Sunteți pe pagina 1din 70

Progress in Reaction Kinetics. Vol. 23, pp211280.

1998
0079-6743 1998 Science and Technology Letters

REACTIONS IN RARE GAS MATRICES MATRIX


AND SITE EFFECTS
Y. Haas and U. Samuni
Department of Physical Chemistry and the Farkas Center for Light Induced Processes,
The Hebrew University of Jerusalem, Givat Ram Jerusalem 91904, Israel

Contents
ABTRACT
1.
INTRODUCTION AND SCOPE
11
Reactions in matrices matrix and site effects
12
How ordered is a matrix?
14
Matrix structure diffraction and scattering methods
14
Molecular adducts in matrices and in supersonic jets

212
213
215
216
217
221

2.

SITE STRUCTURE MODELING


21
Models using inter atomic potentials
22
Force field calculations
23
Molecular dynamics based simulations
231 Fixed shell models
232 Models using periodic boundary conditions

224
226
230
231
231
235

3.

REACTIONS IN MATRICES
31
Exclusions
32
Unimolecular reactions
321 Direct photodissociation
322 Sensitized bond fission
33
Isomerization reactions
331 General comments
332 Matrix effects
3321 Tunneling
3322 Thermal and infrared induced conformational isomerization
3323 Photochemical isomerizations
34
Site effects in IR induced isomerizations
341 Site specific rates in the isomerization of 1,2-difluoroethane

237
237
241
241
245
247
247
248
248
249
250
254
255

211

212

Y. Haas and U. Samuni

4.

BIMOLECULAR REACTIONS
41
Diffusion and reactions of migrating atoms in matrices
42
Reactions between neighbors site and matrix effects
421 Proton transfer: the naphthol/ammonia system
422 Oxidation of olefins by NO2
43
Long range interactions
431 Electron transfer and exciplex formation
4311 Photogeneration of ionic species of the form (RG)nH+

257
257
261
262
263
264
264
265

5.

MODELING REACTIONS IN MATRICES


51
Photodissociation
52
Isomerization of polyatomic molecules
521 The cistrans HONO isomerization reaction
53
Bimolecular reactions
531 The ozone+olefin reaction
532 The F2+cis-D2-ethylene reaction

267
267
269
271
271
271
271

6.

SUMMARY

273

7.

ACKNOWLEDGMENTS

276

8.

REFERENCES

276

ABSTRACT
Recent advances in the studies of chemical reactions in cryogenic rare gas matrices are
surveyed, with a special emphasis on site and matrix effects. The reactions considered include
direct and sensitized photodissociation, thermal and photo-induced isomerization of molecules
and molecular clusters, bimolecular reactions between neighbors and reactions involving
diffusing species. Methods for simulating the structure of trapping sites and the dynamics and
kinetics of chemical reactions in this low temperature medium are discussed. Even in the case
of these weakly interacting solvents, a detailed theoretical analysis based on gas phase derived
potentials can be made only for very elementary reactions, such as the photo dissociation of
diatomic molecules. Nonetheless, approximate simulations succeed in providing an insight into
the factors affecting the outcome of chemical reactions in matrices and are already being used
for planning and analysing reactivity patterns and rates.
Prog React Kinet 23: 211 280 1998 Science and Technology Letters

KEYWORDS cryogenic rare gas matrices, site and matrix effects, photodissociation, photoinduced isomerization, bimolecular reactions

Reactions in rare gas matrices matrix and site effects

1.
INTRODUCTION AND SCOPE
This review is about reactions in rare gas solids (RGS). The original idea of isolating molecules
in RGS, and particularly in rare gas matrices, was to prevent them from reacting. The term
matrix isolation, coined by Pimentel [Whittle 1954] implies the isolation of a species, reactive
as it may be, from interacting, let alone reacting, with other species. Due to their inertness, rare
gases are commonly used for this purpose, the isolated species being presumably sufficiently
diluted to have only rare gas atoms as nearest neighbors. Furthermore, the temperature is kept
low enough, so that diffusion is negligibly slow, and even rotation is often inhibited.
Nevertheless, reactions have been observed in RGS almost from the very beginning.
Infrared spectroscopy is one of the main tools used to probe the trapped species; the small energy
of an infrared photon was found in some cases to raise the absorbing molecule beyond a reaction barrier. Thus, conformational isomerization under the influence of a globar source radiation
was reported as early as 1960 [Baldeschwielder 1960]. In a similar fashion, visible and UV
photons can easily induce unimolecular reactions in matrices, and have been used extensively
to probe the mechanisms of many complex reaction systems. In recent years, unimolecular reactions are increasingly being studied in rare gas matrices, for instance in relation to atmospheric
and stratospheric chemistry. Bimolecular reactions are also reported in matrices, either between
neighbors, or as a result of migration. Thus, atoms (produced by photolysis) are found under
certain conditions to diffuse in matrices, or even in rare gas crystals, at rates rapid enough to be
of chemical interest.
The main diagnostic techniques employed in the study of matrix isolated molecules are
spectroscopic methods, mostly infrared, electron spin resonance (ESR, for free radicals), UV
absorption and fluorescence and, to a lesser extent, Raman spectroscopy. A general finding for
all spectra is some small change (up to a few percent) in the frequency of the spectral lines as
compared with gas phase spectra. This change is termed matrix shift, originating from a differing interaction of the matrix environment with the initial and final energy levels involved in an
optical transition. These interactions with the matrix are generally considered to be small compared to intra-molecular interactions, so that spectra can be analysed as if they were obtained in
the gas phase. However, early on many spectral lines were found to be split in a matrix, indicating different environments near the matrix isolated molecules. Other possible causes for the
matrix shift and line splittings could be some impurities or dimer formation, but it is well established that in some cases the splitting is due to the existence of distinct trapping sites that differ
only in the interaction of the guest with the matrix. The existence of differing trapping sites in
matrices is a manifestation of a more general phenomenon: most molecular solutions contain a
large range of trapping sites, yielding very complicated, inhomogeneously broadened spectra
even at very low temperatures (Shpolskii matrices are a notable exception). Rare gas matrices
are unique since for a given molecule one obtains only a few distinct absorption bands, which
possibly indicates the presence of only a small number of distinct trapping sites. This makes
matrices an interesting testing ground for the hypothesis that the rate or product distribution of

213

214

Y. Haas and U. Samuni

a chemical reaction may be environment (site) dependent. Indeed, the topic has been of interest
ever since trapping sites were identified and is still being actively explored.
Another aspect of matrix isolation that has emerged in the last decade or so is the recognition that with proper doping the matrix may become a more active solvent. An ionic character can be imparted to the matrix by doping it with polar molecules, or with powerful electron
acceptors, such as halogen atoms. Molecular polar species (such as (RG)+nBr) form between
the rare gas (notably xenon) and the atom, which are of interest on their own but also seem to
affect the chemistry of other guests trapped in the same matrix. A recent discovery that will be
discussed is the activation of relatively inert bonds such as CC in these doped solid xenon
matrices. This seemingly surprising result is believed to be related to long range electron transfer that can be induced in these solids.
For years, matrix isolation was the only practical way of spectroscopically observing
weak interactions between complex systems, such as found in van der Waals adducts. In the last
20 years, experimental and theoretical understanding of weak interactions between rare gas
atoms and other molecules, as well as between two or more molecules, was enormously
advanced through the studies of jet cooled clusters. Microwave, infrared and UV/visible studies
of gas phase molecular clusters (including many rare gasmolecule clusters) of various sizes are
now routinely performed. The results are often interpreted in terms of potential energy surfaces
whose form may be derived from the experiments. These potential energy surfaces may be
applied also to analyse matrix isolation data, as matrices are sometimes thought of as an infinite
size cluster.
A recent important development in the field (based largely on these and other experimental advances) is the growing number of theoretical papers trying to simulate the structure,
dynamics and kinetics of these system. From the theoretical point of view, rare gases, having no
internal structure (in contrast with molecules), are the simplest solvents. Interactions (amongst
themselves as well as with other atoms and molecules) have been extensively studied in the gas
phase using molecular beam scattering and spectroscopy, leading to the construction of fairly
accurate empirical potentials which were subsequently used in dynamic studies. The applicability of gas phase derived potentials for condensed phase data analysis can be most readily
checked using the rare gas matrices as a model. These efforts are still at a preliminary stage with
relatively simple approximations such as sums of atomatom pair-wise potentials being almost
exclusively used; more refined models are expected to be developed in the future. These theoretical efforts have already raised renewed interest in the concept of a trapping site, providing
in principle a means to describe the immediate environment of the guest molecule. Structures
arising from the simulations are used as starting points for the dynamic and chemical behavior
of the trapped molecules, leading to predictions relating the trapping site structure to reactivity.
These predictions can than be put to experimental tests, leading eventually to a microscopic
detailed description of the matrixguest interaction.
In rare gas matrices, as in any solid solvent, mixing is practically arrested after the

Reactions in rare gas matrices matrix and site effects

matrix is formed. Thus, matrix kinetics are likely to resemble more those of non-stirred systems,
such as surface or membrane reactions, than common solution kinetics. Their inherent simplicity
may actually make them ideal models also for more complex unconventional environments,
which appear to be of crucial importance in biological and many industrial applications. In this
context, the possible utilization of fractal kinetics is an interesting prospect. Some efforts in
these directions will be described in the review.
Most of the material covered in this review relates to matrices in which the major component is a rare gas atom. In some special cases, reference is made to other low temperature
solids, such as N2, O2, or CO2 , when it was felt that the data are relevant to the topic at hand.
The literature survey, restricted to optical monitoring techniques (excluding, for instance
magnetic resonance work) was conducted until December 1996.
11
Reactions in matrices matrix and site effects
The aim of this review is to survey some of the new discoveries and trends in the matrix isolation field, with a particular emphasis on the role of trapping sites and on their experimental and
theoretical characterization. It is necessary to distinguish clearly between site effects and matrix
induced effects. Matrix effects are defined as changes in the spectroscopic or kinetic properties
of a trapped molecule as one matrix is replaced by another. Sites effects are recognized in a
given matrix, often expressed as unexpected spectral line splittings [Gunde 1991].
Unexpected in the sense that all other possible sources for closely lying spectral lines
(oligomerization, adjacent impurities, conformational isomers) have been ruled out. Figure 1
shows an example of the site splitting observed in the UV spectrum of anthracene in an argon
matrix [Wolf 1994, Zilberg 1994]. Anthracene has only one conformer, making the assignment
of the splitting to site effects rather straightforward. Subsequent laser induced fluorescence
work confirmed the presence of several trapping sites. The high sensitivity of laser induced fluorescence allows the observation of trapped molecules at concentrations that are often below
the detecting limits of IR absorption spectroscopy [Hill 1995]. When inert gases are used as
matrix materials, the chemical reactivity of trapped molecules might be expected in general to
be affected by the matrix to a small extent except for the cage effect. In practice, however,
matrix effects on chemical reactivity are fairly common, while site effects are still very rare.
It should be noted in this context that it may be difficult in practice to determine whether
a splitting of spectral lines is due to different chemical species (particularly different conformers,
Section 3322) or to different trapping sites. Changing the matrix host is a common
method for distinguishing between these options since spectral lines arising from different sites
are usually much more sensitive to the nature of the matrix than the conformers relative population [Barnes 1984]. Another conclusive method is by changing the temperature of the gaseous
mixture before deposition: a change in the relative gas phase population of conformers follows,
leading to a corresponding change in the intensity of the spectral lines in the matrix.

215

216

Y. Haas and U. Samuni

Figure 1. The lower trace shows part of the UV absorption spectrum of anthracene in an argon matrix near the
origin of the S0S1 transition at about 370 nm. The three main peaks are due to three distinct trapping sites. The
upper trace shows the spectrum with a small amount of ammonia added. Three main bands due to
anthraceneammonia adducts trapped in argon are indicated by asterisks [Fraenkel 1993].

12
How ordered is a matrix?
Rare gases can solidify in a number of ways; the two most common lead to crystals (usually via
the liquid phase) or to a matrix produced by rapid condensation directly from the gas phase. The
rare gas crystals are well characterized but are rarely used for trapping guests. They would have
been a much better model system than the less well defined matrices had their use not been
severely limited in practice by their extreme reluctance to accept most guests. A good crystal
can be grown only under equilibrium (or near equilibrium) conditions, yet, under equilibrium
conditions, rare gases do not tend to mix well with most systems of chemical interest.
Matrices, typically prepared under non-equilibrium conditions, help in circumventing
the problem. The guest is forced into the rare gas solid by rapidly cooling the system, not allowing it to approach equilibrium. Figure 2 presents a schematic description of the matrix deposi-

Reactions in rare gas matrices matrix and site effects

Figure 2. A schematic description of the matrix isolation technique: a guest molecule is mixed at an elevated temperature with a carrier gas (host), and the mixture is allowed to condense onto a cold surface, held below the freezing point of the host. With argon as a carrier gas, typical values for the high and low temperatures are 300 and
15 K, respectively.

tion technique. A gaseous mixture of the molecule of interest (the guest) and a large excess of
the rare gas atoms (the host) is prepared in a separate chamber. The gaseous mixture is allowed
to effuse unto a cold substrate held typically in the temperatures range of 450 K. On reaching
the surface the gaseous mixture freezes immediately and a matrix is formed.
As will be detailed below, matrices are rarely amorphous. Rather, they seem to consist
of an ensemble of small crystallites, that are packed together randomly. A major advance in the
field in the last 15 years has been the systematic study of matrix structure by solid state methods
such as X-ray and neutron diffraction and scattering. These data can now be combined with
spectroscopic and other methods, to provide a working model for matrix isolation systems.
It should be added from the start that while the trapping itself is a non-equilibrium
process, if an equilibrated mixture of gaseous isomers is deposited, the high temperature concentration ratio is generally found to be preserved in the matrix. This has been directly demonstrated for many systems (see for instance ref. Bodenbinder 1994 for the cis/trans methyl-nitrite
system). Only in cases of very low barriers, the equilibrium may be disturbed during the deposition; it has been estimated that the barrier needs to be smaller than 1011 kJ/mole for that to
happen [Braathen 1990].
13
Matrix structure diffraction and scattering methods
Long range order in solids is studied using standard methods, such as X-ray or neutron diffraction and scattering, which are complementary since X-ray scattering probes mostly heavy
atoms, while neutrons are affected mostly by light atoms. The equilibrium structure of the

217

218

Y. Haas and U. Samuni

Figure 3. Experimental properties of rare gas matrices containing silver atoms, as a function of the deposition
temperature (adapted from [Schulze 1978]). a, The percentage of trapped individual atoms as the deposition
temperature increases, their percentage decreases due to cluster formation. b, The density of the matrix
it increases as the deposition temperature increases, due to a decrease in the number of vacancies and lattice
imperfections.

crystalline form of common matrix materials (rare gases, N2, etc.) has been studied for many
years and is well known [Pollack 1964]. The structure and morphology of matrices, typically
prepared under non-equilibrium conditions, may be different and, only recently, has it begun to
be studied quantitatively. Neutron inelastic scattering (NIS), neutron diffraction and X-ray
diffraction, were applied for this purpose, and the results were compared to data obtained by IR
spectroscopy, density measurements and sorption capacity (that measures porosity and surface
area). Early work showed that the properties of a solid formed by direct condensation of gases
onto a cold surface, may be very different from those of a solid made by slowly freezing a
liquid. The hydrogen (H2) sorption capacity of solid rare gas layers formed by gas condensation
was found [Abe 1979] to depend strongly on the condensation temperature. Other parameters,
such as the layers thickness or the rate of deposition were much less important. Highly porous
layers could be grown only if the deposition temperature was less than a certain value, Tc, that

Reactions in rare gas matrices matrix and site effects

was found to be 40 K, 29 K and 18 K for xenon, krypton and argon, respectively. At higher
temperatures, denser solids were formed, similar to those made from the liquid. This was compared to a previous study of the density (and refractive index): both were found to strongly
decrease when the deposition temperature was below Tc [Schulze 1974].
In a matrix isolation study of silver atoms it was found that the deposition temperature
strongly affected the yield of monomeric silver as compared to clusters. Only at very low temperatures (typically 013 of the matrix triple point temperature), was it possible to achieve high
monomer yield, Figure 3. At higher temperatures self aggregation was severe even at very low
concentrations (<1:1,000).
An example of recent X-ray diffraction work aimed at elucidating the gross structural
properties of matrices is the study of vapor-deposited molecular nitrogen [Becker 1993]. The
material was deposited at 14 K under conditions typical of matrix isolation studies. After a predetermined thickness was obtained, argon was added for X-ray diffraction, or methane for IR
analysis. Diffraction patterns or IR spectra were recorded at different times during the addition
of the second gas. The X-ray diffraction of the initial N2 layer showed the typical pattern of a
face centered cubic (fcc) nitrogen crystal, except that the lines were broadened due to the small
size of the crystallites formed this way. The average crystallite diameter was determined from
the line-widths to be 400 ; this value is small for a macroscopic crystal but large compared to
molecular diameters. Addition of argon caused the gradual diminution of the N2 peaks, and
growth of argon ones, also broadened by comparison with regular crystals. Beyond a certain
molar ratio, the intensity of the argon peaks remained constant. In the IR experiments, the
intense 1,300 cm1 band of methane was used as a probe. At molar ratios smaller than 005, only
one peak was observed. Beyond that value, the peak broadened and finally split into two.
These results were contrasted with two extreme possibilities: two completely separated
layers and a homogeneous mixture. It was concluded that the data are compatible with the situation schematically depicted in Figure 4. The suggested model is that under the conditions of the
experiment, nitrogen forms a very porous structure consisting of small crystallites (average linear dimension, 400). The added argon atoms first fill up the space between the nitrogen crystallites, forming a crystalline structure, since the temperature is high enough (>14 K in this case)
to allow the formation of solid argon in its equilibrium structure. An analog result was obtained
with methane, when deposited at 32 K. The diffraction peaks stop growing when the spaces
between the N2 crystallites are completely filled. This is explained by the much better heat conductivity of the compact solid formed, so that eventually newly deposited argon atoms are
quickly cooled, forming a disordered (possibly amorphous) layer that does not contribute to the
X-ray diffraction lines intensity. Again, corresponding data were obtained when methane
was deposited on the filled, compact solid (where the high wavenumber band appearing at
13052 cm1 is typical for solid non equilibrium methane). It is interesting to note that for both
procedures (argon or methane addition) the characteristics attributed to formation of an amorphous structure begin at surface coverage of 005.

219

220

Y. Haas and U. Samuni

Figure 4. A schematic presentation of the structure of a mixed matrix formed by deposition of a nitrogen substrate
followed by an argon layer on top of it (all depositions at 14 K). The N2 layer is porous so that at first the added
argon first fills the pores. This leads to the formation of small argon crystallites, as revealed by the increase in the
X-ray reflection. After the pores are filled, the heat conductivity of the substrate increases, and further addition of
argon atoms is accompanied by rapid cooling. The newly deposited argon atoms do not contribute to the Bragg
reflections, as the layer is disordered, possibly amorphous. Adapted from Becker 1993.

