Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

The effect of large amplitude motions on the

transition frequency redshift in hydrogen


bonded complexes: A physical picture
Cite as: J. Chem. Phys. 140, 184309 (2014); https://doi.org/10.1063/1.4873420
Submitted: 20 February 2014 . Accepted: 15 April 2014 . Published Online: 09 May 2014

Kasper Mackeprang, Henrik G. Kjaergaard, Teemu Salmi, Vesa Hänninen, and Lauri Halonen

ARTICLES YOU MAY BE INTERESTED IN

The effect of large amplitude motions on the vibrational intensities in hydrogen bonded
complexes
The Journal of Chemical Physics 142, 094304 (2015); https://doi.org/10.1063/1.4913737

Calculation of OH-stretching band intensities of the water dimer and trimer


The Journal of Chemical Physics 110, 9104 (1999); https://doi.org/10.1063/1.478832

Anharmonic vibrational properties by a fully automated second-order perturbative approach


The Journal of Chemical Physics 122, 014108 (2005); https://doi.org/10.1063/1.1824881

J. Chem. Phys. 140, 184309 (2014); https://doi.org/10.1063/1.4873420 140, 184309

© 2014 AIP Publishing LLC.


THE JOURNAL OF CHEMICAL PHYSICS 140, 184309 (2014)

The effect of large amplitude motions on the transition frequency redshift


in hydrogen bonded complexes: A physical picture
Kasper Mackeprang,1 Henrik G. Kjaergaard,1,a) Teemu Salmi,2 Vesa Hänninen,2
and Lauri Halonen2,b)
1
Department of Chemistry, University of Copenhagen, Universitetsparken 5,
DK-2100 Copenhagen Ø, Denmark
2
Laboratory of Physical Chemistry, Department of Chemistry, P.O. Box 55, A.I. Virtasen aukio 1, FI-00014,
University of Helsinki, Helsinki, Finland
(Received 20 February 2014; accepted 15 April 2014; published online 9 May 2014)

We describe the vibrational transitions of the donor unit in water dimer with an approach that is based
on a three-dimensional local mode model. We perform a perturbative treatment of the intermolecular
vibrational modes to improve the transition wavenumber of the hydrogen bonded OH-stretching
transition. The model accurately predicts the transition wavenumbers of the vibrations in water dimer
compared to experimental values and provides a physical picture that explains the redshift of the
hydrogen bonded OH-oscillator. We find that it is unnecessary to include all six intermolecular modes
in the vibrational model and that their effect can, to a good approximation, be computed using a
potential energy surface calculated at a lower level electronic structure method than that used for the
unperturbed model. © 2014 AIP Publishing LLC. [http://dx.doi.org/10.1063/1.4873420]

INTRODUCTION for the stretching of the hydrogen bonded OH-oscillator.


This transition is perhaps the most interesting vibrational
Hydrogen bonded complexes are widely studied in litera-
transition for any hydrogen bonded system, as it changes
ture due to their impact on atmospheric reactions, nucleation,
significantly relative to the corresponding transition for the
and radiative transfer processes, where especially complexes
non-hydrogen bonded monomer unit. The VPT2 method,
containing water have received a lot of attention.1–9 One of
which is a full-dimensional anharmonic normal mode cal-
the complexes, which has a large impact on our atmosphere,
culation, gave a good agreement relative to the reported
is water dimer, as it is predicted to contribute significantly
experimental value for the fundamental hydrogen bonded
to the absorption of solar radiation.8, 10, 11 A commonly used
OH-stretching transition.14, 32 The difference between the
tool in the study of hydrogen bonded complexes is vibrational
experimental value and the value predicted by the VPT2
spectroscopy.12–18 However, the vibrational spectra of water
method at the coupled-cluster method with single, double,
dimer and hydrogen bonded complexes in general are difficult
and pertubative triple excitations (CCSD(T))/AVTZ level of
to simulate accurately due to the large amplitude motions
theory was 10 cm−1 . The corresponding differences for the
involved. Especially, the hydrogen bonded OH-oscillator
VSCF and the HCAO local mode model were approximately
gives rise to discrepancies between experiment and theory if
100 and 35 cm−1 , respectively. Extending the local mode
the effect of large amplitude motions is ignored in the calcu-
model by using a close to exact kinetic energy operator and
lations. Many experimental14, 19–25 and theoretical11, 23, 26–39
potential energy surface, Salmi et al. obtained a value that
studies on the vibrational spectroscopy of water dimer exist
differs by 60 cm−1 .30
in the literature.
A full-dimensional anharmonic normal mode calcula-
Previously, fundamental and overtone transition
tion, such as a VPT2 calculation, is computationally demand-
wavenumbers and oscillator strengths of water dimer were
ing, and consequently for larger hydrogen bonded complexes
calculated, with various different vibrational models.32 The
and hydrogen bonded complexes with larger monomers than
methods used were the VPT2 method, where anharmonic
water, it is unfeasible to perform a VPT2 calculation at a
corrections to a normal mode calculation is calculated
high level of theory. An alternative approach is a reduced
through Rayleigh-Schrödinger vibrational second-order
dimensionality vibrational calculation like the local mode
perturbation theory, the vibrational self-consistent field
model.40–46 The local mode model generally gives good re-
(VSCF) method limited to two-mode coupling in the normal
sults, if the studied vibrations are anharmonic and have small
mode picture, and a three-dimensional harmonically coupled
coupling to other vibrational modes, which often is the case
anharmonic oscillator (HCAO) local mode model. The three
for XH-stretches, where X is typically O, N, or C. The dif-
models predict the vibrational transition wavenumbers of
ference between an experimental and a local mode calcu-
water dimer with varying accuracy compared to experimental
lated value for the fundamental transition wavenumber of
values, especially when considering the predicted value
non-hydrogen bonded XH-oscillators is typically less than 10
cm−1 .47 However, for hydrogen bonded XH-oscillators, large
a) Electronic mail: hgk@chem.ku.dk discrepancies between the experimentally observed values
b) Electronic mail: lauri.halonen@helsinki.fi
and results calculated by a local mode model are found. For

