The photoisomerization of aqueous ICN studied by subpicosecond transient absorption spectroscopy

Article abstract of DOI:10.1063/1.1467897

The photolysis of aqueous ICN at 266 nm was studied using transient absorption spectroscopy. It was observed that the caging of the I and CN photoproducts using the surrounding water molecules limited the I and CN quantum yield to 37% after 1 picosecond (

Full text of DOI:10.1063/1.1467897

The photoisomerization of aqueous ICN studied by subpicosecond transient
absorption spectroscopy
Jane Larsen, Dorte Madsen, Jens-Aage Poulsen, Tina D. Poulsen, Søren R. Keiding, and Jan Thøgersen
Citation: The Journal of Chemical Physics 116, 7997 (2002); doi: 10.1063/1.1467897
View online: http://dx.doi.org/10.1063/1.1467897
View Table of Contents: http://scitation.aip.org/content/aip/journal/jcp/116/18?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Ionization and dissociation dynamics of vinyl bromide probed by femtosecond extreme ultraviolet transient
absorption spectroscopy
J. Chem. Phys. 140, 064311 (2014); 10.1063/1.4865128
The dissociation dynamics of HeI 35 Cl (B,v ′ =2,3) complexes with varying amounts of internal energy
J. Chem. Phys. 122, 044318 (2005); 10.1063/1.1829971
Photodissociation of ICN at the liquid/vapor interface of water
J. Chem. Phys. 121, 2253 (2004); 10.1063/1.1765093
Molecular dynamics study of the photodissociation and photoisomerization of ICN in water
J. Chem. Phys. 119, 2127 (2003); 10.1063/1.1585019
Hybrid quantum/classical study of ICN in an Ar matrix: Photofragmentation and cage exit
J. Chem. Phys. 113, 1027 (2000); 10.1063/1.481882
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.113.111.210 On: Wed, 24 Dec 2014 10:23:40

JOURNAL OF CHEMICAL PHYSICS
VOLUME 116, NUMBER 18
8 MAY 2002
The photoisomerization of aqueous ICN studied by subpicosecond
transient absorption spectroscopy
Jane Larsen, Dorte Madsen, Jens-Aage Poulsen, Tina D. Poulsen, Søren R. Keiding,a)
and Jan Thøgersen
˚
Department of Chemistry, Aarhus University, Langelandsgade 140, DK 8000 Arhus C, Denmark
͑
Received 20 December 2001; accepted 13 February 2002͒
The photolysis of aqueous ICN is studied by transient absorption spectroscopy covering the spectral
range from 227 to 714 nm with 0.5 ps time resolution. The experimental data show that when
2
2
ICN͑aq͒ is photolyzed at 266 nm, it dissociates into I and CN and both the I( P3/2) and I( P1/2)
channels are populated. Approximately half the fragments escape the solvent cage while the
remainder recombines within the solvent cage during the first picosecond. The majority of the
recombinations form ICN while only a minor fraction produces the metastable INC isomer. INC and
ICN relax to the vibrational ground state within 1 ps in good agreement with theoretical estimates
based on the golden rule formalism as well as molecular dynamics simulations. Diffusive
recombination involving fragments that have escaped the solvent cage further reduces the quantum
yield of I and CN to 10% during the following 100 ps. This recombination produces exclusively
ICN. © 2002 American Institute of Physics. ͓DOI: 10.1063/1.1467897͔
9
3
1
I. INTRODUCTION
indicated in Fig. 1. The ⌸ and ⌸ states correlate with
1
3
1
2
ground state I( P ), while ⌸0
ϩ
correlates with the excited
3
/2
One of the most prominent features of liquid-phase re-
action dynamics is the solvents ability to inhibit reactions by
caging the products. This phenomenon is readily observed in
many photodissociation processes where a substantial frac-
tion of the photoproducts is kept within the confines of the
first solvent shell and forced to recombine in a cage back
reaction. Triatomic molecules are the smallest species that
may recombine in a configuration different from that of the
parent molecule, while still being sufficiently simple to allow
for a detailed calculation of the solvent induced recombina-
tion process. In the triatomic systems where geminate recom-
bination has been studied in detail, fast geminate recombina-
tion to the parent molecule on the electronic ground state is
by far the dominating recombination channel within the sol-
2
I( P1/2) channel. A considerable fraction of the excess en-
ergy liberated by the photodissociation of ICN͑g͒ is trans-
ferred to the CN fragment as rotational energy, whereas the
9
CN fragment is left vibrationally cold. Detailed molecular
dynamics ͑MD͒ simulations of the photolysis of ICN in liq-
uid chloroform performed by Benjamin show that the high
angular momentum enables the CN fragment to rotate rap-
idly, thus facilitating the formation of INC within the solvent
cage. More specifically: If the ICN molecule is excited to
the ⌸ state, the MD simulations predict a 75% recombina-
tion yield on the electronic ground state after 1 ps. The re-
combination of I and CN produces ICN and INC. ICN and
INC are formed vibrationally excited and the simulations
predict several interconversions between the two isomers
during the first couple of picoseconds resulting in a relative
yield of 73% ICN and 27% INC. Of the remaining ICN
molecules excited to the ⌸1 state, 2/3 escape the cage in 0.5
ps while 1/3 stay excited as ICN* or INC* for more than 5
ps. If, on the other hand, the ICN molecule is excited to the
state, 85% of the molecules recombine on the excited
state with a 7:3 ratio between the yield of ICN* and
INC*, while the rest escape the cage within the first pico-
second. Nonadiabatic transitions from ⌸0ϩ to the ⌸ and
⌸1 states are not considered by the simulations, but are
expected to lead to a behavior similar to that observed for
dissociation on the ⌸ state. Photodissociation following
excitation of the ⌸ state is not included in the MD simu-
lations and for reasons described in Sec. II excitation to this
state turns out not to be important to the present work.
The high isomerization yield predicted by the MD simu-
lations and the possibility of observing the formation of both
ICN and INC make ICN a promising candidate for studying
the details of geminate recombination processes in liquids
1
0
3
1
1
–6
vent cage.
These studies indicate that upon excitation to
the dissociative potential energy surface the molecule never
really dissociates but rather performs a few oscillations
around a highly distorted geometry dictated by the excited
state potential and the solvent barrier before it nonadiabati-
cally transfers to the ground state.
