An advanced passive diffusion sampler for the determination of dissolved gas concentrations

Transcription

1 WATER RESOURCES RESEARCH, VOL. 45, W06423, doi: /2008wr007399, 2009 An advanced passive diffusion sampler for the determination of dissolved gas concentrations P. Gardner 1 and D. K. Solomon 1 Received 28 August 2008; revised 10 February 2009; accepted 27 March 2009; published 18 June [1] We have designed and tested a passive headspace sampler for the collection of noble gases that allows for the precise calculation of dissolved gas concentrations from measured gas mixing ratios. Gas permeable silicon tubing allows for gas exchange between the headspace in the sampler volume and the dissolved gases in the adjacent water. After reaching equilibrium, the aqueous-phase concentration is related to the headspace concentration by Henry s law. Gas exchange between the water and headspace can be shut off in situ, preserving the total dissolved gas pressure upon retrieval. Gas samples are then sealed in an all metal container, retaining even highly mobile helium. Dissolved noble gas concentrations measured in these diffusion samplers are in good agreement with traditional copper tube aqueous-phase samples. These significantly reduce the laboratory labor in extracting the gases from a water sample and provide a simple and robust method for collecting dissolved gas concentrations in a variety of aqueous environments. Citation: Gardner, P., and D. K. Solomon (2009), An advanced passive diffusion sampler for the determination of dissolved gas concentrations, Water Resour. Res., 45, W06423, doi: /2008wr Introduction [2] Noble gas concentrations have been used for recharge temperature calculation [e.g., Mazor, 1972; Aeschbach- Hertig et al., 1999; Manning and Solomon, 2003], tritium helium age dating [e.g., Tolstikhin and Kamensky, 1969; Solomon et al., 1992], geothermal studies [e.g., Hearn et al., 1990], and mantle geodynamics [Turcotte and Schubert, 1982]. Dissolved noble gases have traditionally been sampled using either extraction or equilibrium techniques [Capasso and Inguaggiato, 1998]. In extraction sampling, a water sample is taken, and brought back to the laboratory where the gases are extracted and then analyzed. Common extraction sampling methods include (1) copper tube bailers [Weiss, 1970], (2) gas tight syringes, and (3) vacuum flasks [Sanford et al., 1996]. In an equilibration technique, the dissolved gas content in the surrounding water is equilibrated with a sample host [Sheldon, 2002]. In general, a gas permeable membrane allows for the exchange of dissolved gases in the surrounding water with the medium in the sampler volume. After equilibrium is achieved the sampler is removed and the equilibrium head space in the sampler can be directly analyzed. Takahata et al. [1997] use a one step method where a gas permeable membrane attached to vacuum line and quadrupole mass spectrometer is submerged in the fluid of interest and dissolved noble gases are extracted and measured in the field. In this paper we describe a sampler that passively collects the dissolved noble gases in water in situ and collects the equilibrium headspace. 1 Department of Geology and Geophysics, University of Utah, Salt Lake City, Utah, USA. Copyright 2009 by the American Geophysical Union /09/2008WR W06423 [3] Extraction sampling has been used with much success, but has some inherent problems. Many dissolved gases have low solubility and rapidly degas when hydrostatic pressure is reduced as must happen during pumping. The atmosphere is a major source of contamination, and even very small air bubbles that are trapped in the sample volume or distribution lines can cause contamination and/or stripping problems. Laboratory gas extraction increases the number of analytic steps and the possibility of contamination and/or degradation. Extraction methods collect the total amount of highly soluble gases like CO 2, while diffusion sampling collects only the equilibrium head space (a much smaller amount if gas is highly soluble). In areas of high dissolved CO 2 concentrations, the large amount of CO 2 in an extraction sample can slow the operation of noble gas cleanup lines. The pumping requirements of extraction sampling also mean that sampling sights must be relatively accessible, and adds logistics to sampling heavily contaminated water. [4] Equilibrium headspace sampling allows for passive, in situ gas extraction avoiding many of the problems associated with extraction sampling. Headspace samplers have been used successfully for a variety of studies. Sanford et al. [1996] use semipermeable silicon tubing to sample for dissolved gases during a tracer experiment. Gascoyne and Sheppard [1993] used ping-pong balls to collect soil gases for 4 He analysis, a technique pioneered by Dyck and Silva [1981] for investigating areas of groundwater recharge to lakes using dissolved 4 He. To our knowledge, none of the previously described headspace samplers are capable of providing an accurate measure of the total dissolved gas pressure (P tdg ) due mostly to the flexibility of the semipermeable membrane and associated changes in volume and hence pressure when the sampler is removed from a water column. Previous studies either assumed a value for P tdg 1of12

