Critical Parameters Influencing Mixed CH4/CO2 Hydrates Dissociation during Multistep Depressurization

Early studies show that multistep depressurization of CH 4 /CO 2 mixed hydrates generates additional CH 4 while storing CO 2 in hydrate-bearing sediments. There are many critical factors that could affect the production and storage efficiency of this method. However, it is unclear how to achieve high efficiency in both CH 4 production and CO 2 storage by controlling these critical parameters. In this experimental work, we identified three critical parameters (CH 4 /CO 2 ratio in mixed hydrates, residual water saturation (S rw ), and shut-in period) and investi- gated their effects on production parameters (CH 4 molar fraction in the gas phase (XCH 4 ), CH 4 recovery percentage (RCH 4 ), and CO 2 storage ratio (SCO 2 )). Experiments were performed on sandstone cores using a high- pressure core flooding system equipped with pressure, temperature, and electrical resistivity measurements. Gas composition was analyzed by gas chromatography. The results showed that the optimal production pa- rameters were determined at low S rw of 43.7 – 47.4% and higher CH 4 /CO 2 ratio of 1.76 – 2.06 in CH 4 /CO 2 mixed hydrates. The optimized values were obtained at the equilibrium pressure of CH 4 /CO 2 hydrate system at a specific reservoir temperature without water production during pressure release. In addition, the period between two pressure releases had a direct effect on the production and storage performances, and the most efficient production and storage was measured at 4-hour shut-in period of pressure release. The measured percent changes in normalized resistivity ( Δ NR 2 ) were dependent on S rw and shut-in period. Positive increase of Δ NR 2 indicated increased hydrate saturation or improved water gas distribution in the sediment during multistep depressurization. The results demonstrated the importance of three critical parameters in designing an effective production and storage scheme after CO 2 injection into CH 4 hydrates.


Introduction
Natural gas hydrates (NGHs) are solid-crystalline compounds consisting of host water molecules and gas molecules (CH 4 , C 2 H 6 , CO 2 and H 2 ). Cage-like structures are formed by hydrogen-bonded water with gas molecules stabilizing inside under low temperature and high pressure [1]. NGHs are widely recognized as promising clean hydrocarbon energy for abundant reserves in marine sediments (0-110 m below seabed) and permafrost regions (130-500 m below underground) [2,3]. The total amount of global hydrate-bonded CH 4 reserve is reported to be within the scale of 10 15 -10 18 standard cubic meters (ST m 3 ), exceeding the known amount of conventional oil and gas resources [4][5][6]. Furthermore, noted that CH 4 is a kind of high-density energy with the combustion heat of approximately 890 kJ/mol and therefore regarded as one of the most excellent alternative energy sources in the 21st century [7]. Due to the vast reserve and high energy density of NGHs, an increasing number of researches have been conducted on NGHs exploitation in recent years.
Exploitation methods refer to those that recover natural gas (the majority are CH 4 ) from NGHs within reservoir, which mainly include depressurization [8,9], thermal stimulation [10,11], inhibitor injection [12,13] and CH 4 -CO 2 swapping [14,15]. The depressurization method reduces the reservoir pressure, thus drawing thermodynamic conditions out of the hydrate stability zone. This thermodynamic instability causes NGHs to decompose and produce CH 4 and water. However, endothermic dissociation of NGHs via rapid pressure drop can cause partial ice or hydrate reformation and then stop CH 4 production. Thermal stimulation is also a thermodynamic method that disrupts phase equilibrium conditions of NGHs. However, large heat loss makes it uneconomical, and more effective heating techniques need to be investigated. Inhibitor injection can moderate the equilibrium of NGHs and thus inducing hydrate dissociation. However, this exploitation method is restricted in application because of high cost and environmental pollution. Compared to other methods, CH 4 -CO 2 swapping is an energy efficient and non-hazardous method, and can achieve a win-win situation of simultaneous CH 4 recovery and CO 2 sequestration [16,17]. The principle is that CO 2 can form hydrate under more moderate thermodynamic conditions, replacing CH 4 in existing hydrate cages by forming new CO 2 hydrates [18]. This kind of swapping process between CH 4 and CO 2 without violent destruction of the hydrate structure can stabilize the stratum that probably collapses during the NGH exploitation process through depressurization and thermal stimulation [19].
Among these exploitation methods, depressurization is demonstrated to be the most commercial method and has been used to develop NGHs in several field tests [20][21][22][23]. Rapid/direct depressurization technique tends to trigger the formation of ice or secondary hydrates, which blocks fluid flow channels and reduces gas production. To avoid these problematic issues, a slow depressurization technique with wellcontrolled pressure drop rate or gradient pressure drop was designed, achieving efficient and continuous gas production. Fig. 1 presents an overview of slow depressurization-based CH 4 production from CH 4 hydrate and CH 4 /CO 2 mixed hydrate together with its technical concerns, which affect factors and efficiency parameters. As shown in Fig. 1, this technique can be mainly classified into constant-rate depressurization [24,25], cyclic depressurization [26,27] and stepwise depressurization [28][29][30]. These well-controlled slow depressurization may simultaneously reduce water production through gas/water migration and avoid stratum instability without sand production. The key parameters that affect gas production efficiencies include operating factors (depressurization pressure/temperature and shut-in period/frequency) and properties of the in situ reservoir (water/hydrate saturation and porous media) [31]. Another aspect that affects slow depressurizationbased gas production is to combine it with other exploitation methods, such as depressurization combined with thermal stimulation/heating, with chemical inhibitor injection, and with CO 2 gas injection.
The single CH 4 -CO 2 swapping method cannot meet the commercial NGH exploitation requirements due to the low CH 4 gas recovery efficiency [32]. The main reason for this low gas production is mass transfer caused by the formation of CO 2 hydrate films around CH 4 hydrates [33][34][35]. And the hydrate films also reduce gas molecules transport and CO 2 concentration around the surface of CH 4 hydrate by decreasing relative gas permeability. To overcome the drawbacks of CH 4 -CO 2 swapping, the combination of swapping with the methods mentioned above have been studied [36][37][38], one of which is CH 4 -CO 2 swapping combined with depressurization. This kind of combination method has been proven to be effective in the production testing of the Inik Sikumi field, where CO 2 -rich gas injection and depressurization were performed accordingly, with a total of 24,410 STm 3 CH 4 recovered from the hydrate-containing reservoir during a whole production period of 30 days [39]. The exploitation patterns of CH 4 -CO 2 swapping combined with depressurization have also been reported in many experimental investigations. Zhao et al. revealed that CH 4 -CO 2 swapping could be divided into two stages, and depressurization could induce partial melting of CH 4 hydrates and create pathways for CO 2 penetration in the second stage [36]. Chen et al. verified the feasibility of enhancing CH 4 recovery and CO 2 storage simultaneously with the assistance of depressurization operation. These two evaluation parameters of performance were integrated and indicated by the single parameter of CO 2 utilization efficiency. The results showed that both outlet and inlet pressures greatly affected CO 2 efficiency [40]. Yang et al. investigated flue gas injection into gas hydrate reservoir with further depressurization to generate CH 4 -rich gas (up to 90 mol%) while storing CO 2 efficiently into hydrate (up to 80 mol%) [41,42]. In another study, Heydari and Peyvandi performed CH 4 -CO 2 swapping followed by pressure reduction with the addition of rhamnolipid. The result of the CH 4 mole fraction was around 52 mol% in the gas phase, with the CH 4 recovery efficiency of 18.02% at the ending point [43]. These studies, however, paid less attention to the simultaneous maximization of CH 4 recovery and CO 2 storage. To do this, strategies can be decided from thermodynamic consideration. Goel [44] proved that the partial pressure of CH 4 increased with the decomposition of CH 4 hydrates, and a small amount of the decomposed CH 4 gas could reform hydrates when the partial pressure of CH 4 reached the phase equilibrium of the mixed CH 4 /CO 2 hydrates. Nevertheless, the ideal operation pressure is that between the equilibrium of CH 4 hydrates and CO 2 hydrates to release as much as CH 4 while storing as much as CO 2 from the perspective of thermodynamic stability [45,46].
It is known that hydrate loss would increase pore pressure and weaken soil particles link, triggering instability of sediments. CH 4 -CO 2 swapping can address this concern by storing CO 2 in the form of hydrates, and CO 2 hydrate-bearing sediments are stiffer than CH 4 hydratebearing sediments [47]. It is anticipated that further depressurization on  hydrates sample would cause a redistribution of CH 4 /CO 2 in hydrate and change of hydrate volume accordingly. Thus, it is needed to quantify the volume change of CH 4 /CO 2 hydrates to consider potential stiffness failure of hydrate-bearing sediments because too much hydrate loss may loosen cementation or skeleton of sediment [48]. From the perspective of electrical properties, NGHs are electric insulator and electrical resistance as one of the crucial electrical properties, can represent hydrate saturation in hydrate-bearing sediment system. It is proved that electrical resistance increased with hydrate formation and decreased with hydrate decomposition [49,50]. And it could be more reasonable to evaluate whether the stability of hydrate-bearing sediment was improved or not by electrical resistivity instead of hydrate volume change, because electrical resistivity was mainly affected by fluid saturation, types of guest molecules and total hydrate volume [51]. Although the study showed that pure CO 2 hydrates have lower resistance than pure CH 4 hydrates [52], the resistivity behavior of CH 4 -rich and CO 2 -rich mixed hydrates during depressurization is still not well understood.
In our preliminary investigation, depressurization combined with CH 4 -CO 2 swapping was designed, i.e. replacing the residual CH 4 within the hydrate cage with CO 2 after the operation of pressure reduction [33]. However, water production and hydrate reformation problems occurred when employing this combination method to exploit NGHs, though CH 4 recovery efficiency was enhanced. In another experimental study, CH 4 -CO 2 swapping combined with three-stage gradient depressurization was proposed to address the problem of mass transfer during CH 4 -CO 2 swapping exploitation of NGHs [53]. According to the phase equilibrium difference between CH 4 hydrates and CO 2 hydrates, depressurization after CH 4 -CO 2 swapping would enhance CH 4 production and CO 2 storage within a hydrate-bearing reservoir at the pressure below equilibrium conditions for CH 4 hydrates as well as above that for CO 2 hydrates. This assumption has been demonstrated by our recent studies, in which CH 4 /CO 2 mixed hydrates dissociation induced by controlled multistep depressurization can increase both CH 4 recovery and CO 2 storage more efficiently [45,46,54]. However, different influencing factors on CH 4 recovery and CO 2 storage performance have not been comprehensively analyzed thus far. More importantly, unknowns on the optimization of both CH 4 recovery and CO 2 storage still exist.
This study extended investigation on CH 4 -CO 2 swapping combined with depressurization by experimentally studying CH 4 /CO 2 mixed hydrate dissociation through controlled multistep depressurization. Assumptions were preset that CH 4 -CO 2 swapping for NGHs exploitation has already occurred and CH 4 /CO 2 mixed hydrate have formed before multistep depressurization. Influencing factors such as hydrate composition, residual water saturation (S rw ) and shut-in period during multistep depressurization were explored. Characteristics of CH 4 mole fraction in gas phase (XCH 4 ) were monitored, CH 4 recovery percent (RCH 4 ) and CO 2 storage ratio (SCO 2 ) during multistep depressurization were calculated in different operation conditions. Additionally, resistivity variation was studied to evaluate the effect of multistep depressurization on mixed hydrates saturation development.