0021-9606/2014/140(18)/184309/9/$30.00 140, 184309-1 © 2014 AIP Publishing LLC


184309-2 Mackeprang et al. J. Chem. Phys. 140, 184309 (2014)

dimethylamine (DMA) dimer, the local mode calculated tran- THEORY AND CALCULATIONS
sition wavenumber of the hydrogen bonded NH-stretching
All electronic structure calculations were carried out us-
vibration is approximately 50 cm−1 smaller than that de-
ing MOLPRO 2010.1.49 The geometry of water dimer was op-
termined by experiment.16, 17 For a stronger complex like
timized and appropriate potential energy surfaces were calcu-
methanol-DMA (MeOH-DMA), the difference between the
lated at the CCSD(T)-F12a/VTZ-F12 and MP2/AVTZ levels
experimental wavenumber and a value calculated by a local
of theory. For the explicitly correlated coupled cluster calcu-
mode model is more than 100 cm−1 .18 The hydrogen bond
lations, we have used the default value of 1.0 for the expo-
strength in DMA dimer is comparable to that in water dimer,
nent in the correlation factor as suggested in the literature.50
but the hydrogen bond in the MeOH-DMA complex is signif-
The geometry optimization thresholds were set to: energy
icantly stronger.17, 18, 47 Thus, the difference between the the-
= 1 × 10−8 a.u., gradient = 1 × 10−7 a.u., and step = 1
oretically calculated transition wavenumber predicted by the
× 10−7 a.u. For all the single point calculations, the
local mode model and the experimentally measured transi-
thresholds were set to energy =1 × 10−9 a.u., orbital = 1
tion wavenumber increases with the strength of the hydrogen
× 10−8 a.u., and coeff = 1 × 10−8 a.u.
bond.
Different effects have been investigated to explain the
discrepancy between the observed and calculated transition
Definition of internal coordinates
wavenumber redshift of the bonded OH-oscillator in wa-
ter dimer.11, 30 It was found that counterpoise corrections We define the vibrations of water dimer in terms of in-
would increase the vibrational wavenumber by 5–10 cm−1 ternal coordinates. There are altogether 12 vibrations. For the
for the fundamental, and the effect of vibrational coupling to donor water unit, there are the bonded OHb -stretch, the free
the two other modes of the donor unit is expected to lower the OHf -stretch, and the donor Hf OHb bend (see Figure 1). For
wavenumber by ∼10 cm−1 . None of these effects fully ac- the acceptor unit, there are the two identical OH-stretches,
count for the discrepancy between experimental values and the OH1 -stretch and OH2 -stretch, and the acceptor H1 OH2
the calculated local mode values in the literature. In water bend. The remaining 6 modes are large amplitude intermolec-
dimer, the intramolecular vibrations of the donor unit are ex- ular motions. We describe these motions in terms of internal
pected to couple little with intramolecular vibrations of the ac- coordinates as the following six vibrations: the OO-stretch,
ceptor unit.32 It has been demonstrated by Wang and Bowman donor rock, donor twist, acceptor wag, acceptor twist, and tor-
that monomer-monomer coupling has an effect for water clus- sion. These vibrations are defined in the following way (see
ters with four or more water molecules, but in the simple case Figure 2 for a visualization of the modes):
of water dimer, this effect can be ignored.33 However, Wang The OO-stretching vector, R, is defined as the Oa −Od
et al. showed that inclusion of some of the intermolecular vector, where Oa is the acceptor oxygen and Od is the donor
modes perturb the OHb -stretch of water dimer significantly.34 oxygen.
In their vibrational configuration interaction calculations, they The donor rocking angle, β, is between the donor Hf OHb
included up to three intermolecular modes and found that angle bisecting vector, q, and the Od −Oa vector, −R:
the wavenumber of the fundamental OHb -stretch changes by
40 cm−1 . It was demonstrated by Kjaergaard et al. and sug- cos β = −uR · uq , (1)
gested by Salmi et al. that a local mode model including some
or all of the intermolecular large amplitude motions of water where uR and uq are unit vectors in the direction of the
dimer would improve the accuracy of the results.30, 32 Garden Od −Oa vector and the donor Hf OHb angle bisecting vector,
et al. found that the effect of the OO-stretch and the acceptor respectively.
wagging motion on the transition wavenumbers of the bonded
OHb -stretch was small.28, 48
Water dimer possesses 12 vibrations, 3 for the donor wa-
ter unit, 3 for the acceptor water unit, and 6 large amplitude
motions. In this work, we have used Rayleigh-Schrödinger
perturbation theory to determine the effect of the 6 large am-
plitude motions on the vibrational transition wavenumbers for
the donor unit of water dimer. The unperturbed system con-
sists of a three-dimensional local mode model of the donor
water unit, a three-dimensional local mode model of the ac-
ceptor water unit, and a one-dimensional treatment of each
of the 6 large amplitude motions. The vibrational energy lev-
els of the donor unit are perturbed by coupling each of the
three modes of the donor water unit with the 6 large am-
plitude modes in turn. From this perturbation, we have ob-
tained first-, second-, third-, and fourth-order corrections to
the energies of the vibrational states of the system. The result-
ing transition wavenumbers are compared with experimental
values. FIG. 1. The water dimer.
184309-3 Mackeprang et al. J. Chem. Phys. 140, 184309 (2014)