The photodissociation and subsequent recombination of
solvated iodine cyanide may progress differently: Gas-phase
iodine cyanide has two bound geometries with a calculated
3
3
⌸0
⌸0
ϩ
3
ϩ
3
3
1.23 eV difference between the binding energy of ground
1
7
1
state ICN ͓D (I–CN)ϭ3.17 eV͔ and its metastable isomer,
0
8
INC. A barrier of 0.8 eV separates the two geometries and
3
while both molecules are linear their interconversion transi-
tion state is virtually an isosceles triangle. The lowest ex-
1
8
1
1
1
ϩ
cited states of ICN, accessible from the ICN(X ⌺ ) ground
3
3
1
state, are ⌸ , ⌸ , and ⌸ . When excited to these states
0ϩ
1
1
with h␯ϳ5 eV, gas-phase ICN dissociates into nearly equal
2
2
ϩ
2
2
ϩ
amounts of I( P3/2)ϩCN( ⌺ ), and I( P )ϩCN( ⌺ ) as
1/2
a͒Electronic mail: Keiding@chem.au.dk
0021-9606/2002/116(18)/7997/9/$19.00
7997
© 2002 American Institute of Physics
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
30.113.111.210 On: Wed, 24 Dec 2014 10:23:40
1

7998
J. Chem. Phys., Vol. 116, No. 18, 8 May 2002
Larsen et al.
FIG. 2. The static absorption spectra of INC͑Ar matrix͒—Ref. 11, CN͑g͒—
2
Refs. 7 and 35, and I( P3/2)͑aq͒—Ref. 18 and ICN. The absorption spec-
FIG. 1. Potential energy surface ͑Ref. 34͒ and absorption spectrum
insert͒—Ref. 15—of gas-phase ICN. Photodissociation of ICN results in
trum of aqueous ICN was measured in this work. The absorption spectrum
of I( P ) was estimated from the I( P3/2) absorption spectrum utilizing the
approach described in Refs. 20 and 17.
͑
2
2
2
2
ϩ
2
2
ϩ
production of I( P )ϩCN( ⌺ ) and I( P )ϩCN( ⌺ ). Only the
1/2
3
/2
1/2
2
2
ϩ
I( P )ϩCN( ⌺ ) channel correlates with the ICN ground state.
3
/2
and solid rare-gas matrices.10–14 Experimentally, liquid phase
photodynamics of ICN has so far only been addressed by
Wan et al., who studied the photolysis of ICN in chloroform
mum is thus blueshifted by 28 nm and the extinction coeffi-
cient is nearly twice of that reported for gas-phase ICN.
9
,15,16
1
ϩ
1
The X ⌺ → ⌸ transition responsible for the short wave-
1
1
2
by monitoring the transient absorption of the CN molecule.
length part of the ICN͑g͒ absorption spectrum has little in-
tensity at the 266 nm photolysis wavelength used in the
present measurements ͑see Fig. 1͒. Hence, assuming the
blueshift of ICN͑aq͒ arises from a common spectral shift of
all three transitions this paper mainly studies the photody-
These measurements showed an instrument limited ͑Ͻ300
fs͒ formation of CN followed by ϳ2–4 ps decay ascribed to
recombination of CN with I within the solvent cage and a 76
ps decay ascribed to H or Cl abstraction from the solvent.
However, the strong UV absorption of chloroform precluded
the observation of the iodine fragments as well as the recom-
bination products, and the process of photoisomerization of
ICN within the solvent cage has therefore yet to be observed
experimentally.
3
3
namics following excitation to the ⌸ and ⌸ states.
1
0
2
The absorption spectrum of aqueous I( P3/2), shown in
Fig. 2, results from a charge transfer complex between
2
17–19
ground state I( P3/2) and water.
The absorption spec-
2
trum of the I( P1 ):H O has, to the best of our knowledge,
/2
2
In this work we present direct experimental evidence for
subpicosecond photoisomerization of aqueous ICN followed
by efficient energy dissipation to the solvent, which cools the
isomer to its lowest vibrational levels in Ϸ1 ps. Second, our
measurements show that while a substantial part of the ICN
reformation occurs by diffusion 5–10 ps after the photodis-
sociation, the INC molecules are exclusively produced by
geminate recombination inside the first solvation shell. The
lack of INC formation by diffusive recombination is ascribed
to steric hindrance by water molecules hydrogen bonding to
the nitrogen atom of the solvated CN radical. The photolysis
of aqueous ICN thus presents a striking example of how
solvents may both assist and inhibit chemical reactions.
not been determined experimentally, nor have we found
any reports on its extinction coefficient. Instead, the spectral
shift relative to that of the ground state complex may, to a
first approximation, be calculated using a semiempirical
approach,1
7,20
which essentially shift the absorption band by
2
the fine-structure splitting. The I( P ):H O spectrum pre-
sented in Fig. 2 is that of the I( P3 ):H O complex red-
1
/2
2
2
/2
2
shifted to 314 nm, while keeping the extinction coefficient of
the ground state complex.
Likewise, INC has never been observed in liquid solu-
tion and the only absorption spectra available are those mea-
sured in solid argon and krypton matrices.11 The extinction
coefficient of INC in Argon presented in Fig. 2 has been
derived from the spectrum of INC calibrated with the known
ICN extinction coefficient using the relative yield of ICN and
INC from the steady state photolysis of ICN in argon
matrices.8 However, just as the extinction coefficient of
aqueous ICN is almost twice the value measured in gas
phase, the extinction coefficient of aqueous INC may differ
from that in solid noble gases, and if so, the quantum yields
reported in the present work will change accordingly.
