Ab initio molecular dynamics study of collective excitations in liquid H₂O and D₂O: Effect of dispersion corrections
The collective dynamics in liquid water is an active research topic experimentally, theoretically and via simulations. Here, ab initio molecular dynamics simulations are reported in heavy and ordinary water at temperature 323.15 K, or 50oC. The simulations in heavy water were performed both with and...
Gespeichert in:
| Datum: | 2016 |
|---|---|
| Hauptverfasser: | , |
| Format: | Artikel |
| Sprache: | English |
| Veröffentlicht: |
Інститут фізики конденсованих систем НАН України
2016
|
| Schriftenreihe: | Condensed Matter Physics |
| Online Zugang: | http://dspace.nbuv.gov.ua/handle/123456789/155810 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| Назва журналу: | Digital Library of Periodicals of National Academy of Sciences of Ukraine |
| Zitieren: | Ab initio molecular dynamics study of collective excitations in liquid H₂O and D₂O: Effect of dispersion corrections / T. Bryk, A.P. Seitsonen // Condensed Matter Physics. — 2016. — Т. 19, № 2. — С. 23604: 1–14. — Бібліогр.: 57 назв. — англ. |
Institution
Digital Library of Periodicals of National Academy of Sciences of Ukraine| id |
irk-123456789-155810 |
|---|---|
| record_format |
dspace |
| spelling |
irk-123456789-1558102019-06-18T01:30:13Z Ab initio molecular dynamics study of collective excitations in liquid H₂O and D₂O: Effect of dispersion corrections Bryk, T. Seitsonen, A.P. The collective dynamics in liquid water is an active research topic experimentally, theoretically and via simulations. Here, ab initio molecular dynamics simulations are reported in heavy and ordinary water at temperature 323.15 K, or 50oC. The simulations in heavy water were performed both with and without dispersion corrections. We found that the dispersion correction (DFT-D3) changes the relaxation of density-density time correlation functions from a slow, typical of a supercooled state, to exponential decay behaviour of regular liquids. This implies an essential reduction of the melting point of ice in simulations with DFT-D3. Analysis of longitudinal (L) and transverse (T) current spectral functions allowed us to estimate the dispersions of acoustic and optic collective excitations and to observe the L-T mixing effect. The dispersion correction shifts the L and T optic (O) modes to lower frequencies and provides by almost thirty per cent smaller gap between the longest-wavelength LO and TO excitations, which can be a consequence of a larger effective high-frequency dielectric permittivity in simulations with dispersion corrections. Simulation in ordinary water with the dispersion correction results in frequencies of optic excitations higher than in D₂O, and in a long-wavelength LO-TO gap of 24 ps⁻¹ (127 cm⁻¹). Колективна динамiка у рiдкiй водi активно дослiджується експериментальними, теоретичними методами та комп’ютерними симуляцiями. Представлено моделювання методом першопринципної молекулярної динамiки для важкої та звичайної води при температурi 323.15 K, чи 50◦C. Моделювання для важкої води були проведенi з та без дисперсiйних поправок. Ми отримали, що дисперсiйнi поправки (DFT-D3) суттєво змiнюють релаксацiю часових кореляцiйних функцiй густина-густина з повiльної, яка є типовою для переохолодженого стану, до експоненцiйного спаду як для звичайних рiдин. Це означає суттєве зменшення точки плавлення льоду для моделювання з DFT-D3. Аналiз повздовжних (L) та поперечних (T) спектральних функцiй потокiв дозволив визначити дисперсiї акустичних та оптичних збуджень та спостерiгати ефект L-T змiшування. Дисперсiйнi поправки D3 зсувають до нижчих частот L та T оптичнi (O) моди та приводять до меншої на майже тридцять процентiв щiлини мiж LO та TO довгохвильовими збудженнями, що може бути наслiдком бiльшої високочастотної дiелектричної проникностi в моделюваннi з дисперсiйними поправками. Моделювання для звичайної води з дисперсiйними поправками дає вищi частоти оптичних збуджень, нiж для D₂O, та щiлину LO-TO порядку 24 пс⁻¹ (127 см⁻¹ ) у довгохвильовiй границi. 2016 Article Ab initio molecular dynamics study of collective excitations in liquid H₂O and D₂O: Effect of dispersion corrections / T. Bryk, A.P. Seitsonen // Condensed Matter Physics. — 2016. — Т. 19, № 2. — С. 23604: 1–14. — Бібліогр.: 57 назв. — англ. 1607-324X PACS: 61.20.Ja, 61.20.Lc, 62.60.+v, 63.50.-x, 78.30.C- DOI:10.5488/CMP.19.23604 arXiv:1603.07144 http://dspace.nbuv.gov.ua/handle/123456789/155810 en Condensed Matter Physics Інститут фізики конденсованих систем НАН України |
| institution |
Digital Library of Periodicals of National Academy of Sciences of Ukraine |
| collection |
DSpace DC |
| language |
English |
| description |
The collective dynamics in liquid water is an active research topic experimentally, theoretically and via simulations. Here, ab initio molecular dynamics simulations are reported in heavy and ordinary water at temperature 323.15 K, or 50oC. The simulations in heavy water were performed both with and without dispersion corrections. We found that the dispersion correction (DFT-D3) changes the relaxation of density-density time correlation functions from a slow, typical of a supercooled state, to exponential decay behaviour of regular liquids. This implies an essential reduction of the melting point of ice in simulations with DFT-D3. Analysis of longitudinal (L) and transverse (T) current spectral functions allowed us to estimate the dispersions of acoustic and optic collective excitations and to observe the L-T mixing effect. The dispersion correction shifts the L and T optic (O) modes to lower frequencies and provides by almost thirty per cent smaller gap between the longest-wavelength LO and TO excitations, which can be a consequence of a larger effective high-frequency dielectric permittivity in simulations with dispersion corrections. Simulation in ordinary water with the dispersion correction results in frequencies of optic excitations higher than in D₂O, and in a long-wavelength LO-TO gap of 24 ps⁻¹ (127 cm⁻¹). |
| format |
Article |
| author |
Bryk, T. Seitsonen, A.P. |
| spellingShingle |
Bryk, T. Seitsonen, A.P. Ab initio molecular dynamics study of collective excitations in liquid H₂O and D₂O: Effect of dispersion corrections Condensed Matter Physics |
| author_facet |
Bryk, T. Seitsonen, A.P. |
| author_sort |
Bryk, T. |
| title |
Ab initio molecular dynamics study of collective excitations in liquid H₂O and D₂O: Effect of dispersion corrections |
| title_short |
Ab initio molecular dynamics study of collective excitations in liquid H₂O and D₂O: Effect of dispersion corrections |
| title_full |
Ab initio molecular dynamics study of collective excitations in liquid H₂O and D₂O: Effect of dispersion corrections |
| title_fullStr |
Ab initio molecular dynamics study of collective excitations in liquid H₂O and D₂O: Effect of dispersion corrections |
| title_full_unstemmed |
Ab initio molecular dynamics study of collective excitations in liquid H₂O and D₂O: Effect of dispersion corrections |
| title_sort |
ab initio molecular dynamics study of collective excitations in liquid h₂o and d₂o: effect of dispersion corrections |
| publisher |
Інститут фізики конденсованих систем НАН України |
| publishDate |
