However, in order to reach a differential
form of the stationary condition, where the forces become visible, the calculus
of variations should be used. It is an important mathematical tool that provides
differential equations in the search of extremal points (we use it to derive, e.g., the
Laplace equation, the Helmholtz equation, and the wave equation, among others,
and also solve problems like the Dido’s isoperimetric problem and more).
The fact that the action function S is stationary indicates that small variations
from the central trajectory precisely end up in the central trajectory, Fig. 6.3.
The correct nature of the perturbed paths is ensured by requiring mathematically
well-behaved functions .αi , together with the introduction of the small parameter .ε.
Suppose that .qi (t, 0) is the central trajectory and .qi (t, ε) is the perturbed trajectory.
Their relation is .qi (t, ε) = qi (t, 0) + εαi (t). The .αi functions are continuous and
non-singular, with continuous first and second derivatives, vanishing at the fixed
boundary points .t1 and .t2 , .αi (t1 ) = 0 and .αi (t2 ) = 0. We deal with the infinitesimal
character of the variations by transforming the condition .δS = 0 onto the condition
.(∂S/∂ε) dε = 0 up to first order, where differential calculus is employed for
finding extremal points. In this connection, the .δS variation is replaced by partial
derivatives.
5 The unknown properties of spacetime in some sense impose the particle dynamics.
Fig. 6.3 Nature favors processes where the mechanical systems evolve along stationary states of
the action.
f This is, the evolution of the particles between two states follows an extremum path,
.δS = δ L(q, q̇, t) dt = 0, in phase space. This constitutes the principle of least action, and is
based on the balance of the kinetic energy and the potential energy present in the expression of the
Lagrangian, L
f t2 E [ ∂(T −V ) ∂qi ∂(T −V ) ∂ q̇i

]
.
∂S
∂ε dε = t1 i ∂qi ∂ε dε + ∂ q̇i ∂ε dε dt
f f
The integration-by-parts technique, . u dv = uv − v du, is applied to the second
term.
f t2 ∂ ∂ ∂qi
t2
f t2 ( )
∂(T −V ) ∂qi ∂qi d ∂(T −V )
t1 ∂ q̇ (T − V ) ∂t ∂ε dt = ∂ q̇i ∂ε − t1 ∂ε dt ∂ q̇i dt
' i '' ' ' '' ' t1
.
u dv
The factors .∂qi /∂ε are the .αi functions in the expansion of .qi (t, ε). The .αi functions
vanish at the instants of time .t1 and .t2 because .αi (t1 ) and .αi (t2 ) are fixed points. The
surviving terms are collected.
f t2 E [ ∂(T −V ) d ∂(T −V )
]
∂qi
.
∂S
∂ε dε = t1 i ∂qi − dt ∂ q̇i ∂ε dε dt
In order to impose the stationary condition of the action, the term .(∂S/∂ε)dε is
replaced by .δS and similarly .(∂qi /∂ε)dε by .δqi , making the expression zero.
f t2 E [ ∂(T −V ) d ∂(T −V )
]
. δS = t1 i ∂qi − dt ∂ q̇i dt δqi = 0
Since the n degrees of freedom .qi are independent of each other, the variations .δqi
are also independent. The way to satisfy the condition .δS = 0 under general terms
is by making each coefficient of .δqi zero.
[ ]
∂(T −V ) ∂(T −V )
.
d
dt ∂ q̇i − ∂qi =0 ; i = 1, · · · , n
6.4 Lagrange’s Equations of Motion 105
It is time to introduce the scalar-valued Lagrangian function .L = T − V . The

equations are abbreviated using this definition, and .i = 1, · · · , n
[ ]
∂L({qi },{q̇i },t) ∂L({qi },{q̇i },t)
.
d
dt ∂ q̇i − ∂qi =0 (6.7)
These are the Lagrange equations of motion which extremize the action. They
represent a set of coupled second-order differential equations for the independent
degrees of freedom. The solutions describe the behavior of the mechanical system.
The term in square parenthesis is known as the generalized momentum, .pi and is
the conjugate variable of .qi (both form a duality under a Fourier transformation).6
. pi = ∂L/∂ q̇i (6.8)
The generalized momentum .pi corresponds to the linear momentum when the
Cartesian coordinates are used, but it may not necessarily represent the linear
momentum when different coordinates are employed. If it happens in Eq. 6.7 that
.∂L/∂qi = 0 for some .qi , the variable .qi is a cyclic variable or an ignorable variable,
since .qi is not present in L, and .pi is conserved along the physical path.
( )
dpi
.
dt = d
dt
∂L
∂ q̇i =0 ⇒ pi = constant (6.9)
The Lagrangian equations of motion differ from the Newtonian equations of

motion in the use of both the scalar function L and the generalized variables. The
Lagrangian equations of motion have general character but become specific when
T and V are proposed for the system under study. For instance, the Lagrangian
equations of a planet moving in the Solar System are different to the Lagrangian
equations of a charged body moving in an electric field because such systems differ
in the types of interaction. The quantum states, like the excited molecular states, are
excluded by the classical Lagrange equations of motion.
6 In order to explain the conjugation of the variables q and p and simultaneously the independence
of each other, suppose that a first observer in an inertial reference frame knows the position q
and momentum p of a particle: q is an absolute quantity and p is a relative one, as it depends on
relative positions and relative clock times (the Newtonian time intervals of different clocks may not
necessarily be the same, yet we suppose equal clock times for the observers). If the first observer
measures p at .q(t), a second observer may not ensure that p corresponds to the particle at .q(t),
because the second observer is supposed to only know the position .q(t) and requires to know the
change of positions and the relative time clocks registered at the location .q(t) by the first observer
to deduce p at .q(t). In fact, the two observers require an equation (of motion) to link not only the
variables .q(t) ⇔ p(t) but also their observations. The initial conditions uniquely determine such
a relation and help synchronize the variables of the two observers, .q(t0 ) ⇔ p(t0 ) at a given instant
of time, say .t0 = 0. Thus, q and p are not only independent variables but also conjugate variables,
as both are required to define the particle motion.
Problem 6.1 A particle with mass m and instantaneous position .r = (x, y, z)

moves freely. The Lagrangian is .L = mṙ2 /2. An observer rotates around the
z axis with frequency .ω = (0, 0, ω). The coordinates of the particle according
to the observer are:
. x ' = x cos(ωt) + y sin(ωt) ; y ' = y cos(ωt) − x sin(ωt) ; z' = z
Deduce the equations of motion in the rotating (primed) reference system.
Solution The Lagrangian in the primed reference system should be proposed

to eventually deduce the equations of motion. We start by differentiating the
primed position coordinates with respect to time.
ẋ ' = ẋ cos(ωt) − ωx sin(ωt) + ẏ sin(ωt) + ωy cos(ωt)

.
ẏ ' = ẏ cos(ωt) − ωy sin(ωt) − ẋ sin(ωt) − ωx cos(ωt) ; ż' = ż
The quantity .ṙ2 = ẋ 2 + ẏ 2 + ż2 is written in terms of the primed variables in

the form:
ẋ ' − ωy ' = ẋ cos(ωt) + ẏ sin(ωt) ; ẏ ' + ωx ' = ẏ cos(ωt) − ẋ sin(ωt)

.
ṙ2 = (ẋ ' − ωy ' )2 + (ẏ ' + ωx ' )2 + (ż' )2
The result is used to build the transformed Lagrangian.
L= ' − ωy ' )2 + (ẏ ' + ωx ' )2 + (ż' )2 ] = m '

+ ω × r' )2
2 [(ẋ
m
2 (ṙ
.
The equation of motion is determined from .d(∂L/∂ q̇ ' )/dt − ∂L/∂q ' = 0.
) ∂L ) ' ' 2 '
d
dt ∂ ẋ ' − ∂x
∂L
' = m(ẍ − 2ω ẏ − ω x ) = 0
. ( ) ( )
' ' 2 ' = mz̈' = 0
∂ ẏ ' − ∂y ' = m(ÿ + 2ωẋ − ω y ) = 0 ; −
d ∂L ∂L d ∂L ∂L
dt dt ∂ ż' ∂z'
We add the previous expressions to have a single equation of motion in vector

form.
m[r̈' + 2ω(−ẏ ' , ẋ ' , 0) + ω2 (−x ' , −y ' , 0)] = m[r̈' + 2ω × ṙ' + ω × (ω × r' )
.
=0
This is the equation of a particle with angular motion.

6.5 Newtonian Equations of Motion from Lagrange Theory 107
6.5 Newtonian Equations of Motion from Lagrange Theory
In order to illustrate the application ofELagrangian theory, consider a system

of particles with kinetic energy .T = 2
i mi vi /2, interaction potential V with
E
dependence on the spatial coordinates of the particles, such as .V = 2
i kxi /2.
The Cartesian coordinates are conveniently considered the generalized variables in
this case. The Lagrangian equation of motion of the ith particle, with position .xi
and speed .vix in the x axis, acquires the form:
E ( )
mk
d ∂L
dt ∂vix − ∂L
∂xi = d ∂
dt ∂vix j 2 vj2x + vjy jz +
2 + v2 ∂V
∂xi =0
. (6.10)
dvix
mi dt = − ∂x
∂V
i
; ∂V
∂xi = kxi
The equation of motion of particle i is that of the harmonic oscillator, .mi d 2 xi /dt 2 =
−kxi . In terms of the energy functions T and V , and the conjugate variables q and
.q̇ of the Lagrangian, it is observed that:
( )
dp
.
d
dt
∂T
∂ q̇ = − ∂V
∂q ⇒ dt = − ∂V
∂q (6.11)
The Lagrange’s equations of motion are originated from a universal principle, and
they are expected to be invariant with respect to different types of coordinates
and applicable to all the mechanical systems. The Newtonian laws appear to
be implicitly included in the principle of stationary action. In this regard, the
Lagrangian equations of motion are presented as an alternative to the Newtonian
equations of motion. The change of formulations, from Newtonian theory to
Lagrangian theory and vice versa, is particularly useful because some problems are
more amenable under certain formulations than others.
Problem ) 6.2 The )Lagrangian of a particle is given in spherical coordinates

L = m ṙ 2 + r 2 θ̇ 2 /2. Obtain the equations of motion from L. Consider the
.
solution .z = reiθ , and show that it corresponds to a free moving particle.
Solution The independent degrees of freedom are the radial distance r

and the angular variable .θ . The equations of motion of these variables are
.d(∂L/∂ ṙ)/dt − ∂L/∂r = 0, and similarly for .θ . After performing the
derivations, the equations of motion are:
. r̈ − r θ̇ 2 = 0 ; θ̈ + 2
r ṙ θ̇ 2 = 0
(continued)
If .z = reiθ then, by considering the general expression .deif (t) /dt =

ieif (t) df (t)/dt, the second derivative of z is .z̈ = (r̈ − r θ̇ 2 ) + i(2ṙ θ̇ + r θ̈).
According to the equations of motion, each term is zero. Thus, .ż = constant,
indicating that the real and imaginary parts of .z = r cos θ + ir sin θ depend
linearly on t.
. r cos θ = αt + β ; r sin θ = α ' t + β '
The variables r and .θ are coupled in the solutions. By doing .x = r cos θ and
y = r sin θ , we obtain .x = αt + β and .y = α ' t + β ' . Therefore, L describes
.
a free particle in motion.
6.6 Non-uniqueness of the Lagrangian
In principle, it is possible to propose different Lagrangians of a single mechanical

system and obtain the same trajectories of the particles. For instance, we may use
L or .αL, where .α is a real number, and the Lagrangian equations of motion are the
same. The Lagrangians may differ by more than a constant factor. Suppose that L is
the Lagrangian of a system of particles and .L' is a different Lagrangian of the same
system of particles. The interest is to demonstrate that the different Lagrangians
are related by the time derivative of a function, which depends on the generalized
coordinates and time. We start from the stationary action.
f t2
. δ t1 L({qi , q̇i }, t) dt = 0
We assume that .L' is related to L in the form:

dF ({qi },t)
. L' ({qi , q̇i }, t) = L({qi , q̇i }, t) + dt
(6.12)
F is a mathematically well-behaved function of the generalized coordinates and

time. If .L' is the Lagrangian of the same particle system, then it also makes the
action stationary.
f t2 f t2
. δ t1 L' ({qi , q̇i }, t) dt = δ t1 L({qi , q̇i }, t) dt + δF ({qi }, t2 ) − δF ({qi }, t1 ) = 0
The F function shows no variation at .t1 and .t2 , and .δF ({qi }, ti ) = 0, with .i = 1, 2.
f t2 f t2
. δ t1 L({qi , q̇i }, t)dt = δ t1 L' ({qi , q̇i }, t)dt = 0 (6.13)
6.6 Non-uniqueness of the Lagrangian 109
It is observed that .L' is equivalent to the Lagrangian L. In short, a system of particles

may be described by different Lagrangians; however, they are related to each other.
The Lagrangian equations, Eq. 6.7, are invariant under the different representations
of L.
[ ] [ ]
d ∂L({qi }, {q̇i }, t) ∂L({qi }, {q̇i }, t) d ∂L' ({qi }, {q̇i }, t)
. − =
dt ∂ q̇i ∂qi dt ∂ q̇i
∂L' ({qi }, {q̇i }, t)
− =0 ; i = 1, · · · , n
∂qi
Problem 6.3 A particle with mass m and charge e moves in an electromag-

netic field. The electric field, .E, and magnetic field, .B, are given in terms of
the magnetic vector potential, .A, and the electric scalar potential, .φ.
. E = −∇φ − 1 ∂A
c ∂t ; B=∇ ×A
Both .A and .φ depend on the coordinate .r and time t. The speed of light is c.
The Lagrangian of the particle is:
. L = mṙ2 /2 − eφ + e ṙ · A/c
Show that the force exerted on the particle depends on the velocity and
corresponds to the Lorentz force.
Solution The potential is a velocity-dependent function; still, the Lagrangian

approach is appropriate to establish the equation of motion. The strategy is to
obtain the equation of motion and, later, the force acting on the particle. We
compute the generalized momentum from the Lagrangian, .p = ∂L/∂ ṙ.
. p = mṙ + eA/c
We are in the position to formulate the Lagrangian equation of motion

dp/dt − ∂L/∂r = 0, taking care of the vectors in the derivation.
.
d e e e E ∂A e ∂A
. (mṙ + A) + e∇φ − ∇(ṙ · A) = mr̈ + · ṙj + + e∇φ
dt c c c ∂rj c ∂t
j
e
− ∇(ṙ · A) = 0
c
(continued)
For two arbitrary vector functions, .F and .G, we have the vector identity .∇(F ·
G) = F × (∇ × G) + (F · ∇)G + G × (∇ × F) + (G · ∇)F. By choosing .F = ṙ
and .G = A, we obtain:
( )
1 ∂A e
.mr̈ + e ∇φ + + [(ṙ · ∇)A − ∇(ṙ · A)]
c ∂t c
( )
1 ∂A e
= mr̈ + e ∇φ + − [ṙ × (∇ × A)] = 0
c ∂t c
The expressions of the fields .E and .B are employed in the equation.
. mr̈ = eE + (e/c) ṙ × B
This is the equation of motion of the particle; it includes the velocity and
is known as the Lorentz force, an important expression in electromagnetic
theory.
An interesting example of the non-uniqueness property of the Lagrangian is

the gauge invariance in electrodynamics. According to electromagnetic theory,
the vector and scalar potentials, .A and .φ, are non-unique, and different potentials
reproduce the same electric and magnetic fields.
. A' = A + c ∇A ; φ ' = φ − ∂A/∂t
The function .A may be a function of space and time; it is the gauge. The transformed
Lagrangian .L' may be expressed in terms of L, but the expression is simplified when
the total derivative of .A with respect to time is used.
) ) ) ∂A )
L' = mṙ2 /2 − e φ − ∂A
∂t + ce ṙ · (A + c∇A) = L + e ∂t + ṙ · ∇A
.
= L + e dA/dt ; L = mṙ2 /2 − eφ + e ṙ · A/c
This takes us back to Eq. 6.12. In short, the gauge invariance is consistent with
the non-uniqueness property of the Lagrangian. Interestingly, the gauge shift in the
electromagnetic field is coupled with the phase shift in the quantum wave function
of a particle with electric charge. Such a coupling establishes the conservation of
electric charge.
6.7 Invariance of the Lagrange Equations of Motion 111
6.7 Invariance of the Lagrange Equations of Motion
The next interest is to show that the Lagrange equations maintain the same
functional form under a transformation of the conjugate variables, thus making
the Lagrangian theory independent of the types of coordinates used to describe the
system. We start by assuming two sets of variables .{qi } and .{xi }, which can be
transformed onto one another.
qi = qi ({xα }, t) ; xα = xα ({qi }, t)
.
dqi E ∂qi ∂qi dxα E ∂xα ∂xα (6.14)
= α ẋα + ; = i q̇i +
dt ∂xα ∂t dt ∂qi ∂t
We require to show the equivalence between the Lagrange equations in the variables
.qi and .xα by making use of the above transformations. To do this, the variables
.xα are assumed to satisfy the Lagrange equations. We now have to show that
the variables .qi satisfy the Lagrange equations, in other words, that the following
expression is zero:
( ) { ( ) [ ]}
d ∂L ∂L E d ∂L ∂ ẋα ∂L ∂xα ∂L ∂ ẋα
. − = α − +
dt ∂ q̇i ∂qi dt ∂ ẋα ∂ q̇i ∂xα ∂qi ∂ ẋα ∂qi
We use the last identity of Eqs. 6.14 in the last contribution where .ẋα is present.
( ) { ( ) ( )
d ∂L ∂L E d ∂L ∂ ẋα ∂L d ∂ ẋα
− = α + −
dt ∂ q̇i ∂qi dt ∂ ẋα ∂ q̇i ∂ ẋα dt ∂ q̇i
. [ ( )]}
∂L ∂xα ∂L ∂ E ∂xα ∂xα
− + q̇j +
∂xα ∂qi ∂ ẋα ∂qi j ∂qj ∂t
By carrying the derivation .d(∂ ẋα /∂ q̇i )/dt in the last term of the first line and
applying the derivative .∂/∂qi in the last term of the second line, we have:
d ∂L ∂L E [ d ( ∂L ) ∂ ẋα ∂L ∂xα
(
∂L d ∂xα
)
. − = − +
dt ∂ q̇i ∂qi α
dt ∂ ẋα ∂ q̇i ∂xα ∂qi ∂ ẋα dt ∂qi
⎤
( )
∂L E ∂ 2 xα ∂ ∂xα ⎦
− q̇j +
∂ ẋα ∂qi ∂qj ∂qi ∂t
j
By using the chain rule in the term .d(∂xα /∂qi )/dt, it is observed that it reproduces
the summation of the last contribution of the above equation but with opposite sign.
( ) [ ( ) ]
d ∂L ∂L E d ∂L ∂ ẋα ∂L ∂xα
. − = α −
dt ∂ q̇i ∂qi dt ∂ ẋα ∂ q̇i ∂xα ∂qi
Once more, we use the last identity of Eqs. 6.14 to show the equivalence .∂ ẋα /∂ q̇i =
∂xα /∂qi . By using this identity in the above expression, we have:
( ) [ ( ) ]
d ∂L ∂L E d ∂L ∂L ∂xα
. − = α −
dt ∂ q̇i ∂qi dt ∂ ẋα ∂xα ∂qi
If the relation between the coordinates .qi and .xα is invertible, then we know that
.∂xα /∂qi /= 0, and it is the other coefficient which should be zero. By considering
that the variables .xα satisfy the Lagrange equations, we conclude that the variables
.qi equally satisfy such equations.
( )
d ∂L ∂L
. − =0
dt ∂ q̇i ∂qi
6.8 Motion with Constraints
When a system of particles is subjected to restraints, the Lagrangian equations of

motion have to consider the imposed conditions to properly describe the particle
behaviors. We assume a number r of restraints denoted by functions .fi , involving
the variables .qi and time t.
. f1 (q1 , · · · , qn , t) = 0 ; f2 (q1 , · · · , qn , t) = 0 ··· fr (q1 , · · · , qn , t) = 0

(6.15)
The constraints .fi may be of holonomic type, and some others may be of non-
holonomic type. An example of common holonomic constraints is the rigid bond,
.|ri (t) − rj (t)| = cij , where the coordinates of selected particles participate.
7
There is a number r of variables .qj which are not independent of the others.
For mathematical manipulations, it is convenient to consider the restraints in their
differential forms.
En
dfi /dt = 0 ⇒ j =1 cij dqj + cit dt = 0
. (6.16)
cij = ∂fi /∂qj ; cit = ∂fi /∂t ; i = 1, · · · , r
The r equations on the variables .qj are linear and integrable expressions. It is
important that the differential forms of the constraints acquire expressions similar
to the last identity of the first line. The constraints expressed with inequalities
are avoided. The strategy to properly describe the time evolution of the particles
consists on incorporating the restraints into the Hamilton’s integral to define an
7 Ina restrained bond like .|ri (t) − rj (t)| = cij , the atoms i and j move according to an equation
of motion, and the distance between them is concurrently maintained constant by satisfying the
constraint that establishes the relation between the variables .ri and .rj .
6.8 Motion with Constraints 113
unconstrained Lagrangian from which the equations of motion should be deduced.

In this regard, the equations of motion derived from the unconstrained Lagrangian
should generate the same phase-space points as those deduced from the constrained
Lagrangian, where the generalized coordinates also satisfy the r constraints .fi , but
in separate form from the ordinary Lagrangian equations of motion.
In this process, the initial independent variables are .n − r due to r constraints.
By introducing the differential forms of the r constraints into the motion equations,
we are able to have n differential equations to solve, with the cost of introducing r
Lagrange multipliers. So we now have .n + r variables and only n motion equations.
However, we also have to satisfy the r constraints in their non-differential forms.
By considering these additional equations, we are able to have a system of .n + r
equations with .n + r unknowns.
The first step to incorporate the constraints into the equations of motion is to deal
with a number r of additional terms to the n variables .qj and add an equal number
of equations with those terms to determine the solutions in unique form. The new
terms are denoted by .λi ; they are linked to the coordinates .qj and time t and are
referred to as the Lagrange multipliers. The number r of multipliers corresponds
to the r constraints. The constraints are introduced in the main equations via their
differential forms 6.16, because such forms can be accommodated in the integral
expression of the action. Since the variations .δqi of the generalized coordinates
produce deviations .δfi , we have:
En f t2 [E En ]
r
. λi δfi = λi j =1 cij δqj = 0 ⇒ t1 dt i=1 j =1 λi cij δqj = 0
(6.17)
The approach to be displayed is known as the Lagrange method of undetermined
multipliers [4]. The constraints in their differential forms are now included (as a
“smart zero”) in the kernel of the Hamilton’s integral to follow a similar treatment
in the deduction of the Lagrangian equations of motion.
f t2 En [ Er ]
.
t1 dt j =1
∂L
∂qj − d ∂L
dt ∂ q̇j + i=1 λ i cij δqj = 0 (6.18)
We have the freedom to choose the coefficients .λi . They are proposed in such a way
to satisfy the equations:
Er
.
∂L
∂qj − d ∂L
dt ∂ q̇j + i=1 λi cij =0 ; j = 1, · · · , r (6.19)
It is with these equations that the multipliers are included in the dynamics. After
making zero the first r terms of Eq. 6.18, the Hamilton’s integral maintains the
functional form for the remaining terms with subindexes .r + 1, r + 2, · · · , n.
f t2 En [ Er ]
.
t1 dt j =r+1
∂L
∂qj − d ∂L
dt ∂ q̇j + i=1 λi cij δqj = 0 (6.20)
Once the Lagrange multipliers are decided, the coordinates .qr+1 , · · · , qn in the
kernel are now independent of each other. Thus, for arbitrary variations of .δqj , the
terms in square brackets are equal to zero.
Er
.
∂L
∂qj − d ∂L
dt ∂ q̇j + i=1 λi cij =0 ; j = r + 1, · · · , n (6.21)
It is convenient to write Eqs. 6.19 and 6.21 together because they have the same
forms.
Er
.
d ∂L
dt ∂ q̇j − ∂L
∂qj = i=1 λi cij ; j = 1, · · · , n (6.22)
At this stage, the equations of motion include a number r of constraints in

their differential forms and using the Lagrange factors .λi . We have n differential
equations of the kind 6.22 but .n + r unknowns. In order to have the number of
equations equal to the number of variables, we consider the r constraints of type
.fi (in their non-differential forms). This procedure establishes a system of .n + r
variables with .n+r equations. Once the solutions of Eqs. 6.15 and 6.22 are obtained
consistently, we have the trajectory of the particles simultaneously satisfying the
constraints. In this regard, the point .(q, λ) of the unconstrained Lagrangian contains
the point .(q, 0) of the constrained Lagrangian in phase space. The equations are
further simplified by using Eq. 6.16 for the coefficients .cij , that is, .cij = ∂fi /∂qj ,
and introducing the generalized forces, .Qj , in Eq. 6.22.
Er ∂fi
.
d ∂L
dt ∂ q̇j − ∂L
∂qj = Qj ; Qj = i=1 λi ∂qj ; j = 1, · · · , n (6.23)
The Lagrange multipliers .λi are identified in terms of the basic variables and the
generalized forces from the Lagrange multipliers. For conservative systems, we
have .Qj = 0, because the Lagrange equations already consider the cases with
conservative forces via the V potential of the Lagrangian. However, in addition to
the constraining forces, the .Qj may equally represent generalized non-conservative
forces as they are restrictive forces on the particle dynamics. The relation of the
.Qj forces to the standard .fj forces is established with the transformations of the
coordinates .ri and .qi .

E ∂ri E E ∂ri E
. δri = j ∂qj δqj ; i fi · δri = i,j fi · ∂qj δqj = j Qj δqj
The .Qj forces are defined in terms of the standard forces .fi and positions .ri in the
form:
E ∂ri
. Qj = i fi · ∂qj (6.24)
The generalized forces may represent not only forces but also torques. In special
cases, the non-conservative forces may be derived from a function, say D, known
as a generalized potential function. In order to illustrate the Lagrange equations of
6.8 Motion with Constraints 115
motion with a generalized potential function, let us consider Rayleigh dissipative

forces with dependence on the speeds of the particles and constant coefficients .kx ,
.ky , .kz .
En ( )
. D= i=1
2 + k v 2 + k v 2 /2
kx vix y iy z iz (6.25)
The generalized potential function may take a linear form on the kinetic energy
or a quadratic form on the speeds. EIt differs from the potential which is quadratic
on the position coordinates, .V = i (kx xi + ky yi + kz zi )/2. The quantity 2D
2 2 2
is understood as the time rate of energy dissipation due to the friction force .ff =
−∇v D.
( )
dE f
.
dt = −f f · dr = −(ff · v) = (∇ D · v) = k v 2 + k v 2 + k v 2 = 2D
dt v x x y y z z
The generalized force associated with the friction force is obtained from Eq. 6.24
and using the fact .∂ ẋα /∂ q̇i = ∂xα /∂qi from Eqs. 6.14.
E f ∂ri E ∂ri E ∂ ṙi
. Qj = i fi · ∂qj =− i ∇v i D · ∂qj =− i ∇v i D · ∂ q̇j = − ∂∂D
q̇j
The equation of motion 6.23 of a particle that includes the Rayleigh dissipative force
D is:
.
d ∂L
dt ∂ q̇j − ∂L
∂qj = − ∂∂D
q̇j (6.26)
By considering Cartesian coordinates in the Lagrangian, .L = T − V , and using the

above expression, we identify equations of motion which include kinematic friction
effects.
[ ]
d ∂ E mk
dt ∂vix − ∂xi = dt ∂vix
d ∂L ∂L 2 + v 2 + v 2 + ∂V = − ∂D
vkx
k 2 ky kz ∂xi ∂vix
. (6.27)
dpix
dt = − ∂V
∂xi − k x vix
The viscous term is negatively proportional to the particle speed. When the vis-
cous term is neglected, the traditional Newtonian equation of motion is reproduced.
A disk rolling down a tilted plane illustrates the importance of the generalized
forces, Fig. 6.4. In conclusion, the constraining forces may be incorporated into the
equations of motion via the Lagrange method of undetermined multipliers. Such
a method is also employed in other fields of science, like in statistical physics to
obtain the maximum entropy of a system with a large number of particles.
Fig. 6.4 A disk of mass m, radius R, and inertia moment .I = mR 2 /2, rolls down a tilted plane
without slipping due to the action of gravity. The plane has length L and makes an angle .φ with
the horizontal axis. The instantaneous position of the disk in the plane is .(L − z). The radius R is
sufficiently small with respect to the instantaneous height h in the plane (.h >> R)
Problem 6.4 A disk with mass m, radius R, and inertia moment .I = mR 2 /2

rolls down a tilted plane without slipping due to the action of gravity. The
plane has length L and makes an angle .φ with the horizontal axis, Fig. 6.4.
Obtain the equations of motion.
Solution The kinetic energy of the disk is due to translation and rotation,
Sect. 5.3, .T = mż2 /2+I θ̇ 2 /2. The position of the disk on the plane is .(L−z),
and the height of the disk mass center is .h = (L − z) sin φ. The corresponding
potential energy is .V = mg(L − z) sin φ. The Lagrangian is:
mR 2
. L=T −V = m 2
2 ż + 4 θ̇ 2 − mg(L − z) sin φ
The non-sliding constraint on the disk motion relates the translation distance
to the rotated distance of the disk by .dz = Rdθ , and thus, the constraint is
.f = z − Rθ = 0. The equations of motion in z and .θ are determined from
Eqs. 6.23.
mR 2
. mz̈ − mg sin φ − λ = 0 ; 2 θ̈ + λR = 0 ; f = z − Rθ = 0
These are three coupled equations with unknown variables z, .θ , and .λ. The
equations are decoupled by using the last identity in the second equation and
solving for .λ, which in turn is used in the equation of z. The final expressions
are:
2 g sin φ
. z̈ = 2
3 g sin φ ; θ̈ = 3 R ; λ = − 31 mg sin φ
The restraining forces, Eqs. 6.23, in terms of the basic variables are:
. Qz = λ ∂f
∂z = λ = − 3 mg sin φ
1
; Qθ = λ ∂f
∂θ = −λR =
1
3 mgR sin φ
6.9 Hamilton’s Function 117
6.9 Hamilton’s Function
In this section, we define the .H function, known as the Hamiltonian. It is a quantity

of significance on the same right as the Lagrangian, because it equally encodes the
relevant information of the system. The Lagrangian and Hamiltonian are related to
each other. In this regard, it is also possible to formulate equations of motion based
on the Hamiltonian.8
In the following lines, we demonstrate that the Lagrangian and Hamiltonian are
linked by a Legendre transformation.9 To do this, we consider a function .f (x)
encoding information in terms of the variable x. The Legendre transform is used
to obtain an equivalent function, say g, that contains the same information of f but
in terms of the variable s, were s is the derivative of f with respect to x, .s = df/dx.
A geometrical interpretation of the Legendre transform is given in Fig. 6.5. In order
to link f and g, the functions .x(s) and .s(x) should be single-valued expressions
with a one-to-one mapping relation, showing the possibility of inverting them. The
Legendre transformation considers a symmetric relation between the functions f
and g and their variables x and s in the form:
. f (x) + g(s) = sx (6.28)
The variables x and s are conjugate to each other, but only one of them is the
independent variable. When x plays the role of the independent variable, we have
.f (x) + g(s(x)) = s(x) x, and for the case of s, it is .f (x(s)) + g(s) = sx(s). By
differentiating expression 6.28 with respect to x and s, the symmetric expressions

in the variables x and s are observed.
⎫
df (x)
dx + dg(s(x)) ds
ds dx = x ds
dx + s(x) ⎬
. ⇒ x(s) = dg(s) ; s(x) = dfdx(x)
dg(s) df (x(s)) dx ⎭ ds
ds + ds = s ds + x(s)
dx
dx
(6.29)
From the results on the right-hand side, the symmetry between f and g goes to
higher derivatives.
⎫
d2g ⎬
dx
ds = ds 2 ) dx ) ) ds ) d2g d2f
. ⇒ =1 ; =1
d2f ⎭ ds dx ds 2 dx 2
ds
dx = dx 2
8 The Lagrangian function stores information of the particle system in the form .L = T − V . From
an initial impression, one may wonder on the possibility of similarly storing that information in a
function, say .H, of the same particle system in the form .H = T + V . Such a possibility motivates
the exploration of an equivalent approach in terms of .H, as an alternative to L. This is a rational
ground for the proposal of a new formulation, which we now recognize as Hamiltonian theory.
9 In addition to the Fourier and Laplace transforms, the Legendre transform presents an alternative
way to characterize a function in terms of its own derivatives: Zia et al. [5].
Fig. 6.5 Inset A shows the plot of a function f in terms of the independent variable x. The
function is convex and smooth. Inset B shows similar curves, they all originate tangent lines equal
to the tangent lines of the curve in A (depicted at the bottom of inset B), but the tangent lines differ
in the intersections with the vertical axis. The function .f (x) of inset A is implicitly determined
when the slopes and the correct intersection points with the vertical axis of the tangent lines are
found, inset C. A tangent line at x is shown in inset D. It has slope s, and the intersection point
with the vertical axis is .−g. The quantity g is visualized as a function of the variable s. The identity
.y = sx is the equation of the line when the origin is at point g. The equation of the tangent line is
described by .f (x) = sx − g(s), where f depends on x and g depends on s
One of the important properties of the Legendre transform resides on the fact
that the transform of g returns the function f and vice versa. The results may be
generalized to any number of independent variables. The Legendre transform of the
Lagrangian takes us to the formal proposal of the Hamiltonian, understood as an
alternative function to code the information of the particle system. In analogy to
Eq. 6.28, the link between the Hamiltonian and Lagrangian is given in terms of the
generalized variables.
E
. H+L= i q̇i pi ; pi = ∂L
∂ q̇i (6.30)
This is the general definition of .H. The derivative of L with respect to the coordinate
q̇i is shown in terms of the generalized momentum .pi , as required by Eq. 6.29. The
.
generalized momentum is an independent degree of freedom that also introduces

flexibility to choose the variables to describe the mechanical system. The basic
variables of the Lagrangian are .qi , .q̇i and time t, while the basic variables of the
Hamiltonian are .qi , .pi , and t.
6.10 Preservation of the Hamiltonian in Time 119
6.10 Preservation of the Hamiltonian in Time
The next interest is to show the preservation of the Hamiltonian with time. This is
especially important since the conserved quantities reduce the number of degrees of
freedom and are useful to monitor the proper evolution of the particle dynamics. In
this respect, we introduce the definition of a constant of motion: a function f with
dependence on the variables .qi , .q̇i and time t is a constant, or integral of motion,
when the total time derivative is zero.
df E ∂f ∂f ∂f
.
dt = i ∂qi q̇i + ∂ q̇i q̈i + ∂t =0 (6.31)
This expression represents a way to find the time invariant observables (a different
method to find integrals of motion is the Hamilton-Jacobi formulation of Sect. 6.18).
If the function .H is invariant with time, its total time derivative should be zero. Let
us assume an isolated system of particles. The goal is to verify that .H is a constant
of motion. By differentiating the Lagrangian with respect to t, we have:
E [ ∂L dqi ∂L d q̇i
]
.
dL
dt = i ∂qi dt + ∂ q̇i dt + ∂L
∂t
According to the Lagrangian equations, the term .d(∂L/∂ q̇i )/dt replaces .∂L/∂qi .
E [ d ( ∂L ) ∂L d q̇i
] E ( )
.
dL
dt = i dt ∂ q̇i q̇i + ∂ q̇i dt + ∂L
∂t = d
i dt q̇i +
q̇i ∂∂L ∂L
∂t
The left- and right-hand-side members are gathered together, and the identity 6.30
is used.
( ) ) ) ∂L
d E d E dH
i q̇i ∂ q̇i − L + ∂t = dt i q̇i pi − L + ∂t = 0 ⇒ dt = − ∂t
∂L ∂L ∂L
.
dt
The requirements to have .dH/dt equal to zero are to: (i) satisfy the Lagrangian
equations of motion for every independent degree of freedom, and (ii) have L with
no explicit dependence on the time, .∂L/∂t = 0. These requirements make the
Hamiltonian a constant of motion in a conservative system.
. H = constant ; conservative system
The form of .H for the system of particles is determined from .L = T − V , with the
kinetic energy given as a homogeneous quadratic function of .q̇i .
E E E
T = ij cij q̇i q̇j ; ∂T
∂ q̇k =2 i cik q̇i ; 2T = ∂T
k q̇k ∂ q̇k
. E (6.32)
H= i q̇i pi − L = 2T − (T − V ) = T + V ⇒ H=E
The Hamiltonian is identified with the energy of the isolated system and is conserved
in the particle dynamics. Historically, the energy and momentum were the first
physical observables that were found to be invariant with time after investigating
an isolated system of colliding particles. If a frictional force is considered in the
potential energy V , then the work done by that force on a particle depends on the
trajectory, the system is not conservative, the process is time irreversible, and the
Hamiltonian is no more a constant of motion. Furthermore, .H does not represent
the energy of the system and is not conserved. This usually occurs when the time
is present in the transformation of coordinates. In the following section, we identify
the symmetries of nature associated with the preservation of the energy, momentum,
and more.
Problem 6.5 A trolley moves, and a particle of mass m is in the interior of

the trolley attached to a spring with constant k. An external force keeps the
trolley moving with constant speed v. The coordinate describing the particle
with respect to a static reference frame is q. The Lagrangian of the particle
and the corresponding equation of motion are:
( )
mq̇ 2
. L= 2 − k2 (q − vt)2 ; d
dt
∂L
∂ q̇ + ∂L
∂q = mq̈ + k(q − vt) = 0
The time-dependent Hamiltonian of the particle is .H = p2 /(2m) + (k/2)(q −

vt)2 . Define the new variable .q ' = q − vt, which describes the position of the
particle in the trolley. Deduce the equation of motion, obtain the transformed
Lagrangian and Hamiltonian in terms of .q ' and the momentum .p' , and derive
a time-independent Hamiltonian.
Solution The equation of motion is obtained by replacing q by .(q ' + vt). It

is .mq̈ ' + kq ' = 0. The transformed Lagrangian is similarly computed.
. L' = (m/2)(q̇ ' + v)2 − (k/2)(q ' )2
The expression of the Hamiltonian requires the calculation of the new

momentum.
. p' = ∂L' /∂ q̇ ' = m(q̇ ' + v)
The primed Hamiltonian is .H' = p' q̇ ' − L' = (mq̇ ' + mv)q̇ ' − m(q̇ ' )2 /2 −
mq̇ ' v−mv 2 /2+k(q ' )2 /2. This expression is changed by using the momentum
'
.p .
. H' = (p' )2 − 2mvp' /(2m) + (k/2)(q ' )2
(continued)
6.11 Conserved Observables and Symmetries 121
The Hamiltonian shows no time dependence, and it is not the energy because
there is an external force on the trolley that affects the particle. Therefore, the
Hamiltonian of a system may not necessarily be the system energy, and the
Hamiltonian is not used to define a conservative system.
6.11 Conserved Observables and Symmetries
There are fundamental variables in molecular dynamics, and many of them are
conserved over time under the appropriate conditions. The interest in this section is
to discuss the conservation of physical variables and their relation to the symmetries
observed in nature. To do this, we present the Noether’s theorem. The theorem is
a generalization of the results obtained on cyclic coordinates, where the absence
of .qi in the Lagrangian is the cause to preserve the generalized momentum .pi ,
Eq. 6.9. The Noether’s theorem is of central importance to establish the intrinsic
links between the conserved observables of particle systems and the associated
symmetries of nature in different fields of physics.
Statement 6.2 Noether’s theorem: for a system with a continuous symmetry,

there is a corresponding time invariant observable determined from the action.
In order to demonstrate the theorem, we obtain the action S after integrating a

Lagrangian L, which is perturbed in the generalized coordinates q and time t by
small amounts:
q ' = q + δq ; t ' = t + δt
.
δq = εQ ; δt = εT
The quantity .ε is a small real factor, Q is a real constant playing the role of a
generator of the generalized coordinates, and T the generator of the time evolution.
t ' = t + εT ; q ' (t ' ) = f (q(t), ε) ; q̇ ' (t ' ) = ∂f (q(t),ε)

∂q q̇(t)
.
f t2 f t2 +εT
S(t) = t1 L(q(t), q̇(t), t) dt → S(ε) = t1 +εT L(q ' (t ' ), q̇ ' (t ' ), t ' ) dt '
(6.33)
In the process, we shall identify the variables playing the roles of the generators
T and Q and the corresponding symmetries. The function f represents the
transformation of the coordinate q in going from time t to .t ' . The strategy is to
integrate the perturbed Lagrangian in .S(ε) and achieve a pair of expressions with
the same functional form and their difference giving zero. One of them shall be
evaluated at time .t1 and the other expression at time .t2 . By equating them, we should
reach the conclusion that the expression under consideration is unique and time
independent. We start from the fact that S is stationary under small perturbations
and then we differentiate S with respect to .ε.
f t2 +δt [ ∂L ∂q ' ∂L ∂ q̇ ' ∂L(q ' (t ' ),q̇ ' (t ' ),t ' ) ∂t '
]
.
d
dε S(ε) = t1 +δt ∂q ' ∂ε + ∂ q̇ ' ∂ε + ∂t ' ∂ε dt ' = 0
The differentiation of S with respect to .ε avoids the direct differentiation of S with

respect t. The differentiation is assisted by the chain rule due to the dependencies
on the mechanical variables with respect to time:
f t2 +δt∂L ∂t '
f t2 +δt
t1 +δt ∂t ' ∂ε dt ' = ∂L
t1 +δt ∂t ' T dt ' = (AL)T
t=t2
. AL = L(q(t), q̇(t), t)
t=t1
f [ ( ) ( )]
t +δt ∂f ∂q ∂f ∂f (q(t),ε)
dε S(ε) = (AL)T + t1 +δt ∂q ' ∂q ∂t ∂ε + ∂ε + ∂ q̇ '
d 2 ∂L ∂t ∂L ∂
∂ε ∂q q̇(t) dt
where the identities of the first line of Eqs. 6.33 were used and are shown in the
round parenthesis above. In order to continue evaluating the integral, we display the
derivatives of the second term in parenthesis.
( ) [( ) ]
∂f (q(t),ε) ∂ ∂f ∂f ∂ q̇
∂L ∂
∂ q̇ ' ∂ε ∂q q̇(t) = ∂∂L q̇ ' ∂ε ∂q q̇ + ∂q ∂ε
. [ ( )( ) 2
]
∂f ∂q ∂t ∂ f ∂f ∂ q̇ ∂t
= ∂L
∂ q̇ '
∂
∂q ∂q ∂t ∂ε q̇ + ∂ε∂q q̇ + ∂q ∂t ∂ε
We also replace derivatives like the following: .q̇ = ∂q/∂t, .q̈ = ∂ q̇/∂t, and .∂t/∂ε =
−T .
f t2 [ ( )
d ∂L ∂f ∂f
. S(ε) = (AL)T + − q̇T +
dε ε=0 t1 ∂q ' ∂q ∂ε
( 2 )]
∂L ∂ f 2 ∂ 2f ∂f
+ ' − 2 q̇ T + q̇ − q̈T dt
∂ q̇ ∂q ∂ε∂q ∂q ε=0
Once the derivatives with respect to .ε are performed, we consider .ε = 0 (which

leads to dropping the primes on the variables). The terms with negative signs are
rewritten using the Lagrangian equations of motion.
6.11 Conserved Observables and Symmetries 123
( ) ( ) ( )
∂L ∂f ∂f d ∂f ∂L ∂f
d
dt ∂ q̇ ∂q q̇T = d ∂L
dt ∂ q̇ ∂q q̇T + ∂L
∂ q̇ dt ∂q q̇T + ∂ q̇ ∂q q̈T
.
( )
∂L ∂f ∂2f ∂L ∂f
= ∂q ∂q q̇T + ∂L
∂ q̇ ∂q 2
q̇ 2 T + ∂ q̇ ∂q q̈T
The integral is simplified by inserting the last result in the integral, keeping
the old terms of the kernel that carry positive signs inside the integral, and
applying the fundamental theorem of calculus on the relation between derivation
and integration.10
f t2 [ ( ) ( )]
d ' ∂L ∂f ∂L ∂ 2f
. S(ε) = AL T − AL T + + q̇ dt ;
dε ε=0 t1 ∂q ∂ε ∂ q̇ ∂ε∂q ε=0
∂L ∂f t=t2
AL' = q̇
∂ q̇ ∂q t=t1
Once more, the Lagrangian equations of motion are employed to reduce the integral.
f t2 [ ∂L ( ∂f ) (
∂2f
)] f t2 (
∂L ∂f
)
.
t1 ∂q ∂ε + ∂L
∂ q̇ ∂ε∂q q̇ dt = t1
d
dt ∂ q̇ ∂ε dt
ε=0
It is now time to collect the final results of the integrals.

[ ]t=t2
∂L ∂f ∂L ∂f
.
d
dε S(ε) = L(q(t), q̇(t), t) T − ∂ q̇ ∂q q̇ T + ∂ q̇ ∂ε t=t =0
1
In the limit .ε = 0, the coordinate q is recovered from f due to the equivalence

f (q, 0) = q, which in turn leads to .∂f/∂q = 1.
.
[ ]
∂L ∂f
L(q(t), q̇(t), t) T − ∂L
∂ q̇ q̇ T + ∂ q̇ ∂ε t=t
2
. [ ]
∂L ∂f
= L(q(t), q̇(t), t) T − ∂ q̇ q̇ T + ∂ q̇ ∂ε
∂L
t=t1
The expressions on the left- and right-hand sides have equal functional forms, but
they are evaluated at different instants of time. In this regard, we have a unique
expression preserved with time.
∂L ∂f
. L(q(t), q̇(t), t) T − ∂L
∂ q̇ q̇ T + ∂ q̇ ∂ε = constant (6.34)
f x functions in the interval .[a, x]. If the

10 Suppose that f and F are continuous and differentiable
function F is obtained from f by integration, .F (x) = a f (y) dy, then f is obtained from F by
differentiation, .f (x) = F ' (x).
This is the main result of the Noether’s theorem using a single degree of freedom and
for a system of classical particles. It leads to the existence of conserved observables
for the symmetries that satisfy the previous result. The integrals of motion are
derived from the type of perturbation produced in the generalized coordinates.
For example, the translation and rotation of the coordinates are associated with
a homogeneous and isotropic space, respectively. The Noether’s theorem changes
forms in the different fields of physics. For example, it changes its form at the
quantum level because the quantized variables are not continuous. The appropriate
expression in quantum field theory is known as the Ward-Takahashi equation. The
following goal is to illustrate the Noether’s theorem by making reference to: (i)
the homogeneity of space, (ii) the isotropy of space, and (iii) the invariance of the
Lagrangian with the passage of time. These cases show the way in which space and
time mold the laws of physics.
6.12 Space Homogeneity
We start by considering the invariance of L with respect to the spatial translation of a

particle (if the space shows no change in the direction of the particle translation, then
no change of the particle momentum is expected). Consider a Lagrangian, which is
symmetric with respect to the translations of the particle coordinates.
. L(ri , ṙi , t) = L(ri + δr ê, ṙi , t) (6.35)
The unit vector .ê indicates the direction of translation. The expression embodies
the homogeneity of the physical space since no portion of it is privileged. In order
to apply the theorem in the form of Eq. 6.34, we observe that there is no time
perturbation, .T = 0; however, .εQ = δr ê. According to the spatial perturbation,
we have .∂f/∂ε = ∂(q + εQ)/∂ε = Q, and the preserved variable is:
∂L ∂f ∂f
L(q(t), q̇(t), t) T − ∂L
∂ q̇ q̇ T + ∂ q̇ ∂ε = constant ; T =0 ; ∂ε =Q
.
p · ê = constant
(6.36)
The application of the theorem gives the conservation of the linear momentum in
the .ê direction. A similar situation applies to the center of mass of a particle system
by summing over the degrees of freedom pointing in the .ê direction, resulting on the
conservation of the total linear momentum of the particle system. The conclusion is
the association of the space homogeneity to the preservation of the linear momentum
of the particle, a characteristic that was anticipated in Sect. 2.1. The preservation of
the linear momentum is broken when, for example, a time-varying external field acts
upon the particle as, in some sense, it is equivalent to alter the space homogeneity
of the system. Yet the preservation of the linear momentum is not a proof of
the emptiness of the physical space (because, in principle, there is a Higgs field
permeating the space).
6.14 Uniform Passage of Time 125
Fig. 6.6 The position of a

particle is rotated by an
infinitesimal angle .δθ around
the axis .ê. The change of the
vector position .r is given by
.δr = δθ ê × r. According to
Noether’s theorem, the
conservation of angular
momentum demands space
isotropy
6.13 Space Isotropy
We analyze the isotropy of space with implications on the rotation of a particle (if
the space shows no change in the region of the particle rotation, then no change of
the body’s angular momentum is expected). Consider the vector .ê as the rotation
axis of the particle. The vector position of the particle is .r (Cartesian coordinates
are used in this discussion). The vector perpendicular to the plane described by the
two vectors is .ê × r. An infinitesimal rotation of the vector position is .r + εê × r,
where .ε is the real number concerned with the infinitesimal rotation, Fig. 6.6.
It is time to apply the Noether’s theorem in the form exposed in Eq. 6.34. There
is no time perturbation, .T = 0. According to the spatial perturbation, we have
.∂f/∂ε = ∂(q + εQ)/∂ε = ê × r.
∂L ∂f
L(q(t), q̇(t), t) T − ∂L
∂ q̇ q̇ T + ∂ q̇ ∂ε = constant
. (6.37)
p · (ê × r) = constant
By using the properties of the cross product .a · (b × c) = b · (c × a), the term on the
left-hand side is written in the form .ê · (r × p), resulting on the conservation of the
angular momentum projected onto the .ê direction of the rotation axis. Therefore, the
conservation of the angular momentum is associated with the space isotropy.
6.14 Uniform Passage of Time
The next application of the Noether’s theorem involves the uniform passage of time
(if the passage of time is uniform, then no change of the particle energy is expected).
Suppose that the Lagrangian L remains constant as time passes. We are interested in
finding the observable that is conserved for the particle under the above condition.
To do this, we follow similar steps to those of the preceding sections. We produce a
shift of the time origin, .T = 1, while the coordinate generator Q is null.
∂L ∂f
L(q(t), q̇(t), t) T − ∂L
∂ q̇ q̇ T + ∂ q̇ ∂ε = constant ; Q=0 ; T =1
.
−L + ∂L
∂ q̇ q̇ = H = constant
(6.38)
where the Hamiltonian .H given according to Eq. 6.30 was used. Since .H is a
constant in time, the Hamiltonian is an integral of motion. In other words, the
uniform passage of time in an isolated particle results on the conservation of the
energy .H = E. Such a result is consistent with that of Eq. 6.32. The Noether’s
theorem indirectly implies the first Newton’s law under the supposition that the
homogeneity and isotropy of the 3D space are altered with the introduction of
an external forcefield, thus affecting the motion state of the body. Moreover, the
Noether’s theorem may be understood as a generalization of the first Newton’s
law under the supposition that the uniform passage of time is changed by the
introduction of an external forcefield capable of changing the spacetime mesh, thus
affecting the energy of the body.
The general conclusion is that the linear momentum, angular momentum, and
energy are preserved by virtue of the invariance of the Lagrangian with respect
to the homogeneity and isotropy of the 3D space and the uniform passage of
time, respectively. The Noether’s theorem is the formal way to indicate that the
laws of physics do not change with the location in space and the passage of time.
The Noether’s theorem is frequently used in the design of new Lagrangians by
demanding the proper symmetries and reproducing the physical laws observed
under experimentation. For instance, the Lagrangian in the theory of general
relativity for a particle in the presence of an electromagnetic field is proposed by:
/
μ dx ν dx μ (t)
. L = −m0 c −gμν (x(t)) dx
dt dt +e dt Aμ (x(t))
The factor .gμν is the metric tensor of the gravitational potential, the variables .xμ are
the spacetime coordinates, and .Aμ is the electromagnetic four-potential.
Problem 6.6 The relativistic Lagrangian in the theory of special relativity of

a particle subjected to the potential U is given in an inertial reference frame
by:
/
. L = −m0 c2 1 − v 2 /c2 − U (x)
Obtain the Lagrangian equation of motion, and compare with the Newtonian
equation of motion. Consider the motion of the particle in the presence of an
electric field .E and magnetic field .B, given in terms of the vector potential .A
and scalar potential .φ, namely, .E = −∇φ − (1/c) ∂A/∂t and .B = ∇ × A,
problem 6.3. Obtain the Lagrangian equation of motion.
(continued)
6.15 Hamilton’s Equations of Motion 127
Solution The generalized coordinates are the Cartesian coordinates.

) ∂L ) /
v2
. d
dt ∂v − ∂L
∂x =0 ⇒ d
dt (mv) = − ∂U∂x(x) ; m = m0 / 1 − c2
The equation of motion in the relativistic frame is similar to the Newtonian

equation of motion, except for the use of the “relativistic mass,” m, defined
in terms of the particle speed. The Lagrangian of the Newtonian particle is
invariant under Galilean transformations, but the Lagrangian of the relativistic
particle is invariant under Lorentz transformations. The Lagrangian of the
relativistic particle may be mixed with that of a particle moving in an
electromagnetic field, problem 6.3, by considering .U = eφ − e ṙ · A/c. The
equation of motion corresponds to the Lorentz force law.
.
d
dt (mv) = eE + ec ṙ × B
The preservation of the energy, linear momentum, and angular momentum is

used to monitor the proper evolution of the molecular dynamics simulations. The
conservation laws (involving observables related to both the space and the time) are
formulated in a single conservation law in general relativity based on the stress-
energy tensor, because space and time are intertwined together. Some of the main
limitations of the Noether’s theorem in the present form involve the inability to
consider discrete symmetries, like the reflection of a system of coordinates, and
the impossibility to deal with discrete variables. Such cases are properly addressed
with a similar theorem in quantum mechanics. On the other side, the electric charge
conservation of a charged particle is properly addressed by an equivalent theorem
in quantum field theory. The extension of the Noether’s theorem to other fields of
physics results in the proposal of continuity equations. In particular, in dissipative
systems, the physical observables fluctuate (i.e., they are not preserved in time
and space), because the space homogeneity and the space isotropy are constantly
changed by external factors.
6.15 Hamilton’s Equations of Motion
The Lagrangian equations of motion are second-order differential equations, and

they are frequently difficult to solve. If we carry an analogy with the Newtonian
equations of motion, we have to solve equations like .f = m d 2 r/dt 2 in Lagrangian
theory; however, this expression is equivalently written in the form of two first-order
differential equations, .p = mdr/dt and .f = dp/dt. In this process, a single equation
of second-order has been transformed into two equations of first order, which may
be easier to solve in terms of .r and .p.
. fi = mi d 2 ri /dt 2 ⇒ pi = mi dri /dt ; f = dpi /dt
In this section, we present the equations of motion of classical mechanics according

to the Hamiltonian theory. In this approximation, the order of the differential
equations of motion is reduced with respect to that of Lagrangian mechanics. To
do this, the conjugate momentum .pi , instead of the velocity .q̇i , and the generalized
coordinate .qi constitute the basic variables in the description of the particle system.
The generalized momentum .pi is defined in terms of the Lagrangian as before,
Eq. 6.8.
. pi = ∂L/∂ q̇i (6.39)
The generalized momentum is consistent with the ordinary knowledge on the

Newtonian Emomentum. Suppose that the kinetic energy of a system of particles
is .T = m
k k k v 2 /2. From Newtonian mechanics, we know that, for example,
in the x direction, .p = m dx/dt = mv. On the other side, by considering the

Lagrangian as .L = T − V , where the interaction potential V has no dependence
on the velocities, we deduce .p = ∂L/∂ ẋ = mv. Therefore, the expression of
the generalized momentum in Lagrangian theory is consistent with our common
knowledge on the Newtonian momentum in Cartesian coordinates. By assuming
a conjugate momentum .pi for every .qi , the set .{pi } is defined. Next, we have to
establish the Hamiltonian equations that rule the motion of the particles in terms of
the conjugate variables .qi and .pi . Since we are dealing with a change of variables
from .q̇i to .pi , the Legendre transformation of the Lagrangian function, Eq. 6.30, is
employed.
E
. H({qi }, {pi }, t) + L({qi }, {q̇i }, t) = i pi q̇i
The transformation is symmetric in L and .H and also in the coordinates .qi and
pi . The Lagrangian and Hamiltonian on the one side, and .q̇i and .pi on the other,
.
classify in the same category according to the Legendre transformation. In order to

reach a differential equation, the following step consists on taking the differential of
the above expression:
E [ ∂H ∂H
]
∂H E
. dqi + dpi + dt = (pi d q̇i + q̇i dpi )
∂qi ∂pi ∂t
i i
E [ ∂L ∂L
]
∂L
− dqi + d q̇i − dt
∂qi ∂ q̇i ∂t
i
The terms .pi d q̇i and .(∂L/∂ q̇i )d q̇i of the right-hand side cancel each other. By using
the definition of the conjugate momentum, .pi = ∂L/∂ q̇i , plus the identity deduced
from the Lagrangian equation .∂L/∂qi = d(∂L/∂ q̇i )/dt − Qi = dpi /dt − Qi =
ṗi − Qi , and considering that the variables are independent, the coefficients from
6.16 Invariance Under Canonical Transformations 129
the left-hand side should match those of the right-hand side.
. q̇i = ∂∂pHi ; Qi − ṗi = ∂∂qHi ; − ∂L ∂H

∂t = ∂t ; i = 1, · · · , n
(6.40)
These are the Hamilton’s equations of motion. The .qi and .pi variables are obtained
by integration of these equations with time. The equations form a set of 2n first-
order differential equations, plus the last expression. If the Hamiltonian is time
independent, we require 2n initial conditions or n conditions at two different
instants of time to determine the solutions in unique form. The equations are in
principle easier to solve than the set formed by the Lagrangian equations, which are
second-order differential expressions. In this formalism, the role of the Lagrangian
is replaced by the Hamiltonian function, .H. The new independent variables are
.qi and .pi . They are shown in symmetric forms (except for a minus sign and a
generalized force, .Qi ) in the equations. The Hamilton’s formulation constitutes

a general approach to study mechanical systems of particles, and in this context,
Eqs. 6.40 are the canonical equations of motion since they are invariant and valid
for any set of conjugate variables .qi and .pi of conservative systems, .Qi = 0.11
The Hamiltonian equations reproduce the E Newtonian equations. If we take a system
of particles with Hamiltonian .H = (p
k kx
2 + p 2 + p 2 )/2m + V , then the
ky kz k
Hamiltonian equations corresponding to the ith particle in the x direction are:
∂H
. ẋi = ∂pix
= pix
mi ; ṗix = − ∂∂xHi = − ∂x
∂V
i
⇒ fi = mi d 2 ri /dt 2
Similar expressions are achieved for the other degrees of freedom. The point
to remark here refers to the Hamiltonian equations reproducing the Newtonian
equations. On the other hand, the procedure to solve the Hamiltonian equations
looks in principle simpler, but the building of the Lagrangian is required to create
the conjugate momenta from the generalized coordinates and insert the momenta
into the expression of the Hamiltonian. This process implies additional work to
that required in the Newtonian and Lagrangian approaches. Yet, all the mechanical
formulations converge to the same solution when the initial conditions are given.
6.16 Invariance Under Canonical Transformations
The generalized coordinates .qi and generalized momenta .pi are used to describe
the system in the Hamiltonian approach. Every pair of conjugate variables .qi and .pi
satisfy the Hamilton’s equations.
. q̇i = ∂H/∂pi ; ṗi = −∂H/∂qi
11 The canonical expressions are part of a set of rules that provide the basis of a theoretical frame-
work. The canonical expressions are simple yet sufficiently general to govern the mathematical
operations and procedures within the theoretical framework.
The choice of using the variables .qi and .pi is arbitrary, as we could have chosen
the variables .Qi and .Pi , different to .qi and .pi , with their own Hamiltonian, K, also
satisfying the Hamilton’s equations.
. Q̇i = ∂K/∂Pi ; Ṗi = −∂K/∂Qi
Under the supposition of equal number of degrees of freedom in the transformations

of the coordinates, there are different sets of generalized variables to describe
the same system. In the case that we require a description in terms of the
variables .(Qi , Pi ), instead of .(qi , pi ), the variables .(qi , pi ) and the Hamiltonian
.H({qi }, {pi }, t) have to be transformed. In general, the transformed Hamiltonian,
K, in terms of the variables .(Qi , Pi ), adopts a different functional form to that of

.H in terms of the variables .(qi , pi ). The variables .(Qi , Pi ) and the transformed
Hamiltonian K do not necessarily satisfy the Hamiltonian equations, since the

proper mechanical requirements on the transformation of coordinates have not been
imposed. The goal of this section is to propose the adequate transformations that
maintain the functional forms of the Hamiltonian equations of motion invariant.
Such types of transformations are known as canonical transformations. Before, it is
important to recall the stationary principle of the action.
f t2
. δ t1 L dt = 0
In place of the Lagrangian, expression 6.30 in terms of the Hamiltonian .H is used.

f t2 f t2 [E ]
. δ t1 L dt = δ t1 i pi q̇i − H({qi }, {pi }, t) dt = 0
If the link between the new variables and old variables is of the kind .Qi =
Qi ({qi }, {pi }, t) and .Pi = Pi ({qi }, {pi }, t), then the Hamilton’s principle should
be satisfied by the variables describing the phase space of the particle system
f [ ]
t2 E
δ
. pi q̇i − H({qi }, {pi }, t) dt = 0 ;
t1 i
f [ ]
t2 E
δ Pi Q̇i − K({Qi }, {Pi }, t) dt = 0
t1 i
In going from the representation .(qi , pi ) to .(Qi , Pi ), we require to build a bridge

through a generating function. We call such a function G, with mixed dependence on
the old and new variables, .G = G({qi }, {Qi }, t). The dependence on the generalized
coordinates .(qi , Qi ) may be equally changed to any other possibility, .(qi , Pi ),
.(pi , Qi ) or .(pi , Pi ), leading to the same results. Based on the discussion of Sect. 6.6,
where the non-uniqueness of the Lagrangian was attributed to the time derivative of
a function, we determine in that context the generating function G. The kernels of
the previous integrals are equal, except for the time derivative of a function, which,
6.16 Invariance Under Canonical Transformations 131
in this case, is the generating function.

E
. ipi q̇i − H({qi }, {pi }, t)
E
= i Pi Q̇i − K({Qi }, {Pi }, t) + dt
d
G({qi }, {Qi }, t)
E E ( ∂G )
= i Pi Q̇i − K({Qi }, {Pi }, t) + i ∂qi q̇i + ∂Q ∂G
i
Q̇ i +
∂G
∂t
By comparing the factors of the independent terms on the left with those on the
right, the conditions on the transformation are obtained.
pi = ∂G/∂qi ; Pi = −∂G/∂Qi
. (6.41)
H({qi }, {pi }, t) = K({Qi }, {Pi }, t) − ∂G/∂t
These conditions involve the generating function, the two Hamiltonians, the general-
ized coordinates of both phase spaces, and time. A transformation that satisfies such
requirements is recognized as a canonical transformation since it preserves the func-
tional form of the Hamiltonian equations. By assuming that the generating function
G is known, the new generalized variables are deduced. In order to illustrate it, we
consider the following points: the first identity (.pi = ∂G/∂qi ) gives n equations
for the .pi momenta in terms of .qi , .Qi , and t, namely .pi = pi ({qi }, {Qi }, t).
From these expressions, the generalized coordinates .Qi are obtained by inverting
the equations, .Qi = Qi ({qi }, {pi }, t). The momenta .{Pi } are obtained from the
second set of equations, .Pi = −∂G/∂Qi , resulting on .Pi = Pi ({qi }, {Qi }, t). When
one substitutes the .{Qi } by the .{qi } and .{pi } variables, the generalized momenta
.Pi are shown in terms of the .{qi , pi } variables, that is, .Pi = Pi ({qi }, {pi }, t).
The Hamiltonian K is written in terms of .{Qi } and .{Pi }, and the transformation
is accomplished.
. Qi = Qi ({qi }, {pi }, t) ; Pi = Pi ({qi }, {pi }, t) ; K = K({Qi }, {Pi }, t)
When the transformation between the variables .{qi , pi } and .{Qi , Pi } is known, but
the generating function is unknown, the generating function G may be obtained. To
do this, it is necessary to have the momenta .pi and .Pi in terms of the coordinates
.qi , .Qi , and time t, because these are the variables of the G function.
. pi = pi ({qi }, {Qi }, t) ; Pi = Pi ({qi }, {Qi }, t)
The final step demands integrating the differential equations 6.41 to determine
G. When it is required to work with other combinations of variables, such as
.(qi , Pi ), .(pi , Qi ) or .(pi , Pi ), instead of the combination .(qi , Qi ) discussed above,
the canonical transformations change accordingly. In those cases, it is convenient to

replace G by one of the generating functions .G' for the respective variables of the
phase space.
E
G' = G({qi }, {Pi }, t) − i Qi Pi ; {qi , Pi }
'
E
.G =
i qi pi + G({pi }, {Qi }, t) ; {pi , Qi } (6.42)
E
G' = i (qi pi − Qi Pi ) + G({pi }, {Pi }, t) ; {pi , Pi }
The equations of the generating functions can be deduced following similar steps
to those illustrated above (where the kernels of the action integrals were related).
The transformation of variables may demand the combination of several generating
functions. In special cases, by using the proper transformation, it may be also
possible to reduce the Hamiltonian to a cyclic form in the coordinates. In general,
the Hamiltonian adopts a different functional form after the transformations of
the conjugate variables are performed. Still, the Hamiltonian equations of motion
remain invariant in the transformed phase spaces.
6.17 Time Reversibility in Hamiltonian Theory
In the motion of particles, the generalized variables change in time. The change
of the variables may be interpreted as a transformation of variables due to a time
shift, .τ . More precisely, the transformation is from the variables .(q(t), p(t)) to the
variables .Q(t) = q(t + τ ) and .P (t) = p(t + τ ). According to Hamilton’s principle
and the canonical transformations .Q(t) ↔ q(t ' ) and .P (t) ↔ p(t ' ), we have the
equivalence:
f t2 [ E ] f t ' [E '
] '
. δ t1 i Pi Q̇i − K({Qi }, {Pi }, t) dt = δ t '2 i pi q̇i − H({qi }, {pi }, t ) dt
1
(6.43)
The equation remains without changes when the variables .(Q(t), P (t)) are replaced
by the variables .(q(t + τ ), p(t + τ )), and the time is .t ' = t + τ . The primed time,
'
.t , on the expression of the right is a dummy variable and is replaced by t.
f [ ]
t2 +τ E
δ
. pi q̇i − H({qi }, {pi }, t + τ ) d(t + τ )
t1 +τ i
f [ ]
t2 E
=δ pi q̇i − H({qi }, {pi }, t) dt (6.44)
t1 i
The expression indicates an invariance with respect to a time shift. Since the
time shift produces the same trajectory .(q(t), p(t)) to that with no time shift, the
motion of the particles may be understood as a canonical transformation. It is also
important to investigate the time reversibility of the solutions of the Hamiltonian
equations. It is assumed that the Hamiltonian is an even function of the momentum,
.H(q, p) = H(q, −p), an acceptable hypothesis in many particle systems. In order
to invert the dynamics with time, we require two steps: (i) to reverse the momenta
6.17 Time Reversibility in Hamiltonian Theory 133
of the particles (e.g., by stopping and pushing the particles backward with the same
momenta they had at the instant of stopping them), and (ii) to reverse the ticks of the
Newtonian clock (for instance, by producing a specular reflection of the clock) [6].
Thus, we consider a negative time, like .τ = −t. The respective equations of motion
are:
. dqi /dτ = −∂H/∂pi ; dpi /dτ = ∂H/∂qi ; τ = −t
The variables .qi and .pi are functions of the time and describe the trajectory of
the ith particle when the initial conditions are given. It is convenient to define a
negative momentum with respect to .pi (such as .πi = −pi ), because the momentum
is a basic variable whose definition involves the time. The Hamiltonian remains
unaltered with the new definition of the momentum.
. dqi /dτ = ∂H/∂πi ; −dπi /dτ = ∂H/∂qi ; πi = −pi
These expressions are equivalent to the Hamiltonian equations of motion 6.40,

with .Qi = 0. The conclusion is that the Hamiltonian equations of motion are
recovered in the time-reversed process and the variables .qi (τ ) and .πi (τ ) describe
the reversed trajectory of the ith particle.
Problem 6.7 Demonstrate that the Lagrangian equations of motion are time
reversible.
Solution The Lagrangian equations of motion are .d(∂L/∂ q̇i )/dt−∂L/∂qi =

0. We change signs in the generalized velocity and time.
( )
.
d
−dt
∂L
−∂ q̇i − ∂L
∂qi =0
After the cancellation of negative signs, the expression reproduces the original
Lagrangian equations of motion. Suppose that .qi and .qi' are the forward and
reverse trajectories described by particle i. When we are at a given point of
the particle trajectory, and a time t elapses, we conclude that .qi (−t) = qi' (t)
and .−q̇i (−t) = q̇i' (t). In this regard, the Lagrangian equations of motion are
time reversible.
6.18 Hamilton-Jacobi Theory
The goal of this section is to describe an alternative approach to the previous

theories. It is the Hamilton-Jacobi formulation of classical mechanics and cor-
responds to a branch of Hamiltonian theory. The Hamilton-Jacobi approach is
consistent with Newtonian mechanics, represents a tool to find integrals of motion
and is considered to be the “closest classical approach” to quantum mechanics.
The Hamilton-Jacobi formulation departs from the Hamiltonian. The purpose is to
employ a canonical transformation, G, to determine the solutions of the Hamiltonian
equations concurrently taking advantage of the system constants of motion. The
Hamilton-Jacobi formulation is an abstract scheme, and for such a reason, it shall
be illustrated with a specific example such as the harmonic oscillator in Sect. 6.19.
The G transformation is supposed to depend on the mixed variables .(qi , Pi , t). It
also relates to the Hamiltonians .H and K by:
. K(Qi , Pi , t) = H(qi , pi , t) + ∂G(qi , Pi , t)/∂t (6.45)
The variables .(qi , pi ) and .(Qi , Pi ) are obtained from both the equations on the
transformation of G (associated with the first identity of Eqs. 6.42, and presented in
the first row below) and the equations of motion on the Hamiltonian K (presented
in the second row below).
pi = ∂G
∂qi ; Qi = ∂G
∂Pi
. (6.46)
Ṗi = − ∂Q
∂K
i
; Q̇i = ∂K
∂Pi
The Hamiltonian of the harmonic oscillator is .H = p2 /2m + kq 2 /2, where the

spring constant k is related to the frequency .ω, with .k = mω2 . The Hamiltonian
contains the basic variables .(q, p) that we shall deal with in Sect. 6.19.
. H = p2 /(2m) + (mωq)2 /(2m)
We are free to choose the transformation G; therefore, it is taken in such a way

that the transformed Hamiltonian satisfies the trivial identity .K = 0. The derivatives
.Q̇i = ∂K/∂Pi and .Ṗi = −∂K/∂Qi involve K, and in this case, they are zero. The
change of .Qi and .Pi in time is null, .Q̇i = 0 and .Ṗi = 0, and they are constants of
motion.
. Pi = αi ; Qi = βi ; H(qi , pi , t) + ∂G(qi , Pi , t)/∂t = 0 ; i = 1, · · · , n

(6.47)
The quantities .αi and .βi are constants. There is a number of n expressions for .Pi
and an equal number for .Qi (in the case of m constraints, the number of degrees
of freedom is reduced to .2n − m). It is time to introduce the expression of the
action S in terms of the function G and the arbitrary constant c. To do this, we
6.18 Hamilton-Jacobi Theory 135
E E
recall from Eqs. (6.12) and (6.30) that . i pi q̇i − H({qi }, {pi }, t) = i Pi Q̇i −
K({Qi }, {Pi }, t) + dG/dt, .Q̇i = Ṗi = 0, and .L = dG/dt, which under time
integration leads to:
. S(qi , αi , t) = G(qi , Pi , t) + c = G(qi , αi , t) + c (6.48)
The action function in this approach is understood as the generating function of

a canonical transformation where the solutions of the equations associated with the
transformed Hamiltonian are constants of motion (in the transformation of variables,
we are implicitly assuming .det(∂ 2 S/∂qi ∂Pi ) /= 0). The S function depends on the
canonical coordinates .qi , time t, and constants of motion .αi , with .i = 1, · · · , n.
The equations of motion in terms of the action function S are deduced from
expressions 6.46 and 6.47.
∂S(qi ,αi ,t)

Qi = βi = ∂αi ; pi = ∂S(q∂q i ,αi ,t)
i
. ( )
∂
∂t S(q1 , · · · , qn ; α1 , · · · , αn ; t) + H q 1 , · · · , qn , ∂S
∂q1 , · · · , ∂S
∂qn ; t =0
(6.49)
The last expression is a first-order nonlinear equation of partial derivatives on the
generating function S. It is called the Hamilton-Jacobi equation . The mathematical
problem now resides on solving the single equation of partial derivatives to obtain
the trajectory of the particles. The way in which the Hamilton-Jacobi equation has
been proposed also becomes of help to find conserved quantities in the dynamics of
the particles. The identity .pi = ∂S/∂qi from Eq. 6.49 is the representation of the
canonical momentum in terms of S and is commonly written in the form:
. pi = ∇i S (6.50)
The Hamiltonian of the harmonic oscillator satisfies .H + ∂G/∂t = 0, Eq. 6.47,

and the Hamilton-Jacobi equation is:
[( )2 ]
.
1
2m
∂S
∂q + (mωq)2 + ∂S
∂t =0 (6.51)
The term .∂S/∂q replaces the momentum p to have the Hamilton-Jacobi expres-
sion written as a partial differential equation for the action function S in terms
of q. In more general terms, the Hamilton-Jacobi approach employs the main
variables .(Qi , Pi ) and the initial conditions .(αi , βi ), to obtain the solutions .(qi , pi )
of the mechanical problem by working out the proper canonical transformation.
Under this perspective, the time evolution of the system is considered a canonical
transformation from .(αi , βi ) = (qi (t0 ), pi (t0 )) to .(qi (t), pi (t)) at any instant of
time t.
The following goal is to introduce an abbreviated action. Firstly, we establish
a relation between the canonical transformation G and the action function S
E an alternative expression to Eq. 6.48. We differentiate G in time, .dG/dt =

as
i (∂G/∂qi )(dqi /dt) + ∂G/∂t. By using Eqs. 6.49, where .pi = ∂S/∂qi and
.∂S/∂t = −H, we reach:
E ∂G dqi E
.
dG
dt = i ∂qi dt + ∂G
∂t = i pi q̇i − H (6.52)
The expression on the right confirms the Lagrangian function of Eq. 6.30. In this
respect, .dG/dt is related to the action function S upon integration over time. The
action function is sufficiently versatile as it is the departure function of different
approaches in obtaining the equations of motion. Now, the abbreviated action, W ,
is introduced by .S = W − αt.
. W (q1 , · · · , qn ; α1 , · · · , αn ) = S + Et (6.53)
Under the supposition of a time-independent Hamiltonian,

f the equation
f is written
in terms of functions L and .H when the expressions .S = L dt and . H dt = Et
are recalled.
f
. W = (L + H) dt (6.54)
The abbreviated action is written in terms of the basic variables with Eq. 6.30.
E f
. W = i pi dqi (6.55)
For a freely moving particle in the .r direction, with momentum .p, the functions
S and W that satisfy the Hamilton-Jacobi equation are:
√
. W =r·p ; S(r, p, t) = r · p − Et ; p= 2mE (6.56)
Problem 6.8 The Hamiltonian principal function S is a power series over

time:
α2 k2 5
. S(t) = αq − 2m t+ k
2 qt 2 − αk 3
6m t − 40m t
The position variable is q, the time is t, and the parameters .α, k, and m are
constants. Find the Hamiltonian and the solutions .(q, p) from the Hamilton-
Jacobi equation that employs the function S.
Solution According to the Hamilton-Jacobi expression, we have .∂S/∂t +

H = 0. Thus, by differentiating S with respect to time, we obtain .H.
( )
k2 t 4
. −H = ∂S
∂t = − 2m
1
α 2 + αkt 2 + 4 + ktq
(continued)
6.19 Illustrating with the Harmonic Oscillator 137
The following step is to find the momentum p to write .H in terms of q and p.