The gas-phase CN radical is known to have a very strong
absorption at 388 nm pertaining to the X→B transition ͑see
Fig. 2͒ and measurements reported by Wan et al.12 also indi-
cate a strong blue absorption in chloroform. Studies of the
spectral shift of the CN radical in noble gas matrices show
II. STATIC ABSORPTION SPECTRA
,11
The experimental technique of transient absorption spec-
troscopy identifies the species involved in the photolysis by
their absorption spectra. Hence, this section presents the
steady state absorption spectra of ICN, INC, I, and CN. The
static absorption spectra of aqueous ICN recorded by a
Uvikon 860 ͑Kontron Instruments͒ spectrophotometer with a
spectral resolution of 2 nm is shown in Fig. 2. The spectrum
peaks at 222 nm with a maximum extinction coefficient of
Ϫ1
Ϫ1
1
81Ϯ5 M cm and has a width of 50 nm ͓full width at
half maximum ͑FWHM͔͒. The ICN͑aq͒ absorption maxi-
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.113.111.210 On: Wed, 24 Dec 2014 10:23:40

J. Chem. Phys., Vol. 116, No. 18, 8 May 2002
The photoisomerization of aqueous ICN
7999
Ϫ1
Iodine cyanide was prepared by dissolving 27 g ͑0.55
mol͒ of sodium cyanide in 100 ml, three times distilled water
at 0° C. A total of 127 g of iodine ͑0.50 mol͒ was added in
small portions while stirring the solution at 0 °C. Subse-
quently, 120 ml of diethyl ether was added and the solution
was stirred for 15 min. The aqueous layer was separated
from the ether layer in a precooled separatory funnel and
extracted six times with cold diethyl ether. The ether was left
standing at room temperature for 1 h in a beaker with a large
surface area allowing most of the ether to evaporate. The
resulting white, needle-shaped crystals were separated from
the residual ether by suction filtration and washed five times
with cold water. The crystalline iodine cyanide was air dried
for 30 min yielding a total of 30–40 g ͑50%–60%͒ iodine
cyanide.
significant redshifts increasing to as much as 1000 cm in
CN–Xe complexes.21 The absorption spectrum of CN in wa-
ter, on the other hand, has not been determined experimen-
tally and is likely to be perturbed by the highly polar envi-
ronment. In order to estimate the spectral shift caused by
hydration we have calculated the excitation energy of the
CN:H O complex. First the X→B excitation energy of gas-
2
phase CN was calculated using multiconfigurational self-
consistent field theory with a correlation consistent one-
electron basis set augmented with one set of diffuse
functions with triple zeta quality ͑/aug-cc-pVTZ͒. The result-
ing gas-phase excitation energy of E(CN)theoryϭ3.31 eV is
in good agreement with the experimental value of
E(CN)expϭ3.20 eV. Having tested the calculations on gas-
phase CN the excitation energy of the CN:H O complex was
2
Immediately before use, 8.5 g of iodine cyanide was
dissolved in 0.5 l of three times distilled water, yielding a
concentration of cϭ0.11 M corresponding to an optical den-
sity of ODϭ0.7 at 266 nm in the 1.8-mm-thick flow cell. The
flow rate was adjusted to ensure a fresh sample of ICN for
each laser pulse. The reproducibility of the transient absorp-
tion data was tested among consecutive scans as well as by
repeating the experiments on different days using different
samples. No measurable degradation of the ICN solution was
observed during a measurement, but the slow buildup of per-
manent photoproducts necessitated frequent replacement of
the ICN solution. From numerous measurements we found
that the data could be measured on a common scale with an
uncertainty of Ϯ10%.
then calculated for a collinear geometry with the OH bond
aligned along the CN bond and for the two separate cases of
H facing C and N, respectively. The intermolecular distance
between H and C ͑and H and N͒ was varied from 3.36 to 3.9
Å. The resulting excitation energy of the X→B transition
was redshifted by 0.02–0.7 eV ͑377–475 nm͒ depending on
the position of the CN radical relative to the water molecule.
In addition to the pronounced spectral shift, the high sensi-
tivity to the solvent geometry suggests a strong broadening
of the CN͑aq͒ absorption spectrum relative to that of CN͑g͒.
Although lacking spectroscopic properties of the reac-
tants necessitate several assumptions and calculations, the
experimental data to be presented give evidence for a con-
sistent picture of the molecular dynamics as well as the static
absorption spectra.
IV. EXPERIMENTAL RESULTS
III. EXPERIMENTAL SETUP
A. Assignment of the absorption transients
The transient absorption spectrometer used in this work
is identical to the one described by Thomsen et al.22 and is
only described briefly: 800 nm pulses emitted by an ampli-
fied titanium:sapphire laser were frequency tripled to gener-
ate the 266 nm pump pulses utilized for photolyzing ICN.
The pump beam was sent through a variable delay line and a
The photoinduced absorption, ⌬A(␭,t), of aqueous ICN
produced by the 266 nm pump pulse and measured at
Ϫ1
500 cm intervals in the spectral range from 227 to 714 nm
is shown in Fig. 3. The measurements are presented in two
graphs for clarity.
Common to all curves in the interval 227–377 nm ͓Fig.
3͑a͔͒ is an initial sharp peak originating from two-photon
absorption in water.22 Hence, we take the position and width
of this peak to mark the point of zero delay and the experi-
␭
/2 wave plate, before it was focused behind the sample cell
by an fϭ50 cm CaF lens. The profile of the pump beam at
2
the sample cell was measured by means of a scanning pin-
hole and the semiaxes of the elliptical beam crosssection
were 0.16 mmϮ0.01 mm and 0.11 mmϮ0.01 mm ͑half
width at half maximum͒. The pump pulse energy was kept at
mental temporal resolution ͑0.5 ps͒, respectively. The two-
Ϫ
aq
photon ionization of water produces hydrated electrons, e ,
and OH and H radicals. As all the absorption transients
across the entire spectrum from 227 to 714 nm are measured
on the same absolute scale, the absorption transients around
700 nm resulting from the hydrated electron set an upper
limit on contributions from the photoproducts arising from
two-photon ionization of water. Accordingly, the absorption
pertaining to ionization of water does at most contribute by
1.5 mOD from 227 to 377 nm in good agreement with ab-
sorption measurements performed at selected wavelengths
when the ICN solution is replaced by neat water. The tran-
55 ␮J in order to reduce contributions to the measured ab-
sorption transients from hydrated electrons and OH and H
radicals produced by two-photon dissociation of water.22 The
probe pulses were generated by a combination of supercon-
tinuum generation, second harmonic generation, and sum-
frequency mixing. A beamsplitter divided the probe beam
into a signal and a reference beam, which were measured by
two matched photodiodes and boxcar integrators. The pump
beam was modulated at 0.5 kHz, locked to the 1 kHz repeti-
tion rate of the laser amplifier, and made to cross the signal
beam inside the sample cell at an angle of Ϸ5°. A ␭/2 wave
plate in the pump beam was adjusted so as to keep the po-
larization of the pump beam perpendicular to that of the
probe beam.
sient absorption data presented in the subsequent figures
Ϫ
aq
have been corrected for the absorption of OH, H, and e .
