Q. Computacional Computational Chemistry Reviews Vol 9 (Leszczynski)

Computational
Chemistry
Reviews of Current Trends
Volume 9
Computational Chemistry: Reviews of Current Trends
Editor-in-Charge: Jerzy Leszczynski,

Dept. of Chemistry, Jackson State University, USA
Published
Vol. 1: Computational Chemistry: Reviews of Current Trends
Edited by Jerzy Leszczynski







Computational
Chemistry
Reviews of Current Trends
Volume 9
editor
Jerzy Leszczynski
Department of Chemistry
Jackson State University
USA
\fp World Scientific

NEW JERSEY • LONDON • SINGAPORE • BEIJING • SHANGHAI • HONGKONG • TAIPEI • CHENNAI
Published by
World Scientific Publishing Co. Pte. Ltd.
5 Toh Tuck Link, Singapore 596224
USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601
UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE
British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library.
COMPUTATIONAL CHEMISTRY: REVIEWS OF CURRENT TRENDS

Volume 9
Copyright © 2005 by World Scientific Publishing Co. Pte. Ltd.
All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means,
electronic or mechanical, including photocopying, recording or any information storage and retrieval
system now known or to be invented, without written permission from the Publisher.
For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance
Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy
is not required from the publisher.
ISBN 981-256-097-1
Printed in Singapore by World Scientific Printers (S) Pte Ltd

PREFACE
There is no doubt that many of the readers of this book series wonder
how the research/computational facilities will look like in the near fu-
ture. It seems that the efficiency and compactness of the silicon-based
computers are almost exhausted, and for significant progress one has to
look for alternative solutions. A very promising future for computing is
based on molecular computers. Among the molecular elements required
for such computers, the developments rely on the existence of nanostorage
devices.
A recent Science article (Z. M. Liu, A. A. Yasseri, J. S. Lindsey, and
D. F. Bocian; Science, 302, 1543, 2003) reveals silicon-tethered porphyrins
that are resistant to heat (up to 400° C) and harsh conditions. Since por-
phyrins possess well known data storage characteristics, the unique prop-
erties of these new derivatives reveal the possibility for their commercial
production. It is assumed that porphyrin-based memory chips can be as-
sembled using facilities devoted to the traditional silicon devices, opening
the way toward nano components of molecular computers.
The research in this area is extensive. To assist the readers, we have
decided to collect more chapters for this book that are devoted to the re-
views of current advances in nano science. As a result, this volume includes
three chapters related to this important research area.
The first chapter, written by J. J. Palacios, A. J. Perez-Jimenez,
E. Louis, E. San Fabian, J. A. Verges, and Y. Garcia, reviews fundamental
issues underlying first-principle quantum theory in atomic and molecular
systems. The authors have recently developed a new approach called Gaus-
sian Embedded Cluster Method that is based on the GAUSSIAN98 and
GAUSSIAN03 codes. The review provides numerous examples that illus-
trate the applicability and reliability of this method. This is supported by
the study and interpretation of various experiments in the field of molecular
electronics.
Apparently, an ultimate goal of nano sciences is the development of
tools for spatial and chemical control of single molecules. Among the ex-
perimental techniques that have emerged in recent years are single molecule
fluorescence, optical and magnetic tweezers, and atomic force spectroscopy.
All of these techniques allow for mechanical manipulations of single
molecules. The proper comprehension of the experimental results requires
adequate theory. W. Nowak and P. E. Marszalek present an overview of
V
vi Preface
recent advances and trends in the applications of computer simulations

aimed at understanding atomic force microscopic experiments. Those vari-
ants of classical molecular dynamics (MD) simulations which are particu-
larly helpful in the interpretation of experimental data obtained for single
biopolymer molecules, such as steered MD or biased MD, are reviewed and
discussed in detail.
In the third chapter, J. Seminario, P. Derosa, L. Cordova, and B. Bozard
discuss the application of combined molecular dynamic simulations based
on ab initio force fields and signal processing techniques to analyze the dy-
namic properties of nanocells. The analysis includes several characteristic
modes at different temperatures. The authors consider the applications of
such techniques for the characterization of new, molecular electronics.
The prediction of the excited state properties of molecular systems
are among the most challenging tasks of computational chemistry. They
are assisted by the development of new methods designed for such tasks.
The most accurate approaches are based on the coupled cluster theory.
D. Mukherjee and his group (S. Chattopadhyay, D. Pahari, and U. S. Ma-
hapatra) review formulations of linear response theory for excited state po-
tential energy surfaces which are based on multi-reference coupled electron-
pair approximation categories of methods. This review includes a thorough
discussion of the genesis of the developed theory. In addition, applications
to the potential energy surfaces of low-lying excited states of the model P4
and Li2 molecules are also provided.
Recently progress has been achieved in the theoretical prediction of the
magnetic exchange coupling constants in compounds of rare earth transi-
tion metals. In the fifth chapter, M. Atanasov, C. Daul, and H. U. Giidel
describe a method developed for such tasks. After a comprehensive dis-
cussion of the calculation procedure, the authors present examples of its
application to various dimers of rare earth metals.
My recent web search of the phrase "hydrogen bonding" resulted in over
50,000 entries. This demonstrates the popularity of this topic which was
formally recognized more than 80 years ago. Over the years new and intrigu-
ing aspects of this phenomenon have been studied. Such studies reveal new
classes of hydrogen bonds. One of such new categories, dihydrogen bonds
(DHB), are discussed in the last chapter of this volume. S. J. Grabowski
and J. Leszczynski present the examples of such interactions existing in
crystal structures. These interactions are compared with the conventional
hydrogen bonds. A comparison shows that DHBs are often similar to typ-
ical H-bonds. The discussion includes the results of ab initio calculations,
Preface vii
decomposition energy, topological parameters, and experimental neutron

diffraction results.
I would like to thank all authors for the excellent contributions and
fine collaborations. The very efficient technical assistance of Dr. Manoj
K. Shukla in putting together this volume is greatly appreciated. As al-
ways, your feedback is very important to me, please feel free to e-mail your
suggestions to jerzy@ccmsi.us.
Jerzy Leszczynski
Jackson, MS
December 2003
CONTENTS
Preface v
1. Molecular Electronics with Gaussian98/03 1

J. J. Palacios, A. J. Perez-Jimenez, E. Louis,
E. SanFabidn, J. A. Verges and Y. Garcia
2. Molecular Dynamics Simulations of Single Molecule 47

Atomic Force Microscope Experiments
W. Nowak and P. E. Marszalek
3. Molecular Dynamics Simulations of a Molecular 85

Electronics Device: The NanoCell
J. Seminario, P. Derosa, L. Cordova and B. Bozard
4. Computation of Excited State Potential Energy Surfaces 121

via Linear Response Theories Based on State Specific
Multi-Reference Coupled Electron-Pair Approximation
Like Methods
S. Chattopadhyay, D. Pahari, U. Mahapatra,
D. Mukherjee
5. Modelling of Anisotropic Exchange Coupling in 153

Rare-Earth-Transition-Metal Pairs: Applications to
Yb 3 +-Mn 2+ and Yb 3 +-Cr 3 + Halide Clusters and
Implications to the Light Up-Conversion
M. Atanasov, C. Daul and H. U. Giidel
6. Is a Dihydrogen Bond a Unique Phenomenon? 195

S. J. Grabowski and J. Leszczynski
Index 237
Content Index 243
ix
Chapter 1: Molecular Electronics with Gaussian98/03
J. J. Palacios ab , A. J. Perez-Jimenez0, E. Louisa'b,

E. SanFabian"'0, J. A. Verges"" and Y. Garcia"
*Departamento de Fisica Aplicada, Universidad de Alicante

San Vicente del Raspeig, Alicante 03690, Spain
b Unidad Asociada del Consejo Superior
de Investigations Cientificas, Universidad de Alicante

c Departamento de Quimica-Fisica, Universidad de Alicante

d Departamento de Teoria de la Materia Condensada
Institute de Ciencia de Materiales de Madrid (CSIC)

Cantoblanco, Madrid 28049, Spain
Abstract
In this review we discuss the fundamental issues underlying first-principles
quantum transport theory in molecular- and atomic-scale systems, making
emphasis on the actual numerical implementation of them. For this pur-
pose we focus on the ab initio method named Gaussian Embedded Cluster
Method, recently developed by the authors, which is based on the popular
quantum chemistry code GAUSSIAN98/03. Various examples that illustrate
the applicability and reliability of this method in the study and interpreta-
tion of a large range of experiments in the field of molecular electronics are
also presented.
1
2 J. J. Palacios et al.
1 The basics
For the past forty years, computers and, more generally, electronic devices
have grown more powerful as their basic subunits have shrunk. Investigators
in the field of next-generation electronics forsee that during the next 10
to 15 years, as the smallest features of mass-produced transistors shrink
below 100 nanometers, the devices will become more difficult and costly
to fabricate. In addition, they may no longer function effectively since the
laws of quantum mechanics may prevent transistors from working under
the same principles as today's. In order to continue the miniaturization of
circuit elements down to the nanometer scale new paradigms need to be
proposed. For instance, unlike today's field effect transistors, which operate
based on the movement of masses of electrons in bulk matter, the new devices
should take advantage of quantum mechanical phenomena that emerge on
the nanometer scale such as the quantum coherence and the discreteness
of the electron charge. There are, essentially, two roads leading towards
nanometer-scale ultra-dense circuits:
• Solid-state quantum-effect devices (silicon-based devices).
• Molecular electronic devices (carbon-based devices).
It is not the purpose of this review to analyze which alternative is more
likely to succeed in the long run. While the first alternative seems the
logical continuation of present technologies, we believe, in our modest and
biased view, that molecular electronics may have a better chance. Even
if this is not the case, basic research in molecular electronics may con-
tribute to advances in multidisciplinary areas such as biotechnology and
medicine. Behind the words "molecular electronics" lies the promising idea
that functional units for electronics can be built out of very stable and well-
characterized carbon-based molecules such as fullerenes, carbon nanotubes,
or polyphenylene chains, to name a few[l]. The advantage of using carbon-
based molecules resides in that they are fabricated identical by nature, can
be obtained in any desired quantity, and can be easily functionalized to meet
any required property by means of standard and cheap chemistry techniques.
Regardless of the final road taken by electronics, there is little doubt that
the functionality of future devices will rely on the electrical and mechanical
properties of nanoscopic regions composed of a number of atoms that could
range from several thousands down to a single one. Hereby the necessity of
quantum chemistry or first-principles computational support to which these
notes are devoted.
Molecular Electronics with Gaussian98/03 3
2 Electronic transport at the nanoscale: A quick

view of Landauer formalism
The first proposal of using molecules for electronic applications was put for-
ward by Aviram and Ratner in 1974[2]. Since then various ideas to use
molecules for electronic applications have been proposed [1]. How to put
into practice these ideas, however, remains the main issue. It is impor-
tant to realize that the coexistence of molecules and metallic electrodes will
probably form the basis of future molecular devices since it is very unlikely
that electronics will be entirely based on carbon. Keeping this in mind, the
simplest molecular electronic unit that can be envisioned is a two-terminal
device consisting of two large metallic electrodes, several nanometers apart,
joined by a molecule or molecules anchored to them. This proposal was
actually realized for the first time in the laboratory by Reed's group us-
ing a Mechanically Controlable Break Junction (MCBJ)[3], although early
measurements go back to the work of Mann and Kuhn[4]. This system is
generically known as a molecular bridge or molecular junction and plays a
central role in molecular electronics. Conventional lithography, on the other
hand, had already been successful in contacting large molecules like car-
bon nanutubes[5]. Before Reed's work, various groups had already managed
to make nanoscale junctions connecting the two metallic electrodes of the
MCBJ by just a chain or even a single atom of the same metallic element [6].
These systems are known as atomic contacts or metallic nanocontacts and,
although they are not expected to be of any practical application, constitute
an excellent test bed to learn about the world of electrical transport at the
atomic scale. It is fair to say that other groups had already been working
on molecular and atomic junctions since the early nineties from a different
perpective using the scanning tunneling microscope (STM)[6, 7, 8].
The actual fabrication of these electronic devices at the molecular and
atomic scale poses new challenges for theoretical physicists and chemists
who must develop new theories and, most importantly, implement existing
ones to address the present and upcoming questions. The basics to calcu-
late the zero-bias, zero-temperature conductance Q of a molecular bridge or
metallic nanocontact were established by Landauer long before the concept
of molecular electronics was commonplace. In Landauer's formalism Q is
simply related to the quantum mechanical transmission probability T for
electrons at the Fermi level to go from one electrode to the other [9] (we will
assume spin degeneracy throughout):
g = goT(EF) = —T(EF) = -^YlT^)- (!)

i
The fact that the zero-bias conductance is proportional to an intrinsically
coherent quantity such as the elastic transmission probability should not
Figure 1: Schematic drawing of the scattering of conduction channels in a

constriction in a semiclassical picture. Quantum conduction channels are
typically associated to specific quantum numbers of the transversal part of
the wavefunctions occupied at the Fermi level when the Schrodinger equation
is separable in bulk.
come to us as a surprise at zero temperature since inelastic scattering events

due to vibrations or many-body interactions are strictly prohibited. The
fact that the concept of elastic transmission probability or elastic scattering
can still be determinant even at finite bias and temperature is what makes
Landauer's formula extremely useful. This is esentially due to the fact that
inelastic scattering lengths are typically much larger than the extent of the
nanoscopic region that determines the conductance of the whole device.
The wide success of Landauer's formula lies in its simplicity. However,
this simplicity is misleading and still is the source of some debate (for a
recent discussion see Ref. [10]). Landauer's formula has been discussed on
numerous occasions[9] and we do not intend to do it again in these notes.
Instead, we will simply give some plausible arguments in favor of it. The
sum in the last expression of Eq. 1 runs over all possible conduction channels
deep in the bulk of the metallic electrodes and not only over the conducting
channels in the actual nanoscopic region of interest. It is customary to
associate quantum conduction channels to specific quantum numbers m, n
of the transversal part of the bulk wavefunctions occupied at the Fermi
level when the Schrodinger equation is separable in bulk; in other words,
\&(f) = ijjm(x)<pn{y)e%kz. Even if there is no elastic scattering affecting the
transmission of electrons in a finite number of these channels, the resistance
Figure 2: Schematic drawing of the alignement between the molecular or-

bitals of the molecule and the Fermi energy of the metallic electrodes for
two extreme cases as discussed in the text.
R = A is still finite due to the electrons in all the other conduction channels
that fully backscatter (see Fig. 1). It is also implicit in Landauer's formalism
that the Joule heating associated to this resistance takes place away from the
nanoscopic region. The factor QQ in Eq. (1) can be easily understood making
use of the Heisenberg's uncertainty principle. We can define the current per
channel as I = e/At, where At is a measure of the time an electron takes
to cross the sample and e is the electron charge. On the other hand we can
define a typical energy scale for the electron, AE, as V = AE/e, where V is
the bias voltage. Knowing that Q = I/V, Heisenberg's uncertainty principle
leads us to Q = e2/(AtAE) « e2/h. Unfortunately, the simple expression in
Eq. (1) can seldom be used as it is since channels cannot always be defined
deep in the electrodes[ll, 12]. This is only the case when translational
symmetry is present and the three-dimensional Schrodinger equation can
be separated into transversal and longitudinal solutions (actual electrodes
do not fall into this category). Instead it is convenient to use a different,
but completely equivalent expression (see below) where the sum does not
run over conduction channels, but over localized basis functions. Still, the
concept of channel can be rescued as will become clear after a few examples
have been seen. Finally, we refer the reader to Refe. [9, 13] for other
issues regarding inelastic effects and many-body corrections to Landauer's
formalism.
The overall transmission probability T(Ep) can be easily estimated on
generic considerations for metallic nanocontacts[14, 15] either by estimating
the section of the nanocontact or by counting the valence orbitals of the
metallic element. It is much more difficult, however, to do so for molecu-
lar bridges. The main reason is the following: One does not know a priori
where the electrode Fermi energy E-p lies with respect to the molecular lev-
6 J. J. Palados et al.
els when metal and molecule are brought into contact. In a simple-minded
picture the positioning of the molecular levels with respect to the electrode
Ep will depend on the relative values of the work function of the metal with
respect to the electron affinity and the ionization potential of the molecule.
In simpler words, if Ep lies above the lowest unoccupied molecular orbital
(LUMO), a charge transfer is expected from the metal to the molecule until
the charge dipole aligns Ep with the LUMO [see Fig. 2(b)]. Thus, elec-
trons will flow through the LUMO. If, on the other hand, E-p lies below the
highest ocuppied molecular orbital (HOMO), charge will transfer from the
molecule to the metal until E-p and the HOMO line up [see Fig. 2(a)]. If
this is the case, the transport will reflect the character of the HOMO. Un-
fortunately, the coupling or hybridization of the molecular levels with the
free-electron levels in the electrodes compounds significantly this picture. In
general, Ep will lie in the middle of the HOMO-LUMO gap (where metal-
induced gap states are likely to appear) and the current will have a weak
tunnelling component. In summary, the conductance is strongly dependent
on the particular molecule, the detailed atomic arrangement of the electrodes
where the molecule binds, and the chemical nature of the various elements
at play. In order to give an answer to these questions one has to abandon
parametrized models and rely on first-principles or ab initio calculations.
3 Numerical implementation of Landauer formula

3.1 Basic caveats
Implementing Landauer formalism requires to know the electronic structure
of a finite region embedded in an infinite system with no particular symmetry
or periodicity and, if a finite bias is applied, out of equilibrium. This is
a non-trivial problem to which a lot of effort has been recently devoted.
However, the difficulties extend far beyond the usual technical problems in
first-principles calculations and, at present, there is no agreement on which
the most reliable implementation is. One has to take into account various
independent aspects that can affect the final outcome. We enumerate them
below in, to our judgement, increasing order of importance:
• The self-consistent field approximation. If one sticks to a stan-

dard density functional theory (DFT) approach, this will be, in most
cases, the ingredient that introduces less uncertainty in the final re-
sults.
• The basis set. Using finite basis sets always gives a certain degree of
uncertainty which can be easily overcome by testing the results against
larger basis sets (whenever feasible).
• The atomic structure. Calculating the possible atomic arrange-

ments in the electrodes and the possible ways the molecule can bind
to them is known to be a major problem in itself. Many standard
techniques already implemented in most popular quantum chemistry
codes can be applied to this purpose and will not be given further con-
sideration in these notes. Unless especified otherwise, we will assume
relaxed atomic structures throughout.
• The model for the bulk electrodes. This is related to the previous
point and, in our opinion, is an important factor in the uncertainty of
any claimed result. There is no obvious way of taking into considera-
tion the bulk of the electrodes in the calculation.
• The experimental evidence. Finally, it is crucial to know the de-

tails on how the experiment has been done to reduce the phase space
of possibilities to consider in the theoretical modelling.
Possibly the main advantage of the numerical implementation that we

review here with respect to similar proposals that have appeared in the
literature[16, 17, 18, 19, 20] is the use of the standard quantum chem-
istry code GAUSSIAN98/03[21] (see Refs. [22, 23, 24, 25]). The GAUS-
SIAN98/03 code provides a versatile method to perfom first-principles or
ab initio calculations of clusters, incorporating the major advancements in
the field in terms of functionals, basis sets, pseudopotentials, etc.. Other
groups also make use of GAUSSIAN98/03 at the core of their numerical
implementations[26, 27, 28], but these differ in a number of technical details
that might deserve attention when a quantitative agreement with experi-
ments is required. We will refer to our method as the Gaussian Embedded
Cluster Method (GECM) from now on.
3.2 Methodology: Zero bias, zero temperature

Our procedure goes as follows. A standard electronic structure calcula-
tion of the region that includes the molecule and a significant part of the
electrodes is performed [see Fig. 3(a)]. Including part of the electrodes
in the self-consistent calculation is essential since, as explained above, this
will guarantee the proper alignement between the Fermi level of the elec-
trodes and the molecular levels. Here we are faced with the first serious
source of uncertainty in the results (not only in ours, by the way). We are
forced to describe at the atomic level the structure of the molecular junction
or nanocontact. There is, however, a statistical uncertainty in the atomic
structure due the many uncontroled factors inherent to most experiments
that is impossible to account for by a single model. A statistical analysis
is, on the other hand, computationally forbidden. The calculation can be
performed within any self-consistent field approach: Hartree-Fock or DFT
Figure 3: (a) A molecular bridge model where a short capped nanotube is

contacted by two (001) parallel surfaces. The standard self-consistent pro-
cedure performed initially by GAUSSIAN98/03 is also shown schematically
on the right, (b) The same system where the first phantom atoms of the
Bethe lattices are also depicted. The self-consistent procedure when the
Bethe lattices are included is also shown schematically on top. (c) A finite
section of a Bethe lattice model in two dimensions with coordination three.
in any of its multiple approximations. As far as transport is concerned,

the self-consistent hamiltonian matrix H (or Fock matrix F) of this cen-
tral cluster or supermolecule contains the relevant information. We will
now make use of the Green's function formalism. The retarded(advanced)
Green's functions associated with F are defined in the standard way[29]:
[(£±i<5)I-F]G(±)=I. (2)
In this expression I is the identity matrix and S is an infinitesimal quantity
(in practice 10~10 will do it). One of the many advantages of the use of Green
functions is that one can incorporate the rest of the infinite electrodes in a
very elegant and simple way:
[(SI - F - S ( ± ) (£)] G (±) (E)=I, (3)

where
SW(£) = SW(E) + 4 ± ) ( E ) , (4)
and S R ( S ) [ S L ( £ 1 ) ] denotes a self-energy matrix that accounts for the part
of the right (left) semi-infinite electrode which has not been included in the
self-consitent calculation. This is another source of uncertainty in the calcu-

lations. A correct choice of this quantity or, in other words, an appropriate
model for the bulk of the electrodes can avoid spurious results and mislead-
ing conclusions. The first problem one encounters is that the self-energy
matrices can only be explicitly calculated in ideal situations. This is why
a number of groups in the field have chosen to model the bulk electrodes
as perfect semi-infinite crystals[30, 18, 26]. In our opinion this choice can
only be justified for mathematical convenience. It is, however, the least
appropriate one since real electrodes are not monocrystals, but polycrys-
talline structures with no expected long-range order beyond a few lattice
constants, particularly near the contact where the atomic structure cannot
possibly be perfect. In these models the perfect crystal band structure neces-
sarily reflects in the final conductance of a molecule and this is not desirable.
Alternatively, jellium models are used in methods based on a description in
terms of scattering states[16, 20,17]. This is also a choice of convenience not
free from controversy [31, 32]. We choose to describe the bulk electrode with
a Bethe lattice parametrized tight-binding model with the coordination and
parameters appropriate for the chosen electrode (part of which has already
been included in the cluster). Figure 3(c) depicts schematically a Bethe lat-
tice of coordination three. The advantage of choosing a Bethe lattice resides
in that, although it does not have long-range order, the short-range order
is captured and it reproduces fairly well the bulk density of states of most
commonly used metallic electrodes. For each atom in the outer planes of
the cluster, we choose to add a branch of the Bethe lattice in the direction
of any missing bulk atom (including those missing in the same plane). In
Fig. 3(b) the directions in which tree branches are added are indicated by
bigger atoms which represent the first atom of the branch in that direction.
Assuming that the most important structural details of the electrode are
included in the central cluster, the Bethe lattices should have no other rel-
evance than that of introducing the most generic bulk electrode for a given
metal. In Appendix A we explain how to calculate Bethe lattice self-energies
for those interested in the mathematical details.
As many other ab initio codes GAUSSIAN98/03 makes use of non-
orthogonal localized orbital basis sets. In a non-orthogonal basis the Green's
function takes the form (see Appendix B)
G W (E) = S [(£S - F - £ W (£) j - 1 S (5)
where S is the overlap matrix. As explained in Appendix B this Green's

function allows us to obtain the density matrix P through the expression
P = _- f F lm\s-lGi-+\E)S-1] dE. (6)

1" J-oo L J
We can now obtain the total charge through the text-book expression
N = Tr[P • S], (7)
where the trace runs over all the basis set elements of the cluster. The upper
limit of the integral in Eq. (6), Ep, is not the Fermi energy of the bulk metal
electrode since these are finite in the actual calculation. (The bulk Fermi
level can only be obtained in the thermodynamic limit). Instead, Ep is set
by imposing overall charge neutrality in the cluster.
The above expression of the Green's function in non-orthogonal basis is
formally correct, but it is not the standard one as commonly appears in the
literature:
G (±) (£) = [tfS - F - E (±) (£)] ~ • (8)
The disappearance of two S matrices in the definition of the Green function
is compensated by a redefinition of the density matrix which now becomes
the "stantard" one:
P =- - f F Im [ G ( + ) ( £ ) 1 &E. (9)
The integral in Eq. (9) or Eq. (6) can be efficiently calculated along a contour
in the complex plane[19,18, 27]. We now force GAUSSIAN98/03 to evaluate
F using the density matrix defined in Eq. (6) instead of the standard one
obtained from filling up the molecular orbitals obtained from diagonalizing
the Fock matrix[33]. The procedure is repeated until self-consistency is
achieved in the density matrix to the desired accuracy. This is, in fact, the
hardest part, computationally speaking and our main achievement. It is
also interesting to note that the applicability of this approach can be easily
implemented in any quantum-chemistry or ab initio code as long as it is
based on a localized orbital expansion of the self-consistent, single-electron
wavefunctions. Finally, as regards the DFT calculations for the core clusters
we will currently use the B3LYP functional[21] and the basis sets and core
pseudotentials described in Refs. [34, 35, 36]. We will check the robustness
of our results against different functional and basis sets when necessary.
Another advantage of using Green's functions is that we can now ob-
tain the transmission probability using a more general expression than that
provided in Eq. (1)[9]:
T{E) = Tr[rh(E)G^(E)rR(E)G^(E)}. (10)
In this expression Tr denotes the trace over all the orbitals of the cluster
the matrices TR^(E) are given by i[S+(L>(£!) - £~, L) (.E)]. Finally, in
order to single out the contribution of individual channels to the current,
one can diagonalize the transmission matrix. While the size of the matrix
FiG^riiG^ can be as large as desired, the number of eigenvalues with
a significant contribution will be typically much smaller and it will be de-

termined by the narrowest part of the nanocontact or the molecule. Notice
that the expression for the transmission probability in Eq. 10 is directly ap-
plicable in non-orthogonal basis sets as long as one uses Eq. 8 to define the
Green's function instead of the mathematically more appropriate definition
of Eq. 5. The transmission can also be found in a slightly different form.
We first rewrite Eq. 10:
T(E) = Tr [TlL/2(E)GW(E)TR(E)1/2TR(E)^2G^(E)r^2(E)\ , (11)
where we have simply made use of the cyclic property of the trace after
splitting the coupling matrices into two. By identifying
t(E) = i f 2 (£)G«(£)rf (£), (12)
we can write
T(E) = Tr[t(E)tf(E)]. (13)
The matrix t and its adjoint tf axe known as the transmission matrices in
scattering theory and their product is a hermitian matrix with real eigen-
values. This confirms that the transmission probability obtained from Eq.
(10) is a real quantity.
Finally, we should stress that in any self-consistent field approximation
only G(E-F) has a strict meaning. In order to obtain the actual dependence
on energy of the zero-bias conductance, Q(E), which would correspond to
the conductance for different values of an external gate potential that can
charge or discharge the system (and change the Fermi energy), one must
perform a new self-consistent calculation for each different value of the Fermi
energy. We have analyzed the extent of this problem on some occasions and
found that the quantity Q(E) evaluated for a charge neutral system can
serve as a good approximation to the real conductance as a function of Ep.
This partially justifies plotting Q(E) for a range of energy values around
the Fermi energy as routinely done in the literature. This problem deserves,
as will be shown below, further consideration when bound or quasibound
degenerate states are present.
3.2.1 Example: Aluminum nanocontacts

Before addressing the subtleties of non-equilibrium transport and moving
onto the detailed analysis of various basic experiments in molecular elec-
tronics, we would like first to present a simple case to illustrate the previous
discussion. An interesting and relatively simple system where our method
can be tested is an Al metallic nanocontact. From all the metallic elements
studied experimentally[6] Al is the least understood despite of the fact that
is one of the most accesible from a computational point of view. A com-
plete theoretical study of electrical transport in metallic atomic contacts
Figure 4: (a) Conductance around the Fermi energy (set to zero) for the
nanocontact shown in the inset, (b) Same as in (a) for the nanocontact
shown in the inset, (c) Transmission of the individual channels as a function
of energy for the nanocontact in (a). The highest trasmitting ones are doubly
degenerate, (d) Same as in (c), but for the nanocontact in (b).
requires a realistic modelization of the formation process of these nanocon-

tacts. Some structural studies for Al using molecular dynamics[15] and 06
initio relaxations[37] have been reported. However, given the illustrative
character of what follows, we will simply consider two archetypal atomic
structures that resemble those appearing in the last stages of the break-
ing process of Al atomic contacts and may be responsible for the last con-
ductance plateau: Single-atom contacts and short atomic wires. The first
structure we consider consists of two opposite pyramids grown in the (001)
direction and "glued" by a single atom [see inset in Fig. 4(a)]. Bulk distances
are taken for the whole structure and it is understood that the appropriate
Bethe lattice has been added to the outer planes of the core cluster.
Single-atom contacts have been studied in the past with modified tight-
binding models[14] which impose local charge neutrality. As shown in Fig.
4(a), we find that, even in this simple case, the ab initio results differ qual-
itatively from those obtained with these models. The value of Q around
the Fermi level is « 3<7o which is remarkably different from the value of
** Go reported in Ref. [14]. This discrepancy is due to a combination of
facts. First, the hopping parameters that reproduce bulk properties in tight-
binding models are not adequate for atoms with low coordination numbers
(like those at the tips), being these typically smaller than the 06 initio ones.
Second, there are non-negligible contributions from next-to-near-neighbor
hoppings in Al, which, are usually neglected in parametrized tight-binding

models[14]. This translates into the fact that the number of conducting
channels at the Fermi energy is larger than 3 as obtained in Ref. [14] [see
Fig. 4(c)].
The second structure we have considered is almost identical to the pre-
vious one, except for having two atoms instead of a single one forming a
short chain (a dimer) at the neck [see inset in Fig. 4(b)]. This structure has
been chosen on the basis of extensive molecular dynamics and ab initio sim-
ulations we have performed which often show the formation of the Al dimer
before the final breaking (a similar conclusion has been recently reached as
reported in Ref. [37]). The distance between tip atoms has been chosen to
sustain a slight stretch. As can be seen in Fig. 4(b), this apparently harm-
less structural modification get us closer to the typical conductance of the
last plateau in the breaking process of an Al nanocontact (G ~ So)- The
number of conducting channels is now three as seen in Fig. 4(d).
These results are fairly robust against changes in the number of planes
forming the pyramids (i.e., the cluster size) and the crystallographic ori-
entation of them. We should mention once again, however, that talking
about an exact value for the conductance has little meaning since this will
always be slightly dependent on the detailed atomic structure close to the
contact which changes from experimental realization to experimental real-
ization. Nevertheless, within this inherent statistical uncertainty, our results
are fairly consistent with the average experimental results. With the hope
that this example has already convinced the reader about the predictive
and quantitative power of this ab initio methodology we move on to more
complex problems.
3.3 Methodology: Finite bias, finite temperature

When one is interested in the response to a finite bias, Q(V), as obtained
from an equilibrium calculation, can still be a good approximation to the
real differential conductance dl/dV. Strictly speaking, however, what one
needs to do is to calculate the current / by integrating the transmission
probability in a energy window eV:
9P f°°
I=~ T(E, V) [fL(E - eV/2) - fK(E + eV/2)]dE, (14)
h J-oo
where / L and / R are the Fermi distribution functions of the left and right
electrodes, respectively. Notice that this innocent-looking expression is not
a trivial generalization of Eq. (1) extended to finite bias voltages and finite
temperature. The transmission coefficient that appears in the integrand
T(E, V) can depend on temperature, but, most importantly, depends on
V. To calculate this transmission coefficient, knowledge of the Fock matrix
in the presence of a bias voltage is required which, in turn, depends on
the density matrix out of equilibrium. Based on standard non-equilibrium

Green function techniques, it has been repeatedly shown in the literature
how to do this [9]
P = - - f°° [S-1G<(E)S"1] dE, (15)
where
G<(E) = tG<">(£) [fL(E)Th(E) + fR(E)TR(E)} G<+>(£). (16)
There are several technical and conceptual issues regarding the computa-
tion of the above equations that need some discussion. We refer the reader
to the literature for a detailed account of them[19, 25, 18]. Here we will
simply mention two issues. Maybe the most important conceptual issue is
that related to the fact that using DFT for non-equilibrium situations is not
fully justified: DFT is a ground-state theory. We will make use of it simply
because, at present, there is no better way to deal with out-of-equilibrium
problems in an operative way. The second issue is a technical one that, in
our view, has not received due attention in the literature and that we find
it worth a more detailed discussion[25]: The out-of-equilibrium electrostat-
ics. In most self-consistent approaches[16, 19, 26, 18] the Poisson equation is
solved with boundary conditions appropriate for the electrode geometries (as
mentioned, typically infinite planes) and the bias applied. In other words,
an external and uniform electric field is superimposed to the self-consistent
field created by the molecule or supermolecule. In the Keldysh formalism,
where one imposes an electrochemical potential difference /J,^C — fj^c = eV,
the electric field does not need to be added if a significant part of the metal-
lic electrodes has already been included in the central cluster. Thus, one
can consider realistic electrode geometries if necessary. To be more specific,
let us separate \iec into two contributions:
Mec = Mchemical + êlectrostatic > (17)
where we define
Mchemical = lim SEcole/SN

6N—>0
êlectrostatic = Vui^SEee/SN.
ECoie is the standard energy contribution to the total energy from the cores
of the atoms and See is the contribution from the electron-electron interac-
tion. Note that this definition is not a standard one in the sense that one
usually thinks of external fields when referring to the electrostatic potential
contribution to the electrochemical potential. Instead we will make use of
this term when referring to any contribution coming from electron-electron
Figure 5: Schematic picture of the development of a potential drop between

electrodes separated by a barrier on imposing an electrochemical potential
difference between them, (a) Before self-consistency begins the chemical
potential in electrodes is different by an amount eV and the bottom of the
conduction bands is the same for both electrodes. This situation corresponds
to a diffusion problem, (b) After self-consistency has been achieved the
chemical potential difference practically vanishes in favor of an electrostatic
potential difference between eletrodes.
interactions. Strictly speaking, however, one should not use the term elec-
trostatic potential since there are exchange and correlation contributions
included in DFT. Specifically,
SEcore = tr[(JP • hcorj
SEn = tr[£P-hee],
where hcore and hee are the core and interaction matrices composing the
Fock matrix. In a self-consistent procedure with overall charge neutrality
one electrode gets charged and the other discharged by the same amount.
Once self-consistency has been achieved, an electrostatic potential difference
V between atoms one or two layers inside opposite electrodes develops while
their chemical potentials are similar, i.e., they remain neutral. On the other
hand, the electrostatic potential drop will be smaller than V between atoms
at the surface of opposite electrodes since they carry the charges. We show
this process schematically in Fig. 5.
3.3.1 Example: Gold nanocontacts

In order to illustrate the previous discussion, we carry out calculations for
Au nanocontacts. It is well known that Au can form atomic chains of up to
7-8 atoms[6]. Two types of fee core cluster models are considered now: (a)
16 J. J. Palacios et cd.
Figure 6: Average on-site energies of the atoms in the Au cluster shown in

the inset for zero bias (white dots) and 5 V (dark dots).
two (111) planes containing 19 atoms each plus a four atom chain labeled
Aul9-Au4-Aul9 (a nanocapacitor can be modelled by taking out the four-
atom chain leaving a wide gap between the planes); (b) two (lOO)-oriented
pyramidal clusters plus a four-atom chain labeled Aul4-Au4-Aul4 [(100)-
oriented double-pyramid nanocontacts like those in previous section, Au5-
Au5 and Aul4-Aul4, have also been considered]. As for the case of Al in
previous section, interatomic bulk distances have been taken for the cluster
(2.88 A). All results correspond to zero temperature and an external bias
voltage in the range 0-5 V. Figure 6 shows the average on-site energies
(5d6s6p) on all atoms of the nanocontact Aul9-Au4-Aul9 for 0 and 5 V.
This magnitude reflects only the potential energy on each atom, not telling
us anything about the chemical potential (charge) on them. As can be
seen, at 5 V bias the major drop in the potential occurs at the chain/plane
contacts. While the total potential drop is 4.44 eV the drop between the
first and the last chain atoms is 1.67. In the case of zero bias the results are
symmetric with respect to the geometric center of symmetry, as expected.
For finite bias the potential drop between the left electrode and the first
atom in the chain is 1.92 eV while it is only of 0.85 eV between the chain
end and the right electrode. This feature seems common to all previously
reported results and reflects not only bulk band structure features, but also
details of the nanocontact geometry [18].
The voltage-dependent transmission T(E, V) for V = 0 and 5 V is shown
in Fig. 7 and changes significantly. For instance, the gap below -0.5 eV
12 " Au
0.9 - tjx/y?^
H 0.6 • f^'
0.3 i l l
[W^ j
-2.5 -1.5 -0.5 0.5 1.5 2.5
Energy (eV)
Figure 7: Tranmission versus energy for V=0 (continuous line) and 5 V

(dashed line) for the Aul9-Au4-Aul9 nanocontact shown in Fig. 6.
Table 1: Fittings of the numerical results for the variation with applied
voltage V of the difference between the average on-site energies of the atomic
orbitals at the outer planes in clusters similar to those of Figures 1 and 2
(see text) and of current through them, with functions AE = aVb and
I/QQ = cVd, respectively (AE, I/QQ and V given in eV).
cluster a b e d
Au5-Au5 0.803 0.935 0.521 1.068
Aul4-Aul4 0.840 0.993 0.668 1.067
Aul4-Au4-Aul4 0.882 0.971 0.805 0.950
Aul9-Au4-Aul9 0.882 1.003 0.886 0.91
Aul9-vacumm-Aul9 0.928 1.001
is partially filled. In both cases the conductance never goes above QQ as

corresponds to an s-like element such as Au. It is worth noting that al-
though the total current calculated by integrating the zero-bias transmis-
sion T(E, V = 0) in an energy window {-eV/2, eV/2] does not differ much
from that obtained with the full non-equilibrium approach (not shown), the
differential conductance is significantly different.
In Table I we report the fittings to the numerical results for the difference
between the average energies of all atoms in the left and right planes in the
previous nanocontact and between the two outer planes in the pyramidal
nanocontacts versus the applied bias V. The fittings were done with aVb,
instead of assuming a linear relationship from the start. All regression co-
efficients were higher than 0.999. Although a and b are always rather close
to 1, some significant features are worth of comment. There is a steady
increase of a and b as the size of the cluster increases. The values closest to
one are found for the capacitors (the nanocontacts with an « 14 A gap). In
the latter case the coefficient a is only 7% smaller than 1, while the exponent
cannot in practice be differentiated from one.
As explained above, the small deviations with respect to the "ideal" be-
havior are due to the finite charge that those planes carry. As expected,
as the size of the cluster increases, the results steadily approach the macro-
scopic behavior where the applied electrochemical potential difference eV
and the potential drop coincide. These results prove our previous state-
ment: No external field needs to be added for sufficiently large electrodes.
The results for the current versus voltage were also fitted to I/Go = bVd (see
Table I). In this case the lowest regression coefficient is 0.99 (in some cases
the results are better fitted with second order polynomia). The significant
deviations from a linear behavior (Ohm's law) are not surprising in systems
as small as those investigated here and have been reported previously [18].
4 Towards a quantitative description of

single-molecule junctions
The following cases illustrate how the technique we have just presented
in previous section can be of great help in the interpretation of various
experiments related to the field of molecular electronics. We have chosen to
focus on the transport properties of three paradigmatic molecules: H2, Ceo,
and a (5,5) metallic carbon nanotube. These molecules have been selected
on the basis of either their practical and basic interest. At the same time
they will take us on a trip through the three main experimental techniques
used nowadays to create single-molecule junctions: MCBJ's, STM's and
conventional lithography.
Figure 8: (a) Conductance as a function of energy (referred to the Fermi en-

ergy) for the (lll)-oriented 6-3-1-3-6 clusters with (dashed line) and without
(dotted line) mirror symmetry and for the (lOO)-oriented 9-4-1-4-9 cluster
(solid line); (b) same as in (a) but for the (lll)-oriented 6-3-1-1-3-6 clusters
with (dashed line) and without (dotted line) mirror symmetry and for the
(lOO)-oriented 9-4-1-1-4-9 cluster (solid line); (c) transmission eigenvalues
for the 6-3-1-3-6 cluster with mirror symmetry [see inset in (a)]; (d) same as
in (c) but for the 6-3-1-1-3-6 cluster [see inset in (b)] with mirror symmetry.
4.1 Platinum nanocontacts with H2 molecules

We start our set of examples by showing how the methodology just de-
scribed can shed light on a recent experiment [38, 39] of electronic transport
in Pt nanocontacts where, apparently, H2 molecules anchor themselves to
the metal and modify the standard conductance of pure Pt. The H2 molecule
is the simplest molecule and is ideal to carry out a detailed structural and
transport analysis. The main conclusion in this section is the following:
Contrary to the authors' explanation for the conductance histograms where
the main peak around Q sw Go (for Go = 2e2/h) is attributed to a config-
uration where a H2 molecule bridges the gap between Pt electrodes (see
also Ref. [40]), we attribute this peak to a situation where the H2 molecule
dissociates at the nanocontact and forms a stable complex Pt2H2 where the
two H atoms chemisorb between electrode tip atoms [see left inset in Fig.
10]. We speculate that the bridge with H in molecular form could account
for the peak around Q « O.lSo reported in Ref. [38], but not commented on
there.
As a starting point we consider the conductance of a clean Pt nanocon-
tact without H2. As mentioned on several occassions in this review, it would

be desirable to simulate the whole breaking (or formation) process of the
nanocontact on stretching (or indentation). This is, unfortunately, computa-
tionally not feasible from first principles for Pt. As we did for Al, we select a
few representative atomic arrangements for the lowest conductance plateau
out of the many possible. There is "visual" evidence for Au[41] that con-
firms the formation of pyramids on both (111) and (001) stretching lattice
directions in mechanical break junction experiments. Assuming the same for
Pt given the lattice similarities, we have chosen to study the conductance
for the following nanocontact atomic configurations: 9-4-1-4-9 (single-atom
contact) and 9-4-1-1-4-9 (chain contact) for the (001) direction and 6-3-1-3-6
(single-atom contact) and 6-3-1-1-3-6 (chain contact) for the (111) direction
(the numbers indicate the number of atoms per plane). We have considered
two possible plane orderings for the (111) direction: 6B-3A-.. .-3A-6B and
6C-3A-. ..-3C-6A (remember that the bulk one is ... A-B-C-A-B-C . . . ) .
These nanocontacts correspond to cases with and without mirror symmetry
with respect to the middle plane perpendicular to the stretching axis z [the
(001) nanocontact has also been chosen with mirror symmetry]. Specifi-
cally, we have taken bulk atomic distances for the atoms in the outer planes,
but we have performed full relaxation in all the other coordinates, includ-
ing the distance between the two outer planes. The resulting structures are
thus expected to be representative of the ones responsible for the last(first)
conductance plateau on stretching(indentation) cycles.
The conductance reaches « 2ft> at the Fermi energy (set to zero) for
the two short-chain cases with mirror symmetry and drops slightly for the
other one [see Fig. 8(b)]. Two conduction channels mainly contribute to the
conductance in the three clusters considered with the two-atom chain [see
Fig. 8(d)]. Given the polycrystalline nature of the electrodes, one should
not expect mirror symmetry to be the dominant situation in the experiment
so we can expect an average conductance around Q « 1.75ft)- This value
nicely coincides with the position of the lowest conductance histogram peak
reported in Ref. [38]. The conductance of the single-atom contacts presents
similar values for the (111) direction [although, as Fig. 8(c) shows, three
channels now contribute to transport], but is appreciable higher in the (001)
direction [see Fig. 8(a)]. Although we cannot afford a statistical analysis,
the fact that the position of the lowest conductance histogram peak shifts to
higher values when single-atom contacts are favoured (on only considering
indentation cycles or higher bias voltages as reported in Ref. [42]) is con-
sistent with our results. The conductance is also here noticeably dependent
on the detailed atomic arrangement of the electrodes (note that the conduc-
tance increases appreciably on the scale of 1 eV below and above the Fermi
energy). This is due to the fact that the Fermi level lies on the edge of the
d band for bulk and the contact atom has almost bulk coordination. This
is also fairly consistent with the fact that Pt histogram peaks are broad in
_274.07 p , , . . . , .
I • CF |5 * • * '
s
U -274.09 - *
- 274 - 12 o • 1 i ' i
Distance Uwstrcnes)
Figure 9: Total energy as a function of the distance of the H2 molecule to

the axis of the nanocontact for the cluster shown in the insets. The relative
orientation of the molecule has been forced to that shown in the insets (along
the axis), but the rest of coordinates are allowed to relax. The upper inset
corresponds the minimum of the curve. This minimum indicates that a
certain amount of strain is necessary in order for the molecule to move in
between the two Pt atoms of the tips.
comparison to, e.g., those of Au[6].

When the same conductance measurements are performed in a H2 atmo-
sphere the histograms change noticeably. Inelastic vibrational spectroscopy
along with theoretical support lead the authors in Ref. [38] to conclude that
a H2 molecule can position itself along the stretching axis of the Pt nanocon-
tact right before it breaks apart. The authors claim that this configuration
accounts for the main histogram peak for conductance where Q » (?o- With
this claim in mind we consider nanocontacts like that shown in the inset
of Fig. 8(b), but with a H2 molecule anchored in between electrode tip
atoms. The results presented below are for the smallest cluster in the (001)
direction, 4-1-H2-1-4, but we have performed similar calculations for (111)
nanocontacts for reassurement. We maintain bulk interatomic distances for
the four atoms in the two outer planes whose variable respective distance d
mimics the effect of the piezo voltage. For every d we let the positioning of
the H atoms and that of the two contacting Pt atoms relax, starting from
a configuration where the H2 molecule sits along the stretching axis. Full
circles in upper curve in Fig. 10 represent the total energy vs. d for all the
stable configurations we have found in which the molecule stays on the z-
axis (see right inset in Fig. 10). Beyond d « 9.5A the molecule sticks to one
of the electrodes or dissociates. Below d « 8.25A the molecule tends to loose
22 J. J. Palacios el al.
-274.001 1
-274.02 - 9
^-274.08- \£^>*^T
•g -274.10 - \ J
-274.12 - \X _/
-274.14 -
55 6.0 6J 7.0 7.5 8.0 8J 9.0 9.5 10.0 10.5
d (A)
Figure 10: Total energy of a H2 molecule in a Pt nanocontact versus distance

between electrodes (outer planes). Upper curve circles correspond to the H
atoms positioned along the stretching axis (2) in molecular form (see right
inset) and lower curve circles correspond to the H atoms positioned on the
plane perpendicular to the stretching axis (see left inset).
its alignment with the 2-axis. (As a guide to the eye we also present with
empty circles the energy for configurations where the molecule is forced to
stay along the z-axis). In all these points the H2 retains its molecular form.
We also observe that a certain amount of stress is necessary in order for the
molecule to position itself in between Pt atoms. This can be more easily
seen in Fig. 9 where we see a minimum in the bonding energy at a position
w 1 A away from the nanocontact. At an intermediate point (d « 9 Ain
Fig. 10) the H-H bond distance and the H-Pt distance turn out to be simi-
lar to those proposed in Ref. [38]. Figure 11 (a) shows Q{E) for d = 9.25A
(we chose the distance that more closely reproduced the vibrational mode
frequency reported in Ref. [38].
The conductance at the Fermi energy turns out to be remarkably smaller
than Qo- A similar calculation for the whole range of stable configurations
shown in Fig. 10 gave a conductance in the range 0.2-0.5C/O- To our ease
this result agrees with the intuitive picture: Closed-shell molecules can-
not conduct if the molecular character is maintained after contact with the
electrodes and the charge transfer is much smaller than one. Both condi-
tions seem to hold here. The charge transfer from the H2 molecule to the
electrodes is typically « 0.15 or smaller and the bonding and antibonding
molecular levels maintain their character since they are clearly visible in the
density of states (not shown here) around -6 eV and 17 eV, respectively. A
value of conductance close to Qo is still possible through hybridization of the
Pt states with those of the molecule[40]. In our calculations, however, this
does not occur at the Fermi energy, but at E « —1.5 eV. We believe that
3i 1 -r p
I (a) A ^ ! (b) •
°-2 - 1 0 1 - 2 - 1 0 1 2
Energy (eV) Energy (eV)
Figure 11: (a) Conductance as a function of energy (referred to the Fermi

energy) for the cluster shown in the inset where a H2 molecule sits along
the stretching axis of the Pt nanocontact (d = 9.25A). Dotted lines in (c)
represent the transmission of the individual eigenchannels for the cluster in
(a), (b) Same as in (a) but for the structure shown in the inset where the
two H atoms form a double atomic bridge (d = 6.5A). Dotted lines in (d)
represent the transmission of the individual eigenchannels for the cluster in
(b).
this H2 bridge actually accounts for a secondary peak around Q =» 0.1ft) n °t

discussed, but clearly present in the reported data[38].
So, what accounts for the main conductance peak at « Go?. There is
little doubt that this peak is H-related, but here we propose an alternative
to the H2 molecular bridge discussed above. From the previous calculations
we observe a "natural" tendency for the molecule to stand up-right in be-
tween Pt atoms for small distances between outer planes. We now perform
calculations similar to the previous ones, but initially placing the molecule
axis perpendicular to the stretching 2-axis. We let the Pt tip atoms and the
H atoms relax towards equilibrium for several values of d without assuming
the H2 to be in molecular form. The total energy of the stable configurations
we have found are represented by full circles in the lower curve in Fig. 10.
(Again, for completeness, we also present results when the H atoms are not
allowed to leave the plane perpendicular to the 2-axis). The two H atoms
are only in molecular state at the upper limit of this curve (d « 8 A). For
smaller values of d the H2 molecule breaks apart and the H atoms position
themselves up to 2.0 A apart from each other forming a double atomic H
bridge as the one shown in the left inset of Fig. 10. The results of conduc-
tance for d = 6.5A (a distance for which vibrational modes are closer to
the experimental ones) are depicted in Fig. ll(b). The conductance at the
Fermi energy approaches closely QQ. The eigenchannel transmission results
_274,04 | . 1 * . , . 1
-274.08 %|j>
8 -274,08 • * •
I .' *'
-874-1*, • i • i • ;
Distance {AascroossS
Figure 12: Total energy as a function of the distance of the H2 molecule to

the axis of the nanocontact for the cluster shown in the insets. The relative
orientation of the molecule has been forced to that shown in the insets, but
the rest of coordinates are allowed to relax, including the H-H distance. The
lower inset corresponds to the molecule far from and perpendicular to the
axis, The only energy minimum is found at zero distance when the H atoms
separate (see upper inset).
are also consistent with the noise analysis in Ref. [38] since, the fact that
fluctuations do not strictly vanish at QQ may indicate that more than one
channel is contributing to the conductance. At this point the distance be-
tween tip Pt atoms is approximately that of bulk Pt and that between tip Pt
atoms in Fig. 8(b). When comparing with the results presented in Fig. 8(b)
we see that the H seems to block out one of the channels. It is important to
finally stress here that the exact value of the conductance is not very depen-
dent on the basis set and the functional. However, we note once again that
the atomic configuration of the electrodes in the cluster and the model used
to represent the bulk of these electrodes introduces a source of uncertainty
in the results. We believe that this is the source of the discrepancy between
our results and those presented in Refs. [38] and [40]. After considering
various other possibilities in terms of cluster sizes and orientations, we can
safely conclude that the stark difference between the conductance of the two
basic bridge structures studied in this work is conclusive.
Dissociation of the H2 molecule at low temperature is not that surprising.
Although it is well known that the molecule is not dissociated upon adsorp-
tion on flat, perfect surfaces (and our procedures were checked to reproduce
this result), it may well be possible in the presence of low-coordinated Pt
(such as in the present case or in strongly stepped surfaces). We have studied
the formation probability of these structures letting the H atoms fully relax
starting from H2 in molecular state far from the nanocontact. Although a
full dynamical adsorption analysis is beyond the scope of this work, we have
found that these novel configurations can be reached starting from molecu-
lar H fax away from the nanocontact without surpassing activation barriers
(see Fig. 12). We have also found higher-energy stable configurations or
metastable local minima; for instance, on some relaxations we have found
that a single atom locates between the Pt electrodes while the other moves
onto the electrode surface. Although these metastable configurations can
play an active role in the formation of the actual nanocontact and in the
histograms, we defer the taxonomy of this problem for future work.
4.2 Scanning tunneling spectroscopy of fullerenes adsorbed

on Au surfaces
Since its discovery in 1982 by Binnig et a/. [43] the Scanning Tunneling Micro-
scope (STM) has proved to be an indispensable tool to study the mechanical,
electrical, and geometrical properties of surfaces, adsorbates and, as men-
tioned in the introduction, to build metallic nanocontacts and nanodevices.
The tremendous power of STM comes both from its highly local character,
which allows close inspection at the atomic level, and its dependence on
the electronic structure of the substrate being scanned. This is reflected by
Tersoff and Hamann's formula[44, 45] who, simplifying Bardeen's first-order
formalism [46] by approximating the electronic structure of the tip, arrive at
the following equation:
fEF+eV
Joe / pt(E-eV)Ps(E,rt)dE. (18)
JEF
As deduced from this formula, the current I increases with the density of
states (per unit volume and energy) in the tip, pt(E), and the density of
states in the sample localized around the radius of curvature of the tip,
ps(E, rt), both lying in the energy window of width eV between the electro-
chemical potentials of the electrodes. Bardeen's approximation consists in
assuming that the states on each side of the energy barrier (built up between
the electrodes by their different work functions) decay in the vacuum, being
their mutual overlap the origin of the exponential decay of the tunneling
current with tip-sample distance. Actually, following Lang[47], the barrier
may be simplified by a square potential well of width s, and the tunneling
current attains the following form:
pEF+eV .
Joe / pt(E - eV)ps(E)e-2s^('t'+EF-E)+eVdE (19)
JEF
where cj> represents the work function of the sample.

