You are on page 1of 9

Energy &

Environmental Science

View Article Online / Journal Homepage / Table of Contents for this issue

Dynamic Article Links <

Cite this: Energy Environ. Sci., 2012, 5, 7033

PAPER

Published on 29 February 2012. Downloaded by University of Hong Kong Libraries on 19/10/2014 07:25:55.

www.rsc.org/ees

Replacement mechanism of methane hydrate with carbon dioxide from


microsecond molecular dynamics simulations
Dongsheng Bai,a Xianren Zhang,a Guangjin Chenb and Wenchuan Wanga
Received 20th July 2011, Accepted 20th February 2012
DOI: 10.1039/c2ee21189k
Replacement of CH4 in hydrate form with CO2 is a candidate for recovering CH4 gas from its hydrates
and storing CO2. In this work, microsecond molecular dynamics simulations were performed to study
the replacement mechanism of CH4 hydrate by CO2 molecules. The replacement process is found to be
controlled cooperatively by the chemical potentials of guest molecules, memory effect, and mass
transfer. The replacement pathway includes the melting of CH4 hydrate near the hydrate surface and
the subsequent formation of an amorphous CO2 hydrate layer. A large number of hydrate residual
rings left after the melting of CH4 hydrate facilitate the nucleation of CO2 hydrate and enhance the
dynamic process, indicating the existence of so-called memory effect. In the dynamic aspect, the
replacement process takes place near the surface of CH4 hydrate rather easily. However, as the
replacement process proceeds, the formation of the amorphous layer of the CO2 hydrate provides
a significant barrier to the mass transfer of the guest CH4 and CO2 molecules, which prevents the CH4
hydrate from further dissociation and slows down the replacement rate.

1. Introduction
Clathrate hydrates are a special class of non-stoichiometric
compounds formed by liquid water and gas molecules.1 The cage
lattice of hydrates can adopt many complex forms, but only three
of them are known to be naturally occurring: structure I (sI),
structure II (sII) and structure H (sH).2 For a unit cell of sI
hydrate, there are six T-51262 cages, two D-512 cages and 46 water
molecules in total.
Natural gas hydrates are abundant in permafrost regions and
seafloor of continental margins.3,4 Kvenvolden5 and Makogon6
pointed out that the amount of gas in known hydrate reserves up
until 1988 was at least twice as much as the energy contained in
a
Division of Molecular and Materials Simulation, State Key Laboratory of
Organic-Inorganic Composites, Beijing University of Chemical
Technology, Beijing 100029, China. E-mail: zhangxr@mail.buct.edu.cn
b
State Key Laboratory of Heavy Oil Processing, School of Chemical
Engineering, China University of Petroleum, Beijing 102249, China.
E-mail: gjchen@cup.edu.cn

the total fossil fuel reserves. Significant amounts of natural gas


hydrates might be considered as a new source of sustainable
energy. Some prospective plans are being studied to develop
reliable extraction schemes from the hydrate sediments, although
exploitation of these sediments in unfavorable circumstances
could drastically modify the marine ecosystem and even generate
underwater gas blowouts. Another risk for the hydrate sediments
comes from global climate change. A slight global warming
would raise the hydrate temperature above the equilibrium point,
and as a result the dissociation of natural gas hydrate would
release a great quantity of CH4,7 which is a so-called greenhouse
gas.
Of all the greenhouse gases, CO2 is responsible for about 64%
of the enhanced greenhouse effect. Given that the greenhouse
effect is undeniably responsible for climate warming, reducing
the quantity of CO2 released into the atmosphere is a major
environmental challenge. In order to exploit the natural gas in
hydrate form and disposal CO2, an original perspective proposed
is to replace CH4 with CO2 from natural gas hydrate.810 If the

Broader context
Natural gas hydrates are abundant in permafrost regions and seafloor of continental margins. In order to exploit the natural gas in
hydrate form and disposal CO2, an original perspective was recently proposed to replace CH4 with CO2 from natural gas hydrate.
Although recent experimental and theoretical investigations indicated that the recovery of CH4 from CH4 hydrate is possible, the
mechanism for the replacement, especially kinetic properties of the process, is not well understood. In this study, microsecond
molecular dynamics simulations show that the replacement process is found to be controlled cooperatively by the chemical potentials
of guest molecules, memory effect, and mass transfer barrier.
This journal is The Royal Society of Chemistry 2012

Energy Environ. Sci., 2012, 5, 70337041 | 7033

Published on 29 February 2012. Downloaded by University of Hong Kong Libraries on 19/10/2014 07:25:55.