The absorption transients from 250 to 377 nm attain
their maximum within 0.5 ps and decay by 2/3 to a nearly
constant level during the first 20 ps. The spectral profile re-
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.113.111.210 On: Wed, 24 Dec 2014 10:23:40

8000
J. Chem. Phys., Vol. 116, No. 18, 8 May 2002
Larsen et al.
the concentration of INC molecules changes by less than
Ϯ10% during the next 100 ps. The fast formation of the INC
molecules indicates that INC is produced by photoisomeriza-
tion of ICN within the solvent cage, but the peak pertaining
to two-photon absorption in water obscures the photodynam-
ics of the first 0.5 ps making it impossible to monitor the
transition state dynamics.
The spectral range from 385 to 714 nm is dominated by
strong absorption peaking at 667 nm with a spectral shape
and temporal development somewhat resembling that re-
ported for hydrated electrons produced by two-photon ab-
sorption at 266 nm.22 However, measurements obtained
when substituting neat water for the aqueous ICN solution
without changing the alignment of the setup displayed a nar-
rower transient absorption spectrum of the hydrated electron
with a slightly faster decay than that observed for the ICN
solution. The difference between the two data sets clearly
reveals the presence of an additional species appearing as the
broad shoulder around 450 nm in Fig. 3͑b͒. The absorption
of this species decays to roughly 45% of its initial value after
1
00 ps, but the nonlinear pump intensity dependence of the
absorption pertaining to the hydrated electron prevents the
accurate determination of its spectrum and temporal devel-
opment. The spectral characteristics of the shoulder are in
good accord with those predicted in Sec. II for the CN radi-
cal in water, whereas none of the other products expected
from the photolysis of ICN absorb in this spectral region.
Hence, we ascribe the 450 nm hump to CN and note that the
decay in the CN͑aq͒ absorption transients follows, within the
uncertainty of its determination, the dynamics of the aqueous
iodine absorption confirming the notion that the iodine and
cyanide concentrations decay due to diffusive recombination.
FIG. 3. Pseudo-three-dimensional plot of the induced absorption, ⌬A(␭,t),
produced when the 266 nm pump pulse photodissociates aqueous ICN. The
transients are shown in two graphs for clarity. ͑a͒ The measurements from
2
27 to 377 nm; ͑b͒ the transients from 385 to 714 nm. The transients per-
taining to two-photon absorption in water have been truncated at 18 mOD.
B. Quantum yields
mains constant at all delays and is well approximated by the
2
2
The induced absorption pertaining to INC completely
masks the photoinduced depletion of ground state ICN, ren-
dering a direct determination of the quantum yields for I and
CN formation and recombination to INC and ICN difficult.
Assuming the transient absorption data represent the equili-
brated absorption of the species, the quantum yields for the
photochemical processes may instead be estimated from the
measured induced absorption, ⌬A, following the approach
described by Thomsen et al.24 Accordingly, the quantum
yield of a species, X, can be expressed as
absorption spectra of I( P3 ) and I( P1/2) assuming a
/2
0
.45:0.55 ratio between the total absorption of the two fine-
structure components. Hence, we ascribe the absorption tran-
sients to atomic iodine and note that both dissociation chan-
nels are populated in aqueous solution. The iodine atoms are
thus observed Ͻ0.5 ps after photoexcitation of ICN and the
slow decay in the iodine concentration indicates diffusive
recombination with the CN fragments. The similarity of the
transient absorption dynamics pertaining to I( P ) and
I( P ) seems surprising considering that only the I( P3/2)
ϩCN( ⌺ ) channel correlates with the ICN( ⌺ ) ground
state. However, studies of the photolysis of, for instance,
2
3
/2
2
2
1
/2
2
ϩ
1
ϩ
⌬
A͑␭,t͒
⌽͑X,t͒ϭ
,
͑1͒
log͑Ano pump͒Ϫlog͑Apump
͒
CS ͑aq͒ show that the solvent–solute interaction greatly en-
2
hances nonadiabatic transitions and facilitates recombination
where the absorption of the probe pulse with overlapping
pump pulse, Apump , and without pump pulse, Ano pump , is
determined from the pump pulse intensity, the length of the
sample cell, and the extinction coefficients of the involved
species.
processes that otherwise would be forbidden.23
The strong induced absorption covering the range from
227 to 240 nm rises to its final value in 1 ps. The spectra of
the possible photoproducts presented in Fig. 2 peak at sig-
nificantly longer wavelengths and cannot readily account for
the absorption transients. However, guided by the 0.63 eV
2
Thus, the quantum yield of iodine atoms ͓I( P3/2) and
2
I( P1/2)͔ after 1 ps determined from the absorption transients
͑
28 nm͒ solvent induced blueshift observed for aqueous ICN
at 294 nm, ⌬A ͑294 nm, 1 ps͒ϭ12.1 mOD, is ⌽(I, 1 ps)
ϭ38%, while subsequent recombination reduces the quan-
tum yield for iodine atoms surviving the first 100 ps
͓⌬A͑294 nm, 100 ps͒ϭ3.1 mOD͔ to ⌽͑I, 100ps͒ϭ10%. Pro-
and the vicinity of the strong INC absorption spectrum mea-
sured in noble gas matrices, we ascribe this spectral feature
to aqueous INC. Accordingly, INC is produced in Ͻ1 ps and
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.113.111.210 On: Wed, 24 Dec 2014 10:23:40

J. Chem. Phys., Vol. 116, No. 18, 8 May 2002
The photoisomerization of aqueous ICN
8001
FIG. 5. Model of the photolysis of ICN͑aq͒ when excited at 266 nm. See the
text for details.
