Injection of CO2/N2 gaseous mixtures into gas hydrates to contemporary perform CH4 recovery and CO2 storage

Carbon dioxide injection into natural gas hydrate reservoirs represents a promising opportunity to predispose a theoretically carbon neutral energy source. This technique allows to replace methane molecules with an equal number of carbon dioxide molecules and, consequently, to balance in advance emissions associated to methane utilization. While the direct CH4/CO2 replacement has been widely investigated, more data and scientific evidences are required to well define the feasibility of recovering methane by replacing it with CO2-based gaseous mixtures. In this sense, the most promising opportunity consists in flue-gas mixtures. In some cases, the presence of nitrogen was found capable to improve the overall efficiency, due to the direct competition between CH4 and N2 molecules to fill small cages characterizing hydrate structures. Moreover, these mixtures are extremely less-expensive than pure carbon dioxide. In this work, a binary CO2/N2 (50/50 vol%) gaseous mixture was used to recover methane contained into hydrate structures. Experiments were carried out in a smallscale experimental apparatus, designed to simulate a natural gas hydrate reservoir and to intervene on it with replacement techniques. Composition of gaseous mixtures present into hydrates and in the gaseous phase present immediately above, where defined via gas-chromatographic analyses. Finally, results were compared with data currently present in literature, in order to validate their consistency.


Introduction
Natural gas hydrates (NGH) are white and ice-like crystalline compounds, mainly formed from methane molecules trapped inside solid cages composed by water molecules [1]. To be stable, these compounds need of relatively high pressures and lower temperatures. For this reason, hydrates mainly occur in terrestrial permafrost and deep oceans sediments [2]. Gas hydrates consist of non-stoichiometric cage-type crystalline

Experimental apparatus
Experiments described in this work were carried out in a small-scale experimental apparatus, designed to simulate an offshore hydrate reservoir. Here a brief description of the whole apparatus is provided; however, more detailed information is available elsewhere in literature [17,18]. The core consists of a cylindrical-volume unstirred reactor, entirely made with 316SS. The top and the bottom are closed with two flanges, equipped with apposite channels used for gas injection and to host sensors. Gas can be injected from the top and from the bottom; in the present work, the injection from the bottom was selected. A digital manometer, model MAN-SD and four Type K thermocouples were used to measure respectively pressure and temperature. The reactor is immersed in a thermostatic bath, filled with water and glycol. The bath is equipped with a double copper coil, in which the refrigerant fluid is continuously circulated with a chiller, model GC-LT, to control the internal temperature. Finally, a further channel was thought for gas ejection. It allows to directly eject the whole gaseous phase present inside the reactor and is also gifted with a pressure reducer and a porous septum to take samples. Figure 1 shows a scheme of the experimental apparatus (at left) and a picture of the ejection channel, as soon described.  (4)).
The ejection channel is equipped with two valves: the first is positioned between the experimental volume and the pressure reducer; conversely, the second valve is placed above the pressure reducer and directly communicates with the external. When the only main valve is opened, the pressure reducer can be used to move part of the gaseous mixtures in a secondary little volume where, thanks to the presence of the porous septum it can be easily taken for further analyses.

Materials
The internal volume was filled with porous sand (744 cm 3 ) and pure demineralized water (236 cm 3 ). Sand consists of quartz porous spheres, having diameter equal to 500 μm; its porosity was evaluated with a porosimeter, model Thermo Scientific Pascal 140. Also space present between grains was considered, because it exercises the same role of intrapores space. The overall porosity is equal to 34%. Experiments were performed with 2 1 3 4 methane and a binary gaseous mixture containing nitrogen and carbon dioxide. Both gases are UHP (ultra-high purity degree): about methane, the guaranteed concentration is at least 99.97%, while the flue-gas concentration is 50 CO2/50 N2, with an uncertainty of ± 0.01 %.

