Professional Documents
Culture Documents
Chemistry
Reviews of Current Trends
Volume 9
Computational Chemistry: Reviews of Current Trends
Published
Vol. 1: Computational Chemistry: Reviews of Current Trends
Edited by Jerzy Leszczynski
Volume 9
editor
Jerzy Leszczynski
Department of Chemistry
Jackson State University
USA
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
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
Jerzy Leszczynski
Jackson, MS
December 2003
CONTENTS
Preface v
Index 237
ix
Chapter 1: Molecular Electronics with Gaussian98/03
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
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.
• 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).
Molecular Electronics with Gaussian98/03 7
• 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.
[(£±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:
We can now obtain the total charge through the text-book expression
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]:
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
Molecular Electronics with Gaussian98/03 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.
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).
where
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:
where we define
^electrostatic = 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
Molecular Electronics with Gaussian98/03 15
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.
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
Molecular Electronics with Gaussian98/03 17
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)
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
18 J. J. Palacios et al.
_274.07 p , , . . . , .
I • CF |5 * • * '
s
U -274.09 - *
- 274 - 12 o • 1 i ' i
Distance Uwstrcnes)
-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)
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
Molecular Electronics with Gaussian98/03 23
3i 1 -r p
I (a) A ^ ! (b) •
°-2 - 1 0 1 - 2 - 1 0 1 2
Energy (eV) Energy (eV)
_274,04 | . 1 * . , . 1
-274.08 %|j>
8 -274,08 • * •
I .' *'
-874-1*, • i • i • ;
Distance {AascroossS
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
Molecular Electronics with Gaussian98/03 25
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.
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
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.
28 J. J. Palacios et al.
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],
Molecular Electronics with Gaussian98/03 29
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
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.
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 - 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.
Molecular Electronics with Gaussian98/03 31
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
32 J. J. Palacios et al.
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]
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.
36 J. J. Palacios et al.
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)].
Figure 23: Same as in Fig. 19, but for an N = 11 (5,5) open metallic
nanotube end-contacted to Pd(lll) surfaces (see inset).
Figure 24: Same as in Fig. 19, but for an N = 10 (5,5) open metallic
nanotube end-contacted to Ti(lll) surfaces (see inset).
38 J. J. Palacios et al.
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):
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,
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
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.
40 J. J. Palacios et al.
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
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
Changing to an orthogonal basis \<j>a > can be done through the equation
|^>=^5^1/2|^> (32)
0
The local charge density is written in terms of the Green's function as [29]
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.
Molecular Electronics with Gaussian98/03 43
[37] P. Jelinek, Ruben Perez, Jose Ortega, and Fernando Flores. arXiv:cond-
mat/0211488.
[43] G. Binnig, H. Rohrer, Ch. Gerber, and E. Weibel. Appl. Phys. Lett,
40:178, 1982.
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.
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 .
<£•••"'•'• v '- 4
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
Molecular Dynamics Simulations 55
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:
F = k(xo+d-x). (2)
F = at(xo-x). (3)
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
Molecular Dynamics Simulations 57
1000 I i
800 I
g- 600 /
-200 \ f
0 20 40 60 80 100 120 140 160 180 200
Extension (nm)
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
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.
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:
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 kcal- mol'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 kcal- mol 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.
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
4600 .
3500^ R > ^ ^ \ - ^ ^ ^ i
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
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.
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).
Molecular Dynamics Simulations 73
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).
74 W. Nowak and P. E. Marszalek
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).
Molecular Dynamics Simulations 75
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).
78 W. Nowak and P. E. Marszalek
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
Molecular Dynamics Simulations 79
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).
Molecular Dynamics Simulations 83
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
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
88 J. Seminario et al.
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 )
Raman Intensity
dr )
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.
90 J. Seminario et al.
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.
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,
92 J. Seminario et al.
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
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.
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
94 J. Seminario et al.
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)
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.
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.
Molecular Dynamics Simulations of a Molecular Electronics Device 95
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 ).