FIG. 4. Graphic representation of the measured quantum yields of ICN,
INC, and iodine after 1, 10, 20, and 100 ps. The values should be taken as
approximate estimates only due to the uncertainties pertaining to the extinc-
tion coefficients and the pulse intensity. See the text for details.
the lack of diffusive reformation of INC is likely a conse-
quence of solvation. After dissociation and ejection of CN
from the water cage the CN fragment thermalizes and be-
comes hydrated. Due to the higher electronegativity of N, the
water molecules principally form hydrogen bonds with the
nitrogen end of the CN radical leaving the C atom more
accessible to reactions with passing iodine atoms. Hence,
ICN is produced by diffusive recombination, while INC is
not. Geminate recombination inside the cage, on the other
hand, is distinctly different: As shown by Benjamin’s MD
simulations, ICN and INC may be formed during the first
encounter of the recoiling I and CN fragments or result from
interconversions during the initial stages of vibrational relax-
ation on the electronic ground state following recombination.
In either case, both isomers are produced and the preferred
geminate reformation of ICN likely reflects the fact that ICN
is the more stable of the two isomers.
vided the concentration of electronically excited ICN mol-
ecules after 1 ps is insignificant, the quantum yield for gemi-
nate recombination to either ICN or INC within 1 ps is thus
⌽
͑cage back, 1 ps͒ϭ62%. The quantum yield of INC was
determined from the induced absorption at 230 nm, ⌬A ͑230
nm, 1 ps͒ϭ12.4 mOD corrected for the minor contributions
from ICN and I using the INC͑aq͒ absorption spec-
trum normalized to the maximum extinction coefficient of
Ϫ1
Ϫ1
INC͑Ar͒ (3739 M cm ). The resulting yield is ⌽͑INC,
ps͒ϭ11% leaving ⌽͑ICN, 1 ps͒ϭ51% for the production of
1
ground state ICN within the first picosecond. As some of the
quantities entering the calculations are vitiated with signifi-
cant uncertainties the relative uncertainty of the above-listed
quantum yields is estimated to Ϯ25%. The quantum yields
are summarized graphically in Fig. 4.
A. Modeling the photolysis
Hence, we conclude that geminate recombination within
the solvent cage during the first picosecond is about five
times more likely to produce ICN than INC. We also note
that the decay in the iodine concentration caused by diffusive
recombination with CN molecules outside the cage occurs on
a 8 ps time scale and thus significantly slower than the for-
mation of the INC molecules. Now, the extinction coefficient
of INC is much larger than that of I potentially leading to a
strong increase in the 230 nm absorption transient if even a
small fraction of iodine atoms were to slowly recombine
with CN and form INC. This is not observed and we there-
fore conclude that the formation of INC occurs inside the
water cage within 1 ps with less than 10% of the INC mol-
ecules produced by subsequent diffusive recombination.
Hence, iodine atoms that have not recombined with CN to
form INC or ICN within the solvent cage and therefore still
exist as free species after 1 ps, diffusively recombine with
CN outside the cage to form ground state ICN on a 8 ps time
scale. The resulting total recombination yield for producing
ICN after 100 ps is thus ⌽͑ICN, 100 ps͒ϭ79%.
In accordance with the assignments made in Sec. IV A,
we have simulated the absorption transients induced by a 150
fs ͑FWHM͒ excitation pulse using a first-order rate equation
description on the reaction scheme depicted in Fig. 5. Here
I––CN is the ICN molecule excited to one of its dissociative
states, CN is the ground state CN radical, I represents
2
2
I( P ) as well as I( P1/2), and INC denotes the ground
3
/2
state INC isomer. The nonexponential behavior of the diffu-
sive recombination process between I and CN is for simplic-
ity simulated by assuming ad hoc that part of the CN and I
fragments separate permanently. While the diffusive recom-
bination process is more correctly described by, for instance,
the Smoluchowski equation, the time-dependent concentra-
tions of I, CN, and ICN governed by diffusive recombination
are well approximated by the exponential evolution resulting
from the rate equation description.
The induced absorption transients from 227 to 377 nm
resulting from the simulations are compared to their experi-
mental counterparts in Figs. 6 and 7͑a͒, while the rate con-
stants resulting from the fit are compiled in Table I. A very
satisfactory fit to the measured evolution of all 36 transients
is obtained confirming the consistency of the reaction
scheme with the available data and the static absorption
spectra. The lack of a CN absorption spectrum prevents the
extension of the simulations to longer wavelengths. The
simulations have utilized the absorption spectra of ICN,
V. DISCUSSION
The very different recombination behavior of ICN and
INC indicates the presence of a barrier for the diffusive ref-
ormation of INC. Since ab initio gas-phase calculations in-
1
1
dicate that the radical reaction IϩCN→INC is barrierless,
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.113.111.210 On: Wed, 24 Dec 2014 10:23:40

8002
J. Chem. Phys., Vol. 116, No. 18, 8 May 2002
Larsen et al.
FIG. 7. ͑a͒ Relative concentrations of ICN, I, CN, and INC as a function of
time derived from the simulation of the absorption transients. ͑b͒ Compari-
son of the measured and the fitted ͑curves͒ transient absorption spectra 5 and
50 ps after the photodissociation pump pulse.
range from 227 to 250 nm. The resulting INC spectrum
peaks at 212Ϯ5 nm with a width of 40Ϯ5 nm ͑FWHM͒. The
blueshift of approximately 38 nm is similar to the blueshift
of ICN͑aq͒ and the width is only a few nanometers wider
than that of INC in solid argon and krypton.11
The concentration dynamics of ICN, I, CN, and INC
shown in Fig. 7͑b͒ reveal that more than half of the ICN
molecules are reformed during the pump pulse and the rest
are produced by recombination on an 8 ps (1/e) time scale.
Geminate recombination forming INC occurs in 1 ps and the
good agreement with the experimental data is obtained as-
suming no additional contributions to the INC concentration.
In particular, the transient absorption spectra show no sign of
HCN resulting from hydrogen abstraction nor does it reveal
FIG. 6. ͑Color͒ Contour plot of ͑a͒ the measured transient absorption from
27 to 377 nm and ͑b͒ the transient absorption fit based on the model
depicted in Fig. 5. The INC isomer is clearly observed as the nearly constant
strong absorption at the shortest wavelengths while the two iodine channels
give rise to the absorption centered at 308 nm.
2
TABLE I. Rate constants used with the model presented in Fig. 5 to simu-
late the transient absorptions.