2016 |
| url |
http://dspace.nbuv.gov.ua/handle/123456789/155810 |
| citation_txt |
Ab initio molecular dynamics study of collective excitations in liquid H₂O and D₂O: Effect of dispersion corrections / T. Bryk, A.P. Seitsonen // Condensed Matter Physics. — 2016. — Т. 19, № 2. — С. 23604: 1–14. — Бібліогр.: 57 назв. — англ. |
| series |
Condensed Matter Physics |
| work_keys_str_mv |
AT brykt abinitiomoleculardynamicsstudyofcollectiveexcitationsinliquidh2oandd2oeffectofdispersioncorrections AT seitsonenap abinitiomoleculardynamicsstudyofcollectiveexcitationsinliquidh2oandd2oeffectofdispersioncorrections |
| first_indexed |
2025-07-14T08:02:37Z |
| last_indexed |
2025-07-14T08:02:37Z |
| _version_ |
1837608634161299456 |
| fulltext |
Condensed Matter Physics, 2016, Vol. 19, No 2, 23604: 1–14
DOI: 10.5488/CMP.19.23604
http://www.icmp.lviv.ua/journal
Ab initio molecular dynamics study of collective
excitations in liquid H2O and D2O:
Effect of dispersion corrections
T. Bryk1,2, A.P. Seitsonen3,4
1 Institute for Condensed Matter Physics of the National Academy of Sciences of Ukraine,
1 Svientsitskii St., 79011 Lviv, Ukraine
2 Institute of Applied Mathematics and Fundamental Sciences, Lviv Polytechnic National University, 79013 Lviv,
Ukraine
3 Institut für Chemie, Universität Zürich, Winterthurerstrasse 190, CH-8057 Zürich, Switzerland
4 Département de Chimie, École Normale Supérieure, 24 rue Lhomond, F-75005 Paris, France
Received January 31, 2016, in final form February 12, 2016
The collective dynamics in liquid water is an active research topic experimentally, theoretically and via simula-
tions. Here, ab initiomolecular dynamics simulations are reported in heavy and ordinary water at temperature
323.15 K, or 50◦C. The simulations in heavy water were performed both with and without dispersion correc-
tions. We found that the dispersion correction (DFT-D3) changes the relaxation of density-density time corre-
lation functions from a slow, typical of a supercooled state, to exponential decay behaviour of regular liquids.
This implies an essential reduction of the melting point of ice in simulations with DFT-D3. Analysis of longitudi-
nal (L) and transverse (T) current spectral functions allowed us to estimate the dispersions of acoustic and optic
collective excitations and to observe the L-T mixing effect. The dispersion correction shifts the L and T optic (O)
modes to lower frequencies and provides by almost thirty per cent smaller gap between the longest-wavelength
LO and TO excitations, which can be a consequence of a larger effective high-frequency dielectric permittivity in
simulations with dispersion corrections. Simulation in ordinary water with the dispersion correction results in
frequencies of optic excitations higher than in D2O, and in a long-wavelength LO-TO gap of 24 ps−1 (127 cm−1).
Key words: collective excitations, optic modes, water, heavy water, van der Waals corrections, ab initio
molecular dynamics
PACS: 61.20.Ja, 61.20.Lc, 62.60.+v, 63.50.-x, 78.30.C-
1. Introduction
Collective dynamics in simple and molecular liquids is so far well understood only on macroscopic
length and time scales. Experiments on Brillouin scattering of light on liquids can be very nicely described
by the hydrodynamic theory, which explains themain relaxation and propagation processes contributing
to the measured scattered intensity of light [1, 2]. However, hydrodynamic theory, which is a collection of
fundamental local conservation laws (balance equations) for a system treated as continuum, is incapable
of describing the atomic-scale dynamics and the collective processes on nanoscale where the atomistic
structure of matter does not make it possible to treat the system as continuum. Atomistic molecular dy-
namics (MD) computer simulations is a very efficient tool in exploration of the dynamic properties of
condensed matter and provide a lot of precious information on time-dependent correlations in liquids
on nano- and atomic-scale resolution. However, the classical computer simulations make use of effective
interparticle potentials, which often do not permit, especially in the case of molecular liquids, to recover
the values of experimental melting points, Tm, or transport coefficients. In the case of water, the differ-
ence between the calculated melting temperature from classical MD simulations and the experimental
© T. Bryk, A.P. Seitsonen, 2016 23604-1
http://dx.doi.org/10.5488/CMP.19.23604
http://www.icmp.lviv.ua/journal
T. Bryk, A.P. Seitsonen
melting point of ice Ih can be over 80 K (with the SPC model) [3]. Depending on the values of effective
ionic charges, massless charges used for correction of a dipole moment, fixed intramolecular O–H bond
length and the H–O–H angle various water models result in different melting points which are usually
lower than the experimental value [3].
Ab initiomolecular dynamics (AIMD) simulations, which use the density functional theory (DFT) for
an electronic subsystem and classical equations of motion for ions, are free of effective interatomic po-
tentials because the input interactions for simulations are electron-ion pseudopotentials. On-the-fly min-
imization of total energy to the Born-Oppenheimer surface and the consequent equilibrium AIMD runs
make it possible to study instantaneous distributions of electron density that contain all the information
about the bonding and polarization fluctuations in a system. The AIMD of water and ice also face some
problems from the point of view of the correct melting point of ice Ih . Recently, two-phase ice/water
AIMD simulations with Becke-Lee-Yang-Parr (BLYP) [4, 5] exchange-correlation functional indicated that
at ambient pressure, the melting point of ice Ih should bemore than 400 K, but taking into account the so-
called (London) dispersion corrections to BLYP, the melting temperature was lowered to about 360 K [6].
Namely, one of the problems of DFT calculations with generalized gradient approximations (GGA), such
as BLYP, is that DFT-GGA does not reproduce the long-range attractive interaction (typical of Lennard-
Jones effective potentials), whose leading term in atom pairs is ∝ −r−6. The semiempirical dispersion
correction proposed by Grimme [7] to account for van der Waals interactions in DFT-GGA calculations
resulted in good agreement with experimental results in binding energies and intermolecular distances
for a variety of molecular complexes. Another approach to account for van der Waals corrections in DFT
calculations was initially proposed by Dion et al. [8]. Such corrections were used to study the effect of
dispersion correction on the structure and self-diffusion in water [9].
One can alsomention another direction in attempts to improve static and dynamic properties of water
from AIMD through application of different hybrid density functionals by making use of short-ranged
Hartree-Fock exchange [10], or the recent truly ab initio MD simulations using the MP2 perturbation
theory for correlations beyond the Hartree-Fock energy [11]. A newer generation (D3) of the dispersion
corrections was suggested in 2010 [12]. No AIMD simulations have been reported so far to check how the
D3 dispersion correction would affect the value of the calculated melting point of ice Ih .