96 J. Seminario et al.
Figure 5. Time and frequency domain plots of the C-H bond vibration at 1000 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.
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.
Molecular Dynamics Simulations of a Molecular Electronics Device 97
98 J. Seminario et al.
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].
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.
100 J. Seminario et al.
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
frequency spectra are on the right. The analyzed bond is boxed on the molecule.
Molecular Dynamics Simulations of a Molecular Electronics Device 101
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
frequency spectra are on the right. The analyzed bond is boxed on the molecule.
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.
Molecular Dynamics Simulations of a Molecular Electronics Device 103
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
frequency spectra are on the right. The analyzed bond is boxed on the molecule.
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.
104 J. Seminario et al.
3
Molecular Dynamics Simulations of a Molecular Electronics Device 105
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.
106 J. Seminario et al.
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.
Molecular Dynamics Simulations of a Molecular Electronics Device 107
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 .
(
108 J. Seminario et al.
Molecular Dynamics Simulations of a Molecular Electronics Device 109
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.
i An
112 J. Seminario et al.
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.
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.
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.
Molecular Dynamics Simulations of a Molecular Electronics Device 117
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.
118 J. Seminario et al.
(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
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
2 Theoretical Developments
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;
^ = ^exp(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
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
Computation of Excited State Potential Energy Surfaces 129
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
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
+ J2(xi\Tl'\MH^cv=0 (7)
V
132 S. Chattopadhyay et al.
Hnv — Hw + ^ ^ + ^ N E P V (8)
(9)
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(r'X°|<M<<Mtf^> + E e x P( r A I )^^^l^>
+ exp(r")S^QH M |^) + exp(T")a^gH M |^>]c M
= Efc^exp(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
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)
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
vm
+ Y.(xlMTl/-T»)\<t>>1)H<iiVdvk
V
+E<^i(r"-ni^><^|SM^t|^>ccfc = ^ 4 * (21)
vm
Computation of Excited State Potential Energy Surfaces 137
\ m,n / (23)
H"v = Hv (24)
%v = {^\Hv\4>u) (25)
H'l = Hv (26)
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.
Computation of Excited State Potential Energy Surfaces 139
A. P4 Model System
-2.000 -j 1
-2.025 -
-2.050 - , s^
• \ 1 A i /
-2.075 - \ /
^ -2.100- \ /
3 • \ /
^ -2.125- \ /
-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
142 S. Chattopadhyay et al.
<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.
-1.60- 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.)
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.
144 5. Chattopadhyay et al.
-14.845 -i 1
-14.850 - ^ * ^ *
-14.855- /
-14.860- /«!/r:rrrTr^
^ -14.865- / /
-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
Computation of Excited State Potential Energy Surfaces 145
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.
~ 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
146 S. Chattopadhyay et al.
-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)
MRCEPA(O)-LRT
.1467. -\ MRCEPA(D)-LRT
" \ FCI
-14.68- \
3 ' \
>£- -14.69- \
5K \
O> \
<5 ' \
UJ -14.70- \
-14.72 A , , , , , , , , , , , , ,
4 6 8 10 12 14 16
4 Summary
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.
Bibliography
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,
Computation of Excited State Potential Energy Surfaces 151
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
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
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
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
156 M. Atanasov, C. Daul and H. U. Giidel
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.
J12=~J- Z ^ (9)
nln2 iel,je2
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
Modelling of Anisotropic Exchange Coupling 159
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.
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
ci^pci
H 2 O©«Mn II ^OH2
C1QJPC1
H2OCHCon8S0OH2
J*(lz,lz')=J(lzMz)
J(1,O)=-J(O,-1) (15)
J(1,1)=J(-1,-1)
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
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:
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/
(Jjos^i) (19)
J-- , i a -2\-2OSx(CO)|2/
\"2Sx2/
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/_
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 )
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)
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
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+:
D = - ^ (JaJ+Jl2) (28)
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:
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
3.1. 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)
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
Modelling of Anisotropic Exchange Coupling 167
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.
t2 <=> (1/2)1
t, »-(3/2)l (35)
^ - - ^ - " - ^
t1(0') = f(0) = fz3
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')+
Modelling of Anisotropic Exchange Coupling 169
scalar term:
+J(a2,a2)
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.