2 W06423 GARDNER AND SOLOMON: PASSIVE DIFFUSION SAMPLER W06423 [e.g., Spalding and Watson, 2006], or this value was independently measured using a total dissolved gas pressure probe [e.g., Sheldon, 2002; Manning et al., 2003; Manning and Solomon, 2003]. Because of changes in water temperature after recharge, and/or the production (e.g., CO 2 )or consumption (e.g., O 2 ) of dissolved gases in the subsurface, total dissolved gas pressure can easily differ from the barometric pressure by 20% or more. As a result, previous methods that assume a value for P tdg are only semiquantitative for providing an absolute value for C i. Total dissolved gas pressure probes can provide an accurate value for P tdg (typically to within 1%), but have depth limitation, can be slow to equilibrate [Manning et al., 2003], and are not readily available. In addition, existing passive samplers continue to exchange with the water column and atmosphere during retrieval, a problem in deep samples and/or samples with a large vadose zone. [5] We have developed a new headspace sampler for use in dissolved noble gas studies. Noble gas thermometry requires a precision and accuracy that is on the order of 1%, and helium has a very high diffusion coefficient and must be stored in either a metal, or highly specialized glass container in order to prevent diffusive exchange. This new passive sampler was designed to meet the stringent requirements of noble gas sampling. Our sampler (1) passively samples the dissolved gases in a water sample, (2) maintains the total dissolved gas pressure, (3) allows for discreet samples, and (4) greatly extends the depth range and precision of passive sampling. 2. Description and Operation [6] Figure 1 show a schematic of the sampler with gas exchange open (a) and closed (b). The sampler is roughly 15 cm in length with a maximum diameter of 3 cm. The major operational features of the sampler are the sample volume (where captured gas is collected and stored), the silicon rubber gas-exchange membrane, the gas-exchange piston and the hydraulic activation mechanism. The gas exchange membrane is 5 cm in length and consists of 0.8 mm thick silicon rubber tubing that is stretched over 8 mm diameter knurled stainless steel tubing giving a total exchange surface area of 150 cm 2. Knurls in stainless steel tubing provide channels for gas movements along the stainless steel tube, to the gas inlet port where it can access the sampler volume depending upon the sampler position. The sample volume is roughly 0.5 cm in diameter with length of 1 cm giving a total internal volume of 0.6 cc. This volume is all metal and vacuum tight and when closed with a pinch clamp is capable of storing gases for period of >1 month. The hydraulic activation mechanism consists of a piston with a one-way valve that allows fluid to fill tubing when the sampler in put in place and closes when pressurized from above allowing the activation piston to move the gas-exchange piston. The gas-exchange piston has a slide valve which either keeps the sample volume in line with the gas-exchange membrane, or isolates the sample volume in a leak tight position. The gas-exchange piston is moved via the hydraulic activation mechanism allowing for gas exchange shut off in situ. [7] The sampling procedure is relatively easy and many samples can be collected in a two day period. Samples are placed on the first day of a sampling trip, and then retrieved the following day in the same order ensuring at least 24 hours to equilibrate. Samplers, initially filled with ambient atmosphere, are placed in the fluid to be sampled with gas exchange open, attached to the ground surface with 0.5 cm diameter polyethylene tubing. Gases exchange across the silicon membrane between the sample volume and the fluid. The hydraulic activation mechanism has one way intake ports which allow the sampled fluid to fill the tubing, preventing the tubing from becoming a large volume for gas exchange that could alter the gas composition of the fluid and eliminating buoyancy problems in water. After 24 hours samplers are retrieved and clamped. To retrieve the sampler, pressure in the tubing is increased at the surface with a hand operated pump which closes the intake valves and moves the hydraulic activation mechanism. The gas exchange piston is pushed forward, moving a slide valve which isolates the gas inlet port from the sample volume without affecting the total sample volume. The sample volume is now isolated in a rigid, nickel, stainless steel, and compressed o-ring bounded volume that prevents gas exchange and decompressional expansion as the sampler is pulled to the surface. At the land surface, the nickel tubing containing the gas sample is clamped shut with a metal pinch clamp, giving a leak tight, very low diffusion container, which is capable of storing gas for extended periods. [8] It is important to consider the diffusion processes that this sampler utilizes and the effects advection, or lack thereof, has on sampler equilibrium. As soon as the sampler is placed in a fluid, gases partition into the sampler volume, which depletes the gas content of the surrounding fluid. If the sampler is placed in a steady state, advecting fluid, the fluid column surrounding the sampler is continually replaced, producing constant gas concentration at the boundary, resulting in fast equilibrium times. For aqueous samples, the maximum amount of water needed for exchange can be calculated as the amount of water needed to achieve equilibrium with helium, assuming a zero initial concentration in the sampler. Since helium is the least soluble of the noble gases, it will require exchange with the most water to achieve equilibrium. Using the internal volume of the sampler (1 cc), assuming