Unfortunately, the above formulae deduced within Bardeen's formalism
are not valid when tip and sample come in closer contact. This occurs, for
Figure 13: Left panel: Normalized conductance derived from CITS mea-
surements by Rogero et al. [48] for the two types of C ^ adsorption sites on
Au(lll). Right panel: CITS image by Rogero et aZ.[48] at a bias of +0.6V,
where the difference in brightness between on-top and bridge sites is clearly
seen. The circle and triangle indicate the points where the curves shown
in the left panel were obtained. The Fermi level is set to zero. [Reprinted
with permission from Ref. 48 (Figs. 4d, 4c)© 2002 American Institute of
Physics].
instance, when the STM tip is used to fabricate nanocontacts (see, e.g., Ref.
[6]). Therefore, a computational method incorporating Landauer's formal-
ism is much more useful if one is interested in studying STM experiments
beyond the tunneling regime. An efficient implementation of such a theo-
retical framework has been explained in Sec. 3 of this work, which has also
the appealing feature of being an ab-initio method, without the need to in-
troduce ad-hoc parameters to describe the tip-sample interaction[47, 49, 50].
In the following we will compare the results achieved with the GECM with
those from a STM study by Rogero et al.[4S] of Ceo adsorbed on Au(lll).
To be more specific, what these authors perform is a spectroscopic study
named Current Imaging Tunneling Spectroscopy (CITS). This technique
measures the tunneling current at each point during a topographic scan for
a range of bias voltages applied to the sample-tip system. The numerical
differentiation of each curve provides the corresponding conductance profile.
From these curves it is possible to construct an image of the conductance
at a given bias for each point of the topographic scan. This type of mea-
surements provide a tremendous amount of information about the electronic
structure of the sample around the Fermi energy. However, taking full profit
of these type of experiments requires a theoretical method able to corrob-
orate the interpretation of the measurements. We give next the details of
Molecular Electronics xuith Gaussian98/03 27
Figure 14: Conductance spectra calculated with our method for the on-top
and bridge adsorption sites in the region around the Fermi level (here set
to zero). The upper left inset shows the same data for the on-top geometry
on a wider energy range, where the fullerene gap can be clearly appreciated.
The other insets show the atomic models we have used in the calculations.
Rogero et al.'s work, starting with a brief overview of the adsorption of Cgo
on metallic surfaces. Later on we will discuss their main achievements as
compared with our own findings using the the GECM[51].
Nucleatkm and growth of Ceo monolayers and thin films on Au(lll)
surfaces has been widely studied[52, 53, 54, 55, 56, 57]. Although bind-
ing between gold and Ceo is weaker than with other metals [55, 58, 57],
it is far from negligible; the adsorption energy is estimated around 40-60
kcal/mol[53, 57]. In fact, adsorption of Ceo is able to lift the well-known
23 x y/3 Au(lll) reconstruction, and photoemission studies of Ceo on poly-
crystalline Au[59, 60] and Au(lll) surfaces{57] revealed energy shifts indica-
tive of LUMO hybridization and charge transfer from Au to the adsorbed
fullerene molecule. At high coverages, closed-packed layers grow with the
thermodynamically most stable adsorbate phase being a (2\/3 x 2\/3) R30°
structure with a nearly perfect lattice matching in which all the molecules
are in equivalent surface sites [53]. Apart from this, another superstructure
forms with crystallographic directions matching those of the substrate, re-
sulting in fullerene molecules sitting on different adsorption sites[53, 55].
The proposed 11 x 11 Cgo coverage[53, 55] for this superstructure has been
recently discovered to be composed of a smaller 2x2 grid by Rogero et al. [48].
The main features that characterize the adsorption of the above-mentioned
Ceo adlayers can be drawn from Fig. 13, where we show the CITS data ob-
tained by Rogero et oZ.[48]. The image appearing in the right panel of Fig.
13 was obtained at a bias of +0.6V[48]. The curves plotted in the left panel
of Fig. 13 correspond to the conductance profiles measured at the points
indicated with the circle and triangle[48]. From the different position and
width of the peaks appearing in the conductance profile Rogero et al. deduce
the following facts:
1. The relatively weak C6o-Au(lll) binding reflects on the sharp form of
the peaks, with the Ceo LUMO resonance shifted towards the Fermi
level by charge transfer from the substrate, yielding a HOMO-LUMO
gap of 2.3 eV.
2. The two curves appearing in the left panel of Fig. 13 are interpreted
as corresponding to a 2x2 superstructure of the adsorbed layer that
places the Cgo molecules on two different adsorption sites (bridge and
on-top) of the underlying substrate surface: the interaction of the
fullerene with the adsorption site being responsible for the two type of
spectra.
We now proceed to check if the interpretation of the experimental facts
by Rogero and coworkers is confirmed by our calculations. With this aim we
calculate the conductance profile corresponding to the two geometries that
characterize the 2x2 Ceo superstructure (see insets in Fig. 14). The results
are plotted in Fig. 14. From this figure it is immediately apparent one of the
main facts found by Rogero et al. and previous authors; namely, that the
interaction between the molecule and the surface is not very strong. This
is reflected in the sharp peaks that reveal the underlying positioning of the
Ceo orbitals, from which we estimate the HOMO-LUMO gap to be 2.9 eV,
in good accordance with the experimental CITS results [48].
On the other hand, the relative positioning and height of the conductance
maxima between the on-top and bridge geometries is also reproduced. The
form of the peaks reflects, as suggested by Rogero et al.,[48] the different
interaction of the molecules with the adsorption sites as deduced from the
Potential Energy Scan (PES) shown in Fig. 15. For each adsorption site
we plot DFT calculations when a Ceo molecule approaches the Au(lll)
surface towards the on-top and bridge sites, respectively, with either a six-
member ring or a five-member ring facing the surface. The same equilibrium
distance (2.75A) is obtained for all the geometries, which coincides with
that found in Ceo-gold nanobridges[61] and is consistent with Altman and
Colton suggestion of no height differences between adsorbed Ceo molecules
on Au(lll)[53, 54]. As seen from Fig. 15 the adsorption energy is smaller
when the molecule sits on top of a gold atom (16 kcal/mol) than when the
fullerene binds to a bridge site (35 kcal/mol), which explains the narrower
and higher form of the conductance maximum corresponding to the on-top
geometry with respect to the bridge site. Once the aforementioned values
are corrected by considering the lateral interaction of the Ceo monolayer [57],
300 r—•—i—•—i—•—i—•—,—•—i—•—i—•—i—•—i
2 5 0 - 1
o\ • • Hexagon facing on-top
• \l • • Hexagon facing bridge :
t oil o o Pentagon facing on-top
*\\ a o Pentagon facing bridge
E/kcalmol ' -JQO \ -
50 - %^
-50 - :
_100 I—• ' ' ' • ' • ' • ' ' ' • ' •—'
1 1.5 2 2.5 3 3.5 4 4.5 5
Distance C,0-surface / A >
Figure 15: DFT Potential Energy Scan of a Ceo molecule approaching the
two types of Au(lll) surface sites (on-top and bridge). Results with an
hexagon and a pentagon of the fullerene facing the surface are included for
completeness. The energy of the separated Ceo + surface is set to zero.
estimated from the Lennard-Jones potential to be of about 25 kcal/mol[57],

they lie within the experimental margins discussed above. The stronger
Au-C6o interaction when the molecule sits on the bridge site is consistent
with a larger amount of charge transferred from the gold surface to the
molecule: 0.8 electron, as compared to the value of 0.5 electron obtained
for the on-top site. This gap accounts for the different alignment of the
Fermi level, which for the bridge geometry lies closer to the LUMO-derived
orbitals, in complete agreement with Rogero et al.'s findings. The amount
of transferred charge calculated by us is in accordance with recent results
derived from photoemission spectra[57], which give 0.8±0.2 electrons per
fullerene molecule adsorbed on Au(lll). Finally, we point out the fact that
there exist minor differences in binding energy with respect to the symmetry
axis that points towards the surface. Actually, this coincides with the fact
that no predominant molecular orientation had been found [48] and is also
indicative of the relative weak C6o-Au(lll) interaction, which allows a large
degree of rotational freedom on the adsorbed fullerene.
As mentioned before, our method is not only applicable to the tunnel-
ing regime but also at tip-sample contact distances. This is reflected in
Fig. 16 where we plot the conductance of the system for the two types of
adsorption sites as the tip moves towards the adsorbed Ceo- The two max-
ima correspond to conductance channels coming from the first three C6o
LUMO-derived orbitals, whose degeneracy has been partially removed due
to the interaction with the gold surface[48, 57] in two sets of two (broader
1.5 | . 1 . . .
1.5 I—.—i—.—i—.—i—.—I
1 B.«
O - 3 - 2 - 1 0 1 D=1E.DA
ft"»'«v :', D.15.5A
: ;ji D = is DA
Ail
-0.5 0 0.5 1
Energy / eV *
1.5 | . 1 . 1 •
1.5 i—. i i i——T—.—i
1 - 3 J/ . .
Q /.I \[« ' 14 D.140A
o n [AL, . , . JtA-J m\ D.I«A
o -» -! -1 ° ' i ="-S:S{
0.5 - | U
-0.5 0 0.5 1
Exergi'UV •
Figure 16: Top panel: Conductance profile from tunneling to contact regime
of a Ceo molecule adsorbed on top of a gold atom in a Au(lll) surface.
Bottom panel: Same as above for the Ceo molecule adsorbed on a bridge
site in a Au(lll) surface. Distance (in Amgstron) between gold surface and
tip displayed in the legend. The Fermi level has been set to zero in both sets
of curves. The insets in both panels plots the same data on a wider energy
range at tip-surface distances of 13.5 and 14.5 A.
1.5 I . 1 . 1 . ,
X \ _4 [—, 1 , 1 ,—Mt—
S A \ 1i 12 13 14
O \ \-, Distance A •
\ \
O—e On-top \ ^
4—A Bridge *J\
11 12 13 14
Tip-surface distance/ A r
Figure 17: Maximum conductance from the second peak of Fig. 16 for both
geometries. The tunneling regime can be distinguished by the exponential
decay in the conductance beyond 13A. The inset shows the same data with
a logarithmic scale for the conductance.
peak) and one (sharper peak) resonances, respectively. These two peaks
also appear at tunnel tip-surface distances in the experimental CITS data
of Rogero et al. (see Fig. 13). We can check that, as the tip comes in
closer contact with the fullerene, their interaction alters the size, width, and
positioning of the peaks. The net result is an increase in the conductance
to non-negligible values and a larger shift of the LUMO-derived C6o orbitals
towards the Fermi level due to the larger amount of charge transferred from
the tip atoms to the fullerene as they approach each other. This fact ex-
plains the slight closure of the HOMO-LUMO gap as the tip approaches
the molecule, which is apparent after inspecting the insets of Fig. 16. On
the other hand, the relative shape of the peaks corresponding to the on-top
and bridge geometries found at tunneling distances is maintained as the tip
approaches the sample. This still reflects the differences in binding strength
commented above.
Figure 17 shows the value of the conductance maxima of the second
peak vs. tip-surface distance. We can see the change of slope due to the
different conductance regimes, with the exponential decay typical of tunnel-
ing appearing beyond 13A, which agrees well with the experimental value
reported by Joachim et al.[50] of 13.2A. The departure from a linear trend
in the logarithmic representation of the conductance is related to the above-
mentioned closure of the Ceo HOMO-LUMO gap. Structural deformation
of the fullerene cage has not been considered in our calculations since we
Figure 18: Two atomistic models for the CNT-electrode contact: (a) an
open (5,5) carbon nanotube end-contacted to (111) surfaces and (b) a similar
nanotube side-contacted.
have focused on the change from tunneling to contact regimes, where the
deformation of Ceo by the tip is negligible [50].
In summary, we have shown that the so-called GECM method is able
to accurately reproduce CITS spectra. The example of Ceo adsorbed on
Au(lll) represents just a starting point, but the method looks promising
as a valuable tool in the interpretation of STM and CITS spectra up to the
contact regime. The main results of this section have been published in Ref.
[51]
4.3 Carbon nanotubes with realistic contacts

The controversy on the observed electrical transport properties of carbon
nanotubes (CNT's) is due to our lack of control and understanding of their
contact to the metallic electrodes. It is clear now that the CNT-electrode
contact influences critically the overall performance of the CNT and that
it is crucial to lower the inherent contact resistance to achieve the definite
understanding of the intrinsic electrical properties of CNT's[62]. In order
to determine the relevant factors behind the contact resistance so that this
can be pushed down to the quantum limit per channel RQ = I/Go, a big
experimental effort has been made both in CNT growth and lithographic
techniques[63, 64]. While considerable progress has already been made, it
is fair to admit that theoretical support has not been determinant in this
regard. A limited number of works have addressed this issue previously [65,
66, 19, 67, 68], but a full analysis of this problem requires the use of the
state-of-the-art ab initio techniques to calculate electrical transport like the
one we are describing in this review and these are still under developement.
As we have already mentioned in previous sections, there are multiple
difficulties beyond the usual computational ones in first-principles calcula-
tions. The main one is that the actual atomic structure of the electrode
(and probably that of the CNT) at the contact are unknown and, most
likely, change from sample to sample even when fabricated under the same
Figure 19: (a) Conductance as a function of energy for an N = 10 (5,5) open

metallic nanotube end-contacted to Al(lll) surfaces [see Fig. 18(a)]. The
Fermi energy is set to zero. Inset: Schematic band structure of the metallic
nanotube showing the four states responsible for the resonances, (b) Trans-
mission as a function of energy for the highest conducting eigenchannels.
conditions. It is our believe that atomic-scale modeling, however, can still be

of guidance in the interpretation of the experiments and to the future design
of operational devices with CNT's. There are two ingredients in this puzzle:
The atomic-scale geometry and the chemical nature of the electrode. In
order to address this issue we study open single-walled metallic (5,5) CNT's
contacted in two representative forms (see Fig. 18) to Al, Ti, and Pd elec-
trodes which are among the most commonly used metals in the experiments.
We find that CNT's contacted to Al in end-contact geometry [see Fig. 18(a)]
the two CNT bands couple weakly to the electrodes. Moreover, we find that
the two bands couple very differently to the electrodes (one of them is almost
shut down for transport) and do not mix. For the side-contact geometry [see
Fig. 18(b)] the coupling is the same for both bands, but equally weak as for
to the end-contact geometry.
Figure 19(a) shows Q{E) for a (5,5) metallic CNT composed of N = 10
carbon layers (diameter « 6.8 A and length «s 11.2 A) that has been end-
contacted [Fig. 18(a)] to Al(lll) surfaces. The end-carbon-layer-surface
distance has been optimized to a value of 1.8A. It is important to obtain a

correct estimate of this distance since the charge transfer and the contact
coupling are fairly sensitive to it and cannot be chosen whimsically. Small
variations of a few per cent away from the optimum contact distance increase
the energy of the contact by a large amount. In particular, a decrease of the
contact distance by a few tenths of an A, which might improve the contact
transmission, is energetically prohibited. Four resonances can be seen in
the conductance around the Fermi energy (set to zero) in Fig. 19. These
resonances can be easily traced back to four extended states of the isolated
finite CNT[69]. Two of them (A;i, fo) originate in the bonding (TT) band of the
CNT and the other two (fc*,^) in the antibonding (TT*) band (see inset in
Fig. 19). The resonances have different widths for different bands indicating
that they couple very differently to the electrodes. Moreover, the two bands
do not mix with each other. This is more clearly seen in Fig. 19(b) where we
show the highest transmission eigenvalues of the transmission matrix. Two
independent channels exhibit resonances in the energy window (« 3.5eV)
around Ep where only the TT and TT* bands can contribute to transport. This
result is consistent with the fact that TT* states, of large angular momentum,
do not couple to the low-angular momentum states of the electrode, while
7T states, of low angular momentum, couple more easily[65, 66]. Notice
that there is a charge transfer from the metal to the CNT, but this mainly
localizes at the end carbon layer (fa 0.2 electrons per carbon atom) and it
does not affect the overall band positioning in the center of the CNT.
The specific band origin of the resonances is nicely confirmed by their
evolution on the length of the CNT presented in Fig. 20. We present the
conductance for TV = 8,9,10,11,12, and 13 carbon-layer CNT's. The oppo-
site signs of the group velocity for the TT and n* bands make the quasi-bound
states belonging to the TT* band shift down in energies while those belonging
to the 7T band shift up as N increases. As expected from a simple particle-
in-a-box argument applied to finite CNT's[69], for N — 3Z, where I is an
integer, we should expect two states with the same wave vector kn but in
different bands to coincide at the Fermi energy. Naively one should thus
expect Q = 4e2/h[70}. Our results for the contacted N = 9 and N = 12
CNT's show otherwise: Two resonances never coincide at the Fermi level.
The reason is that Coloumb blockade prevents two (band and/or spin) de-
generate quasibound states to be filled up at the same time and degeneracies
are removed. From Figs. 20(b) and (e) we estimate the charging energy to
be w 0.3 eV in these CNT's which is smaller than the single-particle level
spacing as confirmed by experiments [63]. We have analyzed this Coulomb
blockade effect in detail for the N = 9 CNT. For a partially discharged CNT
the two resonances labeled fci and kl coincide in energy above the Fermi en-
ergy and the conductance reaches there Ae2/h (see Fig. 21). For the neutral
system this degeneracy is partially removed and the conductance drops.
The spin degeneracy removal due to Coulomb blockade requires technically
Figure 20: Conductance as a function of energy for an N = 8(a), N = 9(b),

N = 10(c), N = ll(d), N = 12(e), and AT = 13(f) (5,5) open metallic
nanotube end-contacted to Al(lll) surfaces [see Fig. 18(a)]. The Fermi
energy
3, , , , 1 , , ,
°-2 - 1 0 1
Energy (eV)
Figure 21: Conductance as a function of energy for an N = 9 (5,5) open
metallic nanotube end-contacted to Al(lll) surfaces [see Fig. 18(a)] for dif-
ferent values of the Fermi energy or charge of the system Q=-3,-2,-1,0,1,2,3.
The Fermi energy has been always set to zero so that curves shift downward
as Q increases.
Figure 22: Same as in Fig. 19, but for an N = 15 (5,5) open metallic
nanotube side-contacted to Al(lll) surfaces [see Fig. 18(b)].
challenging open shell calculations and is currently under study.

If the interpretation of the different coupling strengths of the CNT bound
states with the Al electrodes is correct and angular momentum considera-
tions are relevant, similar couplings should be expected for both bands if
no axial symmetry is present. This is the case for the other contact geom-
etry considered in this work [see Fig. 18(b)]. Figure 22 shows results for
an AT = 15 CNT side-contacted to Al(lll) surfaces (the CNT-surface dis-
tance has been optimized to 2.3A). Conductance resonances come in pairs
in the relevant energy window which is what is expected for an N = 15
CNT. More importantly, all of them present similar widths, confirming our
expectations. Contrary to the previous geometry, localized end states[69]
influence the coupling around leV for this contact geometry where mixing
with the CNT extended states takes place. Our results for the coupling
strength with Al contacts are consistent with previous studies where jellium
models were considered as contacts[66], and with those in Ref. [19], but we
do not subscribe previous ab-initio results presented in Ref. [71] based on
what it seems to be more realistic contact models similar to ours.
We now complete our study for end-contacted CNT's considering Ti, and
Pd electrodes (see Figs. 24 and 23). Pd behaves very much like Al, except
that the coupling to the CNT is clearly stronger and the overall conduc-
nanotube end-contacted to Pd(lll) surfaces (see inset).
nanotube end-contacted to Ti(lll) surfaces (see inset).
tance higher. Apart from this difference in the coupling strength, the bands
do not mix and couple, as well, very differently to the electrode [see Fig.
23(b)]. In contrast to Al and Pd electrodes, where Q exhibits resonances, Q
presents an oscillatory behavior for Ti around E-p [see Fig. 24(a)], with the
two bands exhibiting similar coupling to the electrodes. This is accompanied
by band mixing as the anticrossings in the transmission eigenvalues reveal
in Fig. 24(b). This is, to our knowledge, the first numerical evidence of what
has been hinted at on the basis of indirect first-principles calculations[68]
and what has recently been observed in experiments[64]: Early 3-d elements
as Ti are probably the best choice for making high-transparency contacts
to CNT's compared to more traditional metals such as Al and Pd (appar-
ently due to the presence of d-states at the Fermi energy). Although perfect
transparency at the contact is never achieved, our calculations indicate that
properly engineered Ti contacts are a good bet for future perfect contacts to
CNT's. At this point, however, we can only speculate on the possibility of
perfect transparency for other Ti electrode geometries. The optimal contact
distance is similar to that of Al but the charge transfer is « 0.4 electrons
per C atom at the end layers. There seems to be a correlation between
charge transfer and contact transparency if we compare Al and Ti, but this
is no longer the case when comparing Al and Pd and we cannot draw a clear
correlation between these two quantities at the present time.
5 Open issues
We would not like to finish this review without enumerating some of the
issues in nanoscale and molecular tranport that we have not addressed here
and that are still open (maybe we should say even more open than those
discussed in these notes):
• Electron phonon coupling. A number of effects derived from the

interaction between the electronic and nuclear degrees of freedom has
been observed in a number of cases[61] and it is nowadays the subject
of experimental and theoretical scrutinity[72]. This is closely related
to how inelastic effects modify elastic transport and how they can be
included in Landauer's formalism.
• Structural relaxation in a non-equilibrium situation. The way

a current can modify the atomic structure of the nanoscopic region that
sustains it has been only recently addressed[73, 74] and, although, it
seems to have a small effect, it is no clear whether or not it can be
completely neglected from the overall picture.
• Beyond a single-particle description. Present theoretical work re-

lies heavily on the single-particle description of the electronic structure.
Many-body physics like the Rondo-effect recently observed in single

molecules [75] cannot be addressed within these simplified methodolo-
gies.
6 Acknowledgements
Financial support from the Spanish MCYT (grants BQU2001-0883, PB96-
0085, and MAT2002-04429-C03), FEDER european funds, and the Univer-
sidad de Alicante is gratefully acknowledged.
A Bethe lattices
In this appendix we discuss how selfenergies for Bethe Lattices (BL) used
to describe the leads are calculated. A BL is generated by connecting a site
with N nearest-neighbors in directions that could be those of a particular
crystalline lattice. The new N sites are each one connected to N — 1 different
sites and so on and so forth. The generated lattice has the actual local
topology (number of neighbors and crystal directions) but has no rings and,
thus, does not describe the long range order characteristic of real crystals.
Let n be a generic site connected to one preceding neighbor n — 1 and N — 1
neighbors of the following shell (n + i with i — l,..,N — 1). Dyson's equation
for an arbitrary non-diagonal Green function is written as,
(EI-Eo)G n , f c =V B , B _iG n _ u + Yl V «,* G a (20)

i=l,...,JV-l
where A; is an arbitrary site, E the energy, and Vjj is a matrix that incorpo-
rates interactions between orbitals at sites i and j (bold capital characters
are used to denote matrices). Eo is a diagonal matrix containing the orbital
levels and I is the identity matrix. Then, we define a transfer matrix as
Ti_i,iGj_ij = Gij (21)
Multiplying Eq. (1) by the inverse of G n _i i n we obtain,
(El - E 0 )T n _ l i n = V n , n _! + I Yl Vn<iTn>i J T B _i, n (22)

\»=1,...,JV-1 /
Due to the absence of rings the above equation is valid for any set of lattice
sites, and, thus, solving the Bethe lattice is reduced to a calculation of a few
transfer matrices. Note that a transfer matrix such as that of Eq. (2) could
also be defined in a crystalline lattice but, in that case it would be useless.
Eq. (3) can be solved iteratively,

-i - l
T B _i, n = El-Eo- Yl ^n,iTn,i Vn.n-1 (23)

i=l,...,JV-l
If the orbital basis set and the lattice have full symmetry (including inversion
symmetry) the different transfer matrices can be obtained from just a single
one through appropriate rotations. However this is not always the case (see
below).
Before proceeding any further we define selfenergies that can be (and
commonly are) used in place of transfer matrices,
Sij = V y T y (24)
Eq. (4) is then rewritten as,
-i -i
£«-!,« = VB_i,B EI-Eo Y, S ».' V»-i,» W

i=l,...,N-l
where we have made use of the general property V n ) n _i = Vn_i „•

As discussed hereafter, in a general case of no symmetry this would
be a set of N coupled equations (2iV if there is no inversion symmetry).
Symmetry can be broken due to either the spatial atomic arrangement, the
orbitals on the atoms that occupy each lattice site, or both. When no
symmetry exists, the following procedure has to be followed to obtain the
selfenergy in an arbitrary direction. The method is valid for any basis set
or lattice. Let T\ be the N nearest-neighbor directions of the lattice we are
interested in and Vn the interatomic interaction matrix in these directions.
To make connection with the notation used above note that the vector that
joins site n - 1 to site n, namely, r n - r n _ ! would necessarily be one of the
lattice directions of the set 71. The selfenergies associated to each direction
have to be obtained from the following set of 2N coupled selfconsistent
equations,
S Ti = V n [El - E o - (ET - E ^ ) ] - 1 V*, (26)
S n = V n [El - E 0 - ( E T - S T i )]" : V t , (27)
where i = 1,..., N and fj = — TJ. VTi is the interatomic interaction in the Tj
direction, and S x and S ^ are the sums of the selfenergy matrices entering
through all the Cayley tree branches attached to an atom and their inverses,
respectively, i.e.,
N
Sr = £ S T i (28)
Sf = E s ^ - (29)
»=1
This set of 2N matricial equations has to be solved iteratively. It is straight-
forward to check that, in cases of full symmetry, it reduces to the single
equation. The local density of states can be obtained from the diagonal
Green matrix,
P j
Gn,n= £ I - E 0 - Y, S n (30)
t=l,..,JV
Finally, as regards the tight-binding parameters, which should include

only nearest-neighbors interactions, there is no unique way of choosing them.
The ones we generically use were obtained through fittings to the electronic
bulk band structures[23], but the reader is free to choose any other criterion
that fits better the problem at hand.
B Green functions and non-orthogonal basis sets

In this appendix we briefly discuss how Eqs. 6 and 7 for the density matrix
and the total number of particles in a nonorthogonal basis are derived.
Working with a nonorthogonal basis requires the use of the metrics
£ > Q > S^1 < %| =/• (31)

a/3
Changing to an orthogonal basis \<j>a > can be done through the equation
|^>=^5^1/2|^> (32)
0
Then an expression which allows to obtain the matrix elements in an or-

thogonal basis of a general matrix A from those in the non-orthogonal basis,
can be easily derived,
Ao = S~1/2ANoS-1/2 (33)
Using the above metrics and the definition of the Green's function operator,
which includes the operator selfenergies,
[(£ ± iS)I -F- tf} - E^}] (?W = /, (34)

Eq. (5) for the matrix elements of the Green's function in a nonorthogonal
basis inmediately follows.
The local charge density is written in terms of the Green's function as [29]
p(r) = - - f F Im < r|G (+) (£)|r >

K J-oo
= - I fEFlmJ2Mr)S^G{+\E)S^;(r), (35)
J-°° afrS
while the total number of particles is in its turn given by [33]
N = f drp(r) = Y, Sa0Pap = Tr [P • S]. (36)

J 0a
References
[1] A. Nitzan and M. A. Ratner. Electron transport in molecular wire
junctions. Science, 300(5624): 1384-1389, 2003.
[2] A. Aviram and M. A. Ratner. Chem. Phys. Lett, 29:277, 1974.
[3] M. A. Reed, C. Zhou, C. J. Muller, T. P. Burgin, and J. M. Tour.
Science, 278:252, 1997.
[4] B. Mann and H. Kuhn. J. Appl. Phys., 3:4398, 1971.
[5] C. Dekker. Physics Today, 52:22, 1999.
[6] N. Agrait, A. Levy Yeyati, and Jan M. van Ruitenbeek. Physics Reports,
377:81-279, 2003.
[7] N. Agrait, J. C. Rodrigo, and S. Vieira. Phys. Rev. B, 47:12345, 1993.
[8] N. Agrait, G. Rubio, and S. Vieira. Phys. Rev. Lett, 74:3995, 1995.
[9] S. Datta. Electronic transport in mesoscopic systems. Cambridge Uni-
versity Press, Cambridge, 1995.
[10] Mukunda P. Das and Frederick Green. In J. da Providencia, editor, Pro-
ceedings of XXVI International Workshop on Condensed-Matter Theo-
ries. Nova Science, New York, 2003.
[11] J. A. Torres and Saenz J. J. Phys. Rev. Lett, 77:2245-2248, 1996.
[12] A. Garcia-Martin, J. A. Torres, and J. J. Saenz. Phys. Rev. B, 54:13448-
13451, 1996.
[13] Y. Meir and N.S. Wingreen. Phys. Rev. Lett, 68:2512, 1992.
[14] J. C. Cuevas, A. Levy-Yeyati, and A. Martm-Rodero. Phys. Rev. Lett.,

80:1066, 1998.
[15] A. Hasmy, E. Medina, and P. A. Serena. Phys. Rev. Lett, 86:5574,
2001.
[16] N. D. Lang. Phys. Rev. B, 52:5335, 1995.
[17] K. Hirose and M. Tsukada. Phys. Rev. B, 51:5278, 1995.
[18] M. Brandbyge, J. L. Mozos, P. Ordejon, J. Taylor, and K. Stokbro.
Phys. Rev. B, 65:165401, 2002.
[19] J. Taylor, H. Guo, and J. Wang. Phys. Rev. B, 63:245407, 2001.
[20] M. Di Ventra, S. G. Kim, S. T. Pantelides, and N. D. Lang. Phys. Rev.
Lett, 86:288, 2001.
[21] M. J. Frisch, G. W. Trucks, H. B. Schlegel, M. A. Robb G. E. Scuse-
ria, J. R. Cheeseman, V. G. Zakrzewski, J. A. Montgomery, Jr., R. E.
Stratmann, J. C. Burant, S. Dapprich, J. M. Millam, A. D. Daniels,
K. N. Kudin, M. C. Strain, O. Farkas, J. Tomasi, V. Barone, M. Cossi,
R. Cammi, B. Mennucci, C. Pomelli, C. Adamo, S. Clifford, J. Ochter-
ski, G. A. Petersson, P. Y. Ayala, Q. Cui, K. Morokuma, D. K. Malick,
A. D. Rabuck, K. Raghavachari, J. B. Foresman, J. Cioslowski, J. V.
Ortiz, A. G. Baboul, B. B. Stefanov, G. Liu, A. Liashenko, P. Piskorz,
I. Komaromi, R. Gomperts, R. L. Martin, D. J. Fox, T. Keith, M. A. Al-
Laham, C. Y. Peng, A. Nanayakkara, C. Gonzalez, M. Challacombe,
P. M. W. Gill, B. Johnson, W. Chen, M. W. Wong, J. L. Andres,
C. Gonzalez, M. Head-Gordon, E. S. Replogle, and J. A. Pople. GAUS-
SIAN98, Revision A.7, Gaussian, Inc., Pittsburgh PA, 1998.
[22] J. J. Palacios, A. J. Perez-Jimenez, E. Louis, and J. A. Verges. Phys.

Rev. B, 64:115411, 2001.
[23] J. J. Palacios, A. J. Perez-Jimenez, E. Louis, E. SanFabian, and J. A.
Verges. Phys. Rev. B, 66:035322, 2002.
[24] J. J. Palacios, A. J. Perez-Jimenez, E. Louis, and J. A. Verges. Nan-
otechnology, 12:160, 2001.
[25] E. Louis, J. A. Verges, J. J. Palacios, A. J. Perez-Jimenez, and E. San-
Fabian. Phys. Rev. B, 67:155321, 2003.
[26] P. S. Damle, A. W. Ghosh, and S. Datta. Phys. Rev. B, 64:201403,
2001.
[27] Y. Xue, S. Datta, and M. A. Ratner. J. Chem. Phys., 115:4292, 1991.
[28] J. Heurich, J. C. Cuevas, W. Wenzel, and G. Schon. Phys. Rev. Lett.,

88:256803, 2002.
[29] E. N. Economou. Green's functions in Quantum Physics. Number 7 in

Springer Series in Solid State Physics. Springer, 1970.
[30] J. Taylor, H. Guo, and J. Wang. Phys. Rev. B, 63:121104, 2001.
[31] Y. Fujimoto and K. Hirose. Phys. Rev. B, 67:195315, 2003.
[32] N. D. Lang and M. DiVentra. arXiv:cond-mat/0307442.
[33] A. Szabo and N. S. Ostlund. Modern Quantum Chemistry. McGraw-

Hill, New York, 1989.
[34] L. F. Pacios and P. A. Christiansen. J. Chem. Phys., 82:2664, 1985.
[35] M. M. Hurley, L. F. Pacios, P. A. Christiansen, R. B. Ross, and W. C.

Ermler. J. Chem. Phys, 84:6840, 1986.
[36] R. B. Ross, J. M. Powers, T. Atashroo, W. C. Ermler, L. A. LaJohn,

and P. A. Christiansen. J. Chem. Phys., 93:6654, 1990.
[37] P. Jelinek, Ruben Perez, Jose Ortega, and Fernando Flores. arXiv:cond-
mat/0211488.
[38] R. H. M. Smit, Y. Noat, C. Untiedt, N. D. Lang, M. C. van Hemert,

and J. M. van Ruitenbeek. Nature (London), 419:906, 2002.
[39] R. H. M. Smit. PhD thesis, Universiteit Leiden, 2003.
[40] F. Pauly J. Heurich, J. C. Cuevas, W. Wenzel, and G. Schon.

arXiv:cond-mat/0211635.
[41] L. G. C. Rego, A. R. Rocha, V. Rodrigues, and D. Ugarte. arXivxond-

mat/0206412.
[42] Y. Noat S. K. Nielsen, M. Brandbyge, R. H. M. Smit, K. Hansen,

L. Y. Chen, A. I. Yanson, F. Besenbacher, and J. M. van Ruitenbeek.
[43] G. Binnig, H. Rohrer, Ch. Gerber, and E. Weibel. Appl. Phys. Lett,
40:178, 1982.
[44] J. Tersoff and D. R. Hamann. Phys. Rev. Lett, 50:1998, 1983.
[45] J. Tersoff and D. R. Hamann. Phys. Rev. B, 31:805-813, 1985.
[46] J. Bardeen. Phys. Rev. Lett, 6:2654, 1961.
[47] N. D. Lang. Phys. Rev. B, 34:5947-5950, 1986.

[48] C. Rogero, J. I. Pascual, J. Gomez-Herrero, and A. M. Baro. J. Chem.

Phys., 116:832-836, 2002.
[49] V. Meunier and Ph. Lambin. Phys. Rev. Lett, 81:5588-5591, 1998.
[50] C. Joachim, J. K. Gimzevski, R. R. Schlittler, and C. Chavy. Phys.
Rev. Lett, 74:2102-2105, 1995.
[51] A. J. Perez-Jimenez, J. J. Palacios, E. Louis, E. SanFabian, and J. A.
Verges. CHEMPHYSCHEM, 4, 2003.
[52] R. J. Wilson, G. Meijer, D. S. Bethune, R. D. Johnson, D. D. Chambliss,
M. S. de Vries, H. E. Hunziker, and H. R. Wendt. Nature, 348:621-622,
1990.
[53] E. I. Altman and R. J. Colton. Surf. ScL, 279:49-67, 1992.
[54] E. I. Altman and R. J. Colton. Surf. ScL, 295:13-33, 1993.
[55] E. I. Altman and R. J. Colton. Phys. Rev. B, 48:18244-18249, 1993.
[56] D. Fujita, T. Yakabe, H. Nejoh, T. Sato, and M. Iwatsuki. Surf. Sci.,
366:93-98, 1996.
[57] C.-T. Tzeng, W.-S Lo, J.-Y Yuh, R.-Y. Chu, and K.-D. Tsuei. Phys.
Rev. B, 61:2263-2272, 2000.
[58] T. Hashizume, K. Motai, X. D. Wang, H. Shinohara, Y. Saito,
Y. Maruyama, K. Ohno, Y. Kawazoe, Y. Nishina, H. W. Pickering,
Y. Kuk, and T. Sakurai. Phys. Rev. Lett., 71:2959-2962, 1993.
[59] T. R. Ohno, Y. Chen, S. E. Harvey, G. H. Kroll, J. H. Weaver, R. E.
Haufler, and R. E. Smalley. Phys. Rev. B, 44:13747-13755, 1991.
[60] S. J. Chase, W. S. Bacsa, M. G. Mitch, L. J. Pilione, and J. S. Lannin.
Phys. Rev. B, 46:7873-7877, 1992.
[61] H. Park, J. Park, A. K. L. Lim, E. H. Anderson, A. P. Alivisatos, and
P. L. McEuen. Nature (London), 407:57-60, 2000.
[62] S. Frank, P. Poncharal, Z. L. Wang, and W. A. de Heer. Science,
280:1744, 1998.
[63] Hyongsok T. Soh, Calvin F. Quate, Alberto F. Morpurgo, Charles M.
Marcus, Jing Kong, and Hongjie Dai. Appl. Phys. Lett, 75:627, 1999.
[64] J. Kong, E. Yenilmez, T. W. Tombler, W. Kim, H. Dai, R. B. Laughlin,
L. Liu, C. S. Jayanthi, and S. Y. Wu. Phys. Rev. Lett, 87:106801, 2001.
[65] H. J. Choi, J. Ihm, Y.-G. Yoon, and S. G. Louie. Phys. Rev. B,

60:R14009, 1999.
[66] M. P. Anantram. Appl. Phys. Lett, 78:2055, 2001.
[67] A. N. Andriotis, M. Menon, and G. E. Proudakis. Appl. Phys. Lett.,
76:3890, 2000.
[68] C. K. Yang, J. Zhao, and J. P. Lu. Phys. Rev. B, 66:041403, 2002.
[69] A. Rubio, D. Sanchez-Portal, E. Artacho, P. Ordejon, and J. M. Soler.
Phys. Rev. Lett, 82:3520, 1999.
[70] D. Orlikowski, H. Mehrez, J. Taylor, H. Guo, J. Wang, and C. Roland.
Phys. Rev. B, 63:155412, 2001.
[71] M. B. Nardelli, J. L. Fattebert, and J. Bemholc. Phys. Rev. B,
64:245423, 2001.
[72] Yu-Chang Chen, Michael Zwolak, and Missimiliano Di Ventra.
[73] Z. Yang and M. Di Ventra. Phys. Rev. B, 67:161311, 2003.
[74] M. Brandbyge, K. Stokbro, J. Taylor, J. L. Mozos, and P. Ordejon.
Phys. Rev. B, 67:193104, 2003.
[75] J. Park, A.N. Pasupathy, J. I. Goldsmith, C. Chang, Y. Yaish, J. R.
Petta, M. Rinkoski, J. P. Sethna, H. Abruna, P. L. McEuen, and D. C.
Ralph. Nature (London), 417:722-725, 2002.
Chapter 2: Molecular Dynamics Simulations of Single
Molecule Atomic Force Microscope Experiments
Wiestew Nowakab and Piotr E. Marszatekb
"Institute of Physics, Nicholaus Copernicus University Grudziadzka 5,

87-100 Torun, Poland
hDepartment
of Mechanical Engineering and Materials Science
CBIMMS, Duke University, Durham, 27708 NC, USA
Abstract
The ultimate goal of nanosciences is to develop tools for spatial and chemical
control of single molecules. A number of experimental techniques have
emerged in recent years which allow for mechanical manipulations of single
molecules including single molecule fluorescence, optical and magnetic
tweezers, atomic force spectroscopy. Proper understanding of experimental
results requires adequate theory. In this paper we present an overview of
recent advances and trends in applications of computer simulations aimed at
understanding of atomic force microscope (AFM) experiments. Special
attention is devoted to those variants of classical molecular dynamics (MD)
simulations, such as steered MD or biased MD, which are particularly helpful
in interpretation of experimental data obtained for single biopolymer
molecules. A short meta-review (review of recent reviews) of advances in the
MD technology is also presented.
47
48 W. Nowak and P. E. Marszalek
Glossary of Terms
AFM Atomic Force Microscope
ATP Adenosine TriPhosphate
ADP Adenosine DiPhosphate
BMD Biased Molecular Dynamics
BPTI Bovine Pancreatic Trypsin Inhibitor
CHARMM Chemistry at Harvard Macromolecular Modeling
DFT Density Functional Theory
DNA Deoxyribonucleic Acid
FF Force Field
FN Fibronectin
FPS Force Probe Microscopy
GTP Guanosine TriPhosphate
IMD Interactive Molecular Dynamics
Ig Immunoglobulin
LCD Liquid Crystal Device
MD Molecular Dynamics
MscL Mechanoselective Ion Channel
NAMD2 Name of Molecular Dynamics Code
PC Personal Computer
PMF Potential of Mean Force
RC Reaction Coordinate
REMD Random Expulsion Molecular Dynamics
QM/SMD Quantum Mechanical Steered Molecular Dynamics
QSAR Qualitative Structure Activity Relationship
RNA Ribonucleic Acid
SCC-DFTB Self Consistent Charge Density Functional Theory
based Tight Binding Method
SMD Steered Molecular Dynamics
TMD Targeted Molecular Dynamics
UV Ultra Violet
WHAM Weighted Histogram Analysis Method
1. Introduction
Single molecule experiments are somehow at variance with the idea of
"standard" chemistry that typically deals rather with myriads of molecules
than a single one (Avogadro number is ~1023). However, one should
remember, that the ultimate limit of miniaturization is perhaps at a single
Molecular Dynamics Simulations 49
molecule level.1 The quest for better and better computers, digital memories,
transducers, sensors, etc. is the driving force of the nanotechnological
revolution. Clearly, single molecule manipulations and chemistry performed
on a couple of molecules are in focus of many research labs' activities.
There are numerous techniques used to observe and manipulate single
molecules. Description of those experimental approaches and new results were
subject of many reviews and will not be repeated here. Optical methods of
fluorescence2"4 based on scanning confocal microscopy, near field scanning
optical microscopy and fluorescence correlation spectroscopy5"7 are very
popular in biologically oriented studies.8"10 Optical tweezers,11"13 where light is
used to manipulate molecules attached to dielectric beads,10 came a long path
from the studies of single myosin molecules binding to a single actin
filament14 to the observation of DNA packing inside lambda phage.15'16
Sometimes magnetic tweezers, known for longer time,17 are better suited for
performing advanced research on single molecules.18'19
The first single-molecule AFM experiments, initiating the field of force
probe spectroscopy (FPS), were performed in 1994 by E.-F. Florin, V. Moy
and H. Gaub.5 In their experiments the unbinding force of the individual ligand
- receptor pair in the important biotin-streptavidin complex was measured
using Atomic Force Microscope (AFM). The principle of FPS is rather simple:
a sharp tip of the AFM cantilever is attached to a single molecule while the
other end of the molecule is firmly attached to a moving support. The motion
of this support is controlled by the application of voltage to the piezo device.
The elongation of the molecule requires a force and the cantilever bends under
this force. Its minute deformation is usually detected by measuring changes in
the intensity of the laser beam reflected from the tip part of the cantilever and
projected onto a split photodiode. Since a proper calibration procedure allows
for a precise determination of the cantilever force constant, and the position of
the piezo stage is strictly controlled, a dependence of the molecule size with
respect to the applied force may be determined with high accuracy.
The basic principles of AFM experiments are presented in Fig. 1

50 W. Nowak and P. E. Marszatek
Principles of tinglt^molerule fartv <pectroscopy. A, Schematic of an atomicfarcemicroscope. The spring constant kc of each individual
AFMcantil'~\cr lS:j.Si,';'; '-'jht'Fal--J in solution, just before each experiment (10-20 mN/m). B. A polymer chains is attached at one end to
thepiezo actuator and at the other end to the force sensor (cantilever). When thepiezo moves down from position 1 to position 2 the
polymer is extended and this extension generates an enlropic restoring force, F, which bends the cantilever. The force is measured by the
deflection of the cantilever (F=kt&zJ and the extension of the molecule is calculatedfrom thepiezo travel &ZP. as x-Lp-Li where LF is the
final, and L, initial length of the polymer. C Force spectrogram of a single cellulose chain (1.4 ft-linked polyglucose). Stretching
cellulose with AFM requires force, F, which is monitored as a deflection of the cantilever, Azr .
Fig. 1. The principle of AFM spectroscopy.
The AFM was used, for example, in studies of unfolding single

molecules20 composed of titin21'22 or fibronectin23 domains and in the
measurement of the elasticity of various polysaccharides. Since piconewton
forces are registered, the technique is sensitive enough to study such basic
chemical processes as chair-boat transitions in glucose24"25 and other
mechanically induced conformational changes26"28 in biopolymers29 or on their
surfaces.30 The AFM applications in nanotechnology are enormous and range
from DNA based ink31 to MILIPEDE - an innovative idea of a terabyte
storage device based on the array of AFM nanotips32.
<£•••"'•'• v '- 4
_^JL&~^~** —«*-« • W~;"

100 Ml
Carrion-Vazquez, Oberhauser, Fisher, Marszalek. Li & Fernandez. (2000}

Prog. Bicphys. MoL Bid 74, 63-91
Fig. 2. An example of a force spectrogram of a protein composed of many identical

repeats of the 127 domain of titin (A). Consecutive peaks on force - extension plot
correspond to unfolding of individual Ig domains (B). The detailed analysis of peak
shapes and SMD simulations reveal that main unfolding events are related to
concerted rupture of several hydrogen bonds between A'-G |3 stands of Ig domains of
titin (C). (Fig. 2 is reprinted from ref. [104], © 2000, with permission from Elsevier.)
The reader interested in experimental details33 is referred to the overview

of AFM studies of titin and polysaccharides and application of this technique
for chemical analysis purposes, presented by Fisher et al.26'34' An excellent
and comprehensive review of experimental biophysical techniques used for
stretching single molecules of DNA35, RNA and proteins has been published
quite recently36. Force spectroscopy of single-biomolecules was also described
by Gaub et al. in two short reviews.37'38 Very recent results on single molecule
studies of protein folding which combine fluorescence and force spectroscopy
were summarized by Rief et al.39'40 A comprehensive review of the
measurements of protein-protein interaction forces was published by
Leckband41 and instrumentation hints were given by Reiner et al.42
Methodological issues on using AFM in protein folding research were
discussed by Best et al.,43 in biochemistry by Ikai et al.44 and in single bond

detection by Williams.45
2. Meta-review on molecular dynamics

Huge progress in computer technology brought MD simulations of
biomolecules to nanosecond and even microsecond time regime. Many
excellent books and papers cover the progress in methodology of molecular
simulations and recent progress in such studies. Thus, here we present only a
brief listing of available readings which indicate good sources of information
on particular topics.
The classical book by Haile,46 explains in detail the technical background
of molecular dynamics methods. The book is rather old, so more recent books
by Frankel47 or Field48 may be recommended, as well as the comprehensive,
interdisciplinary oriented textbook by Schlick,49 written primarily for graduate
students. For those who prefer a physicists' point of view on MD, a detailed
description of algorithms by Kholmurudov et al.50 might be of great interest.
An important subject of enzymatic reactivity and its theoretical description in
the framework of stochastic methods and nonlinear dynamics was recently
covered by authors contributing to the book edited by Werner Ebeling et al.51
In addtion to monographs or textbooks, there are also numerous journal
review articles describing the applications of computer modelling to chemistry
and biology. Perhaps all researchers active in the field would appreciate the
historic overview of MD development presented by Karplus.52 Van Gunsteren
gave a personal perspective on the recent progress (2002) in molecular
dynamics simulations.53 The emerging and fast growing field of "ab initio"
molecular dynamics calculations was described in detail by Hutter.54 Parinello
et al.55 focused their description of the "ab initio" MD field on applications of
the popular Carr-Parrinello method. Karplus and McCammon published a brief
account on the role of the modeling of flexbility in uderstanding functions of
biomolecules.5 Goodfellow at al.57 presented the MD methodology in simple
terms and put forward some representative examples of applications. In
particular, a review of studies of medically important proteins, such as p53 and
HIV related, is included in this paper, along with research on protein folding-
unfolding phenomena, ligand-target interactions, and a brief description of the
methods of analysis of the MD results.
Protein folding has been studied quite extensively in recent years, since it
is perhaps one of the greatest unsolved general problems of structural
molecular biology. Among numerous reviews of this subject, the most recent
onces were published by Karplus et al.58'59 Vendruscolo and Paci discussed the
interplay between theoretical approaches and folding experiments.60 Fersht and

Daggett gave a detailed, atomistic picture of folding (and unfolding) of
selected proteins from ns to microsecond.61'62 The search for a unifying
mechanism of protein folding was at least partially succesful - it seems that
the preformed secondary structure elements of protein colapse to the transition
state which eventually make the final fold.63'64
Another large area of applications of MD methods are studies of
biochemical reactions.65'66 The important field of enzymatic catalysis was
critically examined by Warshel.67 Karplus tried to put forward arguments that
proteins behave differently than simple chemicals.68 Dynamical aspects of
chemical catalytical reactions may be modeled computationally,69 but the
rigorous description of the processes of creating and breaking of the chemical
bonds requires full quantum-mechanical treatment.70
Reaction kinetics is related to the changes in free energy (AF or AG) of
reactants and products. Computer methods are mature enough to give realistic
numbers of AF for many systems,71 though indeed the energy landscapes of
proteins or biopolymers are particularly complex.72 The AFM type of
experiments, in which an additional external force is added to the molecule,
make these hypersurfaces even more complex and difficult to calculate. This is
because typical parametrizations of forcefields are tuned to equilibrium
structures of the standard benchmark molecules, and these structures are
strongly disturbed by the external force.
To get reasonable convergence in studies of biologically relevant
conformations or realistic values of free energy estimates, one has to rely on
the ergodic hypothesis, i.e. to assume that a simulation is long enough to cover
all of the important regions of the phase space of the dynamical system under
study. The progress in simulation time was spectacular: from a mere 60 ps
simulations of BPTI by Levitt, done in 1983, through 5 ns simulations of
chymotrypsin inhibitor by Dagett (1995) and the famous 1 microsecond
simulation of villin fragment by Duan and Kollman in 1998, to the current
distributed computing projects, which use thousands of processors to collect
information for a single protein folding on a milisecond timescale.73 However,
long time MD simulations are still rather expensive and rare.74'75 There is a
need for approximate, but effective methods able to grasp the basic features of
conformational transition on such long times. One approach, reviewed here, is
to use an external, controlled force to brake a computer simulation time
barrier. The stochastic approach proposed by Elber is also a promising
alternative with respect to this goal.76
The scope of the applications of computational MD techniques is huge,
and encompasses different topics such as high preassure simulations of
biomolecules,77 drug design,78 QSAR application of DFT based methods in
medicinal chemistry,79 studies of metalloproteins,80 ion channels,81 membrane

transporter proteins,82 just to name a few. In this review we focus on the
possible applications of modeling in interpretation of AFM single molecule
experiments. The most recent literature (2000 - 2003) related to forced
conformational transitions will be reviewed in greater detail.
3. Variants of MD methods used in force spectroscopy of

single molecules
Molecular dynamics simulation, due to its very nature, is particulary suitable
for modeling single molecule experiments. Perhaps the first application of the
MD method in interpretation of AFM experiments on streptavidin-biotin
complex rupture forces was done by Grubmiiller et al.83 in 1996.
However, the strategy of realizing "in silico" a transition from one
predefined (and known) conformation to the other one is not new and may be
traced back, at least, to the classical reaction path studies by Elber and
Karplus84 published in 1987. The review of such computational approaches
was published by Elber85 and now this research has evolved into advanced
long time stochastic studies of biological processes.86 The original idea of
using an externally applied force to induce a "controlled" conformational
change was put forward by Schlitter et al. in 1993.87 They proposed a method
called targeted molecular dynamics (TMD) to enforce a transition between the
initial state and a target state defined as points in multidimensional space. The
first application was to the T -> R transition in hemoglobin. The TMD paved
the road to more popular techniques - steered MD (SMD)88'89 and biased MD90
(BMD).
3.1 Steered MD
The initial development of SMD was triggered by the needs of Duke
Univeristy (Durham, NC, USA) biochemist Jane Richardson, who wanted to
study "interactively" cysteine connecting loops in a model protein SScorin.
The group of J. Hermans (UNC, Chapell Hill, USA) modified their computer
codes in 1996 to make such forced manipulations on protein structure
possible.88 Next year the first paper on SMD came from the K. Schulten
group (UIUC, Illinois, USA).89 In this work the authors investigated the biotin-
avidin complex using a wide range of forces and various pulling speeds, much
more related to real AFM experiments than those used by Grubmiiller et al.83
in their seminal study of the same molecular sytem. The conclusion of this
paper89 was rather pesimistic: since simulation and experiment timescales vary
in some six orders of magnitudes, calculated forces are always too high in
comparison to experimental ones. Later Galligan et al.91 studied the same
system using Brownian dynamics and successfully reproduced experimental
logaritmic dependence of a rupture force on the loading rate.
The technique of umbrella sampling92 is commonly used to drive the
system studied by MD simulations along the selected reaction coordinate.
Adding an external potential keeps the desired value of the coordinate as long
as necessary to collect conformations required for the calculation of free
energy change (potential of mean force, PMF). The effect of external potential
should be as small as possible in order to take away its bias from the
Hamiltonian. In SMD the external force, usually time-dependent, pulls the
system along some choosen degrees of freedom. The situation now is
somewhat different: we do seek strong perturbations and large conformational
changes, with external force changing rapidly in time. In contrast to classical
free energy computations, termodynamic nonequilibrium is involved in this
type of simulations.
In SMD there are two typical protocols: SMD with constant pulling
velocity and constant force. The standard way of applying external forces is to
restrain some selected part of the system studied (for example one atom, a
ligand or a group of atoms) by a harmonic (or other) potential. The restrain
potential is localised in space, in the sense that the point in x0 e R3 is "a
source" of this potential U, and there is a certain conformation of the molecule
which gives the lowest energy for the constrained system. The harmonic
constrain may be defined as:
U = k(x-x o -d) 2 /2, (1)
where k is the stiffnes of the external spring (restraint), d = vt, and v is a

constant velocity of the motion of the restraint point. One can easily see that
the corresponding external force exerted on the system is:
F = k(xo+d-x). (2)
The meaning of this force is simple - it corresponds to a molecule being pulled

by a harmonic spring with its end moving with velocity v. This is exactly the
situation one has in AFM experiments, where the molecule is being pulled by
the tip of the cantilever.
Sometimes another variant of SMD is used. The restraint point is selected
far from the atoms of the molecule, the point remains fixed in space but the
spring constant changes linearly with time:
F = at(xo-x). (3)
There are several reviews of applications of SMD and technical details

related to the best simulation strategy published by scientists collaborating
with the Schulten group.93'94 Special attention should be devoted to a very
recent review by Tjakhorshid et al.95 The authors describe the results of the
simulations of nearly 20 molecular systems - as different as a bilayer
composed of 200 lipids, membrane-water interface, binding of estrogen
receptor to DNA, apolipoprotein, calmodulin, Rieske subunit motion in
cytochrome bcl complex,96 bacteriorhodopsin and purple membrane,
aquaporin-1, photosynthetic light harvesting system, fibronectin type IIIi0 or
CD2-CD58 complex. New results of very advanced calculations on F0-ATPase
and FpATPase complexes presented in this review are of particular interest,
since SMD type calculations were used to study these molecular engine
molecules by adding an external torque to the carefully selected parts of the
ATPase assambly.
It is interesting to note, that the hardware requirements have changed
immensly in the last 5 years: from one 4-processor HP computer used in early
SMD studies to two 512-processor clusters applied recently to 327 K atoms of
the F,-ATPase complex.95 With an inexpensive (below $20 K ) home built 16-
processor linux cluster and free academic software, one can run SMD
simulations for 30-40 K atoms systems at a reasonable rate of 1-2 ns/day. With
the advent of LCD displays having stereo capabilities, perhaps new tools of 3-
D modeling, similar to virtual labs proposed by Prins et al.,97 will allow for
collective work in virtual reality using, for example, haptic devices. Such a
device, within the framework of interactive molecular dynamics98 (IMD) was
successfully applied in studies of the GlpF membrane channel and glycerol
kinase." However, as far as we know, no SMD studies based on a virtual
reality lab and related to AFM experiments were reported so far.
Problems with matching SMD and AFM timescales may be aleviated by
calculating the potential of mean force from SMD trajectories. Gilllingsrud et
al.' 00 presented three time series analysis techniques: based on the weighted
histogram analysis method (WHAM), van Kampen's Q-expansion and
minimization of the Onsager-Machlup action. As an example, the PMF of
extraction of lipids from a membrane was calculated.l00'10'
3.1.1 Titin
It seems that a huge (1 urn long) muscle protein titin is the favorite object of
both AFM and SMD single molecule studies. Titin is responsible for passive
elasticity of muscle. It consists of hundreds of immunoglobulin-like domains
(Ig, seven-stranded P-sandwich) and fibronectin (FN) domains and less

structured sequences, such as PEVK. The Gaub group102 studied single titin
molecules using AFM, and they have obtained a picture of the unfolding of
individual domains. The Fernandez group, investigated the dependence of
unfolding forces in titin models upon stretching velocity1 The mechanism of
Ig domains unfolding involves rupture of two sets of hydrogen bonds.104
Mutagenesis work proved this hypotesis, by demonstrating that the "hump"
which is clearly visible in the first peak of the force spectrogram of the 127
module is absent in its K6P mutant. The mutant is not able to make the first set
of hydrogen bonds that link A and B strands of I2721'105 (see Fig. 3).
1279-K6P mutant polyprotein
1000 I i
800 I
g- 600 /
-200 \ f
0 20 40 60 80 100 120 140 160 180 200
Extension (nm)
Mamzalek, Lu, Li, Carrion-Vazquez, Oberhauser, Schulten, Fernandez (1999).