View Article Online

CH4 hydrate could be converted into CO2 hydrate, large


amounts of solid CH4 hydrate on continental margins and in
permafrost regions would serve a dual duty, i.e. as CH4 sources
and CO2 storage sites. Smith et al.11 studied the possibilities of
the replacement process, and Ohgaki et al.8 illustrated that the
equilibrium pressure of CO2 hydrate is lower than that of CH4
hydrate at the same temperature. Furthermore, it is also
demonstrated that the formation heat of CO2 hydrate can meet
the heat demand to decompose CH4 hydrate,8 which resolves the
problem of lacking a heat source to exploit natural gas hydrate.
Recent experimental studies indicated that the recovery of CH4
from CH4 hydrate is possible through its reaction with CO2.1215
Thermodynamic calculations also support the replacement
scheme in hydrates at appropriate conditions.16
Note that molecular simulation has become an important
method in understanding the microscopic behavior of hydrates.
Molecular dynamics (MD) simulations have been performed to
investigate the properties of gas hydrate, such as their dependence on potential models,1719 determination of various hydrate
properties,1924 formation and melting,2534 and nucleation of gas
hydrate.3539 Recently, Geng et al.40 in their MD simulations
demonstrated that the CH4/CO2 mixed hydrates are more stable
than CH4 hydrates or CO2 hydrates alone.
Although the above investigations provide important insights
to the hydrate replacement process, the mechanism for the
replacement, especially kinetic properties of the process are not
well understood. Thus, the purpose of this work is to study the
kinetic properties of the replacement of CH4 in the hydrate form
by CO2 molecules in molecular scale. With computer simulations, we also attempt to address the questions how the CO2
hydrate forms and CH4 hydrate melts during the replacement
process.
In this paper, we performed a series of microsecond molecular
dynamics (MD) simulations to investigate the mechanism of the
replacement process. We found that the memory effect (in this
case, it is hydrate residual rings, rather than residual cage-like
structures suggested by Sloan et al.41), the chemical potentials of
guest molecules, and the barrier of mass transfer play important
roles in the replacement process.

2. Model and simulation method


In our work, MD simulations were performed using
LAMMPS,42 an open source program for massively parallel
simulations. To obtain the initial configuration, we firstly set
a perfect crystal structure of sI CH4 hydrate into a simulation
box with the size of 3.56 nm  3.56 nm  11.88 nm (3  3  10
unit cells). The positions of the oxygen atom in H2O molecules
and carbon in CH4 were obtained from the data of the space
group of the crystal structure.43 As is shown in Fig. 1, we divided
the simulation box along the z-direction into several regions
according to the crystal cell length, which are denoted by regions
A, B, C and D. Then, the crystal units from the 4th to 7th unit
cells along the z-direction (region D in Fig. 1a) were replaced
randomly with the same amount of H2O and CH4 molecules.
In our simulations, the TIP4P water model44 was used and the
rigidity of the model molecules was implemented with SHAKE
algorithm,45 while CO2 molecules were represented by EPM2
model.46 The set of potential parameters were used in our earlier
7034 | Energy Environ. Sci., 2012, 5, 70337041

Fig. 1 (a) The initial configuration (top) and (b) the final configuration
(bottom) for a typical simulation run. To simplify the description, the
system was divided into several regions along the z-direction, including
regions A, B, C, and D. In this figure, magenta spheres represent CH4
molecules, and green spheres represent CO2 molecules. H2O molecules
are shown in the stick model, and dashed lines represent hydrogen bonds.
At the initial state, the system contains totally 4140 H2O, 455 CH4 and 94
CO2 molecules; while in region D, the number of each component is 1656
H2O, 23 CH4 and 94 CO2 molecules.

study to investigate the nucleation of hydrate.47 A single point


model48 was used in this work to model CH4 molecules. The
unlike parameters of Lennard-Jones potentials were obtained
by the Lorentz-Berthelot mixing rule, except the parameters
for H2OCO2 interaction taken from ref. 49. A cutoff radius of
 was utilized for the short-ranged interactions, while the
12.0 A
long-ranged interactions were evaluated by using the pppm
algorithm.50 Periodic boundary conditions were imposed in all
the three Cartesian directions.
A relaxation process was performed firstly at 270 K and 50
bar. During the relaxation process, we adjusted the chemical
potential of CH4 by adding/deleting CH4 molecules in region D
to ensure that the interfaces between the hydrate crystal (region
C) and liquid phase (region D) did not undergo any melting in
5 ns. In region D, the final configuration contained 1656 H2O
molecules and 117 CH4 molecules. Then, some, e.g. 80%, of the
CH4 molecules in region D were replaced by the same number of
CO2 molecules (Fig. 1a). This is the initial configuration for
further constant temperature and constant pressure (NpT) MD
simulations, in which the perturbation on the chemical potentials
of guest molecules would cause the melting of CH4 hydrate and
the subsequent formation of the CO2 hydrate. The simulation
runs in the microsecond time scale were performed at 270 K and
20 bar by using the Nose-Hoover algorithm51 and with a time
step of 2 fs. The choice of the thermodynamic conditions is based
on our simulation results that with the interaction models we
used, the temperature is above the melting point of CH4 hydrate
but below that of CO2 hydrate at 20 bar.

3. Results and discussions


To investigate the replacement mechanism of CH4 hydrate by
CO2 molecules, we performed MD simulations to model this
process. Our computer simulations reveal that, as the melting
This journal is The Royal Society of Chemistry 2012

View Article Online

process of CH4 hydrate proceeds, the CO2 hydrate nucleates at


the interface between the CH4 hydrate and liquid phase, and as
a result, the in situ replacement of CH4 molecules in the hydrate
form by CO2 molecules takes place (Fig. 1b).

Published on 29 February 2012. Downloaded by University of Hong Kong Libraries on 19/10/2014 07:25:55.

