Marine sands highly saturated with gas hydrates are potential energy resources, likely forming from methane dissolved in pore water. Laboratory fabrication of gas hydrate-bearing sands formed from dissolved-phase methane usually requires 1–2 months to attain the high hydrate saturations characteristic of naturally occurring energy resource targets. A series of gas hydrate formation tests, in which methane-supersaturated water circulates through 100, 240, and 200,000 cm3 vessels containing glass beads or unconsolidated sand, show that the rate-limiting step is dissolving gaseous-phase methane into the circulating water to form methane-supersaturated fluid. This implies that laboratory and natural hydrate formation rates are primarily limited by methane availability. Developing effective techniques for dissolving gaseous methane into water will increase formation rates above our observed (1 ± 0.5) × 10−7 mol of methane consumed for hydrate formation per minute per cubic centimeter of pore space, which corresponds to a hydrate saturation increase of 2 ± 1% per day, regardless of specimen size.
 Though hydrate-bearing sands contain only ~10% of the world's gas hydrates [Collett et al., 2009], their high hydrate saturations and methane concentrations have made them energy resource targets, with notable marine resource studies in the Nankai Trough offshore Japan [Fujii et al., 2008; Yamamoto et al., 2012] and northern Gulf of Mexico [Boswell et al., 2012; Frye, 2008]. Forming methane hydrate-bearing sands and measuring their mechanical properties in the laboratory is important for developing safe, effective procedures for extracting methane as a resource.
 Gas hydrates generally form from biogenic methane [Collett et al., 2009], and as reviewed by Waite et al. , this methane, usually dissolved in pore water, cools as it migrates up through the sediment. In the presence of hydrate, cool water cannot hold as much methane as warm water, and this excess dissolved methane can be used to form added hydrate. We seek to duplicate this dissolved-phase hydrate formation process when characterizing gas hydrates in marine sands because acoustic, electrical, and nuclear magnetic resonance analyses suggest gas hydrates forming in this fashion are load bearing, strengthening sediment less effectively than when hydrates form in the presence of excess gaseous methane and can cement sand grains together (reviewed in Waite et al. ).
Priest et al.  demonstrate how forming hydrate in sands from gaseous methane and an excess of water produces specimens with load-bearing mechanical properties. Though relatively rapid, the technique is limited to hydrate saturations Sh < 40%. Higher hydrate saturations are essential for mimicking resource-grade hydrate-bearing sands in nature, which can have Sh as high as 60 to 90% (reviewed in Collett et al. ). It is possible to attain Sh ~ 95% by forming hydrate directly from methane dissolved in water [Spangenberg et al., 2005], though this is slower than the Priest et al.  method.
 Shortening specimen fabrication times by accelerating hydrate formation from dissolved-phase methane can improve experimental laboratory research efficiency. Here we show the rate-limiting step is dissolving gaseous methane into the water from which hydrate grows. The implication, both for laboratory systems with relatively high fluid-flow rates and elevated methane concentrations and for natural systems with relatively low fluid-flow rates and methane concentrations, is that hydrate formation is efficient enough to be limited by the methane availability. Dissolved-methane concentrations can be increased in the laboratory to increase hydrate formation rates.
2 Apparatus and Procedure
 To isolate key factors limiting observed hydrate formation rates in dissolved-phase systems, we use a broad range of chamber sizes, circulating-water flow rates, pressure, and temperature conditions (Table 1). All tests conducted employ the Spangenberg et al.  hydrate formation strategy based on the ability of warm water to hold more dissolved methane than cold water can in the presence of hydrate [Zatsepina and Buffett, 1997] (Figure 1). In our systems, gaseous methane dissolves into warm water circulating through a chamber with a gas/water interface. As this circulating water cools in the cold specimen chamber, excess methane that can no longer be held in the dissolved phase comes out of solution and forms additional hydrate. The amount of excess methane is the supersaturation SS, calculated as shown in Figure 1b. In the absence of hydrate, the methane solubility curve continues rising at lower temperatures rather than having a peaked shape (Figure 1b), meaning water circulating through a cool, hydrate-free specimen chamber would be unable to form gas bubbles but would be able to form additional hydrate once hydrate nucleated in the specimen chamber.
Table 1. Specimen Chamber Size and Test Condition Summarya
Chamber Diameter (cm)
Chamber Length (cm)
Pore Volume (cm3)
Operating Pressure (MPa)
Interface Temperature (°C)
Specimen Temperature (°C)
Flow Rate (cm3/min)
aApparatus-specific schematics and procedures are given for GFZ1 by Spangenberg et al. , DPC by Waite et al. , and LARS by Schicks et al. .