To do this, we recall expression 6.50, .p = ∂S/∂q = α + kt 2 /2.
. p2 = α 2 + αkt 2 + k 2 t 4 /4
The Hamiltonian is now written in terms of the conjugate variables q and p

as .H = p2 /2m − ktq. It is possible to deduce the Hamiltonian equations of
motion.
. q̇ = ∂∂pHi = p/m ; ṗ = − ∂∂qHi = kt
The integration of the last expression gives .p = p0 + kt 2 /2. This result is

inserted in the first expression to integrate the position over time, giving .q =
q0 + p0 t/m + kt 3 /6m. More results are found in [7].
6.19 Illustrating with the Harmonic Oscillator
The abbreviated action W is of particular help to study the harmonic oscillator. In

order to obtain the coordinates q and p of the oscillator in terms of the time, we
require to determine the constants of motion Q and P and use W as the canonical
transformation to get q and p. Some of the constants of motion are found from the
initial conditions, when .t = 0. These are, for example, the oscillation amplitude, the
phase factor, and energy. The conjugate variables q and p are described in terms of
such constants of motion. We start from the Hamilton-Jacobi equation 6.51, taking
.S = W − Et into consideration.
[( ) ]
2
.
1
2m
∂W
∂q + (mωq) = α
2
The .α term is the constant energy. Among the many types of variables that are
used to formulate the equations of the harmonic oscillator, there is a particular set
that leads to the solution in a relatively simple manner. We take .α as the variable
P (it could equally be Q). The function W defines the canonical transformation in
going from the variables .(Q, P ) to the variables .(q, p). According to the second
identity of Eqs. 6.49 and 6.53, the canonical momentum is .p = ∂W/∂q, and this
term is present in the above equation of W . By solving for the term .∂W/∂q, we
obtain p in terms of q.
[ ]1/2
. p= ∂W
∂q = 2mα − (mωq)2 (6.57)
In order to solve the equation for W , we integrate over the variable q.

f[ ]1/2
. W = 2mα − (mωq)2 dq (6.58)
The following step is to find the variable Q from the expression of S by

considering .Q = β = ∂S/∂α, which corresponds to the first identity of Eqs. 6.49,
and .W (q, α) = S(q, α, t) + αt, Eq. 6.53.
∂(W −αt) f[ ]−1/2

. β= ∂S
∂α = ∂α =m 2mα − (mωq)2 dq − t (6.59)
f
The integral is done according to the result . dx/(a 2 − x 2 )1/2 = sin−1 (x/a).
/
. β= 1
ω sin−1 mw 2 q 2 /(2α) − t (6.60)
At this stage, we have the variables .(Q, P ) = (β, α). We now invert the above
identity for the variable q and use it in the expression of p, Eq. 6.57.
/ √
. q(t) = 2α/(mω2 ) sin(wt + ωβ) ; p(t) = 2mα cos(ωt + ωβ)
(6.61)
The solutions describe the oscillatory character of the motion. The terms involv-
ing .α are the sinusoidal amplitudes, and the terms involving .β are the phase factors.
In this treatment, the P and Q conjugate variables are the .α and .β constants of
motion, respectively. Thus, by dealing with the equation for W , which relates the
variables .P = α and q, it was possible to obtain the variables q and p with their
explicit time dependence.
6.20 Contact Between Quantum and Classical Mechanics
The Hamilton-Jacobi approach is particularly useful to establish “contact” with

the theory of quantum mechanics. In order to illustrate it, we start from quantum
mechanics because it is a more general theory. A particle in quantum mechanics is
represented by a wave, and the equation ruling the dynamics of the particle is the
time-dependent wave equation.
. i h̄ ∂ψ
∂t = Ĥψ
(6.62)
The function .Ĥ is the Hamiltonian operator , and .ψ is the wave function that
describes the particle. For a particle with a given momentum and subjected to
potential V , the wave equation is:
2 ∂ 2 ψ(q,t)
. i h̄ ∂ψ(q,t)
∂t = − 2m
h̄
∂q 2
+ V (q) ψ(q, t) (6.63)
6.20 Contact Between Quantum and Classical Mechanics 139
where the Hamiltonian operator was identified by .Ĥ = −h̄2 /(2m) ∂ 2 /∂q 2 + V (q).
Due to the time derivative and the spatial derivative, the wave function .ψ is expected
to be related to an exponential function. Hence, a wave function where the action S
plays the role of a phase factor is proposed.
f∞
. ψ = ψ0 exp(iS/h̄) ; −∞ |ψ(q, t)|
2 dq =1 (6.64)
The wave function is normalized and is called the action wave when it is
given in that form. The factor .h̄ is the Planck constant that divides S in .h̄ units,
making the exponent unitless. Such a factor is characteristic of quantum mechanics
(like the Boltzmann constant .kB to statistical mechanics or the speed of light c
to relativistic mechanics). However, .h̄ is very small in classical systems. Since
the particle propagates as a wave, we consider the similarity between the action
.S = p · r − Et of a freely moving particle, Eq. 6.56, and the phase .k · r − ωt of a
plane wave, where the wave vector is .k and the angular frequency is .ω = E/h̄. In
fact, when we divide .(p · r − Et) by .h̄, we obtain .(k · r − ωt), with the definitions
.k = p/h̄ and .ω = E/h̄.
The following purpose is to show that S is the action function of classical

mechanics. We insert .ψ given by Eq. 6.64 in Eq. 6.63, using the expression
.∂ψ/∂q = (i/h̄)ψ ∂S/∂q.
( )2
∂2S
. − 2m
i h̄
∂q 2
+ 1
2m
∂S
∂q + V (q) + ∂S
∂t =0 (6.65)
By assuming that the terms without .h̄ are much bigger than those containing .h̄,
we neglect the first term of the equation.
( )2
. 1
2m
∂S
∂q + V (q) + ∂S
∂t =0 ⇒ H(q, ∂S/∂q) + ∂S
∂t =0 (6.66)
f
Problem 6.9 The action function is proposed in the form .S = ± pdr + c,
where c is a constant. The de Broglie relation is .λ(r) = h/p(r). Show that
.|∂λ(r)/∂r| << 2π when .|h̄(∂ S/∂r )| << (∂S/∂r) .
2 2 2
Solution Based on the form of the above action function, and assuming
the inequality .|h̄(∂ 2 S/∂r 2 )| << (∂S/∂r)2 , we obtain (with .p = ∂S/∂r) an
expression in terms of the momentum p.
. |h̄ ∂p/∂r| << p 2
The inequality is written in the equivalent form .|∂(h̄p−1 )/∂r| << 1, where
.h̄ = h/(2π ). The de Broglie relation is .λ(r) = h/p(r); therefore,
(continued)
|∂λ(r)/∂r| << 2π . This inequality is one of the conditions for quantum

.
mechanics to transition into classical mechanics.
The last expression is the Hamilton-Jacobi equation 6.49. Thus, we have

evidence of a close contact between classical mechanics and quantum mechanics
under the conditions discussed previously.12 If we proceed in the reverse direction
to meet quantum mechanics from the classical formulation, then discrepancies are
found. By considering the Hamilton-Jacobi expression 6.66, and using the identity
.ψ = ψ0 exp(iS/h̄), we obtain:
2
( )2
. i h̄ ∂ψ
∂t = − 2m
h̄ 1
ψ
∂ψ
∂q +Vψ (6.67)
This expression differs from the wave equation 6.63. Historically, there was no
reason to create a new formulation in the years previous to the birth of quantum
mechanics because classical mechanics appropriately explained the experiments of
that epoch.
6.21 Poisson’s Brackets
In this section, we discuss the brackets approach of Poisson, understood as an

alternative representation of the equations of motion in classical mechanics. From
the Poisson brackets, the Liouville operator is introduced to perform the time
evolution of the physical observables, which are expressed in terms of the conjugate
variables and time.
The Poisson approach departs from an anti-commutative operator, denoted with
brackets for the arbitrary functions u and v.
. {u, v} = ∂u ∂v
∂q ∂p − ∂v ∂u
∂q ∂p (6.68)
We are interested in applying the Poisson brackets to the Hamiltonian, .H, and an
arbitrary function, F , both in terms of the generalized variables q and p.
12 The Hamilton-Jacobi equation is a particular case of the Hamiltonian formulation, which in turn
is linked to the Lagrangian formulation by a Legendre transformation. The Lagrange equations are
the result of the motion of particles satisfying the principle of stationary action. In this respect, the
wave equation of quantum mechanics is believed to implicitly contain the principle of stationary
action by reaching the Hamilton-Jacobi equation.
6.21 Poisson’s Brackets 141
[ ]
∂H ∂ H
. {F, H} = ∂p ∂q − ∂∂q ∂
∂p F = L̂F ; L̂F = {F, H} = −{H, F }
(6.69)
This expression defines the Liouville operator, .L̂, over the function F . The
imaginary symbol i may be employed under convention in the definition of the
Liouville operator, such as .i L̂F instead of .L̂F . We require to differentiate F with
respect to time to eventually introduce the equations of motion of the particles into
the structure of the Poisson brackets.
∂F ∂q ∂F ∂p
.
d
dt F (q, p, t) = ∂F
∂t + ∂q ∂t + ∂p ∂t
The time derivatives .q̇ and .ṗ are replaced, in that order, by the Hamiltonian
derivatives of .H with respect to the variables p and q. Thus, the change of F in
time is computed using Hamilton’s equations of motion.
∂F ∂ H ∂F ∂ H
.
d
dt F (q, p, t) = ∂F
∂t + ∂q ∂p − ∂p ∂q (6.70)
The use of the Poisson brackets simplifies the above expression.
.
d
dt F (q, p, t) = ∂F
∂t + {F, H} (6.71)
When .F = H, we have .dH/dt = ∂F /∂t. Expression 6.71 is consistent with the

Hamiltonian equations of motion for .F = q and .F = p.
H H
.
d
dt q = {q, H} = ∂∂p ; d
dt p = {p, H} = − ∂∂q
In this regard, Eq. 6.71 is a general equation of motion in the Poisson formulation.
Such an equation is the classical version of the quantum mechanical counterpart
shown below.
. d
dt F (t) = <ψ(t)| ∂ F̂∂t(t) |ψ(t)> + 1
i h̄ <ψ(t)| [F̂ , Ĥ] |ψ(t)>
The quantity .F̂ is the quantum mechanical operator associated with the physical
observable F , and .ψ is the wave function of the particle system. The above equation
gives the time evolution of the observable at the quantum mechanical level. In
principle, there is a correspondence between quantum and classical terms in the
classical limit, which is symbolized by .h̄ → 0.
.
1
i h̄ <ψ(t)| [F̂ , Ĥ] |ψ(t)> −→ {F, H}
h̄→0
The Newtonian equations of motion are recovered from the quantum mechanical
approach when the proper conditions are provided (these are discussed in a later
chapter).
Problem 6.10 Deduce the Newtonian equations of motion from Eq. 6.71.
Solution Consider the dynamic variables x and p. We write .F = x on the

one hand and .F = p on the other, in the Poisson brackets of Eq. 6.71.
∂x ∂ H ∂x ∂ H dp ∂p ∂ H ∂p ∂ H
.
dx
dt = ∂x ∂p − ∂p ∂q ; dt = ∂x ∂p − ∂p ∂q
each other, .∂x/∂p = 0, .∂p/∂x = 0. By

Variables x, p are independent of E
assuming the Hamiltonian as .H = i pi2 /2mi +V ({qi }), the above equations
reduce to:
dxi pi dpi
.
dt = mi ; dt = − ∂q
∂V
i
Thereby, the Newtonian equations of motion are reproduced in the Poisson

formulation.
In the classical situation, the time evolution of F is obtained from Eq. 6.71, and
after taking the Liouville operator into consideration.
[∂ ] [ ]
.
d
dt F (q, p, t) = ∂t − {H, ·} F = ∂t∂ + L̂ F (6.72)
The total time derivative of a constant of motion, say F , satisfies the expression
dF /dt = 0 and, consequently, .∂F /∂t = {H, F }, which provides an alternative
.
way to find constants of motion. If the constant of motion shows no explicit time
dependence, then .∂F /∂t = 0, and F commutes with the Hamiltonian, .{H, F } = 0.
At the quantum mechanical level, the quantities conserved over time also commute
with the Hamiltonian operator, thus maintaining consistency between the quantum
and classical theories.
There are additional properties satisfied by the Poisson brackets. According to
the definition based on the partial derivatives, we obtain for .u = qi and .v = pi in
Eq. 6.68 the binary operation rules.
{qi , qj } = 0 ; {pi , pj } = 0 ; {qi , pj } = δij

. (6.73)
{qi , u} = ∂u
∂pi ; {u, pi } = ∂u
∂qi
Other features of the Poisson brackets are the antisymmetric, bilinear, non-
commutative product, and permutation properties, which are illustrated with the
arbitrary functions u, v, g, and the constants .c1 and .c2 .
6.22 Classical Time Propagator 143
{u, u} = 0 ; {u, v} = −{v, u}

. {c1 u + c2 v, g} = c1 {u, g} + c2 {v, g} ; {uv, g} = u{v, g} + v{u, g}
{u, {v, g}} + {v, {g, u}} + {g, {u, v}} = 0
(6.74)
These properties define a Lie algebra for the Poisson brackets. The Poisson
brackets are invariant under canonical transformations and also include properties
of the time derivatives.
∂t {u, v} = { ∂u
∂t , v} + {u, ∂t }
.
∂ ∂v (6.75)
6.22 Classical Time Propagator
One of the advantages of the Poisson scheme over the other approaches is the
introduction of a classical time propagator to advance a system of particles forward
in time. The time evolution is performed on the individual positions and velocities
of the particle system. In contraposition, the time propagator of quantum mechanics
evolves the wave function with time. The outcome of evolving a particle system in
time is the detailed knowledge of its history. The goal is to propose an arbitrary
function, like .F ({qi }, {pi }, t), with dependence on the generalized coordinates and
time, and find its evolution with time. Some examples of a function like this are
the distribution function in phase space, the mechanical pressure, system volume,
temperature, and, in general, any quantity given in terms of the basic variables of
position and momentum.
In order to find the classical time propagator, it is supposed that the equation of
motion:
. ∂F /∂t = 0 ⇒ dF /dt = {F, H}
is satisfied, where the Hamiltonian is time independent. The formal solution of

this equation is obtained in iterative form (in the same way as it is performed in
the quantum situation for the Dyson time propagator). The result contains nested
Poisson brackets and is written as a series expansion.
t2 t3
. F (t) = F (t0 ) + {F, H}t=t0 t + {{F, H}, H}t=t0 2! + {{{F, H}, H}, H}t=t0 3!+ ···
(6.76)
Problem 6.11 Obtain the series expansion 6.76 of function F by assuming

that .{F, H} is time indifferent.
(continued)
Solution We depart from the equation .dF /dt = {F, H}. It is integrated over
time.
ft ft
.
d
t0 dt F dt = t0 {F, H} dt ⇒ F (t) = F (t0 ) + {F (t0 ), H} At
The identity on the left is the expression that we take as a reference point to
work out the iteration process. This process starts by inserting .F (t), given on
the right, into the identity on the left.
ft
F (t) = F (t0 ) + t0 dt {F (t0 ) + {F, H} At, H}
.
= F (t0 ) + {F (t0 ), H} At + {{F (t0 ), H}, H} (At)2 /2
By inserting this identity into the expression considered as our reference point,
we have:
ft 2
F (t) = F (t0 ) +
.
t0 dt {F (t0 ) + {F (t0 ), H} At + {{F, H}, H} (At)
2 , H}
2
= F (t0 ) + {F (t0 ), H} At + {{F (t0 ), H}, H} (At)
2
(At)3
+ {{{F (t0 ), H}, H}, H}
3!
where the properties 6.74 of the Poisson brackets were used. When we
continue with the iterations, the series expansion 6.76 is obtained. The same
result is achieved by continuous differentiation of .dF /dt = {F, H}, and using
the fact that the nth derivative is the derivative of the .(n − 1)th derivative,
namely, .d n /dt = d(d n−1 /dt)/dt, which implies an iteration process in the
Poisson method. For a third derivative, we have:
{ }
dF d 2F d dF
. = {F, H} ; 2
= {F, H} = , H = {{F, H}, H}
dt dt dt dt
{ } { }
d 3F d dY dF
= {{F, H}, H} = ,H = { , H}, H = {{{F, H}, H}, H};
dt 3 dt dt dt
Y = {F, H}
Higher derivatives are similarly computed. A Taylor expansion of F is:
t 2 d2F
. F (t) = F (t0 ) + t dF
dt + 2! dt 2 + ···
t=t0 t=t0
(continued)
6.22 Classical Time Propagator 145
When the derivatives are replaced by their respective Poisson brackets, the
result 6.76 is reached.
The series expansion based on the Poisson brackets plays the role of a generator
in the approach of infinitesimal canonical transformations, leading to the time
evolution of the dynamical variables [8]. It is convenient to write Eq. 6.76 in a
compact form by taking the function F as the common factor of the Poisson brackets
from the left side.
[ ]
t2 t3
. F (t) = F (t0 ) 1 + { , H} t + {{ , H}, H} 2 + {{{ , H}, H}, H} 3! + · · ·
t=t0
(6.77)
The terms in square brackets are symbolically represented altogether by an expo-
nential operator, .exp(H t), acting on the F function from the right-hand side.
. F (t) = F (t0 ) exp(H t) (6.78)

t=t0
The time propagator, .Û , is now defined to advance the function F from .t0 to t.
. F (t) = F (t0 ) Û (t0 , t) ; Û (t0 , t) = exp(H t) (6.79)
This expression is the analog of a quantum mechanical propagator after carrying

a Wick rotation, where the Newtonian time is replaced by the imaginary time .−it/h̄.
We now illustrate the use of the Poisson brackets as a time propagator. Consider the
Hamiltonian of a particle moving in the z dimension and subjected to gravity effects,
.V = −mgz.
. H = p2 /2m − mgz
We require to calculate the time evolution of the z coordinate. By making .F =

z, and using Eqs. 6.73 to compute the first Poisson brackets, .{qi , F } = ∂F /∂pi ,
.{F, pi } = ∂F /∂qi , we implement the time propagator.
. {z, H} = p/m ; {{z, H}, H} = {p, H}/m = g
The following Poisson brackets in the series expansion of the exponential are
null because they commute with the constant that results from the application of
the previous brackets. Such a procedure represents a way to generate constants
of motion using the Poisson approach. We collect the non-zero terms in a single
expression according to Eq. 6.77.
z(t) = z(t0 ) + {z(t0 ), H} t + {{z(t0 ), H}, H} t 2 /2

.
z(t) = z(t0 ) + p(t0 ) t/m + gt 2 /2
The equation describes a particle falling in a gravitational field with constant

acceleration g.
In this chapter, we have discussed the various formulations of analytical mechan-
ics, each of which leads to different equations of motion. However, it is important to
note that these different formulations and their corresponding equations of motion
are equivalent, resulting in the same time evolution of the particle system and
providing a unique solution. While some formulations may be more suitable for
solving certain types of problems, it is often necessary to use numerical methods to
solve the equations of motion when no analytical solution is available. This will be
the focus of a future chapter.
References
1. H. Goldstein, Classical Mechanics, 2nd edn. (Addison-Wesley, London, 1980)
2. V.I. Arnold, Mathematical Methods of Classical Mechanics, 2nd edn. (Springer-Verlag, New
York, 1989), translated from the Russian by K. Vogtmann, A. Weinstein
3. M. Levi, Classical Mechanics with Calculus of Variations and Optimal Control (AMS,
Providence, Rhode Island, 2014)
4. W.F. Trench, Introduction to Real Analysis, books and monographs, book 7 (2013). http://
digitalcommons.trinity.edu/mono/7
5. R.K.P. Zia, E.F. Redish, S.R. McKay, Making sense of the Legendre transform. Am. J. Phys. 77,
614–622 (2009)
6. J.L. Lebowitz, Boltzmann’s entropy and time’s arrow. Phys. Today 49, 32–38 (1993), p. 37
7. G.F. Torres del Castillo, Generation of solutions of the Hamilton-Jacobi equation. Rev. Mex.
Física 60, 75–79 (2014)
8. H. Goldstein, Classical Mechanics, 2nd edn. (Addison-Wesley, London, 1980), Chapt. 9, Sect.
5, pp. 405–416
Part II
Basics of Quantum Mechanics
Wave-Particle Duality of Matter
7
Quantum mechanics is an essential part of our science to understand nature. It

emerged in the turn of the 20th century and continues growing up to explain
phenomena at the atomic and subatomic levels. The purpose of this chapter is to
briefly expose the experiments from which the novel ideas emerged and gave birth
to quantum mechanics. The first section describes the works that demonstrated the
wave character of both the light and the microscopic particles as in the Young’s
experiment, and contrariwise, the experiments that unveiled the particle character of
the electromagnetic waves as in the photo-electron and the Compton experiments,
thus establishing the dual nature of matter. In the second section, the symmetry
between the particles and the waves is confirmed with the Davisson-Germer’s
experiment and the de Broglie hypothesis. It is rationalized that the particles may
be described as waves and the waves may be described as particles. All these
experiments point out the need to reformulate the conceptual world to have a wave
theory of matter.
7.1 Young’s Experiment
In the first centuries of science, the light was conceived as a particle. However, in the
Young’s experiment, an interference pattern was produced, showing the wave nature
of light. The experiment consisted on emulating the undulations and interference
observed in water waves and sound pulses. Two coherent light waves were produced
from a single source when the original light wave was diffracted by two slits.
The pair of waves interfered in constructive and destructive forms, creating white
and black fringes on a screen as a result of traveling different optical paths and
interfering with relative phases, Fig. 7.1.
The diffraction pattern is not the smooth curve that one expects from the
dispersion of small particles and deduced from the addition .I1 + I2 of the individual
light intensities, .Ii . In fact, the pattern is understood by assuming a probability
density obtained from a wave, .ψ, describing the particles. At the instant of time
150 7 Wave-Particle Duality of Matter
Fig. 7.1 The double-slit experiment demonstrates the wave nature of light, as well as that of the
electron, proton, neutron, and sub-atomic particles. Inset A: a single slit makes the beam parallel,
and a couple of slits located further ahead produce two sources of coherent beams. Inset B: the
beams are described by spherical wavefronts and treated as plane waves at far distances. Inset C:
in essence, a wave is made to interfere with a copy of it to produce a diffraction pattern consisting
of dark and bright fringes onto a screen. A dark band is due to waves interfering destructively, and
a bright band is due to waves interfering constructively
t in the position .r of the screen, the total wave function is the superposition of
the individual wave functions, .ψ(r, t) = ψ1 (r, t) + ψ2 (r, t), where the position
.r is a continuous variable and the subindexes of the wave functions distinguish
the beams coming from the two slits of the experiment. The probability density
is .I (r, t) = |ψ(r, t)|2 [1]. The Young’s experiment is considered a proof of the
wave behavior of light and in general of other microscopic particles capable of
producing a diffraction pattern in that type of experiment. For instance, it has been
possible to obtain the quantum interference of large objects with masses exceeding
4
.10 amu in the Young’s experiment [2]. In this connection, the Young’s experiment
is also important to differentiate the behavior of particles acting in the quantum

domain from the behavior of particles acting in the classical domain. The Young’s
experiment demonstrates the need to reformulate our conceptual world at the
microscopic level since classical mechanics fails to provide the correct description
on the behavior of particles at the atomic and sub-atomic levels.
7.2 Interference of Waves
The double-slit experiment is an experiment that provides information on the wave

character of the particles. The beam coherence is required to induce the stationary
states of the waves. In principle, the coherence of waves is estimated by evaluating
the correlation of functions involving the phase, frequency, amplitude, and intensity.
7.3 Photo-Electron Experiment 151
The key aspect of the Young’s experiment is the formation of a diffraction pattern
as a consequence of the interference of waves.
The interference represents the addition of waves to create a new wave with
an amplitude that depends on the relative phase of the interfering waves. The
interference depends on the coherence of the waves, but the coherence is not directly
measured in the lab due to the fast oscillations of the electromagnetic waves;
nevertheless, it can be indirectly estimated using the correlation between waves.
There are spatial and temporal correlations that establish the coherence of waves
at different spatial points and instants of time. By making a wave to interfere with
a copy of itself using a beam splitter, it is possible to induce a self-interaction at
different spatial positions and time instants.
A monochromatic wave is characterized by owing a single frequency, like the
light emitted by a laser with a high degree of correlation. A wave composed
by waves with different frequencies (and possibly with different amplitudes and
phases) as, for example, the white light emitted by an ordinary bulb, has a poor
degree of correlation. The interference of waves consisting of different frequencies,
but with a fixed phase difference, produces localized pulses. If the waves have dif-
ferent frequencies and no fixed phase difference, then their superposition produces
a continuous beam of white light. It is referred to as white noise because it is formed
by incoherent and poorly correlated waves.
The precise trajectories of the waves before they form the interference pattern
are unknown. The interference pattern may be sampled by changing the distance of
the screen from the slits.
Finally, other similar phenomena, like the interference of laser light, supercon-
ductivity, and superfluidity, represent cases of coherence. They are observed at
the macroscopic level. In this connection, the forms in which the waves behave,
interfere, and build a diffraction pattern are important for the formulation of
quantum mechanics, where probability fields, probability densities, and more have
to be considered.
7.3 Photo-Electron Experiment
The metals are characterized by atoms containing a large number of electrons. Many
of the electrons orbit the nucleus relatively far away and are weakly bonded. Such
electrons can produce an electric current when an electric field is applied to the
metal. However, under the action of an intense external force, the electrons may
abandon the metal. When light impacts a metal, some electrons are released from
the metal surface since the light transports energy, which is transformed onto kinetic
energy for the electrons. The released electrons receive the name of photo-electrons.
In the original photo-electron experiment, the light strikes a metal of sodium atoms
and detaches electrons, Fig. 7.2.
More energy is spent on releasing electrons from the internal material layers
than from the surface. The released electrons show different values of the kinetic
energy .Ekin , but the surface electrons achieve the maximum kinetic energy. The
Fig. 7.2 In the photo-electron experiment, a light striking a metal produces a flux of electrons after
being detached. The electron detachment depends on the characteristics of the impacting light and
the material. For instance, electrons are released from an alkali metal surface when visible light
strikes the surface. X rays, ultraviolet light, and the infrared beams can detach electrons from other
types of surfaces as well
energy distribution of the detached electrons is independent of the light intensity;

nevertheless, it depends on the light frequency because a stream of electrons is
produced when the light frequency exceeds a threshold value, .ν0 , particular of the
surface. Moreover, the maximum kinetic energy of the ejected electrons is observed
to be proportional to the light frequency .ν, where the constant of proportionality is
the Planck’s constant, h, with units of angular momentum.
.
max = h(ν − ν )
Ekin 0 ; h = 6.626, 07 × 10−34 J s (7.1)
The term .hν0 is the minimum energy required by an electron to escape from the
surface. It is identified as the photo-electron work function. If we direct ultraviolet
light of 3500 Å wavelength onto a sodium surface, .hν0 = 2.2 eV , then the maximum
max = 1.4 eV . Interesting phenomena is observed in
kinetic energy registered is .Ekin
the photo-electron experiment:
• The number of released photo-electrons increases with the number of photons

(or light intensity) for a specific light frequency.
• The maximum kinetic energies of the electrons are the same for high-intensity
light and low-intensity light of the same frequency.
• A high-intensity light produces less energetic electrons than a low-intensity light
of higher frequency (the high-intensity red light produces the ejection of a large
number of electrons with relatively small kinetic energies; in contrast, a low-
intensity violet light produces the ejection of a smaller number of electrons with
comparatively higher kinetic energies).
• No absorption of a fraction of a photon occurs, and all the photon energy is
entirely transferred to the electron.
7.3 Photo-Electron Experiment 153
• The ejection of electrons occurs with no time delay (i.e., there is not a gradual
energy accumulation).
The theory of classical electromagnetism is incapable to explain the above observa-

tions as it assumes several points:
• A wave is a uniform and continuous distribution of energy with spatial extension.

The light spreads uniformly over the surface and is difficult to be absorbed by a
spatially localized (tiny) electron.
• For any wave frequency, by increasing the light intensity (the intensity is related
to the amplitude E of the electromagnetic wave), the electric force eE is believed
to induce large amplitude vibrations of the electron, eventually producing its
release from the material.
• The electron emission is presumed to occur after gathering sufficient energy from
the illuminating beam. The energy gathering is independent of the light frequency
and the threshold frequency of the material.
• It is speculated that a time lag should be registered due to the time it takes an
electron to accumulate energy from the striking electromagnetic waves before its
ejection. The time lag is supposed to depend on the wave intensity, since low-
intensity light provides less energy than a high-intensity light, in a given time
interval.
According to Einstein, the phenomena observed in the photo-electron experiment

are explained by considering light as a localized-energy entity rather than a
continuous distribution of energy (like a wave) in space. The entity transfers energy
in the form of discrete energy bundles to the surface. The name photon or quanta is
given to light that is considered in discrete energy packets and acts as a particle. It is
also understood as a moving electromagnetic ripple. The photo-electron experiment
indicates that frequency is the main variable in the electron detachment. The
emission occurs instantaneously, by overcoming a threshold frequency which is
specific to the material.
The photo-electric effect is not the only technique to eject electrons from a
metal, because we also have the thermionic emission. In this process the metal
is subjected to intense heat. When the thermal vibrations exceed the thermionic
work function, known to be close in value to the work function of the photo-
electric effect, the electrons are ejected (the functioning of incandescent bulbs is
based on the thermionic emission process). In this respect, the photo-electric and
thermionic emission effects represent two sources for ejecting electrons. Some of
the important applications of the photo-electric effect are the photomultiplier tubes
and photometers, which work as “electric eyes” to detect ionizing radiation and
measure the intensity of light emitted by materials.
Concluding, the usual behavior of light in free space may be that of a wave;
however, the photo-electron experiment suggests that a beam of light behaves as a
stream of particles, which is distinguished by a frequency (an attribute originally
thought to be exclusive of waves). Clearly, we require a new formulation capable
of explaining the interaction of light with matter and, in particular, the energy
transfer between light and the microscopic particles, still, without abandoning the
classical electromagnetic theory because it provides the appropriate description of
light propagation, as well as the interference and diffraction of electromagnetic
waves.
7.4 Compton’s Experiment
Another work that fundamentally changed our understanding of nature was Comp-
ton’s experiment. In such an experiment, a beam of X rays is created in a vacuum
tube. The electromagnetic radiation impacts a target made of carbon atoms and is
scattered by the electrons of the carbon atoms in every direction. The electrons have
mass .m0 at rest. The scattered wavelength .λ' is larger than the wavelength .λ of the
incident beam and depends on the scattering angle .φ, Fig. 7.3.
The shift in the wavelength with the scattering angle is explained by employing
the energy and momentum conservation laws in the collision of the incident light
with the electron, where the light is assumed as a particle. In this context, the shifted
wavelength is given by:
. λ' = λ + h/(m0 c) (1 − cos φ) (7.2)
Compton’s experiment additionally confirms the theory of special relativity. In

contrast to Compton’s experiment, there are experiments that reveal the wave nature
of microscopic particles. They are discussed below.
Fig. 7.3 In the Compton’s experiment, an electromagnetic wave of high energy, such as an X
ray with wavelength .λ, impacts an electron of an atom. The electron recoils with angle .θ, and the
wave is scattered out of the original direction with angle .φ and wavelength .λ' . The interaction is
described as a collision between a pair of particles, that is, between a photon and an electron. The
experiment illustrates the particle character of the electromagnetic wave in this case
7.6 de Broglie’s Hypothesis 155
7.5 Davisson-Germer’s Experiment
Davisson-Germer’s experiment starts with a thermionic emission, where a flux of

electrons is emitted from a hot filament and is used as an electron gun. The electrons
are accelerated by applying a voltage gradient of a battery and scattered by the
atoms of the crystal planes of a nickel-chloride metal. Since the crystal atoms
simulate a kind of reflection and regular grating, the electrons are expected to
produce a uniform distribution of particles; however, they are observed to produce
a distribution of alternating intensity peaks at specific scattering angles, namely,
a diffraction pattern. The pattern is explained by Bragg’s law, which was in
principle believed to only apply to the diffraction of waves. The Bragg condition
for constructive interference is:
. 2d sin θ = nλ ; n = 1, 2, · · · (7.3)
The distance between contiguous lattice planes is d, the variable .θ is the angle
between the incident beam and the crystal planes of the metal, and .λ is the associated
wavelength of the particles. The angle between the incident beam and the reflected
beam is the scattering angle, denoted by .φ, which is not present in Eq. 7.3. The factor
n gives the different intensity peaks of the diffraction pattern. By measuring the
relevant parameters in Davisson-Germer’s experiment (.d = 0.91 Å, .θ = 65◦ ), the
maximum intensity (.n = 1, with direction .φ = 50◦ ) is obtained at an accelerating
voltage of .V = 54 eV . The resulting wavelength is .λ = 1.649 Å. There are
complications in applying Bragg’s law to the electrons diffracted by the inner layers
of the crystal as they interfere in complex forms. Similar experiments to Davisson-
Germer’s experiment have been performed using other types of particles, such
as hydrogen and helium, thus determining the wave character of those particles.
Summing up, Davisson-Germer’s experiment gives testimony of the wave properties
of matter particles and also confirms the de Broglie hypothesis, discussed below.
7.6 de Broglie’s Hypothesis
The fact that matter particles also act as waves permits to assign wave properties to
the microscopic particles. Louis de Broglie hypothesized a wavelength for matter
waves by considering both: (i) the relativistic energy .Ep = (p2 c2 + m20 c4 )1/2 of a
particle, where the particle’s rest mass is .m0 , the momentum is p, and the light speed
is c, and (ii) the energy .Ew = hν of a wave, where .ν is the wave frequency and h
is the Planck’s constant. For small kinetic energies of the particle with respect to
the rest mass energy, .m0 c2 , the relativistic effects are neglected after doing .m0 = 0.
By recalling that .λν = c, we have two representations of the wave-particle energy,
.Ep = pc and .Ew = hc/λ. The equivalence of the energies, .Ep = Ew , gives the
main result .λ = h/p. The particle has momentum .p = mv, and the wavelength of
matter waves is inversely proportional to the momentum.
Table 7.1 de Broglie’s Particle Mass (Kg) Speed .λ (Å)

wavelength of different
Electron .9.11× 10−31 .10
6 m/s 7.3
bodies
Proton .1.6726 × 10
−27 5
.10 m/s 0.04
Tennis ball 0.056 .260 Km/ hr .1.6 × 10
−24
. λ = h/(mv) (7.4)
For macroscopic bodies, the mass is large, and the wavelength is sufficiently small;
for sub-atomic particles, the mass is small, and the wavelength is large. Table 7.1
shows the wavelength of different particles. The particles with large wavelengths
have wave character. The electron is the particle with largest wave character in
the table, while ordinary bodies, such as a tennis ball, have essentially null wave
character. Thus, Davisson-Germer’s experiment is not the only one to exhibit the
wave character of the microscopic particles.
An alternative form of writing the matter wavelength is in terms of the kinetic
energy of the particle, .Ekin = p2 /(2m). By solving for the momentum p in this
expression, the de Broglie relation is shown in terms of the kinetic energy.
. λ = h (2m Ekin )1/2 (7.5)
For a macroscopic particle with values .m = 1 Kg, and .v = 1 m/s, the de

Broglie wavelength is .λ = 6.63 × 10−24 Å. For a microscopic electron, which
is accelerated with a voltage .V = 54 eV in Davisson-Germer’s experiment, the
de Broglie wavelength is .λ = 1.669 Å, and it compares well with .λ = 1.649 Å
obtained from the Bragg’s law of the previous section. These results have two
important consequences: (i) Davisson-Germer’s experiment confirms the de Broglie
hypothesis, and (ii) bodies with small de Broglie wavelengths like a dust particle, a
tennis ball, and other macroscopic objects are properly described by the theory of
classical mechanics. However, bodies with large de Broglie wavelengths, like the
helium atom, hydrogen atom, electron, proton, and neutron, lead to interference and
diffraction patterns, and in this respect, they are properly described by a wave theory
of matter.
The fact that microscopic particles have a wave character has led to technological
applications in unexpected areas, like optics. For instance, by conceiving the
electrons as waves, it was possible the invention of the electron microscope.
Summing up, the experiments firmly establish the dual aspect of matter: the particles
may be characterized using wave variables, and the waves may be characterized
using particle variables. In this connection, the de Broglie’s hypothesis represents a
symmetry of nature since a wave and a particle are rationalized as equivalent entities.
References 157
7.7 Bohr’s Complementary Principle
The wave-particle duality is an intrinsic property of matter. It is mainly observed in

the microscopic particles and light beams, to which we refer in generic form as the
entities. The wave and particle aspects have equal validity on an entity; nevertheless,
such a type of symmetry is destroyed when there is an interaction with a measuring
device because the device registers either the wave character or particle character of
the entity. For instance, if we determine the slit where the entity travels through
in the Young’s experiment, then the entity is revealed with a particle character.
In contrast, when no measurement on any of the slits is performed, the entity is
permitted to interfere, and its wave aspect is disclosed on the screen. In general,
when one of the two behaviors is displayed, the other complementary behavior
is inhibited. In this respect, Neils Bohr proposed the complementary principle to
indicate that it is possible to determine either the wave character or particle character
of an entity in a given experiment, but it is impossible to observe both behaviors of
the same entity simultaneously. The wave-particle duality of an entity is implicitly
considered in the uncertainty principle, yet the uncertainty principle should not be
confused with the complementary principle.
References
1. R.P. Feynman, A.R. Hibbs, Quantum Mechanics and Path Integrals (McGraw-Hill, New York,
2005), Emended by Daniel F. Styer, Sect. 1.1, pp. 2–9
2. S. Eibenberger, S. Gerlich, M. Arndt, M. Mayor, J. Tuxen, Matter-wave interference of particles
selected from a molecular library with masses exceeding 10000 amu. Phys. Chem. Chem. Phys.
15, 14696–14700 (2013). M. Arndt, O. Nairz, J. Vos-Andreae, C. Keller, G. van der Zouw, A.
Zeilinger, Wave-particle duality of C60 molecules. Nature 401, 680–682 (1999)
Quantization of the Energy
8
Quantum mechanics describes phenomena which in principle are beyond our

common sense. It represents a general subject from which the theory of classical
mechanics is derived under special circumstances. The general aspects of quantum
mechanics are originated from a scientific frame based on a rigorous mathematical
formulation, which allows to realize not only the wave-particle nature of matter
but also the energy quantization in the microscopic world. Energy quantization was
revealed in a few important experiments in the beginning of the 20th century. They
now represent landmarks in the history of science, as they provide the foundations
of quantum mechanics. We introduce in the following sections additional concepts
and principles to those already discussed previously. It is with quantum mechanics
that we are able to reformulate the old ideas, and establish a new philosophy to
understand the world.
8.1 Planck’s Energy Equation
The maximum kinetic energy .Ekin max of the electrons emitted in the photo-electron
experiment shows a linear relation with the frequency .ν of the striking radiation,
Eq. 7.1. The linear relation is independent of the beam intensity, and the slope of the
max
line in this relation is the Planck constant, h. In this regard, the connection of .Ekin
to .ν is ruled by the general Planck’s energy equation.
. E = hν (8.1)
After overcoming a threshold energy, the above expression agrees with the energies
reported in the photo-electron experiment. It also represents the basic hypothesis
to explain the distribution of frequencies in the blackbody-radiation experiment.
160 8 Quantization of the Energy
Planck’s expression may be written in terms of the angular frequency, .ω, using the
expression of .h̄.
. E = h̄ω ; h̄ = h/(2π ) ; ω = 2π ν (8.2)
8.2 Blackbody Radiation Experiment
When a piece of matter is heated, it reaches a temperature above absolute zero, and
there is a transformation of the body’s thermal energy onto electromagnetic energy,
which is dispersed as electromagnetic radiation. This radiation is known as thermal
radiation. However, a body equally absorbs thermal radiation. When the body or
system absorbs and eventually emits all the incident radiation with no dependence
on the type of the incident radiation, then the body is known as a blackbody. The
blackbody is a perfect body made of non-reflective material that absorbs and radiates
energy in equal amounts and in all directions in the state of thermal equilibrium,
Fig. 8.1. The emission factor of a blackbody is 1. The blackbody becomes a gray
body when the emission factor is lower than one. The body is a white body when it
reflects all the incident radiation. In contrast, a perfect transparent body transmits
all the incident radiation.
When an ordinary body is burned, the energy emitted per unit area and per unit
time determines the body’s radiation intensity. In the special case of a blackbody
at temperature T , the radiation shows a characteristic distribution of continuous
frequencies. Moreover, for a given frequency of emitted radiation, the radiation
intensity exclusively depends on the body’s temperature, regardless of the body’s
composition and shape. The typical colors of the emitted radiation start from low-
frequency invisible radiation (given at room temperature), following red visible light
(colloquially referred as heat), passing by bright yellow, dazzling blue-white color,
Fig. 8.1 A blackbody is a perfect absorber and radiator of electromagnetic radiation. The figure
shows that a tiny hole in the surface of a box can be used to characterize a blackbody. The irradiance
of the blackbody illustrates the quantization of the energy, confirms the dual character of light, and
satisfies the Planck formula from which the Rayleigh-Jeans law, Stefan-Boltzmann expression, and
Wien’s displacement law are obtained
8.3 Rayleigh-Jeans Law 161
high-frequency blue visible light, reaching ultraviolet light, and X rays. The order of
the colors corresponds to increasing temperatures and leads to a chromatic diagram
in terms of the temperature. A model frequently used to emulate the radiation
features of a blackbody is an idealized hole located in one of the walls of an insulated
cavity with reflective walls in the interior, Fig. 8.1. The hole is sufficiently small with
respect to the cavity size but larger than the wavelength of the incident radiation
to avoid reflections. A small hole also ensures negligible perturbations once the
thermal equilibrium is achieved inside the cavity. It is through the hole that radiation
comes in and goes out of the cavity, simulating the radiation absorption-emission
process of a blackbody.
8.3 Rayleigh-Jeans Law
The energy in the cavity takes the form of electromagnetic waves.1 The waves
are unpolarized and take specific forms since the electric and magnetic fields are
perpendicular to each other and must satisfy boundary conditions of zero electric
field and zero magnetic field at the cavity surfaces. The waves are adjusted to the
cavity dimensions. The electromagnetic waves satisfy the wave equation:
∂ 2E ∂ 2E ∂ 2E 1 ∂ 2E
. + + = (8.3)
∂x 2 ∂y 2 ∂z2 c2 ∂t 2
For a cubic cavity with side length L, in thermal equilibrium, the electromagnetic
waves of different frequencies in the cavity are represented by periodic functions
forming standing waves, also recognized as the resonant vibrational modes of the
cavity.
n πx n πy n πz
1 2 3 2π ct
. E(x, y, z) = A sin sin sin sin (8.4)
L L L λ
By dealing with the escaping energy from the cavity in perpendicular way to the
hole, a radiated energy with dependence on the frequency .ν is found. From a
classical perspective, each vibrational mode represents an independent degree of
freedom and has the possibility to exchange energy with the other modes. According
to the equipartition theorem of classical statistical mechanics, the vibrational modes
have the same energy .kB T in thermal equilibrium. The number of cavity modes per
frequency and per volume is proportional to .ν 2 .
1 An electromagnetic wave is understood as an electric field and a magnetic field in interaction with
mutual perturbations on each other, jointly oscillating and travelling in space with the speed of
light. The electric field may been considered as the source of the magnetic field from a relativistic
perspective, but both of them are treated as the electromagnetic field, governed by the Maxwell
equations.
8π 2
. ρ(ν) = ν (8.5)
c3
The total radiated energy per volume and per frequency corresponds to the number
of modes times the energy per mode.
. u(ν, T ) = 8π
c3
ν 2 kB T (8.6)
According to this expression, the waves in the cavity with short wavelengths
(or high frequencies) contribute more to the energy than the waves with large
wavelengths (or low frequencies), as the energy contribution scales quadratically
with the frequency, Fig. 8.5. If we base our predictions on classical mechanics,
there is no restriction on the radiated energy in acquiring any value. In consequence,
for increasing frequencies, the intensity spectrum predicted by classical mechanics
increases without limits (and the spectrum rapidly diverges when the frequency
approaches the ultraviolet region). The corresponding expression of the radiation
intensity or spectral radiance is known as the Rayleigh-Jeans law, presently known
to be valid at low frequencies solely.
. I (ν, T ) = (2/c2 ) ν 2 kB T (8.7)
In Sect. A.1, expressions 8.5–8.7 are derived.
8.4 Wien’s Displacement Law
The experiments on objects that approximately exhibit features of a blackbody show

a curve of the radiation intensity with a bell shape that eventually decays in the
ultraviolet region of the electromagnetic spectrum. For increasing temperatures, the
prominence moves to smaller wavelengths (or higher frequencies). The wavelengths
of the peaks are related to each other by the Wien’s displacement law, which
indicates that the product of a wavelength .λmax registered at the maximum of the
radiation spectrum with the equilibrium temperature .Teq is constant.
. λmax Teq = 2.898 × 10−3 mK (8.8)
There is an analog expression in terms of the maximum frequency, .νmax T −1 =

5.879 × 1010 .H z K −1 . The Wien’s displacement law in its different forms is derived
in Sect. A.1.5.
8.6 Planck’s Law 163
8.5 Ultraviolet Catastrophe
The experimental observations contradict the predictions of classical mechanics

referent to the shape of the blackbody spectrum, especially in the region where
an infinite number of vibrational modes is conceived from the classical perspective.
Such a discrepancy is catastrophic in the sense that it is of fundamental importance
in physics, receiving the name of ultraviolet catastrophe. However, we find the
appropriate explanation of the blackbody radiation in the wave theory of matter
and, in particular, using the Planck’s law on the energy.
8.6 Planck’s Law
In order to explain the maximum radiation intensity for every temperature of the
blackbody, Planck postulated an energy quantization based on the vibrational modes
in the cavity. Specifically, the electromagnetic waves in the cavity only take discrete
packets of energy.
. E = hν, 2hν, 3hν, · · · (8.9)
The basic energy unit is .hν, and it represents a minimal energy and is called the
quantum (from the Latin word which means “how much”). According to the energy
quantization, the energy states can only acquire integral multiples of the energy .hν,
where h is the Planck’s constant and .ν is the frequency of the vibrational mode.
The radiation in the cavity and emitted by the cavity is quantized and comes in
energy packets defined by the frequency and Planck’s constant. One of the main
consequences of assuming the energy quantization is found at high frequencies (or
short wavelengths), because high frequency modes have lower probability to occur,
imposing a probabilistic upper limit to the radiation intensity. The blackbody’s
radiative intensity was proposed from experimental results and fitted by Planck in
the form:
. I (ν, T ) = (2hν 3 /c2 ) [exp(hν/kB T ) − 1]−1 (8.10)
The speed of light in the medium (it may be the vacuum) is c, and the Boltzmann
constant is .kB . The intensity I is the energy radiated in the normal direction per
unit surface area, per unit solid angle, per unit time, and per unit frequency at the
given temperature T .2 The radiated energy is emitted in integral multiples of .hν,
becoming more intense at high temperatures for a given frequency, but with upper
bounds on the intensity. Planck’s law explains why the Rayleigh-Jeans and Stefan-
Boltzmann laws are partially valid, Fig. 8.5. The energy quantization is in agreement
2 Planck’s constant is determined by adjusting its value to reproduce the blackbody spectrum in the
blackbody expression.
with the experimental facts and eludes the ultraviolet catastrophe (the Planck and
Stefan-Boltzmann laws are derived in Sect. A.1).
Problem 8.1 By using Planck’s law, Eq. 8.10, referent to the radiation
intensity, I , and integrating over the variables .(ν, θ, φ), deduce the Stefan-
Boltzmann law expressed by .J = σ T 4 , where J is the total radiated power, T
is the absolute temperature, and .σ is the Stefan-Boltzmann constant. Discuss
the possible applications of such a law. The following integrals will be
required.
2π π/2 ∞ x3 π4
. J1 = dφ dθ sin θ cos θ = π ; J2 = dx =
0 0 0 exp(x) − 1 15
Solution The blackbody radiates energy with all frequencies in every direc-
tion. The total radiated power is obtained by integrating over the variables .ν,
.φ, and .θ .
∞ 2π π/2
J =
. dν dφ dθ sin θ cos θ I (ν, T ) ;
0 0 0
2hν 3 1
I (ν, T ) =
c2 exp(hν/kB T ) − 1
The function .cos θ in the kernel of the integral derives from Lambert’s cosine
law; it considers the deviations in the radiation from the normal direction to
the surface. By making the change of variables .x = (h/kB T )ν, we have:
4
2h kB T 2π 5 kB4
. J = 2 J1 J2 = σ T 4 ; σ =
c h 15 c2 h3
This is the Stefan-Boltzmann law, where .σ is a constant and the radiated

power is proportional to .T 4 . The law is frequently used to estimate the
temperature of interstellar objects, such as stars and planets, by measuring
their radiated power, Fig. 8.2. The temperature may be similarly assessed by
knowing the wavelength at the maximum of the intensity spectrum according
to the Wien’s displacement law.
The blackbody radiates more energy when the energy supply is increased. The
blackbody is an idealized object as the energy emission is totally compensated by
the energy absorption, establishing a perfect pattern of the radiation spectrum. Any
other body in thermodynamic equilibrium cannot radiate more energy than that of a
blackbody. The blackbody spectrum is commonly used to determine the deviations
8.7 Franck-Hertz Experiment 165
Fig. 8.2 Many interstellar objects like stars, planets, and more radiate light in all directions in the
photosphere envelope. Some light is internally reflected (black arrows), and other light escapes to
outer space (red arrows). By monitoring the light emitted to outer space and assuming that such
interstellar objects behave like a blackbody, their temperatures are determined from the equations
of the blackbody
of radiation intensity of real or “non-perfect” objects, generically referred to as gray

objects. The deviations from the ideal radiation spectrum are attributed to molecular
composition, surface roughness, energy diffusion in the material, and more, which
modify the radiation emission of the object.
One of the objects that is close to reproduce the properties of a blackbody is
a black hole. The black hole acts as a perfect absorber since it absorbs all the
electromagnetic waves that come in contact with it, and without any reflection
or transmission. The black hole is also a blackbody radiator in a process called
Hawking radiation. The investigation on the blackbody radiation is old but far from
being over. According to the Fulling-Davies-Unruh effect, a non-inertial observer
registers a non-zero temperature, whereas an inertial observer registers a zero
temperature. If a is the acceleration of the non-inertial observer, the Fulling-Davies-
Unruh temperature is .T = h̄a/(2π ckB ), a quantity similar to Hawking’s radiation,
where a replaces the gravitational acceleration. In this context, the inertial and non-
inertial observers in principle measure different energy states, unexpectedly leading
to different numbers of particles in the same vacuum.
8.7 Franck-Hertz Experiment
The quantization of the energy is a fundamental aspect of microscopic matter. The

Franck-Hertz experiment gives evidence of the energy quantization in atoms. In
such an experiment a flux of electrons is produced to induce the collision of the
moving electrons with the electrons bonded to atoms, which are in the gas phase
state. The collisions of interest are of inelastic type as they modify the internal
atomic structure. When an accelerated electron has the “appropriate energy” and
collides with a bonded electron of the atom, the incident electron gives its kinetic
energy, and the bonded electron is excited to a higher electronic orbit. The excitation
is possible due to the fact that the supplied energy is a quantum of energy that
matches the energy difference of the electronic orbits. The external electrons that
come to a halt give no signal of an electric current in the electrodes of the Franck-
Hertz experiment, until they are again accelerated by the external voltage. The
changes of voltages and the changes of electric currents are related to the excitation
energies of the atoms. The excitation energies using the electron bombardment
technique reproduce the excitation energies registered in the absorption of photons
by atoms in other experiments. In this regard, the change of electronic states is
independent of the technique employed, as long as the energy provided to the atom
(either by an external particle in collision or by the absorption of an electromagnetic
wave) corresponds to an energy quantum of the atom.
The atom returns to the ground state when the electron de-populates the high
energy state with energy .Ef and decays to the energy state with lower energy .Ei .
The transition has radiative nature as a photon with frequency .ν is emitted in the
decay process to preserve the energy.
. ν = (Ef − Ei )/ h (8.11)
The spectral radiation is characteristic of the atom. The emission and absorption
spectra of an atom are the same (except, for instance, in the fluorescence phe-
nomenon where a molecule may show a non-radiative transition due to vibrational
relaxation, dispersing energy to the environment). A refined Franck-Hertz experi-
ment is required to determine the excitation energies of the inner electronic orbits,
because they are masked by the electronic transitions of the first energy levels. Not
all the electronic transitions are permitted due to symmetry rules in the angular
momentum and the spin of the electrons, Sect. 10.7. Still, among the possible
excitations of an atom, the one of general interest is that where the electron makes
a transition to the continuum, namely, a transition to the electronic state where the
binding energy of the electron is zero, as the electron is released from the atom and
the atom is left with a positive charge in this case. The required energy to knock the
most weakly bonded electron out of the atom is an energy quantum and corresponds
to the ionization potential of the atom. Thus, the atoms can be ionized by giving the
appropriate kinetic energy to the external electrons in the Franck-Hertz experiment.
When the electron density is sufficiently high, the ionized atoms recover their
charges by capturing electrons from the dense environment. On the other hand, when
the atomic density is low, there is a possibility to produce avalanche effects since not
only the external electrons but also the released electrons have large mean free paths,
that is, they can travel large distances achieving high speeds and ionizing more
atoms by collision. The control of the temperature is equally important to minimize
the thermal velocities of the atoms in the experiment. As an example, the formation
of an Aurora Borealis is due to the collisions of particles in the atmosphere, resulting
in a burst of photons. Concluding, the Franck-Hertz experiment is relevant because
it demonstrates the energy quantization of the electron states in atoms.
8.8 Heisenberg’s Uncertainty Principle 167
8.8 Heisenberg’s Uncertainty Principle
The theory of classical mechanics is based on what we perceive at macroscopic

scales from the physical world. According to classical mechanics, the trajectory
of a particle is well-defined when the initial conditions are given, and in conse-
quence, the particle’s trajectory is predictable. However, this is not the case of a
microscopic particle. The inherent wave aspect of a microscopic particle leads to
an indeterminacy in the position, . x, and momentum, . p. In this regard, there is a
probabilistic factor in the prediction of the particle’s trajectory. The indeterminacy
in the predictions is not related to the inaccuracy of the measuring devices but to
the intrinsic wave behavior of the microscopic particles. This is usually illustrated
as follows. Suppose we are interested in determining the particle’s position in the
laboratory. We point a beam of light to the particle. The radius of the particle is
r, a non-zero value, and the wavelength of the light is .λ. We choose .λ ∼ r. The
position of the particle is registered by the detectors when the light is deviated by
the particle. However, an electromagnetic wave is a type of energy that can transfer
momentum, even though it does not have mass. The energy of a particle, whether
it has mass or not and whether it is moving or not, is described using the equation
.E = pc+mc . The last term of this equation applies only to particles that have mass
2
and in a steady state. Photons, on the other hand, have no mass but have energy due
to their momentum, which is described by the equation .E = pc. This energy can
be transferred to the particle, and we cannot accurately establish the momentum of
the microscopic particle. On the other side, if the beam is less energetic and .λ r,
the momentum is accurately determined at the expense of an indeterminacy of the
particle’s position, Fig. 8.3. In general terms, we disturb a particle by measuring
its position while the conjugate variable, p, is instantaneously affected and vice
versa. Thereby, it is not possible to simultaneously determine the exact position and
momentum of the particle.
Fig. 8.3 The detection of a particle with radius r is done using a light pulse or wave packet with
wavelength .λ ∼ r. The pulse is the superposition of many waves. The larger the number of waves
forming the pulse, the more localized the pulse is. A small pulse width implies a small uncertainty
in detecting the particle’s position; nevertheless, the many waves that build the pulse imply a
large uncertainty in the momentum according to the theory of Fourier transforms. In contrast, a
large pulse width implies a large uncertainty in the particle’s position but a small spread of the
momentum
If we consider a mass of 1 Kg hanging from a rope with length 10 cm, and

oscillating in a gravitational field, .g = 10 m/s 2 , then the root mean square
deviations of the position and momentum computed quantum-mechanically are
.10
−18 m and .10−17 kg m/s, respectively. These values are negligible with respect to
the dimensions of ordinary macroscopic bodies. This is not the case for microscopic
particles [1].
Problem 8.2 A pulse, represented by .ψ, is a wave packet formed by the

superposition of a large number of waves. The waves have amplitudes .Ai
and discrete
wave numbers .ki . The waves superposition is described by
.ψ(x) = i Ai cos(ki x). However, by working in the continuum of wave
numbers, the wave superposition is represented by a Fourier transform,
.ψ(x) = (2π )
−1/2 A(k) cos(kx) dk. Propose the wave-amplitude function
.A(k) to reproduce the shape of a wave pulse.
Solution The Fourier transform may be written in abbreviated form using

complex numbers as .ψ = (2π )−1/2 A(k) exp(ikx) dk, which is a suitable
expression to perform the integrations. In this case .ψ(x) = Re[ψ (x)], where
Re indicates the real part of the complex wave function .ψ . Next, with the
purpose of generating a single and spatially concentrated wave that represents
the pulse, the superposition of waves requires modulation. Since we have the
freedom to choose the waves amplitude .A(k), we choose a Gaussian envelope
to only consider those parts of the waves that are inside the Gaussian envelope.
A normalized Gaussian distribution, with mean .x and standard deviation .σ ,
is .(2π σ 2 )−1/2 exp{−[x −x]2 /(2σ 2 )}. We take the normalized Gaussian dis-
tribution to introduce the waves modulation, .A(k) = (2α/π )1/4 exp[−α(k −
k)2 ]. By inserting .A(k) in the previous Fourier transform and integrating, we
obtain .ψ (x) = (2π α)−1/4 exp(ikx) exp[−x 2 /(4α)]. For constant values of

.α and .k, the function .ψ renders a pulse shape.
The conclusion on the position and momentum of a microscopic particle is sum-

marized in an important principle with name Heisenberg’s uncertainty principle.
It indicates the impossibility of simultaneously determining the particle’s position
and particle’s momentum in an accurate way. In mathematical form, the principle
introduces a lower limit in the product of the uncertainties of the particle’s position
and the particle’s momentum.
. x p ≥ h̄/2 (8.12)
Next, consider a particle represented by the superposition of many waves with a

Gaussian modulation to produce a wave packet.
8.8 Heisenberg’s Uncertainty Principle 169
∞
−1/2
ψ = (2π ) A(k) exp(−ikx) dk
. −∞
A(k) = (2α/π ) 1/4
exp[−α(k − k)2 ]
When the integral is performed, we obtain the representation of the particle by a

wave packet.
. ψ = (2π α)−1/4 exp(ikx) exp[−x 2 /(4α)] (8.13)
where the real part of .ψ should be considered. The probability distribution function
of the position x is computed from the square of .ψ, that is, .|ψ|2 = ψ ∗ ψ =
(2π α)−1/2 exp[−x 2 /(2α)]. It corresponds to a Gaussian function with standard
deviation .σx = α 1/2 . On the other hand, the Gaussian envelope of the wave packet
is .A(k) = (2α/π )1/4 exp[−α(k − k)2 ]. The probability distribution function of
the wave number is .|A|2 = (2α/π )1/2 exp[−2α(k − k)2 ]. The standard deviation
in this case is .σk = (4α)−1/2 . The standard deviations .σx and .σk give the dispersion
extents of x and k with respect to the central values .x and .k. In other words,
the standard deviations give the uncertainties . x and . k in the position and wave
number. From the expression of the momentum, .p = h̄k, the dispersion of the
wave number is equivalent to the dispersion of the momentum, . k = p/h̄. The
product .σx σk = 1/2 is now transformed onto . x p = h̄/2. This is the Heisenberg
uncertainty principle derived from a wave-packet description of the particle. For
other types of probability distribution functions, the dispersion increases. Therefore,
the general expression which involves the uncertainties is . x p ≥ h̄/2.
We have assumed that the particle is described by a wave packet, and we know
that it satisfies initial conditions on the Heisenberg uncertainty principle, . x0 p0 =
h̄/2. The uncertainty in the momentum may be assigned to an uncertainty in
the particle’s velocity, . p0 = m v0 . The Heisenberg uncertainty principle in
terms of the velocity now reads . x0 (m v0 ) = h̄/2. By solving for . v0 , we
have . v0 = h̄/(2m x0 ). In general, we may write . x(t) = vt to find that
. x(t) = h̄t/(2m x0 ). We observe from this expression that the uncertainty in the
position increases with time, that is, there is a spatial spread of the wave packet. The
Heisenberg principle also provides a qualitative picture on the size of subatomic
particles, such as the electron and proton. The proton has a large momentum due
to its mass, and the electron has a lower momentum due to its small mass. In this
respect, the proton is distributed in a smaller spatial region than that of the electron.
Heisenberg’s uncertainty relation is a fact that establishes a general principle and
applies to any other pair of conjugate variables, like energy and time.
. E t ≥ h̄/2 (8.14)
The expression indicates the impossibility of exactly determining the energy in a

short time interval and vice versa. Expressions 8.12 and 8.14 have applications in
high energy physics where, for example, the virtual particles of high energy have
short lifetimes, thus relating the energy, time, and distance ranges of the fundamental
forces. Yet the point to be aware of is the impossibility of simultaneously observing
both aspects with infinite precision.
In order to use the uncertainty principle and illustrate the way in which it suggests
to abandon concepts of classical nature, let us consider an electron in an atom. This
system represents a bound state of the electron; however, it could be also the case of
a microscopic particle confined in a box. The electron moves in the neighborhood
of the nucleus, and it is not localized at the atom center; otherwise, it would have
no momentum, and the uncertainties in the position and the momentum would be
zero, . x = 0, and . p = 0, contradicting Heisenberg’s uncertainty principle. In
this respect, the uncertainty relation implies a non-zero momentum of the electron
in the atom. The minimal energy associated to this motion receives the name of
zero-point energy, and it constitutes a property of confined microscopic particles,
even under low-temperature conditions. In classical mechanics, the minimal motion
of a particle has zero momentum, and the particle is at rest. In consequence, there is
neither a factor of likelihood in determining the particle’s position and momentum
nor a zero-point energy at the classical level of theory.
Another important result on applying the uncertainty principle in the form
. E t ≥ h̄/2 involves a spatial region of confinement. Under the classical
perspective, the vacuum is a region of space empty of matter and energy. This is
not the case at the quantum-mechanical level, because the vacuum under confining
conditions may have no matter, but there is background energy there. According
to the equation .E = mc2 , matter is a form of energy, and a particle represents an
energy-condensed structure occupying a small space region (a photon, in contrast,
is understood as a structured energy traveling in space). In this respect, there is
a non-zero probability on the transformation of energy onto sub-atomic particles
with time. Such sub-atomic particles, known as virtual particles, come in pairs
of the particle-antiparticle type, and their existence takes just a small fraction of
time, making difficult the observations before they annihilate each other. Those
vacuum fluctuations are believed to be responsible for the Lamb shift, referent to
the spontaneous emission of energy quanta by the nuclei and the atoms when they
return to their ground states, and the Casimir effect as well. The contemporary
understanding of vacuum is complex, and unsurprisingly, it is related to the
spacetime structure of the Universe.
A.1 Appendix: Planck’s Radiation Intensity Law
The goal of this section is to provide a formal discussion of the Planck’s radiation
intensity, and use it to derive other related laws initially proposed in the frame of
classical mechanics. The derivation of the classical results from an initial quantum
proposal that introduces energy quantization exemplifies the general character of the
quantum theory to obtain the classical theory.
A.1 Appendix: Planck’s Radiation Intensity Law 171
A.1.1 Average Energy
Planck hypothesized that the modes and energy spectrum in the blackbody cavity
correspond to the modes and energy spectrum of an ensemble of thermalized
harmonic oscillators, with a harmonic oscillator for each possible frequency. The
photons may be radiated or absorbed in the cavity, achieving an equilibrium at
temperature T . The cavity has non-movable walls, is cubic of side length L, and
is in thermal equilibrium with the oscillators. In contraposition to a classical energy
that changes in continuous form, the energy of the oscillators is supposed to be
quantized. We require the probability distribution function, P , of the energy levels
of the particle system to obtain the average energy of the photons.
∞
P = exp(−En(m) /kB T ) / exp(−En(m) /kB T )
. m=0 (8.15)
En(m) = (m + 1/ 2)(hc) / (2L) n21 + n22 + n23
The quantum number in the expression of the energy is m. The positive integral
numbers .n1 , .n2 , and .n3 form the vector .n = (n1 , n2 , n3 ) with magnitude .n =
(n21 +n22 +n23 )1/2 . They are related to the wavelengths .λi in the three different cavity
axes using the condition of stationary waves, that is, waves that oscillate with time
but without traveling in space. When a wave is reflected on a surface, the reflected
wave is the wave that would have travelled further ahead with no reflecting surface.
The superposition of waves creates a stationary wave when .ni λi /2 = L. The 1/2
factor is the minimum length of the wave to show a .180◦ phase shift.
. λi = 2L/ni ; ni /(2L) = ν/c ; i = 1, 2, 3 (8.16)
The purpose is to simplify the expression of both the energy and probability
distribution function by following the same procedure applied to a particle in a rigid
box. We introduce the .ε energy and the Z function; this last one is the normalization
factor of P in Eq. 8.15.

1
En(m) = m + ε ; ε = (hc)n / (2L)
2
∞ ∞ m
. 1 ε −ε (8.17)
Z= exp(−En(m) /kB T ) = exp(− ) exp( )
2 kB T KB T
m=0 m=0
Z = exp[−ε/(2kB T )] / {1 − exp[−ε/(kB T )]}

The geometric series . ∞ m = 1/(1 − r) was used in the last identity to reach
m=0 r
the closed expression .exp(−x) ∞ m=0 [exp(−y)] = exp(−x) / [1 − exp(−y)]. The
m
average energy .E is computed from the probability distribution function.

