Department of Civil and Environmental Engineering, Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Korea
Corresponding author: Gye-Chun Cho, Graduate Student, Department of Civil and Environmental Engineering, Korea Advanced Institute of Science and Technology (KAIST), Daejeon 305–701, Korea. (firstname.lastname@example.org)
 This study examines the effect of gas hydrate formation on seismic wave velocities of fine-grained sediments. Synthesis of gas hydrates in fine-grained sediments has proved to be challenging, and how hydrate formation would affect the seismic wave velocities and stiffness of clay-rich sediments has not yet been fully understood. In this study, CO2 hydrate was synthesized in remolded and partially water-saturated clayey silt sediments that were originally cored from a hydrate occurrence region in the Ulleung Basin, East Sea, offshore Korea. After achieving excess water conditions, compressional wave and shear wave velocities were measured for different hydrate saturations and under different vertical effective stresses. The results reveal that the compressional wave velocity VP and shear wave velocity VS increase, and the stress-dependency of VP and VS decreases as the hydrate saturation SH increases from 0% to ~60%. In particular, the VS-SH trend lies between the grain-cementing model and the load-bearing model, suggesting that gas hydrate formation in clayey silt sediments causes weak cementation from a hydrate saturation less than ~28%. The weak cementation in fine-grained sediments can be explained by the breakage of hydrate bonds that are cementing grains during sediment compression and/or the innate weakness in bonding between hydrate crystals and fine mineral grains owing to the presence of unfrozen water films on clay mineral surfaces. In addition, it is found that at low SH, the cementation effect on VP is masked by the high stiffness of pore-filling phases, but it becomes pronounced at SH greater than 47%.
 Gas hydrates are ice-like crystalline solid compounds composed of hydrogen-bonded water cages that encapsulate guest molecules such as light hydrocarbons or carbon dioxide. Since their discovery in permafrost regions and in deep marine sediment formations, natural gas hydrates have aroused interest as potential new energy resources. However, these compounds — primarily, methane (CH4) hydrate — have been identified as potential geohazards and contributors to global warming. Therefore, understanding of the physical properties of hydrate-bearing sediments has become an important research topic for interpretation of in situ geophysical data, stability analyses of boreholes and seafloors, and reservoir simulations for gas production from hydrate deposits. Considerable interest continues in carbon dioxide (CO2) hydrate and CO2 hydrate-bearing sediments, both of which are potential byproducts of geologic carbon storage in cold aquifers. The self-sealing capability of CO2 during its hydrate formation within cold sediment pores can be exploited to provide an extra seal to natural geologic seals [Koide et al., 1997; House et al., 2006; Kvamme et al., 2007; Tohidi et al., 2010].
 However, for successful applications of gas production from gas hydrate-bearing deposits and CO2 hydrate formation in CO2-injected deposits, detection and monitoring of the regions undergoing those processes are critical. Techniques based on seismic waves (e.g., seismic survey methods and sonic logging) appear to be one of the most appropriate options for monitoring these processes in sediments. Seismic wave velocities are also analogous to soil stiffness [Clayton et al., 2005] and used as essential input parameters for numerical simulations of CH4 hydrate production and CO2 sequestration using CO2 hydrate.
 Naturally occurring gas hydrates exhibit a wide range of formation morphologies from disseminated, pore-filling hydrates in sandy sediments to vein- or lens-shaped grain-displacing hydrates in fine-grained sediments [Waite et al., 2009; Rees et al., 2011]. Meanwhile, previous studies have demonstrated that seismic wave velocities of hydrate-bearing sediments are strongly influenced by the hydrate-forming methods and the resulting hydrate loci in sediment pores, e.g., pore-filling, load-bearing, grain-cementing, and grain-displacing hydrate [Winters et al., 2004; Santamarina and Ruppel, 2008; Priest et al., 2009; Waite et al., 2009]. For instance, when gas hydrate forms under an excess gas condition, the grain-cementing hydrate formation is known to be predominant [Priest et al., 2005, 2009]. On the other hand, the hydrate formed from dissolved gas in an aqueous phase or under an excess water condition in sandy sediments can be characterized with a pore-filling model when hydrate saturation is less than 20–40%, and with load-bearing or grain-cementing models when hydrate saturation exceeds ~40% [Winters et al., 2004; Kleinberg and Dai, 2005; Yun et al., 2005, 2007; Spangenberg et al., 2005, 2008; Dai et al., 2008; Lee and Waite, 2008; Priest et al., 2009; Waite et al., 2009; Lee et al., 2010].
