- © 2004 American Mineralogist
This paper presents results of shear strength and acoustic velocity (p-wave) measurements performed on: (1) samples containing natural gas hydrate from the Mallik 2L-38 well, Mackenzie Delta, Northwest Territories; (2) reconstituted Ottawa sand samples containing methane gas hydrate formed in the laboratory; and (3) ice-bearing sands. These measurements show that hydrate increases shear strength and p-wave velocity in natural and reconstituted samples. The proportion of this increase depends on (1) the amount and distribution of hydrate present, (2) differences in sediment properties, and (3) differences in test conditions. Stress-strain curves from the Mallik samples suggest that natural gas hydrate does not cement sediment grains. However, stress-strain curves from the Ottawa sand (containing laboratory-formed gas hydrate) do imply cementation is present. Acoustically, rock physics modeling shows that gas hydrate does not cement grains of natural Mackenzie Delta sediment. Natural gas hydrates are best modeled as part of the sediment frame. This finding is in contrast with direct observations and results of Ottawa sand containing laboratory-formed hydrate, which was found to cement grains (Waite et al. 2004). It therefore appears that the microscopic distribution of gas hydrates in sediment, and hence the effect of gas hydrate on sediment physical properties, differs between natural deposits and laboratory-formed samples. This difference may possibly be caused by the location of water molecules that are available to form hydrate. Models that use laboratory-derived properties to predict behavior of natural gas hydrate must account for these differences.
Laboratory acoustic and geotechnical measurements performed on sediment samples that contain gas hydrate help us better understand the interaction between gas hydrate and host sediment. Acoustic properties can dramatically change as gas hydrate forms, and this relationship may allow remote identification and even quantification of hydrate (Chand and Minshull 2003; Lee et al. 1993).
The presence of gas hydrate can significantly increase the shear strength of sediment. Hydrate dissociation, on the other hand, can lower effective stresses and induce slope failures. The Storegga slide, which caused tsunamis to impact both Scotland and Norway about 6000–8000 years ago, may have been influenced by gas hydrate dissociation (Andreassen et al. 2000; Buenz et al. 2003; Hovland et al. 2001; Mienert et al. 1998). Concerns exist that gas hydrate dissociation related to slope failures may release additional methane into the atmosphere and have abrupt climatic effects (Kennett et al. 2003).
This paper presents results of shear strength and acoustic (p-wave) measurements performed on specimens (1) from the Mackenzie Delta that contain natural gas hydrate and (2) of reconstituted Ottawa sand that contain laboratory-formed gas hydrate. We also measured the physical properties of frozen sediment for comparison because ice and gas hydrate coexist in some parts of the Earth’s geosphere (Collett and Dallimore 2000; Skorobogatov et al. 1998). Furthermore, some properties of ice and gas hydrate are similar; therefore study of frozen sediment properties may enhance our knowledge of gas hydrate behavior.
Geologic setting, location of gas hydrate, and field methods
We tested samples containing natural gas hydrate from the Mallik 2L-38 well drilled in the Mackenzie Delta region during 1998. The samples were recovered from beneath 640 m of permafrost (Collett et al. 1999). The deltaic sediment was deposited in numerous transgressive-regressive stratigraphic sequences on top of the lower Cretaceous structural basement (Dixon and Dietrich 1988; Dixon et al. 1992). Natural gas hydrate was observed within coarse-grained sand and gravel deposits between 897 and 922 m occupying an average of 47% of the void space in situ (Collett et al. 1999; Dallimore et al. 1999b; Winters et al. 1999a)
The temperature at 900 m is about 7 °C (Wright et al. 1999), and in situ sediments contained a mix of gas hydrate and water in their pores. However, pore water was converted to ice during core recovery from endothermic cooling induced by gas hydrate dissociation (Dallimore et al. 1999a). The self-preservation phenomena described by Ershov and Yakushev (1992) and Stern et al. (2001) may be responsible for preserving hydrate, however the reason for the preservation is not fully understood. Samples were stored in vessels pressurized with methane gas to about 9 MPa, and held at −6 to −10 °C for the overland trip to Woods Hole, Massachusetts. Core recovery, transportation, and storage may have affected sediment properties, however, some changes are unavoidable without measurements performed inside an in situ pressurized core barrel.
Laboratory equipment and methods
The Gas Hydrate And Sediment Test Laboratory Instrument (GHASTLI) simulates in situ conditions of pressure and temperature that preserve natural gas hydrate in sediment or will form gas hydrate in reconstituted sediment (Winters et al. 1994, 2000). A maximum pressure of 25 MPa can be applied to a 70 mm diameter ×130 mm test sample within a silicone-oil-filled main pressure vessel.