monly used in the literature: |if | jb |k for the donor unit and
|lm± |n for the acceptor unit.32, 44, 51 For the donor unit, the
three quantum numbers represent the free OH-oscillator, OHf ,
the bonded OH-oscillator, OHb , and the bending vibration
of the donor unit, respectively. We denote the displacement
of the OHb -oscillator as qb = rb − rb, 0 , the displacement of
the OHf -oscillator as qf = rf − rf, 0 , and the displacement of
donor bending vibration as θ = ϕ d − ϕ d, 0 , where rb , rf , and
ϕ d are the coordinate values of the OHb -bond, the OHf -bond,
and the donor HOH angle at a specific geometry, respectively,
and rb, 0 , rf, 0 , and ϕ d, 0 are the corresponding values at the
equilibrium geometry. We use a general notation to denote
the displacement of the six low-frequency modes as su , where
u = 1–6.
The Hamiltonian for the donor unit is
Ĥ3D,donor = T̂3D,donor + V̂ (qb , qf , θ ). (6)
We use the exact kinetic energy operator of the water
molecule as described in the literature with the exception of
the pseudo-potential, which has been excluded.30
Recently, the method of frozen mode correction, which
incorporates the presence of the rigid constraints into the ki-
FIG. 2. Visualization of the intermolecular modes. netic energy operator, has been developed.52, 53 This type of
treatment is beyond the scope of our study. Moreover, it can
be safely assumed that this effect is negligible for the vibra-
The donor twisting angle, x, is between the plane defined tional transitions of water dimer. The potential energy surface
by the donor water unit and the plane defined by the donor is separated into a one-, two-, and three-dimensional part.
Hf OHb angle bisecting vector and the Od −Oa vector: For the two stretches, the one-dimensional part of the po-
(ub × uf ) · (uR × uq ) tential energy is represented by a spline fit to the energies of
cos x = − , (2)
|ub × uf ||uR × uq | 23 points obtained by displacing the respective OH bond by
−0.40 to +0.70 Å in the steps of 0.05 Å from its optimized
where ub is the unit vector in the direction of the OHb - value (1 Å = 10−10 m). The fits were generated using the
oscillator of the donor unit, and uf is the unit vector in the spline fit function spline in Matlab.54 For the bending mode,
direction of the OHf -oscillator of the donor unit. a spline fit to the energies of 27 points is obtained by displac-
The acceptor wagging angle, α, is between the acceptor ing the water donor unit from −50◦ to +80◦ in the steps of
H1 OH2 angle bisecting vector, r, and the Oa −Od vector: 5◦ along the bending mode. The donor HOH bisecting vector
cos α = uR · ur , (3) is kept fixed relative to the acceptor unit when the donor unit
is displaced along the bending coordinate. The Schrödinger
where ur is the acceptor H1 OH2 angle bisecting unit vector. equation was solved numerically for these one-dimensional
The acceptor twisting angle, τ , is between the plane de- oscillators.
fined by the acceptor water unit and the plane defined by The two- and three-dimensional potential energy surfaces
the acceptor H1 OH2 angle bisecting vector and the Oa −Od are represented by second, third, and fourth order force con-
vector stants. Higher order corrections are expected to be small. The
(u1 × u2 ) · (uR × ur ) force constants are determined as derivatives of potential en-
cos τ = , (4)
|u1 × u2 ||uR × ur | ergy surfaces, which are found using Matlab.54 For the two-
where u1 and u2 are unit vectors of the symmetrical dimensional parts, the potential energy surfaces are repre-
OH-oscillators of the acceptor water unit. sented by two-dimensional grids obtained by simultaneously
The torsional angle, ξ , is between the plane defined by displacing two given modes and calculating the energies at the
the donor Hf OHb angle bisecting vector and the Od −Oa vec- different geometries. For the two stretches, the displacements
tor and the plane defined by the acceptor H1 OH2 angle bisect- are from −0.20 to +0.20 Å in the steps of 0.05 Å and, for
ing vector and the Oa −Od vector the bending mode, the displacements are from −20◦ to +20◦
in the steps of 5◦ . The two-dimensional force constants were
(uR × uq ) · (uR × ur ) then determined as follows. The derivatives of the potential
cos ξ = − . (5)
|uR × uq ||uR × ur | energy surface with respect to one coordinate were computed
by the coefficients of an 8th order polynomial fitted to the 9
points at each displacement of the second coordinate. Subse-
Vibrational model
quently, the force constants were determined by fitting these
To describe the vibrations of the donor unit and the ac- one-dimensional derivatives to 8th order polynomials, which
ceptor unit of water dimer, we use the notation, which is com- are functions of the second coordinate. The coefficients of
184309-4 Mackeprang et al. J. Chem. Phys. 140, 184309 (2014)

these polynomials provide the necessary force constants. For parts, one for each mode of the donor unit, we obtain
the three-dimensional part, the potential energy surfaces are
6  
represented by a three-dimensional grid obtained by displac- 1
Ĥ1 (qb ) = qb fqb su su + fqb su su su
2
ing the two stretches from −0.10 to +0.10 Å in the steps of u=1
2
0.05 Å and the bend from −10◦ to +10◦ in the steps of 5◦ .
6  
A similar approach to that used for the two-dimensional grid 1  1
+ qb2 fqb qb su su + fqb qb su su su2
was applied to find the three-dimensional derivatives. The un- 2 u=1 2
perturbed states are calculated variationally. We perform basis
set contraction using one-dimensional potential energy sur- 6  
1
faces to obtain the basis functions for the two OH-stretches Ĥ1 (qf ) = qf fqf su su + fqf su su su2
2
and the bend using the method described by Meyer.55 To ob- u=1
(9)
tain the basis functions, we use the g-matrix elements eval- 6 
 
uated at the equilibrium geometry. Thus, the basis functions 1 1
+ qf2 fqf qf su su + fqf qf su su su2 ,
we use in the variational calculation are the products of eigen- 2 u=1
2
functions of the one-dimensional Schrödinger equations for
6 
 