Nature 402, 100-103.
Fig. 3. Experimental force-extension curve of the engineered polyprotein composed of

9 repeats of the K6P mutant of the 127 domain of cardiac titin. Reproduced with
permission of ©Nature Publishing Group.
The interpretation of AFM experiments as sequential unfolding of

consecutive domains was supported by SMD simulation by Lu et al., which
was performed for a single 127 domain of titin in water.106 Two sets of
hydrogen bonds were indicated in the domain: while A-B ani-parallel strands
could be easily separated (force <50 pN), the parallel A'-G stands could be
ruptured in a concerted way only under the action of a large force.107 Later on,
it has been found that this key event of 127 unfolding must be accompanied by
extensive rearrangements of water molecules. These events impose conditions
on the time length of SMD simulations with explicit solvent. m > m - ]10
Further studies (2001) performed by the Schulten group were devoted to

the mechanism of refolding of the single 127 domain. During 2 ns SMD
simulations two refolding patterns have been observed: seperated A'-G
strands reformed the majority of their native hydrogen bonds, while for A and
B strand 2 ns time was too short to recover their hydrogen bonding.''' Since
not all Ig units of titin are identical, SMD was used to study the differences in
the unfolding mechanism between some domains, namely II and I27.112 SMD
suggests that, in contrast to 127, II does not have unfolding intermediates.
Details of folding-refolding dynamics of the II domain were described by Gao
et al. in a separate paper.113 Interestingly, in this work the important
stabilizing effect of the disulfide bridge was easily discovered by modeling
oxidized and reduced forms of the protein. The S-S bridge limits the maximum
extension of the II titin unit. However, very recent AFM results published by
Li and Fernandez114 put in doubt the special role of S-S bridges in II modules
of titin.
More analytical models of protein elasticity probed by AFM and related to
Ig domains were presented by Zhang and Evans.115 Best et al.116 used
engineered titin modules to test their method of searching for the unfolding
mechanism. Some summary information on titin SMD studies may be found
in the paper by Isralevitz et al.' 17
Very recently Best et al. have claimed,118 based on MD simulations and
analysis of protein mutagenesis data, that a stable intermediate is indeed
involved in 127 unfolding,119 but it is related rather to side chain interactions
than to hydrogen bond ruptures, as was previously assumed. The unfolding
intermediate has also been found in AFM experiments by Schweiger et al.120
done for the Ig rod domain of filamin from Dictyostelium discoideum. The
report on AFM experiments performed on the isolated, huge titin molecule
was published by Kellermayer et al.121
3.1.2 Fibronectin and other elastic proteins

Fibronectin (FN) is another interesting protein extensively studied both by the
AFM and MD methods. It is found in the extracellular matrix in all
vertebrates. Elastic FN fibrils connect cells, they may stretch four times as
long as their relaxed length, and play a role in signalling. Two identical
subunits form a dimeric molecule consisting of more than 20 modules on each
part. There are three types of monomers (I-III), a module FN-III10 was most
often studied computationally.122 The first AMF measurements done by Rief et
al.123 have shown that FNIII modules of titin exhibit 20% lower unfolding
forces than Ig domains. At the same time Oberhauser et al., who studied FNIII
modules from tenascin124, have found that these forces are 50% smaller, and
that tenascin FN modules refold twice as fast as those from fibronectin. Since
the overall fold of both types of FNIII modules is very similar, observed
differences in rupture forces (75-200 pN) are hard to explain without atomistic
simulations of dynamics. This hierarchy in elasticity of individual FN modules
may have a physiological role.23
Early simulations of FN suffered from limited simulation times and a
rather crude treatment of the solvent. Kramer et al.125 used the SMD method to
study the behavor of the FNIIIio module under an external force. A small
water sphere was used in the simulations. Since timescales were much shorter
than in the experiments, a few hundred pN forces had to be applied to unfold
this P-structure and to observe that unfolding starts from the rupture of the G
strand.125 Simulations have shown that residues in the loop 78-80 (RGD) are
"strategically" located close to a membrane protein integrin, and act as a
mechanoswitch during unfolding. Further studies by the same group gave a
better picture of the mechanism of FNIII unfolding for different modules.126
Paci and Karplus90 prefered to use an implicit solvent model together with
the biased MD method (yi). A few dozen of the unfolding computer
experiments (0.2-1 ns) were carefully analyzed. Differences between FNIII9
and FNIII 10 folding pathways were highlighted. For the first time an unfolding
path for the FIII9 - FNIII I0 dimer was presented.
Perhaps the most advanced simulations of FNIII modules were performed
by Gao et al.122 They used a huge (370 A long) water box with 130 K atoms to
avoid artifacts caused by a spherical solvent shell which undergoes a large
deformation during the stretching of the protein. This study suggests that there
are at least three distinct unfolding pathways, but their discrimination, based
solely on fast SMD simulations, is not possible. A good discussion of FN
dynamics may be found in the recent review by Tajkhorshid et al.95
In 2003 Li and Makarov suggested127 that the current time scale gap
between AFM experiments and MD simulations (simulations are typically 106
times too fast) may be eliminated by describing the unfolding dynamics in
terms of diffusive motion along a predefined unfolding coordinate R in the
presence of external driving potential and the potential of mean force G(R).
The other degrees of freedom are hidden in a viscous force. The PMF G(R) is
computed from a series of equilibrium MD runs with the fixed values of R,
and friction coefficients are extrated from a series of SMD simulations. It is
notable, that the authors were able to calculate the unfolding free energy
barrier of titin at zero force, which was almost identical to that obtained
experimentally. Two distinct unfoding pathways of titin are predicted by this
model.
Mechanical properties of elastin, yet another protein found in the
extracellular matrix, especially in relation to the temperature factor, were also
computationally studied,128"132 but more traditional techniques of MD
60 W. Nowak and P. E. Marszatek
simulations were used in this case. Spectrin repeats and its two mutants were
recently studied by AFM and SMD techniques as well.133
3.1.3 SMD simulations indirectly related to AFM experiments

The SMD method was used in several important studies of ligand-protein
interactions134 for systems which have not yet been subject to any AFM type
experiments. Since the number of such applications will grow, we present a
brief account of the recent papers.
In 1999 Kosztin et al. presented unbinding pathways of the retinoic acid
hormone from its nuclear receptor h-RAR. Three hypothetical paths were
selected. The SMD calculations indicated that one of those is not probable and
that alternative pathways may be used by the hormone for binding and
unbinding. A detailed analysis of possibile close contacts, salt bridges and
hydrogen bonds along the path has improved our understanding of this
medically important receptor and may help to develop new drugs.
In 2001 Heymann and Grubmiiller136 performed a detailed study of forced
unbinding of spin-labeled dinitrophenyl hapten from the monoclonal antibody
AN02 using a variant of the SMD method called force probe molecular
dynamics (FPMD). Large flexibility of the binding pocket region has been
found. A large (6 kcal/mol) entropic contribution was calculated to the ligand-
protein bond. Interestingly, no differences in unbinding forces were found in
two Y33F and I96K mutants of the antibody.
Mechanosensitive channel with large conductance (MscL) can be gated by
tension in the cellular membrane. Gullingsrud et al, following their initial,
classical 3 ns MD study of the gating mechanism,137 used SMD in a recent 10
ns study138 to add radial external forces (70 pN) in the critical part of the
channel to induce its opening. In contrast to the previous calculations by Ma
et. al. in which TMD was used139, the authors have found that the iris-like
motion is probably involved in the process of the channel opening. Both TMD
and SMD confirm tandem-like mechanism of MscL sensitivity to the
membrane tension.
GTP hydrolizing G proteins are the part of the signal transduction system
that control proliferation, differentiation and metabolism. Among them Ras
p21 attracted particular attention, since it works like a molecular switch,
changing states between the GDP-bound inactive form and the GTP-bound
active form. The conformational changes related to this process were studied
by Ma and Karplus140, Diaz et al.141 using TMD. Recently Kosztin et al.142
added an "artificial" force attached to two carefully selected atoms, in order to
estimate the work related to the transition between tense (T) and relaxed (R)
states of the so called switch II region in h-RAS protein. It has been found
that the force generation proceedes in two steps via a short lived (< Ins)
metastable T state formed just after the GTP hydrolysis, and the amount of
work related to this process is around 1 kcal/mol.
Cell-cell adhesion is mediated by proteins. The interaction between the T-
lymphocyte adhesion receptor CD2 and its ligand CD58 was studied by Bayas
et al.143 using the SMD method. Two regimes - constant force and constant
velocity - were used to enforce detachment of these two proteins. The
mechanism of the forced detachment depends strongly on the loading rate. At
faster loading rates (70 pN/ps) the proteins unfold before detachment, at
slower rates (7 pN/ps) they just separate without significant initial distortions.
These calculations clearly show that multiple conditions in the SMD protocol
should be tried during studies of protein complexes, before conclusions on
interactions mechanisms are reached.
Elastic properties of a P-selectin construct consisting of a lectin domain
and epithelial growth factor-like domain was studied using SMD by Lu and
Long144. Groups of hydrogen bonds were identify to rupture during forced
dissociation of this complex, which is important in inflammatory proceses.
Best et al.145 posed an interesting question whether proteins, which unlike
titin, are not designed for mechanical function can withstand large forces
before unfolding. An artificial complex of barnase with the 127 domain of titin
was studied by AFM and computationally (3-6 ns SMD trajectories). The
results indicate that barnase unfolds at much lower forces than the titin
fragment. Since unfolding rates of both proteins in solution are comparable,
this indicates that the mechanism of mechanically forced unfolding may be
different than the chemically induced one.
3.2 Biased JMD

The first application of an interesting variant of enforcing conformational
transitions was proposed by Paci and Karplus.90 As the authors say, their
method is based on the unpublished work by P. Ballone and S. Rubini, who
used biased forces to study the crystallization process. The idea of applying
very small perturbations for inducing desirable conformational changes, also
related to the earlier work by Harvey and Gabb,146 is the following. The
reaction coordinate (RC) has to be selected first. Sometimes it is very complex
and difficult to determine or even guess. Usually we do not know how the
studied process proceedes.
Let us assume that the unfolding of a protein induced by a force attached
to C-terminus of the protein backbone is to be studied, and RNc is the distance
between the N-terminus and the C-terminus. For example, Paci and Karplus,
in their study of the relation of the unfolding process to the topology of the
protein,147 used as the reaction coordinate p(t) = R2NC • It is worth to mention,
that these authors used the implicit Gaussian model for the solvent. They claim
that the adiabatic relaxation of the solvent is necessary for fast unfolding
studies, and that such a simplified approach is closer to the real AFM
experiments. Time dependent bias force is:
Fb = 7(P -Pmax), (4)
but only when p(t) < pn,ax(t), otherwise the force is set to zero. The parameter
pmax (t) is the maximum value of p reached by the reaction coordinate at times
less or equal to t. Thus, the force is nonzero only when thermal fluctuations
tend to decrease the selected distance (RC). One can guess that if an energy
barrier appears on the selected structural transformation, then the bias force is
large, in the other regions its value is not high. The values of y parametres are
adjusted by a trial and error procedure, depending on the process studied and
assumed timescale, suitable to the forced structural transition under study. In
this paper147 values of y around 0.0012 kcalmol'1 A"4 were used in biasing
forces applied in order to unfold a range of P-sandwich and a-helical proteins.
It is worth to note that 1 kcalmol 1 A"1 = 69.4786 pN.
BMD has been used in several iluminating studies. Recently, Paci et al.148
have applied BMD (together with the constant force MD simulation) to reveal
atomistic details of unbinding of fluorescein from hepten-antibody complex.
The same system was simulated by Heynman and Grubmixller,'49 using the
SMD method, so the two approaches may be compared.
The biased MD technique is useful in generating molecular states, which
are not easy amenable to experiment. For example, Paci et al.150 used this
method to obtain nearly 400 clusters of partially unfolded states of human <x-
lactalbumin. A clever selection of the radius of gyration as a reaction
coordinate, based on the idea of Marchi and Ballone,151 and using an implicit
solvent made the effective generation and analysis of so many semi-stable
protein conformations possible.
4. Free energy profiles in pulling experiments

Relating single molecule measurements to thermodynamics is not an easy task,
since AFM experiments are not performed on ensambles of molecules and the
processes studied are not in equilibrium. Free energy profiles along selected
reaction coordinate or potential of mean force (PMF) have a special
importance at a single molecule level.152'153
In 1997 Balsera et al.154 presented a one-dimensional stochastic model for
the reconstruction of PMF along the path of force induced conformational
transitions in a molecular system. The model, based on suggestions by Evans

and Ritchie,155 is useful in the analysis of ligand unbinding. Dissipation is
related to a friction coefficient. The accuracy of the potential energy function
evaluated is determined from fluctuations which are proportional to the
amount of irreversible work deposited into the system by external forces. In
this paper uncertainties 8F in the calculated forces are estimated to be about 50
pN for a 10 ns simulation.
Let us assume that the initial state / is in equilibrium with the heat bath at
temperature T (unextended molecule). The parameter X, related to a
controllable degree of freedom, describes a transition of the system from state
/ to state / (extended molecule). Changes in X result in expenditure of some
amount of work W, changes of free energy of the system AG and the exchange
of heat with the bath. In 1997 C. Jarzynski156 proved an equality that relates
the irreversible work W, performed during /->/transition of the system, to the
equilibrium free energy difference between states / a n d / , AG:
<e-W/kT> = e-AG/kT ( 5 )
This equation, further analysed by Jarzynski,157'158 is very important and

became the basis of estimates of thermodynamic quantities in single molecule
experiments. In 2001 Hummer and Szabo159 presented a simple recipe on how
to reconstruct free energy surfaces from multiple nonequillibrium pulling
experiments. Perhaps one of the first applications of this formalism to SMD
computer experiments was the work done by Jensen et al.160, who studied
conduction of glycerol through the membrane aquaglycerolporin channel.
PMF calculated from Jarzynski's equality gives a reasonable understanding of
molecular events in this highly selective system. The calculated highest energy
barrier of 7.3 kcal/mol compares well with the experimental Arrhenius
activation energy 9.6 ±1.5 kcal/mol.160 A similar methodology has been used
by Amaro et al.161 in their research of ammonia transfer in the HisF enzymatic
protein, which is important in histidyne synthesis. A total of 40 ns of the
simulation time (many trajectories of 1.2 ns each) was used in order to collect
reasonable statistics.
Park et al.' 62 very recently applied Jarzynski's equality to free energy
calculations from SMD of helix-coil transitions of deca-alanine. The limited
number of 10 trajectories was generated and the second order cumulant
expansion of rigorous formulas yielded the most accuarate estimates of free
energy changes. The authors notice, that the average of the exponential work
(Eq. 5) is dominated by trajectories corresponding to small W values, that only
arise rarely. Practical applications are limited to slow processes, for which the
fluctuations of W are comparable to temperature. A related paper by
Zuckerman and Woolf163 gives a new theory of errors inherent to nonlinear

averages, such as those used in free energy calculations.
It is worth to note that Jarzynski's equality has not been proven
experimentally until recently.164 The work by Liphardt et al.164 that verifed
Jarzynski's equality was named one of the top 10 scientific breakthroughts in
2003 by "Science" magazine. The authors tested Jarzynski's equality by
performing irreversibly and reversibly mechanical stretching of a single RNA
molecule between two conformations. Application of this equality to the
irreversible work trajectories recovers the AG profile of the stretching process
to within kT/2 of its best independent estimate, i.e. the mean work of
reversible stretching.
Other approaches relating thermodynamics to SMD were suggested in the
literature. As we mentioned before, in 1999 Gullingsrud et al.100 proposed
three time series analysis methods, aimed at the reconstruction of PMF from
SMD data on forces and positions. One technique, based on WHAM, views
stretching as a quasi-equilibrium process. It has been found not to be very
promising in the interpretation of AFM experiments. The other two, one based
on van Kampen's Q-expansion and the second on the least squares
minimization of the Onsager—Machlup action with respect to the choice of
PMF, assume a Langevin description of the dynamics in order to account for
the non-equilibrium character of SMD data. A MD trajectory is equivalent to a
particle moving in one-dimensional potential PEM W(z) which has to be
found in the course of analysis from the equation:
7z = F(z,t)-^^- + cr{(t). (6)

oz
Here y is a friction coefficient, calculated from the velocity autocorrelation
function, F(z,t) is the applied external force, and o£(t) is the Gaussian white
noise; the fluctuation-dissipation theorem gives a relation o2 =2ykT, where T
is temperature and k is Boltzmann's constant. The methods were tested on
artificial potentials and a phospholipid membrane monolayer system.100
In AFM experiments a cantilever tip picks up a protein molecule attached
to the surface in a random fashion. Therefore some force spectrograms are
difficult to interpret, since the mode of coupling is not known "a priori". The
same difficulty is related to the problem of selecting RC in SMD or BMD
simulations. In principle one may try all possible (and reasonable) models of
performing the desired conformational transition, but this is not feasible. In
certain studies, such as ligand dissociation from the site buried in an enzyme,
the random expulsion molecular dynamics (REMD) scheme proposed by
Wade et al. may help.165 The random force is acting on the center of mass of
the ligand. The direction of the force is kept constant for a chosen time
interval, unless the ligand encounters rigid parts of the cavity. Then a new
direction of the force is chosen. The substrate probes different regions of the
protein until it finds the exit pathway. Ludemann et al.166 used the REMD
method to determine the most probably exit path of camphor from the
cytochrom P450cam active site. The points determined from REMD, located
outside the protein, were then used to attach a force, which extracts the ligand
from the protein. The magnitude of the SMD unbinding forces is reasonable
(500-1000 pN) and compares well with the forces obtained in earlier studies of
similar systems.83'135
5. The most recent results

The progress in MD simulations of single molecule experiments and forced
conformational transitions is very fast. In 2003 Brockwell et al.167 have shown
that pulling geometry to a large extent determines forces observed during
unfolding. In all-P protein, E21ip3, unolding forces differ by an order of
magnitude, depending on the force direction chosen. Such an anisotropy may
have physiological role.
Lorenzo et al.168 calculated molecule-membrane interaction forces. To
illustrate the principle of the method a small aminoacid alanine was modeled
on a us timescale in an explicit solvent environment. Non-specific interactions
at water-membrane interface were determined.
Hummer and Szabo169 developed a simple procedure for extracting kinetic
data from pulling experiments. This approach was tested on computer
simulation data for titin.
Special attention should be devoted to simplified mechanical models of
proteins. For example, Cieplak et al.170171 used Go-like models to perform
comparative forced unfolding of various secondary structures. Thermal and
mechanical unfolding may be studied in that way. The sequence of events
during the unfolding of titin is similar to those observed in fully atomistic
simulations.171
However, availability of large thousand-processors clusters, make record
breaking, huge SMD simulations possible. Brockman et al.172'173 have
performed ns simulations (280 000 h CPU time) of dynamics of an extremely
interesting system - a dual-domain molecular motor Fô-ATPase. This
enzyme may work in two modes. The off-membrane Fi domain utilizes the
transmembrane proton gradient to synthetize ATP from ADP and P;.
Alternatively, it may hydrolize ATP and release energy. In both modes
torsional motion of the central part of the enzyme is involved.172 It is known
that the isolated ATPase may work as a molecular nanomachine. Brockman et

al. studied a sequence of conformational transitions in Y\ subunit induced by
an external torque applied to the cental "y-stalk". It has been found that some
parts of the enzyme ((3E subunit) behave as a prestretched "springs" during the
enzymatic cycyle.173
Schulten at al.95 reported recently SMD simulations of both model Fo (112
K atoms) membrane and Fj (327 K atoms) off-membrane subunits of this
ATPase. The principal goal of the work was to understand events that couple
proton translocation across the membrane with the mechanical rotation of the
Cio subunit oligomer, which is in the contact with the central y-stalk. This
simulation was not able to predict all steps in the activity cycle of Fo, but
helped to check numerous hypotheses on the mechanism of the ATPase
activity.
The TMD method was used by Ma et al.174'175 to generate trajectories for
the bovine mitochondrial Fl-ATPase, related to elementary 120 degrees
rotations of the central y subunit. This part has a special "ionic track" of Arg
and Lys residues that plays a role in the guiding motions of |3 subunits.
Excited state dynamics opens new challenges and opportunities for
computer simulations.176 The first application of ab inito MD to isomerisation
phenomena in bacteriorodopsin was done by Hayasi et al.177 According to the
simulations, the protein plays a crutial role in the kinetics and selectivity of the
photoisomerisation of the retinal. It discriminates from many
photoisomerization paths the one which initiaties the process of light signal
transduction.
We stress the future role of the modeling of excited states, since
optomechanical experiments on single molecules have been already performed
in H.Gaub's laboratory.178'179 A peptide with multiple photoactive azobenzene
groups incorporated in the backbone was synthetized. Cis to trans-azo
switching was induced by UV irradiation and the changes in single molecule
length were registered by the AFM technique. The molecule was shortened by
light and at the same time it was able to perform some mechanical work. Thus,
optomechanical single molecule energy conversion was demonstrated.179 It
seems that performing SMD simulations of this very interesting system is only
a matter of time.
6. MD results for stretching of polysaccharides

Polysaccharides are not so popular in biophysical or chemical studies as
proteins or nucleic acid, albeit their physiological role is enormous.180
Mechanical properties of simple polysaccharides were studied by AFM

single molecule spectroscopy. In 1997 Rief et al,181 studied dextran linked to a
gold surface and tried to explain the force-extension curves using MD
simulations for 5 ring model systems and CHARMM force field. Li et al.182
measured by AFM the difference in a-(l,4) and P-(l,4) linked polysaccharides
(amylose vs. cellulose). Heymann and Grubmuller183 performed MD
simulations of stretching 16-mer models of these two polysaccharides, using
EGO code and constant velocity pulling regime both in vacuo and in TIP3P
water. In order to correctly reproduce the experimental plateau on the force-
extention curves of amylose, these authors had to introduce artificial
conformational constraints to set up the anti-parallel orientation of individual
a-(l,4)-glucose units. In 1998 Marszaiek et al.24 demonstrated that
polysaccharide elasticity is governed by chair-boat transitions of
gluocopyranose rings. AFM experiments have shown that oxidation of the ring
nearly eliminates the enthalpic component of the elasticity of dextran and
amylose. The same group put forward a hypothesis that stretching of the
pyranose ring may serve as a part of molecular signaling: it is pleausible that
the fragment of pectin, namely a-(l-4)-D-galactouronan, acts as a two state
elastic switch.25 Quantum chemical optimization of geometries of standard
4 Ci, boat and 'C 4 conformations of galatouronic acid, performed on the
B3LYP/6-31* level, supported this hypothesis. The AFM force spectroscopy
technique may be used for analytical purposes. In 2001 Marszaiek et al.34 have
shown that individual molecules of floridean starch may be effectively
identified as contaminats in the solutions of agarose or ^.-carrageenan due to
distinct molecular fingerprints of these polysaccharide species.
Chair-boat transitions in a single polysaccharide molecule may be
observed in force-ramp AFM experiments.184 In this regime a pulling force is
linearly increasing and sudden geometrical changes of the pulled molecule
may be easily identified. An analytical two-state model developed by
Marszaiek et al.184 allows for the quantitative characterization of fundamental
ring transitions. Extensive quantum mechanical studies of barriers to forced
transitions in sugar rings were performed by O'Donoghue and Luthey
Schulten.185 The HF calculations show that the twist-boat conformation of a-
D-glucose is the most probable stable minimum in a stretched polysaccharide
and the barrier for the chair to boat transition in a-D-glucose ranges from 14 to
15 kcal/mol. Recently, results of single molecule AFM measurements of
elasticity of heparin have been published.186
Elastic properties of polysaccharides with 1-6 links are more complex than
those with 1-4 links. Recenly we have found by stretching single pusrulan
molecules in the AFM instrument that P-l—>6 linkages produce an unusual
hookean-like elasticity.187 Pusrulan [a P~(l—>6)-linked glucan, (see Fig. 4)] is
similar to dextran [an oc-(l—>6)-linked glucan] but equatorial position of

glycosidic bonds (CpOi) in pustulan, and the axial one in dextran, lead to clear
differences in mechanical properties of a and P-(l-6)-glicosidic linkages. In
contrast to cellulose whose 1-4 links allow for fairly free rotations of the sugar
rings in the polymer chain pustulan experiences rotational restrictions during
the stretch, and therefore its force-extension relationships strongly deviate
from the simple freely rotating chain-like relationship for cellulose.
Additional work is used to rotate the aglycone bond (O6-C6) about the C6-C5
bond, because this rotation contributes to the extension of pustulan but is
absent in cellulose. The elasticity of dextran is different from both cellulose
and pustulan and suggests that its oc-(l—^-linkages induce not only the
rotation about the C6-C5 bond but also trigger various ring instabilities. We
applied SMD simulations to investigate the deformations of the pyranose ring
induced by forces attached to the ring at carbon atom d and through a
rotatable C6-C5 bond. Using NAMD2 code188 running on a linux cluster in
parallel mode we stretched pustulan and dextran decamers immersed in water
boxes of such a size that at least a 6 A layer of TIP3P water molecules was
present at each step of the simulations. After a 1-5 ns equilibration at 300K
(dielectric constant e = 1) the structures were stretched by applying force
(spring constant of 68 pN/nm) to the Oi atom of the last sugar and moving this
spring at a constant velocity of 10"4 A /fs. To check the effect of the pulling
speed on the results a control 200 ns simulation in vacuum with e = 80 were
also performed
Selected results187 are presented in Fig. 4. A comparison of AFM data and
the detailed atomistic picture emerging from computer modeling, indicates
that one can control the orientation of an atomic "crank": the O6-C6 bond and
reorient it mechanically from the gauche-trans (gt) or gauche-gauche (gg)
position to the less favourable energetically but more extended trans-gauche
(tg) position. These rotations produce a hookean-like elasticity of pustulan but
do not affect the chair conformation of the pyranose ring. However, deforming
the pyranose ring by oc-1—>6 linkages, using the O6-C6 cranks and axial C r O |
levers, such as those in dextran, not only reorients the O6-C6 bond to the tg
position but also produces ring instabilities, which flip the pyranose structure
to a boat-like conformation.
Studies of polysaccharide elasticity and mechanical properites of organic
rings expand single molecule mechanochemistry and are important for
future nanotechnological applications.
4600 .
3500^ R > ^ ^ \ - ^ ^ ^ i
^ 2500 • R7 Ug-75* < 1 tg-!74» jmrnsf
£ 0,-(yA 5.42 ^ 6.12 ^ j l l

O 1600 • TT
1000 • ^ ^ ^
500 • ,^"^1^^
-500 f . . . . . . . . . i . . . . T . . . . i . . . . . . . . . . . i,.,..,,,.!.........
3.5 4.0 4.5 5.0 5.5 6.0 6.5
B :::r^ ._^^*"111" ;
o -
H -12O -;
5 -iao J A- ^ m iTirtfli- • j
2
t3 ISO -I Ptms r 'MHitPIIH^
.•S 1 2 O :
° oo:
o -
-18O - -^BSi*Wfifli*
3.S **.O >*.S 5.O S.3 ©.O 6.S
Normalized Extension per Ring/A

Fig. 4 (A) Results of AFM measurements and SMD simulations for pustulan, a p-
(I—^-linked glucan, dark gray trace - 5 ns in explicit water, gray trace - 200 ns in
vacuum e = 80, black curve - experimental AFM data. (B) The torsional angle (black
dots) changes indicate a gt —> tg flip of pustulan rings in the region exhibiting
"hookean" elasticity. Constant values of tj and t2 torsional angles (horizontal lines in
the middle) show that the chair to boat transition do not occur during the stretching of
pustulan.
1. Quantum molecular dynamics of amylose stretching

Quality data may be obtained from forced MD simulations of single molecules
only when the force field (FF) is not flawed. The results of modeling are force
field dependent. The development of FFs was centered around equilibrium
benchmark structures so the quality of the potentials applied to molecules
stretched far from energy minima is not necessarely the best. Thus, there is
high demand for a fully quantum-mechanical, forcefield free, calculation for
streteched biopolymers. In a recent paper189, which is a result of a
collaborative effort among Yang, Marszalek, Nowak and their students, Lu et
al189 have presented a practical solution to this long lasting problem.
We used for the first time a quantum mechanics based methodology to
investigate how a polysaccharide immersed in water, responds to a stretching
force. The amylose fragment composed of ten sugar rings was placed in a
TIP3P water box (30 A x 30 A x 80 A; 8004 atoms) and was treated by the
self consistent charge density functional theory-based tight-binding method
(SCC-DFTB).190 Explicit long range van der Waals forces were used. The total
energy of the system including the interactions between amylose and water
was calculated with the divide-and-conquer linear-scaling QM/MM model
developed by the Yang group191 Our tests performed for amylose dimers
indicate a very high similarity of potential energy surfaces calculated by the
SCC-DFTB method and the DFT B3LYP method (6-31 lg++** basis set was
used). The results of the amylose stretching are presented in Fig. 5.
OH OH
1 , 1 , 1 , ! , 1 , , , 1 1 1 -60
0,70 0.75 0.80 0.85 0.90 0.95 100 1.05
Normalized extension (x/IJ
Fig. 5 Results of SMD simulations of the amylose stretching. Black trace: the
experimental force-extension relationship obtained by AFM. Dark gray trace: the
force-extension relationship obtained by QM/SMD simulations. Light gray trace: the
O1-C1-C2-O2 dihedral angle averaged over the glucose units. The extension, x, was
normalized by the molecule length, lc, determined at the force of 1620 pN.
The QM/SMD simulations represent the mechanical properties of amylose

in water remarkably well. The force-induced chair to boat transitions
(represented by the dihedral angle shown in Fig. 5) coincide with the plateau
region of the experimental force-extension curve, as detailed analysis of
trajectories shows. The increasing tension forces the neighboring glucose rings
to rotate around the O1-C4' bonds. The rings orient themselves in a parallel
fashion, even though they were initially in an anti-parallel orientation.
The elasticity of amylose to a large extent depends on a complex pattern of
the breakage/creation of hydrogen bonds. Without a proper solvatation it is not
possible to reproduce the value of the plateau force. These calculations,189
though quite time-consuming, bring a promise for doing realistic simulations
of nanomechanical interactions in any biologically important glycoproteins on
a routine basis, without the need of tedious optimization of force field
parameters.
Concluding remarks
The new methods and protocols of performing MD simulations - TMD, SMD,
BMD - greatly help to interpret single molecule experiments. The progress is
very fast and trends are clear: forced transitions may bridge mechanics of
molecules with chemistry. Better simulation protocols and better hardware will
reduce the gap in timescales between SMD or BMD simulations (ns) and AFM
experiments (ms). The findings of computational efforts, sometimes
unexpected, lead to new ideas, sugest new mutagenesis work and new
experimental tests. Piconewton forces observed in AFM or optical tweezer
setups are ultimate mechanical signals sent up from the nanoworld to keen
observers who are now well prepared to receive and interpret them.
Acknowledgements
This work was supported by the NSF (WN and PEM), KBN and GRUMK
(WN).
References
1. Robertson, N. & McGowan, C.A. A comparison of potential molecular wires as
components for molecular electronics. Chem. Soc. Rev. 32, 96-103 (2003).
2. Plakhotnik, T., Donley, E.A. & Wild, U.P. Single-molecule spectroscopy. Ann.
Rev. Phys. Chem. 48, 181-212 (1997).
3. Kuhnemuth, R. & Seidel, C.A.M. Principles of single molecule multiparameter
fluorescence spectroscopy. Single Molecules 2, 251-254 (2001).
4. Orrit, M. Single-molecule spectroscopy: The road ahead. J. Chem. Phys. 117,
10938-10946(2002).
5. Florin, E.L., Moy, V.T. & Gaub, H.E. Adhesion Forces between Individual
Ligand-Receptor Pairs. Science 264, 415-417 (1994).
6. Medina, M.A. & Schwille, P. Fluorescence correlation spectroscopy for the
detection and study of single molecules in Biology. Bioessays 24, 758-764 (2002).
7. Bohmer, M. & Enderlein, J. Fluorescence spectroscopy of single molecules under
ambient conditions: Methodology and technology. Chemphyschem 4, 793-808
(2003).
8. Haran, G. Single-molecule fluorescence spectroscopy of biomolecular folding. J.
Phys.-Cond. Mat. 15, R1291-R1317 (2003).
9. Forkey, J.N., Quinlan, M.E., Shaw, M.A., Come, J.E.T. & Goldman, Y.E. Three-
dimensional structural dynamics of myosin V by single-molecule fluorescence
polarization. Nature 422, 399-404 (2003).
10. Mehta, A.D., Rief, M., Spudich, J.A., Smith, D.A. & Simmons, R.M. Single-
molecule biomechanics with optical methods. Science 283, 1689-1695 (1999).
11. Molloy, J.E. & Padgett, M.J. Lights, action: optical tweezers. Contemporary
Physics 43, 241-258 (2002).
12. Coirault, C , Pourny, J.C., Lambert, F. & Lecarpentier, Y. Optical tweezers in
biology and medicine. M S-Medecine Sciences 19, 364-367 (2003).
13. Schwille, P. & Kettling, U. Analyzing single protein molecules using optical
methods. Curr. Opin. Biotech. 12, 382-386 (2001).
14. Finer, J.T., Simmons, R.M. & Spudich, J.A. Single Myosin Molecule Mechanics -
Piconewton Forces and Nanometer Steps. Nature 368, 113-119 (1994).
15. Purohit, P.K., Kondev, J. & Phillips, R. Mechanics of DNA packaging in viruses.
Proc. Natl. Acad. Sci. 100, 3173-3178 (2003).
16. Bustamante, C , Smith, S.B., Liphardt, J. & Smith, D. Single-molecule studies of
DNA mechanics. Curr. Opin. Struct. Biol. 10, 279-285 (2000).
17. Smith, S.B., Finzi, L. & Bustamante, C. Direct Mechanical Measurements of the
Elasticity of Single DNA-Molecules by Using Magnetic Beads. Science 258,
1122-1126(1992).
18. Zlatanova, J. & Leuba, S.H. Magnetic tweezers: a sensitive tool to study DNA and
chromatin at the single-molecule level. Biochemistry and Cell Biology-Biochimie
Et Biologie Cellulaire 81,151-159 (2003).
19. Romano, G., Sacconi, L., Capitanio, M. & Pavone, F.S. Force and torque
measurements using magnetic micro beads for single molecule biophysics. Optics
Comm. 215, 323-331 (2003).
20. Basche, T., Nie, S. & Fernandez, J.M. Single molecules. Proc. Natl. Acad. Sci.
USA 98, 10527-10528 (2001).
21. Marszalek, P.E., Lu, H., Li, H.B., Carrion-Vazquez, M., Oberhauser, A.F.,
Schulten, K. & Fernandez, J.M. Mechanical unfolding intermediates in titin
modules. Nature 402, 100-103 (1999).
22. Li, H.B., Linke, W.A., Oberhauser, A.F., Carrion-Vazquez, M., Kerkviliet, J.G.,
Lu, H., Marszalek, P.E. & Fernandez, J.M. Reverse engineering of the giant
muscle protein titin. Nature 418, 998-1002 (2002).
23. Oberhauser, A.F., Badilla-Fernandez, C , Carrion-Vazquez, M. & Fernandez, J.M.
The mechanical hierarchies of fibronectin observed with single-molecule AFM. J.
Mol. Biol. 319, 433-447 (2002).
24. Marszalek, P.E., Oberhauser, A.F., Pang, Y.P. & Fernandez, J.M. Polysaccharide
elasticity governed by chair-boat transitions of the glucopyranose ring. Nature
396,661-664(1998).
25. Marszalek, P.E., Pang, Y.P., Li, H.B., El Yazal, J., Oberhauser, A.F. & Fernandez,
J.M. Atomic levers control pyranose ring conformations. Proc. Natl. Acad. Sci.
(USA) 96, 7894-7898 (1999).
26. Fisher, T.E., Marszalek, P.E. & Fernandez, J.M. Stretching single molecules into
novel conformations using the atomic force microscope. Nat. Struct. Biol. 7, 719-
724 (2000).
27. Zhang, L , Wang, C , Cui, S.X., Wang, Z.Q. & Zhang, X. Single-molecule force
spectroscopy on curdlan: Unwinding helical structures and random coils. Nano
Letters 3, 1119-1124(2003).
28. Onoa, B., Dumont, S., Liphardt, J., Smith, S.B., Tinoco, I. & Bustamante, C.
Identifying kinetic barriers to mechanical unfolding of the T-thermophila
ribozyme. Science 299, 1892-1895 (2003).
29. Hansma, H.G., Pietrasanta, L.I., Auerbach, I.D., Sorenson, C , Golan, R. &
Holden, P.A. Probing biopolymers with the atomic force microscope: A review. J.
Biomaterials Sci.-Polymer Ed. 11, 675-683(2000).
30. Jandt, K.D. Atomic force microscopy of biomaterials surfaces and interfaces. Surf.
Sci. 491, 303-332 (2001).
31. Demers, L.M., D. S. Ginger, S.-J. Park, Z. Li, S.-W. Chung & Mirkin, C.A. Direct
Patterning of Modified Oligonucleotides on Metals and Insulators by Dip-Pen
Nanolithography. Science 296, 1836-1838 (2002).
32. Durig, U., Cross, G., Despont, M., Drechsler, U., Haberle, W., Lutwyche, M.I.,
Rothuizen, H., Stutz, R., Widmer, R., Vettiger, P. "Millipede" - an AFM data
storage system at the frontier of nanotribology. Tribology Letters 9, 25-32 (2000).
33. Ludwig, M., Rief, M., Schmidt, L., Li, H., Oesterhelt, F., Gautel, M. & Gaub,
H.E. AFM, a tool for single-molecule experiments. Appl. Phys. - Mat. Sci. &
Proc. 68, 173-176 (1999).
34. Marszalek, P.E., Li, H.B. & Fernandez, J.M. Fingerprinting polysaccharides with
single-molecule atomic force microscopy. Nat. Biotech. 19, 258-262 (2001).
35. Bustamante, C , Bryant, Z. & Smith, S.B. Ten years of tension: single-molecule
DNA mechanics. Nature 421, 423-427 (2003).
36. Strick, T.R. et al. Stretching of macromolecules and proteins. Rep. Prog. Phys. 66,
1-45 (2003).
37. Clausen-Schaumann, H., Seitz, M., Krautbauer, R. & Gaub, H.E. Force
spectroscopy with single bio-molecules. Curr. Opin. Chem. Biol. 4, 524-530
(2000).
38. Engel, A., Gaub, H.E. & Muller, D.J. Atomic force microscopy: A forceful way
with single molecules. Curr. Biol. 9, R133-R136 (1999).
39. Zhuang, X.W. & Rief, M. Single-molecule folding. Curr. Opin. Struct. Biol. 13,
88-97 (2003).
40. Rief, M. & Grubmuller, H. Force spectroscopy of single biomolecules.
Chemphyschem 3, 255-261 (2002).
41. Leckband, D. Measuring the forces that control protein interactions. Ann. Rev.
Biophys. Biomol. Struct. 29, 1-26(2000).
42. Riener, C.K., Stroh, CM., Ebner, A., Klampfl, C , Gall, A.A., Romanin, C ,
Lyubchenko, Y.L., Hinterdorfer, P. & Gruber, H.J. Simple test system for single
molecule recognition force microscopy. Analytica Chimica Ada 479, 59-75
(2003).
43. Best, R.B., Brockwell, D.J., Toca-Herrera, J.L., Blake, A.W., Smith, D.A.,
Radford, S.E. & Clarke, J. Force mode atomic force microscopy as a tool for
protein folding studies. Anal. Chim. Ada 479, 87-105 (2003).
44. Ikai, A., Afrin, R., Sekiguchi, H., Okajima, T., Alam, M.T. & Nishida, S. Nano-
mechanical methods in biochemistry using atomic force microscopy. Curr. Prot.
& Pept. Sci. 4, 181-193 (2003).
45. Williams, P.M. Analytical descriptions of dynamic force spectroscopy: behaviour
of multiple connections. Analyt. Chim. Ada 479, 107-115(2003).
46. Haile, M. Molecular Dynamics Simulation: Elementary Methods, (John Wiley,

1992).
47. Frankel, D. & Smit, B. Understanding Molecular Simulation, (Academic Press,
2001).
48. Field, M.J. A Practical Introduction to the Simulation of Molecular Systems,
(Cambridge University Press, 1999).
49. Schlick, T. Molecular Modeling and Simulation - An Interdisciplinary Guide.,
(Springer, New York, 2002).
50. Kholmurodov, K.T., Altaisky, M.V., Puzynin, I.V., Darden, T. & Filatov, F.P.
Methods of molecular dynamics for simulation of physical and biological
processes. Physics of Particles and Nuclei 34, 244-263 (2003).
51. Werner Ebeling, L.S.-G., Yuri M. Romanovsky. Stochastic dynamics of reacting
biomolecules, (World Scientific, Singapore, 2003).
52. Karplus, M. Molecular dynamics of biological macromolecules: A brief history
and perspective. Biopolymers 68, 350-358 (2003).
53. Hansson T, O.C., van Gunsteren W. Molecular dynamics simulations. Curr. Opin.
Struct. Biol. 12, 190-196(2002).
54. Marx, D. & Hutter, J. Ab initio molecular dynamics: Theory and Implementation,
in Modern Methods and Algorithms of Quantum Chemistry, Vol. 1 (ed.
Grotendorst, J.) 301-449 (John von Neumann Institute for Computing, Julich,
2000).
55. Carloni, P., Rothlisberger, U. & Parrinello, M. The role and perspective of a initio
molecular dynamics in the study of biological systems. Ace. Chem. Res. 35, 455-
464 (2002).
56. Karplus, M. & McCammon, J.A. Molecular dynamics simulations of
biomolecules. Nat. Struct. Biol. 9, 646-652 (2002).
57. Moraitakis, G., Purkiss A. G, Goodfellow J. M. Simulated dynamics and
biological molecules. Rep. Prog. Phys. 66, 483-406 (2003).
58. Zhou, Y.Q. & Karplus, M. Interpreting the folding kinetics of helical proteins.
Nature 401, 400-403 (1999).
59. Lazaridis, T. & Karplus, M. Thermodynamics of protein folding: a microscopic
view. Biophys. Chem. 100, 367-395 (2003).
60. Vendruscolo, M. & Paci, E. Protein folding: bringing theory and experiment
closer together. Curr. Opin. Struct. Biol. 13, 82-87 (2003).
61. Mayor U., Guydosh, N.R., Johnson, CM., Grossmann, J.G, Sato S., Jas G.S.,
Freund S.M., Alonso D.O., Daggett V., Fersht A.R. The complete folding pathway
of a protein from nanoseconds to microseconds. Nature 421, 863-867 (2003).
62. Fersht, A.R. & Daggett, V. Protein folding and unfolding at atomic resolution.
Cell 108, 573-582(2002).
63. Daggett, V. & Fersht, A.R. Is there a unifying mechanism for protein folding?
Trends inBiochem. Sci. 28, 18-25 (2003).
64. Daggett, V. Molecular dynamics simulations of the protein unfolding/folding
reaction. Ace. Chem. Res. 35, 422-429 (2002).
65. Warshel, A. & Parson, W.W. Dynamics of biochemical and biophysical reactions:
insight from computer simulations. Quart. Rev. Biophys. 34, 563-679 (2001).
66. Warshel, A. Molecular dynamics simulations of biological reactions. Ace. Chem.
Res. 35, 385-395 (2002).
67. Warshel, A. Computer simulations of enzyme catalysis: Methods, progress, and

insights. Ann. Rev. Biophys. Biomol. Struct. 32, 425-443 (2003).
68. Karplus, M. Aspects of protein reaction dynamics: Deviations from simple
behavior. J. Phys. Chem. B 104, 11-27 (2000).
69. van Speybroeck, V. & Meier, R.J. A recent development in computational
chemistry: chemical reactions from first principles molecular dynamics
simulations. Chem. Soc. Rev. 32,151-7 (2003).
70. Bala, P., Grochowski, P., Nowinski, K., Lesyng, B. & McCammon, J.A.
Quantum-dynamical picture of a multistep enzymatic process: Reaction catalyzed
by phospholipase A(2). Biophys. J. 79, 1253-1262 (2000).
71. Simonson, T., Archontis, G. & Karplus, M. Free energy simulations come of age:
Protein-ligand recognition. Ace. Chem. Res. 35,430-437 (2002).
72. McCammon, A.J. & Wolynes, P.G. Theory and simulation - Enlarging the
landscape - Editorial overview. Curr. Opin. Struct. Biol. 12, 143-145 (2002).
73. Pande V.S., Baker, I., Chapman, J., Elmer, S.P., Khaliq, S., Larson, S.M., Rhee,
Y.M., Shirts, M.R., Snow, C D . , Sorin, E,J., Zagrovic, B. Atomistic protein
folding simulations on the submillisecond time scale using worldwide distributed
computing. Biopolymers 68, 91-109 (2003).
74. Daggett, V. Long timescale simulations. Curr. Opin. Struct. Biol. 10, 160-164
(2000).
75. Straub, J.E., Guevara, J., Huo, S.H. & Lee, J.P. Long time dynamic simulations:
Exploring the folding pathways of an Alzheimer's amyloid A beta-peptide. Ace.
Chem. Res. 35, 473-481 (2002).
76. Elber, R. Novel methods for molecular dynamics simulations. Curr. Opin. Struct.
Biol. 6,232-235(1996).
77. Paci, E. High pressure simulations of biomolecules. BBA-Protein Struct, and Mol.
Enz. 1595, 185-200(2002).
78. Wong, C.F. & McCammon, J.A. Protein flexibility and computer-aided drug
design. Ann. Rev. Pharm. Toxicol. 43, 31-45 (2003).
79. Sulpizi, M., Folkers, G., Rothlisberger, U., Carloni, P. & Scapozza, L.
Applications of density functional theory-based methods in medicinal chemistry.
Quant. Struct.-Act. Relat. 21, 173-181 (2002).
80. Banci, L. Molecular dynamics simulations of metalloproteins. Curr. Opin. Chem.
Biol. 7, 524 (2003).
81. Giorgetti, A. & Carloni, P. Molecular modeling of ion channels: structural
predictions. Curr. Opin. Chem. Biol. 7, 150-156(2003).
82. Dwyer, D.S. Molecular modeling and molecular dynamics simulations of
membrane transporter proteins. Methods Mol. Biol. 227, 335-50 (2003).
83. Grubmuller, H., Heymann, B. & Tavan, P. Ligand binding: Molecular mechanics
calculation of the streptavidin biotin rupture force. Science 271, 997-999 (1996).
84. Elber, R., Karplus, M. A method for determining reaction paths in large
molecules: Application to myoglobin. Chem. Phys. Letters 139, 375-381 (1987).
85. Elber, R. Reaction Path Studies of Biological Molecules, in Recent developments
in theoretical studies of proteins., Vol. 7 (ed. Elber, R.) (World Scientific,
Singapore, 1996).
86. Elber, R., Ghosh, A. & Cardenas, A. Long time dynamics of complex systems.
Ace. Chem. Res. 35, 396-403 (2002).
87. Schlitter, J., M. Engels, P. Kruger, E. Jacoby & A. Wollmer. Targeted molecular
dynamics simulation of conformational change - application to the T to R
transition in insulin. Mol. Sim. 10, 291-308 (1993).
88. Leech, J., Prins, J.F. & Hermans, J. SMD: Visual Steering of Molecular Dynamics
for Protein Design. IEEE Comp. Sci. Eng. 3, 38-45 (1996).
89. Izrailev, S., Stepaniants, S., Balsera, M., Oono, Y. & Schulten, K. Molecular
dynamics study of unbinding of the avidin-biotin complex. Biophys. J. 72, 1568-
1581 (1997).
90. Paci, E. & Karplus, M. Forced unfolding of fibronectin type 3 modules: An
analysis by biased molecular dynamics simulations. J. Mol. Biol. 288, 441-459
(1999).
91. Galligan, E., Roberts, C.J., Davies, M.C., Tendler, S.J.B. & Williams, P.M.
Simulating the dynamic strength of molecular interactions. J. Chem. Phys. 114,
3208-3214 (2001).
92. Torrie, G.M. & Valleau, J.P. Nonphysical sampling distribution in Monte Carlo
free energy estimation: umbrella sampling. J. Comput. Phys. 23, 187-199 (1977).
93. Izrailev, S., Stepaniants, S., Isralewitz, B., Kosztin, D., Lu, H., Molnar, F.,
Wriggers, W. & Schulten, K... Steered molecular dynamics, in Algorithms for
Macromolecular Modelling (eds. Deuflhard, P. et al.) (Springer Verlag, New
York, 1998).
94. Isralewitz, B., Baudry, J., Gullingsrud, J., Kosztin, D. & Schulten, K. Steered
molecular dynamics investigations of protein function. J. Molec. Graph. Model.
19, 13-25(2001).
95. Tajkhorshid, E., A. Aksimentiev, I. Balabin, M. Gao, B. Isralewitz, J. C. Phillips,
F. Zhu & Schulten, K. Advances in Protein Chemistry, in Protein Simulations,
Vol. 66 (eds. Eisenberg, D. & Kim, P.) (Elsevier, 2003).
96. Izrailev, S., Crofts, A.R., Berry, E.A. & Schulten, K. Steered molecular dynamics
simulation of the rieske subunit motion in the cytochrome bc(l) complex.
Biophys. J. 77, 1753-1768 (1999).
97. Prins, J.F., Hermans, J., Mann, G., Nyland, L.S. & Simons, M. A virtual
environment for steered molecular dynamics. Future Generation Computer
Systems 15, 485-495 (1999).
98. Stone, J., Gullingsrud, J., Grayson, P. & Schulten, K. The 2001 ACM Symposium
on Interactive 3D Graphics: A system for interactive molecular dynamics
simulation, pp 191-194 (ACM SIGGRAPH, New York, 2001).
99. Grayson, P., Tajkhorshid, E. & Schulten, K. Mechanisms of selectivity in
channels and enzymes studied with interactive molecular dynamics. Biophys. J.
85, 36-48 (2003).
100.Gullingsrud, J.R., Braun, R. & Schulten, K. Reconstructing potentials of mean
force through time series analysis of steered molecular dynamics simulations. J.
Comp. Phys. 151, 190-211 (1999).
101.Stepaniants, S., Izrailev, S. & Schulten, K. Extraction of lipids from phospholipid
membranes by steered molecular dynamics. J. Mol. Model. 3, 473-475 (1997).
102.Rief, M., Gautel, M., Oesterhelt, F., Fernandez, J.M. & Gaub, H.E. Reversible
unfolding of individual titin immunoglobulin domains by AFM. Science 276,
1109-1112(1997).
103.Card on-Vazquez, M., Oberhauser, A.F., Fowler, S.B., Marszalek, P.E., Broedel,
S.E., Clarke, J. & Fernandez, J.M. Mechanical and chemical unfolding of a single
protein: A comparison. Proc. Natl. Acad. Sci. (USA) 96, 3694-3699 (1999).
104. Carrion-Vazquez, M., Oberhauser, A.F., Fisher, T.E., Marszalek, P.E., Li, H.B. &
Fernandez, J.M. Mechanical design of proteins-studied by single-molecule force
spectroscopy and protein engineering. Prog. Biophys. & Mol. Biol. 74, 63-91
(2000).
105.Li, H.B., Carrion-Vazquez, M., Oberhauser, A.F., Marszalek, P.E. & Fernandez,
J.M. Point mutations alter the mechanical stability of immunoglobulin modules.
Nat. Struct. Biol. 7, 1117-1120 (2000).
106.Lu, H., Isralewitz, B., Krammer, A., Vogel, V. & Schulten, K. Unfolding of titin
immunoglobulin domains by steered molecular dynamics simulation. Biophys. J.
75,662-671(1998).
107.Lu, H. & Schulten, K.. Steered molecular dynamics simulation of conformational
changes of immunoglobulin domain 127 interpret atomic force microscopy
observations. Chem. Phys. 247, 141-153 (1999).
108.Lu, H. & Schulten, K. The key event in force-induced unfolding of titin's
immunoglobulin domains. Biophys. J. 79, 51-65 (2000).
109.Lu, H., Krammer, A., Isralewitz, B., Vogel, V. & Schulten, K. Computer
modeling of force-induced titin domain unfolding, in Elastic Filaments of the Cell,
Vol.481 143-162(2000).
110.Isralewitz, B., Gao, M., Lu, H. & Schulten, K. Forced unfolding or mutant titin
immunoglobulin domains with steered molecular dynamics. Biophys. J. 78, 28A-
28A (2000).
111.Gao, M., Lu, H. & Schulten, K. Simulated refolding of stretched titin
immunoglobulin domains. Biophys. J. 81, 2268-2277 (2001).
112.Gao, M., Lu, H. & Schulten, K. Unfolding of titin domains studied by molecular
dynamics simulations. J. Muscle Res. Cell. Motil. 23, 513-21 (2002).
113.Gao, M., Wilmanns, M. & Schulten, K. Steered molecular dynamics studies of
titin II domain unfolding. Biophys. J. 83, 3435-3445 (2002).
114.Li, H. & Fernandez, J.M. Mechanical Design of the First Proximal Ig Domain of
Human Cardiac Titin Revealed by Single Molecule Force Spectroscopy.
J. Mol. Biol. 334, 75-86 (2003).
115.Zhang, B. & Evans, J.S. Modeling AFM-Induced PEVK Extension and the
Reversible Unfolding of Ig/FNIII Domains in Single and Multiple Titin
Molecules. Biophys. J. 80, 597-605 (2001).
116.Best, R.B., Fowler, S.B., Toca-Herrera, J.L. & Clarke, J. A simple method for
probing the mechanical unfolding pathway of proteins in detail. Proc. Natl. Acad.
Sci. (USA) 99, 12143-12148 (2002).
117.Isralewitz, B., Gao, M. & Schulten, K. Steered molecular dynamics and
mechanical functions of proteins. Curr. Opin. Struct. Biol. 11, 224-230 (2001).
118.Best, R.B., Fowler, S.B., Herrera, J.L., Steward, A., Paci, E. & Clarke, J.
Mechanical unfolding of a titin Ig domain: structure of transition state revealed by
combining atomic force microscopy, protein engineering and molecular dynamics
simulations. J. Mol. Biol. 330, 867-77 (2003).
119.Fowler SB, Best RB, Toca Herrera JL, Rutherford TJ, Steward A, Paci E, Karplus
M, Clarke J. Mechanical unfolding of a titin Ig domain: Structure of unfolding
intermediate revealed by combining AFM, molecular dynamics simulations, NMR

and protein engineering. J. Mol. Biol. 322, 841-849 (2002).
12O.Schwaiger, I., Kardinal, A., Schleicher, M., Noegel, A.A. & Rief, M. A
mechanical unfolding intermediate in an actin-crosslinking protein. Nat. Struct.
Mol. Biol. 11, 81-85 (2004).
121.Kellermayer, M.S.Z., Bustamante, C. & Granzier, H.L. Mechanics and structure
of titin oligomers explored with atomic force microscopy. BBA-Bioenergetics
1604,105-114(2003).
122.Gao, M., Craig, D., Vogel, V. & Schulten, K. Identifying unfolding intermediates
of FN-III10 by steered molecular dynamics. J. Mol. Biol. 323, 939-950 (2002).
123.Rief, M., Gautel, M , Schemmel, A. & Gaub, H.E. The mechanical stability of
immunoglobulin and fibronectin III domains in the muscle protein titin measured
by atomic force microscopy. Biophys. J. 75, 3008-3014 (1998).
124.0berhauser, A.F., Marszalek, P.E., Erickson, H.P. & Fernandez, J.M. The
molecular elasticity of the extracellular matrix protein tenascin. Nature 393, 181-
185(1998).
125.Krammer, A., Lu, H., Isralewitz, B., Schulten, K. & Vogel, V. Forced unfolding
of the fibronectin type III module reveals a tensile molecular recognition switch.
Proc. Natl. Acad. Sci. (USA) 96, 1351-1356 (1999).
126.Craig, D., Krammer, A., Schulten, K. & Vogel, V. Comparison of the early stages
of forced unfolding for fibronectin type III modules. Proc. Natl. Acad Sci. (USA)
98, 5590-5595 (2001).
127.Li, P.C. & Makarov, D.E. Theoretical studies of the mechanical unfolding of the
muscle protein titin: Bridging the time-scale gap between simulation and
experiment. J. Chem. Phys. 119, 9260-9268 (2003).
128.Li, B., Alonso, D.O.V., Bennion, B.J. & Daggett, V. Hydrophobic hydration is an
important source of elasticity in elastin-based biopolymers. J. Am. Chem. Soc.
123, 11991-11998(2001).
129.Li, B., Alonso, D.O.V. & Daggett, V. The molecular basis for the inverse
temperature transition of elastin. J. Mol. Biol. 305, 581-592 (2001).
130.Li, B., Alonso, D.O.V. & Daggett, V. Stabilization of globular proteins via
introduction of temperature-activated elastin-based switches. Structure 10, 989-
998 (2002).
131.Li, B. & Daggett, V. Molecular basis for the extensibility of elastin. J. Muscle
Res. CellMotil. 23, 561-573 (2002).
132.Li, B. & Daggett, V. The molecular basis of the temperature- and pH-induced
conformational transitions in elastin-based peptides. Biopolymers 68, 121-129
(2003).
133.Altmann, S.M., Grunberg, R.G., Lenne, P.F., Ylanne, J., Raae, A., Herbert, K.,
Saraste, M., Nilges, M. & Horber, J.K.H Pathways and intermediates in forced
unfolding of spectrin repeats. Structure 10, 1085-1096 (2002).
134.Verkhivker, G.M., Bouzida, D., Gehlhaar, D.K., Rejto, P.A., Freer, S.T. & Rose,
P.W. Complexity and simplicity of ligand-macromolecule interactions: the energy
landscape perspective. Curr. Opin. Struct. Biol. 12, 197-203 (2002).
135.Kosztin, D., Izrailev, S. & Schulten, K. Unbinding of retinoic acid from its
receptor studied by steered molecular dynamics. Biophys. J. 76, 188-197 (1999).
136.Heymann, B. & Grubmuller, H. Molecular dynamics force probe simulations of

antibody/antigen unbinding: entropic control and nonadditivity of unbinding
forces. BiophysJ. 81, 1295-1313 (2001).
137.Gullingsrud, J., Kosztin, D. & Schulten, K. Structural determinants of MscL
gating studied by molecular dynamics simulations. Biophys. J. 80, 2074-2081
(2001).
138.Gullingsrud, J. & Schulten, K. Gating of MscL studied by steered molecular
dynamics. Biophys. J. 85, 2087-2099 (2003).
139.Kong, Y., Y. Shen, T. E. Warth & Ma, J. Conformational pathways in the gating
of Escherichia coli mechanosensitive channel. Proc. Natl. Acad. Sci. (USA) 99,
5999-6004 (2002).
14O.Ma, J. & Kaplus, M. Molecular switch in signal transduction: Reaction paths of
the conformational changes in ras p21. Proc. Natl. Acad. Sci. (USA) 94, 11905-
11910(1997).
141.Diaz, J.F., Wroblowski, B., Schlitter, J. & Engelborghs, Y. Calculation of
pathways for the conformational transition between the GDP and GTP bound
states of the Ha-ras-p21 protein. Calculations with explicit solvent simulations and
comparison with calculations in vacuum. Proteins Structure Function and
Genetics 28, 434-51 (1997).
142.Kosztin, I., Bruinsma, R., O'Lague, P. & Schulten, K. Mechanical force
generation by G proteins. Proc. Natl. Acad. Sci. (USA) 99, 3575-3580 (2002).
143. Bayas, M.V., Schulten, K. & Leckband, D. Forced detachment of the CD2-
CD58 complex. Biophys. J. 84, 2223-2233 (2003).
144.Lu, S. & Long, M. Forced extension of P-selectin construct using steered
molecular dynamics. Chinese Sci. Bull. 49, 10-17 (2004).
145.Best, R.B., Li, B., Steward, A., Daggett, V. & Clarke, J. Can non-mechanical
proteins withstand force? Stretching barnase by atomic force microscopy and
molecular dynamics simulation. Biophys. J. 81, 2344-2356 (2001).
146.Harvey, S.C. & Gabb, H.A. Conformational transition using molecular dynamics
with minimum biasing. Biopolymers 33, 1167-1172 (1993).
147.Paci, E. & Karplus, M. Unfolding proteins by external forces and temperature:
The importance of topology and energetics. Proc. Nat. Acad. Sci. (USA) 97, 6521-
6526 (2000).
148.Paci, E., Caflisch, A., Pluckthun, A. & Karplus, M. Forces and energetics of
hapten-antibody dissociation: A biased molecular dynamics simulation study. J.
Mol. Biol. 314, 589-605 (2001).
149.Heymann, B. & Grubmueller, H. AN02/DNP-hapten unbinding forces studied by
molecular dynamics atomic force microscopy simulations. Chem. Phys. Letters
303,1-9(1999).
150.Paci, E., Smith, L.J., Dobson, C M . & Karplus, M. Exploration of partially
unfolded states of human alpha-lactalbumin by molecular dynamics simulation. J.
Mol. Biol. 306, 329-347 (2001).
151.Marchi, M. & Ballone, P. Adiabatic bias molecular dynamics: a method to
navigate the conformational space of complex molecular systems. J. Chem. Phys.
110,3697-3702(1999).
152.Keller, D., Swigon, D. & Bustamante, C. Relating single-molecule measurements
to thermodynamics. Biophys. J. 84, 733-738 (2003).
153.Fox, R.F. Using nonequilibrium measurements to determine macromolecule free-

energy differences. Proc. Natl. Acad. Sci. (USA) 100, 12537-12538 (2003).
154.Balsera, M., Stepaniants, S., Izrailev, S., Oono, Y. & Schulten, K. Reconstructing
Potential Energy Functions from Simulated Force-Induced Unbinding Processes.
Biophysical! 73, 1281-1287(1997).
155.Evans, E. & Ritchie, K. Dynamic strength of molecular adhesion bonds. Biophys.
J. 72, 1541-55 (1997).
156.Jarzynski, C. Equilibrium free energy differences from nonequilibrium
measurements: a master equation approach. Phys. Rev. E 56(1997).
157.Jarzynski, C. Equilibrium free energies from nonequilibrium processes. Ada
Phys. Pol. B 29, 1609(1998).
158.Jarzynski, C. How does a system respond when driven away from thermal
equilibrium? Proc. Nat. Acad. Sci. (USA) 98, 3636 (2001).
159.Hummer, G. & Szabo, A. Free energy reconstruction from nonequilibrium single-
molecule pulling experiments. Proc. Natl. Acad. Sci. (USA) 98, 3658-3661 (2001).
16O.Jensen, M.O., Park, S., Tajkhorshid, E. & Schulten, K. Energetics of glycerol
conduction through aquaglyceroporin GlpF. Proc. Natl. Acad. Sci. (USA) 99,
6731-6736(2002).
161.Amaro, R., Tajkhorshid, E. & Luthey-Schulten, Z. Developing an energy
landscape for the novel function of a (beta/alpha)8 barrel: ammonia conduction
through HisF. Proc. Natl. Acad. Sci. (USA) 100, 7599-604 (2003).
162.Park, S., Khalili-Araghi, F., Tajkhorshid, E. & Schulten, K. Free energy
calculation from steered molecular dynamics simulations using Jarzynski's
equality. J. Chem. Phys. 119, 3559-3566 (2003).
163.Zuckerman, D.M. & Woolf, T.B. Theory of a systematic computational error in
free energy differences. Phys. Rev. Lett. 89, 180602 (2002).
164.Liphardt, J., Dumont, S., Smith, S.B., Tinoco, I. & Bustamante, C. Equilibrium
information from nonequilibrium measurements in an experimental test of
Jarzynski's equality. Science 296, 1832-1835 (2002).
165.Ludemann, S.K., Lounnas, V. & Wade, R.C. How do substrates enter and
products exit the buried active site of cytochrome P450cam? 1. Random expulsion
molecular dynamics investigation of ligand access channels and mechanisms. J.
Mol. Biol. 303, 797-811 (2000).
166.Ludemann, S.K., Lounnas, V. & Wade, R.C. How do substrates enter and
products exit the buried active site of cytochrome P450cam? 2. Steered molecular
dynamics and adiabatic mapping of substrate pathways. J. Mol. Biol. 303, 813-830
(2000).
167.Brockwell, D.J., Paci, E., Zinober, R.C, Beddard, G.S., Olmsted, P.D., Smith,
D.A., Perham, R.N. & Radford, S.E. Pulling geometry defines the mechanical
resistance of a beta-sheet protein. Nat. Struct. Biol. 10, 731-7 (2003).
168.Lorenzo, A.C., Pascutti, P.G. & Bisch, P.M. Nonspecific interaction forces at
water-membrane interface by forced molecular dynamics simulations. J. Comp.
Chem. 1A, 328-339 (2003).
169.Hummer, G. & Szabo, A. Kinetics from nonequilibrium single-molecule pulling
experiments. Biophys J 85, 5-15 (2003).
17O.Cieplak, M., Hoang, T.X. & Robbins, M.O. Thermal folding and mechanical
unfolding pathways of protein secondary structures. Proteins-Structure Function
and Genetics 49, 104-113 (2002).
171.Cieplak, M., Hoang, T.X. & Robbins, M.O. Folding and stretching in a go-like
model of titin. Proteins-Structure Function and Genetics 49, 114-124 (2002).
172.Bockmann, R.A. & Grubmuller, H. Nanoseconds molecular dynamics simulation
of primary mechanical energy transfer steps in Fl-ATP synthase. Nat. Struct. Biol.
9, 198-202 (2002).
173.Bockmann, R.A. & Grubmuller, H. Conformational Dynamics of the F(l)-ATPase
beta-Subunit: A Molecular Dynamics Study. Biophys. J. 85, 1482-91 (2003).
174.Ma, J.P., Flynn, T.C., Cui, Q., Leslie, A.G.W, Walker, J.E. & Karplus, M. A
dynamic analysis of the rotation mechanism for conformational change in F-l-
ATPase. Structure 10, 921-931 (2002).
175.Ma, J.P, Flynn, T.C., Cui, C , Leslie, A.G.W., Walker, J.E. & Karplus, M. A
dynamic analysis of the rotation mechanism for conformational change in F-l-
ATPase (vol 10, pg 921, 2002). Structure 10, 1285-1285 (2002).
176.Doltsinis, N.L. & Marx, D. First principles molecular dynamics involving excited
states and nonadiabatic transtitions. J. Theoret. Comput. Chem. 1, 319-349 (2002).
177.Hayashi, S., Tajkhorshid, E. & Schulten, K. Molecular dynamics simulation of
bacteriorhodopsin's photoisomerization using ab initio forces for the excited
chromophore. Biophys. J. 85, 1440-1449 (2003).
178.Hugel, T., Holland, N.B., Cattani, A., Moroder, L., Seitz, M. & Gaub, H.E.Single-
molecule optomechanical cycle. Science 296, 1103-1106 (2002).
179.Holland, N.B., Hugel, T., Neuert, G., Cattani-Scholz, A., Renner, C , Oesterhelt,
D., Moroder, L., Seitz, M. & Gaub, H.E. Single molecule force spectroscopy of
azobenzene polymers: Switching elasticity of single photochromic
macromolecules. Macromolec. 36, 2015-2023 (2003).
180.Ramesh, H.P. & Tharanathan, R.N. Carbohydrates - the renewable raw materials
of high biotechnological value. Crit. Rev. Biotechnol. 23, 149-73 (2003).
181.Rief, M., Oesterhelt, F., Heymann, B. & Gaub, H.E. Single molecule force
spectroscopy on polysaccharides by atomic force microscopy. Science 275, 1295-
1297(1997).
182.Li, H., Rief, M., Oesterhelt, F., Gaub, H.E., Zhang, X. & Shen, J. Single-molecule
force spectroscopy on polysaccharides by AFM - nanomechanical fingerprint of
alpha-(l,4)-linked polysaccharides. Chem. Phys. Letters 305, 197-201 (1999).
183.Heymann, B. & Grubmueller, H. Chair-boat transitions and side groups affect the
stiffness of polysaccharides. Chem. Phys. Letters 305, 202-208 (1999).
184.Marszalek, P.E., Li, H.B., Oberhauser, A.F. & Fernandez, J.M. Chair-boat
transitions in single polysaccharide molecules observed with force-ramp AFM.
Proc. Natl. Acad. Sci. (USA) 99, 4278-4283 (2002).
185.O'Donoghue, P. & Luthey-Schulten, Z.A. Barriers to forced transitions in
polysaccharides. J. Phys. Chem. B 104, 10398-10405 (2000).
186.Marszalek, P.E., Oberhauser, A.F., Li, H.B. & Fernandez, J.M. The force-driven
conformations of heparin studied with single molecule force microscopy. Biophys.
J. 85, 2696-2704 (2003).
187.Lee, G., Nowak, W., Jaroniec, J., Zhang, Q. & Marszalek, P.E. Nanomechanical
control of glucopyranose rotamers. J. Am. Chem. Soc, 126, 6218-6219 (2004).
188.Kale, L., Skeel, R., Bhandarkar, M., Brunner, R., Gursoy, A., Krawetz, N.,
Phillips, J., Shinozaki, A., Varadarajan, K., Schulten, K. NAMD2: Greater
scalability for parallel molecular dynamics. J. Comp. Phys. 151, 283-312 (1999).
189.Lu, Z., Nowak, W., Lee, G., Marszalek, P.E. & Yang, W. The elastic properties of
single amylose chains in water: a quantum mechanical and AFM study. J. Am.
Chem. Soc, (2003) in press.
19O.Cui, Q., Elstner, M., Kaxiras, E., Frauenheim, T. & Karplus, M. A QM/MM
implementation of the self-consistent charge density functional tight binding
(SCC-DFTB) method. J. Phys. Chem. B 105, 569-585 (2001).
191.Liu, H., M. Elstner, E. Kaxiras, T. Frauenheim, J. Hermans & Yang, W. Quantum
mechanics simulation of protein dynamics on long time scale made possible.
Proteins: Structure, Function and Genetics 44, 484-489 (2001).
Chapter 3: Molecular Dynamics Simulations of a Molecular
Electronics Device: The NanoCell
Jorge Seminario, Pedro Derosa\ Luis Cordova6 and Brian Bozardc
Chemical Engineering Department, Texas A&M University,