The test sample is surrounded by flexible membranes, and top and bottom end caps incorporate acoustic transducers and gas or water flow ports. The bottom end cap rests on an interchangeable internal load cell. The top end cap is contacted by a variable-temperature bath-controlled heat exchanger that imparts a unidirectional cooling front downward through the specimen.
Five 500 mL capacity syringe pumps control pressure and flow (0.001 mL/min to over 80 mL/min) during testing. Confining pressure is externally applied to the specimen; internal specimen pressures include pore pressure, back pressure, and methane-gas pressure. The back-pressure system contains a collector that is capable of separating and measuring water and gas volumes that are pushed out of the specimen at test pressures by the dissociation of gas hydrate. A separate syringe pump controls the movement of the load ram during the shear phase and is used to determine the height of the specimen, which is necessary to calculate velocity. Temperature is measured with four thermocouples and four thermistors placed against the outside of the specimen at different heights.
The Mallik 2L-38 test samples were kept frozen prior to testing in GHASTLI. Samples were prepared within a walk-in freezer maintained at about −30 °C. The test sample was placed into GHASTLI and pressures (between 8 and 12 MPa) and temperatures (near 2 °C) conducive to hydrate stability were quickly established to let the ice melt while preserving the hydrate. An effective consolidation stress, σ′c, (typically 0.25 MPa) was imposed on the sample to partially recreate in situ effective stress. Time-related increases in velocity or gas consumption, indicative of secondary hydrate growth, were not observed during the melting and subsequent test phases.
After all of the ice melted, the temperature was slowly raised in 1 to 2 °C increments until gas hydrate dissociation began. The dissociated gas, specimen pore water, and water released from hydrate entered the back-pressure collector where water and gas were separated and their volumes measured. After dissociation, the sample was removed from GHASTLI, flushed with water, and retested to obtain baseline water-saturated sample velocities.
Laboratory-formed hydrate was created within sediment samples using two methods. The first method began with initially water-saturated sediment (this study). Methane was slowly flowed into the test specimen until a predetermined amount of water was pushed out, and then the temperature was lowered from about 22 to 4 °C, thereby inducing gas hydrate formation. Partly water-saturated sediment was used in the second method (Waite et al. 2004). Initially the pore spaces contained a combination of water and air at atmospheric pressure. As methane pressure was increased to 12 MPa in the initially partly water-saturated samples, the air was greatly compressed. Hydrate formation was again induced by lowering the sample temperature. The currently used procedures to form gas hydrate in the laboratory mimic natural gas hydrate formation in the presence of free gas (Suess et al. 2001). Tests to form gas hydrate using dissolved methane are planned. Frozen sediment was created within GHASTLI by reducing the temperature to subfreezing values.
P-wave velocity (Vp) is measured by through transmission using 0.5–1 MHz (natural frequency) wafer-shaped crystals that are located on the backside (away from the specimen) of each end cap. A 400 volt pulse is sent to the transmitting transducer and the received signal is amplified, digitized, displayed on a digital oscilloscope, and recorded by a computer. Vp is calculated from the specimen length and measured travel time corrected for system delays. The error margin for Vp was estimated to be 0.02 km/s.
Four parameters were measured during triaxial strength tests: load, axial deformation, confining pressure, and pore pressure. Movement of the ram, which can vary from 0.0001 mm/min to 2 mm/min, is measured using a linear displacement transducer. Strain rates were approximately 0.07%/min. Interchangeable load cells can be varied according to the anticipated strength of the specimen.
Gas hydrate saturation
Gas hydrate dissociation causes a volume expansion, which we used to calculate the amount of gas hydrate in the samples. The volume of both water and gas that exited from the sample during dissociation was accurately measured in a collector and a back-pressure syringe pump. We corrected the volume in the collector for the temperature and pressure differences between collector and sample.
The sample porosity remains unchanged during dissociation. This assumption is based on thermodynamic calculations for coarse-grained sediments (Clennell et al. 1999), which suggest that in sandy sediments, such as the Mallik samples, the shape of gas hydrate “crystals” accommodates the shape of the existing pore space. Gas hydrate does not form its own pore space by “pushing aside” sand grains. Then, the sum of the adjusted gas volume and the water volume is equal to the excess volume of gas generated during gas hydrate dissociation.