the two stretches and the bend of the donor water unit. With 1
Ĥ1 (θ ) = θ fθsu su + fθsu su su2
these basis functions, we setup a Hamiltonian matrix includ- 2
u=1
ing vibrational states with up to 7 quanta in the stretches and
6  
5 quanta in the bend. The Hamiltonian is diagonalized using 1 2 1
the matrix diagonalization function Eig in Matlab.54 + θ fθθsu su + fθθsu su su
2
2 u=1 2
We treat the six intermolecular modes as one-
dimensional oscillators. All oscillators except the one for the where the force constants are the derivatives of the potential
OO-stretch are modelled as harmonic oscillators. The har- energy surface at the equilibrium geometry. Some higher or-
monic frequency is determined from a one-dimensional po- der terms were tested and found to have insignificant effects.
tential energy surface obtained by displacing the respective The kinetic energy coupling can be neglected, because we
intermolecular mode from −20◦ to +20◦ in the steps of 5 de- evaluate the g-matrix elements of the intermolecular modes in
grees from the equilibrium structure. The OO-stretching mo- the equilibrium structure and because the expectation values
tion is treated as a Morse oscillator. The Morse parameters ω̃ of the momentum operator for all bound states of a harmonic
and ω̃x are determined from second, third, and fourth order and an anharmonic oscillator disappear. Details of the pertur-
derivatives of the potential energy surface as described in the bation theory model and the associated equations are in the
literature.56 These derivatives are determined from a potential supplementary material.58
energy surface obtained by displacing the vibrational coordi- The force constants used in the perturbation calculations
nate from −0.20 to +0.20 Å by 0.05 Å. are determined as the derivatives of two-dimensional poten-
The g-matrix elements associated with the kinetic energy tial energy surfaces similar to the approach, where we deter-
of the intermolecular vibrations are calculated as mined the force constants involving only the modes of the
donor unit. The two-dimensional potential energy surfaces are

N
1
g (qi ,qj )
= (∇α qi ) · (∇α qj ), (7) represented by a 9 × 9 grids. For all stretching vibrations in
α
m α these grids, the displacements are from −0.20 to +0.20 Å in
the steps of 0.05 Å. For all other vibrations, the displacements
where the sum is over all nuclei and mα is the mass of the are from −20◦ to +20◦ in the steps of 5◦ .
αth nucleus.57 However, the g-matrix element of the torsional
motion is approximated as a sum of the inverse of the mo-
ments of inertia of two linear rotors. The matrix elements are RESULTS
evaluated at the equilibrium geometry and the gradients of the The calculated wavenumbers obtained using our unper-
coordinates are in the supplementary material.58 Our Hamil- turbed 3D model (Eq. (6)) for the donor and acceptor units
tonian, Ĥ0 , describing the donor unit and the intermolecular in water dimer with a potential energy surface calculated at
modes of the unperturbed system then becomes the CCSD(T)-F12a/VTZ-F12 level of theory are summarized
6   in Table I. Our results with the unperturbed model are sim-
 1
Ĥ0 = Ĥ3D,donor + gs(0) p 2
+ V (s ) , (8) ilar to those previously calculated.30 The wavenumbers of
u ,su su u
u=1
2 the donor unit transitions obtained with our perturbation the-
ory model are also given and these transitions are compared
where gs(0)
u ,su
is the Wilson g-matrix elements evaluated at the with observations available from neon matrix, helium droplet,
equilibrium geometry. For the vibrations of the acceptor unit, and gas phase experiments. Overall the agreement is good up
we use a similar three-dimensional Hamiltonian as employed to about 10 000 cm−1 . The most interesting transition is the
for the donor unit. OHb -stretching vibration. The wavenumber of the fundamen-
We perturb the vibrational energy levels, including the tal OHb -stretching vibration has been observed at 3601 cm−1
ground state, by allowing the low-frequency modes to cou- in the gas-phase, and at 3597 and 3591 cm−1 in helium-
ple through the potential energy with the OHb -oscillator, droplet and neon-matrix experiments, respectively.14, 23, 25
OHf -oscillator or HOH-bending mode. Separating these three With our unperturbed model, the corresponding calculated
184309-5 Mackeprang et al. J. Chem. Phys. 140, 184309 (2014)

TABLE I. Vibrational transition wavenumbers (in cm−1 ) of the donor and acceptor unit of water dimer.

