Tracing injected CO2 in the Cranfield enhanced oil recovery field (MS, USA) using He, Ne and Ar isotopes

The He, Ne and Ar isotopic composition of gases collected in 2009 and 2012 from 13 production wells, injection wells and the CO2 supply pipeline at the Cranfield CO2-enhanced oil recovery field (MS, USA) have been measured in order to determine the extent to which they trace the fate of injected CO2 in the reservoir. In the absence of samples of CO2 pre-injection reservoir gas we use the Ne isotope composition of the production and injection well gases to determine the isotopic composition of the natural gas. The noble gas isotopes display binary mixing trends between the injected CO2 and a CH4-rich natural gas that is characterised by radiogenic He, Ne and Ar isotope ratios. 3He/4He and 40Ar*/4He ratios (where 40Ar* represents the non-atmospheric 40Ar) display coherent relationships with CO2 concentrations that can be used to trace and quantify the injected CO2 in an engineered site over a sustained period of injection. The presence of a small amount of air-derived Ar, from a non-atmospheric source, in many gas samples rules out using 40Ar/36Ar to track the injected CO2. The noble gases identify the loss of a significant proportion of the CO2 from the gas phase sampled by five production wells in 2009. Using 3He/4He and 40Ar*/4He ratios to reconstruct the major gas composition, it appears that between 22% and 96% of the CO2 has been lost in individual wells. This study demonstrates that the naturally occurring noble gases have the potential to trace the fate and quantify the sequestration of CO2 at injection sites.


Introduction
Geological storage of CO 2 is the only means to directly reduce man-made CO 2 emissions to the atmosphere from industrial point sources (Scott et al., 2013). Developing techniques for monitoring and verifying the secure storage of injected CO 2 is essential if carbon capture and storage (CCS) is to become a realistic climate mitigation option. The verification process requires an understanding of the mechanism of storage and the ability to track the fate of CO 2 in the subsurface. Numerical simulations and laboratory experiments that model fluid flow, physical-chemical processes acting on the injected CO 2 (Ennis- King and Paterson, 2007;Knauss et al., 2005;White et al., 2005;Xu et al., 2005) and trapping mechanisms within engineered fields (Arts et al., 2004;Emberley et al., 2005) have been developed. These models require validation by field measurements.
Baseline studies at CCS test sites provide constraints on the movement of injected CO 2 (Stalker et al., 2015). Injection of CO 2 into depleted hydrocarbon fields has long been used for enhanced oil recovery (EOR). Given the large volumes of CO 2 transported and injected in these sites they can serve as analogues for geological CO 2 storage (Baines and Worden, 2004;Haszeldine et al., 2005). Additional insights on the storage and movement of anthropogenic gas can be acquired by studying naturally occurring CO 2 accumulations in the shallow crust (e.g. Gilfillan et al., 2008).
Several geochemical techniques have been developed for tracing the movement and storage of CO 2 in the crust (e.g. Humez et al., 2014 and references therein). The stable carbon isotope composition of CO 2 (␦ 13 C CO 2 ) has been widely and successfully used to trace injected gas, and track its interaction with reservoir fluid and rocks (Assayag et al., 2009;Johnson et al., 2011;Lu et al., 2012b;Myrttinen et al., 2010), as well as constrain the storage mechanism (e.g. Raistrick et al., 2006). However, the technique is less powerful if the ␦ 13 C of the injected and in-place reservoir CO 2 are similar (e.g. Sherwood Lollar et al., 1997;Wycherley et al., 1999), or when CO 2 is liberated from carbonate reservoir rocks during reaction with the acidified formation water generated by CO 2 dissolution (Shevalier et al., 2013). Chemical tracers such as SF 6 , perfluorocarbons and CD 4 have been added to CO 2 and used in CCS analogue studies to determine the timing of breakthrough of the injected gas (Stalker et al., 2015 and references therein). The noble gases (He, Ne, Ar, Kr and Xe) are chemically inert and have several isotopes that allow them to be used to trace the physical processes that have affected crustal fluids, and to constrain gas residence time-scales. They have been used, for instance, to resolve the origin and movement of CO 2 in natural gas reservoirs (Gilfillan et al., 2009;Lafortune et al., 2009;Sathaye et al., 2014) as well as determine the long-term storage of CO 2 in natural gas fields (Ballentine et al., 2001;Gilfillan et al., 2008;Zhou et al., 2012). The addition of Kr and Xe isotopes to injected gas has been used to trace the movement of injected CO 2 in CCS trials (Lu et al., 2012a;Nimz and Hudson, 2005;Stalker et al., 2015).