While the above approach allows accurate measurement of gas produced during dissociation, the volume of hydrate can only be estimated within relatively large error margins, because of the unknown gas-hydrate cage occupation number. This number is a measure of the fraction of the cavities in the gas hydrate structure that are actually filled with methane molecules and is important for calculating the amount of gas released during gas hydrate dissociation. We assumed this value to be 0.9 (Sloan 1990), but also used fractions of 0.8 and 1.0 as low- and high-end bounds. The uncertain cage occupation number caused by far the largest error margin in our modeling. The gas hydrate saturation of pore space was calculated from gas hydrate volume, sample volume, and porosity.
Rock physics modeling of Mallik samples containing natural gas hydrate
Rock physics modeling was performed to investigate the microscopic distribution of gas hydrates in the sediments, similar to the approach of Helgerud et al. (1999). Three fundamental models were created. Gas hydrates may (1) be “floating” in the pore space without grain contact, (2) be part of the sediment frame, or (3) act as cement between sediment grains (Fig. 1⇓). Pore and frame models are sometimes described as “dissemination” models.
The Gassmann (1951) equation predicts the compressional modulus K of a porous medium from compressional and shear moduli of the grains Kg and μg, the pore fill (Kf, μf equal 0) and the dry or matrix moduli (Kd, μd). Although the moduli of most fluids and minerals are well known from the literature, the Reuss (1929) average allows for calculation of the properties of a mix of pore filling material, including, for the dissemination model, gas hydrates. Grain properties are usually “mixed” using the Hill (1952) average [see Mavko et al. (1998) for an overview].
The dry moduli Kd and μd depend on the geometry of the grains and how they are in contact. We determined Kd and μd based on the Hertz-Mindlin (HM) contact theory (Mindlin 1949), which assumes a medium that consists of randomly packed elastic spheres, slightly deformed under differential pressure. The HM-theory is valid for sediments at critical porosity (c), at which the sediment grains start to lose contact. A Φc value of 40% is realistic for sandy sediments (Nur et al. 1998). The modified Hashin-Shtrikman (HS) lower bound theory (Dvorkin and Nur 1996; Hashin and Shtrikman 1963) allowed adjustment for the porosity deviation from Φc. Density of the porous medium (ρ) was computed as the average of grain and pore-fill density (ρg, ρf).
Moduli values of water-saturated, gas-hydrate-free sediments were determined first. It was possible that small amounts of gas were present in the sediment after hydrate dissociation, despite attempts to flush the samples with water. Even miniscule amounts of gas significantly lower Vp (Domenico 1977). To ensure that the modeling parameters are not distorted by the presence of gas, a sample (GH073) was reconstituted by mixing sediments from other samples into a water column. The sample was compacted and Vp measured under differential pressures. These results were used to calculate the elastic moduli of gas-hydrate-free host sediment.
Vp in the reconstituted sample was measured as 1.89 ± 0.02 km/s. The sediment consisted mainly of quartz sand. Using Kg and μg values of quartz Vp was estimated to be 1.96 km/s. The rock physics model needed to be adjusted to reach 1.89 ± 0.02 km/s (see Table 1⇓ for elastic properties and references). Log data indicate 15–30% of non-quartz material, much of it illite (Dallimore et al. 1999c). Therefore clay was added (Han 1986) as a second mineral phase to our modeling in order to lower Kg and μg. Between 25 and 44% clay was added in the model to achieve Vp = 1.89 ± 0.02 km/s. These numbers are somewhat higher, but still within the range of non-quartz material suggested from logging.
Using moduli from reconstituted samples implies that the gas-hydrate-free sediments were not cemented, which we knew from visual observation and also from other studies (Dallimore et al. 1999c). Vp was calculated for samples GH060 and GH062 after dissociation of gas hydrate using the moduli from the reconstituted sample, corrected for porosity, as 1.82–1.87 and 1.76–1.80 km/s, respectively. These values are similar to the measured range after dissociation, 1.75–1.80 km/s. Therefore, we are confident that the predicted values for the moduli of the gas-hydrate-free sediment are reliable. Finally, Vp was modeled as a function of gas hydrate saturation applying the three fundamental types of gas hydrate distribution. With this semi-deterministic approach, reference velocities were used from a gas-hydrate-free sample at slightly different porosity from the gas-hydrate-bearing sediments.
Pore model, mixing with the fluid phase (Fig. 1b⇑): In this model, gas hydrates are “floating” in the pore space and do not affect the stiffness of the sediment μ. Kf is the average (Reuss 1929) between the compressional moduli of water and gas hydrates whereas μf equals 0. It may appear unrealistic that solid hydrates do not affect the shear modulus; however, if gas hydrates are not connected to the sediment frame they do not assist in shear energy transmission.