∞ (m) (m)
. E = m=0 En exp(−En /kB T ) / Z = −∂ ln(Z) / ∂(kB T ) (8.18)
By using function Z in terms of .ε, Eq. 8.17, in the last identity, the average energy
is:
. E = −∂ ln(Z) / ∂(kB T ) = ε/2 + ε / {exp[ε/(kB T )] − 1} (8.19)
A.1.2 Density of States
The average energy of Eq. 8.19 does not introduce the cavity size, neither the energy
stored in the blackbody cavity, yet, it is possible to calculate the number of energy
states in which the electromagnetic energy of the blackbody is distributed.
The energy states are given in terms of the numbers .n1 , .n2 , and .n3 , which take
(m)
part in the expression of the energy .En , Eq. 8.15. There are states that are different,
like .(n1 = 2, n2 = 1, n3 = 1) and .(n1 = 1, n2 = 2, n3 = 1), but they give the
same energy. The integers .n1 , .n2 , and .n3 form a 3D grid. The blackbody cavity is
supposed to contain a large number of energy states, and in order to determine that
number, we visualize the number of states as building a big sphere whose center is
the origin of the three orthogonal axes. The number of states is proportional to the
sphere volume (the particular spherical shape alluded here is not important because
an integration over an increasing spherical shell thickness shall be carried out later
to cover the whole 3D space). However, the numbers .n1 , .n2 , and .n3 are positive,
and this characteristic takes us to concentrate our attention on the positive quadrant
of the eight possible quadrants formed by the three orthogonal axes, Fig. 8.4.
The volume of the quadrant corresponds to 1/8 of the spherical volume. If the
radius of the sphere is given by .n = (n21 + n22 + n23 )1/2 , then the volume of the
spherical quadrant is .V = (1/8)(4π n3 /3), and the differential volume element is
.dV = π n dn/2. On the other hand, the direction of the electric field oscillation
2
is twofold polarized (as demonstrated in the Barkla experiment), and it should be

considered in the expression of the number of available energy states. Thus, the
volume is multiplied by a factor of 2 to take the two polarization states of a photon
Fig. 8.4 Left sketch: a 3D sphere is shown. The quadrant in the sphere and in color is the only
one that contains a grid of positive numbers. Right sketch: a magnification of such a grid is shown.
An arbitrary vector .n is expressed as a set of integral numbers .(n1 , n2 , n3 ). The volume of the
quadrant is 1/8 the volume of the sphere
into consideration. The differential volume element is now .dV = π n2 dn. The next
step links dV to the energy .ε using the relation between n and .ε of Eq. 8.17 with
the purpose of dealing with the energy .ε. This is a linear relation, .dε = (hc/2L)dn,
with a scale factor that introduces the cavity length, L. The volume element in terms
of .ε is:
2
2L 2L 8L3
. dV = π n dn = π
2
ε dε = π 3 3 ε2 dε (8.20)
hc hc h c
The identity given above introduces the volume of the cavity and gives the density
of states, that is, the number of energy states in the interval .[ε, ε + dε]. By using the
second expression of Eq. 8.16, .ni /(2L) = ν/c, the energy .ε is .hν, and the number
of states in terms of the frequency is obtained.
. dV = (8π L3 /c3 ) ν 2 dν (8.21)
This expression may be compared with Eq. 8.5 of the main text referent to the
number of cavity modes per frequency and per volume.
A.1.3 Energy per Volume per Frequency
It is important to determine the electromagnetic energy, U , in the cavity with the

purpose of deriving the radiated power of the blackbody. In order to obtain the
energy per volume, the number of available states .dV = π(8L3 )/(h3 c3 ) ε2 dε
should be multiplied by the average energy; this last quantity introduces the
occupation probability of the photon states, Eq. 8.19.
∞
U 1 ε 8π ∞ ε3
. = 3 E − dV = 3 3 dε (8.22)
L3 L 0 2 h c 0 exp(ε/kB T ) − 1
The zero-point vibration energy .ε/2 was subtracted because it is a constant and
we require to determine energy differences. By taking the expression .ε = hν into
consideration, the energy u per unit volume and unit frequency is:
∞
U
. u(ν, T ) = (8π ν 2 /c3 ) hν / [exp(hν/kB T ) − 1] ; = u(ν, T )dν
L3 0
(8.23)
The energy u was proposed by Planck; however, it is presently recognized as a
particular expression of statistical mechanics, which is an approach based on a
family of distribution functions. Expression 8.23 uses the Bose-Einstein energy
distribution function of a gas of Bose particles at maximum entropy and in
thermodynamic equilibrium with the environment. The energy u is indeterminate
when .ν = 0. On the other side, the modes with high frequencies are less likely to
Fig. 8.5 The energy emitted

by a blackbody per unit
volume per unit frequency is
shown in terms of the
frequency. The prediction
based on classical mechanics
is given by the
Rayleigh-Jeans law, Eq. 8.24,
and the one based on
quantum mechanics is given
by the Planck’s law, Eq. 8.23.
The two predictions only
match at low frequencies
occur as the energy u goes to zero when the frequency becomes large. According to
classical theory, the frequency .ν may take any value and grow without restrictions.
The classical limit, which reproduces Eq. 8.6 and is related to the Rayleigh-Jeans
law, is obtained in the limit of low frequencies, .hν kB T , from Eq. 8.23. Thus,
by expanding the exponential function in the expression of u in a power series and
retaining the first order term, we obtain:
. u(ν, T ) = (8π/c3 ) ν 2 kB T (8.24)
The classical and quantum mechanical versions of the energy u coincide at low
frequencies. The classical version of the energy grows without limits, leading
to the ultraviolet catastrophe at high frequencies, Sect. 8.5. In contraposition, the
quantum mechanical version of the energy shows a maximum value of a function
that eventually decreases and fades away at large frequencies, Fig. 8.5.
In order to determine the energy in the blackbody box, we perform the integral of
Eq. 8.23 over .ε. To facilitate the integration, the dimensionless variable t is defined
by .t = ε/kB T .
∞ t3
dt = π 4 /15 ; L3 = V
0 exp(t) − 1
. (8.25)
U 8π 5
= (kB T )4
V 15(hc)3
The energy per volume is found to be proportional to .T 4 , where T is the temperature

of equilibrium. By introducing the Stefan-Boltzmann constant .σ , Sect. 8.6, we
obtain:
U
. = (4σ/c) T 4 ; σ = (2π 5 /15) kB4 / (c2 h3 ) (8.26)
V
Fig. 8.6 A surface that radiates energy in the form of electromagnetic waves is in a vertical
position. A parallel screen registers the radiated energy at a distance x from the surface. The
detecting area of the screen is A. Some of the radiation detected by the screen makes an angle
.θ with respect to the horizontal direction
The energy U should not be confused with the Stefan-Boltzmann law, .J = σ T 4 ,

referent to the radiated power by the blackbody discussed in Sect. 8.6.
A.1.4 Irradiance
The irradiance or power is the energy per unit time emitted by a surface of unit area.
Occasionally, the irradiance is referred to as the luminous intensity. Our interest is
to compute the irradiance produced by a blackbody, which escapes through the tiny
hole located in one of the cavity walls. To do this, let us consider an internal wall of
the blackbody cavity. It is a surface in thermal equilibrium with the electromagnetic
radiation contained in the cavity. Suppose that a screen of area A is positioned at a
distance x from the wall surface with the purpose of detecting the radiation emitted
by the wall. The radiation travels with speed c and makes an angle .θ with respect
to the horizontal direction, Fig. 8.6. The irradiance R registered on the screen is
geometrically established.
. R = u cos θ / (A t/ cos θ ) (8.27)
The factor .t/ cos θ is the time it takes the radiation to travel from the emitting surface
to the screen, and .u cos θ/A is the corresponding energy per unit area registered
on the detecting screen. When .θ = 0◦ , that is, when the screen directly faces the
emitting surface, the traveling time is .t = x/c, and the recorded energy per unit area
is .u/A. In this case, we observe the minimal time of travel and maximal energy per
unit area.
The radiation detected on the screen comes from every direction. However,
in thermal equilibrium, one half of the total number of waves travels toward the
surface, and the other half travels away from the surface. Therefore, a factor of 1/2
should be considered in the expression of the irradiance. The total radiated power
by the surface and detected on the screen is determined after integrating over all
values of .θ . By using the identity . cos2 θ dθ = θ/2 + (1/4) sin(θ/2), integrating
from .−π/2 to .+π/2, and normalizing with respect to the integration interval, .π , the
result is .1/2. The expression of the irradiance is:
. Rtot = u/(4tA) = cu/(4V ) (8.28)
where the change of variables .t = x/c and .V = Ax was used. The luminous
intensity I is related to .Rtot by .I = Rtot /π . In this regard, by using the energy u
in the forms given by Eqs. 8.23 and 8.24, the intensity in quantum theory (qt) and
classical theory (ct) is:
Iqt (ν, T ) = (2ν 2 /c2 ) hν [exp(hν/kB T ) − 1]−1

. (8.29)
Ict (ν, T ) = (2ν 2 /c2 ) kB T
The first equation is the Planck’s radiation law, Eq. 8.10, and the second is the
Rayleigh-Jeans radiation law, Eq. 8.7. We now assume a radiator whose luminous
flux changes by direction in the form .cos θ , where .θ is the angle between the
normal of the surface and the direction of the radiation. Such a dependence is
known as the Lambert’s cosine law. The integration of Planck’s expression over
all the frequencies .ν, and over the angles .θ and .φ describing the hemisphere above
the surface, gives the Stefan-Boltzmann law, Sect. 8.6. The following purpose is to
write the Planck’s equation in terms of the wavelength .λ. Here, it is important to
establish the relation between the spectral variables .λ and .ν in the form .λν = c and
.dν/dλ = −c/λ . Since the energy u of Eq. 8.23 is an energy per unit volume and
2
per unit frequency, we depart from the equation:
. u(λ, T ) dλ = −u(ν, T ) dν (8.30)
The negative sign is due to the inverse relation between the spectral variables .ν and
λ. The final expression of the energy per unit volume and per unit wavelength is:
.
. u(λ, T ) = (8π hc/λ5 ) {exp[hc/(λkB T )] − 1}−1 (8.31)
A similar transformation is obtained for the wave number k, where .k = 1/λ. The
next step consists on obtaining the maximum intensity from the above Planck’s
expression to demonstrate the Wien’s displacement law, Eq. 8.8.
A.1.5 Maximum Intensity
The Planck’s radiation expression gives the maximum radiation that a body can
radiate in thermal equilibrium. The radiated intensity reaches a peak emission at
a finite wavelength, .λmax and only depends on the temperature T . Planck’s law in
terms of the wavelength is given in Eq. 8.31. The maximum of the energy is obtained
by differentiating u with respect to .λ, making the resulting expression equal to zero
and solving such an equation for .λ = λmax . The process leads to the result:
. x [1 − exp(−x)]−1 = 5 ; x = hc/(λkB T ) (8.32)
The first expression is a transcendental equation whose solution is found numeri-

cally, .x = 4.965, 114. Such a value implies a constant for the product:
. λmax T = 2.898 × 10−3 m K (8.33)
The above expression is the Wien’s displacement law, Eq. 8.8. By doing the same
process for u in terms of the frequency, we obtain .3[1 − exp(−x)] − x = 0,
with .x = hν/kB T = 2.821, 439, and the maximum frequency is .νmax T −1 =
5.879 × 1010 H z K −1 , Fig. 8.5. Thus, by supposing radiation of blackbody type,
the temperature of an object, like that of a star, is determined from the wavelength
or frequency of the object’s electromagnetic emission. The process of establishing
the source temperature in this way is known as the color-assigned temperature.
A.1.6 Time Interval of Temperature Cooling
It is of interest to estimate the time interval of radiative cooling of a blackbody

after the energy supply is switched off. The energy of the system is supposed to be
exclusively due to the temperature and given by .E = 3NkB T /2, corresponding
to the energy of an ideal gas where N is the number of moving particles. The
dissipation rate is established by the Stefan-Boltzmann law, .dE/dt = σ AT 4 ,
Sect. 8.6. The time of heat loss is calculated by applying the chain rule, .dE/dt =
(dE/dT )(dT /dt). This expression is equal to .σ AT 4 . Thus, the time of cooling
depends on the way in which the body emits radiation.
dE dT
. = (3NkB )/2 = σ AT 4 (8.34)
dt dt
The second identity is a differential equation of the temperature in terms of t. The
solution is obtained after integration.
. t = [NkB /(2σ A)] Ti−3 − Tf−3 (8.35)
The fact that a real body conducts heat from its inner part to its outer surface makes
the time interval of temperature cooling longer than that of a perfect blackbody.
The time delay observed on the cooling of real objects may be approximated by
replacing .σ by .σ , where . is an emission factor, . = 1 for a blackbody and . < 1
for a real object.
. t = [NkB /(2σ A)] Ti−3 − Tf−3 (8.36)
In this regard, it is possible to establish a lower limit on the cooling time of an object
from the blackbody results.
Reference
1. C. Cohen-Tannoudji, B. Diu, F. Laloe, Quantum Mechanics, ed. 2005, vol. 1 (Wiley, London,
1977), pp. 573.
Quantization of the Angular Momentum
9
According to classical mechanics, the angular momentum of the particles may take
arbitrary values. However, the experiments have shown an opposite feature, that is,
the angular momentum and the spin of the microscopic particles are quantized. The
wave theory of matter accounts for the quantization of the angular momentum and
spin, yet the spin remains as an enigmatic property of the elementary particles.
This chapter is structured as follows: in the first stage, the orbital angular
momentum and the spin are briefly characterized from the theoretical and experi-
mental perspectives. Particular attention is given to the Stern-Gerlach experiment
on the spin of a particle. The spin value, which a particle may take, is used to
classify it as a fermionic or bosonic particle and identify the collective behavior
of such types of particles as well. In the second stage, we introduce new principles,
like Pauli’s exclusion principle and Hund’s rule, which emphasize the difference
between quantum mechanics and classical mechanics. A discussion on the basic
relations of the spin with the magnetic moment and of the magnetic moment with
both the angular momentum and the external magnetic field is presented. Finally, a
brief introduction to spin functions and spin operators is provided.
9.1 Orbital Angular Momentum and Spin
A particle that rotates has angular momentum. The total angular momentum is sep-
arated onto an orbital angular momentum, .L, and an intrinsic angular momentum,
.S. The intrinsic angular momentum is known as the particle spin. In the classical
treatment, the vectors .L and .S are given by:
. L=r×p ; S = Īω (9.1)
The variables .r and .p are the position and linear momentum of the particle, while .Ī
and .ω are the inertia moment tensor and angular velocity of the particle. The total
angular momentum constituted by the vectors .L and .S of an isolated system of
180 9 Quantization of the Angular Momentum
particles is conserved. Such vectors are permitted to take any magnitude and any
spatial direction in classical mechanics, but this is not so in quantum mechanics.
In the case of an atom with N electrons, the total electronic angular momentum
.L
corresponds to the vector sum of their individual contributions .L = N L i , and

i=1
similarly for the total electronic spin angular momentum .S, that is, .S = N i=1 Si .
The experimental observations on microscopic particles indicate that the orbital
angular momentum and spin are quantized. The quantized values are denoted by
2 2
. and s. They take part in the eigenvalues of the operators .L̂ and .Ŝ .
h̄2 ( + 1) ; h̄2 s(s + 1)

. (9.2)
= 0, 1, · · · , (n − 1) ; s = 0, 1/2, 1, 3/2, · · ·
The factor . is the azimuthal quantum number of the orbital angular momentum, and
s is the principal spin quantum number, commonly referred to as the particle spin.
Like other quantum numbers, . and s adopt discrete values. The allowed values of s
are both half-integer and integer numbers, and those of . are integer numbers with
limit .n − 1, where n is the principal quantum number.The atoms have electronic
configurations, and these are classified according to the quantum numbers. The
principal quantum number takes integral values and establishes the energy state and
energy level of an electron in the atom, while the other quantum numbers define the
energy sub-levels. The energy level with .n = 1 has more negative energy than the
.n = 2 energy level and so on. The Aufbau principle imposes a relative order by
pointing out that the electronic energy levels with low energy are populated before
the energy levels with high energy. For the electrons of atoms, the orbital quantum
number . defines the sub-shell type, leading to an orbital diagram of sub-shells with
nomenclature [1]:
s( = 0 : 2) ; p( = 1 : 6) ; d( = 2 : 10) ; f ( = 3 : 14)

.
g( = 4 : 18) ; h( = 5 : 22) ; i( = 6 : 26) ; ···
The value of the orbital quantum number and the maximum number of electrons
permitted in the atomic sub-shell are shown in parenthesis. The maximum number
of orbitals in sub-shell . is given by the sub-shell multiplicity, .2 + 1. We have a
single s orbital with two electrons, one with spin up and one with spin down, three
p orbitals with six electrons, three with spin up and three with spin down, etc. A
closed sub-shell gives zero contribution to the total electronic angular momentum
and total electronic spin. The sub-shells s, p, .· · · , adopt different spatial shapes to
accommodate the different numbers of electrons.
The quantization of .L and .S is also known as space quantization to distinguish
it from the energy quantization, both observed on the microscopic particles in
confinement. Space quantization should not be confused with a quantization of the
spacetime framework. Space quantization and energy quantization are associated
with the wave-particle nature of the microscopic particles. Any interpretation of
9.2 Characterizing a Particle with Spin 181
the intrinsic spin of the microscopic particles using classical theory contradicts the
experimental results. Some of the contradictions are illustrated below when, for
example, we consider a particle with spin as a rotating sphere of radius r, uniform
mass m, and inertia moment .I = 2mr 2 /5.
• The radius of a tiny microscopic particle, such as that of an electron, is

sufficiently small, .r → 0. In this case, the inertia moment (.I = 2mr 2 /5)
approaches zero in quadratic form, which in turn produces a spin tending to
zero. Thus, for a point particle, we essentially have .S = 0. Such a deduction
contradicts the experimental results, where the electron is observed to have a
non-zero spin.
• If we consider that .S is sufficiently small but different from zero, then we require
a very large angular velocity .ω to keep the product .S = Īω with finite value.
However, a very large angular velocity implies a mass on the sphere surface
that rotates with speed greater than the speed of light, a situation which is in
contradiction with relativistic mechanics.
• If we assume a relatively small angular velocity (to keep the angular speed of
the sphere within bounds), then we mistakenly end up with the prediction of an
electron radius that is much greater than the radius of the atom.
The classical picture of an electron as a small sphere with charge and spin is
not appropriate because it leads to contradictions in quantum mechanics. The
spin of the elementary particles cannot be explained using the existence of even
smaller particles, since the elementary particles have no inner structure. The spin
of elementary particles is a true physical attribute associated with pure quantum
mechanical effects and only defined in terms of what is observed and measured:
specifically, the spin is a physical property that makes the elementary particles
act as unconventional magnets. In the following sections, the important features of
the space quantization are distinguished after experimental observations employing
magnetic fields on microscopic particles with spin. Yet no relativistic effects are
considered in this frame of the wave theory of matter.
9.2 Characterizing a Particle with Spin
An electron orbiting around the nucleus resembles an electric current in a circular

loop and generates a magnetic field of dipolar type. The induced magnetic moment
is proportional to the angular momentum .L in the form .μL = gL qL/(2m),
Sect. A.1.1. The quantities q and m are the charge and mass of the particle, and
.gL is the g factor, Sect. A.1.2; it takes the value 1 when the mass and charge of
the particle are homogeneously distributed, and the particle is in orbital motion.
Similarly, a charged particle with spin .S induces a magnetic moment, in principle
described by an electric current in a circular loop, Sect. A.1.5.
. μs = gs qS/(2m) (9.3)
The factor .gs q/(2m) is the gyro-magnetic ratio, and the dimensionless number .gs
is the g factor for the type of particle and that type of motion. According to the
theory of quantum electrodynamics, the value of the g factor of a spinning electron is
.gs = 2.002,319,304,3622. It is the result of the self-interaction of the electron using
an exchange photon. The self-interaction smears the electron distribution by a small

fraction of a Fermi unit in the distance, resulting on a slightly weaker interaction
with the proton.
9.3 Stern-Gerlach Experiment
The Stern-Gerlach experiment is important as it gives evidence of the spin and the
spin quantization of the electron. Stern and Gerlach worked with neutral silver atoms
in their experiment to avoid a Lorentz force, .F = −qv×B, acting on charged atoms.
The silver atom is constituted by a core formed by the nucleus and 46 electrons,
plus a single electron in the outermost shell. The net angular momentum and the
net spin of the core are zero. The electron out of the core also has zero orbital
angular momentum as it is in a s (. = 0) state; however, it has non-zero spin. In this
connection, a measurement on the magnetic moment of the silver atom is in a first
approximation a measurement on the magnetic moment of the electron spin.
In the Stern-Gerlach experiment, the silver atoms are ejected from a furnace with
random orientations of their magnetic moments. In particular, the z component of
the magnetic moment may take any value between .−|μz | and .|μz | before reaching
the magnetic field of the experiment. The magnetic field is inhomogeneous with
the purpose of creating different magnetic forces on the outermost electrons of the
silver atoms. The beam of particles is collimated by a small hole on each of the
frontal screens to produce a fine particle beam. The particles enter in the Stern-
Gerlach apparatus traveling along the y axis. The magnetic field .B of the experiment
is axisymmetric in the z axis, Fig. 9.1. If the potential energy of a particle in the
magnetic field is .Ep , then the force .F = −∇Ep exerted by the magnets on an atom
in the z direction is, Sect. A.1.3:
Fig. 9.1 The Stern-Gerlach experiment gives evidence of the spin and the spin quantization of the
electron. The microscopic particles with spin behave as unusual tiny bar magnets in the presence
of a non-uniform magnetic field
9.3 Stern-Gerlach Experiment 183
∂B
Ep = −μ · B
. ; F = μz k̂ (9.4)
∂z
The force on an atom depends on .μz and the intensity B of the magnetic field.
The atoms that form the beam are deflected differently due to the different atomic
magnetic moments. The experimental magnetic field attempts to align the magnetic
moment in the z direction, thus creating a torque on the magnetic dipole and
inducing the precession of .μ in the .(x, y) plane with frequency .ω, Sect. A.1.4. In
this respect, the x and y components of the .μ vector change with time.
. ω = μz B/h̄ (9.5)
The observed pattern on a screen consists of two separated blots in the z axis. In
more detail, the magnetic field in the Stern-Gerlach experiment defines the spin
inclination relative to the z axis with only two discrete states:
. Sz = ±h̄/2
The electrons are half-spin particles. Under the consideration of .+ and .− signs, the
electron spin is pictured as a clockwise rotation or spin up and an anti-clockwise
rotation or spin down, around its rotation axis. By inserting the value of .Sz in the
magnetic moment of Eq. 9.3, with .g = 2, the z component of the magnetic moment
associated with the electron spin is:
. μz = ± eh̄/(2me ) (9.6)
The results of the Stern-Gerlach experiment contradict the classical predictions

based on the behavior of small and ordinary magnets in a magnetic field.
Problem 9.1 The mass and velocity of the particles in the Stern-Gerlach
experiment are m and v. The particles travel a distance L and impact the
screen at the spots .±z, measured from the center of the observation screen,
Fig. 9.1. Find the value of z in terms of the magnetic force, the distance L,
and the kinetic energy .Ek of the particles.
Solution In their journey to the screen, the particles are subjected to a

magnetic force .Fz = μz ∂B/∂z, Eq. 9.4, in the z direction, similar to a particle
with a given mass and traveling in a gravitational field. The height is computed
from the force .Fz .

1 1 ∂B 2
z=
. (Fz /m) t 2 = μz t
2 2m ∂z
(continued)
The travel time is determined by the velocity v of the particles and the distance
L in the form .t = L/v. The kinetic energy of a particle is .Ek = mv 2 /2. By
using the expression .t 2 = L2 /v 2 = mL2 /(2Ek ), the distance z is:
L2 ∂B
z=
. μz
4Ek ∂z
The experimental observations are independent of the reference z axis defined

in the previous analysis, and whichever axis is imposed in the experiment, two
spin values are always detected in the respective axis. In short, the Stern-Gerlach
experiment exhibits the quantization of the electron spin with specific values and
spatial orientations of the spin components. Similar experiments have shown the
quantization of the orbital angular momentum and spin of other particles, thus
determining the expressions of the .Lz and .Sz components (the z projection is
conventionally used here).
Lz = m h̄ ; Sz = ms h̄
. (9.7)
m = 0, ±1, · · · , ± ; ms = 0, ±1, · · · , ±s
The allowed values of the azimuthal quantum number, ., are those given in
expression 9.2. The .m quantity is the magnetic quantum number, which defines
the projection of the angular momentum. On the other hand, the allowed values
of the principal spin quantum number, s, are those given in expression 9.2. The
.ms quantity is the magnetic quantum number of spin, or spin projection quantum
number. A photon has integral spin, .s = 1, and the only values of the spin
component are .Sz = ±h̄ corresponding to right circularly polarized light and left
circularly polarized light; however, an electron has half-integer spin, .s = 1/2, and
the values of the spin components are .Sz = ±h̄/2. The total number of states that
create the same total electronic spin of a system is referred to as the spin multiplicity,
given by .2S + 1. An atom or molecule with one or more electrons may be in one
of the .2S + 1 possible spin states. A system is in a singlet spin state .(2 × 0 + 1)
when the total spin of the system is zero (.S = 0), it is in a doublet spin state
.(2 × 1/2 + 1) with possible spin states .±1/2, when the total spin of the system
is one half (.S = 1/2), and so on. The spin states that integrate a multiplet are
energy degenerate. Such a degeneracy in atoms may unfold into different energy
sub-levels due to the electromagnetic fields of nearby atoms in a molecule. In the
wave theory of matter, the spin states are formally described by spinors, Eq. 9.40.
The quantum number .ms has no dependence on any other quantum number. There
is not a general wave function of the spin solely due to its nature, and thus, the
spin of every particle in a molecule is frequently considered in the molecular wave
function. More information on the way the spin is introduced in quantum mechanics
is provided in Sects. B.1 and 10.9.
9.4 Wave-Particle Duality and Spin of a Particle 185
9.4 Wave-Particle Duality and Spin of a Particle
Spin quantization in the Stern-Gerlach experiment is in principle considered the

aftermath of the particle-wave character of the microscopic particles. According to
Cresser [2], such a statement is illustrated by displaying an analogy with the Young’s
double-slit interference experiment. Consider a set of Stern-Gerlach (SG) devices
arranged in series. The first SG apparatus has the magnetic field oriented in the
z direction. The beam of particles comes out of the magnetic field with the spin
components .Sz = ±h̄/2. The particles with spin down, .Sz = −h̄/2, are blocked,
and the particles with spin up, .Sz = +h̄/2, are allowed to travel to a second SG
device, preserving the orientation of the magnetic field of the first SG apparatus. The
particles that exit from the second device are the spin-up particles solely. Nothing
has changed with the presence of the second apparatus. On the other hand, when
the magnetic field of the second SG device is re-oriented, say in the y direction,
the emerging beam of the second device shows with equal probability the spin-up
and the spin-down particles in the y direction, .Sy = ±h̄/2. A third SG apparatus in
series with the second device, and the magnetic field oriented in the z direction, like
in the case of the first apparatus, shows a beam of particles with spin components
.Sz = ±h̄/2. Surprisingly, the intermediate device randomized the spin orientations
of the particles coming out of the first apparatus in the z direction, since particles
with both spins emerge from the third device.
The fact that a Stern-Gerlach device randomizes the spin orientations indicates
the loss of the particle history in the experiment and, thereby, the impossibility of
measuring any pair of spin components in different directions, such as the particles
with spin up in the y direction and the particles with spin up in the z direction. The
impossibility to simultaneously have knowledge of the spin components in different
directions is not due to the imperfection of any technique in the experiment but to
the intrinsic randomness ascribed to a spin interference effect which is beyond our
control. In this respect, the Stern-Gerlach spin experiment shows resemblance to
Young’s double-slit experiment, in spite of the different spectra. The similarities
between the two experiments is illustrated after conceiving that an interference
pattern occurs when, in general, a beam of light or a beam of microscopic particles
is presented with the option of two or more paths to travel. Let us consider several
Stern-Gerlach devices in series. In the first stage, a SG apparatus plays the role of a
beam collimator producing particles with a given spin value, say .Sx = h̄/2. In the
second stage, the beam is presented with different options in the spin of the particles,
such as .Sz = ±h̄/2, in resemblance to the paths presented by the two slits in Young’s
experiment. In the third stage, the two beams interfere. It is after the third SG device
that we perform the measurements on the spin of the particles to determine the spin
component, for example, .Sy = h̄/2, without knowing the spin components of the
particles in the previous devices, and in the same way as it occurs in the Young
experiment where we have no knowledge of the precise pathways and states taken
by the particles in traveling toward the screen. Also, there is no formation of a
diffraction pattern when we monitor any of the intermediate paths of the particles.
In conclusion, the Stern-Gerlach and Young experiments give evidence of the

different properties of the microscopic world. Still, the observations share a common
factor based on the general laws of nature, as such laws are in essence indifferent
to the types of measuring devices; however, they determine the way in which the
particles at the microscopic level manifest themselves.
9.5 Fermions and Bosons
The spin has a direction and a magnitude; therefore, it is a vector. However, it

is impossible to simultaneously determine every component of the spin vector. A
single spin component is quantized, and we must use its quantum number to describe
it. The direction of the spin may be changed, but it is impossible to increase or
decrease the spin magnitude in any experiment.
The spin is a physical observable, and we use it to classify the particles and
characterize some common behaviors. The particles with half-integer spin, 1/2,
3/2, 5/2, .· · · , are called fermions, and the particles with integral spin, 0, 1, 2, .· · · ,
are called bosons. No elementary fermions have been discovered with spin values
of .3/2, 5/2, · · · , neither elementary bosons with high spin values (the graviton is
assumed to have a spin value of 2). The Higgs boson is an elementary particle with
zero spin. The composite particles integrated by two or more microscopic particles
such as the helium-4 nucleus (.s = 0), helium-3 nucleus (.s = 1/2), or silver atom
(.s = 1/2) of the Stern-Gerlach experiment exhibit a net spin. In this respect, a
composite particle may classify as a fermion or a boson. A boson may be made of
several fermions, but a fermion cannot be made of bosons. The electrons and the
leptons have half-integer spin and classify as fermions. The photon (.s = 1) and the
force-carrier particles classify as bosons. The neutron and neutrino are fermionic
particles without electric charge, but surprisingly, the neutron has a magnetic dipole
moment because it is a composite particle formed by quarks. The fermions and
bosons have different behaviors and follow different statistical rules, originating
from different particle distributions. In particular, the fermions are ruled by the
Fermi-Dirac statistics, and the bosons obey the Bose-Einstein statistics. Still, at high
temperature, the Fermi gases and the Bose gases with a large number of particles
behave classically and are ruled by the Boltzmann statistics [3].
9.6 Pauli’s Exclusion Principle and Hund’s Rule
The quantum numbers required to describe the electronic state of an atom are n, .,
m , and .ms . They are the principal or total quantum number, n; the orbital quantum
.
number, .; magnetic quantum number, .m ; and spin magnetic quantum number, .ms .
Their possible values are presented in Table 9.1.
The order in which the quantum numbers n, ., .m , and .ms are shown in the
filling of atomic shells by electrons depends not only on the contribution of each
electron to the atomic energy but also on other physical principles and rules. The
9.6 Pauli’s Exclusion Principle and Hund’s Rule 187
Table 9.1 Main quantum Designation Quantum numbers

numbers of fermionic
Principal quantum number .n = 1, 2, 3, · · ·
particles
Orbital quantum number . = 0, 1, 2, · · · , (n − 1)
Magnetic quantum number .m = 0, ±1, ±2, · · · , ±
Magnetic quantum number of spin .ms = ±1/2

. .2 + 1 (electrons) .m
.s ( = 0) 1 (2) 0
.p ( = 1) 3 (6) .−1, 0, +1
.d ( = 2) 5 (10) .−2, −1, 0, +1, +2
.f ( = 3) 7 (14) .−3, −2, −1, 0, +1, +2, +3
Orbital quantum number, ., subshell nomenclature, .s, p, . . .,

multiplicity, .2 + 1, and the respective .m permitted values
Pauli exclusion principle indicates that two or more identical fermions in a common
system are prohibited to have the same set of quantum numbers. As a result, the
fermions must occupy different physical states. In contrast, the bosons are allowed to
occupy the same physical state. In addition to the Pauli exclusion principle, we have
the Hund’s rule, an important ingredient in the characterization of the electronic
configurations, as well. The Hund’s rule is based on a regular pattern, which has
been observed in experimental works and concluded from theoretical calculations
on atoms. It states that configurations with closed shells are equally populated with
spin-up electrons and spin-down electrons. That is, the most convenient electron
distribution in the energy levels, which are below the one being filled up, is the
double occupancy of the energy levels. However, the non-closed levels of equal
energy are populated by single electrons with similar spins. In this regard, the
fundamental state of an atom is, according to the Hund’s rule, a high-spin state;
specifically, it shows the largest number of electrons with parallel spins in the
partially filled shell.
In order to explain Hund’s rule, let us consider a set of electrons occupying
equal-energy levels with one electron in each level; they all have parallel spins.
Hund’s rule is explained by the reduction of the screening effects that electrons with
parallel spins have on the nucleus. The reduced screening of the nucleus brings the
electrons closer to the nucleus. Such a distribution increases the nucleus-electron
energy, leading consequently to an energetically stable atom. Summing up, the
electrons follow the Pauli exclusion principle and show an occupation order of the
atomic shells under the Hund’s rule and Aufbau’s principle. In contraposition, the
bosons follow no exclusion principle, and as the temperature approaches zero, they
condense in the same energy state. The different properties between fermions and
bosons is the reason that fermions satisfy the Fermi-Dirac statistics and the bosons
satisfy the Bose-Einstein statistics.
Electrons and nuclei are important components in the formation of atoms and
molecules. Electrons are negatively charged particles and the nuclei positively
charged particles. The incomplete/complete electron population of the atomic
orbitals results in unpaired/paired electrons, originating a non-zero/zero net electron
spin. This endows some materials with paramagnetic/diamagnetic characteristics.

The paramagnetic materials intensify an external magnetic field, and the diamag-
netic materials reduce an external magnetic field. The electrons tend to oppose the
compression due to both the Coulomb repulsion and the Pauli exclusion principle
[4]. In the inner core of a collapsing star, the electrons resist the gravitational
confinement of the outermost shells of the star. However, at sufficiently large
compression rates, an instability between the forces pointing in the inward radial
direction and the forces pointing in the outward radial direction eventually builds
up, resulting in the explosion of the star. On the other hand, the bosons, with no
restriction on the population of the same energy states, can behave in collective form
as in the case of a laser, a superfluid, or a superconductor. The applications based
on the spin of the particles are different, Sect. A.1.7. There is a spectroscopy based
on the magnetic properties of the electron spin with name electron spin resonance,
ESR, where the spin precessions in the analyzed sample are employed as a source of
information, and a spectroscopy based on the magnetic properties of the nuclei with
name nuclear magnetic resonance, NMR, where the signal intensity of the sample is
proportional to the nuclear magnetic moment. In addition to the imaging techniques
using the particle spins, there are applications in solid-state physics (referent to
materials with ferromagnetic, paramagnetic, and diamagnetic domains leading, e.g.,
to the scientific field of spintronics), optics (involving polarized light), etc. The
fundamental investigation on the electrodynamics of the spin and the proposal of
spin models, such as the Ising model with spin-up and spin-down states to explain
the ferromagnetism of certain materials, are active fields of research in basic science.
A.1 Appendix: Magnetic Moment
This section presents the basic relations of the magnetic moment with other physical
variables. The purpose is to clarify different aspects concerned with the Stern-
Gerlach experiment of Sect. 9.3. The departure point is the existence of atoms that
respond to an external magnetic field due to their magnetic moments. In this respect,
we consider that our atom has a dipolar magnetic field simulating that of a small
bar magnet and capable of interacting with an external magnetic field. The dipolar
magnetic field is described from a classical perspective using an electric current
in a circular loop. In a later stage, the quantum character of the “rotations” of the
microscopic particles is introduced by establishing a link to the quantization rules
of both the orbital angular momentum and the spin of the particle.
A.1.1 Electric Current in a Circular Loop
A charged point particle in motion represents an electric current, and according to

the Maxwell equations, it generates a magnetic moment .μ.
μ = (r × j) / 2 = qr × v/2
. (9.8)
A.1 Appendix: Magnetic Moment 189
The variable .r is the instantaneous position of the particle, .v is the velocity, and
.j is the current density. The 1/2 factor is chosen by convention. In the case of a
large number of particles, the magnetic moment is expressed as an integral over
the volume of the particle distribution. From the classical version of the atom, an
electron with an orbit around the nucleus describes an electric current i in a circular
loop and creates a magnetic moment .μ. The magnetic field generated by the electric
current is equivalent to that of a small bar magnet. The magnetic dipole moment
of the electron in orbit is .μ = iA, where the area enclosed by the trajectory is
.A = π r . The current is given by .i = −e/T = −eω/(2π ), with T and .ω the period
2
and angular frequency of the orbital motion. The magnetic dipole may be written as
.μ = −(eωr /2)n̂, where .n̂ is the normal vector to the area A with direction defined
2
by the right-hand rule using the electric current direction. On the other hand, the
angular momentum is .L = r × p. The magnitude of .L in terms of the particle speed
is .L = r(mv) = mωr 2 , where the velocity is .v = ωr. The relation between .μ and
.L is .μ = −eL/(2m). Therefore, the electron magnetic moment is anti-parallel to
the electron angular momentum.
. μ = −eL/(2m) (9.9)
The magnetic dipole may be described as that created by either an electric current
in a circular loop or a pair of magnetic poles relatively close to each other, Fig. 9.2.
At far distances, the magnetic fields are the same when we use any of the above
descriptions. Yet in the region close to the magnetic sources, the magnetic field
produced by the electric current loop is different from the magnetic field produced
by the pair of magnetic poles. We deal with the model using the electric current in
a circular loop as the expressions derived from it have a general aspect and describe
the magnetic moments of all the particles of an atom (the electron in orbital motion,
the electron with spin, and the nucleus with spin) with similar simplicity.
The magnetic moment .μ may be additionally understood as the property of a
system which establishes a relation between the torque .τ exerted on the system and
Fig. 9.2 Sketch on the left: an atom has a dipolar magnetic field, which is similar to some extent
to that of a tiny magnet. Sketch on the middle: the dipolar magnetic field may be described by
means of an electric current i in a circular loop. Sketch on the right: the dipolar magnetic field may
be also described by two magnetic monopoles, where one is positive and the other is negative. In
the far-field limit, the magnetic field is the same using any of the models that create the dipolar
magnetic field. In the near field, all the magnetic fields differ among one another
the external magnetic field .B, when the system is immersed in such an external
magnetic field.
. τ =μ×B (9.10)
A.1.2 Magnetic g Factor
The fact that a magnetic moment is originated by a charged particle in angular

motion brings the expression of the angular momentum .L of a particle into
consideration.
. L = r × p = mr × v (9.11)
For a many-particle system such as an atom or molecule with many electrons and
nucleons, the vector of total angular momentum is obtained after adding the total
orbital momentum of the electrons, the total spin of the electrons, and the total spin
of the nucleons. Expressions 9.8 of .μ and 9.11 of .L also share some resemblance in
those cases. The ratio of the magnetic moment to the angular momentum is known
as the gyro-magnetic ratio, .γ . The .γ ratio is the proportionality factor between the
magnetic moment .μ of the particle and the physical variable that describes the
motion type of the particle, in this case .L. If the particle has uniform mass and
uniform charge distributions, the gyro-magnetic ratio is:
. γ = gq |r × v| / (2m|r × v|) = gq/(2m) (9.12)
The number g is the magnetic g factor. The value of g depends on the type of motion
and the type of particle under investigation. For example, the value of g depends on
either the type of momentum (angular momentum or spin) or the type of particle
(electron, proton, or neutron). In the case of an electron moving in circular orbit,
the charge is .q = −e, and the magnetic moment is anti-parallel to the angular
momentum.
. μL = −eL/(2m) (9.13)
The .μL and .L quantities are parallel for a positron. The magnetic moment of a
molecule is determined by the molecular energy state. The orbital angular momenta
of the electrons and the number of unpaired electron spins in the atoms lead to
the different magnetic moments of the atoms. In turn, the magnetic properties of a
macroscopic material are derived from the magnetic moments .μi of the individual
particles that build the material. The magnetization density .M of the material with
volume V is defined by:

M = μ/V =
. μi /V (9.14)
i
The materials where a spin resonance can be observed are many. In most instances,
those materials contain spin unpaired electrons and spin unpaired nucleons, giving
non-zero magnetic moments. The magnetic materials are of organic and inorganic
types like the free radical molecules, polymers, liquid molecules, and matter
containing transition metal atoms. By using the spin resonance techniques, it is
possible to determine the inner electronic and nuclear structures, as well as the
magnetic susceptibilities of different compounds.
A.1.3 Magnetic Energy and Magnetic Work
A magnetic moment in the presence of an external magnetic field has potential

energy .Ep . This energy is computed from the work, W , performed on the magnetic
dipole by the external magnetic field in the process to align the magnetic dipole
along the external magnetic field.
rf
W =
. F · ds (9.15)
ri
The infinitesimal small displacement vector is .ds. The deviation of the vector .μ from
the direction of the external magnetic field .B produces a torque on the magnetic
moment, Fig. 9.3. A force couple is built when two equal forces separated by a
distance act in opposite directions. In our situation, a force couple is created to align
the magnetic moment, without translational motion. It is convenient to introduce the
Fig. 9.3 Sketch on the left: a bar magnet with north and south poles is immersed in an external
magnetic field .B. The external magnetic field exerts forces, F , at different points of the magnet
producing a force couple. Sketch in the middle: a particle with charge q moves in a circular
trajectory producing an electric current i. The area enclosed by the trajectory is A. The magnetic
moment induced by the electric current is .μ. The angular momentum vector .L and magnetic
moment .μ have opposite directions for the electron, .q = −e. Sketch on the right: The external
magnetic field produces a torque on the magnetic moment, as in the case of the bar magnet. The
angle between .μ and .B is .θ
angle .θ between the vectors .μ and .B and deal with the expressions of both the arc
length .ds = −n̂θ dθ × r, where .n̂θ is the unit vector perpendicular to the plane
that contains the displacement .ds, and the radius vector .r. The kernel of the integral
describing W becomes .−F·(n̂θ dθ ×r). It can be written in terms of the torque using
the circular shift rules of the scalar triple product, .n̂θ dθ · (r × F) = τ · n̂θ dθ . Under
the consideration that the torque and the angular displacement vector are parallel,
the integral describing W is:
θ θ
W =−
. τ dθ = − μB sin(θ )dθ = μB cos(θ ) = μ · B (9.16)
π/2 π/2
The work is measured for angle changes with respect to .θ = π/2. The potential
energy .Ep is the work exerted against the magnetic field, which tries to align the
magnetic dipole moment, thus .Ep = −W . The lowest and the highest potential
energies, .Ep = ±μB, are achieved when .μ is parallel and anti-parallel to .B.
When a sample in the Stern-Gerlach apparatus is subjected to an orthogonal
electromagnetic wave of high frequency, the magnetic component of the electro-
magnetic wave flips the magnetic moments of the particles in the sample. The
required radio-frequency energy to induce a transition of the magnetic moment is
. Ep = μB − (−μB), being proportional to the strength of the magnetic field.
. Ep = 2μB (9.17)
By using the Planck relation . Ep = h̄ω, we obtain the frequency of the

electromagnetic wave that flips the magnetic moment in an external magnetic field.
. ω = 2μB/h̄ (9.18)
The frequency .ω may be tuned by changing the magnetic strength B in continuous

manner. The force on the magnetic moment is obtained from the potential energy.
. F = −∇Ep = ∇(μ · B) (9.19)
If .μ is constant and the external magnetic field changes in the z direction, then the
non-zero force component is .Fz = μz ∂B/∂z. A magnetic moment in an external
magnetic field shows a deflected trajectory according to the value of .μz , Sect. 9.3.
It is known that diamagnetic materials create a magnetic moment to repel an
external magnetic field. Such a property may be used to levitate a diamagnetic
sample. To illustrate this, the force of gravity .−mg ẑ has to be counteracted by the
magnetic force to reach an equilibrium. A diamagnetic material is characterized by
a negative magnetic susceptibility, .χ = −|χ |. The magnetization .M of a sample
is given in terms of .μ, Eq. 9.14. By solving for .μ, we obtain .μ = MV . In the
case of a weak external magnetic field, the magnetization is proportional to the
external magnetic field, .M −|χ |B/μ0 , where .|χ | 1 and .μ0 is the free space
permeability. This leads to a magnetic moment of the sample.
. dμ = −(|χ | / μ0 ) V dB (9.20)
The sample has no potential energy when it is on the ground, and there is no external
magnetic field. We have to do work against all the forces acting on the sample to
provide it with potential energy.
z B B
Ep =
. mg dz − dμ · B = mgz + (|χ | / μ0 ) V B dB
0 0 0
Ep = mgz + (|χ | / 2μ0 ) V [B(z)]2 (9.21)
The sample gathers energy by its position in both the gravitational field and the
magnetic field. The force exerted on the sample is defined by the potential energy,
.F = −∇Ep (r). The force has rotational symmetry around the z axis, and it must be
zero in the z direction to achieve the equilibrium of the sample.
|χ | ∂B(z)
mg = −
. V B(z) (9.22)
μ0 ∂z
Such a result has been applied to hover different materials, including small living
organisms such as a frog [5]. In order to levitate a human being, one requires a
magnetic field slightly smaller than 45 T (in Tesla units), with .1 × 109 Watts to cool
the system to operate.
A.1.4 Zeeman Effect
It is time to introduce the quantum version of the angular momentum in both the
magnetic moment and the potential energy.
√ This is done by assuming a total angular
momentum with magnitude .L = h̄ ( + 1), Eq. 9.2, and a z component of the
form .Lz = m h̄, Eq. 9.7. By using the expression .μz = −eLz /(2m), Eq. 9.13, then
the potential energy of Eq. 9.16 and the magnetic moment are transformed into:
. Ep = m μB B ; μz = −m μB ; μB = eh̄/(2m) (9.23)
where the Bohr magneton .μB , with units .eV /T , is defined. The main result indicates
that an external magnetic field splits the orbital energy into energy sub-levels of
magnitude .m μB B. For example, when the magnetic field is applied, the energy of
an electron in a p level, . = 1 and .m = 0, ±1, unfolds into three equally spaced
energy sub-levels.
eh̄ eh̄
EB=0 −
. B ; EB=0 ; EB=0 + B
2m 2m
The energy in the absence of the magnetic field, .EB=0 , is the reference energy.
The energy splitting may be increased or decreased in continuous manner since it
is directly proportional to the external magnetic field and it may be tuned up. In
short, the energy of an electron of an atom immersed in a magnetic field depends
on the quantum numbers n and . and also on the magnetic quantum number .m
and the strength B of the magnetic field. The splitting of the energy spectrum with
dependence on B is known as the Zeeman effect.
An atom or molecule is a multi-particle system with angular momentum charac-
terized by the total orbital angular momentum, the total spin, and the corresponding
eigenvalues.

L=
. Li ; S= Si
i i

h̄ ( + 1) ; h̄ s(s + 1)
Such vectors produce magnetic moments .μL and .μS , and precess in a magnetic
field.
. μL = −eL / (2m) ; μS = −g eS / (2m)
On the other side, the √ .J vector is identified with the total angular momentum. The
eigenvalues of .J are .h̄ j (j + 1), where j satisfies .| − s| ≤ j ≤ | + s|. The .Jz
component is quantized in the form .Jz = mj h̄, with .mj = 0, ±1, · · · , ±j . The
electron magnetic moment and magnetic energy are written in terms of the orbital
angular momentum contribution and the spin contribution.
. μ J = μL + μS ; Ep = e[L + 2S] · B / (2m)
The coefficient 2 takes the g factor of the electron spin into consideration. By
introducing the quantization of .J and .Jz , the overall magnetic moment and Zeeman
splitting energy are written in terms of the quantum numbers j and .mj .
√
. μJ = gμB j (j + 1) ; Ep = gmj μB B
The g factor depends on J , L, and S:
. g = 1 + [J (J + 1) + S(S + 1) − L(L + 1)] / [2J (J + 1)]

The g factor is usually computed using perturbation theory for the energy of an
atom in a weak magnetic field, adopting the general form of a second-rank tensor.
The g factor of an electron with orbital angular momentum and spin is known as the
Lande factor. On the other hand, the selection rules determine the possible electronic
transitions and frequency shifts among the new energy sub-levels of the atom or
molecule in a non-zero magnetic field.
Historically, the total magnetic moment of the electron was conceived without
the spin contribution. Now, it is known that the spin also contributes to the
total magnetic moment of the electron. In the past, the deviations between the
hypothesized results, without consideration of the spin, and the lab measurements
originated the name anomalous Zeeman effect. Nonetheless, the anomalous Zeeman
spectrum is presently realized as the Zeeman effect with the spin contribution.
There is a torque on the magnetic moment of an electron moving in orbital
motion around the nucleus and in the presence of a magnetic field. The torque
tends to align the magnetic moment with the external magnetic field, producing
a change of the angular momentum with time, similar to that of a spinning top
or gyroscope in a gravitational field. The frequency of precession is known as the
Larmor frequency. It is computed by considering the angle .θ between the angular
momentum .L and the z axis, this last axis taken as the direction of the external
magnetic field. The component of .L in the plane of precession is .L sin θ , Fig. 9.4.
In the time interval .[t, t + t] there is a directional change . L of the angular
momentum. The corresponding angular change in the plane of precession is . φ.
The torque is the change of angular momentum with time and has a perpendicular
direction to the angular momentum.
.τ = lim L / t = lim φ L sin θ / t (9.24)

t→0 t→0
Fig. 9.4 An external magnetic field exerts a torque .τ on the magnetic moment associated with the
orbital motion of an electron in an atom. The magnetic moment is anti-parallel to the orbital angular
momentum .L of the electron. Vector .L makes an angle .θ with the vertical axis, which corresponds
to the direction of the magnetic field. In a time interval .[t, t + t], the vector .L describes an arc
segment . L. The torque is computed from the expression .dL/dt = lim L/ t
t→0
The torque involves the magnetic moment, .τ = μ × B = μB sin(θ ) n̂, Eq. 9.10,
where .n̂ is the unit vector perpendicular to the plane that contains the vectors .μ and
.B. The magnitude of .μ is .eL/(2m), Eq. 9.13. By comparing the magnitude of .τ
with that of Eq. 9.24, and solving for .dφ/dt, we obtain:
ωL = dφ / dt = lim
. φ / t = eB / (2m) (9.25)
t→0
The frequency .ωL is the Larmor frequency of the magnetic moment in precession.
A.1.5 Electron Spin
The Stern-Gerlach experiment gives evidence of the spin of the electron and the
quantization of the spin by using silver atoms, where the outermost electron of
every silver atom is in a S state characterized by the quantum numbers . = 0 and
.m = 0. A couple of deflections in the atomic trajectories are observed, and two
projection states, .ms = ±1/2, are hypothesized for the electron spin. The existence
of the electron spin makes possible to have states doubly occupied, that is, we have
fermionic states characterized by the set of quantum numbers .(n, l, m ), which may
be occupied by one spin-up electron and one spin-down electron, in line with the
Pauli exclusion principle.
The electron spin is identified as a kind of intrinsic angular momentum of a tiny
charged ball, but with particular properties of quantum nature, which make impos-
sible to provide a clear classical description. For instance, the magnetic moment of
the electron involves the factor g, which is equal to 1.0 for the orbital motion of
the electron and .2.002, 319, 304, 92 for the spin of the electron.1 According to our
old suppositions, the electron is extremely small and demands an exceedingly high
angular velocity to compensate for the observed spin magnetic moment. However,
such a picture is in contradiction with the principles of relativistic mechanics,
as explained in Sect. 9.1.2 In spite of that, the intrinsic angular momentum of
the electron is modeled using an electric current in a circular loop, taking the
quantization rules of the spin into consideration to account for the values of the
magnetic moment, Fig. 9.5. This approach provides a “reasonable” quantitative
description of the electron spin, and for such a reason, it is also applied to the nuclear
spins with modifications of the g-factor magnitudes and the magnetons.
We have the factors .gs , .gp , and .gn for the spin of the electron, proton, and
neutron, respectively (the g factor of the electron in circular orbit is .gL = 1).
1 The value of the g factor is predicted by quantum electrodynamics. It has been verified in the
Lamb shift experiment with high accuracy. Feynman [6].
2 In principle, any incompatibility vanishes in the relativistic zitterwegegun model.
Fig. 9.5 Two spin projections of the electron are observed in the Stern-Gerlach experiment. Such
states are hypothesized to involve an intrinsic angular momentum of the electron, referred to as the
electron spin. The electron spin produces a dipolar magnetic moment. We use a classical approach,
in conjunction with the quantization rules of the spin, in an effort to explain the dipolar magnetic
moment
gs = 2.002, 319, 304, 92 2 ; gp = 5.585, 694, 713

. (9.26)
gn = −3.826, 083, 7
The magneton factors are .μB and .μn for the electron spin and nuclear spin.
μB = (eh̄) / (2me ) = 5.788,382 × 10−5 eV /T

. (9.27)
μn = (eh̄) / (2mp ) = 3.152,451 × 10−8 eV /T
The magnetons are the standard units to gauge the magnetic moments of the
microscopic particles. The proton mass and proton charge are used as the quantities
of reference in the magnetons of the nucleons. There are differences in the values of
the magnetons, .μB μn , due to differences in the masses of the electron, .me , and
proton, .mp . The magnetic moments .μL and .μs of the orbital motion and the spin of
the electron, Eqs. 9.13 and 9.3, and the magnetic moment .μp of the proton spin are,
in that order:

μL = −gL 2me e L ; μs = −gs 2me e S ; μp = gp 2me p I
.
μLz = −gL 2me e h̄m ; μSz = −gs 2me e h̄ms ; μIz = gp 2me p h̄mI
(9.28)
The proton spin is denoted with the symbol .I, and the corresponding z component
is .h̄mI , where .mI takes one of the .2I + 1 possible values of the nuclear spin.
The magnetic moment of the proton spin is smaller than the magnetic moment of
the electron spin. On the other hand, the magnetic moment of the proton spin is
.2.792,847, 350 μn , and that of the neutron spin is .−1.913, 041, 8 μn . The directions
of the electron spin and its magnetic moment are anti-parallel, and those of the
proton are parallel to each other. The proton spin precesses clockwise, and the
electron spin precesses counter-clockwise. The Zeeman splitting of the nuclear
energy levels is computed similarly to the splitting of the electronic energy levels
but using the corresponding nuclear variables.
A.1.6 Paschen-Back Effect
In real samples, particularly those containing transition metals, there are electric
fields and dipolar magnetic fields of the electrons and nuclei. The Zeeman energy
splitting depends not only on the external magnetic field imposed in the lab but
also on the magnetic moments of the nearby particles, which are capable of
modifying the dynamics of a particle due to the proximity. For example, consider
an electron in orbital motion around the nucleus. In the reference frame of the
electron, such an electron is subjected to a magnetic field produced by a charged
nucleus, which is in relative motion with respect to the electron. The magnetic
field interacts with the spin magnetic moment of the electron, generating a spin-
orbit (LS) coupling with dependence on the angular momentum of the electron. In
general, the interactions among the magnetic moments of different particles lead to
a cross-energy term .Een = amI ms , where a is known as the hyperfine coupling
constant. The cross-energy term is determined from the g factors of the electron
and nucleus and the distances and orientations between the interacting magnetic
moments. That interaction has analogy with the interaction of two or more electric
dipole moments.3 The Zeeman splitting of the energy level of an unpaired electron
in the field of the nuclei may be approximated by .E = gs ms μB (B + i ai mIi ),
where the summation is a correction to the imposed magnetic field B. Therefore, in
this approach, an effective magnetic field is considered in the energy expression.
The splitting of energy levels has been discussed in the Zeeman regime, that is,
where the external magnetic field maintains the LS coupling between the angular
momentum of the orbital motion and the angular momentum of the spin motion.
In the Zeeman regime, the splitting of the energy levels is small with respect to
the energy difference among unperturbed levels due to a relatively weak external
magnetic field. However, when the external magnetic field is strong, it modifies
the LS coupling to generate a strong individual coupling of L with the external
magnetic field and similarly for the spin S. As a result, we have the independent
precession of L and S and the intertwining of the energy sub-levels of the particles.
This response receives the name Paschen-Back effect. In such a domain, the energy
splitting depends not only on the atom type but also on the energy-level type.
Moreover, it may occur that for a given strength of the magnetic field, some energy
levels split according to the Zeeman effect, but others split in anomalous form
according to the Paschen-Back effect, all in the same atom. With an approximate
3 For the interaction of two magnetic dipoles of fermionic particles, refer to: Cohen-Tannoudji et
al. [7]. For the interaction of two linear charge distributions, refer to: Maitland et al. [8].
external magnetic field of .3 T , the Zeeman effect is observed in Na atoms, but a

Paschen-Back effect in Li atoms. The Paschen-Back effect is used in astronomy to
determine the magnetic fields of astronomical objects by measuring the radiation
emitted by the atoms subjected to strong magnetic fields.
A.1.7 Applications of the Spin Resonance Technique
In the experiments with electron spin resonance (ESR), electron paramagnetic

resonance (EP R), and nuclear magnetic resonance (N MR), the magnetic field is
tuned, and the adequate frequency to induce the flip of the spin states of electrons
and protons is chosen. The magnetic field applied in N MR spectroscopy may reach
up to .22 T (for reference, the geomagnetic field at our planet surface is between
25 and 65 .μ-Tesla). The spin relaxation emits an electromagnetic signal after the
device is switched off. The emitted signal depends on the external magnetic field.
Thus, under a magnetic field gradient which coils around the sample, the different
frequencies coming from the separate spatial areas of the sample are identified
and analyzed, resulting in well-resolved images of the sample. The spin resonance
techniques are specially effective on materials with radicals, where electrons with
unpaired spins are abundant. Such spectroscopic techniques are non-destructive. By
following a similar derivation of Eq. 9.17, and using the expressions of .μIz and .gp ,
Eqs. 9.28 and 9.26, the energy to flip the proton spin in a .1 T magnetic field is:
. Ep = 2μIz B = gp μn B = 1.7608 × 10−7 eV (9.29)
This energy is small with respect to the average room thermal energy of .0.026 eV
at .300 K. Therefore, the thermal energy randomizes the populations of the proton
spin states. In the case of an electron, the required angular frequency of the
electromagnetic wave to produce a spin transition is derived from the expression
.ωe = Ep /h̄ = 2μSz B/h̄, Eq. 9.18. In the case of a proton, the frequency for
the spin transition is .ωp = 2μIz B/h̄, Eq. 9.29. By assuming an external magnetic
field of .1 T , the respective frequencies, .ν = ω/(2π ), are .νe = 27.995 GH z and
.νp = 42.574 MH z.
On the other hand, the frequency to induce the spin transition of the neutron
is similar to that of the proton. Yet it is surprising that the neutron has a large
and negative magnetic moment since it is a neutral particle. The magnetic moment
of the neutron is indirectly determined by measuring the magnetic moment of the
deuteron, which is formed by a neutron and a proton with parallel spins and with
previous knowledge of the magnetic moment of the proton. The magnitude of the
magnetic moment of the proton and the magnitude and sign of the magnetic moment
of the neutron suggest that both the proton and the neutron have internal structure
and are not elementary particles. They are made of charged quarks, with magnetic
moments resulting in those of the proton and neutron. On the other hand, the
neutrino is hypothesized to have an extremely small magnetic moment, but contrary
to the neutron, the neutrino is an elementary particle with zero mass. The neutrino’s
Fig. 9.6 The Earth has a large magnetic field of dipolar type, created by the geodynamo of the
core particles. The far-field limit is similar to that produced by the electron spin; nevertheless, the
magnetic field strengths differ by many orders of magnitude
magnetic moment is believed to be associated with quantum fluctuations, which

make the neutrino show up as a mixture of two interacting particles, .W + and .e− ,
with the coupling of a magnetic field, like in the case of the electron self-energy
with involvement of a photon, Sect. 9.2. The muon (a lepton 207 times heavier than
the electron but with equal negative charge) is a particle that shows an anomalous
magnetic moment. The experimental measurements and the theoretical predictions
differ by a factor of 0.1%, a relatively large number. By refining the research on
the muon magnetic moment, the experiments may point to either the existence of
new types of particles or the necessity to propose a more exotic theory beyond the
Standard Model and based, for example, on the dark photon (a possible component
of dark matter in the universe). Such conclusions are expected to be derived in the
frame of quantum electrodynamics. Finally, at the other end of the spectrum, many
macroscopic objects are characterized by their magnetic moments. The planet Earth
is an example with an enormous magnetic moment of .∼8 × 1022 Coulomb m2 /s;
contrariwise, the Moon has no dipolar magnetic field, Fig. 9.6.
B.1 Appendix: Spin Functions and Operators
The spin is understood as a pure quantum-mechanical property with no classical

analogue. Therefore, it is important to discuss the notation and algebra that the spin
demands to be formally described. The information may be complemented with the
work of Cohen-Tannoudji et al. [9]. In the first stage, a linear Hermitian operator
designated with the letter .Ŝ is assigned to the spin. The operator .Ŝ classifies as an
angular momentum operator with similar quantization rules to the orbital angular-
momentum operator, .L̂. The relevant spin operators are .Ŝ2 and the components .Ŝi ,
where .i = x, y, z. The relation between these operators is given by .Ŝ2 = Ŝx2 + Ŝy2 +
Ŝz2 . They satisfy commutation relations with cyclic permutations.
[Ŝx , Ŝy ] = i h̄Ŝz ; [Ŝy , Ŝz ] = i h̄Ŝx ; [Ŝz , Ŝx ] = i h̄Ŝy

. (9.30)
[Ŝ2 , Ŝi ] = 0 ; i = x, y, z
B.1 Appendix: Spin Functions and Operators 201
The spin operators act on functions that belong to the space formed by the spin states
of the particle. Since the spin operator .Ŝ2 and .Ŝz commute, they have common
eigenfunctions which we distinguish with the symbols .α and .β. For example, by
using the eigenfunction .α, we have:
Ŝz α(s, ms ) = ms h̄ α(s, ms ) ; ms = −s, −s + 1, · · · , s − 1, s

. (9.31)
Ŝ2 α(s, ms ) = s(s + 1)h̄2 α(s, ms ) ; s = 0, 1/2, 1, 3/2, · · ·
In this particular case, the functions .α and .β are not eigenfunctions of the spin
operators .Ŝx and .Ŝy because these operators do not commute with .Ŝz . The arguments
s and .ms of the eigenfunctions may be considered the degrees of freedom of the spin
functions .α and .β. An elementary particle owns a unique value of s. The photon with
value .s = 1 breaks the rule of .ms stated in Eq. 9.31 as it only takes values .ms = ±1,
and no value .ms = 0. For the electron, we only have the spin .s = 1/2 and the
following eigenvalue expressions:
1 3 2
Ŝz α(s, ms ) = + h̄α(s, ms ) ;
. Ŝ2 α(s, ms ) = h̄ α(s, ms ) ; ms = +1/2
2 4
1 3 2
Ŝz β(s, ms ) = − h̄β(s, ms ) ; Ŝ2 β(s, ms ) = h̄ β(s, ms ) ; ms = −1/2
2 4
(9.32)
The nomenclature spin-up and spin-down is associated with values .ms = 1/2 and
.ms = −1/2 of the electron. The functions .α and .β form an orthonormal basis
set and generate the spin-state space of the electron with dimension 2. A general
expression of a spin function is .ψspin = cα α + cβ β where the coefficients may be
complex numbers. Before discussing additional features of the spin, it is convenient
to introduce a new notation to simplify the expressions. The expansion of a function
.ψ in the basis set w is displayed below for discrete and continuous functions.

. ψ(r) = i ci wi (r) ; ψ(r) = dα c(α) wα (r) (9.33)
Here, the continuous variable .α should not be confused with the .α spin function.
The expansions may be represented under the corresponding Dirac bra-ket vector
notation (the nomenclature follows the word bra-ket).

.|ψ = ci |wi ; |ψ = dα c(α) |wα
i

ψ| = ci∗ wi | ; ψ| = dα c∗ (α) wα | (9.34)
i
The closure relations displayed below involve the identity operator in terms of the
basis functions.

.Iˆ = |wi wi | ; Iˆ = dα |wα wα | (9.35)
i
The closure relation is important to project a bra or ket onto the basis functions.

|ψ = |wi wi |ψ = ci |wi
. i i (9.36)
|ψ = dα |wα wα |ψ = dα c(α) |wα
The space of functions that completely describe the particle state, including the spin,
is constructed from the tensor product of the space formed by the spatial functions
and the space formed by the spin functions. If a space-vector state is represented
by the ket vector .|r , and a spin-vector state is represented by the ket vector .|ε ,
corresponding to either the .α or .β spin function, then the vector that defines the
particle state is given by the tensor product .|r, ε = |r ⊗ |ε . The vector .|r, ε is a
ket vector, and the corresponding complex conjugate vector is the bra vector . r, ε|.
The spin operators commute with the spatial operators, and thus, they have
common eigenfunctions. The eigenvalue equations that involve such operators are:
X|r, ε = x|r, ε ; Y |r, ε = y|r, ε ; Z|r, ε = z|r, ε

.
3 2 h̄ (9.37)
Ŝ2 |r, ε = h̄ |r, ε ; Ŝz |r, ε = ± |r, ε
4 2
The kets .|r, ε form a complete basis sets. The orthonormal condition on the kets
is . r , ε |r, ε = δε ε δ(r − r), where the symbols .δ denote Dirac-delta functions.
The general expression of a state function is written as a linear combination of basis
functions.

.|ψ = |r, ε r, ε|ψ d 3 r (9.38)
ε
The components of .ψ in the spin states .ε = α and .ε = β are in general indicated by

ψε (r) = r, ε|ψ .
.
ψα (r) = r, α|ψ
. ; ψβ (r) = r, β|ψ (9.39)
Such functions define a two-dimensional spinor.

ψα (r)
. [ψ](r) = (9.40)
ψβ (r)
B.1 Appendix: Spin Functions and Operators 203
When a state vector is expressed as a tensor product, .|ψ = |ψ1 ⊗ |ψ2 , the spinor
becomes:

ψα (r) r|ψ1 α|ψ2 ψ1 (r) cα
. [ψ](r) = = = (9.41)
ψβ (r) r|ψ1 β|ψ2 ψ1 (r) cβ
where .cα and .cβ are the expansion coefficients of the spin function. The spin
operators under the spinor representation correspond to matrices with dimension
.2 × 2. Moreover, any matrix of such a dimension may be written as a linear
combination of the three Pauli spin matrices and the unit matrix. The application
of an operator .Â, with matrix representation .[Â], on the spinor .[ψ] takes the form:
. [ψ ](r) = [Â][ψ](r) (9.42)
On the other hand, the matrix element of the operator .Â in terms of the .[ψ] and .[φ]
spinors is given by:

. ψ|Â|φ = [ψ]† (r) [A] [φ](r) d 3 r (9.43)
where .[ψ]† is the adjoint vector of .[ψ].
Problem 9.2 By starting from a closure relation, like those of Eqs. 9.35, show
that the scalar product between the state vectors .|ψ and .|φ is given in terms
of their spinors.

. ψ|φ = [ψ]† (r) [φ](r) d 3 r
Obtain the normalization of the spinor .[ψ] when the kets are equal, .|φ = |ψ .
Solution The appropriate closure relation for the particle involves integration
over the space variable and summation over the possible spin states.

. d 3 r |rε rε| = 1
ε
We introduce the closure relation in the bra vector . ψ| without affecting such
a vector because the closure relation is equivalent to the identity operator.

. ψ| = d 3 r [ ψ|rα rα| + ψ|rβ rβ|]
(continued)
The scalar product between . ψ| and .|φ is:

ψ|φ = d 3 r [ ψ|rα rα|φ + ψ|rβ rβ|φ ]

= d 3 r ψα∗ (r)φα (r) + ψβ∗ (r)φβ (r)
.
∗ ∗ φα (r)
= d r (ψα , ψβ )
3
φβ (r)

= d 3 r [ψ]† (r) [φ](r)
The scalar product of the state vectors .|ψ and .|φ corresponds to the scalar
product of the spinors, which proofs the first requirement. When .|φ = |ψ ,
we have the normalization of the state vector .|ψ from the second expression
presented previously.

. ψ|ψ = d 3 r |ψα (r)|2 + |ψβ (r)|2 = 1
This satisfies the second requirement.
References
1. I.N. Levine, Quantum Chemistry, 4th edn. (Prentice Hall, New Jersey, 1991), Sect. 11.5, pp
297–309
2. J.D. Cresser, Quantum Physics Notes (Department of Physics, Macquarie University, Sidney,
2009), Chapt. 6
3. G. Cook, R.H. Dickerson, Understanding the chemical potential. Am. J. Phys. 63, 737–742
(1995)
4. R. Santamaria, J. Soullard, R.G. Barrera, The electron bubble and the H e60 fullerene: a first-
principles approach. J. Low Temp. Phys. (2019). https://doi.org/10.1007/s10909-018-02134-x
5. M.V. Berry, A.K. Geim, Of flying frogs and levitrons. Eur. J. Phys. 18, 307–313 (1997). M.D.
Simon, A.K. Geim, Diamagnetic levitation: flying frogs and floating magnets. J. Appl. Phys. 87,
6200–6204 (2000)
6. R. Feynman, QED: The Strange Theory of Light and Matter (Princeton University Press,
Princeton, 1985), pp. 7, 115 and 118
1977), pp. 1120–1129.
8. G.C. Maitland, M. Rigby, E.B. Smith, W.A. Wakeham, Intermolecular Forces: Their Origin and
Determination. International Series of Monographs on Chemistry 3 (Clarendon Press, Oxford,
1987), Sect. 1.4, pp. 10–16
1977), Chapt. 9, pp. 965–981
Postulates of Quantum Mechanics
10
The experiments at the beginning of the 20th century revealed unexpected results on
the behavior of the light and the particles at the atomic level. In those experiments a
flux of particles showed interference, diffraction, and dispersion patterns originally
attributed to waves, and, on the other hand, the light showed behavior of a matter
particle by colliding with other particles in different experiments. From a classical
perspective, a particle is an entity with mass, occupying a position in space, and the
light is a wave with spatial extension, no mass, and no fixed position. The particles
and waves represent opposite aspects, and thus, the waves are not expected to behave
as particles, and the particles are not expected to behave as waves. In other words,
the properties of particles and the properties of waves are irreconcilable from our
classical understanding. Still, the experimental observations on the particles and the
waves undeniably showed counterintuitive aspects because an entity, such as a wave
or particle, may behave in some instances as a wave, and in others as a particle, thus
regarding a dual character of the entity, which is in conflict with the expectations
based on classical mechanics.
The diverse aspects of the particles and the waves are nowadays considered
under a new philosophy, which carries the name quantum mechanics. The theory of
quantum mechanics introduces unorthodox ideas; however, it provides the proper
explanations of the microscopic phenomena. It is a theory that we do not fully
understand, with unconventional interpretations. The best way to approach quantum
mechanics is to create analogies with the behaviors of the particles and the waves
in the classical world, but with the appropriate caution since a microscopic entity,
such as a wave or a particle, may behave as a wave under given circumstances
and as a particle under different circumstances. For macroscopic particles, quantum
mechanics assertively reproduces the laws of classical mechanics. In the following
sections, we provide the postulates that give foundations to quantum mechanics;
they engulf the key results of our observations on the microscopic world and
condense the fundamentals of quantum mechanics in a formal mathematical way.
206 10 Postulates of Quantum Mechanics
In addition to the postulates, we discuss the principles that emerge from a new
and intertwined characterization of the microscopic particles and waves. We pay
attention to the wave equation and wave function, the variation principle of the
energy, and other rules that give frame to quantum matter.
10.1 Reformulating the Conceptual World
A new philosophy at looking nature emerged from the experiments on microscopic

bodies. The new philosophy included the particle-wave duality of matter, the
energy quantization, and the consideration of a new degree of freedom, the spin.
It was given the name wave mechanics, also recognized as quantum mechanics.
Quantum mechanics constitutes a general scientific philosophy that is not possible to
rationalize, as it is beyond our habitual comprehension. Yet, it was to revolutionize
the way to describe nature since we may deal with a particle as a wave or with a
wave as a particle. The theory of quantum mechanics combines aspects which in
principle were considered as opposite aspects to us into a single description, and
after a crude reality that remains unclear to humans. Nevertheless, it is capable to
provide the basic features of nature.
Quantum mechanics is not in contradiction with the theory of classical mechan-
ics, but it generalizes it by introducing: (i) the particle-wave character of the
microscopic entities, (ii) the quantization of the energy, and (iii) an intrinsic spin.
Quantum mechanics unifies the theories of particles and radiation (this last one
embodies electricity, magnetism, and optics) to provide a unique conception of
nature. By using quantum mechanics, there are no more differences between what
is theorized and observed in the experiments, since the wave-particle duality,
energy quantization, and spin constitute the foundations of the modern theory.
Wave mechanics has been under test in many different forms, and it has been
successfully confirmed in all of them. In this regard, quantum mechanics is a general
and fundamental theory. Still, some limitations persist. According to relativistic
mechanics, the relativistic effects are important when the speeds of the particles are
close to the speed of light. Relativistic mechanics also considers matter as a form
of energy, and vice versa. However, it has not been possible to properly include
relativistic mechanics in quantum mechanics, and therefore, there is not a general
theory (with general consensus) unifying both fields up to this time.
In the following sections, the main postulates of quantum mechanics are
presented. They establish the physical and mathematical foundations of the theory.
The comments after each postulate provide the details on the formulation of the
postulate. The consideration of the postulates in the simulations makes possible to
unfold the dynamics of particles and suitably explain their behaviors at the atomic
and subatomic levels.
10.2 Postulates of Quantum Mechanics 207
10.2 Postulates of Quantum Mechanics
Classical mechanics is a branch of physics formulated according to our everyday

experience, mostly based on observations at the macroscopic level. Nevertheless, the
experiments performed at the microscopic scale at the beginning of the 20th century
uncovered aspects of particles and waves not observed before. The experiments
demonstrated the quantization of the energy, with microscopic matter absorbing and
emitting radiation in discrete bundles of energy, and revealed the existence of the
spin, a kind of intrinsic angular momentum of the particles. The predictions based on
classical mechanics were found in contradiction with the experimental observations
in the atomic domain. In this regard, the classical understandings of a particle and a
wave, incompatible with each other and mutually exclusive at macroscopic scales,
had to be reconciled into a new scientific theory to describe a particle and a wave as
a single entity at the microscopic realm.
The postulates of quantum mechanics are provided in the sections below to
establish the formal basis of quantum mechanics. We follow the common practice to
state them concisely, with details provided after the formulation of each postulate.1
10.2.1 First Postulate
The particles show a wave character in the microscopic world leading to probabilis-
tic predictions. In this regard, the quantum state of a particle system is expected to be
described by a wave function. The first postulate establishes a relation between the
wave function and the quantum-mechanical state of the system by associating the
wave function to a vector which belongs to the space of state vectors of the system.
When the state vector .|ψ(t) is considered, then we can deal with a representation
of it by employing a wave function .ψ(t) of the system.
Statement 10.1 There exists a function .ψ(t) with name wave function,
which describes the quantum-mechanical state of the system at the instant
of time t. The wave function is a state vector .|ψ(t) which belongs to the
space of state vectors associated with the system.
The wave function depends on the coordinates and spins of the particles and is a
function of the time, t.
. ψ(r1 , σ1 , r2 , σ2 , · · · , rn , σn ; t) (10.1)
1 More information can be found in: Cohen-Tannoudji et al. [1], Chapt. III.
The variables .ri and .σi represent the spatial coordinates and spin of the ith particle.2
The wave function is a state function. It is mathematically well behaved, this is, it
is univaluated, continuous, with continuous derivatives, and quadratically integrable
at any instant of time and satisfies the superposition principle.

ψ ∗ ({ri , σi ; t}) ψ({ri , σi ; t}) dv = 1
. (10.2)
dv = d 3 r1 dσ1 d 3 r2 dσ2 · · · d 3 rn dσn
The expression .ψ ∗ is the complex conjugate of .ψ (the spin variables .σi are discrete,
and summations should be additionally used in Eq. 10.2 for the spins; nevertheless,
the integrals over the different degrees of freedom should be understood to also
include the summations, for simplicity). The product .ψ ∗ ψ dv has a probabilistic
interpretation. If we consider a single particle, the probability of finding it in the
space region limited by the vectors .(x, y, z) and .(x + dx, y + dy, z + dz) at the
instant of time t is given by .ψ ∗ (x, y, z; t) ψ(x, y, z; t) dxdydz. The quadratically
integrable requirement ensures a finite numerical value of the spatial probability
function at any instant of time. The set of all wave functions .{ψ1 , ψ2 , · · · }, referent
to the different states of the system, forms a complex vector space which satisfies
the algebra rules of a Hilbert space. In contraposition to a particle which is described
in deterministic form in classical mechanics, the description of a particle by a wave
function, which may now be written in terms of the basis functions .{ψ1 , ψ2 , · · · },
acquires a probabilistic nature in quantum mechanics.
There are cases in nature where it is impossible to apply the first postulate due
to the extreme physical conditions, for example, in the very first moments of the
Universe creation, when a strong and short-lived singularity was responsible for the
birth of the Universe. The spacetime singularities, also produced by black holes,
impede the use of quantum mechanics due to the lack of a smooth wave function,
free of discontinuities in the derivatives. Also, it is neither possible to describe
phenomena smaller than the Planck scale in quantum mechanics, as the Planck
constant h imposes the smallest quantization scale. Fortunately, there is a myriad
of other physical cases where quantum mechanics may be applied.
10.2.2 Second Postulate
We perform measurements in the laboratory to extract information of the particle

system. In this respect, we require a mathematical tool in quantum mechanics to act
upon the wave function to reproduce the results measured in the lab.
2 The electronic spin is introduced in ad hoc manner in the wave function of the electrons of a
molecule, Sect. 10.9, like in the Hartree-Fock-based methods or the Density Functional Theory,
and after considering Pauli’s exclusion principle, Sect. 10.8. The incorporation of the electronic
spin leads to unique solutions of the quantum-mechanical eigenvalue equations. Unfortunately, the
origins of the spin remain as a conundrum.
Statement 10.2 There is an associated operator, .Â, to every physical observ-

able, .A, which acts upon the vectors that belong to the state-vector space of
the system. It must be Hermitian to render real eigenvalues.
There are two important aspects of the quantum operators that should be considered.
The first aspect includes the way to build the operators, and the second one involves
the mathematical requirements on the operators. The purpose of such aspects is to
guarantee both, the correct mathematical behavior of the operators and the proper
access to the physical information.
The first aspect demands operators that represent the physical observables. To
do this, there are rules to associate an operator .Â of quantum mechanics to every
physical observable .A. The rules consist of replacing the classical variables .r and .p
that describe the physical observable .A, by the respective operators .R̂ and .P̂.
. A(r, p; t) ⇒ Â(R̂, P̂; t) (10.3)
In most situations, we demand a symmetric expression in terms of the operators .R̂

and .P̂. The symmetrization is a common requirement in the proposal of operators
corresponding to those physical observables where the product of the fundamental
variables .r and .p is present. However, care should be exercised, as the dynamical
variable .p is understood as a conjugate momentum and should not be confused
with other types of momenta. For example, the conjugate momentum of a particle
in a vector potential .C is .p = mv + qC, while the mechanical momentum is
.mv. The operator .Â should additionally ensure a faithful representation of the
physical observable .A at the quantum-mechanical level by being mathematically

well behaved. The operators are expected to be linear, a property that is linked
to the superposition principle. They should avoid the breaking of the canonical
commutation relations of quantum mechanics among the Cartesian components of
.R̂ and .P̂.
. [Ri , Rj ] = [Pi , Pj ] = 0 ; [Ri , Pj ] = i h̄ δij (10.4)
The quantum-mechanical operators are Hermitian or self-adjoint to: (i) render

real eigenvalues, in agreement with the results obtained from the experimental
measurements, and (ii) generate vectors that form a basis set in the corresponding
Hilbert space. The Hermitian property renders the possibility to apply the Hermitian
operator, .Â, in any of the two functions .ψ1 or .ψ2 .