atmospheric mixing ratios in equilibrium gas phase, and water with gas content given by equilibrium with the atmosphere at 10 C, the maximum amount of water needed to achieve equilibrium is 150 cc. This volume is generally more than what is needed in practice since our samplers start out with atmospheric composition initially, and thus do not need as much gas as this calculation assumes. Any environment where 150 cc of water flows by the sampler over the period of 24 hours will give complete equilibrium. [9] In environments where fluid movement is very low, diffusion of gas through the fluid will be the dominant mechanism to transport gas to the sampler. Harrington et al. [2000] discuss equilibration times in diffusion dominated environments for diffusion type samplers in slow moving groundwater in great detail and give analytic solutions for equilibration time dependent upon borehole geometry, sampler geometry and aquifer materials. As a rough estimate of times to equilibrium pffiffiffiffiffiffiffiffiffi in a diffusive environment, the diffusion length ( 4D o t) can be used to calculate the distance helium can diffuse in free water in one day. Using the freewater diffusion coefficient of cm/s for helium 2of12

3 W06423 GARDNER AND SOLOMON: PASSIVE DIFFUSION SAMPLER W06423 Figure 1. Schematic of passive diffusion sampler with piston in (a) the exchange position and (b) gas exchange closed position. in water [Jähne et al., 1987] and a time of one day, the diffusive length is 5 cm, giving the sampler access to the gas in a sphere of water radius 5 cm over a day of diffusion. This volume of water equals 523 cc and contains more than enough gas to equilibrate the sampler if the water is close to atmospheric equilibrium. In any environment of close to atmospheric concentrations with free water of radius 5 cm, diffusion will be suitable to equilibrate the sampler. If the radius of free water in the sampling environment gets much smaller than 5 cm, the diffusive transport must occur through the porous media and any wellbore material reducing the diffusion coefficient [Harrington et al., 2000]. In these cases diffusion may not allow equilibrium in 24 hours and a longer equilibration time or a means of inducing advective movement is required. [10] In any rapidly advecting fluid, not at steady state, concentration at the sampling point will very with time, possibly over many timescales. A sample taken in this fluid field at a single point represents some type of average concentration over the time the sample is taken. In the case of the classic copper tube sample, the sample is taken over minutes to seconds and represents something close to an instantaneous concentration. On the other hand, the diffusion process used by our sampler acts as a low-pass filter that captures the average concentration at that point over a 4 hour timescale, removing fluctuations on more rapid timescales. Thus this sampler could not be used to determine variability in concentration in a turbulent environment on any timescale less than one day, but is perfectly suited for capturing an average concentration over that timescale. The longer timescale of averaging by this sampler must be taken into consideration when sampling any fluid field in which concentration can vary on timescales of less than about 4 hours. When the concentration is steady at one point over at least 24 hours (most laminar flow), this sampler will collect an identical sample to a copper tube. [11] The small size and durable construction of these samplers allow them to work in diverse sampling environments, and to date they have been successfully used in springs ranging in temperature from 3 to 92 C, and in groundwater wells with hydraulic depths up to 260 meters. In general, this device can be used to moderately high hydrostatic pressure in any environment in which it fits if there is a moderate amount of advection. In low advection fields, the sampler can be used in environments which the free water radius is greater than 5 cm. For small diameter sampling in low flow situations, an advective field must be induced (e.g., by periodically pumping a small amount of water from the top of the water column) to insure equilibrium in 24 hours. 3. Theory [12] Equilibrium sampling utilizes Henry s law as follows: p i ¼ K i C i ; ð1þ 3of12

4 W06423 GARDNER AND SOLOMON: PASSIVE DIFFUSION SAMPLER W06423 where p i is the in situ partial pressure (atm), C i is the aqueous concentration (ccstp/g), and K i is the Henry s coefficient of the ith gas component in the head space and is a function of the temperature. At equilibrium the total pressure in the sampler is equal to the total dissolved gas pressure in the water and is the sum of the partial pressures of all dissolved gases: P total ¼ P tdg ¼ Xn where P tdg is the in situ total dissolved gas pressure in the water and sampler at equilibrium. In the laboratory the mixing ratio x i for a gas is measured. The equilibrium partial pressure is related to this mixing fraction by: p i ¼ P tdg x i : In order to calculate the aqueous concentration C i using equation (1), the measured values of x i must be converted to equilibrium partial pressures (p i ) and an estimate of P tdg is essential. [13] The partial pressure of a gas in equilibrium with the water can be calculated if we know the mixing ratio of that gas in the sample and the total in situ pressure. Consider the general case of a passive sampler that is not rigid and permits gas diffusion until final closure. Assuming ideal