The message of this work is that fast cooling and slow deposition lead to a very open
structure, in agreement with the earlier sorption and density results. Such a structure would
probably allow much more facile diffusion of any trapped species, and possibly also enhance
the chance of finding guest molecules in grain boundaries and large vacancies. The fact that
matrices prepared in this way are found to contain only a small number of spectroscopically distinct trapping sites may indicate that even under these conditions only a small fraction of the
trapped molecules reside in such locations. Alternatively, different trapping locations may
appear to be spectroscopically indistinguishable.
Several papers deal directly with site characterization using NIS [Prager 1988,

Reactions in rare gas matrices matrix and site effects

Havighorst 1992, Langel 1993], attempting to probe the different trapping sites of methane and
ammonia in krypton, argon, nitrogen and their mixtures. The methane concentration was high
by matrix isolation standards (1:100), probably resulting in the formation of dimers and
oligomers; however, the authors consider the signal due to dimers too weak to be observed. In
pure argon, several types of sites were assigned: a single substitution (SS) in fcc structure the
most stable one. It was the only one observed when deposition was at 40 K. Deposition at lower
temperatures (525 K) led to the appearance of NIS peaks due to different types of sites, such
as fcc near a stacking fault, a SS hexagonal close packed (hcp) site, and distorted hcp sites. In
addition, broad peaks are due to methane trapped in much smaller clusters (or crystallites) with
a very small coherence length. They could not be removed by annealing to 40 K. Corresponding
sites were also observed in argon/nitrogen matrices. An investigation [Langel 1993] that
involved a major effort (four laboratories from three countries participated) demonstrated the
difficulties encountered in attaining quantitative structural data on dilute matrices by scattering
methods. Their results correlated reasonably well with IR data for major trapping sites, though
it seems that dimers appear more conspicuously in IR spectra than monomers trapped in minor
sites, while single molecules, even if trapped in secondary sites are more easily observed in NIS
than dimers. The lower sensitivity of NIS dictates the use of guest/host ratios of 100 for the most
dilute samples. At these ratios, a sample probed by optical methods will reveal a considerable
contribution from dimers and higher clusters.
14
Molecular adducts in matrices and in supersonic jets
Molecular adducts between potential reactants are sometimes precursors to bimolecular
chemical reactions, or appear as intermediates in the course of the reaction. As stated in a recent
paper [Bell 1990] matrix isolation alerts us to the presence of complexes that would otherwise
have remained unobserved. This paper studied the photoisomerization of metalethene
complexes to metalvinyl and metalhydrogen ones (Scheme 1, in which M stands for the
metal atom, M=Rh, Ir or Os). Light absorption loosens the metalethene complex leading to
the formation of a loosely bound adduct, held in place by the matrix atoms. One of the
hydrogen atoms can now bond directly to the metal with a concurrent formation of a
vinylmetal bond (Scheme 1(a)). The presence of a loosely bound adduct is indicated also by
the substitution of ethylene with another ligand bound to the metal (Scheme 1(b)) and by insertion of the metal atom into a XH molecule (Scheme 1(c)). The structure of isolated loosely
bound molecular aggregates is most readily derived from microwave studies in supersonic jets.
(see [Legon 1986] for a review). However, jet expansion is ill-suited for following through the
entire course of a bimolecular reaction since products cannot usually be monitored spectroscopically due to their low concentration and the broad distribution of quantum states resulting
from the reaction. A combination of jet and matrix isolation studies of molecular aggregates
therefore suggests itself as the best way of studying the role of these adducts in bimolecular
reactions.

221

222

Y. Haas and U. Samuni

(a)

(b)

(c)

Scheme 1
A question that naturally arises is whether the same structures that are stabilized in the
free jet will also be preferentially populated in inert matrices. This would be in line with the general premise that the matrix retains the conformation of stable molecules deposited in it.
However, since the binding energy of van der Waals clusters is often quite small, it is possible
that changes (even dissociation) will be incurred upon deposition in a matrix.
It turns out that in some cases the clusters structure in a matrix and in a jet are remarkably similar. For instance, the matrix infrared study of the benzenewater adduct by Engdahl
and Nelander [Engdahl 1985], which was later supplemented by studies of water adducts with
olefins and substituted benzenes [Engdahl 1986, Engdahl 87], was interpreted as indicating that
water is hydrogen bonded to the flat side of benzene with considerable freedom to move. It was
suggested that the two hydrogen atoms of water are equivalent as far as the interaction with the
benzene molecule is concerned. This can be the case only if the barrier between the two possible isomers is very low, namely, the cluster is floppy. These predictions were verified by a
microwave study of the same system [Suzuki 1992], showing that benzene acts in this adduct as
a hydrogen bond acceptor for both H2Os hydrogen atoms and that the potential is extremely flat
over the C6H6 surface. It is interesting to note that the binding energy determined for the gas
phase adduct ( 178 [Suzuki 1992], 225 [Cheng 1995] kcal/mole) is very close to that estimated for the matrix adduct (2 kcal/mole based on the vibrational red shifts of the OH (or OD)
stretch vibration in the matrix).
In some other cases, significant differences between the free and trapped cluster structures are indicated, as found for instance in the case of the SO2 dimer trapped in argon and N2.
According to microwave studies, the SO2 dimer consists of two inequivalent units and is flexible due to a very flat potential surface near the minimum [Taleb-Bendiab 1991]. The two subunits are not in the same plane: one monomer lies in the plane of symmetry, while the two
oxygen atoms of the second monomer straddle it. High level ab initio calculations [Bone 1992]

Reactions in rare gas matrices matrix and site effects

also indicate a weakly bound, flexible structure. In particular, the geometry of the dimer was
found to be very different from the structure of two nearest neighbors in solid SO2 [Post 1952]
in the crystal the symmetry axes and the dipole moments of the molecules are almost aligned,
while in the gas phase they are nearly anti parallel. This difference indicates that a matrix environment might affect the structure and properties of such a dimer.
Indeed, infrared studies [Schriver-Mazzuoli 1995] indicate that the structure of the
dimer in both argon and N2 is different from the gas phase one, reflecting the small energy difference between various possible gaseous structures, and the shallowness of the potential.
Furthermore, the structure appears to be matrix dependent: It was found that in argon, 32S16O2
displays only one band of the 3 mode (SO stretch), while two bands are found in N2. This led
to the suggestion that in argon the two subunits are equivalent, while in N2 they are not. The IR
data are not sufficient for a complete determination of the structure, but indicate in this case a
stronger interaction between the two subunits in argon than in nitrogen. This is also supported
by the fact that the 3 dimer absorption is red shifted by 134 cm1 with respect to the monomer,
while the two bands observed in nitrogen are shifted by only 25 and 5 cm1 respectively.
A stronger interaction between polar subunits in argon matrices (compared to nitrogen)
was also observed in other systems such as SO2H2O [Schriever 1988, Nord 1982]. This is
claimed to be a general trend of nitrogen matrices, which are more active than rare gas ones,
probably due to electrostatic interactions which tend to promote blue shifts. An even more reactive matrix is solid oxygen which is sometimes used in studies aimed at elucidating the mechanism of acid rain formation. The SO2 monomer is not oxidized to SO3 in an O2 matrix, while the
dimer is [Schriver 1991]: no explanation has been offered for this difference which may be due
to thermodynamic constraints.
In this context, it is tempting to probe what happens when a dimer (or a larger cluster)
is initially prepared in the gas phase and then deposited in a matrix. This might provide a means
to study the stepwise transition from isolated molecules to the solid state. Since dimers (and
other clusters) are easily generated in a supersonic expansion, it would appear that this transition could be probed by combining the supersonic expansion technique with matrix isolation.
Several recent papers describe efforts in that direction [Knzinger 1993, Knzinger 1995]. A
mixture containing the molecule seeded in a large excess of rare gas was expanded into high
vacuum through a pulsed nozzle, leading to the formation of dimers and larger oligomers,
entrained in a jet. The jet was skimmed and mass analysed spectrometrically after being ionized
by electron impact, providing a measure of the relative abundance of the different sized clusters.
The neutral beam was allowed at the same time to deposit on a cold surface forming a matrix,
which was subsequently analysed by IR spectroscopy. Monomers and dimers were easily
observed in the IR spectrum of the resulting matrix. However, the stepwise transition from
dimer to larger units (and ultimately to the solid) could not be explored explicitly for the molecules studied (weakly bound CH3CN or CO2 clusters). The spectra (and the interpretation) were
complicated by the copious formation of mixed molecule-rare gas clusters. In the case of HCN,

223

224

Y. Haas and U. Samuni

this complication was by far less severe, showing again that hydrogen bonding in HCN clusters
is stronger than HCN-RG interactions. A large number of clusters of the type (HCN)n with n>3
were observed, which were assigned to cyclic and linear isomers based on comparison with abinitio calculations. Even in this case, however, analysis was complicated by the presence of
HCNRG clusters in the jet, and the presence (in the IR spectrum) of unresolved bands due to
large clusters of indeterminate size. The experimental procedure is still being developed; however, it appears that direct deposition of jets onto a cold surface leads to a complicated, unwieldy
structure. The method of choice may require expansion of the molecules seeded in the lighter
gases (helium or neon, which do not form clusters) combined with trapping in argon or krypton
coming from a separate effusive beam.
In summary, the gas phase structure of a dimer is a proper guide to the matrix structure
only when the intra-dimer interactions are appreciably stronger than the interaction with the
matrix atoms, as expected for hydrogen bonded systems. For complexes that are more weakly
bound, the examples given and many others indicate that trapping in a matrix can affect the
structure. In many cases, these matrix effects remained unaccounted for. This demonstrates
the need for a better understanding of the matrix effect. Currently, simulations and calculations
are being initiated in order to formulate a consistent interpretation as discussed in Sections 2
and 5.
2.
SITE STRUCTURE MODELING
The existence of different trapping sites in a matrix may in principle be manifested in various
ways: spectroscopically, by the appearance of several absorption features in the spectrum,
[Jones 1986] or chemically, by a different reaction characteristics (rates, product distribution).
Experimentally, spectroscopy is the most obvious and common method for detecting different
sites. In the absence of specific distinct interactions, a single species trapped at infinite dilution
in a matrix is expected to exhibit a single spectral line for each gas phase transition. The observation of several bands corresponding to one gas phase transition consequently implies the presence of different trapping sites. Their disparate frequency shifts are believed to be due to somewhat different microscopic environments around the trapped species. This is a demonstration of
the non-equilibrium character of matrices the system is kinetically, rather than thermodynamically controlled during deposition. Therefore in general some metastable sites are formed, in
addition to the most stable trapping one.
Some of the attempts to relate the different spectroscopic transitions to specific structures will be surveyed in this section. To the best of our knowledge, the problem has not yet been
solved satisfactorily even for the simplest systems. This state of affairs reflects the many difficulties involved: no exact potential function is available for this multi body problem, and no
exact solution exists even for the approximate potentials.
A proper reference system is required for the structure of a matrix consisting of the neat
solvent. Two extreme options suggest themselves: one assumes perfect order, the other com-

Reactions in rare gas matrices matrix and site effects

plete disorder. The first alternative is dominant in the literature, since experiment shows that gas
phase spectral bands tend to split in the matrix only to a small number of narrow lines, suggesting a limited number of distinct sites. This assumption, as discussed in Section 13, is supported
by structural studies implying that rare gas matrices tend to consist of an ensemble of microcrystals. Therefore, a widespread approach is to hypothesize that the rare gas solid assumes its
most stable crystal structure (for instance, face centered cubic (fcc) for argon [Pollack 1964,
Niebel 1976]) and tries to embed the guest in it. Embedding is attempted in either a substitutional or an interstitial site. In the former case, a cavity is formed in the crystal by removing one
or more rare gas atoms from the lattice and the guest is inserted in it. In the latter, no atom is
removed and the guest is squeezed in between lattice atoms. The probability of populating a
given trapping site is estimated either intuitively, using geometrical arguments, or by some
quantitative method (see below). Since in most cases guests are bulkier than the matrix atoms,
the substitutional option is more frequently used.
An example of the intuitive approach is the case of HCN in several rare gas matrices
[Abbate 1985a, Abbate 1985b]. In these papers the authors claim that while multiple trapping
sites have often been proposed to rationalize multiple peaks in absorption spectra in matrix
isolated molecules, there have been no structural explanations of these trapping sites. Having
experimentally observed the splitting of the 2 (bending) mode of HCN in some rare gas
matrices, they offer a structural explanation.
The following trends have been observed in the IR spectra of HCN (several isotopomers
were studied). The bending frequency is blue shifted with respect to the gas phase value, the
largest shift (~9 cm1) is in argon, and it decreases as the size of the host increases. The CH
stretch (3) mode is red shifted, the shift increasing in the Ar<Kr<Xe sequence. For the CN
stretch (1) mode there is a small blue shift in argon, and the shift becomes smaller as the rare
gas atomic number increases, leading to red shifts for krypton and xenon. In xenon, 2 splits into
three peaks, 1 remains single and 3 splits in two. Table 1 summarizes some of the data. In
argon no splitting was observed (within the experimental 024 cm1 resolution), while in krypton and xenon several bands were split. Analysis of the data showed that the most likely reason
Table 1. IR absorption frequencies (cm1) of HCN in rare gas matrices [Abbate 1985a], and in gas phase
[Nakagawa 1969].
Band

HCN
Gas

Ar

Kr

Xe

2 HCN bend

71198

7209
7199

7186
7184
7198

7167

1 CN stretch

209685

20977

20941

20895

3 CH stretch

331148

33064
32948

32920
32800

32762

225

226

Y. Haas and U. Samuni

for the split in xenon is the existence of two trapping sites, one of which removes the degeneracy of the 2 mode. (Other possible causes for line splitting, such as dimer formation or appearance of several rotational bands, were checked and found absent; thus, HCN does not rotate in
xenon in the 930 K range). A plausible explanation for the results was given in terms of two
different SS sites in a fcc xenon lattice. In one the molecule is oriented along the 111 axis, and
in the other along the 110 axis (Figure 5). The lower symmetry of the latter removes the degeneracy between the two bending modes, and causes the site splitting. However, the dynamics of
the system, as tested by the vibrational relaxation measurements and IR fluorescence kinetics
were not affected appreciably by the sites structure.
The paper concluded: It is pleasing to have a complete and concrete physical explanation for the multiple peaks. This demonstrates that multiple peaked spectra can be a source of
information on the structure of impurity trapping sites rather than an annoyance ascribed to
mysterious site splittings. However, the proposed structures were based on purely intuitive
reasoning.
21
Models using inter-atomic potentials
Intuitive guesses based largely on size and shape fitting of the guest into the crystal lattice have
been quite useful in qualitatively explaining site structures, but cannot provide a systematic and
quantitative means for simulating the properties of trapping sites. Early on, Pimentel [Pimentel
1963] proposed a relationship between the sites size and the spectral shift by distinguishing
between tight and loose sites that lead to blue and red shifts, respectively, of IR absorption
bands. However, early attempts to calculate solvent shifts in perfect crystal lattices used models

Figure 5. The proposed structure of two trapping sites of HCN in an argon matrix (adapted from ref. Abbate
1985a).

Reactions in rare gas matrices matrix and site effects

227

[Friedman 1965, Mirsky 1980] that retained the perfect rigid structure of the host crystal. In
these models, the molecule occupied an undistorted substitutional site and the crystal was considered static. The shift was calculated as the sum of three major contributions due to induction,
dispersion and repulsion [Friedman 1965].
More recent papers employ an approach that seeks to reconstruct static and dynamic
properties of matrix isolated molecules by using approximate potential functions for the whole
system. This method is related to force field calculations that were remarkably successful
in reconstructing the vibrational spectra of polyatomic molecules [Lifson 1968, Huler
1974, Warshel 1970] and to methods developed for the analysis of molecular beam scattering
experiments.
In these models, potential surfaces that control the energetics and dynamics of the solid
phase system are constructed from potentials derived from gas phase data or, recently, from
ab initio calculations. Early work sought to minimize the total energy of the system, assuming
thermodynamic control. More recently, the calculation of local minima is emphasized, assuming that the matrix preparation does not necessarily lead to the overall energy minimum.
Classical, rather than quantum, mechanics is used in general, since with the heavy atom matrices quantum effects may often be neglected and the classical equations of motions are more
economically adapted to computer codes. Since there is no analytic solution to the equation of
motion of a system containing more than two bodies, approximate solutions are sought even
within classical mechanics. Different methods vary in the approximations made but the overall
procedure usually conforms with the following pattern:
(1) A potential function describing the interaction energy of all particles in the system
is constructed. This function is usually set up as a combination of three main contributions:
V= V(host)+V(guest)+V(interaction)

(1)

Where V(host) is the potential of the host, often taken as the sum of atomatom pairwise contributions which were parametrized to reproduce the properties of the neat crystal. V(guest) is
the gas phase internal potential of the guest molecule and V(interaction) is the interaction potential of the guest with the host, which is also usually taken as the sum of pairwise potentials. The
most commonly used potentials are Lennard-Jones for the interatomic one (equation 2) and
Morse potential (equation 3) for the intramolecular potential.
Vij=4[(/Rij)12-(/Rij)6]
ij

Lennard-Jones potential

(2)

Vij=D{1-exp[(Rij-Re)]}2

Morse potential

(3)

A listing of some commonly used potentials and their parameters is provided in Table 2.

228

Y. Haas and U. Samuni


Table 2. Lennard-Jones and Morse parameters used in various matrix simulations.
Lennard-Jones potential parameters (Eq. [2])
atom-atom pair
Ar-Ar

(cm1)

()

Reference

834

340

Fraenkel 1995

827

340

Guo 1995

98

335

Winter 1992

Xe-Xe

155

403

Guo 1995

Ar-Xe

124

365

Fraenkel 1994

127

314

Agrawal 1995

18

321

Samuni 1996

Ar-H

Ar-C

Ar-N

Ar-O
Ar-Br

22

311

Samuni 1996

542

338

Fraenkel 1995

420

336

Samuni 1996

389

309

Fraenkel 1995

459

336

Fraenkel 1995

390

309

Winter 1992

390

307

Agrawal 1995

434

291

Agrawal 1995

114

346

Fraenkel 1995

114

351

Fraenkel 1995

114

346

Winter 1992

Ar-I

117

361

Fraenkel 1995

H-O

16

288

Samuni 1996

C-O

39

315

Samuni 1996

Morse potential Parameters ([Eq. 3])


Molecule

D(cm1)

(1)

Re()

Reference

N2

77752

2732

1097

Fraenkel 1994

Br2

24557

1586

2281

Fraenkel 1994

NBr

25335

1824

1769

Fraenkel 1994

Cl2

19962

2036

1988

Alimi 1989

986

1745

3810

Raff 1990

992

Rare gas pair


Ar-Ar

1596

3760

Alimi 1990

Kr-Kr

140

1598

4007

Raff 1990

Xe-Xe

195

1468

4362

Raff 1990

Reactions in rare gas matrices matrix and site effects

(2) In practice, an infinite crystal cannot be simulated without certain simplifications.


The simulation should reconstruct, in the case of the pure host, the regular crystal lattice, since
the matrix is believed to be an ensemble of small crystallites (Section 14). In addition, a way to
avoid edge effects is required. The use of pairwise potentials is mathematically convenient but
is known to lead to a cluster construct whose structure is not face centered cubic, as in the solid,
an icosahedral or dodecahedral structure is preferred [Farges 1988]. This difficulty can be
circumvented by imposing the desired crystal structure as a prerequisite. A template having the
appropriate structure is constructed and the infinite nature of the system (thus avoiding edge
effects) is approximated either by fixing the outer domains of the structure in space, or by applying periodic boundary conditions (see below for details). In addition, since in practice the simulated system is of finite size, some truncation of the potentials at large distances is required
resulting in a significant decrease in the number of terms in the potential calculation. This is
particularly true when using periodic boundary conditions, otherwise multiple summation of the
same terms will occur.
(3) The system is subjected to a procedure that establishes the final size and structure
of the trapping site. Two basic schemes are most frequently used: in the first, a crystal slab is
constructed and the guest replaces a given host atom (or group of atoms) or is squeezed into the
slab interstitially. The total energy of the system is calculated for a large number of randomly
selected structures (Monte Carlo (MC) method [Metropolis 1953]). In the second, the matrix is
constructed by adding host atoms (and the guest) successively onto the template, until the
desired number of particles has been reached. Many trajectories are calculated, leading to a variety of structures (molecular dynamics (MD) approach [Allen 1987]). The final stage of the simulation is comparison with experiment in an attempt to calculate measurable properties of the
lattice, of the trapped molecules and of their dynamics within a matrix.
A justification for the use of the pairwise potentials approximation is the successful
reproduction (to within a few percent) of the structure, cohesive energy, bulk modulus and
Debye temperature of the pure crystals (e.g. the Debye temperature of Kr and Xe crystals
[Niebel 1976, Barker 1974]). Further tests for the physical validity of the simulations such as
reproduction of the crystal unit cell length, its temperature dependence, size convergence, and
stability of the temperature and volume of the simulated crystal with time (i.e. no disintegration
at long simulation times) are routinely performed. Commonly used means for the characterization of the simulated crystals are plots of the density as a function of radial distance or of the
distribution of the number of neighbors in the immediate shell. For example, to verify that the
properties of the simulated matrix are not size dependent the density of the neat crystal as a
function of radial distance from its center is computed (Figure 6).
Once the validity of the method in respect to the pure crystal is established, the simulation of matrices with trapped molecules can be attempted. The ultimate goal is the reproduction
of experimentally observable properties such as the number of distinct trapping sites as judged
from spectral line splittings, their relative abundance (from the integrated area of the bands),

229

230

Y. Haas and U. Samuni

Figure 6. The density of an argon matrix as a function of the radial distance from the central argon atom in two
simulations of an argon matrix. The horizontal line at 165 gm/cm2 represents the experimental density of a perfect argon crystal. Adapted from [Raff 1992]. (a), Simulation of a crystal, showing that as the size of the simulated crystal increases, its density approaches the density of a perfect crystal. (b), Simulation of a 454-atom cluster
formed by the procedure described in the caption to Figure 7 (in which the term pristine matrix is defined). The
matrix is porous, containing many lattice vacancies. Its density increases with the radial distance from the central
atom and then decreases due to the finite size of the cluster. In this simulation, the density is always smaller than
that of a pure argon crystal, even after annealing.