. ψ1∗ Âψ2 dv = ψ2 (Âψ1 )∗ dv (10.5)
The generation of basis vectors with Hermitian operators is important to describe

any state of the system as a linear combination of the basis vectors in the domains
that the physical observable .A demands. There are more required properties on the
operators. Still, in concise manner, when the operators are correctly proposed and
they are in line with all the other quantum mechanical postulates, they form a key
part of the mathematical tool box of quantum mechanics to obtain the information
of the physical observables of the system.
10.2.3 Third Postulate
In the laboratory, the appropriate measuring device is applied to the system under
investigation to obtain the required information. By following an analogy, the third
postulate involves the appropriate operator .Â to extract the information of the
system.
Statement 10.3 The value of a measurement of the observable .A corre-

sponds to an eigenvalue of the operator .Â associated with the observable .A.
The laboratory measurements of the physical quantity .A are definite and correspond
to the eigenvalues of .Â. The operator .Â is Hermitian to guarantee that the set of
eigenvalues of operator .Â are real numbers. A discrete spectrum .{a1 , a2 , a3 , · · · }
deduced from the operator .Â indicates quantization of the results measured on the
system referent to the physical observable .A.3
10.2.4 Fourth Postulate
When the spectrum of the operator .Â is discrete, the eigenvectors .ψi and the
eigenvalues .ai carry a discrete subindex, i, with .ψi and .ai satisfying the eigenvalue
equation of .Â.
. Âψi = ai ψi (10.6)
The eigenvectors are orthonormal with respect to each other and expand the Hilbert
space of the system under investigation. The theory of quantum mechanics is
structured in such a way that the quantization of the energy, the angular momentum,
and the spin is obtained after solving the corresponding eigenvalue equations, giving
solutions with discrete values. There are also operators such as .R̂ and .P̂, representing
3 The discussion provided after every postulate not only complements the postulate but also
provides clarity.
the physical observables .r and .p, that unfold a continuous spectrum. Since the
eigenvectors .ψi generate the Hilbert space, an arbitrary function .ψ of such a space
is described by a linear combination of eigenfunctions, which all together form a
complete orthonormal basis set.

. ψ= i ci ψi (10.7)
The coefficients .ci indicate the weights of the respective state functions .ψi in the
production of .ψ. The weights .ci may also have time dependence. In this context, the
coefficients .ci with large magnitudes infuse a large character of the corresponding
state functions .ψi to the wave function .ψ. The projection of .ψ onto .ψi gives the
factor .ci due to the orthogonality of the eigenvectors.

. ci = ψi |ψ = ψi∗ ψ dv (10.8)
The integration is over all the degrees of freedom, and the Dirac bra-ket notation is
introduced, Sect. B.1 of Chap. 9.
Problem 10.1 Show that the set of functions .{ψi } constitutes a complete
basis set. Every .ψi is normalized and is solution of Eq. 10.6.
Solution In order to demonstrate that .{ψi } is a complete basis set, we show

that it is a set of orthonormal functions. The functions .ψj and .ψk are solutions
of expression 10.6. In this case, the equation of .ψj is multiplied by .ψk∗ and
the complex conjugate equation of .ψk by .ψj :

ψk∗ Âψj − ψk∗ Â∗ ψj = (aj − ak )ψk∗ ψj
. ⇒ 0 = (aj − ak ) ψk∗ ψj dv
V
The left-hand side is zero upon integration because .Â∗ = Â. Since the left-
hand side is 0, and .aj = ak , then the integral is zero. In this respect, .ψj and
.ψk are orthonormal, and the set of functions .{ψi } constitutes a complete basis
set.

. ψk∗ ψj dv = δj k
V
The spectral decomposition may also include degenerate functions, or continuous

functions, or their combination. In the case of degenerate functions, several
(i)
eigenvectors .ψn reproduce the same eigenvalue.
Âψn(i) = an ψn(i)
. ; i = 1, 2, · · · , gn (10.9)
The degeneracy degree is .gn . The description of .ψ in terms of the eigenvectors also
includes all the degenerate states.

gn
ψ=
. cn(i) ψn(i) ; cn(i) = ψn(i) |ψ (10.10)
n i
(i)
The eigenvectors .ψn constitute an orthonormal set of vectors and reproduce the
vector space associated with the eigenvalue .an . In the continuous case, the expansion
of .ψ in terms of the eigenvectors .ψα , which form a continuous basis set, demands
the replacement of the summation of Eq. 10.7 by an integral over the continuous
variable .α.

.ψ = c(α) ψα dα (10.11)
Based on the mathematical formalism that has been introduced with the eigenvalue
equations, we are now in the position to make contact with the probabilistic
predictions of quantum mechanics. To do this, we consider the probability of finding
the system in a certain state, say .ψn , with eigenvalue .an of the operator .Â, in the
following postulate.
Statement 10.4 If the system is described by the normalized state vector .|ψ,
and the corresponding function .ψn is a normalized eigenvector of the operator
.Â (associated with the physical observable .A) with eigenvalue .an , then the
probability .P(an ) of obtaining the value .an in the discrete case, when the
physical observable .A is measured, is:
. P(an ) = |ψn |ψ|2 (10.12)
In the degenerate case with degeneracy degree .gn , the probability of

(i)
obtaining the value .an involves all the eigenfunctions .ψn with eigenvalue
.an .

gn
P(an ) =
. |ψn(i) |ψ|2 (10.13)
i=1
In the continuous case, the probability of obtaining the value .α involves

the eigenfunction .ψα , whose eigenvalue is .α in the extension .dα.
dP(α) = |ψα |ψ|2 dα

. (10.14)
ψ is normalized to deal with probabilities equal or lower than 1.

.
10.2.5 Fifth Postulate
The following postulate deals with the state vector that corresponds to the system
after carrying a measurement of a physical observable .A. In order to establish the
postulate, it is convenient to introduce the projection of a ket vector .|ψ onto the
subspace associated with the eigenvalue .ai of the system. We start by considering
(j )
the function .ψi as a linear combination of the degenerate functions .{ψi }, with
(j )
expansion factors .ci .

gn
(j ) (j ) (j ) (j )
ψi =
. ci ψi ; ci = ψi |ψ (10.15)
j
The projection operator .Pi onto the subspace associated with the eigenvalue .ai is
defined in terms of the degenerate functions using the bra-ket notation as:

gn
(j ) (j )
Pi =
. |ψi ψi | (10.16)
j
By applying the projector to .|ψ, the individual ket .|ψi is recovered.

gn
(j ) (j )

gn
(j ) (j )
Pi |ψ =
. |ψi ψi |ψ = ci |ψi = |ψi (10.17)
j j
The ket vector .|ψi is the projection of .|ψ onto the subspace associated with the
eigenvalue .ai . Next, suppose that the system is described by the wave function
.ψ(t0 ) at the instant .t0 . It evolves with time and is described by .ψ(t). Eventually,
the measurement of the physical observable .A is performed on the system. The

result of the measurement is .ai , an eigenvalue of operator .Â, and the system is in
state .ψi immediately after the measurement. Such a state also evolves with time
after the measurement.
Statement 10.5 When the system is described by the state function .ψ,
and the value obtained by measuring the observable .A on the system is
.ai , then the state function that describes the system immediately after the
measurement is given by the projection of .|ψ onto the subspace associated

with the eigenvalue .ai , namely, .Pi |ψ / ψ|Pi |ψ1/2 , where .ψ|Pi |ψ1/2 is
the normalization constant.
10.2.6 Sixth Postulate
The microscopic particles show wave character, and thus, we are required to propose
an equation that governs such a behavior. A dynamic equation is proposed by
assuming a wave function .ψ with dependence on both the spatial variables .{ri }
and the time t. The change of .ψ with time replaces the trajectory concept of a
classical particle. By following a relative analogy to the expression that describes
the propagation of electromagnetic waves in vacuum, an equation for the particles
is postulated.
Statement 10.6 The time evolution of the state function .ψ is ruled by the
wave equation:
∂ψ({ri }, t)
i h̄
. = Ĥ(t)ψ({ri }, t) (10.18)
∂t
where .Ĥ is the operator corresponding to the Hamiltonian of the system.
The wave equation is a central expression in quantum mechanics since it rules the
behavior of .ψ in space and time, avoiding in that aspect any indeterminacy of .ψ. It
receives the name time-dependent wave equation or Schrödinger equation in honor
of Erwin Schrödinger, who proposed it. Section A.1 derives the wave equation 10.18
for matter waves. The Hamiltonian operator introduces the characteristics of the
system in the Schrödinger equation, like the number of nuclei and electrons, their
masses and charges, the position variables, the kinetic energy operator, .T̂ =
−h̄2 ∇ 2 /(2m), and the interaction types among the particles through the potential
energy operator, .V̂ . For a single particle with mass m, the time-dependent wave
equation is:
∂ψ h̄2 2
i h̄
. = Ĥψ ; Ĥ = − ∇ + V̂ (10.19)
∂t 2m
The Schrödinger equation is linear and homogeneous and accounts for the spatial
and temporal changes of .ψ. When the wave equation is solved, the state function
not only acquires information of the system from the Hamiltonian but also dictates
the temporal evolution of .ψ, which should be compatible with the term .Ĥψ.
This process is similar to that of classical mechanics, where the solutions of the
Hamiltonian equations of motion acquire the information of the system from the
Hamiltonian function. Yet, the wave equation is difficult to solve due to the complex
interactions in the microscopic world. The exact solutions are found in the cases
of model systems and the simplest atomic and molecular systems with a few
10.3 Time Reversibility in Quantum Mechanics 215
degrees of freedom. In the cases of more complex or realistic systems, usually

with a large number of particles, we require to introduce approximations, and the
solutions are frequently achieved in numeric form. The current methodologies that
solve in approximate form the wave equation are classified according to both the
simplifications in their formulations and their accuracy levels.
The most accurate approaches are in principle applicable to all kinds of particle
systems, but practical limitations in the speed and storage of the computational
devices make those approaches only applicable to small systems. By introducing
further approximations, less elaborate schemes are created. These are used to
investigate large complexes; however, some accuracy is lost when an accurate
methodology is simplified. There is a trade-off in such a simplification process
between the system size and the computational efficiency. In this regard, based
on a balance between accuracy and computational efficiency, one may decide the
approach to employ for a system with a given size. When the wave equation
is solved, the state function .ψ is determined, and the molecular system may be
characterized by using .ψ according to the quantum-mechanical procedures.
The purpose of this work is to describe the mainstream quantum-mechanical
equations at different levels of theory, which have practical applications in molecular
systems. Such methods are widely used to study the dynamics of atoms, molecules,
clusters, and extended matter like the crystals, fluids, and gases, in the present
days. We provide their mathematical foundations, paying special attention to the
formulation of the suppositions, because these not only define the accuracy level and
limitations of the approaches but also point out pathways for future improvements.
10.3 Time Reversibility in Quantum Mechanics
The goal of this section is to demonstrate the time reversibility in quantum

mechanics. We follow similar steps to those given by Tolman [2]. Let us deal
with an isolated conservative system of particles. We assume two versions of
the same system: (i) the non-primed system, where the time evolution is in the
forward direction, and (ii) the primed system, where the time evolution is in the
backward direction. We are required to demonstrate that, by reversing the signs
of the momentum and time, we obtain the same predictions from the equations of
motion, probability densities, and expectation values of the physical observables, in
the two systems.
The probability of finding the particle’s position r in the interval dr at the instant
of time t is .P (r, t) dr, while the probability of finding the particle’s momentum p
in the interval dp at the instant of time t is .P (p, t) dp. The expectation value of a
physical variable .A is represented by .Â(r, p, t), where the operator .Â depends on
the conjugate variables r and p. In principle, by reversing the momenta and times
in the terms .P (r, t), .P (p, t), and .Â(r, p, t) of the non-primed system, we should
obtain the respective primed terms, .P (r, t), .P (p, t), and .Â (r, p, t).
P (r, t) = P (r, −t) ; P (p, t) = P (−p, −t)

. (10.20)
Â (r, p, t) = Â(r, −p, −t)
The quantum Hamiltonian .Ĥ is obtained from the classical Hamiltonian .H of

the system by replacing the conjugate momentum p by the momentum operator
.−i h̄∂/∂r. The classical Hamiltonian has quadratic dependence on p; therefore,
the quantum Hamiltonian is real and remains without changes when the complex
conjugate of .Ĥ is considered. On the other hand, the wave function .ψ depends on
the conjugate coordinate r and time t and satisfies the wave equation 10.18, shown
below for easy access.
∂ψ(r, t)
i h̄
. = Ĥ(t)ψ(r, t) (10.21)
∂t
The complex conjugate of the wave equation is obtained by replacing .ψ with .ψ ∗

and changing signs of the imaginary number i.
∂ψ ∗ (r, t)
. − i h̄ = Ĥ(t)ψ ∗ (r, t) (10.22)
∂t
The functions .ψ and .ψ ∗ deal with the time in the forward direction. We work
with the primed system and its corresponding wave functions .ψ (r, t) and .ψ ∗ (r, t).
These functions satisfy the following wave equations.
∂ψ (r, t) ∂ψ ∗ (r, t)
i h̄
. = Ĥ(t)ψ (r, t) ; −i h̄ = Ĥ(t)ψ ∗ (r, t) (10.23)
∂t ∂t
A sign change in the times establishes the relation between primed and non-primed
functions.
ψ (r, t) = ψ ∗ (r, −t)

. ; ψ ∗ (r, t) = ψ(r, −t) (10.24)
The insertion of these functions in the wave equations 10.23 results in modified
expressions for the primed functions.
∂ψ ∗ (r, −t) ∂ψ(r, −t)

i h̄ = Ĥ(t)ψ ∗ (r, −t) ; −i h̄ = Ĥ(t)ψ(r, −t)
∂t ∂t
∂ψ ∗ (r, −t)
.
∂ψ(r, −t)
−i h̄ = Ĥ(t)ψ ∗ (r, −t) ; i h̄ = Ĥ(t)ψ(r, −t)
∂(−t) ∂(−t)
The first expression may be compared with Eq. 10.22, and the second expression
with Eq. 10.21. For the time-independent Hamiltonian, the set of equations become
the same when we replace t by .−t and vice versa. In reference to the probability
∗
.P (r, t) = ψ (r, t)ψ(r, t), it is possible to replace t by .−t, to obtain .P (r, −t) =
10.3 Time Reversibility in Quantum Mechanics 217
ψ ∗ (r, −t)ψ(r, −t). By using Eqs. 10.24, one of the Eqs. 10.20 is confirmed.
. P (r, t) = ψ (r, t)ψ ∗ (r, t) = ψ ∗ (r, −t)ψ(r, −t) = P (r, −t)
The spatial probability .P (r, t) is equivalent to the spatial probability .P (r, −t),
which has a reversed time with respect to the probability .P (r, t). A similar situation
is observed when dealing with the momentum functions.
. φ (p, t) = φ ∗ (−p, −t) ; φ ∗ (p, t) = φ(−p, −t)
In this case, we have an equivalence between momentum probabilities.
. P (p, t) = φ (p, t)φ ∗ (p, t) = φ ∗ (−p, −t)φ(−p, −t) = P (−p, −t)
The momentum probability .P (−p, −t) has arguments with reversed signs with
respect to those of the probability .P (p, t). This result confirms another identity
of Eqs. 10.20.
Problem 10.2 The function .ψ(r, t) is the spatial representation of the wave
function, while .φ(p, t) is the momentum representation of the wave function.
These functions are related by means of Fourier transforms.

−n/2
ψ(r, t) = h φ(p, t) exp[i/h̄ (p1 r1 + · · · + pn rn )] dp
. (10.25)
φ(p, t) = h−n/2 ψ(r, t) exp[−i/h̄ (p1 r1 + · · · + pn rn )] dr
Show that .φ (p, t) = φ ∗ (−p, −t) and .φ ∗ (p, t) = φ(−p, −t).
Solution Function .φ (p, t) is related to function .ψ (r, t) by means of a

Fourier transform.

−n/2
. φ (p, t) = h ψ (r, t) exp[−i/h̄ (p1 r1 + · · · + pn rn )] dr
We use Eqs. 10.24 to substitute .ψ (r, t) by .ψ ∗ (r, −t). In this connection, we

have:

−n/2
. φ (p, t) = h ψ ∗ (r, −t) exp[−i/h̄ (p1 r1 + · · · + pn rn )] dr
On the other hand, we also know the expression of .φ ∗ (−p, −t).
(continued)

∗ −n/2
. φ (−p, −t) = h ψ ∗ (r, −t) exp[i/h̄ (−p1 r1 − · · · − pn rn )] dr
The conclusion to this part is .φ (p, t) = φ ∗ (−p, −t). Similar steps are taken
to show .φ ∗ (p, t) = φ(−p, −t).

φ ∗ (p, t) = h−n/2 ψ ∗ (r, t) exp[i/h̄ (p1 r1 + · · · + pn rn ] dr
.
= h−n/2 ψ(r, −t) exp[i/h̄ (p1 r1 + · · · + pn rn ] dr
On the other hand, we have:

−n/2
. φ(−p, −t) = h ψ(r, −t) exp[−i/h̄ (−p1 r1 − · · · − pn rn )] dr
Thus, we conclude .φ ∗ (p, t) = φ(−p, −t).
The last expression to demonstrate involves the equivalence between expectation

values of physical observables .A in the time forward and time backward directions.
. Â (r, p, t) = ψ (r, t)| Â(r, p, t) |ψ (r, t)
We employ Eqs. 10.24, .ψ (r, t) = ψ ∗ (r, −t) and .ψ ∗ (r, t) = ψ(r, −t) in the
expression of .Â .
∂
. Â (r, p, t) = ψ(r, −t)| Â(r, −i h̄ , t) |ψ ∗ (r, −t)
∂r
We may apply the complex conjugate property to the product .(Â ψ).
∗
∂
. Â (r, p, t) = ψ(r, −t)| Â(r, −i h̄ , t) ψ(r, −t)
∂(−r)
By applying the Hermitian property, Eq. 10.5, of the operators, we have:
∂
Â (r, p, t) = ψ(r, −t)| Â(r, −i h̄ , t) |ψ(r, −t)
. ∂(−r)
= Â(r, −p, −t)
In this respect, we have the required equivalence between expectation values in the
forward and backward time directions in quantum mechanics.
10.4 Stationary States 219
10.4 Stationary States
We are interested in obtaining a solution of the time-dependent wave equation from

the eigenvectors of the Hamiltonian, .Ĥ, which is supposed to be time independent.
We follow the work of Cohen et al. [3]. The eigenvectors .ψij and eigenvalue .Ei of
.Ĥ are time independent. The additional index j of the state function .ψij represents
in generic form the other numbers that are required to quantum mechanically define
the system. The eigenvectors and eigenvalues satisfy the stationary expression:
. Ĥψij = Ei ψij (10.26)
The equation is known as the time-independent wave equation. The set .{ψij } of
solutions forms a complete orthonormal basis set, and any vector of the Hilbert
space associated with the system can be expressed as a linear combination of these
eigenvectors. Let us consider the vector .ψ(t0 ) which describes the system at the
instant of time .t0 , omitting the writing of spatial variables for simplicity, but writing
.ψ(t0 ) in terms of the eigenvectors.

.
ψ(t0 ) = cij ψij (10.27)
i,j
ψ . The coefficients
At a later instant of time t, we similarly have .ψ(t) = i,j cij ij
.c
of .ψ(t0 ) are different from the coefficients .c
of .ψ(t). Instead of using primes,
ij ij
it is convenient to introduce
time-dependent coefficients. Equation 10.27 is written
in the form .ψ(t 0 ) = i,j cij (t0 ) ψij , while the wave function at the instant of time
t is .ψ(t) = i,j cij (t) ψij . In order to establish the time evolution of .ψ from .t0
to t, we now look for a relation between the factors .cij (t0 ) and .cij (t), because the
functions .ψij of the expressions of .ψ(t0 ) and .ψ(t) are the same over time. To do this,
we recallthat .ψ satisfies the time-dependent wave equation. Hence, the expression
.ψ(t) = i,j cij (t) ψij is introduced into the time-dependent wave equation 10.18,
reaching a differential equation for the coefficients, which is similar to that of .ψ(t).
. i h̄ ∂cij (t) / ∂t = Ei cij (t) (10.28)
By considering the initial condition at the instant of time .t0 , the solution of the above
expression is .ckj (t) = ckj (t0 ) exp[−iEk (t −t0 )]/h̄. The wave function at any instant
of time t is written in terms of the parameters computed at the initial instant of time
.t0 .

.
ψ(t) = ckj (t0 ) e−iEk (t−t0 )/h̄ ψkj (10.29)
kj
This expression gives the time evolution of .ψ from .t0 to t. The functions .ψ(t0 ) and
ψ(t) are different due to the participation of different phase factors shown in the
.
exponential. On the other hand, the time evolution of a physical variable, .A(t), is
expressed in terms of the operator .Â and the state vectors .ψkj .

.
ψ(t)|Â|ψ(t) = e−i(Ek −Ek )(t−t0 )/h̄ ck∗ j (t0 )ckj (t0 ) ψk j |Â|ψkj
kj k j
(10.30)
Let us suppose the special case where .ψ(t0 ) is a linear combination of the state
vectors .ψkj , with a fixed value of k.

.
ψ(t0 ) = ckj (t0 ) ψkj (10.31)
j
The function .ψ(t0 ) is an eigenvector of .Ĥ with eigenvalue .Ek . By multiplying the
coefficients .ckj (t0 ) by the phase factor .exp[−iEk (t − t0 )/h̄], the wave function at a
later time is:

.
ckj (t0 ) e−iEk (t−t0 )/h̄ ψkj = ψ(t) ⇒ ψ(t) = e−iEk (t−t0 )/h̄ ψ(t0 )
j
(10.32)
The functions .ψ(t0 ) and .ψ(t) are physically equivalent because they differ by a
single phase factor. The energy of the system at the time instants .t0 and t is .Ek ,
indicating the energy conservation. In this regard, the wave function .ψ describes a
stationary state.
The following situation considers another facet of a stationary state function. It
involves the case of a wave function expressed as the product of two functions; the
first function depends on the spatial variables and the second one on the time.
. ψ(r, t) = φ(r) χ (t) (10.33)
In order to determine .ψ, we need to know .φ and .χ . The equation that each
function satisfies is proposed by inserting .ψ in the time-dependent wave equation
and expressing .Ĥ as the addition of a kinetic energy operator, .−h̄2 ∇ 2 /(2m), and a
potential energy term, .V (r).
i h̄ ∂χ (t) h̄2 ∇ 2 φ(r)

. =− + V (r) (10.34)
χ (t) ∂t 2m φ(r)
The expression on the left depends on t and is independent of the expression on

the right, which depends on .r. In this situation, the two expressions are equal
to a constant, which we write as .h̄ω, where .ω is a constant angular frequency.
The solution of the equation holding the time is .χ = A exp[−i(h̄ω)t/h̄], and the
function .φ satisfies the time-independent equation:
10.5 Superposition Principle of Quantum States 221
h̄2 2
. − ∇ φ + V (r)φ = Eω φ ; Eω = h̄ω (10.35)
2m
This expression is the eigenvalue equation of the Hamiltonian, .Ĥφ = Eω φ, with

quantized energy. The function .ψ is written in terms of the amplitude A, function
.φ, and energy .Eω .
. ψ = A φ(r) e−iEω t/h̄ (10.36)
Next, we are interested in computing the particle density, .ρ(r), from .ψ with the
purpose of showing that it is a physical variable with stationary character for a
stationary state. To do this, we consider the general definition of the particle density
operator of a system with n particles:

n
. ρ̂(r) = δ(r − ri ) (10.37)
i=1
The most probable value of .ρ̂ gives the particle density.
. ψ(r1 σ1 , r2 , σ2 , · · · , rn , σn ; t)| ρ̂(r) |ψ(r1 σ1 , r2 , σ2 , · · · , rn , σn ; t) (10.38)
By inserting the operator .ρ̂ in this expression and displaying the integrals explicitly
using the notation of Eq. 10.2, we have the spatial distribution of the indistinguish-
able particles.

n
ρ(r; t) = δ(r − ri ) ψ ∗ ({ri σi }; t) ψ({ri σi }; t) dv; i = 1, · · · , n
.
s1 ,··· ,sn
i=1
=n |ψ({ri σi }; t)|2 dv ; i = 2, · · · , n
s2 ,··· ,sn
(10.39)
If we consider .ψ as given by Eq. 10.36, the particle density is .ψ ∗ ψ = |A φ(r)|2 . In
this respect, .ρ is time indifferent, and .ψ and .ρ describe a stationary state.
10.5 Superposition Principle of Quantum States
It is important to discuss the superposition principle since it is a key aspect in

quantum mechanics. According to the superposition principle,
the state vectors
.{ψi } may be superposed to create a new wave function, . i i i = ψ, where the
c ψ
coefficients .ci are in general complex numbers. Since the functions .ψi are solutions
of the equation of motion 10.18, their linear combination is also a solution of the
wave equation 10.18 because such an equation is a linear differential equation.
Contrariwise, any wave function of the Hilbert space associated with the system,
say .ψ , may be reproduced by the superposition of state vectors, .ψ = i di ψi .
The wave function .ψ, which results from a linear combination of state vectors
.ψi , is a state with intermediate or collective character. In this connection, it is
possible to conceive that the particle whose state is .ψ may be partially found in
every configuration state. The superposition of state vectors introduces a factor of
randomness, which is inherent in quantum mechanics.4
In the measurement process, the superposition of states collapses into the most
likely state vector, say .ψk , which owns a well-defined eigenvalue, .ak . For the
given physical conditions of the system in the experiment, the most likely state
is determined from the relative weights that the different states .ψi have in the
composition of .ψ. It is in the context of the superposition principle of quantum
states that we are able to describe the diffraction and interference phenomena of
matter waves and fields, like the diffraction of electrons and the entanglement of
quantum computing qbits, with no analog in classical mechanics. Yet, all the random
factors and the probabilistic nature of quantum mechanics fade away at macroscopic
scales (the derivation is given in a later chapter).
10.6 Bohr’s Correspondence Principle
The correspondence principle indicates in general terms that a new and more
general theory in science replaces a less general theory when: (i) the general
theory is reduced to the less general theory under specific limiting conditions,
and (ii) the general theory grants the correct results in the domains where the
less general theory is unsuccessful. The wave theory of matter accounts for the
behavior of the microscopic particles, and additionally, it reproduces the theory of
classical mechanics by describing the behavior of the macroscopic objects. There
are several situations that illustrate the general character of the correspondence
principle. For example, statistical mechanics leads to thermodynamics by dealing
with statistical averages of systems with a large number of particles. Wave optics
leads to geometrical optics. The theory of general relativity leads to the Newtonian
theory of gravity when the gravitational fields are weak. On the other side, the theory
of special relativity leads to Newtonian theory when the velocities of the bodies are
small with respect to the speed of light.
When a particle system increases in size, it is transformed from a microscopic
system of particles to a macroscopic object. This is one of the important conditions
to describe the objects of the macroscopic world from a microscopic theory. The
condition to reach a macroscopic system is known as the classical limit. In this
4 Itis important to recall the uncertainty principle, which points out the impossibility to know with
high accuracy and in simultaneous form the momentum and position of a particle. The position
and momentum operators do not commute, and it is not possible to find common eigenfunctions
of both operators.
10.6 Bohr’s Correspondence Principle 223
limit, the Planck constant .h̄ is sufficiently small, and the potentials acting on the
objects show slow changes over the objects’ characteristic wavelengths. The wave
theory of matter reproduces Newtonian theory in the classical limit. However, the
wave theory of matter has statistical nature and provides probabilistic results, while
classical mechanics gives non-probabilistic values. Also, the wave theory of matter
provides quantized results, whereas classical mechanics gives continuous values
of the physical observables. According to Bohr, the correspondence between wave
mechanics and Newtonian theory is achieved in the limit of large quantum numbers,
specifically, in the limit of large energies and large orbits. Such a correspondence
is confirmed when we compare the energy and the most probable position of
a quantum oscillator with the equivalent quantities of a classical oscillator. For
instance, an energy value of the classical oscillator may be identified with an energy
described by a large quantum number of the quantum oscillator. The energy levels
of the quantum oscillator are additionally found to be extremely close among each
other in the classical limit (the gap energy is small, namely, of order .h̄ω), in
such a way that we essentially deal with continuous values of the energy of the
quantum oscillator, with little possibility of distinguishing the individual quantum
energy values in that limit. In this regard, the quantized energies are blended into a
continuous energy spectrum.
Problem 10.3 The energy of an atomic oscillator is .Eat = (n + 1/2)h̄ω. The

quantum number n takes the values .0, 1, 2, · · · , and the angular frequency
is .ω. On the other hand, a classical harmonic oscillator has an energy that
follows Hooke’s law, .Eclass = kA2 /2, where A is the amplitude at maximum
extension and k is the force constant. Estimate the value of the quantum
number n for the values .A = 0.1 m, mass= 0.3 Kg, and .ω = 1 rad/s of
a macroscopic oscillator, and estimate the gap energy between contiguous
energy levels.
Solution The strategy to find the quantum number of a classical oscillator is

to identify the atomic and classical energies, .E = Eat = Eclass , and solve
the resulting equation√ for the quantum number, n. The oscillation frequency is
related to k by .ω = k/m, where m is the mass of the oscillating particle. The
classical energy is .Eclass = mω2 A2 /2. After taking the values of A, m, and .ω,
the quantum number is found to be sufficiently large, .n ∼ 1 × 1030 . The gap
energy of contiguous energy levels is .En+1 − En = h̄ω = 1.054 × 10−34 J ,
which is extremely small from a macroscopic perspective.
In attention to the postulates of quantum mechanics and the correspondence

principle, it is possible to establish a link between the variables of quantum
mechanics and their counterparts in classical mechanics. Such a correspondence
is possible after recalling the probabilistic character of the wave theory of matter,
where we use probabilities to determine the values of the physical observables.

According to the wave theory of matter, the expectation value .ψ|Â|ψ of a physical
variable .A must correspond, in the limit of large quantum numbers (abbreviated in
the form .n → ∞), to the Newtonian value of that physical observable measured on
the macroscopic object.
. ψ|Â|ψ =A (10.40)
n→∞
The correspondence between both results is confirmed, once more, for a classical
oscillator and a quantum oscillator with large quantum numbers. For example, when
the most probable locations of both oscillators are computed, the value obtained
from Newtonian mechanics and the highest probable value obtained from quantum
theory are found in good agreement [4].
In conclusion, the classical limit of the wave theory of matter takes us to the
theory of classical mechanics. This fact favors the use of classical mechanics over
quantum mechanics in the domains where the two theories overlap, since Newtonian
mechanics is computationally less demanding than quantum mechanics. By dealing
with classical mechanics, we just require the positions and velocities of every
particle to describe any other physical quantity, like the linear momentum, angular
momentum, and energy of the particle system, even those variables of macroscopic
thermodynamics. The evolution of the position and the evolution of the velocity of
a given particle are defined by the respective Newtonian equations of motion, which
are first-order differential equations with time dependence, Sect. 6.15.
10.7 Selection Rules
The atomic excitations are phenomena which cannot be explained in classical

mechanics. However, from the quantum-mechanical perspective, the excitation of
an atom to a high energy level is due to either the absorption of electromagnetic
radiation or the collision with another particle. The transit from a low energy level
to a high energy level depends on the initial and final atomic states, and the energy
of the electromagnetic radiation as well. When the energy of the electromagnetic
radiation fits the energy difference between the initial and final states, the interaction
of the atom and the electromagnetic field is resonant, and the transit to a high energy
state occurs. Eventually, the atom decays to a low energy level (due to the interaction
of the atom with the vacuum fluctuations) with the radiation of a photon by the
atom. In this decay, the emitted radiation once more matches the energy gap of
energy levels. In this context, quantum mechanics considers the possibility of an
atom radiating a photon, among other processes.5
5 Hecht [5], Sect. 3.4.4. In some sense, the impact of a photon, which may be visualized here as a
wave with specific characteristics, on an electron, which may be visualized also as another wave,
is capable to produce new waves.
10.7 Selection Rules 225
The electronic transitions in quantum mechanics are common and important. In

an electronic transition to a lower energy state, there is a sudden distribution change
of the electron cloud. This process emulates a radiating antenna: for instance, if the
electron changes from the 2s state down to the 1s state, a spherical change of the
electron cloud is observed, and the atom radiates a photon. In contrast, an atom in a
stationary state, Eq. 10.36, shows no radiation because the average electron position,
.ψ|r|ψ, is time independent.
∗
. ψ|r|ψ = φ(r)e−iEt/h̄ r φ(r)e−iEt/h̄ d 3 r = φ1∗ (r) r φ2 (r) d 3 r (10.41)
Interestingly, there is a time interval where the electron simultaneously coexists in

the two energy levels of a transition. The state function in this case is given as a
linear combination of the two states, .ψ = c1 ψ1 + c2 ψ2 . The average position with
such a function is:

ψ|r|ψ = c12 φ1∗ r φ1 d 3 r + c22 φ2∗ r φ2 d 3 r +
.
+ c1∗ c2 φ1∗ eiE1 t/h̄ r φ2 e−iE2 t/h̄ d 3 r + c1 c2∗ φ1 e−iE1 t/h̄ r φ2∗ eiE2 t/h̄ d 3 r
(10.42)
The average electron position is time dependent. By using the identity .cos θ =
[exp(iθ ) + exp(−iθ )]/2, the average position of the electron is:

. ψ|r|ψ = c12 φ1∗ r φ1 d 3 r + c22 φ2∗ r φ2 d 3 r + 2c1∗ c2 cos(2π νt) φ1∗ r φ2 d 3 r
(10.43)
The vibrational frequency of the electron is .ν = (E2 −E1 )/ h, which corresponds to
the frequency of the emitted electromagnetic radiation [6]. A short time polarization
occurs in the time interval of electron density changes. If .ψ2 represents the state
function of the system in state 2, .ψ1∗ is the complex conjugate of the state function
of the system in state 1, and .μ̂ is the dipole moment operator, then the value of the
transition dipole moment is computed as [7]:

. ψ1 |μ̂|ψ2 = ψ1∗ μ̂ ψ2 dv (10.44)
We now deal with the wave functions associated with the different degrees of
freedom of a molecule [8]. In many instances the system is in the energy ground
state, and there is a subtle correlation between the dynamics of the electrons and the
vibrational and rotational motions of the nuclei. For a heavy atom where the mass
of the electron is substantially smaller than the mass of the nucleus, it is possible
to (parametrically) separate the dynamics of the electrons from the nuclear motion
without a major loss of accuracy. The state function is approximated by the product
of the electronic state function, .ψe , the nuclear vibrational state function, .ψv , and
the nuclear rotational state function, .ψr .
. ψ(ri , Rj ) ψe (ri , Rj ) ψvr (Rj ) ; ψvr (Rj ) ψv (Rj ) ψr (Rj ) (10.45)
The approximation on the right assumes a vibrational motion uncoupled from

the other motions. According to such separations, the transitions between the
respective energy levels are classified as electronic, vibrational, and rotational
transitions. There is an electronic transition when an electron changes energy levels,
a vibrational transition when an oscillator jumps to a different energy level, and a
rotational transition when a rotor changes energy levels. The coupled transitions
are observed when an electronic transition is coupled to a vibrational mode, or
a vibrational transition is coupled to a rotational mode, leading to spectra with
superimposed fine structure.
In most instances, the electronic transitions occur rapidly with the nuclei being
practically standstill. Yet, for a given electronic state, the nuclei may be excited
to a high molecular vibrational state. The off-diagonal transition matrix element
.ψ1 |μ̂e + μ̂n |ψ2 , where .μ̂e and .μ̂n are the electronic and nuclear dipole operators,
is unfolded in terms of the electronic and nuclear-vibration wave functions.

∗ μ̂ ψ
∗
ψ1 |μ̂|ψ2 = ψe,1 3
e e,2 d r ψv,1 ψv,2 d 3 R
∗ ∗
. + ψe,1 ψe,2 d 3 r ψv,1 μ̂n ψv,2 d 3 R (10.46)
∗ ∗
ψ1 |μ̂|ψ2 α ψe,1 μ̂e ψe,2 d 3 r ; α = ψv,1 ψv,2 d 3 R
The rotational motion was dispensed, and the orthogonality of the electronic
state functions was used to simplify the above expression. The integral over the
vibrational wave functions shown in the first line is denoted with .α. The square
of .α is known as the Franck-Condon factor. The electronic transition induces a
vibrational transition when the vibrational state functions of the Franck-Condon
factor exhibit a significant overlap. On the other hand, when the electronic motion
is considered to be practically uncoupled of the nuclear motion (based, e.g., on the
large difference in the motion time scales of the electrons and nuclei), then the Born-
Oppenheimer method of quantum mechanics may be applied.
Problem 10.4 A system of atoms has constant energy. The coordinates of the
electrons and nuclei are .xi and .Xα , with masses m and .Mα , and momenta .pi
and .Pα , respectively. The system satisfies the time-independent wave equation
.Ĥψ = Eψ, with:
h̄2 h̄
. Ĥ = T̂ + V̂ = − p̂i2 − P̂α2 + V̂ (10.47)
2m α
2M α
i
(continued)
10.7 Selection Rules 227
The potential energy of the electrons is .V̂e , of the nuclei is .V̂n , and .V̂ =
V̂e + V̂n . Find the equations and conditions that .ψe and .ψn should satisfy to
write the function .ψ of the system of atoms in a first approximation as the
product .ψe ({xi }; {Xα }) ψn ({Xα }).
Solution The momentum operator in quantum mechanics is .p̂x =

−(i/h̄) ∂/∂x.
h̄2 h̄2 h̄2 ∂ 2 ψ

. − p̂i2 − P̂α2 + V̂ ψ = − + V̂e ψ
2m α
2Mα 2m ∂xi2
i i
h̄2 ∂ 2 ψ
α
− 2
+ V̂n ψ = Eψ
α
2M α ∂X α
When the function .ψ = ψe ψn is introduced through .ψ, and we identify

2 ψ/∂X 2 = ψ ∂ 2 ψ /∂X 2 + 2 (∂ψ /∂X ) (∂ψ /∂X ) + ψ ∂ 2 ψ /∂X 2 , we obtain:
.∂ α n e α n α e α e n α
h̄2 h̄2
ψn −
. p̂i2 + V̂e ψe + ψe − P̂α2 + V̂n ψn
2m α
2M α
i
h̄2 ∂ 2 ψe ∂ψn ∂ψe

− ψn +2 = Eψe ψn
α
2Mα ∂Xα2 ∂Xα ∂Xα
A time-independent expression for .ψe , and another for .ψn , is reached when
.ψe changes slowly with respect to .Xα , as the crossed terms .∂ψe /∂Xα and
2 2
.∂ ψe /∂Xα may be in this case neglected.
h̄2 h̄2
. − p̂2 + V̂e ψe = Ee ψe ; − P̂α2 + V̂n ψn = (E − Ee )ψn
2m i α
2M α
i
(10.48)
There exist selection rules that determine whether a transition between energy
levels is possible or not. More precisely, the selection rules define the transition
probabilities among the different atomic and molecular states. The selection rules
on the dipole moment indicate the allowed and forbidden energy transitions
based on the results obtained from Eq. 10.44. The transitions involving high-order
multipoles also satisfy the selection rules, but they are less probable to occur from
an energetic point of view. The selection rules are frequently established from
symmetry considerations of the state functions using group theory and examining
the conservation of the total angular momentum because, in the absorption and
emission processes by atoms and molecules, a photon carries a unit of quantized
angular momentum, in spite of its zero mass. Thereby, the selection rules take
the change of the total angular momentum and the preservation of the energy
into consideration. Finally, the selection rules are expressed in simple forms using
good quantum numbers, this is, the quantum numbers that show no change due
to a negligible coupling among the different degrees of freedom like, for instance,
when the electronic angular momentum is weakly coupled to the nuclear spin of the
molecule.
10.8 Pauli’s Principle in the Electronic Wave Function
There are additional rules on the wave function to those imposed by the quantum-
mechanical postulates, especially for the fermionic particles. The electrons belong
to this category, and thus, we are interested in the constraints that the electronic
wave function should satisfy. Pauli’s principle is one of such constraints.
Statement 10.7 Pauli’s principle or exclusion principle: two or more

fermions of a system of particles cannot occupy the same quantum state
defined by the quantum numbers n, l, .ml , and .ms .
Pauli’s principle is also known as the exclusion principle. The implications of

the exclusion principle in the wave function are understood by considering, for
example, a system of two indistinguishable particles, 1 and 2, and the states a
and b. Consider the wave function .ψ(a, b), where the particles 1 and 2 occupy
the a and b states, respectively, and the wave function .ψ(b, a), where the particles
1 and 2 occupy the b and a states, respectively. Both of them satisfy the identity
.|ψ(a, b)| = |ψ(b, a)| , because .|ψ| is a physical observable independent of the
2 2 2
labeling of the particles. In this respect, .|ψ|2 is unique and represents the probability
density of the indistinguishable particles to be found at a given space region. Yet,
the result .|ψ|2 may be derived from two possible symmetries of the wave function:
ψ(a, b) = ψ(b, a) ; symmetric wave function

. (10.49)
ψ(a, b) = −ψ(b, a) ; antisymmetric wave function
When the wave function .ψ is expressed in terms of the individual state functions .ψa
and .ψb , two equally probable versions of .ψ are obtained.
. ψI = ψa (1)ψb (2) ; ψI I = ψa (2)ψb (1) (10.50)

10.9 Wave Function of the Electrons in a Molecule 229
The way to build a symmetric wave function (sym) and an antisymmetric wave
function (ant), satisfying Eqs. 10.49, is to linearly combine .ψI and .ψI I .
. ψsym = ψa (1)ψb (2) + ψa (2)ψb (1) ; ψant = ψa (1)ψb (2) − ψa (2)ψb (1)
(10.51)
In order to determine which of the above wave functions satisfies the exclusion
principle and, thereby, is the appropriate wave function to describe the system of
indistinguishable fermions, let us consider that particle 1 occupies the same quantum
state as particle 2. In this case, the wave function .ψant gives zero, leading to a
zero probability of finding the particles with the same quantum numbers, in line
with Pauli’s principle. Thus, the wave function .ψant is the appropriate function to
describe the pair of particles. The general conclusion is that the fermionic particles,
which cannot exist in the same quantum state at the same instant of time, must be
described by an antisymmetric wave function. According to .ψant of Eq. 10.49, the
antisymmetry principle requires that an exchange of particles yields a change of
signs of the wave functions. On the other hand, with respect to the bosonic particles,
they are not required to satisfy Pauli’s exclusion principle, and they are permitted to
occupy the same quantum state.
10.9 Wave Function of the Electrons in a Molecule
If a system of n identical fermions is described by the state function .ψ, then Pauli’s
exclusion principle ensures that the exchange of any two particles, with variables
.x1 , .x2 , results in the same wave function by reversing the sign.
. ψ(x1 , x2 , · · · , xn ; t) = (−1)2s ψ(x2 , x1 , · · · , xn ; t) (10.52)
The nomenclature .xi = (ri , σi ) is used, where .ri and .σi are the position and spin
of the ith particle. We have the identity .(−1)2s = −1 for fermions (.s = 1/2) and
.(−1)
2s = 1 for bosons (.s = 1), as these last particles may occupy the same energy
state.
One of the simplest representations of an antisymmetric wave function describing
the electrons, fermions, of a molecule is a determinant. The elements that build
the determinant are the spin orbitals .φ(r, σ ). The spin orbitals are mathematical
functions employed to describe the individual electrons and define their spatial
extensions and spin states. The spin orbitals are orthogonal to each other to avoid
linear dependencies and are individually normalized to keep the number of electrons
described by the wave function constant. The spin orbitals form a complete basis set
and guarantee the proper expansion of the Hilbert vector space associated with the
system. In other words, such a feature ensures the description of the state of the
system with a linear combination of spin orbitals. Still, the spin orbitals should not
be understood as physical observables because they are descriptive mathematical
functions with not counterpart in the real world.
If the function .φi (xj ) represents the ith spin orbital occupied by the j th electron,
then the wave function of the molecular system with n electrons may adopt the form
of a determinant [9].

φ1 (x1 ) . . . φ1 (xn )

.
φ2 (x1 ) .. φ2 (xn )
. ψ(x1 , x2 , . . . , xn ) = (10.53)
. ..
.. φi (xj ) .

φ (x ) . . . φ (x )
n 1 n n
The above expression is a Slater determinant. In a Slater determinant, all the

electrons are distributed in all the orbitals; there is not a one-to-one relation between
electrons and orbitals. For instance, electron 1 is located in orbitals .φ1 , .φ2 , .· · · ,
.φn , and similarly for the other electrons, refer to the columns of the determinant.
In this respect, the electrons are spatially delocalized under a Slater-determinant

representation. On the other side, every electron is given equal treatment. For
example, the orbital .φ1 provides “equal treatment” to all the electrons, because it
allocates electrons 1, 2, .· · · , n, and similarly for the other orbitals, refer to the rows
of the determinant. Thereby, in spite of labeling the electrons, the orbitals do not
discriminate them, and the electrons, free to occupy any spin orbital, are treated as
indistinguishable fermions under a determinantal expression of the wave function.
In addition, an exchange of electrons is equivalent to an exchange of columns, and,
in turn, the exchange of columns demands the change of the determinant sign. The
antisymmetry principle is implicitly considered in the Slater determinant following
the quantum-mechanical requirements on the electrons. The Slater determinant is
frequently abbreviated using the bra-ket notation, only making reference to the
diagonal elements (.φi = φii = φi (xi )).

φ1 (x1 ) . . . φ1 (xn )

.. ..
. ψ(x1 , x2 , . . . , xn ) = |φ1 (x1 ) · · · φn (xn ) =
. φi (xi ) . (10.54)
φ (x ) . . . φ (x )
n 1 n n
Problem 10.5 By assuming the orthonormality of the spin orbitals, find the
normalization constant of the wave function 10.54.
Solution The Slater determinant contains n orthonormal spin orbitals and n

electrons. In order to determine the normalization coefficient of the determi-
nant, we consider that the spin orbitals may be populated by any electron of
the system. We have n possibilities to populate the first spin orbital, .n − 1
possibilities to populate the second spin orbital, etc. In this connection, the
normalized wave function is:
(continued)
10.9 Wave Function of the Electrons in a Molecule 231
. ψ(x1 , x2 , . . . , xn ) = (n!)−1/2 |φ1 (x1 ) · · · φn (xn ) (10.55)

The integral . ψψ ∗ dv1 · · · dvn gives unity.
An improved representation of the electron wave function is achieved when .ψ is

written as a linear combination of Slater determinants, with more spin orbitals than
electrons, where the spin orbitals are fractionally occupied by the electrons. The
improvement of a multi-determinant wave function over a single determinant is
associated with the greater spatial relaxation of the electrons due to the availability
of more spin orbitals.
In order to illustrate the construction of a wave function of a 2-fermion system,
let us consider that the set of spin orbitals .{φi } form a basis set. The function
.ψ(x) is
written in terms of such spin orbitals with the coefficients .ci as .ψ(x) = i ci φi (x).
If .ψ(x1 , x2 ) is a function of two variables, then the variable .x2 may
be fixed, this is
.x2 = constant, to write the expansion in the form .ψ(x1 , x2 ) = i ci (x2 )φi (x1 ) for
that particular value of .x2 . The same process may be repeated for every value of .x2 .
However, in these cases, .ci is understood as a function of .x2 , and .ci is expanded in
terms of the same basis set: .ci (x2 ) = j dij φj (x2 ). By inserting this expression in
.ψ(x1 , x2 ), we obtain:

.
ψ(x1 , x2 ) = dij φi (x1 )φj (x2 )
i,j
The antisymmetric condition should be imposed in .ψ(x1 , x2 ) to define the appro-

priate .dij factors. The function .ψ is displayed in terms of diagonal and off-diagonal
elements.

.
ψ(x 1 , x2 ) = dii φ i (x1 )φ i (x2 ) + [dij φi (x1 )φj (x2 ) + dj i φi (x2 )φj (x1 )]
i=1 i<j
According to the antisymmetric condition, when .x1 = x2 , we should have

.ψ(x1 , x1 ) = 0. This condition demands for every pair ij of indexes the following
identity .(dij + dj i )φi (x1 )φj (x1 ) = 0, which is satisfied when .dj i = −dij . Also, for
.i = j , we deduce .dii = 0. We may illustrate the process of building the correct .ψ
for the case of indexes .i, j = 1, 2, 3
ψ(x1 , x2 ) = d12 φ1 (x1 )φ2 (x2 ) + d13 φ1 (x1 )φ3 (x2 ) +

. − d12 φ2 (x1 )φ1 (x2 ) + d23 φ2 (x1 )φ3 (x2 ) +
− d13 φ3 (x1 )φ1 (x2 ) − d23 φ3 (x1 )φ2 (x2 )
In general, the resulting function .ψ with pairwise terms is abbreviated by means of

determinants.

φi (x1 ) φi (x2 )
. ψ(x1 , x2 ) =

dij
φj (x1 ) φj (x2 ) (10.56)
i<j
.ψ is an antisymmetric wave function in terms of two variables and expressed

as a linear combination of two-dimensional determinants built from the set of
spin orbitals .{φi }. The construction of a wave function of n variables leads to
determinants of .n × n dimensions. Thus, the wave functions associated with the
ground state and the excited states of a n-fermion system may be written in terms
of all the unique determinants formed from the one-variable spin orbitals .φi (x),
.i = 1, · · · , m. The number of unique determinants is given by the binomial
coefficient .Cnm , corresponding to the number of ways in which n objects are taken
from a set constituted by m objects and .n < m. Section B.1 provides a method to
expand a Slater determinant because it is useful to evaluate matrix elements of the
operators.
Problem 10.6 Write the wave function .ψ(x1 , x2 , x3 ) of a 3-fermion system

in terms of a linear combination of .3 × 3 determinants following the approach
of the main text.
Solution The factor .dij of Eq. 10.56 is a function of .x, andit is equally
expanded in terms of the complete basis set .{φk }: .dij (x) = k αij k φk (x),
where .αij k are the expansion coefficients of .dij .

dij (x3 ) [φi (x1 )φj (x2 ) − φi (x2 )φj (x1 )]
.
i<j
= αij k φk (x3 ) [φi (x1 )φj (x2 ) − φi (x2 )φj (x1 )]
i<j,k
The permutation of the variables .x1 , x2 , x3 should be considered in the

expression of .ψ to favor no single variable.
1 , x2 , x3 )
ψ(x
= αij k [φi (x1 )φj (x2 )φk (x3 ) − φi (x2 )φj (x1 )φk (x3 )] + ← aei − ibd

i<j,k
. + αkij [φi (x2 )φj (x3 )φk (x1 ) − φi (x3 )φj (x2 )φk (x1 )] + ← gbf − ceg

i<j,k
+ αj ki [φi (x3 )φj (x1 )φk (x2 ) − φi (x1 )φj (x3 )φk (x2 )] ← dhc − f ha
i<j,k
(continued)
10.10 Variational Principle of the Energy 233
In order to satisfy Pauli’s exclusion principle, we separately do .x1 = x2 = x,

x1 = x3 = x, and .x2 = x3 = x, reaching the conclusion .αij k = αkij = αj ki .
.
The above elements of .ψ are identified with the elements of a determinant.

a b c

. aei + dhc + gbf − ceg − f ha − ibd = d e f

g h i
Thereby, .ψ is written as a linear combination of .3 × 3 determinants.

φi (x1 ) φi (x2 ) φi (x3 )

ψ(x1 , x2 , x3 ) =
. i<j <k αij k φj (x1 ) φj (x2 ) φj (x3 )
φ (x ) φ (x ) φ (x )
k 1 k 2 k 3
The possibility of dealing with identical rows or identical columns in every

determinant is avoided with the restriction .i < j < k on the subindexes.
10.10 Variational Principle of the Energy
In this section we demonstrate that .Et , the energy associated with an arbitrary trial
wave function .ψt , is an upper bound on the ground-state energy .E0 of the system.
If the trial wave function is solution of the Hamiltonian .Ĥ, then .ψt satisfies the
time-independent equation.
. ψt |Ĥ|ψt = Et ψt |ψt (10.57)
The goal is to compare the energy .Et with the ground-state energy .E0 . Firstly, we
take an eigenvalue of .Ĥ, say .Ek , as a reference energy. The eigenfunction .ψk has
eigenvalue .Ek .
. Ĥψk = Ek ψk (10.58)
The eigenfunctions of .Ĥ satisfy the orthonormality condition .ψk |ψl = δkl . We
compute the energy difference between .Et and .Ek to work with the deviation
originated by the use of the trial wave function .ψt .
. Et − Ek = [ψt |Ĥ|ψt − ψt |ψt Ek ] / ψt |ψt = [ψt |Ĥ − Ek |ψt ] / ψt |ψt
When the trial function .ψt deviates little from the eigenfunction .ψk , the relation
between .ψt and .ψk is written in linear form .ψt = ψk + δψ.
Et − Ek = ψk + δψ|Ĥ − Ek |ψk + δψ / ψt |ψt

.
= [ψk |Ĥ − Ek |ψk + ψk |Ĥ − Ek |δψ + δψ|Ĥ − Ek |ψk
+ δψ|Ĥ − Ek |δψ] / ψt |ψt (10.59)
The first term in the numerator is zero because .ψk is an eigenfunction of .Ĥ,
Eq. 10.58, and .Ĥ is Hermitian. The second and third terms also vanish. The last
term involves the second-order contribution, which is non-zero.
. Et − Ek = δψ|Ĥ − Ek |δψ / ψt |ψt
The energy difference in the use of the trial wave function .ψt and the wave function
ψk depends on the term .δψ. If we wish to minimize the energy difference .Et − Ek ,
.
which we denote with .δE, then we are required to impose the condition:
. δE = 0
The way to obtain .δE = 0 is by tuning the trial wave function, because .δE depends
on the difference .δψ = ψt − ψk and, in turn, the trial wave function .ψt depends on
parameters, which can be in principle adjusted to produce .δE = 0. The condition
.δE = 0 unfolds a variational principle as it imposes the condition to find the best
trial wave function that minimizes the energy.

It is our next purpose to demonstrate that the trial energy .Et is an upper bound
on the ground-state energy .E0 . To do this, it is convenient to expand .ψt in terms of
the eigenfunctions .ψk of the Hamiltonian.

. ψt = k ck ψk
The energy corresponding to this function is obtained from the expectation value of
the Hamiltonian .Ĥ.

.
ψt |Ĥ|ψt = ck∗ cm ψk |Ĥ|ψm = ck∗ cm Em ψk |ψm = ck2 Ek
k,m k,m k
If we replace .Ĥ by the identity operator .Iˆ and perform the same operations, we
immediately obtain the orthonormality condition in terms of the .ck coefficients.

.
ψt |ψt = ck2 = 1 (10.60)
k
A.1 Appendix: Proposing the Wave Equation for Matter Waves 235
We still require to show that .Et is equal to or greater than the ground-state energy
E0 . Since .E0 represents the lowest value among all possible eigenvalues .Ek , then:
.

.
Et = ck2 Ek ≥ E0 ck2 ≥ E0
k k
The trial energy is an upper bound on the ground-state energy.
. Et ≥ E0 (10.61)
The variational principle favors the trial wave function that gives the lowest energy
value among many other wave functions with higher energy values of the particle
system, because the trial wave function that exhibits the lowest energy is the wave
function that better approximates the ground state of the system. The electronic
wave function expressed as a single Slater determinant of the spin orbitals, in
conjunction with the variational principle of the energy, gives foundation to the
Hartree-Fock method. Other Hartree-Fock based methods deal with many Slater
determinants and give foundation to the multi-configuration methods. Yet, they all
use the variational principle to find the lowest energy of the electronic system [10].
A.1 Appendix: Proposing the Wave Equation for Matter Waves
A particle in the quantum-mechanical world has dual nature and may behave as a
wave or a particle according to the circumstances. In this regard, it is necessary to
establish an equation that accurately describes both possible aspects of the particle
in the quantum realm. Specifically, we need to find an equation that governs the
behavior not only of particles but also of matter waves. To accomplish this, we start
with the one-dimensional equation that rules the behavior of waves in the classical
world.
∂ 2 y(x, t) 1 ∂ 2 y(x, t)
. = (10.62)
∂x 2 v2 ∂t 2
This expression rules the behavior of the wave described by y, traveling with speed
v along the x direction, and with dependence on time t. The changes of y in the
positions are related to changes of y with times. One of the simplest solutions of the
wave equation is:
. y(x, t) = A cos(kx − wt) ; k = 2π/λ ; ω = 2πf (10.63)
The terms A, .λ and f are the amplitude, wavelength, and frequency of the wave,
respectively. By inserting the trigonometric solution in Eq. 10.62, we obtain the
relation .ω = kv or .v = λf between the parameters of the classical wave. There
is consistency between the classical wave equation and both the traveling wave and
the wave parameters, which all together establish the possibility to characterize an
entity as a wave. Still, since a particle has dual character, the equation of matter
waves is expected to also include mechanical parameters to characterize the particle
as a mechanical object.
The behavior of the mechanical particle is defined by the particle’s energy E,
which is integrated by the kinetic energy .Ek = p2 /(2m) and the potential energy
.V (x, t), where the particle is immersed. In this respect, the equation governing
the particle behavior should contain information of the linear momentum p, the
potential energy V , and the energy E. However, before formulating the wave
equation of matter waves, it is important to recall the de Broglie relations linked
to the linear momentum and energy:
p = h/λ ; E = hf
. (10.64)
p = h̄k ; E = h̄ω
The identities .h̄ = h/(2π ) and .k = 2π/λ were used. The particle’s energy is written
in terms of the wave number k and angular frequency .ω.
E = p2 /(2m) + V (x, t)
. (10.65)
= h̄2 k 2 /(2m) + V (x, t) = h̄ω
In the following we consider a plane wave as given by a complex exponential

function. Such a function considers the complex number i and is relatively different
to the trigonometric wave 10.63.
. yp (x, t) = exp[i(kx − ωt)] (10.66)
The quantum wave equation is expected to contain some ingredients of the classical
wave expression, Eq. 10.62, like the differentials of the matter-wave function with
respect to the position and time because they provide the correlated changes of the
wave in space and time. Thereby, we differentiate the plane wave with respect to the
position and time.
∂ ∂2
yp (x, t) = ik exp[i(kx − ωt)] ; yp (x, t) = −k 2 exp[i(kx − ωt)]
. ∂x ∂x 2
∂
yp (x, t) = −iω exp[i(kx − ωt)]
∂t
(10.67)
In order to reproduce the energy 10.65 in terms of the above derivatives, we link
p = h̄k and .E = h̄ω, Eqs. 10.64, to the position and time derivatives.
.
−h̄2 ∂ 2 yp ∂
. = (h̄k)2 yp / (2m) ; i h̄ yp (x, t) = h̄ω yp (10.68)
2m ∂x 2 ∂t
B.1 Appendix: Expansion of a Determinantal Wave Function 237
We now have the possibility of proposing the equation for matter waves. The
energy 10.65 is .E = h̄2 k 2 /(2m) + V (x, t) = h̄ω. It can be written in terms of
the position and time derivatives.
∂ −h̄2 ∂ 2 yp (x, t)
. i h̄ yp (x, t) = + V (x, t)yp (x, t) (10.69)
∂t 2m ∂x 2
This is the non-relativistic 1D wave equation of matter waves. It is different to the
classical equation 10.62 as a single time derivative is considered in Eq. 10.69 while
a double time derivative is considered in Eq. 10.62, thus breaking the symmetry
between spatial and time derivatives, and resembling a kind of diffusion expression.
Equation 10.69 also includes the imaginary number, i. It is common to represent the
wave function with the Greek letter .ψ instead of .yp . The corresponding expression
in 3D space for many particles is:
−h̄2
. i h̄ ∂t∂ ψ({ri }, t) = i 2m ∇i2 ψ({ri }, t) + V ({ri }, t) ψ({ri }, t) (10.70)
The equation of matter waves is an equation of mathematical operators acting on .ψ.

The operators are engulfed by the Hamiltonian operator, .Ĥ.
∂ −h̄2
. i h̄ ψ({ri }, t) = Ĥψ({ri }, t) ; Ĥ = ∇i2 + V ({ri }, t)
∂t 2m
i
(10.71)
The equation of matter waves satisfies the principle of superposition, since the linear
combination of solutions .ψi of the matter-wave equation is also a solution of that
equation.

n
. ψ({rk }, t) = ci ψi ({rk }, t) (10.72)
i=1
The superposition principle is required to have mathematical well-behaved solu-

tions, capable to unfold the Hilbert space associated with the system, in addition to
make possible the dispersion and interference phenomena of waves in the physical
world.
B.1 Appendix: Expansion of a Determinantal Wave Function
From a mathematical point of view, a determinant is an abbreviated description of

a sum of products of the determinantal elements without repetition. The addition
of determinants also represents a summation of products, where all the elements
of the determinants are considered under equal circumstances. The expansion of
a Slater determinant allows to write the electronic wave function in an alternative
form and demonstrate its properties. We use the Laplace expansion approach to
expand the Slater determinant [11]. According to this procedure, a set of smaller
determinants are produced after systematically removing rows and columns of the
original determinant, which is assumed to have size .n×n. Let us consider an element
of the determinant, say .aij , where the i and j indexes represent a row and a column
of the determinant. A set of smaller determinants with dimension .n−1 are produced
after deleting selected rows and selected columns of the original determinant. In the
example provided below, a determinant of size .3 × 3 is expanded in terms of three
determinants of size .2 × 2. The expansion coefficients are the elements (2,4,6) that
integrate the second column of the original determinant.
1 2 7
3 8 1 7 1 7
3 4 8 =−2 + 4 − 6 = −4
5 11 5 11 3 8
5 6 11
The signs of the expansion coefficients are determined from a checkerboard

pattern whose elements are given in the general form .(−1)i+j , where i is the row
and j the column of the element .aij . The checkerboard pattern of a determinant of
dimension 3 alternates signs.
⎛ ⎞ ⎛ ⎞
+−+ (−1)1+1 (−1)1+2 (−1)1+3
. ⎝ − + − ⎠ = ⎝ (−1)
2+1 (−1)2+2 (−1)2+3 ⎠ (10.73)
+−+ (−1)3+1 (−1)3+2 (−1)3+3
The Laplace expansion may be performed using the elements of a row instead of the
elements of a column. Such an alternative is illustrated using our example with the
same checkerboard pattern of signs and taking the elements (1,2,7) that integrate the
first row of the original determinant as the expansion coefficients.
1 2 7
4 8 3 8 3 4
3 4 8 = 1 − 2 + 7 = −4
6 11 5 11 5 6
5 6 11
When a column of the determinant is exchanged with a neighbor column, the sign
of the determinant should be changed according to the previous checkerboard. For
instance, after exchanging the first and second columns, we have the antisymmetric
property of the determinant.

