A capillary high-pressure optical cell (HPOC) (Lu et al., 2007; Caumon et al., 2014; Ou et al., 2016) in combination with a Linkam CAP500 heating–cooling stage was used to simulate and observe the hydrate growing processes, supplied by gas migration in aqueous solution from the dissolution of free gas. The HPOC was constructed from a fused-silica capillary tube with an outer diameter (OD) of 375 μm, an inner diameter (ID) of 50 μm, and a length of around 25 cm. To prepare the sample, we loaded a section of ultrapure water (about 10 mm in length) into the sealed end of the capillary cell, then connected the other end to the pipeline with the feed gases and pressure pump. The capillary cell was inserted into the sample chamber of the heating–cooling stage for temperature control and Raman spectroscopic measurements, where the temperature could be maintained with an accuracy of ± 0.1 K from 273 to 373 K. The pressure in the cell could be adjusted by the pressure generator and measured by a full-scale pressure gauge (Setra 206 digital pressure transducer with a Datum 2000 manometer, accurate to ± 0.14%). The experimental apparatus has been reported in previous studies. (Lu et al., 2008; Guo et al., 2013; Lu et al., 2013).

The water used was ultra-purified in the laboratory with a resistivity of 18.24 MΩ·cm. Three different mixed CH₄ and C₂H₆ gas cylinders, obtained from Wuhan Iron & Steel Group Gas Co. Ltd, were used and the compositions (5.08 mol% C₂H₆ + 94.92 mol% CH₄, 10.18 mol% C₂H₆ + 89.82 mol% CH₄, and 20.12 mol% C₂H₆ + 79.88 mol% CH₄) were analyzed by gas chromatography before the experiment.

Figure 4 shows the relationship between the peak height ratio (HR) of methane and ethane and mole fraction ratios of the gases, and the calibration curve for the CH₄ and C₂H₆ system at 30 MPa and different temperatures that was then constructed. All data used in this figure were obtained at 273.15, 278.15, 283.15, 288.15, 293.15, and 298.15 K. The peak HRs of CH₄ and C₂H₆ were well fitted by a linear function of vapor mole fraction ratios (X(C₂H₆)/X(CH₄)), with a correlation coefficient over 0.9978, which showed good agreement with the results obtained by Subramanian (Subramanian et al., 2000). The calibration curve was further used to determine the composition of feed gas and the hydrate decomposition gases.

For each hydrate crystal, at least three locations were selected for Raman spectra measurements. Figure 5 shows the corrected representative Raman spectra of the mixed gas hydrates in the regions between 2,800 cm^–1 and 3,000 cm^–1 covering the C–H stretching vibrations in hydrates. The Raman spectra are reliable for the identification of hydrate structures (Subramanian et al., 2000), and the measured and assigned Raman band positions of C–H symmetric stretching are listed in Table 2 .

Logically, all the Raman characteristic signatures of C–H symmetric stretching for hydrates can be used as the judgment of hydrate structure, but the difference in the Raman shifts characteristics of CH₄ in sI and sII hydrates is very small, with only one difference of wavenumber. Therefore, the Raman stretching vibration band frequencies of C₂H₆ molecules in hydrates can be a useful index for determining the hydrate structures, with four differences of wavenumber (Subramanian et al., 2000). The ratio of the number of large cages (LC) to the number of small cages (SC) in sI and sII hydrates is 3:1 and 1:2, respectively. If all the cages are occupied, the Raman peaks area intensity ratio (PAR) of CH₄ in the LC to SC (I_MLC/I_MSC) should be 3 and 0.5 for sI and sII hydrates, respectively (Wei et al., 2021). Therefore, the above two Raman spectra characteristics can mutually support the hydrate structures (Klapp et al., 2010b).