 Host sediments containing natural gas hydrates have been often reported as fine-grained sediments [Nimblett and Ruppel, 2003; Francisca et al., 2005; Kwon et al., 2011]; however, synthesizing gas hydrates in fine-grained sediments in the laboratory has proved to be challenging because of the low permeability and small pore sizes of the sediments. As hydrate formation is mostly controlled by gas diffusion, the conditions encountered in fine-grained sediments often lead to grain-displacing hydrate formation, even at the core scale (e.g., X-ray CT images of pressure core samples in Holland et al., 2008 and Rees et al., 2011). These heterogeneously distributed hydrates in sediments pose an important challenge for the correct interpretation of geophysical measurements. Studies on geophysical and strength properties of hydrate-bearing fine-grained sediments were carried out by synthesizing tetrahydrofuran (THF) hydrate in silts and kaolinite clay [Yun et al., 2007; Santamarina and Ruppel, 2008; Lee et al., 2010] and in natural cores from the Gulf of Mexico [Lee et al., 2008] as host sediments. However, little effort has yet been made to measure seismic velocities of gas hydrate-bearing fine-grained sediments in a low-to-medium hydrate saturation regime. Moreover, the cementation effect of hydrate formation on the seismic velocities and mechanical stiffness of fine-grained sediments remains poorly understood. This knowledge gap is hampering the reliable estimation of gas hydrate saturation in such sediments and the calibration of logging and seismic exploration results acquired in hydrate occurrence regions. Considering the complex nature of the hydrate formation process in fine-grained sediments, a key challenge appears to be the design of well-controlled laboratory experiments that can control hydrate quantity, achieve excess water conditions, and monitor seismic responses.
 This study presents the seismic wave velocities of fine-grained sediments containing CO2 hydrate. CO2 hydrate was synthesized with three different hydrate saturations from sediment cored from a hydrate occurrence region in the Ulleung Basin, East Sea, offshore Korea. The remolded and prewetted specimens were brought into hydrate stability conditions to form hydrates and were saturated with water in an attempt to achieve an excess water condition. CO2 hydrate is presumed to be a suitable analogue for CH4 hydrate in terms of seismic velocity because different guest molecules in the hydrates have minimal impact on the mechanical stiffness of gas hydrates and hydrate-bearing sediments [Kiefte et al., 1985; Helgerud et al., 2003, 2009]. On the other hand, hydrate growth morphology has significant impact on the seismic velocities of hydrate-bearing sediments. In addition, the structural similarity between CH4 hydrate and CO2 hydrate as a structure I hydrate [Lee et al., 2003; Sloan and Koh, 2008] also supports the analogy between CO2 hydrate and CH4 hydrate. The compressional wave velocity (P-wave velocity; VP) and shear wave velocity (S-wave velocity; VS) of specimens were measured at ultrasonic frequency ranges (~several hundreds kHz) during the loading process, and the effects of loading and hydrate saturation on seismic wave velocities in fine-grained sediments were discussed.
2 Experimental Program
2.1 Description of the Sediment Used
 The 2007 UBGH1 hydrate drilling expedition program selected three sites — UBGH1-4, UBGH1-9, and UBGH1-10 — for drilling in the Ulleung Basin, East Sea, Korea. The Ulleung Basin is a deep, bowl-shaped, back-arc basin located in the southwestern part of the East Sea. The water depth of the basin ranges from 1500 m to 2300 m [Lee and Kim, 2002]. Natural gas hydrates were found to occur between the seafloor and 141 mbsf at hole UBGH1-10B [Bahk et al., 2009; Kwon et al., 2011]. Natural sediment retrieved from a depth of 121 mbsf at hole 10B (referred to as 10B-15H-2b) was chosen for this study. The conventional Fugro Hydraulic Piston Corer was used to recover the sediment core such that no hydrate was preserved. The sediment core sample that had been initially preserved in a plastic liner after being retrieved from the site was tested for basic index and geotechnical properties [see Kwon et al., 2011]. In this study, the same sediment sample was remolded and used for the experiments.