Frame model, mixing with the solid phases (Fig. 1c⇑): It was assumed that gas hydrate mixed with the mineral components of the sediment, i.e., that gas hydrate become part of the sediment frame. The Hill average (Hill 1952) was computed to predict the seismic properties of two mixed solid phases, as were the upper and lower HS-bounds which describe the theoretically possible limits of these properties. Sediment porosity was decreased accordingly because in this model, gas hydrate formation transfers liquid pore water into the solid sediment frame.
Cementation (Fig. 1d⇑): The two principal cementation models are (1) that gas hydrates are evenly coating the surface of grains, or (2) that they form between grains. The approximation of Dvorkin and Nur (1996) from the cementation theory (Dvorkin et al. 1991) was used to compute the grain and dry properties of the cemented material.
Results and discussion
Salinity-corrected water contents, porosity, and other index properties were determined for specimens dried at 90 °C for at least 24 hours or from initial sample mass measurements (Table 2⇓). The water content calculations include all water in the sample, both free water and water that forms hydrate cages.
The porosities of the intact Mackenzie Delta samples (GH060, GH062) vary from 33.1 to 36.3% and the reconstituted sample (GH073) is slightly lower at 30.4%. Medium sand is the predominant size fraction, but some silt is also present. A closer spread of porosity (32.6–33.0%) is exhibited by the sieved and reconstituted Ottawa sand (0.25–0.50 mm diameter) specimens (GH066, GH069, GH079).
Results of the acoustic measurements are summarized in Table 2⇑. The presence of substantial gas hydrate or ice in sediment pores has a significant effect on acoustic characteristics. The properties measured from samples containing gas hydrate and water are the most relevant to in situ conditions at Mallik 2L-38. However, Vp for samples containing ice are included because of their potential similarity to gas-hydrate-bearing sediments thought to occur within the permafrost itself (Dallimore and Collett 1995).
Typically the baseline P-wave velocity (Vp) of water-saturated Ottawa sand and Mallik sand is 1.77 to 1.94 km/s (Table 3⇓). Vp in Ottawa sand (GH069) changes from 1.9 km/s for water-saturated sediment to 3.95 km/s after nearly complete hydrate saturation of pores (an increase of about 2.05 km/s). The maximum Vp of Mackenzie Delta sediment containing natural gas hydrate at partial saturation is 2.91 km/s. Frozen sediment specimens have final velocities of 4.23 to 4.33 km/s (Table 3⇓), corresponding to velocity increases of 2.33–2.47 km/s (Table 4⇓). Samples that already contained gas hydrate and were then frozen had a maximum 2.11 km/s increase, which is still lower than that of completely frozen samples. This information agrees with the results presented by Lee and Collett (1999) for field measurements of Vp performed at the Mallik 2L-38 well which indicate that practically all field Vp values are less than the values for frozen Mallik sand obtained in the laboratory.
Because of the paucity of strength measurements in gashydrate-bearing sediments, ice is often used as a proxy for the effect of gas hydrate on the shear strength of sediments. However, pure gas hydrate is stronger than ice (Durham et al. 2002, 2003). Frozen sediment exhibits a wide range in strength properties because strength is influenced by a number of factors: strain rate, temperature, consolidation stress, grain size, and density. The strength of sediment containing gas hydrate is probably influenced by these and other factors such as gas hydrate cage occupancy.
Samples from the Mallik 2L-38 well were tested in GHASTLI both before and after natural-gas hydrate was dissociated (Fig. 2⇓) (Winters et al. 1999b, 2002). The strength of the gas hydrate-containing sample (GH062) is much higher than similar samples recovered from the well that did not contain hydrate during the shear phase (GH058, GH059, GH060). The large difference in strength between the specimen containing gas hydrate and the other samples is mainly due to pore filling by gas hydrate, which increases the tendency for dilation during shear. An important feature of the stress-strain curve for GH062 is that its low initial slope, which is almost identical to the non-cemented GH079 test, indicates plastic deformation. This observation suggests that the naturally formed gas hydrate in this sample does not cement sediment grains. Hence, the results from our shear strength tests agree with the acoustic results.