Materials and apparatus
The mixed gas containing CH 4 (70 mol%) and CO 2 (30 mol%) and  the other mixed gas of CH 4 (30 mol%) and CO 2 (70 mol%) were purchased from Air Liquide (Denmark). The deionized water was produced in the laboratory. A schematic of the experimental apparatus of mixed hydrate formation and exploitation is presented in Fig. 2. The significant parts of the apparatus were the core holder with a maximum working pressure of 200 bar. The prepared sandstone sample was wrapped by a rubber sleeve (Fig. 3) and placed at center of the core holder, sealed by an inlet and outlet cap, respectively. The core holder had a thermocouple at the outlet and two differential pressure indicators. Two alligator clips measuring the electrical resistance were placed at the inlet and outlet, respectively. Our previous work can refer to more information on electrical resistance measurement [45,55]. Gas cylinders injected different gases into sandstone samples through relief valves, control valves and supply lines. A high precision syringe pump (Teledyne ISCO) provided the sandstone sample with confining pressure by injecting deionized water to surround the rubber sleeve. A cooling bath controlled the system temperature by circulating antifreeze at the cooling jacket. Temperatures, pressures and electrical resistances were recorded by the data collector (Agilent 34972A, Agilent Tech.) at 10 s period. The gas samples were collected at the outlet of the core holder and analyzed by the in-line gas chromatograph (Micro-GC 490).

Core sample preparation
The preparation of the core sample included two parts: cleaning and saturating. The sandstone core was first washed with organic solvent (methanol and toluene) and then dried in an oven at 100 • C for 24 h. The dry weight of the sandstone core was subsequently recorded. For the core sample of low saturation, the dry sandstone core was immersed with ultrapure water in a beaker for 24 h. For the core sample of high saturation, the immersed sandstone core in a beaker was vacuumed simultaneously for 48 h, ensuring that the pores within the sandstone core were sufficiently degassed and thus saturated with water adequately. A prepared sandstone sample can be seen in Fig. 3. Core samples of different saturations using the preparation method above were obtained and summarized in Table 1.

Hydrate formation
The hydrate formation process in this study is consistent with our previous work [45]. The water-saturated core sample was wrapped tightly with a rubber sleeve and then packed into the core holder, with the inlet cap and the outlet cap pressing inwards. After connected the core holder with tubes and valves, the cooling bath was set to the desired temperature value. The core holder was flushed with target gas at 5 bar for 2 min to ensure the absence of air. Pressurization operation followed the following two steps: firstly, the desired confining pressure was achieved by water injection through ISCO pump; secondly, mixed gas was injected continually with the core pressure around 30 bar less than confining pressure. These two procedures were repeated for multiple injections until the confining pressure of 110 bar and core pressure of 70-90 bar were achieved. When the core pressure stabilized for 2 h at 22 • C, the temperature of the cooling bath was decreased to the experimental value. A noticeable pressure drop can be seen as the process of hydrate formation started. When the pressure of the core holder remained unchanged for 12 h, the hydrate formation process was determined to be completed. To improve water conversion into hydrate and hydrate distribution, a temperature ramping (multiple cooling and heating cycles within the temperature range of 1 • C and 22 • C) was employed [56,57]. Finally, if no pressure difference was seen between the inlet and outlet after temperature ramping operation, then the redistribution was regarded as completion totally inside the core sample due to adequate gas/liquid migration. Electrical resistance (R) between the alligator clip 1 (inlet) and 2 (outlet), as well as pressure and temperature, were measured and recorded during the whole process. The R values were normalized during hydrate formation as the first normalization resistance (NR1); a detailed description can be found in the Ref. [45].

Multistep depressurization
After CH 4 /CO 2 mixed hydrate formation, the following R values were normalized as second normalization resistance (NR 2 ). The position of the thermocouple was placed at the outlet, which can monitor the temperature change at the production site when employing multistep depressurization. Pressure drop at each depressurization stage was controlled by opening and closing the outlet value quickly, with the gas samples simultaneously collected and analyzed through the in-line micro gas chromatographer. Change of gas composition and resistivity were determined to investigate the dissociation characteristics of CH 4 / CO 2 mixed hydrates, i.e. RCH 4 , SCO 2, and percentage change of hydrate volume (Δ V) were calculated at the end of each depressurization according to mass balance. Percentage change of NR 2 before and after multistep depressurization (Δ NR 2 ) was calculated as well to analyze the variation of hydrate saturation during the multistep depressurization. The procedure of multistep depressurization proceeded with the pressure firstly decreased to CH 4 hydrate equilibrium pressure and then below it. This operation ceased until the value close to or under CO 2 hydrate equilibrium pressure. Noted that hydrate equilibrium was obtained from CSMGem for the bulk-only phase and it may differ in porous media. The hydrate equilibrium was assumed unchanged considering the reservoir temperature remained constant (small fluctuations in core temperature was ignored) during slow depressurization. Finally, the temperature was increased to 25 • C to dissociate all the hydrates.