spectral shifts, line widths and their temperature dependence. Even the very number of different
sites is usually not easily compared with experiment, since line splitting is usually different for
different vibrational modes in a polyatomic molecule and since different sites may give rise to
the same spectral shift within experimental uncertainty.
Within the general simulation schemes mentioned, various detailed approaches were
proposed, each with its advantages and limitations. Some commonly used methods for simulating the trapping sites structure will be briefly described (Lennard-Jones pair-wise potentials are
used throughout, unless otherwise stated).
22
Force field calculations
This conceptually simple approach was proposed by Gnthard and coworkers as early as 1982,
and was elaborated on in a series of papers. The first paper in the series dealing with the calculation of the IR spectrum of 1,2 difluoroethane (DFE) [Gunde 1982], appears to be the first
major quantitative effort to calculate site effects and matrix shifts of infrared spectra. The paper

Reactions in rare gas matrices matrix and site effects

sought to reproduce an experimentally established generalization that for many polyatomic


molecules, matrices tend to induce a red shift for high frequency modes (higher than
400500 cm1), while for lower frequencies, blue shifts are more likely. The matrix is treated as
a supermolecule, consisting of a single guest inside a crystal fragment of finite size. Infrared
matrix shifts were calculated by solving the normal mode problem of all vibrational modes. The
gas phase potential of the molecule (V(guest)) was used in the harmonic approximation while
V(host) and V(interaction) were taken as sums of Buckingham pairwise potentials. The molecule was placed in a single substitutional site, and energy minimization was carried out either
by the steepest descent method or by determining local stationary points using Newton
Raphsons method and the Hessian matrix.
The consistent force field (CFF) method used allows the calculation of all normal
modes frequencies and of the self consistent packing of molecules in a crystal lattice [Gunde
1982]. The number of argon atoms was necessarily finite and the minimum number that faithfully represents a matrix was sought. For a small sized crystallite containing 62 argon atoms the
CFF process led to physically unrealistic results such as severe structural distortions in the
molecule (up to 20 in the torsional angle). In addition, argon atoms were found to be severely
displaced from their perfect crystal position by up to 2! Moreover, some normal modes were
calculated to have imaginary frequencies. These artifacts were partially removed by using a
larger crystallite with 364 argon atoms. The computational effort was considerably increased as
1,116 modes had to be calculated for 1,2-difluoroethane molecule in argon. In the larger
crystallite the calculated distortions are smaller (~07) but still unrealistically large, mainly in
the first coordination sphere.
A satisfying result was that different shifts were calculated for different trapping sites,
distinguished by the direction of the two-fold axis of DFE with respect to crystal axes. It was
concluded that the notion of sites as giving rise to unexpected line splittings is confirmed. In
contrast with experiment, all molecular normal mode frequencies were calculated to be blue
shifted (230 cm1). A possible reason for the failure to reproduce red shifts is the lack of any
electric polarization effects in the CFF scheme used.
23
Molecular dynamics based simulations
In this method, all lattice (host) atoms and any molecules or atoms added to the system are
allowed to move under the assumed potential field. Their motion is calculated by numerically
solving Newtons equations of motion (Schroedingers, if quantum effects are sought, but we
shall restrict the discussion to classical systems only). As outlined in Section 21, two principal
methods were used to simulate the infinite size of a matrix. They are briefly described in the
following two subsections.
231 Fixed shell models
One way to avoid possible edge effects due to the finite size of crystal fragment (as encountered

231

232

Y. Haas and U. Samuni

in the CFF model) is to add a rigid shell of motionless atoms around the guest/host system. An
example is the model offered by Raff [Raff 1990]. The starting point is a perfect crystal, represented by a small, arbitrarily chosen section. A typical simulation uses a section containing
5 5 5 unit cells (666 atoms). This structure is partitioned into three parts: the primary (P)
zone consisting of the inner 62 atoms; the secondary (Q) zone the 302 atoms lying outside the
P zone but within one unit cell of the outer boundary of the system; finally, the boundary (B)
zone consisting of the remaining 302 atoms. The latter zone is the shell, whose atoms are fixed
rigidly in their ideal lattice positions. In this way, the B zone atoms ensure the fcc structure of
the lattice for atoms far removed from the trapping site. Infinite crystal behavior is simulated
using the molecular dynamics (MD) approach in the following way: atoms in the Q and P zones
are allowed to move under the assumed potentials at any given temperature.
The motions of the P atoms are determined solely by the forces due to the potentials,
using strict classical mechanics. The motions of the Q atoms are also calculated using
Hamiltons equations but are modified by the presence of reset functions associated with each
Q-zone atom, which ensure the maintenance of a constant temperature in the system [Riley
1988]. The velocity reset functions are of the form:
vx(i)new(tn)= (1-w)1/2vx(i)old(tn)+w1/2vr(x,T)

(4)

where vx(i)new(tn) is the x component of the velocity of atom i at time tn, vx(i)old(tn) the old velocity,
and vr(x,T) a random velocity selected from a Boltzmann distribution at temperature T by the random number x. The strength of the reset is determined by the w parameter that takes values
between 0 (no reset) and 1 (complete reset). Similar expressions hold for the y and z components.
The reset functions, together with the stationary B-zone atoms, maintain the fcc structure.
The simulation procedure is as follows: a perfect crystal is constructed and then the
molecule is inserted into its center in the P zone in an interstitial site. This necessarily introduces
pressure in the crystal, which is relieved by allowing the system to equilibrate at the desired
temperature. However, the system is constrained to a fixed volume, while most experiments are
run at a fixed pressure. This leads to a distortion of the lattice in the innermost P zone.
It turns out that the effect is negligibly small at distances greater than 10 from the
guest. The model must be size convergent, namely that the properties of the matrix will not be
size dependent. This is checked by calculating the density of the neat crystal as a function of
radial distance from its center [Raff 1990]. Figure 6a shows that the density of a fcc argon
crystal indeed converges to the experimental value as the size of the crystal increases.
A variant of the method [Raff 1992] can be employed to simulate disorder, that was at
times assumed important for data concerning real matrices. The same procedure and potentials
are used but the molecule is deposited in the following way. A deposition chamber with fixed
dimensions (large enough 40 sides) is constructed (see Figure 7). The guest is placed in the
center of the chamber and 40 argon atoms are placed around it (but not too close) on the

Reactions in rare gas matrices matrix and site effects

Figure 7. A sketch of the model used by Raff for simulating matrix isolation of co-deposited ethylene and fluorine. The molecular pair is placed at the center of a parallelepiped box of fixed dimensions, and a number of
argon atoms (black circles) are placed in a plane that intersects the box in the middle, at the calculated positions
of a perfect argon crystal in the form shown. Other atoms are added (from above or below) at randomly chosen
coordinates within the box, and allowed to move till they reach a minimum energy. After an added atom reaches
that position, the whole ensemble is subjected to a thermalization procedure using a canonical Markov walk consisting of small steps and only then another atom is added. The final (pristine) matrix is obtained after additional Markov moves are made on the lattice, and the whole process repeated till the desired number of atoms has
been added. The pristine matrix usually contains many vacancies and defects, some of which may be removed by
annealing. (After Raff 1992).

233

234

Y. Haas and U. Samuni

z=0 plane at positions of a perfect fcc crystal. Other atoms are then added at random positions
within the box one at a time, and allowed to assume a local minimum. The system is allowed to
thermalize into a minimum energy configuration of the ensemble, with the guest held in place
and the atoms not allowed to move out of the box. This procedure is then repeated until
the desired total number of atoms is deposited. The resulting pristine matrix is found to be
amorphous with numerous vacancies and other imperfections. Figure 6b shows the density of
this imperfect lattice as a function of the radial distance from the center it is clear that density
is much lower than in the filled lattice (Figure 6), even after annealing.
Conceptually similar models were offered by other workers [Winter 1992, Molnar
1995]. Schurath and coworkers [Winter 1992], in a model designed to simulate the vibrational
frequency change upon trapping of diatomic molecules in a rare gas solid, used the concept of
a finite radius of interaction R0 around a given point: atoms outside of this sphere were fixed in
place, while those within were allowed to move under a potential. A guest molecule was inserted
into the center of the sphere, occupying an interstitial or substitutional site. (For NBr, the subject of the paper, only substitutional sites were considered). The number of mobile atoms was
kept small for computational reasons; a typical value was 200300. Only pairwise interactions
were taken into account, all higher order ones being neglected. For the argonargon and
argonguest atom, a Lennard-Jones potential with an added induction term Vind was used, while
for the intramolecular motion a Morse potential was used. Energy was minimized by letting the
mobile atoms move around and the most stable local configurations obtained were assumed to
represent trapping sites.
In the case of NBr in argon, the single substituted (SS) site was found to be the most
stable one; in it the molecule preferentially lies along the C4 axis of the crystal. Nonetheless,
once trapped in the less stable DS site, spontaneous transformation to a SS site does not take
place. (Annealing to high temperatures might cause the conversion.)
These results were confronted with laser induced fluorescence experiments on NBr
[Seranski 1992]. Spectral features assignable to three main trapping sites were identified, two of
which were nearly degenerate. The intensity of these two peaks was found to decrease on
annealing, showing that they are due to metastable sites They were assigned to DS sites, while
the third feature was assigned to a SS site. The annealing mechanism postulated was the conversion of a DS site to a SS + one vacancy, followed by rapid diffusion of argon atoms and
irreversible annihilation of the vacancy. The vibrational matrix shifts were calculated using an
effective potential, of a Morse form, for which an analytical solution to the Schroedinger equation exists. The Morse potential was derived through fitting the combination of the gas phase
intramolecular potential and the contribution of the guest-host potential in the direction of the
molecular axis. Agreement with experiment was very good for the SS site, but less so for the DS
site: a small blue shift with respect to the gas phase was obtained, while the experiment showed
a small red shift. The difficulty in reconstructing red shifts may have the same origin as in
Gnthards work cited above omission of electrostatic interactions.

Reactions in rare gas matrices matrix and site effects

232 Models using periodic boundary conditions


An infinite bulk crystal may be approximated in finite parallelepiped slab calculations by employing periodic boundary conditions. In this procedure, edge effects are eliminated by allowing the
atoms residing on one of the slabs surfaces to interact with atoms on the opposite surface. The lattice is thus folded back on itself, mimicking an infinite structure. This method was criticized as
it may not produce size convergence and suffer from artifacts due to back reflection of shock
waves that can be formed upon photodissociation [Raff 1990]. Indeed, care should be exercised to
avoid these difficulties: the slab used must be big enough to prevent these artifacts. A finite radius
of interaction must be imposed to avoid multiple counting of interacting pairs.
This model was used by Gerber and coworkers in their studies of the photodissociation
dynamics of molecules in rare gas crystals, see Section 4 [Alimi 1989, Alimi 1988, Krylov
1994, Krylov 1994a]. In these papers, a crystal slab was constructed and one or two of the host
atoms were replaced by the guest in this case the intuitive approach was used for the initial
choice of the trapping site.
In an attempt to simulate directly the deposition process, Fraenkel and Haas [Fraenkel
1994] offered a molecular dynamics based model employing periodic boundary conditions. The
process was begun by constructing a small seed crystal serving as a template and held at the
desired deposition temperature. It could be of either fcc or hcp structure, ultimately determining
the approximate structure of the final matrix. The guest molecules and more host atoms were
selected from a room temperature reservoir and were added one by one to the template by bringing them to within the radius of interaction from the surface atoms. They were allowed to interact with the growing solid under the sole influence of the potentials, except for a cooling procedure added to simulate the continuous cooling as in the real experiment. Only one guest molecule
was inserted simulating infinite dilution. No arbitrary site structure was assumed, the guest
being allowed to settle in the geometry arising from the deposition process. Many independent
runs with randomly chosen initial conditions were performed. The final structures probability
was determined from the frequency at which it appeared. A similar simulation method was
recently described by Cruz and Lopez [Cruz 1996].
It turned out that the structure of the trapping site is strongly influenced by the potentials, the deposition temperature and the cooling procedure. At relatively elevated template temperatures (for instance, 25 K for argon), the resulting matrix was well organized closely resembling a perfect crystal. In fact, with pure argon the perfect crystalline structure was recovered.
The trapping of a guest molecule induced some distortions of the crystalline structure in its
immediate vicinity. These distortions were found to be correlated with changes in the vibrational frequencies of the guest. If the molecule displaced nearby atoms away from their lattice
positions, the site is termed a tight one, as the host atoms tend to exert an inward force on the
molecule, mimicking the application of external pressure. A loose site arises when the nearest
neighboring atoms are found to be shifted towards the guest: in this case they will exert a
pulling force on the molecules atoms away from each other. For diatomic molecules, a correla-

235

236

Y. Haas and U. Samuni

tion between the sites structure and the frequency shift (with respect to the gas phase) was
found. A tight trapping site leads to a blue shift, and a loose one, to a red shift. This is in agreement with qualitative ideas proposed by Pimentel [Pimentel 1963] mentioned above. For a
given molecule, these deposition simulations resulted in essentially a single trapping site,
presumably the thermodynamically preferable one.
Simulation of the deposition at a much lower temperature and under more drastic cooling conditions, led to a less ordered matrix containing vacancies and various distortions. This
trend is in agreement with empirical findings, that optically transparent and high quality matrices
are best prepared at relatively high temperatures (1/3 of the melting temperature).
The method was recently extended to the deposition of polyatomic molecules [Fraenkel
1995]. These were treated as rigid bodies during the deposition process. In contrast with the
case of atoms or small diatomic molecules, in this case a large number of rare gas atoms may be
displaced from their perfect lattice conditions, leading in principle to a large number of different sites. However, when mild conditions were used (relatively high temperature, slow cooling),
the thermodynamically most stable site appears to form predominantly. The relative populations
of different trapping sites were found to depend also on the crystal plane exposed to the incoming gaseous molecules.
The same method may be applied to the simulation of cluster formation in rare gas
matrices (see Section 531). The two components are deposited successively in the usual way;
the clusters formed are found to occupy several different sites.
The calculation of matrix vibrational shifts may be demonstrated by the simulation of
the trapping of ICN in solid argon. The experimentally observed IR frequencies shifts in the
matrix with respect to the gas phase ones are listed in Table 3. It is seen that the matrix shifts for
the three normal modes are quite different: the CN stretch exhibits a red shift of about 1%, the
bend a blue shift of comparable magnitude, while the relative change in the IC stretch frequency
is significantly smaller.
In the simulation, all interatomic potentials were of the Lennard-Jones form and the
ICN intramolecular potential was the sum of harmonic potentials including a stretchstretch
coupling [Fraenkel 1995 p. 413]. Figure 8 presents the calculated structure of an ICN trapping
site. ICN occupies a disubstitutional site, two neighboring argon atoms being replaced by the
molecule. The positions of most of the argon atoms are not appreciably affected by the guest:
Only some of the argon atoms in the nearest solvation shell are displaced from their equilibrium
positions as depicted in Figure 8. The outward movement of the argon atoms lying broadside
to the molecule, adjacent to the carbon atom results in a counter pressure of the argon atoms.
This restricts motions transverse to the molecular axis impeding the bending and leading to an
expected blue shift. The argon atoms situated along the molecular axis are pulled inwards,
one on the iodine atoms side and five argon atoms on the nitrogen atom end. They are expected
to be exerting a pulling force on the ICN molecule along its axis leading to an expected red shift
for the CN stretch, and a rather minor shift for the IC one.

Reactions in rare gas matrices matrix and site effects


Table 3. Calculated vs experimental matrix shift for ICN vibrational frequencies. Data from ref. Fraenkel 1995.
Mode description

Experimental matrix shift


(%)

Calculated matrix shift


(%)

I-C stretch

+021

00

ICN bend

+099

+30

C-N stretch

082

005

The vibrational frequencies of the guest molecules were obtained by Fourier transforming the calculated trajectories of their interatomic separation as a function of time. The frequency calculation was done twice for each mode: Once for the isolated molecule and again for the
matrix trapped one. The difference between the two calculated frequencies yields the matrix
shift.
The calculated shifts, also listed in Table 3, show that the model provides a qualitative
account for the experimental matrix shifts. The bending mode frequency is calculated to be blue
shifted in the matrix by 9 cm1 (+3%). The IC stretch frequency, again in agreement with the
qualitative discussion, is essentially unchanged. However, the CN stretch is calculated with a
red shift of a mere 1 cm1 (005%) as compared to the experimental 082%. The recurring
failure to reproduce the red shifts, found also for the simulation of NBr [Fraenkel 1994], is
thought to be a shortcoming of the pair-wise potentials used.
3.
REACTIONS IN MATRICES
31
Exclusions
This review does not attempt to cover comprehensively the large volume of work on reactions
in matrices. In recent years, experimental work increasingly highlights the importance of trapping sites and matrix induced effects. These developments led to considerable theoretical activity which has already resulted in important cross-fertilization between theory and experiment.
We shall focus on these developments but realize that even within this limited scope, the choice
of topics and subjects is necessarily limited, arbitrary and personally biased. We apologize for
having to omit much important and fascinating work but hope that even this incomplete survey
will help in spurring further progress.
Thus, we shall not touch upon the vast literature dealing with the use of matrices as
media for novel reactions. In some of these studies, the reactant itself is part of the matrix
material (for instance O2 in oxidation reactions [Downes 1995]). The detailed structure of the
reactive sites was often of little importance to the authors who were trying to prepare novel
compounds, or to develop new ways of making known materials. These largely synthesis-driven studies are discussed in a number of recent reviews [Perutz 1985, Almond 1989, Almond
1994].
Free radicals (or biradicals) are often produced in cryogenic matrices, by photolysing a
precursor containing an easily photo-detachable group. Typical precursors include azo and

237

238

Y. Haas and U. Samuni

Figure 8. The structure of a trapping site of ICN trapped in argon, as simulated by MD [Fraenkel 1995]. The model
predicts a DS site causing a blue shift for the bending vibration and a red shift for the stretching ones (see text for
details). Shaded atoms are those that were found to be most severely displaced from their perfect lattice positions.

diazo compounds, whose use was largely due to the stability of the N2 product, equations 3
and 4, or dioxinones and other oxodiones that easily extrude CO or CO2 yield ketene derivatives
(Scheme 2).
R-N=N-R h
R-N=N-R R + N2+ R

(4)

R-N2h
R-N2R + N2

(5)

Reactions in rare gas matrices matrix and site effects

Since it is believed that only atoms are capable of translational motion in matrices, the
fragments formed, even the small ones such as N2, CO and CO2 will not move away from the
trapping site of the parent molecule. A likely scenario is that the original molecule is photodissociated and the products are prevented from geminate recombination by the rapid vibrational relaxation of the very stable smaller fragments. In that case, one expects possible spectroscopic manifestations of the fact that the newly born larger species is not isolated in a pure
rare gas cage. We shall not discuss these subtleties in the present review.
Even when produced in the gas phase by rapid thermolysis or photolysis, and quickly
frozen into a cold matrix, subsequent photolysis may reveal site rearrangement rather than photochemical changes. Matrices are often used to assist in the elucidation of reaction mechanisms,
but care must be exercised to avoid possible wrong assignments due to site effects. An example
was provided by Wentrup and co-workers extensive photochemical studies of the s-Z and s-E
conformers of acetyl ketene in cryogenic matrices [see Kappe 1995 for a typical example].
Various precursors were used to generate these conformers in the gas phase, some of which are
shown in Scheme 2. UV irradiation of the acetyl ketenes formed in this way in nitrogen or
xenon matrices led to interconversion of the conformers. In an argon matrix, UV irradiation led
to the formation of a new series of IR bands that disappeared upon annealing. This evidence and
deuteration experiments reported in a subsequent paper [Zuhse 1996] were used to show that
the photochemically generated new absorption bands are due to different sites, rather than new
chemical species. The exact nature of the site was not reported. This is an example of photo
induced site exchange, which will not be further discussed in this paper.
Another important application of matrix isolation is the characterization of novel, highly
reactive species. Andrews and co-workers developed in recent years a method for synthesizing
compounds of metals, including those with a very high boiling point, by using a laser ablation
metal vapor source. Many oxides, hydrides, nitrides and other compounds of metals such as calcium, strontium and barium [Andrews 1994], zinc and cadmium [Greene 1994], iron [Chertinin
1995, Chertinin 1996, Andrews 1996], zirconium and hafnium [Chertinin 1995a] were investigated for the first time. The quoted references are just a small part of this important effort which
will also not be pursued further in this review. Likewise, the studies of interesting intermediates,
such as metal complexes able to activate selectively CH bonds, continue to benefit greatly from
matrix isolation methods. Some recent examples may be found in ref. Hall 1992a, Hall 1992b,
Hill 1995 and Cronin 1995. In recent years, the use of laser induced fluorescence methods for
these relatively large molecules leads to improved spectral resolution and to the detection of distinct sites and/or conformations. These developments will certainly be soon followed by more
refined theoretical treatments. Another field of intense current activity is the laboratory study of
atmospheric chemistry particularly related to the ozone depletion problem and to pollution. Apart
from a short discussion of the (ClO)2 system, this excellent application of matrix isolation must
be consulted elsewhere [Nelander 1977, Lugez 1993, Jacobs 1994].
In addition the survey is restricted to optical monitoring techniques (excluding, for
instance magnetic resonance work (NMR or ESR).