gas behavior, and no fractionation during sampling, the equilibrium pressure is given by equation (2), and we can write P tdg ¼ P si i¼1 p i ; ð2þ ð3þ ¼ n srt w V si ; ð4þ where P si is the total pressure in the sampler at the sample location, n s is the total number of moles of gas in the initial sampler volume V si. As the sampler is retrieved, there are several processes that affect the pressure inside the sampler. We have expansion due to a reduction in the hydrostatic confining pressure (V exp1 ), a change in pressure due to the temperature change from the water temperature (T w )tothe temperature at closure (T c ), and gas loss from continuous diffusion across the membrane until closure (4P exch ). Thus we have P s2 the pressure just before closing as a result of the above processes: P s2 ¼ P si V si Tc þ4p exch : V si þ V exp1 T w During closure we (1) decrease the volume by the closure mechanism (in our case by clamping the tubing closed) V clamp and (2) expand the total volume due to a pressure increase from closing V exp2. After clamping the pressure P s3 in the sample volume is given by ð5þ P s3 ¼ P s 2 V si þ V exp1 þ V exp2 V clamp : ð6þ V si þ V exp1 Finally there is a pressure change within the sample volume due to temperature changes from the clamp temperature to the temperature at the time the sample is analyzed is T L. The final pressure P s4 at inlet is then P s4 ¼ P s 3 T L T c : ð7þ Combining equations (5) (7), and solving for P tdg = P si we get the expression for P tdg as a function of the pressure inside the sampler at the time of analysis: T w 1 P tdg ¼ P s4 V si þ V exp1 þ V exp2 V clamp T L V si V si þ V exp1 4P exch : ð8þ V si Equation (8) shows that for the general case of a passive diffusion sampler, we must know the initial volume of the sampler, the volume of expansion due to hydrostatic pressure relaxation, the volume taken by closure and the subsequent expansion due to the increase in pressure, the temperature of the water, the temperature in the laboratory, and the pressure loss due to diffusion of gas until final closure. For noble gas work, where we need sampling accuracy approaching 1% it is almost impossible to know all these factors for a flexible, continuous diffusion device, and consequently for these type of devices we require an independent measure of P tdg. Our sampler was designed to simplify the relationship shown in equation (8) such that we can calculate P tdg upon inletting the sample in the laboratory. [14] The gas permeable membrane in our sampler is wrapped around a rigid stainless steel tube for hydrostatic pressure resistance making our sampler effectively rigid during removal. Gas exchange is shut off in situ in our sampler preventing gas exchange during retrieval from the sample position. Eliminating the pressure change terms due to expansion and continuous diffusion from equation (8), the expression for P tdg from the inlet pressure P s4 is T w Vsi V clamp P tdg ¼ P s4 : ð9þ T L V si To calculate the in situ total dissolved gas pressure we need to know the pressure inside the sampler at inlet, the temperature of the water being sampled, the temperature in the laboratory, the initial volume and the volume taken by the final closure (clamping) of the device. If the volume of closure is sufficiently small the right hand fraction of equation (9) is unity, and we need only know the pressure in the sampler at the time of analysis and the temperatures T w and T L. The internal volume of our sampler is 1 cc. In order to reduce the closure volume, the nickel tubing is precrimped such that there is a 0.5mm gap between the tubing wall, with a clamped area width of 2 mm and a tubing width of 5 mm. The total volume of closure is: cm * 0.2 cm * 0.5 cm = cc which is a 0.25% change in total volume due to closure. Since V clamp is sufficiently small in this case, we do not need to take it into account and from equations (9), (1) and (3) the dissolved concentration of the ith gas can be calculated from the measured mixing ratio x i, the gas pressure in the sampler, and the temperatures of the sampled water and the laboratory. 4of12

5 W06423 GARDNER AND SOLOMON: PASSIVE DIFFUSION SAMPLER W06423 the University of Utah Noble Gas Laboratory are given in Appendix A). From the pressure inside the sampler, equation (9) gives the total dissolved gas pressure. 4. Testing and Results 4.1. Gas Exchange [16] Passive gas exchange is accomplished via a semipermeable silicon membrane while the exchange piston is in the open position, keeping the sample volume in line with the exchange membrane. If the sampler membrane is thin and there is a zero initial concentration of gas in the sampler, the concentration of a gas within the sampler can be described by Ficks second law of diffusion, the solution of which is given by Sanford et al. [1996]: p i ¼ K i C i 1 exp DAt ; ð10þ 8L Figure 2. Observed and modeled pressure inside a sampler. Prior to starting the test, the sampler was evacuated and then placed in the atmosphere while gas exchange occurred. Pressure is gauge pressure with a small offset (0.142 MPa) from actual pressure MPa is the equilibrium gauge pressure at the elevation of Salt Lake City (1000 m). Calculated results are for the best fit diffusion coefficient using equation (10). [15] We determine the sample volume for each sampler manometrically. Using the sampler volume, the pressure within the sampler is easily obtained from the inlet pressure in the laboratory (details for the pressure determination at where p