Modelling of Anisotropic Exchange Coupling 171
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:
i
172 M. Atanasov, C. Daul and H. U. Oiidel
(45)
p-orbitals belonging to the same bridging ligand and summation being taken
over all bridging ligands:
[d,fv|d^fv]=XX44fi:; (46)
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
Modelling of Anisotropic Exchange Coupling 173
Ja= 2 <l/2 | Osz' 11/2>= -a2 J(l ",1 ")+b2 J(0',0')-c2 J(1 ',1') (50)
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 ^
176 M. Atanasov, C. Daul and H. U. Giidel
f z fz
CsMnBr3:Yb3+
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
CO
•i
I
179
Modelling of Anisotropic Exchange Coupling 177
178 M. Atanasov, C. Daul and H. U. Giidel
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
Modelling of Anisotropic Exchange Coupling 179
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
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
180 M. Atanasov, C. Daul and H. U. Giidel
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.
(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 n^i
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.
Modelling of Anisotropic Exchange Coupling 181
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\^) •
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( ^ *^i2 > 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
Modelling of Anisotropic Exchange Coupling 183
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 - \*
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.
12 - /
/ 7 exc.state
g . / /MnClYb
4" Jl
- x* o vu y MnCl2Yb
MnBr3Yb S
MnCl3Yb
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,
Modelling of Anisotropic Exchange Coupling 185
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 ^u)- 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.
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
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:
Appendix 2
(1,2,3,4,5,6,7,8,9,10,11,12,13,14)=
((Psa^spAa^sP^a.nsp^a^p^a^p^ca^ep^ca^ep) 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:
190 M. Atanasov, C. Daul and H. U. Giidel
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
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
_A2 [ * 1 1 (A.3.7)
a2\hE(a2-*b) AE(b^>a2)
First order corrections for *F^, and W^, are given by:
(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.
K*2b=-4h2bPJ 1 + l- 1 (A.3.13)
a2bya2b[AE(a2^b) AE(b^a2)\
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+ -a^a+2b+) =
Modelling of Anisotropic Exchange Coupling 193
-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)
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.
194 M. Atanasov, C. Daul and H. U. Giidel
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
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.
Is a Dihydrogen Bond a Unique Phenomenon? 199
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
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.
Is a Dihydrogen Bond a Unique Phenomenon? 201
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,
Is a Dihydrogen Bond a Unique Phenomenon? 203
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
Is a Dihydrogen Bond a Unique Phenomenon? 205
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.
206 S. J. Grabowski and J. Leszczynski
2.8 I — — — ~
2.4 -1
i \
1.6- * ^ ^
1.2 J • 1 • • 1 ~
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;
Is a Dihydrogen Bond a Unique Phenomenon? 207
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).
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
Is a Dihydrogen Bond a Unique Phenomenon? 209
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).
210 S. J. Grabowski and J. Leszczynski
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
Is a Dihydrogen Bond a Unique Phenomenon? 211
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.
&(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.
212 S. J. Grabowski and J. Leszczynski
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
Is a Dihydrogen Bond a Unique Phenomenon? 213
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).
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;
214 S. J. Grabowski and J. Leszczynski
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.
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.
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
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
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
Is a Dihydrogen Bond a Unique Phenomenon? 219
0.028 - Jt
r 0.026- * /
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.
220 S. J. Grabowski and J. Leszczynski
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).
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.
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).
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.
Is a Dihydrogen Bond a Unique Phenomenon? 227
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.
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-
Is a Dihydrogen Bond a Unique Phenomenon? 229
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.
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:
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),
232 S. J. Grabowski and J. Leszczynski
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.
Is a Dihydrogen Bond a Unique Phenomenon? 233
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 '
", 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
237
238 Index
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
Jarzynski 63, 64
Keldysh formalism 13
kinetic exchange 153, 156, 182, 185
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
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
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
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
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