Furthermore, no analysis yet regarding the time correlation functions and regarding the collective
dynamics of water from AIMD studies in which dispersion corrections would be used in simulations. The
collective dynamics of water itself is not completely understood, although many features such as the “fast
sound” phenomenon [13, 14], rotational and structural relaxation in water [15], and the gap existing be-
tween the longitudinal and transverse optic modes have been known for years. To date, there has been
no generalized hydrodynamic approach to the collective dynamics in water capable of recovering the
simulation results with full account for rotational and structural relaxations, although some successful
attempts were made mainly to describe polarization and dielectric relaxation in water [16–19]. As con-
cerns heavy water D2O, only a few studies were devoted to the analysis of collective dynamics [20–22].
In reference [21], the findings for heavy water were compared to the previous experimental scattering
studies on H2O and the existence of a very flat dispersion curve at ∼ 6 meV was attributed to optic modes.
A very interesting finding for heavy water was reported from classical MD simulations using SPC/E
model [22]: the analysis of longitudinal (L) and transverse (T) current spectral functions CL/T(k,ω), cal-
culated directly from the trajectories and the corresponding velocities of particles, showed a striking
feature which later on was referred to as a mixing between L and T dynamics in water: Starting from
wave numbers ∼ 0.5 Å−1, the L- and T-current spectral functions showed the emergence of a shoulder
at a frequency corresponding to the T- and L-excitations, which become well defined peaks at higher
wave numbers. These peaks gave evidence of both (L and T) excitations contributing to L- and T-current
spectral functions. Later on, the inelastic X-ray scattering (IXS) experiments on water [23] completely
supported the MD findings on the observation of the L-T coupling in reference [22]. Although there is a
clear hint that rotational motion of molecules can cause the L-T coupling, so far there has not appeared
either any derivation of a generalized hydrodynamic theory for water with coupled equations for L and
T dynamics capable of explaining the MD or experimental observations on L-T mixing in water. We can
mention here that the issue of the L-T coupling is not solely restricted to water or molecular liquids in
general: The L-T coupling effects were also observed while analysing the IXS scattering experiments on
liquid Hg [24] and Sn [25], as well as reported from the observation of a two-peak structure of current
23604-2
Collective excitations in water
spectral functions obtained in ab initio simulations of liquid Sn [26] and liquid Li at high pressures [27].
No L-T coupling effects were studied so far in water by ab initio simulations. Recently, there was a re-
port on classical simulations of supercooled water at temperature 180 K using TIP4P model [28]. It was
found from a fit of several Lorentzians that for supercooled water, transverse current spectral functions
contained contributions from four collective modes, while for longitudinal dynamics, only three contri-
butions from collective excitations were estimated. Furthermore, it was shown that for the three studied
systems (with density 0.939 g/cm3, 1.029 g/cm3 and 1.179 g/cm3) the high-frequency speed of longitudinal
acoustic modes increased from 3313 m/s to 3679 m/s.
Very recently, a classical simulation study of high-frequency dynamics of water using two TIP4P/2005f
(with 512 molecules) and TTM3F (with 256 molecules) water models was reported [19]. The main focus
was on calculations of polarization time correlation functions and estimation of dispersion of L and T
optic (LO and TO, respectively) modes. The gap between the longest-wavelength LO and TO excitations
obtained at the smallest accessible wave numbers in those simulations was larger than the experimental
values of LO-TO gap at the corresponding temperatures. One of the shortcomings of classical MD simu-
lations for ionic and polar liquids with non-polarizable and even polarizable effective potentials is the
neglect (for non-polarizable models) or a modelling (for polarizable models) of high-frequency polariz-
ability of electron subsystem. It was shown in reference [29], from a comparison of classical simulations
using a rigid ion model for molten salt NaCl and ab initio simulations at the same thermodynamic point,
that the high-frequency polarizability leads to a “softening” of the long-wavelength LO frequency by a
factor ∝
p
ε∞, while the transverse TO frequency becomes slightly higher (TO “hardening”) in compari-
son with the non-polarizable case. Here, ε∞ is the high-frequency dielectric permittivity. These findings
were supported by sequential AIMD simulations for molten salts NaI [30], RbF [31] and LiBr [32]. There-
fore, it would be interesting to study the LO-TO gap for most long-wavelength excitations in water from
ab initio simulations.
In this study, we aimed at calculating, from AIMD, the L- and T-current spectral functions and esti-
mating the dispersion of collective excitations. Initially, we intended to simulate only heavy water D2O
both with and without dispersion corrections to DFT in order to estimate the effect of dispersion cor-
rections on the decay of density-density time correlation functions, which is very sensitive to the reg-
ular liquid/supercooled/glassy state of the studied system, as well as to observe the effect of dispersion
corrections on the high-frequency speed of sound and LO, TO frequencies. Additional simulations were
performed in light, or “normal” water, with dispersion corrections applied, which enabled us to compare
our results on the dispersion of LO and TO excitations with the recently reported ones [19]. Another im-
portant issue was to estimate whether the L-T coupling effects are observed in ab initio simulations of
water and heavy water, and to estimate in general in what exactly consists the difference in dispersion
of collective excitations between heavy and regular water.
The remaining Paper is organized as follows: In the next section we supply details of our ab initio
simulations, while the third section contains results for static structure of D2O obtained from simulations
with andwithout dispersion corrections; the static structure in light water is practically identical to that in
heavywater. Further, wewill show the results for the density-density time correlation functions, L- and T-
current spectral functions, and dispersions of collective excitations. The last section contains conclusions
of this study.
2. Molecular dynamics simulations
We performed ab initio simulations having a collection of 128 molecules in a periodic box with the
box length of 15.7459 Å for three systems: heavy water D2O with D3 dispersion correction and without
it, and ordinary water H2O with D3 dispersion correction. The details of our simulations are very similar
to those used in an earlier work [33]. We performed Born-Oppenheimer molecular dynamics simulations
using the density functional theory to provide the forces into the Verlet algorithm. We integrated the
equations of motion with a time step of 0.5 fs in the canonical, NV T , ensemble, and used a Nosé-Hoover
thermostat chain with a time constant of 2.4 ps to equilibrate the average temperature at 323.15 K. The
slightly elevated temperature is motivated by the known deficiency of the approximations in the details
of the DFT treatment discussed below [33–35]. Either the mass of hydrogen or deuterium was used in
23604-3
T. Bryk, A.P. Seitsonen
simulations of light and heavy water, respectively. All the trajectories are at least 50 ps long, and the first
10 ps were neglected as a period of equilibration.
We used the BLYP generalized gradient approximation as the exchange-correlation term in the Kohn-
Sham equations of DFT. Further, the D3 approximation [12] was added to the Kohn-Sham expression of
total energy in order to incorporate the London dispersion forces missing in the BLYP, or (semi-)local
approximations in general.