College Station, Texas 77843, USA
"Institute for Micromanufacturing/Department of Physics,

Louisiana Tech. University, Ruston, LA71272, USA
^Department of Electrical Engineering,

University of South Carolina, Columbia, SC 29208, USA
(Former affiliation for all authors)
cSpace and Naval Warfare (SPAWAR) Systems Center, Charleston,

SC 29419-9022, USA
Abstract
Molecular dynamics simulations using ab initio force fields and signal
processing techniques are combined to analyze the dynamic properties of a
minimum unit of a field programmable random molecular array also called
nanocell. We analyze stretching, angular, and torsional internal modes at
several temperatures of operation and their correlations as well as the
possibility of signal modulation and processing at terahertz frequencies. The
proposed scenario is another alternative for molecular electronics, which is
being developed at very low frequencies of operation using the electronic
states of molecules, clusters and other nanoscopic systems.
85
86 J. Seminario et al.
1. Introduction
The use of small molecules to complement silicon-based electronic devices is
under intensive development [1-7]; each molecule occupies an area between
105 to 106 [2, 5] times smaller than conventional electronic devices and
therefore provides hopes to go beyond the limits set by the use of bulk silicon
devices. On the other hand, molecules would allow us to develop several
other applications directly related or reminiscent of biological systems. All
this impetus has created the new field of molecular electronics (moletronics).
Moletronics focuses on the development of electronic devices using small
molecules with feature sizes on the order of one or a few nanometers [2, 5].
These very small feature sizes are well below the minimum feature size of
present electronic devices (-45 nm in the year 2004). Notice that a high
doping of 1018 cm"3 corresponds to an average separation between nearest
impurities of 10 nm thus a radical change of technology would be needed
when approaching such a size. The ultimate goal of miniaturization is to use
the minimum amount of atoms per electronic function. However, this goal
poses a strong challenge since molecules with dimensions smaller than 10 nm
cannot be assembled as standard semiconductor devices using modern
lithographic techniques; the alternative is the use of other techniques such as
molecular self-assembly and vapor deposition, which bring a strong random
component into the field. Deposition techniques yield patterns resulting in
minimum units that can be thoroughly analyzed but it will be required that
they can be reconfigured (programmed) in order to compensate for their lack
of addressability [8, 9]. Thus, programmability of devices at the molecular
level compensates for their lack of addressing. Addressing is one of the most
important properties of standard microelectronics. Fortunately,
programmability of non addressable molecular devices can be achieved using
molecules featuring highly nonlinear current-voltage characteristics like
negative differential resistance (NDR) [1]. Experiments showing NDR and
their use as memories have been already reported [10, 11], and a strong effort
is presently underway in order to demonstrate programmable molecular
devices [1,6, 12].
Addressability during the construction and operation of molecular devices
is a challenge. Both self-assembling [5, 13] and direct assembling [14] have
been proposed for the construction and they depend on the chemical affinity of
terminal groups attached to the "electrically active" part of the molecule. Due
to the non-deterministic characteristics of these processes, molecules are
expected to be arranged in quasi-random configurations, which together with
their small size, make them no directly addressable. The conversion,
reconfiguration, or reprogramming during operation of a randomly constructed
Molecular Dynamics Simulations of a Molecular Electronics Device 87
circuit relies on the use of molecules with highly nonlinear features that allow
the system to have alternative states able to be set externally after construction
to perform a specific function [1, 8,9, 12].
Molecular circuits will be created by chemical means inside externally
addressable units; however the inside of these units cannot be addressed
except by the external programming; these small units are called nanoCells.
The nanoCell is one of the most important and viable concepts for the
development of electronics beyond the deterministic or lithographic approach
of present C-MOS integrated circuits. Very few scenarios have been proposed
to the use of molecular-based devices in order to bypass problems
undermining miniaturization or scaling-down processes for integrated circuits
through the introduction of a revolutionary approach in which molecules or
nanoclusters of atoms are used to perform logical operations complementing
or extending the capabilities of deterministic ones. The nanoCell concept
takes advantage of the great ability of chemistry to synthesize molecules with
specific characteristics with tolerances of fractions of one nanometer; such a
tolerance is far from the reach of any IC fabrication technique, which is
limited to practically two dimensions. However molecules are created in a
single flask in quantities in the order of Avogadro's number ~6xlO23 and they
are not individually addressable or interconnectable thus it is impossible to
read or write any information in a deterministic manner. A way out of this
problem is to address a molecular circuit constructed by chemical means, of a
size such that is addressable by deterministic techniques (like present
lithography), and indirectly program the individual molecules for performing
specific and more complex functions than their microelectronics counterparts.
This almost unthinkable approach is possible if the molecules (assembled
almost randomly) can be chosen with programmable features. It has been
already shown in the literature that even two single molecules with strong non
linear characteristics can perform programmable functions thus compensating
for the lack of addressability of single molecules [1, 8, 9, 12].
The applications of these nanoCells can be easily extended
interconnectiong long biomolecules like DNA or one dimensional wires. This
macroscopic systems in one dimension (or backbone) and nanoscopic in the
other two allow the transmission of vibrations along the molecule backbone
due to the presence of actual bands where vibrational modes become part of a
one-dimensional continuum and thus signals entered in one place of a long
molecule can be read and processed in other places far away in the molecule as
it was already suggested ealry in 1941 [15].
In this work, we study one of the possible minimum programmable
structures, a square consisting of four active molecules interconnected by four
Au clusters. This arrangement is one of the two most likely allowed to take
place when self-assembling on a substrate (the other one being a triangle) [12].
We focus this study on the thermal stability of the system and the correlation
among vibrations in the molecules. Ab initio calculations, molecular dynamics
simulations, and signal processing techniques are combined to achieve this
aim. MD simulations provide time dependant information of the system, and
signal processing techniques are used for the analysis of the time-domain
signal in order to show the strong correlations among the dynamics of internal
degrees of freedom and to suggest a new scenario for the use of molecular
arrangements as signal processing devices. A detailed analysis of the
frequency spectra yields correlations between several local vibrations that can
be used to include vibrational signals externally introduced through the
contacts of the molecular device. Although these correlations between internal
modes are not surprising, the question is if using the natural amplitude
oscillations of the vibrational modes, which can be experimentally detected,
will allow the detection of mixed signals in localized sites. If this were
possible, several single operations could be implemented, enabling molecular
processors dealing with signals in the THz range of frequencies. We propose
that signals can be encoded as the movement of atoms that can be detected
observing the changes in the electrical characteristics of the molecule. For
instance, when the displacement, ? f of atoms associated to a vibrational mode
changes the dipole moment of the molecule ^, a signal proportional to the so
called infra red (IR) intensity is obtained from
IR Intensity <* —
\dr )
and when atoms displacements affect the polarizability, «, a signal

proportional to the so called Raman intensity is obtained,
Raman Intensity
dr )
These two intensities which correspond to the most common experimental

spectroscopic techniques, can be unambiguously determined theoretically, and
are potential candidates to be used as means for encoding information because,
on the other hand, induced external changes in the dipole and polarizability by
external fields yield changes in the displacements of the atoms in the
molecule. Thus, signals are internally represented by atom displacements and
transduced to IR and Raman intensities.
2. Procedure and Methods

This study focuses on a molecular unit consisting of four A11405 clusters
interconnecting four active molecular devices (Figure 1). The Au clusters are
roughly spherical with diameter o f - 2 . 3 nm. The molecule, 2,5-dinitro-l,4-
diethynylphenylthiolate-benzene, dinitro hereafter, is a phenyl-based oligomer
known to perform negative differential resistance NDR, a very important
feature for molecular electronics applications [10, 16, 17]. We combine
molecular dynamics simulations, able to simulate time evolution of the system
at different temperatures, with digital signal processing techniques, able to
analyze the time-domain signals obtained by the MD simulations.
The initial geometry of the system is obtained minimizing the energy of
the whole structure using the fast minimizer tool [18] from the Cerius2 [19]
suit of programs. The MD runs solve Newton's equations simultaneously for
all atoms in the system using the leap frog algorithm [20] as implemented in
Cerius2 [19]. We use the universal force field (UFF) [21], which is
parameterized for all the elements in the periodic table and validated using
experimental data for organic molecules [22], main-group compounds [23],
and metal complexes [24]. UFF uses combination rules based on elements,
hybridization, and connectivity to determine interaction parameters from a set
of parameters characterizing each atom type. In this work, modifications are
performed for Au-Au and Au-S interactions with respect to those in the UFF
[21, 25]. We chose a 12-10 potential
with XAU-AU - 2.884 A and Xs.Au - 2.350 A. For a more general description of
the UFF, the reader is referred to the original work of Rappe [21], and to the
compilation and implementation done by Senn [25].
The Au-Au interaction with a 12-10 potential, which is an approximation
from a more realistic Sutton-Chen many-body potential [26, 27]. The
equilibrium distance Au-Au is set to 2.884, which corresponds to the Au-Au
distance between nearest neighbors in an fee Au lattice,[28] and it is within the
equilibrium geometries of 2.6 and 3.0 A, depending of the size and shape of
the cluster obtained from ab initio calculations [29]. The interaction energy DQ
of 14.7 kcal/mol is obtained from the experimental cohesive energy of 87.5
kcal/mol [28], which accounts for 12 bonds per atom, but each bond involves
two atoms; thus, the interaction energy is six times smaller than the cohesive
energy per atom.
Figure 1. (Caption given on the next page)
Figure 2. (Caption given on the next page)

Figure 1. Minimum programmable unit in square conformation consisting of four Au

clusters interconnected with four "dinitro" molecules consisting of four A11405 clusters
interconnecting four active molecular devices. The Au clusters are roughly spherical
with diameter of ~2.3 nm. The molecule, 2,5-dinitro-l,4-diethynylphenylthiolate-
benzene, also called dinitro is a phenyl-based oligomer known to perform negative
differential resistance NDR.
Figure 2. Time and frequency domain plots of the C-H bond vibration at 10 K for a)
the C-H in the top ring closer to the central ring and b) the C-H in the central ring.
Time domain signals are shown on the left and frequency spectra are on the right. The
analyzed bonds are boxed on the molecule.
The S-Au is also chosen to be a 12-10 potential in order to allow the S

alligator clip to attach to or to dissociate from the Au clusters. The
equilibrium bond length is set to 2.35 A, which is based on several density
functional theory calculations on similar systems, some of them already
reported[30], however, Do is set to 14.7 kcal/mol in order to make this bonded
interaction simulated by nonbonded fields and compatible with calculated
geometrical structures. Therefore, the Au-S interaction is chosen based on a
precise reproduction of geometrical features, rather than energetic ones.
Nevertheless, the bond energy of a S-Au bond is 38 kcal/mol, which can be
directly associated with roughly three Au atoms, thus corresponding to an
average of 13.0 kcal/mol per interaction, which is consistent with the 14.7
kcal/mol chosen for geometrical purposes. This is simply an average to have
an idea of the magnitude of the energies involved; it does not mean that the S
atom bonds equally to three Au atoms as already shown theoretically [31-34]
[16] and experimentally [35-37].
The vibrational analysis is performed using a series of simple filters
applied to the time dependant atomic displacement signals. Briefly, a
Hamming window is first applied to reduce spectral leakage. Lowpass filters,
bandpass filters, and signal rectifiers are then used to isolate selected
vibrations of interest. A very detailed description of the methods used in this
paper can be found elsewhere [38].
3. Results
Table 1 shows average bond length and vibrational frequencies obtained for all
the bonds analyzed in this study, when available, comparison to experimental
values of the corresponding bond stretching modes are provided. We focus our
study on the top and central rings of the three-ring molecule (Figure 1). It is
expected that the bottom ring behave similarly to the top ring. Two CH bonds,
one located on the top ring and the other in the central ring, are chosen for this
study since they represent two different environments and can provide
complementary information about other vibrations in the molecule. The C-H
on the top ring (C-Hcc) is located nearby the ethynyl group separating the two
rings; the C-H bond in the central ring (C-HNo2) can provide information about
the NO 2 substituents.
The other vibrations considered in this work are the C-C triple bond in the
ethynyl group joining the top ring to the central ring, two resonant bonds on
the top ring, the C-S bonds and C-S-Au angle, the scissoring and rotation of
the NO 2 , and the relative rotation between the top and the central ring. Table 2
summarizes the results obtained for NO 2 modes (scissoring, rotation).
Table 1. MU4 molecular dynamics results for bond stretchings. We show more than
one value of frequency when more than one peak of high intensity is observed in the
frequency spectrum.
Bond T(K) Average Vibrational Vibrational Experimental
bond Frequency Frequency Frequency
length (A) (THz) (cm'1) (cm'1)
10 1.084 92.4 3082
C H 100 1.084 92.4 3082
CC 300 1.085 92.5 3085 3O51i
1000 1.089 92.6 3089 3002 [39]

10 1.084 92.3 3079 J^m
C R 100 1.084 92.3 3079 3000[4°]
NO2 300 1.085 92.5 3085
1000 1.090 92.4 3082
10 1.206 54.9 1831
rr-r-i 10 ° L206 6L1 2038 , 2 1
C "CTnple 300 1.206 61.1 2038 2I00[4 °]
1000 1.208 61.1 2038
10 1.400 34.3/47.6 1144/1588
C-C 100 1.400 47.6 1588
Resonant 1,6 300 1.401 47.6 1588
1000 1.404 27.7/47.4 924/1581 1600.
10 1.401 49.2/55.9 1431/1865 1500[40]

C-C 100 1.402 48.9 1631
Resonant 4,5 300 1.402 55.7 1858
1000 1.406 55.7/48.9 1858/1631
C-S 10 1.789 —2 1434 794 [41]
Bond T(K) Average Vibrational Vibrational Experimental

bond Frequency Frequency Frequency
length (A) (THz) (cm' 1 ) (cm' 1 )
100 1.789 m2 1434 624 [42]
300 1.791 —2 1441/1144 582 [43]
1000 1.796 —2 1194/530/1438
' Average value for the C-H vibration is 3035 cm"1, 3044 cm"1, 3059 cm"', and 3067
cm"1, corresponding to the Wilson vibrational modes 2, 7, 13, and 20, which belong to
the C-H stretching region in a benzene molecule [44].
2' There are too many peaks, several of them around the experimental frequencies;
however, it is not possible an unambiguous determination.
Table 2. MU4 molecular dynamics results: NO2 scissoring and rotational modes. We
show more than one value of frequency when more than one peak of high intensity is
observed in the frequency spectrum.
Mode Average Vibrational Vibrational Experimental

T (K) Value Frequency Frequency Values
(degree) (THz) (cm"') (degree)
10 118.6 27.4/26.3 914/877
O-N-O 100 118.4 48.5/27.4 1618/914 124.4
scissoring 300 118.3 26.3/27.4/48.5 877/914/1618 (angle ONO)
1000 118.0 27.4/26.3 914/877
10 23.6 1.6 53.4

NO 2 100 24.1 1.3/1.6 43.4/53.4
rotational 300 26.7 1.3/1.6 43.4/53.4
1000 77.5
4. C-H Bonds
These bonds only differ on their van der Waals (vdW) interactions showing
only different magnitudes around the same frequency in the spectra. The C-
HCc and C-HN02 time and frequency domains plots at 10 K are shown in Figure
2a and b, respectively. Their equilibrium bond length and a vibration
frequency are very similar and not affected by their environment; however, the
maximum displacement, which is 0.016 A for C-HCc is only 0.010 A for C-
HNO2 because the NO2 restricts the H movement. In fact, the frequency
spectrum for C-HN02 (Figure 2c) shows a strong line located at about 27 THz,
which is the characteristic frequency for the scissoring modes of the NO 2
(Table 2). Two other strong lines at 49 and 55 THz are also common to both
the C-H vibrations and the scissoring mode; however, these lines are
associated with the C-C resonant bond vibration (vide infra).
Figure 3. Time and frequency domain plots of the C-H bond vibration at 100 K. for a)
analyzed bonds are boxed in the molecule.
The characteristic spectra at 100 K are shown in Figure 3 for the two C-H
bonds. Frequency and bond length do not change with respect to those at 10
K; however, the oscillation amplitude increases almost three times for C-H cc
(0.044 A), and more than three times for C-HN02 (0.036 A) with respect to the
observed amplitudes at 10 K.
Figure 4. Time and frequency domain plots of the C-H Bond vibration at T = 300 K
for a) the C-H in the top ring closer to the central ring and b) the C-H in the central
ring. Time domain signals are shown on the left and frequency spectra are on the
right. The analyzed bonds are boxed in the molecule.
At 300 K (Figure 4), both the equilibrium bond length and the oscillation
frequency slightly increase (0.001 A and 0.1 THz respectively) and the
maximum displacement increases three times with respect to that at 100 K
(0.095 for C-H cc and 0.087 A for C-H N02 ).
Figure 5. Time and frequency domain plots of the C-H bond vibration at 1000 K for a)
analyzed bonds are boxed on the molecule.
At 1000 K (figure 5), the peak at the characteristic frequency of the C-H
vibration spreads out and increases in intensity. All other peaks in the spectra
are very small; as a consequence, modulation is small. The equilibrium bond
length shows the largest increase but still limited due to the harmonic force
field used to simulate this bond, which does not allow bond dissociations only
allowed for Au-Au and Au-S bonds [38]. The vibrational frequency did not
change with the temperature. The maximum displacement duplicates with
respect to the case at 300 K (0.186, and 0.205 A, respectively, for C-HCc and
C-HNo2)- 1000 K is likely beyond the operating temperature for this system.
Figure 6. Time and frequency plots of the upper C-C bond in the top benzene ring at a)
10, b) 100, c) 300, and d) 1000 K. Time domain signals are shown on the left and
frequency spectra are on the right. The analyzed bond is boxed on the molecule.
The dominant frequency of each of the C-H vibrations we analyzed for the
10, 100, 300, and 1000 K simulations ranges from 92.3 THz (3079 cm-1) to
92.6 THz (3089 cm"1) (Table 1). For these cases, the stretching vibrational
frequencies lie very close to the experimental values 3035 cm"1, 3044 cm"1,
3059 cm"1, 3067 cm"1 corresponding to the Wilson vibrational modes no. 2, 7,
13, and 20, of the C-H stretching region in a benzene molecule [44].
5. C-C Resonant Bonds

The time and frequency domain spectra for two C-C resonance bonds in the
upper benzene ring are shown in Figure 6 and 7. Each figure shows the
vibration spectra of the 10, 100, 300, and 1000 K simulations for each of these
bonds. These vibrations are much less symmetric across their mean value than
the C-H bond vibrations.
Figure 6 shows the results for the upper C-C bond in the benzene ring.
The average length of this bond ranges from 1.400 A to 1.404 A (Table 1) as
the temperature increases from 10 K to 1000 K. The bond length is almost the
same up to 300 K (1.400 A for 10 and 100 K and 1.401 A for 300 K) and goes
to 1.404 at 1000 K. The frequency spectra for these vibrations have many
well-defined peaks that are present at all temperatures, but with varying
magnitudes. The peak that is consistently dominant relative to other peaks
over the entire range of temperatures is located at 47.6 THz (1586 cm"1) and
could be assigned to the characteristic frequency for this vibration; however,
this is not the highest peak for all the cases. A second peak at ~49 THz is also
present and could be due to the bottom C-C resonant bond as explained in the
following paragraph. The presence of several dominant peaks at all
temperatures is an indication of a strong correlation that the C-C resonant bond
has with other vibrations in the system. There is an isolated peak located
above 90 THz, which comes from the C-H vibration.
Figure 7 shows the results for the lower C-C bond in the top benzene ring.
The average length of this bond ranges from 1.401 A to 1.406 A as the
temperature increases from 10 K to 1000 K. The frequency content of this
bond vibration spectrum is similar to that of the upper C-C bond shown in
Figure 6. A well-defined peak at 49 THz (1633 cm"1) dominates the spectra
for the lower C-C resonant bond. This frequency was also observed in the top
C-C resonant bond with less intensity than the one at ~47 THz. Conversely, a
small peak at ~47 THz is also observed in the spectra of the bottom C-C
resonant bond. We can conclude, based on the relative intensity of these two
peaks that the top C-C bond vibrates with a characteristic frequency of 47.6
THz while the bottom C-C bond characteristic frequency is slightly shifted to
49 THz.
The C-C resonance bond vibration spectra vanishes after 60 THz, and a
well-defined peak takes place in the neighborhood of the C-H vibration, which
is around 92.5 THz. For the most part, the C-C vibration spectra are
characterized by several well-defined peaks in the range of about 0 to 70 THz.
These characteristics peaks can also be observed in the C-H vibration spectra.
This confirms the correlation between the C-H bond vibrations and the
vibrations of their neighboring C-C bonds. We can conclude that within the
C-C vibrations, we also find the contribution of the C-H bond vibrations, and
vice versa.
Figure 7. Time and frequency plots of the lower C-C bond in the top benzene ring at a)
10, b) 100, c) 300, and d) 1000 K. Time domain signals are shown on the left and
Figure 8. Time and frequency domain plots of the C-C triple bond vibration at a) 10,
b) 100, c) 300, and d) 1000 K. Time domain signals are shown on the left and
6. C-C Triple Bond

The C-C triple bond located between the upper and middle rings in the
molecular device is analyzed in Figure 8. The average length is 1.206 A up to
300 K, and 1.208 A for 1000 K. These values are really close to the
experimental bond length for ethyne (1.203 A [45]). At 10 K (Figure 8a), the
dominant peak in the frequency spectrum is located at 54.9 THz (1831 cm"1),
which is intermediate between the expected vibrational frequency of C-C
resonant bond (-48 THz) and C-C triple bond of about 61 THz (Table 1)
showing the strong effect that the ring vibrations have on this bond. For the
100, 300, and 1000 K spectra shown in Figure 8b to d, the dominant peak is
located at ~61 THz (2033 cm"1), which is in perfect agreement with the
experimental C-C vibrational frequency of 2011 cm"' [46] for ethyne HCCH.
A line at ~61 THz is also present at 10 K, but is not dominant.
Figure 9. Time and frequency domain plots of the C-S bond vibration results at a) 10,
b) 100, c) 300, and d) 1000 K. Time domain signals are shown on the left and
7. C-S Bond
The vibrational spectra for the C-S bond located at the top of the molecular
device are shown in Figure 9. The dominant frequency for 10, 100, and 300 K
is around 43 THz (1434 cm"1), which is certainly not due to the intrinsic C-S
expected to be -20 THz [41-43] (Table 1), but rather to contributions from
other vibrations that have been mixed into the C-S bond vibration. The
resonant C-C bond on the top of the ring has a vibration around this value, for
instance. A line at around 17 THz is present in the frequency spectrum at all
temperatures and it corresponds to the characteristic frequency for this
vibration. The spectrum in this range of frequencies shows several peaks from
neighbor atoms vibrations that makes difficult to characterize it.
3
Figure 10. Time and frequency domain plots of the Au-S-C angle bending at a) 10, b)
100, c) 300, and d) 1000 K . Time domain signals are shown on the left and frequency
spectra are on the right. The analyzed bond is boxed on the molecule.
8. Angle Au-S-C
The results of the Au-S-C angle bending mode are shown in Figure 10. The
average value for this angle increases from 98° to 120° as the temperature
increases from 10 to 1000 K. Ab initio calculations of the C-S-Au predict for
this angle a value in that range (-105°) [16, 33], and a temperature increase is
expected to yield a wider distribution of angles due to thermal agitation.
Notice that we have not calculated ab initio values for this angle within the
force field; S is not ligand to Au in our simulation thus a Lennard-Jones (LJ)
type of interaction is responsible for this grouping. LJ force field potential is a
two-body interaction, which is a particular case of the many-body Sutton-Chen
many-body potentials [26, 27]. Then it follows that the Au-S-C equilibrium
angle is a result mainly of the one to one interactions with surrounding atoms.
It can be inferred, when comparing to ab initio values, that the model used for
the S to Au atom interactions reproduces similar physical parameters of the S-
Au contact.
The vibration of the S to Au distance yields a spectra showing low
frequencies contributions. These contributions have several very close peaks
of similar amplitude indicating that the angle along Au-S-C is affected by low
frequency vibrations of nearby atoms, like the Au-Au vibration in the contact.
As for the C-S bond, the abundant low frequency contributions from the
contacts makes difficult to characterize this vibration.
Figure 11. Time and frequency domain plots of the O-N-O angle bending at a) 10, b)
100, c) 300, and d) 1000 K. Time domain signals are shown on the left and frequency
spectra are on the right. The analyzed bond is boxed on the molecule.
9. Angle O-N-O
The results for the O-N-0 angle bending are shown in Figure 11. The average
value for this angle decreases from 118.6° to 118.1° as the temperature
increases from 10 to 1000 K. The spectra for this angle-bending mode have
several well-defined peaks. At all temperatures there are three peaks located
around 25, 26, and 27.5 THz from which the peak at 27.5 THz dominates at all
temperatures (except at 300 K where the one at 26 THz is slightly more
intense). The strong peak located at 48.5 THz is due to the stretching mode of
the C-C bonds in the aromatic ring and modulates the scissoring vibrational
mode of the NO 2 .
(
Figure 12. Time and frequency domain plots of the C-C-C-C dihedral at a) 10, b) 100,
c) 300, and 1000 K. Time domain signals are shown on the left and frequency spectra
are on the right. A zoom to the 0-5 THz region is shown below the corresponding
time and frequency spectra. The analyzed bond is boxed on the molecule.
10. Relative Rotation Between Rings (Dihedral)

In order to study the relative rotation between two consecutive benzene rings,
we define an "improper" dihedral angle between the planes of the top and
central rings, which represents the relative orientation of the planes of each
ring. The fact that the two C atoms, one in each ring, defining this dihedral are
not formally bonded does not affect this definition since they are practically
co-linear with the two C atoms connecting them. We study the cosine of the
dihedral angle instead of the angle itself, in order to avoid discontinuities
inherent to the +180 to -180 jump.
The results obtained for the C-C-C-C dihedral angle rotation at 10, 100,
300, and 1000 K are shown in Figure 12. For the 10 and 100 K dihedral
rotations, shown in Figure 12a and b, the cosine of the angle stays close to 1
indicating that the dihedral in both cases oscillates around its initial value of
approximately 0°. The amplitude of oscillation increases when going from 10

to 100 K; however, it is unlikely that the rings will complete a rotation with
respect to one another at these temperatures. For the 300 and 1000 K dihedral
rotations shown in Figure 12c and d, several complete ring rotations are
observed.
The frequency spectrum of the relative rotation of the rings, is
characterized by several low frequency (less than 1 THz) peaks located very
close to each other as shown in the high resolution plots shown below the
corresponding time and frequency spectra in Figure 12 (notice the different
vertical scale between the low and high frequency detail plot). The magnitude
of the low frequency peaks is clearly much larger than the high frequency
peaks. Peaks between 0.5 and 1 THz reduce their intensity with T; whereas
for 10 K, the region in the spectra with important peaks goes from 0 to 1 THz,
for 1000 K this region goes from 0 to 0.5 THz.
The high frequency detail is very noisy and it is very difficult to obtain
information from it.
i An
3500JT ' --• ' ' : : ' :

i High Frequency Detail
2500 . . .
•S 2000. y •
a 1500 I : • . '
5Ot l ^ v '^
S ioo( I t - • • •
0 1 2 3 4 5
Frequency(Hz) xl0i!
(d)
Figure 13. Time and frequency domain plots of the O-N-C-C dihedral at a) 10, b) 100,
c) 300, and d) 1000 K, a zoon to the 0-5 THz regime is provided for this case. Time
domain signals are shown on the left and frequency spectra are on the right. The
analyzed bond is boxed on the molecule.
11. NO2 Rotation

To analyze the rotation of the NO 2 , we defined a dihedral angle between the
planes defined by O-N-C and N-C-C, respectively. We perform the analysis in
the same way as for the C-C-C-C dihedral. The results for the O-N-C-C
dihedral rotation are shown in Figure 13. At 10, 100, and 300 K the NO 2 does
not perform a complete rotation as it can be seen in Figure 13a, b, and c, it
oscillates around its initial position of approximately 0°, though the amplitude
of oscillation increases with temperature. Only at 1000 K does the NO2 make
complete rotations Figure 13d.
The spectrum for the NO2 rotation is less noisy than the one for the rings
relative rotation; however, the main frequency components spread and shift to
lower values when T increases, as it was observed for the C-C-C-C rotation.
For 10 K, the higher intensity components are around 1.6 THz; for 100 and
300 K they are below 1.6 THz, whereas for 1000 K, they are below 0.5 THz.
12. Correlation Between Vibrations

Several correlations between vibrations can be observed; for instance, the C-H
bond affects neighboring vibrations, C-C resonant bonds, sharing one C atom
with a C-H bond on the top ring, show a lone peak above 90 THz at all
temperatures (Figure 6 and 7). The influence of the CH vibration is instead
very small on the C-S bond, where a very small peak appear about 90 THz
(Figure 9), and it is inexistent on the C-C triple bond (see Figure 8)
The effect of the scissoring mode of the NO 2 in the central ring seems to
be more delocalized along the molecule than other vibrations; it affects the
dynamics of other bonds, and angles even if they are not close by. A peak
around -27 THz appears in several other vibrations located even on the top
ring. Both C-H resonant bonds, one on the top ring and the other on the
central ring, at all temperature, show a distinguishable peak at 27 THz (Figures
2 to 5). A very small peak at 27 THz can be observed on the C-C triple bond
(Figure 8) and the C-S vibration (Figure 9), although the peak in those
vibrations is just too small to conclude beyond doubt that NO2 is influencing
these vibrations. The C-S-Au angle seems not to be affected by the NO 2
scissoring mode (Figure 10).
The two C-C resonant bonds we treat in this work, both located on the top
ring, show a well defined, and of a relatively high intensity, peak located at
~28 THz (Figure 6 and 7), a small isolated peak at 28 THz is also observed in
the spectra of the NO 2 rotation (Figure 13), It is expected that the NO2 rotation
have some correlation to the scissoring mode, we can conclude then, that the
peak at 28 THz is due to the NO2 scissoring mode, slightly shifted to higher
frequencies. Several peaks around 27 THz exist in the relative ring rotation
spectrum that could correspond to the three peaks observed in the NO 2
scissoring mode (Figure 12), although the spectra in that region is very noisy
and it is difficult to assure such a correlation.
Two main frequencies, close to each other, are found for the C-C resonant
bonds on the top ring; however, the relative amplitude depends on the position
of the bond in the ring, the bond on the top has a main frequency of -47 THz
whereas for the C-C at the bottom, the frequency at 49 THz dominates. The
NO 2 scissoring mode also shows a peak at 48 THz corresponding to a nearby
C-C resonant bond.
It is difficult to follow the relative ring rotation and the NO 2 rotation in
other vibrations due to the low frequency the former two rotations have and
the very noisy nature of the spectra at low frequency.
13. Discussion and Conclusions

Frequency analysis applied to individual bonds, angles, and dihedrals, is
proposed as an alternative to the usual vibrational mode analysis. The analysis
we carried out in this work has several advantages, it is simpler than
identifying normal modes that in a system as big as the one in this work is a
nearly impossible task. Normal modes are perpendicular to each other, thus
correlations between different vibrations in the molecule cannot be obtained

from that analysis. Normal modes usually involve several vibrations along the
molecule, all vibrating at the same frequency; conversely, each vibration can
be part of several normal modes. The vibration of a bond, angle, or dihedral is
projected onto several normal modes thus analyzing normal modes only partial
aspects of the dynamics of each vibration can be seen.
The analysis we perform here, provides the entire frequency spectra
involved in each vibration; then correlating vibrations by identifying similar
frequency components, the effect that a particular vibration has on one part of
the molecule can be inferred.
Of particular interest is the correlation between a vibration in the molecule
that can be controlled externally and other vibrations. For instance the relative
rotations between the central ring and the other two, can be controlled by an
external field if, for instance a dipole moment is created in the central ring by
adding adequate substituents.
We found that some vibrations, such us the C-H vibration, affects strongly
neighboring vibrations, however, the rotation of the NO 2 group also affects
vibrations relatively further from it, C-C resonant bond, C-S vibration and C-H
vibration are affected by NO 2 rotation.
In addition to the vibrational properties, MD is also used to obtain
operational information, we determine that 1000K is most probably beyond
the operational T for this device.
The procedure used in this work is very promising although simple. By
analyzing spectral information on one bond, we can infer how to excite
another bond; thus, the dynamics of a different bond can be modified and
therefore applications in molecular electronics, and signal processing, among
others, are possible.
A potential scenario for the implementation of molecular electronic
circuits is introduced by applying digital signal processing techniques to
results from classical molecular dynamics simulations of a molecular system
interconnected by nanosize gold clusters. Under this new scenario, signals can
be introduced, processed, and read through interactions with the internal
vibrational modes of the small molecular unit. At least modulation operations
are intrinsically inherent to any molecular system. We demonstrated how
molecular dynamics techniques can be used for the analysis of molecular
electronic circuits; this is important since it will provide ideas as how the
molecular circuits can be implemented to perform signal processing. A
modified force-field is implemented for Au-Au and S-Au reproducing very
well geometrical features of these interactions, and yielding calculated
frequencies in very good agreement with available experimental information.
On the other hand, analyzing the vibrational spectra of each bond, angle or
dihedral of interest provides an alternative to normal mode analysis, which can

be complicated for a system as big as the one treated in this work. Since
normal modes are orthogonal to each other, an internal vibrational mode, such
as a bond stretching, may be part of more than one normal mode; thus the
analysis of normal modes only yields partial information for each internal
vibration. The frequency analysis focuses on oscillations to obtain all the
frequencies involved, rather than focusing on normal modes where several
oscillations of the same frequency are combined. Depending on a proper
selection of molecules, temperatures of ~300 K are acceptable, and the
possibilities for analog signal processing using molecules are wide open. Low
frequency oscillations can be externally controlled by electric fields, thus
modulating higher frequency oscillations.
Thus, we demonstrate numerically how molecular dynamics techniques
can be used for the analysis of molecular electronic circuits, and how a
molecular circuit can be implemented to perform signal processing. It is also
shown that the level of coupling between modes is enough to allow its
decoupling using standard techniques.
Analyzing the vibrational spectra of bonds using digital signal processing
techniques provides an alternative to normal mode analysis. A particular
internal vibrational mode, such as a bond stretching, may have contributions
from more than one normal vibrational mode. The analysis of normal modes
only yields partial information for each particular internal vibration; however,
the analysis reported in this brief finds correlations between different
vibrational modes, allowing signal modulation and processing.
The application of an input signal in one region of the molecular circuit
yields the modulation of bonds in other regions. Depending on the range of
the vibrations wavelengths, we can find out if they can be excited (optically,
bias potential, or in other ways), thus we find that modulation can be
performed using standard IR and Raman spectroscopies. A near-future
application is the development and design of molecular-size filters and the
implementation of cellular automata techniques. Thus chemical derivatives of
particular molecules can be proposed in order to use a characteristic of a
particular bond to perform signal processing; therefore, a carefully study of the
effect of several types of bonds when more than one signal is applied needs to
be determined. This detailed analysis of the frequency spectra yields
correlations between several local vibrations that can be used to include
vibrational signals externally introduced through the contacts of the molecular
device. Although these correlations between internal modes are not surprising,
the question is if using the natural amplitude oscillations of the vibrational
modes, which can be experimentally detected, will allow the detection of
mixed signals in localized sites. If this were possible, several single operations
could be implemented, enabling molecular processors dealing with signals in

the THz range of frequencies. We propose that signals can be encoded as the
movement of atoms that can be detected observing the changes in the electrical
characteristics of the molecule. For instance, when the displacement, f, of
atoms associated to a vibrational mode changes the dipole moment of the
molecule <", a signal proportional to the so called infra red (IR) intensity is
obtained from
IR Intensity «= —
KorJ
and when atoms displacements affect the polarizability, <*, a signal
proportional to the so called Raman intensity is obtained,
Raman Intensity«
V& )
These two intensities correspond to the most common experimental
spectroscopic techniques, and can be unambiguously determined theoretically,
and are potential candidates to be used as means for encoding information
because, on the other hand, induced external changes in the dipole and
polarizability by external fields yield changes in the displacements of the
atoms in the molecule. Thus, signals are internally represented by atom
displacements and transduced to IR and Raman intensities.
Depending on a proper selection of molecules, temperatures of ~300 K are
acceptable, and the possibilities for analog signal processing using molecules
are wide open. Low frequency oscillations can be externally controlled by
electric fields, thus modulating higher frequency oscillations. For example,
inducing a motion pattern on the Au clusters (input signal), can modulate the
C-H oscillation amplitude (output signal).
Acknowledgments
We highly appreciate the support of ARO under grants DAAD19-00-1-(0154,
0592, 0634), DARPA under grant N00014-01-1-0657. We also thank David
McClard, Nagendra Vadlamani, Jimena Bastos, and Sridhar Bingi for their
assistance with the details of the manuscript.
References
(1) Seminario, J. M.; Cordova, L. E.; Derosa, P. A., "An ab initio approach to the
calculation of current-voltage characteristics of programmable molecular
devices," Proc. IEEE, vol. 91, pp. 1958-1975, 2003.
(2) Tour, J. M.; Kosaki, M.; Seminario, J. M, "Molecular Scale Electronics: A
Synthetic/Computational Approach to Digital Computing," J. Am. Chem. Soc,
vol. 120, pp. 8486-8493, 1998.
(3) Ellenbogen, J. C; Love, J. C, "Architectures for Molecular Electronic
Computers: 1. Logic Structures and an Adder Designed from Molecular
Electronic Diodes," Proc. IEEE, vol. 88, pp. 386-426, 2000.
(4) Joachin, C; Gimzewski, J. K.; Aviram, A., "Electronics Using Hybrid-
Molecular and Mono-Molecular Devices," Nature, vol. 408, pp. 541-548, 2000.
(5) Kwok, K. S.; Ellenbogen, J. C, "Moletronics: Future Electronics," Mater.
Today, vol. 5, pp. 28-37, 2002.
(6) Tour, J. M.; Reed, M. A.; Seminario, J. M.; Allara, D. A.; Weiss, P. A.,
"Molecular Computer," in US Patent 6,430,511, 2002.
(7) Joachim, C, "Bonding more atoms together for a single molecule computer,"
Nanotechnology, vol. 13, pp. R1-R7, 2002.
(8) Seminario, J. M.; Cordova, L. E., "Toward Multiple-Valued Configurable
Random Molecular Logic Units," Proc. IEEE Nanotechnology Conf, vol. 1, pp.
146-150,2001.
(9) Seminario, J. M.; Cordova, L. E.; Derosa, P. A., "Search for Minimum
Molecular Programmable Units," Proc. IEEE Nanotechnology Conf., vol. 2, pp.
421-424,2002.
(10) Chen, J.; Reed, M. A.; Rawlett, A. M.; Tour, J. M., "Large On-Off Ratio and
Negative Differential Resistance in a Molecular Electronic Device," Science,
vol. 286, pp. 1550-1552, 1999.
(11) Reed, M. A.; Chen, J.; Rawlett, A. M.; Price, D. W.; Tour, J. M., "Molecular
Random Access Memories," Appl. Phys. Lett., vol. 78, pp. 3735-3737, 2001.
(12) Tour, J. M.; VanZandt, W. L.; Husband, C. P.; Husband, S. M.; Wilson, L. S.;
Franzon, P. D.; Nackashi, D. P., "NanoCell Logic Gates for Molecular
Computing," IEEE Trans. Nanoteck, vol. 1, pp. 100-109, 2002.
(13) Tour, J. M.; Jones, L., II; Pearson, D. L.; Lamba, J. S.; Burgin, T.; Whitesides,
G. W.; Allara, D. L.; Parikh, A. N.; Atre, S., "Self-Assembled Monolayers and
Multilayers of Conjugated Thiols, a,w-Dithiols, and Thioacetyl-Containing
Adsorbates. Understanding Attachments Between Potential Molecular Wires
and Gold Surfaces," J. Am. Chem. Soc, vol. 117, pp. 9529-9534., 1995.
(14) Harriott, L. R., "Limits of Lithography," Proc. IEEE, vol. 89, pp. 366-374,
2001.
(15) Szent-Gyorgyi, A., "Towards a New Biochemistry," Science, vol. 93, pp. 609-
611,1941.
(16) Seminario, J. M.; Zacarias, A. G.; Derosa, P. A., "Analysis of a dinitro-based

molecular device," J. Chem. Phys., vol. 116, pp. 1671-1683, 2002.
(17) Chen, J.; Wang, W.; Reed, M. A.; Rawlett, A. M.; Price, D. W.; Tour, J. M,
"Room-Temperature Negative Differential Resistance in Nanoscale Molecular
Junctions," Appl. Phys. Lett, vol. 77, pp. 1224-1226, 2000.
(18) Molecular Simulations Inc., User Guide. San Diego: Molecular Simulations,
Inc, 1997.
(19) Molecular Simulations Inc., "Cerius2." San Diego, 1999.
(20) Allen, M. P.; Tildesley, D. J., Computer Simulation of Liquids. Oxford:
Clarendon Press, 1990.
(21) Rappe, A. K.; Casewit, C. J.; Colwell, K. S.; Goddard III, W. A.; Skiff, W. M.,
"UFF, a Full periodic Table Force Field for Molecular Mechanics and
Molecular Dynamics Simulations," J. Am. Chem. Soc, vol. 114, pp. 10024-
10035, 1992.
(22) Casewit, C. J.; Colwell, K. S.; Rappe, A. K., "Application of a Universal Force-
Field to Organic Molecules," J. Am. Chem. Soc, vol. 114, pp. 10035-10046,
1992.
Field to Main Group Compounds," J. Am. Chem. Soc, vol. 114, pp. 10046-
10053, 1992.
Field to Metal-Complexes," Inorg. Chem, vol. 32, pp. 3438-3450,1993.
(25) Senn, H. M., "Ph.D. thesis, Diss. ETH No. 13972, ETH, URL: http://e-
collection.ethbib.ethz.ch/show?type=diss&nr= 13972." Zurich, 2001.
(26) Sutton, A. P.; Chen, J., "Long-Range Finnis Sinclair Potentials," Phil. Mag.
Lett, vol. 61, pp. 139-146, 1990.
(27) Liem, S. Y.; Chan, K., "Simulation study of platinum adsorption on graphite
using the Sutton-Chen potential," Surface Science, vol. 328, pp. 119-128, 1995.
(28) Lide, D. R., "CRC Handbook of Chemistry and Physics," vol. 79th Edition
1998-1999. Boca Raton: CRC Press, 1998.
(29) Seminario, J. M.; Tour, J. M., "Systematic Study of the Lowest Energy States
of Aun (n=l-4) Using DFT," Int. J. Quantum Chem., vol. 65, pp. 749-758, 1997.
(30) Derosa, P. A.; Zacarias, A. C; Seminario, J. M., "Application of Density
Functional Theory to the Study and Design of Molecular Electronic Devices:
The Metal-Molecule Interface," in Reviews in modern quantum chemistry, Sen,
K. D., Ed. Singapore: World Scientific, 2002, pp. 1537-1567.
(31) Seminario, J. M.; Zacarias, A. G.; Derosa, P. A., "Analysis of a Dinitro-Based
Molecular Device," J. Chem. Phys., vol. 116, pp. 1671-1683, 2002.
(32) Seminario, J. M.; De La Cruz, C. E.; Derosa, P. A., "A Theoretical Analysis of
Metal-Molecule Contacts,"./. Am. Chem. Soc, vol. 123, pp. 5616-5617, 2001.
(33) Derosa, P. A.; Seminario, J. M., "Electron transport through single molecules:
Scattering treatment using density functional and green function theories,"
Journal ofPhysical Chemistry B, vol. 105, pp. 471-481, 2001.
(34) Seminario, J. M.; De la Cruz, C. E.; Derosa, P. A., "A theoretical analysis of
metal-molecule contacts," J. Am. Chem. Soc, vol. 123, pp. 5616-5617, 2001.
(35) Ulman, A., "Formation and Structure of Self-Assembled Monolayers," Chem.
Rev., vol. 96, pp. 1533-1554, 1996.
(36) Heister, K.; Rong, H.-T.; Buck, M.; Zharnikov, M.; Granze, M., "Odd-Even
Effects at the S-Metal Interface and in the Aromatic Matrix of Biphenyl-
Substituted Alkanethiol Self-Assembled Monolayers," J. Phys. Chem. B, vol.
105, pp. 6888-6894,2001.
(37) Haag, R.; Rampi, M. A.; Holmlin, R. E.; Whitesides, G. M., ".Electrical
Breakdown of Aliphatic and Aromatic Self-Assembled Monolayers Used as
Nanometer-Thick Organic Dielectrics," J. Am. Chem. Soc, vol. 121, pp. 7895-
7906, 1999.
(38) Seminario, J. M.; Derosa, P. A.; Bozard, B. H.; Chagarlamudi, K., "Vibrational
Study of a Molecular Device using Molecular Dynamics Simulations," J.
Nanoscience Nanotech, vol. 5, pp. 1-11, 2005.
(39) Li, J.; Li, B.; Zhang, X. C; Su, R. Z., "The study of flame retardants on thermal
degradation and charring process of manchurian ash iignin in the condensed
phase," Polymer Degradation and Stability, vol. 72, pp. 493-498, 2001.
(40) "http://wwwchem.csustan.edu/Tutorials/INFRARED.HTM," 1998.
(41) Lu, G. W.; Xia, H. R.; Wang, X. Q.; Xu, D.; Chen, Y.; Zhou, Y. Q., "Raman
scattering investigation of the zinc cadmium tetrathiocyanate single crystals,"
Materials Science and Engineering B-Solid State Materials for Advanced
Technology, vol. 87, pp. 117-121, 2001.
(42) Kang, Y. U.; Hwang, T. S.; Song, H. Y.; Son, W. K.; Park, J. K, "The synthesis
and characteristics of soluble PAES/PEES copolymers for heat resistant ion
exchange membrane," Polymer-Korea, vol. 23, pp. 1-7, 1999.
(43) Tada, H.; Ito, S., "Conformational change restricted selectivity in the surface
sulfonation of polypropylene with sulfuric acid," Langmuir, vol. 13, pp. 3982-
3989, 1997.
(44) Alcolea, M., "Scaling Factors for the Prediction of the Frequencies of the Ring
Modes in Benzene Derivatives," J. Phys. Chem. A, vol. 103, pp. 11366-11377,
1999.
(45) Hehre, W. J.; Radom, L.; Schleyer, P. v. R.; Pople, J. A., Ab Initio Molecular
Orbital Theory. New York: Wiley, 1986.
(46) Strey, G.; Mils, I. M., J. Mol. Spectrose, vol. 59, pp. 103, 1976.
Chapter 4: Computation of Excited State Potential Energy
Surfaces via Linear Response Theories Based
on State Specific Multi-Reference Coupled
Electron-Pair Approximation Like Methods
Sudip Chattopadhyay, Dola Pahari, Uttam Mahapatraa

and Debashis Mukherjee6
Department of Physical Chemistry

Indian Association for the Cultivation of Science
Calcutta 700 032, India
* Department of Physics, Darjeeling Government College

Darjeeling 734 101, India
hAlso at Jawaharlal Nehru Centre for Advanced Scientific Research Bangalore

560 064, India
Formulation and pilot numerical implementation of linear response theo-

ries (LRT) for excited state potential energy surfaces (PES) are presented,
which are based on a class of state specific (SS) multi-reference coupled
electron-pair approximation (CEPA)-like methods (SS-MRCEPA). These
theories,termed as MRCEPA-LRT, are suited for the excited states whose
ground state either have intruders at some region of the PES, or displays
real or weakly avoided crossings. Genesis of the SS-MRCEPA approxi-
mations is traced from the parent full-blown state specific multi-reference
coupled cluster theory (SS-MRCC), and its efficacy to provide accurate
PES for the ground state in an intruder-free manner is emphasized. Two
CEPA versions are discussed, and the MRCEPA-LRT emerge as the LRT
based on the two SS-MRCEPA. They provide size-intensive energies in
the sense that they yield excitation energies for the fragments in the limit
of noninteracting fragments, just like in the single-reference CCLRT. The
set of MRCEPA-LRT developed in this article bears close structural re-
semblance to our LRT based on the SS-MRCC method (MR-CCLRT),
just as the SS-MRCEPA does to the parent SS-MRCC method. Pilot
applications to the PES of some low-lying excited states of the P4 model
and Li2 molecule, whose ground states possess quasi-degeneracy or face
intruders indicate very satisfactory performance, comparing quite favor-
ably with the results of the full-blown MR-CCLRT, despite the neglect
of a host of complicated terms.
121
122 S. Chattopadhyay et al.
Glossary of Terms
ACPF Averaged Coupled Pair Functionals
AQCC Averaged Quadratic Coupled Cluster
BW Brillouin-Wigner
CAS Complete Active Space
CAS-CI Complete Active Space based Configuration Interaction
CC Coupled Cluster
CCSD(T) A specific scheme incorporating non-iterative inclusion of
triples
CEPA Coupled Electron-pair Approximation
CEPA(D) CEPA with diagonal dressing
CI Configuration Interaction
EPV Exclusion Principle Violating
FCI Full CI
HF Hartree-Fock
LRT Linear Response Theory
MBPT Many-body Perturbation Theory
MR Multi-reference
MR-CISD Multi-reference Configuration Interaction Single and
Double
MRCC Multi-reference Coupled Cluster
MR-ACPF Multi-reference Averaged Coupled Pair Functionals
MR-AQCC Multi-reference Averaged Quadratic Coupled Cluster
MRCEPA Multi-reference Coupled Electron-pair Approximation
MR-MBPT Multi-reference Many-body Perturbation Theory
MR-CCLRT State Specific Multi-reference Coupled Cluster based
LRT
MRCEPA-LRT State Specific Multi-reference Coupled Electron-pair
Approximation based LRT
mH milli Hartree
PES Potential Energy Surface
RS Rayleigh-Schrodinger
SCF Self Consistent Field
SD Single and Double
SR Single Reference
SRCC Single Reference Coupled Cluster
SR-CCLRT Single Reference Coupled Cluster based LRT
SR-CISD Single Reference Configuration Interaction Single and
Double
SS State Specific
SS-MRCC State Specific Multi-reference Coupled Cluster
SS-MRCEPA State Specific Multi-reference Coupled Electron-pair
Approximation
Computation of Excited State Potential Energy Surfaces 123
1 Introduction
There are now well-established theories of electron correlation for closed-

shell systems such as configuration interaction (CI) method [l], many-body
perturbation theory at low orders [2], coupled electron-pair approxima-
tion (CEPA) methods [3], and their different variants [4; 5; 6], and single-
reference coupled cluster (SRCC) theory at various levels of truncations [7;
8]. While the truncated CI's, like the single and double(SD) excitations on a
single reference function (SR-CISD), do not respect the important property
of size-extensivity [9], the other methods mentioned above explicitly satisfy
this criterion and have thus become increasingly popular; in particular the
SRCC method with SD [7; 8] and approximate noniterative inclusion of
triples, SR-CCSD(T) is being widely used nowadays [10]. The CEPA-like
approximations can be viewed both as Cl-like methods corrected for inex-
tensivity by deleting the disconnected terms in the CISD equations and as
suitable approximation of the SRCC theory where only the EPV (Exclusion
Principle Violating) terms stemming from powers of the cluster operators
are retained. On the other hand, development of correlation theories for
excited or ionized states, for biradicals or for studying potential energy sur-
faces (PES) involving quasi-degeneracy of varying degrees-involving real or
weakly avoided crossings-has been a real challenge. Because of its struc-
tural simplicity, the variational approach was the most straightforward to
generalize, leading to the multi-reference CI method with the inclusion
of linearly independent single and double excitations out of the reference
functions (MR-CISD) [ll]. This has been used quite extensively. The
MR-CISD method suffers from lack of size-extensivity, and some multi-
reference generalizations of Davidson corrections have been proposed [12;
13; 14] to correct for this empirically. Several CEPA-like modifications
have also been formulated, where all the disconnected terms are eliminated
by inspection, and only the EPV terms are retained to various degrees [6;
15]. These modifications usually start with a reference function ip0, denned
in a complete active space (CAS), and include single and double excitations
from the CAS function. In almost all the methods, the combining coef-
ficients Cf, accompanying the model functions <^ in the final function are
kept frozen in their values as computed while generating tpQ via a CAS-SCF
or CAS-CI function. For real or weakly avoided crossings in a PES, there
would generally be a strong influence of the crossing state on the target state
of interest, which would lead to substantial revision of the coefficients c^ in
any theory based on an MR function generated from a set of <p^ spanning a