 The sediment sample was tested to determine its physical properties, as summarized in Table 1. Atterberg limits (i.e., plastic and liquid limits; ASTM D-4318) were obtained from an oven-dried sample, and the sample was classified with respect to the unified soil classification system (USCS). Specific surface area was measured by the methylene blue absorption method [Santamarina et al., 2002]. Particle size distribution was determined by a laser diffraction particle analyzer (Beckman Coulter LS 13 320). The sample was characterized as a highly plastic silty soil (MH by USCS distinction), and it had a high specific surface area of 98 m2/g and a mean particle size of 9.4 µm. Particle size distribution analysis revealed that the sediment mainly consisted of silt and clay and contained sand of ~5% as a mass fraction (Figure 1). The clay fraction (~20%), identified as kaolinite and illite, was determined through X-ray diffraction (XRD) analysis (see Supporting Information,Table S1) In addition, scanning electron microscopy images revealed the presence of diatoms (Figure 2). Small platy clayey particles, such as kaolinite and illite, were observed as being attached to moscovite and quartz minerals (Figure 2a). Diatomatious minerals were also observed as shown in Figure 2b. Small pores on these diatom minerals and their skeletal structures resulted in high specific surface area, which led to high levels of water absorption, compressibility, and plasticity [Shiwakoti et al., 2002; Hong et al., 2006; Kwon et al., 2011]. Further information on physical and geotechnical properties of the sediment core sample are detailed in Kwon et al. .
Table 1. Physical Properties of the Sediment Sample Tested
USGS represents the unified soil classification system. Particle sizes mostly range 1–20 µm (see Figure 1).
Specific surface area was determined by the methylene blue absorption method [Santamarina et al., 2002]. A XRD analysis revealed that the sediment sample consists of 24.1% quartz, 24% of muscovite (also known as mica), 17.8% albite (sodium feldspar, NaAlSi3O8), 9% orthoclase (potassium feldspar, KAlSi3O8), 9.4% kaolinite, and 15.6% illite (see Table S1).
 The apparatus was designed and built to synthesize CO2 hydrate in sediments and to measure geotechnical and geophysical properties under zero lateral strain conditions (Figure 3). All experiments were conducted in a cylindrical, rigid-wall high-pressure cell with a volume of 160.8 cm3, internal diameter of 66 mm, and internal height of 140 mm. Effective stress, which is an overburden pressure on the soil skeleton, was applied by a loading plate, the force of which was controlled by an external air cylinder. The cell hosted one T-type thermocouple for measuring the temperature of the specimen interior, one pressure transducer for measuring pressure, one pair of bender elements for shear wave (S-wave) measurement, one pair of piezoelectric ceramic (lead zirconate titanate, PZT) disks for compressional wave (P-wave) measurement, and one pair of electrodes for electrical resistivity measurement. Digital pictures of the experiment apparatus are provided in Figure S1 in the Supporting Information.
 The high-pressure reaction cell was submerged in a bath with temperature controlled by circulating fluids from a refrigerator. Five-millimeter-long bender elements were used for S-wave measurement, while 10 mm-diameter PZT disks were used for P-wave measurements (see Figure S1). Square-shaped signals with amplitude of 10 V were used for excitation, and the input frequency ranged from 1 kHz to 10 kHz.
2.3 Experimental Procedure
 In nature, most of the gas hydrates found in marine settings are presumed to form from dissolved gases in an aqueous phase [Hyndman and Davis, 1992, Soloview and Ginsburg, 1994; Xu and Ruppel, 1999; Garg et al., 2008]. However, development of laboratory techniques that can efficiently form gas hydrates from dissolved gases and consistently produce gas hydrates in fine-grained sediment samples at desirable saturations presents challenges. Although there have been a few successful attempts to form hydrates from dissolved gases in sands or glassbead packs [Buffett and Zatsepina, 2000; Spangenberg et al., 2005, 2008] or from excess water conditions in sands [Priest et al., 2009], no data has been reported for fine-grained porous media, with the exception of THF hydrate [Santamarina and Ruppel, 2008; Lee et al., 2010 The present study documents the first successful attempt to form CO2 hydrate in fine-grained sediments. CO2 hydrate was formed from sediments partially saturated with water, and further water was injected to achieve natural excess water settings. The detailed preparation procedures for remolding the sediment sample, synthesizing hydrate, and preparing the hydrate-bearing water-saturated sample are reported in the following subsections.