2
2
I( P ), and I( P1/2) depicted in Fig. 2 with the 0.45:0.55
3
/2
Rate constant (psϪ1
Reaction
Dissociation
)
ratio between the extinction coefficientϫconcentration prod-
uct for the two iodine fine-structure components determined
in Sec. IV A. The maximum extinction coefficient of INC is
fixed at 3739 M cm , while the wavelength of maximum
absorption and spectral width have been adjusted to give the
best overall agreement with the absorption transients in the
3.8
Geminate reformation of ICN
Geminate isomerization
5.2
1.0
Ϫ1
Ϫ1
Diffusive reformation of ICN
Permanent cage escape
0.125
0.038
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.113.111.210 On: Wed, 24 Dec 2014 10:23:40

J. Chem. Phys., Vol. 116, No. 18, 8 May 2002
The photoisomerization of aqueous ICN
8003
the presence of any vibrationally excited species. The vibra-
tional relaxation of ICN and INC formed by geminate re-
combination is addressed in the Sec. V B.
B. Vibrational relaxation of INC and ICN
Geminate recombination through nonadiabatic transfer
to the electronic ground state leaves the molecules in highly
excited vibrational states close to the dissociation limit. In
the triatomic molecules ClO ͑aq͒ and CS ͑aq͒ the absorption
2
2
spectra of the vibrationally excited species are more than
three times as wide as that of the vibrational ground state
resulting in wide transient absorption spectra during the first
10 ps after recombination. These spectra narrow and ap-
proach that of the vibrational ground state as the molecules
relax. In contrast, the good agreement between the ICN͑aq͒
photolysis experiment and model is obtained using equili-
brated INC and ICN absorption spectra at all times. This
indicates that INC and likely also ICN ͑although its absorp-
tion is barely detectable͒ have only little vibrational energy
after 1 ps, implying that most of the ϳ3 eV excitation energy
remaining after recombination is dissipated to the solvent on
a subpicosecond time scale.
To further investigate the vibrational relaxation process
of INC͑aq͒ we have calculated the relaxation time using both
the semiclassical golden rule formalism and classical mo-
lecular dynamics simulations. The two calculations show that
subpicosecond vibrational relaxation is possible only if it
occurs in the bending or IN stretching modes.
FIG. 8. Force autocorrelation function of ClO in water ͑Ref. 25͒.
2
The fundamental vibration frequencies of INC calculated
8
Ϫ1
Ϫ1
by Samuni et al. are 479.2 cm
͑IN stretch͒, 247 cm
Ϫ1
͑
bend͒, and 2026.5 cm
͑CN stretch͒. The relatively low
vibrational frequencies of the IN stretch and INC bending
modes result in force correlation function amplitudes, which
are nearly two orders of magnitude higher than that of the
fundamental asymmetric stretch of OClO͑aq͒, while the am-
plitude of the force autocorrelation function associated with
the high frequency CN stretch is an order of magnitude
lower ͑see Fig. 8͒. Owing to the harmonic ground state po-
tential at low energies, the dominating coupling matrix ele-
ments of the rate limiting transitions between the lowest vi-
brational states of INC are approximately equal to the
harmonic coupling elements of OClO͑aq͒. Consequently, if
the vibrational relaxation predominantly proceeds via the
bending or IN stretch modes, the relaxation time of INC͑aq͒
is expected to be of the order of 0.1–1 ps in accord with the
fast relaxation observed experimentally. If, on the other
hand, the rate limiting vibrational transitions close to the
bottom of the ground state potential involves the CN stretch
The two theoretical treatments are now discussed in de-
tail beginning with the golden rule approach described by,
for instance, Poulsen et al.25 According to this description
the transition rate kij between two adjacent levels, i and j,
can be expressed as
3
បϪ2
2
k ϭ
ij
͚ ͗i
ϩexp͑Ϫប␻ /k T͒ ␣,␤ϭ1
ij B
͉q␣͉j
͘
1
ϱ
ϫ
͗
j
͉
q␤
͉
i
͘
͵
dt exp͑i␻ t͒
͗
F ͑t͒F ͑0͒
͘
,
͑2͒
ij
␣
␤
Ϫϱ
we estimate a relaxation time in excess of 10 ps. The vibra-
Ϫ1
where ␻ ϭE ϪE is the energy spacing between the two
tional frequencies of ICN are 486 cm
͑C–I stretch͒,
ij
i
j
Ϫ1
Ϫ1
8
vibrational levels, q and q are dimensionless normal coor-
305 cm
͑bend͒, and 2188 cm
͑CN stretch͒ and the
␣
␤
dinates, and F is the classical bath force along normal mode
same arguments therefore also suggest that vibrationally ex-
cited ICN͑aq͒ relax to the vibration ground state on a subpi-
cosecond time scale.
␣
␣
. Accordingly, the relaxation time ␶ ϭ1/kij is inversely
ij
proportional to both the vibrational coupling elements and
the classical bath force autocorrelation spectrum evaluated at
the vibrational energy gap. Force autocorrelation functions
calculated for the relevant normal modes in the aqueous sys-
The MD simulations of the vibrational relaxation em-
ploy a fixed number, volume, and energy (NVE) ensemble of
3
0
one INC molecule and 511 flexible SPC H O molecules in
2
2
6
Ϫ
27
Ϫ
28
tems of H O in H O, N3 in water, CN in water, and
a cube with a side length of 24.86 Å corresponding to a
2
2
2
5
3
ClO in water differ by less than an order of magnitude.
density of 0.997 g/cm . The INC geometry is optimized by a
2
Hence, the force spectrum depends only moderately on the
solute and is for the present estimate of vibrational relaxation
DFT/B3LYP calculation using the LANL2DZ basis set while
the intramolecular force field of INC is derived from elec-
tronic structure calculations utilizing the GAUSSIAN 98 pro-
of aqueous INC approximated by that of ClO in water cal-
2
culated by Poulsen et al.25 and depicted in Fig. 8. Adjusting
gram package. The resulting bond lengths and frequencies,
31
2
5
expression ͑2͒ for quantum mechanical corrections leads to
listed in Table II, compare favorably to theoretical values
8
a 14 ps vibrational relaxation time (1/e) of the first excited
obtained by Samuni et al. as well as experimental values
Ϫ1
asymmetric stretch mode (v ϭ1130 cm ) of OClO͑aq͒ in
measured in argon matrices. A normal mode analysis based
on GAUSSIAN 98 calculations leads to the following approxi-
mate valance bond intramolecular potential:
3
good agreement with the 10 ps (1/e) relaxation time ob-
served experimentally.29
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.113.111.210 On: Wed, 24 Dec 2014 10:23:40

8
004
TABLE II. Spectroscopic properties of INC.