3.1 Melting of CH4 hydrate


To give an example of the replacement process, we set the initial
CO2 concentration to 80% on a water free basis, in which 80% of
the CH4 molecules in the liquid phase (region D) were chosen
randomly and replaced with the same number of CO2 molecules
at the beginning of the simulations. Our simulations indicated
that CH4 hydrate in region C would gradually melt due to the
difference of the chemical potential of CH4 between the hydrate
and the liquid phase.
Fig. 2 shows the evolution of hydration number of CH4
molecules in different regions. The hydration number is defined
as the average number of H2O molecules around one guest
 In a perfect sI crystal of
molecule within a distance of 4.33 A.
CH4 hydrate the average hydration number of CH4 in a unit cell
is approximately equal to 23 (six large cages with a hydration
number of 24 and two small cages with 20). From Fig. 2, we can
find clearly that both the hydration numbers for regions C and B
decrease with time, which indicates the melting of CH4 hydrate in
these regions.
Interestingly, a two-step process is found for the melting of
CH4 hydrate cages in region C (Fig. 2) at a very initial stage of
the cage broken process. Firstly, from the beginning of the
simulation to 0.1 ms, the hydration number of CH4 molecules
decreases very quickly. Then, after a short relaxation period from
0.1 ms to 0.4 ms, the hydration number decreases again until
the whole region is totally melted at 1.0 ms. Within the first step
of the melting process (0.1 ms), Fig. 3a demonstrates that most
of the broken cages are large 51262 cages, while small (512) cages
are maintained within a significantly longer time (0.4 ms). This
is because large cages contain some unfavorable angles of H2O
molecules for the planar hexagonal rings.1,35,52
Note that this observation does not conflict with the conclusion from MD simulations53 that the uniform melting of CO2

Fig. 2 The evolution of hydration number of CH4 molecules in different


processes, for which initial CO2 concentrations are set to 50%, 80% and
100%, respectively, on a water free basis. Note that the average hydration
number for a CH4 molecule in perfect sI hydrate crystal is approximately
equal to 23.

This journal is The Royal Society of Chemistry 2012

Fig. 3 (a) The number of cages in region C as a function of time. The


symbols T and D used in the figure represent 51262 cages and 512
cages, respectively. (b) The number of rings in region C as a function of
time.

hydrate across the hydrate interface is a collective phenomenon.


The algorithm we used allows us to consider a cage as destroyed
(and hence disappeared) even though only one of its edges is
broken. Therefore, we concentrated on the early stage of
a melting process, at which cages begin to break. There should be
plenty of partial cages or residual rings left near the melting
interface although Fig. 2 and Fig. 3 show that the cages were
broken.
Fig. 3b shows the number of rings in region C as a function of
time. The figure indicates that a large number of rings remain in
the freshly formed liquid water in region C although most of the
hydrate cages are broken (Fig. 3a). The residual water structures
after the melting of CH4 hydrate provide a possible rationale for
the common observation of a so-called memory effect.54 In
addition, we note that the structure of the six-membered rings
surviving in region C (with root mean square (RMS) deviation
 is not planar, but adopt a chair conformation as that
0.998 A)
 For
of water in ice Ih lattice (RMS deviation 1.277 A).
comparison, the RMS deviation for the planar hexagonal rings in

large 51262 cages is 0.225 A.
The memory effect has been studied for several years. Some
authors explained the observations of the effect by a hypothesis
that hydrate structure (residual structure41,5456 or persistent
Energy Environ. Sci., 2012, 5, 70337041 | 7035

Published on 29 February 2012. Downloaded by University of Hong Kong Libraries on 19/10/2014 07:25:55.

View Article Online

hydrate crystallites57) remains in a solution after a dissociation


process. Rodger et al.27,58 investigated the properties of the
interface between CH4 gas phase and sI hydrate crystals, and
they ascribed the existence of memory effect to the fact that
some dissolved CH4 molecules remain in the solution after the
hydrate has decomposed. The recent study by Vatamanu and
Kusalik59 suggested that the memory effect can be explained in
terms of the increased possibility of observing regions with high
local concentration of guest molecules. However, the artificial
temperature pulse applied in their work destroyed the residual
structures.
In our simulations, we try to explain the memory effect in
term of residual ring and semi-cage structures left after the
melting of hydrate (Fig. 4b and 4c). As indicated in the following
section, the memory effect plays an important role in the
replacement process: the hydrate residual structures in the liquid
water phase facilitate the nucleation of CO2 hydrate, and thus the
replacement process is significantly accelerated.
3.2 Replacement of CH4 hydrate by CO2 molecules
Our simulations in this work also confirm the in situ replacement
process of gas hydrate: the melting of CH4 hydrate in region C is
followed by the formation of amorphous CO2 clathrate within
the region (Fig. 4). Moreover, hydrate residual rings after
melting are found to facilitate the nucleation of CO2 hydrate, and
are responsible for the appearance of the memory effect.
To investigate the memory effect, we developed a topological algorithm (see appendix A for details) to identify which rings
forming a large CO2 cage are original ones (the residual rings
used to form CO2 hydrate cages) and which rings are newly
formed ones (those freshly formed by independent water molecules). Fig. 5 shows the number of original and total rings in
a large CO2 cage of region C during the replacement process with
an initial CO2 concentration of 80% in region D (on a water free
basis). The figure shows that the total number of pentagonal and

Fig. 4 Several typical snapshots during a replacement process with an


initial CO2 concentration of 80% on a water free basis. The amorphous
cages are denoted using a wire frame model with different colors: blue for
small cages and red for large cages. The color code for different molecules
is the same as described in Fig. 1. Along the direction denoted by the
arrows, the snapshots correspond to (a) 0 ms, (b) 0.1 ms, (c) 0.5 ms,
(d) 0.9 ms, (e) 1.2 ms and (f) 1.8 ms, respectively. For clarity, only the
part of the system around region C is shown.