2.3.1 Remolding Procedure
 A hydrate-free sediment core sample, retrieved from hole UBGH1-10B, was desalinated with distilled water to exclude the effect of salinity on hydrate formation, and it was oven-dried for more than 2 days at 80°C before being remolded. The dry sediment sample was thoroughly mixed with distilled water in a sealed plastic bag. The masses of dry sediment and water were determined to achieve preselected water saturations and to eventually control gas hydrate saturations. More than 24 h were allowed for the sediment–water mixture to become homogeneous in the sealed plastic bag. The remolding procedure was carried out under atmospheric pressure and ambient temperature, such that the pore spaces were filled with air and distilled water. Partially water-saturated samples with water contents (i.e., mass of water divided by mass of dry soil) of 0.16, 0.23, and 0.31 were prepared. Thereafter, the prewetted sediment was compacted into the cell by hand-tamping, and a vertical effective stress of ~100 kPa was applied. The specimen height was monitored using a dial gauge. The resulting porosity and water saturation (Sw) of the pore space, evaluated by measuring the specimen volume, were 0.64, 0.61, and 0.61 and 22.7%, 38.2%, and 51.0%, respectively.
2.3.2 Procedure for Hydrate Formation
 CO2 gas was introduced into the partially water-saturated specimens up to a pressure of 2.8–3.1 MPa. Because the initial air saturations of the three specimens before injecting CO2 were larger than 49%, the air phase is presumed to have percolated throughout the specimens while the water phase tends to form menisci at grain contacts [Fredlund and Rahardjo, 1993; Sahimi, 1994; Leverson and Lohnes, 1995]. Accordingly, injected CO2 gas was expected to occupy the rest of the pores, homogeneously distributed in the specimens. The cell was subsequently cooled to approximately 1–2°C to form hydrate. This induced hydrate nucleation in the samples, with pore pressure kept at ~3 MPa through supplying CO2 gas. The temperature, pressure, and electrical resistivity were recorded every 10 s by a data logger over the course of the experiments (Figure 4). The seismic wave signatures for both VP and VS were recorded every 60 min during the cooling process. After a temperature peak which marked an exothermic reaction of hydrate formation was observed, seismic wave signatures were captured every 30 min until VP and VS showed no change. The completion of hydrate formation was inferred when there was no change in the captured geophysical signatures. 24 h to 60 h were allowed for the samples to form hydrate at their respective pressures and temperatures. Because CO2 molecules are trapped in water-bonded cages, which expand the volume during phase transformation, the resulting hydrate saturation is generally larger than the initial water saturation. The hydration number, or stoichiometric number, determines the volume expansion and the density of formed hydrates. Assuming a 100% phase transformation of water to hydrate, hydration number of 6.6, and CO2 hydrate density of 1.1 g/cm3 [Aya et al., 1997], the hydrate saturations of the specimens (SH), defined as hydrate volume divided by total pore volume, were expected to be approximately 28%, 47%, and 63%, respectively.
2.3.3 Procedure for Postwater Saturation
 The specimen was flushed with nitrogen (N2) gas for 0.8–1.2 pore volume at the sample temperature, which displaced the CO2 gas remaining in the pores of the sediment and tubes of the testing system, in order to avoid additional CO2 hydrate formation during the following water injection. N2 gas was injected at the base of the specimen, and at the same time, a CO2 and N2 gas mixture was released through an additional fluid port at the top of the cell. The flow rate of the CO2 and N2 gas mixture at the outlet port of the top of the cell was controlled at a rate of ~1 mL/min by using a precision metering valve. This CO2 displacement with N2 was continued for approximately 60 min, and the pressure inside the cell was maintained at ~3 MPa. Although a temporary decrease in partial pressure of CO2 in the specimen could trigger hydrate dissociation, the electrical resistivity, VP, and VS values were fairly constant during this procedure. Thus, it was presumed that the time taken from the N2 injection to the following water injection was insufficient for dissociating considerable amounts of hydrate. In addition, because the cell pressure (~3 MPa) was twice as large as the equilibrium pressure of CO2 hydrate at 1–2°C (i.e., approximately 1.41–1.58 MPa; Sloan and Koh, 2008), such N2 flushing for 0.8–1.2 pore volume would not reduce the partial pressure of CO2 far less than the equilibrium pressure. It is also possible to have CO2 gas locked in pores, which would not be affected by the N2 flushing.