Methods
Before each test, an initial gas stream was provided through the reactor for air removal (if present). At the same time, temperature was lowered and brought to 1 -3 °C. Two types of experiments were made: in the first, hydrates were directly formed with flue-gas and methane was not involved in the process. This experiment was made to detect thermodynamic conditions feasible for the process in presence of such mixture and also to compare experimental values with what present in literature. The second typology of experiments comprises two phases: firstly, methane hydrates were formed; then the gaseous methane was removed and replaced with flue-gas, in order to favour the exchange between these two gases in previously formed water cages.
In both cases, gas injection was slow and gradual. Once the target pressure was reached, the reactor was closed and the system started operating in batch conditions. Hydrates formation is exothermic and the formation of first nuclei, at the beginning of the process, may generate a peak in temperature (it explains why, in some tests, the initial temperature is higher than 1 -3 °C).
When hydrates were directly formed with flue-gas, the test was considered finished as soon pressure stabilized and remained constant at the current temperature. The same assumption was made for methane hydrates; however, once they were formed, the replacement process started. The ejection valve was opened and the gaseous phase was entirely removed from the reactor. As soon as it was completely ejected, the flue-gas mixture was inserted, at pressures close to 38 bar. This phase was extremely fast, less than 40 seconds were spent to complete it, in order to maintain the quantity of hydrates, dissociated during this phase, negligible. Then, the replacement phase begun. Here, any variation in pressure can be considered to evaluate the process with accuracy, because fluegas hydrates formation can occur together with methane hydrates dissociation and the contemporary occurrence of two processes, having an opposite effect on pressure, makes this latter parameter useless. By considering previous works focused on replacement strategies, the system was kept free to operate for a time period enough extended to ensure the completion of the whole process [19,20]. The effectiveness of the replacement was defined via gas-chromatographic analysis, which allowed to establish the percentage of methane recovered (as a function of the whole amount present into hydrates) and the amount of carbon dioxide captured (as a function of the initial quantity injected inside the reactor).
Despite pressure and temperature, which were directly measured, experiments were described by evaluating further relevant parameters. Moles of hydrate formed were calculated according to Equation 1: (1) Where VPORE is the volume available for hydrates formation, inside and between sand pores; "R" and "Z" describe respectively the gas constant and the compressibility factor; "ρHYD" is the ideal molar density, calculated considering 100% volume occupancy, according to Aregba [21]. Finally, subscripts "i" and "f" indicate the beginning and the ending of the specific period during which nHYD was calculated.
To correctly evaluate the quantities of gas injected in the reactor, stored into hydrates and recovered during replacement, moles of water involved into hydrates and, consequently, the free volume present inside the reactor at every step, were evaluated. To do this, the hydration number, which indicates the ratio between water and guest molecules involved into hydrates, was considered equal to 6.