The noble gases are present in trace amounts in natural CO 2 (e.g. Zhou et al., 2012). Where the isotopic composition of the noble gases in injected natural CO 2 are significantly different to those of the in-place reservoir gas they have the potential to trace the movement and mixing of the injected gas (Nimz and Hudson, 2005). Here we present the first comprehensive study to apply this technique in an EOR field that is an analogue for geologic CO 2 storage sites. We report the He, Ne and Ar isotope composition of well gases from the Cranfield EOR field, Mississippi, USA, over a period of three years after the start of CO 2 injection. The data allow us to trace the presence of injected CO 2 in the reservoir and the loss of a component of this injected CO 2 .

The Cranfield oil and gas field
Cranfield is located in southwest Mississippi, ∼20 km east of the town of Natchez (Fig. 1). The reservoir is an anticline of highly permeable fluvial sediments of the Upper Cretaceous Lower Tuscaloosa Formation. The reservoir rock is largely composed of 15-25 m thick conglomerates, and occurs at depth of ∼3000 m Lu et al., 2012b). It was discovered in 1943 and oil and gas production commenced in 1944. The field is a simple dome structure (∼6.4 km diameter), with a gas cap and associated down-dip oil ring. Pressure was maintained by gas recycling until the gas cap was depleted in 1959. It was abandoned in 1966 after unsuccessful water injection tests in 1958-59. The base and cap rock of the injection zone are low permeability mudstones. A SE-NW trending normal fault bisects the reservoir .
CO 2 injection for EOR started in July 2008 by Denbury Resources Incorporated. High purity (99%) CO 2 is transported ∼160 km from the Jackson Dome deposit in central Mississippi (Fig. 1). CO 2 injection started at the north part of the field (Fig. 1) and extended to the southeast over the following three years. There was a corresponding development of injection and production wells to produce the oil mobilised by the CO 2 injection. Water and gas produced with the oil are separated at the surface on site. Produced water is disposed of via a deep water injection well, while the produced gas (CO 2 and CH 4 ) is mixed with Jackson Dome CO 2 and re-injected (this is termed recycled gas) .
In April 2010 the injection rate was approximately one million metric tonnes per year. By January 2013 6.8 million metric tonnes of CO 2 had been injected, of which 4 million had been retained in the reservoir Hovorka et al., 2013). Between July 2009 and January 2013 the proportion of recycled gas in the injected gas had increased from ∼9 to ∼30% .
Prior to injection the reservoir re-equilibrated to hydrostatic pressure by groundwater infiltration over the ∼40 years of abandonment. Continuous CO 2 injection without water injection means that Cranfield can be considered as an analogue for injection into a saline aquifer (Hovorka et al., 2011Lu et al., 2012b). This is significant as saline formations offer the vast majority of available global CO 2 storage capacity (Scott et al., 2013). Consequently CO 2 monitoring and flow, sequestration modelling, trapping mechanism constraining and leak detection research have been undertaken extensively in the past few years Hovorka et al., 2013;Lu et al., 2013;Nicot et al., 2013;Zhang et al., 2013Zhang et al., , 2014. The binary mixing of the injected CO 2 with the naturally-occurring methane-rich reservoir gas is supported by ␦ 13 C CO 2 measurements of produced gases (Lu et al., 2012b).