 Specimen composition and test procedures are summarized here (additional experimental details are in the Table 1 references). The GFZ1 specimen is a 250–500 µm diameter glass bead pack with 38% porosity. The Dissolved Phase Cell (DPC) uses 90–250 µm grain size F110 quartz sand packs with 37–39% porosity. Large reservoir simulator (LARS) tests 1 and 2 have 2–3.15 mm grain size gravel specimens with 37% porosity. LARS tests 3–5 have 0.5–1 mm grain size sand specimens with 36–37% porosity. The porosities are all within the ~35–40% range observed in naturally occurring hydrate-bearing sands [Collett et al., 2009].
 The specimen pore space in each apparatus is initially evacuated, then pressurized with an NaCl solution of 0.1 and 0.07 mol/kg for GFZ1 and LARS, respectively. These salinity levels are too low to inhibit hydrate formation (Figure 2), but salinity provides a means of tracking hydrate formation over time as discussed below. DPC is pressurized with deionized water and methane gas as described below.
 GFZ1 and LARS form methane hydrate solely from methane dissolved in water, beginning with a hydrate saturation Sh = 0. DPC specimens are initially seeded with methane hydrate Sh = 15–22%, using the excess water method of Priest et al. : the specimen is first pressurized to ~4.25 MPa with methane gas, then deionized water is injected through the specimen's top and bottom until the operating pressure is attained. Specimen cooling to 3.5°C at 11.7 MPa initiates hydrate formation from methane bubbles, and this pressure is maintained by injecting water during the 4–5 day hydrate formation period. Hydrate seeding is considered complete when the pump no longer injects water into the specimen. The specimen is then temporarily warmed to ~13°C, within 1°C of the hydrate stability temperature, to help the initial circulation of ~22°C methane-rich water break through any clogs formed during seed-hydrate formation.
 Prior to entering the porous specimen in all three systems, water circulates through a gas/water interface chamber designed to create a large interfacial area through which methane can dissolve into the circulating water. The GFZ1 and DPC interface chambers are filled with 250–300 µm diameter glass beads. Water dripping through the chamber's methane headspace spreads over the glass beads, dramatically increasing the water's surface area. In the LARS system, water is sprayed into the interface chamber's methane headspace to increase the gas/water interfacial contact. Water is circulated repeatedly through the methane-pressurized gas/water interface to charge the water with methane prior to circulating water through the complete flow loop (Figure 1a). A Contros brand dissolved-methane sensor measures the methane concentration in the circulating water as it passes into, and out of, the LARS growth chamber.
 To limit clogging at the specimen inlet and outlet ports due to hydrate formation, heater tape or a heater wire is placed around or through the line where fluid contacts the cold region surrounding the specimen. Care must be taken not to heat fluid above the interface temperature, TI, prior to entering the specimen chamber, as doing so reduces the methane solubility of the circulating water (Figure 1b), causing excess methane to form free gas bubbles. These bubbles can be transported to the specimen inlet, where rapid hydrate growth on the bubble surfaces can quickly clog the inlet port.
 Clogs forming in the GFZ1 and DPC specimens are cleared by raising the specimen temperature until the resumption of flow is able to break through the clog. Clogging is not an issue in LARS because the inlet flow line to the specimen is kept at a temperature below that of the gas/water interface but slightly above hydrate stability and is monitored by a temperature sensor in the specimen directly at the inlet port. Once steady flow is established, the flow rate and specimen temperature (Table 1) can be adjusted to obtain the desired formation conditions (Figure 2).
Spangenberg et al.  demonstrate how hydrate formation can be tracked throughout the GFZ1 and LARS experiments by measuring the electrical conductivity of the circulating water: as hydrate formation consumes water, the salt concentration increases in the remaining water, increasing the electrical conductivity. Over each measurement interval, the conductivity change indicates the water volume consumed for hydrate formation and thus the hydrate volume, which we divide by the pore volume to get a hydrate saturation increase. Assuming methane hydrate has a composition of CH4·5.99H2O [Circone et al., 2005], we also calculate the moles of methane consumed for hydrate formation at each measurement interval.