Profile of CH 4 /CO 2 mixed hydrate formation
A detailed description of CH 4 /CO 2 mixed hydrate formation can be referred to our previous publication [45]. In this work, similar strategies were used to generate mixed CH 4 /CO 2 hydrates. The profiles of temperature, outlet pressure (P out ), inlet pressure (P in ), and first normalized resistance (NR 1 ) in different experiments have a similar trend. Fig. 4 shows the profiles of CH 4 /CO 2 mixed hydrate formation in Exp1 and Exp2, respectively. It should be noted that the hydrate morphology is pore filling for water saturation higher than 35% [58]. A summary of moles stored in hydrate (total, CH 4 and CO 2 ), the value of NR 1 , and the valve of NR 1 to hydrate saturation after mixed CH 4 /CO 2 hydrate formation were showed in Fig. 5 (a) and Fig. 5 (b). More detailed information can be tracked in Table S1 of Appendix A.
In Fig. 4, sudden pressures were observed in the initial stage of hydrate formation for each cooling cycle. This was the sign of massive hydrate formation consuming CH 4 /CO 2 mixed gas. Comparably, the following gentle pressure drops and final flat pressure lines were seen in the final stage of hydrate formation, which were caused by mass transfer barriers of hydrate films between water and gas molecules [33][34][35]. In Table S1 of Appendix A, the range of S wi was 53.75-88.62% and it seemed that lower initial water saturation (S wi ) induced a more significant total amount of gas stored in hydrate, i.e. higher hydrate saturation (S h ). Noted that water diffusion plays a dominant role during hydrate formation and growth, in which hydrate films formed around water

Fig. 5.
A summary after mixed CH 4 /CO 2 hydrate formation at different experiments: (a) total gas stored in hydrate, CH 4 in hydrate and CO 2 in hydrate, as well as NR 1 ; and (b) NR 1 to hydrate saturation.

Table 2
Information on multistep depressurization of mixed CH 4 /CO 2 hydrates. S rw and S h are the residual water saturation and hydrate saturation before multistep depressurization. P CH4 , P (CH4/CO2) and P CO2 refer to the hydrate equilibrium pressures of CH 4 , mixed CH 4 /CO 2 right before multistep depressurization and CO 2 , calculated from CSM [62]. Exp  phase acting as diffusion barrier [34,59]. Comparably, CO 2 gas diffusion dominates the mass transfer during CH 4 -CO 2 swapping on existing CH 4 hydrate [60,61]. Nevertheless, mixed hydrates of pure CH 4 , pure CO 2 and CH 4 /CO 2 coexisted in both cases in which hydrates formed from either mixed gases or CH 4 -CO 2 swapping, although heterogeneity and micro properties differed. It can be seen from Fig. 5 (a) that the mole of CH 4 in hydrate phase was higher than that of CO 2 in Exp1, Exp3-4 and Exp7-8, i.e., the mole ratio of CH 4 to CO 2 in the hydrate phase (RH) was greater than 1 shown in Table S1. This showed that mixed hydrates are rich in CH 4 and therefore classified as Type A. Comparably, the mole of CH 4 in hydrate phase was less than that of CO 2 in Exp2 and Exp5-6, which indicated that the CO 2 -rich hydrates formed and thus were referred as Type B. Noted the values of RH for Type A hydrates (1.51-2.10) and Type B hydrates (0.15-0.27) were consistent with those calculated from CSMGem in Table S1. Additionally, a range of NR 1 (1.07 to 1.50) was observed, indicating an increase of hydrate saturation during CH 4 /CO 2 mixed hydrate formation. Generally, it can be speculated from Fig. 5 (b) that higher hydrate saturation triggered higher NR 1 regardless of hydrate type. Nevertheless, the different amounts of CH 4 and CO 2 moles stored in mixed hydrates and heterogeneous hydrate distribution contributed to the difference of NR 1 within a range of similar hydrate saturation.

CH 4 /CO 2 mixed hydrate dissociation
After synthesis of CH 4 /CO 2 mixed hydrates of two types, different depressurization strategies were introduced. The detailed information on well-controlled manners is listed in Table 2. Next, different parameters including hydrate compositions, residual water saturation (S rw ), and shut-in period, were investigated for their correlations with CH 4 recovery and CO 2 storage. Specifically, Type A and Type B hydrates were synthesized in Exp1 and Exp2, respectively, to examine the effect of hydrate compositions on production and storage parameters during multistep depressurization. The effect of higher and lower S rw was then studied in Exp3-6 to determine the value of S rw that benefited CH 4 recovery and CO 2 storage for Type A and Type B hydrates. Finally, shut-in durations of 2 h (Exp4), 4 h (Exp7) and 8 h (Exp8) were employed to find a relatively beneficial selection for multistep depressurization on CH 4 / CO 2 mixed hydrates.