Sampling and analytical methods
The samples analysed for this study were collected in December 2009 and March 2012, 19 and 45 months after the start of CO 2 injection. Only the northeast part of the field was in operation during the first sampling trip. CO 2 injection and production was extended to the southeast by March 2012 (Fig. 1). The incoming pipeline from Jackson Dome to Cranfield (termed the bulkline) was sampled in 2012. Injection wells were sampled in both years with 22 gas samples being collected from production wells: 10 in 2009 and 12 in 2012. Nine of the 2009 wells were re-sampled in 2012 whilst three of the wells sampled in 2012 were not in use in 2009. Detailed well information is shown in Table 1.
The CO 2 in the bulkline and in the injection wells were maintained at around 21 MPa and 35 • C. Production wells were operated at between 0.8 and 9.1 MPa and around 60 • C. Oil, water and gas were present in the production wells therefore gas samples were taken from the top of the pipeline in order to avoid sampling liquid. Samples were collected in 8 mm outside diameter copper tubes that were connected to a pressure regulator using flexible pipe. The tube and gauge were flushed with gas for 5 min and then clamped at the end using a purpose built steel clamp, forming a cold weld. A similar clamp at the upstream end of the copper tube was closed when the pressure reached 2-3 bar, again forming a cold weld and maintaining a helium tight seal. Copper sample tubes were connected to an all-metal vacuum line maintained at ∼35 • C and pumped to ultra high vacuum (p < 10 −7 mbar) using a turbo-molecular pump. The gas from the tube was expanded into a known volume and the absolute pressure was measured using a MKS 615-A Baratron capacitance manometer. The uncertainty on the pressure measurements is typically ±0.5% (1 ). A small fraction of the gas sample was used to determine the concentration of CO 2 and CH 4 using a Pfeiffer Vacuum QMS 200 quadrupole mass spectrometer. The mass spectrometer was calibrated by series of analyses of binary mixtures of CO 2 and CH 4 over the pressure range typical of samples. In several samples a small proportion of a sample gas was trapped into ∼10 cm 3 glass tubes and CO 2 absorbed at −160 • C using liquid nitrogen and iso-pentane. Hydrocarbon gases were pumped off and ␦ 13 C CO 2 was determined using a VG Optima dual inlet isotope ratios mass spectrometer in dynamic mode.
The remaining gas was expanded into a VG Scienta ST22 titanium sublimation pump that ran sequentially between ∼900 • C and 15 • C. This purified gas was exposed to a SAES GP50 ZrAl alloy getter, held at 250 • C, for 20 min and subsequently expanded into a 0.5 litre reservoir. The concentration of 4 He, 20 Ne and 40 Ar were determined on the residual gas that was not trapped in the reservoir using the quadrupole mass spectrometer in order to guide subsequent isotopic analysis from the reservoir.
Typically He isotopes were measured in a ∼1 cm 3 aliquot of gas from noble gas reservoir. This was purified by exposure to four hot GP50 ZrAl alloy getters and the heavy noble gases (Ar, Kr, Xe) were absorbed onto liquid nitrogen-cooled charcoal prior to analysis. Neon isotopes were typically measured in a larger volume of gas than for helium. Purification followed the procedure for He isotopes, except that Ne was trapped on charcoal at -243 • C for 20 min, helium was pumped off and the Ne released at −173 • C for 15 min prior to analysis. The gas was desorbed from the nitrogen-cooled charcoal by heating to room temperature for 15 min and the Ar isotope composition was determined.
The isotopic composition of the noble gases was measured by a MAP 215-50 mass spectrometer in static mode following established procedures (Codilean et al., 2008;Williams et al., 2005). 4 He, 36 Ar, 38 Ar and 40 Ar were measured on a Faraday detector with 10 11 amplification, and 3 He, 20 Ne, 21 Ne, 22 Ne on a Burle channeltron electron multiplier in pulse counting mode at 2.5 kV. The hydrogen partial pressure was minimised by a room temperature GP50 getter in the mass spectrometer source volume. During He and Ne isotope analysis liquid nitrogen-cooled charcoal was used to minimise the heavy noble gases, hydrocarbons and CO 2 in the mass spectrometer.