 Only an average formation rate can be obtained in the DPC system: DPC specimens are dissociated at the conclusion of each test, with the volume change during dissociation tracked using an Isco syringe pump. Given the equation of state for methane [Duan et al., 1992; Duan and Mao, 2006] and assuming a hydrate composition of CH4·5.99H2O [Circone et al., 2005], the volume of hydrate and moles of methane consumed are calculated. The total DPC hydrate volume is the sum of the initial seed hydrate and the subsequent hydrate formation from methane dissolved in the circulating water. We assume all gas initially in the specimen is consumed to form hydrate surrounded by fully methane-saturated water prior to circulating methane-rich water through the specimen and subtract this from the total measured hydrate volume. This provides a lower bound for the estimated volume of hydrate formed from dissolved-phase methane. To get formation rates, we divide the volume of hydrate formed, or equivalently the moles of methane consumed, by the time spent circulating water through the specimen.
 Two key parameters for hydrate formation from dissolved-phase methane are methane supersaturation, SS, and the residence time that excess methane is given to come out of solution and form hydrate in the specimen. Ideally, circulating water is fully methane saturated in the gas/water interface chamber, meaning SS can be calculated from the experimental pressures and temperatures (Figure 1b). As discussed below, full saturation is not always achieved, and these idealized calculations can overestimate the true SS. The residence time is experimentally controlled by the circulating water's flow rate. In these experiments, the rate at which methane is consumed to form hydrate (Figure 3a) or equivalently the rate at which hydrate grows in the pore space (Figure 3b) is largely independent of both the ideal calculated SS and measured residence time.
 Formation rates vary instead with alterations to the gas/water interface chamber. The Figures 3a and 3b results motivated installing a dissolved-phase methane sensor in LARS. The homemade nozzle used in the LARS interface chamber for spraying circulating water through the methane headspace was also replaced with a Bete brand nozzle that generates a fine mist of water droplets that facilitates methane uptake by increasing the water's surface area through which methane can dissolve. Figure 3c shows that the new nozzle nearly methane saturates circulating water, which did not occur with the nozzle used while obtaining the Figures 3a and 3b data. Using the new nozzle doubles the average hydrate formation rate relative to the tests in Figures 3a and 3b. Similarly, for tests with SS > 10%, DPC formation rates scale nearly linearly with the methane headspace height in the interface chamber through which circulating water percolates over glass beads (Figure 3d). The additional headspace volume improves methane transfer from the gaseous to dissolved phase, and that added methane is being consumed to form hydrate in the DPC growth chamber.
 We view dissolving gas into water as our rate-limiting step, but there are other possibilities. Here we examine four general mechanisms potentially limiting the hydrate growth rate in dissolved-phase systems (reviewed by Sloan and Koh ): (1) reaction kinetics—the molecular-level hydrate formation rate at the growth front, (2) the rate at which the heat of hydrate formation is dispersed, (3) the rate at which methane molecules are transported to the growth front through the surrounding water, and (4) the rate at which methane is dissolved into water at the gas/water interface.
4.1 Reaction Kinetics
 If hydrate formation in our experiments is rate limited only by the kinetics of building hydrate from abundantly available methane and water, observed formation rates should increase as the number and size of hydrate growth fronts increase during a formation test. The formation rates measured during the GFZ1 and LARS experiments, however, show no systematic increase with increasing pore-space hydrate saturation, and there is no dependence of the average dissolved-phase growth rate on whether hydrate is initially seeded in the specimen chamber (DPC) or not (GFZ1 and LARS).
4.2 Heat Dissipation
 Hydrate formation generates heat, liberating ~54.44 kJ·(mol CH4)−1 [Gupta et al., 2008]. Given our methane consumption rate of ~1 × 10−7 mol min−1 cm−3 of pore space and a 350 min residence time (Figure 3b), even if the water instantaneously absorbs all heat released from hydrate formation, the 4.18 J (g K)−1 heat capacity for water [Weast, 1987] limits temperature increases to < 0.5°C per cm3 of water. Temperatures actually increase far less because hydrate forms over time. Sensors in the LARS and DPC chambers indicate specimen temperatures remain well within the hydrate stability field (Figure 2), though localized fluctuations away from the sensors are possible.
4.3 Methane Diffusion From Circulating Water to the Hydrate Growth Front
 If methane mobility is the limiting factor in the hydrate formation process, the formation rate should scale with the residence time because slower flow allows time for more methane migration to hydrate formation fronts. Such a trend is not apparent (Figure 3b). Figure 3c shows that essentially all excess methane is stripped from the circulating water to form hydrate, suggesting that formation rates are not limited by any processes related to incorporating dissolved-phase methane into the hydrate structure.