CAS. Two CEPA like approaches proposed rather recently in the MR situ-
ation alleviate this inflexibility by allowing the coefficients c^ to be relaxed
in the presence of the virtual functions included in the correlated state [16;
17]. Since in the CEPA formulations only the EPV terms are retained, the
inextensivity-correcting terms do not in general display the invariance with
respect to separate unitary transformations within the doubly occupied,
active and virtual orbitals defined in the CAS. To bypass this difficulty,
most MRCEPA schemes average the EPV contributions with respect to
the orbitals entering these terms [6] though this unfortunately confer on
the methods some inextensivity errors. It also turns out that the proper
way to correct for the inextensivity from an MR-CISD approach is harder
to discern in the MR case, while in a CC setting it is somewhat suggested
in a natural manner [16; 17]. To develop MRCEPA schemes from multi-
reference coupled cluster (MRCC) methods, which are size-extensive and
also potentially possessing the invariance property mentioned above is thus
a very worthwhile pursuit.
In contrast to the MRCEPA strategy, which is state specific (SS), the
traditional MR-based many-body methods like MR-MBPT [18] or MRCC
theories [19; 20] usually are based on effective hamiltonians and thus gen-
erate N target states at a time for an N-dimensional model space. While
monitoring size-extensivity in these theories is somewhat easier as compared
to state specific theory, and even the orbital invariance is also maintained
in the MRCC theories, their practical implementation remains problem-
atic due to the ubiquitous intruder problem [2l]. Intruders are virtual
functions which come energetically close to some model functions at some
regions of the PES, which spoil the convergence of the equations for the
amplitudes of the wave operator and hence of the effective hamiltonian
leading finally to poor values of the associated eigenvalues. Intermedi-
ate hamiltonian methods have been suggested [22; 23; 24], which target
fewer (M < N) states for an N-dimensional model space, and this flex-
ibility is exploited to improve convergence. As an extreme case of the
intermediate strategy, which however does not partition the model space
into main and intermediate subspaces as in the original formulation [22],
one may envisage developing state specific MRCC theories, or perturbative
and CEPA-like approximants thereof where one targets only one state at
a time. Three allied methods have been suggested in recent times [25; 26;
27; 28]. The method of Malrieu and co-workers [25] is closer in spirit to
a MRCEPA approach couched in a dressed CI form, which makes system-

atic inclusion of other terms stemming from the powers of cluster opera-
tors somewhat difficult to discern. The method of Mahapatra et al. [27;
28] is a full-blown state specific MRCC (SS-MRCC) , displaying mani-
fest size-extensivity and size-consistency. The formalism of Masik and
Hubac [26] leads to simpler equations than in the formulation of Maha-
patra et al. [27; 28], though it is not fully size-extensive. We have sug-
gested recently appropriate size-extensive MRCEPA like approximants [16;
17] to the parent SS-MRCC formalism of Mahapatra et al. [27; 28], and
preliminary numerical applications indicate effectiveness of these methods.
They avoid intruders naturally whenever the target state energy remains
rather away from the virtual function energies. While this is always achiev-
able for the ground states of a system, even those with pronounced MR
character and showing real or avoided crossings, a straightforward appli-
cation of the SR-MRCC or the associated SS-MRCEPA approximants de-
rived therefrom, would lead in general to intruder problem. To get around
the difficulty, we have recently developed a linear response version [17;
29] of the parent SS-MRCC theory, where the excited states are gener-
ated by solving an eigenproblem generated by an excitation operator on
the ground state. The linear response theory (LRT), termed by us as MR-
CCLRT, led to a very good performance to generate several PES for excited
states of different symmetries for difficult systems such as BeH2. Since the
SS-MRCEPA methods proposed by us provide rather accurate energies as
compared to the parent SS-MRCC method, despite the neglect of a plethora
of nonlinear terms, it seems worthwhile to develop linear response theories
based on our recently formulated SS-MRCEPA. We present in this article
formulation of such an approach, alongwith some pilot numerical appli-
cations to the excited PES for systems showing quasi-degeneracy in their
ground states to indicate the efficacy of this approach.
The success of the linear response approach based on a given form of
the ground state is predicated by the quality of the ground state function.
Although the LRT based on SRCC theories are in vogue for quite some
time [30; 31; 32; 33; 34], not much has been attempted to generalize this
to the MR case. Pal et al. [35] has developed MR-based LRT based on
effective hamiltonian approach in the frame work of the valence-universal
formalism. Response to a time dependent field and its static limit yield-
ing time-independent equations has been derived by Ajitha and Pal [36].
Another MRCC based LRT has been proposed by Tenno et al. [37] in the
state-universal setting. Szalay [38] has developed a constrained variational

approach for the MRCC method, which is relevant in the context of the
linear response method, where the first derivative amplitudes are naturally
not required in the expression for the first derivative energy.
A linear response extension of the state specific MRCEPA methods such
as MR-ACPF (multi-reference averaged coupled pair functionals) and MR-
AQCC (multi-reference averaged quadratic coupled cluster) was formulated
by Szalay and co-workers [39]. Gadanitz and Ahlrichs [4] introduced the
MR-ACPF based on the CEPA-type arguments. Latter on, Szalay and
Bartlett developed a new method called MR-AQCC [5; 6] which in sprit and
level of approximations is closely related to MR-ACPF. The size-consistent
self-consistent complete active space singles and doubles CI [(SC)2CAS-
CISD] has been applied very recently by Martin et al. [40] to computation of
vertical excitation as well as the ionization energies. As mentioned earlier,
Chattopadhyay et al. [29] have recently proposed a LRT based on the SS-
MRCC theory (MR-CCLRT) using the same strategy as that of the SRCC
based LRT (SR-CCLRT) [30; 31; 32; 33; 34] for computing the excited state
energies. It has been shown that the MR-CCLRT formalism [29] is size-
intensive in nature [34; 41], in the sense of providing excitation energies of
fragments in the limit of non-interacting fragments.
The organization of the paper is as follows. In Sec.2, after a brief dis-
cussion of the structure of our SS-MRCEPA and our SS-MRCC based lin-
ear response theory, MR-CCLRT [29], we discuss two physically motivated
approximations of graded complexity, based on our parent MR-CCLRT
generated from the SS-MRCC guided by the same type of approximations
we made in arriving at the SS-MRCEPA schemes. Issues such as size-
intensivity and orbital invariance will also be addressed to. In Sec. 3, we
present the numerical examples to test the potentiality of our MRCEPA-
LRT as compared to the FCI values. Sec. 4 summarizes the main thrusts
of the article, alongwith the concluding remarks.
2 Theoretical Developments
2.1 The SS-MRCC Theory and Its CEPA-like Approximants

Before presenting the formulation of the linear response method (MRCEPA-
LRT), based on our SS-MRCEPA, we would first introduce and discuss the
structure of the parent SS-MRCC equations. This would serve both to fix
the notations for the key ingredients entering our formalism, and also to
motivate towards the approximations warranted in the CEPA-like methods.
We would discuss then emergence of two CEPA-like approximants to the
SS-MRCC method. We would next introduce the essentials of the full-
blown MR-CCLRT derived as the appropriate linear response version, and
then proceed to approximate them to generate MRCEPA-LRT in a manner
entirely analogous to the steps taken to reach the SS-MRCEPA schemes
from the SS-MRCC method.
We start out with a CAS spanned by a set of model functions <£M, which
also specifies the set of doubly occupied core orbitals, the set of active
orbitals and also the set of virtual orbitals. We write the ground state
function t/j0 in a cluster expansion representation of the component of a SS
wave operator defined for each (^ it acts on;
^ = êxp(r")^cM (1)
Our theory is quite flexible regarding the choice of the coefficients cM.
We may, if we so desire, keep them frozen at their initial values obtained
by the CAS-SCF or CAS-CI procedure, or we may determine them self-
consistently alongwith the cluster amplitudes of TM. Since each virtual
function \i c a n be reached from every model function <^>M, there is a re-
dundancy in the number of cluster amplitudes. In SS-MRCC theory, this
redundancy is exploited by positing sufficiency conditions to attain the
twin desirable goals: (a) maintenance of manifest size-extensivity and (b)
avoidance of intruders in a natural manner. The working equations for de-
termining the cluster amplitudes leading to excitations to \i from </v a r e
given by
(Xi\HM»K + £<X<|exp(-T<i)exp(rO|<^)iVc1, = 0 V/,M (2)

V
where H^ — i7exp(J>) and Htlv=((j>ll\Hv\<j)v). Xi's are all the excited

states reached by the action of TM on 4>\i •
From now on, we assume that we solve for the coefficients c^ self-
consistently by diagonalizing the effective matrix H^:
128 5. Chattopadhyay et al.
^HpVcv = Ecll (3)

V
We note that the set cM and T** are coupled through Eq.(2) and Eq.(3).
Solving these coupled set of equations, we obtain both the cluster ampli-
tudes and the converged coefficients from the diagonalization.
We now discuss the development of coupled electron-pair like approxi-
mations of our SS-MRCC theory leading to the SS-MRCEPA methods.
There could be several routes to the CEPA-like approximations starting
from the SS-MRCC theory which, while capturing the physical essence of
the CEPA-like schemes for the single-reference situation, emphasize differ-
ent aspects of technical details. The fundamental idea in any CEPA-like
method is to eliminate the size-inextensive terms coming from the energy
containing term of a MR-CISD. They are not cancelled automatically in a
MR-CISD if multiple excitations beyond the CISD space are not included.
One may imagine that the simplest way to eliminate them is to identify the
disconnected terms and subtract them explicitly in the MR-CISD equations.
As we would emphasize below, in contrast to the single-reference situation,
this analysis does not exactly carry through in the multi-reference case.
It turns out that the disconnected looking terms in the cluster expansion
representation of the function ipo have both connected and disconnected
components, and some counter terms which automatically arise in the clus-
ter expansion eliminate just the disconnected portion. The difference of the
term with the disconnected component and the counterterm is connected,
but these two entities contain different cluster operators TM and Tv. This
kind of cancellation with counterterms would have been very hard to guess
in a MR-CISD setting. What should obviously be a common character-
istic is the inclusion of the singles and doubles space interacting via the
hamiltonian. In the multiple cluster expansion of the wave-function, what
constitutes the SD-space is not unique, and it is here that the SS-MRCEPA
methods have a freedom of choice. One may, for example, emphasize that
one should include only the single and double excitations {xf} reached
by the action of TM on <^>M in Eq.(2) for every \f used in the projection.
Another possibility could be to include all the virtual functions which are
obtained from single and double excitations from the entire model space.
In this latter situation, a virtual function which is singly or doubly excited
with respect to a given cp^ may be more than doubly excited with respect
to another model function <f>Vy and more than quadratic powers of cluster
operator are generally needed for inducing excitation to all functions in the
CISD space. There are structural differences in the CEPA-like schemes ob-
tained with the above two choices. In the former (scheme I), upto quadratic
powers of one-body operators Tf can be included, and the quadratic powers
of T% should include only the EPV terms. In the latter scheme (scheme II),
upto quartic powers of TM leading to the singles-doubles excitation space
with respect to the model space can appear. In the most general equation
of the Scheme II, as in Eq.(4), the projector QSD corresponds to the sin-
gles and doubles projector of the entire model space figures. However, in
our numerical implementation, we would approximate our equations by ne-
glecting more than quadratic powers of every TM, and retain only the EPV
terms coming from the powers of T2M. In this case, the projector reduces to
QM, the projector to the virtuals which are singles and doubles with respect
to the model function (p^, and the CEPA approximation reduces to what
we called scheme I above.
The most general CEPA-like equations for both the schemes can gener-
ically be written as
(xt\H^(P + QsD)exp(T^\^) + £< X f | exp(-T") exp(T")|^)
<^|ffM (P + QsD)exp{T")\tf>u)cv = 0 V/,/i (4)
As we emphasized above, we would rather use the less general QM in our

present article in the numerical implementation.
To see the explicit form of the CEPA-like approximations, we write the
above equation in long hand as follows:
\{XiW\^) + to|[#,m^> + !<Xj|[[ff,TV]|^>]c M

I II III
+ T,AXi\exp(-T>l)exp(Tl/)\(j>tt)H^Cv + other t e r m s =0 (,s
IV K '
There are several features of the equations which should be emphasized

here: (a) the combining coefficients cM are not frozen at some preassigned
values, but are iteratively updated to their relaxed values; (b) the first three
terms have explicitly connected algebraic structure in case T^ is connected;
(c) among the quadratic terms appearing in III, the terms with one-body ex-
citations are still doubles while the product of doubles or of singles and dou-
bles lead to virtual functions \i which are more than doubly excited with re-
spect to (j)^; (d) the fourth term contains two pieces, one containing T" and
the other containing TM, which may be individually disconnected but the
entire term containing the matrix element {xi\ exp(—TM)exp(T")|0M)H'A1^
is a connected entity [27; 28]. The aspect (d) is what we emphasized ear-
lier as generating a connected entity from two individually disconnected
pieces, and which could not have been derived from the MR-CISD equa-
tions. For a proof of this assertion, we refer to our earlier publications [27;
28]. This feature of the SS-MRCC equation, retained in generating the
SS-MRCEPA equations is very important for our purpose, since in any
approximation or truncation scheme, each term of the two pieces must be
treated on the same footing to ensure connectivity. Thus the term IV has
to be computed in a manner which uses the same approximation for the
matrix elements H^.
Using the same kind of considerations as were used to generate the var-
ious CEPA-like approximations from the SRCC-SD equations, we confine
the rank of cluster operators TM to at most two-body excitations. Thus for
each <j>n, the x;' s used in the projection in Eq.(5) above must be at most
doubly excited with respect to <^M. Moreover, the term III should include
product of single excitations leading to the doubles, and should exclude all
other product terms which are more than doubly excited with respect to
</>M. However, depending on the sophistication of the CEPA-like approxi-
mation, we will include the EPV terms of III coming from the product of
doubles to various degrees. For every xi reached by a T/* from 0M, these
EPV terms of III must include at least one Tf operator while the other TM
can be any one of the possible excitations which have at least one orbital
in common with those involved in T/\ To show this, we express the leading
terms of Eq.(5) explicitly in the following form
l(Xi\H\<f>») + Em(Hlm - H^Slm)t^ - Arf

I II III(a)
HE'mnd'm^t^ + Ev{Xl\Tv\<l>ll)HllvCl/
III(b) IV(a)
+other terms = Stfcu ,cs
IV(b) ^
where we have shown the genesis of all the terms by the same labels by
which they were indicated in Eq.(5). All CEPA schemes neglect the terms
not explicitly shown in Eq.(6). The leading terms of the two pieces of the
term IV containing T" and Tu are now separately indicated as IV(a) and
IV(b) respectively. The 'energy' £ is taken as a parameter, whose actual

value depends on the approximation used in H^. AM is the EPV correction
in III(a) coming from product of doubles or of singles and doubles, whose
actual form again depends on the approximation used in our SS-MRCEPA.
The negative sign of this term in Eq.(6) is to keep conformity with the
analogous correction term used in the SR case. The term III(b) denotes
the doubles contribution confined to the product of two singles (which is
indicated by the prime in the sum). We note at this stage that the terms
I-III are entirely similar to those in the SRCEPA method. In fact if we
neglect the terms IV(a) and IV(b), then we get precisely SRCEPA-like
methods for each cp^. But this approximation is too drastic in the sense
that it entails intruders (because of the implicit presence of (Hu - H^)
in Eq.(6)), and thus belies one of the very basic goal which motivated our
SS developments. We recall at this stage the feature which we highlighted
earlier: while two pieces of IV, i.e. IV(a) and IV(b) may be disconnected,
the entire term IV taken together is a connected entity. The two pieces
IV(a) and IV(b) serve two specific and distinct purposes in this regard. A
portion of the term in IV(a) for v = /i cancels both H^îm and A^tf terms
of II and III(a), while the rest of the expressions for v ^ [i of IV(a) correct
for the lack of extensivity coming from the terms appearing in Et^c^.
We now discuss two specific types of CEPA-like approximations. En-
tirely in line with the CEPA(O) scheme for the SR case, if we ignore
AM entirely in Eq.(6), then we should start generating an analogous SS-
MRCEPA(O) scheme. We still have the freedom of approximating H^v
independently, but in the spirit of quasi-linearization, we now approximate
Hfj,v as just H^v Then £ becomes just EQ, the CAS energy itself. This set
of approximations leads to what we call the SS-MRCEPA(O). Written in
long hand, it takes the form
{(Xi\H\^) + ]T(ff,m - E08lm)C + ^ Xy r o > n «]c,.
+ J2(xi\Tl'\MH^cv=0 (7)
with Eo satisfying the equation
V
We now discuss the most sophisticated approximation of AM, by retain-

ing all the EPV terms. This is analogous to what was suggested by Malrieu
and co-workers [15] in the SR case, which we denote as CEPA(D). The qual-
ifier D indicates that all such terms can be incorporated in the exact CEPA
by a diagonal dressing of the SDCI equations [15]. We approximate A^ by
all the terms containing all TMs with at least one orbital in common with
those appearing in t*. A^ in each equation for t% is thus /-dependent, and
we indicate this by A^. As emphasized earlier, we have to treat the two
pieces of IV on the same footing in their handling of H^,. We approximate
this by its quadratic expansion involving double commutators. The diag-
onal term H^ appearing in IV(a) of Eq.(6) for a given I can be written
as
Hnv — Hw + ^ ^ + ^ N E P V (8)
where A^ is the correction for TM with at least one orbital in common

with those labeling tf. The rest of the term has no orbital in common
with those appearing in tf. With this observation, it is clear that the
term (H^^ + AjJ<$jm in Eq.(6) is cancelled by the first two terms of Eq.(8).
Denoting the energy £ as £QEPA(D)> Eq.(6) for this scheme, referred to as
SS-MRCEPA(D), takes the form:
KXl\H\^) + Y,Wlm ~ (£CEPA(D) + £|,NEPv)Mtf +

m
(9)
where £ satisfies the equation
To see the avoidance of the intruders, we rewrite the SS-MRCEPA equa-

tions in the following form
_ (Xl\H\<t>,*} + I Em,n ff*m.n« + E ^ W I ^ M ^ / C M ) /1n ,

*' ~ S-Hu-D^ (10)
H^v — H^ for SS-MRCEPA(O), and has a multi-commutator structure

for SS-MRCEPA(D). £ = Eo for SS-MRCEPA(O), and is £CEPA(D) for
SS-MRCEPA(D). D^ = 0 for SS-MRCEPA(O) and has the value as given
by Eq.(8) for SS-MRCEPA(D). Since D^ is small, unlike in the effective
hamiltonian based theories, (£ — Hu — Djj will always be robust if £ is
well-removed from any Hu, even when H^ is close to this Hu. Intruders
are thus avoided naturally.
Structurally, the SS-MRCEPA(O) and the SS-MRCEPA(D) bear close
resemblance to the recently developed size-extensive SSMR perturbation
theory (SS-MRPT) [42; 43] of the Rayleigh-Schrodinger (SS-MRPT(RS))
and the Brillouin-Wigner (SS-MRPT (BW)) types, respectively, generated
from the SS-MRCC theory, but are generally more accurate than the per-
turbative schemes.
There is a real advantage of approximating the SS-MRCC theory in the
SS-MRCEPA form if we can retain the bulk of the correlation energy cap-
tured by the full-blown SS-MRCC theory, while we save a lot of computer
time since we neglect a host of nonlinear terms. Our results, discussed in
Sec.3, will amply demonstrate the accuracy of the SS-MRCEPA schemes.
2.2 Linear Response Theory Based on SS-MRCC (MR-

CCLRT) and Its Approximants Based on SS-MRCEPA
As discussed earlier, the major thrust in this paper is the development
and pilot numerical implementation of the linear response theories based on
the SS-MRCEPA methods, discussed in the preceding subsection. For the
excited states whose ground states are multi-reference in nature, this is a
very convenient and compact approach. Before embarking on the detailed
considerations leading to the MRCEPA-LRT based on the response func-
tion derived from SS-MRCEPA, it is pertinent to discuss at this stage the
essential issues of the linear response theory based on the parent SS-MRCC
(MR-CCLRT) itself.
In the CC based linear response theory, the excited functions are written
in terms of the ground state function via the action of a suitable excitation
operator. In the spirit of the SR-CCLRT [30; 32; 33; 34; 31] we posit on
our excited state, \tpk)-, the following Ansatz:
h/>*> = $ > x p ( T " ) 5 £ | < ^ (11)

\tl>k) satisfies the Schrodinger equation
H\il>k) = Ek\il>k) (12)
In close analogy with the SR-CCLRT, we include in each excitation operator

S% all the virtual excitations from |0M), and retain the cluster components
TM of the multi-reference ground state (eq(l)) in our Ansatz for \tpk}- In
addition, we also want to change the relative weights of the various |^M)'s
in the function 1^*)- With this choice, we can represent S% as
SZ = *fl + SZ with S» = Y,*?kY?' (13)

m
where Y™^ creates a virtual function \Xm) from \4>n)- It is evident that
there are several S£'s which excite to the same \xm) from several | ^ ) ' s .
Thus, we have to invoke a set of sufficiency conditions to determine them.
We have
£[exp(T^)|^><^|^S£|^> + exp(r")5f |^>(^|ff M |^)
+exP(r'X°|<M<<Mtf^> + E e x P( r A I )^^^l^>
+ exp(r")S^QH M |^) + exp(T")a^gH M |^>]c M
= Efcêxp(T^^|^)^ (14)
where Q is the full virtual space projector. Following the same manipula-
tions leading to the SS-MRCC equations [27; 28], we interchange the labels
H and v in the above equation in the first three terms and equate terms
with same fi. For each virtual space projection onto (x;| exp(-T^) we have
(Xl\%S%\<f>^ + (x^QH^K + {xilHMsfcp

+ ^{Xi\ exp(-T") e x p C n S j ^ X ^ l i ^ l ^
V
+ (xi\ exp(-T") eMTv)\^)(^\HvSvk\4>v)cv

+{Xi\ exp(-T*) exP(Tn\^)(^\Hu\<Pu}sfcu
= Ek(xi\S^K (15)
The sufficiency conditions imply that each virtual function (x;| can be
reached from several <£M's and the projection for each (xi\ has to be done
for each |(/>A). Also, each ket \xm) appears with a different coefficient,
depending on how it is generated via various 5£'s acting on the associated
\<f>^)'s. To keep this information as a mnemonic, we would from now on
denote (x;| as (xjj if it is used for projection from Eq.(15) for a certain
\(j),j) and function \xm) as |x™) if it is obtained by Y™^ acting on |^M)
From now on, we shall use the coefficients d^'s and d™fc's defined by
<*„* = 4 % (16)
<^k = *HW (17)

Using the above equations, we can rewrite Eq. (15) with the notation |x™)
for \xm) reached from | ^ ) :
+ Y,(xU exp(-r") eXp(T'')Yvmi\<i>li)(<t>r\Hv\<t>v)d%t

vm
+ JîxU exp(-r^) ex P (T-)|^)(^|F^t|^>C fc

vm
+ E<^l ex P(- r ") exp(T")|^)(^|^|^)^ fc

= Ekd\k (18)
The third term in Eq. (18) involves QH^ acting on \<f>^), which implies
that QH^\<p^} is an excited function (henceforth denoted by \Hf?\ )•
In the same manner we project Eq. (14) onto the active space manifold
(0 A |exp(-r' i ) and get
Y,{<l>\\HvS^\<t>v)cv + J2(4>x\HA<f>v)sfcv = Ekcxsf (19)
which on simplification generates,

£<0A|j?,*;mti6,><e* + ^ ( ^ I ^ I ^ ) ^ = EkdXk (20)
vm v
The above MR-CCLRT equations can be written compactly as a matrix

eigenproblem.
Once the cluster amplitudes of TM for all /i's have been found out by
solving the SS-MRCC equations, we can get the associated excited state
energies relative to the state \ip) by solving the MR-CCLRT equations.
An attractive feature of the MR-CCLRT is that the computed excitation
energies are size-intensive [29] in the sense that they become asymptotically
equal to the sum of the fragment excitation energies in the limit of non-
interacting fragments.
We will now develop the corresponding MRCEPA-LRT using exactly the
same approximations on MR-CCLRT as has been used in the SS-MRCC
theory to arrive at the SS-MRCEPA theories. To derive the working equa-
tions for the SS-MRCEPA based response theories, we start with the parent
MR-CCLRT equations (18) and (20) for the excited states and make ap-
proximations akin to what have been used to arrive at the SS-MRCEPA(O)
and SS-MRCEPA(D) theories from the SS-MRCC theory for the ground
state. Both the MRCEPA-LRT schemes can be subsumed in the same
type of working equations, with appropriate definitions of certain dressed
operators.
In our SS-MRCEPA-like methods, we ignore more than double excita-
tions out of any given </>^ leading to a xjj a n d quasi-linearize the terms
in [exp(—TM)exp(T")] accordingly. The working equations of the virtual
space projections of the CEPA-based MR-CCLRT Eq. (18) for both the
CEPA-based schemes can be simplified as follows
vm
+ Y.(xlMTl/-T»)\<t>>1)H<iiVdvk
V
+E<î(r"-ni^><^|SM^t|^>ccfc = ^ 4 * (21)
vm
The corresponding model space equation is
X ^ I H ^ I ^ C * + Y,{^\HAMdvk = EkdXk (22)

vm v
In the above working equations of the MRCEPA-LRT, some appropriate
approximant to the operator H^ figures whose actual form depends on
whether we are considering SS-MRCEPA(O) or SS-MRCEPA(D).
We give below the explicit forms of H^, and the dressed operators H^
and H"^ for different SS-MRCEPA methods:
For SS-MRCEPA(O) based response equations
\ m,n / (23)
H"v = Hv (24)
However, for the SS-MRCEPA(D) based response equations
%v = {^\Hv\4>u) (25)
H'l = Hv (26)
H^L + HTt + Hfi+lY, oljWin + \YS ^JUL) (27)

\ m,n m,n /
It should be mentioned here that there is at least one orbital in common

between t2m and t£n in the term £ m , n < 4 , « O L -
Just as in the parent MR-CCLRT [29], we can cast the working equa-
tions of MRCEPA-LRT also in the following matrix form:
/ Hvu 'Hlim\(dvk\_ ( d»k \
{ nlv nlm ){dfk )-Ek\d^k )
The elements of the operator H for the two MRCEPA-LRT are given by
(^Wfa) = (^[Hvfo) (28)
{^\n\xT) = {îWrFlM (29)

(xlM<t>») = <x^|<M<W + (ti\(Tv - T»)\^)H> »v (30)
(timix?) = ixUiT" - ^ i r ' i ^ i g y .

+{xlii\Yv^Qli[Hl,}\<t>v)8liV + ( x U f ^ l - M V
+(X1MT" - T^)\^)(^\WVY^\4>V) (31)
Once the cluster amplitudes of TM for all /z's have been found out by
solving the SS-MRCEPA equations, we can get the associated excited state
energies relative to the state \ip) by solving the MRCEPA-LRT equations
by constructing the matrix of % via Eqs.(21-31). The 5 amplitudes are
confined to single and double excitations only.
It is evident that if one drops the excitation operator, Y™^ from the
working equations of MRCEPA-LRT they naturally reduce to the corre-
sponding ground state CEPA theories, SS-MRCEPA. This implies that
SS-MRCEPA are the completely unperturbed versions of MRCEPA-LRT,
which is why we coin the term, MRCEPA-LRT, for the linear response the-
ory based on the SS-MRCEPA method. We not only get the excited state
energies but also the ground state energy by solving the MRCEPA-LRT
equations.
It is important to mention that, just as in the full-blown MR-CCLRT
[29], one can prove in the same manner that the underlying operators in
the eigenvalue problem of MRCEPA-LRT are connected and the possible
excitation energies also satisfy the size-intensivity criteria [29; 34; 41].
We conclude this section with the following comments. In general, the
CEPA-like methods do not possess the invariance property with respect
to separate unitary transformation of orbitals within the manifold of dou-
bly occupied, active and the virtual orbitals. A rigorous formulation of
MRCEPA theories with this property is rather hard. Interestingly, the SS-
MRCEPA(O) discussed in this paper does possess this property. The more
elaborate MRCEPA(D) unfortunately does not. We are recently exploring
a somewhat less elaborate SS-MRCEPA method, where this invariance is
restored, but at a price of having to include in the method a small set of
more than doubly excited CI space functions to maintain the invariance
[44]. These and the use of CAS-SCF orbitals to study their performance in
the MRCEPA context are some of the aspects of study for us in the near
future.
3 Results and Discussions
In our two applications, we employ a two-electron/two orbital, designated

as (2 x 2) model space, consisting of two active orbitals, denoted as a and
b and they belong to two different symmetries. This ensures the complete
model space is effectively two-dimensional, spanned by fa = [core]a2 and
<f>2 = [core]b2. The spin-adaptation of the working equations pertaining to
both SS-MRCEPA and MRCEPA-LRT is a rather simple exercise, since
the reference determinants are closed shell singlets.
In order to examine the efficacy and performance of our methodology,
we have applied our MRCEPA-LRT to study the PES of the excited states
of two molecules whose ground states display some typical difficulties of
treating states with quasi-degeneracy: (i)the ground state possesses a pro-
nounced MR character at some geometry in the PES, (ii) for these illustra-
tive molecular systems, the configurational and orbital quasi-degeneracies
vary continuously with geometry, (iii) there are intruders to some reference
determinant at some regions of the PES of the ground state.
The performance of our MRCEPA-LRT is strongly dependent on the
ability of the parent SS-MRCEPA methods to tackle both quasi-degeneracy
and the intruder problem. In our numerical applications, we have studied
PES of some low-lying excited states of the P4 model and Li2 molecule.
Our formalism treats both these determinants (fa and fa>) o n the same
footing, and we have two distinct options to choose our orbitals: (a) we
may choose either of the two determinants to generate the HF orbitals, or
(b) we may use the orbitals obtained from a CAS-SCF calculation. It turns
out that for the molecules studied by us in this paper, there is the following
rough general trend: the PES of the ground state obtained either by the
SS-MRCEPA or as the lowest root from the MRCEPA-LRT is improved
generally, by going over the CAS-SCF orbitals. For P4 model, the results
of the ground state obtained by CEPA(O) method show improvement by
around 0.3-0.9 mH and for CEPA(D) this varies from around 0.03-0.45
mH. At the square geometry (highly degenerate case), this improvement
is maximum. The performance of the MRCEPA-LRT for the excited PES
with the CAS orbitals is not uniformly better however, over the entire
PES. For the P4 model, the results obtained by either of the two variants
of MRCEPA-LRT are altered by 1-3 mH relative to the actual choice of the
orbitals, as described above. Similarly, for L12 molecule, the ground state
energies obtained using CAS orbitals are improved by around 0.4-1.5 mH
and 0.42-0.8 raH in case of CEPA(O) and CEPA(D) methods respectively.

In the dissociation region, the improvement in either method using CAS
orbitals over the HF one is negligibly small (around 0.006 mH). However, for
Li2 molecule, the performance of excited PES exhibited by either of the two
methods of MRCEPA-LRT are changed between 0.87-3.4 mH relative to the
actual choice of the orbitals. Thus, there is no significant loss in accuracy
even if we use the HF orbitals, but it saves the time for optimization. Also,
the CAS-SCF orbitals for the ground state tend to over-emphasize the
importance of the nondynamical correlation for the ground state, which
can get altered substantially for the excited states. This is why, we will
present here the results obtained with HF orbitals using one of the two
model functions. The actual model function used by us is described below
for each system. To assess the performance of our methods for studying
the low-lying PES, we compare our results obtained with those from full CI
(FCI) calculations in the same basis. The required molecular integrals and
the different CI results were generated by the electronic structure package,
GAMESS.
The two sets of unknowns, the cluster amplitudes and the combining
coefficients, are coupled in the SS-MRCEPA equations. They are solved
in a nested iterative loop. In the micro-iteration the cluster amplitudes
are solved for fixed coefficients till a local convergence is reached. The
coefficients are then updated by diagonalizing the matrix H^ in the macro-
iteration.
Once these unknowns in the MRCEPA are determined, we construct
the matrix of Ji, and get the required eigenvalues and vectors from the
MRCEPA-LRT eigenproblem.
It should be noted that, unlike in the case of SR-CCLRT, our MRCEPA-
LRT equations are defined in a space of overcomplete basis: each virtual
function \xm) appears several times-generated from various | ^ ) ' s , with
attendant coefficients d™k. If the number of linearly independent functions
in the Hilbert space is Nh, then only Nh roots are meaningful. The rest of
the roots are extraneous. These are easily diagnosed from the very small
value of the norm of these functions.
A. P4 Model System
Our first test case is the simultaneous stretching of both H — H bonds

in P4 (rectangular i/ 4 model). The P4 model introduced by Jankowski and
-2.000 -j 1
-2.025 -
-2.050 - , s^
• \ 1 A i /
-2.075 - \ /
^ -2.100- \ /
3 • \ /
^ -2.125- \ /
> ' \ / MRCEPA(0)-LRT

g -2150" \ / MRCEPA(D)-LRT
g -2,75: \ / — F C |
-2.200 - . / \
-2.225 - ^ / \ ,
-2.250 - ^ ^ ^ ^ - ^ ^ ^ ^ ^ ^ ^ ^
-2.275 -I , , 1 , , , , , ,
2 3 4 5 6
R (a.u.)
Fig. 1 : PES of some low-lying states of the P4 model systerr
(Energy of the 11A2 state shifted down by 0.25 a.u.)
Paldus [45; 46] and later also studied extensively by several workers [28; 42;
47; 48; 49] is a rather simple test system and thus well suited to explore
the validity and efficacy of our formalism. The geometries considered here
are the same as in Ref.[45]. The geometry of P4 model is specified by
the distance of approach of the two H2 molecules, R, which is varied from
the pure square geometry and then gradually elongated to the rectangular
geometry. The configurational quasi-degeneracy involving <fii : la\lb\ and
<f>2 : la\2a\ becomes more pronounced with lowering of the values of R. For
R=2.0 a.u. we reach the situation where the two configurations become
exactly degenerate. With increasing R, the degeneracy is lifted very rapidly.
We have used the DZP basis taken from Ref.[46] and the HF orbitals of
the function fa in our calculations. In this model system, all molecular
calculations are carried out using C ^ symmetry.
-1-55-1 1Ba .^p.
-1.60- 1 1B\ \
• \ \
-1.65- \ ^v
-1.70- \
= -1.75- ^ * * » « - ^ ^ ^ _ _ _ ^ ^
O) -1-80- >'/^
O • //
m "185" if MRCEPA(0)-LRT
190 / MRCEPA(D)-LRT
2'A,/ FCI
-1.95- /
-2.00 -
2 3 4 5 6
R (a.u.)
Fig. 2 : PES of some low-lying states of the P4 model systerr
The results for SS-MRCEPA(O) and SS-MRCEPA(D) based linear re-

sponse methods for the PES of states of various symmetries are plotted
in Figs. 1 and 2. The corresponding FCI results are also shown. In or-
der to show PES of several states in the same graph, we have sometimes
shifted down the energies of some higher-lying states by a fixed amount, as
explained in the captions for these graphs. From this we can see that:
(i) The FCI values and the results obtained by the MRCEPA-LRTs
for the ground state, l 1 ^ ! as well as excited state, 11A2 are very close
and uniform for all points on the PES. For the I1.02 state, the scenario is
slightly different. For this state, the performance of our methods is also
quite good. However, for very small values of R, our results are found to
be slightly away from the corresponding FCI values.
(ii)At small values of R, the performance of our MRCEPA-LRT for
I1.©! state is quite uniform and the deviation is fairly small with respect to
the FCI results. But the deviation of MRCEPA-LRT results from the FCI
values is not that small for this state at larger values of R.
(iii) Although the performance of MRCEPA-LRT for 21Ai state is gen-
erally good at small and intermediate values of R, it becomes somewhat
worse with increase in the values of R.
Both Figs. 1 and 2 show that at larger values of R the deviation of
MRCEPA(O)-LRT from the FCI for the 2xAx and l 1 ^ states is smaller
than the MRCEPA(D)-LRT, the reason for which is not clear to us at this
point.
B. Li2 Molecule
The Li2 molecule is a well studied system [28; 42; 50; 51] to test the
efficacy of the various SR and MR methods. It is now well known that the
ground state of the Li2 molecule requires a two-determinantal description in
the zeroth order. In this calculation we have used a two-dimensional CAS
defined by the two determinants (pi = lojlo^2o| and fa — \a'1g\a2u2u'2u.
This is a good example to test the performance of our methods within
a small reference space, since there is a quasi-degeneracy of the reference
determinants at large internuclear distances in the PES of the ground state
and there are intruders to <j>2 at the intermediate distances. We have used
a small basis (DZ) [52] for the study of the Li2 PES of some low-lying
electronic states, using D2h point group at each step of calculation. The
configuration <j>\ has been chosen as the HF function in our calculations.
Using these, we can compare the potentiality of our MRCEPA-LRT with
the FCI values for all the states studied.
-14.845 -i 1
-14.850 - ^ * ^ *
-14.855- /
-14.860- /«!/r:rrrTr^
^ -14.865- / /
>, -14.870- \ " I I

35 \ /
5 \ / / MRCEPA(0)-LRT
•j - 1 4 - 8 7 5 " 1 / / MRCEPA(D)-LRT
1 / / FCI
-14.880- \ I /
-14.885- \ \l/
•v'v
-14.890-
' i ' i • i • i • i ' I ' I '
4 6 8 10 12 14 16
R (a.u.)
Fig. 3 : PES of some low-lying £ states of Li2 molecule
(Energy of the £u* state shifted down by 0.06 a.u.)
We plot the results in Figures 3-5 to check the uniformity and accuracy
of the PESs of various low-lying excited states obtained using our newly
developed MRCEPA-LRT alongwith the corresponding FCI values. Again,
as in the case of P4, we have sometimes shifted down the energies of the
PES for some higher-lying various states by a fixed amount, as explained in
the captions of the corresponding figures. We have tagged the symmetries
of the different energy states according to the I?ooft point group, though
we have performed the calculation within the £>2/i framework. It is evident
that the MRCEPA-LRT results are in close proximity to the FCI values
and quite uniform throughout the PES, including the points of intruders
and degeneracies. The performance of our SS-MRCEPA based linear re-
sponse methods for Li2 molecule is quite good for low-lying £ states as
is evident from the Figs. 3 and 5. From Fig. 4, it is clear that the per-
formance of our newly developed MRCEPA-LRT to compute the PESs of
Ug and n u states is not so satisfactory as that of the performance for the
previously mentioned low-lying £ states. To underline the performance of
the MRCEPA-LRT vis-a-vis the parent MR-CCLRT in predicting the ex-
citation energies, we present in Table-I the vertical excitation energies of
the Li2 molecule at the equilibrium ground state geometry obtained from
both MRCEPA(O)-LRT and MRCEPA(D)-LRT as well as the full-blown
MR-CCLRT methods alongwith with the FCI results.
Table-I FCI, MRCEPA(O)-LRT, MRCEPA(D)-LRT and MR-

CCLRT vertical excitation energies (in mH) for a few low-lying
singlet excited states of Li2 at the equilibrium distance (R=5.0512
a.u.) of the ground state. Values within the parentheses indicate
the deviation from FCI.
~ Methods £+ II M ng S~ S~
FCI 68.203 109.723 112.709 173.4 206.726
MR-CCLRT 68.23 109.56 112.524 173.598 206.896
(-0.027) (0.163) (0.185) (-0.198) (-0.17)
MRCEPA(0)-LRT 69.397 110.633 113.451 174.455 207.755
(-1.194) (-0.91) (-0.742) (-1.055) (-1.029)
MRCEPA(D)-LRT 68.221 109.543 112.502 173.713 207.144
(-0.018) (0.183) (0.207) (-0.313) (-0.418)
From the table, it is evident that the numerical accuracy achieved by the
MRCEPA-LRT is quite close to that of the parent MR-CCLRT approach.
This indicates that SS-MRCEPA based LRT (MRCEPA-LRT) is a very
useful and good approximant of the full-blown SS-MRCC based LRT (MR-
CCLRT), although SS-MRCEPA neglects most of the complicated non-
linear terms of the parent SS-MRCC method using a series of physically
motivated graded approximation schemes. The method works very well to
compute the PES of various low-lying exited states for ground states with
-14.760 -i 1
.14765. I MRCEPA(O)-LRT
i MRCEPA(D)-LRT
I FCI
-14.770- u !l
3. -14.775- \l
>
^ -14.780-
' 1 \
-14.785- \ "'^:^?'
-14.790- \. s^^"***^
-14.795 -I 1 , , , 1 , , , , , , , ,
4 6 8 10 12 14 16
R (a.u)
Fig. 4 : PES of some low-lying n states of Li2 Molecule
varying degrees of MR character and displaying real or avoided crossings.

The general trend of the results for both P4 and Li2 systems is very
encouraging indeed. We do not want to draw definite conclusions regard-
ing the relative performances of MRCEPA(0)-LRT and MRCEPA(D)-LRT
from these pilot calculations. Both perform quite well. As Table-I indicates,
the excitation energies computed by MRCEPA(D)-LRT are of better qual-
ity at the ground state equilibrium geometry of Li2, though MRCEPA(D)-
LRT sometimes show greater variations at some other geometries.
-14.66 -i 1
MRCEPA(O)-LRT
.1467. -\ MRCEPA(D)-LRT
" \ FCI
-14.68- \
3 ' \
>£- -14.69- \
5K \
O> \
<5 ' \
UJ -14.70- \
-14.71 - j ;' i >v
-14.72 A , , , , , , , , , , , , ,
4 6 8 10 12 14 16
Fig. 5 : PES of someRlow-lying

(a.u.) 2 ' states of Li2 Molecule
4 Summary
In this paper, we proposed two variants of linear response methods devel-

oped for excited states from the state specific multi-reference CEPA theo-
ries for the associated ground state. These theories are designed to generate
excited state energies of a system whose ground state is multi-reference in
character and also show intruders. Around the quasi-degenerate configura-
tions, the effective hamiltonian based methods are expected to work well,
but they will fail in the regions of the PES where intruders are present.
But our SS-MRCEPA based LRT performs with equal efficacy in the entire
range of the PES for the various excited states. Numerical test calcula-
tions on P4 model and Li2 molecular systems support the above facts. A
comparison of our results with those obtained from the FCI in the same
basis indicates that the computed excited state energies are very accurate
and uniform in nature throughout the entire range of the PES, and this
is generally valid for states with various spatial symmetries. Thus our
MRCEPA-LRT serves as a good model to compute accurate and uniform
PES of the excited states. This fortifies our belief that the MRCEPA-LRT
methods are useful and good approximants of the parent MR-CCLRT, de-
spite the neglect of a host of terms present in the parent theory. The results
are encouraging for the systems we have studied using our MRCEPA-LRT,
which leads us to express cautious optimism about its general potentiality.
5 Acknowledgments
DM wishes to thank the CSIR, INDIA for financial support [Proj. No. 01
(1624)/EMR-II)]. DP thanks the CSIR, INDIA for the research fellowship.
This article is dedicated to P Ca rsky, I Hubac and M Urban on the

happy occasion of their reaching sixty.
Bibliography
1. I. Shavitt (1975) in Methods of electronic structure theory, Vol. 3, H. F. Schae-

fer III (ed.) Plenum Press, NY.
2. R. J. Bartlett and W. D. Silver (1975) Int. J. Quantum Chem. S9, 183; J. A.
Pople, J. S. Binkley and R. Seeger (1976) Int. J. Quantum Chem. S1O, 1.
3. W. Meyer (1973) J. Chem. Phys. 58, 1017; W. Meyer (1989) Theor.Chim. Ada
35, 277; A. C. Hurley (1976) in Electron Correlation in Small Molecules,
Academic Press, London; R. Ahlrichs, H. Lischka, V. Staemmler and W.
Kutzelnigg (1975) J. Chem. Phys. 62, 1225; W. Kutzelnigg (1977) in Meth-
ods of Electronic Structure Theory, H. F. Schaefer III (ed.) Plenum Press,
NY.
4. R. J. Gadanitz and R. Ahlrichs (1988) Chem. Phys. Lett. 143, 413.
5. P. G. Szalay and R. J. Bartlett (1995) Chem. Phys. Lett. 214, 481; L. Fusti-
Molnar and P. G. Szalay (1996) J. Phys. Chem. 100, 6288.
6. P. G. Szalay (1997) in Recent Advances in Coupled-Cluster Methods, Ed. R. J.

Bartlett, World Scientific, Singapore and references therein.
7. J. Cizek (1966) J. Chem. Phys. 45, 4256; J. Cizek (1969) Adv. Chem. Phys.
14, 35; J. Paldus, J. Cizek and I. Shavitt (1972) Phys. Rev. A 5, 50.
8. R. J. Bartlett (1989) J. Phys. Chem. 93, 1697; R. J. Bartlett (1995) in Modern
Electronic Structure Theory, D. R. Yarkony (ed.) World Scientific, Singa-
pore.
9. R. J. Bartlett (1981) Ann. Rev. Phys. Chem. 32, 359.
10. J. Noga and R. J. Bartlett (1987) J. Chem. Phys. 86, 7041; E. E. Scuseria
and H. F. Schaefer III (1988) J. Chem. Phys. 152, 382; K. Raghavachari,
J. A. Pople, E. S. Replogle and M. S. Head-Gordan (1990) J. Chem. Phys.
94, 5579; P. Piecuch, S. A. Kucharski and R. J. Bartlett (1999) J. Chem.
Phys. 110, 6103.
11. R. J. Buenker and S. D. Peyerimhoff (1974) Theor. Chim. Acta 35, 33; P. E.
M. Siegbahn (1980) J. Chem. Phys. 72, 1674; R. J. Buenker and S. Krebs
(1999) in Recent Advances in Multireference Methods K. Hirao (ed.) World
Scientific, Singapore and references therein.
12. S. R. Langhoff and E. R. Davidson (1974) Int. J. Quantum Chem. 8, 61; E.
R. Davidson (1974) in The World of Quantum Chemistry, R. Dauel and
B. Pullman (eds.) Reidel, Dodrecht; J. P. Francois and R. Gybels (1990)
Chem. Phys. Lett. 172, 346.
13. J. P. Malrieu (1982) Theor. Chim. Acta 61, 163.
14. M. R. A. Blomberg and P. E. M. Seigbahn (1983) J. Chem. Phys. 78, 5682;
W. D. Laidig and R. J. Bartlett (1984) Chem. Phys. Lett. 104, 424.
15. J. P. Daudey, J. L. Heully and J. P. Malrieu (1993) J. Chem. Phys. 99
1240; I. Nebot-Gil, J. Sanchez-Marin, J. P. Malrieu, J-L. Heully and D. J.
Maynau (1995) J. Chem. Phys. 103, 2576. J. P. Malrieu, I. Nebot-Gil and
J. Sanchez-Marin (1994) J. Chem. Phys. 100, 1440.
16. S. Chattopadhyay, U. S. Mahapatra, B. Datta and D. Mukherjee (2002) Chem.
Phys. Lett. 357, 426.
17. S. Chattopadhyay, U. S. Mahapatra, P. Ghosh and D. Mukherjee (2002) in
Low-Lying Potential Energy Surfaces, Eds. M. R. Hoffmann and K. G. Dyall
ACS Symposium Series No. 828, Washington, DC.
18. B. Brandow (1967) Rev. Mod. Phys. 39, 771; I. Lindgren (1974) J. Phys. B,
7, 2441.
19. D. Mukherjee, R. K. Moitra, and A. Mukhopadhyay (1977) Mol. Phys. 33,
955; I. Lindgren (1978) Int. J. Quantum Chem. S12 33,; I. Lindgren, and
D. Mukherjee (1987) Phys. Rep. 151, 93; D. Mukherjee, and S. Pal (1989)
Adv. Quantum Chem. 20, 561.
20. B. Jeziorski, and H. J. Monkhorst (1981) Phys. Rev. A 24, 1668; B. Jeziorski,
and J. Paldus (1988) J. Chem. Phys. 88, 5673.
21. T. H. Schucan and H. A. Weidenmiiller (1972) Ann. Phys. 73, 108; G. Hose
and U. Kaldor (1979) J. Phys. B 12, 3827.
22. J. P. Malrieu, Ph. Durand and J. P. Daudey (1985) J. Phys. A: Math. Gen.
18, 809; J. L. Heully and J. P. Daudey (1988) J. Chem. Phys. 88, 1046.
23. D. Mukhopadhyay, B. Datta and D. Mukherjee (1992) Chem. Phys. Lett.
197, 236; B. Datta, R. Chaudhuri and D. Mukherjee (1996) J. Mol. Struct:
THEOCHEM 361, 21.
24. A. Landau, E. Eliav, Y. Ishikawa and U. Kaldor (2000) J. Chem. Phys. 113,
9905.
25. J. P. Malrieu, J. P. Daudey and R. Caballol (1994) J. Chem. Phys. 101, 8908;
J. Meller, J. P. Malrieu and J. L. Heully (1995) Chem. Phys. Lett. 244, 440;
J. Meller, J. P. Malrieu and R. Caballol (1996) J. Chem. Phys. 104, 4068.
26. J. Masik, and I. Hubac (1997) in Quantum Systems in Chemistry and Physics:
Trends in Methods and Applications, R. McWeeny et al. (eds.) KA, Dor-
drecht; I. Hubac, J. Prittner and P.Carsky (2000) J. Chem. Phys. 112,
8779.
27. U. S. Mahapatra, B. Datta, B. Bandyopadhyay and D. Mukherjee (1998)Adv.
Quantum Chem. Vol. 30 D. Hanstrop and H. Persson (eds.) Academic
Press Inc. San Diego.
28. U. S. Mahapatra, B. Datta and D. Mukherjee (1998) Mol. Phys. 94, 157;
U. S. Mahapatra, B. Datta and D. Mukherjee (1999) J. Chem. Phys. 110,
6171.
29. S. Chattopadhyay, U. S. Mahapatra and D. Mukherjee (1999) in Recent Ad-
vances in Multireference Methods K. Hirao (ed.) World Scientific, Singapore;
S. Chattopadhyay, U. S. Mahapatra and D. Mukherjee (2000) J. Chem.
Phys. 112, 7939.
30. H. J. Monkhorst (1977) Int. J. Quantum Chem. S l l , 421; E. Dalgaard and
H. J. Monkhorst (1983) Phys. Rev. A28, 1217.
31. H. Nakatsuji and K. Hirao (1978) Chem. Phys. Lett. 59, 362; H. Nakatsuji
and K. Hirao (1988) J. Chem. Phys. 68, 2053; H. Nakatsuji and K. Hirao
(1979) Chem. Phys. Lett. 67, 329, 334; K. Hirao (1993) J. Chem. Phys. 79,
5000.
32. D. Mukherjee and P. K. Mukherjee (1979) Chem. Phys. 39, 325; B. Datta ,
P. Sen and D. Mukherjee (1995) J. Phys. Chem. 99, 6441.
33. H. Sekino and R. J. Bartlett (1984) Int. J. Quantum Chem. S18, 255; P.
Piecuch and R. J. Bartlett (1999) Adv. Quantum Chem. Vol. 34, 295 and
references therein.
34. H. Koch and P. J0rgensen (1990) J. Chem. Phys. 93, 3333; H. Koch, R.
Kobayashi, A. Sanchex de Mears and P. J0rgensen (1994) ibit 100, 4993.
35. S. Pal (1989) Phys. Rev. A 39, 39; S. Pal (1992) Int. J. Quantum Chem. 41,
443.
36. D. Ajitha and S. Pal (1997) Phys. Rev. A 56, 2658. D. Ajitha, N. Vaval and
S. Pal (1999) J. Chem. Phys. 110, 2316.
37. S. Tenno, S. Iwata, S. Pal and D. Mukherjee (1999) Theor. Chem. Ace. 102,
252.
38. P. G. Szalay (1995) Int. J. Quantum Chem. 55 , 15.
39. P. G. Szalay, T. Muller and H. Lischka (2000) Phys. Chem. Chem. Phys. 2,
2067.
40. I. Martin, C. Lavin, Y. Perez-Delgado, J. Pitarch-Ruiz and J. Sanchez-Marin
(2001) J. Phys. Chem. A 105, 9637.
41. D. Mukhopadhyay, S. Mukhopadhyay, R. Chaudhuri andD. Mukherjee (1991)
Theor. Chim. Ada 80, 441.
42. U. S. Mahapatra, B. Datta and D. Mukherjee (1999) Chem. Phys. Lett. 299,
42; U. S. Mahapatra, B. Datta and D. Mukherjee (1999) J. Phys. Chem. A
103, 1822.
43. P. Ghosh, S. Chattopadhyay, D. Jana and D. Mukherjee (2002) Int. J. Mol.
Set. 3, 733.
44. A. Deb, S. Chattopadhyay, D. Pahari and D. Mukherjee (2003) (to be com-
municated).
45. K. Jankowski and J. Paldus (1983) Int. J. Quantum Chem. 18, 1243.
46. J. Paldus, P. E. S. Wormer and M. Benard (1988) Coll. Czech. Commum. 53
1919.
47. S. Zarrabian and J. Paldus (1990) Int. J. Quantum Chem. 38, 761; K.
Jankowski, J. Paldus and J. Wasilewski (1991) J. Chem. Phys. 95 , 3549.
48. J. P. Finley, R. Chaudhuri and K. F. Freed (1995) J. Chem. Phys. 103, 4990.
49. K. Jankowski and I. Grabowski (1995) Int. J. Quantum Chem. 55, 205.
50. U. Kaldor (1991) Theo. Chim. Ada 80, 427; U. Kaldor (1990) Chem. Phys.
Lett. 140, 1.
51. I. Garcia-Cuesta, J. Sanchez-Marin, A. Sanchez de Mears and N. Ben Amor
(1997) J. Chem. Phys. 107, 6306.
52. T. H. Dunning Jr. (1970) J. Chem. Phys. 53, 2823.
Chapter 5: Modelling of Anisotropic Exchange Coupling in
Rare-Earth -Transition-Metal Pairs:
Applications to Yb3+-Mn2+ and Yb^-Cr34
Halide Clusters and Implications to the
Light Up-Conversion
M. Atanasova, C. DauP, H.U. GiideP
a Institute of
General and Inorganic Chemistry
Bulgarian Academy of Sciences
Acad. G. Bontchev Str. Bill, 1113 Sofia, Bulgaria
b Departement
de Chimie, Universite de Fribourg,
Ckdu Musee 9, CH-1700 Fribourg, Switzerland
c Departement fur Chemie und Biochemie, Universitdt Bern, Freiestrasse 3,

CH-3000 Bern 9, Switzerland
Abstract
A procedure for calculating magnetic exchange coupling constants in rare
earth (RE)-transition metal (TM) dimers is presented. In a first step RE-TM
transfer (hopping) integrals between orbitals carrying unpaired (magnetic)
electrons on RE and TM are determined from a MO-calculation utilising the
concept of orbital exchange pathway and effective Hamiltonian theory. In a
second step, many-electron wave-functions for ground or excited electronic
states on RE and TM are constructed in which spin-orbit coupling (which
dominates for the RE) is fully (variationally) taken into account. Finally, the
data from steps 1 and 2 are combined to obtain numerical values for kinetic
exchange integrals (within Anderson's superexchange theory) utilising a non-
Heisenberg (orbital dependent) exchange operator. The MO's corresponding
153
154 M. Atanasov, C. Daul and H. U. Giidel
to the d(Mn) and f(Yb) magnetic orbitals have been used to also calculate
ferromagnetic contributions to the exchange coupling. The model is applied to
Yb3+-Mn2+ dimers with corner, edge and face shared octahedra and Cl" and Br"
ligands. For that purpose we utilise a spectroscopically adjusted (to spectra of
single nucleous octahedral Mn2+, Cr3+ and Yb3+ complexes) parameterisation
of the Extended Huckel theory. The expediency of the approach is discussed
based on a comparison between calculated and experimental (available from
inelastic neutron scattering experiments) anisotropic exchange parameters in
Br3Cr3+-Ji(Br3) Yb3+Br3 dimers. Ferromagnetic exchange parameters for the
ground state and the lowest emitting excited state increase from corner to
edge to face Mn-Cl-Yb sharing thus varying in the same way as the efficiency
of the up-conversion found for such species. This lends support for an
exchange mechanism for this phenomenon postulated previously.
1. Introduction
Interactions between transition metals (TM) and rare earths (RE) in solids
have been of great interest because of their ability to combine photons of low
energy to photons of higher energy (up-conversion, UC) '"2. It was discovered
that pairs of Mn2+ and Yb3+ bridged by three or one Cl" ions are able to
convert IR radiation at about 10300 cm"1 ( corresponding to the 2F7/2->2F5/2
excitation at Yb3+ ) into red (14700 cm"1, CsMnBr3:Yb - face sharing 3) or
orange (16200 cm"1, Rb2MnCl4:Yb - corner sharing 4) luminescense, both
corresponding to the "^Tig—>6Aig transition within the Mn2+ ion. Based on
spectra in high resolution an exchange UC mechanism has been postulated.
Surprisingly, no theoretical work in that direction have been done yet, in spite
of the fact, that theories of excitations in pairs of exchange coupled TM are
well documented5"10 and tested by experiment n"18.
In this study we describe a procedure allowing to treat exchange
interaction within a cluster of one TM and one RE, bridged by common
ligands. Utilising MO wavefunctions and energies we base our model on the
concept of localized magnetic orbitals allowing to treat both cluster ground
and excited states and to account for both antiferromagnetic (kinetic) and
ferromagnetic (potential) exchange. After an introduction into the modelling
of exchange interactions using Anderson's exchange theory including ions in
non-degenerate (Section 2.1) and degenerate (Section 2.2) electronic states,
we focus on one example known from literature 19, that of CoCl2(H2O)2:Mn2+
with Mn2+ and Co2+ ions in the pair possessing ground states without and
Modelling of Anisotropic Exchange Coupling 155
with orbital degeneracy, respectively. For didactical purposes and including

mathematical details as well, we derive expressions for the anisotropic
exchange coupling tensor of a Mn2+-Co2+ pair for a general (without
assuming any symmetry) bridging geometry. In Section 3 we further
introduce a model for treating exchange coupling in TM(Mn2+)-RE(Yb3+)
pairs. To approximate the model parameters we make use of a
spectroscopically adjusted extended Hueckel theory which we introduce in
Section 3.3. As a first application an elaborated example - that of MnC^4"
and YbCl63" octahedra sharing one Cl" ligand will be presented in detail
(Section 3.6.). In Section 4 we present additional applications to edge and face
shared octrahedra of the same ions both in their ground state and in the lowest
emitting excited state. In Section 4.2 the relevant aspects of our results
regarding the mechanism of up-conversion will be discussed.
2. Modelling Exchange Interactions: Anderson's Exchange

Theory
2.1. Exchange interactions for orbitally non-degenerate states

Electrons of d (TM) and f(RE) orbitals are quite localized. Therefore, a good
starting point in describing exchange coupling between electrons on different
centers is to consider unperturbed d and f orbitals and to correct their
wavefunctions affected by the surrounding ligands. Taking two TM centers
with unpaired electrons occupying d^ orbitals on each (Fig.l) we consider
magnetic orbitals dli and dl2 composed of d^1 and d^2 orbitals on centers 1
and 2 and (smaller) tails of the ligand (pz) (pL1 and <pL2 functions and on the
neighbour TM ions:
dlr-N^d^+aV-pd*2)
dl2-N2(dz22-a2(pL2 -Pdrf 1 ) (1)
OCl-XdlLl+SdiLl
0C2 ~ Xd2L2 + Sd2U
p~s dd
Overlap integrals d-L (SdL ) and d-d (Sdd) are small with a covalency
parameter %<jL, typically in the range 0.1 - 0.5. Within the basis of the four
Slater determinants |dli+dl2+|, |dli+dl2"|, |dlfdl2+|, |dlfdl2"| we can construct one
triplet and one singlet functions (eqs. 2). They are eigenfunctions of the
Heisenberg exchange operator (eq.3) with the eigenvalues given in eq.4. The
exchange coupling constant JJ2 describes the exchange splitting of the two
multiplets Es-Er=J12. It is a phenomenological parameter, to be deduced from
experiment (magnetic susceptibility, spectroscopy, inelastic neutron
scattering) or calculated theoretically.
T T : Idlx'd^l,(l/V2)[|dl,+dl2|+|dlrdl2+|], |dlfdl2"l (2.1)

¥ s = (l/V2)[|dl1+dl2|-|dl1dl2+j] (2.2)
Hexc=-Jl2SlS2 (3)
ET=-(1/4)J 1 2 (4.1)
Es= (3/4)J12 (4.2)
Fig. 1 Localized magnetic orbitals in Anderson's super exchange model.
Within the exchange theory of Anderson20 and using perturbation theory,

Ji2 is expressed as a sum of two terms (eq.5). Ji 2 ^2Ki 2 , the Heisenberg
exchange integral (eq.6) is a positive quantity, tending to stabilize the triplet
over the singlet state. It reflects the gain of potential energy due to the Fermi
hole, which excludes an electron of a given spin in the vicinity of a reference
electron with the same spin (potential exchange). The Ji 2tf = -4T]22/U term
describes the gain of kinetic energy due to the delocalization (described by the
tails in eqs 1), tending to stabilize a singlet ground state (kinetic exchange).
As is seen, it is counteracted by the energy U, i.e. the increase of Coulomb

repulsion when moving from singly (cuVcuV ) to doubly occupied (dli2 or dl22
) orbitals dli and dl2. U is a large quantity, typically in the range 5-8 eV. The
crucial quantities K12 and the hopping integral Ti2 can be expressed using the
explicit forms of dli and dl2 (eq.7 and 8). As expected, T !2 is intrinsically
connected with the covalency of the metal-ligand bond. Eq.(8) describes an
useful approximation for Ki2, expressing this quantity in terms of the intra-
atomic exchange integral of the bridging ligand, weighted by orbital
reductions factors (Xi2 and OC22 The expression of the exchange integral (Eq.5)
reflects the configurational mixing of the singlet function (Eq.2.2) with the
charge transfer excited states |dli+dlf| and |dl2+dl2"| which stabilize *FS against
T T . In this form the equation for Jn can be easily generalized for magnetic
centers (like high-spin Mnn (d5)) with more than one magnetic electrons. In
this case a sum of terms over different exchange pathways, Jf2 (Eq.9) have to
be included in the expression for Ji2, with nj and n2 accounting for the number
of unpaired electrons on center 1 and 2, respectively.
Ji2 3Ji2f+Ji2at2K12-4T122/U (5)

K12=JJdlI(r1)*dl2(r1)(l/rI2)dl1(r2)dl2rr2)*dT:dT2=[dl1dl2|dl,dl2] (6)
T 12 ~ N , N 2 [CC2 (d 1 1 h I 9 L 2 ) + a,(d 2 I h I <pL') ] (7)
K 12 ~ N , 2 N 2 2 Cd2 a 2 2 [<PL'9L 2 I <PL V ] (8)
J12=~J- Z ^ (9)
nln2 iel,je2
2.2. Exchange interactions in the case of orbital degeneracy

The description of exchange coupling in terms of a single (scalar) parameter
Ji2 (isotropic Heisenberg-Dirac-van Vleck Hamiltonian, Eq.3) is only valid if
electronic states on ions 1 and 2 are orbitally non-degenerate. This is
generally the case for half filled shells of TM (octahedral d3, d8, d5 (high-
spin)) and if spin-orbit coupling is negligible. For orbitally degenerate states
and/or cases of strong spin-orbit coupling the isotropic model is generally not
valid; orbital degrees of freedom intermix with spin ones to lead in general to
an anisotropic exchange tensor, j = |j.. j (i j=x,y,z). The Hamiltonian for the
spin coupling now takes the form:
158 M. Atanasov, C. Daul and H. U. Gildel
Uxx J*y JASA

Hexc=-SlJS2=-[six,Sly,Slz_Jyx J^ Jyz S2y = (10)
~^xx^lx^2x ~Jyy^\y^2y ~^zz^\z^2z ~
~ Jxy^lx^2y ~^yx^ly^2x ~^xzÛ^2z ~^zx^\z^2x ~^yz^ly^2z ~^zy^lz^2y
2.2.1. A prototype example: CoCl2(H2O)2:Mn2+ 19.