The CP2K suite of programs [36] was used in the simulations, in particular its QuickStep [37] module.
The Gaussian Plane Wave (GPW) method [38] was used with the triple-zeta valence doubly polarized
(TZV2P) basis set of Gaussians and plane waves up to 400 Ry to expand the Kohn-Sham orbitals and
the electron density, respectively. We also employed the NN50 smoothing algorithm [37] to reduce the
errors in the evaluation of the energy and forces of the BLYP term, since tests [33] have shown that this
combination delivers a good convergence in the simulations. The action of the nuclei and core electrons
on the valence electrons was replaced by the Goedecker-Teter-Hutter (GTH) norm-conserving pseudo-
potentials [39, 40].
At each configuration, we calculated Fourier-components of the following dynamic variables: partial
particle densities
nα(k, t) =
1
p
Nα
Nα∑
i=1
e
−ikrα
i
(t )
, α=H/D, O (2.1)
densities of partial longitudinal momentum
JL
α
(k, t) =
mα
k
p
Nα
Nα∑
i=1
(kv
α
i )e
−ikrα
i
(t )
, α=H/D, O (2.2)
and densities of partial transverse momentum
J
T
α
(k, t) =
mα
k
p
Nα
Nα∑
i=1
[kv
α
i ]e
−ikrα
i
(t )
, α=H/D, O. (2.3)
Here, NO and NH/D are 128 and 256, respectively, in our simulations. The smallest accessible wave num-
ber from our AIMD was 0.399 Å−1. We studied the range of wave numbers up to 4.5 Å−1. All the wave
number-dependent quantities calculated from AIMD were averaged over all possible directions of wave
vectors corresponding to the same absolute value. The partial densities were used for calculations of
collective Bhatia-Thornton time correlation functions and corresponding static structure factors either
in “T–C” (total number–concentration) or in “M–X” (total mass–mass-concentration) representations [41].
All these partial and collective representations contain the full information about the same collective
mode, although, as it was shown in [42, 43], the mass-concentration currents are the most useful ones in
the studies of optic collective modes.
3. Results and discussion
Partial pair distribution functions from simulations with and without the D3 dispersion correction
are shown in figure 1. The simulations of D2O and H2O both with dispersion correction resulted in prac-
tically identical static, or average, structure. One can see in figure 1 that the D3 correction led to a strong
reduction in the depth of the first minimum of partial O–O distribution compared to the regular BLYP
simulations, which resulted in an overbonded water molecules; this is related to a very high melting tem-
perature of ice Ih from AIMD with BLYP exchange-correlation functional, suggested in reference [6]. The
height of all the first intermolecular peaks in the partial distribution functions was reduced in BLYP+D3
simulations, and first intermolecular minima shifted to larger values. This means that there is more ac-
tive exchange of molecules in the first coordination shell with molecules from the bulk, being consistent
with an enhanced diffusion constant found in reference [33]. The inclusion of the D3 dispersion correc-
tion slightly reduced the values of intramolecular O–H distance and H–O–H angle, as in the right-hand
frames of figure 1 one can see that the corresponding distributions were shifted towards smaller values
in simulations with BLYP+D3. In simulations with classical equations of motion for ions, there is prac-
tically no difference in the results obtained for H2O and D2O, which appears in quantum treatment via
23604-4
Collective excitations in water
Figure 1. (Color online) Partial pair distribution functions gi j (r ) and distributions of intramolecular O–
D lengths and D–O–D angles for D2O at T = 323.15 K obtained from BLYP (red solid line) and BLYP+D3
(dashed green line) simulations. The amplitude of intramolecular peak in gOD(r ) is 33.6 and 34.6 with
BLYP and BLYP+D3, respectively. Partial pair distribution functions and distributions of intramolecular
bond lengths and angles in H2O with BLYP+D3 were practically identical to D2O-BLYP+D3 results. The
experimental data for water at T = 298 K and pressure 1 bar were taken from reference [45].
path integral simulations [44, 46]. In figure 1, by blue short-dash line we show the experimental data
for partial distribution functions in water at T = 298 K [45], which are in agreement with our BLYP+D3
simulations, especially this being clearly seen in the O–O pair distribution function.
The angular O–O–Odistribution functions in the first coordination shell are shown infigure 2. One can
compare these distributions with similar ones reported from classical simulations of SPC/Ewater [47, 48].
There is a resemblance in the angular O–O–O distributions from BLYP simulations with the results from
SPC/E model in the temperature range 200–220 K that is in an supercooled state, because the estimates
for the melting point of SPC/E model from two-phase classical simulations resulted in 225 K [47, 49];
thermodynamic integration technique for SPC/E model resulted in a somewhat lower value of 215 K [3].
The large peak at about 105◦ corresponds to tetrahedral ordering of the nearest neighbors. One can see
that the application of D3 dispersion correction significantly reduces the contribution from tetrahedral
configurations of nearest neighbors, which is actually in line with a suggestion of the reduction of the
melting point of ice in simulations with D3 dispersion correction. In the literature one can find a similar
23604-5
T. Bryk, A.P. Seitsonen
Figure 2. (Color online) Angular O–O–O distribution functions in D2O at T = 323.15 K fromBLYP (red solid
line) and BLYP+D3 (dashed green line) simulations, obtained within the radial cut-off of 3.375 Å. Results
in H2O with BLYP+D3 were practically identical to D2O ones with BLYP+D3. Path integral simulations
analyzed in reference [46] yield in H2O and D2O results in reasonable agreement with our BLYP+D3
simulations.
analysis of O–O–O angle distributions for H2O andD2O, obtained from path integral simulations [46]. Both
O–O–O angle distributions in [46] show pronounced maxima at about 100◦ −105◦ and a small shoulder
close to 60◦, which is in agreement with our BLYP+D3 simulations.
Partial static structure factors are of huge importance in understanding the role of partial contribu-
tions to the measured X-ray or neutron diffraction intensities. Here, we report the Bhatia-Thornton total
density Stt(k) and concentration Scc(k) static structure factors. The total density structure factor shown in
figure 3 has themain peak at∼ 2 Å−1 and a well-defined shoulder in wave number range k ∼ 2.5−2.8 Å−1,
which is in agreement with experimental results [50] and ab initio calculations of X-ray spectra of liquid
water [51]. The concentration structure factor Scc(k) has a well-defined maximum at k ∼ 3.1 Å, which
is related to short-range order of hydrogen bonds. The concentration structure factor should have the
same long-wavelength asymptote of ∼ k2 as the charge structure factor. In figure 3 (b), we show the long-
wavelength asymptotes of concentration structure factors in simulations with and without D3 dispersion
correction, which nicely recover the ∼ k2 behaviour. The coefficient at the k2 term depends on the value
of the high-frequency dielectric permittivity ε∞, which makes it possible to estimate ε∞ if classical sim-
ulations with non-polarizable effective interaction models were available at the same thermodynamic
Figure 3. (Color online) Bhatia-Thornton static structure factors in H2O at T = 323.15 K in BLYP+D3 ap-
proximation (a) and comparison of the long-wavelength asymptotes of the concentration structure fac-
tors from BLYP and BLYP+D3 simulations of D2O (b). The straight lines in (b) correspond to a fit with k2
dependence.