i = partial pressure of gas i in the sampler [P], P K i = Henry s constant from equation (1) [ VðSTPÞM 1 ] C i = dissolved gas concentration in volume (STP) per mass water [V (STP) M 1 ], D = effective diffusion coefficient of gas in the membrane [L 2 T 1 ], A = area of membrane exposed to the solution, t = time [T], 8 = sampler internal volume [L 3 ], L = membrane thickness [L]. [17] Equilibrium within the sampler is a function of the ratio DA/8L, and in order to speed equilibration, a sampler Figure 3. Fraction of equilibration for all noble gases from samplers submerged in air-equilibrated water removed after increasing periods of submersion. Dissolved gas concentrations were normalized by the theoretical concentrations at the laboratory temperature and pressure. Error bars are given from the 1s relative standard deviation of the standard runs for the day the samples were analyzed. 5of12

6 W06423 GARDNER AND SOLOMON: PASSIVE DIFFUSION SAMPLER W06423 Figure 4. Fraction of equilibrium for He from samplers placed in a flowing well near Tooele, Utah, removed after increasing periods of submersion. Error bars are given from the 1s relative standard deviation of the standard runs for the day the samples were analyzed. Figure 5. Total dissolved gas pressure (P tdg )determined from the pressure of gas inside the sampler versus that independently measured with a total dissolved gas pressure probe. 6of12

7 W06423 GARDNER AND SOLOMON: PASSIVE DIFFUSION SAMPLER W06423 Figure 6. Total dissolved gas pressure (P tdg ) precision. P tdg determined from the pressure of gas inside of the sampler from groups of samplers equilibrated with the same water. Relative standard deviations in precision for each measurement are given in Table 1. should be designed with a thin membrane and large surface area to volume ratio. [18] To test gas exchange for a gas phase sample, we evacuated the gas from a sampler and then recorded the increase in pressure in the sampler over time with the sampler immersed in the atmosphere. The pressure for a given gas inside the sampler P ir is given by the right hand side of equation (10) with H cc C w replaced with p o the constant partial pressure of the gas outside the sampler. We fit the modified form of equation (10) to the observed data and obtain the effective diffusion coefficient of the silicon tubing in atmosphere. The effective diffusion coefficient is a function of the solubility of the gas in silicon and the diffusion coefficient of the gas in the silicon. Figure 2 shows the total gauge pressure inside the sampler for runs with gas exchange open and closed, along with the calculated pressures using the best fit effective diffusion coefficient. The equilibrium gauge pressure at the elevation of Salt Lake City, Utah is 0.07 MPa. The observations are modeled well with a best fit diffusion coefficient of cm 2 /s. From Figure 2, the internal pressure is close to the equilibration gauge pressure on the order of a few hours after opening. Manning et al. [2003] derive the following expression from equation (10), t E ¼ 8L lnð1 EÞ; ð11þ AD i which can be used to calculate the time (t E ) needed to achieve a fraction of equilibrium (E) if we know the diffusion coefficient (D i ) in silicon. The fraction of equilibrium in is written E =(p i t p i o )/(p i p i o ) where p i t is the partial pressure at time t, p i o is the initial partial pressure in the sampler, and p i is the equilibrium partial pressure of the ith gas. For E = 0.99, t E = 2.5 hours. If we assume all gases have diffusion coefficient in silicon of roughly the same magnitude, we should expect our sampler to accurately contain the total dissolved gas pressure and gas content on the order of several hours after emplacement in a gas phase fluid. [19] We tested aqueous gas exchange in the laboratory for all noble gases. Samplers were filled with N 2 and then placed in air equilibrated water and retrieved after increasing periods of time. The aqueous concentration for each gas species, normalized by the theoretical equilibrium concentration at the temperature and pressure of the laboratory water bath are show in Figure 3. Because the internal volume (dead) volume of the sampler was filled with air, Table 1. Relative Standard Deviations for Total Dissolved Gas Pressure Precision Determined With Diffusion Sampler Group n =4 n =3 n =6 Precision Error of12

8 W06423 GARDNER AND SOLOMON: PASSIVE DIFFUSION SAMPLER W06423 Figure 7. there were nonzero concentrations of all measured gases at time zero. All noble gases, spanning atomic mass from 4 to 136 amu, equilibrate in roughly the same time frame. Neon appears to equilibrate slowest and xenon a bit faster, but all of the gases were close to full equilibration after about 16 hours (Figure 3). Differences in diffusion rates for gases would be due to differences in the effective diffusivity of the gases. If xenon does equilibrate faster, then the solubility of xenon in the silicon must be much higher than the other gases since the diffusion coefficient on xenon would be the lowest. [20] An aqueous effective diffusion coefficient of cm/s, was determined by fitting equation diffusion to the argon data for the laboratory experiment, and the resulting normalized curve is shown in Figure 3. The aqueous effective diffusion coefficient is on the same order, but lower than the atmospheric effective diffusion coefficient, and the time to 99% equilibrium (t e ) is 6.3 hours for this diffusion coefficient which is similar to that reported by Diffusion