Between 2,800 cm^–1 and 3,000 cm^–1, two patterns of Raman spectra can be distinguished. Figure 5A shows the first pattern of Raman spectra, that for hydrate samples A1, A2, A3, A4, and B2. There are two distinct strong peaks for CH₄ and a weak peak (ν₂) for C₂H₆ molecules (at about 2,946 cm^–1). The Raman spectra signature (ν₁), located at 2,891 cm^–1, was very feeble and could be fitted only with the use of software. Figure 5B shows the second pattern of Raman spectra, that for hydrate samples B1, B3, and B4. In this case, there are two distinct strong peaks for CH₄ molecules and two weak peaks (ν₁ and ν₂) for C₂H₆ molecules (at 2,886 cm^–1 and 2,942 cm^–1, respectively), with lower wavenumbers. Note that the Raman peaks for C₂H₆ molecules located at 2,891 cm^–1 and 2,946 cm^–1 are assigned to the C₂H₆ engaged in the large cavities of the sI hydrate, whereas the Raman peaks for the C₂H₆ molecules that occur 2,887 cm^–1 and 2,942 cm^–1 were reported by previous researchers to be an indication of C₂H₆ engaged in the large cavities of the sII hydrate (Murshed and Kuhs, 2009; Klapp et al., 2010b; Sum et al., 1997). Therefore, we considered hydrate samples A1, A2, A3, A4, and B2, exhibiting the first pattern of Raman spectra, to be sI hydrates, and hydrate samples B1, B3, and B4, exhibiting the second pattern of Raman spectra, to be sII hydrates. The weak signature of C–H shifts of ethane for sI hydrates might be due to the relatively low cage occupancies of ethane in large cages, which could lead to the deviation of the Raman spectra signature (ν₁) in the fitting procedure. The shapes of the Raman spectra obtained for sII hydrates in this work are consistent with that of the methane and ethane hydrate formed in Monterey Bay (1,024 m, 278 K) (Hester et al., 2006).

In addition, Raman spectra of the ν₁ and ν₂ C–H stretch of C₂H₆ for mixed gas hydrates and Raman PARs of CH₄ in the LC to SC (I_MLC/I_MSC) are demonstrated in Figure 6 , which corresponds to the Raman spectra of the first and second patterns in Figures 5A, B . The Raman peak positions of ethane in C–H regions for sII hydrates were lower than those of ethane for sI hydrates ( Table 2 ). The Raman PARs of CH₄ in the LC to SC (I_MLC/I_MSC) for hydrate samples A1–A4 and B2 were 3.41, 3.26, 3.11, 3.02, and 3.13, respectively, and I_MLC/I_MSC for hydrate samples B1, B3, B4, and C1 was 0.41, 0.24, 0.24, and 0.17, respectively. Hence, we concluded that the hydrate samples A1–A4 and B2 were sI hydrates, and hydrate samples B1, B3, B4, and C1 were sII hydrates. The Raman PAR of CH₄ in the LC to SC for sI hydrates (red point in Figure 6 ) was greater than 3.0, which might be attributed to relatively high occupancies of methane in large cages for sI hydrates. Meanwhile, I_MLC/I_MSC for sI hydrates decreased from 3.4 to 3.0 with a rise in temperature, which indicated that ethane had a preference to enter the large cages for sI hydrates with increasing temperature. By contrast, I_MLC/I_MSC for sII hydrates was smaller than 0.5; a reasonable explanation is that the relatively high cage occupancies of ethane for sII hydrates led to a decrease in the occupancies of methane in large cages, and the small cages were almost fully occupied by methane. From the limited data available for sII hydrate samples, no distinct conclusions could be drawn on the effect of temperature on the I_MLC/I_MSC for sII hydrates, but it could be seen that there was a significant decrease in I_MLC/I_MSC, from 0.41 to 0.24, when the temperature changed from 278.15 to 288.15 K, which was probably caused by the decrease in cage occupancies for methane in small cages.

We calculated the cage occupancies of hydrates and hydration numbers ( Table 3 ) using Raman spectroscopic analysis (Sum et al., 1997; Subramanian et al., 2000; Kumar et al., 2008). The calculated hydration numbers are in agreement with previously reported values, which range from 5.8 to 6.3 (Ripmeester and Ratcliffe, 1988).