The mass spectrometer sensitivity and mass fractionation was determined by repeated analysis of aliquots from reservoirs of HESJ international standard (Matsuda et al., 2002) for He, and air for Ne and Ar, during each analytical period. The reproducibility of noble gas isotope ratios and amounts presented in Table 2 are typically ±0.5% (1 ). Blanks for all isotopes were negligible.

Results
A sample of the bulkline gas taken in 2012 contains >99% CO 2 with a trace amount of CH 4 and noble gases. ␦ 13 C CO 2 is −2.9‰, which is similar to previous determinations (Lu et al., 2012b;Zhou et al., 2012). The noble gas isotope ratios (Table 2) are similar to those measured from the Jackson Dome CO 2 deposit (Zhou et al., 2012).
The injected gases sampled in 2009 and 2012 are significantly different from the bulkline gas. The CO 2 content is lower; 96% in 2009 and 87% in 2012, with correspondingly higher CH 4 contents. 3 He/ 4 He (4.73-4.0R A , where R A is the atmospheric value of 1.39 × 10 −6 ) and 40 Ar/ 36 Ar (2906-2763) also decrease with time (Table 2). This is qualitatively consistent with the increased proportion of (methane-rich) reservoir gas in the injected gas as a result of recycling. 20 Ne/ 22 Ne and 21 Ne/ 22 Ne ratios do not change with time (10.61 and 10.69, and 0.035 and 0.037 respectively) taking experimental uncertainties at the 2 level into account.
The gases from the production wells typically show an increase in the proportion of injected gas with time. This is apparent as an increase in the proportion of CO 2 in the nine wells that were sampled in both 2009 and 2012 (e.g. from 0.9 to 81.9% CO 2 in well 28F-2). This is mirrored by increases in 3 He/ 4 He, 20 Ne/ 22 Ne, 40 Ar/ 36 Ar and decreases in 21 Ne/ 22 Ne as well as 4 He, 20 Ne and 40 Ar concentrations (Table 2). In all the produced (and injected and bulkline) gas samples, 4 He/ 20 Ne ratios are at least five orders of magnitude higher than the atmospheric value ruling out any contamination from the atmosphere during field sampling or gas extraction from the Cu tube. ␦ 13 C CO 2 measured in several gas samples (Table 2) is rather constant (−2.5 to −3‰). When plotted against CO 2 content (Fig. 2) the data are consistent with mixing of injected gas and the in-place reservoir gas with ␦ 13 C CO 2 = −10.5 ‰ determined previously (Lu et al., 2012b).

Noble gas isotope systematics
Neon in natural gases is typically a mixture of three isotopically distinct sources: the crust, the mantle and the atmosphere (Ballentine and O'Nions, 1991). The Ne isotopic composition of CO 2 Table 2 Major gas, carbon isotope and noble gas isotope data from bulkline, injected and production well gases from the Cranfield EOR field.  (1) 16452 (36) 930 (1) Uncertainties (1 ) in last significant figures are in parentheses. Relative uncertainties of CO2 measurements are 1%, while the absolute uncertainty of ␦ 13 CCO 2 is ±0.17‰. The composition of air is after Eberhardt et al. (1965), Mamyrin et al. (1970) and Mark et al. (2011). Standard conditions are after Ozima and Podosek (2001). Noble gas concentrations are fraction of volume and are given in cm 3 (STP)/cm 3 .