239

240

Y. Haas and U. Samuni

Scheme 2

Reactions in rare gas matrices matrix and site effects

32
Unimolecular reactions
As noted above, the cage effect is one of the dominant features of matrix isolation. Therefore,
dissociation reactions are often strongly inhibited in matrices, in contrast with gas phase or
liquid solution systems. In matrices, a large energy excess over the gas phase dissociation energy
is usually required for a permanent dissociation to take place. This practically rules out thermal
dissociations since the matrix is melted before dissociation is observed. Even infrared laser
induced dissociations in matrices have not been reported, to our knowledge. This section will
therefore focus on photodissociation reactions.
321 Direct photodissociation
A central question in this field concerns the way the fragments escape from the matrix cage. The
large numbers of molecules that were dissociated in matrices in spite of the obvious restrictions
on translational motion led to speculations that in many cases matrix faults are involved. An
early empirical observation that smaller fragments have a higher escape probability than larger
ones indicated that this is not always the case. In the systematic study of XCN photolysis (X=H,
F, Cl, Br) [Miligan 1967], the yield of CN was high when X=H, and was reduced substantially
as X increased. This was further substantiated recently [Fraenkel 1993a, Samuni 1994] by the
failure to detect CN upon photolysis of ICN in an argon matrix. In this study, the photolysis of
ICN lead to a decrease of only 26% in the ICN concentration (as measured from the decrease
in the 2171 cm1 band intensity). Concomitantly new absorption lines appeared at 20575, 2019
and 494 cm1. These lines were assigned to INC: the first being the C-N stretch of INC, the
second the CN stretch of the natural abundance 13C isotopomer (an assignment supported by a
calculation yielding exactly the same frequency 2019 cm1) and the last is assigned to the IN
stretch. The assignment is further supported by an ab initio calculation with good agreement
with all the experimentally observed frequencies as well as the IR intensities. It was concluded
that the photolysis fragments, the iodine atom and CN radical, cannot escape the cage; most collisions between the two lead to recombination to ICN and only a small fraction form INC. The
final product ratio, obtained after an elongated irradiation, appears to be due to a photostationary state. Earlier ESR work reported detection of CN radicals in a similar experiment [Easley
1970]. A possible reason for this apparent discrepancy is the much higher sensitivity of ESR
detection as compared to IR reaction in lattice imperfections or some minor trapping sites may
be overlooked by IR, and still detected by ESR for which quantitative yield data were not
reported.
The least barrier for cage escape is expected to be found for the smallest atom. The
photo-formation of H atoms appears, therefore, as the simplest and most unhindered photodissociation reaction in a matrix. However, even in this case a substantial cage effect was found for
systems such as HI in xenon [Lawrence 1992] and H2O in argon or krypton [Schriever 1989,
Schriever 1990]. VUV synchrotron radiation was used to excite the H2Orare gas systems, the
reaction progress being followed by the accumulation of OH radicals in the solid. The matrix

241

242

Y. Haas and U. Samuni

induced barrier was about 18 eV in excess of the gas phase barrier (5118 eV), for both rare
gases. The data were fit with the empirical expression:
q(h)=A(h-ETh)n

(6)

where q(h) is the dissociation quantum efficiency at the photon energy h, ETh the threshold
energy and A an efficiency parameter. The dissociation efficiency was found to increase substantially when the temperature was raised from 5 to 30 K. Empirical fits to equation 6 showed
that this was due to the increase of A with temperature, rather than to a smaller threshold at the
higher temperature. Deuteration led to a small (01 eV) increase in the threshold energy, and a
40% decrease in A.
These data may be accounted for by assuming that water occupies a single substitutional site as shown in Figure 9 (position 1) and that dissociation is completed when the H atom is
stabilized in a nearby Oh interstitial site (position 3 in Figure 9). The trajectory to the Oh interstitial site position 2 is the easiest one, yet it has no barrier to prevent recombination. The
trajectory to position 3, on the other hand, requires a jump to the Td interstitial site (involving an
energy barrier of about 1 eV) followed by a movement through the basis of the tetrahedron via
the D3 interstitial site (with an additional barrier of 0103 eV), as depicted in Figure 9b. The
overall barrier prevents the recombination of the H atom after it thermalized in the well of the
Oh interstitial site at position 3. It is noteworthy that the calculated barrier is very close to the

Figure 9. Photodissociation of water in an argon matrix. (a) The molecule is assumed to occupy a tetrahedral SS
site (position 1). The H atom leaving it may end up in one of two octahedral interstitial sites 2 and 3. (b) The
trajectory to site 2 involves no barrier, making recombination very likely. The path to site 3 involves two intermediate sites: a tetrahedral one TD and a D3 one. After crossing through these sites, the H atom is thermalized in
the Oh site and has to overcome a barrier to recombine (After [Schriever 1989]).

Reactions in rare gas matrices matrix and site effects

experimental value (18 eV). Thus, the data indicate a prompt cage escape beyond a certain barrier. This study is an example of the strong influence that a trapping site may have on photodissociation probability. It is also remarkable that the data appear to support the notion of a single
predominantly occupied site. Had a significant fraction of the water molecules resided in lattice
imperfections, one would expect to observe a measurable yield below the threshold. A search
for below threshold yield showed that it amounts to less than 103 of the saturation value above
threshold.
The escape of a dissociated H atom may be more facile when larger molecules are
photodissociated, as the matrix in the immediate neighborhood of the trapping site is strongly
perturbed. The formation of metallocenes (M(5-C5H5) 2), where M=Re,Mo, W [Hill 1995] by
photolysis of M(5-C5H5) 2 H is a typical example. It was found that the metallocenes are easily
formed upon irradiating the hydrides by a high pressure mercury arc in argon and N matrices.
The visible absorption bands of the products were found to consist of several distinct progressions. It is still not clear, however, whether these progressions are due to the coexistence of
several trapped conformers or to distinct sites.
Facile photodissociation was observed for other systems such as H2S (to H+HS) [Zoval
1994], F2 [Kunttu 1990, Kunttu 1991], N2O (to N2+O(1D)) [Lawrence 1992, Ryan 1994], O3 (to
O2 and O(1D)) [Benderskii 1994] and even OCS (to CO and S(1D)), where sulfur, a second row
atom, seems to escape the cage [Tanaka 1995]. These results were taken to indicate that small
atoms can easily leave the cage, while larger ones are inhibited. The 1D states of oxygen and sulfur are apparently small in this context based on calculations using pair wise potentials [Kizer
1995, Dunning 1977]. The very facile dissociation of F2 in crystalline argon and krypton (quantum yield ~1 at an excess energy of 24 eV [Feld 1990]) is explained by a model similar to that
used for the water photodissociation. The reaction coordinate is assumed to involve crossing of
a plane connecting the three nearest neighbors in a single substitutional trapping site. The barrier is alignment dependent and smaller when the p electron orbital is parallel to the plane than
when it is perpendicular to it. The two fluorine atoms fly apart along the FF bond, and since F2
molecules may be oriented in different angles with respect to the window, a distribution of
barrier heights is obtained. In addition, for a given p orbital orientation, the transmission of the
atom through the window depends on the size of the window which varies due to the thermal
motion of the three atoms, leading to a distribution of barrier heights shown in Figure 10b.
An attempt to estimate how far a relatively large fragment travels before the molecule
can be considered to be fully dissociated was taken up in [Clemitshaw 1988]. CF3X (X=Br, I
and NO) molecules were chosen, the assumption being that a large fragment will not move far
due to the cage effect. Highly diluted matrices of up to 1:105 were used to ensure monomer reactions. The infrared spectra of the CF3 product indicate that they are formed in a range of sites,
each leading to a somewhat different IR shift. In particular, the absorption band due to 3 (the
CF stretching mode) was found to split into two major peaks, at 12478 and 12517 cm1. They
were assigned to a proximity radical of the structure [CF3I] and to a fully dissociated one,

243

244

Y. Haas and U. Samuni

Figure 10. (a) The structure of the F2 SS trapping site in a krypton matrix. The arrow depicts the minimum energy
path for dissociation. The F atom has to pass through the indicated (shaded) triangle, to reach the nearest trapping
site, an interstitial octahedral one. (b) The potential energy surface along the reaction coordinate, Q. The height
of the barrier fluctuates due to the thermal motion of the three cage atoms (shown) leading to a distribution of
barriers, depicted schematically on the right hand side. (Adapted from [Kizer 1995]).

Reactions in rare gas matrices matrix and site effects

respectively. The fully dissociated species can be identified by comparison with literature data
on the infrared spectrum of the isolated fragments, in this case of the CF3 radical, in an argon
matrix [Jacox 1994]. The long range motion of the heavy iodine atom that is implicated appears
to be incompatible with the previously mentioned high barriers to dissociation in matrices. In
particular, iodine was demonstrated to be unable to escape the cage in an argon matrix, as discussed above for the case of ICN photolysis [Fraenkel 1993a, Samuni 1994]. Local annealing
was proposed as a possible mechanism enabling the separation of the heavy iodine atom from
the site in which the reaction took place: the immediate neighborhood of the absorbing molecules is heated following energy transfer from the electronically excited species. The local
annealing hypothesis is that even though the matrix bulk remains solid throughout the experiment, the immediate neighborhood of the reacting molecule(s) may become liquid-like, when a
large fraction of the absorbed light energy is converted to heat. Further work appears to be
required in this case, since local annealing is usually observed under conditions of high light
intensities or high guest concentrations, conditions apparently not holding for the experiments
reported in this paper [Clemitshaw 1988].
Rare gas matrices are capable of hosting reactive species such as atoms or free radicals
for long periods of time. Apkarian and coworkers have studied the fate of the photodissociated
halogen atoms in considerable detail [Fajardo 1986, Kunttu 1990]. It was found that charge
transfer often takes place, to form ionic species such as RG+X, which are stabilized by charge
delocalization. In particular, complete photodissociation of F2 can be achieved by extended UV
irradiation, due to the facile migration of F atoms mentioned above and further discussed in
Section 41. For instance, photodissociation of F2 in krypton leads to the formation of Kr6+F
[Kunttu 1990], that localizes on a subpicosecond time scale to Kr2+F, observed by its characteristic fluorescence spectrum. In xenon doped with atomic fluorine [Fajardo 1986], the vertically accessed charge transfer states decay to form charge transfer configurations that involve
energy storage by permanent charge separation. This unique feature of halogen doped xenon
matrices provides an opportunity for utilizing the photon energy for chemical conversions, as
discussed below.
322 Sensitized bond fission
Molecules containing the cyclopropane ring are known to be photo-stable under UV irradiation
at wavelengths exceeding 220 nm, since they do not absorb light in that wavelength range. It
was recently discovered that certain matrix environments can be used to induce a photo-assisted
ring opening at longer wavelengths [Maier 1994]. Thus, 254 nm irradiation of methylene cyclopropane in a halogen doped xenon matrix leads to trimethylene methane biradical (scheme 3a),
and of cyclopropane, to propene and allyl radical (scheme 3b).
It is noted that no reaction takes place in a neat xenon matrix. Several different halogen
atoms (Cl, Br or I) lead to reactivity, but bromine appears to be the most often used one. Different
methods may be used to introduce the halogen atom into the matrix, all leading to the same effect:

245

246

Y. Haas and U. Samuni

(1) Irradiation of halogen molecules in the matrix (not useful for I2, whose dissociation
yield is negligible).
(2) Pyrolytic generation of the atoms and immediate trapping at 10 K.
(3) Irradiation of HX (X=Cl or Br) in the matrix. In this case it is known that, in addition to the halogen atoms, ionic species are formed: Xe2H+ and HBr2, [Kunttu 1992, Kunttu
1994], see Section 4311

Scheme 3
The proposed mechanism is based on the results of Apkarian et al., quoted above, concerning the formation of a delocalized exciplex of the form {[Xen]+Br} in which n can be a fairly
large number. This means that while the negative charge is localized on the bromine atom, the
positive charge is delocalized over a large volume in the matrix. Fajardo and Apkarian [Fajardo
1986, Fajardo 1988] showed that the exciplex can be localized to a smaller emitting species of
the form Xe2+Br, but Maier and coworkers propose that in the presence of a suitable substrate,
such as cyclopropane (CP), the delocalized hole may act as a acceptor to form a partial charge
transfer complex [XenCP]+. This complex, in turn, is neutralized by electron transfer from the
Br ion, which may be separated from the CP molecule by several xenon atoms. Electron transfer, which can take place over large separations [Calcaterra 1982], may well be efficient under
these conditions. The resulting charge neutralization releases a large amount of energy (typically
34 eV), some of which is channeled to the cyclopropane molecule, which eventually decomposes thermally.
This proposed mechanism is admittedly not without flaws. It is difficult to see why energy
localizes efficiently in the CP molecule, and if it does, how is it that vibrational relaxation does
not compete effectively with the reaction. Furthermore, the charge transfer complex
{[Xen].+Br} absorbs strongly at 310 nm, but irradiation at this wavelength leads to no reaction
whatsoever.

Reactions in rare gas matrices matrix and site effects

Nevertheless, the fact stands that the reaction proceeds readily upon irradiation at
254 nm, and that it takes place only when the xenon matrix is doped with a halogen atom. The
actual mechanism may turn out to be more complicated.
In summary, it now appears that small atoms, such as H, F or O can be prepared
by photodissociation. The cage escape is facile when the p orbitals are properly aligned, as
shown by detailed simulations starting from a well defined site (see Section 51). When potent
electron acceptors such as halogen atoms and oxygen are formed, photodissociation often
results in the formation of charge transfer states in which the matrix atoms themselves act as
electron donors. A non uniform charge distribution which imparts an ionic character to the
matrix is created which in turn may lead to secondary molecular dissociations, as found for
cyclopropane.
If photodissociation leads to larger atoms, or molecular fragments, cage escape is in
general restricted to a very short distance, possibly leading to a proximity pair. Complete separation of the fragments or a considerable rearrangement of the immediate neighborhood of the
reactant appear to involve at times local melting, particularly when high light intensities or high
concentration are used.
33
Isomerization reactions
331 General comments
Isomerizations, which in the context of this section include conformational transformations, are
quite common in rare gas matrices the cage effect reduces the probability of potentially competing dissociation processes but often does not impede isomerization. The barriers to isomerizations span a wide range from less than 1 kcal/mole for some conformational changes to tens
of kcal/mole for valence isomerizations. Therefore, in contrast with dissociations, thermal excitation may induce an isomerization without necessarily damaging the matrix. This provides an
opportunity for observing subtle effects in reaction patterns and for the identification of exotic
reaction products. The prospects of selective excitation were one of the major driving forces in
this field and led to extensive efforts to use monochromatic radiation (mostly infrared lasers) for
the initiation of matrix isomerization. The ever present vibrational relaxation due to coupling to
matrix phonons is an efficient energy dissipation process. As a result, even seemingly small
changes in the reaction rate may significantly affect yields or branching ratios.
These properties of the isomerization reactions make them a most suitable test ground
for observing matrix or site effects on chemical reactivity. As stated in Section 12, we restrict
the term site effects to effects assignable to the coexistence of two or more well defined sites in
a given matrix. Matrix effects are defined as changes observed in the kinetic behavior upon
switching from one matrix material to another, or at times, as specific kinetic effects noted in a
matrix and not in other phases.
In this section we discuss several cases where matrix or site effects appear to be important in isomerization reactions, highlighting some unique possibilities of this medium.

247

248

Y. Haas and U. Samuni

332 Matrix effects


3321 Tunneling
One of the simplest unimolecular reactions is intramolecular hydrogen transfer along a hydrogen bond. At low temperatures, when thermal activation is prohibited, it often proceeds by a
tunneling mechanism. Tunneling rates may be readily and accurately measured spectroscopically from the tunnel splitting providing, in principle, a sensitive means to check matrix effects.
Intramolecular hydrogen transfer reactions are an example of reactions for which tunneling was
often observed; its rate may be expected to be affected by the nature of the matrix provided the
transferred atom is not well shielded from the environment.
In a system such as malonaldehyde shown in Scheme 4a, the potential surface along the
reaction coordinate is characterized by a double minimum due to two degenerate structures.
When a molecule is embedded in a matrix, the symmetry of the system is generally reduced,
possibly affecting the observed tunnel splitting. This provides a simple gauge for the matrix
effect, and the relative importance of intra- vs intermolecular interactions. It was found that
whereas for 9-hydroxyphenalone (9-HP, Scheme 4b), the tunnel splitting was hardly changed
when the molecule was transferred from the gas phase to a neon or argon matrix [Rosetti 1980,
Rosetti 1981, Bondybey 1984], a dramatically different situation was encountered with the
smaller malonaldehyde: no tunnel splitting was observed in an argon matrix in the 1530 K
temperature range while the gas phase value is 21 cm1 [Firth 1989].
The result was interpreted as indicating a strong localizing effect by the argon matrix,
effectively stabilizing one of the two hydrogen atoms structures that are equivalent in the gas
phase (Scheme 4). The different stabilization was proposed to be due to local interactions within the trapping site (local in the sense that they originate primarily from argon atoms in the close
vicinity of the guest molecule) but no details were offered. Elucidation of the nature of these
interactions remains a challenge for future simulations. The preservation of the tunnel splitting
in an argon matrix in the case of 9-HP was speculated to be due to effective shielding of the
transferred proton from the nearby argon atoms.

Scheme 4

Reactions in rare gas matrices matrix and site effects

3322 Thermal and infrared induced conformational isomerization


A strong matrix effect was also observed in conformational isomerization reactions. An example that could be expressed in quantitative terms is the interconversion of the cis and gauche
forms of 2,3-difluoropropene (DFP, [Knudsen 1991]) in several matrices at 12 K using IR laser
excitation. The quantum yield was found to strongly depend on the matrix material used, being
as low as 107 for N2, as compared to 105 for argon and 104 for krypton. This result is by no
means unique a similar effect was observed for many other molecules. Isomerization rates for
HONO [Shirk 1983], nitroethanol (CH2OHCH2NO2, [Rsanen 1983]), and allyl alcohol
(CH2=CHCH2OH) were reported to depend strongly on the host matrix. Possible causes such as
a matrix dependent barrier [Felder 1984], or different vibrational relaxation efficiencies
[Rsanen 1983] were suggested. The much reduced reactivity in a nitrogen matrix may be due

249

250

Y. Haas and U. Samuni

to stronger interaction with the matrix resulting in a more efficient coupling to lattice modes.
This is another manifestation of the active nature of nitrogen matrices (Section 15).
A further observation was that the IR induced isomerization of 2,3-DFP appeared to proceed more readily upon excitation of the CF stretch mode (near 1080 cm1) than of the higher
energy C=C stretch mode. It was speculated [Knudsen 1991] that the former is better coupled to
the torsional reaction coordinate but the issue is still open, as is the cause of the strong matrix
effect on the quantum yield. Apparent mode selective dependence of reaction rates was also
found in other systems: one of the early and much debated examples is the cis-trans isomerization of HONO. This was the first IR-induced isomerization reported in a matrix
[Baldeschwielder 1960], that raised much interest, particularly in the context of mode selective
chemistry [Shirk 1983]. In a small molecule such as nitrous acid, the intramolecular redistribution of vibrational energy is expected to be slow relative to the coupling between certain modes
and the reaction coordinate, (torsion around the NO bond). Should the efficiency of this coupling be mode dependent, a mode selective rate enhancement may be observed. Indeed, Shirk
and co-workers [Shirk 1983], using a tunable IR laser, reported that excitation of the 22 mode
(N=O stretch) induced isomerization much less efficiently than excitation of the 1 (OH stretch)
that is only 200 cm1 higher in energy (factor of 20 for the cistrans, and factor of 100 for the
reverse reaction). As will be discussed in Section 521, mode selectivity in this case can be
accounted for by molecular dynamics simulations.
3323 Photochemical isomerizations
The subtle effects expected to arise from the unique properties of a low temperature matrix are
likely to be less important when irradiation is in the UV/vis region, involving a much larger
energy input. Energetically, dissociation can take place as discussed in Section 321, but irradiation at this range leads often to isomerization, sometimes with unexpected results (expectations being based on the conventional reaction methods). Selective optical excitation is in general impractical for photoreactive molecules, as absorption bands are broader than site splitting.
However, matrix and perhaps even site effects may still be observable, provided reaction parameters, such as barrier heights, are site dependent. A recent example is provided by the rich
chemistry of the (ClO)2 system [Jacobs 1994].
Chlorine oxides have drawn much attention in the last few years due to their role in the
depletion of the ozone layer [Molina 1996], particularly in polar regions. Dimeric forms of ClO
were proposed as possible important intermediates [Molina 1987]. Matrix isolation is the method
of choice to study such systems, as several different isomeric forms can be prepared, isolated and
characterized in the same medium. Furthermore, irradiation of these species in the gas phase
normally leads to dissociation which is arrested in matrices. Rare gas solids thus provide an
experimentally accessible model for aerosol and ice particle environments which are of potential
interest in polar atmospheric chemistry. We shall mention only a small number of studies, that
relate to the central theme of this review, emphasizing the effect of the matrix environment.