In theory, CH₄ can occupy all the large cavities in all hydrate structures (Sum et al., 1997) whereas C₂H₆ molecules can be included only in large cages (Uchida et al., 2007). In the sI hydrate samples A1–A4 and B2, θ _L–M,sI (CH₄ cage occupancies for sI LC) was equal to, or was close to, θ _S–M,sI (CH₄ cage occupancies for sI SC) and was higher than θ _L–M,sII (CH₄ cage occupancies for sII LC) in sII hydrate samples B1, B3, and B4, which indicated that CH₄ had an advantage over C₂H₆ in the competition for inclusion in the large cage for sI hydrates. In sII hydrate samples B1, B3, and B4, the small cages were almost fully occupied by CH₄, and θ _S–M,sII (CH₄ cage occupancies for sII SC) was higher than θ _S–M,sI . Meanwhile, the cage occupancies of C₂H₆ for the sII hydrates samples suggested that θ _L–E,sII (C₂H₆ cage occupancies for sII LC) was also higher than that for sI hydrates, which might be the reason for the weak Raman peaks (ν₁ and ν₂) for C₂H₆ in sI hydrates. This cage occupancies phenomenon could reasonably be explained by the guest-to-cavity ratio proposed by Lederhos et al. (Lederhos et al., 1993). As temperatures increased, the cage occupancies of C₂H₆ gradually increased; the cage occupancies of CH₄ in large cages were the opposite, which can be attributed to the contribution of ethane to cavity stabilization.

Under the dissolution–migration–hydration processes, the feeding gas source components affect both the composition and structure of the hydrates. The sII hydrates were formed only in the experiment in which the feed gas components contained at least 20.12 mol% C₂H₆. With this composition of feed gas, sample B2 formed sI hydrate, and other samples, B1, B3, and B4, all formed sII hydrate. At a constant temperature and pressure, we found that the higher the proportion of C₂H₆ in the feed gas, the higher the concentration of C₂H₆ in the hydrates ( Table 3 ).

An important discovery was that the C₂H₆ concentration in hydrate formed at 30 MPa was depleted compared with the feed gas. For example, in samples A1–A4 and B1–B4, the proportion of C₂H₆ in the hydrates was lower than that in feed gas; methane-to-ethane ratios (C₁/C₂) in hydrates formed in the dissolution–migration–hydration processes were 1.2 to 2.1 times that of the gas source; this contrasts with the findings of previous studies reporting heavy hydrocarbon enrichment in hydrates crystallized directly from free gases (Kumar et al., 2008; Seo et al., 2009).

Under the dissolution–migration–hydration processes, temperature has a significant effect on hydrate composition. For example, by comparing sI hydrate samples A1 (hydrated at 278 K) and A4 (hydrated at 293 K), and sII hydrates B1 (hydrated at 278 K), B3 (hydrated at 288 K), and B4 (hydrated at 293 K), we found that the proportion of C₂H₆ in sI hydrate and sII hydrates increased with temperature by 49.6% and 24.6%, respectively ( Table 3 ). The trend of the proportion of C₂H₆ in hydrate increasing with temperature is consistent with the core data, where temperature increases with the depth under geothermal gradients. For example, core data from samples from the South China Sea show that the C₁/C₂ ratio in hydrates gradually tends toward a low value with depth (Paganoni et al., 2016), and core data in many ocean drilling programs (ODP)/International Ocean Discovery Program (IODP) sites (Milkov et al., 2004).

As the core geochemical data indicated (Milkov et al., 2004; Paganoni et al., 2016), an abrupt decrease in C₁/C₂₊ ratios in headspace gas in sediments occurred around the BSR, and C₁/C₂₊ ratios in the hydrates, sampled at the South China Sea (Zhang et al., 2019; Wei et al., 2021), tended to be lower at greater depth. Previous researchers have suggested two possible mechanisms to explain this phenomenon. One possible mechanism is the biodegradation of heavy hydrocarbons by microorganisms as sediment depth decreases (Pape et al., 2010), and the other is the addition of content of heavy hydrocarbons in the deep strata (Sassen et al., 2001). Neither mechanism can explain why there was an abrupt decrease in C₁/C₂₊ ratios in headspace gas in sediments occurring around the BSR. There should be a specific process only near the depth around the BSR, and the process should be related to the BGHSZ, the boundary of the hydrate-occurring zone, and the free gas zone. According to our experimental observation, the molar fraction of C₂H₆ in hydrate is depleted compared with gas sources. Gases partitioning occurred throughout the reaction transport process. We propose that the dissolution–migration–hydration of gases is the mechanism that enriches ethane near the BSR.