Assignment ν̃ (0),a ν̃corr b ν̃tot c ν̃exp,matrix d,23 ν̃exp,gas ν̃exp,H e e,25

|00|1 1599 1599 160119


|0f |0b |1 1620 − 2.0 1618 1617 161519
|00|2 3163 3163
|0f |0b |2 3204 − 9.6 3195 3194
|0f |1b |0 3557 49.9 3607 3591 360114 3597
|10+ |0 3652 3661 366014 3654
|1f |0b |0 3726 0.2 3727 3734 373514 3730
|10− |0 3747 3757 374520 3749
|0f |1b |1 5160 48.6 5209 5191 521922
|10+ |1 5232 5243
|10− |1 5325 5333 532922
|1f |0b |1 5326 0.0 5326 5336 534522
|0f |1b |2 6722 44.2 6766 6750
|10+ |2 6776 6790
|10− |2 6870 6881
|1f |0b |2 6891 − 3.9 6887 6895
|0f |2b |0 6943 103.8 7047 7018
79%|20+ |0 − 17%|11|0 7193 7207 719321
57%|2f |0b |0 + 37%|1f |1b |0 7215 20.1 7235 7237 724021
|20− |0 7238 7253 725021
57%|1f |1b |0 − 40%|2f |0b |0 7332 24.8 7357 7362
80%|11|0 − 18%|20+ |0 7429 7442
|0f |2b |1 8530 103.2 8633 8603
75%|20+ |1 − 15%|11|1 8754 8771
|20− |1 8797 8810
53%|2f |0b |1 + 34%|1f |1b |1 8800 18.9 8819 8821
55%|1f |1b |1 − 37%|2f |0b |1 8912 27.7 8939 8942
76%|11|1 − 16%|20+ |1 8988 9000
73%|0f |3b |0 + 19%|0f |2b |2 10 154 148.9 10 302 10 267
|30+ |0 10 588 10 611
42%|3f |0b |0 − 32%|1f |2b |0 10 599 46.3 10 646
+18%|2f |1b |0
|30− |0 10 601 10 620
48%|1f |2b |0 + 44%|3f |0b |0 10 657 51.9 10 708 10 798
|21+ |0 10 855 10 876
76%|2f |1b |0 + 12%|1f |2b |0 10 863 43.1 10 906 10 912
|21− |0 11 011 11 033
|0f |4b |0 13 111 222.6 13 334
|40+ |0 13 815
|40− |0 13 818
63%|1f |3b |0 + 10%|1f |2b |2 13 827 129.3 13 957
|4f |0b |0 13 856 10.6 13 866
40%|2f |2b |0 − 39%|3f |1b |0 14 125 73.4 14 198
63%|31+ |0 + 25%|22|0 14 210
46%|3f |1b |0 + 45%|2f |2b |0 14 288 61.9 14 350
|31− |0 14 301
67%|22|0 + 26%|31+ |0 14 512
|0f |5b |0 15 905 344.4 16 250
47%|5f |0b |0 − 26%|4f |0b |2 16 874 5.5 16 880
|50+ |0 16 887
|50− |0 16 887
a
Calculated transition wavenumbers with the unperturbed model at the CCSD(T)-F12/VTZ-F12 level of theory.
b
Correction to the transition wavenumbers calculated using the perturbative treatment described in the text for the donor unit. Potential energy surfaces calculated at the CCSD(T)-
F12/VTZ-F12 level of theory.
c
Transition wavenumbers for the donor unit of water dimer calculated as the unperturbed value plus the perturbative correction.
d
From a neon-matrix experiment.23
e
From a helium droplet experiment.25
184309-6 Mackeprang et al. J. Chem. Phys. 140, 184309 (2014)

TABLE II. The perturbative energy correction (in cm−1 ) calculated with
different potential energy surfaces.a

Assignment F 12
ν̃corr MP 2
ν̃corr MP 2 -ν̃ F 12
ν̃corr corr

|0f |0b |1 − 2.0 − 2.3 − 0.3


|0f |0b |2 − 9.6 − 10.5 − 1.1
|0f |1b |0 49.9 53.6 3.7
|1f |0b |0 0.2 0.2 0.0
|0f |1b |1 48.6 52.2 3.5
|1f |0b |1 0.0 − 0.1 − 0.1
|0f |1b |2 44.2 47.5 3.3
|1f |0b |2 − 3.9 − 4.6 − 0.7
|0f |2b |0 103.8 110.2 6.4
57%|2f |0b |0 + 37%|1f |1b |0 20.1 21.6 1.5
57%|1f |1b |0 − 40%|2f |0b |0 24.8 26.8 2.0
|0f |2b |1 103.2 109.8 6.6
53%|2f |0b |1 + 34%|1f |1b |1 18.9 20.2 1.3
FIG. 3. Visualization of the energy convergence with respect to the pertur- 55%|1f |1b |1 − 37%|2f |0b |1 27.7 29.7 2.0
bation order for the states |0f |vb |0, where v = 0, 1, and 2. 73%|0f |3b |0 + 19%|0f |2b |2 148.9 161.8 12.9
42%|3f |0b |0 − 32%|1f |2b |0 46.3 49.4 3.1
+18%|2f |1b |0
48%|1f |2b |0 + 44%|3f |0b |0 51.9 55.3 3.4
wavenumber is significantly lower at 3557 cm−1 . However,
76%|2f |1b |0 + 12%|1f |2b |0 43.1 46.2 3.1
by applying the perturbative corrections of our model, we ob- |0f |4b |0 222.6 274.3 51.7
tain a value of 3607 cm−1 , which is in good agreement with 63%|1f |3b |0 + 10%|1f |2b |2 129.3 138.6 9.3
the experimental results. |4f |0b |0 10.6 11.2 0.6
For the first overtone, the perturbation shifts our calcu- 40%|2f |2b |0 − 39%|3f |1b |0 73.4 77.7 4.3
lated transition wavenumber to 7047 cm−1 . This transition 46%|3f |1b |0 + 45%|2f |2b |0 61.9 66.2 4.3
has not been observed in gas-phase experiments, but has been |0f |5b |0 344.4 565.2 220.8
identified in a neon-matrix experiment at 7018 cm−1 . How- 47%|5f |0b |0 −26 %|4f |0b |2 5.5 5.1 − 0.4
ever, the matrix is known to redshift transitions. In the fun- a
For the unperturbed model CCSD(T)-F12a/VTZ-F12 potential energy surfaces were
damental region, the transition wavenumber is redshifted by used. The two-dimensional surfaces between the intermolecular vibrational modes and
10 cm−1 in the neon matrix relative to the reported gas phase the intermolecular modes were calculated at both the CCSD(T)-F12a/VTZ-F12 (first
column) and the MP2/AVTZ (second column) level of theory and results using either
value, and, for the first overtone, we expect a redshift of type of surfaces are shown.
∼20 cm−1 , which brings the reported neon-matrix experimen-
tal value in good agreement with our calculated result.
The contributions of the different orders in the pertur-
bation are shown in the supplementary material (Tables S2– calculation.30, 47 For example, the fundamental transition of
S4).58 We observe that the corrections become smaller as the OHb -stretching mode in water dimer is calculated at
the order of the perturbation increases. It is clear that states 3522 cm−1 with the MP2/AVTZ level of theory, i.e., 35 cm−1
involving OHb -stretching are mostly affected by the pertur- lower than the result obtained with the CCSD(T)-F12a/VTZ-
bation. In Figure 3, we illustrate the clear convergence for F12 level of theory. Considering possible future applications
the |0f |vb |0 states with v = 0, 1, and 2. For the majority to larger bimolecular systems, we investigated if it was possi-
of the vibrational states and all states with energy less than ble to use potential energy surfaces calculated at a high level
10 000 cm−1 , the fourth order perturbation theory correction of electronic structure calculation for the 3D model of the
is less than 1 cm−1 . In higher overtones of the bonded OHb - donor unit and at a lower level of electronic structure cal-
oscillator (more than 3 quantas), the convergence is slower culation for the potential energy surface needed for the per-
with respect to the order of the perturbation. The fourth or- turbation. We show the calculated transition wavenumbers of
der correction to the energy of the vibrationally excited state the donor unit of water dimer, where we used an MP2/AVTZ
|0f |4b |0 and |0f |5b |0 is 6.7 cm−1 and 37.2 cm−1 , respec- potential energy surface to obtain the perturbative terms, but
tively. Clearly, our model becomes unsuitable for these higher used the CCSD(T)-F12a/VTZ-F12 potential energy surface
overtone transitions. for the unperturbed part in Table II. Again we find that states
To obtain the perturbative corrections to the OH- involving OHb -stretching are the most affected ones. For the
stretching and HOH-bending vibrational energy levels from fundamental, the difference is 3.7 cm−1 and, for the first over-
all six large amplitude modes, we need more than 1000 sin- tone, the difference is 6.4 cm−1 . For states involving higher
gle point electronic structure calculations. At a high level, overtones of the OHb -oscillator, we observe larger deviations,
like accurate coupled cluster methods, this is expensive and but as we mentioned, the energies of these states are not con-
time consuming, especially if we want our model to be ap- verged as indicated by large fourth order perturbation contri-
plicable to systems with larger monomers. The energy lev- butions. It appears that we can calculate vibrational transitions
els in the unperturbed local mode model depend a lot on up to the first overtone of the OHb -stretch using our fourth or-
the ab initio method employed in the electronic structure der perturbative treatment of the intermolecular modes and
184309-7 Mackeprang et al. J. Chem. Phys. 140, 184309 (2014)