sampler versus copper tube determined 40 Ar. Sanford et al. [1996]. The equilibration in an aqueous solution differs substantially from the laboratory experiment in air. Diffusion into the sampler is a 3 step process that includes (1) transport through a boundary layer on the outside of the membrane, (2) transport through the membrane, and (3) transport in the gas phase into the internal volume of the sampler. The difference between the aqueous and atmospheric equilibration times likely reflects the role of aqueous diffusion through a boundary layer as a rate limiting transport mechanism. In water at 20 C, these samplers appear to be well equilibrated on the order of 10 hours after emplacement. [21] To test the time to equilibrium for samplers in small wells in the field, we immersed samplers for increasing periods of time in a flowing well from Six Mile Spring near Tooele, Utah that has a helium concentration that differs significantly from air. Figure 4 shows the calculated aqueous concentration of helium (normalized to fully equilibrated) after various periods of immersion. After 6 hours helium was Figure 8. Diffusion sampler versus copper tube determined R R a. Head space samples have been corrected for fractionation. 8of12

9 W06423 GARDNER AND SOLOMON: PASSIVE DIFFUSION SAMPLER W06423 Figure 9. Diffusion sampler versus copper tube determined 84 Kr concentration. only about 75% equilibrated, but appears to be equilibrated within 24 hours. The difference between the calculated helium concentration between samplers pulled at 24 and 120 hours is just above the 1s error on the run day, but is within the 2s error and probably represents an artifact either due to instrumental or field sampling error. We also measured the other noble gases during this test, but since the sampler was initially filled with air and the noble gas content of Six Mile Spring is close to air saturation, there is a very small difference between the initial and final gas compositions for these gases. The diffusion of gas through the silicon membrane will be temperature dependent. The well water temperature at the time of sampling was 10 C, which could result in slower diffusion and a longer equilibrium time. Under field conditions this sampler is equilibrated in 24 hours and does not differ significantly from samples equilibrated at longer time intervals (Figure 4) In Situ Gas Exchange Shut Off [22] In order to obtain discrete depth samples, and/or samples from wells with a long vadose zone, the gas exchange shut off mechanism must be capable of containing the gas in the sampler during retrieval. As discussed previously and shown in Figure 2, in the absence of a shut off mechanism the pressure inside the sample can quickly change when it is exposed to the atmosphere. Our sampler disconnects the membrane from the sample volume in situ, but diffusion of highly mobile gases such as He could still occur across the elastomer seals. We tested the integrity of these seals by waiting 15 and 30 minutes with the sample closed but exposed to the atmosphere before installing the metal clamps on the sample volume. No measurable change was observed, even in the helium isotope ratio. Our previous experience with containing helium using valves with elastomers seals suggests that samples cannot be maintained for an extended period of time unless all-metal seals are used Total Dissolved Gas Pressure [23] In order to calculate aqueous concentrations from the measured mixing ratios of the gas phase sample we must know the total dissolved gas pressure. Our sampler should allow for the determination of the total dissolved gas pressure by (1) knowing the volume of the sampler and (2) measuring the total pressure upon inlet to the analytical system. Figure 5 shows the total dissolved gas pressured determined using laboratory measurements versus the pressure measured with an independent total dissolved gas probe. The two measurements are reasonably well correlated with an R 2 of The repeatability of the P tdg calculated from the sampler is less than 1%. The total dissolved gas pressure and 1% error bars, for three groups of samplers, where each group was equilibrated at a constant dissolved gas pressure, is shown in Figure 6. The relative standard deviation for each group is shown in Table 1 and is less than 1% in all cases. [24] When interpreting the total dissolved gas pressure determined by either our sampler, or a total dissolved gas probe, the time to equilibrium must be considered. Discrepancies between the total dissolved gas pressure measured Table 2. Noble Gas Concentrations and Modeled Recharge Temperature for a Single Well Near Tooelle, Utah a Sample No. 40 Ar 84 Kr 129 Xe 20 Ne Recharge Pressure Recharge Salinity Recharge Temperature ccstp/g ccstp/g ccstp/g ccstp/g MPa % C Error a Errors calculated as the relative standard deviation. 9of12

