The Inherent Tracer Fingerprint of Captured CO2.

Carbon capture and storage (CCS) is the only currently available technology that can directly reduce anthropogenic CO 2 emissions arising from fossil fuel combustion. Monitoring and veri ﬁ cation of CO 2 stored in geological reservoirs will be a regulatory requirement and so the development of reliable monitoring techniques is essential. The isotopic and trace gas composition − the inherent ﬁ ngerprint − of captured CO 2 streams is a potentially powerful, low cost geochemical technique for tracking the fate of injected gas in CCS projects; carbon and oxygen isotopes, in particular, have been used as geochemical tracers in a number of pilot CO 2 storage sites, and noble gases are known to be powerful tracers of natural CO 2 migration. However, the inherent tracer ﬁ n-gerprint in captured CO 2 streams has yet to be robustly investigated and documented and key questions remain, including how consistent is the ﬁ ngerprint, what controls it, and will it be retained en route to and within the storage reservoir? Here we present the ﬁ rst systematic measurements of the carbon and oxygen isotopes and the trace noble gas composition of anthropogenic CO 2 captured from combustion power stations and fertiliser plants. The analysed CO 2 is derived from coal, biomass and natural gas feedstocks, using amine capture, oxyfuel and gasi ﬁ cation processes, from six di ﬀ erent CO 2 capture plants spanning four di ﬀ erent countries. We ﬁ nd that δ 13 C values are primarily controlled by the δ 13 C of the feedstock while δ 18 O values are predominantly similar to atmospheric O 2 . Noble gases are of low concentration and exhibit relative element abundances di ﬀ erent to expected reservoir baselines and air, with isotopic compositions that are similar to air or fractionated air. The use of inherent tracers for monitoring and veri ﬁ cation was provisionally assessed by analysing CO 2 samples produced from two ﬁ eld storage sites after CO 2 injection. These experiments at Otway, Australia, and Aquistore, Canada, highlight the need for reliable baseline data. Noble gas data indicates noble gas stripping of the formation water and entrainment of Kr and Xe from an earlier injection experiment at Otway, and inheritance of a distinctive crustal radiogenic noble gas ﬁ ngerprint at Aquistore. This ﬁ ngerprint can be used to identify un-planned migration of the CO 2 to the shallow subsurface or surface.


Introduction
Global carbon dioxide emissions must be drastically reduced to avoid global warming (IPCC, 2013).Carbon capture and storage (CCS), increasingly combined with bio-energy (BECCS), has potential for emission mitigation (Azar et al., 2013;International Energy Agency, 2014;IPCC, 2013;Scott et al., 2012).CCS is the only currently available technology that can reduce emissions from various industrial processes, such as cement and steel manufacture and many forms of chemical synthesis.A lack of accurate and cost effective methods for monitoring the geologically stored CO 2 , and unambiguously identifying the source of CO 2 in a suspected leak, remains a barrier to the roll out of CCS at commercial scales.A potential solution is to use the geochemical properties of the CO 2 itself as a geochemical tracer.Carbon (δ 13 C) and oxygen (δ 18 O) stable isotope compositions have proven to be useful tracers and monitoring tools at a number of pilot storage sites (Boreham et al., 2011;Emberley et al., 2005;Johnson et al., 2011aJohnson et al., , 2011b;;Kharaka et al., 2009Kharaka et al., , 2006;;Lu et al., 2012;Martens et al., 2012;Mayer et al., 2013;Myrttinen et al., 2010;Nowak et al., 2014;Raistrick et al., 2006;Shevalier et al., 2013Shevalier et al., , 2009) ) and noble gases have the potential to be powerful tracers for CCS monitoring, based on their use in tracing natural CO 2 and hydrocarbon migration, tracing CO 2 during enhanced oil recovery, and ruling out a geologically deep source of fluids during a suspected leak (Gilfillan et al., 2014(Gilfillan et al., , 2017(Gilfillan et al., , 2011(Gilfillan et al., , 2009;;Gilfillan and Haszeldine, 2011;Güleç and Hilton, 2016;Györe et al., 2017Györe et al., , 2015;;Holland and Gilfillan, 2013;Mackintosh and Ballentine, 2012;Shelton et al., 2016).
Recent review papers collate the available information on the inherent tracer fingerprint of captured CO 2 , make predictions regarding the likely fingerprint when empirical data is not available, and note the need for more direct measurements of captured CO 2 (Flude et al., 2016;Serno et al., 2017).Captured CO 2 stream generation was reviewed for conventional combustion, oxyfuel combustion, gasification (syngas production), fermentation, cement and steel manufacture, and CO 2 purification methods (chemical absorption/amine capture and physical absorption in organic solvents).It was predicted that the carbon isotope fingerprint will be controlled by the carbon-rich feedstock (e.g.fossil fuels, biomass, limestone) used in the CO 2 -generating process (e.g.combustion, gasification, cement and steel manufacture, fermentation) with small isotope fractionations expected for both conversion of the feedstock to CO 2 and the CO 2 purification process (Flude et al., 2016).Oxygen isotopes were expected to be controlled by the dominant sources of oxygen in the CO 2 generation and purification processes, including water (Serno et al., 2017).Noble gases were expected to be of low concentration for most capture techniques, with the noble gases being decoupled from the CO 2 during purification; radiogenic noble gases ( 4 He and 40 Ar) were expected to be enriched relative to nonradiogenic noble gases in fossil fuel derived CO 2 streams, and an enrichment of Kr and Xe relative to air was expected for oxyfuel combustion CO 2 capture, introduced via cryogenically separated oxygen (Flude et al., 2016).
Here we present new carbon and oxygen isotope, and noble gas concentration and isotope data from samples collected from captured CO 2 streams from six different capture plants, including CO 2 derived from coal, biomass, and natural gas combustion with amine capture, coal oxyfuel combustion, and gasification of methane.This work aims to contribute to answering four questions: 1) What is the inherent tracer fingerprint of captured CO 2 2) What controls this fingerprint?3) Will this fingerprint change en route to the storage reservoir?4) Will this fingerprint be retained during subsurface migration in the storage reservoir?