Effect of hydrate composition
In realistic hydrate exploitation, CH 4 /CO 2 mixed hydrates would exist at the CH 4 hydrate reservoir after CO 2 injection. In this case, kinetic inhibitors and anti-agglomerate chemicals are recommended to be added together with CO 2 injection to eliminate CO 2 plugging at the injection site. This could further improve CO 2 injectivity and allow a high CO 2 sweep range [37,63]. The following CO 2 migration from the injection site depends on permeability of hydrate-bearing sediment and relative permeability of CO 2 , resulting in CO 2 -water interface and CO 2hydrate interface moving forward. This further induces formation and growth of both pure CO 2 hydrate and CH 4 /CO 2 mixed hydrate. The CO 2 concertation decreases as the distance increases from injection site, causing CO 2 -rich hydrates to enrich nearby the injection well and CH 4rich hydrates to stay closer to the production well. As more CH 4 is produced from production well, CO 2 migration is driven from injection well to production well. This may alter the previous CH 4 -rich hydrate zone into CO 2 -rich one because of water and gas redistribution. Therefore, we investigated and compared the dissociation characteristics of CH 4 -rich (Exp1) and CO 2 -rich (Exp2) mixed hydrates, respectively. Additionally, the mole ratio of CH 4 to CO 2 in hydrate was 2.10 in Exp1 and 0.15 in Exp2, and the residual water saturation in Exp1 (77.3%) and Exp2 (73.8%) were similar. The shut-in period was controlled at 4 h between multistep depressurization. More information can be referred in Table 2 above and Table S1 of Appendix A. Fig. 6 presents the profiles of CH 4 and CO 2 mole fraction in collected gas samples (XCH 4 & XCO 2 ), pressure (P in & P out ), temperature (T) and second normalized resistance (NR 2 ) during the multistep depressurization on Type A hydrate in Exp1. It should be noted that P out refers to depressurization pressure because pressure release was conducted in the outlet. Table S2 of Appendix A lists the values of those evaluation indicators at the end of different stages. The shut-in period was set to 4 h with a total of 13 pressure drops (Stage A to Stage M). It can be seen that the first drop was 6 bar, which brought the pressure to equilibrium pressure of mixed CH 4 /CO 2 hydrate (P CH4/CO2 ) at 0.7 • C. The following pressure drops were controlled at 3-4 bar (stage B to stage J) to explore the characteristics of multistep depressurization. The final pressure drops were made at 8 bar to approach the equilibrium pressure of CO 2 hydrate (P CO2 ).
It was inspected that from Fig. 6 initial XCH 4 was 73.7 mol% before the multistep depressurization. This value decreased to 66.7 mol% at the end of Stage A without pressure rebound. Then XCH 4 increased as well as fluctuated around 70.0 mol% from Stage B and Stage D with pressure rebound. The increase of XCH 4 was attributed to the dissociation of CH 4rich hydrate. It was noted that a pressure differential existed between outlet and inlet until stage G, in which a noticeable spike was detected on the outlet pressure curve. Pressure differential between outlet and inlet mainly redistributed gas and water in the pores. Subsequently, continuous pressure fluctuation was observed during Stage H and XCH 4 peaked at 78.6 mol% at the end of Stage H. This was caused by CH 4 -rich hydrate dissociation and continuous fluid movement and gas-water mixing in the pore. It was noted water production occurred at the end of Stage L with obvious pressure rebound, which were produced by a large amount of CO 2 -rich hydrates dissociation. This should be avoided because CO 2 release would reduce CO 2 storage efficiency and massive CO 2 -rich hydrate loss may trigger instability of hydrate-bearing sediment. According to the Table S2 of Appendix A, a continuous increase of RCH 4 was observed at Exp1 from 72.0% at Stage A to 79.7% at Stage M. This was caused by gradual CH 4 hydrate dissociation when depressurization pressure was reduced gradually. However, actual CH 4 recovery efficiency would be decreased because of obvious XCH 4 decrease in the later steps of depressurization. The problematic issues of water production and dramatic XCH 4 decrease were therefore co-considered to cease multistep depressurization at selected stage.
In terms of enhanced CO 2 storage in Type A hydrates, SCO 2 experienced an increase before the end of Stage H with the highest value of 84.0% and then decreased slightly until the end of multistep depressurization. This was because CO 2 initially formed hydrate with the depressurization pressure far beyond P CO2 . It was noted that SCO 2 started to drop since Stage I in Table S2. This was observed because those CO 2 -rich hydrates started to dissociate even the depressurization pressure was 10 bar higher than P CO2 . Considering water production was observed during gas sample collected at the end of Stage L. Therefore, multistep depressurization was suggested to be terminated before this pressure release to avoid uncontrolled massive hydrate dissociation.
As can be seen in the Table S2 of Appendix A, negative value of ΔV in Exp1 hydrate meant a loss of hydrate mass volume during multistep depressurization. This was caused by the fact that the amount of CO 2 hydrate storage was fewer than that of CH 4 hydrate recovery. A relative minimal of ΔV (− 56.28%) was realized at the end of Stage H, accompanied by the largest XCH 4 and SCO 2 . This pressure point should be marked because enhancement of both CH 4 recovery and CO 2 storage was maximized. Nevertheless, it seems from Table S2 of appendix A that ΔNR 2 increased with more steps of depressurization. Although a decrease of hydrate saturation was observed according to the negative ΔV, the increased ΔNR 2 implied that multistep depressurization on CH 4 /CO 2 mixed hydrates between P CH4 and P CO2 improved water-gas distribution along the core. Fig. 7 shows the gas composition, pressure, temperature and NR 2 during the multistep depressurization on Type B hydrate in Exp2. Supplementary data can be referred to the Table S2 in Appendix A. It can be seen from Fig. 7 that the number of pressure drops was 6 in total (Step A to Step F). When the depressurization was performed from 50.8 bar to 41.2 bar, it can be seen that XCH 4 increased from 35.2 mol% at the beginning to 37.0 mol% at the end of Stage A. Since the pressure (41.2 bar) during Stage A was still higher than P CH4 , the slight increase of XCH 4 above was induced by CO 2 -rich hydrate reformation, which was supported by a minimal pressure drop during Stage A. Noted that significant hydrate reformation was shown as sharp pressure drop just before the end of Stage B. The amount of CH 4 and CO 2 forming hydrate were equivalent, and therefore, the XCH 4 at the end of Stage B (36.6 mol %) was nearly equal to that at the end of Stage A (37.0 mol%).
It was noticed from Fig. 7 that the pressure differential between outlet and inlet occurred throughout the multistep depressurization. At the beginning of Stage D, a sudden pressure drop induced dissociation of CO 2 -rich hydrates to release water and gas. On the one hand, this caused a 1.2 mol% increase of XCO 2 . On the other hand, mass transfer was alleviated because of partial hydrate dissociation. Water and gas redistributed from outlet to inlet with time, causing hydrate dissociation and reformation along with the core, showed as pressure fluctuation during Stage D. The value of XCH 4 decreased from 34.8 mol% to 31.6 mol% from Stage D to Stage F. This was caused by the fact that the existing CO 2 -rich hydrates tend to be unstable when depressurization pressure was close and below P (CH4/CO2) . As a result, partial CO 2 -rich hydrates dissociated and released more CO 2 than CH 4 and decreased XCH 4 . Nevertheless, as seen in Table S2, RCH 4 increased from initial 30.4% as more CH 4 recovered with multistep depressurization. The highest RCH 4 was 81.2% acquired at the end of Stage F.
As for CO 2 storage for Type B hydrate in Exp2, an increasing trend of SCO 2 was observed with stages, and the highest SCO 2 was 82.8% nearly at the end of Stage E. Since the depressurization pressure was above PCO 2 , CO 2 -rich gas tended to form hydrate. The process of multistep depressurization was suggested terminated at the end of Stage E because massive water production occurred at the end of Stage F. The positive ΔV in Exp2 of Type B hydrate in Table S2 indicated a gain of hydrate mass volume during multistep depressurization on CO 2 -rich hydrates, i. e. the amount of CO 2 reformation was larger than total loss of CH 4 hydrates. A maximal of ΔV (109.57%) was realized at the suggested ceasing point. Additionally, NR 2 increased with multistep depressurization indicating hydrate saturation in sediment increased.
According to the Table S2, the highest XCH 4 for Type A hydrate (78.6 mol% at 18.48 bar) was more than twice larger than those any value of XCH 4 for Type B hydrate, although RCH 4 at 18.48 bar for Type A hydrate (74.8%) was a lower than that for Type B hydrate (79.4%). This is important because CH 4 -rich mixed gas can be applied after purification through an economical separation method, while CO 2 -rich one was costly to be used [64,65]. In terms of CO 2 storage, the highest SCO 2 for Type A hydrate (84.0%) was slightly more than that for Type B hydrate (82.8%). It seemed that the Type A hydrate-bearing reservoir was dominated by CH 4 recovery while Type B by CO 2 storage through multistep depressurization according to the negative ΔV in Exp1 and positive ΔV in Exp2, respectively. However, ΔNR 2 values were both positive in Exp1 and Exp2, showing that sediment was stabilized with multistep depressurization operation. This can be understood that water gas transportation and distribution in the pore mainly affect the resistivity of hydrate-bearing sediment [51], although hydrate volume increased in Exp2 was mainly created by CO 2 hydrate reformation, whose resistivity is lower than CH 4 hydrate [52]. The results of negative ΔV in Exp1 and positive ΔV in Exp2 while both have increased NR 2 also indicated resistance was not directly correlated with mass volume but distribution that connected with fluid flow and water saturation. Therefore, the results above collectively indicated that it is better to perform multistep depressurization at the CH 4 -rich hydrate zone to acquire and utilize CH 4 gas. This means the operation of multistep depressurization should cease before CH 4 -rich hydrate changes into CO 2 -rich one.

Effect of residual water saturation
Two types of hydrate deposits can be formed in porous media with different initial water saturation (S wi ), i.e. excess gas (S wi ＜50%) and excess water (S wi > 50%). And excess water deposit may hinder gas production from depressurized hydrate reservoir [9]. It is noted that sandstones of high water saturation (>35%) generate hydrate morphology of pore-filling [58]. Therefore, Exp3-6 of excess water deposits (S wi = 53.75-88.62%) were created with mass transfer barrier among pore water/gas and pore-filling hydrate affecting CH 4 recovery and CO 2 storage. During gas production from hydrate reservoir with higher water saturation, multiphase exits in the porous media and residual water saturation (S rw ) may be the dominant factor affecting heat transfer and mass transfer [66], and thus deciding CH 4 recovery and CO 2 storage by multistep depressurization. Therefore, the effect of different S rw on multistep depressurization of CH 4 /CO 2 mixed hydrate was investigated by performing experiments on Type A (Exp3 & Exp4) and Type B hydrate (Exp5 & Exp6), respectively. The mole ratio of CH 4 to . The shut-in period was designed 2 h between multistep depressurization for Exp3-6, as described in Table 2 and Table S1 of Appendix A.