Forming gas hydrate in Ottawa sand increases its strength significantly (GH069 in Fig. 2⇑), similar to freezing (GH066). The shear strengths measured for the gas hydrate-bearing and frozen samples are in agreement with the wide range encountered for frozen sands (Andersland and Ladanyi 1994). The frozen specimen is somewhat stronger, but the gas-hydrate-containing sample has a higher Young’s modulus (slope of the initial stressstrain curve). Neither the frozen nor the gas-hydrate-bearing Ottawa sand had an initial plastic phase. The ice-free Ottawa sand (GH079), which obviously is not cemented, has an initial plastic segment before the elastic phase. All three Ottawa sand samples exhibited strain softening after peak strength.
Cementation appears to be common in (1) sandy sediments containing laboratory-formed gas hydrate and (2) frozen Ottawa sands. We have inspected a number of samples with different mineralogy (but always predominantly sandy) by taking them out of the pressure chamber after forming gas hydrates (including GH069). The samples appeared cemented because they were intact and hard, did not crumble when handled, required great physical force to break, and produced significant amounts of methane gas during dissociation. The pore fillings were a whitish color indicative of gas hydrate, rather than ice formed by hydrate dissociation.
Rock physics modeling
Velocities predicted from the cementation theory are much higher than the observed velocities (Fig. 3⇓). Within the error margins related primarily to hydrate cage occupancy, our results suggest that natural gas hydrates act as part of the sediment frame and do not cement grains. We caution that if gas hydrate cage occupancy is considerably lower than 80%, a pore model without any grain contact would also be possible (Fig. 3⇓).
Our modeling approach contains various simplifications, starting with the assumption that the sediment consists of spheres. The Gassmann equation is a low-frequency approximation and therefore is not ideal to model laboratory measurements. A similar simplification is often made implicitly when using seismic laboratory measurements for calibrating field data. Critical porosity is not well constrained. However, the relative changes with respect to the hydrate-free reference sediment should not be significantly affected by these approximations.
An accurate determination of gas hydrate saturation, however, remains the key uncertainty for our analysis despite having well-controlled laboratory conditions. Determining gas hydrate saturations based on gas expansion during dissociation requires a good estimate of cage occupancy and an estimate of sediment contraction during dissociation.
Significant improvements for studying the elastic properties of gas-hydrate-bearing sediments are expected from planned modifications in GHASTLI. Measurement of shear-wave velocity (Vs) will enable calculation of all three parameters that describe elastic media (compressional modulus, shear modulus, and density). Pore (almost no change of Vs with gas hydrate saturation) and frame models (change of Vs in the same order of magnitude as Vp) would also be better distinguished.
Vs within the gas hydrate zone in the Mallik 2L-38 borehole is markedly higher than in surrounding sediments. This rules out a pore model for gas hydrate distribution (Sakai 1999). Compressional-to shear-converted waves on the Blake Ridge can also be best explained by changes of Vs due to gas hydrates (Pecher et al. 1997). On the other hand, results from an ocean bottom cable survey off of Norway (Andreassen et al. 2003) suggest that Vs is not affected by gas hydrate in the study area. However, this may be partly due to relatively low gas hydrate concentrations.
Our main finding from acoustic modeling, that naturally formed gas hydrates in the Mallik cores do not cement sediments, is consistent with field results and laboratory strength measurements (previous section). Laboratory-formed gas hydrates in Ottawa sands do appear to cement sediment grains. The cause of the apparent contradiction between the cementing behavior of laboratory-formed gas hydrates in Ottawa sand and the probable frame-forming natural gas hydrate in Mallik sands is not yet fully known. It appears from other studies that gas hydrate in different natural settings is also disseminated in the pore space.
These results are intriguing because they suggest that yet-unknown factors, such as formation mechanism or mineral phases, may control the interaction between gas hydrate and sediment grains. These factors need to be understood because they appear to cause vastly different responses of sediment physical properties to gas hydrate saturation. Perhaps it is the location of water molecules adhering to sediment grains that causes cementation at grain contacts during hydrate formation in the laboratory. Because gas hydrate formed within porous media in the laboratory tends to cement grains, models using laboratory results must be adjusted to account for the difference in behavior between lab and natural gas hydrate.
The authors thank M. Lee, S. Circone, B. Dugan, J. Grozic, J. Kliner, and an anonymous reviewer for their helpful comments. This work was supported by the Coastal and Marine Geology and Energy Programs of the U.S. Geological Survey and funding was provided by the Gas Hydrate Program of the U.S. Department of Energy.
Manuscript handled by Bryan Chakoumakos
- Manuscript Received November 15, 2003.
- Manuscript Accepted May 11, 2004.