from the Jackson Dome field is a mixture of mantle-derived Ne with air-derived Ne that has been isotopically fractionated prior to mixing (Zhou et al., 2012). Previous measurements of Jackson Dome gas are plotted on Fig. 3, with a grey dashed line showing the mixing trend. The Ne isotope composition of gases from the injection and production wells from Cranfield define a trend between a point on the Jackson Dome mixing line that is indistinguishable from the measured bulkline gas ( 20 Ne/ 22 Ne = 10.92; 21 Ne/ 22 Ne = 0.037), and a point along a mixing line between air and crustal (radiogenic) Ne that must represent the in-place Fig. 2. Plot of ␦ 13 CCO 2 (‰) vs. CO2 concentration for production well gases from the Cranfield field. The mixing curve in inset figure is that generated by Lu et al. (2012b) for mixing of a natural gas with 4% CO2 and ␦ 13 C = −10.5‰ and bulkline composition of 99.3% CO2 and ␦ 13 C = −2.6‰. 1 uncertainties are smaller than symbols.
natural gas (Fig. 3). Following Ballentine et al. (2005), a best-fit line through the data, using the least square method can be used to determine the Ne isotope composition of the natural gas endmember ( 20 Ne/ 22 Ne = 9.62 ± 0.02 and 21 Ne/ 22 Ne = 0.0384 ± 0.001).  The identification of this end-member allows us to determine the He and Ar isotopic composition of the in-place natural gas by plotting values against 20 Ne/ 22 Ne (Fig. 4). These values are insensitive to natural variation in the crustal 21   1997) the 20 Ne/ 22 Ne of the reservoir gas is 9.68 ± 0.02. This minor shift has no significant impact on the He and Ar isotope composition of the reservoir gas (Fig. 4) and the interpretation of noble gas-CO 2 data (Fig. 5). Fig. 4A shows the 3 He/ 4 He ratio of the injected and production well gases plotted against 20 Ne/ 22 Ne ratios. The data, in particular the production well gases from 2012, define a trend that is consistent with a binary mixture of injected gas and the inplace reservoir gas. Using the Ne isotope composition of the end-members defined above, the best-fit mixing curve defines the 3 He/ 4 He of the natural gas of 0.05 ± 0.005R A . This value is close to the average ratio of continental crust (∼0.02R A ; after Andrews, 1985) and is lower than the lowest measured 3 He/ 4 He in the production well gases (28F-2 2009; 0.2R A ). The hyperbolic nature of the mixing curve indicates that the 3 He/ 4 He of the natural gas endmember is rather poorly defined. The degree of the curvature of the mixing line is defined by 4 He/ 22 Ne Jackson Dome / 4 He/ 22 Ne natural gas (0.488). Combining this value with the measured 4 He/ 22 Ne of the Jackson Dome gas (12,452;Zhou et al., 2012), the 4 He/ 22 Ne of the natural gas is 25,524. The lowest CO 2 sample has measured 4 He/ 22 Ne = 28,178 which is consistent with the calculated value.
Using a similar technique, the 40 Ar/ 36 Ar of the natural gas end-member can be obtained (Fig. 4B). Although the Ne-Ar isotope data is not as coherent as the He-Ne isotope data, the data is also consistent with mixing between the injected gas and reservoir gas. The best-fit line defines the natural gas 40 Ar/ 36 Ar of 836 ± 75. This is higher than the atmospheric 40 Ar/ 36 Ar (298.6; Mark et al., 2011) and is consistent with the presence of a significant concentration of crust-derived radiogenic 40 Ar in the natural gas. The curvature of the mixing line is defined by 36 Ar/ 22 Ne Jackson Dome / 36 Ar/ 22 Ne natural gas = 0.2. Combining this value with the measured 36 Ar/ 22 Ne of the Jackson Dome gas (8.36), the 36 Ar/ 22 Ne of the natural gas is 41.7.