1 2 7 2 1 7

.3 4 8 = −4 3 8

5 6 11 6 5 11
If two columns of the determinant are repeated, for example, the column (1,3,5) is
replaced by the column (2,4,6) in the numeric example, we in principle have two
B.1 Appendix: Expansion of a Determinantal Wave Function 239
fermions populating the same spin orbital, and the determinant is zero, satisfying
the exclusion principle.
2 2 7
4 8 4 8 4 4
4 4 8 = 2 − 2 + 7 =0
6 11 6 11 6 6
6 6 11
The small determinants resulting from the previous expansions are reduced
determinants, known as minors. The Laplace expansion has general character and is
applied to determinants of any dimension. The general expansion in terms of the ith
row of a determinant A has the form:
. det A = (−1)i+1 ai1 det Ai1 + · · · + (−1)i+n ain det Ain (10.74)
And the expansion in terms of the j th column of a determinant A is:
. det A = (−1)1+j a1j det A1j + · · · + (−1)n+j anj det Anj (10.75)
The indexes of the coefficients .ai1 , .· · · , .ain in Eq. 10.74 and the indexes of the
coefficients .a1j , .· · · , .anj in Eq. 10.75 indicate the rows and columns that should be
omitted from the original determinant A to build the minors of dimension .(n − 1).
Thus, the identities 10.74 and 10.75 show an expansion of the original determinant
in terms of determinants with reduced dimension. The Laplace expansion may be
again applied to the minors to achieve further reduction of the determinants. The
recursive expansion of the minors leads to a final expression where only sums of
products are shown (with no presence of a determinant).
The computational cost of the Laplace expansion is .n!, where n is the dimension
of the original determinant. The Laplace expansion is not an efficient technique;
nevertheless, the efficiency may be improved by carrying expansions in terms
of rows or columns that contain zeros (since a zero multiplied by its respective
minor gives a null contribution and simplifies the expansion process). The Laplace
expansion is used to obtain the matrix elements of operators representing the
physical observables:
. A = φ1 (x1 ) · · · φn (xn )| Â |φ1 (x1 ) · · · φn (xn ) (10.76)
where .Â is the operator associated with the physical observable .A and where
we have used the bra-ket notation, Eq. 10.54, to describe the determinantal wave
function of the particle system. In conclusion, the Laplace expansion shows that a
Slater determinant can be represented as a summation of smaller determinants, and
such a summation satisfies Pauli’s exclusion principle.
References
1977)
2. R. Tolman, The Principles of Statistical Mechanics (Dover Publications, New York, 1979),
Sect. 95, pp 395–399
1977), pp. 244–250
4. A. Beiser, Conceptos de Física Moderna, 2nd edn. (Mc Graw-Hill, Edo. de México, 1977), pp.
160–165
5. E. Hecht, Optics, Global edn., 5th edn. (Pearson Education Limited, Essex, 2017)
6. A. Beiser, Conceptos de Física Moderna, 2nd edn. (Mc Graw-Hill, Edo. de México, 1977),
Sect. 6.9-6.10
7. G. Herzberg, Molecular Spectra and Molecular Structure: Infrared and Raman Spectra of
Polyatomic Molecules, vol. 2 (Van Nostrand, New York, 1945)
8. G. Herzberg, Molecular Spectra and Molecular Structure: Electronic Spectra and Electronic
Structure of Polyatomic Molecules, vol. 3 (Van Nostrand, New York, 1966)
9. A. Szabo, N. Ostlund, Modern Quantum Chemistry (Dover, New York, 1996), Sect. 2.2.3
10. T. Helgaker, P. Jorgensen, J. Olsen, Molecular Electronic-Structure Theory (Wiley, Chichester,
2000)
11. S. Lang, Linear Algebra, 4th Printing (Addison-Wesley, London, 1969), Chapt. VI, Sect. 4, pp.
96–102
Part III
First-Principles Molecular Dynamics
Dynamics of Electrons and Nuclei
11
Quantum mechanics is a formulation of general nature created to describe the

microscopic world. It carries the attribute of quantum to distinguish it from classical
mechanics, which in principle is understood as a sub-branch of quantum mechanics.
At the microscopic scale, the particles have probabilistic manifestation, and it is
not possible to assign well-defined positions and momenta. On the other side, at
the macroscopic level, the particles have well-defined positions and momenta and
exhibit trajectories with time. The energy at the atomic scale is quantized, but
has continuous character at the macroscopic level. The quantum particles display
features like spin, non-local entanglement and can do tunneling through high energy
barriers, among other exotic features, but these features fade away at macroscopic
dimensions. The principles of quantum mechanics establish the behavior of the
particles and should be applied at the microscopic domain.1
There is a universal time-dependent wave equation in quantum mechanics that
should be solved to determine the wave function in terms of the time and, in turn,
obtain the time evolution of the system of particles. Yet, the time-dependent wave
equation is a complex equation that defies exact solution, especially for real systems.
In this respect, we need to introduce approximations to achieve the solution. The
purpose of this chapter is to present a relatively general and flexible method to
simplify the time-dependent wave equation of a system of atoms. We depart from
the fact that most complications in the process of solving the time-dependent wave
equation are attributed to the complexity of the Hamiltonian. Thus, we analyze the
nature of the molecular Hamiltonian and proceed to uncouple the dynamics of the
electrons from that of the nuclei by considering physical properties such as the
individual mobility of the particles. In this enterprise, we reach a time-dependent
wave equation for the electrons, and another for the nuclei. The equations maintain
a correlation between the two types of particles in an average manner, in principle
1 The words “macroscopic” and “microscopic” have relative meanings in our discussion, but they
shall be defined in the proper context later.
244 11 Dynamics of Electrons and Nuclei
facilitating the finding of the solution, which is now defined as the product of the
nuclear wave function and the electronic wave function.
11.1 The Electronic and Nuclear Dynamics Are Coupled
The behavior of the electrons is correlated with that of the nuclei due to the
crossed interactions between the two sets of particles. The crossed interactions lead
to the formation of atoms and molecules; however, they complicate the quantum
mechanical description of the particle system. In the next sections, we follow the
work of Tully and Marx on the treatment of the electrons and nuclei [1]. The
approximations to simplify the quantum mechanical description of the molecular
system permit a relative separation of the electron dynamics from the nuclear
dynamics, resulting in a time-dependent wave equation for the electrons and another
for the nuclei. The equations maintain an average correlation between the electrons
and nuclei, and their solutions determine the electronic wave function, .ψelec , and
the nuclear wave function, .ψnuc . Thus, the mathematical complexity in solving a
single wave equation for the two types of particles forming the molecular system
is reduced to two simpler equations. The wave function .w of the system is the
product .ψnuc ψelec , describing both types of particles at equal level of theory.
Depending on the molecular system and the conditions imposed, the approach
presented in this chapter may oversimplify the correlations between the electrons
and nuclei. Thereby, for strongly correlated systems, the wave equation that portrays
the behavior of the fully coupled system of particles must be solved. Currently, this
topic represents a challenge to understand the behavior of particles and is beyond
the scope of this work [2].
11.2 Molecular Hamiltonian
The time-dependent wave equation is the fundamental expression that rules the time
evolution of molecular systems. It is stated as a differential equation on the wave
function .w over time.
( )
∂w {ri }ni=1 ,{Rα }N
α=1 ;t
( )
. i h̄ ∂t = Ĥw {ri }ni=1 , {Rα }N
α=1 ; t
(11.1)
The .ri and .Rα symbols represent the electron and nuclear coordinates, and t is
the time. In principle, there is no need to differentiate between nuclear and electronic
coordinates, because they all have equal hierarchy; nevertheless, we anticipate such
a nomenclature for the next sections. The spin variables of electrons and nuclei may
be equally considered in the nomenclature of the coordinates (in this case, e.g., the
variable .ri also includes the spin of the ith particle). The Hamiltonian .Ĥ specifies
the number and types of particles participating in the system and introduces the
kinds of interactions among the particles. For the case of a molecule constituted by
electrons and nuclei subjected to no external forces, the non-relativistic Hamiltonian
11.2 Molecular Hamiltonian 245
contains the nuclear and electronic kinetic energies, and the electron-electron,
nucleus-electron, and nucleus-nucleus Coulombic interactions. By assuming n
electrons with mass .me and charge e, and N nuclei with mass .Mα and charge .Zα e,
the Hamiltonian terms in the previous specific order are:
wN h̄2 wn h̄2 wn w e2
Ĥ = − α=1 2Mα ∇α2 − i=1 2me ∇i2 + i=1 j >i rij
.
wN wn wN w (11.2)
Zα e 2 Zα Zβ
− α=1 i=1 riα + α=1 β>α Rαβ e2
The electron-electron, electron-nucleus, and nucleus-nucleus interparticle dis-

tances are:
. rij = |ri − rj | ; riα = |ri − Rα | ; Rαβ = |Rα − Rβ | (11.3)
The time-dependent wave equation 11.1 applies at any instant of time and
depends on the Hamiltonian because it contains the interactions that rule the particle
dynamics.
Problem 11.1 (i) Build the Hamiltonian of formaldehyde (.O = CH2 ) and
display the different sets of terms. (ii) Determine the number of Hamiltonian
terms of a protein with N nuclei and n electrons, when N is large. (iii)
A computer is capable to deal with .m2 particle interactions in a period of
time .AT . According to Moore’s law, every 18 months the computer power
is duplicated. Determine the number of particles that the next generation
computer may handle in the same period of time.
Solution
(i) The masses and electric charges of oxygen, carbon, and hydrogen carry
the subindexes O, C, and H . The Hamiltonian consists of many sets of
terms:
h̄2 h̄2 2 h̄2 h̄2 E h̄2 16

.Ĥ = − ∇O − ∇C − ∇H 1− ∇H 2− ∇2
2 2 2
2MO 2MC 2MH 1 2MH 2 2me i
i=1
E
16 E
16
e2 E
16
ZO e2 E
16
ZC e2 E
16
ZH 1 e2
+ − − −
rij |ri − RO | |ri − RC | |ri − RH 1 |
i=1 j >i i=1 i=1 i=1
(continued)
E
16
ZH 2 e2
−
|ri − RH 2 |
i=1
ZO ZC e2 ZO ZH 1 e2 ZO ZH 2 e2 ZC ZH 1 e2
+ + + +
ROC ROH 1 ROH 2 RCH 1
ZC ZH 2 e2 ZH 1 ZH 2 e2
+ +
RCH 2 RH 1H 2
(ii) The protein has N nuclei and n electrons. The respective Hamiltonian,
Eq. 11.2, has .N + n + n(n − 1)/2 + N(N − 1)/2 + nN = N (N + 1)/2 +
n(n + 1)/2 + nN terms. When N is large, n is also a big number, and
the number of terms in the Hamiltonian scales quadratically with the total
number of particles, .∼ (n + N)2 /2. This shows the complexity to obtain
.w, a function of the .3(n + N) spatial variables plus time.
(iii) We use the symbol .cpu1 for the present computer and .cpu2 for the next
generation computer. From Moore’s law, the power of .cpu2 is .2 cpu1 .
The number √ of2interactions calculated
√ in .AT with the new computer is
.2m = ( 2m) . If we take .n =
2 2m, then the number of particles that
the new computer may deal with is .n = 1.414 m. We require one further
generation, .cpu3 , to double the initial number of particles (.1.4142 = 2).
The efficiency can be improved with several computers working in parallel
(.cpu2A , cpu2B , · · · , cpu2N ) or employing novel computing hardware
(.gpuA , gpuB , · · · , gpuN ). The scaling on the number of particles with
computer power is more complex than our discussion.
11.3 Approximating the Total Wave Function
In this section we introduce a wave function that simplifies the time-dependent wave
equation of the molecule.
},{Rα };t)
. i h̄ ∂w({ri ∂t = Ĥ w ({ri }, {Rα }; t) (11.4)
w wn
The correlations between electrons and nuclei are due to the term .− N α i
Zα e2 /riα , which is present in the Hamiltonian as it accounts for the electron-nucleus
Coulomb interactions. However, the crossed Coulombic interactions complicate the
finding of the solution of the wave equation. It is convenient to separate the kinetic
energy terms from the potential energy terms in the Hamiltonian.
11.3 Approximating the Total Wave Function 247
wN h̄2 wn h̄2
Ĥ = − α=1 2Mα ∇α2 − i=1 2me ∇i2 + V̂ ({ri }, {Rα })
V̂ ({ri }, {Rα }) = V̂ee + V̂Ze + V̂ZZ

. wn w w wn (11.5)
e2 Zα e 2
V̂ee = i=1 j >i rij ; V̂Ze = − N α=1 i=1 riα
wN w Zα Zβ
V̂ZZ = α=1 β>α Rαβ e2
It is important to recall the form in which a wave function is evolved in quantum

mechanics with time, .w(t) = Û w(t0 ). The time evolution operator .Û takes one of
the forms:
⎧ [ ft ]
⎪
⎪ Tˆ exp − h̄i t0 dt ' Ĥ(t ' ) Ĥ does not commute at different times
⎪
⎪
⎨ [ ft ]
.Û (t, t0 ) = exp − h̄i t0 dt ' Ĥ(t ' ) Ĥ commutes at different times
⎪
⎪
⎪
⎪ [ ]
⎩ exp − h̄i (t − t0 ) Ĥ Ĥ is time independent
(11.6)
ˆ
.T is the time-ordering operator, and the Hamiltonian may or may not commute with
itself at different instants of time. The function .w(t) = Û w(t0 ) represents a solution
of the time-dependent wave equation when the time evolution operator .Û is known.
Nevertheless, it is of little help in our case due to the mathematical complications in
obtaining .Û .
Problem 11.2 Suppose that a Hamiltonian, .Ĥ, is constant during short time
intervals, .[t0 , t1 ], .[t1 , t2 ], .· · · , .[tn−1 , tn ], and commutes at different times,
.[Ĥ(ti ), Ĥ(tj )] = 0. Show that .Û (t, t0 )ψ(t0 ) satisfies the wave equation.
Solution The expression of .Û is given in terms of an integral of the

Hamiltonian over time. The integration from .t0 to .t = tn is computed by
intervals.
ft ' ft ft f tn
dt Ĥ(t ' ) = t01 dt ' Ĥ(t ' ) + t12 dt ' Ĥ(t ' ) + · · · + tn−1 dt ' Ĥ(t ' )
.
t 0 wn
= Ĥ(t1 )At1 + Ĥ(t2 )At2 + · · · + Ĥ(tn )Atn = i=1 Ĥ(ti )Ati
Where it was used .t = tn . The time evolution operator .Û adopts a new form
based on the constant nature of the Hamiltonian during short time intervals.
[ f ] [ ]
i E
t n
i
.Û (t, t0 ) = exp − dt ' Ĥ(t ' ) = exp − Ĥ(ti )Ati
h̄ t0 h̄
i=1
(continued)
We require to show that .i h̄ ∂[Û (t, t0 )ψ(t0 )]/∂t is equal to .Ĥ Û (t, t0 )ψ(t0 ).
We start by .Û (t, t0 ) in .i h̄ ∂[Û (t, t0 )ψ(t0 )]/∂t and recalling that
w inserting
.exp(x) = x k /k!.
k
∞
[ ]k ( )
∂ E 1 i E
n
i
.i h̄ Û (t, t0 )ψ(t0 ) = i h̄ − Ĥ(ti )Ati − Ĥ(t)ψ(t0 )
∂t k! h̄ h̄
k=0 i=1
The Hamiltonian evaluated at different times commutes with itself, and the
previous power series is again identified with an exponential function, then:
[ ]
i E
n
∂
.i h̄ Û (t, t0 )ψ(t0 ) = Ĥ(t) exp − Ĥ(ti )Ati ψ(t0 ) = Ĥ Û (t, t0 )ψ(t0 )
∂t h̄
i=1
Thereby, the expression .Û (t, t0 )ψ(t0 ) satisfies the wave equation.
A molecule consists of atoms, where an atom is understood in our description as

a particle formed by electrons and a nucleus, and such a nucleus is supposed to have
no internal structure. The total wave function is approximated by the product of an
electronic wave function, .ψelec , and a nuclear wave function, .ψnuc . This ansatz on
the wave function also considers an exponential function in terms of the energy, .E ' ,
and the time, t.
w ({ri }, {Rα }; t) = A(t) ψelec ({ri }; {Rα }, t) ψnuc ({Rα }; {ri }, t)

. [ f ( ) ] (11.7)
A(t) = exp h̄i t0 E ' t ' dt '
t
The electronic wave function .ψelec depends on the nuclear coordinates in

parametric form, and the nuclear wave function .ψnuc depends on the electron
coordinates in parametric form.2 The electronic and nuclear functions also include
the time. The amplitude A introduces an exponential in terms of .E ' and t, instead of
the Hamiltonian operator .Ĥ which is present in .Û , Eq. 11.6. Thus, the complications
2 The concepts of variable and parameter apparently play similar roles, but there is a subtle
difference between them in mathematics. A variable is an entity which can take different numerical
values after an equation is defined. On the other hand, the parameters take constant values in an
equation, yet they are not constants with perpetual values, like the number .π . A parameter defines
a family or class of equations, or equivalently, it is an entity that reduces a general equation to a
specific one. The usefulness of a parameter in an equation is related to the connection of variables,
allowing the exchange of information.
11.3 Approximating the Total Wave Function 249
related to the time evolution operator acting on the total wave function are avoided
with the incorporation of an exponential factor, which only considers average effects
through the introduction of .E ' .
f ∗ ψ ∗ Ĥ
. E ' (t) = ψnuc elec elec ψelec ψnuc dr dR
(11.8)
The symbols .dr and .dR abbreviate the products of the infinitesimally small
volumes .d 3 r1 · · · .d 3 rn and .d 3 R1 · · · d 3 RN , for the electronic, .ri , and nuclear
variables, .Rα . The expression of the energy .E ' (t) is originated from the electronic
Hamiltonian operator.
E
n
h̄2 2 E
n E 2
e E
N E
n
Zα e2
Ĥelec = −
. ∇i + −
2me rij rαi
i=1 i=1 j >i α=1 i=1
E
N E
Zα Zβ 2
+ e (11.9)
Rαβ
α=1 β>α
The nucleus-nucleus repulsion term is included in the electronic Hamiltonian

operator to maintain the electrostatic stability, because without the nuclear repul-
sions in the electronic Hamiltonian, the nuclei may coalesce. The electronic and
nuclear wave functions are mathematically well-behaved, as they are quadratically
integrable and normalized at every instant of time.
f ∗ ψ
f ∗ ψ
. ψelec elec dr = 1 ; ψnuc nuc dR = 1 (11.10)
After inserting the total wave function .w = A ψelec ψnuc into the expression
i h̄ ∂w/∂t = Ĥw, we reach an equation which displays the electronic and nuclear
.
functions.
( )
∂A ∂ψelec ∂ψnuc
.i h̄ ψelec ψnuc + Aψnuc + Aψelec
∂t ∂t ∂t
E
N
h̄2
= −A ψelec ∇ 2 ψnuc + A Ĥelec ψelec ψnuc
2Mα α
α=1
∗ ψ ∗ , inte-
Multiplication from the left by the complex conjugate functions .ψelec nuc
grating over all electronic and nuclear coordinates, making use of the normalization
constraints, leads to:
( f f )
∂A ∗ ∂ψelec ∗ ∂ψnuc
.i h̄ + A ψelec dr + A ψnuc dR
∂t ∂t ∂t
E
N f
h̄2 ∗
= −A ψnuc ∇α2 ψnuc dR
2Mα
α=1
E
n f f
h̄2 ∗ ∗ ∗
−A ψelec ∇i2 ψelec dr + A ψelec ψnuc V̂ ψelec ψnuc dr dR
2me
i=1
where the terms building the electronic Hamiltonian are displayed. It is important to
focus attention on the energy contribution from the potential .V̂ , in particular on .V̂Ze ,
Eq. 11.5, as it is the term that introduces the coupling between electrons and nuclei
(it is also the term responsible for coupling the fast and slow particles in the Born-
Oppenheimer approximation). Due to the exponential form, the time derivative of
A is proportional to itself.
. ∂A/∂t = (i/h̄)AE ' (11.11)
The insertion of this result and the expression of .E ' in terms of the electronic and
nuclear functions in the wave equation, let us avoid the use of the amplitude A.
(f f )
∗ ∂ψelec ∗ ∂ψnuc
i h̄
. ψelec dr + ψnuc dR
∂t ∂t
f E
N
∗ ∗ h̄2
= ψnuc ψelec Ĥelec ψelec ψnuc dr dR −
2Mα
α=1
f E
n f
∗ h̄2 ∗
∗ ψnuc ∇α2 ψnuc dR − ψelec ∇i2 ψelec dr
2me
i=1
f
∗ ∗
+ ψelec ψnuc V̂ ψnuc ψelec dr dR (11.12)
This equation is equivalent to the total wave equation because it has been only
rewritten in terms of .ψelec and .ψnuc . Still, it is difficult to solve due to the terms
mixing the electronic and nuclear degrees of freedom.
11.4 Time-Dependent Self-Consistent Field Equations
The next step consists on deriving an equation for the electrons and another for
the nuclei. To do this, it is important to introduce the constraints that guarantee the
solutions of the equations. In addition to the normalization constraints, Eqs. 11.10,
which involve spatial variables, the new constraints should demand well-behaved
functions with time. In particular, we should ensure that integrations of products
∗ ∂ψ
like .ψelec ∗
elec /∂t and .ψnuc ∂ψnuc /∂t are not undetermined.
11.4 Time-Dependent Self-Consistent Field Equations 251
f ∗ f ∗
. i h̄ ψelec ∂ψelec
∂t dr = E ' ; i h̄ ψnuc ∂ψnuc
∂t dR = E (11.13)
The quantities .E ' and E have dimensions of energy. It is necessary to take the
physical attributes of the particles into consideration to achieve the separation of
equations. In most common processes, the nuclei are the heavy-weight and slow-
moving particles, and the electrons are the light-weight and fast-moving particles.
The electrons and nuclei exhibit different physical properties and time scales in
their motions, giving basis to the proposal of a ruling equation for the electrons and
another for the nuclei.3
In order to establish the time-dependent wave equations that determine the
electronic and nuclear wave functions, we consider Eq. 11.12 and compare the terms
of the left side with the terms of the right side. The resulting equations are:
wn [f ]
h̄2 ∗ V̂ ψ
i h̄ ∂ψ∂telec = − i=1 2me ∇i2 ψelec + ψnuc nuc dR ψelec
.
wN [f ] (11.14)
h̄2 ∗ Ĥ
i h̄ ∂ψ∂tnuc = − α=1 2Mα ∇α2 ψnuc + ψelec elec ψelec dr ψnuc
The first expression is the wave equation that regulates the dynamics of the
electrons, and the second one is the wave equation that regulates the dynamics of the
nuclei. The electrons and nuclei have equal hierarchy as the electronic and nuclear
wave functions are shown in symmetric forms in both equations and are treated
at the same level of quantum theory. The presence of .ψnuc in the electronic wave
equation, and the presence of .ψelec in the nuclear wave equation, is the manner in
which such equations maintain the correlation between the different sets of particles.
According to such expressions, the electrons do not interact with any nucleus in
particular, and the nuclei do not interact with any electron in particular. However,
the electrons are understood to be immersed in an effective field or average potential
created by the nuclei and, in turn, the nuclei are immersed in an effective field
created by the electrons.
After introducing the concept of an effective field, the time-dependent wave
equations are simplified by using the effective Hamiltonians.
eff wn h̄2
f ∗ V̂ ({r }, {R }) ψ
Ĥelec ({ri }; t) = − i=1 2me ∇i2 + ψnuc i α nuc dR
.
eff wN h̄2
f ∗ Ĥ
Ĥnuc ({Rα }; t) = − α=1 2Mα ∇α2 + ψelec elec ({ri }; {Rα }) ψelec dr
(11.15)
3 The separation of variables discussed in the text is not a common separation of the independent
degrees of freedom, because the position variables of the electrons and nuclei are equally
important, all of them have equal priority, and satisfy the same conditions. The separation of
variables proposed in the text is based on the phenomenological features of the particles and are
similar to those employed in the Born-Oppenheimer approach, with the difference that time is not
present in the Born-Oppenheimer approach.
Both Hamiltonians change with time by means of the wave functions. The time-
dependent wave equations of the electrons and nuclei are simplified by taking the
notation of the average potentials into consideration.
eff eff
. i h̄ ∂ψ∂telec = Ĥelec ({ri }; t) ψelec ; i h̄ ∂ψ∂tnuc = Ĥnuc ({Rα }; t) ψnuc
(11.16)
These equations are difficult to solve analytically due to the particle correlations
and, thereby, have to be solved numerically in a self-consistent way. For example,
an initial nuclear wave function is proposed with the purpose of building the
average potential acting on the electrons. Thus, the electronic wave equation may be
numerically solved to obtain a first approximation of the electronic wave function.
Next, the electronic wave function is inserted into the nuclear wave equation for
building the average potential acting on the nuclei. Now, the nuclear wave equation
may be numerically solved to obtain an approximate nuclear wave function. The
required feedback demands a self-consistent iteration process to improve the wave
functions at every step. The self-consistent process is terminated when the electronic
and nuclear wave functions and their energies are accurate and show no significant
changes from one cycle to the next one of the iterative process. Due to computational
efficiency, one or another wave function may be favored to start the process. Still,
the final wave functions are independent of the initial function that was used to
start the self-consistent process. The expressions 11.16 are recognized as the time-
dependent self-consistent field equations. The equations satisfy the normalization
∗ ∂ψ
conditions 11.13 imposed on the products .ψelec ∗
elec /∂t and .ψnuc ∂ψnuc /∂t, which
'
lead to the energies .E and E, respectively.
Problem 11.3 (i) Show that the plane wave .ψ(x, t) = A exp[i(kx −
wt)] reaches a finite value when the kernel .ψ ∗ ∂ψ/∂t is integrated. (ii)
Demonstrate that the wave equations of electrons and nuclei, Eqs. 11.14, are
consistent with the constraints 11.13 imposed on the functions .ψelec and .ψnuc ,
resulting in the energies .E ' and E, respectively.
Solution
fb fb
(i) The required integral is . a (ψ ∗ ∂ψ/∂t)dx = AA∗ a e−i(kx−wt)
.(−iwt) e
i(kx−wt) dx. When .a = −∞ and .b = ∞, the integral is
undetermined; however, if the integration limits are finite (.a = −l and

.b = l), the integral achieves a finite value. Thus, the function .ψ acquires
physical meaning for finite integration limits.
(continued)
References 253
∗
(ii) If the first expression of Eqs. 11.14 is multiplied from the left by .ψelec
and integrated, then we obtain:
f f E
n
∗ ∂ψelec h̄2 ∗
i h̄
. ψelec dr = − dr ψ ∇ 2 ψelec
∂t 2me elec i
i=1
f
∗ ∗
+ ψelec ψnuc V̂ ψnuc ψelec dr dR
If we use the orthonormality condition 11.10 of the nuclear wave function

∗ ψ
to introduce both the product .ψnuc nuc on the first term of the right and the
electronic Hamiltonian 11.9, then the energy .E ' of Eqs. 11.13 is obtained.
f f
∗ ∂ψelec ∗ ∗
i h̄
. ψelec dr = ψelec ψnuc Ĥelec ψelec ψnuc dr dR = E ' (t)
∂t
The nuclear equation is symmetric to the electronic equation, and therefore, a

similar process on the nuclei leads to the energy E.
f f
∗ ∂ψnuc ∗ ∗
i h̄
. ψnuc dR = ψnuc ψelec Ĥ ψelec ψnuc dr dR = E
∂t
Interestingly, we started from a single wave equation in the overall simplification

process and ended up with two correlated wave equations, apparently doubling the
difficulty in obtaining the final solution of the total wave equation of the molecular
system. Nevertheless, the pair of equations are in principle easier to solve than
the initial one due to the average potentials in which the electrons and nuclei are
immersed. Finally, the wave functions may be propagated with time by building the
time evolution operators .Ûnuc and .Ûelec from the effective Hamiltonians. Yet, we
avoid such operators in later sections by further simplifying the wave equations and
determining the time evolution of the electrons and nuclei.
References
1. J.C. Tully, Mixed Quantum-Classical Dynamics. Faraday Discuss. 110, 407–419, 1998. D.
Marx, J. Hutter, Ab initio molecular dynamics: theory and implementation, in Proceedings of
the Modern Methods and Algorithms of Quantum Chemistry, ed. by J. Grotendorst. Jülich, NIC
Series, 2nd edn., vol. 3 (John von Neumann Institute for Computing, 2000), pp. 329–477
2. P. Fulde, Electron Correlation in Molecules and Solids. Springer Series in Solid-Sate Sciences,
2nd edn. (Springer, Berlin, 1993)
Classical Limit of the Nuclear Motion
12
In going from the microscopic world to the macroscopic one, it is possible to have
a smooth transition from quantum mechanics to classical mechanics in different
ways. According to the Bohr’s correspondence principle, there is a connection
between quantum theory and classical theory. Such a principle demands large
quantum numbers to describe the particles as classical objects. The corresponding
principle has been given different interpretations, yet it is difficult to have practical
use of it due to the requirement of establishing the degree of quantization of the
particles [1]. There is an equivalent way to determine a connection between quantum
mechanics and classical mechanics. It is referred to as the classical limit. After
perceiving no major obstacles in the transition from quantum mechanics to classical
mechanics, the classical realm is reached from quantum mechanics by introducing
the observations of the macroscopic phenomena as limiting mathematical conditions
in the quantum-mechanical equations.
The quantum nature of the nuclei is important in processes such as the nuclear
dispersion, the nuclear spin transitions, the tunnel effect to traverse high energy
barriers, and in other phenomena. These all processes can be solely studied from
the quantum-mechanical perspective. However, as we have observed in previous
sections, the application of quantum mechanics is difficult in many instances. For
example, we face problems to apply the quantum-mechanical theory to the nuclei
of polyatomic molecules due to the many degrees of freedom required to describe
these molecules. The vast demand of efforts and computational resources implied
in an approach that considers electrons and nuclei at the same level of quantum-
mechanical theory is only justified when a precise and detailed picture of the particle
system is crucial. On the other side, the approximation of the nuclei as classical
particles may be used to simulate the dynamics of particles in experiments involving,
for example, low-energy collisions, temperature effects, chemical reactions, and
more, where the quantum-mechanical nature of the nuclei is less relevant. Once the
nuclei are described classically, their quantum character is lost in the description.
The aim of the present chapter is to provide a classical description of the
nuclei while maintaining the quantum character of the electrons. In the following
256 12 Classical Limit of the Nuclear Motion
sections, we reformulate the wave equation of the nuclei and migrate the quantum
observables to classical variables to reach a description in terms of classical theory.
The discussion exhibits a hierarchy of approximations, starting from a quantum
mechanical treatment of the nuclei and ending up on a classical one, because there
are phenomena where the nuclei act as quantum particles, or the quantum nature of
the nuclei plays a second role, favoring the nuclei to act as classical objects. In these
last cases, a description of the nuclear particles based on analytical mechanics is
possible, rendering results that are close to those obtained with quantum mechanics
but with a lower computational cost.
The chapter follows the main ideas of Tully [2], and is structured as follows.
We depart from the wave equation of the electrons and the wave equation of the
nuclei of the previous chapter. Such equations deal with the electrons and nuclei
at the same quantum-mechanical level of theory, maintaining a good degree of
correlation among the particles. However, with the purpose of further reducing the
complexity of the expressions, we turn the attention to the nuclear wave function.
By assuming a polar form for that function, we are able to simplify the nuclear
potential and make contact with classical mechanics, where the nuclear particles
become classical bodies with motions governed by Newtonian mechanics. This
gives formal birth to mixed quantum-classical dynamics. Later, by parameterizing
the atomic interactions, a full classical description of the atoms is achieved, thus
establishing classical molecular dynamics. By dealing with the classical particles,
we are able to establish an equation of continuity and discuss the diffusion and flux
of the classical particles by deriving the first and second Fick’s laws, and linking the
results to the chemical potential and the diffusion constant. On the other side, the
electronic wave function is expanded employing a complete basis set of functions,
with implications on the equations of the nuclei, since all the particles maintain a
degree of correlation. Finally, a section is dedicated to the Bohm potential, which is
one of the important intermediate results in the transition from quantum mechanics
to classical mechanics. Based on the Bohm potential, a discussion is oriented to an
alternative interpretation of quantum mechanics.
12.1 Polar Form of the Nuclear Wave Equation
In order to simplify the description of the nuclei in polyatomic molecules, we

depart from the time-dependent wave equation of the nuclei and the time-dependent
wave equation of the electrons, Eqs. 11.15 and 11.16. In the time-dependent self-
consistent field method, the nuclei and electrons are described at the same level of
theory by their respective wave equations.
eff eff
. i h̄ ∂ψ∂tnuc = Ĥnuc ({Rα }; t) ψnuc ; i h̄ ∂ψ∂telec = Ĥelec ({ri }; t) ψelec (12.1)
12.1 Polar Form of the Nuclear Wave Equation 257
E
N f
eff h̄2 ∗
.Ĥnuc ({Rα }; t) =− ∇ + ψelec ({ri }; t)
2
2Mα α
α=1
× Ĥelec ({ri }, {Rα }) ψelec ({ri }; t) dr

En f
eff h̄2 2 ∗
Ĥelec ({ri }; t) = − ∇i + ψnuc ({Rα }; t)
2me
i=1
× V̂ ({ri }, {Rα }) ψnuc ({Rα }; t) dR (12.2)
The wave functions of nuclei and electrons are .ψnuc and .ψelec . The effective
Hamiltonians involve the kinetic and potential energies of the corresponding parti-
cles. The potential .V̂ and Hamiltonian .Ĥelec are shown in the kernels of the integrals
and are given by Eqs. 11.5 and 11.9. The averaged potentials where the nuclei and
electrons are immersed may be written in bra-ket notation .<ψelec |Ĥelec |ψelec > and
.<ψnuc |V̂ |ψnuc >, respectively.
f ∗
<ψelec |Ĥelec |ψelec > = ψelec ({ri }; t) Ĥelec ({ri }, {Rα }) ψelec ({ri }; t) dr
.
f ∗
<ψnuc |V̂ |ψnuc > = ψnuc ({Rα }; t) V̂ ({ri }, {Rα }) ψnuc ({Rα }; t) dR
In the simplified description of the nuclei, the quantum character of the electrons
is retained because it is of primary importance. Following similar steps in the
transition from wave optics to geometrical optics, this is, from waves to rays, a
function of polar type for the nuclei is proposed.
. ψnuc ({Rα }; t) = D ({Rα }; t) exp [iS ({Rα }; t) / h̄] (12.3)
This equation is recognized as the action wave, Eq. 6.64. The factor D represents
the amplitude and the term S is to be identified. Both functions are real and depend
on the nuclear coordinates .Rα and time t. The introduction of this expression in the
nuclear wave equation produces an identity involving both functions:
E h̄2 [ ]
N
∂
i h̄
. D+ 2 ∇α (iS/h̄) · ∇α D + D ∇α2 (iS/h̄)
∂t 2Mα
α=1
EN [ ]
∂ h̄2 D
=D S− − 2 |∇α S| + ∇α D + D Velec
2 2
(12.4)
∂t 2Mα h̄
α=1
where it was convenient to introduce the function .Velec to simplify the above
expression.
f ∗ ({r }; t) Ĥ
. Velec = ψelec i elec ψelec ({ri }; t) dr (12.5)
When the separation of the real and imaginary parts of Eq. 12.4 is performed,
two coupled expressions which contain the functions D and S are obtained.
EN EN
∂D
∂t + 1
α=1 Mα ∇α D · ∇α S + 1
α=1 2Mα D ∇α2 S = 0
.
EN f ∗ Ĥ
EN h̄2 ∇α D
2
∂S
∂t + α=1 2Mα |∇α S|
1 2 + ψelec elec ψelec dr − α=1 2Mα D =0
(12.6)
Problem 12.1 Deduce the expression 12.4 of the main text.
Solution The time-dependent wave equation of the nuclei is:

eff
. i h̄ ∂ψ∂tnuc = Ĥnuc ψnuc
eff
When the polar form of .ψnuc and the form of .Ĥnuc are inserted in this
equation, an expression in terms of the functions D and S is obtained.
[ ] [ E h̄2
f ∗ Ĥ
]
i h̄ ∂t∂ D exp(iS/h̄) = − Nα=1 2Mα ∇α2 + ψelec elec ψelec dr
.
×D exp(iS/h̄)
The divergence of a continuously differentiable vector field .F = F1 i + F2 j +

F3 k is defined as .∇ · F = ∂F1 /∂x + ∂F2 /∂y + ∂F3 /∂z, while the Laplacian
.∇ is the divergence of the gradient or .∇ = ∇ · ∇. In this connection, the
2 2
time and spatial differentiations give:

{ } E
N
∂ ∂ h̄2 (iS/h̄)
i h̄ e
.
(iS/h̄)
D+D iS/h̄ = − e [∇α (iS/h̄) · ∇α D
∂t ∂t 2Mα
α=1
+ ∇α2 D + ∇α D · ∇α (iS/h̄) + D∇α (iS/h̄) · ∇α (iS/h̄)

]
+D∇α2 (iS/h̄) + De(iS/h̄) Velec
where Eq. 12.5 was used. The exponential function represents a common
factor and is dismissed.
{∂ [} E h̄2
i h̄ ∂t D+D ∂
∂t iS/h̄ = − N
∇α (iS/h̄) · ∇α D + ∇α2 D
α=1 2Mα
.
]
+ ∇α D · ∇α (iS/h̄) + D ∇α (iS/h̄) · ∇α (iS/h̄) + D ∇α2 (iS/h̄) + D Velec
(continued)
12.2 Continuity and Hamilton-Jacobi Equations 259
The real and imaginary terms are separated from each other, thus reproducing
Eq. 12.4. The resulting expression is only possible when each member of
Eq. 12.4 equals zero. Thus, an equation for D and another for S is obtained,
leading to Eqs. 12.6.
12.2 Continuity and Hamilton-Jacobi Equations
In the following lines, the first identity of Eq. 12.6 is shown to be equivalent to
an equation of continuity, with no source and sink terms, thus establishing the
conservation of the probability, and the second identity of Eq. 12.6 is shown to lead,
under specific conditions, to a Hamilton-Jacobi classical equation. In order to unfold
the first demonstration, let us multiply the nuclear wave equation and its complex
∗ and .−ψ
conjugate by .ψnuc nuc , respectively, adding the two resulting expressions to
have a single one.
∗ ψ
∂(ψnuc nuc ) ∗ Ĥ
= ψnuc
eff ∗
eff
. i h̄ ∂t nuc ψnuc − ψnuc Ĥnuc ψnuc
The equation is simplified when the nuclear Hamiltonian consists of a kinetic

energy operator and a potential energy operator. The latter operator contributes
nothing to the equation because, contrary to the kinetic energy operator that involves
differentiation over the wave function, the potential energy acts as a multiplicative
factor affecting no wave function, thereby cancelling the difference.
∗ ψ
∂(ψnuc nuc )
EN h̄2
[ ∗ ∗
]
. i h̄ ∂t =− α=1 2Mα ψnuc ∇α2 ψnuc − ψnuc ∇α2 ψnuc
With the purpose of simplifying the equation, we define the probability density
ρ and probability flux .J, in terms of .ψnuc and .∇α ψnuc .
.
∗ ψ
ρ (R; t) = ψnuc nuc
. EN [ ∗ ∗
] (12.7)
J (R; t) = α=1 2iMα ψnuc ∇α ψnuc − ψnuc ∇α ψnuc
h̄
The functions .ρ and .J depend on the variable .R as these functions represent

probability densities. The direction of the .J vector is defined by the difference of the
gradients on the wave functions. .J is responsible for producing probability currents,
in analogous manner to a current of electrons in a metal [3]. The equation that
establishes the conservation of the probability is now stated in terms of the functions
.ρ and .J.
∂ρ(R;t)
.
∂t + ∇ · J (R; t) = 0 (12.8)
Such an expression is commonly referred to as the equation of continuity. On the

other hand, the insertion of .ψnuc = D exp(iS/h̄), Eq. 12.3, in the definitions of .ρ
and .J makes these variables to show dependencies on D and S.
E
N
1
ρ(R; t) = D 2
. ; J (R; t) = D 2 ∇α S
Mα
α=1
When they are introduced in the continuity equation 12.8, the differential equa-
tion in the amplitude D, Eq. 12.6, is obtained, thus demonstrating the equivalence
between equations that apparently look different.
EN EN
∂D
∂t + 1
α=1 Mα ∇α D · ∇α S + 1
α=1 2Mα D ∇α2 S = 0
. | (12.9)
∂ρ(R;t)
∂t + ∇ · J (R; t) = 0
It is time to work with the differential equation in the function S, Eq. 12.6, and
show that under limiting conditions it leads to the second Newton’s law. To do this,
we consider the term containing Planck’s constant, .h = 2π h̄. It is a reference
parameter that determines the size of quanta. When the term of an expression
carrying h is substantially smaller than the other terms of the expression carrying
no h, we conclude that we are in the classical realm. Such a situation is frequently
abbreviated in the form .h → 0 (it is a symbolic form because h is a constant and it
never goes to zero and is similar to the transition from relativistic to non-relativistic
mechanics, when we write .c → ∞). The term that carries h in Eq. 12.6 is the Bohm
quantum potential, defined by:
EN h̄2
. Vbohm = − α=1 2Mα ∇α2 D / D (12.10)
This is an important quantity that deserves special attention, Sect. A.1, because
the quantum potential is capable to take us to an interpretation of quantum
mechanics different to the traditional one. By assuming that .Vbohm is sufficiently
small with respect to the other terms present in Eq. 12.6, we step into a classical
description framework.
∂S({Rα };t) EN f ∗ Ĥ
.
∂t + α=1 2Mα |∇α S({Rα }; t)|
1 2 + ψelec elec ({ri }, {Rα }) ψelec dr = 0
(12.11)
This result is isomorphic to the Hamilton-Jacobi equation obtained in classical

mechanics.
12.2 Continuity and Hamilton-Jacobi Equations 261
∂S({Rα };t)
.
∂t + H ({Rα }, {Pα }) = 0 (12.12)
The expression 12.12 is the first-order differential equation of the classical action
function, S, Eq. 6.49, which in turn is defined in terms of the Lagrangian, L [4].
f
. S= L ({Rα }, {Pα }; t) dt (12.13)
The last two terms of 12.11 constitute the classical Hamiltonian, .H, defined in
terms of the kinetic and potential energy contributions.1
E
N f
1 ∗
. (∇α S)2 + ψelec Ĥelec ψelec dr
2Mα
α=1
−→ H = T ({Pα }) + Velec ({Rα }; t) (12.14)

h̄→0
The projection of the terms on the top onto the terms on the bottom is done
under physical basis: the integral represents the potential .Velec , Eq. 12.5, created
by the electron cloud where the nuclei are immersed, and the term involving the
summation represents the kinetic energy T of the nuclei. The kinetic energy is
a quadratic function of the momentum .P and is identified in terms of the action
function. Thus, the classical terms are defined by:
f ∗ ({r }; t) Ĥ
Velec ({Rα }; t) = ψelec i elec ({ri }, {Rα }) ψelec ({ri }; t) dr
. (12.15)
Pα = ∇α S
We now follow the ideas of the Hamilton-Jacobi theory to link the momentum to
the potential energy. By applying .∇α to each term in .∂S/∂t + H = 0, we obtain:
dPα
.
dt = −∇α Velec ({Rα }; t) (12.16)
The change rate of the momentum is due to a gradient of the potential .Velec
created by the electrons and the other nuclei. Thus, the particle evolves according
to the second Newton’s law with time. Concluding, it has been demonstrated the
relation between the following equations:
EN f ∗ Ĥ
EN h̄2 ∇α D
2
∂S
∂t + 1
α=1 2Mα |∇α S|2 + ψelec elec ψelec dr − α=1 2Mα D =0
. ⇓ h̄ → 0
dPα
dt = −∇α Velec ({Rα }; t)
(12.17)
1 We discriminate terms by following a similar procedure as that of the W KB method, where an
expansion in terms of .h̄ is performed.

Due to the fact that the particles have wave-particle duality, there is not a concept
of force at the quantum-mechanical level. Still, the force is a consequence of
quantum theory by reaching after special considerations the classical limit. In such
a limit, the Newtonian force is associated with the action of a field on a particle
that shows a sufficiently localized mass in space. In the second Newton’s law, the
force is defined implicitly as that entity that produces the change of inertia of a
particle with mass, giving no explanation on the origin of the force. Therefore, in
the Newtonian framework, the force is manifested by its action on a particle with
sufficiently localized mass.
Problem 12.2 A particle with mass m and electric charge e is subjected to a

potential V and a magnetic field, which is described by the vector potential
.A. The vector potential satisfies the equation .∇ · A = 0. The wave equation
of the particle is:

( )2
h̄2 ie ∂ψ
. − ∇− A ψ + V ψ = i h̄
2m h̄c ∂t
A solution is proposed by expanding S as a power series.

( ) ( )2
h̄ h̄
.S = S0 + S1 + S2 + · · ·
i i
Show that .ψ = exp(iS/h̄) is an exact solution of the wave equation when S

satisfies the Hamilton-Jacobi equation.
Solution The wave function .ψ = exp(iS/h̄) is supposed to be an exact

solution, and it is inserted into the wave equation. The result, using .∇ · A = 0,
is an equation in the function S.
1 [ e e ] i h̄ 2 ∂S
. (∇S)2 + ( )2 A · A − A · ∇S − ∇ S+V + =0
2m c c 2m ∂t
By inserting the power series of S and using the Laplace equation, the above
expression leads to a system of equations, which are defined after comparing
the coefficients with the same power of .h̄ in both sides of the equation. The
expression of interest is the zero-order term .S0 , as it is a Hamilton-Jacobi
equation in the function .S0 .
∂S0 1 [ e ]2
. + ∇S0 − A + V = 0
∂t 2m c
(continued)
12.3 Conditions to Describe the Nuclear Particles Classically 263
Interestingly, there are forms of defining the potential V with dependence

on the coordinates and time, in such a way that a solution S of the Hamilton-
Jacobi equation gives an exact solution, .exp(iS/h̄), of the corresponding wave
equation [5]. Similarly, there are forms of defining the refractive index, in
such a way that a solution S of the eikonal equation gives an exact solution,
.exp(ik0 S), of the corresponding Helmholtz equation [6].
It is convenient to summarize the main equations on the dynamics of the nuclei

and electrons.
dPα f ∗
dt = −∇α ψelec ({ri }; t) Ĥelec ({ri }, {Rα }) ψelec ({ri }; t) dr
.
E [f ]
h̄2 ∗ ({R }; t) V̂ ψ
i h̄ ∂ψ∂telec = − ni=1 2m e
∇i2 ψelec + ψnuc α nuc ({Rα }; t) dR ψelec
(12.18)
Care should be exercised in not confusing the potential .Velec , Eq. 12.15, with the
potential .V̂ = V̂ee + V̂Ze + V̂ZZ , Eq. 11.5, which contains operators without being
averaged. The equation in the first line describes from a classical point of view the
behavior of the nuclei immersed in an effective potential created by the electrons.
Instead, the equation in the second line describes the behavior of the electrons
immersed in an effective potential created by the nuclei. In this last expression,
the wave function of the nuclei is used, in spite of the fact that the nuclei are
described classically in the first expression. In this regard, Eq. 12.18 is considered
a transitional expression in going from a quantum-mechanical description toward
a classical-mechanical description of the nuclei. In order to be consistent with a
description of the nuclei as classical particles, we must avoid the use of the nuclear
wave function in the final equations. This is done later.
12.3 Conditions to Describe the Nuclear Particles Classically
In the previous section, the classical limit was achieved by considering .h̄ → 0. Let
us now be more specific on determining the physical conditions to reach the classical
limit. When the objects are sufficiently massive or the particles increase in size, their
quantum numbers are big, and the probabilistic outputs of quantum mechanics give
a single and definite outcome. In this limit we are in the domain where the results
obtained from quantum mechanics correspond to those of the classical world. Since
the procedure to reach classical mechanics from quantum mechanics follows similar
steps to the procedure where geometrical optics is reached from wave optics, it is
convenient to briefly document this last procedure [7].
When the Hamiltonian of the system is time indifferent, Hamilton’s principal

function, S, of classical mechanics is written in terms of both Hamilton’s charac-
teristic function, W , which depends on the conjugate variables .{qi , pi }, and a term
which is linearly proportional to the time, Eq. 6.53.
. S({qi , pi ; t}) = W ({qi , pi }) − αt
In principle, it is a consequence of considering the wave function of quantum

mechanics in a stationary state, Eq. 10.36:
. ψnuc ({Rα }, t) = ψnuc ({Rα }) exp(−iEt/h̄)
where the coordinates and time are separated into a pair of functions. The substitu-
tion of S in the Hamilton-Jacobi equation defines the value of .α, this is, .α = H = E,
where E is the integration constant associated with the energy. The separation of the
conjugate variables from the time in the expression of S makes the Hamilton-Jacobi
approach useful: if the value of the Hamilton’s characteristic function at .t = 0 is
.W0 , then the corresponding value of S is .S(0) = W0 . Suppose that we work with the
value .S(0). At a later instant of time, say .ti , the value of W is .Wti = W0 + Eti , and
thus, the value of S is preserved, .S(ti ) = Wti −αti = [W0 +Eti ]−αti = W0 = S(0).
We have a surface S with fixed value that moves from the surface .W0 at the instant
of time .t = 0 to the surface .Wti at the instant of time .ti . It is visualized as a wave
front propagating in phase space with a velocity computed from the properties of the
particles under consideration. The velocity can be shown to point perpendicularly
to the direction of the wave front. The surface with constant value of S satisfies a
wave equation. A procedure like this characterizes the transition from wave optics
to geometrical (ray) optics establishing, among other things, the eikonal equation
of geometrical optics, with ray trajectories pointing in the perpendicular direction
to the wave fronts, and giving foundations to Fermat’s principle and the Huygens
propagation of wave fronts.
Now, we start from Eq. 12.4, and assume the separation of positions and time in
the action function, .S = σ ({Rα })−Et, where .σ is a function of the positions solely.
It is also supposed that the amplitude D of Eq. 12.4 is a slowly varying function of
the positions and time. In this last case, the spatial and time derivatives of D are
negligible, and Eq. 12.4 is simplified.
∂ E 1
N E i h̄ N
. S+ |∇α S|2 − ∇ 2 S + Velec = 0
∂t 2Mα 2Mα α
α=1 α=1
It is time to introduce the expression of the action function in the above equation.
EN EN
.
1
α=1 2Mα |∇α σ |2 + Velec − i h̄
α=1 2Mα ∇α2 σ = E (12.19)
12.3 Conditions to Describe the Nuclear Particles Classically 265
If we use the power series expansion .σ = σ0 + (h̄/i)σ1 + (h̄/i)2 σ2 + · · · , which is a

consequence of writing the wave function of a particle system in the stationary state,
.ψ({Rα }, t) = exp(iσ {Rα }/h̄), problem 12.2, then we recover the classical equation
of the energy. This process is equivalent to neglecting the term of Eq. 12.19 that
contains .h̄, because it is sufficiently small with respect to the other terms containing
no .h̄.
E
N
1
|∇α σ0 |2 >> h̄|∇α2 σ0 | = h̄|∇α · ∇α σ0 |
. ⇒ |∇α σ0 |2 + Velec = E
2Mα
α=1
By identifying .Pα with .∇α S = ∇α σ0 , Eq. 12.15, the above expressions are
written in terms of the momentum.
EN
. P2α >> h̄ |∇α · Pα | ⇒ α=1 Pα /(2Mα ) + Velec
2 =E (12.20)
The first expression is equivalent to .1 >> h̄|∇α · Pα |/P2α . The second expression is
consistent with the Hamiltonian of Eq. 12.14. When the .∇ operator is applied to the
expression of the energy E, Eq. 12.20, we deduce:
. h̄Mα |∂Velec /∂Xα | << Pα3 (12.21)
Therefore, the classical limit of the quantum motion of a particle is achieved for
large values of the particle momentum with respect to the gradient of the potential
.Velec . An alternative form of establishing the classical behavior is by considering
an action, say .S0 , where .h̄ << S0 (as discussed in Sect. 12.2, where .h → 0 was
employed, Eq. 12.17).
Let us now deal with a single particle in one dimension and use de Broglie’s
relation of the momentum with the wavelength, .P = 2π h̄/λ, to transform the
inequality 12.21 written in terms of the particle momentum to an expression in terms
of the wavelength .λ. By differentiating the de Broglie’s relation with respect to the
particle position, we obtain:
. ∂Pα /∂Xα = −(2π h̄/λ2 ) ∂λ/∂Xα
It is deduced that .|∂Pα /∂Xα |/Pα2 = 1/(2π h̄)(∂λ/∂Xα ). Since .1 >> h̄|∇α ·
Pα |/P2α , the quantity .∂λ/∂Xα is found to be very small.
(2π )−1 ∂λ/∂Xα << 1

.
If .∂λ/∂Xα is approximated by .λ/L, with L a characteristic dimension of the

system, then an alternative condition to have a classical description is:
. λ << L (12.22)
The degree of quantum character of the system is equally determined from the
de Broglie thermal wavelength, .λT .
. λT = h/(2π mkB T )1/2 (12.23)
The factors h and .kB are the Planck and Boltzmann constants, and T is the
temperature. By following similar steps in the transition from quantum optics to ray
optics, when .λT is smaller than the characteristic spatial variations of the interaction
potential between particles, the condition of classicality on the particle system is
observed [8].
Problem 12.3 (i) Suppose a classical harmonic oscillator with mass m,

spring stiffness k, and angular vibrational frequency .ω. The energy .Ec is
given by:
.Ec = kx 2 /2 ; ω2 = k/m
On the other hand, the energy of a quantum oscillator is .Eq = (n+1/2)h̄ω,

where n is the quantum number (.n = 0, 1, · · · ), .h̄ = h/(2π ), and h is
the Planck constant. Estimate the quantum number of a grain of pollen
with mass .2.5 × 10−10 Kg, vibrating with frequency .ω = 1 rad/s, and
amplitude .1 mm. (ii) Obtain the wavelength of a proton with thermal energy
.3kB T /2 when the temperature is 300 K.
Solution
(i) The quantum number is solved in terms of the energy .Eq , giving .n =
Eq /(h̄ω) − 1/2. The classical energy of the particle is .Ec = kx 2 /2. By
using the expression .k = mω2 , the energy becomes .Ec = mω2 A2 /2. If
the energy value is associated with the quantum energy, the corresponding
quantum number is .n = mωA2 /(2h̄) − 1/2. The insertion of the numeric
values gives .n = 1 × 10+20 . Thus, the pollen grain may be treated as
a classical particle due to the big quantum number. (ii) According to de
Broglie’s relation, the wavelength of the particle is .λ = h/p, where p is
the particle’s momentum. The kinetic energy is the thermal energy.
. E = p2 /(2m) = 3kB T /2
/ momentum is written in terms of the wavelength, to receive .λ =

The
h2 /(3mkB T ). The higher the momentum, or temperature, the smaller .λ.
(continued)
12.4 Simplification of the Nuclear Potential 267
By inserting values (from the National Institute of Standards and Tech-

nology webpage) of h, .kB , and the proton mass m, the wavelength is
determined.
. λ ∼ 1.45Å
Depending on the characteristic spatial variations of the interaction potential

between particles, the proton may be considered a classical or quantum
particle at 300 K.
12.4 Simplification of the Nuclear Potential
The next goal is to avoid the presence of the nuclear wave function in the nuclear
potential which acts on the electrons, Eq. 12.18, to fully reach the classical limit of
the nuclei.
f ∗ ({R }; t) V̂ ψ
f ∗ [V̂ + V̂
. ψnuc α nuc ({Rα }; t) dR = ψnuc ee Ze + V̂ZZ ]ψnuc dR
The small spatial extension of the nucleus with respect to that of the electron,
with a difference of 4 to 5 orders of magnitude, shows the nucleus as a point
particle. In this regard, the spatial profile of a nucleus may be considered sufficiently
localized, and the particle density may be represented by a product of Dirac-delta
functions.
. |ψnuc ({Rα (t)}; t)|2 = ||α δ (Rα − Rα (t)) (12.24)
It is with this particle density that the complexity of the nuclear potential is
reduced. We start by evaluating each term of the potential. Firstly, we deal with
the nucleus-nucleus .V̂ZZ term.
f [E ] E f [ ||γ δ (Rγ −Rγ (t)) e2 ]
∗ Zα Zβ e 2
. ψnuc α<β |Rα −Rβ | ψnuc dR = α<β Zα Zβ |Rα −Rβ | dR
It is convenient to take the properties of the delta function into consideration for
the integral evaluation.
f
. F (R) δ (R − R(t)) dR = F (R(t)) (12.25)
The integration gives the same function F of the kernel, but evaluated on the
variable .Rα which shows time dependence. In the case of the nucleus-nucleus
function, we obtain a similar result. The integrations over the variables different
to .Rα and .Rβ give, by orthonormality conditions, factors that produce unity, so we
are left with the integrals that only involve the .Rα and .Rβ variables.
E f [ ( )]
2 δ (Rα − Rα (t)) δ Rβ − Rβ (t)
. Zα Zβ e d 3 Rα d 3 Rβ
|Rα − Rβ |
α<β
By integrating over the .β coordinate, and later over the .α coordinate, we obtain
an expression where the nuclear coordinates show time dependence, inherited from
the particle density.
E f [ ] E
δ (Rα − Rα (t)) Zα Zβ e2
. Zα Zβ e 2
d 3 Rα =
|Rα − Rβ (t)| |Rα (t) − Rβ (t)|
α<β α<β
A similar integration is done for the .V̂Ze contribution, which considers the
electron-nucleus interaction. The result equally exhibits time dependence through
the nuclear coordinates.
f ∗ ({R ; t}) V̂ ({r }, {R }) ψ
. ψnuc α Ze i α nuc ({Rα ; t}) dR = VZe ({ri }, {Rα (t)})
(12.26)
Problem 12.4 Demonstrate that the electron-nucleus interaction is simplified

by:
f
∗
. ψnuc ({Rα ; t}) V̂Ze ({ri }, {Rα }) ψnuc ({Rα ; t}) dR = VZe ({ri }, {Rα (t)})
when the particle density corresponds to a product of delta functions.
Solution The demonstration follows similar steps to those given for the
nucleus-nucleus interaction. Let us include the forms of the potential and the
density in the equation.
f E
N E
n f
∗ ||γ δ(Rγ − Rγ (t))
. ψnuc V̂Ze ({ri }, {Rα }) ψnuc dR = − Zα e dR
|ri − Rα |
α=1 i=1
For simplicity, we consider the .β term of the summation over the nuclei and
unfold the product of delta functions.
(continued)
12.4 Simplification of the Nuclear Potential 269
En f ||γ δ(Rγ −Rγ (t)) En f f

i=1 |ri −Rβ | dR = i=1 δ(R1 − R1 (t)) dR1 · · · δ(Rβ−1 − Rβ−1 (t)) dRβ−1 ∗
.
f δ(Rβ −Rβ (t)) f f
∗ Zβ e |ri −Rβ | dRβ δ(Rβ+1 − Rβ+1 (t)) dRβ+1 · · · δ(RN − RN (t)) dRN
All the integrals give unity, except the one containing the electron-nucleus
potential which, by using Eq. 12.25, gives the potential at the instant of time
t.
En f δ(Rβ −Rβ (t)) En Zβ e
. i=1 Zβ e |ri −Rβ | dR = i=1 |ri −Rβ (t)|
When we reinstall the sum over the nuclei, we recover the expression of the
potential .V̂Ze , but with a time dependence through the nuclear coordinates.
EN En Zβ e ( )
. − β=1 i=1 |ri −Rβ (t)| ≡ V̂Ze {ri }, {Rβ (t)}
With the consideration of classical point-like particles, the final potential acting
on the electrons is independent of the nuclear wave function.
f E
n E 2
∗ e
. ψnuc ({Rα }; t) [V̂ee + V̂Ze + V̂ZZ ] ψnuc ({Rα }; t) dR =
rij
i=1 j >i
E
N E
n
Zα e2 EE N
Zα Zβ e2
− +
|ri − Rα (t)| |Rα (t) − Rβ (t)|
α=1 i=1 α=1 α<β
By paying attention to the coordinates .Rα and their time dependence, the above
result is summarized using bra-ket notation in a single line.
. <ψnuc ({Rα }; t) |V̂ ({ri }, {Rα }) | ψnuc ({Rα }; t)> = V̂ ({ri }, {Rα (t)})
h̄→0
(12.27)
The new set of equations ruling the electron and nuclear dynamics changes to:
dPα f ∗
dt = −∇α ψelec ({ri }; {Rα (t)}, t) Ĥelec ψelec ({ri }; {Rα (t)}, t) dr
. [ E ]
h̄2
i h̄ ∂ψ∂telec = − ni=1 2m e
∇ 2 + V̂ ({r }; {R (t)}) ψ
i i α elec ({ri }; {Rα (t)}, t)
(12.28)
In the description of the nuclei as classical particles, the electronic wave function
ψelec has acquired dependence on the instantaneous nuclear positions .Rα (t),
.
because the potential .V̂ in the last equation has dependence on such coordinates.
This is the reason to write the electronic wave function with arguments in the
form .ψelec ({ri }; {Rα (t)}, t). The electronic Hamiltonian is defined from the above
expression by:
En h̄2
. Ĥelec ({ri }; {Rα (t)}) =− i=1 2me ∇i2 + V̂ ({ri }; {Rα (t)}) (12.29)
h̄→0
This Hamiltonian takes the same functional form of the electronic Hamiltonian,
Eq. 11.9, but with nuclear coordinates showing time dependence. The behavior of
the electrons and nuclei persists correlated because, after the previous approxima-
tions, the Newtonian nuclei remain in an average electronic potential, while the
electrons are immersed in an average nuclear potential. The averaged correlation
between particles permits the redistribution of the energy between the electrons and
nuclei. Still, the correlation is only approximately accounted for because, in working
with average potentials, only the averaged collective behavior (not the individual
behavior of each particle) is considered in the interaction between the different sets
of particles. The average correlation demands that the equations ruling the dynamics
of both types of particles are self-consistently solved.
dPα
dt = −∇α Velec ({Rα (t)})
f ∗ ({r }; t) Ĥ
Velec ({Rα }; t) = ψelec i elec ({ri }, {Rα }) ψelec ({ri }; t) dr
En h̄2 En E e2 EN En Zα e 2
. Ĥelec = − i=1 2me ∇i2 + i=1 j >i rij − α=1 i=1 rαi (12.30)
EN E Zα Zβ
+ α=1 β>α Rαβ e2
};{Rα (t)},t)
i h̄ ∂ψelec ({ri∂t = Ĥelec ({ri }; {Rα (t)}) ψelec ({ri }; {Rα (t)}, t)
In the approximation of point-like particles, we may also write the results with
bra-ket notation:
.
dPα
dt = −∇α <ψelec |Ĥelec |ψelec > ; i h̄ ∂ψ∂telec = Ĥelec ψelec (12.31)
We may use the fact that the electrons are the tiny and fast particles, and the nuclei
the heavy and slow particles to invoke the stationary state of the electrons. This is,
the electrons are supposed to rapidly reach the stationary state with nuclei moving
slowly. Under such a situation, the time-independent wave equation replaces the
time-dependent wave equation of the electrons, Sect. 10.4.
dPα
.
dt = −∇α <ψelec |Ĥelec |ψelec > ; Ĥelec ψelec = Eelec ψelec
(12.32)
12.5 Parameterizing the Potential Function 271
This is one of the most common sets of equations to perform mixed quantum-
classical molecular dynamics. It also receives the name quantum molecular dynam-
ics because, whenever the nuclei meet the conditions to satisfy the classical limit, a
quantum approach on the nuclei would in principle render the same results to those
obtained with Newtonian mechanics on such a nuclei. In this regard, the theory of
classical mechanics can be now employed to describe the nuclei of the microscopic
world.
12.5 Parameterizing the Potential Function
Eq. 12.32 of the nuclei is further simplified by working with .<ψelec |Ĥelec |ψelec >,
which is the electronic energy .Eelec of the molecule and works simultaneously as
the potential energy for the nuclei. This energy depends in parametric form on the
instantaneous atomic positions, .{Rα (t)}. When the energy .Eelec is computed for
different molecular conformations, .{Rα (t1 )}, .{Rα (t2 )}, .{Rα (t3 )}..., we are in fact
scanning the potential energy surface where the nuclei are immersed. In this context,
the energy .Eelec corresponds to the potential energy function .Velec for the nuclei,
with dependence on the nuclear coordinates.2
. Velec ({Rα (t)}) = Eelec (12.33)

{Rα }
The potential energy is frequently computed on the fly by solving the quantum
mechanical electronic wave equation:
. Eelec = <ψelec |Ĥelec |ψelec >
Eelec is obtained at the current conformation .{Rα (t)} of the nuclei in the molecular
.
dynamics simulation. The forces over the nuclei are obtained by differentiation of
the potential energy function .Velec , and evaluating the forces at the current nuclear
positions:
.Fα = −∇α Velec {Rα (t)}
Once the set of forces .{Fα } are obtained, these are applied to the nuclei to reach
their new positions .{Rα (t)}. The process is repeated after this step is completed.
The major time-consuming aspect of the simulation involves the solution of the
wave equation. An alternative approach to improve the time-consuming process of
the simulation consists on using a parameterized version of the potential energy
2 In the equations of the text, the nuclear coordinates play the role of parametric variables. Note that
the parametric expression of a function, say .y = y(x), introduces a third variable, t, referred to as
the parametric variable, in terms of which the independent and dependent variables are expressed,
reaching their values in the form .y = g(t), .x = f (t).
.Velec , with the purpose of avoiding the process of continuously solving the equation
.Eelec = <ψelec |Ĥelec |ψelec > at every time step. Such a procedure has important
limitations (to be pointed out later), but, on the other hand, it allows to investigate
structural properties of molecules in solvation and under conditions of temperature
and pressure, which is a difficult task using the wave function approach, especially
for large molecules, like the proteins.
In order to illustrate the parameterized approach, consider, for example, a
molecule with three atoms: a, b, and c, like water. Their positions are .Ra , .Rb , .Rc ,
and the equilibrium distance is .R0 for the bonds ab and ac. The .3N − 6 = 3
corresponding degrees of freedom, dof ’s, may be represented by the pair distances
.Rab and .Rac , plus the angle .θabc formed by the vectors .Rab and .Rac with atom
a located at the vertex. The angle of equilibrium is .θ0 . After computing a set of
energies .Eelec for the elongation and shrinking of the ab bond, with the other dof ’s
fixed at their equilibrium values, .Rac = R0 and .θabc = θ0 , the potential energy of
this dof may be written as:
. Vab = kab (Rab − R0 )2 /2 ; Rab = |Rab | (12.34)
A similar process for the other dof ’s gives .Vac = kac (Rac − R0 )2 /2 and .Vabc =
kabc (θabc −θ0 )2 /2. The force constants, equilibrium distance, and equilibrium angle,
.kab , kac , kabc , R0 , θabc , are parameters determined after fitting the potential energies
.Vab , .Vac and .Vabc to the molecular energies .Eelec , simultaneously reproducing the
structure of the reference molecule. The force constants of other molecules, different
from the water molecule, are in general different. By using Eqs. 12.32–12.34 and
considering the total potential energy function as the addition of individual terms,
.Velec = Vab + Vac + Vabc , we obtain:
dPα
.
dt = −∇α Velec = −∇α [Vab (Rab , R0 ) + Vac (Rac , R0 ) + Vabc (θabc , θ0 )]
(12.35)
The model potential is acceptable in the domains of the parameterized dof ’s solely.
When the potential function is used to extrapolate the conformations of the molecule
to other unexplored conformations, there is a risk of obtaining nonphysical behavior
and wrong energies. For instance, when the atom c is separated from the atoms ab
that integrate the example molecule, we are left with an isolated atom c and an
isolated dimer ab. Yet, the parameterized potential retains the molecular bond of
c to the dimer ab due to the parabolic shape of the harmonic potential function
employed in this case. Such a model potential is incapable to deal with the breaking
of the .c − ab bond.
For molecules with many atoms, the .Velec may be represented by a general n-
body form [9].
[E E E E
(1) (2) (3) (4)
.Velec ({Rα (t)}) = Ei + Eij + Eij k + Eij kl
12.5 Parameterizing the Potential Function 273
E E ]
+ Eijcoul + EijvdW (12.36)
{Rα }
The expression shows the first additive contributions of a general n-body potential.
E (1)
The first term, . Ei , gives the addition of atomic energies. When the number of
atoms does not change or the atoms do not mute onto different atoms in the particle
system, the first term is an energy constant and may be neglected. The following
E (2)
contribution, . Eij , introduces the pair interactions of neighbor atoms of the same
molecule. The potentials employed to describe the pair interactions are frequently of
E
harmonic type in the separation distance, .Rij . The next term, . Eij(3)k , represents the
potential associated with trimers, comprised by three neighbor atoms of the same
molecule. The potentials used to describe the trimers are frequently of harmonic
E (4)
type in the trimer angle, .θj ik . The term, . Eij kl , is the contribution of tetramers,
formed by four neighbor atoms of the same molecule. The potentials employed for
tetramers frequently consist of a series of cosine functions in the tetramer torsion
angle, .θij kl . The first four contributions of Eq. 12.36 describe the internal dof ’s
of the molecule, because they involve bonded atoms in the separation distance,
the trimer angle, and the tetramer torsional angle. The first four potential terms
apply to small or large molecules and, due to their nature, constitute all together the
molecular internal energy.
Finally, the last two energy terms of the model potential, Eq. 12.36, represent
the non-bonded interactions like the Coulomb, coul, and van der Waals, vdW ,
interactions. They consider the interactions among the i and j atoms, which are
distant apart from each other, usually belonging to different molecular groups.
The non-bonded atomic interactions introduce the external forces or environmental
forces acting on the molecule. For example, given a specific water molecule, the
non-bonded atomic interactions introduce the forces acting on that specific water
molecule by the other water molecules. In short, a model potential like that proposed
in Eq. 12.36 represents a parameterization of the energy landscape of the nuclei. The
model potential accounts for both the molecular dof ’s and the atomic interactions
by defining groups of atoms and proposing the internal and external potential
functions of the atomic groups. The force constants, the equilibrium distances, and
the equilibrium angles represent all together a way of introducing the information
referent to the structure and rigidity of the molecule in the force functions previous
to the simulations. The method displayed here is known as molecular mechanics. It
constitutes a branch of classical molecular dynamics, while the proposal of model
potentials takes us to the area of forcefields, to be unfolded later. In particular,
when the potentials of harmonic type are employed, and the force constants have
relatively large values or the bond distances and bond angles are fixed in the
simulations, Sect. 15.7, we may transit from molecular mechanics to a ball-and-
stick approximation.
12.6 Total Energy of the Molecular System
The level of theory employed in the description of the electrons is different to that
used in the description of the nuclei, because the electrons are treated as quantum
particles and the nuclei as classical particles. The addition of the kinetic energy plus
the potential energy gives the energy of the particle system.
E P2α f ∗ ({r }; {R (t)}, t) Ĥ
. E= α 2Mα + ψelec i α elec ψelec ({ri }; {Rα (t)}, t) dr
(12.37)
When the energy is preserved over time, the time derivative of the energy is zero.
[ ]
dE E Pα dPα f ∗
dψelec dψelec d Ĥelec
. = · + ∗ Ĥ
Ĥelec ψelec + ψelec ∗
+ ψelec ψelec dr = 0
elec
dt α
Mα dt dt dt dt
If the following three conditions are met: (i) .ψelec satisfies the wave equation,
∗
(ii) .ψelec satisfies the opposite-sign wave equation, and (iii) the Hamiltonian is a
Hermitian operator, then the first two terms in square brackets eliminate each other.
By recalling that .Ĥelec depends on t by means of the nuclear coordinates, the time
derivative .d Ĥelec /dt is evaluated by applying the chain rule.
E Pα dPα f E
∗ dRα
. · = − ψelec ∇α Ĥelec · ψelec dr
α
Mα dt α
dt
The factor .dRα /dt is identified with .Pα /Mα , and the force on each nucleus is
deduced.
dPα f ∗
.
dt = − ψelec ∇α Ĥelec ψelec dr (12.38)
This expression reproduces the first expression of Eqs. 12.28. The electronic
Hamiltonian is in turn divided into kinetic energy and potential energy terms,
Eq. 12.30. The electron kinetic energy operator may be considered as a function with
the purpose to carry the derivation; nevertheless, it contains no nuclear coordinates
and is neglected. On the other side, the potential energy terms of the electronic
Hamiltonian are, Eq. 12.27.
E
n
h̄2 2 E
n E 2
e
. V̂ ({ri }, {Rα (t)}) = − ∇i +
2me rij
i=1 i=1 j >i
E
N E
n
Zα e2 E
N E
Zα Zβ 2
− + e
riα Rαβ
α=1 i=1 α=1 β>α
12.6 Total Energy of the Molecular System 275
By defining a force like in classical theory, this is, as the negative of the gradient
acting upon the potential function, it is possible to compute the force over the .αth
nucleus.
. Fα = −∇α V̂ ({ri }, {Rα (t)}) (12.39)
This expression is the general Hellmann-Feynman theorem. It states that the

force acting upon the .αth nucleus is due to the electrostatic potentials and the
electronic wave function [10]. Care should be exercised when .ψelec is not the
exact wave function but an approximate one, because the previous expression may
produce spurious forces. By paying attention to the specific contributions of .V̂ with
dependence on the nuclear coordinates, we have:
f [ E E ]
Fα = ∗ − ni=1 ∇α Vαi − N
ψelec β=1 ∇α Vαβ ψelec dr
.
Vαi = −Zα e2 /riα ; Vαβ = Zα Zβ e2 /Rαβ ; α /= β
From the fact that the nuclear and electronic coordinates in the .Vβi potential
are shown in symmetric form, it is possible to use the equivalence .∇α Vβi =
−δαβ ∇i Vβi , with .δαβ = 0 for .α /= β and .δαβ = 1 for .α = β, in the expression of
the force. In particular, .∇α Vαi = −∇i Vαi . The force shall be further simplified by
carrying integrations over selected electronic coordinates.
f En [f ∗ ψ
] 3
Fα = i=1 ∇i Vαi ψelec elec d r1 · · · d ri−1 d ri+1 · · · d rn d ri +
3 3 3 3
.
E
− N β=1 ∇α Vαβ
Problem 12.5 Demonstrate the following identities in the potentials .Vβi and
Vβγ .
.
E E
∇α Vβi = −δαβ ∇i Vβi
. ; ∇α Vβγ = ∇α Vαβ
β,γ β
γ >β β/=γ
Solution The constants of the potentials are omitted for simplicity in the
derivations. Firstly, we recall the definition of the attractive potential .Vαi and
perform the differentiation with respect to the .α coordinate.
∇α Vαi = (∂Xα , ∂Yα , ∂Zα ) [−1/|ri − Rα |] = −(ri − Rα )/|ri − Rα |3

.
By similarly operating over the coordinates of particle i, we have:
(continued)
∇i Vαi = (ri − Rα )/|ri − Rα |3

.
The result is .∇α Vαi = −∇i Vαi . If the .β coordinate is considered, a delta
function is required, and the conclusion is .∇α Vβi = −δαβ ∇i Vβi . In order to
proof the second identity, we exemplify a particular case to rationalize the
general case. By showing the components of the summation, we have:
E
∇α β,γ Vβγ = ∇α (V12 + V13 + V14 + · · · + V1N +
. γ >β )
+ V23 + V24 + · · · + V2N + V34 + · · · + V(N −1)N
Suppose .α = 3. The non-zero terms are those containing subindex .α = 3.

E
∇α=3
. Vβγ = ∇α=3 (V13 + V23 + V34 + V35 + · · · + V3N )
β,γ
γ >β
E
N
= ∇α=3 Vα=3 β
β=1
α/=β
It was considered .Vαβ = Vβα in the last identity. The generalization

E of the
result is done by replacing .α = 3 with .α = κ, namely, .∇κ β,γ Vβγ =
EN γ >β
β=1 ∇ κ Vκβ , and .β =

/ κ.
The term in square brackets of the previous expression is the electron density,
ρelec , a function of a single variable.
.
f ∗
ρelec (ri ) = n ψelec ψelec d 3 r1 · · · d 3 ri−1 d 3 ri+1 · · · d 3 rn
.
[f f ] E
Fα = n1 ∇1 Vα1 ρelec (r1 )d 3 r1 + ∇2 Vα2 ρelec (r2 )d 3 r2 + · · · − N β=1 ∇α Vαβ
Variable .ri , with .i = 1, 2, · · · , n, is a dummy variable due to the indistinguisha-

bility of the electrons, and instead, we may use .r in every term. On the other hand,
the factor .∇i Vαi is understood as the electric field .Eα (ri ) produced by the electron
distribution, which acts on the .α nucleus. The Hellmann-Feynman force is defined
by:
[f f ] E
Fα = 1
n Eα (r1 ) ρelec (r1 )d 3 r1 + Eα (r2 ) ρelec (r2 )d 3 r2 + · · · − Nβ=1 ∇α Vαβ
.
Eα (ri ) = ∇i Vαi
(12.40)
12.7 Establishing the Accuracy of Atomic Forces 277
The force on a nucleus is the classical electrostatic force exerted by the other
point-like nuclei and the electron charge density. We additionally observe that
force changes over the nuclei are related to electron-distribution changes and vice
versa. The n individual terms of the above expression are all the same, and they
are simplified to a single one with factor n. The final expression of the force is
referred to as the electrostatic theorem, because the force on a nucleus is due to
both a continuous electron distribution and a discrete nuclear distribution. In concise
manner, we write:
f EN
. Fα = Eα (r) ρelec (r)d 3 r − β=1 ∇α Vαβ (12.41)
The computation of the force in the Hellmann-Feynman approach is described by

functions with dependence on 3 spatial degrees of freedom instead of the 3n spatial
degrees of freedom associated with n electrons.
12.7 Establishing the Accuracy of Atomic Forces
After considering several simplifications in the equations of the nuclei, it is

important to establish the accuracy of the atomic forces, Eq. 12.38, with the purpose
to reach reliable values of the physical observables. To do this, we are required to
estimate the deviations of the theoretical forces from the experimental forces. There
are different experiments whose results may be used for assessing the accuracy
of the atomic forces of the simulations. The experiments using the atomic force
microscope or optical tweezers may directly provide the atomic forces; nevertheless,
those experiments are not abundant. On the other side, there is a relatively large
number of experiments that provide the vibrational spectra of the individual
molecules. The accuracy of the atomic forces may be indirectly determined from
the vibrational spectra [11].
Suppose a molecule with N atoms moving with no constraints. The number of
degrees of freedom is .3N − 6. In the calculation of the vibrational spectrum of the
molecule, the Hessian matrix, .hij , is proposed.
⎛ ∂2V ∂2V ∂2V
⎞
···
⎜ ∂x12 ∂x1 ∂x2 ∂x1 ∂xN
⎟
⎜ ⎟
⎜ ∂2V ∂2V
··· ∂2V ⎟
⎜ ∂x2 x1 ∂x22 ∂x2 ∂xN ⎟
⎜ ⎟
. hij = ⎜ ⎟ (12.42)
⎜ .. .. ⎟
⎜ . . ⎟
⎜ ⎟
⎝ ⎠
∂2V ∂2V ∂2V
∂xN x1 ∂xN x2 ··· 2
∂xN
The Hessian matrix depends on the second derivatives of the potential with
respect to the atomic positions, .xi . It gives information of the multidimensional
topography of the potential energy surface associated with the molecule. The
Hessian matrix is diagonalized to obtain the eigenvectors, known as the normal

vibrational modes, and the eigenvalues, known as the vibrational frequencies.
The strength of each normal vibration mode gives the corresponding vibrational
intensity. The .3N − 6 normal vibration modes, and equal number of vibrational fre-
quencies and intensities, constitute all together the molecular vibrational spectrum.
Such a spectrum may be determined from experiments in the lab. By estimating the
deviations between the theoretical and experimental spectral data, it is possible to
indirectly establish the reliability of the atomic forces employed in the simulations.
The atomic forces represent the first derivatives of the potential with respect to the
atomic positions, and the Hessian matrix introduces the second derivatives of the
potential with respect to the atomic positions. Therefore, the proper prediction of the
molecular vibrational spectrum indicates acceptable second derivatives, indirectly
pointing out bona fide first derivatives.3
The experiments with atomic force microscopes or optical tweezers provide
accurate values of the atomic forces. The number of such experiments is not as large
as those giving the vibrational spectra of molecules. Still, the atomic forces from
those experiments or alternative experiments should be used whenever available for
an additional check of the atomic forces [14].
12.8 Diffusion from the Continuity Equation
The goal of this section is to deduce an equation for the statistical behavior of a set
of classical particles that diffuse in a fluid. The diffusion of particles is important in
processes that involve the doping of materials, the homogenization of substances in
liquids, the sedimentation of particles, etc. The discussion departs from the equation
of continuity 12.8, which is a conservation law on the number of particles since it
ensures that the particles penetrating an imaginary volume in the fluid are precisely
the ones that come out of that volume. Thus, any density change of the diffusing
particles in a volume of the fluid is only attributed to the inflow and outflow of
particles. The set of particles have an initial concentration in a spatial region of
the fluid, and, after some time, the particles spread evenly with random trajectories
over a larger volume. The equation of continuity 12.8 is reproduced below for ease
reference.
∂ρ(R;t) EN
.
∂t + ∇ · J (R; t) = 0 ; J (R; t) = 1
α=1 Mα D 2 ∇α S (12.43)
The function .ρ is the density of diffusing particles in terms of .R at the instant of

time t. The function .J is the particle flux passing through a unit area per unit time.
.J is also referred to as the particle current. The expression .Pα = ∇α S was given in
3 Santamaria et al. [12]. Also, for an example of a vibrational spectrum where the position of
a single atom of a molecule is differentiated from the others, refer to: Mollinedo-Rosado and
Santamaria [13].
12.9 Diffusion Equation and Particle Flux 279
Eq. 12.15; thereby, the flux is proportional to the velocities .Vα of the particles. To
be consistent with the delta profile of the particle density of Eq. 12.24, and using the
identity .ρ = D 2 of Sect. 12.2, a delta profile is also assigned to the particle current.
EN
. J= α=1 Vα δ(R − Rα (t)) ; Vα = dRα /dt (12.44)
Initially, the high particle concentration tends to disperse, and after some time,
the particles achieve a uniform concentration in a homogeneous fluid. For slowly
varying distributions of the particles, it is possible to provide an expression for the
change of the particle concentration. A subtle change of concentrations in the z axis
may be described as a linear relation between the concentration .ρ and the position z,
this is .ρ = αz, where the constant of proportionality is .α. In mathematical terms, .α
is a slope, and in the general case, it is a gradient, .α = ∇ρ(R). This is, .α represents
the rate of concentration change with respect to the position. The particle flux per
unit area is related to the diffusion constant, d, by .J = −αd. In this regard, we have
an equation relating the particle flux to both the rate of concentration change and
the particle diffusion constant.
. J = −d ∇ρ(R) (12.45)
This equation relates the particle current .J to the gradient of .ρ using the diffusion
constant d. It is known as first Fick’s law. When the gradient of the density is zero,
the average density is constant (there is a uniform concentration of the diffusing
particles), and no particle currents or fluxes .J are perceptible [15].
12.9 Diffusion Equation and Particle Flux
The goal of this section is to create consistency among the equations that involve
parameters referent to the diffusion of the particles. It is convenient to start by
introducing the expression of .J given by the first Fick’s law, Eq. 12.45, into the
continuity equation, Eq. 12.43, and under the assumption of a diffusion coefficient
d with no dependence on the density of the diffusing particles.
. ∂ρ(R; t)/∂t = d ∇ 2 ρ(R; t) (12.46)
This is a partial differential equation which involves key ingredients of the

diffusion process. It receives the name of diffusion equation or second Fick’s law,
as it describes the spatial dispersion of the particles in terms of the distribution
function, .ρ. The interactions of the particles are not explicitly considered in the
diffusion equation. Also, the particle dispersion is time irreversible, problem 15.3.
The diffusion factor has units of square length over time. The solution of the
equation is unique when the conditions on the spatial distribution at the instant of
time .t0 are imposed. The diffusion equation may be mutated onto an equation for the
heat diffusion: if the temperature plays the role of the particle concentration, and the
thermal conductivity replaces the diffusion factor, then the new expression becomes
the heat equation. It is equally possible to deduce the wave equation from the
continuity equation [16]. The diffusion equation has wide applications in different
fields of science.
When the system of particles achieves a uniform density, .ρ = constant, there
are no spatial and temporal changes of the concentration. The first Fick’s law is
recovered from the second Fick’s law due to the fact that after using Eqs. 12.43
and 12.46, we reproduce the first Fick’s law, .J = −d∇ρ. More precisely, we have
.∂ρ (R; t) /∂t + ∇ · J (R; t) = 0, and on the other hand, .∂ρ(R; t)/∂t = d ∇ ρ(R; t).
2
By comparing terms, we obtain .d∇ ρ = −∇ ·J. Thus, the first Fick’s law represents
2
a first integration of the second Fick’s law. A solution of the diffusion equation is
.ρ(R; t) = exp[i(kR − ωt)] with frequency:
ω = −id (2π/λ)2
. ; k = 2π/λ
The .λ factor describes the periodic variation of the density. It is possible to

express a general solution of Eq. 12.46 as a linear combination of plane waves.
More general solutions of the diffusion equation than these described here are found
elsewhere.4
Problem 12.6 The one-dimension diffusion equation is .∂F /∂t =

d (∂ 2 F /∂x 2 ). Find the solution using the method of Fourier transforms.
Solution We have to integrate the diffusion equation to obtain the solution

F . The Fourier transform is applied on both sides of the diffusion equation.
f ∞ f ∞
1 ∂F d ∂ 2F
. √ eiωx dx = √ eiωx dx
2π −∞ ∂t 2π −∞ ∂x 2
We
f ∞ define the transformed
√ function in the form f (ω, t)
. =
e iωx F (x, t) dx/ 2π to have:
−∞
∂f (ω,t)
.
∂t = −dω2 f (ω, t)
where the property of Fourier transforms on derivatives was used to replace

the second-order derivative in the variable x with the multiplicative factor
2
.ω . The equation is now an ordinary differential equation of first order in the
(continued)
4 For instance, refer to: Tveito et al. [17], Chapt. 7 and 8, Caretto [18] notes on engineering analysis,
and Ursell [19].