Samuni et al. Argon matrix ͑expt͒
J. Chem. Phys., Vol. 116, No. 18, 8 May 2002
Larsen et al.
Spectroscopic property This work
͑Ref. 8͒
͑Ref. 8͒
rIN ͑Å͒
rCN ͑Å͒
2.004
1.203
3.57
273
459
2073
0.992
0.182
0.0135
1.991
1.203
3.50
227
479
2073
␮
͑D͒
␻bend (cmϪ1)
Ϸ200
494
2057.5
Ϫ1
␻IN (cm
␻CN (cmϪ1
kCN (hartree/bohr )
kNI (hartree/bohr )
)
)
2
2
2
k␪ (hartree/rad )
1
2
eq
CN
2
1
2
eq
IN
2
1
2
2
VINCϭ kCN͑rCNϪr ͒ ϩ k ͑r Ϫr ͒ ϩ k ͑␪Ϫ␲͒
IN IN
␪
͑3͒
FIG. 9. Simulated vibrational energy relaxation of the IN stretch excited by
Ϫ1
2080 cm . The vibrational excitation energy decays in 1.2 ps (1/e).
with the force constants listed in Table II. Consistent with
Eq. ͑3͒, r and rIN become actual vibrational modes of the
CN
3
2
molecule. The electrostatic charges on I(qϭ0.16 e), C(q
ϭϪ0.27 e), and N(qϭ0.11 e) are calculated from the
Merz–Kollman scheme, while the van der Waals interactions
between solute and solvent are described by Lennard-Jones
of energy relaxation within the first 5 ps likewise supporting
the golden rule estimate. Vibrational relaxation in the bend-
ing mode was not investigated, but from the force correlation
amplitude depicted in Fig. 8 it is expected to be even faster
than relaxation in the IN stretch. In accordance with the con-
clusions based on the golden rule estimates, the MD simula-
tions thus lead us to infer that the CN stretch is not signifi-
cantly involved in the vibrational relaxation of INC.
interactions between any atom in INC and H O. The particu-
2
lar ␧ and ␴ parameters of the H O–INC Lennard-Jones po-
2
tential are derived by applying standard Lorentz-Berthelot–
mixing rules to single atom–atom ␧ and ␴ values obtained
from Refs. 33 and 34 and listed in Table III.
The system is propagated in steps of 0.5 fs using the
3
3
Verlet algorithm and standard periodic boundary conditions
with half box-length spherical cutoff. First, the initial 20 ge-
ometries used in the simulations are extracted from a 20 ps
trajectory run equilibrated at 300 K by sampling one system
configuration for every picosecond. The INC molecule is
then excited by deforming the bond in question and propa-
gated for 5 ps, while monitoring the total energy of the INC
molecule. Initial excitation energy of 10 kT is chosen in or-
der to follow the rate limiting transitions close to the bottom
of the potential while keeping the potential energy within the
valence bond approximation.
C. Comparison with MD simulations of ICN in
chloroform
The detailed experimental information about the
ICN͑aq͒ photodynamics now enables the comparison with
the MD simulations performed by Benjamin.10 However, due
to the strong blueshift of the ICN͑aq͒ absorption spectrum
relative to that in chloroform and the fact that hydrogen
bonded water constitutes a stronger solvent cage than chlo-
roform, perfect agreement with theory is not to be expected.
Nevertheless, the estimated 62% quantum efficiency for
geminate recombination of ICN in water compares favorably
with the simulations and the Ϸ5:1 ratio between the quantum
yield of ICN and INC measured in aqueous solution is also
in fair agreement with the simulated recombination yields.10
The reaction dynamics are faster in water, though. According
to the MD simulations the vibrationally excited molecule
formed by geminate recombination performs several inter-
conversions between the ICN and INC configurations before
energy dissipation to the solvent freezes the molecule in its
metastable or stable geometry after Ϸ3 ps. This is slower
than the 1 ps upper limit for the isomerization measured in
aqueous solution suggesting a more efficient energy dissipa-
tion to water.
Figure 9 shows the excess vibrational energy relaxation
Ϫ1
of INC when a potential energy of 10 k Tϭ2080 cm is
B
stored in the IN bond. The energy relaxation decays expo-
nentially with a time constant of 1.2 ps (1/e) leaving an
Ϫ1
excess vibration energy of only Ϸ250 cm
after 3 ps.
Hence, the fast relaxation of the IN stretch is in good agree-
ment with the result of the golden rule estimate. The corre-
sponding results for the CN bond dynamics showed no sign
TABLE III. Lennard-Jones parameters describing the interaction between
INC and H O.
2
Atom pair
␴ ͑Å͒
␧/kB ͑K͒
I–H
I–O
C–H
C–O
N–H
N–O
3.31
3.38
3.08
3.15
3.06
3.13
45.5
121.4
20.9
55.9
17.9
49.9
MD simulations of the geminate recombination of ICN
in chloroform show that the CN stretch becomes excited by
Ϸ2100 cm when ICN is formed on the electronic ground
state. The CN vibration persists for more than 5 ps after
geminate recombination and Benjamin infers that the vibra-
tional relaxation predominantly occurs in the bending and
Ϫ1
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.113.111.210 On: Wed, 24 Dec 2014 10:23:40

J. Chem. Phys., Vol. 116, No. 18, 8 May 2002
The photoisomerization of aqueous ICN
8005
1
2
3
CN stretch modes. The long CN stretch relaxation time de-
termined by Benjamin agrees with the calculated
relaxation time of the CN stretch of INC presented
previously, but the subpicosecond relaxation time observed
experimentally suggests that the CN stretch is not
significantly involved in the relaxation process. Further ex-
perimental as well as theoretical studies are obviously
needed to uncover the finer details of the vibrational relax-
ation of ICN and INC.
J. Thøgersen, C. L. Thomsen, J. A. Poulsen, and S. R. Keiding, J. Phys.
Chem. A 102, 4186 ͑1998͒.
C. L. Thomsen, J. Thøgersen, and S. Keiding, J. Chem. Phys. 114, 4099
͑
2001͒.