 Distilled water was then introduced at the same temperature to saturate the specimen and to achieve excess water conditions without a free gas phase in the pores (i.e., postwater saturation). A bottle filled with distilled water was connected to a water pump and the reaction cell. As the precision metering valve at the drainage fluid port at the top of the cell gently unlocked, the N2 and CO2 gas mixture remaining in the sediment pores was drained, and at the same time, water began to flow into the cell, filling the pores from the bottom of the specimen. When injecting water, the pressures of the water bottle and reaction cell were maintained at ~2.8–3.1 MPa by the water pump. The flow rate of the water was controlled by the precision metering valve. Total 200 mL of water was injected for 10–12 h at a flow rate of ~0.3 mL/min, thus circulating approximately 2.7, 3.8, and 7.5 pore volumes for the specimens with SH = 28%, 47%, and 63%, respectively. As soon as water was introduced from the bottom, a decrease in electrical resistivity was observed because of capillary rise of water (Figure 4b). Thereafter, as water displaced or dissolved gas, VP increased significantly, which confirmed the elevated water saturation in the specimens (Figure 4c). No or minimal vertical displacement was measured during the water injection. After the water injection, the specimen was held for more than 10 h until the geophysical signatures showed no change (Figure 4). Dissolution and re-formation of CO2 hydrate in sediments were presumed to occur because the injected water was under-saturated with CO2. The effect of these processes on geophysical signatures will be discussed further in the following section. Thereafter, a loading process was commenced, in which the vertical effective stress was increased while monitoring the pressure, temperature, vertical displacement, seismic wave velocities, and electrical resistivity.
 A complimentary experiment was performed to calculate the amount of hydrate formed. The specimen with SH = 47% was thermally dissociated while CO2 gas was collected. The hydrate saturation was estimated to be approximately 38%, which is smaller than the initial target hydrate saturation of 47%. This ~9% difference in the hydrate saturation estimation is expected to be caused by the unfrozen water absorption on clay surfaces and the partial dissolution of CO2 hydrate into water under-saturated with CO2 during water saturation. Accordingly, the hydrate saturations estimated above, based on the assumption of 100% conversion of water into hydrate (i.e., 28%, 47%, and 63%), can be considered as the upper limits that can be achieved. Given the unavoidable uncertainties, approximately 20% error in the estimation of hydrate saturation was assumed and considered throughout the analyses, such that a range of hydrate saturation was presented rather than a single value (e.g., Figure 6). To avoid confusion, the specimens were referred to their initial target hydrate saturations (SH = 28%, 47%, and 63%), which are the upper limits.
3 Results and Analyses
3.1 VP, VS, and Electrical Resistivity during Hydrate Formation and Postwater Saturation
3.1.1 Hydrate Formation
 Figure 4 shows the evolutions of temperature, electrical resistivity, and seismic wave velocities (VP and VS) during the hydrate formation process and the postwater saturation process for the specimen with SH = 63%. As the temperature decreased, the initiation of hydrate formation was observed by the first exothermic temperature jump (Figure 4a). The hydrate formation began to stiffen the host sediments, increasing VP and VS (Figures 4c and 4d). The frequency contents of P-wave signatures were observed to consistently range from 1 kHz to 200 kHz before and after hydrate formation (Figures S4a and S4b). However, the frequency range of S-wave signatures shifted from 1–30 kHz to 1–200 kHz after hydrate formation (Figures S4c and S4d). Meanwhile, the electrical resistivity increased because the hydrate crystals reduced charge passage in the porous media (Figure 4b). Similar trends were observed in the other specimens with SH of 28% and 47%. (Figures S2 and S3).
 The geophysical signatures (VP, VS, and electrical resistivity) showed a gradual increase within a time scale of hours, suggesting that the hydrate forms gradually throughout the specimen rather than instantly. For SH = 47% and 63%, ~10 h was required for VP and VS to reach plateaus (Figures 4c, 4d, S3c, and S3d), while ~5 h was required for that of SH = 27% (Figures S2c and S2d). The amplitude of the exothermic temperature jump and the time required for the temperature of the specimen interior to stabilize and reach equilibrium also support this gradual hydrate formation. The formation rate is governed by the amount of water necessary for hydrate formation and the diffusion rate of the hydrate-forming gas (herein CO2) into water; as more hydrate forms, higher temperature jumps and longer periods are required to reach equilibrium. However, electrical resistivity can be easily altered by the local blockage of charge passages. That is, even a small amount of local hydrate formation that happens to surround an exposed electrode acting as a receiver or a source would lead to a huge increase in electrical resistivity (Figure S3b). On the contrary, VP and VS represent the property and state of a wave propagation path and give average values for the elastic waves propagating in the soil.