10 W06423 GARDNER AND SOLOMON: PASSIVE DIFFUSION SAMPLER W06423 Figure 10. Noble gas recharge temperature precision. Noble gas recharge temperature determined from four samplers equilibrated with water from Six Mile spring. Recharge temperature was calculated using the gas content and P tdg determined from the pressure inside the sampler. from a total dissolved gas probe and that measured using the pressure within the sampler (Figure 5) could result from the different timescales over which the measurement is made. The total dissolved gas pressure inside the sampler or probe represents an integration of any transient in the total dissolved gas pressure over the interval of sampling; however, the sampler is generally in place for a period of one day, while the probe samples for 20 min. Deviations in the total dissolved gas pressure on diurnal timescale will be recorded by the diffusion sampler and not by the probe, and could be the cause of some of the discrepancy between the two measurement methods. Nevertheless, the largest difference between the probe and the diffusion sampler values is 3% Copper Tube Comparison [25] Dissolved noble gas samples have traditionally been made by extracting gases from water collected in clamped copper tubing. Aqueous concentrations of noble gases measured by equilibrium headspace samplers were compared to those determined from traditional copper tube samples, for samples taken from wells and springs with a wide range of noble gas compositions. Diffusion samplers give very similar dissolved concentrations for noble gases over the mass spectrum. Figure 7 shows the concentration of 40 Ar determined from diffusion samplers to that determined from copper tube samplers with measurement error bars of ±3% and the one to one line. For all but one sample on Figure 7, the dissolved gas concentration determined by diffusion samplers agree to within 3% of the copper tube sample. [26] As a comparison of diffusion sampling s ability to capture and retain more mobile gases, we compare the 3 He/ 4 He ratios (as R/R a ) for samples collected using diffusion and copper tube extraction sampling techniques. The R/R a determined by copper tube samples and diffusion samplers (with error bars of 1%) and the one to one line is shown in Figure 8. All samples fall very close to the one to one line and all but one are within 1% error. The equilibrium headspace ratio of 3 He/ 4 He is fractionated from the aqueous value by a known value due to differences in the solubility of 3 He and 4 He [Weiss, 1970]. In Figure 8 we have corrected the headspace value for this fractionation, back to the original dissolved concentration allowing for direct comparison. [27] Diffusion sampling is capable of collecting heavier noble gases. Figure 9 shows the comparison of Krypton values determined by copper tube extraction and diffusion samplers along with the one to one line and error bars of 5%. The Kr concentration determined from both methods compare favorably and agree to within 5% (Figure 9). In general, diffusion samplers agree well with classic extraction sampling techniques over a wide range of aqueous concentrations and atomic masses (Figures 7, 8 and 9) Noble Gas Thermometry [28] Noble gas thermometry requires great precision and accuracy. We examine the precision in calculated noble gas recharge temperature for four samplers equilibrated in the same water. The solubility of noble gases in groundwater is a function of the temperature and salinity of the water at the water table at the time of recharge. If recharge pressure and salinity are assumed, the measured aqueous concentration of all noble gases except helium, can used to determine the best fit recharge temperature, using the closed equilibrium model by Aeschbach-Hertig et al. [1999]. Noble gas concentrations measured using our diffusion sampler, the assumed recharge pressure and salinity, and the calculated recharge temperatures are shown in Table 2. Error for all measurements and the calculated noble gas recharge temperature are given by the 1s relative standard deviation. 10 of 12

11 W06423 GARDNER AND SOLOMON: PASSIVE DIFFUSION SAMPLER W06423 Aqueous concentrations agree to within measurement error (Table 2). Calculated noble gas recharge temperatures agree within 1 C as shown by Figure 10 with a relative standard deviation of 10%. This comparison integrates all noble gas concentrations and the total dissolved gas pressure determination giving proof of the repeatability of measurements made with this sampler. 5. Summary and Conclusion [29] We have designed a passive sampler that collects the equilibrium dissolved gas headspace for a water of interest using gas permeable silicon tubing. The unique feature of this sampler is that it preserves the in situ total dissolved gas pressure, which allows for a precise calculation of the aqueous concentration without an independent measurement of the P tdg. An in situ gas shut off mechanism allows for discreet sampling, and retrieval of samplers from great depths and long vadose zones, with no exchange. [30] We have designed our sampler so that the total dissolved gas pressure can be easily determined from the pressure inside the sampler at the time of analysis. We tested the total dissolved gas pressure determined from the sampler against independent measurements from a total dissolved gas pressure probe with the agreement being better than 3%. The precision of the total dissolved gas pressure determined by groups of samplers collecting gas from the same water is within 1% and shows that measurements made with our sampler are highly reproducible. [31] This sampler equilibrates within 24 hours for all noble gases. Samples are sealed in metal tubing and have a long shelf life, even for helium isotopes that readily exchanges through containers with elastomer seals or pyrex glass. Our sampler provides a simple and robust way of obtaining all the necessary information for the accurate calculation of the aqueous concentration of dissolved noble gas species in a broad range of environments and applications. Appendix A [32] Equation (9) gives the relationship to calculate the total dissolved gas pressure