23604-6
Collective excitations in water
Figure 4. (Color online) Total density autocorrelation functions for three wave numbers fromBLYP (a) and
BLYP+D3 (b) simulations of D2O at T = 323.15 K. The time unit τ= 0.39475 ps. The Ftt(k, t) in (a) contain
long tails which are typical of slow stretched-exponential relaxation in undercooled liquids, while in (b)
the relaxation of Ftt(k, t) is typically a simple exponential one as for regular liquids.
point, like it was done in reference [29] in liquid NaCl. Here, figure 3 (b) implies that the system simulated
with D3 dispersion correction has a larger value of (effective) ε∞. This fact is important in analysing the
frequencies of longitudinal optic modes that experience a softening for a system with larger values of ε∞.
Electron structure simulations further permit to perform calculations of more general total charge struc-
ture factors SQQ (k), where Q(k, t) are the total charge fluctuations composed of point positive ions and
instantaneous distributions of electron density in AIMD. Recently, it was shown that the long-wavelength
asymptote of SQQ(k) in water recovers the ∼ k2 behaviour [52].
The total density-density time correlation functions Ftt(k, t) at three wave numbers, calculated from
simulations of D2O with and without D3 dispersion correction, are shown in figure 4. In general, the de-
cay of density-density time correlation functions is very sensitive to the studied thermodynamic point
on the phase diagram. At low wave numbers, the hydrodynamic theory predicts leading contributions
to Ftt(k, t) coming from longitudinal collective excitations (and represented by a damped oscillator) and
from thermal relaxation having an exponential tail. The exponential decay of the tail of density-density
time correlation inherent to an ordinary liquid state is replaced by a slow stretched-exponential decay
if the liquid becomes supercooled. By approaching the glass transition, the tail of the Ftt(k, t) shows fea-
tures connected with α-relaxation, and ultimately, at the glass transition, there appears a non-decaying
plateau Ftt(k, t →∞) = f (k) known as the non-ergodicity factor [53]. Moreover, for supercritical fluids it
was recently shown that the analysis of density-density correlations allows one to discriminate between
liquid-like and gas-like types of fluid in the supercritical regime [54, 55].
In figure 4, one can see that the total density time correlation functions obtained from simulations
without D3 dispersion correction show slow relaxations typical of supercooled liquids, while our AIMD
simulations with D3 dispersion correction give evidence of regular exponential tails of Ftt(k, t). This is a
further proof that the simulated thermodynamic point is below the melting line in simulations without
accounting for dispersion corrections, while BLYP+D3 simulations result in density-density correlations
typical of a regular liquid state.
Longitudinal and transverse total current spectral functions CL/T
tt (k,ω) obtained from simulations of
D2O (figure 5) and H2O (figure 6) with D3 dispersion correction contain contributions from propagating
L and T collective excitations as well as from intramolecular normal modes. Due to a larger mass of D
with respect to H, the intramolecular part in D2O is located in the range of frequencies 180–600 ps−1,
while in light water, the intramolecular vibrations are at higher frequencies, 300–800 ps−1. The L and T
total current spectral functions have a high-frequency peak corresponding to dispersionless (localized)
vibrations at frequencies ∼ 226 ps−1 in D2O (figure 5) and ∼ 310 ps−1 in H2O (figure 6). Simulations with
and without D3 dispersion correction in D2O resulted practically in the same frequency. These peaks
correspond to intramolecular bending normal modes, which have nearly two times lower frequencies
than those of the symmetric stretch normal modes.
The frequencies of L and T acoustic collective excitations can be observed as the low-frequency peaks
23604-7
T. Bryk, A.P. Seitsonen
in the corresponding total current spectral functions CL/T
tt (k,ω). In figures 5 and 6 one can see the emer-
gence of another low-frequency peak at k > 1 Å−1 in C L
tt(k,ω), which practically coincides with the fre-
quency of the transverse excitation. This is the effect of L-T mixing reported previously from classical MD
simulations [22] and X-ray scattering experiments [23].
Figure 5. (Color online) Total longitudinal (L) and
transverse (T) current spectral function CL/T
tt
(k,ω) at
three wave numbers from BLYP+D3 simulations of
D2O at T = 323.15 K.
Figure 6. (Color online) Total longitudinal (L) and
transverse (T) current spectral function CL/T
tt
(k,ω) at
three wave numbers from BLYP+D3 simulations of
H2O at T = 323.15 K.
One can easily observe the optic-like modes as peaks in the mass-concentration current spectral func-
tions CL/T
xx (k,ω), shown for light water in figure 7. A nice property of CL/T
xx (k,ω) spectral function is that
the contributions from acoustic modes in it are suppressed and the high-frequency optic modes are well
observable, in contrast to the CL/T
tt (k,ω) spectral function. The L and T optic modes clearly appear at
different frequencies in figure 7, that is typical for ionic melts and polar fluids, and result in a LO-TO
gap in the long-wavelength region. For ionic crystals, there is a famous Lyddane-Sachs-Teller relation for
the difference between long-wavelength LO and TO phonons (non-damped propagating modes of small
23604-8
Collective excitations in water
Figure 7. (Color online) Mass-concentration current spectral function CL/T
xx (k,ω) at three wave numbers
from BLYP+D3 simulations of H2O at T = 323.15 K.
displacements of positive/negative ions from equilibrium positions with opposite phases) [56]
ω
2
LO
ω
2
TO
=
ε(0)
ε∞
, (3.1)
where ε(0) is the macroscopic static dielectric permittivity of a non-conducting system. Although in non-
conducting liquids there should be some corrections to the Lyddane-Sachs-Teller due to the damping of
optic modes, which in its turn renormalizes the observed LO and TO frequencies, this relation gives a
hint of how the high-frequency screening by the electron density reduces the LO-TO gap.