Type A mixed hydrate.
In Exp3 of Type A hydrate sample, the higher S wi of 82.2% triggered higher S rw of 78.9% and lower S h of 3.29% after hydrate formation. Fig. 8 shows the profiles of Type A mixed hydrate dissociation during multistep depressurization of 8 steps in Exp3. When performing multistep depressurization above P CH4 in Type A hydrate during Stage A and Stage B, no drastic change of XCH 4 was detected (only from 76.9 mol% to 77.9 mol%). This was caused by by slight CH 4 -rich hydrate dissociation, as shown a pressure rebound appeared just after shut-in operation. A noticeable pressure drop occurred across the P CH4 line and the XCH 4 decreased from 77.9% to 75.8% during Stage C, indicating massive reformation of CH 4 -rich hydrates. Subsequently, a pressure differential occurred between outlet and inlet, and XCH 4 was nearly unchanged during Stage D, showing nearly no gas composition change at the outlet. Water production was observed during gas collection at the end of Stage E, which can be predicted based on the pressure rise during Stage E as a result of massive hydrate dissociation. It was interesting to observe both dissociation and reformation of mixed hydrate with pressure increasing and decreasing during Stage G, while only hydrate dissociation dominated during Stage H with pressure rising.
When multistep depressurization was conducted between P CH4 and P CO2 , most CH 4 -rich hydrates dissociated and contributed to RCH 4 recovery and CO 2 -rich hydrate formed to enhance CO 2 storage. Practically, depressurization should controlled before Stage E in which water production started. To explore the highest values of RCH 4 and SCO 2 with the depressurization reduced close to P CO2 , further multistep depressurization was introduced after water production at the end of Stage E. It should be noted that pressure differences existed again during Stage G. This may be attributed to extreme inhomogeneity along the sandstone sample. Table S3 in the Appendix A illustrates the data of XCH 4 , RCH 4 , SCO 2 , ΔV, and ΔNR 2 in Exp3. It can be seen that RCH 4 increased to peak values of 87.6% and SCO 2 increased to 86.2% at the end of Stage G. However, CO 2 -rich hydrates occurred to dissociate significantly with depressurization pressure decreasing close to P CO2 at the end of Stage G, which was indicated as an obvious increase in pressure and a decrease in XCH 4 in Stage H, resulting in both decreases in RCH 4 and SCO 2 . It was noted that water was continuously produced from Stage E to Stage H proving massive hydrates continuously dissociated, suggesting depressurization should be stopped before the end of Stage E. It can be seen from Table S3 that ΔV in Exp3 was negative, and its value decreased with time. This indicated that the amount of CH 4 recovered was greater than that of CO 2 stored over time. A negative range of ΔNR 2 (− 2.6% to 9.9%) was also observed during multistep depressurization, confirming hydrate loss in hydrate-bearing sediment.
Comparatively, lower S rw (43.7%) corresponded to higher S h (12.32%) in the sandstone core was generated in Exp4 of Type A hydrate sample, the shut-in period was fixed at 2 h, same as in Exp3. A total of 9 steps of depressurizations was introduced, and each pressure drop was controlled with 8-12 bar. Fig. 9 describes the changes of hydrate dissociation characteristics for multistep depressurization in Exp4. More information can be tracked in Table S3 of the Appendix A. No change in XCH 4 was detected during Stage A because pressure was above P CH4/CO2 and only free gas was removed. Noted that a pressure recovery was observed at the outlet in stage C due to the pressure difference existing between the outlet and the inlet. This pressure rebound allocated CH 4 and CO 2 between phases, causing noticeable declines in RCH 4 (from 77.2% to 67.6%) and SCO 2 (from 83.5% to 77.7%). Similar slight pressure increases were observed during Stage D-F, caused by continuous hydrate dissociation during shut-in period. The peak values of XCH 4 (78.6 mol%) and SCO 2 (85.1%) appeared at the end of Stage G, and then decreased gradually as a result of CO 2 -rich hydrate dissociation caused by depressurization pressure performed below P CO2 . The phenomenon of water production appeared at the end of Stage G, which was the indicator of unwanted massive dissociation of hydrates and herby multistep depressurization needed to be terminated.
In Table S3 of Appendix A, the negative value of ΔV showed that the amount of CO 2 hydrate storage was less than that of the recovery of CH 4 hydrate during the whole depressurization. It was interesting ΔNR 2 that was initially positive at Stage A-D, indicating that multistep depressurization on lower S rw hydrate sample improved gas water distribution at these stages. Subsequently, ΔNR 2 was negative since stage E, and the sediment became less stable with loss of hydrate saturation than before employing multistep depressurization. Nevertheless, the suggested ceasing point was selected at the end of Stage G with a relative lower ΔNR 2 of − 1.2%.
It was inspected that Exp3 of higher S rw was not suitable for multistep depressurization because more hydrates dissociated and involved high risk of water production. Comparably, the drop in NR 2 was lower in Exp4 than Exp3, denoting that hydrate saturation was much more reserved and gas water distribution was improved in Exp4 of low S rw . Fig. 10 presented the correlation between RCH 4 and SCO 2 with depressurization pressure (P out ) in Exp3 and Exp4. A depressurization pressure window was proposed for the controlled pressure release operation. Specifically, the upper pressure window (P upper ) was equal to PCH 4 and the lower pressure window (P lower ) was the ceasing pressure where water started to produce. When the depressurization pressure was reduced below P upper , even though a higher RCH 4 of 79.0% was obtained in Exp3 at P lower of 17.6 bar, much larger negative value of Δ NR 2 indicated huge loss of hydrate saturation in Exp3. This limited the application of multistep depressurization in such hydrate reservoir. Although lower RCH 4 of 75.2% was achieved in Exp4 at P lower of 22.5 bar, less loss of hydrate mass and lower drop in Δ NR 2 allowed more effective depressurization than Exp3. The highest CO 2 storage ratio values were nearly the same in Exp3 (82.9%) and Exp4 (82.6%) at P lower , indicating that both scenarios were potential selections for CO 2 storage. Collectively, multistep depressurization was highly proposed to improve CH 4 recovery from the CH 4 -rich hydrate reservoir of lower S rw and most of the CO 2 can be stored safely within the pressure window.

Type B mixed hydrate. The effect of different S rw on dissociation manners of multistep depressurization was also investigated in Type B
hydrates. Exp5 was conducted in Type B hydrate sample of higher S rw = 86.6%. Fig. 11 presents the changes of CH 4 /CO 2 in gas phase, pressure, temperature, and NR 2 in Exp5. A total of 9 depressurization steps was performed with each pressure drop of 6-10 bar, and the shut-in period was fixed at 2 h. More information can be found in the Table S4 in Appendix A. Initially, the pressure was brought below P CH4 by performing 6 steps of depressurization at the outlet. It was observed that the inlet pressure was first above P out and then plummeted below the corresponding P out . This was caused by a slow water-gas distribution process along the length of sandstone. It was interesting that XCH 4 peaked at 38.3 mol% at stage C, where the P out was equal to the P in . It seemed that there was a blockage between the outlet and the inlet after stage E, since the pressure differential came and existed within the rest time. At the early stage of Stage G, two obvious pressure drops were inspected, indicating that CO 2 -rich gas reformed hydrates as the pressure below P (CH4/CO2) . This was supported by the indicators of SCO 2 , which reached the highest values of 88.7% at the end of Stage G. The value of XCH 4 decreased significantly with more steps of depressurization at the pressure close to P CO2 . Nevertheless, it was observed that water production occurred at the end of Stage G and thus ceasing point was selected accordingly.
In Table S4 of Appendix A, the positive value of ΔV showed that Exp5 achieved more CO 2 hydrate storage than CH 4 hydrate recovery during the whole multistep depressurization. It was understood that CO 2 -rich mixed gas tended to form CO 2 -rich mixed hydrate with the pressure above P CO2 . ΔNR 2 was within the range of (− 4.9% to 1.9%), except for a sharp decrease caused by short-term insulation failure in Stage E weakening the ΔNR 2 (− 6.6% at the end of Stage F). The value of ΔNR 2 above illustrated that multistep depressurization of Type B hydrate of higher S rw overall increased hydrate saturation and significantly improved gas water distribution along the core.
Exp.6 was the investigation of multistep depressurization characteristics in Type B hydrates of lower S rw = 42.0%. A total of 9 stages of mixed hydrate dissociation were studied, with each pressure drop being   6-10 bar. The 2-hour shut-in period was used to perform the stepwise operations. Fig. 12 listed the detailed description of gas composition, pressure, temperature and NR 2 during multistep depressurization, and more data can be tracked in Table S4 of Appendix A.
In Fig. 12, the initial XCH 4 was 39.8 mol% before stage A. Subsequently, the depressurization pressures in Stage A and Stage B was decreased close to PCH 4, and thus the XCH 4 values were elevated to 42.0 mol% and 43.7 mol%. A significant pressure drop was observed during stage C, and XCH 4 was reduced to 38.5 mol%. This was caused by CH 4rich hydrate reformation within stage C. The operating pressure fluctuated around P CO2 during Stage D and Stage E, having a marginal effect on the stability of CO 2 -rich hydrates without obvious change of XCH 4 . However, following depressurization was carried out during Stage F-I below P CO2 , causing CO 2 -rich hydrates to dissociate and reduced XCH 4 . It was noticed that water appeared at the end of Stage H and therefore depressurization should be closed before it. Exp6 also witnessed a positive ΔV during whole stages, indicating that more CO 2 was stored than CH 4 hydrate was recovered. ΔNR 2 was observed to be increasing with steps after Stage C, indicating an increase of hydrate saturation in terms of multistep depressurization on Type B hydrate of lower S rw .
It was concluded from Table S4 that both ΔV were positive and ΔNR 2 increased with small fluctuations in Exp5 and Exp6, implying that Type B hydrate reservoirs of both higher S rw and lower S rw can be potential selection for multistep depressurization with slow manners. Fig. 13 showed the correlation between RCH 4 and SCO 2 with P out in Exp5 of higher S rw = 86.6% and Exp6 of lower S rw = 42.0%. Regarding the depressurization pressure window of P upper and P lower , it can be seen that the highest SCO 2 was 88.7% for Exp5 of higher S rw at P lower of 15.97 bar, with a relative higher RCH 4 of 87.4%. The corresponding ΔV was 108.27% and ΔNR 2 was − 1.9% at P lower , which indicated hydrate saturation gain during multistep depressurization. Comparably, P lower in Exp6 was 13.10 bar which was just above P CO2 , with the relative higher RCH 4 of 81.9% and SCO 2 of 88.3%. ΔNR 2 was only − 0.8% and ΔV was 54.65% at this P lower denoted increased hydrate saturation after multistep depressurization. As a reslut, Type B hydrate reservoirs of either higher S rw or lower S rw is able to achieve high efficiency of enhanced CH 4 recovery and CO 2 storage via multistep depressurization. According to the Table S4 in Appendix A, it may increase the economic competitiveness in Exp6 of low S rw hydrate reservoir considering its higher XCH 4 of 39.8 mol% than 32.2 mol% in Exp5 of higher S rw . Therefore, multistep depressurization was highly proposed for Type B hydrate of lower S rw . And the multistep depressurization was suggested to be conducted within the pressure window to efficiently enhanced CH 4 recovery and CO 2 storage. Fig. 14 summarizes the RCH 4 and SCO 2 in combination studies on CO 2 swapping with depressurization at different S rw and compared the results with those in Section 3.2.2 in this work. It can be seen from Fig. 14 that the RCH 4 was 60% and 76% in Sun's work [67] conducting one-step depressurization and CO 2 /H 2 injection. It was found that lower S rw induced higher RCH 4 with the SCO 2 was nearly same at 40%. With a relatively lower S rw of 19.2-28.5%, the range of RCH 4 was 17.1-74.4% and SCO 2 was 7.4-41.7% in Shi's [53] performing multistage depressurization after CO 2 /N 2 swapping. The large differences between CH 4 recovery and CO 2 storage were caused by different multistage depressurization pressures. Overall, the RCH 4 and SCO 2 in Section 3.2.2 were was 75.2-87.4% and 82.6-88.7%, higher than those in the references. And no obvious trend was observed on CH 4 recovery and CO 2 storage as a function of S rw . The optimized depressurization scheme was determined in Section 3.2.2 based on higher XCH 4 in produced gas without significant loss of hydrate volume, i.e. 78.6 mol% obtained at Type A hydrate of lower S rw = 43.7%. Fig. 13. The correlation between RCH 4 and SCO 2 with P out in Exp5 of higher S rw = 86.6% and Exp6 of lower S rw = 42.0%. SCO 2 and RCH 4 were calculated using equations described in Appendix B.2.