the MP2/AVTZ method for potential energy surfaces of the


perturbation part.
One possibility of reducing the number of calculations
required for the perturbation is to include only some of the
intermolecular modes.28, 48 To illustrate which of the inter-
molecular modes have the largest effect on the vibrational en-
ergy levels, we compare the contributions from each mode
in turn to the energy of the vibrational ground state. In
Figure 4, we show the contribution to the energy from per-
turbation by each of the six intermolecular modes in turn to
either the OHb -stretch, OHf -stretch, or the HOH-bend vibra-
tional ground state. It is clear from Figure 4 that the six modes
are not equally important perturbers and that the perturbations
to the OHb -oscillator are most important. This was as ex-
pected from the physical picture of the vibrational modes, and
because the transition wavenumbers of the OHb -stretching
states have been found to deviate most from experimental
FIG. 4. Perturbation theory contributions to the ground vibrational state en-
values. All perturbations to the free OH-oscillator can be ig- ergy. Each high-frequency mode has been coupled to each large amplitude
nored, as the contribution is less than 1% of the total cor- motion in turn. Couplings of the donor OHb -stretch, OHf -stretch, and HOH-
rection to the energy of the |0f |0b |0 vibrational ground bend with the intermolecular modes are represented by blue, green, and red
state. The clearly non-zero contributions from the couplings bars, respectively. The definitions of the intermolecular coordinates are in the
text and are visualized in Figure 2.
to the donor bend were unexpected as earlier reported local
mode values of the bending transition wavenumbers are in The effect of the dominant intermolecular modes can be
good agreement with experiment.30 However, the contribu- rationalized from the physical picture of these vibrations. The
tions from the different low-frequency modes are small and two largest contributions arise from the coupling of the OHb -
partly cancel each other. oscillator with the donor twisting vibration (x in Figure 2)

TABLE III. Transition wavenumbers (in cm−1 ) for the fully deuterated water dimer complex.

Assignment ν̃ (0),a ν̃corr b ν̃tot c ν̃exp,matrix 60 ν̃exp,gas 59

|00|1 1181 1182


|0f |0b |1 1196 − 1.3 1195 1192
|00|2 2343
|0f |0b |2 2372 − 4.7 2368
82% |0f |1b |0 + 16% |1f |0b |0 2612 24.2 2636 2626 2632
|10+ |0 2668 2673
83% |1f |0b |0 + 16% |0f |1b |0 2759 2.2 2761 2762 2765
|10− |0 2782 2790 2783
80% |0f |1b |1 + 14% |1f |0b |1 3801 22.8 3824
|10+ |1 3840
80% |0f |1b |1 + 14% |1f |0b |1 3944 2.1 3946
|10− |1 3952
75% |0f |1b |2 + 12% |1f |0b |2 4967 20.0 4987
|10+ |2 4993
|10− |2 5104
79% |0f |1b |2 + 13% |1f |0b |2 5110 0.6 5110
81%|0f |2b |0 + 12%|1f |1b |0 5146 51.6 5198
66%|20+ |0 + 31% |11|0 5285
49%|2f |0b |0 + 37%|1f |1b |0 5320 15.4 5335
− 10%|0f |2b |0
|20− |0 5365
49%|1f |1b |0 − 46%|2f |0b |0 5456 10.3 5466
66% |11|0 + 32%|20+ |0 5518
|0f |2b |1 6329 50.1 6379
62%|20+ |1 + 29% |11|1 6449
46%|2f |0b |1 + 36%|1f |1b |1 6500 14.5 6514
|20− |1 6527
a
Calculated transition wavenumbers with the unperturbed model at the CCSD(T)-F12/VTZ-F12 level of theory.
b
Correction to the unperturbed transition wavenumbers calculated using the perturbative treatment described in the text for the
donor unit. Calculated at the CCSD(T)-F12/VTZ-F12 level of theory.
c
Transition wavenumbers for the donor unit of water dimer calculated as the unperturbed value plus the perturbative correction.
184309-8 Mackeprang et al. J. Chem. Phys. 140, 184309 (2014)