In CoCl2.2H2O the cobaltous ions are surrounded by a nearly perfect square
arrangement of four Cl ions and two water molecules along the b-axis. The
crystalline field symmetry around Co2+ is nearly tetragonal with a principle
axis (z) parallel to the b-axis (Fig.2). Co2+ possesses a high-spin *Ti ground
state with orbital components which can be classified according to the three
eigenstates of a fictitious orbital angular momentum of (1=1, lz= ±1,0). The
spin-orbit interaction and the tetragonal crystal field in the 4T1 basis can be
calculated making use of angular momentum operators 1 and Iz:
H=A/I.S0-8(Iz2-2/3) (11)
Here So is the spin operator acting on the s=3/2, ms=±3/2, ±1/2 spin-
functions of Co2+ and A,'=-(3/2)kX, - the spin orbit constant (X, Co2+ free ion=-
180 cm'1, k - a reduction factor). In CoCl2(H2O)2:Mn2+ Mn2+ substitutes for
Co2+ and the 6Ai ground state of Mn2+ is exchange coupled with the ground
state of a neighbour Co2+ ion (Fig.3). The latter is orbitally degenerate and
the spin coupling becomes orbital dependent, since different orbitals on Co
overlap in a different way with d-orbitals on Mn2+. In this sence the present
example closely resembles that of a Yb3+-Mn2+ pair, however the orbital
degeneracy here is only 3 (^T{) instead of 7 (one hole in the 4f-sub-shell of
Yb3+). For didactic purposes it is useful to consider this example in some
detail, before going to the more complicated cases of Mn2+-Yb3+ pairs. The
exchange interaction between Co2+ and Mn2+ can be written in the general
form of Eq.12; here J(lz,lz') is the exchange integral between Mn2+ and the lz-
and lz' are the 1=1 orbital components of Co2+ (Eq.13).
Hex=-OSo.S = - E J C U / ^ S r S (12)
lz,lz'
Piziz- is an operator connecting the lz and lz' d-functions of Co2+ and So and
S are spin operators for Co2+(So=3/2) and Mn2+(S=5/2). In the expression for
J(lzJz') (eq 13) singly occupied d^-orbitals of Mn2+ are subsummed and
integrals (Uhld^) and (djh|l z ') of electronic hopping between Mn2+ and Co2+
describe various exchange pathways, allowing for off-diagonal (lz*lz') terms
(non-diagonal exchange). Such terms may arise in ligand fields of low
symmetry.
J ( W z ' ) = - £ (UhldpXdJhllzVCSU) (13)

M
i—^ r—'• T cW
JD \ j I
. T. / 1 A b(z)
T Z*y)
° O Co-Cl 2.45,2.48A
"•^ Co Co-0 2.04 A
° Mn
Fig. 2 Coordination geometries and magnetic ordering structure in the low-

temperature modification of CoCl2.2H2O.
A
As is derived in Appendix 1, the orbital operator O can be expressed in

terms of angular momentum (Ij, i=x,y,z) and the identity 1 operators as
follows:
6 = [J(l,l)-J(0,0)]lz2 + V2 J'(l,0) ( U Z + U , ) - V2J"(l,0) (Iylz+lzly) +
j'(i,-i)(ix2-iy2) - J"(i.-i) (Uy+M0+J(0,0) i (14)
ci^pci
H 2 O©«Mn II ÔH2
C1QJPC1
H2OCHCon8S0OH2
Fig. 3 The Mn2+-Co2+ dimer units in Mn2+ doped CoCl2.2H2O.
From the 9 components of J(lz,lz') (lz,l2'=±l,0) tensor only 6 are

independent; the following relations hold (see Appendix A.I.):
J*(lz,lz')=J(lzMz)
J(1,O)=-J(O,-1) (15)
J(1,1)=J(-1,-1)
The J(12,1Z) parameters (l^V) are generally complex numbers. In eq 14

we denote their real and imaginary parts by J'(lz»lz) and J"(l z ,l z ),
respectively. Since crystal field and spin-orbit coupling energies (Eq.ll) are
much larger than J(lz,lz') we take the eigenstates of this Hamiltonian and focus
on the ground state Kramers doublet of Co2+ which can be written in the
following form:
H)-HH*H'4 (16)
Here, mj and nis in |m,,m s ) stay for Co2+ orbital and spin quantum
numbers. Values of 8 (Eq.ll) and a,b,c (Eq.16) have been determined by
Oguchi in such a way, that theoretical g-values match the experimental ones:
8 =5kX , a=895, b=-0.288, c=0.346.22 The separation between the ground and
the first excited state (260 cm"1) is much larger than the exchange energy.
Therefore excited states can be neglected and exchange tensor components
can be obtained applying the operator (Eq. 12 and 14) on the right hand side of
A
eq.16. Let us calculate as an example J^ and Jxx. Applying the operator O on

the function +_L\ of Co2+ we notice that the effect of the operators lz2 and
2/
lx2-ly2 of relevance for J^ and JU is just to produce the following orbital
replacements:
lz2: P( 1 -> 1), P(-1 ->-1) and multiplying by 1, P( 0-»0) and multiplying
byO
lx2-ly2: P(l->-l) and P(-1->1) and multiplying by 1 (17)
Keeping this in mind expressions for J^ and JM are obtained for example
taking:
-JzzSzSz I \ 5l\=_6Sz(Co)Szl\ 5\\

2/ 2 2/ 2/ 2 2/
(18)
-JxxsA I\ i,l\ =-6 S,(Co)Sx i\ i l \

2/ 2'2/ 2/ 2'2/
In eq.18 s ^ and S^Co) are spin-operators acting on the left and right
sides of eq.16. We notice that the operators S^ for Mn2+ lead to the same
result on both sides of eq 18 and can thus be dropped out. Keeping in mind
that sx leads to changes of mz by ±1 and integrating we obtain:
J -- /I 1\ /=2 (20S*(C0)l)
\2 Sz 2/
(Jjosî) (19)
J-- , i a -2\-2OSx(CO)|2/
\"2Sx2/
Here we make use of the relations s z i \ = i i \ and sx J_\_! _ i \ to give

2/ 2|2/ 2/ 2 2/
(1/2) for the denominators in Eq.19. Now J^ and J^ are derived as follows;
we apply first the Sz and Sx operators on the right hand side of eq.16 and
make the substitutions21 :
Sz 2,2\ = 3 3 3\ Sz 3 A= I 3 1 U 2 l\ 1 3 A. (2(u)
2 2/ 2 2 2/ 2 2/ 2 2 2/ 2 2/ 2 2 2/
sx 2 , 2 \ = ^ 2 , I \ ; s x 2 , I \ = ^ 2 , 2 \ + 2,_I\ ; (20.2)
2 2/ 2 2 2/ 2 2/ 2 2 2/ 2 2/
s,2,_I\ = 2,I\ + ^2,_2\;
2 2/ 2 2/ 2 2 2/
We obtain:
6sz(Co)|^=6[|a|-l,mb|o)^-Ic|l,-^j=[J(l,l)-J(0,0)].
Tla _ lf 2\_I c i,_I\]+J(O,o)[2a _!,2\ + I b Of I\-Ic i,-l\"
[2 2/ 2 2/J |_2 2/ 2 2/2 2/_
Jzz=2/- OSz(Co) I \ = (3a2-c2)J(l,l)+b2J(0,0) (21a)
6 sx(co) I\= 6 [ ^ fl _ 1 ,I\ + ô,2\ +6 o i _l\ + ^J lf _2\ + J lf I\ ]=

2/ 2 2 / 2 2/ 2 / 2 ( 2 / 1 2/
=[J(I,D-J(O,O)] r^ a |_ 1 ) i^ c | u _|^ + c | 1 ; î+
^ = 2 / - ! OS x (CoJi\= 2V3acJ(l,l)+2b2J(0,0)+2c2J'(l,-l) (21b)
Expressions for the Jy (ij=x,y,x) parameters are listed in Table 1. The

J(l,l) and J(0,0) terms lead to an axially symmetric exchange tensor
(Jzz^Jxx=Jyy)5 while the off- diagonal term J'(l>-1) introduces additionally
orthorombicity (J^Jyy). The J"(l,-1) , J'(l,0) and J"(l,0) parameters lead to
the off-diagonal exchange [ Jy (i*j=x,y,z)] . The non-diagonal part of the

exchange interaction (Eq. 10) can be written using expressions in Table 1 as:
2(V3ab-3bc)J'(l,0)(s x S z + s z S x ) + 2(V3ab-bc)J"(l,0)(s y S z + s z S y ) +
(22)
2(bc-V3ab)J'(l,0)(s x S z - s 2 S x ) + 2(3bc-V3ab)J"(l,0)(s y S z - s z S y )
Table 1. Expressions for the exchange-tensor components for a Co2+-Mn2+ exchange

pair in a general coordination geometry around Co2+ and Mn2+ of having no
symmetry.
Jxx=2V3acJ(l,l)+2b2J(0,0)+2c2J'(l,-l)
Jyy=2V3acJ(l,l)+2b2J(0,0)-2c2J'(l,-l)
J2Z=(3a2-c2)J(l,l)+b2J(0,0)
J xy =-2c 2 J"(l,-l)
J^c'r'Cl,-!)
J^bcJ'O.O)
Jzx=2(2bc-V3ab)J'(l,0)
Jyz=2bcJ"(l,0)
J^VSab^bcygO)
As seen from (Eq.22) the non-diagonal exchange interaction can be

decomposed into a symmetric part and an antisymmetric part (Dzyaloshinski-
Moriya exchange). We note that in eq.22 quantization axes have been adopted
to make the g-tensor diagonal (main axes of the g-tensor).
These anisotropic exchange interactions have important consequences for
the EPR spectra of Mn2+, when doped into the CoCl2.2H2O lattice. Below
T=17.5K COCI2.2H2O is an antiferromagnet with sub-lattice magnetization
directed along the b-axis and longitudinal components (along the b-axis ,
Fig.2) of the exchange integrals much larger than the transversal ones. A
succession of antiferro- ferri- and ferromagnetic transitions appear upon
increasing magnetic field directed along the the b-axis. Below the ordering
temperature internal fields due to the anisotropic J-tensor can lead to splittings
of the S=5/2 levels of Mn2+. It has been shown 19, based on symmetry
arguments, that exchange coupling between Mn2+-Co2+ pairs along the c(x)
axis and within the ab(zy) plane are described by the Hamiltonians:
Hexc(c)= - - C sxSx - j ; sySy- J°B s2Sz (23)
H eX c(ab)=-/i c s x S x - / ^ sySy-Jlzz s 2 S z - / ^ s x S y -J^ SyS,-/^ sxSz-
-J'y, sySz (24)
Below the ordering temperature the spins of Co2+ are ordered, with an
easy direction along the z-axis (Fig.2) i.e. they all are in a — —\ state.
2*2/
Applying the sx and sy operators on the —,—) state leads to the spin-flip
state 1 _ l \ at a higher energy. Let us denote its energy with respect to

2' 2/
1 _L\ by Eo (HeXc(c)) and Ei (HeXC(ab)). The following matrix elements
2'2/
mixing the ground magnetic state with the spin-flip state are derived:
(HKHH) = -{ y » 5 '4 J - 5 ' (25)

(i4h-Hi-i) = -i J1 - s «-i J ^ (26)
We focus now on H<>xc (c) and notice that E o » J^,J° (see below).
Applying perturbation theory, and rearranging terms we obtain the following
spin Hamiltonian for Mn2+:
Heff(c)= -^(J°J +/;2)[5(5 +l)-S/] (27)
The contribution to the D-parameter from this term is obtained by

summing the contributions from the two Co2+ ions (above and below the Mn2+
ion), i.e. multiplying Eq(27) by 4 (=22). We get
D = - ^ (JaJ+Jl2) (28)
Infrared resonance absorption of clustered spins and critical magnetic

fields for antiferro-, ferri- and ferromagnetic transitions 23 allow to obtain
EO=22.1 cm'1, 7^=0.59 cm"1 and ^ = ^ = a J ° , a < l . Substituting into

eq 28 we obtain:
D=0.032 a 2 cm"1 (29)
This value of D is much smaller than the experimental one and of opposite
sign (D=-0.13 cm"1) . Let us turn now to Heff within the ab-plane.
Contributions to D from the diagonal part of HeXC(ab) are again negligibly
small and we focus here at the off-diagonal part (the last two term of the
matrix element eq 25). We obtain:
Heff(ab)= _ J ^ 5 ( 5 + 1) + ( ^ + r ; -j^)Sl\ (30)
The contribution to the D-parameter from this term is obtained by

summing the contributions from the four Co2+ ions in the plane surrounding
each Mn2+, i.e. multiplying Eq.30 by 16(=42). Substituting expressions of
J\zijXyz md Jly from Table 1 we get
D=
-^r(JlJ +Jlyz2-Jl2) = -^r{4b2c2[J'm2 +/"(l,0) 2 ]-4cV"(l -I) 2 }
Ex El
(31)
This is negative and matches the sign from experiment if
0 0 0
J\z + J\z > J\y • Concluding this section, we can state, that anisotropic
exchange coupling can produce measurable effects in the EPR spectra of the
Mn2+ ion, when entering as impurity in CoCl2.H2O. However, the large
number of model parameters and lack of numerical estimates for these
parameters did not allow to make more conclusive statements. In Section 3 we
describe a procedure allowing to make this possible.
166 M. Atanasov, C. Daul and H. U. Gudel
3. Magnetic Exchange in Mn 2+ (Cr 34 )-Yb 3+ Pairs
3.1. Electronic structure and energy levels ofYb3+ in cubic and lower
symmetric ligand fields
With one hole in the Yb3+ f-shell (f13-configuration) we can deal with
electronic states of Yb3+ in a one electron description, keeping in mind that
electron-hole equivalence requires to turn signs of all one-electron operators
(ligand field and spin-orbit coupling). When treating energy levels of f-
elements one usually starts from real f-functions - fLy^x^y2)], f[xyz], ffyz2],
ftz3], flxz2], f M x V ) ] and ffxCx^y2)] which for a well oriented Yb-Cl(Br)
bond direction ( along the z-axis) and a cylindrical symmetry transform as the
irreducible representations <ps, 8S, Ks, a, JI^, 5C and <pc of the C^, point group.
Octahedral ligand field splits the 7-fold degenerate f-orbitals into a2u, t2U and
tiu species. Within the angular overlap model (AOM) we have the following
expressions for their energies24:
a2u: 0
t 2u : (5/2K (32)
tlu: (l/2)(4eo+3ejt)
Estimations of the AOM parameters of a and n -type - eCT and eK (see

Section 3.3) yield values for ea(eK) of 338(64) and 208(37 cm"1) for Yb-Cl and
Yb-Br respectively, leading to the order of increasing energy from a2U to t2u to
t]u. Symmetry lowering from Oh to C4v, C2v and C3v in solid networks
composed of octahedra sharing one, two or three ligands, respectively, leads
to corresponding symmetry splittings:
Oh C4V C2v C3V

ti u ai+e ai+bi+b 2 ai+e
t2u bi+e a 2 +b!+b 2 a 2 +e
a2u b2 ai ai
We note that AOM parameters for RE are one order of magnitude smaller
compared to their typical values in 3d metals complexes, and also, in the case
of Yb 3+ , compared to the spin-orbit coupling constant (£(Yb 3+ )=-2900 cm"1). It
follows that spin-orbit coupling leads to a strong deviation from the regime of
a strong ligand field with f-orbitals optimally aligned for metal-ligand
overlap; it is thus reasonable to start from the free-ion levels of Yb3+ - 2F7/2 <
2Fs/2separated in energy by (7/2)£ =10000 cm"1 and to consider ligand field
as small perturbation. However for practical purposes we consider all effects
on the same footing starting from real f-orbitals but diagonalising the full
14x14 matrix of spin-orbit coupling (Appendix 2) and ligand field (see
Sections 3.2,4). In cubic symmetry the 2F7/2 ground state of Yb3+ splits into
twb Kramers doublets F 6 and F 7 and a quartet F8, while the 2F5/2 excited state
of Yb3+ gives rise to Eg and F 7 . From the spectrum of Cs2NaYbCl6 containing
the octahedral YbCU3" chromophore 25, the positions of these multiplets have
been located at (in a n 1 ) 362 (r 8 ), 701(T7) 10243 (r 8 ) and 10708 (r 7 ) with
F6 - the ground state. These become additionally split at sites with a lower
symmetry as found in Cs3Yb2Br9 (C3v site ) and MnCl2:Yb3+ (C2v site), see
Section 3.2.
3.2. Exchange Hamiltonian and configuration state basis functions for

Mn2+(Cr3+)-Yb3+pairs
Kramers doublets and quartets of Yb3+ in octahedral ligand field can be
regarded as fictitious spins of 1/2 (F6, F 7 ) and 3/2 (F8) respectively. Let us
denote fictitious spin-operators acting on these functions by sx, sy and sz and
the corresponding spin-operators for the S=5/2(3/2) spin functions of
Mn2+(Cr3+) by Sx,Sy,Sz. In the magnetic dimers considered in this work, we
are concerned with site symmetries C2v, C3v and C4v (for Yb3+ and the TM
octahedra sharing common edge, face and corner, respectively). In these
symmetries the exchange Hamiltonian takes a diagonal form (Eq.33). The
exchange parameters J^, J ^ , Jyy can be calculated by letting
H e x ^ -JzzSzSz-JXX SxSj-Jyy SySy (33)
Hexc act on product functions, = l \ c = l\ (Yb3+-Mn2+) or „ = i\\ s = 2\

2/ 2/ 2/| 2/
(Yb3+-Cr3+) and equating the result with the effect of the spin
Hamiltonian Hexc' (Eq.34) acting on linear combinations (due to spin-
A
orbit mixing) of orbital (via the orbital operator O) times spin

functions (via the spin operators s z ', s x ' and s y ' of the genuine spin 1/2)
representing the Kramers state of Yb3+. We utilize wave functions for
Kramers doublets and quartets with proper symmetry with respect to
the time-reversal operator to ensure correct phases. It is practical to
express O in terms of angular momentum operators (see Section 2.2.1). In

octahedral field the orbital momenta of the f-orbitals (1=3) are only partly
quenched; t2, ti orbital functions behave as angular momenta for 1=1 (eq.35)
and a2 as a pseudo-scalar. In order to utilize this equivalence we make use of
A
f-functions described by Griffith26 (Eq.36) to allow to set up the O operator

in the form given by (Eq.37); we denote by ± l " , 0 " and ±l',0' the
components of X2 and ti which obey relations (35). The splitting of the f-basis
into two types of orbital momentum functions leads to corresponding sets of
orbital exchange parameters, which we group in Eq.37 into an isotropic
J(a2,a2) term and anisotropy parameters for axial - J ( l " , l " ) , J(0",0") (t2-
type); J(1\1'X J(0',0') (t r type) and, additionally (in two-fold symmetries)
orthorhombic J(l",-1") (t2-type) and J(l',-1') (ti-type) exchange coupling
terms.
HeXc = - 6 (sz'Sz+s,'SI+sy'Sy) (34)
t2 <=> (1/2)1
t, »-(3/2)l (35)
^-° = -i| f( - 3)+ i| f(1)

t 2 ( 0 > 7? f { - 2 ) + 7? f ( 2 )
t2(O = -^|f(3)+3|f(-1) (36)
^ - - ^ - " - ^
t1(0') = f(0) = fz3
tl(1') = -3| f ( 1 ) -i| f ( - 3 )

a2= "i f( - 2)+ i f(2)
A
O= axial term:
(l*2(t2)[J(l ",1 ")-J(0",0")]+0«2(ti)[J(l M >J(0',0')]+ IJ(0",0")+I.
J(0\0')+
orthorhombic term: (37)

( l ^ H y 2 ^ ) ) J(l",-l")+(lx2(tl)-ly2(tl)) J(l ',-1')+
scalar term:
+J(a2,a2)
5.5. Parameterization of the ligand field in TM and RE complexes: a

spectroscopically adjusted EH-theory
Exchange integrals in TM-RE pairs depend on the electronic states of the TM
and RE involved and on a large number of orbital exchange parameters
(Eq.37) which need to be calculated in advance. Parameterizations of the LF
in TM complexes using density functional theory (DFT) have been quite
successful in recent time allowing to calculate almost parameter-free ground
and excited state multiplets of TM complexes . In contrast, DFT
applications on RE complexes turn out to present severe problems,
presumably due to basis sets and/or functionals for 4f-orbitals which are not
well established yet. Thus, a DFT calculation 29 on the octahedral YbCl63"
unit using triple zeta basis of 4f and a LDA 30 plus PW96 31 functional and
average occupancy (13/7 electrons on each f-orbital) yields a a2u<t2u<tiu
splitting which exceeds by far the experimental one (see Table 2).
Table 2. Energies of the 4f-orbitals (in cm"1) in octahedral YbC^3" and YbBr63"
complexes as calculated using EHT and their comparison with AOM energy
expressions to allow getting e o and e* -bonding parameters for Yb-Cl and Yb-Br
bonds
Orbital (O h ) I YbCl63" I YbBr63" I AOM-energy
AOM parameters EHT DFT 3 exp. EHT expression
ti» 773 2793 799 472 (l/2)(4e q +3e 1I )
t2u 161 1508 221 93 (5/2) eK
a2u 0 0 0 0 0
eo 338 944 333 208
e^ 64 603 88 37
e^/ep 1 0-19 0.64 0.26 | 0.18 |
PW96 functional; composition of f-MOs: t lu 70%, t2u 74% a2u 100% Yb;
aiu-tig(non-bonding Cl 3p) energy difference -0.075 eV.
In addition, mixing of Cl (3p) and Yb(4f) functions in ti u (70%) and

t 2u (74%) is calculated anomalously large, which we attribute to a nearly
vanishing 4f(Yb)-3p(Cl) energy separation. The problems become even worse
when trying to calculate exchange coupling in Yb3+-Mn2+ pairs; the attempt to

let such calculations converge failed already at the SCF step; scrambling of
4f(Yb), 3d(Mn) and 3p(Cl) orbitals leads to oscillations in the charge density,
between spatially separated but otherwise energetically close ranges. In
contrast, less sophisticate Extended Hiickel (EH) calculations have lead to
reasonable result as is seen by the results included in Table 2. In the EH-
method 32, diagonal elements of the one-electron operator for a given orbital
(HH) are approximated by the valence state ionization potential from that
orbital (VSIP);; (Eq.38), while off-diagonal elements Hy are taken to be
proportional to diatomic overlap integrals Sy (Eq.39). The parameter k can be
adjusted such, as to fit spectroscopic lODq values; we take k=1.25(1.75) for
Mn2+(Cr3+) leading to lODq parameters 6700(15000) cm"1 close to experiment
7500(12400)cm"1 . Adopted values of the VSIP for 4f (Yb) and 4p(Br) (-
12.36 eV and -14.20 eV) are altered compared to tabulated ones (-13.86 and -
13.10 eV 33) to ensure a larger 4f(Yb)-3p(Cl) and 4p(Br) energy separation.
As for the spin-orbit coupling constant of Yb3+ we take the value of-0.353 eV
from a relativistic ZORA calculation of the free Yb3+ ion. Using this
spectroscopically adjusted EH-theory a reasonable agreement between
calculated and spectroscopic ground and excited energy levels for YbCl63" and
YbBr63" chromophors has been achieved (Table 3).
Hfi=KVSIP)fi (38)
Hij = k.Sij(Hii+Hij)/2 (39)
Table 3. Spectroscopic data on the ground and excited state energy levels (in cm"1) of
YbCl/" and YbBr63" chromophores and their values calculated using EHT.
Energy level I MnCl2:Ybj+ C2v 1J I Cs3Yb2Br9 C 3 / ; I CszNaYbC^CV
exp. calc.EHT exp. calc.EHT exp. calc.EHT
2F7/2 gr.state
0 0 0 0 0 0 0 T6(2)
1 268 362 114 218 362 385 T8(4)
2 371 435 140 276 701 703 T7(2)
3 608 733 441 439
F5/2 exc.state
0* 10179 10193 10119 10102 10243 10203 Tg(4)
1* ...? 10242 10146 10135 10708 10657 T7(2)
2* I 10764 10670 \ 10590 10402 |
Experimental values taken from: 1) P.Gerner, H.U.Guedel, personal communication;
2) M.P. Hehlen and H.U.Gudel, J.Chem.Phys. 98(3) (1993) 1768; 3) R. Schwartz,
Inorg. Chem. 16 (1977) 1694.
3.4. Integrals of electronic 4f(Yb3+)-3d(Mn2+ or Cr3*) hopping

EH calculations yield d- and f-orbitals which can be easily identified by their
dominating percentage of d- and f-character, respectively. For the purpose of
evaluating effective integrals of electron d<=>f hopping we cut the ligand part
from these MO's and take the 12x12 matrix containing MO coefficients (in
columns) for d and f-orbitals only. Let us denote this matrix by U and the
diagonal matrix containing the respective MO energies by A. From these two
matrices we can generate an effective one-electron Hamiltonian which
contains on the block diagonals 5x5 and 7x7 ligand field matrices for d
(Mn(Cr)) and f (Yb) electrons and the 5x7 matrix of integrals of d<=>f
hopping on the off-diagonal. This is accomplished as follows; we introduce
the ovelap matrix S(Eq.4O). One can easily find S~1/2 using the following
recipe; get eigenvalues A,(diagonal matrix) and eigenvectors (in columns) U,.
S"1/2 is then given by eq.41. As has been shown by DesCloizeaux 34 the
effective matrix h is calculated following Eq.42.
S=U.UT (40)
S- 1 / 2 =U S A; 1 / 2 U S T (4i)
h = S"1/2UAUTS-1/2 (42)
Calculations on RE-TM dimers have been done (see Section 3.3) using a
basis of real d(TM ), f(RE) and s and p(ligand) functions. For the sake of
applying the operator 6 we get the matrix h' (Eq.43), by using the matrix T
(see Appendix A.2) which transforms the 1=3, mi=0,±l, ±2, ±3 functions into
t 2 (0", ±1"), t!(0',±r) and a2 (eq.36), while leaving the basis of real d-
functions (d^, d^yz and d^.^xy) unchanged:
h'= T"'hT (43)
5.5. Two electron exchange integrals involving 4f(Yb3+) and 3d(Mn2+

or Cr3*)
In calculating two-electron exchange integrals we made use of approximation
(Eq.8). In other words, from the MO's dominated by the d- and f-orbitals we
take only that part residing on the bridging ligand (1). Let d^ and fv be
representative for two such functions; we have:
i
172 M. Atanasov, C. Daul and H. U. Oiidel
fv= Y^PJ < 44 >

Then the exchange integral is approximated by a sum of intra atomic
exchange integrals on the bridging ligands (1) times squares of the coefficients
c and c y , the sum extending over
(45)
p-orbitals belonging to the same bridging ligand and summation being taken
over all bridging ligands:
[d,fv|d^fv]=XX44fi:; (46)
3.6. An elaborated example: exchange coupling in ClsMnClYbCls6'

(corner sharing)
As an example of application of our method we consider the Cl5MnClYbCl56"
dimer with a linear Mn2+-Cl-Yb3+ bridging geometry. We consider here the
exchange coupling between the ground states of Mn2+ (6Ai) and Yb3+ (T6)
(C4v symmetry).
Diagonalizing the Hamiltonian of the spin-orbit coupling plus the ligand
field we get the following expressions for the components of the Kramers
doublet F6:
KH-^ffH) (47)
|-iH'iH«4)-H>
With a=0.020 b=-0.563 c=0.826; we note the dominance of the
ti(±l',0') orbitals involved in stronger a-interactions and the negligible
participation of the weaker ^-interacting t2(±l",0") orbitals. In Table 4 we
list Mn2+ -Yb3+ hopping integrals ( d j h | f v ) and two-electron exchange
integrals [dM fv| dM fv] (in parenthesis), which we calculate following the
recipes in Sections 3.4 and 3.5, respectively. From these data we calculate
orbital exchange parameters, (J(v,v)), where we differentiate between
contributions from kinetic (Jk(v,v), Eq.(48)) and potential (Jp(v,v),Eq(49))

exchange; values of these parameters are listed in Table 5.
h(y^= ^l(*Mf>u\hM (48)

W,v)=}0j^\dM (49)
Derivation of expressions for J^ and Jxx=Jyy is straightforward:
Jzz: 11/2>= a(-l ",-l/2)+b (O',l/2) +c (1 ',-1/2)

sz' 11/2>= -(l/2)a(-l",-l/2)+(l/2)b (O',l/2) -(l/2)c (1 ',-1/2)
Osz' | l/2>=-(l/2)aJ(l",l")(-l",-l/2)+(l/2)b J(0',0') (O',l/2) -(l/2)c
J(l',l')(l',-l/2)
Ja= 2 <l/2 | Osz' 11/2>= -a2 J(l ",1 ")+b2 J(0',0')-c2 J(1 ',1') (50)
sx' I l/2>=(l/2)a(-l",l/2)+(l/2)b(O',-l/2) +(l/2)c (l',l/2)
Os x '|l/2>=(l/2)aJ(l",l")(-l",l/2)+(l/2)b J(0',0') (O',-l/2) +(l/2)c

J(l',l')(l',l/2)
inT 2 <-l/2 | Osx' | l/2>= b2 J(0',0') (51)
Using these expressions, we obtain after substitution of a,b,c and Jk(v,v')

and Jp(v,v) values of J^ and Jxx=Jyy listed in Table 6 (last column). Following
the same procedure exchange coupling constants for Mn2+-Yb3+ clusters with
face (Cl, Br), edge and comer (Cl) sharing geometries (Figure 4) have been
calculated. They are included in Table 6.
O
Table 4. Orbital exchange constants (hopping integrals) (d^ | h | fv) and two-electron exch exchange integrals [d^ fv| d^ fv] (in ^
parenthesis) of a Cl5MnClYbCl56" dimer with linear Mn2+-Cl-Yb3+ sharing (corner shared share MnCl64" and YbCl^3") octahedra. b
Energies of the ligand field splitted d- and f-orbitals are also listed (all energies in eV). |_
, ,
I .
I t2-i" .
I ,
t 2 -0" I t21" ,
1 — ,I
trr ti-o , . . i"
t2-l"
-12.337 t 2 -0"
-12.337 t21"
-12.337 tpT
-12.267 tpOJ
-12.26 UV_ a2 °-
xy(t2g) -11.483 -12.337
0.0 -12.337
0.0 -12.337
0.0 -12.267
0.0 -12.269
0.0 -12.267 -12.360 ^
_xy(t
^2g) -11.483 0.0
(oo) 0.0
(oo) 0.0
£ao) 0.0
(o.o) 0.0
(o.o; 0.0 0.0 ^
_x2^-y2(eg) -10.571 (O0)

0.0 (O0)
0.0 (O0)
0.0 (0.0)
0.0 (0.0)
0.0 (0.0) (0.0) S
x2-y2(eg) -10.571 0.0
(0.0008) 0.0
(0.0002) 0.0
(0.0008) 0.0
(0.0005) 0.0
(0.0; 0.0 0.0 &
^(eg) -10.617 (0.0008)
0.0 (0.0002)
0.0 (0.0008)
0.0 (0.0005)
0.0 (0^0)
-0.066 (0.0005) (0.0002)
zêg) -10.617 0.0
(0^0) 0.0
(£0) 0.0
(O0) 0.0
(OO) -0.0666
(O0 0.0 0.0
yz(t2g) -11.486 (0^0)
0.0089i (£0)
0.0 (O0)
0.0089i (O0)
-0.0098i (O0)
0.0 (O0) (0.0)
yz(t2g) -11.486 0.0089i
(0.0013) 0.0
(0.0023) 0.0089i
(0.0013) -0.0098i
(0.0012) 0.0
(OXV -0.0098i 0.0
xz(t 2 g ) -11.486 (0.0013)
-0.0089 (0.0023)
0.0 (0.0013)
-0.0089 (0.0012)
-0.0098 (OX)]
0.0 (0.0012) (0.0024)
xz(t 2g ) | -11.486 I -0.0089
(0.0009) 0.0
1 (0.0007) [ -0.0089
(0.0009) [ -0.0098
(0.0007) | 0.0
(0.0 -0.0098 0.0
| I (0.0009) 1 (0.0007) [ (0.0009) | (0.0007) | (0.0) | (0.0007) | (0.0008)
f z fz
MnO 2 :Yb 3+ Rb2MnCl4:Yb3+
CsMnBr3:Yb3+
Fig.4 Mn-ligand-Yb sharing geometries of Yb-Mn dimers in Yb3+ doped Mn2+

halogenide solids; for RbMnCl3:Yb3+ and CsMnCl3:Yb3+ both comer and face sharing
is encountered.
Table 5. Kinetic and potential (in parenthesis) exchange parameters (in units 10"4
eV2/3U[5U] and \QAI5 eV, respectively) for Mn2+-Yb3+ and Cr^-Yb3* halide
bioctahedral complexes with face, edge and corner sharing bridging geometries
Octahedra I CrBr3Yb I MnBr3Yb I MnCl3Yb I MnCl2Yb I MnClYb
Sharing: face face face edge corner

t2:
J(l",l") -4.31 -5.26 -12.90 -3.64 -1.60
(6.49) (3.22) (5.21) (6.70) (30.00)
J(0",0") -0.31 -5.68 -6.48 -16.20 0.00
(1.83) (1.58) (4.02) (1.00) (32.00)
J(l",-1") 0 0 0 3.35 0.00
(O0) (O0) (O0) (0.0) (0.0)
ti:
J(l',l') -7.14 -4.35 -9.58 -8.60 -1.88
(7.10) (3.04) (8.80) (6.70) (24.00)
J(0',0') -74.65 -47.33 -59.02 -7.34 -44.36
(3.81) (1.24) (4.30) (1.80) (0.00)
J(l',-1') 0 0 0 7.19 0.00
(O0) (O0) £00) (0.0) (0.0)
a2:
J(a2,a2) -16.72 -5.68 -5.55 -2.28 0.00
I (16.93) I (6.77) I (18.90) | (5.40) | (34.00)
4. Applications, Results and Discussion
4.1 Exchange integrals in Mn +-Yb and Cr +-Yb + pairs
4.1.1 Exchange coupling in Br3CrBr3YbCl33" with Cr3"1" and Yb3+ in their

groimd states
With three unpaired electrons in the t2g shell, CrBr63' possesses a 4A2 ground
state, while YbCl63" is in a T4 ground state (notations for the C3v site
symmetries). Exchange constants (Jzz= -0.0064 eV and Jxx=Jyy=0.0052 eV)
deduced from neutron scattering experiments 35 indicate exchange coupling to
be highly anisotropic - antiferromagnetic along the Yb-Cr chain and
ferromagnetic in perpendicular to this direction. In a qualitative agreement
with this interpretation of the neutron scattering data, our calculations show
/3U['
Table 6. Kinetic and potential (in parenthesis) exchange contributions (in units 10"4eV2/3U[5U] and lO"4^ eV, respectively) of
the anisotropic exchange Hamiltonian Hexc=~Jxx s^-Jyy sy Sy - J^ sz Sz for CrI"[Mnn]-Ybin 1halide face, edge and corner shared
bi-octahedra; Sj, Sj (i=x,y,z) - spin -operators acting on the (fictious) (1/2) and (3/2[5/2]) spin-functions on Yb and the TM
Cr[Mn], respectively.
I Br3CrBr3YbBr33" I Br3MnBr3YbBr33~ I Cl3MnCl3YbCl33' I CUMnC^YbCL,

CUMnC^ 5" I Cl5MnClYbCl55" I |
calc. exp.(meV) «;
gr.st. exc.st. gr.st. exc.st. gr.st. e:
exc.st. gr.st. exc.st. 3'
Jm -11.20 -0.64 -5.34 -0.55 -6.97 1.31 -1.19 -I-0.80 -12.77 -1.35 ^
(0.30) (0.20) (0.40) (0.10) (0.60) (4.43) i(3.94) (16.00) (10.00) a*
Jxx -11.39 0.52 -7.28 -2.00 -9.95 -2.25 -1.98 (0.40 -14.06 0.00 I
(1.45) (0.60) (0.60) (1.75) (1.40) (0.23) ((1.74) (0.00) (8.00) |
Jyy -11.39 0.52 -7.28 -2.00 -9.95 -2.25 -6.94 -4.74 -14.06 0.00 |
I (1.45) | (0.60) (0.60) | (1.75) (1.40) | (0.83) (2.74)
( (0.00) (8.00) ts
?r
CO
•i
I
179
(Table 6) J^ to be mainly antiferromagnetic; ferromagnetic coupling in the z

direction is calculated about four times smaller than that in the x-y plane, the
latter compensating perhaps with excess the negative J^Jyy. With one
unknown parameter U and with different approximations inherent in
calculating kinetic and potential contributions to J, we are not able to calculate
total values of J ( a sum of two times the ferromagnetic term plus four times
the antiferromagnetic term). One can speculate and get U=2.35 eV - not quite
unreasonable but still smaller than expected based on experience and take a
scaling of the ferromagnetic terms by a factor of 5 in order to bring numerical
data quantitatively agree with experiment. A good agreement between
calculated and experimental exchange integrals using an EH-approximation
for the integrals of Cr-Yb hopping, but accounting for electronic correlation
within a CI treatment has been reported by Mironov et al 36 .
4.1.2. Exchange constants for Mn2+ ground state - Yb3+ ground and excited
states of Mn2+-Yb3+ dimers with corner, edge and face sharing.
Comparing data for the Cl- clusters we conclude that antiferromagnetic
(kinetic) exchange parameters with Yb3+ in the ground state are strongest in
the Mn-Yb dimers with corner sharing, about twice weaker for face sharing
and the smallest for edge charing, with a disctinct anisotropy in the latter case.
Calculated values for the antiferromagnetic coupling with Yb3+ in the emitting
excited Kramers doublet are much smaller than those for dimers with Yb3+ in
the ground state. Going to the potential (ferromagnetic) contributions for Yb3+
in its ground state, we notice the large anisotropy with J^Jyj^Jzz for face
sharing , J ^ J y ^ O for corner sharing and Jzz^JxxJyy for edge sharing.
Focussing on the ferromagnetic term Jzz, we notice the tremendous increase
of this parameter from face to edge to corner sharing topology. Again
concerning J^ it is also worth to note that exchange coupling between
6Aj(Mn2+) and the lowest emitting Kramers doublet of Yb3+ is not as much
smaller compared to the coupling in the 6A](Mn2+)-Yb3+ ground state, as is the
case for antiferromagnetic coupling.
4.2. Exchange coupling and its implications for the mechanism of the
up-conversion of light in Mn +-Yb + dimers
Electronic transitions accompanied by a spin-flip within a TM are spin-
forbidden, however they can gain intensity by exchange coupling to other
ions. Focusing on a Mn2+-Yb3+ pair, we take Mn2+ in its ground state 6Ai and
Yb in a fictitious spin Vz state to represent its ground state or some excited
doublet state. These states can couple with spins parallel or antiparallel to
each other to give pair states with total spins S=3 and 2, respectively. The flip
of one spin within the d5 configuration of Mn2+ site leads to *Ti and *T2
excited states. A parallel coupling between the S=3/2 spins of Mn2+ and that
of s=l/2 of Yb3+ leads now to the same total spins S=2 as the one for the
antiferromagnetically coupled S=5/2(6Ai) - s=l/2 pair; the transition between
these two states corresponds formally to a 6Ai—»4Ti, ^ 2 excitation within a
single nuclei Mn2+ complex; it is not spin-forbidden with respect to the cluster
total spin (Fig.5). As has been shown by experiment 6 ' n and theory 5 ' 6 ' 7 , such
transition can gain intensity by a mechanism which is not identical, but
closely related to that of the exchange coupling between magnetic ions.
T S=l
l l o r 2 , ^
S=3/2 / \
—5 C S=2 \
Mn Yb
Fig.5 Exchange mechanism of intensity gain for intra-configurational spin-forbidden

transition at octahedral Mn2+ site in Mn2+-Yb3+ halide bridged dimers.
In general terms, the interaction between the radiation field and a pair of
paramagnetic ions a and b is described by the operator H' (Eq.52). E is the
electric field, n^y are vector parameters describing the dipole moment of the
transition between the singly occupied orbitals i and j on centers a and b,
respectively; Saj and Sy are vector operators which act on the genuine spins of
electrons on orbitals i ang j . We should note that the spin part of H' is
identical with the
H^SS^-EXs^) (52)
iea jEb
one of the exchange operator Eq.3; indeed the latter equation can be rewritten
from a form involving total spins Si and S2 for ions a and b into one involving
contributions from separate orbitals:
H-'-SIW'-^) (53)
tea jeb
It has been shown7 that the parameters n^bj equal the derivative of Jaibj
with respect to E, i.e.
naibj ~ (~^~)W K '
In other words, to consider the mechanism of pair-transitions implies to

reconsider the mechanism of superexchange in the presence of an electric
field. This leads to the following expression for n ^ 6 :
4(ai|p|bj)h(bj->ai) ^(uJplaiXaibjIJbju)^-^) |
K**~ AE(bj->ai) & AE(ai->u)
2(ai|p|v)(vbj[|bjai)nv
ti AE(v-âi)
[expressions obtained by interchanging ai and bj in the above] (55)
Here p is the dipole moment vector, n^, the number of electrons in the
fi,v-th orbital, AE(bj—»ai) the energy required to transfer one electronfrombj
to ai, h(bj—»ai) - the transfer integral between bj and ai and {aibj\bj/j) the
two-electron integral:
(aibj\\bJu)=j\ai(l)bj(2)^-bj(l)/l(2)dTldT2 (56)
All summations over (j. and v are to be carried out over all the empty,
singly and filled orbitals except for ai and bj. The operator H' and the nî
parameters defined by Eqs.52 and 55 are quite general and allow to obtain the
intensity of any transition involving local excitations either of a or b or local
excitations in which both electronic states on a and b change (see below).
Appications of this formalism has been precluded due to the large number of
model parameters and the lack of any practicable computational scheme to
allow calculation of these quantities from fist principles. That is why,
applications of the model has been confined only to qualitative interpretations
of spectroscopic data6'13"18.
In what follows, we study in some detail the correlation between J]2 and
the intensity of a pair transition. We take as a simple model example now a
pair TM(d2)-RE(f1) (Fig.6). The S=l (triplet, T) and ^(doublet, D) states of
TM and RE couple to yield pair states of total spin 3/2 (quartet, Q) and V2,
with the wavefunction of the latter (Ms=l/2 component) given by:
¥ £ =^T(MS =l)D(ms = - l / 2 ) - ^ 7 W , =0)D(ms =1/2) =
(57)
We consider the transition from this state into the excited doublet state in
which one spin of TM has flipped to give a singlet state (S) TM(S=0)-
RE(s=l/2); it is described by(Ms=l/2 component):
V&=S(M, =0)D(ms=V2) = ^\d;d-r\-^\d;d;r\ (58)
tfl
S=0 S=l/2
S=3/2
(^Q MS=l/2
S^ll S=l/2 / D
T D
Fig.6 A simple model example of intensity enhancement of the spin-forbidden
S=l—>S=0 (spin-flip) transition of a transition metal with two unpaired electrons(such
as Ni2+ in octahedral coordination) and S=l/2 ion (such as Cu2+ or Yb3+(pseudo-spin
>/,)).
182 M. Atanasov, C. Daul and H. U. Gildel
The two states interact with S=l/2 charge transfer states in which one
electron has jumped from TM to RE, L/ 2+ / + /~|,L/, + / + /i,or from RE to TM,
d*d-d*\, d*dXd' • This mixing is the cause both for the exchange
1 1 2| 1 2 2
stabilization of »p£ over the »?£, = d*d^f* (described by the kinetic
exchange integral jkn ) and the intensity of the *F^, - » ¥ ^ , pair transition.
As shown in Appendix A.3, using both perturbation theory and applying
explicitly the operator H'(eq.52), the ^F^, - ^ ^ transition probability, is
(neglecting higher order two-electron terms) proportional to the square of the
matrix element (yfD \H\^) •
W*->-^^W^ + W^>]~ (59)
Let us compare this result with Eq.(60) for the exchange integral Jn:
Jk =-h 2[ ! , * ] h 2f 1 | 1 1
(60)
We see that in addition to the sum over squares of the hd2fand hdlf,
common to \ fofi, H^SD) '2 an( ^ *î2 > t ' i e r e ^s an interference term
(hd2f.hd}f) which appears in the intensity expression. This can add or
subtract from the first one, to show that, in general, | (w°D //j*F^) 12 and J\2
will not be interrelated. However, as it frequently appears, one of the two
terms, that for hd2f or hdxfy is small or zero and both | Nf^ M Y ^ , ) I2
and Jj2 are dominated just by one leading term. It then follows that for such
cases the intensity of pair transitions will be proportional to J\2. Our results
show, that this is the case for corner (see Table 5) and face sharing; in both
cases, the exchange Mn2+-Yb3+ is governed by one single dominant
contribution - the one between the d^ (Mn2+) - ti(O') f^ (Yb3+) orbitals: in the
case of edge sharing several competing terms occur. In Fig.7 we have plot the
Jkzz values for the complexes studied; we see that Mn2+-Yb3+
antiferromagnetic coupling is largest for corner sharing; it is tempting to relate
this result with the experimental finding, that up-conversion is most efficient
in this case; the underlying electronic excitation involves a change of the
electronic state created after the first absorption 6Ai(Mn2+)-2F7/2(Yb3+) - »
6Ai(Mn2+)-2F5/2(Yb3+) to a state where simultaneously Yb 3+ turns back into the
ground state 2F5/2, while Mn 2+ (6Ai) becomes excited into the ''Tig state.
• Jzz y .
A y ^ \ MnCl7Yb
/\ exc. state
-4 - / \
MhBrj^Ns/ \ MnClYb
-8 ' MnCl.Yb \
\ gr. state
-12 - \*
Fig.7 Antiferromagnetic J^ (kinetic) exchange coupling constants for Mn +-Yb3+

dimers with octahedrally coordinate Mn2+ and Yb3+ and three, two and one sharing
halide atoms (face-, edge- and corner shared octahedra).
Turning to the bilinear dependence of the intensity of pair transitions on

terms of the form hlf,h\fwn& hlf.h2f we notice the following; integrals of
TM-RE hopping, hxf and hlf, are expressed in terms of hdl and hfl metal-
ligand hopping integrals as follows:
hdf~_htL==_±£== (61)
J I.
\Q - ^ZZ t gr. state
12 - /
/ 7 exc.state
g . / /MnClYb
4" Jl
- x* o vu y MnCl2Yb
MnBr3Yb S
MnCl3Yb
Fig.8 Ferromagnetic J^ (potential) exchange coupling constants for Mn2+-Yb3+ dimers

with octahedrally coordinate Mn2+ and Yb3+ and three, two and one sharing halide
atoms (face-, edge- and corner shared octahedra).
Since coefficents cM and cfl (Eq.46) depend linearly on hdi and hji as well,
one can expect that integrals of potentials exchange Jf2 and the transition
probability | (w^ H^¥^ \2 will show-up again in a symbathic behaviour,
they both increase with h. Here cross terms, hlf .h2f with contributions from
different ligand orbitals are automatically included (which is the missing
terms in the comparison of | (*F£ I t f 1 ^ ) ^ with û)- R is striking, that
values of J?z plotted in Fig. 8 follow the same trend as the one found
experimentally for the efficiency of light conversion; they increase from face
to edge and corner sharing (cf with the ./,* plot, Fig 7) . We attribute this
difference to mixed terms (h d l f . hdlf) which do not appear in J\2, but
which are important in the case of edge sharing. The correlation does not
change qualitatively when comparing exchange integrals for the ground and
for the lowest emitting excited state of Yb3+ (see potential exchange terms in
Table 6).
5. Conclusions
1) Exchange coupling between TM(Mn2+) and RE(Yb3+) has been studied
experimentally with the aid of understanding the mechanism of up-conversion
of light [3,4]. The analysis of the experimental data up to date has lead to the
proposition that the phenomenon is governed by exchange coupling between
the TM and RE. The modelling of exchange coupling between Mn2+ and Yb3+
lends support to this hypothesis; we find that both kinetic and potential
contribution to the exchange coupling constant are largest for a corner
sharing Mn2+-Yb3+ dimer. We further get the interesting result, that
ferromagnetric exchange integrals exactly mirror trends born out by
experiment: efficiency of up-conversion increases from face to edge to corner
shared Mn-Yb ions.
2) We propose a procedure of calculating exchange coupling constants based

on the following steps: (i) In a first step RE-TM transfer (hopping) integrals
between orbitals carrying unpaired (magnetic) electrons on RE and TM are
determined from a MO-calculation utilising the concept of orbital exchange
pathway and effective Hamiltonian theory, (ii) In a second step, many-
electron wave-functions for ground or excited electronic states on RE and TM
are constructed in which spin-orbit coupling (which dominates for the RE) is
fully (variationally) taken into account, (iii) Finally, the data from steps 1 and
2 are combined to obtain numerical values for kinetic exchange integrals
(within Anderson's super exchange theory) utilising a non-Heisenberg (orbital
dependent) exchange operator. The MO's corresponding to the d(Mn) and

f(Yb) magnetic orbitals have been used to also calculate ferromagnetic
exchange integrals as well. Our method has been implemented at the level of
an extended Huckel model (EHT), where a restricted number of parameters
has been adjusted to fit spectra in high resolution of separate MnCl64" and
YbC^3" chromophores. However, the procedure is quite general and can be
used in combination with any method (DFT or ab-initio) which is able to
yield reasonable account for the magnetic orbitals involved. In this respect
we have to state, that available basis functions and/or functionals within the
DFT methodology are not sufficient yet to do this. We motivate further work
for improving DFT for rare earths.
Appendices
Appendix 1
Derivation of the orbital operator (O) for exchange coupling in the case of
orbital degeneracy.
The orbital operator O, representing the exchange coupling tensor J within
the 1=1 basis
"J(U) J(l,0) J(l-l)"