C. L. Thomsen, P. J. Reid, and S. R. Keiding, J. Am. Chem. Soc. 122,
12795 ͑2000͒.
S. Hess, H. B u¨ rsing, and P. V o¨ hringer, J. Chem. Phys. 111, 5461 ͑1999͒.
4
5
P. K. Walhout, C. Silva, and P. F. Barbara, J. Phys. Chem. 100, 5188
͑
1996͒.
6
7
8
T. K u¨ hne and P. V o¨ ringer, J. Chem. Phys. 105, 10788 ͑1996͒.
D. D. Davis and H. Okabe, J. Chem. Phys. 49, 5526 ͑1968͒.
U. Samuni, S. Kahana, R. Fraenkel, Y. Haas, D. Danovich, and S. Shaik,
Chem. Phys. Lett. 225, 391 ͑1994͒.
W. P. Hess and S. R. Leone, J. Chem. Phys. 86, 3773 ͑1987͒.
I. Benjamin, J. Chem. Phys. 103, 2459 ͑1995͒.
J. Helbing, M. Chergui, S. Fernandez-Alberti, J. Echave, N. Halberstadt,
and J. A. Beswick, Phys. Chem. Chem. Phys. 2, 4131 ͑2000͒.
C. Wan, M. Gupta, and A. H. Zewail, Chem. Phys. Lett. 256, 279 ͑1996͒.
I. Benjamin and R. Wilson, J. Chem. Phys. 90, 4176 ͑1989͒.
A. I. Krylov and R. B. Gerner, J. Chem. Phys. 100, 4242 ͑1994͒.
E. A. Coronado, V. S. Batista, and W. H. Miller, J. Chem. Phys. 112, 5566
9
VI. CONCLUSION
10
11
In summary, our study of the photolysis of aqueous ICN
at 266 nm shows that caging of the I and CN photoproducts
by the surrounding water molecules limits the I and CN
quantum yield to 38% after 1 ps. Diffusive recombinations
involving I and CN fragments that do escape the solvent
cage further reduces the yield to 10% during the subsequent
1
1
1
1
2
3
4
5
͑
2000͒.
1
1
1
6
7
8
W. M. Pitts and A. P. Baronavski, Chem. Phys. Lett. 71, 395 ͑1980͒.
A. Treinin and E. Hayon, J. Am. Chem. Soc. 97, 1716 ͑1975͒.
P. Fornier de Violet, R. Bonneau, and J. Joussot-Dubien, Chem. Phys.
Lett. 19, 251 ͑1973͒.
The spectrum of iodine is reported by Fornier de Violet and the extinction
coefficient is reported by Treinin et al. The basis for the extinction coef-
ficient measurement is not known.
R. Foster, Molecular Association ͑Academic, London, 1979͒.
A. Thoma, G. Schalmoser, A. M. Smidt, B. E. Wurfel, and V. E. Bondy-
bey, J. Chem. Phys. 100, 5387 ͑1994͒.
C. L. Thomsen, D. Madsen, S. R. Keiding, J. Thøgersen, and O. Chris-
tiansen, J. Chem. Phys. 110, 3453 ͑1999͒.
C. L. Thomsen, D. Madsen, J. Thøgersen, J. R. Byberg, and S. R. Keiding,
J. Chem. Phys. 111, 703 ͑1999͒.
C. L. Thomsen, D. Madsen, J. A. Poulsen, C. Rønne, J. Thøgersen, S. J. K.
Jensen, and S. R. Keiding, J. Chem. Phys. ͑to be published͒.
J. Poulsen, T. M. Nymand, and S. R. Keiding, Chem. Phys. Lett. 343, 581
100 ps. Diffusive recombination produces only ICN—not
INC. This product distribution is explained in terms of the
higher electronegativity of N than of C. Thus hydrogen
19
bonding between CN and H O will involve complexes like
2
CN¯H–OH, which polarize the orientation of CN and fa-
2
2
0
1
cilitate the reaction leading to ICN.
The majority of the dissociating ICN molecules retained
within the solvent cage recombine to form ICN in Ϸ1 ps,
while a minor fraction isomerizes to INC. ICN and INC relax
to the vibrational ground state in Ϸ1 ps in good agreement
with theoretical estimates based on the golden rule formal-
ism and MD simulations assuming the vibrational relaxation
proceeds via the IC/IN stretch and ICN/INC bending modes.
However, the details of the cage back reaction and the sub-
sequent vibrational relaxation process are not experimentally
observable with the present time resolution. Hence, the fun-
damental question of to what extent isomerization of ICN in
solution is due to rotational reorientation of the dissociating
fragments prior to recombination and to interconversion dur-
ing the vibrational relaxation is still unsettled and remains
open to further investigation.
2
2
2
3
24
25
͑
2001͒.
26
J. A. Poulsen ͑private communication͒.
2
2
2
7
8
9
A. Morita and S. Kato, J. Chem. Phys. 109, 5511 ͑1998͒.
R. Rey and J. T. Hynes, J. Chem. Phys. 108, 142 ͑1998͒.
J. A. Poulsen, C. L. Thomsen, S. R. Keiding, and J. Thøgersen, J. Chem.
Phys. 108, 8461 ͑1998͒.
K. Toukan and A. Rahman, Phys. Rev. B 31, 2643 ͑1985͒.
M. J. Frisch, W. Trucks, H. B. Schlegel et al., GAUSSIAN 98, Pittsburgh
30
31
͑
1998͒.
32
By adopting this particular form of the potential a mixed term of the type
eq
eq
kCN,IN(rCNϪr )(rINϪr ) has been neglected. However, this term is
CN
IN
2
small, kCN,INϭϪ0.97 eV/Å .
3
3
3
3
4
5
M. P. Allen and D. J. Tildesley, Computer Simulations of Liquids ͑Oxford
University Press, Oxford, 1987͒.
S. F. Alberti, J. Echave, V. Engel, N. Halberstadt, and J. A. Beswick, J.
Chem. Phys. 113, 1027 ͑2000͒.
ACKNOWLEDGMENT
The many fruitful discussions with Svend Knak Jensen
are highly appreciated.
J. U. White, J. Chem. Phys. 8, 79 ͑1940͒.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
30.113.111.210 On: Wed, 24 Dec 2014 10:23:40
1