3.1.2 Postwater Saturation
 Postwater saturation after hydrate formation in sediments invokes the following processes. (1) Hydrate dissolution by injected water: Because the injected water is undersaturated with CO2 gas, it dissolves the hydrate. (2) Hydrate reconfiguration by Ostwald ripening: Small pores in fine-grained sediments generate freezing-point depression, which leads to preferential formation of the hydrate in large pores [Clennell et al., 1999]. Therefore, hydrate re-formation in large pores can concomitantly occur during hydrate dissociation or dissolution in small pores [Kwon et al., 2008]. In addition, the hydrate can dissociate at one location and re-form at another to minimize surface energy as water can be transported by electro-osmosis or capillarity in porous media (i.e., Ostwald ripening; e.g., Tohidi et al., 2001; Katsuki et al., 2006, 2007). Therefore, the gas hydrate may reconfigure its morphology by redistributing itself in a sediment specimen as the additional water injection enhances transports of water and CO2 within sediments.
 VP is a function of properties of both grains and pore fluids while VS is controlled by shear stiffness of the granular skeleton. As water replaces gas bubbles in pores during the postwater saturation process, VP increases owing to the increased bulk stiffness of pore fluids. Meanwhile, VS decreases owing to the hydrate dissolution. As expected, VP increased in all cases during the postwater saturation processes (see Figures 4c, S2c, and S3c). A minor decrease in VS was observed in the tests with SH = 61% (from 682 to 649 m/s; Figure 4d) and SH = 47% (from 596 to 501 m/s; Figure S3d). This result is likely attributed to the hydrate dissolution by fresh water injection. On the other hand, the test with SH = 28% showed a slight increase in VS (from 356 to 384 m/s; Figure S2d), which may be due to reconfiguration of the hydrate. Therefore, the postwater saturation procedure resulted in variation in VS by 10–20%, owing to hydrate reconfiguration or hydrate dissolution. Despite our limited estimation on hydrate saturation, the postwater saturation process allows for comparison of measured VP with the values of hydrate-bearing sediments under excess water conditions where little or no free gas exists in pores. The frequency contents of P-wave and S-wave signatures were observed to increase during postwater saturation. The frequency range of P-wave signatures shifted from 1–200 kHz to 10–1000 kHz while that of S-wave signals shifted from 1–200 kHz to 10–300 kHz (Figure S4). In addition, as free water was introduced, electrical resistivity decreased because free electron passages were widened (Figure 4b). In all cases, electrical resistivity decreased by one order of magnitude following the postwater saturation procedures (Table S2).
3.2 VP and VS during Loading Processes: Implications to Cementation Mechanism
 Seismic velocities of unconsolidated sediments subjected to loading are affected by the sediment skeletal stiffness that can be described with the Herzian-type contact model between particles. Thus, it is known that the relation between velocities (VP and VS) and vertical effective stress (σv′) can be described by a power law [Hardin and Black, 1968; Hardin and Drnevich, 1972; Cascante et al., 1998; Santamarina et al., 2001; Priest et al., 2005, 2009; Lee et al., 2010]:
where α and β are fitting parameters, and kPa is included as a divisor to render the expression in parentheses unitless. The parameter α is the wave velocity at σv′ = 1 kPa, and the β exponent reflects the nature of contact behaviors and fabric changes as stress conditions change [Santamarina et al., 2001]. A straight line can be drawn from this power relation in log-log space. The slope of this line (the β exponent) represents the extent of dependency of the velocity on σv′. The measured seismic wave velocities during the loading processes, where σv′ increased from 0.1 to 2.5 MPa, are plotted in log-log space, as shown in Figure 5. The fitting curves based on equation (1) are superimposed, and the corresponding β exponents extracted for different hydrate saturation levels are shown in Figure 5. If hydrate formation induces cementation between grains (i.e., grain-cementing hydrate), the stress-dependency of the seismic wave velocities, or the β exponent, is expected to decrease with increasing SH. Thus, the obtained β exponents can be interpreted in relation to a cementing mechanism by hydrate formation in fine-grained sediments.