Reactions in rare gas matrices matrix and site effects

Scheme 5
Three stable (ClO)2 isomers are predicted by ab initio calculations [McGrath 1988, Lee
1992, Luke 1995]: In order of decreasing stability they are: dichlorine peroxide ClOOCl,
chloryl chloride ClClO2 and chlorine chlorite ClOClO. Matrix isolation spectroscopic evidence for these species was obtained as early as 1967 [Rochkind 1967, Alcock 1968] but definitive infrared assignments of all three isomers were reported only a few year ago [Jacobs 1994].
ClClO2 was prepared in the gas phase by the reaction between FClO2 and BCl3, and immediately trapped in either a neon or an argon matrix. Visible light irradiation of this compound gave

251

252

Y. Haas and U. Samuni

rise to instantaneous formation of ClOClO in argon, indicating that this product is not formed
via a long-lived intermediate. In neon, no photoisomerization product was observed. The reactivity difference between the two matrices was assigned to different sizes of the matrix cages,
although no attempt was made to define better the trapping sites. Irradiation of ClClO2 in neon
resulted in slow decrease of its concentration without formation of any new infrared active
products. This was taken to indicate that the only important products are Cl2 and O2 which are
infrared inactive.
The mechanism of the reactions is not yet fully understood. It was proposed (see
Scheme 5) that in both matrices the first step is the dissociation of the ClCl bond. The different cage size then determines the reaction pattern: in the smaller neon cage, recombination dominates and, occasionally, the Cl atom which cannot move away abstracts a chlorine atom from
the ClO2 radical. Since ClO2 does not photoisomerize to OClO, the reaction can take place by a
one step mechanism in which the two ClO bonds elongate together (dashed lines in b1), the
OClO angle decreasing as the two diatomic molecules form. Another alternative is a two step
mechanism shown in b2. One Cl-O bond elongates and the OClO angle decreases, and the
molecule decomposes to Cl2+O2 product via a three center intermediate. This implies that the
rotational motion of the product OClO is severely hindered in the neon matrix, preventing
the isomerization to ClOOCl, leading to the observed matrix effect.
The cage effect is manifested also by the relatively high photon energies required to
initiate the reaction in a matrix as compared to the gas phase. Table 4 compares the experimentally determined minimum photon energies required for (ClO)2 photolysis in an argon matrix
with the calculated gas phase dissociation energies.
This result is reminiscent of the large excess energy required for the photodissociation
of small molecules such as H2O, HCl, F2 and Cl2 in matrices, Section 321 [Schriever 1990,
Schriever 1989, Feld 1990, Alimi 1991]. However, cage escape for a species as large as ClO is
extremely unlikely. Another possible cause of the large excess energy needed in the present case
is related to the fact that the fragments tend to recombine unless the cage is rearranged to accept
them. This rearrangement can be achieved by collisions with the cages wall, if the energy available is large enough.
The matrix photochemistry of XOClO (X= Cl, Br or I) was also explored by Johnsson

Table 4. Gas phase dissociation energies (Ediss) of some (ClO)2 isomers and the minimum photon energy (Ephoton)
required in an argon matrix to observe the reaction.
Process

Ediss(kJ/mole)
(Calculated, gas)

Ephoton(kJ/mole)
(Experimental, matrix)

ClOClO 2ClO

25

180

ClClO2 Cl+ OClO

80

198

ClOOCl 2ClO

70

332

Reactions in rare gas matrices matrix and site effects

et al. [Johnsson 1995]. These compounds were prepared in the gas phase by the addition of
X to OClO followed by trapping in an argon matrix; the authors admit that it is not clear why
the more stable Y shaped molecule XClO2 is not formed, but offer no mechanistic explanation.
All three molecules easily rearrange to the Y isomer upon visible irradiation; UV irradiation
causes re-formation of the XOClO isomers for Cl and Br, while in the case of iodine the
thermodynamically stable ICl+O2 pair is formed. These examples suggest that the matrix plays
a major role in determining the results of photochemical transformations, in which the products
geometry is different than the gas phase one.
A further example for the formation of a less stable isomer in a matrix is found in the
reaction between H atoms and OClO [Johnsson 1996]. Both HOClO and HClO2 are formed,
even though the latter is calculated to be less stable by 443 kcal/mole. Evidently the reactions
are not controlled by thermodynamic stability of the final product. The authors note that the
spherically symmetric H atom (2S1/2) does not discriminate between the two possible reaction
channels, whereas the halogen (2P3/2) and O (3P3/2) atoms prefer exclusively to attack the O atom
side.
A conceptually similar example concerns the photochemistry of benzene in rare gas
matrices. Photoisomerization of benzene is one of the classical examples of organic photochemistry [Gilbert 1991]. In solution, benzvalene (I) and fulvene (II) are produced upon irradiation at 2537 nm, namely at the S0S1 transition. Dewar benzene (III) can also be formed but

was previously reported to be produced only at shorter wavelength, 204 nm (the S0S2 transition). Benzene was photolysed [Johnstone 1991] in an argon matrix (1:1000) at 2537 nm at
42K. IR bands due to benzvalene, fulvene and Dewar benzene were observed after six hours.
In a xenon matrix, no products were observed. The production of benzvalene and fulvene is
expected, but that of Dewar benzene is in contradiction with previous work. A triplet state
mechanism is ruled out by the absence of a reaction from the xenon matrix. The authors propose
that Dewar benzene may be formed due to a reduction of the symmetry in the trapping site from
D6h to a lower one, in which the S1 state (B2u in D6h) can mix with S2 (B1u in D6h). It follows that
the matrix can provide new photochemical routes, not found in other environments. Evidence
for the existence of different trapping sites for benzene in argon is available from a previous
study of phosphorescence lifetimes [Johnson 1972]. Further work on this interesting proposition

253

254

Y. Haas and U. Samuni

is highly desirable, for instance, correlation of the data with IR spectroscopy and theoretical
simulations.
An analogous reaction was observed for pyridine [Johnstone 1991], but in this case the
product bands were actually enhanced when argon was replaced by xenon. Thus, the reaction
was believed to be triplet induced. No firm identification of the product(s) was proposed but it
was suggested that the reaction is isomerization.
In summary, site effects in UV induced photoisomerization are, in principle, possible,
since the reaction pattern appears to depend strongly on the intimate environment of the reacting species. The large energy excess usually accompanying this excitation mode tends to conceal the subtle reactivity differences that may be expected between different sites, and local
annealing may wipe them out altogether. Milder excitation modes are by infrared irradiation or
gentle heating by which the energy input to the system may be controlled to meet closely the
necessary minimum. Some attempts in that direction are discussed in the next subsection.
34
Site effects in IR induced isomerizations
Since different trapping sites are easily distinguished through infrared absorption spectroscopy,
and, since isomerizations in rare gas matrices are quite common, it appears that site specific
reactions in matrices can be induced by IR absorption. Several cases in which site effects are
indicated have been discussed in the literature, though in most cases the actual site structure is
not yet known and no attempt to establish it was reported. We begin with cases where site
effects are only hinted at and conclude with an apparently clear experimental demonstration of
a kinetic effect assignable to distinct spectroscopically identified sites.

Scheme 6.
The first example deals with hydrogen bonded structures that are of great current interest,
and are being studied in the gas phase [Pugliano 1992, Huisken 1991] as well as in matrices.
Methanol, as water, spontaneously forms oligomers upon cooling. The condensation of
methanol/nitrogen mixtures at 15 K [Coussan 1994] yields IR absorption bands in the OH
stretch region (32003600 cm1) that are assignable to hydrogen bonded oligomers, among

Reactions in rare gas matrices matrix and site effects

them the cyclic methanol trimer. It was found that the most stable cyclic trimer (peak absorption
at 3450 cm1) was converted by IR absorption (using the globar source of the spectrometer) to
the less stable open structure (see Scheme 6). The opening of the ring was deduced from the
spectral changes in the OH and CO stretch regions and supported by a force constant calculation and by kinetic analysis. Increasing the temperature led to disappearance of the new absorption bands and recovery of the original spectrum. Inspection showed that two sets of open trimer
bands could be discriminated based on their temperature behavior. The main (more intense) set
progressively decreased in intensity in the temperature range of 1015 K, and a minor set,
whose bands vanished completely for a temperature rise of only 2 K. The latter was assigned to
a different form of the open trimer, although no detailed structural forms were offered. Based on
the accumulated evidence discussed in this review, a likely candidate for the species giving rise
to the minor band set is an open trimer trapped in a different site than the major one. The
extreme temperature sensitivity indicates a smaller barrier for ring closure. A possible cause for
the different barrier heights for the two sites is the different hindrance to the opening of the ring
by neighboring nitrogen molecules
341 Site specific rates in the isomerization of 1,2-difluoroethane
One example that appears to fulfil unambiguously the requirements of site selective reactions is
the conformational isomerization of 1,2-difluoroethane (DFE) in an argon matrix by Benderskii
and Wight [Benderskii 1996]. The reaction, which was studied earlier experimentally [Gunde
1982] and theoretically [Raff 1990], was mentioned in relation to the simulation of the site
structure in Section 22.
Of the two possible rotamers, the gauche conformer is more stable than the trans one by
335 kJ/mole [Durig 1992] and dominates the IR spectrum in solid argon under normal deposition conditions. Heating the gas mixture to 800C prior to its deposition [Huber-Wlchli 1975]
raises the trans conformer population sufficiently to yield a good quality spectrum of this
species as well. IR radiation was found readily to cause rotamer inter-conversion [Gunde 1982]
in the matrix.
Benderskii and Wight [Benderskii 1996] found an ingenious way of preparing the molecule in either a DS or SS site. Both rotamers can be prepared in the matrix by co-deposition of
ethylene and F2, followed by UV irradiation. Under these conditions, in addition to the most
stable trapping site (presumably DS site) another one is populated. This less stable site, not
observed before, was proposed to be a SS site [Misochko 1996], which is formed in the following way: F atoms produced in the low temperature photolysis of F2 become mobile at 26 K [Feld
1990], and diffuse over to the SS trapping site of ethylene, reacting with it to form initially the
fluoroethyl (CH2CH2F) radical. This radical, still in a SS site, reacts with another F atom to form
the SS trapped DFE. It was suggested that both rotamers are formed in this fashion. This mechanism is supported by kinetic evidence (a sigmoidal growth curve) and by the blue shift of the
bands assigned to the SS trapped species compared to the DS trapped ones. Figure 11 shows

255

256

Y. Haas and U. Samuni

Figure 11. Part of the infrared spectrum of 1,2-difluoroethane formed in a single substitutional site (marked (1))
and a disubstitutional site (2). Both gauche (g) and trans (t) conformers are present [Benderskii 1996].

part of the IR spectrum of the system prepared in this way. It is seen that in this case the site
effect can be easily discerned from the conformer splitting.
The transgauche isomerization reaction was initiated thermally, and was found to
occur over a narrow temperature range (3036 K for the DS site and 3439 K for the SS one).
The activation energy in the DS site was found to be 97 kJ/mole similar to the gas phase value
of 89 kJ/mole [Durig 1992]. For the SS site a considerably higher barrier was measured
at 15 kJ/mole.
The properties of the system were simulated using a fcc cluster model, with the potentials used by Raff [Raff 1987, Raff 1991]; according to the model the SS site is only 2 kJ/mole
higher in energy than the DS site, whereas a trisubstitional or interstitial sites are much higher
in energy. This accounts for the experimental observation of only two sites. The MD simulated
barrier to isomerization was 162 and 126 kJ/mole for the SS and DS sites, respectively, for

Reactions in rare gas matrices matrix and site effects

fixed argon atom positions. When the matrix atoms were allowed to relax during the reaction to
the nearest local minimum, the barriers obtained were practically the same for the two trapping
sites. At the experimentally used temperature range, some matrix atoms relaxation is expected
during the reaction. Therefore, the reaction barrier difference found in the simulation cannot be
taken as a conclusive explanation for the experimentally observed one.
The reaction is site specific in the sense that a different barrier is found for different
trapping sites in a given reaction. A site dependent reaction path has not yet been reported to our
knowledge. This remains a challenging prospect for future work in the field.
In summary, isomerizations are very common in RGS. Subtle and special effects that
are often masked in liquid solutions, such as mode selectivity, unusual products and site specific
rate constants are detectable in this medium. These effects can be traced to the much more rigid
environment provided by the matrix (as compared to liquid solvents). Rigid and semi-rigid
environments are of cardinal importance in molecular biology, heterogeneous systems and solid
state chemistry; the level of detail attainable in the RGS systems, together with the more accurate theoretical simulation, make these systems ideal test grounds for novel ideas.
4.
BIMOLECULAR REACTIONS
Having repeatedly stated that translational motion is inhibited in matrices it is expected that collision induced bimolecular reactions are rarely observed. The relative large number of bimolecular reactions reported in the matrix literature has been generally accounted for by assuming
that they take place between neighboring reactants, formed during deposition. However, diffusion controlled reactions cannot be ruled out; as we shall see in Section 421, some experimental data have been interpreted by assuming that even molecules as big as ammonia may diffuse
in an argon matrix during annealing. New methods have been used in recent years to study
quantitatively diffusion in matrices as will be briefly summarized in this section. Some deductions from these studies have already been used in the interpretation of matrix reactions, as
discussed in Sections 341 and 51.
41
Diffusion and reactions of migrating atoms in matrices
Although thermal diffusion in rare gas solids, be they crystals or matrices, is usually negligibly
slow, the initial kinetic energy of small fragments formed by photodissociation can be quite
large and, if they are properly aligned with respect to a window in the cage, some small atoms
are found to diffuse away relatively rapidly.
Significant experimental progress in this field has recently been made by several
research groups, notably Weitzs and Apkarians. The results, along with theoretical interpretations, have been used to explain some unique properties of bimolecular reactions in matrices, in
which atoms (particularly H, F and O) participate. Only a brief survey of the main results will
be given, and for a more extensive coverage, the reader is referred to the original papers: Feld
1990, Kizer 1995 (F in Ar), Kunttu 1990, Alimi 1991, Krylov 1994 (F in crystalline Kr),

257

258

Y. Haas and U. Samuni

Krueger 1992, Ryan 1993, Ryan 1993a, Ryan 1994, Danilychev 1993 (O in Xe and Kr),
LaBrake 1993 (H in Xe), LaBrake 1995 (Br in Xe).
H, F and O atoms can be easily detected in rare gas matrices by the characteristic exciplex emission mentioned above ([Fajardo 1986, LaBrake 1995], see Section 431). O(1D)
atoms are cleanly produced by irradiating a N2O doped xenon matrix at 193 nm [Ryan 1993,
Ryan 1994]. They rapidly relax to the ground 3P state and can be trapped in the matrix for a long
time (presumably as a charge transfer complex with xenon as the donor). Significant amounts
up to 8 103M were reported to be semi-permanently stored at 15 K [Ryan 1993]. Their relative concentration can be followed by laser induced fluorescence (LIF), excited by a weak
probe laser at 193 or 248 nm. The LIF signal was stable for several hours or even days (depending on the matrix temperature). A temperature dependent slow decrease of the signal was
assigned to diffusion controlled recombination of O atoms; however, attempts to account quantitatively account for the data with a single diffusion coefficient were unsuccessful.
This result may be explained by assuming that several different types of O atoms are
trapped in the matrix, for instance in different sites. Indeed, it was found that two different diffusion coefficients were sufficient for a satisfactory reproduction of the data. The values derived
for the diffusion constants (14 1015 and 54 1018 cm2/s at 32 K, and 73 1015 and 20
1017 cm2/s at 40 K in xenon [Krueger 1992]) appear to be reasonable. However, the usual diffusion controlled mechanism [Smoluchowski 1917], which was developed for liquid solutions,
is not applicable. Instead, it was proposed that, as in other non-stirred systems, it may required
the application of fractal kinetics. That involves time dependent diffusion coefficients [see e.g.
Kopelman 1988] of the form:
k(t) = k0t

(7)

Where is a constant. In such systems, atoms that are close to each other react with a larger rate
constant than those that are further away. If this mechanism will be found to apply in rare gas
solids, they may be used as model systems for other non-stirred systems, such as biological
membranes, or surface reactions.
A further interesting finding is that highly mobile O atoms are formed upon photolysis
of a xenon matrix containing stored oxygen atoms. These mobile atoms are calculated to
have a diffusion coefficient of about 84 107 cm2/s, almost as high as in normal liquids
(105 cm2/s). A possible explanation is a local melting of the region near the trapped atom site.
These mobile atoms are probably produced from the photolysis of XeO molecules, that are
formed in the trapping sites of O atoms.
Danilychev and Apkarian [Danilychev 1993] studied the formation, mobility and
recombination of O atoms in krypton and xenon crystals. O atoms were formed by photolysis of
N2O as above, or of O2, which is followed by recombinant emission: apparently, O atoms
formed this way (O(3P)) cannot escape the cage and tend to recombine to form electronically

Reactions in rare gas matrices matrix and site effects

259

excited O2 in several states: (b(1+g), A(3u) and c(1u)). The recombinant emission is temperature dependent and appears to indicate the existence of several distinct trapping sites in free
standing crystals. No attempt was made to further characterize these sites.
When O atoms were formed from N2O, yielding (O(1D)), they readily migrated away
from the initial site. It was found that even in very dilute crystals (1:30,000), thermoluminescence of O2 could be observed upon warming the Xe crystal to 374 K. The emission decayed
with time due to depletion of the O atoms obeying first order decay kinetics. This seemingly surprising result was accounted for in the following way. O atoms isolated in the lattice are indefinitely stable with respect to recombination at 10 K due to the existence of a barrier (see below
for its nature). Warming the matrix imparts mobility to the atoms, designated as Omob (equation
8), by surmounting the barrier In a crystal, any barrier appears periodically, so that the mobile
atom may be retrapped after losing energy to the lattice Eq. [9]
O Omob

k1

(8)

Omob O

k1

(9)

Since at any given time only a small fraction of the atoms are mobile, the encounter of two
mobile ones is extremely unlikely, and recombination is due to a reaction between a mobile
atom and a trapped one Eq. [10]:
O + Omob O2

k2

(10)

Assuming a steady state for Omob, whose concentration at any time is much smaller than that of
O, equations 810 lead to the rate expression (Eq. 11) for the disappearance rate of the O atoms:
d[O]/dt= k1k1[O]2/(k2[O]+k1)

(11)

This will lead to pseudo first order kinetics if k2[O] >> k1. This inequality is equivalent to the
requirement that the volume swept by a mobile atom prior to trapping (k2/k1) will be much larger than the volume occupied by the stationary atoms 1/[O]. In a 1:30,000 matrix, even if all
molecules are dissociated, this implies a volume of 1018 cm3, or a migration range of at least
100 . Considering the approximations made, the authors conclude that a range of 300 is
a realistic estimate for the average distance traveled by the thermally activated O atom in a
krypton or xenon crystal!
The physical origin of this very long migration range is not quite clear. The authors
assume that the O atoms are trapped in their ground state (3P), for which the barriers for cage
escape are calculated from pairwise potentials to be 14 and 125 eV in Kr and Xe, respectively.
These are more than an order of magnitude larger than the experimentally observed barriers.