7036 | Energy Environ. Sci., 2012, 5, 70337041

Fig. 5 Different origins of the rings forming large CO2 cages in region C
(columns) and the distribution of survival time (life time) for hydrate
residual rings (line). In the figure, original rings represent the number
of the residue rings used to form CO2 large cages. total rings represent
the total number of rings per CO2 large cage, which is equal to14.0 for an
ideal large cage. The distribution of life time for the hydrate residual rings
is in an arbitrary unit (a. u.).

hexagonal rings per large cage is about 14.0, being consistent


with that for an ideal large cage. More importantly, for newly
formed cages, 47% of total rings, i.e., 6.6 original rings in the
total 14.0 rings, are found to be the residual water rings (Fig. 5).
The observation indicates that these persistent rings nucleate
a rapid formation of CO2 hydrate, and consequently, the
memory effect accelerates the formation of CO2 cages.
The nucleation of CO2 hydrate begins with the formation of
several empty small cages and large cages in region C (Fig. 3a and
4b), similarly to the nucleation pathway for the formation of gas
hydrate suggested by Jacobson et al.60 The small cages are rarely
occupied by CO2 molecules (Fig. 3a) because the size of a CO2
molecule is slightly larger than the effective diameter of small
cages.1
Even though some of CO2 molecules can occupy small cages,
they may cause significant distortion of the filled cages. To
characterize the degree of structural distortion and hence to
evaluate the stability of the amorphous CO2 clathrate, we
calculated the distortion factor, fnorm, which is defined as the sum
of displacement of all oxygen atoms in a distorted cage relative to
its perfect structure (see appendix B for details). The calculated
 for small
 for small cage filled with CO2, 0.18 A
values are 0.96 A
 for empty small cage, respeccage filled with CH4, and 0.24 A
tively. As is expected, the distortion for small cages filled with
CO2 molecules is more pronounced than that for the cages filled
with CH4 molecules and than that for empty cages. The fact
indicates that the small cages tend to be filled by CH4 molecules
or keep empty rather than filled by CO2 molecules. More
evidences can be seen in Fig. 3a, which demonstrates the existence of a large number of empty small cages in region C in the
final configuration. We also calculated the average occupancies
of large cages and small cages filled with guest molecules in the
final configuration. The values are 88.2% for large cages and
18.7% for small cages, respectively. Note that the values are
This journal is The Royal Society of Chemistry 2012

Published on 29 February 2012. Downloaded by University of Hong Kong Libraries on 19/10/2014 07:25:55.

View Article Online

roughly estimated, because there are incomplete cages and cages


with catastrophic distortion.
As the replacement process proceeds, the number of the newly
formed cages increases until the complete formation of an
amorphous CO2 hydrate layer (Fig. 3a and 4b). After the
formation of amorphous CO2 hydrate layer in region C (Fig. 4d),
it inhibits the diffusion of CH4 molecules from the interface
(region B) to the bulk region (region D), slowing down the
melting of CH4 hydrates. In addition, it also blocks the diffusion
of CO2 molecules from the bulk region to replacement region
(region B). As a result, a mixed liquid layer, containing liquid
water with residual structures and guest molecules, is found in
region B (see Fig. 1b and 4f). Because of the difficulty of mass
transfer of the guest molecules, the melting process of CH4
hydrate in region B is terminated after 1.0 ms (Fig. 2) even
though a lot of residue structures remains in the region, and
accordingly, the replacement process also slows down and finally
stops (Fig. 6).
After the replacement process in region C completed at 1.0 ms
(Figs. 2, 4d and 6), the amorphous clathrate begins to grow into
region D and the gas molecules congregate into a bubble in the
middle of the simulation box (Fig. 1b). In this case, the CO2
hydrate formed in region C can act as nuclei and causes the growth
of CO2 hydrate into region D. The growth of CO2 hydrate in
region D differs from the nucleation of CO2 hydrate in region C:
with hydrate residual structures for the nucleation process and
without hydrate residual structures for the growth process.
3.3 Effects of initial CO2 concentration
To study the effects of the initial CO2 concentration on the
replacement process, different numbers of CH4 molecules in
region D were replaced by CO2 molecules at the beginning of the
simulations. As a result, different initial CO2 concentrations in
region D were obtained. For region D, Table 1 summarized
initial compositions of different simulations. Our simulations
show that the initial CO2 concentration, namely the chemical
potential of CO2, plays an important role in the replacement
process.

Fig. 6 The evolution of hydration number for CO2 molecules at


different initial CO2 concentrations. Note that the average hydration
number for CO2 molecules in perfect sI hydrate crystal is approximately
equal to 24.

This journal is The Royal Society of Chemistry 2012

Table 1 The number of guest molecules and their concentration in the


initial configuration of region D for different simulations
Number of CO2a

Number of CH4

CO2%b

CH4%b

0 (0%)
12 (10.3%)
23 (19.7%)
35 (29.9%)
47 (40.2%)
58 (49.6%)
70 (59.8%)
82 (70.1%)
94 (80.3%)
105 (89.7%)
117 (100%)

117
105
94
82
70
59
47
35
23
12
0

0%
0.677%
1.297%
1.974%
2.651%
3.271%
3.948%
4.625%
5.302%
5.922%
6.599%

6.599%
5.922%
5.302%
4.625%
3.948%
3.328%
2.651%
1.974%
1.297%
0.677%
0%

a
The values in parentheses represent the concentrations of guest
molecules in region D on a water free basis, i.e. n(CO2)/[n(CO2) +
n(CH4)]. b The values represent the molar fractions of guest molecules
in region D, i.e. n(guest)/[n(H2O) + n(CO2) + n(CH4)]. Note that the
molar fraction of CH4 in perfect sI hydrate is 14.8% (both large and
small cages are filled) and that of CO2 in perfect sI hydrate is 11.5%
(only large cages are filled).