from the pressure inside the sampler at the time of analysis. We determine the pressure inside the sampler upon inlet using a capacitance manometer on the inlet branch of our mass spectrometer s clean up line. Just before inlet, the pressure inside the sampler is the sum of the dry pressure P sd and the water vapor pressure P H2 O by equation (3): P s4 ¼ P sd þ P sh2 O : ða1þ At the noble gas laboratory, the gas inside the sampler is expanded into the inlet arm of the mass spectrometer clean up line with volume V line, the water vapor in the sampler is frozen down, and the dry gas pressure P b is recorded. The total volume (V t ) is now given as V s + V line and the dry pressure inside the sampler is given by P sd ¼ P bðv s þ V line Þ : ða2þ V s The water vapor pressure in the sampler at inlet can be written as a function of the in situ vapor pressure P sih2 O using equation (9), which is valid for any partial pressure of a gaseous species if we assume no gas fractionation. The total pressure inside the sampler at the time of inlet can then be written P s4 ¼ P sih2 O T L T w Vi þ V clamp þ P bðv s þ V line Þ : ða3þ V s V i Finally we can write the equation for the total dissolved gas pressure using equations (9) and (A3): T L Vi þ V clamp P tdg ¼ P sih2 þ P bðv s þ V line Þ Tw O T w V i V s T L V i ; ¼ P bðv s þ V line Þ T w V i V i þ V clamp V s T L V i þ V clamp þ P sih2 O : ða4þ We use equation (A4) to calculate the in situ total dissolved gas pressure from the inlet pressure. V line is known, V s is measured manometrically in the laboratory, P b is measured, T w and T L are known, and V i /(V i + V clamp ) the clamping factor is assumed to be one. The in situ vapor pressure of water (P sih2 ) is a function of the sampling temperature and O is given by Gill [1982]: " 0:7859þ0:03477ðTwÞ# P ¼ 10 1þ0:00412ðTwÞ sih2 O 1013:25 * 760: ða5þ Combining equations (A4) and (A5) we can calculate the total dissolved gas pressure from the pressure measured on the inlet branch of the mass spectrometer. [33] Acknowledgments. Many people contributed to the development of the device described in this paper. Andy Manning, Amy Sheldon, and Phil Gardner (among others) each had a hand in this device s evolution. An especially big slice of credit goes to Alan Rigby whose patience in laboratory is unparalleled. Will Mace and Tom Marston provided invaluable support in the laboratory. References Aeschbach-Hertig, W., F. Peeters, U. Beyerle, and R. Kipfer (1999), Interpretation of dissolved atmospheric noble gases in natural waters, Water Resour. Res., 35(9), Capasso, G., and S. Inguaggiato (1998), A simple method for determination of dissolved gases in natural waters: An application to thermal waters from Vulcano Island, Appl. Geochem., 13(5), Dyck, W., and F. D. Silva (1981), The use of ping-pong balls and latex tubing for sampling the helium content of lake sediments, J. Geochem. Explor., 14, Gascoyne, M., and M. I. Sheppard (1993), Evidence of terriestrial discharge of deep groundwater on the Canadian Sheild from helium in soil gases, Environ. Sci. Technol., 27(12), Gill, A. E. (1982), Atmosphere-Ocean Dynamics, Academic, San Diego, Calif. Harrington, G., P. Cook, and N. Robinson (2000), Equilibration tines of gas-filled diffusion samplers in slow-moving ground water systems, Ground Water Monit. Rem., 20(2), Hearn, E. H., B. M. Kennedy, and A. H. Truesdell (1990), Coupled variations in helium isotopes and fluid chemistry Shoshone Geyser Basin, Yellowstone National Park, Geochim. Cosmochim. Acta, 54, Jähne, B., G. Heinz, and W. Dietrich (1987), Meaurement of the diffusion coefficients of sparingly soluble gases in water, J. Geophys. Res., 92(C10), 10,767 10, of 12

12 W06423 GARDNER AND SOLOMON: PASSIVE DIFFUSION SAMPLER W06423 Manning, A. H., and D. K. Solomon (2003), Using noble gases to investigate mountain-front recharge, J. Hydrol., 275, Manning, A. H., D. K. Solomon, and A. L. Sheldon (2003), Applications of a total dissolve pressure probe in ground water studies, Groundwater, 41(4), Mazor, E. (1972), Paleotemperatures and other hydrological parameters deduced from noble gases dissolved in groundwaters: Jordan Rift Valley, Israel, Geochim. Cosmochim. Acta, 36, Sanford, W. E., R. G. Shropshire, and D. K. Solomon (1996), Dissolved gas tracers: Simplified injection, sampling and analysis, Water Resour. Res., 32(6), Sheldon, A. L. (2002), Diffusion of radiogenic helium in shallow groundwater: Implications for crustal degassing, Ph.D. thesis, Univ. of Utah, Salt Lake City, Utah. Solomon, D. K., R. J. Poreda, S. L. Schiff, and J. A. Cherry (1992), Tritium and Helium 3 as groundwater age tracers in the Borden Aquifer, Water Resour. Res., 28(3), Spalding, B., and D. Watson (2006), Measurmement of dissolved H 2,O 2, and CO 2 in groundwater using passive samplers for gas chromatographic analyses, Environ. Sci. Technol., 40, Takahata, N., G. Igarashi, and Y. Sano (1997), Continuous monitoring of dissolved gas concentrations in groundwater using a quadrupole mass spectrometer, Appl. Geochem., 12(4), Tolstikhin, I. N., and L. Kamensky (1969), Determination of groundwater age by the T 3 -He method, Geochem. Int., 6, Turcotte, D. L., and G. Schubert (1982), Geodynamics, John Wiley, New York. Weiss, R. F. (1970), Helium isoptope effect in solution in water and seawater, Science, 168, P. Gardner and D. K. Solomon, Department of Geology and Geophysics, University of Utah, William Browning Building 711, Salt Lake City, UT 84112, USA. (w.gardner@utah.edu; kip.solomon@utah.edu) 12 of 12