260

Y. Haas and U. Samuni

The most plausible mechanism calls for intersystem crossing to the singlet surface, forming
O(1D). Ab initio calculations [Dunning 1977] show that while the main interaction between
O(3P) and rare gases is repulsive, O(1D) forms one bound state (1+) and two repulsive states. In
the bound state, the internuclear distances are much smaller than in the equilibrium position of
O(3P), so that the oxygen atom can be considered as shrinking (by as much as 15 ) upon
becoming O(1D). Therefore O(1D) is expected to be much more mobile than ground state oxygen. The barrier for the O(1P) O(1D) conversion is not given, but it is assumed that it is low
in view of the strong charge transfer character of the RGO (1+) state and the strong spinorbit
coupling in the heavy atom environments. These qualitative considerations were checked further by considering adiabatic potential energy surfaces, based on angular anisotropic pair interaction known from the gas phase [Danilychev 1994]. Some properties of the system could be
reproduced by this model, for instance, the O(1S)O(1D) emission and the charge transfer
spectra. It was also found that the singlet path is more likely to be involved in the thermal diffusion, but the barrier for the triplet to singlet crossing was calculated to be an order of magnitude higher than the observed one. Clearly, a more detailed model needs to be worked out.
Very recently, a more direct measurement of the average distance traveled by photogenerated F atoms was reported [Schwentner 1996]. In this experiment, a sandwich configuration was used: three layers were sequentially deposited. The bottom was a pure krypton layer,
used for detection; the middle one was pure argon, serving as a spacer of variable thickness d
and the top one F2 doped argon. The fluorine molecule was photodissociated with an excess
energy of 85 eV. The F atoms, having a kinetic energy of 425 eV, began to travel, some
traversed the spacer layer, reaching the krypton layer. Exciplex emission of Kr2F immediately
followed, and served to estimate the number of F atoms reaching the krypton layer. It was found
that the exciplex emission approached an asymptotic maximum value as a function of the radiation dose. This value was found, in turn, to decrease exponentially with the spacer layers
thickness, d:
I = I0exp(-d/)

(12)

The parameter , representing the mean penetration depth of the photo-generated F atoms, was
found to be 28 nm. This is a direct confirmation of the long migration range of atoms moving
in a matrix, although it is smaller than deduced by the less direct methods which may have
actually probed the maximum value.
In summary, some photogenerated small atoms are found to undergo relatively long
range diffusion, enabling them to be involved in bimolecular reactions in RGS. Thermal diffusion is slow, but one must recall that trapping of photogenerated atoms by a charge transfer
adduct with the rare gas atoms creates a reservoir of potent reactants. Subsequent irradiation,
even by room lights, may induce photodecomposition and produce highly mobile atoms.
Therefore, any thermal diffusion experiments must be conducted in the dark [LaBrake 1995]

Reactions in rare gas matrices matrix and site effects

and the history of the matrix must be taken into account when using potentially mobile atoms
precursors. In contrast, there is no evidence for long range diffusion (or migration) of heavier
species such as Cl and Br atoms, or of molecules. This result is at odds with reports of an apparent long range motion of some molecules upon annealing (see Section 421).
42
Reactions between neighbors site and matrix effects
Rather than depend on diffusion, bimolecular reactions may be made possible by co-depositing
the two reactants: the resultant matrix is likely to contain sites in which the two reactants are
close to each other. This procedure could in principle result in a thermal reaction between the
gaseous reactants during the deposition process. However, by a proper choice of the deposition
temperature, the concentrations and the deposition rate, negligible amounts of products can be
formed during the deposition. After deposition, a reaction can be initiated thermally for low
barrier processes, (e.g. at 1020 K, no products are expected over a reasonable laboratory time
scale, for reactions whose gas phase activation energies exceed ~10 kJ/mole) and photochemically for any reaction.
An intriguing aspect of the matrix effects is that for some inert hosts, reactions take
place even though the gas phase activation energy is high. An early example is the oxidation of
NO by ozone in a N2 matrix [Pimentel 1978]: the product (NO2) was observed to form in solid
N2 in the temperature range of 1220 K. It is noted that the matrix was deposited at 12 K
throughout in all experiments under these conditions the matrix is of a rather open structure
and may allow slow diffusion. The interpretation offered was that the reaction is controlled in
the cage by an orientational activation barrier which opens an atom transfer reaction channel, not
observable under normal high temperature conditions. This could be due, for instance to a highly
angular dependent activation energy for the reaction. In the gas phase, the particular orientation
favored by the matrix is statistically unlikely and is therefore overwhelmed by other channels.
An apparently similar situation was encountered for the oxidation of ethylene by ozone
in matrices [Kohlmiller 1981, Hawkins 1982, Nelander 1977, Samuni 1996]. In argon, clear
evidence for the formation of an ethylene/ozone adduct is obtained by detecting a new IR
absorption peak. Nonetheless, no reaction is taking place up to the evaporation temperature of
the matrix (~40 K). In a CO2 matrix deposited at a high temperature (65 K), no reaction occurs
below 77K. However, when the CO2 matrix is deposited at low temperatures (15 K), a reaction
is observed at a temperature as low as 26 K. The matrix in this case is much more porous than
the high temperature deposit, and presumably the two reactants can approach each other. Yet the
gas phase activation energy is about 20 kJ/mole, so a reaction should be arrested below 6070 K.
Here again, some matrix specific orientation may make an otherwise improbable reaction route
more likely than those that dominate at high temperatures.
Another plausible explanation for the ethylene ozonation reaction in the low temperature deposited CO2 is related to the amorphous nature of the CO2 matrix. The amorphous CO2
matrix undergoes a phase transition which begins at 26 K and converts the amorphous solid to

261

262

Y. Haas and U. Samuni

a crystalline one. The heat released in the process may be the main energy source for the
observed reaction. The ozonolysis of ethylene in solid CO2 may thus be an example of a reaction driven by an exothermic solid state phase change.
421 Proton transfer: the naphthol/ammonia system
Kelley and coworkers [Brucker 1989, Brucker 1989a, Swinney 1990] studied the proton transfer reaction from electronically excited naphthol to ammonia. These papers report that excited
state proton transfer (ESPT) in these systems takes place only if several ammonia molecules are
in close contact with the naphthol molecule. In the gas phase, the reaction was studied by forming clusters of the type naphthol/(NH3)n in supersonic jets; and proton transfer took place only
for n4 [Cheshenovsky 1988, Droz 1990], while in an argon matrix only three ammonia molecules were required. In both cases, it was assumed that the ammonia molecules are bound to the
OH group by hydrogen bonds. That ESPT took place was deduced from the change in the emission spectrum observed as the cluster size was varied. The need to have several ammonia molecules near the proton donor is accounted for by energy considerations in the smaller clusters
the process is endothermic.
According to picosecond measurements ESPT is a rapid process its average duration
being 22 and 24 ps, for the and -naphthol isomers, respectively. The tunneling mechanism
which was proposed to account for the reaction rate, was supported by deuteration experiments:
no ESPT was observed on deuteration, in agreement with the calculated rate which was 200
times smaller [Brucker 1989, Brucker 1989, Swinny 90]. In the matrix a spectral feature
assigned to a tri-solvate species that is ESPT inactive is also reported the authors assume that
its structure is different from that of the active one (one ammonia interacting with the aromatic
ring, rather than with the OH group).
A remarkable feature of the matrix work is the fact that very few ammonia/naphthol
clusters are observed right after deposition at 10 K. Annealing is required to produce the fluorescence excitation and emission peaks assigned to the ESPT products. The authors suggest that
even at the lowest annealing temperature (24 K) new ammonia/naphthol clusters are formed by
diffusion of ammonia.
The data were simulated by a model assuming that ammonia monomers can diffuse.
The model was found to apply only for low annealing temperatures (2426 K), and only for
samples that were annealed a few minutes. The authors propose that discrepancies between the
models predictions and experimental results at higher annealing temperatures or longer times
are due to dimer diffusion, that sets in under these conditions. The simulation predicts that only
tri-solvates undergo ESPT, while clusters with 4 or 5 NH3 molecules do not. The authors propose that with larger ammonia aggregates, a charge transfer may take place even in the ground
state.
An interesting aspect of these experiments is that their interpretation depends on the
diffusion of a fairly large species an ammonia molecule and even ammonia dimers. It is noted

Reactions in rare gas matrices matrix and site effects

that deposition was carried out at 10 K, a temperature that is known to lead to a porous structure.
The authors do not comment on this aspect nor do they disclose whether deposition at higher
temperatures, in which a more compact matrix is formed, yielded different results. In view of
the present knowledge concerning diffusion, long range diffusion of ammonia in just a few
minutes appears highly unlikely. An alternative possibility is that ammonia molecules were actually
deposited nearby the naphthol site, but not in the appropriate configuration for ESPT. The main
effect of annealing would then be to restructure the trapping site to the reactive configuration.
422 Oxidation of olefins by NO2
This topic was thoroughly studied by Frei, Nakata, Shibuya and coworkers. The results were
summarized in a series of papers [Harrison 1994, Tanaka 93, Tanaka 95, Nakata 1989a, Nakata
1989b, Nakata 1992, Fitzmaurice 1992] and reviews [Frei 1991, Frei 1992]. It is an example of
a controlled reaction which is apparently possible only in low temperature matrices: NO2 is a
strong oxidizing agent, and can be used to oxidize olefins to the corresponding acids no intermediate oxidation products are observed at room temperature. Illumination with visible or near
UV light enhances the reactions rate, since the electronic absorption spectrum of NO2 extends
over the entire UV and almost all the visible range. The dissociation threshold (to NO+O) is at
398 nm providing in principle an easy way of comparing oxidation by an electronically excited
molecule with that initiated by nascent oxygen atoms.
It was found that in rare gas matrices, the reaction proceeds under visible light illumination but leads to an epoxide product being much milder than the room temperature reaction
that yields carbonyl compounds or even complete oxidation to CO2 and water. For instance,
trans or cis 2-butene were found to form adduct pairs with NO2 and to yield 2,3-epoxy butane.
[Nakata 1989a, Nakata 1989b]. Moreover, this was the only final product observed upon direct
photolysis. The reaction was stereo selective: in the case of the trans compound, complete stereochemical retention was found, while for the cis 70% of the product epoxide retained the initial
configuration. The proposed reaction mechanism for the cis isomer is illustrated in Scheme 7.
Following electronic excitation, an oxygen atom is transferred within the olefin/NO2 pair, to
form a biradical intermediate, which is paired with the NO fragment. This biradical can be
either cis (BRc) or trans (BRt) and leads to the final epoxide by ring closure. A side reaction was
the formation of a butyl nitrite (Ic or It) which was also detected spectroscopically. Further irradiation led to the scission of the O-NO bond (via an electronically excited intermediate indicated
by an asterisk in the scheme) leading to either a carbonyl compound or to the epoxide.
The key factor permitting the observed high stereoselectivity is the availability of a spin
conserving path for the epoxidation by NO2(2B2), a doublet state. This makes ring closure competitive with conformational scrambling of the oxirane biradical intermediate. It is also noted
that photodissociation of both cis and trans butyl nitrites is completely stereoselective.
The higher stereoselectivity of the trans butene was attributed to the fact that O atom
transfer to trans-2-butene does not require a change of the dihedral angle about the central CC

263

264

Y. Haas and U. Samuni

Scheme 7.
bond of more than 10200 to form the oxirane biradical intermediate in its most stable form. In
contrast, O atom transfer to the cis isomer without angle change would produce the radical in a
conformation corresponding to a maximum along the torsional angle coordinate. This in turn
requires a large amount of energy to be put in that coordinate, allowing isomerization to take
place.
The method was thus successful in obtaining selective, mild controlled oxidation. It has
not proved yet to be of practical value, since the products converted to the usual oxidation product
upon warming the matrix.
43
Long range interactions
431 Electron transfer and exciplex formation
Electron transfer is one of the most extensively studied processes in liquid solutions and the
gas phase. Apart from energy requirements that need to be satisfied, one often considers the

Reactions in rare gas matrices matrix and site effects

265

importance of orientation, as was demonstrated with rigidly linked donoracceptor pairs [Closs
1986, Speiser 1996]. The rigid environment of a matrix offers in principle an opportunity to
study such effects, but at present, only a small number of papers have been published on the
subject. A major difficulty lies in the unambiguous structural determination of the donoracceptor adduct in the matrix. Even the stoichiometry of the charged species formed upon electron
transfer is not always well established, as discussed in the next subsection.
4311 Photogeneration of ionic species of the form (RG)nH+
An interesting offshoot of the detailed studies of halogen molecules and halogen halides photoexcitation in RGS is the discovery of new rare gas compounds, such as hydrides [Kunttu 1994,
Pettersson 1995, Pettersson 1995a]. The isolation properties of rare gas matrices make them a
suitable medium for storing not only neutrals, but also ions. It has been found quite early on
[Bondybey 1972, Milligan 1973] that species such as (RG)2H+ (RG=Ar, Kr) can be prepared by
depositing gases that were subjected to electric discharge, or VUV irradiation. However, it is
now known, that these species (as well as the corresponding xenon one) can be prepared under
much milder conditions [Kunttu 1994]. When a matrix containing a halogen halide is irradiated
at 193 nm (64 eV) or even 248 nm (498 eV), the characteristic absorption of (RG)2H+ appears.
The reaction proceeds readily with conventional lamps (no lasers are required), excluding the
involvement of non linear processes. The proposed mechanism (Eqs. [1316]) does involve the
absorption of two photons in a consecutive manner: one photon photo-dissociates HX, the halogen atoms form a charge transfer type RG+X compound (Section 321), which absorbs another photon to yield a charge separated ion pair (with possible delocalization, see Section 41 for
the discussion of mobility). Trapping of the loose H atom formed in the first step (Eq. [13])
completes the synthesis.
HX+h H + X

(13)

X + RG {X...RG}nRG

(14)

{X...RG}nRG + h {RG+ +X}nRG

(15)

{RG+ +X}nRG + H {(RG)2H+...X}(n1)RG

(16)

In Eq. [1316] the symbol {}nRG indicates solvation by n rare gas atoms and RG+ is essentially
a hole which may be as mobile as an electron. The fact that the infrared spectra of all RG2H+
ions are identical, regardless of the nature of X, is taken as a proof that the hydrogen atom travels a long way away from the initial location of HX, before being trapped. Comparison with theoretical calculations indicates that the structure of these species is linear centro-symmetric
(RGHRG)+.

266

Y. Haas and U. Samuni

The decay kinetics of these ionic species were found to be neither first order nor strictly
second order. There is a very strong H isotope effect, the protonated species decaying much
faster than the deuterated one, and the reaction persisted at rather low temperatures: 12 K or
lower. These characteristics are similar to those observed for O atom diffusion, discussed in
Section 41, indicating the presence of a highly mobile H atom. It appears that the experimental
facts concerning low temperature kinetics of highly mobile atoms is way ahead of theoretical
understanding.
In addition to the RG2H+ cation, negative ions such as X and XHX are also formed in
irradiated rare gas matrices containing HX. The latter are probably produced from HX dimers,
that capture a photo-generated electron to form (HX)2, and then lose a hydrogen atom to form
XHX. These anionic species are immobile and do not interact with the cationic ones discussed
above. However, in subsequent work it was found that UV irradiation of matrices containing
these cations followed by annealing, yields extremely intense new IR absorption bands due to
neutral RGHX molecules (for instance XeHBr) [Pettersson 1995, Pettersson 1995a]. The new
species are believed to be formed by the interaction of the X anions with a hole of the form
(RGn)+ which becomes mobile on annealing. In his mechanism, the H atom is captured by the
X center to form a loosely bound {X...RG...H} species, which is a trap for the mobile hole.
The RGHX species is unstable with respect to dissociation to HX and RG. Ab initio calculations
indicate a barrier of about 02 eV for the HArF system, high enough to assure stability at
cryogenic temperatures. However, the very fact that the system finds this local minimum in a
thermal process (annealing) is remarkable.
Even more remarkable is the finding that in xenon matrices neutral molecules of the
type HXeH are also formed. The binding in this new type of rare gas neutral molecules is similar to that of XeF2. The mechanism of their formation is not clear. It may involve a hydride (H)
intermediate, though independent evidence for its presence is not available. These recent results
indicate that rare gas matrices, and particularly xenon ones, may provide a medium for new
types of chemical reactions. For instance, these H atom containing species are expected to be
powerful reducing agents possibly opening the way to new synthetic routes.
The special properties of xenon matrices indicate that solid xenon behaves in a quasimetallic fashion it allows long range migration of electrons and holes. Some of the ionic
species formed in it must be viewed as being delocalized. At the same time, solid xenon is similar to solid argon or krypton in not allowing completely free motion of electrons, at least under
conventional pressure conditions. Future work on the photochemistry of xenon trapped species
can be fascinating; for instance, it might be possible to tune a photochemical reaction from a
primarily covalently controlled one to an ionic controlled one either by slight doping (e.g.
[Maier 1994]) or perhaps by increased pressure. This topic is worthy of more extensive future
work.

Reactions in rare gas matrices matrix and site effects

5.
MODELING REACTIONS IN MATRICES
The unique reactivity patterns found in many matrix reactions often cannot be easily accounted
for by conventional kinetic equations. Some of the attempts to account for reaction rates and
their dependence on temperature and concentrations were mentioned in previous sections. In
this section we focus on attempts to model the dynamics of molecules entrapped in rare gas
solids, emphasizing reaction dynamics. Such models are believed to be the first step towards a
more comprehensive theory of matrix kinetics.
All models are based on the simulation of a trapping site, as discussed in Section 2,
combined with the calculation of motions on potential energy surfaces (PES). Accurate potential surfaces in the bulk are essentially unknown, the usual practice is to use gas phase
potentials. Even those are rarely available except for very simple systems so that simulations
depend largely on empirical expressions or model potentials. Recently, ab initio calculations
were being used to compute the structures, vibrational frequencies and IR transition intensities
of stable molecules and of molecular van der Waals adducts. The quantum chemical methods
are expected to be instrumental for estimating potential energies for matrices in the very near
future.
A truncated PES, that includes the matrix/guest interactions and a one dimensional
potential along the reaction coordinate is usually used. The equations of motion are solved by
the molecular dynamics method within the classical mechanics approximation, except in some
special cases, such as reactions involving hydrogen atoms, for which a mixed classical-quantal
model may be used [Gerber 1990].
51
Photodissociation
The simplest case appears to be the dissociation of diatomic molecules. This problem was taken
up by Gerber and coworkers [Alimi 1988, Alimi 1990], who developed a model designed to
simulate the photodissociation of small molecules in a rare gas crystal. A perfect crystal slab
(usually fcc for the rare gas solids) was the starting point. A Morse potential and a repulsive
exponential one were used for the ground and excited states of a diatomic molecule, respectively. Either periodic boundary conditions or a static shell (Section 2) were used to mimic the
infinite nature of the lattice.
A photodissociation experiment (for instance HI in xenon) was simulated as follows:
the guest was introduced into a substitutional site in an arbitrarily chosen (or intuitively motivated) location. Prior to photoexcitation, the system was allowed to evolve in time long enough
(typically 25 ps) to achieve relaxation of the originally prepared structure. Several runs were
made providing randomly chosen initial configurations for the photodissociation experiment.
Excitation was assumed to be instantaneous, transforming the system from the ground to the
excited state potential without changing the coordinates and momenta (except for the diatomic
vibration) of the atoms. The equations of motions were solved with the new potentials for
another lengthy period (1020 ps). The trajectories of the atoms constituting the molecule were

267

268

Y. Haas and U. Samuni

followed as a function of time. The final result may be dissociation (atoms found in separate
final cages), recombination or isomerization (for larger species).
Typical results were as follows: for hydrides, final site selectivity was found; the hydrogen atom was always trapped in an interstitial site in the nearest neighbor cell. An octahedral
one was occupied for low excess energies and a tetrahedral one for high energies. The H atom
escape was usually not direct the atom collides with the initial cage walls many times losing
much of its energy before escaping. For F2 in argon, the photodissociation yield rose sharply
from the threshold energy of 06 eV to nearly unity at excess energy of 2 eV. The two atoms
were typically ejected in opposite directions from the initial site and travel ballistically for ~3
without being deflected. In contrast with most other atoms, fluorine atoms were found to
migrate over large distances. They, and other heavy atoms escape through windows in the trapping site (cf. Figures 9 and 10), the escape being often a direct one. A unique property of fluorine,
resulting from the FAr potential, is that after cage escape its trajectories are curvilinear, as
opposed to sharp cornered turns calculated for other atoms. Also, geminate recombination is
rare, again in contrast to other systems (Cl2 or HI in Xe, [Alimi 1988, Alimi 1989]). Final destination site selectivity was found: the F atoms ended up mostly in an Oh interstitial site, and
sometimes in a newly created SS site (at the expense of the originally occupied one) but never
in a Td interstitial site. These results are in complete accord with the gross features of experimental photo-dissociation data [Feld 1990, Katz 1991], and suggest that diffusion controlled
bimolecular reactions between photogenerated F atoms and suitable substrates could be
observed in rare gas matrices (Section 41).
This prediction was later borne out: the reaction between ethylene and a fluorine molecule trapped in the same site yields as the major product the C2H3FHF pair (Section 341). An
explanation, based on the molecular dynamics simulation by Alimi et al. [Alimi 1990] was
offered by Benderskii and Wight [Benderskii 1996]: F atoms (formed upon the photodissociation of F2) escape easily from a SS site of argon, in which they rotate essentially freely.
Consequently, the atoms can leave the cage in opposite directions through triangular windows
almost unhindered, and they are unlikely to meet again. In the presence of ethylene, however,
this rotation is severely hindered, and the nascent fluorine atoms are trapped, increasing the
chances of a bimolecular reaction. The initial product is a vibrationally excited adduct, C2H4F2,
which relaxes predominantly to the fluoroethylene-HF pair, and less frequently to the two addition products:
C2H4F2 C2H3F+HF