In this work, the amount of CH4 molecules recovered, which is


defined as the difference of free CH4 molecules between the final
and the initial configurations, and the amount of CO2 captured
in region C were used to characterize the degree of substitution of
the replacement process. The effects of initial CO2 concentration
on the recovery amount of CH4 and the storage amount of CO2
molecules in region C are shown in Fig. 7a and 7b, respectively.
Fig. 7a shows that the net amount of CH4 molecules recovered
increases with the initial concentration of CO2 molecules in
region D. Similarly, the storage amount of CO2 molecules in
region C (the main replacement region) also increases with the
initial CO2 concentration (Fig. 7b).
At a low initial concentration of CO2 (<40% on a water free
basis), both the melting of CH4 hydrate and the formation of
CO2 hydrate take place very slowly, leading to a low recovery
amount of CH4 molecules and a low storage amount of CO2
molecules in region C. With the increase of the concentration
from 40% to 80%, the number of residual rings used to form large
CO2 cages increase slowly (Fig. 5). This observation demonstrates a stronger memory effect at a higher initial CO2
concentration (Fig. 5), and as a result, both the amount for CH4
recovered and that for CO2 captured increase slowly with the
initial concentration of CO2 molecules (Fig. 7).
In contrast, at a very high CO2 concentration (>80% on
a water free basis), the faster melting process of CH4 hydrate
(Fig. 2) leads to a more disordered liquid phase. Furthermore,
the local concentration of CH4 around residual rings may
become lower since the average concentration of CH4 in liquid
decreases, which then weaken the memory effect. The weakening of memory effect can be proved by the much lower
percentage of residual rings in the newly formed CO2 cages
(Fig. 5). In this case, it is the high CO2 concentration that
causes the shortening of survival time (life time) of the residual
rings (Fig. 5), and consequently decreases the number of residue
rings in the formed CO2 cages. However, a much higher initial
CO2 concentration not only melts the CH4 hydrate more
thoroughly (Fig. 2), but also provides a strong driving force for
the formation of CO2 hydrate. Therefore, both the amount for
Energy Environ. Sci., 2012, 5, 70337041 | 7037

Published on 29 February 2012. Downloaded by University of Hong Kong Libraries on 19/10/2014 07:25:55.

View Article Online

Fig. 8 Snapshots during a simulation process with an initial CO2


concentration of 100% on a water free basis. The amorphous cages are
denoted using a wire frame model with different colors: blue for small
cages and red for large cages. The color code for different molecules is the
same as described in Fig. 1. Along the direction denoted by arrows, the
snapshots correspond to (a) 0 ms, (b) 0.1 ms, (c) 0.5 ms, (d) 0.8 ms,
(e) 1.3 ms and (f) 1.8 ms, respectively. For clarity, only part of the
system is shown.

Fig. 7 (a) The amount of CH4 molecules recovered as a function of


initial CO2 concentrations in region D (on a water free basis). In the
figure, the red columns represent the amount of CH4 molecules released
from hydrate crystal to liquid water; the blue columns represent the
amount of CH4 molecules from liquid water to newly formed amorphous
hydrate (actually it is a negative recovery process); the black line represents the net amount of CH4 molecules recovered. The error bars in the
figure represent the maximum and minimum values of CH4 molecules
recovered from the five independent simulation runs. (b) The amount of
CO2 captured in region C as a function of initial CO2 concentration in
region D.

CH4 recovered and that for CO2 captured increase rapidly


(Fig. 7).
Fig. 8 also gives a series of typical snapshots around region C
for the replacement process with an initial CO2 concentration of
100% (on a water free basis). In comparison with Fig. 4, the
pathway of replacement process does not change: the melting of
CH4 hydrate in region C is followed by the formation of amorphous CO2 clathrate within the region.
In general, our simulations demonstrate that the initial CO2
concentration, namely the chemical potential of CO2, plays an
important role in controlling the replacement process. As the
initial concentration of CO2 molecules increases, both the net
amount of CH4 molecules recovered and the amount of CO2
molecules captured in region C increase.
3.4 Mass transfer in the replacement process
So far, we discussed how the memory effect of the melted
hydrate and the concentration of guest molecules affect the
7038 | Energy Environ. Sci., 2012, 5, 70337041