and the donor rocking vibration (β in Figure 2). The nature donor acceptor interaction, also give rise to the largest con-
of the two intermolecular vibrations is such that the hydrogen tributions to the OHb -stretching wavenumber. The net result
bonded H-atom is displaced significantly relative to the in the physical picture is such that the bonded hydrogen atom
acceptor unit along these two vibrations resulting in a weaker movement is really three-dimensional.
hydrogen bond and a large change in the OHb -stretching Finally, we have also made a pleasing observation that
transition wavenumber. This results in a larger transition high level electronic structure calculations such as CCSD(T)-
wavenumber, which we see from our perturbation theory F12, which are required to obtain accurate vibrational fre-
model. The third largest contribution arises from coupling of quencies, are not required in determining potential energy
the OHb -oscillator to the acceptor twist (τ in Figure 2), but coupling terms between high- and low-frequency vibrational
the picture is not as intuitive as the two previous ones. The modes. The MP2 method with adequate basis sets is well
fourth largest contribution arises from the coupling between suited. This, in combination with the fact that the main contri-
the donor bending motion and the donor rocking motion (β bution arises from only two intermolecular modes, will make
in Figure 2). We found it counter intuitive that the OO-stretch it possible to extend our approach to larger, more demanding
has such a limited effect on the OHb -stretch. All remaining bimolecular complex systems.
couplings are at most a few wavenumbers. Previously, Garden
et al.28, 48 investigated the effect on the OHb -stretch in water
ACKNOWLEDGMENTS
dimer from either the OO-stretch (R in Figure 2) or the accep-
tor wag (α in Figure 2). The OO-stretch was found to shift the We are grateful to Elina Sjöholm for helpful discussions.
OHb -stretching wavenumber by 1.7 cm−1 and the acceptor We thank the Danish Council for Independent Research—
wag to shift by 2.6 cm−1 , in good agreement with our results. Natural Sciences, and the Danish Center for Scientific Com-
To test our model further, we computed the transition puting (DCSC) and the Academy of Finland for funding.
wavenumbers of the fully deuterated water dimer and com- 1 P. G. Sennikov, S. K. Ignatov, and O. Schrems, ChemPhysChem 6, 392
pared with available experimental values. The results are
(2005).
shown in Table III. The agreement between our calculated 2 V. Vaida, J. Chem. Phys. 135, 020901 (2011).

wavenumbers and the experimentally measured values is 3 V. Vaida, J. Phys. Chem. A 113, 5 (2009).
4 S. Aloisio and J. S. Francisco, Acc. Chem. Res. 33, 825 (2000).
again good. We find that our vibrational model can predict the 5 H. G. Kjaergaard, T. W. Robinson, D. L. Howard, J. S. Daniel, J. E.
fundamental transition wavenumbers of the deuterated wa- Headrick, and V. Vaida, J. Phys. Chem. A 107, 10680 (2003).
ter dimer to within a few wavenumbers. For the fundamental 6 J. S. Daniel, S. Solomon, R. W. Sanders, R. W. Portmann, D. C. Miller, and

ODb -stretch, the unperturbed local mode calculation predicts W. Madsen, J. Geophys. Res. D 104, 16785, doi:10.1029/1999JD900220
the wavenumber to be 2612 cm−1 , and the correction from (1999).
7 P. Chýlek and D. J. W. Geldart, Geophys. Res. Lett. 24, 2015,
the perturbation is 24 cm−1 , which gives a total wavenum- doi:10.1029/97GL01949 (1997).
ber of 2636 cm−1 , in good agreement with the experimen- 8 J. S. Daniel, S. Solomon, H. G. Kjaergaard, and D. P. Schofield, Geophys.

tal gas-phase value of 2632 cm−1 .59 For the fundamental Res. Lett. 31, L06118, doi:10.1029/2003GL018914 (2004).
9 V. Vaida, H. G. Kjaergaard, and K. J. Feierabend, Int. Rev. Phys. Chem. 22,
ODf -stretch and DOD-bend, the calculated values after per-
203 (2003).
turbation are 2761 cm−1 and 1195 cm−1 , respectively. The 10 V. Vaida, J. S. Daniel, H. G. Kjaergaard, L. M. Goss, and A. F. Tuck, Q. J.
corresponding experimental gas-phase value is 2765 cm−1 R. Meteorol. Soc. 127, 1627 (2001).
11 D. P. Schofield, J. R. Lane, and H. G. Kjaergaard, J. Phys. Chem. A 111,
for the ODf -stretch and the neon-matrix observation for the
DOD-bend is 1192 cm−1 .59, 60 567 (2007).
12 S. Chung and M. Hippler, J. Chem. Phys. 124, 214316 (2006).
13 M. Hippler, J. Chem. Phys. 127, 084306 (2007).
14 U. Buck and F. Huisken, Chem. Rev. 100, 3863 (2000).
CONCLUSION 15 D. L. Howard and H. G. Kjaergaard, J. Phys. Chem. A 110, 9597 (2006).
16 L. Du and H. G. Kjaergaard, J. Phys. Chem. A 115, 12097 (2011).
We have investigated the effects of the large amplitude 17 L. Du, J. R. Lane, and H. G. Kjaergaard, J. Chem. Phys. 136, 184305
intermolecular motions on the high-frequency intermolecular (2012).
modes of water dimer donor unit. We have shown using per- 18 L. Du, K. Mackeprang, and H. G. Kjaergaard, Phys. Chem. Chem. Phys.