3.2.1 VP Analysis
 While a general trend of decreasing β with increasing SH was observed in VP–σv′ curves, as shown in Figure 5a, two features were observed: (1) the β exponents were fairly constant over the SH range of 0%–47%, ~0.02–0.025; and (2) the β exponent significantly decreased at SH = 63%. In the low SH regime, the β of ~ 0.02, higher than the β at SH = 63%, suggests that the cementation effect by hydrate was not fully revealed in VP because the contribution by porosity reduction and SH increase during loading masked that due to a cementation effect. When the stiffness of the skeleton is much smaller than the stiffness of sediment components, such as minerals, water, and hydrates, in a water-saturated unconsolidated sediment, VP is mainly affected by the bulk moduli and composition of the pore-filling fluids [Gassmann, 1951; Mavko et al., 1998; Berryman, 1999; Santamarina et al., 2001]. During loading, the reduction of porosity and corresponding increase of SH were considerable. For an example, when σv′ = 2.31 MPa applied, SH increased from 28% to 32% and the porosity decreased from 64% to 56%. (Table S2). On the other hand, in the high SH regime, the low β value observed at SH = 63% implies that the cementation by hydrate formation became significant enough to manifest itself in VP over the other changes in porosity and SH during loading at this hydrate saturation level. This observation is similar to the results presented by Lee et al.  in spite of the different hydrate-forming method, where only minor increases in VP were observed when SH increased from 0 to 50% for kaolinite specimens and precipitated silt specimens containing THF hydrate formed from an aqueous phase with no free gas. This is also consistent with the analysis results in the later section 3.3 (Figure 6a).
3.2.2 VS Analysis
 S-wave velocity (VS) is sensitive to the vertical effective stress as well as hydrate saturation because S-wave propagation is not affected by pore fluids but by grain contact phenomena between soil particles, such as contact area, contact stiffness, bonding, and cementation. Thus, the β exponent from a VS-σv′ relation can be a better indicator than that from a VP-σv′ curve to examine the cementation behaviors in soils. For examples, β is 0 in solids, β approaches 0 for cemented soils, the Hertzian contact exhibits β = 1/6 for elastic spherical particles and β = ~0.25 for sands; and β can be larger than 0.3 for clayey soils and increases with specific surface areas and resultant plasticity of soils [Santamarina et al., 2001]. As shown in Figure 5b, while the VS of hydrate-free sediment specimen showed the most dependence on σv′ (β = 0.283), VS of the specimen with SH = 63% was the least sensitive to σv′ (β = 0.018). In particular, a significant reduction in β was observed when SH increased from 0% to 28%. It appears that a skeletal stiffening mechanism by cementation began to manifest itself in VS at SH = ~28% or lower than 28% in clayey silts. While the β value is expected to decrease with an increase in SH, the β value at SH = 47% was slightly larger than the β value at SH = 28%, which is an outlier and contrary to our expectation. This peculiarly weak cementation effect at SH = 47% can be explained by the breakage of hydrate crystals that were cementing grains when the sediment specimen with high initial porosity (~0.61) was compressed.
3.3 Effect of Hydrate Saturation on VP and VS: Implications to Hydrate Pore Morphology
 Figure 6 shows the effect of hydrate saturation on seismic wave velocities under a vertical effective stress σv′ of ~1 MPa. Approximately 20% of uncertainty for estimating hydrate saturation was considered by marking an error bar toward lower hydrate saturation. Our experimental results were plotted with grain-cementation [Ecker et al., 1998], load-bearing, and pore-filling models [Lee and Waite, 2008] for comparison, even though these theoretical models assume uniform-sized spherical grains. In addition, the semi-empirical VS and VP models suggested by Lee et al., 2010, which are based on measurements in fine-grained silty and clayey sediments containing high hydrate saturations (50% and 100%), are superimposed in Figure 6. Although all of the theoretical models compared in this study are based on the low-frequency Gassmann equation [Gassmann, 1951; Berryman, 1999], the different frequency from Hz to kHz is estimated to cause less than 7% variation in seismic velocities (VP and VS) according to Biot theory [Biot, 1956a, 1956b]. In this study, the changes in VP and VS as a function of hydrate saturation were as high as ~80% for VP and ~300% for VS (Figure 6). Thus, the frequency effect on VP and VS can be assumed to be relatively insignificant compared to the effect of hydrate saturation.
 VP measured in this study exhibited a minor increase as SH increased from 0% to 47%, where the contribution through increased SH and stiffer solid hydrate in pores was more significant to VP than that by cementation. VP at SH = 63% seemed to approach the values of the load-bearing hydrate and the grain-cementing hydrate. Note that the cementation was observed to take place from SH = 28%, as can be seen from the stress dependency of VS (Figure 5b) and the VS-SH trend (Figure 6b); thus, the obtained VP-SH trend does not necessarily imply the hydrate morphology transition from the pore-filling hydrate to the load-bearing hydrate. However, the fast VP value at SH = 63% indicates that the cementation became more pronounced. This result is consistent with the observation of the stress dependency of VP, as discussed in the section 3.2.