process of replacement of CH4 in hydrate form with CO2. In


what follows, we turn to the influence of mass transfer on the
dynamic properties of the replacement process.
At the beginning of our simulations, the large density gradient
of CH4 molecules at the interface (Fig. 9a) provides a driving
force to melt the hydrate. The slope of the curve also indicates no
significant barrier for the mass transfer of methane molecules
from the melting region to region D. After 0.7 ms, however, the
slope of the curve decreases (Fig. 9a), which can be explained as
the gradual formation of the amorphous hydrate layer in region
C (Fig. 4). After the complete formation of the amorphous
hydrate layer at 1.0 ms, the molecule exchange between region B
and region D stops even though there still exists a significant
density gradient (Fig. 9a). The accumulation of CH4 in region B
reduces the difference of the chemical potentials of CH4 in region
A and the thin liquid layer in region B. As a result, the melting of
region B slows down and even stops at 1.0 ms (Fig. 2).
It is found from Fig. 9b that the same problem exists for CO2
molecules. The decrease of the slope shows that the difficulty of
mass transfer becomes evident with the increase of simulation
time, i.e. the increase of the thickness of the amorphous hydrate
layer. In general, an amorphous hydrate layer gradually forms as
the replacement process proceeds, and the hydrate layer provides
a significant mass transfer barrier for the exchange of guest
molecules. The difficulty of mass transfer slows down the melting
of CH4 hydrate and thus the hydrate replacement.
Hence, the replacement process is likely to lead to the substitution near the surface of CH4 hydrate. However, as the
replacement process proceeds, an amorphous layer of CO2
hydrate forms. The amorphous hydrate layer at the hydrate
surface apparently inhibits the molecule exchange between CO2
source and CH4 released from the melting hydrates, thus slowing
down the melting process. On the other hand, in order to reach
the interior sites of the hydrate sample, CO2 molecules have to
diffuse through the amorphous hydrate layer, and as a result, the
replacement process would become very inefficient, as showed in
experimental work.13 It seems that the sample morphology of the
This journal is The Royal Society of Chemistry 2012

Published on 29 February 2012. Downloaded by University of Hong Kong Libraries on 19/10/2014 07:25:55.

View Article Online

Fig. 9 The density evolution for (a) CH4 and (b) CO2 molecules in
different regions (in the case with an initial CO2 concentration of 80% on
a water free basis).

initial CH4 hydrate plays an important role in the replacement


process, for example, a large specific surface area would increase
the replacement rate. In general, our results reveal that the
formation of amorphous CO2 layer slows down the replacement
rate, which is quite similar to the mechanism of the self-preservation effect of methane hydrate.
We must note that in our simulations the temperature is
maintained uniformly across the simulation box, while in real
situations the formation of CO2 hydrate will provide some
energy for further melting the vicinal methane hydrate. However,
the resistance of mass transfer still exists as the hydrate layer
forms. Therefore, the CO2 hydrate layer inevitably leads to
a barrier for transport of methane to the liquid phase and CO2 to
the hydrate phase.

4. Conclusions
In this work, microsecond molecular dynamics simulations were
performed to study the mechanism for the replacement of CH4
molecules in the hydrate form with CO2 molecules. Our simulations reveal that the melting of CH4 hydrate shows a two-step
manner, in which large 51262 cages melt firstly and then small 512
cages melt. The melted CH4 hydrate brings a large number of
This journal is The Royal Society of Chemistry 2012

hydrate residual rings (both pentagonal and hexagonal rings),


which are responsible for the memory effect.54 The hydrate
residual rings facilitate the nucleation of CO2 hydrate and are
used for the subsequent formation of the amorphous layer of
CO2 hydrate.
Furthermore, the chemical potential of guest molecules can
also affect the replacement process. We chose the initial CO2
concentration in the liquid water as a controllable initial
parameter to study how the chemical potential of guest molecules
affects the process. Our simulations indicate that, as the initial
concentration of CO2 molecules increases, both the net amount
of CH4 molecules recovered and the amount of CO2 molecules
captured in region C increase.
From a dynamic aspect, the replacement process would induce
the substitution of CO2 amorphous hydrate near the surface of
CH4 hydrate rather easily. As the replacement process proceeds,
an amorphous layer of CO2 hydrate forms on the surface of the
CH4 hydrate. The amorphous layer provides a significant mass
transfer barrier to the further melting of CH4 hydrate and the
formation of CO2 hydrate. The replacement process therefore
becomes very difficult. Hence, the formation of amorphous CO2
layer slows down the replacement rate and prevents the CH4
hydrate from further dissociation.
In summary, the recovery of CH4 and the storage of CO2 are
controlled cooperatively by the chemical potentials of guest
molecules, memory effect, and mass transfer. Especially, our
work reveals that it is possible to replace methane in hydrate lattice
with carbon dioxide; however, measures should be taken to eliminate the mass transfer barrier resulted from the formation of
amorphous layer of CO2 hydrate on the surface of the CH4 hydrate
during the replacement process and therefore increase the replacing efficiency. Our results imply that formation of porous CO2
hydrate would be helpful to realize this object. In our next work, we
will try to simulate the replacement process in the presence of
surfactants determining whether porous CO2 hydrate could form.

Appendix A
In the present work, a topological algorithm was developed to
identify the structure evolutions during a replacement process,
especially whether a new formed cage contains hydrate residual
rings. The algorithm is based on the ring perception algorithm
developed by Matsumoto et al.61 and the cage identification
algorithm developed by Jacobson et al.52
Since the connectivity index for one H2O molecule is equal to
four in a perfect hydrate crystal (both sI and sII), H2O molecules
can be represented by a four-dimensional linking vector
0 1
p
BqC
~
B
bi @ C
, which indicates that the ith H2O molecule connects
rA
s
with the pth, qth, rth and sth H2O molecules. If the connectivity
index is less than four (e.g. three), the last element in the linking
vector is set to zero. In our algorithm, isolated H2O molecules
0 1
0 1
0
p
B0C
B0C
~
~
B
B
C
with bi @ A and isolated H2O molecule pairs with bi @ C
0
0A
0
0
were not considered, because they could not form a ring.
Energy Environ. Sci., 2012, 5, 70337041 | 7039

Published on 29 February 2012. Downloaded by University of Hong Kong Libraries on 19/10/2014 07:25:55.