12.9 Diffusion Equation and Particle Flux 281
variable t. The integration gives .f (ω, t) = C exp(−dω2 t), with the factor
.C = f (ω, 0) imposing the initial conditions. We apply the inverse Fourier
transform to recover F .
f ∞
1
e−iωx C(ω) e−dω t dω
2
.F (x, t) = √
2π −∞
The expression is simplified when the coefficient C is constant. The goal is

to write an expression in the variable .ω by squaring terms and proceed to
integrate.
f ∞ [√ √ ]2
C
F (x, t) = √ e−x /(4dt)
2
. exp{− dt ω + ix/(2 dt) } dω
2π −∞
f∞
The integration is done using the result . −∞ dω exp[−(α/2) (ω − iy/α)2 ] =
√
2π/α.
( )
C x2
.F (x, t) = √ exp −
2dt 4dt
The diffusion equation in more dimensions is solved similarly, with a

result that corresponds to the product of Gaussian functions in the different
dimensions.
The following step is to establish an equation of .J when there is an external force

acting on the diffusing particles, thus inducing a drift current. This step shall allow
to find the diffusion constant in terms of the temperature and medium viscosity. We
follow the strategy of Chaikin and Lubensky [20]. Suppose a particle concentration
.ρ. An external force generated from the external potential .Vext induces a drift of
the diffusing particles, pushing them into a given direction. The particles are also
subjected to the random viscous forces of the fluid. Such forces are supposed to be
negatively proportional to the particle’s velocity.
.f = fext − ξ V ; fext = −∇Vext
The constant .ξ is the viscosity of the fluid, and the inverse of .ξ is the particle
mobility. The velocity .V should not be confused with an energy potential. The total
force in the steady state is .f = 0. By solving for the particle velocity, we have .V =
−∇Vext /ξ . Therefore, the external force has effect on the particle current 12.45.
Such an effect is accounted for by the term .−ρ ∇Vext /ξ .
. J = −d ∇ρ(R; t) − ρ(R; t) ∇Vext /ξ (12.47)
There is equilibrium when the particle current is zero.
∇ ln ρ(R; t) = −∇Vext / (dξ )

.
A solution of this equation involving the temperature of the fluid is proposed

as an exponential function, .ρ = exp(−Vext /kB T ). It is a Maxwell-Boltzmann
expression. By using the exponential in the last equation, we obtain a diffusion
coefficient which linearly depends on the equilibrium temperature T and is inversely
proportional to the viscosity .ξ .
. d = kB T /ξ (12.48)
This expression is the diffusion constant initially proposed by Einstein. Equa-

tion 12.48 is one of the important fluctuation-dissipation relations of stochastic
dynamics.
The final purpose is to connect the particle current .J to the chemical potential .μ to
further characterize the diffusion process. The expression of the chemical potential,
.μ, is related to the change rate of the densities.
( )3/2
h2
μ = kB T ln(ρ/ρ0 )
. ; ρ0−1 = = λ3T
2π MkB T
The term .λT is the de Broglie thermal wavelength, Eq. 12.23. By considering
the particle current 12.47, and the expression of the diffusion constant d, the new
equation of .J is:
. J = −(ρ/ξ ) ∇(kB T ln ρ + Vext ) = −(kB T /ξ ) ∇ρ − (ρ/ξ ) ∇Vext (12.49)
By inserting the expression of the chemical potential in the first identity of .J

presented above, the particle current becomes the gradient of the chemical potential
difference.
. J = −(ρ/ξ ) ∇[μ − μext ] ; μext = −Vext (12.50)
By considering the free energy in the form .F = NkB T [ln(ρλ3T )−1], the particle
current is also written in terms of the free energy changes, .δF , with respect to the
change in the particle number, .δN, while maintaining the temperature and confining
volume constants.
( δF )
. J = −(ρ/ξ ) ∇ δN
(12.51)
12.10 Expansion of the Electronic Wave Equation 283
Eqs. 12.50 and 12.51 are expressions of the particle current .J in terms of the
chemical potential .μ and the free energy F , respectively. On the other side, the
mobility of a particle depends not only on the fluid viscosity factor .ξ but also on the
particle’s geometry. According to the Stokes dragging-force law, a spherical particle
with radius a has an effective viscosity factor .6π ξ a [21]. By using the expression
of the particle flux 12.49 with .Vext = 0, we have:
. J = −kB T ∇ρ / (6π ξ a) (12.52)
The first Fick’s law, Eq. 12.45, indicates that .J = −d∇ρ; thereby, the diffusion
factor for the spherical body is:
. d = kB T / (6π ξ a) (12.53)
The diffusion factor can be derived from the velocity autocorrelation function
and shown to be related to the mean square displacement, .<r2 (t)>.
12.10 Expansion of the Electronic Wave Equation
It is time to return our attention to the electronic wave function, since the nuclear
wave function does not fit into the classical limit. In the following, the electronic
wave function is expanded using a linear combination of complete basis set
functions. Under this approximation, the solution of the electronic wave equation,
Eq. 12.31, is relegated to find the optimal coefficients that satisfy such an equation.
We start by writing the electronic wave function as a linear combination of the basis
functions .{ψelec
k }, maintaining the dependence on the electronic coordinates and the
parametric dependence on the nuclear coordinates.

E
. ψelec ({ri }; {Rα (t)}, t) = k
k ck (t) ψelec ({ri }; {Rα (t)}) (12.54)
The amplitudes or coefficients .ck exhibit time dependence solely. The strategy
consists in separating the quantities that depend on the time from those that depend
on the spatial variables. The wave function .ψelec satisfies the wave equation of the
electrons, Eq. 12.31.
∂ψelec ({ri }; {Rα (t)}, t)

i h̄
. = Ĥelec ({ri }; {Rα (t)}) ψelec ({ri }; {Rα (t)}, t)
∂t
By introducing the expansion of the wave function and using the chain rule on
the time derivatives, the wave equation is presented in terms of the basis functions
and their amplitudes.
E [ k ∂ck (t) E ∂Rα

] E
. i h̄ k ψelec ∂t + ck (t) α ∇α ψelec ·
k
∂t = k ck (t) Ĥelec
k
ψelec
(12.55)
We multiply from the left by .(ψelec )∗ and integrate over the electronic coordi-
j
nates.
[ ]
Ef ∂ck (t) E j ∂Rα
j ∗ k ∗
.i h̄ dr (ψelec ) ψelec + ck (t) (ψelec ) ∇α ψelec ·
k
∂t α
∂t
k
Ef
dr ck (t) (ψelec )∗ Ĥelec ψelec
j
= k
Further simplification is achieved when the basis functions are the adiabatic
j j
functions of the Hamiltonian, .Ĥelec ψelec = Ej ψelec , and after applying the
f j ∗ k
orthonormality property of the basis functions, . (ψelec ) ψelec dr = δj k .5
E [f ]
∂cj (t) ∂Rα
dr (ψelec )∗ ∇α ψelec
j
.i h̄ + i h̄ ck (t) k
· = cj (t) Ej
∂t ∂t
k,α
The integral in parenthesis is a coupling term between the basis functions. It

is designated by the function .dαjk , which depends parametrically on the nuclear
coordinates.
f j ∗
. d ({Rα (t)}) = (ψelec ) ∇α ψelec
α k dr (12.56)
jk
By using this coupling term, the final equation for the .cj amplitude is established.
∂cj (t) E ∂Rα

. i h̄ ∂t = cj (t)Ej − i h̄ α
k,α ck (t) dj k · ∂t
(12.57)
There are as many coupled differential equations of the type 12.57 as there are
k
functions .ψelec in the expansion of the electronic wave function .ψelec , Eq. 12.54.
The above equation is not compelled, in principle, to satisfy the time reversibility
property of dynamic systems, especially in circumstances where the particle
system moves in a multi-branched energy surface because, in this case, the energy
surfaces may have sufficiently similar energies and wave functions. Such similarities
introduce a probabilistic factor in the equations of motion due to the presence of
the coupling terms, Eq. 12.56, leading to particle trajectories in the time-forward
direction which are not necessarily the same as those in the time-backward direction.
Other types of expansions of the electronic wave function and total wave function
are found elsewhere [22].
5 In the adiabatic approximation, the time evolution of the potential is sufficiently slow in such a
way that the wave function of the electrons practically changes instantaneously with respect to the
slowly changing potential of the nuclei.
12.10 Expansion of the Electronic Wave Equation 285
Problem 12.7 Consider the following expansion of the electronic wave

function:
E f t
.ψelec ({ri , Rα , t}) = cj (t) ψj ({ri , Rα }) exp[(−i/h̄) Hjj dt ' ]
f
The basis functions .ψi are orthonormal, . dr ψi∗ ψj = δij , and may be
adiabatic
f or∗ diabatic functions. The matrix elements of the Hamiltonian are
.Hij = ψi Ĥψj dr. The expansion is different to that of Eq. 12.54. Obtain
the time evolution of the amplitudes .cj (t).
Solution In order to obtain the time evolution of the coefficients .cj , we

introduce the electronic wave function in the time-dependent wave equation.
f t f t
∂ E E
i h̄
. cj ψj exp[−i Hjj /h̄ dt ' ] = cj Ĥψj exp[−i Hjj /h̄ dt ' ]
∂t
j j
The partial time derivative is applied to every term of the left-hand

f t side, and
the result is multiplied by the complex-conjugate term .ψi∗ exp[i Hii /h̄ dt ' ]
and integrated over the whole space.
E f f t
i
. i h̄ċi + cj Ṙα · ψi∗ ∇α ψj dr exp[−i (Hjj −Hii )/h̄ dt ' ] + ci (− )Hii
h̄
j,α
E f t
= cj Hij exp[−i (Hjj − Hii )/h̄ dt ' ]
j
The chain rule was applied in the way it was done to obtain Eq. 12.55. By
solving for .ċi and considering that: (i) the diagonal term which includes .Hii
is cancelled
f with the respectivef diagonal term considered in the summation,
and (ii) . ψi∗ ∇α ψi dr = ∇α (ψi )2 dr = 0, the final result is obtained.
[ ] f t
E E i
ċi (t) = −
. cj (t) Ṙα · dij + Hij exp[−i
α
(Hjj − Hii )/h̄ dt ' ]
h̄
j /=i α
This expression corresponds to an equation of the surface hopping approach

and is important to work with transitions between energy surfaces. The
Landau-Zener transition between two energy states is an approximation to
the above expression.
12.11 Expansion of the Newtonian Equation of the Nuclei
The following step is to write the Newtonian equation of the nuclei in terms of the
expansion amplitudes of the electronic wave function following the discussion of
Tully [23]. We start from the expression of the total energy, Eq. 12.37, and recall the
electronic wave function in terms of the basis functions, Eq. 12.54.
[ ] ⎡ ⎤
E P2 f E E
ci∗ (t) (ψelec )∗ Ĥelec ⎣ cj (t) ψelec ⎦ dr
α j
.E = + i
α
2Mα
i j
The conservation of the energy demands a time derivative of the total energy
equal to zero.
⎡ ⎤
E ∂Rα ∂Pα f
∂E ∂ ⎣E ∗
)∗ Ĥelec ψelec dr⎦ = 0
j
. = · + i
ci cj (ψelec
∂t α
∂t ∂t ∂t
i,j
The adiabatic functions are eigenfunctions of the electronic Hamiltonian, and it

j j
is possible to replace the term .Ĥelec ψelec by .Ej ψelec . The orthonormal property of
the adiabatic eigenfunctions is used to reduce the summation with two indexes to a
summation with one index.
E ∂Rα ∂Pα E [ ∗ ∂Ej (
∗c
)]
. α ∂t · ∂t = − j cj c j ∂t + E j
∂
∂t cj j (12.58)
The electronic energy was differentiated because it depends on the nuclear

coordinates, which show time dependence. On the other hand, by multiplying
Eq. 12.57 by .cj∗ , the complex-conjugate equation by .cj and subtracting one equation
from the other, we have:
∂ ∗ E[ ] ∂R
cj∗ ck dαjk + ck∗ cj (dαjk )∗ ·
α
. (cj cj ) = −
∂t ∂t
k,α
The insertion of .d(cj∗ cj )/dt in the last member of the expression that imposes
the conservation of the energy, Eq. 12.58, leads directly to:
E E [ ∗ ∂Ej E ( ) ]
.
∂Rα
α ∂t · ∂Pα
∂t =− j cj cj ∂t − Ej k,α cj∗ ck dαjk + ck∗ cj (dαjk )∗ · ∂Rα
∂t
E
By using the chain rule, the coefficient .∂Ej /∂t is replaced by . α ∇α Ej ·∂Rα /∂t.
The substitution leads to an expression for each element .α of the summation.
∂Pα E E [ ]
. =− cj∗ cj ∇α Ej + Ej cj∗ ck dαjk + ck∗ cj (dαjk )∗
∂t
j j,k
12.11 Expansion of the Newtonian Equation of the Nuclei 287
By redefining indexes, the equation is transformed.

E ∗
E
.
∂Pα
∂t =− j cj cj ∇α Ej + j,k (Ej − Ek ) cj∗ ck dαjk (12.59)
Problem 12.8 Deduce the force presented in Eq. 12.59.
Solution We depart from the expression previous to Eq. 12.59.
∂Pα E E [ ]
. =− cj∗ cj ∇α Ej + Ej cj∗ ck dαjk + ck∗ cj (dαjk )∗
∂t
j j,k
By recalling the definition of the coupling term .dαjk , Eq. 12.56, and using the
identity between .(dαjk )∗ and .dαkj , we have:
f k ∗ j f j f
k )∗ dr + (ψ k )∗ ∇ ψ j dr = 0
∇α (ψelec ) ψelec dr = ψelec ∇α (ψelec elec α elec
.
f j f
k )∗ dr = − (ψ k )∗ ∇ ψ j
ψelec ∇α (ψelec elec α elec dr
The last identity indicates .(dαjk )∗ = −dαkj . If we use this expression and
exchange the j and k indexes, then Eq. 12.59 is recovered.
∂Pα E E( )
. =− cj∗ cj ∇α Ej + Ej cj∗ ck − Ek cj∗ ck dαjk
∂t
j j,k
When the equations that rule the time evolution of the electrons and nuclei in
terms of the basis set coefficients are written together, the coupling between such
equations is shown.
E j
ψelec ({ri }; {Rα (t)}, t) = j cj (t) ψelec ({ri }; {Rα (t)})
∂cj (t) E ∂Rα
. i h̄ ∂t = cj (t) Ej − i h̄ α
k,α ck (t) dj k · ∂t (12.60)
2 R (t) E E ∗
Mα ∂ α
∂t 2
=− j |cj (t)|2 ∇α Ej + j,k cj (t) ck (t) [Ej − Ek ] dαjk
The first term on the right-hand side of the last equation represents a classical force
term, as the force over particle .α is deduced from the gradient of a potential energy
function, and the second term, with no classical analog, contributes to the force
on particle .α by taking the coupling of the different potential energy surfaces into
consideration. The equation of the total energy, Eq. 12.37, is equally written in terms
of the basis set functions .{ψelec
k }, and the respective probability weights.
E P2α E ∗
f
({ri }; {Rα })]∗ Ĥelec ψelec ({ri }; {Rα }) dr
j
E= α 2Mα + i,j ci (t) cj (t) [ψelec
i
.
E P2α E
E= α 2Mα + j |cj (t)|2 Ej
(12.61)
In principle, the equations of motion 12.60 together with the total energy 12.61
may be solved numerically. The results are independent of the basis set .{ψelec
k } when
the basis set is complete. The solutions of the equations depend on the capability to
solve the electronic equation .Ĥelec ψelec
k = Ek ψelec
k for the many values that the
subindex k may take. Unfortunately, this is a main problem of technical character as
it is difficult to solve the equation when one works with real systems. However, there
are exceptional cases where the solution may be achieved as, for instance, in model
atomic systems. If it happens that in the expansion of the electronic function, the
term corresponding to the ground state is the dominant one, this is .ψelec ≈ c0 ψelec
0 ,
then the equations deduced above in terms of the amplitudes .cj (t) become more
amenable to deal with.
A.1 Appendix: Bohm’s Quantum Potential
The purpose of this section is to discuss the implications of the Bohm quantum
potential, Eq. 12.10, in the interpretation of quantum mechanics.
There are different interpretations of quantum mechanics in physics. The most
common one is based on a probabilistic description of the particles, recognized as
the Copenhagen interpretation, proposed by Niels Bohr and Werner Heisenberg in
1925–1927. However, there are other equally interesting interpretations like that
proposed by Bohm in 1952 [24]. In this approximation a particle is driven by
a quantum field, .ψ, which satisfies the wave equation. In Bohm’s approach of
quantum theory, the particle is acted not only by the classical potentials but also
by the quantum potential, .Vbohm of Eq. 12.10, in turn determined by the quantum
field .ψ.
EN h̄2 ∇α D
2
. Vbohm = − α=1 2Mα D ; D = |ψ|2 (12.62)
In Bohm’s approach, a particle such as the electron is supposed to follow a

continuous and causally determined trajectory. The particle is not regarded as an
object with wave-particle dual character. Yet, the wave-like features are derived
from the quantum field. More precisely, the quantum field is manifested with the
particle’s behavior, and the particle manifests the observed reality. The quantum
field and the particle are never separated from each other.6
6 In order to clarify the roles of the particle and the quantum field, let us suppose a radio-controlled
boat. The boat plays the role of the particle. It sails on a lake and follows instructions of a radio-
frequency, or an electromagnetic wave. The radio-frequency wave plays the role of the quantum
A.1 Appendix: Bohm’s Quantum Potential 289
Some important properties distinguish the quantum potential. According to our

derivation, the quantum potential has no presence in the Hamiltonian, as it is the
outcome of using a polar function in the description of the particle. The quantum
potential suffers no change by introducing a scaling factor, because the function .ψ
is present in the numerator and denominator of the quantum potential, producing
the cancellation of the scaling factor. Therefore, contrary to classical waves, like the
ocean waves which have dependence on the field intensity, the quantum potential is
independent of the field strength, but it depends on the properties of .ψ.
Since the quantum potential shows no dependence on the quantum field intensity,
the distant environmental effects may have repercussions on the properties of the
particle. In order to illustrate this last point, let us consider the Bohm potential 12.62
as part of the force exerted on the particle [25].
dPα
.
dt = −[∇α Velec + ∇α Vbohm ] (12.63)
The first term is the classical force and the second one is the quantum force. In
the absence of the quantum force, the classical behavior of the particle satisfies
the Newtonian equation .dPα /dt = −∇α Velec [26]. On the other hand, in the
absence of the classical force, the behavior of the particle satisfies the equation
.dPα /dt = −∇α Vbohm . In a region of empty space, it is possible to have a particle
traveling in such a region in a nonlinear path, because it may respond to quite distant
effects, giving evidence of non-local interactions between particles, such as the
action-at-a-distance, also called spooky action. In general, the non-local quantum
connection is broken with small disturbances, such as those corresponding to small
temperature gradients. Special conditions are required to maintain the control of
a fragile quantum connection. The non-locality of Bohm’s quantum mechanics is
consistent with relativistic quantum field theories by avoiding the transmission of
signals faster than the speed of light [27].
In the superconducting state, the quantum potential is responsible for the strong
correlations that keep the atoms in an organized collective state. Nevertheless,
when the temperature is increased, the particles are scattered, breaking their
organized behavior, and the multidimensional character of the quantum field is
partitioned. A bifurcation of the information occurs in this case, which results in
the particles taking different information channels and trajectories. The quantum
field is also capable to give sensible explanations of the quantum transitions, as well
as explaining the fusion and fission of subatomic particles.
In the double-slit experiment, the wave travels through both slits and interferes
with itself, always driving the particle toward the screen, but the particle only goes
through a single slit and is deflected according to the initial conditions imposed
in the experiment. The result of dealing with many particles is an ensemble of
trajectories forming a diffraction pattern. When the quantum field is defined, for
field. The instructions given to the boat are independent of the wave intensity, but depend on the
wave frequency.
example by solving the wave equation, the quantum field carries information of
both slits, and thus, the experimental results concerning the particle are determined.
According to Bohm’s theory, there is a cause, with the observer playing no relevant
role in such a situation. The interpretation of quantum mechanics based on the
existence of a quantum potential indicates that, in a measurement process, the
measuring device and the object being measured constitute an indivisible system.
The quantum field is understood as the entity that carries information, while the
particle is the object that carries the main portion of the energy.
The quantum potential integrates the information of all the particles by consider-
ing the non-local connection among them. In Bohmian mechanics, the state of the
system is specified by the set .{ψ, Q}, where .ψ is the wave function of the many
particle system, and Q is the set of particle positions in 3D space. No classical
limit is required in Bohm’s interpretation of quantum mechanics. However, when
the classical potential is larger than the quantum potential, the classical behavior
of the particles dominates the dynamics. According to the traditional interpretation
of quantum mechanics, the classical limit is achieved in the limit of large quantum
numbers, problem 12.3, and when the potential changes little with respect to the
wavelength of the particles, Eq. 12.22. On the other hand, the quantum potential
is relatively large in the case of short wavelengths, indicating that a system under
special conditions may still retain a dominant quantum character at high quantum
numbers, thus contradicting the Bohr’s correspondence principle. In other words,
the classical limit is not always achieved at the macroscopic scale due to the
possibility to have a significant contribution of the quantum potential.
Bohm’s interpretation of quantum mechanics is a formulation that deals with
the reality at every scale level, avoiding difficulties at the overlap domain of
classical mechanics and quantum mechanics, which have been observed in other
interpretations. Still, the physical origins of the quantum potential are unknown.
The quantum field and the particle together constitute a complex structure (recall
that measurable distances in physics can reach a resolution of .10−18 m, and the
Planck distance is .10−35 m, giving opportunity to discover new structures at smaller
dimensions). The main goal of Bohm’s interpretation is to provide an intuitive basis
of quantum mechanics after following some similarities with classical mechanics.
The new philosophy on quantum mechanics is only possible when we are open to
consider new ways of thinking on the properties of matter and embrace ideas and
concepts of different nature to those given by previous interpretations of quantum
mechanics.
References
1. A. Bokulich, Bohr’s Correspondence Principle, The Stanford Encyclopedia of Philosophy,
Spring 2014. ed. by E.N. Zalta
2. J.C. Tully, Mixed quantum-classical dynamics. Faraday Discuss. 110, 407–419 (1998)
3. J.R. Reitz, F.J. Milford, Foundations of Electromagnetic Theory (Addison-Wesley, Boston,
MA, 1969), Sect. 7.2
References 291
4. A.S. Davydov, Quantum Mechanics. International Series in Natural Philosophy, 2nd edn., vol.
1 (Pergamon, Oxford, 1985), Sect. 21, pp. 71–73
5. G.F. Torres del Castillo, C. Sosa Sánchez, Solutions of the Schrödinger equation given by
solutions of the Hamilton-Jacobi equation. Rev. Mex. Fis. 62, 534–537 (2016)
6. G.F. Torres del Castillo, I. Rubalcava-García, Solutions of the Helmholtz equation given by
solutions of the eikonal equation. Rev. Mex. Fis. 63, 211–213 (2017)
7. H. Goldstein, Classical Mechanics, 2nd edn. (Addison-Wesley, London, 1980), Sect. 10.8, pp.
484–492
8. T.L. Hill, Statistical Mechanics, Principles and Selected Applications (Dover, New York,
1987), Appendix 6, p. 407.
9. Cerius2 , Forcefield Based Simulations: General Theory and Methodology (MSI, San Diego,
1997), p. 31
10. I.N. Levine, Quantum Chemistry, 4th edn. (Prentice Hall, New Jersey, 1991), Sect 14.3–14.4,
pp. 445–453
11. G. Herzberg, Molecular Spectra and Molecular Structure: Infrared and Raman Spectra of
Polyatomic Molecules, vol. 2 (Van Nostrand, New York, 1945), Chapt. II
12. R. Santamaria, E. Charro, A. Zacarías, M. Castro, Vibrational spectra of nucleic acid bases and
their Watson-Crick pair complexes. J. Comput. Chem. 20, 511–530 (1999)
13. P. Mollinedo-Rosado, R. Santamaria, Structural and electronic changes of cytosine upon
addition of atomic sulfur. J. Molec. Struct. (Theochem) 805, 111–117 (2007)
14. O.J. Gutíerrez-Varela, R. Santamaria, Molecular nature of the drag force. J. Molec. Liq. 338,
116466 (2021)
15. L.A. Girifalco, Statistical Mechanics of Solids (Oxford University Press, Oxford, 2000), Chapt.
17
16. Research Analysis Group, National Research Council, Physics of Sound Waves, Part 1:
Transmission (Murray Printing Company, Wakefield, 1949), Chapt. 2, Eqs. 2, 25
17. A. Tveito, H.P. Langtangen, B.F. Nielsen, X. Cai, Elements of scientific computing (Springer,
Heidelberg, 2010)
18. L. Caretto, Solution of the diffusion equation (California State University, College of Engineer-
ing and Computer Science, Mechanical Engineering Department, Northridge, 2009)
19. T.S. Ursell, The diffusion equation, a multi-dimensional tutorial, Tech. Rep. (California
Institute of Technology, Pasadena, 2007)
20. P.M. Chaikin, T.C. Lubensky, Principles of Condensed Matter Physics (Cambridge University
Press, Cambridge, 2000), Chapt. 7, Sect. 7.4, Subsect. 1, 4
21. O.J. Gutíerrez-Varela, R. Santamaria, Molecular nature of the drag force. J. Molec. Liq. 338,
116466 (2021)
22. J.C. Burant, J.C. Tully, Nonadiabatic dynamics via the classical limit Schrödinger equation.
J. Chem. Phys. 112, 6097–6103 (2000). D.S. Sholl, J.C. Tully, A generalized surface hopping
method. J. Chem. Phys. 109, 7702–7710 (1998)
23. J.C. Tully, Mixed quantum-classical dynamics: mean-field and surface-hopping in Classical
and Quantum Dynamics in Condensed Phase Simulations, ed. by B.J. Berne, G. Ciccotti, D.F.
Coker (World Scientific, Singapore, 1998), Chapt. 21, pp. 489–514
24. D. Bohm, B.J. Hiley, P.N. Kaloyerou, An ontological basis for the quantum theory. Phys. Rep.
144, 321–375 (1987)
25. D. Dürr, S. Goldstein, N. Zanghì, Seven steps towards the classical world, in Quantum Physics
without Quantum Philosophy (Springer, Berlin, Heidelberg, 2013), pp. 171–182
26. B.F.E. Curchod, I. Tavernelli, On trajectory-based nonadiabatic dynamics: Bohmian dynamics
versus trajectory surface hopping. J. Chem. Phys. 138, 184112 (2013)
27. D. Bohm, B.J. Hiley, Measurement understood through the quantum potential approach.
Found. Phys. 14, 255–274 (1984)
Part IV
13
The complexity of quantum theory prevents its application to systems with a

large number of particles. The most important problems to reach a quantum-
mechanical description of molecules are concerned with the solution of difficult
differential equations of motion, the optimization of a great number of degrees of
freedom to determine the proper structures, the diagonalization of big matrices to
establish the energies, the integration of complex functions to determine the average
potentials, and other algebraic and computational complications. Fortunately, in
many instances the atoms behave in the classical sense, and it is possible to
use classical mechanics in such instances. The possibility to deal with classical
atoms circumvents many situations where the application of quantum mechanics
is difficult. In fact, under such a perspective, a new branch of molecular dynamics is
created. It carries the name classical molecular dynamics. In this context, classical
molecular dynamics is shown as an alternative to quantum molecular dynamics to
describe many-body systems at the atomic level. Classical molecular dynamics was
derived from quantum molecular dynamics under limiting conditions in the previous
sections; thereby, we apply Newtonian mechanics infused with important concepts,
methods, and algorithms of Lagrangian and Hamiltonian mechanics to unfold the
classical molecular dynamics theory.
One of the central points in classical molecular dynamics is the approximation
of the forces. The static and dynamic correlations of all the particles complicate the
formulation of the atomic forces. The particle correlations, not fully understood to
present, constitute a general problem known as the many-body problem. The theory
of classical molecular dynamics employs model interatomic potentials to mimic the
interactions of quantum character and derive the forces from such potentials. Thus,
classical molecular dynamics (CMD) reaches simplicity by implementing model
interaction potentials and classical forces.
The CMD approach is capable to deal with a large number of particles because it
has been designed to be computationally efficient. In consequence, CMD allows the
evaluation of different statistical and thermodynamic observables. By using CMD
we are able to compute the time evolution of physical properties of many-particle
296 13 Classical Molecular Dynamics
systems and, for example, design new compounds like proteins, paying attention to
their folding state to fit pharmacological purposes. The CMD approach is nowadays
an integrated part of most software codes. Still, the application of less accurate
methods than quantum mechanics should be always examined in detail and wise
judgment in advance.
In the previous chapters, we discussed the different methods of classical mechan-
ics and formulated their respective equations of motion. The forces were assumed
to be known there with the purpose of focusing attention on the derivation of the
motion equations. The interest in this chapter is to provide, in a first stage, a general
classification of the model potential functions. To do this, we separate atoms of
biological nature from metallic atoms for the building of the respective potential
functions. In a second stage, the model potential functions are discussed for the
two-body, three-body, and four-body interactions. These all functions are applied
to bonded atoms. The potential functions of non-bonded atoms are also introduced
to simulate the electrostatic and van der Waals distant interactions. Other types of
potential functions for different purposes of the atomic phenomena are presented.
The parametrization of the potential functions is discussed as well.1
13.1 Model Interaction Potentials
Classical molecular dynamics is an approach integrated by model interaction

potentials, considered to be functional energy expressions. They emulate the
interactions among atoms, which are used to investigate the behavior of molecules.
The functional expressions of the energy represent one of the fundamental aspects
to reduce the computer time and the computer memory in the simulations. The
CMD approach is capable to reproduce the collective features of the atoms,
reaching consistency with the laboratory observations on the physical phenomena
of molecules. The model interaction potentials are formulated after solving the
electronic wave equation of a system with a number of atoms.
. Ĥ ψ(r1 , · · · , rN ) = E ψ(r1 , · · · , rN ) (13.1)

r1 ,··· ,rN
The electronic energy eigenvalue, E, depends in parametric form on the atomic

coordinates .ri , .i = 1, ..., N. Thus, by departing from a quantum energy, it is
possible to propose a model interaction potential, V [3].
. E = V (r1 , · · · , rN ) ; model interaction potential (13.2)

r1 ,··· ,rN
1 We follow the works: Maitland et al. [1], Chapt. 1 and Rappe and Casewith [2].
13.1 Model Interaction Potentials 297
The model interaction potentials, MI P s, allow to carry simulations of molecules

in solution, large clusters, extended materials like polymers, ceramics, solids, fluids,
etc. There are no formal rules in the classical modeling of the quantum interactions;
nonetheless, the quantum interactions suggest some functional expressions of the
MI P s. It is common to start working with simple MI P s, whose parameters are
tuned to reproduce different experimental facts of the compounds under investi-
gation. The MI P s are designed according to the degrees of freedom which are
characteristic of the molecules. The degrees of freedom frequently correspond to
the motion of pairs, triplets, and quadruplets of bonded atoms. The simplest model
interaction potentials that describe the interactions of atoms like H , C, N, O, P ,
S, etc., of biological compounds, are collected in a scheme that carries the name
forcefield approach. This is, a forcefield represents a collection of MI P s for atoms
of biological nature. There are also MI P s of non-biological compounds, which are
mainly used in materials science. The set of MI P s that simulate the behavior of
atoms like Cu, Ag, Au, P d, P t, etc., in materials are called many-body potentials.
The discussion and methods given below apply equally well to both biological
and non-biological compounds. The incorporation of the MI P s in the Newtonian
equations of motion generates the forces that unfold the dynamics of the atoms.
Thus, the motion equation of the ith atom employs the model interaction potential
V.
2
. m ddtr2i = −∇i V (r1 , · · · , rN ) (13.3)
The potential function has dependence on all the atom coordinates and the
parameters for the interaction of the ith atom with the other atoms of the system.
Classical molecular dynamics deals with the proposal and evaluations of the
potential V . The potential function is frequently approximated in piecewise form.
E E E
. V (r1 , · · · , rN ) = i V1 (ri ) + i<j V2 (ri , rj ) +
V3 (ri , rj , rk ) + · · ·
i<j <k
(13.4)
The first term contains the individual atomic energies, the second represents
the two-body energies, the third contribution gives the three-body energies, and
so on. The summation indexes run over the atoms of the particular interaction
groups. The external potentials, like a confining potential and a gravity potential,
may be equally included in the expression of V . The model interaction potentials
.V1 , .V2 , .V3 , etc., are originated by the attraction and repulsion of the atoms and
are frequently proposed after employing first-principles calculations. This process

implies knowing the solution of the wave equation of small systems. However, the
proposal of MI P s may be also achieved by observing the behavior of atoms in
the molecules of the lab and establishing the parameter values of MI P s from the
experimental measurements.
One of the main drawbacks of using model interaction potentials is the averaging
of electron effects. More precisely, the parameters of the MI P s include the elec-
tronic effects in average form, restricting the use of classical molecular dynamics to
the domains of the parametrized molecular conformations. In this regard, the CMD
simulations overlook the breaking of bonds, or other effects where a substantial
redistribution of the electron cloud occurs with the molecular structural changes.
This is the case of the electronic transitions, the tunnel effect, electric currents, the
interaction of radiation with matter, chemical reactions, etc. In the following, the
MI P s are discussed within the forcefield approach.
13.2 Forcefields
A forcefield is a collection of model interaction potentials for a given set of

molecules of biological nature, in an effort to simplify the atomic interactions
for computational purposes, but without missing the key ingredients and general
features of the atomic interactions. The success of a forcefield, F F , is to a
great extent based on the faithfulness of the MI P s on reproducing the atomic
interactions, with significant reduction of the computational times and memory.
Each MI P of a F F is associated with a degree of freedom characteristic of the
molecule; however, the number of degrees of freedom described by a F F does not
necessarily correspond to the number of degrees of freedom of the molecular phase
space. A F F is also constituted by parameters, which are specific to both the model
interaction potentials and the molecular system. The parameters are determined by
reproducing particular properties of the real system. The distinct F F differ in the
types and numbers of the MI P s. Thereby, a F F has its own particular composition,
with its own parameters.
The model interaction potentials are classified in terms of bonded, .Vb , and
non-bonded, .Vnb , interactions, or intramolecular and intermolecular interactions,
respectively. The terms .Vb and .Vnb contain the parameters determined from
experimental and theoretical sources of information.
. V = Vb + Vnb (13.5)
The bonded interactions consider linked atoms and are modeled using sets of
two, three, and four atoms. The bonded interactions deal with bond elongation and
bond compression, changes of bond angles, and torsion effects of groups of atoms
with up to four atoms. On the other side, the non-bonded interactions consider
long-range forces due to charged atoms. The non-bonded interactions include the
electrostatic interactions and the van der Waals forces. The long-range forces are
the main cause of molecules to change conformations to reach structural stability
in a given environment. The MI P s and the motion equations coil together to
make classical molecular dynamics a concerted approach in the investigation of the
structural and energetic changes of molecules, with linkage to molecular reactivity.
These all aspects are important, for example, in elucidating the structure-reactivity
relationship of proteins.
The next step is to discuss the individual interaction potential functions. They
are given in terms of bond distances, bond angles, atom charges, and more. It is
13.3 Atom Types 299
important to remark that there are no formal rules in the proposal of the MI P s.
For such a reason, it is common to find a variety of expressions describing the
same type of interaction in the different software codes implementing classical
molecular dynamics. Thereby, classical molecular dynamics opens the opportunity
for creativity.
13.3 Atom Types
The forcefield approach introduces the concept of atom type, where every atom
of the molecule is classified not only according to the periodic table but also with
respect to the close surrounding environment. For example, a carbon atom bonded to
an oxygen atom, .C − O, classifies differently in reference to a carbon atom bonded
to a nitrogen atom, .C − N , due to the neighbors O and N of C, respectively. The
classification process considers not only the different elements that an atom may link
to but also the types of bonds that it may establish with other atoms. For instance,
there are different types of bonds, single, double, and triple bonds, among carbon
atoms, and from a sharp perspective, the atoms are additionally classified according
to their chemical environment, Table 13.1. The concept of atom type in the F F
approach enlarges the classification provided by the periodic table of the elements.
It is in such a fashion that a F F approach accounts for the behavior of atoms in
molecules. In this regard, CMD is sufficiently efficient and relatively accurate.
A main advantage of classical molecular dynamics over the quantum-mechanical
calculations is the no necessity of re-computing the electron cloud distribution at
every time step of the simulation to retain good accuracy.
In most situations, the atom types are assigned by the user based on both the
classification and the nomenclature rules of the computational program in use.
The atom types which are automatically assigned by some programs are an error-
prone process. The assignation process of atom types is one of the initial steps
in the construction of the molecular model; it is based on the correct chemical
identification of the molecular fragments and represents a critical point to obtain
acceptable results.
Table 13.1 Atom types of C Atom type Assignment

according to the bond type
C Carbonyl C, peptide backbone
with other atomsa
CA Aromatic C
CT 1 Aliphatic .sp 3 C for CH
CT 2 Aliphatic .sp 3 C for .CH2
CT 3 Aliphatic .sp 3 C for .CH3
CS thiolate C
a The atom-type nomenclature changes
from one software code to another
13.4 United Atom
The concept of atom type may be generalized to introduce the united atom approach.
In this approximation, the terminal carbons and the terminal nitrogens, together
with their respective hydrogens, are redefined as either a single atom, united atom,
or super atom. It is usually applied to atomic groups like .CH3 , .NH2 , .CH2 , and
carbonyl and hydroxyl groups, treating each group as a single atom, with its own
model potential function for the interaction with other backbone atoms. The purpose
of the united atom is to save computer time on the atom groups that many times play
second roles in the simulations. The united atom approximation is a coarse-grained
approach and, in principle, is simpler to implement than the all-atom forcefield
approach.
13.5 Bond Elongation and Compression
It is time to discuss the individual terms of the forcefield, Eq. 13.4. The first term,
.V1 , involves the independent atomic energies. It constitutes a constant energy when
the numbers and types of atoms have no change in the simulation and may be
neglected since the important energies are the relative energies. The second term,
.V2 , represents the total energy of the two-body interactions.
rij = |ri − rj |
. E E E ∂V2 (rij ) rij
(13.6)
V2 (r1 , · · · , rN ) = i j >i V2 (rij ) ; Fi = − j ∂rij rij
The term .V2 (rij ) is a pairwise potential which applies to bonded atoms. To avoid
confusions, it shall be denoted by .Vb . The force .Fi is derived from the pair potential.
It is the force exerted by all the other atoms on atom i. According to .3rd Newton’s
law, the force .Fij over the ith atom due to the j th atom is the same as the force .Fj i
exerted over the j th atom by the ith atom, but with opposite sign, .Fij = −Fj i . The
use of .3rd Newton’s law reduces the number of computations on the atomic forces.
The pair potential of a forcefield is based on the natural elongation and
compression (vibration) of a molecular bond. Such a potential is usually of harmonic
type. Thus, we have the harmonic potential, Fig. 13.1.
. Vb = k(rij − r0ij )2 /2 (13.7)
The term .Vb represents the bond-stretching interaction between the ith and
j th atoms. The elongation and compression are measured with respect to the
equilibrium distance .r0ij . The spring stiffness or force constant is k; it depends on
the types of atoms taking part in the interaction (the indexes ij in the notation
of the force constant k are dropped for simplicity). The constant computations
of the electron cloud re-arrangements of a bonded atomic dimer in a quantum-
13.5 Bond Elongation and Compression 301
Fig. 13.1 The figures of the top and bottom illustrate the bond-distance and bond-angle vibra-
tions. Such degrees of freedom are frequently simulated with the harmonic potential functions
'
.Vb = k(rij − rij ) /2 and .Va = k (θij k − θij k ) /2. The parameters .rij and .θij k are the reference
0 2 0 2 0 0
'
bond-distance and the reference bond-angle, and k and .k are the spring constants
mechanical calculation are greatly simplified in the classical approach with the
harmonic potential, which may be corrected by introducing anharmonic effects.
[ ]
. Vb = k (rij − r0ij )2 − α(rij − r0ij )3 /2 (13.8)
The constant .α is the weight of the anharmonic term gauged with respect to the
harmonic one. The addition of the anharmonic term makes the potential more steep
at small distances and less steep at large distances, breaking the symmetry imposed
by the harmonic term at large distances and, in principle, producing a more realistic
behavior. The parameters k and .r0ij of potential 13.8 are not necessarily the same
as those of Eq. 13.7. The potential 13.8 may be further improved with a quartic
term, .β(rij − r0ij )4 , due to the fact that a cubic anharmonic term may overvalue the
negative contribution of the potential. In most instances of molecular behavior, no
extra contributions are required beyond the quartic term in the expression of .Vb .
There are alternative proposals to the harmonic potential. An important proposal
describes covalent bonds with pronounced elongations, like the Morse potential.
( )2
. Vb = V0 1 − exp[−α(rij − rij0 )] − V0 (13.9)
The parameters in the above expression are the reference energy, .V0 , the
reference distance, .rij0 , and the exponential factor, .α. Another potential for covalent
bonds is the Lennard-Jones potential with parameters .e, .σ .
[ ]
. Vb = 4e (σ/rij )12 − (σ/rij )6 (13.10)
The Buckingham potential, with parameters .V0 , .ρ, and .C6 , is another expression
that may be used to describe covalent bonds.
. Vb = V0 exp[−rij /ρ] − C6 /rij 6 (13.11)
Once the interaction potentials are defined, the force exerted on the ith atom by
the j th atom is obtained by differentiating .Vb with respect to the position of atom i,
and employing the chain rule.
( )
∂Vb ∂rij ∂rij ∂rij ∂Vb rij
. fi = −∇i Vb = − ∂rij ∂xi , ∂yi , ∂zi = − ∂rij rij (13.12)
When the potential cannot be described analytically, it is possible to deal with

numeric potentials, whose values are tabulated. There are more expressions to
describe the covalent bonds in the literature; nevertheless, the ones presented
previously are well-known and have been tested in many occasions, rendering
acceptable results on the vibration of atom pairs.
13.6 Combination Rules
In many instances, the model interaction potentials are applied to systems integrated
by different types of atoms. In these cases, we are required to deal with parameters
for the interaction of atoms of different species. One of the straight ways to
obtain such parameters without much additional work is to use the combination
rules. According to those rules, the interaction parameters of atoms of type a and
the interaction parameters of atoms of type b may be combined to generate the
parameters of the interaction of unlike atoms, ab. There are no formal proposals
in the combination of the parameters, because they are intertwined with their
MI P s. Nevertheless, there are some empirical strategies. If the atomic species
a and b constitute the system, and the interaction potentials of the like species
are of Lennard-Jones type, Eq. 13.10, then the crossed parameters ab may be
approximated in any of the following ways.
eab = (ea eb )1/2 ; σab = (σa + σb )/2
eab = 2 eeaa+e
eb
b
; σab = (σa + σb )/2
. (13.13)
eab = (ea eb )1/2 ; σab = (σa σb )1/2
rij = (ri rj )1/2 ; rij = (ri + rj )/2
More combination strategies are found elsewhere [4].

13.7 Bond Angle Vibration 303
13.7 Bond Angle Vibration
The model bond functions, .Vb , and the forces derived from them are not enough to
describe the many degrees of freedom of a polyatomic molecule in most instances.
We require to introduce further terms like the .V3 and .V4 , which take part in Eq 13.4
to guarantee the molecular stability.2 The purpose of this section is to discuss the
term .V3 of Eq. 13.4.
When atom i is linked to atom j and also to atom k, the bond angle between the
bonds ij and ik is formed. The bond angle vibrates describing a kind of scissoring
motion, Fig. 13.1. The respective potential function is similar to the bond-stretching
potential; however, the bond angle takes part in this case.
( )
rij ·rik
. Va = k(θij k − θij0 k )2 /2 ; θij k = arccos rij rik (13.14)
The above function is known as the bond-angle harmonic potential. The

reference angle formed by vectors .rij and .rj k of the atoms in equilibrium is
0
.θ
ij k . It is possible to introduce correction terms similar to those of the bond
elongation-compression potential by considering higher-order terms up to the
quartic contribution.
. Va = k(θij k − θij0 k )2 /2 + k ' (θij k − θij0 k )3 /3 + k '' (θij k − θij0 k )4 /4 (13.15)
We may also insert terms with different character, such as those that mix a bond-
vibration term with a bond-angle vibration term in Eq. 13.14.
. Vmix = V0 (θij k − θij0 k ) (rij − rij0 ) (rik − rik

0)
(13.16)
The expression 13.16 is a crossed-terms equation which couples a bond vibration

to an angle vibration. The constants of the potential are the reference distances .rij0 ,
0
.r , reference angle .θ
0 ' ''
ik ij k , force constants k, .k , .k , and factor .V0 . An alternative form
of the potential .Va is the harmonic cosine expression.
. Va = k[cos(θij k ) − cos(θij0 k )]2 /2 (13.17)
There are many more expressions for the bond-angle potential, frequently
formulated in the general form .Va = V0 (θij k ) F (rij ) F (rik ) F (rj k ). The proposal
of bond-angle potentials in numeric form is equally possible.
2 For the number of terms to achieve the stability of a cluster, refer to: Kaplan et al. [5] and Kaplan
et al. [6].
13.8 Plane Bending
There are groups of atoms located in planes like in the aromatic rings and other
conjugated systems in many molecules. The dynamics of the molecule may induce
the bending vibration of the planar atoms. It is possible to propose a potential
function that describes the vibration of sets of four atoms, where three atoms of them
form a plane, and other three atoms of them form a different plane, Fig. 13.2. The
vibration of a plane with respect to the other plane can be described by a dihedral-
angle harmonic potential.
[ ]
(rij ×rj k ) · (rj k ×rkl )
. Vp = k(θij kl − θij0 kl )2 /2 ; θij kl = arccos |rij ×rj k | |rj k ×rkl | (13.18)
The variable .θij kl is the instantaneous angle between the two planes, and .θij0 kl
the angle of equilibrium. When .θij0 kl = 0, the potential imposes the planarity of the
four atoms. This type of potential is frequently applied to the planar peptide bonds
of proteins. In all the previous model potential functions, the deviations from the
reference values produce a 3D conformational change and, consequently, an energy
change. In this regard, the forcefield approach always requires an initial reference
conformation of the molecule and the corresponding energy of reference. Other
potentials describing the bending of planes are the cosine potential and harmonic
cosine potential.
. Vp = V0 [1 + cos(αθij kl − β)] ; Vp = k[cos(θij kl ) − cos(θij0 kl )]2 /2

(13.19)
We may also deal with different short series of cosine functions to describe the
torsional effects.
Fig. 13.2 Two different molecules with four bonded atoms are shown. In the left figure, the atoms
ij k form a plane, and the atoms j kl form a different plane. In the right figure, the atoms ij k form
a plane, and the atoms ij l form a different plane. In each case, the angle between planes is .θij kl .
The plane-bending degree of freedom demands a model interaction potential in a forcefield. Such
a potential may adopt any of the forms 13.18–13.20
13.9 Angle Inversion 305
Vp = (k/2)[1 + cos(θij kl )] + (k ' /2)[1 + cos(2θij kl )] + (k '' /2)[1 + cos(3θij kl )]

. Vp = (k/2)[a0 + a1 cos(θij kl ) + a2 cos2 (θij kl ) + a3 cos3 (θij kl )
+ a4 cos4 (θij kl ) + a5 cos5 (θij kl )]
(13.20)
The four-body terms are computationally more expensive than the three-body
and two-body terms. The expressions of the forces that have to be derived
from complicated model potential functions may be obtained using a symbolic-
algebra software or an artificial-intelligence software, to directly translate onto a
programming language.
13.9 Angle Inversion
We frequently have an atom surrounded by three other atoms in symmetric form,

like a nitrogen atom in ammonia, .NH3 . The central atom is in many occasions
able to invert positions, producing a chiral change. The inversion of positions is
described by an inversion potential, which may adopt simple forms like those of
harmonic type.
. Vin = k(θij kl − θij0 kl )2 /2 ; Vin = k[cos(θij kl ) − cos(θij0 kl )]2 /2 (13.21)
The inversion process of the positions may be equally described by a planar

potential.
. Vin = k2 [1 − cos(nθij kl − θij0 kl )] (13.22)
If we consider that atoms j , k, l surround atom i, then the angle .θij kl corresponds
to the angle described by the vector .rij and the ikl plane.
Problem 13.1 Determine the angle .θij kl of Eqs. 13.21 and 13.22 under the
supposition that atoms j kl surround atom i, in the way it happens with the
hydrogens of nitrogen in the ammonia molecule.
Solution The strategy to determine .θij kl consists of determining two copla-

nar vectors in the plane ikl. To do this, we define the basis vectors:
r̂ik + r̂il r̂ik − r̂il

ûkl =
. ; v̂kl =
|r̂ik + r̂il | |r̂ik − r̂il |
(continued)
The unit vectors .ûkl and .v̂kl are described as the sum and the difference of unit
vectors .r̂ik and .r̂il . The basis vectors lie in the ikl plane. From the projection
of vector .rij onto the basis vector .(ûkl , v̂kl ), we define the new vector .skl on
the ikl plane.
. skl = (rij · ûkl )ûkl + (rij · v̂kl )v̂kl
The dot product between the coplanar vector .skl and the non-planar vector .rij
gives the angle between the vector .rij and the ikl plane.
( )
rij · skl
θij kl
. = arccos
rij skl
13.10 Torsional Motion
In a molecule the rotation of atomic groups around a central bond is commonly

observed. For example, for the group of atoms .X − CH3 , we have the central bond
.X − C. The hydrogens of the .CH3 group face low energy barriers in the rotation
around bond .X − C. When a rotation of the hydrogen bonds around the central bond
.X−C occurs, the group of atoms achieves a conformation which is indistinguishable
from the initial one. In this regard, the potential function of a rotating atomic group
may be described by a periodic potential.
. Vt = (k/2)[1 + cos φ] + (k ' /2)[1 + cos(2φ)] + (k '' /2)[1 + cos(3φ)] (13.23)
The parameters of the potential are the constants k, .k ' , and .k '' . The angle .φ is a
dihedral angle. The second term has a greater periodicity than the first one; in turn,
the third term has a greater periodicity than the second one. The terms introduce the
number and the height of peaks and subpeaks between major peaks, according to
the .spi -bond hybridization, Fig. 13.3. The potential only applies to groups of atoms
linked to a central single bond, since the group of atoms show no rotation around
double and triple bonds.
The model potential functions of the different degrees of freedom presented so far
produce forces and energies of different magnitude, because the potentials describe
interactions of different nature, contributing in different amounts to the molecular
energy. The energy contributions increase from the soft to the hard degrees of
freedom. It takes more energy to distort a bond distance than a bond angle; in turn,
it is energetically more costly to distort a bond angle than a torsion angle with small
energy barriers. Such observations have been confirmed by establishing the infrared
and Raman vibrational spectra of the DN A nucleic acid bases and their Watson-
Crick pair complexes, where different types of individual and collective motions
have been assessed and classified with different frequencies [7].
13.11 Electrostatic Interaction 307
Fig. 13.3 The plots in red, green, and blue colors represent the functions .cos(φ), .[1 + cos(φ)] +
[1+cos(2φ)] and .[1+cos(φ)]+[1+cos(2φ)]+[1+cos(3φ)] in terms of the angle .φ, respectively.
The last two functions introduce one and two subpeaks between the main peaks of .cos(φ). In
general, for n-fold rotational barriers, there are n additional functions for the cosine function
13.11 Electrostatic Interaction
The intramolecular and short-range interactions have been modeled using potential
energy functions in the previous sections. It is time to deal with the intermolecular
and long-range interactions. We start by considering that the atoms of a molecule are
partially charged as they invest part of their electron charge in forming bonds (it is
common to speak of polar bonds). The charged atoms interact with other relatively
far apart charged atoms of the molecule and also with the solvent atoms according
to the electrostatic Coulomb potential.
1 qi qj
. Vel = 4π ε rij ; rij = |rij | = |ri − rj | (13.24)
The parameters of the potential are the partial charges .qi and .qj of the ith and j th
atoms. They are maintained constant in the course of a simulation. The interatomic
separation distance is .rij . The constant .ε is the relative permittivity of the medium.
It is a term representing the screening effects of the medium in the electrostatic
interactions of atoms. In manyinstances, .ε is modified to account in average manner
for the interactions with distant atoms.
Vel = [4πg(rij )]−1 qi qj /rij ; g(r) = αεr

. (13.25)
fi = (2π αε)−1 qi qj rij /rij4
In this expressions, the dielectric g has linear dependence on the distance, as .α

is a constant. The force .fi exerted on the ith atom derived from the potential is also
shown. The expressions reduce the computational times because the electrostatic
energy is made to converge faster by considering the average effect of distant
charged atoms.
The monopole interactions 13.24 represent one of the computational expensive

parts of a simulation due to the pairwise nature and slow decay of the electrostatic
interaction. Such effects, together with a poor computational efficiency of the
mainframes, constitute some of the main reasons to create novel methods, and even
new hardware, for the rapid convergence of the electrostatic energy, especially for
large systems such as a molecule immersed in a liquid. The novel methods are
classified within the implicit-solvent approach or explicit-solvent approach. The
implicit-solvent approaches are, for instance, the continuum dielectric and static
dipolar field approximations, where the presence of the liquid molecules is avoided,
and only their effects on the compound under investigation are considered. The
explicit-solvent approaches involve the Ewald summation, and the approximations
of truncated potentials, where the explicit presence of the liquid molecules is
considered, and the interactions are simplified. The different approaches may be
combined to lead, for example, to the reaction field scheme together with the fast
multipole approximation. Some of these relevant methods are presented later.
13.12 van der Waals Forces
The van der Waals forces are important when the interacting molecules are not too
close but also not too distant apart from each other. The van der Waals forces
are the result of the mutual distortions of the electron clouds of the interacting
molecules at intermediate distances, because in every molecule the electrons are
in constant motion and are susceptible to a spatial re-arrangement for the electric
fields of the other molecules. In any particle system, the molecular interactions
introduce a correlation of the fluctuating electron clouds, resulting in the formation
of instantaneous multipoles (dipole, quadrupole, and more). The van der Waals
interaction potential is due to the induced multipoles of the molecules interacting
at intermediate distances. The interaction potential is of quantum nature as the zero-
point vibration energy of a molecule provides a contribution to the dispersion energy
(the calculation of the dispersion energy of a pair of model interacting molecules is
illustrated in Sect. A.1). This fact cannot be explained using a classical description
of the electron cloud [8].
The van der Waals interactions are non-bonded forces and lead to the dispersion
energy, usually of attractive character at semi-large distances. The van der Waals
potential also contains a repulsive contribution at short distances. It arises from
the partial overlap of the electron clouds, which have quantum character, and
with origins on the Pauli exclusion principle. The short-range and semi-long-
range contributions of the van der Waals forces are, in a few words, due to the
electron correlations which are not considered by the classical electrostatic forces.
In this connection, the electrostatic force due to point charges or electric monopoles
classifies as a first-order energy because it involves no deformation of electron
clouds; however, the dispersion energy classifies as a second-order energy because
it involves a small deformation of electron clouds. The proper behavior of the
13.12 van der Waals Forces 309
interactions at the intermediate range of distances is a complicated problem to solve

since, at short distances, the antisymmetric character of the wave function plays
an important role and, at far distances, the asymmetric part of the wave function
is the dominant contribution. Thus, at the intermediate region, where there is an
overlap of the intramolecular and intermolecular forces, the mixed antisymmetric-
asymmetric nature of the wave function complicates the description of the molecular
interactions. A rough simplification, with the purpose of modeling the forces in
the intermediate region, consists in averaging the short-distance and large-distance
interactions.
The van der Waals forces are weaker than the hydrogen bonds, but they increase
with the delocalization of the electron clouds, particularly for atoms with a big
number of electrons or molecules with significant polarization. The van der Waals
forces are anisotropic, because the polarizability depends on the orientation of the
interacting molecules. This is the reason to deal with the polarizability as a second
rank tensor, where the electric field component in one direction and the dipole
moment component in a different direction are, in fact, related [9].
The inflection point in going from the attractive to the repulsive part of the van
der Waals interaction is an equilibrium point, say .r0 , in the van der Waals potential.
The equilibrium point defines the sum of the van der Waals radii of the interacting
atoms. A common representation of the van der Waals interactions between two
atoms is given by the Lennard-Jones potential [10].
. VW = α/rij12 − β/rij6 (13.26)
The parameters .α and .β depend on the atom types and are determined from
energy values at intermediate atomic distances. The term .rij is the atomic separation.
The repulsive part of the potential is deduced from empirical observations and
corresponds to the first term of Eq. 13.26. The attractive part of the potential is
designed according to the attraction of the dipoles and corresponds to the second
term of Eq. 13.26. The repulsive part grows faster than the attractive part at
short distances. At large distances the attractive part dominates over the repulsive
contribution. The van der Waals potential given above is spherically symmetric
and pairwise additive. The interactions exclusively based on the Lennard-Jones
potential lead to compact structures, like the close-packed hexagonal and face-
centered cubic structures. There are equivalent forms of writing the Lennard-Jones
expression in terms of different pairs of parameters, making it more appealing from
the computational point of view, since one of the terms is calculated as the square
of the other. If we make the transformation .α = 4E0 σ 12 and .β = 4E0 σ 6 , then the
Lennard-Jones expression is:
[( )12 ( )6 ] [( )12 ( )6 ]
r0 r0
. VW = 4E0 σ
rij − σ
rij = E0 rij −2 rij ; r0 = 21/6 σ
(13.27)
The reference energy .E0 is the interaction strength at the equilibrium distance
.r0 and corresponds to the depth of the potential well: when .rij = r0 , the potential
is .VW = −E0 . The quantity .σ is the coordinate where the potential is null: when
.rij = σ , the potential is .VW = 0 and, for such a reason, .σ is interpreted as the
hard-core van der Waals radius.
Problem 13.2 Determine the relation .r0 = 21/6 σ between the equilibrium
distance .r0 and the hard-core radius .σ of the van der Waals potential energy
function, Eq. 13.27.
Solution At the equilibrium distance, .r0 , the force .−∂VW /∂r over the
particles is zero. Thus, by differentiating .VW with respect to r, we have:
[ 12 6
]
∂VW
. ∂r = 4E0 −12 σr 13 + 6 σr 7 =0 ⇒ 12 σ 12 /r013 = 6 σ 6 /r07
r=r0
From the last expression, we obtain .r0 = 21/6 σ .
The smoothness of the repulsive part, first term of Eq. 13.27, may be also
modulated using a decaying exponential function, which is in turn based on the
short-range energy behavior of an anti-bonding molecular orbital, for example, of
the .H2 molecule.
[ ( )6 ]
. VW = γ exp(−r0 /rij ) − r0 /rij (13.28)
The above equation is the Buckingham expression of the van der Waals potential.
The van der Waals forces strongly depend on the environment due to the re-
arrangements of the electron cloud with the nearby molecules, in consequence
provoking the induction and extinction of local dipole moments. The dipoles create
a screening effect, resulting in the reduction of the magnitude of the electrostatic
interactions. Thus, it is common to incorporate additional terms in the forcefield to
take the electronic polarization into consideration and improve the reliability of the
simulations.
The van der Waals and electrostatic contributions are non-bonded potentials of
pairwise nature, and their calculations represent one of the highest computational
costs in the simulations. If we consider that an atom has non-bonded interactions
with all the other atoms of the system, then for a total number of N atoms, there is a
number of .N (N −1)/2 interactions. For a large number N of atoms, the calculations
scale quadratically, .N 2 . For such a reason, the non-bonded interactions are in many
times calculated just for the nearest neighbors. This last process is implemented
by introducing a cutoff distance, with the creation of a list of atom neighbors.
13.13 Interaction Potential Functions of Water 311
The purpose of the list is to keep track of the atom neighbors and compute the
atomic interactions according to such a list. The list assigned to each atom should
be updated on a regular basis because the atoms of the list may diffuse, getting out of
the neighbor limits. Scale factors of the Coulombic and van der Waals interactions
may be also included in the simulations to implicitly consider distant atoms. The
scale factors are estimated by averaging the interactions with distant atoms that
belong to the background. The introduction of scale factors is expected to reduce
the number of non-bonded calculations in the simulations.
In many forcefields, the non-bonded interactions between atoms of a covalent
pair are discarded, since such interactions are already introduced by potential
functions of the type .V2 , which describe the pair interactions, and similarly the
functions .V3 and .V4 .
13.13 Interaction Potential Functions of Water
The dynamics of molecules many times take place in a fluid. The fluid may be
water, benzene, chloroform, acetone, ethanol, etc. The interaction of the fluid with
the molecule may be considered explicitly or implicitly. In the explicit form, the
atoms of the fluid are considered, and a model potential energy function is assigned
to simulate the interactions among atoms, following the rules of the forcefield
approach.
In the case of water, which is the most common fluid, the common energy
potential functions are the T I P nP models, with .n = 3, ..., 6. The atoms of the
water molecule are bonded by distance and angle potentials of the types discussed
in Sects. 13.5 and 13.7. The T I P nP models also introduce n partial charges on the
water atoms and in centers that may be massless but symmetrically located in the
neighborhood of the water atoms. The fictitious charged centers are due to that part
of the electron cloud that has no participation in bond formation. The water molecule
has .C2v symmetry or isosceles-triangle geometry, and the fictitious charged centers
are strategically located, depending on the T I P nP model, around the water atoms.
In the T I P 3P potential energy function, there are no fictitious centers, and all the
charges are situated on the atoms. In the T I P 4P function, three of the charges are
on the atoms, and the extra charge is located on a point of the line bisecting the
angle H OH , in proximity to the oxygen atom. In the T I P 5P function, the two
charged fictitious centers are positioned in such a way that the water atoms plus
the two fictitious centers resemble the geometry of the .CH4 compound with .Td or
tetrahedral symmetry and the oxygen atom positioned at the center of the geometry,
like the carbon of .CH4 . The T I P 6P function includes the renormalized charged
centers of the T I P 4P and T I P 5P models together. In short, the purpose of the
extra charged sites is to delocalize the electronic charge of the atoms in an attempt
to reproduce the quantum-mechanical charge distribution in a kind of coarse-grained
approach.
In addition to the T I P nP water models, other representations like the SP C
(simple point charge) three-center water model are available. The parameters of all
these models are proposed after reproducing the dipole moments of the individual
water molecule, and the collective properties of bulk water, such as the density,
radial distribution function, relative permittivity, vaporization heat, static dielectric
constant, diffusion constant, fluid shear viscosity, and the free energy of hydration,
at a given pressure and temperature. The parametrization process is particular of the
model interaction potentials, which may be also designed for specific purposes, for
example, to investigate the different water phases.
The interactions between the solvated molecule and water atoms are of non-
bonded type, due to the electrostatic and van der Waals forces.
qi qj β
. Vwat = γ + α
− (13.29)
rij rij12 rij6
The Coulomb interactions also consider the offset distances introduced by the
fictitious charged centers. When the structure of the water is of second importance in
the simulation, the pair of bonds OH and the angle H OH may be maintained rigid
using holonomic constraints in the equations of motion. The rigid approximation of
the water molecules allows for computational efficiency, but reduces the accuracy
of the simulations. Further improvements referent to the computational cost are
obtained by: (i) considering a cheap T I P nP water model where .n = 3 or 4,
and (ii) including van der Waals interactions between the oxygen atoms solely,
neglecting the van der Waals interactions among the water hydrogens. A coarse-
grained approach where one, two, or more water molecules are described by just one
or two charged centers, using a kind of united atom strategy, may be also applicable.
A radical strategy consists on avoiding the whole water molecules and introducing
their effects implicitly on the solvated molecule with the assumption of a polarizable
model. However, the approaches introducing the individual water molecules are
accurate and more realistic, since the solvated molecule has the opportunity to
interact with the neighbor atoms of the solvent. Such approaches come with a high
computational cost, as the charged centers of the molecules forming the fluid scale
the number of interactions. A similar discussion applies to solvents different to
water by introducing the proper atom types, interaction potential functions, and their
parameters.
13.14 Atom Polarization
When an atom is immersed in an electric field (it may be an external field or the
field produced by the surrounding molecules), the electron cloud is deformed. The
positive nucleus or ion is displaced toward the electric field .E, while the electrons
move away from the source of the electric field. The net effect of the field on the
atom is the charge redistribution, producing an internal polarization that in turn
builds an electric field to counteract the external field, Fig. 13.4. There are atoms
with different degrees of polarization; this property defines the induced dipole, .μ
[11].
13.14 Atom Polarization 313
Fig. 13.4 In the top figure,

an electron cloud (area
depicted with many negative
signs) surrounds the nucleus
(area depicted with a positive
sign). In the bottom figure,
the atom is immersed in an
electric field .E, and as a
result, the electron cloud is
distorted. A dipole is created
due to a shift d of the
negative charge with respect
to the positive charge
. μ = αE + β E 3 /3! + · · · (13.30)
The expression is acceptable for small distortions of the electron cloud, when the
field strength is small. We usually deal with the first term, where the factor .α is the
electric polarizability. The energy of the dipole in the electric field is:
fE
. V =− 0 μ · dE = −αE 2 /2 (13.31)
In a molecule, the atoms are polarized by the electric field of the other atoms.
The polarizability of an atom in a forcefield is emulated by considering the atom
as an entity with internal structure. The atom polarizability is approximated using
fictitious charged centers. They are, in fact, Drude particles playing the roles of a
central core and an external shell. The core and shell mimic the ion and the electron
cloud, respectively. The core and shell split the atom charges and are supposed to be
harmonically linked to each other.
. Vpol = krij2 /2 (13.32)
There is no other interaction between the core and shell units as, otherwise,
an additional interaction potential function has to be designed. There are neither
rotational energy and individual kinetic energy contributions from the core and shell
to the total energy, because the core and shell are considered to be a single unit. An
instantaneous equilibrium is produced when the mechanical force kd equals the
electric force qE, where d is the maximum displacement of the electron cloud. In
this case, .kd = qE. On the other hand, if we consider that the dipole is written in
the form .μ = qd, together with Eq. 13.30, then we can establish identities between
the mechanical and electrical variables .kd = qE = qμ/α = q 2 d/α. In this
regard, the polarizability is written in terms of the basic constants of the atom; more
precisely, the polarizability is inversely proportional to the restoring force constant
k and proportional to the square of the charge.
. α = q 2 /k (13.33)
The core and shell may additionally divide the atom mass. The masses linked
by a harmonic force lead to an internal vibration, with a frequency inside the atom
that is substantially larger than the vibrational frequency of the whole atomic unit
observed in the molecule. The internal vibration may get correlated with the motion
of the other atoms, but fortunately, the correlation is frequently small. The position
of the core is usually this of the atom, and the position of the shell is determined
after minimizing the molecular energy with respect to the shell position at each
time step of the simulation to have a self-consistent dynamic process with the other
atoms. The higher-order polarization terms of the atom may be equally simulated
by dealing with spherical shells around the core, which may perform breathing
motions. In this case, the interaction of the polarized atom with the other atoms
of the molecule depends on the shell radii.
For a couple of molecules interacting through their dipoles, the Stockmayer
potential may be used as the model interaction potential.
[( ) ( σ )6 ]
σ 12 μ1 μ2
V (r) = 4ε r − r − 4π ε0 r 3
η(θ1 , θ2 , φ)
. (13.34)
η(θ1 , θ2 , φ) = 2 cos θ1 cos θ2 − sin θ1 sin θ2 cos φ
The dipole moments of the molecules are .μ1 and .μ2 . The characteristic length
and energy are .σ and .ε. The quantities .θ1 and .θ2 are the angles of the first dipole axis
and the second dipole axis with respect to the line linking the two dipole centers,
and .φ is the azimuth angle between the two dipole axes.
In summary, the spring constants, partial electric charges, and positions of the
Drude particles, these positions determined at every time step of the simulation,
define the way in which the electrically polarized atoms respond to the local
and external electrostatic fields. The polarizability model demands additional
computational resources. The Drude particles are not expected to correctly respond
to intense electrostatic fields as, in those cases, the electron cloud suffers severe
distortions.
13.15 External Fields and Potentials
There are situations when we require to investigate the behavior of a molecule in

the presence of an external field. The forces exerted on the ith atom by an electric
field, magnetic field, and gravitational field are, in that order:
. Fi = fi + qi E ; Fi = fi + qi (vi × B) ; Fi = fi + mi g (13.35)
where the other atoms of the molecule are supposed to exert a net force .fi on
the ith atom. By introducing a sinusoidal function, a pulsating electrostatic field
is simulated.
13.16 Parameterization of Forcefields 315
. Fi = fi + qi E sin(2π ωt) (13.36)
There is always the possibility to use other types of external potentials, for
example, to simulate the external forces that help in the folding of a protein [12].
In this case, the external potential is of mechanical type. The effects of laser
optical tweezers and those of an atomic force microscope to fold or unfold a
protein can be equally simulated using mechanical potentials [13]. The possibility of
creating external potentials with different purposes, for example, to stress a polymer,
accelerate its relaxation, induce a particular reaction [14], confine the molecular
system, examine certain regions of energy landscapes, dissociate a substrate from
the receptor, follow a trajectory along a specific pathway, induce the transport of ions
through membrane channels, and more, opens a new branch recognized as steered
molecular dynamics.
Problem 13.3 Design a potential that simulates a spherical confinement.
Solution A model potential that may simulate a spherical confinement

should in principle depend on the radius .R0 of the spherical confinement.
In this respect, it is possible to propose an exponential function in terms of r,
which is the particle’s radial position.
.V = V0 exp(αr)/ exp(αR0 )
The term .V0 is the potential strength. The factor .exp(αR0 ) is a normalization
term. When .r = R0 , the potential V achieves the maximum value, .V = V0 .
The factor .α determines the potential steepness. When .α is large, the potential
increases rapidly and strongly repels the particles close to the spherical
surface, this is, the repulsive potential plays the role of a rigid wall. On
the other hand, when .α is small, the potential decreases slowly and softly
repels the particles close to the spherical wall. By using several potentials of
exponential type, the confinement of particles in a rectangular box or in a
toroidal circuit can be equally designed.
13.16 Parameterization of Forcefields
A forcefield is a set of empirical functions with the required parameters to simulate

the atomic interactions of a family of molecular compounds. The parameters are
defined according to the atom types of the molecule and correspond to the constants
present in the bonded and non-bonded interaction potential functions. In the bonded
contributions, the parameters are, for example, the lengths and angles of reference
and the force constants. In the non-bonded contributions, the parameters are the
effective atomic charges of the electrostatic interactions and the reference radii and
energy strengths of the van der Waals interactions at the equilibrium distance. The
parameters may be initially proposed from accurate first-principles calculations on
small reference molecules and refined by matching the experimental measurements
on bulk systems, particularly to reproduce the X-ray structures, the spectroscopic
vibrational patterns, the thermodynamic energies, phonon energies, specific heats,
elastic constants from compression measurements, etc. [15]. The parameters may be
equally constrained to reproduce a database of well-known properties of a family of
compounds. An optimal fitting process demands a global search of the parametric
space to determine the best parameters.
A model interaction potential improperly parameterized in a F F may result in the
inaccurate behavior of the molecule in the simulation. Thus, it is important to work
with properly parametrized forcefields and pay attention to the following topics.
The set of parameters are specific to that part of the phase space and molecular
conformations where the parameters were obtained for. The parameters are not
transferable to different molecules or applicable to (very) different conformations of
the same molecule. They are neither transferable to different forcefields, although
different forcefields correctly parameterized should predict similar conformations
and energy changes of the same molecule. For a given forcefield, some parameters
may be reused when specific molecular fragments of a large molecule are identical
to the molecular fragments of another molecule that has been already parameterized.
This process enlarges the list of molecules that a F F may be applied to and keeps
the F F self-consistent. Any model interaction potential with its set of parameters
should be extensively tested before it can be applied.
Most software packages implementing classical molecular dynamics include
their own data banks, with the set of parameters for the particular molecular
structures they deal with. In such data banks, there is a file containing the parameters
of the MI P s. The file is frequently referred to as the parameter file; it comes
together with the topology file, where the atom types and their connectivity in the
molecule are indicated. In this regard, the proposal of new structures with their atom
types and parameters requires the modification of such files in the software codes.
Problem 13.4 Consider a pair of interacting atoms AB, with masses .mA =
1.2 and .mB = 1.4 amu. After carrying first-principles calculations, as
given in Eq. 13.1, the energies of the molecule AB at different distances are
registered.
Distance(Å) Energy(au) Distance(Å) Energy(au)

R1 = 1.20 E1 = 0.062 R4 = 2.00 E4 = 0.023
R2 = 1.40 E2 = 0.022 R5 = 2.40 E5 = 0.123
R3 = 1.60 E3 = 0.002 R6 = 2.80 E6 = 0.303
(continued)
13.17 Model Potentials of Non-biological Systems 317
Determine the values of the constants k and .R0 of a model harmonic potential
from the numerical values, and predict the vibrational frequency of the
molecular bond. Discuss the validity of the model potential when the molecule
is immersed in water.
Solution A model harmonic potential has the form .V (R) = k(R − R0 )2 /2,
the force constant is k, the equilibrium distance is .R0 , and R is the instan-
taneous separation between atoms A and B. By plotting the values of the
table, we obtain a parabolic curve, which is reproduced by the expression
.E(R) = k(R − R0 ) /2. The constants that best fit the set of values are
2
2
.k = 0.50 au/Å and .R0 = 1.70 Å. According to Eq. 13.2, we write
.V (R) = E, and the model harmonic potential that describes the oscillations
of the AB bond is .V (R) = 0.50 (R − 1.70)2 /2 au for .R e [1.20, 2.80] Å.