Sampling strategy
In order to determine the inherent tracer fingerprint of captured CO 2 and what controls this fingerprint, we collected CO 2 samples produced at six different capture plants, using three different feedstocks (coal, gas, and biomass) and three different capture technologies (amine capture, oxyfuel and gasification).Carbon isotopes in coal, biomass and gas feedstocks were also analysed, where available.At UKCCSRC's Pilot Scale Advanced Capture Technologies (PACT) facility (see site details below) we also collected samples of gas direct from the combustion flue (i.e.before capture) and from the exhaust of the amine absorber tank (i.e.residual gas, not absorbed during amine capture), to assess how the capture process alters the fingerprint.These samples are summarised in Table 1.
To ascertain if the fingerprint will change en route to the storage reservoir, we collected a sample from the CO 2 injection wellhead at the SaskPower Aquistore storage site, located in Saskatchewan, Canada, for comparison with our Boundary Dam sample (Section 2.3.2.).CO 2 from the Boundary Dam capture plant is compressed and transported to Aquistore via ∼4 km of underground pipeline (Rostron et al., 2014).
In order to establish if this inherent fingerprint is retained during subsurface migration in the storage reservoir, we collected post-injection samples from the reservoir fluid sampling system in the well of the Otway 2BX-1 pilot storage project for comparison with the CO 2 injected from the Callide oxyfuel plant.A sample from the Aquistore monitoring well downhole fluid recovery system was collected for comparison with the injected Boundary Dam CO 2 .Sampling site descriptions are provided in Section 2.3.

Sampling
Samples for noble gas analysis were collected using the well-established Cu-tube method (e.g.Györe et al., 2015).Gas samples were collected in 8 mm outer diameter, refrigeration grade copper tubing, connected to sampling ports (via a pressure regulator, where appropriate) with flexible high pressure hosing.An exhaust hose connected to the tube prevented turbulent back-mixing of air into the sample.The copper tube was purged with the gas stream being sampled for at least 2 min in order to flush out atmospheric contamination.The tube was then sealed with metal clamps, creating a cold-weld which is impermeable to helium.Gas samples were also collected in Tedlar gas sample bags, for stable isotope analysis, by attaching the gas bag either onto the sample port/regulator or on to the end of the copper tube exhaust pipe after the minimum flushing period.The occasions where the sampling process was modified from this standard procedure are described in the sample context descriptions below.Sample context is summarised in Table 1.