View Article Online

With all the vectors, a linking table representing the hydrogen


bond networks can be obtained.
If a pair of connected molecules is separated during a simulation process, the vector of them will be changed, e.g., setting the
value of the last non-zero element to zero. When one of the
hydrogen bonds in a ring is broken, all the molecules in the ring
will change their vectors. To improve the efficiency of the algorithm, a molecule will be excluded from the linking table if the
connectivity index for the molecule is less than two.
To investigate whether a newly formed cage contains hydrate
residual rings, we used the ring perception algorithm61 and the
cage identification algorithm52 firstly to find all of the rings
within a cage, and then searched the linking vector of each H2O
molecule within a ring. If the ring contains at least one H2O
molecule whose linking vector has three or four zero elements (or
at least one H2O molecule not belonging to the linking table), the
ring must be a newly formed ring. Otherwise, it must be a hydrate
residual ring.

Appendix B
In order to evaluate the distortion of cages, we defined distortion
P
ri  ~
r0 jj, where ~
factor, f, as f min j~
r(i) and ~
r0(i) represent
j J

the coordinates of the ith oxygen atoms in a newly formed hydrate


cage and those in a geometrically perfect cage, respectively. The
origin of the coordinates is located at the center of mass for the
cage. The summation is done over all the oxygen atoms in one
cage relative to a perfect cage. The set J includes all possible maps
of the oxygen atoms between the newly formed cage and the
perfect cage, but only the minimum sum of the displacements is
adopted. Thus, the value of f can be used to characterize the
degree of distortion of the cage.
For comparison of deformation, a normalized distortion
f
factor is defined as fnorm , where N is the total number of
N
oxygen atoms in one cage with N 24 for large cages and N 20
for small cages.

Acknowledgements
This work is supported by National Natural Science Foundation
of China (Nos. 20736005, 20876004 and 20925623). Generous
allocations of computing time by Chemical Grid Project of
BUCT are acknowledged.

References
1 E. D. Sloan and C. A. Koh, Clathrate Hydrates of Natural Gases, 3rd
ed.; CRC Press: Boca Raton, FL, 2008.
2 B. A. Buffett, Annu. Rev. Earth Planet. Sci., 2000, 28, 477.
3 F. Ning, Y. Yu, S. Kjelstrup, T. J. H. Vlugt and K. Glavatskiy, Energy
Environ. Sci., 2012, DOI: 10.1039/C2EE03435B.
4 R. Boswell and T. S. Collett, Energy Environ. Sci., 2011, 4, 1206.
5 K. A. Kvenvolden, Chem. Geol., 1988, 71, 41.
6 Y. F. Makogon, Natural gas hydrates: the state of study in the USSR
and perspectives for its use, 3rd chemical congress of North America:
Toronto, Canada, 1988.
7 P. G. Brewer, Ann. N. Y. Acad. Sci., 2000, 912, 195.
8 K. Ohgaki, K. Takano, H. Sangawa, T. Matsubara and S. Nakano, J.
Chem. Eng. Jpn., 1996, 29, 478.
9 R. P. Warzinski and G. D. Holder, Energy Fuels, 1998, 12, 189.
10 Y. T. Seo and H. Lee, J. Chem. Eng. Data, 2001, 46, 381.