Fig. 14.
Comparison of RCH 4 and SCO 2 in CO 2 swapping combined with depressurization from references and this work at different residual water saturation [53,67]. More details can be referred to Table S6 in Appendix A.

Effect of shut-in period
Extension of shut-in period can allow enough time for water and gas migration, which is dependent on capillary forces. This migration of water and gas helps to avoid the blockage of fluid flow channels and thus promoting hydrate dissociation and reformation. Meanwhile, multistep depressurization with sufficient shut-in period can allow sensible heat to transfer from reservoir to hydrate dissociation front and thus preventing unwanted ice formation. However, valuable production time can be wasted to pursue uneconomically enhanced CH 4 recovery and CO 2 storage using excessive shut-in period. The following work, therefore, extended the shut-in period to 4 h (Exp7) and 8 h (Exp8) in the CH 4 -rich hydrate reservoir and compares the results with that of 2 h (Exp4). The aim was to find the highest XCH 4 and SCO 2 , consider the total production time, and verify the resistance variation. It was known that S rw was similar among experiments with 43.7% in Exp4 and Exp7, and 47.4% in Exp8. The mole ratio of CH 4 to CO 2 in hydrate was identical with 1.76 in Exp4, 2.06 in Exp7 and 1.80 in Exp8.
The shut-in period of 4 h was firstly selected, and a total of 14 steps were performed on Type A hydrate in Exp7. Profile of multistep depressurization on Type A hydrate in Exp7 can be analyzed in Fig. 15 and Table S5 of Appendix A. An increase in XCH 4 from 72.9 mol% to 74.5 mol% was inspected in Stage A with slight pressure decrease, induced by CO 2 -rich hydrate reformation. Pressure rebounds were observed during Stage B-F with depressurization pressures just above P (CH4-CO2) , indicating CH 4 -rich hydrate dissociation. This brought the highest XCH 4 (81.3 mol%) at the end of Stage F. Afterwards, the depressurization pressure was below P (CH4/CO2) and approached P CO2 . This induced CO 2 -rich hydrates unstable and started dissociating, causing XCH 4 to decline gradually until the final pressure below P CO2 . Water production was detected at the end of Stage J and ceasing point should be marked. It was noted that pressure fluctuation happened between outlet and inlet during Stage L. This implied that gas and water redistributed and transported as well as hydrate dissociation and reformation.
In terms of CO 2 storage evaluation, SCO 2 gradually increased from 74.6% at Stage A to the highest value of 85.3% at stage F because of CO 2rich hydrate formation according to Table S5. After the depressurization pressure approached P CO2 , the existing CO 2 -rich hydrates were more likely to dissociate, and therefore SCO 2 decreased since stage F. The exploitation time was considered 24 h for the highest XCH 4 and SCO 2 by counting 4-hour steps of 6 total. In Table S5 of Appendix A, the negative values of ΔV showed that the amount of CO 2 hydrate storage was fewer than that of CH 4 hydrate recovery. However, ΔNR 2 was positive throughout multistep depressurization, indicating that gas and water movement along the core significantly improved hydrate-containing sediment stability. Fig. 16 shows the profiles of gas composition, pressure, temperature and NR 2 during the multistep depressurization in Exp8. The shut-in period was extended to 8 h with 12 steps of depressurization performed on Type A hydrate. Detailed information on XCH 4 , RCH 4 , SCO 2 , ΔV, and ΔNR 2 can be found in Table S5 of Appendix A. The initial XCH 4 was 75.1 mol% at 44.2 bar before performing multistep depressurization. An immediate pressure rebounded and a slight pressure decrease were detected during stage A, with XCH 4 increasing to 78.6 mol%. This may be attributed to CO 2 -rich hydrate forming initially. As the pressure reduced close to P (CH4/CO2), XCH 4 decreased to 76.7 mol% and 74.3 mol % at the end of stage B and stage C, respectively. This may be due to the fact that CH 4 -rich hydrates were more likely to form than dissociation within a longer shut-in period. It was noted that a pressure differential existed during stage B and stage C, causing possible interactions of water and gas and heterogeneity of hydrate along with the sample.
When dissociation pressure was further decreased below P (CH4/CO2) and the pressure drop touched P CO2 since Stage F, a significant increase in XCH 4 was observed with a maximum value of 79.5 mol%, accompanied by the highest SCO 2 of 84.9% at the end of Stage F. Therefore, the operation time of 48 h was marked at the end of Stage F with the maximum XCH 4 and SCO 2 . The explanation for the above phenomenon was that massive CH 4 -rich hydrates dissociated CH 4 together with many CO 2 -rich hydrates formed at the pressure below P (CH4/CO2) . The following values of XCH 4 and SCO 2 declined with the next depressurization steps. RCH 4 , however, still grew to the highest value of 90.3% throughout continuous hydrate exploitation until the end of multistep depressurization. It was noted that water production started coming at the end of Stage H and multistep production should be creased. The negative value of ΔV was the indicator of CH 4 hydrate recovery overnumbered CO 2 hydrate storage. Additionally, the negative value of ΔNR 2 showed that overall hydrate saturation was decreased correspondingly. This may be because too much CH 4 was recovered while CO 2 was not stored equivalently with a longer shut-in period of 8 h. Table 3 summarizes the depressurization schemes and exploitation performances in Exp4, Exp7 and Exp8. It can be seen that different shutin periods of 2 h-4 h-8 h affect the exploitation performance of CH 4