60%

(17)

C2H4F2 trans C2H4F2

20%

(18)

C2H4F2 gauche C2H4F2

20%

(19)

Reactions in rare gas matrices matrix and site effects

269

In an extension of the method, the photochemistry of the triatomic molecule ICN was studied
[Krylov 94a]. It was found that in contrast with the gas and liquid phase reactions, no dissociation occurred in the solid even when the molecule was excited beyond the dissociation limit.
Instead of the frustrated dissociation, isomerization to INC took place, in accord with experimental results ([Fraenkel 1993a] Section 321).
52
Isomerization of polyatomic molecules
These reactions, involving a large number of internal degrees of freedom, pose a much more
difficult challenge to theory. The current approach is to define a reaction coordinate, and construct a one dimensional potential surface along it, a task made easier with the help of Ab initio
calculations. Other degrees of freedom are often treated as harmonic oscillators, which may be
coupled internally (i.e. in the isolated molecule) or via interaction with the matrix atoms, with
motion along the reaction coordinate.
An energy dissipation process which relaxes the vibrationally excited molecules after
crossing the barrier to a local energy minimum is an essential part of isomerization and addition
reactions. An important part of the simulation therefore involves the study of vibrational relaxation, and energy transfer to the matrix. Some recent examples are surveyed.
521 The cistrans HONO isomerization reaction
The experimental study of this reaction, briefly summarized in Section 3322, raised interest as
mode selectivity was indicated [Hall 1963, Shirk 1983, McDonald 1982]. Theoretical studies
[Qin 1994] showed that for this reaction, mode selectivity is actually expected in the gas phase,
apparently since some vibrational modes are better coupled to the reaction coordinate (the torsional motion around the ONOH dihedral angle ) than others, and since in this small molecule,
IVR rates are slow with respect to the reaction rates for some modes. In an extension of previous gas phase MD simulations, Agrawal et al. [Agrawal 1994, Agrawal 1995] assessed the
effect of the matrix environment on the rate constants of reaction in both directions (cis to trans
(kct) and trans to cis (ktc)). The simulation started with the trapping model suggested by Raff
(Section 231), with the HONO molecule occupying an interstitial site in a fcc xenon lattice.
The effect of lattice vacancies was checked by random removal of 1020% of the rare gas
atoms. Mode selectivity was probed by starting each trajectory with the same total initial energy
of 17 eV, put initially into different normal modes. The reaction progress was followed by running many trajectories for each initial mode under randomly selected starting conditions which
mimic different trapping sites in the matrix The reaction time was defined as the time needed to
reach the transition state (=850) for the reactive trajectories. Rate coefficients were obtained
from the slopes of logarithmic decay plots of the form
N=N0exp(-kt)

(20)

270

Y. Haas and U. Samuni

where N0 is the total number of trajectories and N of those not achieving isomerization by time t.
Table 5 summarizes the computed results for the gas phase and xenon matrix (12 K) reactions.
Mode selectivity was evident in the gas phase simulations (rate coefficients for trajectories
starting from different initial vibrational modes vary by more than an order of magnitude)
and was found to be preserved, though to a lesser extent, in the matrix. A notable result is that
in 12 K xenon the rate coefficients increase for all modes in such a way that selectivity is
reduced.
The matrix induced increase of the rate constants was accounted for by the opening of
a phonon-assisted energy transfer pathway from other modes to the torsional one. The process
can be described as a vibrationlattice phonon moderotationtorsion mechanism. The role
of rotation in promoting the reaction was found to be important in previous gas phase simulations [Guan 1989]. This mechanism was supported by a second paper [Agrawal 1995] in which
an argon matrix was used, and the effects of isotopic substitution and the matrix temperature
were examined. Qualitative agreement with experiment was found but serious quantitative
discrepancies remain. Thus, the experimental isomerization rate ratios kct /ktc are about 500
times larger than the calculated ones. Also, the reaction was simulated at much higher initial
energies than used experimentally; no attempt was apparently made to simulate more closely
the real experimental conditions possibly since the lower reaction probability involved would
have made the calculations excessively time consuming.
The successful reproduction of non-RRKM behavior by the MD simulation is noteworthy.
The small number of vibrational modes appears to be of crucial importance in this respect, as
for the isomerization of larger nitrites no mode selectivity was reported.

Table 5. Calculated isomerization rate constants for HONO in the gas phase and in a xenon matrix [Agrawal
1994]. Different vibrational modes were excited to the same total initial energy of 17 eV in each case. Single
mode excitation was used in the first five entries, while the last entry, reports the result of simultaneous excitation of
5 vibrational modes. kct and ktc are the rate constants for the cistrans and transcis isomerization, respectively.
Mode

kct (s1)
Gas
Xenon matrix

ktc (s1)
Gas
Xenon matrix

O-H stretch

023

080

0006

037

N-O stretch

015

053

009

022

HON bend

211

271

010

040

O-H stretch

027

079

011

025

ONO bend

020

059

007

035

combination of
the 5 modes

072

095

003

031

Reactions in rare gas matrices matrix and site effects

53
Bimolecular reactions
531 The ozone+olefin reaction
The MD simulation method of Fraenkel and Haas was applied to the ethyleneozone bimolecular reaction system [Samuni 1996]. The thermal reaction does not take place in an argon matrix
but does proceed in a xenon or CO2 matrices. Ozone and ethylene were found to form an adduct
in an argon matrix, as deduced from the appearance of a new peak in the IR spectrum. No reaction was observed even upon irradiation with a pulsed CO2 laser tuned to the adducts absorption
band. This negative result was supported by the MD simulation in which the two guest molecules
were added one after the other (see Section 232) when simulating the deposition process. In
some runs, the two guest molecules were eventually found to end up within the same site. It was
found that whereas the most likely trapping sites for ethylene and ozone are DS and a SS, respectively, when co-deposited they often form a DS site, and sometimes three substitutional (TS) ones.
The results of two typical runs are shown in Figure 12. Figure 13 shows the trapping site
of the reaction product the primary ozonide (POZ) which was found to occupy a triple substitutional site. Comparison of the trapping sites of the ethyleneozone pair with that of the
product POZ reveals that in order to react and form the primary ozonide, a major rearrangement
of the argon atoms in the matrix must take place. For instance, the closely lying ethylene and
ozone occupying the DS site can react to form POZ, only by displacing an argon atom far
enough to form a TS site. This result implies that the simulation predicts that a reaction will not
take place within an argon matrix, in agreement with experiment [Nelander 1977, Kohlmiller
1981, Hawkins 1982, Samuni 1996].
532 The F2+cis-D2-ethylene reaction
The trapping site model developed by the Raff group (Section 231) was used [Raff 1990]. Both
substitutional and interstitial sites were examined. The reaction is addition and subsequent decomposition of the hot 1,2-difluoroethane-d2 (DFE-d2) product. (see Eq. [1719]) The propensity of
DFE-d2 to undergo orientational exchange increases as the available free space in the lattice
decreases. [Orientational exchange: the immediate neighborhood of the trapped molecule (DFE in
this case) can be heated locally due to relaxation of the vibrationally excited molecule. This in turn
sometimes leads to rotation of the trapped molecule to a different metastable site]. Thus, it is more
likely to take place in an interstitial site than in a substitutional one, and in an argon than in a xenon
matrix. The probability of HF or DF elimination increases as the space available decreases (This
is due to the fact that strong interaction with the matrix was found necessary for any reaction to
start). Five possible different mechanisms were considered for the elimination reaction.
Comparison with experiment showed that elimination yield ratios were well reproduced, the
cis/trans ratio of fluoroethylenes formed after HF elimination in argon were a factor of 27 lower,
and the calculated ratios of stabilized DFE-d2 to fluoroethylenes were much larger than observed
experimentally in N2 [Frei 1983]. These results were interpreted as showing that the experimental
matrix is much more open and spacious than the crystal structure used in the simulation.

271

272

Y. Haas and U. Samuni

Figure 12. The structures of two commonly obtained trapping sites for the ozoneethylene adducts in a fcc argon
lattice, as computed by the MD simulations (deposition temperature 25K). Only a small portion of the actually calculated structure is shown to facilitate the exposition of the geometry near the trapped molecule. Each site is shown
from top and side view of a cut taken along the lattice planes in which some argon atoms were replaced by the guest
molecules. Argon atoms are shown as open large circles, hydrogen atoms as smaller open circles, oxygen atoms are
black, and carbon atoms are gray. One (a) is a disubstitutional site in which the two molecules replace two adjacent
argon atoms in a 100 plane of the lattice. The other (b) is a three-substitutional site, in which the ethylene and ozone
molecules substitute three argon atoms lying in a row in a 111 plane of the crystal lattice [Samuni 1996].

Reactions in rare gas matrices matrix and site effects

Figure 13. The structure of the most frequently obtained trapping site of POZ deposited in an argon fcc lattice at
25 K as calculated by the molecular dynamics simulation. The molecule is seen to displace three argon atoms
from a 111 plane of the crystal lattice, the displaced atoms forming a triangle. See Figure 12 for drawing details
[Samuni 1996].

6.
SUMMARY
This paper discussed some recent advances made in the application of the matrix isolation technique to the study of chemical reactivity with an emphasis on matrix and site effects. The advent
of modern experimental techniques (commercial closed cycle cryostats, Fourier transform IR
spectrometers, powerful UV lasers), along with the wide-spread use of quantum chemical calculations, have gradually transformed the field. Further experimental advances, for instance
ultra-fast techniques probing the sub-picosecond time scale, will undoubtedly provide deeper
insight into the dynamics of matrix isolated molecules. The refined data thus obtained will
require an even closer interaction between theory and experiment than hitherto attempted.
Phenomenological descriptions of chemical transformations are giving way to efforts at detailed
mechanistic interpretation on the microscopic level. Matrix isolation spectroscopy may now be
at the same stage as gas phase dynamics thirty years ago, when molecular beams began to be
widely used. This paper summarized some of the recent theoretical efforts, mainly based on the
MD method, to simulate and predict experimental results.

273

274

Y. Haas and U. Samuni

Modeling of matrices and matrix reaction dynamics is still at its infancy. In spite of considerable progress, none of the models described in this review provides a satisfactory explanation for all experimental data. The molecular dynamics method, which appears to dominate the
field at this time, is appropriate for the eventual description of dynamics and chemical reactivity
in matrices. Sums of pair-wise atom-atom potentials are widely used, even though it is well
known that they do not provide a strictly correct description of even a simple rare gas crystal.
However, since a major ingredient of the total interaction energy can be represented by pair
wise potentials, this appears to be an acceptable first approximation. The extreme added computational requirements incurred by the addition of three body (and higher) interactions is a
major obstacle for further progress. It is hoped that this difficulty will be partially alleviated by
the rapid advances in computer technology. Even the currently used rather crude model has led
to a better understanding of matrix properties and dynamics, as detailed in Sections 2 and 5, and
helped the design and successful execution of novel experiments.
Matrix isolation is widely used for the preparation and/or isolation and characterization
of new, often exotic, species. This application will certainly continue to be of central importance
in the field. The elucidation of microscopic details may help in advancing this aspect as well. In
that connection, it appears that many synthetic reactions did not take place in really matrix isolated conditions. Local heating may have played an important role, in allowing some of the
photochemical reactions. An intriguing prospect for future study is the role of matrix structural
changes, for instance in cases where solidsolid phase transitions can take place (Section 42).
It seems appropriate to enumerate some of experimental and theoretical challenges of
current interest in the field:
Development of better potential functions for simulating the properties of matrix isolated systems. Matrix shifts remain largely unaccounted for, except for rare cases. The difficulty in
reproducing the red shifts in IR spectra was repeatedly referred to (Section 2). Introduction of
many body interactions is beginning to be implemented [Grigorenko 1996]; the use of electrostatic interaction is also a promising way of obtaining a more realistic view of matrix guest
interactions.
Once realistic potentials are found, the simulation of trapping site structure may be used
to probe the relationship, if any, between it and the sites chemical reactivity. Computer simulated sites will be tested for different reactivity patterns helping in understanding observed
trends, and to predict novel ones.
The relationship between diffusion and chemical reactivity in matrices is still not completely settled. Considerable evidence supports the view that only small atoms can move relatively
freely, and even that only if they are small enough, as in the case of O(1D) (Section 41). However,
the motion of larger species (cf. ammonia in the NH3/naphthol experiments) is clearly indicated in
careful experiments. There is a need for experiments probing the motion of larger molecules in
matrices, including the effects of site restructuring upon temperature changes.
A related problem is the extent and applicability of electron (and hole) migration in

Reactions in rare gas matrices matrix and site effects

matrices. The formation mechanism of new, rather unstable species such as XeH2 is not yet
understood, and may involve charge transfer interactions. The mechanisms of newly discovered
reactions, such as the sensitized bond fission of cyclopropanes (Section 322) are still not completely understood. It may require the development of new theoretical methods to describe the
properties of rare gas solids containing charge separated species.
Experimental and theoretical elucidation of the detailed changes taking place in the
matrix upon the photochemical dissociation of a large molecule (Section 321). How far do the
fragments travel, is there a preferred configuration for the products and what are the factors
determining it?
The role of matrix (or site) structure in determining the outcome of photochemical
isomerization reactions is only beginning to emerge. The example of benzene photolysis
(Section 3323) shows that various products may be formed in the same matrix, a result that
apparently can be explained only as due to site effects. However, no microscopic description of
the process is available at present.
Developments of new experimental techniques aimed at site selective excitation. The
existence of different trapping sites offers in principle the possibility of site selective chemistry.
The low temperature environment practically prevents collisional scrambling, but energy transfer may take place between molecules trapped in different sites. One possibility is the use of
tunable infrared lasers to excite the IR absorption band at a given site, followed by picosecond
UV excitation of the IR excited molecules. Details of vibrational relaxation routes need to be
elucidated before such an experiment can be planned.
New experimental opportunities opened by the presence of reactive species such as F
atoms have begun to emerge. This trend is expected to continue. For instance, the chemical
reactions of species such as XHX and HXH may prove to be most interesting.
In conclusion, matrix isolation research covers now a much broader range than the
freezing of reactive intermediates for allowing their characterization. It forms the natural link
between gas phase work, via cluster research in supersonic jets, to a macroscopic solid. Various
theoretical predictions are most readily tested for matrix isolated species. Ab initio calculations
are routinely used for predicting and verifying the identity of novel molecules; their ubiquitous
use in the assignment of infrared spectra is impressive there are hardly any matrix experimental papers that do not compare their spectroscopic results with these calculations.
The current interest in site effects is partly based on the hope that a better understanding of these relatively simple systems will help in constructing models for more practical purposes, such as used in industry or biology. Subtle reactivity patterns can already be understood,
at least qualitatively, by the use of molecular dynamics simulations. At times, we have seen that
many unexpected effects are still being discovered even in this relatively simple medium, and
that in addition to considerable past efforts, new theoretical tools must still be developed before
a more quantitative account of the observations can be formulated. At the present level of
activity in the field this state of affairs is not expected to last long!

275

276

Y. Haas and U. Samuni

7.
ACKNOWLEDGMENTS
We thank our coworkers Ruchama Fraenkel and Smadar Kahana for many stimulating discussions, and for their substantial contribution to the work done by the Jerusalem group. Parts of
this review were written during a visit of YH to Oxford University, supported by the Visiting
Research Professorship Scheme of the Royal Society. We thank Professor John Simons and the
Laboratory staff for their warm hospitality. The Farkas Center for Light Induced Processes is
supported by the Minerva Gesellschaft fr die Forschung, GmbH, Munich. Parts of the authors
work were supported by Grant number HRN-5544-G-00-2078-00 Program in Science and
Technology Cooperation, Human Capacity Development, Bureau for Global Programs, Field
Support and Research USAID.
8.

REFERENCES

Abbate 1985a. A. Abbate and C. B. Moore, J. Chem. Phys. 82, 1263 (1985).
Abbate 1985b. A. Abbate and C. B. Moore, J. Chem. Phys. 83, 975 (1985).
Abe 1979. H. Abe and W. Schulze, Chem. Phys. 41, 257 (1979).
Agrawal 1994. P. M. Agrawal, D. L. Thompson and L. M. Raff, J. Chem. Phys. 101, 9937 (1994).
Agrawal 1995. P. M. Agrawal, D. L. Thompson and L. M. Raff, J. Chem. Phys. 102, 7000 (1995).
Alcock 1968. W. G. Alcock and G. C. Pimentel, J. Chem. Phys. 48, 2373 (1968).
Alimi 1988. R. Alimi, R. B. Gerber and V. A. Apkarian, J. Chem. Phys. 89, 174 (1988).
Alimi 1989. R. Alimi, A. Brokman, R. B. Gerber and V. A. Apkarian, J. Chem. Phys. 91, 1161 (1989).
Alimi 1990. R. Alimi, R. B. Gerber and V. A. Apkarian, J. Chem. Phys. 92, 3551 (1990).
Alimi 1991. R. Alimi, R. B. Gerber and V. A. Apkarian, Phys. Rev. Lett. 66, 1295 (1991).
Allen 1987. M. P. Allen and D. J. Tildesley, Computer Simulations of Liquids (Clarendon, Oxford, (1987).
Almond 1989. M. J. Almond and A. J. Downs, Adv. Spectrosc. 17, 1 (1989).
Almond 1994. M. J. Almond, Chem. Soc. Rev. 23, 309 (1994).
Andrews 1994. L. Andrews, J. T. Yustein, C. A. Thompson and R. D. Hunt, J. Phys. Chem. 98, 6514 (1994).
Andrews 1996. L. Andrews, G. V. Chertinin, A. Citra and M. Neurock, J. Phys. Chem. 100, 11235 (1996).
Baldeschwielder 1960. J. D. Baldeschwielder and G. C. Pimentel, J. Chem. Phys. 33, 1008 (1960).
Barker 1974. J. A. Barker, R. O. Watts, K. Lee, T. P. Schaefer and Y. T. Lee, J. Chem. Phys. 61, 3081 (1974).
Barnes 1984. A. J. Barnes, J. Mol. Str. 113, 161 (1984).
Becker 1993. A. Becker, W. Langel, S. Maass and E. Knoezinger, J. Phys. Chem. 97, 5525 (1993).
Beichert 1995. P. Beichert, D. Pfeiler and E. Knzinger, Ber. Bunsenges. Phys. Chem. 99, 1469 (1995).
Bell 1990. T. W. Bell, D. M. Haddleton, A. McCamley, M. G. Partridge, R. N. Perutz and H. Willner, J. Am. Chem.
Soc. 112, 9212 (1990).
Benderskii 1994. A. V. Benderskii, C. A. Wight, Chem. Phys. 189, 293 (1994); J. Chem. Phys. 101, 292 (1994).
Benderskii 1996. A. V. Benderskii and C. A. Wight, J. Phys. Chem. 100, 14958 (1996).
Bodenbinder 1994. M. Bodenbinder, S. E. Ulic and H. Willner, J. Phys. Chem. 98, 6441 (1994).
Bondybey 1972. V. E. Bondybey and G. C. Pimentel, J. Chem. Phys. 56, 3832 (1972).
Bone 1992. O. R. G. A. Bone, C. R. Le Sueur, R. D. Amos and A. J. Stone, J. Chem. Phys. 96, 8390 (1992).
Braathen 1990. G. O. Braathen, A. Gatial, P. Klaboe and J. Nielsen, J. Mol. Struct. 218, 67 (1990).
Brucker 1989. G. A. Brucker and D. F. Kelley, J. Chem. Phys. 90, 5243 (1989).
Brucker 1989a. G. A. Brucker and D. F. Kelley, Chem. Phys. 136, 213 (1989).
Cheng 1995 B. M. Cheng, J. R. Grover and E. A. Walters, Chem Phys. Lett. 232, 364 (1995).
Cheshenovsky 1988. O. Cheshenovsky and S. Leutwyler, J. Chem. Phys. 88, 4127 (1988).
Chertinin 1995. G. V. Chertinin and L. Andrews, J. Phys. Chem. 99, 12131 (1995).
Chertinin 1995a. G. V. Chertinin and L. Andrews, J. Phys. Chem. 99, 15004 (1995).