 A pore-filling growth habit of gas hydrates in coarse-grained sediments such as sands when formed from dissolved methane [Spangenberg et al., 2005, 2008] or THF hydrate [Yun et al., 2005; Lee et al., 2010] has been reported for hydrate saturations less than 40%. Hydrate formed under excess water conditions [Priest et al., 2009] exhibited load-bearing behavior for hydrate saturation less than 40%. Therefore, it is presumed that the pore-filling or load-bearing mechanism is dominant in the low hydrate saturation regime (e.g., < 40%) in coarse-grained sediments when hydrate forms under a water-saturated condition (excess water condition). In the higher saturation regimes, the load-bearing growth habit may be preserved, or the cementation mechanism may prevail [e.g., Yun et al., 2005; Priest et al., 2009; Waite et al., 2009; Lee et al., 2010].
 VS values measured in this study were higher than those predicted by the pore-filling model, the load-bearing model, and the model by Lee et al. . In addition, the shear stiffness of sediments increased rapidly as SH increased from 0% to 63%. As shown in Figure 6b, the measured VS-SH trend lies between the grain-cementing model and the load-bearing model. This is a clear indication of the weak cementation behavior caused by hydrate formation. A plausible explanation for the observed weak cementation in fine-grained sediments can be (1) the breakage of hydrate crystals that are cementing grains due to the sediment compression by the loading increase, and/or (2) the innate weakness in bonding between hydrates and fine mineral grains due to the presence of unfrozen water films on clay mineral surfaces, which can act as lubricant.
 It is found that the manifestation of skeletal stiffening due to the hydrate formation differs in VP and VS. Because an increase in VP with increasing SH was mostly attributed to the increased stiffness of the pore-filling phases at low SH (i.e., < 47%), the cementation effect on VP was not distinct. However, at higher SH than 47%, the cementation became more pronounced, and the relative contribution of the increased shear modulus of the skeleton to VP became significant compared to that of the increased bulk moduli of pore-filling phases. Meanwhile, we observed the cementation effect on VS exerting from SH = 28%.
 It should be noted that heterogeneous hydrate formation in fine-grained sediments as well as in coarse-grained sediments, such as grain-displacing hydrates or patchy hydrate distribution, is an innate process [e.g., Dai et al., 2012]; therefore, theoretical pore-scale models that assume uniform hydrate distribution may be less applicable to fine-grained sediments with such spatial distribution of the hydrate phase. Furthermore, deploying three-dimensional computed tomographic imaging would be beneficial for identifying the morphology of hydrate formation in fine-grained sediments and relating the results to geophysical properties of hydrate-bearing sediments.
 This paper presented seismic wave velocities measured in fine-grained sediments (highly plastic, clayey silt) containing CO2 hydrate with different hydrate saturations under different vertical effective stresses. CO2 hydrate was formed from partially water-saturated samples under excess gas conditions, with additional water injected into the specimen, resulting in excess water conditions. The cementation effect by hydrate formation in fine-grained sediments and its effect on seismic wave velocities were explored by analyzing the stress dependency of VP and VS during loading processes and by comparing the VP- and VS-SH trends with pore-scale hydrate growth models and published data.
 In general, both of the diminished stress dependency of VS and the VS-SH trend revealed that gas hydrate formation in fine-grained sediments caused weak cementation as SH increased from 0% to 63%, placing VS-SH trend between the grain-cementing model and the load-bearing model. A skeletal stiffening mechanism by weak cementation was observed at SH = ~28% in hydrate-bearing clayey silts. A plausible explanation for the observed weak cementation effect in fine-grained sediments is the breakage of hydrate crystals that are cementing grains during sediment compression and/or the innate weakness in bonding between hydrates and fine mineral grains due to the presence of unfrozen water films on clay mineral surfaces.
 Meanwhile, the manifestation of skeletal stiffening due to the hydrate formation is found to differ in VP and VS because VP in water-saturated hydrate-bearing sediments is mainly influenced by the stiffness of the pore-filling phases. Therefore, at low SH (i.e., < 47%), the cementation effect on VP was observed to be indistinct and masked by the effect of increased SH and stiff solid hydrate in pores. However, in a high SH regime (i.e., > 47%), the cementation effect on VP became more pronounced.
 We are grateful to anonymous reviewers for their valuable comments and suggestions. This research was supported by a grant from the National Research Foundation of Korea (NRF) funded by the Korean Government (MEST) (No. 2011–0027581) and by the National Gas Hydrate Project, “Gas Hydrate Development and Production” under Korean Ministry of Knowledge Economy.