Table 3
Summary of depressurization schemes and exploitation performances in experiments with shut-in period of 2 h, 4 h and 8 h at the targeted ceasing point, P out is the depressurization pressure, XCH 4 is the mole fraction of CH 4 in gas phase, RCH 4 is the CH 4 recovery percent, and SCO 2 is the CO 2 storage ratio. ΔV is the percentage change of hydrate volume and ΔNR 2 is the percentage change of normalized resistance. , which were nearly same and indicated that most CO 2 could form hydrates and remained stable during multistep depressurization at the pressure close to P (CH4/CO2) . This meant that the most suitable depressurization schemes achieved simultaneous the highest efficiency of CH 4 recovery and CO 2 storage at the targeted ceasing pressure, with the total production time were 10 h, 24 h and 48 h for Exp4, Exp7 and Exp8, respectively. The ΔV values for these three were almost same (− 65.71%, − 63.86% and − 73.91% for 2 h, 4 h and 8 h), indicating a total loss of hydrate volume. However, the value of ΔNR 2 was 10.2% for 4 h, which was higher than the negative values for 2 h (− 1.2%) and 8 h (− 6.7%). This means that hydrate-bearing sediment was improved through efficient gas water movement and distribution within core samples after multistep depressurization with 4-hour shut-in period. Fig. 17 compares the RCH 4 and SCO 2 in combination studies of CO 2 swapping with depressurization from references and Section 3.2.3 in this work. It can be inspected from Sun's work [31] that higher numbers of pressure drops with shorter shut-in period was more beneficial to CH 4 production and CO 2 retention. The SCO 2 in Yang's [41] was lower than those in Pandey's [45], which was resulted from CH 4 /CO 2 /N 2 mixed hydrate in Yang's and CH 4 /CO 2 mixed hydrates in Pandey's. This indicated that multistep depressurization was suggested to be performed after CO 2 injection in CH 4 hydrates for higher efficiency of CO 2 storage. It seemed that reducing shut-in period can improve CH 4 recovery and CO 2 storage slightly in terms of 24-hour and 12-hour cases in Pandey's. Nevertheless, the RCH 4 can be further enhanced from 71.0% to 73.8% to 75.0-79.4%, and SCO 2 from 81.0% to 82.9% to 82.8-85.1%, respectively through reducing shut-in period from 12 h to 24 h to 2-8 h. This indicated that sufficient water and gas migration can be completed within a much shorter shut-in period (2-8 h) for efficient CH 4 recovery and CO 2 storage. But too long shut-in period (12-24 h) without depressurization may cause less efficiency on CH 4 -rich hydrate decomposition and CO 2 -rich hydrate reformation than those with shorter ones (2-8 h). Fig. 18 summarizes the changes of XCH 4 , SCO 2 , and RCH 4 with P out during multistep depressurization at Exp1-8. The assumptions and equations used to calculate SCO 2 and RCH 4 were described in the Appendix B. Noted that the lowest P out was the point at which the multistep depressurization should be ceased as discussed above. As seen in Fig. 18 (a-c), the maximum values of XCH 4 for Type A hydrate were 74-82 mol %, while those for Type B hydrate were 37-44 mol%. The maximum value of XCH 4 (81.3 mol%) was observed in Exp7, accompanied by the peak SCO 2 at the end P out of 20.74 bar. No obvious differences in peaked SCO 2 values were observed between Type A hydrate (84-87%) and Type B hydrate (83-89%) in Fig. 18(d-f), indicating that majority of CO 2 in all scenarios can be stored as hydrate in sediment by multistep depressurization. And SCO 2 can be enhanced generally with the reduced depressurization pressure at the conditions that CO 2 -rich hydrates can form thermodynamically. According to Fig. 18(g-i), the maximum value of RCH 4 seemed to depend mainly on P out . When P out was decreased close to 10 bar, the maximum value of RCH 4 can be greater than 80% in Exp2, 3, 5-6. Compared to Exp1, 4, 7-8, when P out was reduced to about 20 bar, the maximum value of RCH 4 was less than 80%. This was following the fact that more CH 4 can be recovered with the increase of depressurization step and the decrease of depressurization pressure.

Comparison of efficiency and practical implication
When hydrates dissociate by depressurization in a porous medium, continuous mixing occurs between the liquid and gaseous phases due to capillary forces, differences in relative permeability, and mobilization of pore water. In a mixed hydrate system with two different hydrates (CH 4 and CO 2 hydrate of different composition), the CH 4 hydrate can remain stable outside its stability zone because it is surrounded by a more stable CO 2 hydrate layer. Therefore, the pressure response and production behaviors are influenced by the mass transfer barrier of the CO 2 hydrate as well as the presence of two stability zones, resulting in the likelihood of hydrate reformation. This gives the mixed hydrate system additional stability compared to a pure CH 4 hydrate system [46].
Generally, four types of dissociation and reformation characteristics during multistep depressurization mainly affecting CH 4 recovery and CO 2 storage can be recognized in this work, as summarized in Fig. 19. When the depressurization pressures were reduced below P (CH4/CO2) and above P (CO2) , obvious wanted CH 4 -rich hydrate dissociation and CO 2rich hydrate reformation can be observed according to the profiles of pressure rebound and sharp decrease. Gas concentration changes regarding XCH 4 and XCO 2 also supported the explanation above. A similar phenomenon of pressure rebound was reported in the schemes of slow stepwise depressurization [30] and intermittent depressurization [68], which significantly contributed to the CH 4 production. In addition, the pressure rebounded to just above P (CH4/CO2) at CH 4 -rich hydrate dissociation, indicating that further depressurization may improve CH 4 recovery continuously. Noted that the gas compositions were collected and acquired at the end of each multistep depressurization while this dynamic hydrate dissociation-reformation, as well as water-gas distribution, may occur at any time during the shut-in period. In contrast, when the depressurization pressures were operated below P (CO2) or above P (CH4/CO2) , unwanted CO 2 -rich hydrate dissociation or CH 4 -rich hydrate reformation were detected based on variations of pressure and CH 4 /CO 2 fraction. The reforming CH 4 -rich hydrates were not problematic because they can be recovered with the depressurization pressure reduced below P (CH4/CO2) . However, the dissociating CO 2 -rich hydrates should be avoided to decrease the efficiency of CH 4 recovery and CO 2 storage. These four types of pressure and gas composition characteristics can be employed to verify the occurrence of hydrate reformation and dissociation, thus enhancing CH 4 production and CO 2 retention via depressurizing stepdown pressure between P (CH4/CO2) and P (CO2) .
In a more realistic production field, CH 4 recovery and CO 2 storage should be considered comprehensively because the former is related to the direct profit of CH 4 product, and the latter can save undirected costs by CO 2 sequestration. These two indicators, therefore, are being maximized. Additionally, the production time for the multistep depressurization is critical concerning the costly operation cost on the platform. Meanwhile, the stability of hydrate-bearing sediment needs to be maintained through increased hydrate saturation or improved gas water distribution. Fig. 20 is the combined analysis of SCO 2 to increase of CH 4 concentration (ΔXCH 4 ), P out to P (CH4/CO2) , and production time to ΔNR 2 at the optimized conditions in this work. From the perspective of higher Fig. 17. Comparison of RCH 4 and SCO 2 in CO 2 swapping combined with depressurization from references and this work at different shut-in period [31,41,45]. More details can be referred to Table S6 in Appendix A. enhancement of CH 4 recovery and CO 2 storage, it can be speculated that Exp1, Exp4, Exp7, and Exp8 should be highlighted with the positive values of ΔXCH 4 (>0 mol%) and SCO 2 (＞84.0%) simultaneously shown in Fig. 20(a). These positive ΔXCH 4 and higher SCO 2 values were obtained by performing multistep depressurization until P out was close to or just below the equilibrium pressure of mixed CH 4 /CO 2 hydrates, shown in Fig. 20(b). Among these best four experiments, Exp7 emerged as the highest value ΔNR 2 (10.2%) with a shorter production time (24 h: 4-h shut-in × 6 stages) in total, as indicated in Fig. 20(c). It should be noted that Exp1 was also marked as its ΔNR 2 was of positive value (2.3%), however, the production time (32 h: 4-h shut-in × 8 stages) was relatively longer. Therefore, the 4-hour shut-in period may be the best option to enhance CH 4 recovery and CO 2 storage simultaneously during multistep depressurization in the CH 4 -rich hydrate reservoir.
The scheme of an injection well and a production well to explore NGHs by swapping CH 4 -CO 2 in the reservoir is highly recommended in the realistic production trial. Specifically, CO 2 is delivered to the reservoir from the injection well, while CH 4 -rich gas is collected from the production well by depressurization. Herein, CH 4 -rich hydrates around the production well tend to become CO 2 -rich by dissociating and reformation hydrates with the risk of water production. Furthermore, further depressurization in the production well may yield lean CH 4 gas effluent, which needs to be costly to acquire a CH 4 gas product. Our work is concerned with the above issues and determines the patterns of different multistep depressurization. A suitable operation strategy of a 4hour shut-in period with a production pressure close to P (CH4/CO2) is Fig. 18. Summary of dissociation characteristics during multistep depressurization at Type A hydrate (Exp1, 3,4,[7][8] and Type B hydrate (Exp2, 5-6) with the lowest P out as the ceasing point: (a-c) XCH 4 with P out ; (d-f) SCO 2 with P out and, (g-i) RCH 4 with P out . SCO 2 and RCH 4 were calculated using equations described in Appendix B.2.  proposed for the original CH 4 hydrate reservoir. After the CH 4 -rich hydrate reservoir is transformed into CO 2 -rich by depressurization exploitation, multistep depressurization on CO 2 -rich hydrates should be avoided from the economic point of view due to the lean effluent of CH 4 . Related to the initial water saturation, hydrate saturation appears to be less critical than operation modes and hydrate compositions, depending on the production performance. Furthermore, water production can signify massive dissociation of CO 2 hydrates, and therefore, multistep depressurization must cease before the pressure decreases close to P CO2 .
As shown in the ΔNR 2 results, either hydrate saturation or water gas distribution was improved overall after multistep depressurization. However, the distribution of water, gas and mixed hydrates needs to be further investigated to know their effects on sediment stability. A further investigation employing magnetic resonance imaging (MRI) or x-ray microcomputed tomography (CT) to characterize fluid distribution in the core needs to be combined to better understand multistep depressurization on mixed hydrates.