J = J(0,l) J(0,0) J(O,-l) (A.1.1)
j ( - u ) j(-i,o) j ( - i - i )
can be expanded into a series of orbital operators l»ly,lz acting on angular
momentum eigen functions for 1=1, mi=0, ±1 and the identity operator 1 as
follows:
6 = alz2 +b(UE+Ux)+c(IyIz+lzly) + d(lx2-ly2) +e (Uy+M*) + f 1 (A. 1.2)
The coefficients a to f can be determined in the following way; we substitute

the matrices of the operators 1*, ly and lz and 1 2 1 :
o -4= o o ^ L o r n r -,
V2 4i [i o ol p oo
l x = — 0 — 1= — 0 — l z = 0 0 0 1 = 0 1 0
4i J i y V 2 4 i o o - i o o i
o -^ o o 4= o
(A.1.3)
into eq (A. 1.2) to obtain, after collecting terms:
a+f -^(b-ic) d-ie

V2
6 = 4=0> + ic) f ^ ( - b + ic) < A - L4 )
V2 V2
d + ie -pr(-b-ic) a+f
V2
The comparison with eq (A. 1.1) yields the following connection between
the parameters a to f and the components of the exchange tensor:
J(l,l)=J(-l,-l) = a+f; J(l,O)=J'(l,O)+iJ"(l,O)= 1 ( b _ i c ) =J( 0 ,l)*;
J(l,-l)=J'(l,-l)+iJ"(l,-l)= (d-ie)=J(-l,l)*; (A. 1.5)

J(O,-I)=J'(O,-l)+iJ"(O,-l)= - L (_b+ic)=J(-l,0)'; J(O,O)=f
V2
We thus obtain
a=J(l,l)-J(O,O); f= J(0,0); b= <2 J'(l,0); c= -<2 J"(l,0); d=J'(U-l);

e=-J"(l,-l) (A.1.6)
After substituting in eq.A.1.2 we obtain eq.14 in Section 2.2.1. From eqs

A. 1.5 we derive the relations between the components of the anisotropic
exchange tensor (eq. 15).
Appendix 2
Spin-orbit coupling matrix elements for f-orbitals and their transformations

Adopting the following standard order of real f-spin-orbitals:
(1,2,3,4,5,6,7,8,9,10,11,12,13,14)=
((Psa^spAa^sPâ.nspâ^pâ^p^caêp^caêp) A.2.1
We have the following non-zero spin-orbit matrix elements (in units of £,
the spin-orbit coupling constant):
(1,4)= JI; (l,12)=-i JI; (l,13)=il; (2,3)=- JI ; (2,11)= -i (1(2,14)=- ii;
V8 V8 2 \8 V8 2
(3,6)= JI; (3,10) =-i JI; (3,11)= i; (3,14)= -i JI; (4,5)=- JI; (4,9)=-iJI
\8 V8 V8 V8 V8
(4,12)=-i; (4,13)= -i JI; (5,8)= -i J I ; (5,9)=il; (5,12)=-iJI; (6,7)=-i J I ;
V8 V2 2 1/8 V2
(6,10)= -il; (6,ll)=-iJI; (7,10)= -JI; (8,9)= JI; (9,12)= -JI; (10,11)=
2 V8 V2 V2 V8
JI;(11,14)=-JI; (12,13)= JI (A.2.2)
V8 V8 V8
In eqs (A.2.2) only matrix elements (ij) with i<j have been listed; the (j,i)
matrix elements can be obtained taking the complex conjugate (i j)*.
The spin-orbit coupling matrix elements (A.2.2) are set up in a real f-
orbital basis. To convert this matrix into the representation of the t 2 (0", ±1"),
ti(0',±l') and a2 basis (eq.36), a two-step transformation has been performed.
We first take:
(-3,-2,-1,0,1,2,3)= (<psA,Tts,c,n,A,^) *
-- 0 0 0 0 0 —]=
2 VI
0 —\= 0 0 0 4- °
° ° i ° i ° ° (A.2.3)
0 0 0 1 0 0 0
o o 4= o —1= o o
V2 V2
o 4= o o o - ) = 0
-^= 0 0 0 0 0 —\=
_& V2.
or in short:
fi = fR.T, (A.2.4)
-Modelling of Anisotropic Exchange Coupling 189
191
followed by:
(-1", 0",l", - r , 0 \ l ' , a2)=(-3,-2,-l,0,l,2,3)*
~-^= 0 0 0 0 ~^= 0
2V2 2V2
0 -J= 0 0 0 0 —j=
V2 A/2
* 0 0 Mr -J* 0 0 0 (A.2.5)
2V2 2V2
0 0 0 0 1 0 0
4= 0 0 00 -^L 0
2V2 2V2
0 -£= 0 0 0 0 - 4 =
1/2 V2
0 0 - ^ L L -JL* 0 0 0
2V2 2V2
or
f=fi.T2
A.2.6)
leading
e=f-.T2 to (
leading to (A.2.7)
for the f-orbital
Tf=T!.T 2 part of T (eq.43) and diagonal matrix elements of 1 for(A.2.7)
the d-
orbital part.
Appendix 3
On the relationship between the transition probability for pair excitations and
the contributions to the exchange coupling constant.
Let us consider two ions (a) and (b) possessing 2 and 1 unpaired electrons on
orbitals ai,a2 and b, respectively. The two electrons on (a) couple to give one
triplet *Fy at lower energy and one singlet *F^ at higher energy, the energy
separation between the two terms being twice the exchange coupling integral
between the two electrons (Fig.6). The single electron on b gives rise to a
doublet, *F^. The wave functions of these terms are given by:
^=^(a!a;+a;a!)U *»«£} (A3.1)
v°=-j=(a;a-2-a;a;)
We assume that the triplet state on (a) couples with the doublet one on (b)
to give one quartet (Q) and one doublet (D) state, 4 ^ , and 4 ^ , , respectively,
while *Fj and *F^ combine to a single doublet state, 4*^,; explicitly we
have
KW, =^) = â+2b- - - L . - ^ ^ a " +a;a+2)b+ (A.3.2)
V&W, =±) = -j=(a;a-2 -a;a;)b+

In addition to these states, there are charge-transfer states resulting from
ai—»b, a2—>b and b-»ai and b ^ a 2 excitations; their energies and wave
functions are described by:
¥,(<*, -*b) = a;b+b- ; AE(a^b)
T 2 (a 2 -> 6) = a\b"b-; AE(a2->b)
^ 3 ( A ^ a 1 ) = a1X«2+; AE(b->ai) (A.3.3)
«P4(6->fl2) = a,+a*aJ; AE(b->a2)
In what follows we derive dipole transition matrix elements for the
^TD ~* ^SD transition (corresponding to a spin-flip local excitation) and for
the exchange coupling constant using to different approaches - perturbation
theory and, alternatively applying the operators H' (eq.52) and HeXC (eq 53) on
the wave functions of the exchange coupled pair states (eq A.3.2).
A.3.1 Perturbation theory

Here we account for the first order corrections to 4 ^ , and ^ ^ (eq
A.3.2)due to their mixing with the charge transfer states (eq A.3.3) and use
these corrected functions to get transition dipole matrix elements; second
order corrections to the energy of «p^,, E(2)( »F£), allows to account for the
stabilization of »p£ over the »p£ and thus for the value of the exchange
coupling constant. We use the well known expressions:
£<2> _ y I Vmn 1 • xpd) _ y Vmn t p(o) • (A.3.4)
» Z^ r(0) _ £<<,) » ^ p(0) _ r(o) i m
For the matrix elements Vnm we obtain:
{v%,\h\V2(a2 - » A ) ) = ^ | A . B ; (^|*|«P 2 ( fl2 ^b)) = -^=halb
(^|Aj«P,(A ->a,)) = - ^ 1 6 i ( i £ 1^3(6->«,)) = - - ^ A . , , (A-3.5)
( ^ 1*1^(6 -»a 2 )) = ^ 2 i J (¥£|A|^ 4 (6 -»a 2 )) = - - ^ ^ 2 4

We thus get:
ro 2AE(a]^b) 2AE(a2^>b) 2AE(6-*a,) 2A£(i->a 2 )

£(2)(^) = 0 (A.3.6)
and, since (S=3/2):
S.JI2=E(S-1)-E(S);
Jl2 ~ 3L ( TD) ( ro)J" albLAE(a,^b) AE(bâ,)_
_A2 [ * 1 1 (A.3.7)
a2\hE(a2-*b) AE(b^>a2)
First order corrections for *F^, and W^, are given by:
tf" =£ ^ *(«, ->b)-M ^ ^(a2 ->6) +

TD i2AE(ai-+b) Kl ' \2AE(a2^b) K2
fl *& ipp_>«,)_ I ^ y(b_>a2) (A.3.8)
¥£(1) =4= — ¥(a, -»*>)+4= — V(a2 ->*>) +
+ 1 K* »F0->fl.)+ - L ^ *P(6-»fl2) (A-3-9>

V2AE(6-+a,) ' V2A£(6â 2 )
To calculate the transition dipole *Fj^—>*F^,, the following matrix

element of the dipole operator p has to be evaluated:
n ( ¥£->¥£)= («?£•> |jP^|vpW)}+(^> 1^1^") (A.3.10)
Direct substitution from eq.A.3.2(for the zero order wave functions <p^)<<>)
and ¥£<o)) and A.3.8,9 yields:
(A.3.11)
Here we have denoted the matrix elements of the dipole moment operator
p for the a2-b and a r b pairs by p^b and paib, respectively.
A.3.2 Effective Hamiltonian approach

In the considered example, the operator H' takes the form:
H'= 7talb E.s al s b +7t a2b E.s a2 s b (A.3.12)

where
' " ^ ' " [ ^ ( o , ->b) AE(b-*a,)]
K*2b=-4h2bPJ 1 + l- 1 (A.3.13)
a2bya2b[AE(a2^b) AE(bâ2)\
When applying the operator H' on the zero-order order wave functions eq
A.3.2 we substitute the saisb and s^Sb spin-operators making use of the Dirac
permutation operator, which replaces SjSj by (1/2)(P;J-1/2), where Py
interchanges spins of electrons occupying orbitals i and j . Here we omit the
constant term -(1/4) (of relevance for diagonal matrix elements only):
H ' ¥ £ = H'.j=(ata-2b+ -aâ+2b+) =
-l( 1
- J V JtalbE.Palb +7ta2bE.Pa2b).-Wfll+a2-&+ -araJA*)-
^[*, ! i I.( f l , + a ! -!> t )-^,£Ka 2 t A-) + f j 2 t i . ( a I t a 2 T ) - ^ I . ( a - i ! 2 t i t ) ]
(A.3.14)
(*£ \H\V&) = -J3XMJZ + V s ^ J f (A.3.15)
This is identical with eq A.3.11 obtained using perturbation theory.
References
1. R.Valiente, O.S.Wenger and H.U.Gudel (2000) Chem.Phys.Lett. 320,639.
2. R.Valiente, O.S.Wenger and H.U.Gudel (200l)Phys.Rev.B 63,165102.
3. P.Gerner, O.S.Wenger, R.Valiente and H.U.Gudel (2001) Inorg.Chem. 40,
4534.
4. CReinhard, R.Valiente and H.U.Gudel (2002) J.Phys.Chem.B 106,
10051.
5. Y.Tanabe, LMoriya and S.Sugano, (1965) Phys.Rev.Lett. 15,1023.
6. J.Ferguson, HJ. Guggenheim and Y.Tanabe (1966) J.Phys.Soc.Japan 21,
692.
7. K.-I.Gondaira and Y. Tanabe (1966) LPhys.Soc. Japan 21, 1527.
8. N.Fuchikami and Y.Tanabe (1978) J.Phys.SocJapan 45, 1559.
9. N.Fuchikami and Y. Tanabe (1979) J.Phys.Soc.Japan 47,505.
10. H.Shigi and Y.Tanabe (1982) J.Phys.SocJapan 51, 1415.
11. J.Ferguson, H.J.Guggenheim and Y.Tanabe (1965) J.Appl.Phys. 36,
1046.
12. J.Ferguson, HJ.Guggenheim and Y.Tanabe (1965) Phys.Rev.Lett. 14,
737.
13. J.Ferguson, H.J.Guggenheim and Y.Tanabe (1966) J.Chem.Phys. 45,
1134.
14. J.Ferguson, HJ.Guggenheim and Y.Tanabe (1967) Phys.Rev. 161,207.
15. J.Ferguson, H.U.Gudel and E.R.Krausz, Mol.Phys. 30,1139.
16. H.U.Gudel (1975) Chem.Phys.Lett. 36, 328.
17. H.U.Gudel (1984) Comments Inorg.Chem. 3,189.
18. H.Riesen and H.U.Gudel (1987) MoLPhys. 60, 1221.
19. M.Tachiki (1968) J.Phys.SocJapan 25, 686.
20. P.W.Anderson (1959) Phys.Rev. 115, 2

21.0.Kahn, Molecular Magnetism (VCH Publishers, Weinheim, 1993)
22. T.Oguchi (1965) J.Phys.Soc Japan 20, 2236.
23. J.Torrance, Jr. and M.Tinkham (1968) J.appl. Phys. 39, 822.
24. W.Urland (1976) Chem.Phys. 14, 393
25. R.W.Schwartz (1977) Inorg.Chem. 16, 1694.
26J.S.Griffith, The Theory of Transition-Metal Ions (Cambridge University
press, Cambridge, 1971)
27.M.Atanasov, C.A.Daul and C.Rauzy (2003) Chem.Phys.Lett. 367, 737.
28.M.Atanasov, C.A.Daul and C.Rauzy (2003) Struct.and Bonding, 106, 97.
29. We made use of the Amsterdam Density Functional (ADF) program;
E.LBaerends, D.E.Ellis and P.Ros (1973) Chem.Phys. 2,42.; P.M.Boerrigter,
G.te Velde and E.J.Baerends (1988) IntJ.Qunatum Chem., 33, 87; G.te Velde
and EJ.Baerends (1992) ComputPhys. 99, 84.
30.S.H.Vosko, L.Wilk and M.Nusair (1980) CanJ.Phys. 58, 1200.
31J.P.Perdew, J.A.Chevary, S.H.Vosko, KAJackson, M.R.Pederson,
D.J.Singh and CFiolhais (1992) Phys.Rev.B, 46,6671.
32. M.Wolfsberg and L.Hehnholz (1952) J.Chem.Phys. 20, 827.
33. P.Alemany and R.Hoffmann (1993) J.Am.Chem.Soc. 115, 8290.
34. J.des Cloizeaux (1960) Nucl.Phys. 20, 321.
35. M.A.Aebersold, H.U.Gudel, A.Hauser, A.Furrer, H.Blank and R.Kahn
(1993)Phys.Rev.B,48, 12723.
36.V.S.Mironov, L.F.Chibotaru and A.Ceulemans (2003) Phys.Rev. B, 67,
14424.
Chapter 6: Is a Dihydrogen Bond a Unique Phenomenon?
Slawomir J. Grabowskiab and Jerzy Leszczynskib
a Department of Crystallography and Crystal Chemistry, University of Lodz,

90-236-todz, ul.Pomorska 149/153, Poland
b Computational Center for Molecular Structure and Interactions,

Department of Chemistry, Jackson State University, Jackson,
MS 39217, USA
Abstract
The dihydrogen bonds (DHBs) are considered. The examples of such
interactions existing in crystal structures are presented. These interactions are
compared with conventional hydrogen bonds. The comparison shows that
DHBs are often similar in nature to O-H...O and other typical H-bonds. One
can observe for DHBs the elongation of the proton donating bond due to
complexation and the correlations between geometrical, energetic and
topological parameters. There are different types of DHBs, such as those with
7t-electron delocalization, similar to resonance assisted hydrogen bonds
(RAHBs). The insight into the nature of DHBs shows that in principle the
interaction energy decomposition scheme for them is similar to that one for
conventional H-bonds showing that the electrostatic energy term is the most
important attractive term. The case of o-bonds as the proton acceptors is also
considered. The relationships derived from the bond valence model are in
agreement with the experimental neutron diffraction results as well as with ab
initio and DFT calculations. The spin-spin coupling constants for dihydrogen
bonded systems are analyzed.
195
196 S. J. Grabowski and J. Leszczynski
1. Introduction
For a long time the nature of hydrogen bonding has been the subject of many
investigations.1'2'3 This is due to the importance of H-bond interactions in
chemical, physical and bio-chemical processes, especially life processes.3'4'5
Among the most pronounced examples are catalysis processes, proton transfer
reactions, enzymatic reactions, cell membrane transport processes and many
others which take place in the gas phase, in liquids and in solids. In addition,
this is an important interaction for phase transitions and the driving force in
crystal engineering.6 The criteria for the existence of hydrogen bonds are not
strictly defined because this interaction is understood as a very broad
phenomenon.4'7 Early works on hydrogen bonds define them as interactions
between X-H donating bonds and acceptor centers Y; hence, the H-bond is
usually designated as X-H...Y where X and Y should be electronegative
atoms such as O, N, Cl, etc.2 Hence early works also consider H-bonds as
mainly electrostatic interactions where there is excess positive charge on the
H-atom and excess negative charge on the X and Y atoms. In principle, there
was no problem to detect hydrogen bonds if the geometries of the complexes
or interatomic contacts were known. The situation was changed drastically
when it was pointed out that the C-H bond may act as a proton donor. The C-
H...O and C-H...N bonds were analyzed by Suttor in the crystal organic
structures.8 The subject of controversy in the 60s and 70s was the issue of
whether or not such interactions fulfil the criteria for the existence of
hydrogen bonding. However it was commonly accepted that there are such H-
bonds after the appearance of the work of Taylor and Kennard.9 The authors
have performed refined statistical analyses justifying the existence of C-H... Y
H-bonds in crystals. Afterwards also the other kinds of H-bonds were
analysed and those with C-atom acceptors (X-H...C) or 7t-electron acceptors
(X-H...71).4 Even the C-H...C hydrogen bonds were detected, for example in
one of the polymorphic forms of acetylene crystals (with the Acam space
group).10
Accorging to Jeffrey,5 H-bonds may be classified in the following way:
those for which H-bond energies are about 15-40 kcal/mol are known as
strong ones; the range of 4-15 kcal/mol is for moderate H-bonds and 1-4
kcal/mol for weak ones. The H-bonds mentioned here with C-donors or/and
C-acceptors usually belong to the last group. However there are exceptions;
for example the C-H...C interactions of moderate strength were analyzed
theoretically, and the binding energy for H3N+-CH2'...HCCH at the MP2/6-
311++G(3d,3p) level of theory (BSSE included) was calculated to be 8.16
kcal/mol.11 One can mention the other kinds of unconventional H-bonds
Is a Dihydrogen Bond a Unique Phenomenon? 197
which are recently the subject of investigations: for example, blue-shifting H-

bonds for which the blue shift of the stretching mode for the donating bond is
observed.12 For conventional H-bonds the stretching vibration of the X-H
proton donating bond changes significantly upon complexation with the
acceptor molecule. The stretching mode can be shifted to the red of the
spectrum by hundreds of cm"1 due to the X-H bond elongation, and the band is
intensified several fold. The red shift may be treated as a measure of the
hydrogen bonding strength since it correlates with the other typical
descriptors such as H-bond energy as well as the proton - acceptor (H...Y)
distance.13 Hobza and Havlas have studied intensively the reverse situation for
which the shortening of the X-H bond is observed due to complexation. It is
accompanied by a shift to the blue of the stretching mode and most often by a
decrease in the intensity of the band. Such interactions with mainly C-H
bonds as donors were named as "improper" or "inverse" hydrogen bonds, but
currently the name blue-shifting H-bonds is commonly accepted.12
In the middle of the 90s, the X-H...H-Y bond (another kind of H-bond
interaction) was reported and named as a dihydrogen bond since for such a
system there is the typical conventional X-H donor, and another hydrogen
atom with excess negative charge which acts as a proton acceptor and is
attached to the heavier non-hydrogen atom Y.14 Recently X-H....H-Y
interactions have been the subject of an increasing number of studies, both
experimental as well as theoretical.15 There are few reviews of dihydrogen
bonds; however, they mainly consider these interactions from an experimental
point of view.15'1647
The aim of this chapter is to briefly summarise the theoretical studies of
different kinds of dihydrogen bonds (DHBs) and the nature of such
interactions. It seems that deeper insight into the details of DHBs will be
useful for the analysis of a number of chemical and biochemical reactions.18 It
was pointed out that DHBs play an important role in many biochemical
processes, similar to the involvement of typical conventional X-H...Y
bonds.18
2. First examples of dihydrogen bonds

The first examples of the existence of dihydrogen bonds (DHBs) were
described in detail by Crabtree and co-workers.16 Two such examples which
prove that M-H bonds (M-metal) may act as proton acceptors were reported in
1990.19'20 The results of neutron diffraction measurements of the structure of
[Ir-(PMe)3(H)(OH)]+ indicate an H...H distance of 2.4 A between the OH
proton and the Ir-H hydride, and the authors have claimed that this probably
represents a weak dipole-dipole interaction.19 The neutron diffraction crystal
structure of [Fe(H)2(H2)(PEt2-Ph)4] displays the molecular hydrogen acting as
a ligand and closely attached to the Fe-H group. This Fe-H...H2 interaction
was termed as a "cis-effect,"20 but Morris and co-workers claimed that this
may be considered to be intramolecular hydrogen bonding.21
In the middle of the 90s two groups, Morris and co-workers 21'22 and
Crabtree's group,23'24'25'26 reported several examples of the metal organic
crystal structures containing DHBs. For the structure of
[IrH2(Me2CO)2(PPh3)2]+ the syn-orientation of the iminol OH was found
which allows the OH proton to closely approach one of the iridium hydrides.
This accounts for coupling between the OH and IrH protons (JHH = 2-4 Hz).23
Spectroscopic studies confirm direct H...H interactions since the v(OH) IR
band of the iminol group is in the range 3310-3409 cm"1 which corresponds to
lower energy than expected for free iminol OH. Similarly, the ! H NMR
spectrum confirms the attractive H.. .H intramolecular interction. The cases of
intermolecular DHBs were also studied in the middle of the 90s; for example
the neutron diffraction solid state studies of the ReH5(PPh3)3 • indole
complex14 reveal N-H...H-Re bifurcated intermolecular interactions. Fig. 1
presents the geometrical details of this type of H-bond.
/H\1.734(8)
l 6 8 3 X 1 1 8.i)\ 1 6 3 - 1 ( 6 )
RB /H ^N
\ / 1.007(6)
\97.2(3)/
1.683(6)\ / 2212 ( 9 )
H
Fig. 1. The geometrical parameters of Re-H...H-N interactions obtained from neutron
diffraction measurements, distances in A, angles in degrees.
One can see that one of the H...H contacts is significantly shorter (1.73
A) than the corresponding sum of van der Waals radii, and the N-H.. .H angle
for this interaction is equal to 163°, close to linearity as for the conventional
intermolecular H-bonds. Fig. 2 shows the packing of molecules for another
crystal structure ([ReH5(PPh3)2(imidazole)]) for which intermolecular
bifurcated DHBs were detected.27
Spectacular evidence for the existence of DHBs was presented for
boraneamine (H3BNH3) by Crabtree and co-workers.16 A comparison of
melting points for H3BNH3 (104°C) and isoelectronic species C2H6 (-181°C)
show that stronger interactions, not typical van der Waals interactions, should
occur for boraneamine. H3BNH3 is a polar compound which may explain such
a situation; but for non H-bonding CH3F species isoelectronic with ethane, the
melting temperature amounts to -141 °C. The lack of lone electron pairs for
H3BNH3 suggests the existence of DHBs. It is worth mentioning that the
crystal structure of H3BNH3 was determined by the neutron diffraction
method, and the existence of H...H interactions was confirmed.28 Fig. 3
displays the closest contacts between three neighbouring boraneamines. As
one can see there are intermolecular bifurcated dihydrogen bonds for this
crystal structure. The two H...H contacts are symmetrically equivalent; the
H.. .H distance is equal to 2.02(3) A.
H2A
Fig. 2. The neighbouring molecules of [ReH5(PPh3)2(imidazole)] in the X-ray crystal

structure. Reprinted with permission from ref 27. © 1995 Royal Society of Chemistry.
In the initial studies of DHBs the complexes for which X-H...H-M

interactions exist were considered. Hence it was pointed out that DHBs occur
200 5. J. Grabowski and J. Leszczynski
if there is the typical X-H proton donor, such as O-H or N-H, where typically
there is excess positive charge on the hydrogen atom. H-M is an accepting
group (M designates transition metal or boron), and the H-atom is negatively
charged. However, further studies have shown that other kinds of dihydrogen
bonds may exist, such as C-H...H-M (M-transition metal) 29 and even C-
H...H-C.30 The term "dihydrogen bond" was proposed for X-H...H-Y
interactions in 1995 by Crabtree and co-workers.31 The authors of the article
stated that "a referee points out that these unconventional hydrogen bonds
require a new name, for which we suggest the term 'dihydrogen bond.' " This
term is currently commonly accepted. Alkorta and co-workers7 claim that the
"protic" or "hydride" nomenclature may be applied for hydrogen bonds.
There are protic X-H^.-.Y HBs, hydric X-H^.-.Y HBs (designated by
Alkorta and co-workers as inverse hydrogen bonds) and protic-hydric X-H1"8
...'^-Y DHBs. The authors pointed out that protic HBs cover more than
99.9% of the studies of H-bonding, but the situation will change in the future
since unconventional hydrogen bonds, in spite of being weak, play an
important role in some cases in many chemical and biological processes.
m
% I
Fig. 3. The closest N-H.. .H-B contacts for the neutron diffraction crystal structure of
boraneamine. Reprinted with permission from ref 28. © 1999 American Chemical
Society.
3. First theoretical studies of DHBs

The first theoretical calculations of dihydrogen-bonded systems were
performed by Liu and Hoffmann.32 The authors considered the structure of
[Ir{H(Ti1-SC5H4NH)(PCy3)}2]+ [BF4]" (where Cy = cyclohexyl) synthesized
and experimentally investigated by Morris and co-workers.22 To simplify
calculations Liu and Hoffman neglected the counter ion [BF4]' as not
influencing the region with dihydrogen bonds. They also replaced the PCy3
group with PH3 and slightly changed the geometry to finally obtain the
[Ir{H(Ti1-SC5H4NH)(PH3)}2]+ ion with C2v symmetry. The authors have
applied the extended Huckel (EH) method in their study.33 The geometrical
parameters for the calculations have been taken from crystallographic data
since the EH method is not suitable for geometry optimizations. The
considered model ion has two intramolecular symmetry equaivalent N-H.. .H-
Ir dihydrogen bonds. The H...H distances are equal to 1.75 A, less than the
corresponding sum of van der Waals radii. The calculated Mulliken net
atomic charges amount to +0.22 and -0.27 for the H(N) and H(Ir) atoms,
respectively, showing that an attractive electrostatic interaction should appear.
The authors claimed that a positive Mulliken overlap population (OP) of
0.016 also indicates stabilization of the H.. .H interaction.
Liu and Hoffman32 have also considered two model complexes (FH.. .HLi
and FH...HMn(CO)5) using the RHF method. The first structure after the
optimization (the 6-3 lG(d) basis set was used) has an H.. .H distance of 1.658
A with Cs symmetry and nearly linear F-H...H and Li-H...H angles. The
binding energy for this complex without BSSE calculations is equal to 9.3
kcal/mol. An approximation of the electrostatic contribution using Mulliken
atomic charges allowed the authors to obtain the electrostatic term of 6.7
kcal/mol. Hence they claimed that this is mainly an electrostatic interaction as
observed for the other typical H-bonds.
For the second FH.. .HMn(CO)5 complex calculated at the RHF/3-21G(d)
level of theory, a minimum energy structure was found for the H.. .H distance
of 1.683 A and the binding energy of 6.6 kcal/mol. The FH and H-Mn bonds
were elongated due to complexation. Taking into account the features of these
two model complexes discussed here, Liu and Hoffman claimed that they are
similar in nature to the other H-bonds.
Almost at the same time as the appearance of the work of Liu and
Hoffman, a theoretical work of Crabtree and co-workers31 was published. In
this study the H3BNH3 dimer was investigated. The authors studied the cyclic
dimer of C2 symmetry with equivalent B-H.. .H-N bonds. The PCI-80/B3LYP
method was applied, and the following geometrical parameters were found for
dihydrogen bonds: the H.. .H contact is equal to 1.82 A; the N-H.. .H angle is
158.7°; and B-H..H is 98.8°.
It is worth mentioning that in the scope of the PCI-80/B3LYP method the
geometries were obtained after optimization with the B3LYP functional34 and
the standard double X, plus polarization basis sets. The energies for the
optimized structures were predicted with the use of the PCI-80 parametrized
method35 which gives greatly improved results compared to unparametrized
methods.
The binding energy calculated for the H3BNH3 dimer amounts to 12.1
kcal/mol which corresponds to 6.1 kcal/mol for each dihydrogen bond. The
Mulliken net atomic charges are equal to +0.27e and -0.09e for the H(N) and
H(B) atoms, respectively. The authors claim that the existence of opposite
charges at the hydrogen atoms of the H...H contact indicates that this
interaction may be classified as hydrogen bonding. The same conclusion may
be derived from the H...H contact energy which corresponds to the range of
the typical H-bond energies (3-8 kcal/mol). The angles of the B-H...H-N
system are in agreement with those found from the Cambridge Structural
Database (CSD).36 Crabtree and co-workers searched CSD for B-H...H-N
contacts and found 26 such species for which the H.. .H distance is less than
2.2 A. This search filter was applied since the authors expected that such a
value (less than the corresponding sum of van der Waals radii at 2.4 A) is
evidence of the attractive interaction. It is of course only a rough assumption
since the crystal packing may be constrained more than the sum of van der
Waals radii interatomic contacts. It was found that the N-H...H(B) angles
vary (an average of 149°, a range of 117-171°, o=17°); they are larger than the
B-H.. .H(N) angles (an average of 120°, a range of 90-171°, o=26°). Hence the
results for the H3BNH3 dimer are in line with the tendencies found for the
crystal structure complexes. The authors claim that the strong bending of the
B-H...H(N) angle is evidence that o-bonds are acceptors of protons, not H-
atoms with a negative net charge. In further studies37 they explain that the
hydrogen bond concept has been extended in recent years from the typical X-
H.ione pair interaction to X-H...71 electron interactions, and the last case
involves the interaction between the proton donating X-H bond and o-bond
acceptor, the X-H...G interaction commonly known as the dihydrogen bond.
Such a concept was also suggested by Cramer and Gladfelter.38
Theoretical calculations up to the MP2/6-31G(d,p) level with the BSSE
correction (counterpoise method of Boys and Bernardi39) were carried out on
a series of A-H...H-X dihydrogen bonds (A= B, Li, Be; X= N, C).40 It was
found that the considered complexes fulfilled the commonly accepted criteria
for the existence of hydrogen bonds. The systems studied were BH4...HCN,
BH4...CH4, LiH...NH4+, LiH...HCN, LiH...HCCH,BeH2... NH,+, BeH2...HCN

and CH4... NHt*. It is worth mentioning that Alkorta and co-workers were the
first to apply the Bader theory41 to analyse dihydrogen bonded systems.
Systematic studies of simple dihydrogen-bonded complexes have been
performed by Remko42 who considered electron-rich hydrogen atoms as
proton acceptors (for LiH and BeH2 molecules) and as proton donors with
positive H-atoms: HF, HCN, and HCCH. For these complexes full geometry
optimization was performed at the MP2(full)/6-31G(d) level. Such obtained
geometries were used to carry out the Gl and G2 energetic calculations.
Generally, the obtained results show that the dihydrogen bonds are
energetically similar to the typical, conventional H-bonds. The H...H
distances are in the range of 1.575 A (for LiH...HF) to 2.271 A (for H-Be-
H...HF).
Orlova and Scheiner have studied43 intermolecular H...H interactions
between HF, H2O and H3O+ donating species and metalloorganic compounds
such as Mo(H)(CO)2(L)2(L')2 (where cis-ligand L=PH3 and NH3, while L'
trans-ligand =NO,C1,H) and W(H)(CO)2(NO)(PH3)2 using the HF and DFT
methods. The results show that the energy of the H...H contact for these
complexes surpasses 10 kcal/mol. The same authors investigated the specific
role of the H...H bonded complexes in reactions of transition metal hydrides
with the proton donors. The ruthenium and rhenium compounds were
considered, and the calculations were performed at the DFT level using
nonlocal correction. The B3PW91 and LANL2DZ basis sets were used.44
Rozas and co-workers have investigated the problem of the influence of
field effects on dihydrogen bonded systems.18 They found, using the HF/6-
31G(d,p) level of theory, that a coordinated H2 molecule within the
Li+...H2...F" system may react yielding the dihydrogen bonded system Li-
H...H-F. And a dihydrogen bonded complex may yield two molecules
coordinated to the H2 molecule as in the case of the following reaction: H3lSr-
H...H-BH3" = NH3...H2...BH3.
The other early calculations of dihydrogen-bonded complexes include
particularly small systems where acceptor molecules involve main group
elements, eg., LiH, BeH2, BH3 and A1H3.3M5'46'47
4. Geometrical and energetic features of DHBs

The equilibrium geometry of molecules may be treated as a source of
chemical information. Especially, geometrical parameters describe the nature
of intra- and intermolecular interactions.3'4 Some of them are sensitive to
environmental effect; terminal bonds are usually elongated due to the H-bond
formation. For example this was observed for O-H...O=C systems.48 For the
neutron diffraction results taken from the Cambridge Structural Database,36'49
it was revealed that not only the proton donating bonds but also the accepting
C=O bonds are elongated.48 The statistical approach has showed that for
shorter H...O contacts the elongation of O-H and C=O bonds is greater. The
proton donating bond elongation is usually considered as strong evidence of
the existence of hydrogen bonding3'4 and it was many times detected for
different types of hydrogen bonds. There are also good correlations between
the H...Y distance and the X-H proton donating bond length for X-H...Y H-
bonds.3'5 Such correlations were observed for O-H...O systems,50'51 for N-
H...O, N-H...N52 and even for species with C-H...O.53 The only criterion that
should be fulfilled for the investigated samples is that they must consist of
related compounds; in other words the samples have to be homogeneous.54
Fig. 4 shows the relationship between the H...O distance and the O-H bond
length for O-H...O bonds taken from accurate crystal structure neutron
diffraction results.55 The open circles relate to the O-H...O systems, while the
full ones correspond to deuterated crystals (O-D...O systems). The solid line
represents the relationship between H...0 and O-H obtained from the Bond
Valence Model (BV model - see Appendix). It is worth mentioning that Fig. 4
shows the correlation obtained for the sample of O-H...O systems taken from
different crystal structures, and in spite of the heterogenity of the sample the
relationship is fulfilled.
Similar relationships for O-H...O species were obtained from the
results of ab initio and DFT calculations.55'56 The presented dependencies are
well fulfilled for typical H-bonds. For example the elongation of the proton
donating bond is not a criterion for the existence of such interactions for blue-
shifting H-bonds for which the shortening of the X-H bond is observed due to
the H-bond formation.12
Similarities between dihydrogen bonds and conventional O-H.. .O bonds were
investigated,57 and a relationship between the elongation of the proton
donating bond and the H...H distance was found for samples of simple
complexes (Fig. 5). The following samples were taken into account. The first
one consists of XH4..HF complexes (where X designates the elements of the
fourth group of the periodic table of elements: C, Si, Ge, Sn and hydrogen
fluoride was chosen as relatively strong proton donor). The CrH2...HF
complex is also included in the first sample to show the proton accepting
abilities of transition metal hydrate since the first crystal structures for which
DHBs were observed represent metal organic compounds of transition metals.
This sample is designated by circles in Fig. 5, and the related complexes were
calculated within the B3LYP/Lanl2DZ level of theory. For two remaining

samples there is the same proton donor molecule (HF) and the following
acceptors: CH4, S1H4, BeH2, MgH2, LiH and NaH. For one of the samples the
optimization of complexes was performed at the MP2/6-31G(d,p) level of
theory (triangles in Fig. 5); for another sample optimization was performed at
the MP2/6-311++G(d,p) level (squares). All calculations were carried out
with the use of the Gaussian98 suite of programs.58 One can see good
correlation between the H...H distance and the proton donating bond length,
similarly as for typical H-bonds. The greatest elongation of the H-F bond and
the shortest H...H contact for the first sample is revealed for the CrH2...HF
complex; for the remaining samples the greatest elongations occur for
LiH...HF and NaH...HF. Fig. 5 also displays the relationship between the
H...H distance and the elongation of the H-F bond obtained from the Bond
Valence model (continuous line - see Appendix). The results of a simple BV
model are in good agreement with the MP2 and DFT calculations. The
presented analysis of DHBs was based on the samples for which there is the
same proton donor (hydrogen fluoride molecule) and different acceptors.
1.26 -j ,
o i
°# |
1.02- °^L i
0.94-I 1 1 1 i
1.1 1.4 1.7 2 2.3
H..0 distance
Fig. 4. The relationship between the H...O distance and the O-H bond length (both
parameters in A). Reprinted with permission from ref 55. © 2000 Elsevier Science
B.V.
2.8 I — — — ~
2.4 -1
i \
1.6- * ^ ^
1.2 J • 1 • • 1 ~
0 0.01 0.02 0.03 0.04 0.05
elongation of H-F
Fig. 5. The dependence between the elongation of the H-F bond and the H...H
distance (both values in A). Reprinted with permission from ref 57. © 1999 Elsevier
Science B.V.
A similar analysis based on MP2 results of DHBs was performed for the
sample where there is the same proton acceptor (the BeH2 molecule), and
there are different donors such as HF, HC1, HBr, HCN, HCCH, HNNN, HBF2
and HOF.59 For such a sample, there is no correlation either between the
elongation of the Be-H donating bond and the H.. .H distance or between this
elongation and the H-bond energy. Practically the Be-H bond length is
constant; there is only one exception: the BeH2.. .HOF complex for which the
H-bond is strongest within the sample (4.8 kcal/mol) and for which a slight
elongation of Be-H is observed due to the H-bond formation (of about 0.03
A).
The following geometrical criteria for the existence of hydrogen bonds are
commonly applied and accepted by crystallographers:9
the distance between the proton and the acceptor atom is
shorter than the sum of their van der Waals radii;
- the conformation of the X-H... Y system should be close to

linearity, at least for intermolecular H-bonds because for
intramolecular H-bonds the strain effects may force greater
deviations from linearity; commonly the systems for which
the X-H.. .Y angles are less than 90° are not considered as H-
bonds;
- sometimes the next geometrical criterion is included which
requires that the proton donating bond is elongated due to the
H-bond formation.3'53
However these criteria are controversial since the hydrogen bonding
interaction is mainly electrostatic, and there is the electrostatic attraction far
beyond the van der Waals cut-off.4'51 It is also difficult to estimate accurate
positions of hydrogen atoms from X-ray crystal structure measurements.
Hence the neutron diffraction results should be applied for a geometrical
analysis of H-bonds in crystals.3 However the number of neutron diffraction
crystal structures collected in CSD is less than 0.5% (about 1000 crystal
structures)60 of the whole database (the remaining are X-ray crystal
structures), and sometimes for the statistical analysis of a required class of
compounds there is not a sufficient number of neutron diffraction results. For
the weaker C-H...Y bonds the elongation of the proton donating bond is not
detectable or is negligible.60 Additionally, as was mentioned before, there are
blue-shifting H-bonds for which the third criterion cannot be applied.
Recently few crystal structures were analyzed where there are repulsive
interactions existing due to crystal packing and fulfilling the H-bond
geometrical criteria.61
In spite of the reservations given above these criteria may be applied for
DHBs shown in Fig. 5. These ab initio results accurately reveal the H-atom
positions, and elongation of HF bonds is detectable. For some cases the H.. .H
distance is much smaller than the sum of van der Waals radii. For example for
the NaH...HF complex the H...H distance is equal to 1.371 A, and the H-F
bond length amounts to 0.958 A (the MP2/6-311++G(d,p) level), while for
HF not involved in H-bond interactions the bond length is equal to 0.917 A.
The analyzed systems have features typical of other conventional H-bonds, a
meaningful difference between the energy of the complex on one hand and
the energies of isolated monomers on the other hand. The binding energy is
often treated as the H-bond energy of intermolecular connections. It amounts
to -13.8 and -12.6 kcal/mol for the NaH...HF and LiR.HF complexes (the
MP2-6-311++G(d,p) level). One can see that selected DHBs are stronger than
the typical conventional H-bond of the linear-trans water dimer (about -5
kcal/mol). And for all DHB systems presented here there are the opposite net
atomic charges on connected H-atoms, positive on H(F) and negative on the

second H-atom. The only exception is the CH4...HF dimer for which both
charges are positive and which is not energetically stable; the binding energy
is positive (repulsive) at the MP2/6-311++G(d,p) level (0.39 kcal/mol). All X-
H.. .H systems included in Fig. 5 are linear or are close to linearity.
The relationships mentioned above were confirmed by the use of high
level ab initio calculations performed at the MP4(SDQ)/6-311++G(d,p),
OCISD/6-311++G(d,p) and QCISD(T)/6-311++G(d,p) levels of theory.62 For
example, the H-bond energy of the NaH...HF complex calculated at the latter
level of theory amounts to -12.66 kcal/mol (correction for BSSE included),
and the corresponding H.. .H distance is equal to 1.426 A. In this high level ab
initio study the Bader theory41 was also applied to analyse DHBs (see next
sections).
5. Application of the "atoms in molecules" theory to study

DHBs
The electron density, p(r), defines a structure as well as ground-state
properties of many-electron systems.63 The topological "atoms in molecules"
(AIM) theory of Bader41 allows quantitative structural information from
electron density to be obtained. The AIM theory is often applied to locate and
characterize critical points. The following relation, which requires that the
gradient of p(r) vanish, stands for the definition of the critical points (CPs):
Vp(rJ = 0 (1)
CPs are classified according to the number of negative eigenvalues of the
matrix of partial second derivatives of p with respect to {x,y,z} (Hessian
matrix).
h = _a^n ( r ) P.q=x,y,z (2)
There are four types of stable critical points: maxima in p(r) correspond to
attractors usually attributed to nuclei, and minima correspond to cage, bond or
ring critical points. The characteristics of bond critical points (BCPs) are often
used in studies of inter-atomic interactions. For example, the negative value of
Laplacian for BCP means that there is an increase in electron density in the
inter-atomic region; such a situation corresponds to the covalent bond. The
positive Laplacian value shows that there is the depletion of the electron
density, and this corresponds to the interaction of closed-shell systems: ionic
interaction, hydrogen bonding, and the van der Waals interaction.64
The characteristics of the bond critical points may be useful to estimate
the H-bond strength. ' ' The correlations between different topological
parameters and H-bond energy were found for various samples of complexes.
It is worth mentioning that the AIM theory provides new measures of H-bond
strength. From a geometrical point of view the proton.. ..acceptor distance is a
rough descriptor of H-bond strength, although it does not always correlate
with H-bond energy.3 The elongation of the proton donating bond may also
reveal the H-bond strength, and for stronger H-bonds the X-H...Y system
should be linear. There are three geometrical parameters often used by
crystallographers; however, they are not accurate indicators of H-bond
strength. The characteristics of BCPs such as the electron density at H...Y (Y
designates acceptor) BCP (PH...Y), its Laplacian (V2pH...vX the electron density
at the proton donating bond length (PXH) and its Laplacian (V2pXH) are often
applied to characterize H-bond interactions. All of these descriptors reveal the
H-bond strength and correlate with the H-bond energy if the homogeneous
sample of complexes is taken into account.54 One can observe that for typical
H-bonds there is the elongation of the proton donating bond due to
complexation; the pXH value decreases in comparison with the free donating
molecule; V2pXH increases. For stronger H-bonds H...Y contacts are shorter,
and hence the PH...Y and V2pH...y values are greater.68
Similar observations were revealed for dihydrogen bonds. Fig. 6 shows
the relief map of the electron density for the Li-H...H-F system. One can see
the differences between the electron densities of hydrogen atoms. For both of
them the densities are much lower than those of the F and Li atoms. However,
the hydrogen atom connected with Li is separated from Li (left side of the
figure), while the second hydrogen atom is strongly connected with the
fluorine atom (right side of the figure).
Fig. 6. The relief map of the electron density of the Li-H...H-F complex; the large
"mounds" correspond to the Li (left) and F (right) attractors, small ones to hydrogen
atoms.
The first rough correlation between H-bond energy and pH...y for DHBs
was found by Alkorta and co-workers (Y in such a case denotes an H-atom
acceptor). Table 1 presents the topological parameters obtained in
subsequent studies from MP2/6-311++G(d,p) wave functions for
homogeneous samples of dihydrogen bonded complexes.57'62 For all DHB
complexes hydrogen fluoride represents the proton donating molecule. Hence
other systems with an HF donor, with conventional H-bonds and calculated at
the same level of theory, are included in Table 1 for comparison. The trans-
linear conformation of the water dimer is also included as the most often
applied reference system in studies of hydrogen bonding.69
Table 1. The topological parameters (in au) and H-bond energies (EHB, corrected for
BSSE and in kcal/mol) for dihydrogen-bonded systems and for other conventional H-
bonds at the MP2/6-311++G(d,p) level of theory. For all cases (except for the trans-
linear dimer of water) the hydrogen fluoride molecule is a donor.
Acceptor pXH V 2 p XH PH...Y V 2 p H ... Y EHB

LiH a 0.323 -2.323 0.042 0.057 -12.62
NaHa 0.314 -2.205 0.046 0.048 -13.81
BeH2a 0.361 -2.763 0.016 0.049 3!94
MgH 2a 0.348 -2.624 0.026 0.060 ^02
CH/ 0.370 -2.830 0.002 0.005 +0.39
SiH," 0.367 -2.810 0.009 0.028 ^085
CH 2 O a a 0.359 -2.787 0.022 0.107 -5.43
H2Oaa 0.347 -2.652 0.037 0.142 ^54
NH 3 a a 0.325 -2.365 0.050 0.120 -11.18
F&aa 0.174 -0.349 0.174 -0.349 -39.87
H 2 O" & * aa 0.356 -2.512 0.023 0.091 A45
&(F...H...F)" system
&& liner-trans water dimer
arefs 57,59
aaref68
The results presented in Table 1 show the tendencies discussed above that the
topological parameters correlate with H-bond energy. For the DHB
complexes the linear correlation coefficient between P H . H and H-bond
energy is equal to 0.995, while it decreases to 0.991 for all systems presented
in the table. This indicates that P H . Y topological parameter may be applied
not only for homogeneous samples but also for heterogeneous samples. This
was also revealed for the other H-bonded systems. Fig. 7 shows the details
of the aforementioned relationships.
5-
0*b . , , ,
c 0 * n 0.05 0.1 0.15 0.2
-5- <*o
|-15- *
•S - 2 0 -
| -25
X
-30
-35-
-40- O
-45-
electron density at H...Y BCP
Fig. 7. The relationship between PH...Y and the H-bond energy for the complexes
collected in Table 1: foil circles correspond to DHBs, while empty circles correspond
to the other conventional H-bonds.
There is also good linear correlation for DHBs between PXH and EHB (R,
the linear correlation coefficient, amounts to 0.997). For the relation between
V2pXH and EHB, R is equal to 0.991; only the correlation between V2pH...y and
EHB is poor (R = 0.687).
Similar relationships between parameters describing DHB systems were
observed in other studies. For example, for the neutral and charged
dihydrogen-bonded complexes with the LiH molecule as a proton acceptor,
good correlations between H-bond energy and H...H distance and H-bond
energy vs. PH...H were found.71 The authors also observed that there is a slight
shortening of the Li-H proton accepting bond for neutral complexes, while for
positive charged complexes there is the elongation of the Li-H bond.
A detailed analysis of the (BH3NH3)2 dimer based on the AIM theory was
performed by Popelier.72 The optimization of the geometry of the dimer was
carried out at the HF and MP2 levels of theory using the 6-31G(d,p) basis set.
In this case a topological analysis based on appropriate wave functions was
performed using the program MORPHY97.73 The positions of the critical
points were found using the eigenvector method.74 There are three B-H...H-N
contacts for the optimized (BH3NH3)2 dimer; because of its symmetry (Cs
symmetry) two of them are equivalent due to the presence of a mirror plane.
The geometrical parameters of these DHB connections for both levels of
theory are presented in Table 2.
One can see that the N-H...H-B connections of the (BH3NH3)2 dimer
roughly fulfil the geometrical criteria for the existence of H-bonds; H...H
distances are less than the corresponding sum of van der Waals radii; N-H and
B-H bonds are elongated due to complexation; N-H...H and B-H...H angles
are greater than 90°. Also the latter angles are smaller than N-H...H angles
which is in accordance with the results known for crystal structures where
DHBs exist.
Table 2. The geometrical parameters of B-H...H-N connections (in A and degrees) for
the (BH3NH3)2 dimer (resultsfromref. 72).
Parameter First N-H...H-B bond I Second N-H...H-B bond &

HF/6-31G(d,p) I MP2/6-31G(d,p) HF/6-31G(d,p) I MP2/6-3 lG(d,p)
H...H L914 1/726 2.324 2.149
N-H...H 1631 168?7 128/7 BOO
B-H...H 144.6 139.2 113.8 112.6
AN-H && 0.004 0.008 0.001 0.002
AB-H** 0.011 0.011 0.007 0.007
& there are two such symmetrically equivalent systems

44 the elongation of the bond due to complexation
For the (BH3NH3)2 complex Popelier has also investigated the topological
criteria for the existence of hydrogen bonds72 which were proposed earlier.75
There are eight topological criteria based on the AIM theory:
- there is the bond critical point and the bond path for the H...Y
interaction;
- the electron density at this BCP ranges from 0.002 to 0.035 au;
- the Laplacian of this electron density ranges from 0.024 to 0.139 au;
- there is the mutual penetration of the hydrogen and the acceptor atoms;
- an increase in net charge of the hydrogen atom is observed;
- energetic destabilization of the hydrogen atom;
- decrease of dipolar polarization of the hydrogen atom;
- decrease of the hydrogen atomic volume.