The dispersion curves of acoustic and optic collective excitations obtained from the L and T current
spectral functions for D2Owith andwithout D3 dispersion correction are shown in figure 8. The small size
of the simulated system does not allow one to observe the transition towards hydrodynamic dispersion
law with adiabatic speed of sound. The dashed straight lines indicate the linear dispersion law with the
apparent high-frequency speed of sound. We obtained a reduction in the value of the apparent high-
frequency speed of sound from 3640 to 3070 m/s when the D3 dispersion correction was switched on. At
the lowest accessible wave number, the value of the LO-TO gap was also lower when the D3 dispersion
23604-9
T. Bryk, A.P. Seitsonen
Figure 8. (Color online) Dispersion of collective excitations obtained from peak positions of the longitudi-
nal (L) and transverse (T) current spectral functions CL,T(k,ω) in D2O at T = 323.15 K from BLYP (a) and
BLYP+D3 (b) simulations. The dashed straight line corresponds to a high-frequency linear dispersion law
with a speed of sound of 3640 m/s (a) and 3070 m/s (b).
correction was applied. The longitudinal LO frequency was reduced from 123 to 107 ps−1, while the
transverse TO frequency changed from 113.8 to 102.8 ps−1, resulting in a reduction of the LO-TO gap
from 16 to 11 ps−1 due to the effect of D3 dispersion correction, which transforms the collective dynamics
typical of an undercooled state to the regular dynamics of liquid as well as increased the high-frequency
dielectric permittivity [see figure 3 (b)]. Both simulations with and without D3 dispersion correction show
the same L-T mixing as was reported earlier from classical MD simulations [22]. In both cases, there is
also a similar behaviour in the TO mode, whose frequency decreases with an increasing wave number
up to k ∼ 1.5 Å−1. At larger wave numbers, it changes the dependence, showing a small increase up to
a wave number k ∼ 2.5 Å−1, where it merges with the LO branch. At higher wave numbers, LO and
TO branches stay at the same frequencies. This behaviour of dispersion implies that there is a similar
crossover between the intrinsically collective (k < 1.5 Å−1) and partial (at k > 1.5 Å−1) types of dynamics
described in detail in reference [42].
In figure 9, we present the dispersion of longitudinal and transverse collective excitations in liquid
H2O from simulations using BLYP+D3 approach. The value of apparent high-frequency speed of sound is
3223 m/s, which is in good agreement with the values ∼ 3100 m/s obtained from the X-ray scattering ex-
periments on water close to the melting point [57] and ∼ 3200 m/s reported in [21]. The optic branches of
collective excitations due to the smaller molecular mass are located at higher frequencies than in the case
of heavy water. At the long-wavelength limit, the LO and TO frequencies are 152.2 and 128.2 ps−1, respec-
tively, yielding our estimate for the LO-TO gap of 24 ps−1. The frequency of the LO long-wavelength exci-
23604-10
Collective excitations in water
Figure 9. (Color online) Dispersion of collective excitations obtained from peak positions of the longitu-
dinal (L) and transverse (T) current spectral functions CL,T(k,ω) in H2O at T = 323.15 K from BLYP+D3
simulations. The dashed straight line corresponds to a high-frequency linear dispersion law with a speed
of sound of 3223 m/s.
tation is slightly lower than the one obtained from classical simulations using a flexible non-polarizable
model TIP4P/2005f [19]. This difference could be due to a softening of the LO mode in AIMD due to the
high-frequency screening effects. Our calculated TO frequency is higher than the one from the classical
simulations [19], thus both effects resulting in a smaller long-wavelength LO-TO gap from AIMD.
4. Conclusions
We performed ab initiomolecular dynamics simulations for heavy water with and without the third
generation dispersion correction of Grimme and co-workers D3 to DFT [12]. We found that the D3 disper-
sion correction strongly affects the static and dynamic structure of heavywater. Very sensitive to the ther-
modynamic state, the behaviour of density-density time correlation functions showed slow, stretched ex-
ponential relaxation, typical of supercooled liquids, in the case of simulations without D3 dispersion cor-
rection, and an exponential relaxation of regular liquids when the dispersion correction was used. This
finding could be a consequence of a significant reduction of the melting point of ice in the DFT/BLYP+D3
simulations compared to BLYP-only approach.
Longitudinal and transverse current spectral functions were analyzed in order to estimate the dis-
persion of acoustic and optic excitations. The application of the D3 dispersion correction resulted in a
significant reduction of the apparent high-frequency speed of sound, as well as in a reduction of the gap
between long-wavelength LO and TO excitations. This fact indicates that the D3 dispersion correction
leads to higher values of the high-frequency dielectric permittivity ε∞. This change is effective, or indi-
rect, since the D3 dispersion correction does not directly involve the electronic structure but only via the
changed static and dynamic structure of the water.
Another simulation with D3 dispersion correction was performed for light water at the same temper-
ature and number density, and it resulted in an identical static structure as in D2Owith the D3 dispersion
correction. Due to a smaller molecular mass, higher frequencies of optic L and T and of intramolecular
modes were obtained.We have a good agreement for the value of the high-frequency speed of sound with
X-ray scattering experiments. The gap between the long-wavelength LO and TO excitations in light water
was obtained as ∼ 24 ps−1, or 127 cm−1. The calculated frequency of the LO excitations from our AIMD is
slightly lower than the one from classical simulations using flexible non-polarizable model TIP4P/2005f
[19], while frequencies of transverse optic excitations in our ab initio simulations with D3 dispersion
correction were much higher than the ones from classical simulations.
In all three simulations, we observed the L-T mixing effect, since the L/T current spectral functions
contained an additional peak, corresponding to the T/L excitations, at relatively large wave numbers.
23604-11
T. Bryk, A.P. Seitsonen
This L-T mixing effect was observed only for acoustic excitations; the optic modes appeared at separated
frequencies at wave numbers k < 2.5 Å−1. So far there is no indication that TO modes can appear in the
longitudinal spectra.
Our study gives indications of a significant reduction of the ice Ih melting temperature in ab initio
simulations with D3 dispersion correction applied. There is thus a quest for large AIMD simulations with
D3 dispersion correction on the melting of ice Ih .
Acknowledgements
APS acknowledges Guillaume Ferlat, A Marco Saitta and Rodolphe Vuilleumier for many illuminating
discussions on simulations of water, and the computing resources at the Universität Zürich, Zurich, and
Centro Svizzero di Calcolo Scientifico, CSCS, Ticino, Switzerland.
References
1. Hansen J.-P., McDonald I.R., Theory of Simple Liquids, Academic, London, 1986.
2. Boon J.-P., Yip S., Molecular Hydrodynamics, McGraw-Hill, New York, 1980.
3. Vega C., Sanz E., Abascal J.L.F., J. Chem. Phys., 2005, 122, 114507; doi:10.1063/1.1862245.
4. Lee C., Yang W., Parr R.G., Phys. Rev. B, 1988, 37, 785; doi:10.1103/PhysRevB.37.785.
5. Becke A.D., Phys. Rev. A, 1988, 38, 3098; doi:10.1103/PhysRevA.38.3098.
6. Yoo S., Xantheas S.S., J. Chem. Phys., 2011, 134, 121105; doi:10.1063/1.3573375.
7. Grimme S., J. Comput. Chem., 2005, 25, 1463; doi:10.1002/jcc.20495.
8. Dion M., Rydberg H., Schröder E., Langreth D.C., Lundqvist B.I., Phys. Rev. Lett., 92, 246401;
doi:10.1103/PhysRevLett.92.246401.
9. Wang J., Roman-Perez G., Soler J.M., Artacho E., Fernandez-Serra M.-V., J. Chem. Phys., 2011, 134, 024516;
doi:10.1063/1.3521268.