Conclusion
The multistep depressurization on CH 4 /CO 2 hydrates was intensively investigated to enhance CH 4 recovery and CO 2 storage. The effects of three critical parameters: hydrate compositions (Type A and Type B hydrate), residual water saturation (S rw ), and shut-in period (2 h, 4 h and 8 h) were studied to evaluate the gas compositions, CH 4 recovery and CO 2 storage efficiency, water gas distribution in hydrate-bearing sediment during multistep depressurization. The conclusions are summarized below.
• To maximize CH 4 recovery and CO 2 storage with higher CH 4 molar fraction in the gas phase (XCH 4 ) and CO 2 storage ratio (SCO 2 ), it was highly recommended to perform multistep depressurization on Type A hydrates (CH 4 /CO 2 ratio of 1.76-2.06) of low S rw (43.7-47.4%). • Depressurization was suggested reduced in steps and controlled at a certain pressure window to achieve wanted CH 4 -rich hydrate dissociation and CO 2 -rich hydrate reformation without water production. The final ceasing pressure was found close to the equilibrium pressure of CH 4 /CO 2 mixed hydrates concerning the initial gaseous CH 4 / CO 2 composition. • With combined consideration of efficiency and production time, the highest XCH 4 reached 81.3 mol% and the corresponding SCO 2 was 85.3% through 6-step depressurization with 4-hour shut-in period,

Table S3
Pressure, XCH 4 , RCH 4 , SCO 2 , ΔV and ΔNR 2 at different stages of multistep depressurization in Exp3-4. XCH 4 is the mole fraction of CH 4 in gas phase. RCH 4 is the CH 4 recovery percent. SCO 2 is the CO 2 storage ratio. ΔV is the percentage change of hydrate volume and ΔNR 2 is the percentage change of normalized resistance. The suggested ceasing stage was marked with bold. Exp     Considering CH 4 /CO 2 mixed hydrates occurs inside the core with a constant gas volume, moles of CH 4 (n CH4,f ) and CO 2 (n CO2,f ) in the gas phase after hydrate formation are calculated: n CH4,f = n mix,f × χ CH4,f (6) n CO2,f = n mix,f × χ CO2,f (7) n mix,f = P f V i Z f RT f (8) where χ CH4,f and χ CO2,f are the mole fraction of CH 4 and CO 2 gas components after hydrate formation. n mix,f Refers to the mole amount of residual mixed gas within the core, calculated from the same gas equation of state using Eq. (8).
In terms of the hydrate composition at the end of hydrate formation, the mole amount of CH 4 in the hydrate phase ( Δn CH4,H ) and that of CO 2 ( Δn CO2,H ) is given by Eqs. (9) and (10): Δn CO2,H = n CO2,i − n CO2,f

B.2. Hydrate dissociation
Three indicators are employed to evaluate the performance of mixed hydrate dissociation, i.e. CH 4 recovery percentage R CH4 (%), CO 2 storage ratio S CO2 (%), hydrate volume percent change ΔV (%), and normalized resistance percent change ΔNR 2 (%). The total CH 4 recovery percentage is calculated using the flowing equations based on different stages of mixed hydrate stability: n CH4,Re is the total CH 4 mole released during multistep depressurization, which can be obtained based on the mass balance and assumption calculation below: where ( Δn CH4,H + n CH4,f ) is the total mole of CH 4 at the beginning of multistep depressurization. Δn CH4,step and n CH4,v1 are the mole amount of CH 4 remaining in the hydrate and gas phases at the end of multistep depressurization, respectively. When the dissociation pressure is below the CH 4 hydrate equilibrium pressure, it can be assumed that all CH 4 are recovered, and no CH 4 are left in the hydrate phase (Δn CH4,step = 0) in terms of the CH 4 released during multistep depressurization. And n CH4,v1 can be obtained by: n CH4,v1 = y CH4,v1 P v1 V i Z v1 RT v1 (13) where y CH4,v1 is the mole fraction of CH 4 in the gas phase after the multistep depressurization. The corresponding total CO 2 storage ratio S CO2 (%) can be given: n CO2 ,St is the total CO 2 mole stored in hydrate during cyclic depressurization, which can also be obtained based on the mass balance calculation: where ( Δn CO2,H + n CO2,f ) is the total mole of CO 2 at the beginning of multistep depressurization.Δn CO2,re is the mole amount of CO 2 recovered together with CH 4 -rich gas. When the dissociation pressure is above the CO 2 hydrate equilibrium pressure, CO 2 gas is enriched in the hydrate phase as CO 2 hydrate reformation. It can be assumed that the recovered amount of CO 2 , together with CH 4 -rich gas released from CH 4 hydrate dissociation, can be ignorable Δn CO2,re = 0. n CO2,v1 is the mole amount of CO 2 in gas phase at the end of multistep depressurization given by: where y CO2,v1 is the mole fraction of CO 2 in the gas phase after the multistep depressurization. The hydrate volume percent change ΔV is introduced to investigate the percentage change of mixed hydrate mass, with a positive value of ΔV indicating hydrate mass increase and negative one means hydrate mass decrease. ) where V H is the hydrate volume before multistep depressurization, ΔV CO2(St) and ΔV CH4(Re) refer to the volumes of CO 2 stored, and CH 4 released, respectively and given by Equation (18)(19)(20): where M w is the mole mass of water (18 g/mol), and N is the hydrate number (6.00). And ρ CO2,H (0.9 g/cm 3 ) and ρ CH4,H (1.1 g/cm 3 ) are the density of CO 2 hydrate and CH 4 hydrate [1]. The increase of CH 4 concentration (Δ XCH 4 ) is calculated as followed: where χ CH4,lower is the mole fraction of CH 4 in the gas phase at the suggested ceasing point, χ CH4,f is the mole fraction of CH 4 in the gas phase before multistep depressurization. The normalized resistance percent change ΔNR 2 is defined as followed: where NR 2,f is the value of normalized resistance at the end of each stage and NR 2,i is the corresponding value before multistep depressurization.