Results and Discussion
The experimental section consists of three experiments: the first described CO2/N2 hydrates formation, while the others characterize the replacement of methane with the same flue-gas mixture used in the first test.
Test 1, where only flue-gas hydrates formation was testes, is described in Table 1 and in figures 2 and 3. Finally, Figure 4, shows equilibrium for CO2/N2 hydrates starting from different initial concentrations.  Table 1 shows pressure and temperature at the beginning and the ending of hydrates formation, the overall amount of flue-gas injected (nINJ), the quantity of CO2 dissolved in water (nCO2DISS), the amount of hydrates produced (nHYD), moles remained in the gaseous phase and their composition. Finally, CO2captured is the ratio between carbon dioxide stored and the overall quantity injected inside. The first term includes moles involved in hydrates formation and moles dissolved in water, while the second is the sum of the gaseous and the dissolved in water quantities. Hydrates formation was intensive and led to a drop in pressure close to 16 bar. In particular, hydrates mainly formed during the first 20*10 3 seconds; however, the pressure required more than 60*10 3 seconds to stabilize. After the initial peak, the decrease in temperature was not linear and two little variations were observed. These peaks were related to an acceleration of hydrates formation.
As expected, the process mainly involved carbon dioxide: the mixture remained above hydrates, once the process finished, contained 24.683% of CO2 and 75.317% of N2. It means that the concentration of carbon dioxide in the gaseous mixture has halved, compared to the initial concentration.   Fig. 4. Pressure-temperature diagram, describing the phase equilibrium for binary CO2/N2 mixtures, shown in function of the initial CO2 concentration, which was expressed in molar fraction (mf) [22 -31].
In this latter diagram, equilibrium conditions for CO2/N2 hydrates were reported for different initial concentrations. All curves were shown in black, except for the equilibrium line obtained for concentrations close to what used in this work. As expected, the mildest thermodynamic conditions were obtained for mixtures mainly containing carbon dioxide. However, it was observed that, also the presence of very low concentrations of carbon dioxide are able to drastically reduce pressure required for hydrates formation (if compared with pure nitrogen).
The experimental curve, showed in Figure 3, is below the red one, showed in Figure 4 and two reason can motivate such difference. Firstly, the red line also includes mixture with lower CO2 concentrations than 0.5 mf. In addition, experiments were carried out in presence of a porous medium made with pure quartz, which does not chemically affect the process, but allows to increase the surface/volume ratio and the number of gas-liquid interfaces present inside the reactor, thus improving the formation process.
Several researchers asserted that reactors used to study hydrates, because of their little size, usually affect the process and, and results may be slightly different in function of the specific reactor used to produce them [32]. As a consequence of it, this preliminary experiment was realized to characterize hydrate formation with a binary CO2/N2 (50/50 vol%) gaseous mixture in the same reactor used to investigate the replacement of methane.
The following table describes Test 2 and Test 3, where the replacement of methane with flue-gas was directly tested. The table is divided in three sections, respectively describing methane hydrates formation, the removal of methane, which was not involved in the process, and the injection of flue-gas and, finally, the replacement process. The first section was described with the initial and final pressure and temperature values, moles of methane injected, moles of hydrates formed, water involved into hydrates, volume occupied by hydrates and the remaining volume containing the gaseous phase.
The second step was characterized with pressure and temperature measured as soon as the injection was finished, moles of flue gas injected and the specific amount of nitrogen and carbon dioxide. The last phase was defined with the final pressure and temperature, measured at the end of the whole process, the variation in the overall quantity of hydrates present inside the reactor, occurred during this phase, water and volume associated to this variation, volume interested by the gaseous phase and the concentration of each compound in the gaseous phase, with the associated number of moles. Finally, carbon dioxide and nitrogen captured into hydrates were calculated with the following equation: (2) Conversely, methane recovered was defined as the ratio between moles present in the gaseous phase at the end of the test and moles initially trapped into hydrates, expressed as percentage.
Both tests led to similar results in terms of methane hydrate formation: in the first case, 0.389 moles of hydrate were formed, while 0.333 moles in the second experiment. The process started at initial pressures close to 50 bar and reached 24 -28 bar as final values.
The main difference was observed in temperature trend: Test 2 showed an immediate peak in temperature and the first values measured were close to 10 °C. Conversely, in Test 3 hydrates formation did not occur immediately and temperature initially remained in the range externally established with the chiller. Once moles of hydrates formed were evaluated, also water involved in the process and volume occupied by hydrates were measured, in order to evaluate with accuracy the free volume feasible for flue-gas injection.
This latter compound was injected at pressures about 36 -37 bar, or higher if compared with Pf values. It means that methane replacement into hydrates was not favoured by thermodynamic conditions, as commonly occurs when pure carbon dioxide is used, but it was only based on the higher tendency of flue-gas (than methane) to be involved into hydrates at the same thermodynamic condition. Considering the high concentration of nitrogen, the equilibrium conditions for the present mixture are more severe than that for methane hydrates and any drop in pressure or increase in temperature can be performed to enhance the process.
Knowing the volume available to host flue-gas and the composition of the mixture, the exact number of carbon dioxide and nitrogen moles inserted into the reactor, was defined. The replacement process did not cause any significant variation in pressure, neither remarkable peaks in temperature. In Test 2, the pressure moved from 37.19 bar to 33.05 bar, while in Test 3, from 36.35 bar, it reached 34.49 bar. This little difference did not provide enough information to define the effectiveness of replacement, which was evaluated with the auxilium of gas-chromatographic analyses. However, this little decrease in pressure proved that, in any case, the amount of flue-gas involved in the process was higher than the methane recovered. The ending of the replacement phase was established as a function of time, according to previous works, where time required to complete the process was defined [33]. Here, a time period significantly higher than this latter one, was considered, to ensure the completion of the whole process. That solution was adopted due to the impossibility of considering pressure and temperature trend to define it.
As soon as the process finished, the gaseous phase present upon hydrates was analysed and revealed the following composition: 18.3 -20.71 % CO2; 50.13 -53.85 % N2; 25.44 -31.57 % CH4. Even if the thermodynamic conditions remained feasible for methane hydrate stability, part of methane was released from hydrates and was found in the gaseous phase. Between the two compounds present in the flue-gas mixture, carbon dioxide was preferentially involved into hydrates and showed a lower concentration in the gaseous phase. In Test 2, the gaseous phase above methane hydrates, initially contained 0.214 moles of both compound; conversely, when the process finished, 0.059 moles of CO2 and 0.162 moles of N2 were found, confirming that CO2 was mainly involved in the process and proving the participation of nitrogen in the replacement. Similar conclusions follow results belonging to Test 3, where the final gaseous phase contained 0.068 moles of CO2 and 0.179 moles of N2.
The last three parameters were used to clearly define the effectiveness of methane replacement with this specific flue-gas mixture. More than two third of carbon dioxide inserted in the reactor was trapped into water structures, proving a high capture efficiency, if considering its relatively low concentration in the initial mixture. The percentage of nitrogen was clearly lower and ranged from 14.76% to 24.3%. If pure nitrogen had been used there would have been no methane recovery, neither nitrogen storage. In a binary gaseous mixture, nitrogen participated to the process and contributed to increase methane recovery. Finally, the percentage of methane recovered was unsatisfactory (about 25.53 -26.22%) and must be significantly increased to make the process attractive. Two main reasons can be attributed to these results: the low content in carbon dioxide of the mixture used to carry out the replacement and low diffusion of flue-gas into the methane hydrated deposit. The initial phase was extremely intensive and methane hydrates formation interested the whole internal reactor, leaving too little free space for flue-gas diffusion. Due to this latter challenge, part of methane hydrates present inside the reactor, remained far from the replacement zone and the percentage of methane recovered remained consequently low.
Finally, temperature and pressure trend over time was shown for both experiments (figures 3 -6).   These diagrams graphically confirmed the previous description, based on results shown in Table 1. Methane hydrates formation was massive in both cases and a relevant temperature peak was consequently found: in Test 2 it occurred immediately, while in Test 3 it was measured after a certain time period. It mainly deepened from the stochastic nature of the reaction, which might not start immediately, even if the thermodynamic conditions are completely feasible. In addition, in Test 2, a secondary peak was observed, proving the formation of further hydrates nuclei in the reactor. Gaseous methane removal and flue-gas injection were performed very quickly and the variation in temperature related to them consists of an almost vertical treat. However, some thermocouples measured an increase in temperature, while the other revealed a local decrease. Firstly, methane was ejected and the gas expansion led to a decrease in pressure. However, gas expanded only in regions left free from hydrates and temperature decreased only in those regions. For the opposite reason, the subsequent flue-gas injection brought to an increase in temperature in the same regions, which was less marked than the previous drop. The presence of free water and feasible thermodynamic conditions, led to an immediate formation of CO2 hydrates (independently from the replacement process) which caused a remarkable increase in temperature. Finally, any significant variation was observed during replacement.
Pressure diagrams revealed how methane hydrates mainly formed during the first hours. In Test 2, the process accelerated immediately after the second temperature peak, while in Test 3 it assumed a constant gradient. Once hydrate formation was completed, pressure remained constant, until the second phase, where it decreased to the atmospheric value, due the removal of the whole gaseous phase present inside the reactor; then it increased again and reached PR. After an initial drop, mainly due to the ex-novo formation of carbon dioxide hydrates, pressure remained almost constant, proving that the leading process was gas exchange into already existing water cages, which did not cause relevant variation in pressure neither in temperature.