10. Todorova T., Seitsonen A.P., Hutter J., Kuo I.-F.W., Mundy C.J., J. Phys. Chem. B, 2006, 110, 3685;
doi:10.1021/jp055127v.
11. Del Ben M., Hutter J., VandeVondele J., J. Chem. Phys., 2015, 143, 054506; doi:10.1063/1.4927325.
12. Grimme S., Antony J., Ehrlich S., Krieg H., J. Chem. Phys., 2010, 132, 154104; doi:10.1063/1.3382344.
13. Rahman A., Stillinger F.H., Phys. Rev. A, 1974, 10, 368; doi:10.1103/PhysRevA.10.368.
14. Ruocco G., Sette F., Condens. Matter Phys., 2008, 11, 29; doi:10.5488/CMP.11.1.29.
15. Ruocco G., Sette F., J. Phys.: Condens. Matter, 1999, 11, R259; doi:10.1088/0953-8984/11/24/202.
16. Bopp P.A., Kornyshev A.A., Sutmann G., J. Chem. Phys., 1998, 109, 1939; doi:10.1063/1.476884.
17. Omelyan I.P., Tokarchuk M.V., J. Phys.: Condens. Matter, 2000, 12, L505; doi:10.1088/0953-8984/12/30/104.
18. Sedlmeier F., Shadkhoo S., Bruinsma R., Netz R.R., J. Chem. Phys., 2014, 140, 054512; doi:10.1063/1.4863444.
19. Elton D.C., Fernandez-Serra M., Nat. Commun., 2016, 7, 10193; doi:10.1038/ncomms10193.
20. Shelton D.P., J. Chem. Phys., 2002, 117, 9374; doi:10.1063/1.1514976.
21. Sacchetti F., Suck J.-B., Petrillo C., Dorner B., Phys. Rev. E, 2004, 69, 061203; doi:10.1103/PhysRevE.69.061203.
22. Sampoli M., Ruocco G., Sette F., Phys. Rev. Lett., 1997, 79, 1678; doi:10.1103/PhysRevLett.79.1678.
23. Cimatoribus A., Saccani S., Bencivenga F., Gessini A., Izzo M.G., Masciovecchio C., New J. Phys., 2010, 12, 053008;
doi:10.1088/1367-2630/12/5/053008.
24. Hosokawa S., Inui M., Kajihara Y., Matsuda K., Ichitsubo T., Pilgrim W.-C., Sinn H., Gonzalez L.E., Gonzalez D.J.,
Tsutsui S., Baron A.Q.R., Phys. Rev. Lett., 2009, 102, 105502; doi:10.1103/PhysRevLett.102.105502.
25. Hosokawa S., Munejiri S., Inui M., Kajihara Y., Pilgrim W.-C., Ohmasa Y., Tsutsui S., Baron A.Q.R., Shimojo F.,
Hoshino K., J. Phys.: Condens. Matter, 2013, 25, 112101; doi:10.1088/0953-8984/25/11/112101.
26. Munejiri S., Shimojo F., Hoshino K., Phys. Rev. B, 2012, 86, 104202; doi:10.1103/PhysRevB.86.104202.
27. Bryk T., Ruocco G., Scopigno T., Seitsonen A.P., J. Chem. Phys., 2015, 143, 104502; doi:10.1063/1.4928976.
28. Jedlovszky P., Garberoglio G., Vallauri R., Phys. Chem. Chem. Phys., 2011, 13, 19823; doi:10.1039/C1CP21850F.
29. Bryk T., Mryglod I., Chem. Phys. Lett., 2008, 466, 56; doi:10.1016/j.cplett.2008.10.045.
30. Bryk T., Mryglod I., Phys. Rev. B, 2009, 79, 184206; doi:10.1103/PhysRevB.79.184206.
31. Bryk T., Mryglod I., Int. J. Quantum Chem., 2010, 110, 38; doi:10.1002/qua.22327.
32. Bryk T., Klevets I., J. Chem. Phys., 2012, 137, 224508; doi:10.1063/1.4770269.
33. Jonchiere R., Seitsonen A.P., Ferlat G., Saitta A.M., Vuilleumier R., J. Chem. Phys., 2011, 135, 154503;
doi:10.1063/1.3651474.
23604-12
http://dx.doi.org/10.1063/1.1862245
http://dx.doi.org/10.1103/PhysRevB.37.785
http://dx.doi.org/10.1103/PhysRevA.38.3098
http://dx.doi.org/10.1063/1.3573375
http://dx.doi.org/10.1002/jcc.20495
http://dx.doi.org/10.1103/PhysRevLett.92.246401
http://dx.doi.org/10.1063/1.3521268
http://dx.doi.org/10.1021/jp055127v
http://dx.doi.org/10.1063/1.4927325
http://dx.doi.org/10.1063/1.3382344
http://dx.doi.org/10.1103/PhysRevA.10.368
http://dx.doi.org/10.5488/CMP.11.1.29
http://dx.doi.org/10.1088/0953-8984/11/24/202
http://dx.doi.org/10.1063/1.476884
http://dx.doi.org/10.1088/0953-8984/12/30/104
http://dx.doi.org/10.1063/1.4863444
http://dx.doi.org/10.1038/ncomms10193
http://dx.doi.org/10.1063/1.1514976
http://dx.doi.org/10.1103/PhysRevE.69.061203
http://dx.doi.org/10.1103/PhysRevLett.79.1678
http://dx.doi.org/10.1088/1367-2630/12/5/053008
http://dx.doi.org/10.1103/PhysRevLett.102.105502
http://dx.doi.org/10.1088/0953-8984/25/11/112101
http://dx.doi.org/10.1103/PhysRevB.86.104202
http://dx.doi.org/10.1063/1.4928976
http://dx.doi.org/10.1039/C1CP21850F
http://dx.doi.org/10.1016/j.cplett.2008.10.045
http://dx.doi.org/10.1103/PhysRevB.79.184206
http://dx.doi.org/10.1002/qua.22327
http://dx.doi.org/10.1063/1.4770269
http://dx.doi.org/10.1063/1.3651474
Collective excitations in water
34. Lin I.-C., Seitsonen A.P., Coutinho-Neto M.D., Tavernelli I., Rothlisberger U., J. Phys. Chem. B, 2009, 113, 1127;
doi:10.1021/jp806376e.
35. Schmidt J., VandeVondele J., Kuo I.-F.W., Sebastiani D., Siepmann J.I., Hutter J., Mundy C.J., J. Phys. Chem. B, 2009,
113, 11959; doi:10.1021/jp901990u.
36. CP2K Developers’ Group, 2016; CP2K is free available via http://www.cp2k.org/+.
37. VandeVondele J., Krack M., Mohamed F., Parrinello M., Chassaing T., Hutter J., Comput. Phys. Commun., 2005,
167, 103; doi:10.1016/j.cpc.2004.12.014.