Niederaussem pilot capture plant, Germany
Sample Nied#3 was collected from RWE Generation's pilot capture plant at a 1000 MW lignite-fired power plant unit at Niederaussem.This captures 90% of the CO 2 in the combustion exhaust (flue) gas and has a capture capacity of 7.2 t per day (Moser et al., 2014).The sample was collected after all washing and filtering processes, from a valve immediately before the CO 2 storage tank.A sample of lignite fuel was also collected for isotopic analysis.

Boundary Dam capture plant and Aquistore, Saskatchewan, Canada
Boundary Dam Power Station's 139 MW coal-fired Unit 3 is attached to the world's first commercial-scale post-combustion capture plant (Stéphenne, 2014).Samples BD1 and BD2 were collected from the Boundary Dam Capture Plant's compressor building.Here, ambient pressure CO 2 is transported from the capture plant and compressed to over 2000 psi (13.8 MPa) in stages.Both samples were collected after the CO 2 was compressed to 400 psi (2.8 MPa), dried, and accessed via a low-pressure (15-25 psi/∼0.1 MPa) sample port.Sample AS_P_2 was collected from the Aquistore injection pipeline a few days after injection commenced (April 13th 2014).The pipeling transports pressurised CO 2 from the Boundary Dam Capture Plant to the Aquistore site.The sample was collected from a valve on the wellhead registering a pressure of 1608 psig (11 MPa).The pressure reduction during sampling caused dry ice to form and flow of CO 2 through the regulator slowed due to icing up; flow was maintained for long enough to collect samples but we note that sampling was under disequilibrium conditions (mixture of dry ice and gas) and so there may be some unexpected sample fractionation.Sample Tanker #1 was collected from the CO 2 used to pressurise the injection well before commencing pipeline injection into Aquistore.This CO 2 was collected from the CO 2 tanker vapour discharge valve, which registered a pressure of 230 psi (1.6 MPa) during sampling.The Tanker CO 2 was derived from the Agrium Fertiliser plant, Fort Saskatchewan, Alberta, which carries out steam reforming of methane (Agrium, 2015) A piece of the coal used as fuel at the Boundary Dam power station (BD coal) and water used to dilute the amine solution in the capture plant were also collected.In April 2014, 20 cm 3 of gas, degassed from water collected from the observation well fluid recovery system (FRS) was supplied for noble gas analysis by the University of Alberta (AQ1).

Ferrybridge power plant, UK
A sample of CO 2 was supplied from the CCPilot100+ pilot capture plant at the Ferrybridge Power Station, West Yorkshire, United Kingdom.The pilot capture plant operated for two years, capturing up to 90% of the CO 2 in Ferrybridge Power Station's coal-fired Unit 4 flue gas, and producing up to 120 t CO 2 per day via post-combustion amine capture (Fitzgerald et al., 2014).The coal being burnt during sampling was a blend of Columbian and Yorkshire coal (Mike Till, pers. comm. 2016).

Otway pilot injection project, Australia
The Otway Project in Victoria, Australia, is a CO 2 storage field test injection site.The Otway 2B Extension project (Otway 2BX) took place over 80 days in October − December 2014.The aim of the project was to study the effect of impurities in the CO 2 stream on water quality (days 1-62: 2BX-1) and to characterise CO 2 residual trapping (days 63-80: 2BX-2) (Serno et al., 2016).CO 2 was injected into the Paaratte Formation, employing the same reservoir and well configuration as the 2011 Otway 2 B project, which used injection and back-production from a single well to investigate CO 2 residual trapping (Paterson et al., 2013(Paterson et al., , 2011)).Pure CO 2 from the Callide oxyfuel plant (Uchida et al., 2013) was injected, along with added impurities, during Otway 2BX-1 on days 11 and 36 (Haese et al., 2016).The CO 2 was co-injected with earlier-produced reservoir water that had been stored in surface tanks and mixed downhole using a gas mandrel; the experiment was designed so that all of the injected CO 2 would dissolve in the formation water on injection (Haese et al., 2016).For Otway 2BX-2, Callide CO 2 was mixed with natural CO 2 produced from the nearby Boggy Creek well (Serno et al., 2016).
Tracing injected CO 2 was not the purpose of the Otway 2BX experiments, but an impromptu opportunity arose to collect samples of CO 2 produced at the end of the Phase 1 experiment, after injection and residence in the subsurface.Subsurface CO 2 samples (L283/2 and L284/2) were collected on 5th December 2014, Day 62 of the Otway 2BX test, at the end of the initial phase (2BX-1).Reservoir water and gas samples were collected into 150 mL stainless steel cylinders at reservoir pressure (140 bar) using a downhole U-tube system (Freifeld, 2005) and degassed into 10 L gas sample bags (Serno et al., 2016).The gas from these bags was then sampled into copper tubes using the standard procedure outlined previously.
Samples of Callide CO 2 were collected in copper tubes by filling a 150 mL stainless steel Swagelok cylinder with liquid CO 2 directly from the Callide tanker, and subsequently depressurising the cylinder to flow gaseous CO 2 through the copper tube.