turbation theory that we are able to improve drastically the 15, 10194 (2013).
19 J. B. Paul, R. A. Provencal, C. Chapo, K. Roth, R. Casaes, and R. J.
results of local mode models where the vibrational motions Saykally, J. Phys. Chem. A 103, 2972 (1999).
in the donor and acceptor units are modelled separately and 20 Z. S. Huang and R. E. Miller, J. Chem. Phys. 91, 6613 (1989).
21 S. A. Nizkorodov, M. Ziemkiewicz, D. J. Nesbitt, and A. E. W. Knight, J.
without taking into account the intermolecular modes. We are
also able to provide a physical explanation why the previous Chem. Phys. 122, 194316 (2005).
22 I. V. Ptashnik, K. M. Smith, K. P. Shine, and D. A. Newnham, Q. J. R.
local mode models have failed in explaining the large redshift Meteorol. Soc. 130, 2391 (2004).
of the OHb -stretching vibrations. 23 Y. Bouteiller, B. Tremblay, and J. Perchard, Chem. Phys. 386, 29 (2011).
24 Y. Bouteiller and J. Perchard, Chem. Phys. 305, 1 (2004).
With our new approach, we reproduce the hydrogen
25 K. Kuyanov-Prozument, M. Y. Choi, and A. F. Vilesov, J. Chem. Phys. 132,
bonded OHb -oscillator fundamental transition wavenumber
014304 (2010).
to within 6 cm−1 both in the normal and fully deuterated 26 G. R. Low and H. G. Kjaergaard, J. Chem. Phys. 110, 9104 (1999).
isotopomers of water dimer. In comparison, the local mode 27 R. Kalescky, W. Zou, E. Kraka, and D. Cremer, Chem. Phys. Lett. 554, 243

models underestimated these transitions as much as 50 and (2012).


28 A. L. Garden, L. Halonen, and H. G. Kjaergaard, J. Phys. Chem. A 112,
20 cm−1 , respectively. The key point in our new results is the
7439 (2008).
observation that the low-frequency modes, donor rock, and 29 D. P. Schofield and H. G. Kjaergaard, Phys. Chem. Chem. Phys. 5, 3100
donor twist, which have the largest weakening effect on the (2003).
184309-9 Mackeprang et al. J. Chem. Phys. 140, 184309 (2014)

30 T. Salmi, V. Hänninen, A. L. Garden, H. G. Kjaergaard, J. Tennyson, and 48 A. L. Garden, L. Halonen, and H. G. Kjaergaard, Chem. Phys. Lett. 513,
L. Halonen, J. Phys. Chem. A 112, 6305 (2008). 167 (2011).
31 V. Hänninen, T. Salmi, and L. Halonen, J. Phys. Chem. A 113, 7133 (2009). 49 H.-J. Werner, P. J. Knowles, G. Knizia, F. R. Manby, M. Schütz et al.,
32 H. G. Kjaergaard, A. L. Garden, G. M. Chaban, R. B. Gerber, D. A. MOLPRO , version 2010.1, a package of ab initio programs, 2010, see
Matthews, and J. F. Stanton, J. Phys. Chem. A 112, 4324 (2008). http://www.molpro.net.
33 Y. Wang and J. M. Bowman, J. Chem. Phys. 136, 144113 (2012). 50 K. A. Peterson, T. B. Adler, and H.-J. Werner, J. Chem. Phys. 128, 084102
34 Y. Wang, S. Carter, B. J. Braams, and J. M. Bowman, J. Chem. Phys. 128, (2008).
071101 (2008). 51 H. G. Kjaergaard, B. R. Henry, H. Wei, S. Lefebvre, T. Carrington,
35 A. Shank, Y. Wang, A. Kaledin, B. J. Braams, and J. M. Bowman, J. Chem. O. Sonnich Mortensen, and M. L. Sage, J. Chem. Phys. 100, 6228 (1994).
Phys. 130, 144314 (2009). 52 J. Pesonen, J. Chem. Phys. 139, 144310 (2013).
36 V. Babin, C. Leforestier, and F. Paesani, J. Chem. Theory Comput. 9, 5395 53 L. Partanen, J. Pesonen, E. Sjöholm, and L. Halonen, J. Chem. Phys. 139,

(2013). 144311 (2013).


37 C. Leforestier, K. Szalewicz, and A. van der Avoird, J. Chem. Phys. 137, 54 MATLAB and Statistics Toolbox Release 2013a, The MathWorks, Inc.,

014305 (2012). Natick, Massachusetts, USA.


38 C. Leforestier, Philos. Trans. R. Soc., A 370, 2675 (2012). 55 R. Meyer, J. Chem. Phys. 52, 2053 (1970).
39 Y. Watanabe, S. Maeda, and K. Ohno, J. Chem. Phys. 129, 074315 (2008). 56 D. L. Howard, P. Jorgensen, and H. G. Kjaergaard, J. Am. Chem. Soc. 127,
40 M. Sage and J. Jortner, Adv. Chem. Phys. 47, 293 (1981). 17096 (2005).
41 O. S. Mortensen, B. R. Henry, and M. A. Mohammadi, J. Chem. Phys. 75, 57 J. Pesonen, J. Chem. Phys. 112, 3121 (2000).

4800 (1981). 58 See supplementary material at http://dx.doi.org/10.1063/1.4873420 for de-


42 M. S. Child and R. T. Lawton, Faraday Discuss. Chem. Soc. 71, 273 (1981). tails of our perturbation model, details of the convergence behaviour of the
43 B. R. Henry and H. G. Kjaergaard, Can. J. Chem. 80, 1635 (2002). model, and the force constants used in the perturbation part.
44 L. Halonen and T. Carrington, J. Chem. Phys. 88, 4171 (1988). 59 J. B. Paul, R. A. Provencal, C. Chapo, A. Petterson, and R. J. Saykally, J.
45 M. S. Child and L. Halonen, Adv. Chem. Phys. 57, 1 (1984). Chem. Phys. 109, 10201 (1998).
46 L. Halonen, Adv. Chem. Phys. 104, 41 (1998). 60 J. Ceponkus, P. Uvdal, and B. Nelander, J. Phys. Chem. A 112, 3921
47 J. R. Lane and H. G. Kjaergaard, J. Chem. Phys. 132, 174304 (2010). (2008).

You might also like