Air-derived noble gases are present to varying extents as a third component that tends to obscure the mixing trend in Ne-Ar isotope space (e.g. well 27-3, Fig. 4B). The 20 Ne/ 36 Ar data close to air-saturated water values are suggesting that the atmospheric noble gases likely originate from degassed formation water rather than directly from air. The low concentration of He in air (5.24 × 10 −6 cm 3 (STP)/cm 3 ) means that air addition has no significant effect in Fig. 4A. The atmospheric 40 Ar contribution can be removed leaving radiogenic Ar, denoted as 40 Ar * (Stuart and Turner, 1992). A plot of 40 Ar * / 4 He vs. 20 Ne/ 22 Ne reveals a coherent trend (Fig. 4C) that yields a 40 Ar * / 4 He = 0.09 ± 0.01 for the natural gas end-member.
The He, Ne and Ar isotope compositions of the in-place gas end-member are radiogenic, and strongly indicative of a crustal source with no resolvable mantle-derived volatiles. This is typical of natural gas accumulations from stable continental cratons (Prinzhofer, 2013). The 40 Ar * / 4 He is significantly lower than produced in normal continental crust (∼0.2; e.g. Torgersen et al., 1989) and likely reflects the preferential release of 4 He from minerals in the reservoir rocks at lower temperature than 40 Ar by diffusion and/or ejection of ␣-particles.

Tracking the fate of injected CO 2
The coherent relationships between noble gas isotope compositions in the produced gas (Figs. 3 and 4) support the hypothetical binary mixture between the injected gas and in-place gas. Consequently the naturally occurring noble gases have the potential to trace the injected CO 2 . A plot of the noble gas isotope ratios against the molar proportion of CO 2 in the gas samples helps to elucidate the relationship (Fig. 5).
The helium isotopic composition of the production well gases is plotted against the CO 2 content (Fig. 5A). Theoretical mixing curves are drawn between the 2009 and 2012 injected gases and a natural gas end-member ( 3 He/ 4 He = 0.05R A ). The curvature of the mixing lines are calculated using the measured 4 He concentration in the injected gas (Table 2) and the natural gas, which is assumed to be that of the lowest CO 2 sample (28F-2 2009; 3 × 10 −4 cm 3 (STP)/cm 3 ). The 2009 and 2012 mixing lines overlap as the evolution of the injected gas mirrors the mixing of in situ and injected gases in the reservoir. With the exception of the five lowest CO 2 concentration 2009 samples, the data plot within the field defined by the mixing curves and are consistent with the predicted two-component mixing. This demonstrates that He isotopes fingerprint the injected CO 2 and have the potential to be used as a tracer of fugitive emissions of CO 2 from CCS sites where there is significant difference between the helium isotopic composition of the injected and in-place gases.
The 40 Ar/ 36 Ar of the production well gases are plotted against CO 2 content in Fig. 5B along with theoretical mixing lines between the 2009 and 2012 injected gas composition with the in-place natural gas. As in Fig. 5A, the mixing curves are defined by data from the injectors and the 40 Ar/ 36 Ar of the natural gas (836) determined from the Ar-Ne isotope relationship (Fig. 4B). The concentration of 36 Ar of the reservoir gas is taken from the lowest CO 2 sample (28F-2 2009; 7 × 10 −8 cm 3 (STP)/cm 3 ). Few data plot on the mixing line, and the systematic relationship observed for 3 He/ 4 He-CO 2 is not apparent. For instance, the four low CO 2 samples plot above the mixing line, as observed in Fig. 5A. Gas samples from well 27-3 taken in 2009 and 2012 plot significantly below the mixing curve. This may be due to the addition of small amount of air-derived Ar that will tend to lower 40 Ar/ 36 Ar ratios but have little effect on CO 2 due to its low concentration in air. In addition, four samples (two from each sampling campaign) have higher 40 Ar/ 36 Ar than the injected gases, yet below the value of Jackson Dome CO 2 from the bulkline. This implies that the production wells are tapping gas that was injected earlier than that sampled at the injectors, when the Ar isotope composition of the injected gas was closer to that of Jackson Dome CO 2 as the presence of atmospheric air contamination has been ruled out by the consistently high 4 He/ 20 Ne ratios.