4.4 Methane Transport From the Gaseous to the Dissolved Phase
 As reviewed in Sloan and Koh , the Skovborg and Rasmussen  analysis of extensive hydrate formation rate experiments by Bishnoi et al. [1985, 1986] suggests the hydrate formation rate in a stirred reactor is limited by methane transport across the gas/water interface. Stirring continually refreshes the interface, transporting methane away from the interface and into the bulk water where it can be consumed to form hydrate at rates on the order of 0.2–1 × 10−5 mol min−1 cm−3 of pore space [Linga et al., 2012; Mohebbi et al., 2012], two orders of magnitude more rapidly than we observe (Figure 3a).
 Based on the results of Su et al. , we attribute our hydrate formation rate to the limited methane transport across our gas/water interface rather than to the relatively large distance between our gas/water interface and the hydrate growth fronts. Su et al.  show how hydrate formation close to a gas/water interface, as occurs in stirred reactors, does not guarantee rapid hydrate growth. Their system circulates water and gas phase methane through a sand pack in a 0.5 m diameter, 1 m long chamber. During the primary hydrate growth phase from 60 to 300 h, the average gas consumption rate is ~1.2 × 10−7 mol min−1 cm−3 of pore space. This agrees with our results (Figure 3a) in spite of the Su et al.  system having a gas/water interface directly within their sand pack.
5 Implications and Recommendations
 Rather than continually refresh the gas/water interface via stirring, our systems are designed to provide a large interfacial area and short gas-diffusion lengths while water passes through the chamber's methane headspace. Figures 3c and 3d show how increased interfacial contact between methane and water increases the dissolved-methane concentration and subsequent hydrate growth rate. LARS experiments 4 and 5 also demonstrate the impact of interface chamber performance on hydrate growth rate. During these experiments, the nozzle spraying water droplets through the methane headspace of the gas/water interface chamber became partly clogged with hydrate, further limiting the nozzle's efficacy in generating small water droplets and a large total interfacial contact area. Average growth rates for these two experiments are the lowest of the five LARS experiments (Figures 3a and 3b).
 Three general implications can be drawn from the strong correlation between the extent of methane/water contact in the interface chamber and hydrate growth rate:
 Increased methane supersaturation is equivalent to increased available methane for hydrate formation and correlates with an increased hydrate formation rate (Figures 3c and 3d). This correlation is obscured in Figure 3a because the gas/water interface chamber does not fully methane-saturate the circulating water, meaning a pressure/temperature-based calculation overestimates, and does not reflect, the true supersaturation.
 In our experiments, hydrate formation is rapid enough to consume essentially all available methane in even the shortest residence times (Figures 3b and 3c).
 Hydrate grows efficiently enough from methane dissolved in water that changing pressure and temperature conditions (Figure 2) and circulation flow rates (Figure 3b) does not improve hydrate formation rates as significantly as designing a gas/water interface chamber that effectively dissolves gaseous methane into water.
 Stirred reactors are clearly effective cells for dissolving gaseous methane into water, but for an uninstrumented cell, a high-efficiency, small droplet spray nozzle can almost fully saturate the circulating water. If low flow rates are required and glass beads must be used instead of a nozzle, maximizing the bead-filled gaseous headspace volume and using the smallest bead size the chamber can retain is suggested.
 Rapid conversion of nearly all available dissolved-phase methane to hydrate implies that hydrate formation rates in laboratory and natural systems are limited by the methane concentration in the water. To more rapidly mimic highly hydrate-saturated, naturally occurring hydrate-bearing sands formed from methane dissolved in water, establishing an efficient means of dissolving methane gas into the circulating water is more important than the specific pressure, temperature, or fluid-flow conditions at the hydrate growth front.
 U.S. Geological Survey contributions were partially supported through an Interagency Agreement DE-FE0002911 with the U.S. Dept. of Energy's National Energy Technology Laboratory. Use of trade names is for descriptive purposes only and does not imply endorsement by the U.S. Government. The GFZ contribution was partly funded by the program GEOTECHNOLOGIEN of BMBF and DFG, grant 03G0605A and by the Federal Ministry of Economics and Technology (BMWi) within the SUGAR project framework, grant FKZ03SX320E. GFZ machine shop staff support during the experimental setup is gratefully acknowledged. We thank Mr. Saturov from CONTROS Systems & Solutions GmbH for his help in installing and calibrating the methane concentration sensor.
 The Editor thanks Jeff Priest and an anonymous reviewer for their assistance in evaluating this paper.