It is now possible to predict interparticle distances with their corresponding
energies and the vibrational frequency .ω of the pair AB in vacuum.
√ 2
ω= k/μ = 0.880 [au/(Å amu)]1/2
.
1/μ = 1/mA + 1/mB = 1.548 amu−1
The oscillations of AB depend on the molecular environment, and the above

model potential V is not acceptable for the case of AB in water, this is, the
results are only valid in the spatial domain .[1.20, 2.80] Å in vacuum, where
the parameters were determined. We require new data of the molecule AB in
water to carry the appropriate parametrization of the MI P .
13.17 Model Potentials of Non-biological Systems
In addition to the forcefields for biological systems, there are model interaction
potentials especially designed for non-biological systems. Such potentials are
referred to as many-body potentials. The metals and the alloys are described by
many-body potentials. A many-body potential is a function which involves N -body
interactions under the premise that the interactions between atoms depend upon
the arrangement of the neighboring atoms. The many-body potentials introduce
the effects of the environment in the molecular bonds, and therefore, they differ
from a two-body potential, where the interaction depends on the distance of the two
concerned particles solely. A many-body potential is approximated by correlating
two-body terms with higher-order effects. The formulation of a many-body potential
relies on the embedded-atom approach, which departs from pair-like terms, taking
the general form:
EN EN EN EN
. V = 1
2 i=1 j /=i Vij (rij ) + i=1 F (ρi ) ; ρi = j /=i ρij (rij )
(13.37)
The first term, .Vij , gives the two-body contributions. In the embedded-atom
scheme, every atom pair is considered immersed in a bulk of atoms described by
the density .ρi . Thus, the second term of the potential introduces a functional F of
the density .ρi . It usually assumes the form .F (ρi ) = αρi , where .α is a constant.
The functional may be split into different shells, like the s and d shells. The density
.ρi is of local nature and depends on the atomic coordination through the functions
.ρij , second expression of Eq. 13.37, which are due to atom pairs. It may be further
corrected for non-local effects by considering three-body or four-body terms, and it

may be also averaged over large distance interactions to achieve higher accuracy.
When there are different species of atoms in the system as, for example, in the
case of alloys, and the required parameters to compute the interactions of unlike
atoms are unknown, the combination rules may be applied to create the two-body
potential functions and the two-body densities of the mixed-species system.
VijAB = VijBA = [VijAA + VijBB ]/2

. [ ]1/2
ρijAB = ρijBA = ρijAA ρijBB
The functions .VijAB and .ρijAB define the interaction between the A and B atoms
of Eq. 13.37. However, it is recommended to carry the appropriate parametrization
process to avoid the use of the combination rules on the parameters of unlike atoms.
In the following lines, some examples of metallic potentials inspired by Eq. 13.37
are presented.
13.18 Sutton-Chen Potential Function
The Sutton-Chen function is a many-body potential designed for metallic atoms,

such as N i, Cu, Rh, P d, Ag, I r, P t, Au, P b, Al, etc. It is used for investigating
clusters and crystals. It has the form [16]:
{ ]1/2 }
EN EN [EN
. V =e 1
2 i/=j (a/rij )n −c i=1 j /=i (a/rij )m (13.38)
The first term is a repulsive one and the second term is an attractive one. The
factor c is a positive dimensionless parameter, and the n and m exponents are
positive integers with the restriction .m < n. The parameters .e and a determine
the energy intensity and a characteristic length of the system. The parameters are
fitted to reproduce the cohesive energies per atom, and the crystal lattice parameters
of different face-centered-cubic metals, with acceptable results in the prediction of
the bulk moduli and elastic constants of the materials. By restricting the values
13.19 Gupta Potential Function 319
of the parameters .(m, n), it is possible to use the scaling properties of the many-
body Sutton-Chen potential to deal with other materials. For example, the potential
function of Cu with numbers .(m, n) = (6, 9) may be transferred to Ni with the
same .(m, n) numbers, and re-scaling the values of the energy intensity and the
characteristic length, and similarly for P t and Au.
13.19 Gupta Potential Function
The Gupta interaction potential is a many-body function designed to describe the

interaction of transition metal atoms of a cluster or crystal [17].
EN EN {EN }1/2
. V =A i/=j exp[−p (rij − r0 )/r0 ] − B i=1 j /=i exp[−2q (rij − r0 )/r0 ]
(13.39)
The form of such a potential is based on Friedel’s tight-binding model of s

electrons. The first term is a short-range pair potential of repulsive character,
and the second one is an attractive term where the function in square brackets
is an embedding or glue potential. The parameters p and q are dimensionless
numbers; they are assumed to change little along a transition metal series. The other
parameters of the potential are the factors A and B with units of energy, and .r0 with
units of length. They all are fitted on the basis of the bulk and surface properties.
Interestingly, the Morse and Lennard-Jones potentials may be applied to a similar
particle system; however, they are not appropriate to describe the behavior of a metal
crystal since, for the topmost surface layers, they predict an increasing interlayer
separation. It is a result in contraposition to experiments, especially for the (1,0,0),
(1,1,0), and (1,1,1) surfaces of a face-centered-cubic transition metal. The Gupta
potential provides the correct picture by giving a contraction in the separation of the
metal topmost surface layers.
Problem 13.5 Obtain the force on the ith particle according to the Sutton-
Chen interaction potential, on the one hand, and the Gupta interaction
potential, on the other.
Solution The force on the ith particle is obtained from the expression
Fi = −∇i V ({rk }). In the case of the Sutton-Chen potential, the function V
.
corresponds to Eq. 13.38. By considering the identity .∂rij /∂xi = xij /rij , the
expression of the Sutton-Chen force is:
(continued)
⎡ ⎤−1/2
e E ( n n+2 ) E
N N
.Fi = n a /rij − cf ma m /rijm+2 rij ; f = ⎣ (a/rij )m ⎦
2
j /=i j /=i
In the case of the Gupta potential, the function V corresponds to Eq. 13.39.
The expression of the Gupta force is:
EN {
Fi = j /=ipA exp[−p (rij − r0 )/r0 ] − qBf ∗
}
.
∗ exp[−2q (rij − r0 )/r0 ] rij /(r0 rij )
{E }−1/2
N
f = j /=i exp[−2q (rij − r0 )/r0 ]
If the two-body terms of the many-body interaction potential, Eq. 13.37, are
avoided, then the perturbation potential is obtained.
[ ]
EN EN { } 1/2
. V = −e i=1
1 a
j /=i 2 rijm 1 + erf[α(rij − r0 )] (13.40)
The presence of the error function in the potential is clarified in the Ewald
approach, Sect. 14.5
13.20 Tersoff Potential Function
The last expression to discuss is the Tersoff potential function. It was initially
proposed for silicon, which is polymorphic due to the different types of bonding
with close cohesive energies under moderate pressure effects, .10–150 GP a. The
Tersoff expression goes beyond the pair potentials, which are useful for describing
close-packed structures, abandoning the typical N -body potential, Eq. 13.4, since
the Tersoff function is designed to describe bonded structures and also open
structures, like those observed in silicon materials. The coordination number, this
is, the number of neighbors of an atom, and the bond order or bond strength are
specially treated in the Tersoff function. The bond order rapidly decreases when the
coordination number increases. The close-packed structures maximize the number
of bonds with a bond order which exhibits a weak dependence on the coordination
number, this is usually the case of metallic systems. The Tersoff potential adopts the
form [18]:
13.20 Tersoff Potential Function 321
EN E [ ]
V = 1
2 i/=j j fc (rij ) aij fr (rij ) + bij fa (rij )
fr (r) = A exp(−λ1 r) ; fa (r) = −B exp(−λ2 r)

⎧ (13.41)
.
⎪
⎨1 ; r <R−D
fc (r) = 2 − 2 sin[π(r − R)/(2D)] ; r − D < r < R + D
1 1
⎪
⎩
0 ; R+D <r
The .fr and .fa functions represent repulsive and attractive pair potentials,
respectively. Their forms are based to some extent on the universal type of bonding
given by the Morse potential. The function .fc is a cutoff function to limit the range
of the potential. It is continuous with continuous derivatives in terms of the distance
r, going from 0 to 1 in a small range centered at the value R. The term D is a cutoff
parameter. The pair of values .R = 2.75 Å and .D = 0.1 Å are frequently chosen
together. The Tersoff potential includes the environment of a bond by dealing with
the function .bij , related to the bond order. Thus, an extra summation in the particle
index k, with .k /= i, j , is introduced.
bij = (1 + β n ξijn )−1/(2n) ; aij = (1 + α n ηij n )−1/(2n)

E
ξij = k/=i,j fc (rik ) g(θij k ) exp[λ33 (rij − rik )3 ] ;
. 2 2 (13.42)
g(θ ) = 1 + dc 2 − d 2 +(h−cos
c
θ)2
E
ηij = k/=i,j fc (rik ) exp[λ3 (rij − rik )3 ]
3
The .bij function has a monotonic decreasing behavior according to the coordina-
tion of atoms i and j . Such a function also introduces an exponential function in the
bond difference, .rij − rik , to account for the separation of bonds. The functions .bij
and .aij impose a limit in the interaction range to just consider the first neighbor shell
atoms, and account for the atom coordination’s relatively weak dependence on the
cohesive energy. The function .bij introduces angular terms to stabilize, for example,
the diamond structure. In general .bij /= bj i , but the symmetry in such terms may
be imposed when necessary. The term .aij is included for completeness; however, its
presence may be omitted or employed to improve the potential.
The secondary functions .ξij and .ηij establish the number of bonds different to
the ij bond. The quantity .θij k is the bond angle between the bonds ij and ik. The
dependence on the angle is not necessarily monotonic; this is the case of arsenic
compounds. The coefficient h in the expression of .g(θ ) may take values out of the
range .[−1, +1], the coefficient d emphasizes or diminishes the dependence on the
angle .θ , and the coefficient c establishes the strength of the angular effect.
The fitting parameters of the potential are A, B, .λ1 , .λ2 , R, D, shown in functions
.fr and .fa of Eq. 13.41. We additionally have the coefficients .α, .β, n, .λ3 , c, d,
h, shown in the bond-order functions .aij and .bij of Eqs. 13.42. It is occasionally
assumed that .λ1 = 2λ2 = 2λ3 for simplicity. The Tersoff function is a short-
range potential frequently acting up to 3 Å. It has been used successfully in the
study of silicon and carbon structures, like the Buckminster fullerenes, nano-tubes,
nano-rods, nano-cones, diamond, graphite, surfaces, etc. Less encouraging results
are obtained in the formation of point defects and vacancies, and in the description
of the liquid state. The combination rules may be used for unlike atoms that interact
in the system.
The parameter values of the Sutton-Chen, Gupta, and Tersoff potential functions
are found in the original works. The parameters depend on the type of system
under investigation. The forces are all obtained after (tedious) differentiation of the
interaction potential functions, .fi = −∇i V ({rj }) for the ith particle. The forces may
be directly derived or obtained after using a symbolic-algebra software package or
an artificial-intelligence software package. There is a large number of potentials in
the literature, many of them are of general use, and others are specific to one or more
elements of the periodic table. In this regard, the research continues in this field.
A.1 Appendix: Harmonic Model of the Dispersion Energy
The dispersion energy is an important contribution at intermediate distances. The

goal in this section is to derive the dispersion energy by using a pair of simplified
molecules separated by a distance r. Each molecule is neutral; however, the ion
charge .+q and electron charge .−q form a dipole in each molecule, because each
molecule is immersed in the electric field of the other molecule. The two molecules
lie along the x axis. The positive charges have fixed positions, and the negative
charges are allowed to vibrate√harmonically in the neighborhood of their ions in the
x axis with frequency .ω0 = k/m, where m is the mass of the vibrating particle,
and k the vibrational force constant. The ions are separated by the distance r, in
such a way that the distance from .−q to .+q is always much smaller than r. We
are interested in finding the dispersion energy of the system from the dipole-dipole
interaction given by [19]:
. V (r) = − 2μ a μb
4π e r 3
; μa = xa q ; μb = xb q (13.43)
The dipole-dipole electrostatic interaction of the molecules is attractive and

decays as .r −3 . The dispersion energy is a non-classical energy and should be
obtained in the quantum mechanical framework. In this regard, we have to solve
the wave equation, taking the above potential function into consideration.
Problem 13.6 Suppose a molecule with a pair of atoms. One of the atoms
has a positive charge q and the other a negative charge .−q. Their separation
is d. The dipole p of the molecule is .p = qd. The origin of the reference
(continued)
A.1 Appendix: Harmonic Model of the Dispersion Energy 323
system is located at the middle point of the line joining the atoms. Obtain the
electrostatic potential at the position .r, far from the atoms, with .r >> d.
Solution Let us consider .r1 and .r2 , with magnitudes .r1 and .r2 , as the vectors
starting at atom 1 and atom 2, respectively, both finishing at point .r. The
electrostatic potential at the position .r is:
( ) ( )
1 q q q r2 − r1
V =
. − =
4π e r1 r2 4π e r1 r2
The difference .r2 − r1 at large distances is approximately .d cos θ , where .θ is

the angle between .r and the line joining the atoms. The product .r1 r2 is also
approximated by .r 2 . By using such approximations, the electrostatic potential
includes the dipole p.
1 p cos θ
V =
.
4π e r 2
The harmonic behavior of the oscillating charges and the interaction 13.43 of the
dipoles are considered in the formulation of the wave equation.
[ ( ) ]
2
h̄ ∂ ψ 2
h̄2 ∂ 2 ψ 2xa xb q 2
. − 2m ∂x 2
− 2m ∂x 2 + 2 xa + xb +
k 2 2
4π e r 3 ψ = Eψ (13.44)
a b
The distance from .+q to .−q in one of the molecules is .xa and in the other
molecule is .xb . The complexity of the potential is reduced by introducing the new
variables .x1 and .x2 and the effective force constants .k1 and .k2 .
√ √
x1 = (xa + xb )/ 2 ; x2 = (xa − xb )/ 2
. ( 2
) ( 2
) (13.45)
k1 = k 1 + 4π1 e 2q
kr 3 ; k2 = k 1 − 4π1 e 2q
kr 3
The potential now avoids crossed terms in the new variables.
k1 k2 ( ) 2xa xb q 2
.
2 x12 + 2 x22 = k
2 xa2 + xb2 + 4π e r 3
(13.46)
The wave equation is given in terms of the variables .x1 and .x2 and the effective
spring constants .k1 and .k2 .
[ ]
h̄2 ∂ 2 ψ h̄2 ∂ 2 ψ k1 k2
. 2m ∂x 2 + 2m ∂x 2 + 2 x12 + 2 x22 ψ = Eψ (13.47)
1 2
The equation corresponds to two independent harmonic oscillators. The energy

and frequencies are given in terms of the old frequency .ω0 and the old force constant
k.
( ) ( )
E = h̄ω1 n1 + 12 + h̄ω2 n2 + 12
.
√ ( )
2 1/2
(13.48)
ωi = ki /m = ω0 1 ± 4π1 e 2q kr 3
; i = 1, 2
When the oscillators are far from each other, .r → ∞, they have no interaction
and the energy of each oscillator in the ground state is .E = h̄ω0 /2. At intermediate
distances, the oscillators interact and the total energy of the system contains the
dispersion energy. We apply the binomial expansion .(1 + x)1/2 = 1 + x/2 − x 2 /8 +
x 3 /16+· · · to the expressions of the frequencies .ωi , where the distance r is present,
to deal with the energy which is now called .Etot .
( )1/2 ( )1/2
1 2q 2 1 2q 2
Etot = 12 h̄ω0 1 + 4π e kr 3 + 12 ω0 1 − 4π e kr 3
[ ]
q2 1 1 q2 2
Etot = 12 h̄ω0 1 + 4π1 e kr 3 − 2 ( 4π e kr 3 ) + · · ·
.
[ ]
q2 1 1 q2 2
+ 21 h̄ω0 1 − 4π1 e kr 3 − (
2 4π e kr 3 ) + · · ·
[ ]2
q2
. Etot = h̄ω0 − 1
2 h̄ω0 1
(4π e) kr 3
(13.49)
The expansion is consistent with Eq. 13.43 referent to the interaction of distant
dipoles. We require to subtract the energy of the infinitely separated oscillators from
the above energy to obtain the dispersion energy.
q4
. Edisp = Etot − E = −h̄ω0 2(4π e)2 k 2 r 6 (13.50)
r→∞
The zero-point energy, .h̄ω0 , is present in .Edisp and exhibits the quantum
character of the interaction. The dispersion energy is written in equivalent form
by defining the factor .C6 .
. Edisp = C6 /r 6 ; C6 = −h̄ω0 q 4 /[2(4π e)2 k 2 ] (13.51)
By using the expression of the polarizability 13.33, the dispersion factor .C6 is
expressed in terms of the molecular polarizability.
. C6 = −h̄ω0 α 2 /[2(4π e)2 ] (13.52)
The dispersion energy depends on the spatial orientations of the interacting

dipoles and also of the individual energy states of the molecules. In many situations,
the dispersion energy is an attractive energy; however, it may also be a repulsive
References 325
energy. The dispersion energy demands, in fact, higher-order multipoles to be

properly described.
. Vdisp = C6 /r 6 + C8 /r 8 + C10 /r 10 + · · · (13.53)
The dispersion energy of three particles is approximated with additive terms plus
a non-additive contribution using the Axilrod-Teller expression.
ij jk αij k
. Vdisp = C6 /rij6 + C6ik /rik
6 + C /r 6 +
6 jk (rij rik rj k )3
(1 + cos θij + cos θik + cos θj k )
(13.54)
The quantities .θij , .θik and .θj k are the internal angles of the triangle described by
the particles i, j , and k. The angle values depend on the triangular shape (equilateral,
isosceles, scalene, including a linear conformation). The cosine functions define the
sign of the non-additive contribution. In this regard, the role of the angular term is
to correct the additive contributions by adding/subtracting a non-additive part.
The van der Waals interaction potential of Eq. 13.26 includes the dispersion
function .−β/r 6 as an attractive term. Still, such a potential additionally includes a
repulsive interaction of the electron clouds at short distances. This last contribution
is simulated by means of the term .α/r 12 to overcome the attractive dispersion
.−β/r at short distances. The discussion on the proper form of the van der Waals
6
interaction potential continues in our days [20].
References
1. G.C. Maitland, M. Rigby, E.B. Smith, W.A. Wakeham, Intermolecular Forces: Their Origin
and Determination. International Series of Monographs on Chemistry 3 (Clarendon Press,
Oxford, 1987)
2. A.K. Rappe, C.J. Casewith, Molecular Mechanics across Chemistry (University Science
Books, Sausalito, 1997)
3. I.G. Kaplan, I.L. Garzon, R. Santamaria, B.S. Vaisberg, O. Novaro, Ab-initio model potentials
and their application to the thermal stability of metal clusters. J. Molec. Struct. (Theochem)
398–399, 333–340 (1997). I.L. Garzon, I.G. Kaplan, R. Santamaria, B.S. Vaisberg, O. Novaro,
Ab-initio model potential and molecular dynamics simulation of Ag6 clusters. Z. fur Phys.
D40, 202–205 (1997). I.L. Garzon, I.G. Kaplan, R. Santamaria, O. Novaro, Molecular
dynamics study of the Ag6 cluster using an ab-initio many-body model potential. J. Chem.
Phys. 109, 2176–2184 (1998)
4. Cerius2 , Forcefield Based Simulations: General Theory and Methodology (MSI, San Diego,
1997), Eqs. 44–46, p. 111.
5. I.G. Kaplan, R. Santamaria, O. Novaro, Non-additive forces in atomic clusters. Molec. Phys.
84, 105–114 (1995)
6. I.G. Kaplan, R. Santamaria, O. Novaro, Nonadditive interactions and the relative stability of
neutral and anionic silver clusters. Int. J. Quant. Chem. 55, 237–243 (1995)
7. R. Santamaria, E. Charro, A. Zacarias, M. Castro, Vibrational spectra of nucleic acid bases and
their Watson-Crick pair complexes. J. Comput. Chem. 20, 511–530 (1999)
1977), pp. 1130–1140
Oxford, 1987), Chapt. 1
1977), pp. 1130–1140
11. E. Hecht, Optics, Global edn., 5th edn. (Pearson Education Limited, Essex, 2017), Sect. 3.5.1
12. R. Santamaria, K. Jones, M. Arroyo, T. González-García, Polymer folding via external
potentials in ab-initio calculations. Comput. Theor. Chem. 1068, 72–80 (2015)
13. B. Isralewitz, M. Gao, K. Schulten, Steered molecular dynamics and mechanical functions of
proteins. Curr. Opin. Struct. Biol. 11, 224–230 (2001). H. Lu, K. Schulten, Steered molecular
dynamics simulations of force-induced protein domain unfolding. Proteins: Struct. Funct.
Genet. 35, 453–463 (1999). C.J. Ordoñez Martínez, D. Martínez-Zapata, R. Santamaria,
Dissociation of the Watson-Crick base pairs in vacuum and in aqueous solution: a first-
principles molecular dynamics study. J. Biomolec. Struct. Dyn. (2021). https://doi.org/10.
1080/07391102.2021.1987988
14. D. Martínez-Zapata, R. Santamaria, The damage of the Watson-Crick base pairs by Nickel
nanoparticles: a first-principles molecular dynamics study. Comput. Biol. Chem. 87, 107262
(2020)
Oxford, 1987), Chapts. 4 and 7
16. A.P. Sutton, J. Chen, Long-range Finnis-Sinclair potentials. Philos. Mag. Lett. 61, 139–146
(1990)
17. R.P. Gupta, Lattice relaxation at a metal surface. Phys. Rev. B23, 6265–6270 (1981). J.P.K.
Doye, Lead clusters: different potentials, different structures. Comp. Mater. Sci. 35, 227–231
(2006)
18. J. Tersoff, New empirical approach for the structure and energy of covalent systems. Phys. Rev.
B37, 6991–7000 (1988)
Oxford, 1987), Eq. 1.7 and Sect. 1.4.3
20. R.O. Watts, I.J. McGee, Liquid State Chemical Physics (Wiley, London, 1976) Sect. 2.2-2.3,
pp 11–22. T.A. Halgren, Representation of van der Waals (vdW) interactions in molecular
mechanics force fields: potential form, combination rules, and vdW parameters. J. Am. Chem.
Soc. 114, 7827–7843 (1992) (1992)
Extended Systems
14
The study of extended systems is important in science. An extended system may

be a gas, fluid, or solid. A crystal, an amorphous material, and a substance with
no homogeneous structure and no periodicity are considered to be 3D extended
systems. Other physical systems such as a surface or ceramic are 2D extended
systems. A large 1D structure, such as a polymer, may be considered an extended
system of reduced dimensionality.
In many instances, we require to investigate the properties and the behavior of
extended matter or the features of a compound immersed in an extended sample,
like a protein in water, with the water extending all over the available space. Yet,
we have a limit on the number of particles that can be simulated in a computer to
study extended systems.1 This limitation compels to implement approximations to
investigate the behavior of extended matter. The purpose of this chapter is to discuss
the methods that allow the characterization of extended systems by dealing with
a relatively small number of atoms of the sample. In a first stage, we consider a
finite material sample enclosed by walls. Later, the walls are avoided to work with
a material with no apparent borders, but with high repetitive structure. In this last
case, the Ewald summation approach for systems with periodic boundary conditions
is introduced. The electrostatic interactions in the Ewald approach are divided
according to short-range and long-range interactions, leading to closed formulae
which are computationally accurate and efficient. In the final part of the chapter,
alternative approaches are discussed to compute the electrostatic interactions from
modified potentials and forces. The required mathematical tools in every case are
presented.
1 A mole of substance is considered the smallest piece of matter representative of a substance; still,
it contains a huge number of particles, specifically .6.0 × 1023 , for a molecular simulation.
328 14 Extended Systems
14.1 Fixed and Flexible Boundaries
When many particles interact, they may scatter in all directions. In a molecular
dynamics simulation, it is important to introduce boundaries to confine the particles
and avoid their dispersion. The confinement of particles is of help to study a small
or large fraction of the atoms of an extended system. The confinement may be
imposed in different ways. The case of rigid walls is one of the straight forms to
confine particles. Such a case evolves to a more realistic model when the boundaries
are simulated using repulsive walls. The repulsive walls may be simulated by
introducing exponential functions with positive arguments.
EN ( )1/2
. V ({ri }) = i=1 exp(αri )/ exp(αr0 ) ; ri = xi2 + yi2 + zi2 (14.1)
This expression gives the spherical confinement of the particles. The summation
is over the confined particles. The parameter .α determines the growing rate of the
exponential potential, V ; big values of .α make the potential rising sharply, and small
values of .α make the potential increasing slowly. The exponential in the numerator is
weighted with a similar exponential in the denominator. This last function represents
a normalization factor because the potential acting on a particle becomes unity
at .r0 , which corresponds to the radial position of the wall. Thus, the increasing
exponential plays the role of a repulsive potential that keeps the particles confined
and away from the confining surface. There are other forms of repulsive potentials.
Such barriers may additionally exhibit texture by using different or changing values
of .r0 ; an oscillating wall is obtained by doing .r0 = Af (t), where A is the position
of the wall and the factor .f (t) is, for example, .[1 + (1/10) sin(θ (t))]. When .θ is
written in terms of the time, .f (t) introduces a pulsating behavior of the wall with
10% fluctuations of the value of A. Other forms of .f (t) may be proposed.
The function playing the role of a repulsive potential may be replaced by particles
linked to each other to form a relatively hermetic container with atomic structure.
The particles at the interior of the container move and interact with the particles of
the container and cannot escape from the confinement. In response to the interaction
with the confined particles, the particles forming the container may be permitted to
vibrate in the neighborhood of their positions, without missing the shape of the
confining cage. The links between the container atoms may be of harmonic type for
simplicity, Fig. 14.1.
( )
. V (ri ) = k rij − rij0 /2 (14.2)
The spring stiffness is k, and the equilibrium distance between the ith and j th
particles is .rij0 . When k or the masses of the boundary particles are sufficiently
large, the fixed-particle boundary scheme is recovered, because the wall particles
barely move. The dynamics of the particles enclosed by rigid walls is different to
the dynamics of the particles enclosed by flexible walls. This last type of approach
is known as the semi-rigid boundary scheme. In principle, a large number of atoms
14.1 Fixed and Flexible Boundaries 329
Fig. 14.1 The inset on the left shows a system of particles; they are confined by rigid walls formed
by particles with fixed positions. The inset on the right shows flexible walls, formed by particles
that vibrate in the neighborhood of their original positions
(that build a sample) may be imprisoned by walls to investigate the properties of

extended matter.
Problem 14.1 Consider a torus formed by sweeping a disk around a circular

axis. The torus lies in the xy plane with radius R and disk radius .rd . Create a
potential energy function that confines N particles in a container with toroidal
geometry from exponential functions.
Solution Suppose that the position of the ith particle is .ri . It is possible to
find a factor .λ in such a way that .λ(xi2 + yi2 )1/2 gives the torus radius R, this
is, .λ = R/(xi2 + yi2 )1/2 . Once the factor .λ for particle i is determined, the
torus vector radius is .R = λ(xi , yi , 0). Next, consider the disk around the ith
particle. The position of the ith particle with respect to the central point of the
disk is .r'i = ri − R = (xi − Rx , yi − Ry , zi − 0). The toroidal confinement is
now built from exponential functions.
E
N
( )
V = V0
. exp αri' / exp(αrd )
i=1
When .ri' reaches the disk radius, this is .ri' = rd , the potential exerted
on particle i is a maximum since .V0 exp(αri' )/ exp(αrd ) = V0 ; otherwise
'
.V0 exp(αr )/ exp(αrd ) < V0 . Therefore, particle i is confined in a toroidal
i
geometry when the potential V is applied.
14.2 Periodic Boundary Conditions
A common approximation to simulate an extended system using a minimum of

computing resources is the periodic boundary conditions (P BC) approach. It
consists in taking a representative portion of the system to build a central unit cell,
and repeat it in every direction to fill the entire space, Fig. 14.2.
A requirement to apply the P BC approach is that the shape of the central cell
covers the 3D space by employing translational operations on the cell. The resulting
system is a perfect periodic system, representing an idealization of the real extended
system. The most common cell shapes are the cubic cell, hexagonal prism, truncated
octahedron, and rhombic dodecahedron, Fig. 14.3. Other shapes may not satisfy the
required conditions, like a cell with spherical symmetry. The favored cell shapes are
those that not only tile the 3D space but also involve the lowest number of particles
per cell, without missing the relevant aspects of the system, because they reduce the
computational costs.
Fig. 14.2 Top inset: in the periodic boundary conditions approach, a central unit cell is indefi-
nitely repeated in every direction, and there are no borders. The repetition of the nanoscopic piece
of matter simulates the crystal. In this 2D periodic lattice representation, the central and image
cells may be enumerated in the form [(-1,-1), (-1,0), (-1,1)], [(0,-1), (0,0), (0,1)], [(1,-1), (1,0),
(1,1)], where the indexes run from the lower-left corner to the upper-right corner. Bottom inset:
the particles of the boxes may travel from one box to another crossing the imaginary borders. A
molecule is located in the central unit cell at the instant of time t. After an interval of time .δt, the
molecule moves out of such a cell, but an image molecule concurrently moves into the central unit
cell to keep the particle density of the cell constant
14.2 Periodic Boundary Conditions 331
Fig. 14.3 The depicted

polyhedra are some of the
geometrical structures that
completely tile the 3D space.
They are commonly used in
the periodic boundary
conditions approach
Fig. 14.4 The rendering of the left illustrates a real crystal with impurities, particle vacancies, and
local irregular forms, randomly located, breaking the symmetry of the crystal. The rendering of the
right shows the central unit cell with different background color to the image cells. The simulated
crystal is a crystal with perfect symmetry in the P BC approach since the unit cell is repeated in
every direction
The central cell contains a finite number of particles and reproduces the particle
density of the real sample. The copies of the central cell are known as the image
cells and have the same size and shape of the central cell. The image cells contain
an equal number of particles and the same positions and momenta of the particles
as those of the central cell. In this regard, the P BC approach simulates a perfect
bulk sample with an infinite number of replicas of the central unit cell. However,
it is important to remark that the simulated crystal is not necessarily the case of
a real crystal because the real crystal may contain impurities, vacancies, localized
irregular forms, and more, with no periodicity, Fig. 14.4.
A particle in the central cell is considered to interact with the other particles of
the central cell, plus the closest particles in the neighbor image cells. The force
exerted on the ith particle by the j th particle, plus the corresponding j th particles
of the neighbor image cells is:
Fig. 14.5 According to the minimum image convention, the particle in blue color and centered
in the gray square interacts with the other particles located in that square solely. In a different
approximation, the same blue particle may only interact with the other particles located inside the
blue sphere, which has a given cutoff radius. In this case, the number of particles in the blue sphere
is lower than the number of particles in the gray square
[ ]
. fij = − ∇i V (rij (0) ) + ∇i V (rij (1) ) + · · · + ∇i V (rij (26) ) (14.3)
The nomenclature .j (k) , for .k = 0, 1, · · · , 26, stands for the j th particle of the
central cell, .k = 0, and the image particles of the neighbor cells, .k = 1, · · · , 26, in
3D space. In order to minimize the number of calculations, the interactions may
be truncated. In this case, the minimum image convention is assumed, where a
particle located in the central unit cell only interacts with the surrounding particles
positioned inside an imaginary cell, .iL3 , with equal dimensions to the central unit
cell, and where the particle of interest, namely, the ith particle is located at the center
of such an imaginary cell. In this case, the force exerted on the ith particle is due to
a reduced number of particles, Fig. 14.5.
E
. fij = − j ε iL3 ∇i V (rij ) (14.4)
The summation runs over the particles of the imaginary cell. It may happen that a
moving particle in the central cell travels from the central cell to a contiguous image
cell during the dynamics. This situation changes, among other things, the particle
density of the simulated material. Hence, in order to maintain the properties of the
material, an equivalent particle with equal velocity must enter the central cell from
an opposite face, Fig. 14.2.
In principle, by using the P BC approach, we work in the thermodynamic limit,
making possible to apply the thermodynamic concepts to the extended sample.
Thus, the P BC scheme is frequently used to investigate the thermodynamic aspects
of extended systems. Still, there are some drawbacks of the P BC approach. One of
these, previously mentioned, is referent to the repetition of a unit cell describing a
perfect material. In fact, the real material is not perfect in composition, neither in the
atomic spatial arrangements. In the case of simulating compounds in solution, like a
DNA molecule or a protein in bulk water, the solvated compound interacts with all
14.2 Periodic Boundary Conditions 333
Fig. 14.6 A solvated protein encapsulated in a truncated octahedron cell is shown in the P BC
approach. The water molecules are not depicted for simplicity, and the unit cells forming the
crystal are separated for better visualization. In a molecular simulation of the crystal, there may
be spurious forces due to interactions between the protein and its images. Such interactions are
negligible when the distance between contiguous proteins is sufficiently large. This demands unit
cells of large sizes
its own solvated images, especially when the central unit cell is small, introducing
spurious interactions that should be avoided. A practical way to deal with such a
situation is to introduce large separations between the solvated compound and its
image compounds, this is, dealing with a big central unit cell to diminish the non-
bonded Coulomb and van der Waals interactions, Fig. 14.6. Also, the multipolar
electric moments are not preserved in the P BC simulation for the introduction of
fictitious electric fluxes. More precisely, the charged atoms move in and out of the
cells, producing charge-distribution changes and inducing uncommon fluctuations
of the multipolar electric moments in the central cell; details are given later. It is
recommended to have a zero total charge to avoid an overcharged material due to the
charges of every image cell. To do this, ions or charged particles may be introduced
to impose the charge neutrality of the central cell.
The long-range forces decay by their own nature slowly in the material. In the
P BC approach, it is convenient to have long-range forces decaying relatively fast
to avoid discontinuities due to the finite size of the central cell.
For example, the van der Waals interactions become null at relatively far
distances. Yet, the van der Waals potential may be forced to decay faster by
subtracting a constant term. If we require the potential to be zero at the cutoff
distance .rc = aσ , where .rc is a scale factor of the hard-core van der Waals radius
t
.σ , and .1 < a, then the truncated van der Waals potential, .V , may be used under
W
the definition:
{
VW (r) − VW (rc ) ; r ≤ rc
. VW =
t (14.5)
0 ; rc < r
The truncated potential reaches zero at the imposed cutoff point .rc . In this
respect, it is convenient to deal with a central cell whose size is larger than .rc to avoid
the interaction of a solvated compound with its image counterparts. The truncated
potential brings a small additional computational cost, Fig. 14.5. On the other side,
measurements on the properties of traveling waves in the substance demand large
box sizes. The collective fluctuations, especially near critical points, are not properly
simulated when the central box contains a small number of particles due to the
existence of long-range correlations. In this regard, the central cell is expected to be
greater than the typical correlation lengths observed on bulk matter.
14.3 The P BC System Is an Open System
In the following lines, it is shown that there is a non-zero flux of mass, angular
momentum, and energy in the central unit cell of the P BC method, and therefore,
such a cell constitutes an open system. The strategy is firstly exemplified with the
no-conservation of the angular momentum, .L, following the work of Kuzkin [1].
For simplicity we consider image cells with cubic shapes of length .l and a single
particle traveling with momentum .p = mv and negligible interactions with the
image particles. The particle moves in the central unit cell and eventually leaves the
cell at the instant of time t. At this precise instant of time, an image particle enters
the central unit cell with the same momentum from the opposite cell face. The linear
momentum of the particle is preserved with time, but the angular momentum .L is
not necessarily preserved with time because .L depends on the origin of coordinates.
If .δt is the time step of the particle dynamics, the change of angular momentum in
this time interval is:
. L(t + δt) − L(t) = mr × v − m(r − l) × v ⇒ AL = ml × v (14.6)
The first term of the first expression corresponds to the image particle that entered
in the central cell at the instant of time .t +δt, and the second term corresponds to the
particle in the central cell that leaves the central cell at the instant of time t, Fig. 14.7.
There is a difference between .L(t + δt) and .L(t) which depends on the cell size and
the particle mass and velocity. When the surface vector .l is parallel to .v, this is, the
particle moves in perpendicular form to the cell surface, the cross product is zero,
and the angular momentum is preserved in time. However, the angular momentum
is discontinuous when the cross product between .l and .v is different from zero,
namely, when .l and .v make an angle different from zero. Therefore, when the
particle stays in the central cell or it travels from one cell to another in perpendicular
way to the cell surface, the angular momentum is conserved. The possibility to have
a discontinuous angular momentum makes the particle system in the central unit
cell of the P BC approach an open system.
14.3 The P BC System Is an Open System 335
Fig. 14.7 Two instances of the dynamics of a molecule are depicted. The size of the square box
is .l. At the instant of time t, the molecule is in the central cell with position .r' , and an image
molecule is located at .r' − l (upper right and upper left insets, respectively). In a later step, .t + δt,
the molecule leaves the central unit cell, and an image molecule enters the central unit cell. The
molecule’s position is .r, and the image molecule is present at .r − l (lower right and lower left
insets, respectively). The angular momenta of the molecule in the central unit cell at the instants of
time t and .t + δt are in general different when .l × v = |l| |v| sin(θ) is different from zero, Eq. 14.6
The angular momentum is additionally changed by the torques exerted on the

particles of the central cell due to the interaction with the image particles. The torque
.T depends on the positions and forces.
E
. T= i,j ri × fij ; i ε L3 (t) ; j ε iL3 (t) (14.7)
The index i considers the particles of the central cell, .L3 , and the index j considers
the particles of the image cells, .iL3 . The change of the total angular momentum in
time is given by .dL/dt = T + Q, where the contribution .T is given by Eq. 14.7, and
the contribution .Q represents the flux of angular momentum due to the incoming and
outgoing particles of the central unit cell according to Eq. 14.6. Instead of writing
the flux of angular momentum in terms of a differential equation, it may be written
as a balance expression on the angular momentum.
f t ' +δt f t ' +δt (E E' E )
. AL = t' (Q + T) dt= t' i ri × mvi − i ri × mvi + i,j ri × fij dt
(14.8)
The first summation runs over the particles entering the central cell, and the
second summation with a prime runs over the particles leaving the central cell in
the time interval .[t ' , t ' + δt]. The last summation considers the interaction with
the image particles. Similar types of fluxes occur for other physical variables. For
example, we deal with a charge flux when the charged particles travel from one cell
to another.
Problem 14.2 Consider a pair of particles p1 and p2 in the central unit cell
and their images i1 and i2, respectively. The particles are allowed to move
and repel each other at short distances. Demonstrate that when the particle
interactions .p1−i2 and .p2−i1 occur, the angular momentum is discontinuous
in time.
Solution Suppose that the particle p1 is close to the image particle i2, and
they interact. At that instant of time, the particle p2 is also close to the image
particle i1 for symmetry, and they also interact. The total torque in that instant
of time is:
T = rp1 × Fp1,i2 + rp2 × Fp2,i1

.
In this case, .T is not zero. Since .T = dL/dt, the angular momentum is not
a constant. However, when the distance between the particles p1 and i2 is
large, they have no interaction, the forces .Fp1,i2 and .Fp2,i1 are zero, and the
angular momentum is preserved with time as .T = 0 and .L = constant.
The flux of angular momentum may be important for systems in the gas
phase state. In liquids, such a flux produces the torques that disrupt the angular
momentum of the particles in the central unit cell. In solids, the torques may break
the conservation of the angular momentum. The P BC approach also induces the
breaking of the equipartition principle of the energy, which applies to particles
in thermodynamic equilibrium. The breaking of that principle is the consequence
of imposing constraints on the position and momentum of the mass center of the
system to keep it in the central unit cell of the P BC scheme. The constraints affect
light-weight and heavy-weight particles in different forms, resulting in different
average energies, .<E>, for the light and heavy particles in the dynamics. In the NV E
ensemble. the average energies are related by the different particle masses [2].
[ ] ( )
−2
.<E>heavy = <E>light 1 − (mheavy − mlight )/Mtot + O Mtot
Care should be exercised when thermodynamic quantities are derived in the

P BC approach.
14.4 Electrostatics in the P BC Approach
The electrostatic interactions are important in the behavior of extended structures. In

the P BC approach, the atoms of the central cell are repeated periodically, building
an endless 3D grid of atoms, and the infinite number of image charges interact
14.5 Ewald Sum Approach 337
with the charged atoms of the central unit cell, because the electrostatic interactions
decay slowly, which complicates the computation of the total electrostatic energy.
In the following sections, we introduce the Ewald summation approach, understood
as a method that permits the computation of the electrostatic energy in the P BC
scheme with the required numerical accuracy [3]. To do this, we base the discussion
on the work of C. Kittel, complemented with the work of Lee and Cai [4].
The purpose of the Ewald sum approach is to compute the electrostatic energy of
the P BC system under the consideration of an ordered array of charged atoms. The
atomic distribution has periodicity, say d, in the interplanar distances of a crystal,
with frequency .1/d in the reciprocal space. The Ewald sum approach employs
Fourier transformations of the electrostatic interactions to work in the reciprocal
space and obtain the total electrostatic energy. The approach is based on key
aspects to simplify the electrostatic computations; the slow decay of the electrostatic
potential is recognized as it is proportional to .r −1 , and it is possible to separate
the short distance interactions from the large distance interactions to individually
deal with each of them. The electrostatic interactions between nearby atoms change
rapidly with distance, becoming singular at .r = 0, but they can be rapidly computed
in the real space. In contrast, the electrostatic interactions between distant atoms
change slowly, but they can be rapidly computed in the reciprocal space.2 The
Ewald sum approach requires the particle system in the central unit cell to be
globally neutral. In order to make the summations of the electrostatic interactions
convergent, Gaussian functions are introduced and located at every grid point to deal
with an atomic charge distribution. The Gaussian functions are manageable under
the Fourier transforms and are implemented in such a way to avoid any spurious
contribution to the final energy. The Gaussian functions are essential to produce the
screening effects and achieve a rapid convergence of the electrostatic summations.
In this respect, the Ewald proposal is important, computationally efficient, and
provides accurate results.
14.5 Ewald Sum Approach
Let us consider the presence of N charged atoms in the central unit cell. The
atoms haveE positions .ri with charges .qi , .i = 1, · · · , N. The central cell is
neutral, . N
i=1 qi = 0; otherwise, there is no convergence in the Ewald electrostatic
summations. The central cell is repeated in every direction. The fundamental
2 The Yukawa approach explains the long-range nature of the Coulomb potential. In such an
approach, the neutron and proton have a short-range interaction by exchanging a particle.
The Yukawa approach predicted the existence of the meson. The Yukawa potential is .VY =
−g 2 exp(−rμc/h̄)/r, where .μ is the mass of the exchanged particle and g is a scalar. The Coulomb
potential is deduced from .VY when the particle of exchange is the photon with zero mass and
.g = qi qj /(4π ε0 ), this is, .VY = qi qj /(4π ε0 )r
2 −1 . The deduction is consistent with the uncertainty
principle, as .1/R = μc/h̄ = μc2 /(h̄c) = AE/(h̄c) and, on the other hand, .1/R = 1/(cAt), which
takes to .AEAt ∼ h̄.
translational vectors of the lattice are .a1 , a2 , a3 . They are not necessarily orthogonal
to each other. The numbers .ni , .i = 1, 2, 3, are integers running without limits. The
positions of the image charges are .ri + n1 a1 + n2 a2 + n3 a3 , which are abbreviated
by .ri + nL, with .n = (n1 , n2 , n3 ) and L the cell length. The central cell is defined
with .n = 0. A charge .qi located at .ri of the central cell interacts with the charges .qj
positioned in the central cell, with the condition .j /= i, plus all the periodic image
charges, including its own periodic image charges. The electrostatic energy of the
particle system of interest is:
∞
E ∞
E ∞
E E N∗
N E
1 1 qi qj
Eel =
.
2 4π ε0 |ri − rj + nL|
n1=−∞ n2=−∞ n3=−∞ i=1 j =1
All the charged atoms of the lattice are included in the summations. The
denominator gives the distances between the different atom pairs, where one atom is
located in the central cell and the other may be located in the same central cell or in
an image cell. The asterisk in the last summation recalls to avoid the self-interactions
in the central cell. In this respect, the diagonal terms .i = j are omitted when .n = 0.
The asterisk may also indicate other restrictions such as the electrostatic interactions
that should be avoided between bonded atoms. The factor 1/2 avoids the double
counting of the interactions. The vacuum permittivity is:
ε0 = 1/(μ0 c2 ) = 8.854 × 10−12 C 2 N −1 m−2

.
where c is the speed of light and .μ0 the vacuum permeability. The energy .Eel con-
verges slowly. For a system with N atoms, the calculation of pairwise interactions
to determine the energy .Eel is computationally costly, scaling in quadratic form,
.N (N − 1)/2. In the periodic boundary conditions approach, a 1D grid of alternating
positive-negative charges separated by a distance d demands 800,000 evaluation

terms to reach an electrostatic energy with accuracy up to six decimal digits. By
using the Ewald summation for such a system, the same accuracy is achieved with
less than a dozen evaluations. The goal is to reduce the calculation cost of the
electrostatic energy of a many-particle system employing the Ewald sum scheme.
A charged atom can be described by a distribution function .ρi centered at the
point .r. In our approach, the ith charge is the distribution positioned at .ri with
charge .qi , described by a delta function.
. ρi (r) = qi δ(r − ri ) (14.9)
The electrostatic potential energy is now written in terms of the distribution

functions.
E EN ∗ ff ρi (r) ρj (r' )
. Eel = 1 1
2 4π ε0 n i,j =1 |r−r' +nL| d 3r d 3r ' (14.10)
14.6 Using the Poisson Equation 339
The electrostatic energy expressed in either of the above forms may be evaluated
for a system with a finite number of particles. However, for an infinite number
of particles, like in the P BC case, the summations converge in relative form,
but not in absolute form. The expression is said to be a conditionally convergent
sum and it is divergent for non-neutral systems. One of the solutions to avoid
complications in the evaluation of the P BC electrostatic energy is to fade the
interactions away in the form proposed by Eq. 14.5 using a truncated potential.
Such an approximation is efficient, but some loss of accuracy is introduced in
the calculation of the electrostatic energy. Another possibility is to consider the
decaying behavior of the electrostatic energy in terms of the interatomic distance,
without the intervention of a truncated potential, but performing the summations
for the charged atoms which are just located within a spherical threshold distance
.rc . A spherical cutoff radius of approximately 8–12 Å may be chosen for every
atom of interest, while the interactions beyond .rc may be neglected. This process
is flexible to impose the required accuracy in the calculation of the electrostatic
energy according to the radius .rc ; nevertheless, some care should be exercised since
the sums are conditionally convergent for big cutoff values.
14.6 Using the Poisson Equation
The lack of efficiency in the calculation of the electrostatic energy is attributed to

the Coulomb expression, because it does not discriminate close atoms from distant
atoms. The goal in this section is to deduce an expression that allows to differentiate
close atoms from distant atoms, maintaining computational efficiency and accuracy,
and making the summations of the electrostatic energy convergent.
The electrostatic potential at point .r created by a source charge .qj positioned at
point .rj , and using the distribution 14.9, is given in the form:
f
1 qj 1 ρj (r' ) 3 '
. ⇒ d r
4π ε0 |r − rj | 4π ε0 |r − r' |
The electrostatic potential .v(i) at point .r and created by all the charged particles,
excluding the ith particle, is:
N∗ f
1 EE ρj (r' )
v(i) (r) =
. d 3r '
4π ε0 n |r − r' + nL|
j =1
⎡ ⎤
N f
EE f
1 ρj (r' ) ρi (r' ) 3 ' ⎦
= ⎣ d 3r ' − d r
4π ε0 n j =1
|r − r' + nL| |r − r' |
f
1E
N
Eel = ρi (r) v(i) (r) d 3 r (14.11)
2
i=1
Fig. 14.8 The Ewald charge distribution functions .ρ s and .ρ l are illustrated in the 1D case,
Eq. 14.13. The up and down vertical lines in the top inset are delta functions representing the
positive and negative charges of the atoms in the 1D grid. The charge density .ρ s is built by both
the delta functions and Gaussian functions, depicted in green color and with opposite signs to the
delta functions. The charge density .ρ l is only built by Gaussian functions, in yellow color and with
equal signs to the delta functions. The addition of positive and negative Gaussian functions gives
zero total charge
The total energy .Eel is written in terms of the interaction potential .v(i) . The
energy excludes the self-interaction energy of the ith atom with the negative term
in square brackets. The above expression of .v(i) establishes an integral relation
between the interaction potential and the charge density. Yet, there is an alternative
relation of differential type between those two quantities using the Poisson equation.
. ∇ 2 v(i) (r) = −ρi (r)/ε0 (14.12)
In order to discriminate close atoms from distant atoms and make the summations
of the energy convergent, it is convenient to introduce distribution functions of
Gaussian type centered at the grid points .ri . They are introduced as a “smart zero,”
by simultaneously adding and subtracting the Gaussian functions, Fig. 14.8, in the
short-range distribution function .ρis and the long-range distribution function .ρil . The
attributes “short-range” and “long-range” shall take sense later in the discussion.
ρis (r) = qi δ(r − ri ) − qi g(r − ri )

. ; ρil (r) = qi g(r − ri )
. g(r) = α exp(−βr 2 ) ; α = 1/(2π σ 2 )3/2 ; β = 1/(2σ 2 ) (14.13)
ρi (r) = ρis (r) + ρil (r)

.
The Gaussian distributions split the charge density .ρi , but they have no con-
tribution to the final electrostatic energy. The role of the Gaussian functions in
14.6 Using the Poisson Equation 341
the charge density .ρis is to screen the point charges as the signs of the Gaussian
functions are opposite to the signs of the delta functions, and their role in the charge
density .ρil is to balance the interactions produced by the Gaussians introduced in
.ρ . The electrostatic potential of a Gaussian function, say .g(r) = α exp(−βr )
s 2
i
with parameters .α and .β, Eq. 14.13, is given in terms of the error function.
√ √ ft
. v(r) = α
4βε0 π/β r −1 erf( βr) ; erf(t) = √2
π 0 exp(−x 2 ) dx (14.14)
Problem 14.3 The Poisson equation is .∇ 2 v(r) = −g(r)/ε0 , where .ε0 is the
vacuum permittivity; the constants may change after using different units.
Show that the electrostatic potential, v, of a Gaussian charge density .g(r) =
α exp(−βr 2 ), with .α = 1/(2π σ 2 )3/2 , .β = 1/(2σ 2 ), is given in terms of the
error function:
/
v(r) = (γ /r) erf( βr)
.
where .γ depends on the parameters .α, .β, and .ε0 . Discuss the asymptotic limit
of the potential at large distances.
Solution The electrostatic potential depends on the distance r and is sym-

metric in the spherical coordinates .θ and .φ. Thus, the Laplacian operator is:
( 2 ) 1 ∂ 2 (rv)
. ∇ 2 v(r) = 1 ∂
r 2 ∂r
r ∂v
∂r = r ∂r 2
We need to integrate twice to obtain v. The first integral is:

f
1 ∂ 2 (rv) α ∂(rv) α r '2
= − e−βr r ' e−βr dr '
2
. ⇒ =−
r ∂r 2 ε0 ∂r ε0 0
f βr 2
α
=− e−t dt
2βε0 0
The change of variables .t = βr '2 was used in the last identity. The integral
of the right involves an exponential, and the result is also an exponential
function.
f
∂(rv) α α 1 r −βr '2 '
e−βr
2
. = ⇒ v(r) = e dr
∂r 2βε0 2βε0 r 0
f √βr
α 1
e−x dx
2
= √
2βε0 β r 0
(continued)
√ '
The change of variables .x = βr was used in the last identity. The
second integral is over the Gaussian function, and it corresponds to the error
function 14.14.
/ /
.v(r) = γ erf( βr)/r ; γ = α/(4βε0 ) π/β
The presence of the error function is due to the impossibility of directly

computing the integral of the Gaussian expression in terms of basic functions.
When the expressions of .α and .β are introduced in .γ , we have .γ = 1/(4π ε0 ).
The error function approaches unity fast enough in the distance. For example,
when the argument of the erf function is 3, the value of the function is over
99% close to the asymptotic limit 1, Fig. 14.9. In such a limit, the potential is
that of a point charge, .v(r) = 1/(4π ε0 r).
Fig. 14.9 The error and complementary error functions are depicted. The addition of such
functions gives 1 (line in blue color). The potential .1/r may be expressed in the form .1/r =
[erf(r) + erfC(r)]/r. The error function tends to 1 in the limit .r → ∞, while the complementary
function tends to 0 in that limit. On the other hand, the product .r −1 erf(r) is finite when .r → 0
and exhibits a long tail when .r → ∞. The product .r −1 erfC(r) is singular when .r → 0 and fades
exponentially away with increasing values of r
14.7 Short-Range Interactions 343
14.7 Short-Range Interactions
The short-range interaction potential 14.11 in terms of the charge distribution .ρis ,
Eq. 14.13, is:
N∗ f ρjs (r' )
1 EE
s
.v (r) = d 3r '
(i)
4π ε0 n |r − r' + nL|
j =1
N∗ f
1 E E qj δ(r' − rj ) − qj g(r' − rj ) 3 '
= d r
4π ε0 n |r − r' + nL|
j =1
The term with the delta function can be integrated. The terms are separated to
reach a simplified expression.
N∗ N∗ f
1 EE qj 1 EE qj g(r − r' ) 3 '
s
.v (r) = − d r
(i)
4π ε0 n |r − rj + nL| 4π ε0 n |r − r' + nL|
j =1 j =1
According to problem 14.3, the potential derived from a Gaussian function is:
(√ )
. v(r) = 1/(4π ε0 ) erf βr /r
This result is employed in the expression of the short-range potential.
s (r) = 1 E EN ∗ q / |r − r + nL| +
v(i) 4π ε0 n j =1 j j
. E EN ∗ q j (√ )
− n j =1 4π ε0 erf β |r − rj + nL| / |r − rj + nL|
The short-range potential and energy .E s are written in terms of the complemen-
tary error function, erfC.
s (r) = E EN ∗ qj (√ )
v(i) 1
4π ε0 n j =1 |r−rj +nL| erfC β |r − rj + nL|
.
EN (14.15)
Es = 1
2
s
i=1 qi v(i) (ri ) ; erfC(x) = 1 − erf(x)
The initial contribution of the delta function in the charge density .ρis is
considered by the unit term that builds the complementary error function. The
s is the electrostatic potential at position .r, avoiding the contribution
function .v(i)
of atom i indicated by the asterisk in the summation. The potential .v(i)s is a short-
range potential due to the fact that the complementary error function of Eq. 14.15
rapidly fades away in the distance, Fig. 14.9. On the other hand, the energy .E s
has equal expression to the electrostatic energy of point charges, except for the
consideration of the erfC function. It enhances the convergence of .E s through the
parameter .β = 1/(2σ 2 ), and for such a reason, it is computed in real space.
14.8 Electrostatic Self-Interactions
The next step is to compute the interaction of a charge with itself. This contribution
is included in the long-range interactions with the purpose to avoid disruption
of the closed expression of the long-range interactions. Yet, the self-interaction
energy is a spurious term that should be dismissed in a later stage from the total
electrostatic energy by direct subtraction to compensate for its insertion in the
long-range interactions. The self-interaction potential and self-interaction energy
are established by: (i) finding the solution of the Poisson equation for the charge
distribution .ρil of Gaussian type, problem 14.3,
[ ] E
qi |r−r
√ i|
. v (ri ) =
l
i
1
4π ε0 |r−ri | erf ; Eself = 12 N l
i=1 qi vi (ri ) (14.16)
2σ
(ii) taking the zero limit in the argument of the error function, .erf(x → 0) =
and √
2x/ π .
( )1/2 E
1 qi 2
. v l (ri ) =
i 4π ε0 σ π ; Eself = 4π1ε0 √ 1 2 N 2
i=1 qi
(14.17)
2π σ
.Eself is a constant energy, which should be subtracted from the total electrostatic
energy.
14.9 Long-Range Interactions

E EN
The last energy to compute is obtained from .ρ l (r) = n i=1 ρi (r + nL),
l
l
originated from the Gaussian functions .ρi , Eq. 14.13. The electrostatic potential of
a Gaussian function is proportional to .erf/r, which is non-singular and contains
the self-interaction. In the limit of large distances, it leads to erf.(r → ∞) = 1,
indicating that the electrostatic potential of .ρ l is a long-range potential. The slow
decay of the potential in real space turns into a rapid decay in reciprocal space. Thus,
it is convenient to work in the reciprocal space. The steps to follow are succinctly
sketched below.
real space reciprocal space

electrostatic potential Fourier transformation electrostatic potential
(Poisson equation) .−→ (Poisson equation)
.↓
solve the equation

.↓
solution inverse Fourier transformation solution

.←−
14.9 Long-Range Interactions 345
We require to perform Fourier transforms to work in the reciprocal space. The

representation of a function in reciprocal space is defined in terms of exponential
functions with imaginary arguments. If the function in real space is .f (r), and
the function in reciprocal space is .f˜(k), then their relation is given by a Fourier
transform.
E f
˜ ˜ 1
.f (r) = f (k) e ik·r
; f (k) = f (r) e−ik·r d 3 r
V V
k
The functions of the reciprocal space are denoted with a tilde. A periodic
condition on f demands .exp(ik · R) = 1, with .R the lattice translational vector.
E E
f (r + R) =
. f˜(k) exp[ik · (r + R)] = f˜(k) exp(ik · r) = f (r)
k k
The periodic condition on the function f is achieved when .k · R = 2π m, with

m an integral number. In this regard, for the lattice in the real space characterized
by the translational vector .R, written in the form .R = n1 a1 + n2 a2 + n3 a3 , we may
generate an alternative geometrical lattice with the conjugate vector .k = m1 a∗1 +
m2 a∗2 +m3 a∗3 . The numbers .mk , .k = 1, 2, 3, are integers similar to the numbers .ni of
the real space grid. The vectors .a∗1 , a∗2 , a∗3 may be found from their dot product with
the primitive real space lattice vectors, .ai · a∗j = 2π δij , where .δij is the Kronecker
delta function. An expression that satisfies this condition is:
( )
∗ ∗ ∗ a2 × a3 a3 × a1 a1 × a2
.(a1 , a2 , a3 ) = 2π , ,
a1 · (a2 × a3 ) a2 · (a3 × a1 ) a3 · (a1 × a2 )
The primitive vectors of the conjugate cell are .a∗1 , a∗2 , a∗3 , obtained from vectors in
real space and vice versa. These primitive vectors have units .1/ length and are called
reciprocal vectors. The scalar triple product .a1 · (a2 × a3 ) in the first denominator
is the volume of the real cell. The other denominators also represent that volume,
since the scalar triple product remains the same with the circular shift of vectors.
V = a1 · (a2 × a3 )
.
The periodic condition .exp(ik · R) = 1 in terms of the lattice vectors in real and
reciprocal space is:
[ (E ) (E )] ( E )
. exp(ik · R) = exp i mi a∗i · nj aj = exp 2π i δij mi nj = 1
The solutions of the differential equations in reciprocal space are of particular

interest. The differentiation of a function in real space corresponds to a multiplica-
tion by the conjugate variable of the reciprocal space, thus facilitating the finding of
the solution in reciprocal space, as shown in the initial sketch. The total distribution
function .ρ l and the Poisson equation in real space, together with their counterparts
in reciprocal space, are:
E EN
ρ l (r) = n i=1 ρi (r + nL)
l ; ∇ 2 v l (r) = −ρ l (r)/ε0
. f (14.18)
ρ̃ l (k) = 1
V V ρ l (r) exp(−ik · r) d 3 r ; k 2 ṽ l (k) = ρ̃ l (k)/ε0
The electrostatic potential .ṽ l in reciprocal space is determined by the charge

density .ρ̃ l divided by .k 2 , Eq. 14.18. Thus, it is convenient to simplify the charge
density .ρ̃ l to latter retake .ṽ l and, after applying a Fourier transform, determine .v l
in real space. The periodic attributes of the exponential function allow to simplify
the expression of .ρ̃ l . Specifically, we use the translational property with the integral
number n and the arbitrary real number .x0 .
ρ(r) = ρ(r + R) ; exp[i(x + 2π n)] = exp(ix)

. (14.19)
if h(x) = f (x − x0 ) ⇒ h̃(k) = exp(−i2π x0 k) f˜(k)
R is the translational vector.

.
Problem 14.4 Consider a periodic charge density .ρ l given in terms of the

individual densities .ρjl .
EE
N
. ρ l (r) = ρjl (r + nL)
n j =1
The sums run over the particles of the central unit cell and the lattice vectors
. n = (n1 , n2 , n3 ) to produce the particle grid. Obtain the density .ρ̃ l in
reciprocal space.
Solution The Fourier transform of the charge density is given as a com-

bination of exponential functions with continuous arguments that demand
integration, and each integral is defined over a unit cell volume.
f
1 EE
ρ̃ l (k) =
. e−ik·r ρjl (r + nL) d 3 r
V n V j
The volume of integration is V ; it may be considered in the reciprocal space,

or in real space, or shared in the form .V −1/2 by both spaces. The expression
is simplified when the change of variables .t = r + nL is done and using the
identity .exp[ik·nL] = 1, which gives the translational invariance requirement
(continued)
14.9 Long-Range Interactions 347
of the unit cell.

f f
1 EE −ik·t l 1 E
.ρ̃ (k) =
l
e ρj (t) d t =
3
e−ik·t ρjl (t) d 3 t
V n V +nL V R3
j j
The summation over integrals covering the image-cell volumes corresponds

to an integral over the 3D space, .R3 .
The previous periodic and translational properties avoid the summation over .n,
which is the vector referent to the image cells in the expression of the charge density,
problem 14.4.
EN f f EN
ρ̃ l (k) = 1
V j =1 R3 e−ik·t ρjl (t) d 3 t = 1
V R3 j =1 qj g(r − rj ) e−ik·r d 3 r
EN f
. = 1
V j =1 qj R3 α exp[−β(r − rj )2 ] exp[−ik · r] d 3 r
α = 1/(2π σ 2 )3/2 ; β = 1/(2σ 2 )

(14.20)
The expression 14.20 is the Fourier transform of a Gaussian function. In order to
carry the integration, we require to apply the result:
f ∞
. exp[−a 2 (x − x0 )2 ] exp(−ikx) dx = (π 1/2 /a) exp(−ikx0 ) exp[−k 2 /(4a 2 )]
−∞
which applies to the 1D case and is generalized to the 3D case.

EN
. ρ̃ l (k) = V −1 j =1 qj exp(−ik · rj ) exp(−σ 2 k 2 /2) (14.21)
E
When .k = 0, we have the neutrality of the system . Nj =1 qj = 0, and .ρ̃ (0) = 0.
l
According to Eq. 14.18, by dividing the above expression by .ε0 k 2 , the electrostatic
potential in reciprocal space is determined.
EN
ṽ l (k) = (V ε0 )−1 2 ) e−ik·rj e−σ
2 k 2 /2
.
j =1 (qj /k (14.22)
There is no divergence in the charge density and electrostatic potential for .k = 0,

since .limk→0 ρ̃ l (k)/k 2 = finite value. We are interested in the expression of the
potential in real space to compute the electrostatic energy. The expression in real
space is obtained by inverse-Fourier transforming the potential from the reciprocal
space to the real space.
E E EN
e−σ k /2
2 2
. v l (r) = k/=0 ṽ
l (k) exp(ik · r) = 1
V ε0 k/=0 j =1 (qj /k
2 ) eik·(r−rj )
(14.23)
The energy due to the potential l
.v in real space is deduced from .E
l =
E l
i qi v (ri )/2.
E EN
. E l = (2V ε0 )−1 k/=0 i,j =1 (qi qj /k
2 ) e−σ 2 k 2 /2 eik·(ri −rj ) (14.24)
The summations show no asterisk as the self-interaction terms are included.