The 40 Ar * / 4 He ratio allows the effect of small contributions of air-derived Ar to be removed (e.g. Stuart and Turner, 1992). The 40 Ar * / 4 He is plotted vs. CO 2 along with theoretical mixing lines between the injected gas and a natural gas end-members in Fig. 5C. The curvature of each mixing line is defined by the concentration of 4 He (see above). The data generally follow the mixing lines to a greater extent than Ar isotopes. As observed previously, the five low CO 2 samples plot above the hypothetical mixing line.
The low CO 2 concentration gases sampled in 2009 are the general exception to the systematic mixing relationships identified in CO 2 vs. 3 He/ 4 He and 40 Ar * / 4 He (Fig. 5). In all cases, including the CO 2 vs. 40 Ar/ 36 Ar plot, the data plot significantly above the mixing lines. The internal consistency of the noble gas isotope data (Fig. 4) implies that this is unlikely to be the result of the addition of another noble gas component. The simplest and most likely explanation is that it reflects the loss of CO 2 from the gas phase during the early phase of CO 2 injection. The proportion of CO 2 that has been lost can be quantified using the measured 3 He/ 4 He and 40 Ar * / 4 He ratios. The distance on the ordinate between the data point and the mixing curve in Fig. 5 represents the absolute change in CO 2 concentration. The proportion lost (%) is calculated by normalising to the predicted CO 2 content. For instance, gas from well 28F-2 2009 has 0.8% CO 2 and 3 He/ 4 He of 0.21R A . The 2009 mixing curve for that 3 He/ 4 He predicts 11.8% CO 2 , equating to a loss of 93%. The predicted CO 2 concentrations and the lost proportion have been calculated using 3 He/ 4 He and 40 Ar * / 4 He ratios (Table 3 and Fig. 6).
The 3 He/ 4 He data require the loss of between 28% (27-5 2009) and 93% (28F-2 2009) of the injected CO 2 from the gas phase. From the 40 Ar * / 4 He this range is 22-96%. The degree of CO 2 loss from each sample calculated from the two tracers appears to be broadly consistent (Fig. 6). Uncertainties have been propagated by taking into account the uncertainty of CO 2 concentration and noble gas isotope ratios of each sample, and the mixing curve uncertainty which originates from the uncertainty of the end-member compositions. It should be noted that any process that causes CO 2 loss may fractionate the 40 Ar * / 4 He ratio of the residual gas. We do not take that into account here as it is likely negligible compared to the changes in CO 2 concentration. Whilst the mechanism by which CO 2 has been sequestered (e.g. dissolution into the formation water, precipitation as carbonate, residual trapping) is not the aim of this study, the data provide a clear indication that the naturally-occurring noble gases have the potential to track the fate of injected CO 2 in future CCS trials. Further work will examine this topic in detail but is outside the scope of the current work.

Conclusions
The unique noble gas isotope composition of the Jackson Dome CO 2 injected in the Cranfield CO 2 -EOR field permits evaluation of the effectiveness of using He, Ne and Ar isotopes to trace injected CO 2 . The isotopic compositions of production well gases sampled in 2009 and 2012, 19 and 45 months after the start of CO 2 injection define binary mixtures between injected gas and the in-place reservoir gas. This allows the noble gas isotopic composition of the reservoir gas to be determined. 3 He/ 4 He and 40 Ar * / 4 He of produced gas samples plot on theoretical mixing lines with CO 2 content indicating that the noble gas isotopes track the presence of the injected CO 2 . The five sample with the lowest proportion of CO 2 sampled in 2009 plot consistently above the theoretical mixing lines. This requires the loss of 22-96% of the injected CO 2 from the gas phase, and suggests that the noble gas isotopes have potential to quantify the storage of CO 2 in CCS sites. The total amount of sequestered CO 2 and the sequestration mechanism is currently unresolved but is the focus of ongoing work.