For the H.. .H contacts of (BH3NH3)2 there are BCPs and bond paths. The
HF/6-31G(d,p) level pH...H values are equal to 0.0126 and 0.0067 au (0.0190
and 0.0096 calculated at the MP2/6-31G(d,p) level). The corresponding
values of Laplacians (V2pH...H) amount to 0.0353 and 0.0247 au (the HF
level) and 0.0464 and 0.0347 au (the MP2 level). One can see that the electron
densities and Laplacians are within the ranges proposed by Koch and
Popelier. Usually for studies of hydrogen bonds with the use of the AIM
theory only the first three criteria are investigated. However, for the
(BH3NH3)2 dimer all eight criteria are fulfilled.
6.rc-electron delocalization-assisted intramolecular

dihydrogen bonds
First experimental works on DHBs have included the metalloorganic crystal
structures for which mainly the intramolecular interactions of this type exist.
Similarly the first theoretical calculations have been often performed for such
systems. For example in the work of Liu and Hoffman32 the intramolecular N-
H...H-Ir dihydrogen bonds have been analyzed. In this section the cases of
intramolecular DHBs for which the effect of TT-electron delocalization occurs
are analyzed. The influence of this effect on the strength and other
characteristics of dihydrogen bonds are presented.
The resonance-assisted hydrogen bonds (RAHBs) are very well known as
those whose properties are strongly influenced by 7t-electron
delocalization.51'76 For example, malonaldehyde (Fig. 8) is an example of a
simple system in which intramolecular resonance-assisted hydrogen bond
exists.77
For malonaldehyde and its simple derivatives (different R-substituents -
Fig. 8) there is equalization of dl and d4 CO bonds on one hand and
equalization of the d2 and d3 bonds on the other hand. For such
intramolecular RAHBs the Q-parameter was introduced as a descriptor of the
delocalization.51
Q = dl-d4 + d3-d2 (3)

R1 d2 C
I R2
Fig. 8. Malonaldehyde (R1=R2=R3=H) is an example of a system with
intramolecular H-bonds assisted by Ji-electron delocalization. Reprinted with
permissionfromref 77. © 2001 Elsevier Science B.V.
The greater is the Jt-electron delocalization, the smaller is the Q-

parameter, and also the H-bonds are stronger. Such dependencies were
observed for experimental RAHBs in organic crystal structures51 as well as
were confirmed theoretically for malonaldehyde and its fluoro- and chloro-
derivatives77 at the MP2/6-311++G(d,p) level of theory. For the border case of
delocalization, when there is full equalization of conjugated CO and CC
bonds, the proton is exactly in the middle of the 0....0 distance. H-bonds in
such a case are very strong, and the H-bond energy amounts to about 30-40
kcal/mol.51 Such systems are often known as low barrier hydrogen bonds.78 It
was also pointed out that the electron withdrawing Rl-substituents and the
electron donating R3-substituents (Fig. 8) should cause greater 7C-electron
delocalization, greater bond length equalization and stronger RAHBs.51 This
conclusion was confirmed by MP2/6-311++G(d,p) calculations.77
The influence of 7t-electron delocalization on intramolecular dihydrogen
bonds was also studied for (lZ)-2-borylethen-l-ol (Fig. 9)79 and its derivative
containing simple R-substituents causing an increase in H-bond energy
(R1=C1, R3=Na).
For these molecules ab initio calculations were performed up to the
MP2/6-311++G(3d,3p) level of theory. The H-bond energies were roughly
estimated as the difference between the closed conformations (such as the one
presented in Fig. 8) and the open conformations. The open conformation may
be obtained from the closed form after the rotation of the O-H bond by 180°
around the dl (C-O) bond. The molecular geometries are optimized before
and after rotation. Such an approach in the evaluation of intramolecular H-
bond energy is often applied.77'80 However, it is a rough estimation of H-bond
energy since other effects come into play such as the unfavourable approach
of the lone pairs of the two oxygen atoms after the changing of
conformations.80
T<-> TH
d1l J B<M
R1 ^ [ C <* R3
• R2
Fig. 9. (lZ)-2-borylethen-l-ol (R1=R2=R3=H) as an example of intramolecular
dihydrogen bonding. Reprinted with permission from ref 79. © 2000 Elsevier Science
B.V.
It is worth mentioning that in the case of typical intramolecular RAHBs

such as malonaldehyde the 7t-electron delocalization occurs not only for the
closed conformation but also for the open one.77 However, for the closed
conformation the delocalization and bond equalization should be greater. For
(lZ)-2-borylethen-l-ol an elongation of the C=C (d2) and B-H (d4) bonds and
a shortening of the C-O (dl) and C-B (d3) bonds can be observed due to the
transformation from the open to the closed conformation. Table 3 partly
supports this, showing the meaningful influence of the formation of an
intramolecular dihydrogen bond on the geometrical parameters of the
molecules. The only exception is for the d2 bond length of the A molecule.
Table 3. Geometrical parameters (in A) of (lZ)-2-borylethen-l-ol (A) and its

derivative, R1=C1, R2=Na (B), at the MP2/6-311++G(3d,3p) level of theory (results
taken from ref. 79).
Molecule I dl I d2 I d3 I d4
A closed conformation 1.340 1.346 1.516 1.200
A open conformation 1.352 1.355 1.528 1.189
B closed conformation 1.333 1.367 1.534 1.225
B open conformation 1.343 1.356 1.547 1.202
The results collected in Table 3 also indicate that the delocalization is

greater for the B molecule where there are Cl and Na substituents which make
this effect stronger. The H...H intramolecular distances amount to 1.933 and
1.647 A for the A and B closed conformations, respectively (the MP2/6-
311++G(3d,3p) level), while the H-bond energies roughly estimated in the
manner described previously are equal to 4.6 and 6.5 kcal/mol. Similarly, the
values of the topological parameters show that the intramolecular DHBs
presented here are of medium strength. The PH.H values are equal to 0.016
and 0.027 au for A and B, respectively, and the V2pH...H values are equal to
0.049 and 0.059 au for A and B (the MP2/6-311++G(2d,2p) level).
A more extended sample of intramolecular dihydrogen bonds was also
considered to investigate if the dependencies known for typical H-bonds are
fulfilled for DHBs.81 Systems similar to (lZ)-2-borylethen-l-ol were
considered, but the -BH2 group was replaced by -BH3\ All possible fluoro
derivatives were included in this study. The calculations have been performed
at the MP2/6-311++G(d,p) level of theory. Again, the well known approach
of roughly estimating H-bond strength as a difference in energy between the
closed and open conformations was applied. However, in this case this
difference does not correlate with the other parameters describing H-bond
strength, not with H...H distance or with the PH...H values. It seems that for
this sample with negative charged species, other effects than H...H
interactions strongly disturb the H-bond interaction. Hence another approach
was applied81 to estimate the H-bond strength. It was found that the local
potential energy density at the proton...acceptor contact, V(rcp), for H-bonds
correlates with the H-bond energy.67 For H-bonds taken from crystal
structures and for their electron densities, the local energy densities, V(rCp),
were calculated, and the following relationship was proposed: E H B = V*
V(rcp).67 The H-bond energies were calculated from the local energy densities
for the intramolecular DHBs mentioned above, and good correlations between
such energy and the other parameters were found. The linear correlation
coefficient for the relationship between the calculated H-bond energy and
H.. .H distance is equal to 0.995 (Fig. 10 shows this dependence).
_ HLHdstsnoe
•on 1 1 1 1
16 1.65 1.7 1.75 1.8
>. -6 <T R=Q995

2>
g-7J
Fig. 10. The relationship between the H-bond energy (kcal/mol) calculated from the
local potential energy density and H...H distance (in A) for intramolecular DHBs.
Reprinted with permissionfromref 81. © 2003 Elsevier Science B.V.
Similarly, the linear correlation coefficient between this energy and the
topological parameter pH...H is equal to 0.988. It is worth mentioning that the
linear correlation coefficient for the relationship between the energy
calculated from the local energy density, and the difference in energy between
the open and closed conformations amounts to 0.226. For this sample of
ionized intramolecular DHBs, the properties of the ring critical point were
also investigated.81 The ring critical point (RCP) is a point of minimum
electron density within the ring surface and a maximum on the ring line.64
For O-H...O intramolecular H-bonds it was found previously that electron
density at RCP correlates with H-bond strength.82 Similar relationships were
found for O-H.. .N and N-H.. .O intramolecular H-bonds83 and also for double
intermolecular H-bonds for which the closed rings of the covalent bonds and
the H...Y contacts are formed.84 For the sample of BH3'-CH-CH-OH and its
fluoro derivatives a good polynomial correlation (the polynomial of the
second degree) was found with a correlation coefficient of 0.973. Fig. 11

displays this relationship which confirms that, for intramolecular dihydrogen
bonds, the topological properties of RCP may be used as descriptors of H-
bond strength as for the other conventional intramolecular hydrogen bonds.
These molecules with intramolecular DHBs were chosen as model
systems because of their simplicity and the fact that they are convenient
species for higher level ab initio calculations. However, similar but usually
much more complex molecules exist in the liquid or solid phase. For example,
for crystals of 2,2,2-tris((Boranato(diphenyl)phosphonio)methyl)ethanol
(YIBTAE refcode in the Cambridge Crystal Structural Database36)
intramolecular B-H'5... +5H-O hydrogen bonds were detected (the BH3" group
as a proton acceptor) with an H...H distance of 2.230 A. Single point ab initio
calculations of the moiety of 2,2,2-tris((Boranato(diphenyl)phosphonio)-
methyl)ethanol (geometry taken from the crystal structure) have been
performed.81 Because of the complexity and the large size of the moiety the
HF/3-21G* level of theory was chosen. The AIM calculations yielded the
value of the electron density of the H"8../5!! bond critical point (pH...H =
0.0060 au) and the value of the electron density at the ring critical point (pRCp
= 0.0055 au). The corresponding Laplacian values, V2pH..H and V2pRCp,
amount to 0.0340 and 0.0282 au, respectively. These results suggest that for
this crystal structure intramolecular dihydrogen bonds may exist since the
topological criteria are fulfilled.75
0.030-I
0.028 - Jt
r 0.026- * /
1=1 0.024- j(*
0.022- ^ ^ ^
0.020 I +^—r 1 1 1 1
0.015 0.016 0.017 0.018 0.019 0.020
PRCP
Fig. 11. The relationship between electron density at H.. .H BCP and electron density
at RCP (values in au). Reprinted with permission from ref 81. © 2003 Elsevier
Science B.V.
7. The nature of DHBs - energy decomposition analysis

It is possible to obtain more detailed insight into the nature of interactions by
decomposition of the total interaction energy into different components. There
are different approaches to perform such a decomposition, but the scheme
which is the most often applied is that of Morokuma and co-workers.85 Briefly
speaking, the electrostatic (ES) term represents the Coulombic interaction
between the charge distributions of the two subunits of the dimer, and the
exchange energy (EX) corresponds roughly to steric repulsion between the
two charge clouds. When the monomers are allowed to influence the charge
clouds between themselves, it is possible to extract polarization (POL) energy
connected with the shifts of electron density occurring within monomers and
charge transfer (CT) energy connected with the density shifts from one
monomer to the other. The effects of electron correlation are included in the
CORR term.
Such a decomposition scheme was applied recently to different complexes
containing dihydrogen bonds.86 Among them the (H3BNH3)2 and (H2BNH2)2
complexes were considered where for the former four equivalent H...H
contacts exist, and for the latter there are two equivalent H....H interactions.
Their binding energies calculated at the MP2/aug-cc-pVDZ level of theory
(BSSE included) are equal to 14.1 and 2.1 kcal/mol, respectively. For the
(H3BNH3)2 complex for each equivalent B(sp3)-H...H-N(sp3) dihydrogen
bond, the H-bond energy amounts to 3.5 kcal/mol, and for each B(sp2)-H.. .H-
N(sp2) interaction the energy is 1.1 kcal/mol for (H2BNH2)2. This is indicative
of the influence of hybridization on the strength of the dihydrogen bond. For
these and other complexes the Morokuma partitioning scheme was applied.85
The results are presented in Table 4 and compared with the other complexes
connected through conventional and unconventional H-bonds.87'88'89
The results presented in Table 4 show that for conventional H-bonds in
trans-linear dimers of water, the electrostatic attractive term is the most
important; it overweighs the repulsive exchange term. The polarization and
charge-transfer terms are much smaller than the electrostatic term, similar to
the electron correlation energy term. For the unconventional C-H.. .0 bond of
the CH4...OH2 complex, the electrostatic energy is compensated by the
exchange energy, but due to the small attractive remaining terms, the total
interaction for this complex is slightly attractive. The substitution of a
methane-donating molecule by electronegative fluorine atoms causes an
increase in the donating characteristics of the C-H bond. Hence, the
percentage contributions of the energy terms for the F3CH.. .OH2 complex are
similar to those of the water dimer. For two complexes with 7t-electrons of the
acetylene molecule as proton acceptors (FH...C2H2 and HCCH...C2H2) the

electrostatic term is compensated by the exchange term, and the remaining
attractive contributions cause stabilization of the complexes. The polarization
and charge-transfer contributions are greater in the case of the FH...C2H2
complex. There is a similar situation for HOH...C6H5OH; however, in such a
case the correlation term is significant.
Table 4. Morokuma's partitioning of Interaction Energy (kcal/mol).
Complex/ref I ES I EX I POL I CT I CORRd

(H 2 O) 2 /87 a ^6 42 4J7 4$ ^03
CH 4 ...OH 2 /87 a ~^0A O4 50 ~4A Ôl
F 3 CH...OH 2 /87 a I7l 4~\ Ô7 To ^03
FH...C2H2/88b 4A 63 T5 22 Ô4
HCCH...C2H2/88b ^2 Ti ^03 Ô5 Ô4
HOH...C 6 H 5 OH/89 a ^ 32 ^09 ^07 2~A
(H3BNH3)2 /86 c ^9 4\4 3?7 T6 4l
(H2BNH2)2 /86 c A3 12 ^5 Ô8 4 l
"6-31+G** basis set

b 6-311++G** basis set
c aug-cc-pVDZ basis set
d CORR=AE(MP2)- AE(HF)
What is a unique feature of the two dihydrogen bonded complexes

presented in Table 4? For (H2BNH2)2 the binding energy and H-bond
interactions are weaker than for the (H3BNH3)2 complex. For the first
complex the exchange energy overweighs the electrostatic term, but finally
the total interaction is attractive due to the remaining attractive terms, the
correlation term being the most important. For the second complex,
(H3BNH3)2, the electrostatic term is slightly greater than the exchange term,
and the remaining terms contribute significantly to the total interaction
energy. Hence one can see that for stronger dihydrogen bonds such as in the
(H3BNH3)2 complex the decomposition components are slightly similar to
typical H-bonds; the electrostatic contribution is slightly greater than the
exchange contribution, and the other attractive terms play an important role.
Another decomposition scheme was applied recently for the sample of

complexes with dihydrogen bonds.90 The interaction energy for the IJH...H2,
LiH.. .CH4, LiH.. .C2H6 and LiH.. .C2H2 complexes has been partitioned using
the IMPPT scheme.91'92'93'94 The IMPPT energy terms are designated by e(ij)
where i and j are the orders of the intermolecular interaction operator and the
intramolecular correction operator, respectively. All e(i j ) terms are calculated
in the basis of full complex to avoid the basis set superposition error. The total
interaction energy at the all-electron MP2 level is decomposed into a Hartree-
Fock (SCF) contribution and a correlation term:
AEMP2 = AE SCF +AE CORR (4)
The Hartree-Fock term is further decomposed:
AESCF = A E SCF d e f + A E HL (5)
Where AESCFdef is the deformation energy, and AEHL is a Heitler-London

term. The first term describes an effect due to the relaxation of orbitals in the
Coulomb field of the partner with the restriction imposed by the Pauli
principle. The second-order approximation to AESCFdef considers this term as
the induction response term e ^ O ) ^ . The Heitler-London contribution is
further decomposed into electrostatic and exchange energies:
AEHL = e(l,0) els + eHLexch (6)
The correlation energy term is also decomposed into:
AECORR = AE(2)exch + 8(2,0)^+ e(l,2)els,r (7)

Where AE(2)exCh is the second-order exchange correlation correction;
e(2,0)disP is the dispersion correlation correction; and E(l,2)eisr is the second-
order electrostatic correlation correction.
The complexes mentioned above have been studied at the MP2 level of
theory using the aug-cc-pVDZ basis set. For smaller complexes the aug-cc-
pVTZ basis set was also used. Table 5 shows the energy contributions for the
complexes. Only results of MP2/aug-cc-pVDZ are given in order to compare
all complexes. The results obtained within the same level and the same
scheme of energy decomposition for the water dimer are also presented for
comparison. Emphasis is put only on the selected terms.
Table 5. Decomposition of the MP2 interaction energy (in kcal/mol) of the LiH...H2,
L1H...CH4, LiH...C2H2, LiH...C2H6 and H2O...H2O complexes; the aug-cc-pVDZ
basis set was used (results taken from ref. 90).
Term I LiH...H2 I LiH...CH4 I LiH...C2H6 I LiH...C2H2 I H2O...H2O

8(1,0)* I H Tl A2 ^2 TS
~^~ O T3 22 6M 7^
AE^drf : O6 5?7 5H 5?7 53
lOOW 5?7 TI T5 53 42
£(2,1)^ 51 51 51 51 51
e(l,2)elv 51 00 00 O6 51
A brief analysis of the results presented in Table 5 indicates that only for
one of dihydrogen bonded complexes (LiH.. .C2H2) the first order electrostatic
term (e(l,0) e i s ) is greater than the exchange term, similarly as for the water
dimer. For the remaining dihydrogen bonded complexes the dispersion term
(£(2,0YjjsP) is comparable with the electrostatic term. The authors of this
paper explain that, for LiH...C 2 H 2 , the dihydrogen bond may be classified
as hydrogen bond interaction; however, the remaining species should be
characterized as van der Waals complexes.
8. Spin-spin coupling constants across dihydrogen bonds

Vibrational frequencies and NMR shielding constants have been applied for a
long time to describe hydrogen bonds and other intermolecular interactions.
However, in recent years data concerning spin-spin coupling constants has
become a new tool in the analysis of interactions, especially hydrogen bonds.
A review of ab initio and DFT scalar coupling constants has appeared
recently, and the usefulness of this new method is herein explained in detail.
The scalar coupling can be defined as the mixed second derivative of the
energy of the molecule regarding the magnetic moments of two nuclei. It can
be divided into four components: Fermi contact (FC), paramagnetic spin-orbit
(PSO), diamagnetic spin-orbit (DSO) and spin-dipolar (SD). The nJ
designation is usually used for nuclear spin-spin coupling constants through n
bonds, where bond means not only the typical covalent bond but also the
inter- or intramolecular interaction such as, for example, the H-bond.

Parentheses or subscripts may be also used to designate the coupled nuclei.
Couplings through hydrogen bonds are usually designated as ***]. Hence the
hydrogen-hydrogen coupling between atoms of two different water molecules
connected through the O-H...O hydrogen bond is designated as 2hJ if one of
the hydrogens belongs to the proton donating O-H bond and as 4hJ if none of
the coupled hydrogen atoms belongs to the O-H.. .O system.
A number of studies dealing with calculations of J-values have been
performed in recent years. For example, the J-values for the water monomer
and for the water dimer were calculated and compared with experimental
results. The scalar coupling constants were analyzed for X-H.. .0 H-bonds
in the following complexes: CH2O...H2O, C2H2...H2O, CH3OH...H2O and
(HCOOH)2.97 The calculations were performed at the MCSCF level, and the
results show that lhJoH intermolecular coupling correlates with the H-bond
strength and is dominated by the FC term. Similarly the 2hJxy couplings
through the X-H...Y hydrogen bonds are dominated by the FC component.
Complexes with N-H...O=C and N-H...N=C systems were also studied, and
exponential decay of 2hJxY upon increasing X... Y distance was observed.
Different kinds of other H-bonds and van der Waals complexes95 were
analyzed in terms of1*! values. However in the latter case only the theoretical
results are available, and the experimental results are not available for
verification. Among ^J values related to this review those concerning
dihydrogen bonds are the most interesting.
Recently ab initio calculations have been performed to determine one-
bond IhJHH and three-bond 3hJHH spin-spin couplings across X-H...H-M
dihydrogen bonds for complexes with 13C-'H, 15N-[H and 17O-!H proton-
donor groups and 7Li-!H and 23Na-!H proton-acceptor metal hydrides.
Coupling constants have been calculated using the equation-of-motion
coupled cluster singles and doubles method (EOM-CCSD) employing Cl-like
approximation.99>10"ll01'102The lhJm couplings containing the PSO, DSO, FC
and SD components mentioned above have been calculated for linear CTOV
dihydrogen bonded complexes of NCH...HLi, NCH...HNa, CNH...HLi,
CNH...HNa and LiNCH+...HLi and for nonlinear Q complexes of
HOH.. .HLi and HOH.. .HNa. The authors have observed that the SD term is
negligible (ranging from -0.04 to -0.20 Hz). The relationships between the
remaining terms and the H...H distance for each of the complexes were
investigated. The FC term decreases exponentially with an increase in the
distance of H.. .H. PSO and DSO depend linearly on the H.. .H intermolecular
distance but have opposite signs and tend to cancel each other. Hence the
relationship between the total coupling and H...H distance is practically the
same as between FC and H.. .H distance. And hence the FC value may be an
indication of the strength of the hydrogen bond. Similar dependencies were
observed between the 3hJHH spin-spin coupling and the C...M (M = Li, Na)
distance for each of the complexes. However, in the case of three-bond spin-
spin couplings, three terms, PSO, DSO and SD, are negligible and this is the
reason why almost the whole contribution to the total coupling comes from
the FC term.
An analysis of the intermolecular coupling constants have been performed
very recently for the following complexes: LiH...H2, LiH.-.CtL,, LiH...C2H6
and LiH...C2H2.90 The intermolecular lhJHH and 3hJXY (3hJuc) constants have
been plotted as functions of internuclear distance. The ^JRH values have been
calculated at the DFT (B3LYP) and CCSD levels of theory using the aug-cc-
pVTZ basis set. The obtained DFT and CCSD curves are similar, and the
following conclusions were derived. 3hJLjC decreases exponentially with an
increase in the Li...C distance for all four investigated complexes. For three
weaker complexes (LiH...H2, L1H...CH4 and LiH...C2H6) lhJHH is negative
and tends to approach zero when the H...H distance increases. A different
behaviour was noticed for the LiH...C2H2 complex. In this case lhJHH is
negative for larger distances, is positive for shorter H...H contacts and is
similar to the other typical covalent bonds. It is worth mentioning that the
authors have classified the LiH...C2H2 complex as dihydrogen-bonded, and
the three remaining complexes as van der Waals complexes since the
exchange energy terms overweight the electrostatic contributions (see
previous section).
9. Borders and limitations - summary

It is difficult to conclude that there is something specific for dihydrogen
bonds in comparison with the other conventional H-bonds. It is certain that at
least one feature is unique: in DHB complexes another hydrogen atom is the
proton acceptor. It was pointed out that such an atom should be negatively
charged as was revealed for some metal or boron hydrides. However even in
early studies of DHBs it was indicated that such interactions may be treated in
the another way, ' namely that a bonds are the proton acceptors following
the trend: conventional X-H....Y hydrogen bonds, H-bonds with Tt-electrons
as acceptor (X-H...Jt) and X-H...0 H-bonds. However Crabtree et al. also
claim that early examples of interactions termed as dihydrogen bonds were
studied for transition metal hydrides. For such cases one cannot be sure if the
M-H a bond is an acceptor since transition metals also possess nonbonding
d^-electrons that could act as alternative H-bond acceptors.
The strength of DHBs is usually similar as predicted for typical H-bonds.
DHBs are usually considered to be medium to strong H-bonds. The results of
the decomposition energy analysis do not indicate unique properties of DHBs.
For stronger DHBs such as for LiH...C2H2 the energy terms are similar to
those of the typical H-bonds. The most important is the electrostatic term
on
which is greater than the exchange energy term (Table 5). For other systems
such as LiH...H2, IJH...CH4 and LiH...C2H2, the exchange term overweighs
the electrostatic contribution. These complexes are weak, and the dispersion
term is important for them (Table 5). Such systems may be classified as van
der Waals complexes.
The clusters of ammonia with hydrogen molecules (NH4+(H2)n) could be
considered an important example of the continuously changing characteristic
of the DHB complex. These clusters have been studied where the number of
H2 ligands ranges from one to eight. Geometry optimizations have been
performed at the MP2/6-311G(d,p) level of theory, and the dissociation
energies were calculated within the MP2 theory using the single point
calculations in the extended aug-cc-pVDZ basis set. ' 5 Corrections for the
basis set superposition error according to the scheme of Boys and Bernardi
were applied. For the studied systems no symmetry constraints were
imposed during the geometry optimization. The energetically lowest structure
for the NH»+ + H2 complex is the vertex bond moiety with the H2 molecule
perpendicular to the line of the N-H bond (Fig. 12).
The binding energy for such a complex amounts to 2.56 kcal/mol, and the
molecular arrangement supports the idea of the proton acceptor abilities of o
bonds. It is worth mentioning that for the other arrangement of molecules for
N H / + H2 (for the face bonded molecules where the H2 molecule is parallel
to the HHH plane of NH/) the transition state is predicted, and such a
complex is higher in energy (by 0.98 kcal/mol) than the equilibrium complex.
It was indicated, using the interaction energy decomposition scheme, that for
NH4+(H2)n complexes the exchange energy usually overweighs the
electrostatic term, and the dispersion energy significantly contributes to the
total energy. The situation is similar to that observed for the LiH...H2,
IJH...CH4 and LiH...C2H2 complexes; however, for NH41" H2 the binding
energy is greater. Therefore, it seems reasonable for the latter case to be
considered the charge-assisted (N-H... o)+ interaction.
Fig. 12. The structure of the NH/ complex of the lowest energy (distances in A,
angles in degrees). Reprinted with permissionfromref 103. © 2001 Elsevier Science
B.V.
Very recently, a comprehensive study of C-H.. .H-C dihydrogen bonds

existing in crystal structures was published. This kind of interaction in
crystals is very important since for a large number of metal organic and
organic crystals the H.. .H contacts are common. Until now they were usually
treated as typical van der Waals contacts. Robertson et al.30 considered the
crystal structures of the tetraphenylborates for which DHBs exist as well as
the other H-bonds such as N-H.. .N, C-H.. .N, N-H.. .Ph and C-H.. .Ph. These
interactions have been characterized using the Bader theory after multipole
refinement of the structures. Different dependencies were studied; the authors
found that the difference between the Mulliken charges on H-atoms in the
uncomplexed X-H and H-Y components correlates well with some of the
parameters of the X-H...H-Y dihydrogen-bonded system. They also found
that the H...H intermolecular distance is within the range of 2.18 - 2.57 A;
the electron density at H...H BCP is within 0.05-0.08 au, and the binding
energy is about 1.5 kcal/mol. The range of the electron density values shows
that the systems analyzed are within the criterion of Koch and Popelier.75 The
authors also pointed out that there is a detectable border between the H2
connections and the H...H contacts. The border is connected with the acute
change of the geometrical and topological parameters. The correlation curves
have characteristic gaps indicating these borders. And there are no
discontinuities in the considered features between the dihydrogen bonds and
the van der Waals complexes.
Interestingly, early studies of the methane dimer show that the H.. .H
interaction is repulsive and that this system is not stable. However, the later,
more accurate post-HF calculations indicate that the methane dimer is
bound in all six examined configurations: H3CH...H2CH2 (Cs),
H2CH2...H2CH2 (D2h and D2d), HCH3...H3CH (D3h and D3d) and
H3CH...HCH3 (D3h). It is worth mentioning that for such weak H...H
interactions the analyzed structures are very sensitive to the level of the
calculations. For example, the minimum (no imaginary frequencies) D3<j
symmetry H3CH...HCH3 dimer not included in the analysis of Novoa et al.106
was obtained at the MP2/6-311++G(3d,3p) level of theory.107 The other lower
basis sets indicate that such a configuration, H3CH...HCH3 (D3d), is a
transition state. Fig. 13 shows a molecular graph of this D3d symmetry
methane dimer obtained within the AIM theory.
Fig. 13. Molecular graph of the H3CH...HCH3 (D3(i) methane dimer (the MP2/6-
311++G(3d,3p) level); big circles correspond to attractors attributed to nuclei, and
small circles correspond to bond critical points.
The electron density at H...H BCP for this dimer is equal to 0.0035 au
and is within the range of Koch and Popelier which suggests that this
interaction may be classified as an H-bond.
The examples presented here show that there are problems with
distinguishing between van der Waals interactions and DHBs. The only not-
so-sharp difference is in the nature of the interactions; for the former species
the dispersion term is the most important; for the latter it is the electrostatic
term that should overweight the exchange contribution.
10. Appendix - the Bond Valence Model

Previously, in 1947, for metal crystal structures, L. Pauling introduced the
concept of the bond number n (Pauling, 1947).
rn-r0 = Ar = -clogn (Al)

In the above equation, rn is the considered interatomic distance for which
the bond number is equal to n; c is the constant; and r0 is the reference
distance for which the bond number is usually equal to unity. The bond
number may be understood as the number of electron pairs attributed to the
considered pair of atoms, and it correlates with the strength of the interaction.
The number of applications for this idea exists such as for example the
analysis of chemical reactions and the construction of reaction paths (Burgi,
1975; Dunitz, 1979). This example concerns the proton transfer reactions.
This concept is also useful for studies of intra- and intermolecular interactions
(Grabowski, Krygowski, 2002).
This idea could be apply to studies of hydrogen bond interactions. For
example, for the O-H bond not involved in any intermolecular interactions,
the bond number should be equal to unity. When the O-H bond is involved in
the O-H...O interaction, the O-H bond is elongated, and n-value decreases
according to eq. (Al). But the decrease of n is compensated by the H...0
contact. The sum of bond numbers of O-H and H...0 should be equal to
unity. It is known as the bond number conservation rule (Dunitz, 1979).
A similar and more often applied idea of bond valence was developed by
Brown (Brown, 1978; Brown, 1992) within the consistent and simple model
known as the Bond Valence (BV) model. According to the BV model, the
bond valences s are usually related to bond distances in the following way:
siJ = expf(ro-riJ)/BJ (A2)

or
Sy = (ro/riJf (A3)
Where, sg designates that the valence bond concerns the pair of i-th and j -
th atoms; r0 is the reference (usually single) bond length.
B and N are constants which may be easily determined if, for the partial
bond, both values S;J and ry are known. For example, for the symmetrical
linear system of O-H.. . 0 with the hydrogen atom in the middle of the O.. .0
distance, the sy value of H.. .O is equal to 0.5, and the H.. .O distance is equal
to 1.22 A (Dunitz, 1979) or even less, 1.2 A (Gilli et al., 1994). Such an
attitude is often applied to evaluations of constant values. However, Brown
proposed to treat all constant values, B, N and r0, iteratively. This means that
all constants may be determined from the large sets of systems taken from the
results of crystal structures (Brown and Altermat, 1985) and by applying the
least squares or one of the other procedures.
One of the most important statements of the BV model is that the atomic
valence may be expressed in terms of the bond valences. This is the so-called
valence sum rule (VSR):
r.-zs, <A4>
The valence V; of the i-th atom is assumed to be shared between the bonds
that it forms. According to the Bond Valence Model the bond between the i-th
and j-th atoms may be the typical covalent bond or the intermolecular contact.
The atomic valences usually correspond to the oxidation states. For the
hydrogen atom within O-H...O system the valence sum rule takes the form:
exp[(r0OH - rOH)/BoHl + exp[(r0OH - rH...o)/BOHI =1 (A5)
This equation simply corresponds to the bond number conservation rule

mentioned above. In other words the valence sum rule is more general than
the bond number conservation rule.
If the constant values r0 and B are known it is possible to derive the
relationship between the rOH bond length and the H...0 distance rH...o
(Grabowski, 1988). This is a theoretical relationship based on the simple BV
model, and it agrees excellently with the experimental results (Gilli et al.,
1994). It is worth mentioning that the r0 and B constant values are attributed
to pairs of atoms; thus they are the same for O-H covalent bonds and the
H...0 contacts.
Similar relationships as the one expressed by eq. (A5) may be constructed

for any atom in the solid, liquid or gas phase. For example, for the proton
acceptor oxygen atom in C=O...H-O systems, the following valence sum rule
equation may be written:
exp[(roc'° - rc=o)/Bc=o] + exp[(r0OH - rK,.o)/Bo«] =2 (A6)
The relationship between the C=O bond length and H.. .O contact distance
obtained from eq. (A6) corresponds well to the experimental neutron
diffraction results (Grabowski, 1998).
The VSR for the +8H-atom within the F-lT8...'5!! dihydrogen bonded
system may be written in the following way:
VH = exp[(ro"-F - rH.F)/BHF] + exp[(roH"H - rH_jJ/BH..jJ (A7)

Where, rH.F and rH...H are the H-F bond length and the H...H distance within
F-H...H system; roHF ,BHF , r0H'H and BH...H are constants.
The dependencies mentioned above obtained from the VSR often
correspond to experimental results and to the relationships obtained from ab
initio or DFT calculations (see the main text of this chapter). They also show
the proper dependencies between the geometrical parameters and are
indicative of the changes in the geometrical parameters during chemical
reactions.
Brown, I.D. Chem. Soc. Rev. 1978,7,359.

Brown, I.D. Acta Cryst. 1992, B48,553.
Brown, I.D. Altermat, D. Acta Cryst. 1985, B41,244.
Burgi, H.-B. Angew. Chem. Internat. Edit. 1975,14,460.
Dunitz, J.D. X-ray Analysis and the Structure of Organic Molecules, Cornell
University Press, Ithaca, 1979.
Gilli, P., Bertolasi, V., Ferretti, V., Gilli, G. J. Am. Chem. Soc. 1994,116,909.
Grabowski, S.J. Croat. Chem. Acta 1988,61, 815.
Grabowski, S.J. Tetrahedron 1998,54,10153.
Grabowski, S.J., Krygowski, T.M. Molecular Geometry as a Source of Chemical
Information - Application of the Bond Valence-Bond Number Models, in
Advances in Quantitative Structure Property Relationships, Eds. B.Charton &
M.Charton, Vol. 3, pages 27-66, Elsevier Science B.V. 2002.
Pauling, L. J. Am. Chem. Soc. 1947, 69, 542.
Acknowledgement
Support for this research is provided by grant 505/675 2003 (University of
L6dz), grant No. 3 T09A 138 26 (State Committee for Scientific Research),
NSF CREST HRD-0318519, NSF EPSCoR 02-01-0067-08/MSU, and ONR

Grant N00034-03-1-0116.
11. References
1 Pauling, L., The nature of the chemical bond. Cornell University Press, Ithaca NY,
2 1939.
Pimentel, G.C., McClellan, A.L. The Hydrogen Bond. Freeman, San Francisco, CA,
1960.
3 Jeffrey, G.A., Saenger, W. Hydrogen Bonding in Biological Structures. Springer-
Verlag: Berlin, 1991.
4 Desiraju, G.R., Steiner, T. The weak hydrogen bond in structural chemistry and
biology. Oxford University Press Inc., New York, 1999.
5 Jeffrey, G.A. An Introduction to Hydrogen Bonding. Oxford University Press, New
York, 1997.
6 Desiraju, G.R. Crystal Engineering - The Design of Organic Solids. Elsevier,
Amsterdam, 1989.
7 Alkorta, I., Rozas, I., Elguero, J. Chem.Soc.Rev. 1998, 27, 163 (and references cited
therein).
8 Suttor, D. J. J. Chem. Soc. 1963, 1105.
* Taylor, R., Kennard, O. J. Am. Chem. Soc. 1982, 104, 5063.
Grabowski, S.J. Polish J. Chem. 1995, 69,223.
11 Platts, J. A., Howard, S.T., Wozniak, K. Chem.Commun. 1996, 63.
12 Hobza, P., Havlas, Z. Chem.Rev. 2000, 100,4253 (and references cited therein).
13 Scheiner, S. Hydrogen Bonding: A Theoretical Perspective. Oxford University
Press: New York, 1997.
14 Wessel, J., Lee, J. C , Peris, E., Yap, G. P. A., Fortin, J. B., Ricci, J. S., Sini, G.,
Albinati, A., Koetzle, T. F., Eisenstein, O., Rheingold, A. L., Crabtree, R. H.
Angew. Chem. Int. Ed. Engl. 1995,34,2507.
15 Custelcean, R., Jackson, J. E. Chem. Rev. 2001, 101,1963.
16 Crabtree, R. H., Siegbahn, P. E. M., Eisenstein, O., Rheingold, A. L., Koetzle, T. F.
Ace. Chem. Res. 1996,29, 348.
17 Epstein, L. M., Shubina, E. S. Coord. Chem. Rev. 2002, 231, 165 (and references
cited therein).
18 Rozas, I., Alkorta, I., Elguero, J. Chem.Phys.Lett. 1997, 275, 423 (and references
cited therein).
19 Stevens, R. C , Bau, R., Milstein, D., Blum, O., Koetzle, T. F. J.Chem.Soc, Dalton
2Q Trans. 1990, 1429.
Van der Sluys, L. S., Eckert, J., Eisenstein, O., Hall, J. H., Huffman, J. C , Jackson,
S. A., Koetzle, T. F., Kubas, G. J., Vergamini, P. J., Caulton, K. G. J. Am. Chem.
21 Soc. 1990,112,4831.
Park, S., Ramachandran, R., Lough, A. J., Morris, R. H. J. Chem. Soc. Chem.
Commun. 1994,2201.
22 Lough, A. J., Park, S., Ramachandran, R., Morris, R. H. J. Am. Chem. Soc. 1994,
,, 116,8356.
Lee, J. C , Jr., Rheingold, A. L., Miiller, B., Pregosin, P. S., Crabtree, R. H. J.
Chem. Soc. Chem. Commun. 1994,1021.
24Lee, Jr., J. C , Peris, E., Rheingold, A. L., Crabtree, R. H. J. Am. Chem. Soc. 1994,
116,11014.
" Peris, E., Lee, Jr., J. C , Crabtree, R. H. J. Chem. Soc. Chem. Commun. 1994,2573.
Peris, E., Lee, Jr., J. C , Rambo, J., Eisenstein, O., Crabtree, R. H. J. Am. Chem.
Soc. 1995,117,3485.
27 Peris, E., Wessel, J., Patel, B. P., Crabtree, R. H. J. Chem. Soc. Chem. Commun.
1995,2175.
28 Kloosrer, W. T., Koetzle, T. F., Siegbahn, P. E. M., Richardson, T. B., Crabtree, R.
H. J. Am. Chem. Soc. 1999,121, 6337.
Z* Richardson, T. B., Koetzle, T. F., Crabtree, R. H. Inorg. Chim. Ada 1996,250,69.
Robertson, K. N., Knop, O., Cameron, T. S. Can. J. Chem. 2003,81,727.
31 Richardson, T. B., de Gala, S., Crabtree, R. H., Siegbahn, P. E. M. J. Am. Chem.
Soc. 1995,117, 12875.
I Liu, Q., Hoffinan, R. J. Am. Chem. Soc; 1995, 117,10108.
Hoffinan, R. J. Chem. Phys. 1963,39,1397.
34 Becke, A. D. Phys. Rev. 1988, A38,3098.
35 Siegbahn, P. E. M., Blomberg, M. R. A., Svensson, M. Chem. Phys. Lett. 1994,
™ 223,35.
Allen, F. H., Kennard, O. Chem. Des. Autom. News 1993, 8,31.
37 Klooster, W. T., Koetzle, T. F., Siegbahn, P. E. M., Richardson, T. B., Crabtree, R.
38 H. J. Am. Chem. Soc. 1999,121, 6337.
Cramer, C. J., Gladfelter, W. L. Inorg. Chem. 1997,36, 5358.
39 Boys, S. F., Bernardi, F. Mol. Phys. 1970,19, 553.
40 Alkorta, I., Elguero, J., Foces-Foces, C. Chem. Commun., 1996,1633
41 Bader, R. F. W., (1990) Atoms in Molecules. A Quantum Theory. Oxford
University Press, New York.
f Remko, M. Molecular Physics, 1998,94,839.
Orlova, G., Scheiner, S. /. Phys. Chem. 1998, 102, 260.
*/ Orlova, G., Scheiner, S. J. Phys. Chem. 1998,102,4813.
* Kulkarni, S. A. J. Phys. Chem. A 1998,102, 7704.
*! Kulkami, S. A. J. Phys. Chem. A 1999,103,9330.
Kulkarni, S. A., Srivastava, A. K. /. Phys. Chem. A 1999,103, 2836.
48 Grabowski, S. J., Tetrahedron 1998, 54, 10153 (and references cited therein).
49 Allen, F. H., Davies, J. E., Galloy, J. E., Johnson, J. J., Kennard, O., Macrave, C.
F., Mitchel, E.M., Smith, J. M., Watson, D. G. J. Chem. Inf. Comput. Sci. 1991,
50
31, 187.
3 1 ' 1 8 7 '
51 Ichikawa, M. Acta Cryst. 1978, B34,2074.

52 Gilli, P., Bertolasi, V., Ferretti, V., Gilli, G. J. Am. Chem. Soc. 1994,116,909.
" Steiner, T. J. Phys. Chem. 1998, 102,7041.
54 Steiner, T. Chem. Soc. Perkin Trans. 2, 1996, 1315.
Grabowski, S. J. J. Phys. Org. Chem. 17,2004,18.
55 Grabowski, S. J. J. Mol. Struct. 2000, 552, 153.

Grabowski, S. J., Krygowski, T. M. Chem. Phys. Letters, 1999,305,247.
"Grabowski, S. J., Chem. Phys. Lett. 1999,312, 542.
Frisch, M. J., Trucks, G. W., Schlegel, H. B., Scuseria, G. E., Robb, M. A.,
Cheeseman, J. R., Zakrzewski, V. G., Montgomery, J. A., Stratmann, R. E.,
Burant, J. C , Dapprich, S., Millam, J. M., Daniels, A. D., Kudin, K. N., Strain,
M. C , Farkas, O., Tomasi, J., Barone, V., Cossi, M., Cammi, R., Mennucci, B.,
Pomelli, C , Adamo, C , Clifford, S., Ochterski, J., Petersson, G. A., Ayala, P. Y.,
Cui, Q., Morokuma, K., Malick, D. K., Rabuck, A. D., Raghavachari, K.,
Foresman, J. B., Cioslowski, J., Ortiz, J. V., Stefanov, B. B., Liu, G., Liashenko,
A., Piskorz, P., Komaromi, I., Gomperts, R., Martin, L. R., Fox, D. J., Keith, T.,
Al.-Laham, M. A., Peng, C. Y., Nanayakkara, A., Gonzalez, G., Challacombe,
M., Gill, P. M. W., Johnson, B., Chen, W., Wong, M. W., Andres, J. L.,
Gonzalez, C , Head-Gordon, M., Replogle, E. S., Pople, J. A. Gaussian 98;
Revision A.6; Gaussian, Inc.: Pittsburgh, PA, 1998.
5 Grabowski, S. J. J. Mol. Struct. 2000, 553,151.
60 Grabowski, S. J. "Simulations of Hydrogen Bonds in Crystals and Their
Comparison with Neutron Diffraction Results, in Neutrons and Numerical
Methods - N2M," Grenoble, France December 1998, AIP Conference
Proceedings 479, (eds.) M. R. Johnson, G. J. Kearley and H. G. Buttner,
American Institute of Physics 1999.
61 Novoa, J. J., Nobeli, I., Grepioni, F., Braga, D. NewJ. Chem. 2000,24, 5.
"Grabowski, S. J. J. Phys. Chem. A2000,104, 5551.
£ Hohenberg P., Kohn, W. Phys. Rev. 1964, B864, 136.
Popelier, P. Atoms in Molecules. An Introduction, Prentice Hall, Pearson Education
Limited 2000.
65 M6,0., Yanez, M. Elguero, J. /. Chem. Phys. 1992,97,6628.
^ Mo, O., Yanez, M. Elguero, J. J. Mol. Struct. (Theochem) 1994,314,73.
Espinosa, E., Molins, E., Lecomte, C. Chem. Phys. Lett. 1998,285,170.
* Grabowski, S. J. J. Phys. Chem. A, 2001, 105,10739.
Scheiner, S. Annu. Rev. Phys. Chem. 1994,45,23.
70Grabowski, S. J. J. Mol. Struct. 2002,615,239.
71 Alkorta, I., Elguero, J., M6, O., Yanez, M., del Bene, J. E. J. Phys. Chem. 2002,
„ 106,9325.
Popelier, P. L. A. J. Phys. Chem. A 1998,102, 1873.
73 MORPHY97, a program written by P. L. A. Popelier with a contribution from R.
G.A. Bone, UMIST, Manchester, U.K., EU, 1997.
74Popelier, P. L. A. Chem. Phys. Lett. 1994,228,160.
"Koch, U. Popelier, P. L. A. J. Phys. Chem. A 1995,99,9747.
Madsen, G. K. H., Iversen, B. B., Larsen, F. K., Kapon, M., Reisner, G. M.,
Herbstein, F. H. J. Am. Chem. Soc. 1998,120,10040.
77 Grabowski, S. J. J. Mol. Struct. 2001, 562,137.
78 Garcia-Viloca, M., Gelabert, R., Gonzalez-Lafont, A., Moreno, M., Lluch, J. M. J.
Phys. Chem. A 1997,101,8727 (and references cited therein).
79 Grabowski, S. J. Chem. Phys. Lett. 2000,327,203.
", Cuma, M., Scheiner, S., Kar, T. J. Mol. Struct. (Theochem), 1999,467, 137.
Wojtulewski, S., Grabowski, S. J. J. Mol. Struct. 2003,645,287.
82 Grabowski, S. J, Monatsh.Jur Chetnie, 2002,133,1373.
83 Rybarczyk-Pirek, A. J., Grabowski, S. J., Malecka, M., Nawrot-Modranka, J. J.
84 Phys. Chem. A 2002,106,11956.
Dubis, A., Grabowski, S. J, Romanowska, D. B., Misiaszek, T., Leszczynski, J. /.
g5 Phys. Chem. A 2002,106,10613.
Morokuma, K., Kitaura, K. in Molecular Interactions, vol. 1, W. J. Orville-Thomas
(ed.), Wiley, New York, 1980, page 21.
I Kar, T., Scheiner, S. J. Chem. Phys. 2003, 119,1473.
Gu, Y., Kar, S., Scheiner, S. J. Am. Chem. Soc. 1999, 121,9411.
* Scheiner, S., Grabowski, S. J. J. Mol. Struct. 2002,615,209.
f Scheiner, S., Kar, T., Pattanayak, J. J. Am. Chem. Soc. 2002,124, 13257.
Cybulski, H., Pecul, M., Sadlej, J. J. Chem. Phys. 2003,119, 5094.
I* Chalasinski, G., Sczeiniak, M. Mol. Phys. 1988,63,205.
Moszyhski, R., Rybak, S., Cybulski, S. M., Chalasinski, G. Chem. Phys. Lett. 1990,
„ 166,609.
Cybulski, S. M., Chaiasinski, G., Moszyhski, R. J. Chem. Phys. 1990,92,4357.
9 Chalasinski, G., Sczfiniak, M. Chem. Rev. 1994,94, 1723.
Alkorta, I., Elguero, J. Int. J. Mol. Sc. 2003,4, 64.
: ! Pecul, M., Sadlej, J. Chem. Phys. Lett. 1999,308,486.
9J Pecul, M., Leszczynski, J., Sadlej, J. J. Chem. Phys. 2000,112, 7930.
* Pecul, M., Leszczynski, J., Sadlej, J. J. Chem. Phys. 2000,104, 8105.
i Perera, S. A., Sekino, H., Bartlett, R. J. J. Chem. Phys. 1994,101,2186.
°, Perera, S.A., Nooijen, M., Bartlett, R. J. J. Chem. Phys. 1996,104,3290.
j°, Perera, S. A., Bartlett, R. J. J. Am. Chem. Soc. 1995, 117,8476.
~T Perera, S. A., Bartlett, R. J. J. Am. Chem. Soc. 1996, 118, 7849.
!°: Urban, J., Roszak, S., Leszczynski, J. Chem. Phys. Lett. 2001, 346,512.
r* Dunning, Jr., T. H. J. Chem. Phys. 1989, 90, 1007.
Kendall, R. A., Dunning, Jr., T. H., Harrison, R. J. J. Chem. Phys. 1992, 96, 6769.
^ Novoa, J. J., Whangbo, M.-H., Williams, J. M. J. Chem. Phys. 1991,94,4835.
Grabowski, S. J., Leszczynski, J. unpublished results.
INDEX
cr-bond acceptor 202, 225, 226

7r-electron delocalization 214-219
ab initio 85, 88, 89, 105

adhesion 61
aglycone bond 68
amylose 67, 70, 71
Anderson's theory 154-157
angle bending 105-107
angular overlap model 166
anisotropic exchange 154, 155, 157, 163, 165, 187
atomic force microscope (AFM) 47-51, 53-62, 64, 66-71
atomic valence 230
atoms in molecules (AIM) theory 208-214
ATPase 56, 65, 66
basis set superposition error 222, 226

Bethe lattice 7, 8, 11, 38
biased molecular dynamics 54, 62, 64, 71
bifurcated dihydrogen bonds 198, 199
binding energy 196, 201, 202, 207, 208, 220, 221, 226, 227
blue-shifting H-bonds 197, 204, 207
bond critical point (BCP) 208, 209, 213, 219, 228
bond number 229-231
bond number conservation rule 229, 230
bond valence 204, 205, 229, 230
bond valence model (BV model) 204, 205, 229-231
Cambridge structural database (CSD) 202, 204

carbon nanotube (CNT) 31-33, 35, 37
chair-boat transition 50, 67
charge transfer 157, 182, 190
charge transfer energy 220
CHARMM 67
conduction channel 3, 4, 19
conventional H-bonds 197, 199, 203, 204, 207, 210-212, 219, 220, 225
correlation energy term 220, 222
coupled electron-pair approximation (CEPA) 121-124, 126-132, 136, 138, 139,
148
237
238 Index
coupling constant 153, 156, 166, 170, 173, 183-185, 188-191

coupling integral 189
criteria for the existence of hydrogen bonds 196, 202, 206, 207, 213, 214, 219
crystal (ligand) field 158, 160
crystal engineering 196
density functional theory (DFT) 5, 6, 9, 13, 14, 27, 28, 53, 70, 169
dextran 67, 68
dihedral 109, 110, 112-115
dihydrogen bond 195-235
dipole moment 88, 114, 116, 179, 180, 192
Dzyaloshinski-Moriya 163
elastin 59
electric field 179, 180
electronic hopping 159
electronic transition 178
electrostatic energy term 201, 207, 220-223, 225, 226, 229
energy decomposition analysis 220-223
enzymatic catalysis 53
EPR-spectra 163, 165
exchange energy 220-223, 225, 226, 229
exchange interaction 154, 155, 157, 158, 163
exchange operator 153, 156, 180, 186
excited state 121, 125, 126, 133, 134, 136, 138-140, 142-144, 145, 148
exclusion principle violating (EPV) terms 123, 124, 129, 130, 132
ferromagnetic 154, 163, 164, 177, 178, 186

fibronectin 50, 56, 58
force probe molecular dynamics 60
force probe spectroscopy 49
free energy 53, 55, 59, 62, 63
frequency spectrum 92, 93, 101, 103, 110
fullerene 1, 24, 26-28, 30
Gaussian Embedded Cluster Method (GECM) 6, 25, 26, 31

GAUSSIAN98/03 1, 69
Go-like models 65
Green's function 7, 9, 10, 40, 41
g-tensor 163
H... H interactions 202, 203, 212

Index 239
H-bond energy 197, 206-212, 215-218, 220

H-bond strength 209, 217-219, 224
Heisenberg-Dirac-van Vleck Hamiltonian 157
heterogeneity of the sample 204
HF calculations 67
homogeneity of the sample 204, 209-211
hydrogen bonding 195-235
implicit solvent 59, 62

IMPPT energy terms 222
intramolecular dihydrogen bonds 198, 201, 214-219
intruders 121, 124, 125, 127, 131-133, 139, 143, 145, 148
Jarzynski 63, 64
Keldysh formalism 13
kinetic exchange 153, 156, 182, 185
Landauer's formalism, formula 2-5, 25, 37

Laplacian of the electron density 208, 209, 213, 214, 219
LDA 169
Lennard-Jones 105
Li2 121, 143
linear response theory (LRT) 125, 126, 133
local energy density 217, 218
localized magnetic orbitals 154, 156
low barrier hydrogen bonds 215
magnetic exchange 153, 166

magnetic field 163, 165
magnetic orbitals 154-156, 186
MBPT 124
mechanically controllable break junction (MCBJ) 2, 17
methane dimer 228
molecular bridge 2, 7, 22
molecular device 86, 88, 89, 91, 101, 103, 115
molecular dynamics 47, 52, 54, 56, 60, 69, 85, 88, 89, 92, 93, 114, 115
molecular junction 2, 6
MRCC 124-126
MRCEPA 124-126, 138-140
MR-CISD 123, 124, 128, 130
MscL channel 60
240 Index
Mulliken atomic charges 201, 202, 227
NAMD2 code 68
nanocell 85, 115
nanocontact 2, 4, 6, 10-12, 14-25
NDR 86, 89, 91
negative differential resistance 86, 89, 91
optomechanical 66
orbital momentum 168
P4 121, 139-143, 145, 146, 148

pectin 67
perturbation theory 156, 164, 182, 190, 193
polarization energy 220, 221
polysaccharides 50, 51, 66, 67
potential energy surface (PES) 121-125, 139-141, 143-148
potential exchange 156, 185
potential of mean force 55, 56, 59, 62
protein folding 51-53
proton acceptors 197, 202-204, 206, 212, 219, 221, 224-226, 231
proton donors 196, 197, 200, 202-205, 207, 209, 210, 224
pustulan 67-69
PW96 169
pyranose ring 67, 68
quantum molecular dynamics 70
radiation field 179

random expulsion molecular dynamics (REMD) 64, 65
rare earth 153, 186
reaction coordinate 55, 61, 62
reaction kinetics 53
resonance assisted hydrogen bonds 214-219
resonant bond 93, 98, 99, 101, 112-114
ring critical point (RCP) 208, 218, 219
scanning tunneling microscope (STM) 2, 17, 24, 25, 31

SCC-DFTB 70
Schrodinger equation 3, 4
size-extensivity 123-125, 127
size-intensivity 138
Index 241
spectrin 60
spin operator 159, 160, 168
spin-forbidden transition 179
spin-orbit coupling 153, 157, 160, 166, 167, 170, 172, 185, 187, 188
spin-orbit interaction 158
spin-spin coupling constants 223-225
S-S bridge 58
SS-MRCC 122, 125-128, 133, 134, 136, 146
streptavidin-biotin 54
targeted molecular dynamics (TMD) 54, 60, 66, 72

temperature 85, 89, 96, 98, 103, 105, 107, 110, 112, 113, 115, 116
the electron density 208-210, 213, 218-220, 227, 228
titin 50, 51, 56-59, 61, 65
topological parameters 209-211, 217, 218, 228
transition metal 153, 154, 181
unconventional H-bonds 196, 200, 220

up-conversion 153-155, 178, 183, 185
valence sum rule (VSR) 230, 231

van der Waals interaction 199, 209, 225-228
van der Waals radii 199, 201, 202, 206, 207, 213
van der Waals 93
vertical excitation energies 145
vibrational modes 87, 88, 92, 98, 114, 115
water dimer 207, 210, 211, 220, 222-224

wave function 210, 212
WHAM 56, 64
CONTENT INDEX
Volume 1
Relativistic Many-Body Calculations on Atoms and Molecules (Y. Ishikawa &
U. Kaldor)
Modern Developments in Hartree-Fock Theory: Fast Methods for Computing the
Coulomb Matrix (M. Challacombe, E. Schwegler & J. Almlof)
Local Shape Analysis of Macromolecular Electron Densities (P. G. Mezey)
Liquid-State Quantum Chemistry: Computational Applications of the Polarizable
Continuum Models (J.-L. Rivail & D. Rinaldi)
Elemental Boron Route to Stuffed Fullerenes (E. D. Jemmis & B. Kiran)
Interactions of DNA Bases and the Structure of DNA. A Nonempirical Ab Initio
Study with Inclusion of Electron Correlation (J. Sponer, P. Hobza &
J. Leszczynski)
Computational Approaches to the Design of Safer Drugs and Their Molecular
Properties (N. Bodor & M.-J. Huang)
Volume 2
The Electron Propagator Picture of Molecular Electronic Structure (J. V. Ortiz)
SAC-CI Method: Theoretical Aspects and Some Recent Topics (H. Nakatsuji)
Quantum Monte Carlo and Electronic Structure (R. N. Barnett & W. A. Lester,
Jr.)
Molecular Structure and Infrared Spectra of the DNA Bases and Their Deriva-
tives: Theory and Experiment (M. J. Nowak, L. Lapinski, J. S. Kwiatkowski &
J. Leszczynski)
Derivation and Assessment of a New Set of Ab Initio Potentials and Its Appli-
cation to Molecular Dynamics Simulations of Biological Molecules in vacuo, in
Crystal and in Aqueous Solution (M. Aida)
Practical Exercises in Ab Initio Quantum Chemistry — the World Wide Web as
a Teaching Environment (H. P. Lu'thi, G. Vacek, A. Hilger & W. Klopper)
Volume 3
Multireference Brillouin-Wigner Coupled-Cluster Theory (/. Hubac, J. Masik,
P. Mach, J. Urban & P. Babinec)
243
244 Content Index
Model Core Potentials: Theory and Applications (M. Klobukowski, S. Huzinaga

& Y. Sakai)
Stratospheric Bromine Chemistry: Insights from Computational Studies (5. Guha
& J. S. Francisco)
Bonding in Gas-Phase Sulfur Radicals (D. C. Young & M. L. McKee)
Chemistry of the Liquid State: Current Trends in Quantum-Chemical Modeling
(L. Gorb & J. Leszczynski)
Chemistry and the Internet (K. Flurchick, M. Hurley, J. K. Labanowski, G. H.
Lushington & T. L. Windus)
Volume 4
Topography of Atomic and Molecular Scalar Fields (5. R. Gadre)
The Ab Initio Model Potential Method: A Common Strategy for Effective Core
Potential and Embedded Cluster Calculations (L. Seijo & Z. Barandiarari)
Continuum Models of Macromolecular Association in Aqueous Solution (M. A.
Olson)
Interactions of Nucleic Acid Bases: The Role of Solvent (M. Orozco, E. Cubero,
B. Hernandez, J. M. Lopez & F. J. Luque)
Recent Advances in Multireference M0ller-Plesset Method (K. Hirao,
K. Nakayama, T. Nakajima & H. Nakano)
Detonation Initiation and Sensitivity in Energetic Compounds: Some Computa-
tional Treatments (P. Politzer & H. E. Alper)
Volume 5
In Search of the Relationship between Multiple Solutions Characterizing Coupled-
Cluster Theories (P. Piecuch & K. Kowalski)
Computational Time-Dependent Two-Electron Theory and Long-Time Propaga-
tors (C. A. Weatherford)
Self-Consistent Field Theory of Weakly Bonded Systems (E. Gianinetti,
I. Vandoni, A. Famulari & M. Raimondi)
Aromatic DNA Base Stacking and H-Bonding (J. Sponer, P. Hobza &
J. Leszczynski)
Direct Ab Initio Dynamics Methodology for Modeling Kinetics of Biological
Systems (T. N. Truong & D. K. Maity)
Content Index 245
Molecular Structure and Vibrational IR Spectra of Fluoro, Chloro and Bromo-

substituted Methanes, Silanes and Germanes: An Ab Anitio Approach (J. S.
Kwiatkowski & J. Leszczynski)
Volume 6
Relativistic Multireference M0ller-Plesset Perturbation Theory (Y. Ishikawa &
M. J. Vilkas)
15 Years of Car-Parrinello Simulations in Physics, Chemistry and Biology
(U. Rothlisberger)
Methods of Combined Quantum/Classical (QM/MM) Modeling for Large
Organometallic and Metallobiochemical Systems (/. B. Bersuker)
A Review of Ab Initio Calculations on Proton Transfer in Zeolites (M. Allavena
& D. White)
Ionic Clusters with Weakly Interacting Components-Magic Numbers Rational-
ized by the Shell Structure (S. Roszak & J. Leszczynski)
Turning Point Quantization and Scalet-Wavelet Analysis ( C R. Handy)
Volume 7
Molecules as Components in Electronic Devices: A First-Principles Study (M. Di
Ventra)
Tackling DNA with Density Functional Theory: Development and Application
of Parallel and Order-N DFT Methods (C. F. Guerra, F. M. Bickelhaupt, E. J.
Baerends & J. G. Snijders)
Low-Scaling Methods for Electron Correlation (5. Saeb0)
Iterative and Non-Iterative Inclusion of Connected Triple Excitations in Coupled-
Cluster Methods: Theory and Numerical Comparisons for Some Difficult
Examples (J. D. Watts)
Explicitly Correlated Coupled Cluster R12 Calculations (J. Noga & P. Valiron)
Ab Initio Direct Molecular Dynamics Studies of Atmospheric Reactions: Inter-
conversion of Nitronium Ions and Nitric Acid in Small Clusters (Y. Ishikawa &
R. C. Binning, Jr.)
Volume 8
Computational Modelling of the Solvent Effects on Molecular Properties: An
Overview of the Polarizable Continuum Model Approach (R. Cammi, B. Men-
nucci & J. Tomasi)
246 Content Index
Electronic and Nonlinear Optical Properties of 2-Methyl-4-Nitroaniline Clusters

(M. Guillaume, B. Champagne, F. Castet & L. Ducasse)
Ab Initio Calculations of the Intermolecular Nuclear Spin-Spin Coupling
Constants (M. Pecul & J. Sadlej)
Base Polyad Motifs in Nucleic Acids - Biological Significance, Occurrence in
Three-Dimensional Experimental Structures and Computational Studies
(M. Meyer & J. Siihnel)
Model Calculations of Radiation Induced Damage in DNA Constituents Using
Density Functional Theory (D. M. Close)
Excited States of Nucleic Acid Bases (M. K. Shukla & J. Leszczynski)

Q. Computacional Computational Chemistry Reviews Vol 9 (Leszczynski)

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

Q. Computacional Computational Chemistry Reviews Vol 9 (Leszczynski)

Uploaded by

Copyright:

Available Formats

Computational

Editor-in-Charge: Jerzy Leszczynski,

Vol. 2: Computational Chemistry: Reviews of Current Trends

Vol. 3: Computational Chemistry: Reviews of Current Trends

Vol. 4: Computational Chemistry: Reviews of Current Trends

Vol. 5: Computational Chemistry: Reviews of Current Trends

Vol. 6: Computational Chemistry: Reviews of Current Trends

Vol. 7: Computational Chemistry: Reviews of Current Trends

Vol. 8: Computational Chemistry: Reviews of Current Trends

\fp World Scientific

British Library Cataloguing-in-Publication Data

COMPUTATIONAL CHEMISTRY: REVIEWS OF CURRENT TRENDS

Printed in Singapore by World Scientific Printers (S) Pte Ltd

recent advances and trends in the applications of computer simulations

decomposition energy, topological parameters, and experimental neutron

1. Molecular Electronics with Gaussian98/03 1

2. Molecular Dynamics Simulations of Single Molecule 47

3. Molecular Dynamics Simulations of a Molecular 85

4. Computation of Excited State Potential Energy Surfaces 121

5. Modelling of Anisotropic Exchange Coupling in 153

6. Is a Dihydrogen Bond a Unique Phenomenon? 195

Content Index 243

J. J. Palacios ab , A. J. Perez-Jimenez0, E. Louisa'b,

*Departamento de Fisica Aplicada, Universidad de Alicante

b Unidad Asociada del Consejo Superior

de Investigations Cientificas, Universidad de Alicante

c Departamento de Quimica-Fisica, Universidad de Alicante

d Departamento de Teoria de la Materia Condensada

Institute de Ciencia de Materiales de Madrid (CSIC)

2 Electronic transport at the nanoscale: A quick

g = goT(EF) = —T(EF) = -^YlT^)- (!)

Figure 1: Schematic drawing of the scattering of conduction channels in a

come to us as a surprise at zero temperature since inelastic scattering events

Figure 2: Schematic drawing of the alignement between the molecular or-

3 Numerical implementation of Landauer formula

• The self-consistent field approximation. If one sticks to a stan-

• The atomic structure. Calculating the possible atomic arrange-

• The experimental evidence. Finally, it is crucial to know the de-

Possibly the main advantage of the numerical implementation that we

3.2 Methodology: Zero bias, zero temperature

Figure 3: (a) A molecular bridge model where a short capped nanotube is

in any of its multiple approximations. As far as transport is concerned,

[(SI - F - S ( ± ) (£)] G (±) (E)=I, (3)

self-consitent calculation. This is another source of uncertainty in the calcu-

G W (E) = S [(£S - F - £ W (£) j - 1 S (5)

where S is the overlap matrix. As explained in Appendix B this Green's

P = _- f F lm\s-lGi-+\E)S-1] dE. (6)

N = Tr[P • S], (7)

T{E) = Tr[rh(E)G^(E)rR(E)G^(E)}. (10)

a significant contribution will be typically much smaller and it will be de-

T(E) = Tr [TlL/2(E)GW(E)TR(E)1/2TR(E)^2G^(E)r^2(E)\ , (11)

3.2.1 Example: Aluminum nanocontacts

requires a realistic modelization of the formation process of these nanocon-

hoppings in Al, which, are usually neglected in parametrized tight-binding

3.3 Methodology: Finite bias, finite temperature

the density matrix out of equilibrium. Based on standard non-equilibrium

P = - - f°° [S-1G<(E)S"1] dE, (15)

G<(E) = tG<">(£) [fL(E)Th(E) + fR(E)TR(E)} G<+>(£). (16)

Mec = Mchemical + ^electrostatic > (17)

Mchemical = lim SEcole/SN

Figure 5: Schematic picture of the development of a potential drop between

3.3.1 Example: Gold nanocontacts

Figure 6: Average on-site energies of the atoms in the Au cluster shown in

Figure 7: Tranmission versus energy for V=0 (continuous line) and 5 V