Paull (Paull et al., 1993) proposed a mechanism by which gas could be concentrated by recycling at the BGHSZ. Progressive burial and subsidence through geologic time shifts the BGHSZ upward, such that deep-seated hydrate decomposes and the free gas zone also shifts upward. Some free gas could rise buoyantly upward (permeating fissures in the overlying hydrate stability layer) and be incorporated into the hydrate layer (Cheng et al., 2020; Bello-Palacios et al., 2022). However, as proposed by Brown (Brown et al., 1996), significant residual free gas remains below the gas hydrate stability zone (GHSZ), trapped there by capillary forces. The bubbles in the lower portion of the free gas zone must continually dissolve, and methane and ethane must be transported in pore fluids. Our experimental observations show that gas fractionation occurred between the hydrate and the gas source under “dissolution–migration–hydration” processes. This may cause a larger C₁/C₂₊ ratio in gas hydrates relative to the residual free gas below the BGHSZ, which causes the abrupt decrease in C₁/C₂₊ ratios in headspace gas in sediments occurring around the BSR.

To test the “dissolution–migration mechanism” and its links to the enrichment of ethane near the BSR and increasing upward C₁/C₂ ratios in the hydrate-occurring zone, our experimental observation simulated these processes in a limited time and space. However, dissolved gas migration and gas hydrate formation in marine sediment usually take place over days, or even hundreds to thousands of years. To study the gas fractionation mechanism during the dissolution–migration–hydration processes, and extend our knowledge to a geologic time and space scale, we first compared the dissolution and diffusion properties of methane and ethane at the typical temperatures and pressures in the GHSZ ( Table 4 ). We then simulated the fractionation during the dissolution–diffusion processes ( Figure 7 ) and at last discussed the fractionation during the hydration.

Dissolution can cause gas fractionation in underwater conditions. Ethane is less soluble than methane at the same temperature and pressure; for example, at 283.15 K and 30 MPa, the solubility of methane and ethane (the mole fraction of gases in the water) is 0.00417 and 0.00132, respectively (Li et al., 2015), which indicates that in the same conditions methane is around 3.2 times more soluble than ethane. For binary-component gas mixtures, the phase equilibrium data and models are very limited. We calculated the C₁/C₂ ratio of methane and ethane in an aqueous solution after dissolution at 30 MPa and at different temperatures with a thermodynamic model (Li et al., 2015): the results show that the C₁/C₂ ratios increased by around 2.3 to 2.5 times in the aqueous solution ( Table 4 ). From these calculations and comparisons, we can see that, under submerged conditions, dissolution causes gas fractionation and increases the C₁/C₂ ratio in aqueous solution approximately 2.3 to 3.2 times relative to the vapor gas source.

Diffusion is a dominated process that causes gas fractionation in the geological system. The dissolution–diffusion of gas in the aqueous phase can be treated as a one-dimensional diffusion process because the gas migrates up from the bottom along the long fluid migration pathway. For ideal solutions, the variation of concentration with time is subject to Fick’s second law. The bottom gas concentration, near the BSR, in the aqueous phase is constant (theoretically equal to the solubility of the gas in the water at the specific temperature and pressure). Under infinite boundary conditions along the diffusion path, the dissolution–migration of gas can be calculated by the diffusion model (Lu et al., 2006; Guo et al., 2013). We simulated the dissolution–migration process of a methane–ethane mixture (C₁ = 90%, C₂ = 10%), and calculated the C₁/C₂ ratio at a depth of around 1 m above the BSR under dissolution–diffusion processes at 30 MPa, and 278.15 and 288.15 K. The entire duration of the numerical simulation was 10 days. Figure 7 shows that the dissolution–diffusion process can produce significant ethane depletion compared with the gas source; this results in the ratio of C₁/C₂ increasing by several orders of magnitude with distance upward. Decreasing temperature increases the difference in diffusion coefficient between methane and ethane ( Table 4 ), and increases gas fractionation during the dissolution–migration processes ( Figure 7 ); the numerical simulation results are generally consistent with our experimental observations. Advection of methane-bearing fluids is an important means of gas migration in hydrate-occurring zones. When the pore is under low fluid flux, diffusion is the dominant mechanism that causes gas fractionation, and can be further enhanced by advection. Multi-component, multi-phase transport–reaction processes related to the hydrate formation are a complex issue; we will investigate the effect of such processes (as well as advection) on hydrate formation and gas fractionation in the near future.