Pilot scale Advanced Capture Technologies (PACT) facility, UK
PACT is a research and development facility for carbon capture and combustion research with pilot scale combustion and solvent carbon capture plants (PACT, 2016) located in Sheffield, UK.Samples were collected while the combustion rig was burning coal or biomass in air.The first sample (PACT1) was collected from the flue pipe leading from the burner rig to the capture plant (Fig. 1) using a peristaltic pump located between the copper tube and gas bag parts of the sampling assembly.This sampling port is located after a candle filter assembly used to clean particulate material from the flue gas.Subsequent samples were collected from different sections of the PACT amine capture facility via the exhaust line of PACT's gas analysis Fourier Transform Infra Red spectroscopy (FTIR) instrument, during line purging (Fig. 1).A piece of hosing connected the FTIR exhaust to the sampling tube assembly.An exhaust hose prevented turbulent back mixing of gas into the sample tube, and allowed venting of the flowing gas outside of the instrumentation hut.During sampling, the FTIR line, exhaust and sample assembly were purged with the desired gas for a total 4 min.After two minutes, a Tedlar bag was attached to the exhaust line of the sampling assembly to collect samples for stable isotope analysis (approximate sampling time of 30 s), followed by clamping of the copper tube to collect samples for noble gas analysis.An extra sample of natural gas combustion flue gas, equivalent to PACT1, was collected via the FTIR (PACT8) because the PACT1 Tedlar bag failed to retain the gas.This sample was collected at the end of natural gas combustion operation, while biomass was being fed into the boiler and it is thus possible that this sample has been contaminated by biomass flue gas.
During sample collection, the absorber column was at a temperature of 35-48 °C and the desorber column was running at 91-99 °C.No sampling ports were available at the PACT facility to collect the mains gas burnt in the rig, so a sample was collected by connecting a piece of hosing to a gas stove top (Sheffield Mains Gas − SMG).A piece of the biomass used as fuel was also sampled.

Stable isotopes
Stable isotope composition of captured CO 2 , feedstocks, and water were analysed in the stable isotope laboratories at the University of Calgary, Canada, and at the Scottish Universities Environment Research Centre (SUERC), East Kilbride, UK.
For the PACT samples, gas compositional analysis was carried out at the Applied Geochemistry Laboratory at the University of Calgary.The analysis was undertaken using a Bruker 450 gas chromatograph (GC), which was calibrated immediately prior to analysis, and drift monitored every ten samples, using a certified formation gas standard.Reproducibility of the samples are typically ∼5.0%.
At the University of Calgary Isotope Science Laboratory, CO 2 samples were purified on a glass extraction line to produce pure aliquots of CO 2 .Water samples were equilibrated with CO 2 for δ 18 O H2O measurements.C-and O-isotopes were then analysed by Dual Inlet Isotope Ratio Mass Spectrometry (DI-IRMS) using a DeltaV+ stable isotope ratio mass spectrometer.Internal laboratory reference materials, that are periodically calibrated to international reference materials, were run at the beginning and end of each sample set and were used to normalize the data relative to V-PDB and V-SMOW and correct for any instrument drift.Precision and reproducibility are better than ± 0.1‰ for both δ 13 C and δ 18 O.
At SUERC, solid samples (coal and lignite) were converted into CO 2 gas for stable isotope analysis.2-5 milligrams of coal were added to a quartz glass tube, along with pre-roasted wire-form copper oxide.The tube was sealed and roasted at 850 °C for 1.5 h, followed by 12 h of cooling.The copper oxide reacted with the carbon in the coal to form CO 2 , which can be analysed to give the C-isotope composition of the bulk coal.All CO 2 samples (oxidised from coal and original gas samples) were purified on a glass, vacuum extraction line to produce pure CO 2 .Samples were then isotopically analysed using a VG Isotech Sira II DI-IRMS using standard techniques.Based on repeat analyses of internal and international standards during the running of these samples, precision and reproducibility are typically around ± 0.2‰ for both δ 13 C and δ 18 O.
For oxygen isotope water analyses, 200 μL of the water sample was decanted into a 10 mL exetainer ® using an adjustable pipette with disposable pipette tips.The exetainers ® were then placed in the autosampler tray of a Thermo Scientific Delta V mass spectrometer, set at 25 °C.Each sample was then over-gassed with a 1% CO 2 -in-He mixture for 5 min and left to equilibrate for a further 24 h before analysis.Oxygen isotopic data were then produced using the method established by Nelson (2000).Reproducibility of the data, based on within-run repeat analyses of at least three standard waters, was around ± 0.2‰ for δ 18 O.All results are presented in standard delta notation, relative to V-PDB (for δ 13 C, and in the supplementary information for δ 18 O) and V-SMOW (for δ 18 O).