7040 | Energy Environ. Sci., 2012, 5, 70337041

11 D. H. Smith, K. Seshadri and J. W. Wilder, The 1st National


Conference on Carbon Sequestration, National Energy Technology
Laboratory, US Department of Energy, 2001, p. 1.
12 C. Moon, R. W. Hawtin and P. M. Rodger, Faraday Discuss., 2007,
136, 367.
13 H. Lee, Y. Seo, Y.-T. Se, I. L. Moudrakovski and J. A. Ripmeester,
Angew. Chem., Int. Ed., 2003, 42, 5048.
14 Y. Park, D.-Y. Kim, J. W. Lee, D.-G. Huh, K.-P. Park, J. Lee and
H. Lee, Proc. Natl. Acad. Sci. U. S. A., 2006, 103, 12690.
15 T. Uchida, S. Takeya and T. Ebinuma, The 5th International
Conference on Greenhouse Gas Control Technology, Collingword,
2001, p. 523.
16 E. M. Yezdimer, P. T. Cummings and A. A. Chialvo, J. Phys. Chem.
A, 2002, 106, 7982.
17 N. J. English and J. M. D. Macelroy, J. Comput. Chem., 2003, 24,
1569.
18 A. A. Chialvo, H. Mohammed and P. T. Cummings, J. Phys. Chem.
B, 2002, 106, 442.
19 H. Jiang, K. D. Jordan and C. E. Taylor, J. Phys. Chem. B, 2007, 111,
6486.
20 S. Alavi, J. A. Ripmeester and D. D. Klug, J. Chem. Phys., 2006, 124,
014704.
21 J. S. Tse, M. Klein and I. R. McDonald, J. Chem. Phys., 1984, 81,
6146.
22 V. V. Sizov and E. M. Piotroskaya, J. Phys. Chem. B, 2007, 111,
2886.
23 E. J. Rosenbaum, N. J. English, J. K. Johnson, D. W. Shaw and
R. P. Warzinski, J. Phys. Chem. B, 2007, 111, 13194.
24 H. Jiang and K. D. Jordan, J. Phys. Chem. C, 2010, 114, 5555.
25 J. A. Ripmeester, C. I. Ratcliffe, D. D. Klug and J. S. Tse, Ann. N. Y.
Acad. Sci., 1994, 715, 161.
26 O. K. Forrisdahl, B. Kvamme and A. D. J. Haymet, Mol. Phys., 1996,
89, 819.
27 P. M. Rodger, T. R. Forester and W. Smith, Fluid Phase Equilib.,
1996, 116, 326.
28 N. J. English, J. K. Johnson and C. E. Tylor, J. Chem. Phys., 2005,
123, 244503.
29 H. Nada, J. Phys. Chem. B, 2006, 110, 16526.
30 J. Vatamanu and P. G. Kusalik, J. Phys. Chem. B, 2006, 110, 15896.
31 J. Zhang, R. W. Hawtin, Y. Yang, E. Nakagawa, M. Rivero,
S. K. Choi and P. M. Rodger, J. Phys. Chem. B, 2008, 112, 10608.
32 S. Alavi, J. A. Ripmeester and D. D. Klug, J. Chem. Phys., 2007, 126,
124708.
33 J. Vatamanu and P. G. Kusalik, J. Phys. Chem. B, 2008, 112, 2399.
34 E. Myshakin, H. Jiang, R. Warzinski and K. D. Jordan, J. Phys.
Chem. A, 2009, 113, 1913.
35 M. R. Walsh, C. A. Koh, E. D. Sloan, A. K. Sum and D. T. Wu,
Science, 2009, 326, 1095.
36 C. Moon, P. C. Taylor and P. M. Rodger, J. Am. Chem. Soc., 2003,
125, 4706.
37 G. J. Guo, Y. G. Zhang and Y. J. Zhao, J. Chem. Phys., 2004, 121,
1542.
38 G. J. Guo, Y. G. Zhang and H. Liu, J. Phys. Chem. C, 2007, 111,
2595.
39 R. Radhakrishnan and B. L. Trout, J. Chem. Phys., 2002, 177, 1786.
40 C. Y. Geng, H. Wen and H. Zhou, J. Phys. Chem. A, 2009, 113,
5463.
41 E. D. Sloan, S. Subramanian, P. N. Matthews, J. P. Lederhos and
A. A. Khokhar, Ind. Eng. Chem. Res., 1998, 37, 3124.
42 S. J. Plimpton, J. Comput. Phys., 1995, 117, 1.
43 M. T. Kirchner, R. Boese, W. E. Billups and L. R. Norman, J. Am.
Chem. Soc., 2004, 126, 9407.
44 W. L. Jorgensen, J. Chandrasekhar, J. D. Madura, R. W. Impey and
M. L. Klein, J. Chem. Phys., 1983, 79, 926.
45 J. P. Ryckaert, G. Ciccotti, H. J. Berendsen and C, J. Comput. Phys.,
1977, 23, 327.
46 J. G. Harris and K. H. Yung, J. Phys. Chem., 1995, 99, 12021.
47 D. Bai, G. Chen, X. Zhang and W. Wang, Langmuir, 2011, 27, 5961.
48 W. L. Jorgensen, J. D. Madura and C. J. Swenson, J. Am. Chem. Soc.,
1984, 106, 6638.
49 Z. H. Duan and Z. G. Zhang, Geochim. Cosmochim. Acta, 2006, 70,
2311.
50 R. W. Hockney and J. W. Eastwood, Computer Simulation Using
Particles, Adam Hilger, NY, 1989.
51 W. G. Hoover, Phys. Rev. A: At., Mol., Opt. Phys., 1985, 31, 1695.

This journal is The Royal Society of Chemistry 2012

Published on 29 February 2012. Downloaded by University of Hong Kong Libraries on 19/10/2014 07:25:55.

View Article Online

52 L. C. Jacobson, W. Hujo and V. Molinero, J. Phys. Chem. B, 2009,


113, 10298.
53 S. Sarupria and P. G. Debenedetti, J. Phys. Chem. A, 2011, 115, 6102.
54 Y. F. Makogon, Hydrates of Natural Gas, Moscow, Nedra,
Izadatelstro, 1974.
55 J. P. Lederhos, J. P. Long, A. Sum, R. L. Christiansen and
E. D. Sloan, Chem. Eng. Sci., 1996, 51, 1221.
56 R. Ohmura, M. Ogawa, K. Yasuka and Y. J. Mori, J. Phys. Chem. B,
2003, 107, 5289.

This journal is The Royal Society of Chemistry 2012

57 P. Buchanan, A. K. Soper, R. E. Westacott, J. L. Creek and


C. A. Koh, J. Chem. Phys., 2005, 123, 164507.
58 P. M. Rodger, Ann. N. Y. Acad. Sci., 2000, 912, 474.
59 J. Vatamanu and P. G. Kusalik, Phys. Chem. Chem. Phys., 2010, 12,
15065.
60 L. C. Jacobson, W. Hlujo and V. Molinero, J. Phys. Chem. B, 2010,
114, 13796.
61 M. Matsumoto, A. Baba and I. Ohmine, J. Chem. Phys., 2007, 127,
134504.

Energy Environ. Sci., 2012, 5, 70337041 | 7041

You might also like