38. Lippert G., Hutter J., Parrinello M., Mol. Phys., 1997, 92, 477; doi:10.1080/002689797170220.
39. Hartwigsen C., Goedecker S., Hutter J., Phys. Rev. B, 1998, 58, 3641; doi:10.1103/PhysRevB.58.3641.
40. Goedecker S., Teter M., Hutter J., Phys. Rev. B, 1996, 54, 1703; doi:10.1103/PhysRevB.54.1703.
41. Bhatia A.B., Thornton D.E., March N.H., Phys. Chem. Liq., 1974, 4, 97; doi:10.1080/00319107408084276.
42. Bryk T., Mryglod I., J. Phys.: Condens. Matter, 2000, 12, 6063; doi:10.1088/0953-8984/12/28/305.
43. Bryk T., Mryglod I., Phys. Rev. E, 2000, 62, 2188; doi:10.1103/PhysRevE.62.2188.
44. Chen B., Ivanov I., KleinM.L., Parrinello M., Phys. Rev. Lett., 2003, 91, 215503; doi:10.1103/PhysRevLett.91.215503.
45. Soper A.K., Chem. Phys., 2000, 258, 121; doi:10.1016/S0301-0104(00)00179-8.
46. Soper A.K., Benmore C.J., Phys. Rev. Lett., 2008, 101, 065502; doi:10.1103/PhysRevLett.101.065502.
47. Bryk T., Haymet A.D.J., Mol. Simul., 2004, 30, 131; doi:10.1080/0892702031000152172.
48. Galicia-Andr’es E., Dominguez H., Pizio O., Condens. Matter Phys., 2015, 18, 13603; doi:10.5488/CMP.18.13603.
49. Bryk T., Haymet A.D.J., J. Chem. Phys., 2002, 117, 10258; doi:10.1063/1.1519538.
50. Hura G., Sorenson J.M., Glaeser R.M., Head-Gordon T., J. Chem. Phys., 2000, 113, 9140; doi:10.1063/1.1319614.
51. Krack M., Gambirasio A., Parrinello M., J. Chem. Phys., 2002, 117, 9409; doi:10.1063/1.1517040.
52. Klevets I., Bryk T., J. Chem. Phys., 2014, 141, 214504; doi:10.1063/1.4902543.
53. Götze W., Mayr M.R., Phys. Rev. E, 2000, 61, 587; doi:10.1103/PhysRevE.61.587.
54. Simeoni G., Bryk T., Gorelli F.A., Krisch M., Ruocco G., Santoro M., Scopigno T., Nat. Phys., 2010, 6, 503;
doi:10.1038/nphys1683.
55. Bryk T., Mryglod I., Scopigno T., Ruocco G., Gorelli F., Santoro M., J. Chem. Phys., 2010, 133, 024502;
doi:10.1063/1.3442412.
56. Lyddane R., Sachs R., Teller E., Phys. Rev., 1941, 59, 673; doi:10.1103/PhysRev.59.673.
57. Monaco G., Cunsolo A., Ruocco G., Sette F., Phys. Rev. E, 1999, 60, 5505; doi:10.1103/PhysRevE.60.5505.
23604-13
http://dx.doi.org/10.1021/jp806376e
http://dx.doi.org/10.1021/jp901990u
http://www.cp2k.org/+
http://dx.doi.org/10.1016/j.cpc.2004.12.014
http://dx.doi.org/10.1080/002689797170220
http://dx.doi.org/10.1103/PhysRevB.58.3641
http://dx.doi.org/10.1103/PhysRevB.54.1703
http://dx.doi.org/10.1080/00319107408084276
http://dx.doi.org/10.1088/0953-8984/12/28/305
http://dx.doi.org/10.1103/PhysRevE.62.2188
http://dx.doi.org/10.1103/PhysRevLett.91.215503
http://dx.doi.org/10.1016/S0301-0104(00)00179-8
http://dx.doi.org/10.1103/PhysRevLett.101.065502
http://dx.doi.org/10.1080/0892702031000152172
http://dx.doi.org/10.5488/CMP.18.13603
http://dx.doi.org/10.1063/1.1519538
http://dx.doi.org/10.1063/1.1319614
http://dx.doi.org/10.1063/1.1517040
http://dx.doi.org/10.1063/1.4902543
http://dx.doi.org/10.1103/PhysRevE.61.587
http://dx.doi.org/10.1038/nphys1683
http://dx.doi.org/10.1063/1.3442412
http://dx.doi.org/10.1103/PhysRev.59.673
http://dx.doi.org/10.1103/PhysRevE.60.5505
T. Bryk, A.P. Seitsonen
Дослiдження колективних збуджень у водi методом
першопринципної молекулярної динамiки: ефект
ван дер вальсiвських поправок на дисперсiю
колективних збуджень
Т. Брик1,2, А.П. Сейтсонен3,4
1 Iнститут фiзики конденсованих систем НАН України, вул. I. Свєнцiцького, 1, 79011 Львiв, Україна
2 Iнститут прикладної математики та фундаментальних наук, Нацiональний Унiверситет ”Львiвська
Полiтехнiка”, 79013 Львiв, Україна
3 Iнститут хiмiї, Унiверситет Цюрiха, Вiнтертурерштрассе, 190, CH-8057 Цюрiх, Швейцарiя
4 Хiмiчний факультет, Вища нормальна школа, 24 вул. Льомон, F-75005 Париж, Францiя
Колективна динамiка у рiдкiй водi активно дослiджується експериментальними, теоретичними методами
та комп’ютерними симуляцiями. Представлено моделювання методом першопринципної молекулярної
динамiки для важкої та звичайної води при температурi 323.15 K, чи 50◦C. Моделювання для важкої води
були проведенi з та без дисперсiйних поправок. Ми отримали, що дисперсiйнi поправки (DFT-D3) суттє-
во змiнюють релаксацiю часових кореляцiйних функцiй густина-густина з повiльної, яка є типовою для
переохолодженого стану, до експоненцiйного спаду як для звичайних рiдин. Це означає суттєве змен-
шення точки плавлення льоду для моделювання з DFT-D3. Аналiз повздовжних (L) та поперечних (T)
спектральних функцiй потокiв дозволив визначити дисперсiї акустичних та оптичних збуджень та спо-
стерiгати ефект L-T змiшування. Дисперсiйнi поправки D3 зсувають до нижчих частот L та T оптичнi (O)
моди та приводять до меншої на майже тридцять процентiв щiлини мiж LO та TO довгохвильовими збу-
дженнями, що може бути наслiдком бiльшої високочастотної дiелектричної проникностi в моделюваннi
з дисперсiйними поправками. Моделювання для звичайної води з дисперсiйними поправками дає вищi
частоти оптичних збуджень, нiж для D2O, та щiлину LO-TO порядку 24 пс−1 (127 см−1) у довгохвильовiй
границi.
Ключовi слова: колективнi збудження, оптичнi моди, вода, важка вода, ван дер вальсiвськi поправки,
першопринципне комп’ютерне моделювання
23604-14
Introduction
Molecular dynamics simulations
Results and discussion
Conclusions
|