Also, the running index i should not be confused with the imaginary number of
the exponential function. The 3D array of points allows to introduce the structure
factor, S.
EN
. S(k) = j =1 qj exp(ik · rj ) (14.25)
The expression of the energy .E l is now simplified in terms of the structure factor
using the identity .|S(k)|2 = S(k)S ∗ (k) = S(k)S(−k).
E
E l = (2V ε0 )−1 k/=0 |S(k)|2 k −2 exp(−σ 2 k 2 /2)
.
E EN (14.26)
|S(k)|2 = | Nj =1 qj cos(k · rj )| + |
2
j =1 qj sin(k · rj )|
2
The final expression of the long-range electrostatic energy is given in real space,
in spite of the fact that the summation runs over the variable .k. The summation is
symmetric in .k and .−k, and only the positive reciprocal variable .k requires con-
sideration. The summation is rapidly convergent because the exponential function
decays with an increasing value of .|k|. The parameter .σ of the Gaussian functions
is used to control the summation convergence. The efficiency of the calculations
improves from .N 2 to .N 3/2 [5].
14.10 Ewald Electrostatic Energy
We are now in the position of gathering all the energy contributions,

Eqs. 14.15, 14.17, and 14.26, and recast the electrostatic energy 14.10 of the system.
E E E EN EN ∗ [ √ ]
qi qj
Es = 1 1
2 4π ε0 n1 n2 n3 i=1 j =1 |ri −rj +nL| erfC |ri − rj + nL|/( 2σ )
E −2
. El = 1
2V ε0 k/=0 k exp(−σ 2 k 2 /2) |S(k)|2
√ E
Eself = 1
4π ε0 ( 2π σ 2 )−1 N 2
i=1 qi
Eel (r1 , · · · , rN ) = E s − Eself + E l (14.27)
The energy depends on the Gaussian standard deviation, .σ , Eq. 14.13. The
efficiency in the computation of .E s and .E l is due to the erfC function and
14.10 Ewald Electrostatic Energy 349
factor .exp(−σ 2 k 2 /2). The value of .σ is a trade-off in the computational cost of

the terms .E s and .E l ; a small .σ produces the rapid convergence of .E s and a
slow convergence of .E l , and vice versa. Physically, there is a weakening of the
electrostatic interactions in one of the contributions while the other is strengthened,
and technically, there is a scaling of order .O(N ) in one of the contributions and a
scaling of order .O(N 2 ) in the other. The .σ value may be decided after weighting
the .E s and .E l energy contributions with respect to .σ .
[ √ ] ( )
r −1 erfC r/( 2σ ) < er
. ; k −2 exp −σ 2 k 2 /2 < ek
It is common to establish the .σ value that satisfies the inequality on the left
by assigning a value to r larger than the cutoff distance required in the .E s term.
In contrast, the inequality on the right imposes a limit on the conjugate reciprocal
variable, k, to rapidly perform the summation of .E l . The tolerances .er and .ek depend
on the charged atoms of the grid, but are expected to be sufficiently small, taking
values from .10−4 to .10−8 depending on a small, medium, or high accuracy. In many
instances, the cell dipole moment may be included in the Ewald sum method. In the
case of considering an exclusion list, where the electrostatic interactions of selected
pairs of atoms, usually separated by a few bonds, are excluded, it is necessary to
subtract the respective contributions from the total energy. This is done in analogy
to the self-interaction energy of Eq. 14.17, eliminating the selected non-diagonal
terms, .i /= j .
[ ]
1 1 E E qi qj |ri − rj |
.Eexcl =− erfC √ ; (i, j ) e {exclusion list}
2 4π ε0 |ri − rj | 2σ
i j
The electrostatic forces over the ith atom are computed by differentiation of the
energies presented in Eqs. 14.27 and under the considerations that .d [erfC(r)]/dx =
−d [erf(r)]/dx, and .d [erfC(r)]/dx = −(2/π )1/2 σ −1 exp(−r 2 ) (xi −xj +nx L) / r
[5].
F = −∇i (E s − Eself + E l ) ; r = |r| = |ri − rj + nL| ; l = x, y, z

E E qj [ ( √ )
Fls = 4π1ε0 qi N
j /=i n r2 (r l /r) erfC r / [ 2σ ]
. √ ]
+ 2/π (rl / σ ) exp[−r 2 / (2σ 2 )]
E EN
Fll = (2V ε0 )−1 qi k/=0 j /=i qj (kl /k 2 ) exp[−σ 2 k 2 /2] sin[k · (ri − rj )]
The expression 14.24 of the energy .E l simplifies the computation of the force
l
.F .The self-interaction energy is constant and gives no contribution to the force
l
components. The total force over the ith atom should additionally include the other
terms derived from the total energy of the system, such as the bond force, angle
force, dihedral force, van der Waals force, etc. When all the forces are computed for
every atom, the time evolution of the system may be unfolded.
14.11 Smooth Particle Mesh Ewald Approach
With the purpose of improving the efficiency of the Ewald sum approach, a
multidimensional piecewise interpolation of the structure factors 14.25 and 14.26
may be employed. The approach applies a Cardinal B-spline interpolation to a
regular grid to produce an approximation to the structure factors, Fig. 14.10. The
B-splines permit analytic differentiation of the energy, leading to smooth changes
of the forces with the positions of the particles. When the energy and forces of
the reciprocal space are written as convolutions and evaluated with fast Fourier
transforms, the calculations are reduced from .O(N 2 ) down to an almost linear
scaling, .O(N log N). The approach preserves the total energy, but not the total
momentum of the particle system since the forces on the atoms are not exactly
zero. The deviations are estimated to be close to the root mean square error of the
forces. The formulation is of general character and receives the name of smooth
particle mesh Ewald method. It equally applies to potentials decaying in the general
form .1/r x , .1 ≤ x. The smooth particle mesh Ewald method is also used in the
calculation of the London dispersion energy, this is, the long-range part of the van
der Waals interaction when the contributions are given by pairwise interactions:
E EN∗ ij
. E= 1
2 n i,j =1 C6 /|ri − rj + nL|6 (14.28)
The summation is absolute convergent, but computationally costly. In this regard,

it is important to apply the Ewald summation approach as in the electrostatic case
[6].
Fig. 14.10 The evolution of a system of particles from the instant of time t to .t + δt is depicted.
The large filled circles indicate the locations of the particles at the instants of time t (inset on the
left) and .t + δt (inset on the right). The large empty circles in the rendering of the right indicate the
locations of the particles at the previous instant of time. The tiny circles represent a regular grid of
points. The atom charges are projected and interpolated onto that grid at every instant of time. The
electrostatic potential is evaluated from such a grid, and the computed forces from the electrostatic
potential are applied to the particles
14.11 Smooth Particle Mesh Ewald Approach 351
E EN EN ∗ ij g(β|ri −rj +nL|) EN

Edir = 1
2 n i=1 j =1 C6 |ri −rj +nL|6
; Eself = (β 6 /12) ii
i=1 C6
E∗ EN EN ij [ ]
. Erec = π 3/2 β 3 /(2V ) k i=1 j =1 f (π k/β)C6 exp −2π ik · (ri − rj )
E = −Edir + Erec − Eself

(14.29)
The dispersion coefficients are approximated using the combination rules. There
ij jj
are several proposals for such rules; one of them is .C6 = (C6ii C6 )1/2 . The .β
parameter plays the role of .σ −1 in Eqs. 14.27. The auxiliary functions f and g
introduce the complementary error function and the Gaussian function.
[ √ ]
f (r) = (1 − 2r 2 ) exp(−r 2 ) + 2r 3 π erfC(r) /3
. (14.30)
g(r) = exp(−r 2 ) (1 + r 2 + r 4 /2)
The order of the B-splines may be chosen from the .3rd up to the 7th order
for small, medium, and high accuracy. The accuracy is also established with the
grid density. If the P BC simulations require constant pressure, then the stress
tensor P should be computed. It is calculated by differentiating the electrostatic
√ with respect to the atomic coordinates, keeping in mind that .d[erf(r)] /dr =
energy
(2/ π ) exp(−r 2 ).
E EN ∗ [ −3 ]
s V =
Pαβ 1
2 n r erfC(βr) + 2βπ −1/2 r −2 exp(−β 2 r 2 ) rα rβ
i,j =1 qi qj
l V = (2π V )−1 E
Pαβ −2 exp[−(π k/β)2 ] |S(k)|2
k/=0 k
.
[ ]
× δαβ − 2[1 + (π k/β)2 ] kα kβ / k 2
s + Pl
P = Pαβ αβ ; r = |r| = |ri − rj + nL|
(14.31)
The stress tensor is given in terms of fractional coordinates, or coordinates

normalized with the system volume. The terms of an exclusion list may be
introduced in the above identity by reproducing the expression of .Pαβ s V with
negative sign and avoiding the summation over the lattice numbers .n1 , n2 , n3 . The
smooth particle mesh Ewald approach is accurate and efficient. A cutoff imposed in
the real space summation of 9 Å results in van der Waals energy deviations lower
than the errors observed in the electrostatic energy. The pressure is more sensitive
to cutoff distances, showing deviations of hundreds atm units. The deviations are
associated with the fact that the pressure is proportional to the negative derivative
of the energy with respect to the volume, making the pressure highly sensitive to
energy changes. Other physical variables sensitive to a cutoff radius are the diffusion
constant, d, and the dielectric constant, .e, approximated by:
E
N [ ]
d(t) =
. di2 (t) / (6Nt) ; e = 1 + 3 4πρN μ2 / (9kB T ) <m2 > / (Nμ2 )
i=1
.di (t) is the distance travelled by the mass center of particle i in the simulation. The
diffusion constant is calculated in long-time simulations, .t → ∞. The diffusion
constant depends on the solvent and its density. For values of the water density close
to 1 . g/cm3 , the diffusion constant takes values from .2.5 × 10−5 to .6.0 × 10−5 cm2 /s
in a simulation of several ps. On the other hand, in the expression of the dielectric
constant, .m is the dipole moment. The dipole moment of the individual molecule
is denoted by .μ, the number of molecules is N, and .ρN is the numeric density.
The dielectric constant depends on the dipole moment variations, and henceforth,
of the relatively long-range interactions. At ambient temperature and density of 1
.g/cm , we may find .e = 60 − 70, also depending on the water model applied in the
3
simulation.
The electrostatic potential and energy of a large system of charged particles may
be also calculated efficiently using the fast multipole algorithm of Greengard and
Rokhlin [7]. Here, a cell with large size is divided in subcells, and the interactions
of the charged atoms located in the same subcell and similar subcells in the
neighborhood are calculated. For charged atoms in distant subcells, a multipole
expansion is employed to determine the long-range electrostatic interactions. The
approach is equally applicable to gravitational forces.
Finally, a common shortcoming of the methods dealing with periodic charges is
the transmission of the electrostatic interaction, considered to occur instantaneously
when, in fact, there is a retardation effect imposed by the speed of light. Fortunately,
the retardation effect is only important at very large interparticle distances, when the
electrostatic interactions are sufficiently small.3
14.12 Shifted Potentials and Forces
The use of the P BC approach is justified by the impossibility of dealing with a

large number of particles and computing their non-bonded interactions. However,
with the advent of powerful hardware, we are now able to deal with a relatively
large number of particles. This fact has implications in the simulation models, as
those exposed by Fennell and Gezelter [8]. For instance, if we consider that the
electrostatic interactions are effectively short-ranged in condensed phases, then we
face the possibility of directly calculating those interactions with minor deviations
using the present mainframes. In a more precise parlance, it is possible to replace
the P BC approach and use a large cutoff radius for the non-bonded interactions,
with the additional consideration of a spherical charged background whose role is to
3 The light speed is .∼ 3 × 108 m/s, and a typical time step of molecular dynamics is .1 fs. In this
time interval the light has traveled a distance of .∼ 300 ηm.
14.12 Shifted Potentials and Forces 353
Fig. 14.11 Top inset: due to the low efficiency of old mainframes, the Ewald sum approach was
implemented for the rapid calculations of the electrostatic forces of extended systems. In this
approach, a crystal is studied using a relatively small number of atoms in a central unit cell and
repeating it indefinitely according to the P BC scheme. Bottom inset: with the advent of more
efficient mainframes, it is possible to deal with a big number of atoms and compute the non-
bonded interactions within a large cutoff radius. Approximations for the residual interactions of
atoms located far away may be used
neutralize the total charge of the particle system contained within the cutoff radius,
Fig. 14.11. In order to unfold such a proposal, it is convenient to start from a general
expression similar to that given in Eq. 14.5, written below for easy reference, with a
spherical cutoff radius .rc in the shifted potential .Vsp and force .Fsp .
{
V (r) − V (rc ) , r ≤ rc ; Fsp = −∇V (r)
. Vsp = (14.32)
0 , rc < r ; Fsp = 0
The force obtained from expression 14.32 has pairwise nature, but it is not a
continuous function. In order to avoid discontinuities in the force, a new potential
.Vsf and new force .Fsf are proposed in the region .r ≤ rc , with a shift in the force.
∂V (r)
. Vsf = V (r) − V (rc ) − ∂r (r − rc ) ; Fsf = − ∂V∂r(r) + ∂V (r)
∂r
r=rc r=rc
(14.33)
The discontinuity in the force of Eq. 14.32 is avoided in the force of Eq. 14.33,
because the potential and the force presented in Eqs. 14.33 introduce derivative
corrections, making the potential and the force going smoothly to zero at .rc ,
avoiding spurious energy drifts. By introducing the form of V in terms of the
distance r and the charges .qi and .qj , the electrostatic potential and the new force,
paying no attention to constant units, are:
[ ] ( )
. Vel = qi qj 1
r − 1
rc + 1
rc2
(r − rc ) ; Fel = qi qj 1
r2
− 1
rc2
(14.34)
Based on our experience referent to the Ewald sum approach, the Ewald
expression that carries the complementary error function should be used because
this term includes the charge screening in the potential. We have two versions, the
first one is derived from Eqs. 14.32 and is referred to as the damped shifted potential
(dsp) approach.
[ ] ( )
erfC(αr) erfC(αrc ) erfC(αr) 2α exp(−αr 2 )
. Vdsp = qi qj r − rc ; Fdsp = qi qj r2
+ π 1/2 r
(14.35)
We have .Vdsp = 0 when .r = rc . The force .Fdsp is applied in the region .r ≤ rc ,

zero otherwise, but it is discontinuous at .r = rc . When we write .α = 0 and use
the fact erfC.(0) = 1 in the above expressions, the dsp equations of the electrostatic
potential and the electrostatic force are:
[ ]
1 1 qi qj
Vdsp (r)
. = qi qj − ; Fdsp (r) = ; r ≤ rc
α=0 r rc α=0 r2
The second version of the electrostatic expressions is derived from Eqs. 14.33.
The expressions attempt to solve the previous shortcoming on the discontinuity at
.r = rc of the force; they are referred to as the damped shifted force (dsf ) approach.
[
Vdsf = qi qj r −1 erfC(αr) − rc−1 erfC(αrc )
.
( ) ]
+ rc−2 erfC(αrc ) + βrc−1 exp(−αrc2 ) (r − rc )
[( ) (14.36)
Fdsf = qi qj r −2 erfC(αr) + βr exp(−αr 2 )
( )]
− rc−2 erfC(αrc ) + βrc−1 exp(−αrc2 )
These two expressions, with .β = 2α/π 1/2 , gradually vanish toward the boundary
.r = rc , thus avoiding discontinuities. When .α = 0, the dsf equations of the
electrostatic potential and the electrostatic force are obtained.

[ ]
Vdsf (r)
. = qi qj 1
r − 1
rc + 1
rc2
(r − rc )
α=0
[ ]
Fdsf (r) = qi qj 1
r2
− 1
rc2
; r ≤ rc
α=0
14.12 Shifted Potentials and Forces 355
The sets of Eqs. 14.35 and 14.36 avoid the summation in the conjugate reciprocal
variable that characterizes the Ewald sum approach. Typical values of .α are 0.1–0.3
Å.−1 , usually combined with a cutoff radius in the range 9–15 Å. The expressions
of the potentials and forces provided in Eqs. 14.35 and 14.36 have been tested
in representative simulations of chemically relevant systems like liquid water,
crystalline water, NaCl crystals, NaCl melts, ionic solutions, etc. The study avoids
any bias by the types of systems that were tested toward favorable conditions on
the application of any expression. The method of reference is the particle mesh
Ewald method. The properties that have been studied using the damped shifted
potentials and the damped shifted forces are the velocity correlation function .Cv (t)
and its Fourier transform .I (ω), recognized as the power spectrum of the velocity
correlation function [9].
f ∞
1
.Cv (t) = <v(0) · v(t)> / <v(0) · v(0)> ; I (ω) = Cv (t) e−iωt dt
2π −∞
(14.37)
In contrast to pair distribution functions and diffusion constants, the .Cv and .I (ω)
quantities are more sensitive to the long-range interactions and the low frequency
behavior of the substance. After monitoring .Cv (t) in short- and intermediate-time
dynamics and .I (ω) in long-time dynamics with large-scale collective motions, it is
found that the results of each trajectory using the damped shifted proposals with
moderate values, .rc = 9 Å and .α = 0.2 Å.−1 , are encouraging, closely matching the
results obtained with the smooth particle mesh Ewald scheme, where a convergence
parameter of 0.3119 Å.−1 was used.
An alternative model to abbreviate the computation of long-range forces, espe-
cially the electrostatic ones in an electrically neutral system, is the reaction field
approach. In this approximation, a cutoff radius .rc is defined around a molecule,
and the interactions are calculated within the cutoff radius. The external particles to
the cutoff radius are assumed to form a dielectric. The dielectric interacts with the
charges enclosed by the cutoff radius and is permitted to be polarized. The reaction
field potential .Vrf is included in the expression of the interaction of the charges:
[10]
E [ ]
Vrf = 1
4π ε0 i<j qi qj 1
rij + (ε − 1) / (2ε + 1) rij2 /rc3
. [ ] (14.38)
E
f irf = 1
j /=i qi qj 1
− 2(ε − 1) / (2ε + 1) rc−3 rij
4π ε0 rij3
The last term of the potential is the contribution of the reaction field, with
dielectric constant .ε, to the electrostatic interactions. The second expression is the
force on atom i calculated from the potential .Vrf . Care should be exercised when
an atom located at the boundary of the cutoff radius interacts with the central atoms.
In this case, an amendment for the pair interactions should be introduced in the
potential.
[ ]
qi qj 1 qi qj
.
4π ε0
1
rc + (ε − 1)/(2ε + 1) 1
rc = 4π ε0 rc [1 + (ε − 1)/(2ε + 1)] (14.39)
The reaction field model simplifies the computations by considering the local
electrostatic interactions explicitly and simplifying the distant electrostatic interac-
tions with implicit terms.
The general conclusion is that the damped shifted approaches and the reaction
field model represent alternative schemes to those employing lattice summations.
They are capable to maintain both the efficiency since the computations scale
linearly and are friendly parallelizable, and the accuracy of the calculations. Finally,
the damped shifted expressions and reaction field model avoid the artificial stabi-
lization of over-compacted proteins, occasionally observed in the P BC approach
and associated with the uniform periodic images in the Ewald summation method.
References
1. V.A. Kuzkin, Z. Angew, On angular momentum balance for particle systems with periodic
boundary conditions. Math. Mech. 95, 1290–1295 (2014)
2. R.B. Shirts, S.R. Burt, A.M. Johnson, Periodic boundary condition induced breakdown of the
equipartition principle and other kinetic effects of finite sample size in classical hard-sphere
molecular dynamics simulation. J. Chem. Phys. 125, 164102 (2006)
3. P.P. Ewald, The calculation of optical and electrostatic grid potential. Ann. d. Physik., Leipzig
64, 253–287 (1921)
4. C. Kittel Introduction to solid state physics, 7th ed. (Wiley, New York, 1996), Chapt. 2 and
App. B. H. Lee, W. Cai, Ewald summation for Coulomb interactions in a periodic supercell,
P DF document, January 10, 2009
5. A.Y. Toukmaji, J.A. Board Jr., Ewald summation techniques in perspective: a survey. Comput.
Phys. Commun. 95, 73–92 (1996)
6. T. Darden, D. York, L. Pedersen, Particle mesh Ewald: an N log(N ) method for Ewald sums
in large systems. J. Chem. Phys. 98, 10089–10092 (1993). U. Essmann, L. Perera, M.L.
Berkowitz, T. Darden, H. Lee, L.G. Pedersen, A smooth particle mesh Ewald method. J. Chem.
Phys. 103, 8577–8593 (1995)
7. L. Greengard, V. Rokhlin, A fast algorithm for particle simulations. J. Comput. Phys. 73, 325–
348 (1987)
8. C.J. Fennell, J.D. Gezelter, Is the Ewald summation still necessary? Pairwise alternatives to the
accepted standard for long-range electrostatics. J. Chem. Phys. 124, 234104 (2006)
9. N.H. March, M.P. Tosi Introduction to Liquid State Physics (World Scientific, New Jersey,
2002), Chapt. 5
10. M. Neumann, The dielectric constant of water: computer simulations with the MCV potential.
J. Chem. Phys. 82, 5663–5672 (1985)
Part V
Time Evolution Operators
Integrating the Equations of Motion
15
There are different formulations of classical mechanics like that of Newton,

Lagrange, Hamilton, Hamilton-Jacobi, and Poisson. They all have the common
goal of providing in one way or another the equations of motion that rule the
behavior of the classical particles. Yet, the different formulations provide no general
or analytical solutions to the equations of motion, because the atomic forces are
intricate and depend on the characteristics of the atoms. In this context, there are
complications in the process of finding the solutions of the motion equations.
The purpose of the present chapter is to discuss the discrete time propagators,
also known as numerical time integrators. By using them, we are able to evolve the
positions and velocities of the particles with time, thus avoiding the commitment of
analytically solving the equations of motion, which in most conditions of realistic
particle systems is impossible. The discussion starts by introducing the Poisson
propagator, because it represents a formal way to evolve the system of particles in
time. However, since the Liouville operator is fundamental to build the basic blocks
of the discrete time propagators, a connection of the Poisson brackets is established
with the Liouville operator. In the simulations, the initial conditions of the particles
are supposed to be known, and the forces over each particle are assumed to be
computable at every instant of time from either first-principles calculations or model
potential functions, with the goal of determining the unique particle trajectories, and
thereby, establishing the history of the particle system.
15.1 Liouville Operator as a Time Propagator
The method based on Poisson brackets, Sect. 6.22, gives the possibility of creating
a classical time propagator to investigate the time evolution of the particles. For
practical reasons, it is convenient to work with the Liouville operator, .L̂, as defined
in terms of the Poisson brackets in Eq. 6.69. If .H is the system Hamiltonian,
and .F ({qi }, {pi }, t) represents an arbitrary function in terms of the generalized
coordinates q, p, and time, then the action of .L̂ on function F is defined by:
360 15 Integrating the Equations of Motion
∂F ∂H ∂H ∂F
L̂F = {F, H} =
. −
∂q ∂p ∂q ∂p
Suppose that F depends on the positions and velocities solely. The time
derivative of F is written in terms of the Liouville operator, Eq. 6.72.
.
dF
dt = {F, H} = −{H, F } = L̂F (15.1)
The solution of the equation is of exponential type.

[ ]
d
dt F (t) = d
dt exp(L̂t) F (t0 ) = L̂ exp(L̂t) F (t0 ) = L̂F (t)
. (15.2)
F (t) = exp(L̂t) F (t0 )
The exponential operator .exp(L̂t) is a symbolic representation of a power series

in terms of .L̂t. The imaginary symbol i may be employed under convention in
the definition of the Liouville operator, like .i L̂ instead of .L̂, and in analogy to the
time propagators of quantum mechanics. The function .F (t0 ) includes the initial
conditions on the positions and momenta at the instant of time .t0 . The expression
.exp(L̂t) is the classical time propagator giving the propagation of F from time .t0 to
time t. If the function .F (t) represents a position or momentum, it may be advanced

forward in time. To illustrate this, we write the Hamiltonian in terms of the kinetic
energy and potential energy, .H = p2 /(2m) + V (ri ). Under such a representation,
the Liouville operator is:
[ ] [ ]
p
. L̂F = {F, H} = m · ∂ ∂ ∂
∂x , ∂y , ∂z F +f· ∂ ∂ ∂
∂px , ∂py , ∂pz F (15.3)
The net force acting on the particle is .f. We deal with the Liouville operator in
the form discussed by Tuckerman et al. [1] The action of .L̂ is divided into .L̂1 to
include the effect of the momenta on the positions, and .L̂2 to include the effect of
the forces on the momenta.
[ ] [ ]
p
. L̂1 = · ; L̂ = ·
∂ ∂ ∂ ∂ ∂ ∂
m , ,
∂x ∂y ∂z 2 f ∂p x , ∂p y , ∂p z (15.4)
The Liouville operators .L̂1 and .L̂2 may adopt different expressions according
to the types of coordinates. A summation is introduced to describe the system of n
particles.
En pi En
. L̂1 = i=1 mi · ∇i ; L̂2 = i=1 fi · ∇pi (15.5)
The first operator evolves the positions and the second one the momenta.
15.2 Discretizing the Time Propagator 361
Problem 15.1 Demonstrate that .L̂1 and .L̂2 , Eqs. 15.5, are non-commuting
operators.
Solution For simplicity, we work with operators expressed in a single

dimension. The commutator of .L̂1 and .L̂2 is .[L̂1 , L̂2 ] = L̂1 L̂2 − L̂2 L̂1 . By
introducing the expressions of the operators in the commutator, we obtain:
( ) ( )
p ∂ ∂ ∂ p ∂ p ∂f ∂ f ∂
.[L̂1 , L̂2 ] = f (x) − f (x) = −
m ∂x ∂p ∂p m ∂x m ∂x ∂p m ∂x
Therefore, .[L̂1 , L̂2 ] /= 0, and .L̂1 and .L̂2 of Eqs. 15.5 do not commute.
15.2 Discretizing the Time Propagator
In principle, the Liouville operator has the capability to evolve the system from
time .t0 to t in a single time step, Eq. 15.2. Nevertheless, we can use a concise series
expansion to advance the system in small time increments.
The result of applying the operators .L̂1 and .L̂2 on a function F depends on
the application order. In this regard, a symmetrical factorization similar to the
symmetrization process of operators in quantum mechanics is used. Based on the
Trotter expansion, the symmetrical factorization is:1
. exp[(L̂1 + L̂2 )t] = lim [eL̂2 t/2N eL̂1 t/N eL̂2 t/2N ]N (15.6)
N →∞
We have an approximation to the Trotter expansion.
. exp[(L̂1 + L̂2 )t] = [eL̂2 t/2N eL̂1 t/N eL̂2 t/2N ]N + O(t 3 /N 2 ) (15.7)
The expression is accurate up to second order. It is convenient to apply the Nth

root on both sides of the expression, with the simultaneous definition of the time
integration step .At = t/N.
. exp[(L̂1 + L̂2 )At] = eL̂2 At/2 eL̂1 At eL̂2 At/2 (15.8)
1 According to the Trotter theorem, the exponential .exp(P + Q) may be decomposed in the form
.exp(P + Q) = limN →∞ [exp(Q/2N ) exp(P /N ) exp(Q/2N )]N when N is a large integer. Trotter
[2].
The time step .At is the time interval defined for the numerical integration of the
equations of motion. Different time steps may be introduced in the particle dynamics
by dealing with the exponential factor in the middle of the expression.
. exp[(L̂1 + L̂2 )t] = eL̂2 At/2 [eL̂1 δt ]M eL̂2 At/2 ; δt = At/M (15.9)
The time step .δt is a fraction of .At, and therefore, .δt < At. When working
with particles whose masses are notably different, or the forces exhibit substantially
different magnitudes, then different time steps are usually required. The different
time steps in the integration of the equations of motion are particularly helpful for
particle systems where we observe multiple time scales [3]. The approach may
be used as many times as required in a simulation, with multiple applications
depending on the type of system under investigation. The time propagator is now
discretized based on the Liouville operator 15.8.
. U (At) = U2 (At/2) U1 (At) U2 (At/2) ; Ui (At) = eL̂i At (15.10)
Problem 15.2 From the general expansion of .exp(Ât) = 1 + Ât + Â2 t 2 /2! +
· · · , show that the following approximations are satisfied up to first- and
second-order terms, respectively.
exp[(L̂1 + L̂2 )t] = exp[L̂1 t] exp[L̂2 t]

.
exp[(L̂1 + L̂2 )t] = exp[L̂2 t/2] exp[L̂1 t] exp[L̂2 t/2]
Solution The common term in both approximations is .exp[(L̂1 +L̂2 )t]. Thus,
we expand the operator .exp[(L̂1 + L̂2 )t].
. exp[(L̂1 + L̂2 )t]

( ) ( )2 ( )3
= 1 + L̂1 + L̂2 t + L̂1 + L̂2 t 2 /2! + L̂1 + L̂2 t 3 /3! + · · ·
( ) [ ]
= 1 + L̂1 + L̂2 t + L̂21 + L̂1 L̂2 + L̂2 L̂1 + L̂22 t 2 /2 + · · ·
On the other hand, the product of individual expansions gives:

[ ][ ]
. exp[L̂1 t] exp[L̂2 t] = 1 + L̂1 t + L̂21 t 2 /2 + · · · 1 + L̂2 t + L̂22 t 2 /2 + · · ·
( ) [ ]
= 1 + L̂1 + L̂2 t + L̂21 + 2L̂1 L̂2 + L̂22 t 2 /2 + · · ·
(continued)
15.3 Evolving Positions and Momenta 363
Thereby, the approximation .exp[(L̂1 + L̂2 )t] = exp(L̂1 t) exp(L̂2 t) holds

up to the first-order term, because the following terms break the sym-
metry between .L̂1 and .L̂2 in the expansion of .exp[(L̂1 + L̂2 )t]. A
similar demonstration is provided in the case of .exp[(L̂1 + L̂2 )t] =
exp(L̂2 t/2) exp(L̂1 t) exp(L̂2 t/2). A symbolic algebra package may facilitate
the demonstration, reaching the conclusion that the approximation holds up to
second-order terms in this last case.
The inverse operator is obtained by replacing .+t by .−t. In this connection, the
classical propagator 15.10 is demonstrated to be unitary, allowing time reversibility.
( )( )
U (At) U (−At) = eL̂2 At/2 eL̂1 At eL̂2 At/2 e−L̂2 At/2 e−L̂1 At e−L̂2 At/2
.
= eL̂2 At/2 [eL̂1 At e−L̂1 At ] e−L̂2 At/2 = eL̂2 At/2 e−L̂2 At/2 = I
(15.11)
The time reversibility is a property of all the particle systems, except in

the dissipative particle approximation, where a portion of the system energy is
transformed into heat and vice versa, making the time evolution irreversible [4].
15.3 Evolving Positions and Momenta
In the previous section, we briefly discussed the Trotter expression and the way it
enriches the time propagator expressions. The next step is to apply the classical time
propagator to implement numerical time integrators. We start from the operators .L̂1
and .L̂2 of Eqs. 15.5.
E
n
pi E
n
.L̂1 = · ∇i ; L̂2 = fi · ∇pi
mi
i=1 i=1
En
The operators are deduced from a Hamiltonian of type .H = i=1 pi /2mi +
2
V (ri ). The time propagator 15.2 with inclusion of Eq. 15.8 evolves the function f
from time .t0 to time t.
[ ]
. F (t) = eL̂At F (t0 ) = eL̂2 At/2 eL̂1 At eL̂2 At/2 F (t0 ) (15.12)
The insertion of the individual operators .L̂1 and .L̂2 in the exponential functions
gives:
[ ( )] [ ( )]
At pi
.F (t) = ||i=1 exp fi · ∇pi · ∇i
n n
||i=1 exp At
2 mi
[ ( )]
At
||ni=1 exp fi · ∇pi F (t0 ) (15.13)
2
where the summations present in the .L̂1 and .L̂2 operators were transformed onto
products of exponential functions in the expression of .F (t). We now evaluate the
action of the exponential operators on function F , which is considered to be either
.ri or .pi , to advance the position and momentum of the ith particle. In this process,
the exponential function .exp[a ∂/∂g(t)] is approximated:
. exp[a ∂/∂g(t)] = 1 + a ∂/∂g(t)
where a is independent of t. The first operator to be applied is the rightmost

expression of Eq. 15.13.
( )
At At
.pi (At/2) = exp fi · ∇pi pi (t0 ) = pi (t0 ) + fi (ri (t0 ))
2 2
The term .(At/2)fi was identified with a and the function g with the momentum
of the particle. The operator in the middle of Eq. 15.13 acts on the positions. The
computations are performed in a similar way as the previous one.
( )
pi
ri (At) = exp At
. · ∇i ri (t0 ) = ri (t0 ) + At pi (At/2)/mi
mi
The leftmost operator of Eq. 15.13 acts on the momentum. It is computed in a

similar form, but shifted in time.
( )
At At
.pi (At) = exp fi · ∇pi pi (At/2) = pi (At/2) + fi (ri (At))
2 2
It is convenient to gather all the results with the proper time order.
pi (At/2) = pi (t0 ) + At
2 fi (ri (t0 ))
. ri (At) = ri (t0 ) + At pi (At/2)/mi ; fi (ri (At)) is evaluated (15.14)
pi (At) = pi (At/2) + At
2 fi (ri (At))
The set of Eqs. 15.14 represents a process in the evaluation of the positions
and momenta in terms of the time in the simulation; given the initial conditions
.{ri (t0 ), pi (t0 )}, the momentum should be first obtained at half time step, .pi (At/2),
to later proceed with the computation of the position .ri (At) at the immediate
full time step, and finish with the momentum at the full time step, .pi (At). The
15.3 Evolving Positions and Momenta 365
force is evaluated at full time steps, after .ri (At) is determined. The strategy
based on the Liouville time propagator to create time integrators is known in
the literature as the RESP A approach, acronym of reversible reference system
propagator algorithm. The RESP A approach is characterized by its stability, with
the possibility of simultaneously dealing with different time scales, ensuring time
reversibility, accelerating the integration of the equations of motion, and sufficiently
versatile to include other integration strategies [5].
The Eqs. 15.14 may be further simplified. To do this, the first identity is inserted
in the second and third expressions, to end up with just a pair of equations, one for
the position and the other for the momentum.
ri (At) = ri (t0 ) + pi (t0 ) At/mi + fi (ri (t0 )) At 2 /(2mi )

. fi (ri (At)) is evaluated (15.15)
pi (At) = pi (t0 ) + [fi (ri (t0 )) + fi (ri (At))] At/2
This integrator is recognized as the velocity-Verlet time integrator. In the next

time step, the position and velocity may be advanced forward in time by a time
amount .At.
(At)2
ri (2At) = ri (At) + At vi (At) + 2 ai (ri (At))
. fi (ri (2At)) is evaluated (15.16)
vi (2At) = vi (At) + At
2 [ai (ri (At)) + ai (ri (2At))]
We are equally interested in traveling in the reverse time direction employing the
velocity-Verlet equations. To do this, we suppose that several time steps have been
performed in the simulation. The goal is to reach point .[ri (At), vi (At), ai (At)]
from point .[ri (2At), vi (2At), .ai (2At)]. The negative time step, .−At, is considered
in Eq. 15.16 to build the new position and velocity, denoted by .ri (t ' ) and .vi (t ' ).
(At)2
ri (t ' ) = ri (2At) − At vi (2At) + 2 ai (ri (2At))
. (15.17)
vi (t ' ) = vi (2At) − At
2 [ai (ri (At)) + ai (ri (2At))]
We require to identify the variables .ri (t ' ) and .vi (t ' ). We insert the velocity-Verlet
Eqs. 15.16 in terms of .ri (At) and .vi (At) in the above expressions.
(At)2
.ri (t ' ) = {ri (At) + At vi (At) + ai (At)}
2
At
− {vi (At) +
[ai (At) + ai (2At)]} At + ai (2At) At 2 /2
2
At
vi (t ' ) = {vi (At) + [ai (At) + ai (2At)]}
2
− [ai (At) + ai (2At)]At/2
The results of such equations are .ri (t ' ) = ri (At) and .vi (t ' ) = vi (At). Therefore,
the equations permit time reversibility because the old variables can be recovered
systematically from the new ones.
ri (At), vi (At), ai (At)

. − At ri (2At), vi (2At), ai (2At)
←−
Problem 15.3 Demonstrate that .F (t + At) is the result of evolving .F (t)

in time after using the operator .exp(At ∂/∂t) on .F (t). On the other hand,
consider the diffusion expression .∂ρ/∂t = d ∇ 2 ρ, Eq. 12.46, and the
Newtonian equation .fx (t) = m d 2 x/dt 2 . They are written in terms of finite
differences:
[ ]
[ρ(t + δt) − ρ(t)] / δt = (d/2) ∇ 2 ρ(t + δt) + ∇ 2 ρ(t)
.
fx = m [x(t + δt) − 2x(t) + x(t − δt)] / δt 2
Are these equations time reversible?
Solution The exponential function is expanded in a power series and applied

to F at the instant of time t.
[ ] [ ]
∂ At ∂ At 2 ∂ 2 At 3 ∂ 3
. exp At F (t) = 1 + + + + · · · F (t)
∂t 1! ∂t 2! ∂t 2 3! ∂t 3
1 (2)
= F (t) + F (1) (t) At + F (t) At 2 + · · ·
2!
= F (t + At)
Therefore, .exp(At ∂/∂t) F (t) advances the function F forward in time by an

amount .At. In reference to the second point, the time evolution proceeds from
t to .t + δt. In order to travel backward in time, we evolve the system from t
to .t − δt in the diffusion equation and from .t + δt to .t − δt in the Newtonian
expression.
ρ(t − δt) − ρ(t) d[ 2 ]

. = ∇ ρ(t − δt) + ∇ 2 ρ(t) ;
δt 2
x(t − δt) − 2x(t) + x(t + δt)
fx = m
δt 2
We rename variables by writing .t = t ' + δt in both equations.
(continued)
15.4 Simplified Time Integrators 367
[ρ(t ' + δt) − ρ(t ' )] d[ 2 ' ]

. − = ∇ ρ(t + δt) + ∇ 2 ρ(t ' ) ;
δt 2
x(t ' + 2δt) − 2x(t ' + δt) + x(t ' )
fx = m
δt 2
After doing the time-reverse process, the initial and final diffusion equations
differ in signs, but the initial and final Newtonian equations are unchanged.
The diffusion equation is time irreversible, and the Newtonian equation is time
reversible.
15.4 Simplified Time Integrators
In the following sections, we derive some integrators of common type from the
equations of the velocity-Verlet approach, maintaining the character of a series
expansion. The differences among the integrators are due to the way the positions
and the velocities are processed. In some of those integrators, the positions and
velocities are shown in synchronous or retarded forms, namely, the dynamical
variables may be defined at the same instant of time, like .r(t) and .v(t), or one of
them may be delayed (retarded) with respect to the other, like the velocity .v(t − At)
from the position .r(t).
In order to abbreviate the notation, we use .r to represent the position variable
of any particle; however, the notation .rα , with a Greek index .α, makes reference
to a particular atom. The simulation time is partitioned in small intervals dt. The
time step dt in a simulation is frequently proposed as the oscillation period of the
fastest particle divided by a factor of ten. For example, if the hydrogen atom of the
H Cl molecule vibrates with a period of .1.13 × 10−14 s ≈ 10f s, then the time step
to choose in the simulation is .dt = 1f s. A variable evaluated at the ith instant of
time means evaluation at .ti = i dt, with .i = 0, 1, · · · . The evaluation at half-integer
numbers is equally possible, .ti+1/2 = i dt + dt/2.
. ti = i dt ; ti+1/2 = i dt + dt/2 ; i = 0, 1, · · · (15.18)
The representation .r(ti ) is replaced by .ri , which should not be confused with the
position of particle i. The same treatment is given to the other dynamical variables,
like the velocity .vi and acceleration .ai . We start with the leapfrog algorithm
and continue with the Verlet algorithm. Their basic characteristics reside on the
simplicity, computational efficiency, accuracy using sufficiently small time steps,
stability, and time reversibility.
15.5 Leapfrog Algorithm
There are alternative time integrators that may be obtained from the expressions of
the velocity-Verlet integrator, Eqs. 15.15. In order to derive one of those integrators,
we start from Eqs. 15.15, which are written with the new notation here.
. ri+1 = ri + vi dt + ai dt 2 /2 ; vi+1 = vi + (ai + ai+1 ) dt/2 (15.19)
All the variables are evaluated at integral time steps. We suppose the equivalences
vi+1/2 = vi + ai dt/2 and .ai+1/2 = (ai + ai+1 )/2. They correspond in that order
.
to the last two terms of .ri+1 and .vi+1 of Eqs. 15.19. Thus, they are inserted in such
equations, but writing .j = i + 1.
rj = rj −1 + vj −1/2 dt
. ; vj = vj −1 + aj −1/2 dt
These expressions may be interpreted as follows: the evolution of the position is

governed by the velocity evaluated at a half-integer time step, while the evolution
of the velocity is governed by the acceleration evaluated at a half-integer time step.
Still, the equation of .rj is given in terms of .vj −1/2 , while the equation of the right
gives .vj . The equations are not linked to each other because the speeds are shifted
in time. In order to make them self-consistent, we change the j index of the velocity
by .j + 1/2.
. rj = rj −1 + vj −1/2 dt ; vj +1/2 = vj −1/2 + aj dt (15.20)
The position and velocity are now coupled through velocities computed at half-
integer time steps. The above equations constitute the leapfrog time integrator. It
is based on a time leap since the position and velocity differ by a half-integer time
step. Also, the acceleration is shown in the expression of the velocity, but it is not
present in the expression of the position. The leapfrog integrator is accurate up to
second order. In the case of accuracy or stability problems, it is always possible to
consider sufficiently small time steps to avoid such problems, with the consequent
computational cost of performing more steps in the simulation. In order to illustrate
the propagation of variables in this model, let us assume that the initial values
.(r0 , v0 , a0 ) are given, Fig. 15.1. An initial expression of the velocity is proposed:
v1/2 = v0 + a0 dt/2
.
There is no general consensus to offset the first velocity .v1/2 ; nevertheless, the
above proposal goes along the initial suppositions on the formulation of the leapfrog
algorithm. Once the first velocity is determined, we obtain the first position .r1 in
terms of the known quantities according to the leapfrog expression 15.20.
r1 = r0 + v1/2 dt
.
15.5 Leapfrog Algorithm 369
Fig. 15.1 The leapfrog algorithm is illustrated. The initial conditions are .(r0 , v0 , a0 ) at the instant
of time .t0 . Once the velocity .v1/2 is obtained, the new position .r1 is determined. At this stage, the
process is repeated, computing the particle acceleration from the force
Table 15.1 Propagation of the position and velocity forward and backward in time in the
leapfrog approach. The force .fi is immediately computed after the position .ri is known. The
initial conditions .(r0 , v0 , a0 ) are supposed to be given.∗
.r0 , .v0 , .a0 .r0 , .v0 , .a0
.. ..
.. ..
.↓ .ri = ri−1 + vi−1/2 dt .ri = ri+1 − vi+1/2 dt .↑
. vi+1/2 = vi−1/2 + ai dt . vi+1/2 = vi+3/2 − ai+1 dt

T .ri+1 = ri + vi+1/2 dt .ri+1 = ri+2 − vi+3/2 dt T
i . vi+3/2 = vi+1/2 + ai+1 dt . vi+3/2 = vi+5/2 − ai+2 dt i
m .ri+2 = ri+1 + vi+3/2 dt .ri+2 = ri+3 − vi+5/2 dt m
e . vi+5/2 = vi+3/2 + ai+2 dt . vi+5/2 = vi+7/2 − ai+3 dt e
..
.ri+3 = ri+2 + vi+5/2 dt ..
.↓ . vi+7/2 = vi+5/2 + ai+3 dt .ri+3 , .vi+7/2 , .ai+3 .↑

.. ..
.. ..
*The leapfrog equations in the forward and backward time directions are similar because in both
situations the previous position, velocity, and acceleration are required, except for the change of
signs of dt
The force is calculated at the new position .r1 , giving .a1 . This gives the
opportunity to compute the next speed from the leapfrog proposal.
v3/2 = v1/2 + a1 dt
.
At this stage the first cycle is completed in the time evolution process because we
know .r1 , .v3/2 , .a1 . The second cycle similarly follows the leapfrog algorithm. The
general time-forward process is sketched in Table 15.1.
The leapfrog algorithm also allows time reversibility. Thus, after advancing a
certain number of time steps forward in time, it is possible to return to the original
position and velocity by reversing the algorithm. The time reversibility is ascribed
to the symmetry in the expressions of both the position and velocity in the leapfrog
approach. In order to exemplify the time reversibility, we suppose that the last
known vectors are (.ri+3 , .vi+7/2 , .ai+3 ), Table 15.1. According to the leapfrog
equations, we firstly solve for the velocity .vi+5/2 and later for the position .ri+2 .
vi+5/2 = vi+7/2 − ai+3 dt

. ; ri+2 = ri+3 − vi+5/2 dt
These operations give the new set of values .(ri+2 , vi+5/2 , ai+2 ), where the
acceleration .ai+2 is computed at the position .ri+2 . This process gives the first
reversed cycle because the old positions and velocities, .(ri+2 , vi+5/2 , ai+2 ), are
recovered from .(ri+3 , vi+7/2 , ai+3 ). By repeating a number of similar steps, it is
possible to reach the initial conditions from which the dynamics was originally
started, Table 15.1. In short, the leapfrog expressions are invariant under time
reversibility, as in going backward in time the functional forms of the equations
are retained, except for the corresponding signs of the time steps.
The time reversibility is an important feature of time integrators because such
a property allows to verify the conservation of physical quantities like the energy,
linear and angular momenta, and more, in the reversed time direction. The lack of
conservation of such variables may be attributed to many possible situations that
are capable of producing a long-term drift of the particle trajectories. However, it
is rarely associated with the leapfrog time integrator. A machine performs millions
of mathematical operations with small rounding errors, which accumulate in the
molecular simulation leading to slightly different energies of the particle system.
In this case, the machine, not the numeric integrator, produces the deviations.
On the other hand, the systematic errors may be also present. These may be
estimated by knowing the number of reliable digits of the energy. Be also aware
of the cancellation of errors which, in principle, may mask important deviations.
Fortunately, this is an infrequent situation.
15.6 Verlet Algorithm
The Verlet algorithm is a time integrator which allows to unfold the dynamics to
create the trajectory and history of each particle with time. In order to derive the
equations of the Verlet approach, we start from the velocity-Verlet expression of
the position given in Eqs. 15.15. It is written in the forward and backward time
directions.
ri+1 = ri + vi dt + 1
2 ai dt 2 + 1
6 bi dt 3 + O(dt 4 )
. (15.21)
ri−1 = ri − vi dt + 1
2 ai dt 2 − 1
6 bi dt 3 + O(dt 4 )
The sum of the previous equations gives the position at the instant of time .ti+1 .
. ri+1 = 2ri − ri−1 + ai dt 2 + O(dt 4 ) (15.22)

15.6 Verlet Algorithm 371
The new position is given in terms of the previous positions and acceleration.
In contraposition to the leapfrog position 15.20, the Verlet position has no direct
dependence on the velocity, but the acceleration. The velocity is required to evaluate
the kinetic energy, and thus, a proposal of the velocity is needed. The velocity is
given by the difference of Eqs. 15.21.
. vi = [ri+1 − ri−1 ] / (2dt) − bi dt 3 / [3 (2dt)] + O(dt 5 ) (15.23)
The term with the contribution .bi may be neglected to deal with a simplified
expression of the velocity.
. vi = (ri+1 − ri−1 ) / (2dt) + O(dt 2 ) (15.24)
In order to compute the velocity at time .ti , we require to have the positions at the
instants of time .ti+1 and .ti−1 . The computation of the velocity .vi is delayed by one
time step from the position .ri+1 . Equations 15.22 and 15.24 constitute the Verlet
time integrator of the position and velocity; they are written below for easy access.
ri+1 = 2ri − ri−1 + ai dt 2 + O(dt 4 )

. (15.25)
vi = (ri+1 − ri−1 ) / (2dt) + O(dt 2 )
The acceleration is derived from the force exerted on the particle, which in turn
is computed on the fly from the first-principles mechanical equations, involving a
quantum electronic-structure method, or derived from a classical model potential.
It is of interest to estimate the accumulation of errors in the Verlet approach in a
time interval .τ of the simulation. We know that the position 15.25 is accurate up to
third order.
. Error (ri+1 ) = O(dt 4 ) (15.26)
The simulation is supposed to be initiated from the position .ri , at time .ti . The
error in the position at this instant of time is null, .Error (ri ) = 0. The error
accumulated in the following step is calculated from expression 15.25, namely,
.ri+2 = 2ri+1 − ri + ai+1 dt + O(dt ).
2 4
Error (ri+2 ) = 2 Error (ri+1 ) − Error (ri ) + O(dt 4 ) = 3 O(dt 4 )

.
No error is attributed to the force. At step .ti+3 , the computation of the error is
similarly worked out. We have .ri+3 = 2ri+2 − ri+1 + ai+2 dt 2 + O(dt 4 ), and there
is an error of .ri+2 , an error of .ri+1 , Eq. 15.26, and the carrying error .O(dt 4 ).
Error (ri+3 ) = 2 Error (ri+2 ) − Error (ri+1 ) + O(dt 4 )

.
= 2[3 O(dt 4 )] − O(dt 4 ) + O(dt 4 ) = 6 O(dt 4 )

We follow a similar procedure at the instant of time .ti+4 , with .ri+4 = 2ri+3 −
ri+2 + ai+3 dt 2 + O(dt 4 ).
Error (ri+4 ) = 2 Error (ri+3 ) − Error (ri+2 ) + O(dt 4 )

.
= 2[6 O(dt 4 )] − 3 O(dt 4 ) + O(dt 4 ) = 10 O(dt 4 )
The generalization of the result to n simulation steps is:
n(n+1)
. Error (ri+n ) = 2 O(dt 4 ) (15.27)
By supposition, n time steps cover the time interval .τ , this is, .ndt = τ . The error
gathered in such an interval of time is obtained by inserting .n = τ/dt in Eq. 15.27.
( )
τ2
. Error (ri+n ) = 1
2 dt 2
+ τ
dt O(dt 4 ) = O(dt 2 ) (15.28)
The error in the velocity, Eq. 15.25, is of order .O(dt 2 ) because the expression
of the velocity is based on the difference of contiguous positions. For example, for
.i + 1 simulation steps, we have:
Error (ri+1 ) − Error (ri−1 )

Error (vi ) =
. + O(dt 2 )
2dt
[ ]
1 (i + 1)(i + 2) (i − 1)i
= − O(dt 4 ) + O(dt 2 )
2dt 2 2
1
= [4i + 2] O(dt 4 ) + O(dt 2 )
4dt
The error of the Verlet velocity remains of order .O(dt 2 ). The Verlet scheme is
one of the most commonly used integrators in molecular dynamics. It is numerically
stable and exhibits time reversibility when small time steps are employed. Other
numerical time integrators are available such as the Runge-Kutta, or predictor-
corrector schemes. They all find wide applications in computer animation to
produce ripples in water, cloth movements, articulated bodies, etc. The utility of
any integrator is recognized when the trajectory of each particle of a many-particle
system is accomplished in an extended time interval, and thus, the history of the
particle system is known.
Problem 15.4 The position of the center of mass, cm, of a particle system is
rcm . The velocity of the mass center obtained from the positions .rcm in the
.
Verlet approach is:
(continued)
15.7 Bond Constraints 373
vcm (t) = [rcm (t + dt) − rcm (t − dt)] / (2dt)

.
Obtain .vcm from the set of particles with positions .rα and velocities .vα .
Solution The speed of the mass center deduced from the particles, p, is:
(p)
E E
vcm (t) =
. mα vα (t)/M ; M= mα
α α
The velocity .vα , Eq. 15.25, of the .α particle is .[rα (t + dt) . −rα (t − dt)] /2dt.
E
(p) mα [rα (t + dt) − rα (t − dt)]
.vcm (t) = α
2dt M
E E
m r
α α (t + dt) mα rα (t)
= α − α
2dt M 2dt M
E
The terms on the right correspond to the cm position, .rcm = α mα rα /M.
(p)
vcm (t) = [rcm (t + dt) − rcm (t − dt)]/(2dt)
.
(p)
The expression .vcm deduced from the set of positions .{rα } is equal to the
(p)
velocity .vcm directly deduced from .rcm , this is, .vcm = vcm .
15.7 Bond Constraints
In many instances, we are required to constrain the bond of a given pair of atoms.
A constraint of such a type demands a force to keep the atoms with the required
separation; however, the force also has effect on the speeds of the atoms. We follow
the algorithm of Andersen, known as the rattle model, to determine the equations
that impose bond constraints in molecular dynamics [6]. A holonomic constraint
between the ith and j th atoms starts from the distance condition .σij and the time
derivative .σ̇ij (t).
σij (t) = [ri (t) − rj (t)]2 − dij2

. (15.29)
σ̇ij (t) = 2[ri (t) − rj (t)] · [ṙi (t) − ṙj (t)]
The conditions .σij (t) = 0 and .σ̇ij (t) = 0 are used to impose the required
distance .dij on the ij -bond at every instant of time in the simulation. The Lagrangian
equation of motion that considers the net intermolecular force .fi exerted by the other
atoms of the molecule on the ith atom, plus the constraining force .gi , is:
mi ∂ 2 ri (t) / ∂t 2 = fi (t) + gi (t) ; fi (t) = −∇i U

. E' E' (15.30)
gi (t) = − j λij (t) ∇i σij (t) = − j λij (t) 2(ri − rj )
The force .fi is obtained from the interaction potential U . The constraints are
introduced using the Lagrange multipliers .λij , which depend on the time. The
summation considers the atoms forming constrained bonds with the ith atom. If
we have 3N equations of the type 15.30 and a total of n constraints, then there
are .3N + n variables of the type (.{ri (t)}, {λij (t)}) to be determined. It is with
the velocity-Verlet integrator approach, Eqs. 15.15, that we unfold the dynamics
of the particles satisfying the previous constraints. In this case, the velocity-Verlet
equations are:
⎡ ⎤
(δt)2 ⎣ E'
.ri (t + δt) = ri (t) + δt ṙi (t) + fi (t) − 2 λrij (t) rij (t)⎦
2mi
j
rij (t) = ri (t) − rj (t)
⎡
δt ⎣ E'
ṙi (t + δt) = ṙi (t) + fi (t) − 2 λrij (t) rij (t) +
2mi
j
⎤
E'
+ fi (t + δt) − 2 λvij (t + δt) rij (t + δt)⎦ (15.31)
j
Under Lagrangian theory, Sect. 6.8, the unknown factors .λrij and .λvij are used
to impose the first and second conditions given in Eqs. 15.29. The velocity-Verlet
integrator is particularly helpful in this task. The bond constraint, first expression of
Eq. 15.29, is maintained at the instants of time .t + δt and t. The last terms of the
first expression of Eqs. 15.31 are written as:
{ [ E ]}
. δt ṙi (t) + δt
2mi fi (t) − 2 'j λrij (t) rij (t) (15.32)
It is convenient to introduce the new variable .q and the new Lagrangian

multipliers .gij and .kij with the purpose of simplifying Eq. 15.32.
E'
qi = ṙi (t) + 2mi fi (t) − mi
δt 1
j gij rij (t)
. (15.33)
gij (t) = δt λrij (t) ; kij (t + δt) = δt λvij (t + δt)
15.7 Bond Constraints 375
The velocity-Verlet expressions are concisely written in terms of .qi , .gij , and .kij .
ri (t + δt) = ri (t) + δt qi
. E (15.34)
ṙi (t + δt) = qi (t) + 2m δt
i
fi (t + δt) − m1i 'j kij rij (t + δt)
A bond constraint is under examination by computing .b2ij and comparing with

.d
2 in the form shown in Eq. 15.29.
ij
. bij = [ri (t) + δt qi (t)] − [rj (t) + δt qj (t)] (15.35)
The positions satisfy the constraint when .b2ij = dij2 . If this last identity is not
satisfied, then a correction term proportional to .rij is introduced to reproduce the
required bond distance.
bij = ri (t) + δt qi (t) − δt mgi rij (t)

. [ ] (15.36)
− rj (t) + δt qj (t) + δt mgi rij (t)
The value of g is defined by the condition .b2ij = dij2 . The modified positions of
the particles are:
g
ri (t + δt) = ri (t) + δt qi (t) − δt mi rij (t)
.
g
(15.37)
rj (t + δt) = rj (t) + δt qj (t) + δt mj rij (t)
A similar procedure is followed for the atom speeds. The velocities in the
velocity-Verlet scheme satisfy the constraint when .[ri (t)−rj (t)]·[ṙi (t)−ṙj (t)] = 0,
Eq. 15.29. When this constraint is not satisfied, a correction term proportional to the
vector .rij is again introduced to reproduce the required constraint.
. ṙi = ṙi (t + δt) − k

mi rij (t + δt) ; ṙj = ṙj (t + δt) + k
mj rij (t + δt)
(15.38)
The value of k is defined by the condition .[ri − rj ] · [ṙi − ṙj ] = 0, where

the modified velocities of the particles at the instant of time .t + δt are the
expressions 15.38.
Problem 15.5 Determine the values of factors g and k of Eqs. 15.37

and 15.38 from the conditions .[ri (t + δt) − ri (t + δt)]2 = dij2 and .[ri −
rj ] · [ṙi − ṙj ] = 0, respectively.
Solution In order to determine g, we deal with the first condition and use the
modified Eqs. 15.37 to compute .dij2 .
[ ( )]2
g g
2
.dij = ri (t) + δt qi (t) − δt rij (t) − rj (t) + δt qj (t) + δt rij (t)
mi mj
The expression is simplified by using .bij of Eq. 15.35.

[ ]2
g 1 1 1
2
.dij = bij − δt rij (t) ; = +
μij μij mi mj
The first expression is solved for g by just considering linear terms in .δt:
[ ]
g = b2ij − dij2 / (2δt) μij / bij · rij
.
In regard to factor k, we deal with the condition .[ri − rj ] · [ṙi − ṙj ] = 0 and
use the modified velocities 15.38.
[ ]
k
.rij (t + δt) · ṙi (t + δt) − ṙj (t + δt) − rij (t + δt) = 0
μij
By solving for factor k, but identifying .r2ij = dij2 , we have:
k = [rij · (ṙi − ṙj )] / dij2 μij

.
where the previous variables are evaluated at the instant of time .t + δt.
References
1. M. Tuckerman, B.J. Berne, G.J. Martyna, Reversible multiple time scale molecular dynamics.
J. Chem. Phys. 97, 1990–2001 (1992)
2. H.F. Trotter, On the product of semi-groups of operators. Proc. Am. Soc. 10, 545–551 (1959)
3. S.J. Stuart, R. Zhou, B.J. Berne, Molecular dynamics with multiple time scales: the selection of
efficient reference system propagators. J. Chem. Phys. 105, 1426–1436 (1996)
4. E. Hernández-Huerta, R. Santamaria, T. Rocha-Rinza, Thermodynamics from Lagrangian
theory and its applications to nanosize sytems. Molec. Phys. 119(14), e1940333 (2021). https://
doi.org/10.1080/00268976.2021.1940333
References 377
5. G.J. Martyna, M.E. Tuckerman, D.J. Tobias, M.L. Klein, Explicit reversible integrators for
extended systems dynamics. Molec. Phys. 87, 1117–1157 (1996). S. Toxvaerd, Comment
on: reversible multiple time scale molecular dynamics. J. Chem. Phys. 99, 2277 (1993). M.
Tuckerman, B.J. Berne, G.J. Martyna, Reply to comment on: reversible multiple time scale
molecular dynamics. J. Chem. Phys. 99, 2278–2279 (1993)
6. H.C. Andersen, Rattle: a “velocity” version of the Shake algorithm for molecular dynamics
calculations. J. Comput. Phys. 52, 24–34 (1983)
Index
A Boltzmann statistics, 186

Abbreviated action function, 136 Bond-angle cosine potential, 303
Accelerated motion, 14 Bond-angle harmonic potential, 303
Acceleration, 13 Bond order, 320
Action function, 98 Bond strength, 320
Action-reaction law, 29 Bose-Einstein statistics, 186
Action at a distance, 289 Bosons, 186
Action wave function, 139 Boyle-Mariotte law, 64
Adiabatic functions, 284 Brachistochrone path, 99
Angular acceleration, 18 Bra-ket notation, 211
Angular momentum balance, 335 Bra-ket vector notation, 201
Angular momentum conservation, 34 Bragg’s law, 155
Angular momentum flux, 335 bra vector, 202
Angular velocity, 18 Breaking the equipartition principle, 336
Anharmonic effects, 301 Brownian particle, 50
Anomalous Zeeman effect, 195 Buckingham potential, 302
Anti-commutative operator, 140
Antisymmetric wave function, 229
Artificial-intelligence software, 305 C
Atom type, 299 Calculus of variations, 100
Aufbau principle, 180 Canonical commutation relations, 209
Autocorrelation function, 76 Canonical equations of motion, 129
Average energy, 171 Canonical transformations, 130
Average of physical variables, 50 Cause-effect relationship, 14, 23, 76, 100
Average potential, 251 Center of mass, 41
Axilrod-Teller potential, 325 Central force, 39
Azimuthal quantum number, 180 Central unit cell, 330
Centripetal acceleration, 18
Characteristic polynomial, 94
B Chemical potential, 282
Ball-and-stick approach, 273 Classical action function, 261
Basis set functions, 211 Classical electromagnetic theory, 154
Binomial coefficient, 232 Classical limit, 222, 255, 262
Blackbody, 160 Classical mechanics, 3
Black hole, 165 Classical molecular dynamics, 256, 295
Body inertia, 5 Classical time, 143
Body spin, 90 Classical time propagator, 360
Bohm quantum potential, 260 Classical wave equation, 235
Bohr magneton, 193 Closed system, 33
© The Editor(s) (if applicable) and The Author(s), under exclusive license to 379
Springer Nature Switzerland AG 2023
R. Santamaria, Molecular Dynamics, https://doi.org/10.1007/978-3-031-37042-7
380 Index
Closure relations, 202 Disorder, 50

Coherent light, 149 Dispersion energy, 308, 322
Cold body, 48 Distribution function, 72
Color-assigned temperature, 177 Doublet spin state, 184
Combination rules, 302, 318 Drude particles, 313
Complementary error function, 343 Duality of matter, 149
Complementary principle, 157 Dynamic equilibrium, 24
Compton’s experiment, 154
Concurrent force, 15
Conditionally convergent sum, 339 E
Configuration space, 102 Effective field, 251
Conjugate momentum, 128 Effective force constants, 323
Conjugate symmetry, 78 Effective Hamiltonian, 251
Conservation of kinetic energy, 46 Eikonal equation, 264
Conservative force, 36 Elastic collision, 48
Constant momentum, 24 Electric polarizability, 313
Constant of motion, 119, 142 Electromagnetic energy, 173
Convolution function, 78 Electromagnetic field, 10
Cooling temperature, 177 Electron angular momentum, 189
Coordinates generator, 121 Electron density, 276
Coordination number, 320 Electronic wave function, 248
Copenhagen interpretation, 288 Electron spin resonance, 188
Coriolis acceleration, 22 Electrostatic theorem, 277
Correlation function, 72 Embedded-atom approach, 317
Correlation matrix, 79 Emission factor, 177
Correspondence principle, 255 Energy balance, 48
Coulomb constant, 10 Energy conservation, 220
Cosine potential, 304 Energy dissipation, 115
Coulomb’s law, 10 Energy flow, 48
Coulomb potential, 307 Energy ground state, 225
Cross-correlation function, 76 Energy packet, 153
Cross-correlation theorem, 80 Energy preservation, 38
Cyclic variable, 105 Energy quantization, 159, 163
Ensemble average, 50
Environmental forces, 273
D Equation of continuity, 260
Davisson-Germer experiment, 155 Equation of motion, 28
de Broglie relation, 139, 236 Equation of state, 66
de Broglie thermal wavelength, 266 Equilibrium state, 33
de Broglie wavelength, 156 Equipartition of the energy, 57
Degenerate function, 211 Equivalence principle, 6
Degrees of freedom, 102 Ergodic hypothesis, 52
Density of states, 172 Ergodic system, 76
Diamagnetic material, 188, 192 Error function, 341
Dielectric, 355 Euler equation, 91
Differential forms, 112 Ewald summation, 337
Diffracted light, 149 Excitation energy, 165
Diffraction pattern, 155 Exclusion principle, 228
Diffusion constant, 282 Explicit solvent, 308
Diffusion equation, 279 External force, 42
Dihedral harmonic potential, 304 External forces, 273
Dirac-delta function, 267 External work, 49
Discrete time propagators, 359 Extremum trajectory, 99
Index 381
F H
Fast multipole algorithm, 352 Hamiltonian operator, 138
Fermi-Dirac statistics, 186 Hamilton-Jacobi equation, 135
Fermions, 186 Hamilton-Jacobi formulation, 134
Fictitious force, 12 Hamilton’s equations of motion, 129
First Fick’s law, 279 Hamilton’s principle, 99
First law of thermodynamics, 49 Hamilton’s characteristic function, 264
First Newton’s law, 24 Hamilton’s principal function, 264
Fluctuation-dissipation relation, 282 Hamiltonian energy, 119
Fluid viscosity, 281 Hamiltonian function, 117
First-order transition, 69 Hard-core van der Waals radius, 310, 333
Force, 262, 275 Harmonic potential, 300
Force concept, 23 Hawking radiation, 165
Force couple, 191 Heat, 48
Forcefield, 37, 315 Heat equation, 280
Forcefield approach, 297 Heat reservoir, 48
Force functions, 273 Heat source, 48
Force vector, 13 Heisenberg’s uncertainty principle, 168
Fourier transform, 80, 168, 217, 345 Hellmann-Feynman approach, 277
Fractional coordinates, 351 Hellmann-Feynman force, 276
Franck-Condon factor, 226 Hermitian operators, 209
Franck-Hertz experiment, 165 Hessian matrix, 277
Free energy, 282 High-intensity light, 152
Free fall principle, 6 Holonomic constraints, 101
Free particle, 24 Homogeneous medium, 11
Frequency, 19 Hot body, 48
Friction, 9 Hund’s rule, 187
Friction force, 115 Hyperfine coupling constant, 198
Fulling-Davies-Unruh effect, 165
Fundamental entities, 4
Fundamental translational vectors, 338 I
Ideal gas, 64
Ideal gas law, 67
G Ignorable variable, 105
Galilean transformations, 11 Image cell, 331
Gauge invariance, 110 Imaginary cell, 332
Gaussian standard deviation, 348 Implicit solvent, 308
General Hellmann-Feynman theorem, 275 Impulse, 34
Generalized coordinates, 76, 98 Induced dipole, 312
Generalized forces, 114 Inelastic collision, 48
Generalized momentum, 105, 128 Inertial mass, 5
Generalized potential function, 114 Inertial reference frame, 11
Generalized velocities, 98 Inertia moment products, 87
Generating function, 130 Instantaneous power, 35
G factor, 181 Integral of motion, 119
Good quantum numbers, 228 Interatomic distance fluctuations, 70
Gravitational mass, 5 Interatomic potential, 28
Gray body, 160 Interference, 149
Green-Kubo approach, 77 Intermolecular interactions, 298, 307
Gupta interaction potential, 319 Internal energy, 46
Gyro-magnetic ratio, 182, 190 Internal energy conservation, 46
382 Index
Intramolecular interactions, 298 Magnetic quantum number, 184

Invariance of the laws of physics, 25 Magnetic quantum number of spin, 184
Invariance of the physical laws, 25 Many-body potentials, 297, 317
Inversion potential, 305 Many-body problem, 295
Ionization potential, 166 Many-body system, 41
Irradiance, 175 Mass distribution, 86
Irreversible process, 9 Matrix diagonalization, 93
Isolated system, 33 Matrix eigenvalues, 94
Isotropic medium, 11 Matrix eigenvectors, 93
Matter waves, 235
Maximum intensity, 176
K Maxwell-Boltzmann expression, 282
Ket vector, 202 Maxwell-Boltzmann function, 56
Kinematic friction, 115 Melting temperature, 70
Kinetic energies, 245 Metastable state, 55
Kinetic energy, 8, 36 Microstates, 55
Kinetic energy components, 92 Minimum image convention, 332
Kinetic expression of the temperature, 58 Minors expansion, 239
Mixed quantum-classical dynamics, 256
Mixed quantum-classical molecular dynamics,
L 271
Lagrange equations of motion, 105 Model interaction potentials, 296
Lagrange multipliers, 113 Model interatomic potentials, 295
Lagrangian, 98 Molecular mechanics, 273
Lagrangian function, 105 Molecular size variable, 56
Lagrangian theory, 101 Moment of inertia, 86
Lambert’s cosine law, 176 Monochromatic wave, 151
Landau-Zener transition, 285 Monopole interactions, 308
Lande factor, 195 Moore’s law, 245
Laplace expansion approach, 237 Morse potential, 301
Larmor frequency, 195 Motion energy, 36
Law of inertia, 25 Motion period, 19
Leapfrog time integrator, 368 Muon magnetic moment, 200
Lennard-Jones potential, 301
Lie algebra, 143
Lindemann index, 70 N
Linear momentum, 24 Negative charge, 9
Linear momentum conservation, 25 Nematic order, 72
Line integral, 35 Neutral particle, 10
Line of action, 15 Neutrino magnetic moment, 199
Liouville operator, 141 Neutron magnetic moment, 199
List of atom neighbors, 310 Newtonian dynamics, 3
Localized pulse, 151 Newtonian momentum, 128
Long-range interactions, 307 Newtonian theory, 23
Lorentz force, 110 Newtonian time, 13
Lorentz transformations, 11 Newton unit, 28
Low-intensity light, 152 Noether’s theorem, 121
Non-additive term, 325
Non-bonded forces, 308
M Non-concurrent forces, 15
Macroscopic stationary state, 51 Non-conservative forces, 37, 48, 114
Macroscopic variables, 33 Non-equilibrium state, 55
Magnetic g factor, 190 Non-holonomic constraints, 102
Magnetic moment, 181 Non-inertial reference frame, 20
Index 383
Non-local interactions, 289 Polar bonds, 307

Normalized Gaussian distribution, 168 Polarized state, 10
Nuclear magnetic resonance, 188 Positive charge, 9
Nuclear wave function, 248 Potential energy, 8, 36
Numerical time integrators, 359 Power spectrum, 81, 355
Numeric potentials, 302 Preservation of total energy, 47
Nyquist-Shannon theorem, 62 Pressure, 63
Principal axes of rotation, 88
Principal moments of inertia, 88
O Principal quantum number, 180
Observable, 5 Principal spin quantum number, 180
Offset speeds, 57 Principle of least action, 99
One-body distribution function, 72 Probability density, 149, 259
Open system, 33 Probability distribution function, 169
Orbital equation, 40 Probability flux, 259
Orientational order parameter, 71 Probability function, 55
Projection operator, 213
Proper energy, 45
P
Paramagnetic materials, 188
Parameter file, 316 Q
Parameter fitting process, 316 Q balance factor, 47
Parametric variable, 271 Quanta, 153
Particle correlation, 24 Quantum, 163
Particle current, 278 Quantum force, 289
Particle density, 267 Quantum mechanics, 205
Particle diffusion, 278 Quantum molecular dynamics, 271
Particle flux, 278 Quantum potential, 288
Particle mass, 5 Quarks, 199
Particle mobility, 281
Particle motion, 11
Particles of light, 5
Particles of matter, 5 R
Particle speed, 13 Rayleigh dissipative forces, 115
Particle spin, 179 Rattle model, 373
Particle trajectory, 11, 101 Rayleigh-Jeans law, 162
Paschen-Back effect, 198 Rayleigh-Jeans radiation law, 176
Pauli exclusion principle, 187, 228, 308 Reaction field approach, 355
Periodic boundary conditions, 330 Reciprocal space, 337, 344
Periodic motion, 19 Reciprocal vectors, 345
Periodic potential, 306 Reduced distribution function, 72
Permittivity of space, 10 Reduced mass, 30
Perturbation potential, 320 Relative motion, 19
Phase space, 102 Relative permittivity, 307
Photo-electrons, 151 Resonant vibrational modes, 161
Photons, 5 RESPA approach, 365
Planck’s constant, 152 Retardation effect, 352
Planck’s energy equation, 159 Retarded motion, 14
Planck’s radiation intensity law, 170 Reversible process, 9
Planck’s radiation law, 176 Rheonomous constraints, 102
Plane-bending cosine potential, 304 Rigid body, 85
Poisson brackets, 140 Root mean square deviation, 71
Poisson equation, 340 Rotation axes, 87
384 Index
S System of coordinates, 11
Scalar triple product, 192 System replicas, 50
Schrödinger equation, 214
Scleronomous constraints, 102
Second Newton’s law, 27 T
Second-order transition, 69 Temperature, 57
Second Fick’s law, 279 Temperature spectrum, 62
Selection rules, 227 Temporal equilibrium, 56
Self-consistent field equations, 252 Temporal stability, 56
Self-consistent procedure, 252 Tersoff potential function, 320
Self correlation, 76 Thermal de Broglie wavelength, 64
Semi-rigid boundaries, 328 Thermal energy, 9
Singlet spin state, 184 Thermal equilibrium, 58
Slater determinant, 230, 237 Thermal radiation, 160
Smooth particle mesh Ewald method, 350 Thermionic emission, 153
Space, 4 Thermodynamically stationary states, 33
Space quantization, 180 Thermodynamic equilibrium, 51
Spacetime, 4 Thermodynamic limit, 51
Spatial correlation function, 75 Thermodynamics, 32
Spin-down, 183, 201 Thermostat, 60
Spin multiplicity, 184 Third Newton’s law, 29, 300
Spinor, 202 Time, 4
Spin orbitals, 229 Time average, 50
Spin projection quantum number, 184 Time-dependent wave equation, 138, 214
Spin states, 201 Time correlation function, 75
Spin up, 183, 201 Time evolution, 11
Spooky action, 289 Time generator, 121
Spurious forces, 275 Time-independent wave equation, 219
Standard deviation, 168 Time integration step, 361
State function, 50, 208 Time propagator, 145
Static equilibrium, 26 Time reversibility, 30, 132, 363, 366
Stationary action, 99 Topology file, 316
Stationary parameters, 33 Torque, 15
Stationary state, 220 Total energy, 47
Stationary waves, 171 Total wave function, 150
Statistical average, 50 Trajectory average, 50
Statistical forces, 50 Transparent body, 160
Statistical mechanics, 32, 50 Trapezoidal integration method, 53
Steered molecular dynamics, 315 Trotter expansion, 361
Stefan-Boltzmann law, 164, 176 Two-body distribution function, 72
Stochastic dynamics, 282 Two-body potentials, 73
Stockmayer potential, 314
Stokes dragging-force law, 283
Stress tensor, 351 U
Structural similarity, 71 Ultraviolet catastrophe, 163
Structure factor, 348 Uncertainty principle, 157
Sub-shell multiplicity, 180 Unconstrained Lagrangian, 113
Super-atom approach, 300 Uniform accelerated circular motion, 19
Superposition of waves, 168 Uniform acceleration, 14
Superposition principle, 28 Uniform angular motion, 20
Sutton-Chen function, 318 Uniform circular motion, 19
Symbolic-algebra software, 305 Uniform translational motion, 20
Symmetrization of quantum operators, 209 United atom, 300
Index 385
V Wave number, 168, 236

Vacuum, 170 Wave packet, 167
van der Waals radii, 309 Wave-particle duality, 157
Variance, 79 Wave pulse, 168
Variational principle, 234 Waves superposition, 237
Vector triple product, 86 Wave theory of matter, 156
Velocity-Verlet time integrator, 365 Weight, 5
Velocity correlation function, 355 White body, 160
Velocity vector, 13 White noise, 151
Verlet time integrator, 371 Wick rotation, 145
Vibrational frequencies, 278 Wien’s displacement law, 162, 177
Vibrational modes, 278 Work, 35
Vibrational spectrum, 278 Work function, 152
Virial equation of state, 74
Virial theorem, 65
Virtual particles, 170 Y
Viscous term, 115 Young’s experiment, 149
W
Wave equation, 161 Z
Wave equation of matter waves, 237 Zeeman effect, 194
Wave function, 138, 207 Zero-point vibration energy, 308
Wave mechanics, 206 Zero law of thermodynamics, 48, 57
Wave modulation, 168 Zero-point energy, 170
Wave nature of light, 149 Zero potential energy, 9, 36

Molecular Dynamics

Uploaded by

Copyright:

Available Formats

Molecular Dynamics

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Molecular Dynamics

Uploaded by

Copyright:

Available Formats

Ruben Santamaria

ISBN 978-3-031-37041-0 ISBN 978-3-031-37042-7 (eBook)

Paper in this product is recyclable.

Mexico City, Mexico Ruben Santamaria

• Basics of Classical Mechanics

Basics of Classical Mechanics

Basics of Quantum Mechanics

This part of the book is an introduction to quantum theory. It explains the

First-Principles Molecular Dynamics

Classical Molecular Dynamics

determining the dynamical and thermodynamic properties of compounds immersed

Time Evolution Operators

Part I Basics of Classical Mechanics

3.8 Breaking the Ergodic Hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

6.20 Contact Between Quantum and Classical Mechanics . . . . . . . . . . . . . 138

Part II Basics of Quantum Mechanics

A.1.4 Zeeman Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193

Part III First-Principles Molecular Dynamics

12.9 Diffusion Equation and Particle Flux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279

Part IV Classical Molecular Dynamics

Part V Time Evolution Operators

Classical Mechanics is the theoretical framework of physics which describes the

1.1 Newtonian Dynamics

Classical mechanics establishes the underlying concepts, mechanical elements,

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 3

1.2 Space and Time

The sub-atomic experiments on particle production exemplify the transformation

Fig. 1.1 According to the

. a = (mg /mi ) g (1.3)

Solution The gravitational force of .Mplanet exerted on body 1 is .f1 =

. amu = 1.660, 538, 92 × 10−27 Kg

3 A quantity of 46.6 mg of mass is required to approximately produce a kiloton of T N T . The

Statement 1.3 The energy is always preserved in isolated systems, and it is

The conservation of a physical observable with time is associated with a symmetri-

speed is frequently considered as the reference point of zero-potential energy. There

1.5 Electric Charge

1.6 Reference System of Coordinates

The description of a set of particles requires a system of coordinates. The purpose

1.7 Newtonian Time

1.8 Linear Motion

. v(t) = lim [r(t + At) − r(t)]/At = dr(t)/dt = ṙ(t) (1.6)

. a(t) = dv(t)/dt = v̇(t) = r̈(t) (1.7)

Acceleration owns units of .distance/time2 , or .velocity/time. A change in the

. f(t) = m dv(t)/dt (1.8)

t is given by the sum of the individual displacements.

A similar expression is displayed Efor the continuous (or infinitesimallyf small)

An alternative expression relating the variables x, v, and a is obtained from the

acceleration a, which in turn depends on the position x.

. x(ti ) = x(ti−1 ) + v(ti−1 )[ti − ti−1 ] + 1

The acceleration produced by an external force acting on the particle is constant

form to describe the torque given by:

. τ = (ry fz − rz fy )ux + (rz fx − rx fz )uy + (rx fy − ry fx )uz (1.15)

The torque is important in investigating the rotation of a particle around an axis

Problem 1.2 Consider a number of individual forces .fi exerted on different

1.9 Angular Motion

When a particle travels along a curvilinear trajectory, the kinematic equations of

Fig. 1.3 A particle travels

The acceleration is a vector usually described in terms of the Cartesian components

. ut = ux cos φ + uy sin φ ; ut · un = (ux cos φ + uy sin φ) · (ux a + uy b) = 0

. un = −ux sin φ + uy cos φ