Reactions in rare gas matrices matrix and site effects


Chertinin 1996. G. V. Chertinin, W. Saffel, J. T. Yustein, L. Andrews, M. Neurock, A. Ricca and
C. W. Bauschlichter, J. Phys. Chem. 100, 5261 (1996).
Clemitshaw 1988. K. C. Clemitshaw and J. R. Sodeau, J. Phys. Chem. 92, 5491 (1988).
Closs 1986. G. L. Closs, L. T. Calcaterra, N. J. Green, K. W. Penfield and J. R. Miller, J. Phys. Chem. 90, 3673,
(1986).
Calcaterra 1982: L. T. Calcaterra, G. L. Closs and J. R. Miller, J. Am. Chem. Soc. 105, 670 (1983), J. Am. Chem.
Soc. 104, 6488 (1982).
Coussan 1994. S. Coussan, N. Bakkas, A. Loutellier, J. P. Perchard and S. Racine, Chem. Phys. Lett. 217, 123
(1994).
Cronin 1995. L. Cronin, M. C. Nicasio, R. N. Perutz, R. G. Peter, D. M. Rodick, M. K. Whittlesey, J. Am. Chem.
Soc. 117, 10047 (1995).
Cruz 1996. A. J. Cruz G. E. Lopez, J. Chem. Phys. 104, 4294 (1996); J. Chem. Phys. 104, 4301 (1996).
Danilychev 1993. A. V. Danilychev, V. A. Apkarian, J. Chem. Phys. 99, 8617 (1993).
Danilychev 1994. A. V. Danilychev, V. A. Apkarian, J. Chem. Phys. 100, 5556 (1994).
Downes 1995. A. J. Downes, T. M. Greene and C. M. Gordon, Inorg. Chem. 34, 6191 (1995).
Droz 1990. T. Droz, R. Knochenmuss and S. Leutwyler, J. Chem. Phys. 93, 4520 (1990).
Dunning 1977. J. D. Dunning Jr. and P. J. Hay, J. Chem. Phys. 66, 3767 (1977).; J. Chem. Phys. 74, 3718 (1981).
Durig 1992. J. R. Durig, J. Liu, T. S. Little and V. F. Kalasinsky, J. Phys. Chem. 96, 8824 (1992).
Easley 1970. W. C. Easley and W. Weltner, J. Chem. Phys. 52,197 (1970).
Engdahl 1985. A. E. Engdahl and B. Nelander, J. Phys. Chem. 89, 2860 (1985).
Engdahl 1986. A. E. Engdahl and B. Nelander, J. Phys. Chem. 90, 4982 (1986).
Engdahl 1987. A. E. Engdahl and B. Nelander, J. Phys. Chem. 91, 2253 (1987).
Fajardo 1986. M. E. Fajardo and V. A. Apkarian, J. Chem. Phys. 85, 5660 (1986).
Fajardo 1988. M. E. Fajardo and V. A. Apkarian, J. Chem. Phys. 89, 4102 (1988).
Farges 1988. J. Farges, M. F. deFeraudy, B. Raoult and G. Torchet, Adv. Chem. Phys. 70, 45 (1988).
Felder 1984. P. Felder and Hs. H. Gnthard, Chem. Phys. 85, 1 (1984).
Feld 1990. J. Feld, H. Kunttu, V. A. Apkarian, J. Chem. Phys. 93, 1009 (1990).
Firth 1989. Q. W. Firth, P. F. Barbara and H. P. Trommsdorff, Chem. Phys. 136, 349 (1989).
Fitzmaurice 1992. D. J. Fitzmaurice and H. Frei, J. Phys. Chem. 96, 10308 (1992).
Lifson 1968. S. Lifson and A. Warshel, J. Chem. Phys. 49, 5116 (1968), 53, 582 (1970).
Huler 1974. E. Huler and A. Warshel, Acta Crysta B 30, 1822 (1974).
Fraenkel 1993. R. Fraenkel, U. Samuni and Y. Haas, Chem Phys. Lett. 203, 523 (1993).
Fraenkel 1993a. R. Fraenkel and Y. Haas, Chem Phys. Lett. 214, 234 (1993).
Fraenkel 1994. R. Fraenkel and Y. Haas, Chem. Phys. 186, 185 (1994).
Fraenkel 1995. R. Fraenkel, S. Kahana, Y. Haas, Ber. Bunsenges. Phys. Chem. 99, 412 (1995).
Frei 1983. H. Frei, J. Chem. Phys. 79, 748 (1983).
Frei 1991. H. Frei, Chimia 45, 175 (1991).
Frei 1992. H. Frei, Vibrational Spectra and Structure, Ed. J. Durig, Vol 20 p. 1, (1992).
Friedman 1965. H. Friedman and S. Kimel, J. Chem. Phys. 43, 3925 (1965).
Gerber 1990. R. B. Gerber and R. Alimi, Chem. Phys. Lett. 173, 393 (1990).
Gilbert 1991. A. Gilbert and J. Baggot, Essentials of Molecular Photochemistry (Blackwell, London, 1991,
p. 357).
Greene 1994. T. M. Greene, W. Brown, L. Andrews, A. J. Downs, G. V. Chertinin, N. Runeberg and P. Ryykko,
J. Phys. Chem. 99, 7925 (1995).
Grigorenko 1996. B. L. Grigorenko, A V. Nemukhin, V. A. Apkarian, J. Chem. Phys.104 5510 (1996);
B. L. Grigorenko, A V. Nemukhin, A. V. Savin, Chem. Phys. Lett. 250, 226 (1996).
Guan 1989. Y. Guan and D. L. Thompson, Chem. Phys. 139, 147 (1989).
Gunde 1982. R. Gunde, P. Felder and Hs. H. Gnthard, Chem. Phys. 64, 313 (1982).
Gunde 1991. R. Gunde, H. J. Keller, T.-J. Ha and Hs. H. Gnthard, J. Phys. Chem. 95, 2802 (1991).

277

278

Y. Haas and U. Samuni


Guo 1995. Y. Guo and D. L. Thompson, J. Chem. Phys. 103, 9024 (1995).
Hall 1963. R. T. Hall and G. C. Pimentel, J. Chem. Phys. 38, 1889 (1963).
Hall 1992a. C. Hall, W. D. Jones, R. J. Mawby, R. Osman, R. N. Perutz, M. K. Whittlesey, J. Am. Chem. Soc.
114, 7425 (1992).
Hall 1992b. C. Hall, W. D. Jones, M. K. Whittlesey, R. J. Mawby, R. Osman, R. N. Perutz, L. D. Field,
M. P. Wilkinson and M. W. George, J. Am. Chem. Soc. 115, 8627 (1992).
Harrison 1994. J. A. Harrison and H. Frei, J. Phys. Chem. 98, 12142 (1994); 98, 12152 (1994).
Havighorst 1992. M. Havighorst, M. Prager and W. Langel, Physica B 180 & 181, 674 (1992).
Hawkins 1982. M. Hawkins, C. K. Kohlmiller and L. Andrews, J. Phys. Chem. 86, 3154 (1982).
Hill 1995. J. N. Hill, R. N. Perutz and A. D. Rooney, J. Phys. Chem. 99, 531 (1995); 538 (1995).
Huber-Wlchli 1975. P.Huber-Wlchli and Hs. H. Gnthard, Chem. Phys. Lett. 30, 347 (1975).
Huler 1974. E. Huler and A. Warshel, Acta Cryst. B 30, (1974) 582.
Huisken 1991. F. Huisken, A. Kulcke, C. Laush and J. M. Lisy, J. Chem. Phys. 97, 3924 (1991).
Jacobs 1994. J. Jacobs, M. Kronberg, H. S. P. Muller and H. Willner, J. Am. Chem. Soc. 116, 1106 (1994).
Jacox 1994. M. E. Jacox, J. Phys. Chem. Ref. Data, Monograph 3, Energy levels of polyatomic transient
molecules 1994.
Johnstone 1991. D. E. Johnstone and J. R. Sodeau, J. Phys. Chem. 95, 165 (1991).
Johnsson 1995. K. Johnsson, A. Engdahl, J. Kolm, J. Nieminen and B. Nelander, J. Phys. Chem. 99, 3902 (1995).
Johnsson 1996. K. Johnsson, A. Engdahl and B. Nelander, J. Phys. Chem. 100, 3923 (1996).
Johnson 1972. P. M. Johnson and L. Ziegler, J. Chem. Phys. 56, 2169 (1972).
Jones 1986. L. H. Jones, S. A. Ekberg and B. I. Swanson, J. Chem. Phys. 85, 3203 (1986).
Kappe 1995. C. O. Kappe, M. W. Wong and C. Wentrup, J. Org. Chem. 60, 1686 (1995).
Katz 1991. A. I. Katz and V. A. Apkarian, J. Phys. Chem. 94, 6671 (1991).
Kizer 1995. K. S. Kizer, V. A. Apkarian, J. Chem. Phys. 103, 4945 (1995).
Niebel 1976. K. F. Niebel and J. A. Venables in M. L. Klein J. A. Venables, Eds., Rare Gas Solids, Vol. I, p. 559,
(Academic Press, London, 1976).
Kohlmiller 1981. C. K. Kohlmiller and L. Andrews, J. Am. Chem. Soc. 103, 2578 (1981).
Knzinger 1993. E. Knzinger, P. Beichert, J. Hermeling and O. Schrems, J. Phys. Chem. 97, 1324 (1993).
Knzinger 1995. E. Knzinger and P. Beichert, J. Phys. Chem. 99, 4906 (1995).
Knudsen 1991. A. K. Knudsen and G. C. Pimentel, J. Phys. Chem. 95, 2823 (1991).
Kopelman 1988. R. Kopelman, Science 241, 1620 (1988).
Krueger 1992. H. Krueger and E. Weitz, J. Chem. Phys. 96, 2846 (1992).
Krylov 1994. A. I. Krylov, R. B. Gerber and V. A. Apkarian, Chem. Phys. 189, 261 (1994).
Krylov 1994a. A. I. Krylov and R. B. Gerber, J. Chem. Phys. 100, 4242 (1994).
Kunttu 1990. H. Kunttu, J. Feld, R. Alimi and V. A. Apkarian, J. Chem. Phys. 92, 4856 (1990).
Kunttu 1991. H. Kunttu, E. Sekret and V. A. Apkarian, J. Chem. Phys. 94, 7819 (1991).
Kunttu 1992. H. Kunttu, J. Seetule, M. Rasanen and V. A. Apkarian, J. Chem. Phys. 96, 5630 (1992).
Kunttu 1994. H. M. Kunttu J. A. Seetula, Chem. Phys. 189, 273 (1994).
LaBrake 1993. D. LaBrakeand E. Weitz, Chem. Phys. Lett. 211, 430 (1993).
LaBrake 1995. D. LaBrake, E. T. Ryan and E. Weitz, J. Chem. Phys. 102, 4112 (1995).
Langel 1993. W. Langel, M. Prager H.-W. Fleger, E. Knzinger, H.-J. Lauter, H. Blank and C. J. Carlyle, J. Chem.
Phys. 98, 4838 (1993).
Lawrence 1988. W. G. Lawrence, F. Okada and V. A. Apkarian, Chem. Phys. Lett. 150, 339 (1988).
Lawrence 1992. W. G. Lawrence and V. A. Apkarian, J. Chem. Phys. 97, 2229 (1992).
Lee 1992. T. J. Lee, C. M. Rohlfig and J. E. Rice, J. Chem. Phys. 98, 9335 (1992).
Legon 1986. A. C. Legon and D. J. Milen, Chem. Rev. 94, 635 (1986).
Lifson 1968. S. Lifson and A. Warshel, J. Chem. Phys. 49, 5116 (1968).
Lugez 1993. C. Lugez, A. Schriever, L. Schriever-Mazzuoli, E. Lasson and C. J. Nielsen, J. Phys. Chem. 97,
11617 (1993).

Reactions in rare gas matrices matrix and site effects


Luke 1995. B. T. Luke, Theochem. J. Mol. Struct. 332, 283 (1995).
Maier 1994. G. Maier, D. Jurgen, R. Tross, H. P. Reisenauer, B. Andes Hess Jr and L. J. Schaad, Chem. Phys. 189,
383 (1994).
Mirsky 1980. K. Mirsky, Chem. Phys. 46, 445 (1980); J. Manz and K. Mirsky, Chem. Phys. 46, 457 (1980).
McDonald 1982. P. A. McDonad and J. S. Shirk, J. Chem. Phys. 77, 2355 (1982).
McGrath 1988. M. P. McGrath, K. Clemitshaw, F. S. Rowland and W. J. Hehre, Geophys. Res. Lett. 15, 883
(1988).
Metropolis 1953. N. Metropolis, A. W. Rosenbluth, M. N. Rosenbluth, A. H. Teller, and E. Teller, J. Chem. Phys.
21, 1087 (1953).
Miligan 1967. D. E. Miligan and M. E. Jacox, J. Chem. Phys. 47, 278 (1967).
Milligan 1973. D. E. Milligan and M. E. Jacox, J. Mol. Spectrosc. 46, 460 (1973).
Misochko 1996. E Misochko, E. Ya. Benderskii and C. A. Wight, J. Phys. Chem. 100, 4496 (1996).
Molina 1987. L. T. Molina and M. Molina, J. Phys. Chem. 91, 433 (1987).
Molina 1996. M. J. Molina, Angew. Chem, Int. Ed. Eng. 35, 1778 (1996).
Molnar 1995. F. Molnar and B. Dick, Ber. Bunsenges. Phys. Chem. 9, 412 (1995).
Nakagawa 1969. T. Nakagawa and Y. Morino, Bull. Chem. Soc. Japan 42, 2212 (1969).
Nakata 1989a. M. Nakata and H. Frei, J. Am. Chem. Soc. 111, 5240 (1989).
Nakata 1989b. M. Nakata and H. Frei, J. Phys. Chem. 93, 7670 (1989).
Nakata 1992. M. Nakata and H. Frei, J. Am. Chem. Soc. 114, 1363 (1992).
Nelander 1977. B. Nelander and L. Nord, Tetrahedron Lett. 32, 2821 (1977).
Nord 1982. L. Nord, J. Mol. Struct. 96, 27 (1982).
Perutz 1985. R. N. Perutz, Chem Rev. 85, 77, 97 (1985).
Pettersson 1995. M. Pettersson, J. Lundell and M. Rasanen, J. Chem. Phys. 102, 6423 (1995).
Pettersson 1995a. M. Pettersson, J. Lundell and M. Rasanen, J. Chem. Phys. 103, 205 (1995).
Pimentel 1963. G. C. Pimentel and S. W. Charles, Pure Appl. Chem. 7, 111 (1963).
Rochkind 1967. M. M. Rochkind and G. C. Pimentel, J. Chem. Phys. 46, 2373 (1967).
Pimentel 1978. G. C. Pimentel, Ber. Bunsenges. Phys. Chem. 82, 2 (1978).
Pollack 1964. G. L. Pollack, Rev Mod. Phys. 36, 748 (1964).
Post 1952. B. Post, R. S. Schwartz and I. Frankuchen, Acta Cryst. 5, 372 (1952).
Prager 1988. M. Prager and W. Langel, J. Chem. Phys. 88, 7995 (1988).
Qin 1994. Y. Qin and D. L. Thompson, J. Chem. Phys. 100, 645 (1994). and references therein.
Raff 1987. L. M. Raff, J. Chem. Phys. 91, 3266 (1987).
Raff 1990. L. M. Raff, J. Chem. Phys. 93, 3160 (1990).
Raff 1991. L. M. Raff, J. Chem. Phys. 95, 8901, 8905 (1991).
Raff 1992. L. M. Raff, J. Chem. Phys. 97, 7459 (1992).
Rasanen 1983. M. Rasanen, A. Aspiala, L. Homanen and J. Murto, J. Mol. Struct. 96, 81 (1983).
Riley 1988. M. E. Riley, M. E. Coltrin and D. J. Diestler, J. Chem. Phys. 88, 5934 (1988).
Rosetti 1980. R. Rosetti and L. E. Brus, J. Chem. Phys. 73, 1546 (1980).
Rosetti 1981. R. Rosetti, R. Rayford, R. C. Haddon and L. E. Brus, J. Am. Chem. Soc. 103, 4034 (1981).
Ryan 1993. E. T. Ryan, E. Weitz, J. Chem. Phys. 99, 1004 (1993).
Ryan 1993a. E. T. Ryan, E. Weitz, J. Chem. Phys. 99, 8628 (1993).
Ryan 1994. E. T. Ryan and E. Weitz, Chem Phys. 189, 293 (1994).
Samuni 1994. U. Samuni, S. Kahana, R. Fraenkel, Y. Haas, D. Danovich and S. Shaik, Chem. Phys. Lett. 225, 391
(1994).
Samuni 1996. U Samuni, R. Fraenkel, Y. Haas, R. Fajgar, and J. Pola, J. Am. Chem. Soc. 118, 3687 (1996).
Pugliano 1992. N. Pugliano and R. J. Saykally, Science 257, 1937 (1992).
Smoluchowski 1917. M. V. Smoluchowski, Z. Phys. Chem. 92, 129 (1917).
Schulze 1974. W. Schulze and D. M. Kolb, J. Chem. Soc. Faraday Trans II 70, 1098 (1974).
Schulze 1978. W. Schulze, H. U. Becker and H. Abe, Ber. Bunsenges. Phys. Chem. 82, 138 (1978).

279

280

Y. Haas and U. Samuni


Schriever 1988. A. Schriever, L. Schriever and J. P. Perchard, J. Mol. Spectr. 127, 125 (1988).
Schriever 1989. R. Schriever, M. Chergui, H. Kunz, V. Stepanenko and N. Schwentner, J.Chem. Phys. 91, 4128
(1989).
Schriever 1990. R. Schriever, M. Chergui, O. Unal, N. Schwentner and V. Stepanenko, J. Chem. Phys. 93, 3245
(1990).
Schriver 1991. L. Schriver, O. Carrere, A. Schriver and K. Jaeger, Chem. Phys. Lett. 181, 505 (1991).
Schriver-Mazzuoli 1995. L. Schriver-Mazzuoli, A. Schriver and M. Wierzejewska-Hnat, Chem. Phys. 199, 227
(1995).
Schwentner 1996. N. Schwentner, Phys. Rev. Lett. 76, 648 (1996).
Seranski 1992. K. Seranski, M. Winter and U. Schurath, Chem. Phys. 159, 247 (1992).
Winter 1992. M. Winter, K. Seranski and U. Schurath, Chem. Phys. 159, 235 (1992).
Shirk 1982. P. A. McDonald, A. Shirk and J. S. Shirk, J. Chem. Phys. 77, 2355 (1982); A. Shirk and J. S. Shirk,
Chem. Phys. Lett. 97, 549 (1983).
Shirk 1983 A. E. Shirk and J. S. Shirk, Chem. Phys. Lett. 97, 549 (1983).
Speiser 1996. S. Speiser, Chem. Rev. 96, 1953 (1996).
Suzuki 1992. S. Suzuki, P. G. Green, R. E. Bumgarner, S. Dasgupta, W. A. Goddard III and G. A. Blake, Science
257, 942 (1992).
Swinney 1990. T. C. Swinney and D. F. Kelley, J. Phys. Chem. 95, 2430 (1990).
Taleb-Bendiab 1991. A. Taleb-Bendiab, K. W. Hillig and R. L. Kuczkowski, J. Chem. Phys. 94, 6956 (1991).
Tanaka 1993. N. Tanaka, Y. Kajii, K. Shibuya and M. Nakata, J. Phys. Chem. 97, 7048 (1993).
Tanaka 1995. S. Tanaka, H. Kajihara, S. Koda and V. A. Apkarian, Chem. Phys. Lett. 233, 555 (1995).
Warshel 1970. A. Warshel and S. Lifson, J. Chem. Phys. 53, 582 (1970).
Whittle 1954. E. Whittle, D. A. Dows and G. C. Pimentel, J. Chem. Phys. 22, 1943 (1954).
Wolf 1994. J. Wolf and G. Hohlneicher, Chem. Phys. 181, 185 (1994).
Zilberg 1994 S. Zilberg, U. Samuni, R. Fraenkel and Y. Haas, Chem. Phys. 186, 303 (1994).
Zoval 1994. J. Zoval and V. A. Apkarian, J. Phys. Chem. 98, 7945 (1994).
Zuhse 1996. R. H. Zuhse, M. W. Wong and C. Wentrup, J. Phys. Chem. 100, 3917 (1996).

S-ar putea să vă placă și