Conclusions
This work deals with methane replacement, into gas hydrates, with a binary flue-gas mixture, equally divided in carbon dioxide and nitrogen. Experiments were carried out in a small-scale experimental apparatus, appositely designed to simulate marine hydrates reservoirs. A first test consisted of direct formation of hydrates with flue-gas, in order to define the conditions required for their formation in the present reactor. Values produced where then compared with equilibrium values present in literature. The second part of this article, describes methane replacement with flue-gas. The replacement was performed by removing all gaseous methane and replacing it with flue-gas, as soon as the methane hydrates formation was completed. Both tests produced similar results: gas exchange, into previously formed water cages, occurred and was more abundant than ex-novo hydrates formation. As expected, carbon dioxide was mainly involved in the process (67 -73% of CO2 was captured and stored in form of hydrate); in addition, a significant percentage of nitrogen participated to the process, about 14 -24%, and provided a relevant contribution in terms of methane recovery. Finally, the percentage of methane recovered was low: it remained below 26% due to the low initial concentration of carbon dioxide in the replacing mixture and to difficulties addressed for flue-gas diffusion through the conglomerate of methane hydrates.
Future works will be focused on involving different flue-gas mixtures, with different initial concentrations; moreover, different strategies will be applied to carry out the replacement process, in order to improve its efficiency.