Hydration is also a process that causes gas fractionation. Hydrates grown from aqueous solutions usually enrich ethane relative to methane, as well as in submarine conditions. Data from ODP site 1245 show that the C₁/C₂ ratios in the pore solution are around 3.3 to 4.4 times higher than the C₁/C₂ ratios in hydrates for the samples at the depths of 50 to 100 mbsf (Milkov et al., 2004). It is clear that gas fractionation between the aqueous and hydrate phases is affected by the temperature–pressure–salinity conditions and dissolved gas composition. For example, the thermodynamic model developed by Velaga (Velaga et al., 2016) for predicting the aqueous solubility of hydrocarbon mixtures at the two-phase hydrate–liquid water equilibrium shows that the presence of propane in the mixture affects the aqueous solubility of hydrocarbon mixtures at the two-phase hydrate–liquid water equilibrium. More experiments are needed to verify such models.

Our study shows that gas fractionation occurring between the hydrate and the gas source under “dissolution–migration–hydration” processes may cause larger C₁/C₂₊ ratios in gas hydrates relative to the residual free gas below the BGHSZ. Such processes cause the abrupt decrease in C₁/C₂₊ ratios in headspace gas in sediments that occur around the BSR ( Figure 8 ).

Hydrate accumulations originating from thermogenic gases containing heavier hydrocarbons have been found in many regions around the world (Hillman et al., 2017; Plaza-Faverola et al., 2017; Thiagarajan et al., 2020). Under the effect of the formation pressure, the thermogenic gas sources with higher-order hydrocarbons gradually migrate upward into the hydrate stability zones through the preferential transport pathway, which includes gas chimneys, unconformity surfaces, faults, fractures, and sandstone channels, and then form a hydrate layer at the suitable pore sizes (Panieri et al., 2017; Portnov et al., 2021; Liang et al., 2022).

However, not all free gas can be transported into the hydrate stability zone. When the gas pressure cannot overcome the capillary pressure, the gas phase will be trapped in sediments in the free gas zone, and must continually dissolve and be transported upward in solution to be incorporated into the hydrate layer. Our experiments of gas hydrate formation in capillary tubes are highly analogous to the process of natural gas hydration in the pore space in submarine sediments. Gas dissolves and migrates along the capillary pore space through short or long distances, and then forms hydrates. Such “dissolution–migration–hydration” processes result in the percentage of ethane in the hydrate increasing with temperature and depth ( Figure 8 ), especially in sites where thermogenic gas is rapidly transported along a gas chimney and the local geothermal gradient is increased by the warm fluids.

Previously, the amount of carbon contained in hydrates formed from heavy hydrocarbons, such as ethane, was ignored, resulting in a significant underestimation of the number of carbon resources. Essentially, the estimated amounts of carbon in gas hydrates are constrained by the volume of pore space available for hydrate formation and the gas composition in the hydrate phase (Dickens, 2011; Wallmann et al., 2012), and 1 m³ of ethane hydrate contains twice the amount of carbon in a similar volume of methane hydrate (Plaza-Faverola et al., 2017).

Dissolution–migration of gas near the BSR enriches ethane near the GHSZ; such changes in gas composition could affect the depth of the GHSZ. Our calculation indicates that an increase of 10% in the mole fraction of ethane could deepen the GHSZ thickness by at least 37 m (compared with pure methane hydrate; Figure 8 ). The higher the proportion of heavier hydrocarbons, the deeper the bottom boundary of the gas hydrate stability zone ( Figure 8 ). Our experiments indicated that the proportion of heavy hydrocarbon in the bottom hydrate (9.77% at 293 K; Table 3 ) is 49.6% higher than the proportion of heavy hydrocarbon in the top hydrate layer (6.53% at 278 K; Table 3 ). Assuming 6.53% ethane in the hydrate, this could lead to the amount of carbon per 1 m³ being underestimated by at least 2.42-fold (considering the 37 m deep thickness of the GHSZ, and comparing with hydrate accumulated from pure methane).