Noble gases
Copper tubes were connected to an all-metal ultra-high vacuum system and prepared for noble gas analysis at SUERC as described previously (Györe et al., 2015).Where necessary, three aliquots of the gas were then trapped into glass ampules for stable isotope analysis, using liquid nitrogen and sealing the ampule with a blowtorch after pumping away the other volatile gases.The remaining gas from the copper tube was cleaned using a titanium sublimation pump and getters, to remove all substances other than the noble gases, which were then stored in a stainless steel cylinder.The concentration and isotopic composition of noble gases was determined on an MAP 215-50 noble gas mass spectrometer.Analytical procedures for He, Ne, Ar, Kr and Xe isotopic analysis are described in previous work (re et al., 2017, 2015; re et al., 2017, 2015) and reproducibility of noble gas concentrations and isotope ratios is typically better than 0.5% for He, Ne and Ar, and better than 5% for Kr and Xe.

Stable isotopes
Full stable isotope data, and gas concentration data when undertaken, are provided in Supplementary Data Tables 1-3 and are summarised in Tables 2-4.Gas concentration results from PACT (Supplementary Table S1) indicate that the samples collected from the FTIR exhaust contain a significant proportion of atmospheric contamination.However, given the low concentration of CO 2 in the atmosphere, this CO 2 contamination is small can be easily corrected for using a simple two end-member mixing model: [CO ] (1) where square brackets indicate fractions of the component where [CO 2- measured ] = 1.The fractions of CO 2 from sample and air were estimated by comparing the expected CO 2 concentrations, where available (∼9% CO 2 for gas rig combustion flue, 99% CO 2 for captured CO 2 , based on usual operation of the PACT capture plant) with the measured values.This indicates that between 70 and 80% of the collected sample is air contamination.The PACT stable isotope results presented in Tables 2  and 3 are corrected values.
The measured CO 2 concentrations for the absorber outlet were higher than the expected values.Assuming an expected flue gas composition of ∼9 vol% CO 2 , and a 90% capture efficiency, the absorber outlet CO 2 concentration would be expected to be ∼0.9 vol%.The measured values were between 1 and 2.4%.Air has a CO 2 concentration of ∼0.04 vol%, lower than that expected in the absorber outlet samples.The higher than expected CO 2 concentrations cannot thus be explained by air contamination and we infer that the amine capture process was not running at full efficiency during our sampling period.
Stable isotope results from different parts of the CO 2 generation and capture process (fuel, flue gases, absorber exhaust gases, solvent) are compared to the captured CO 2 results in Table 3. Solid fuels (biomass, lignite, coal) range in δ 13 C from −26.7 to −23.5‰ while natural gas is −40.6‰.Flue gases range from δ 13 C = −27.3 to −25.3 and δ 18 O = +24.0 to +29.0‰.
Stable isotope data from the post-injection subsurface CO 2 samples collected from the Otway 2BX-1 experiment are compared to the injected CO 2 in Table 4. δ 13 C is ∼−24‰ and δ 18 O is ∼+35‰.

Noble gases
Noble gas data are presented in Tables 5 and 6.Noble gases were not analysed in the air-contaminated PACT samples.
Noble gas results are reported as concentrations (cm 3 per cm 3 at standard temperature and pressure: STP − 273.15K and 1 atmosphere) and isotope ratios in Tables 5 and 6.Helium isotope ratios are expressed as R/R A , where R is the 3 He/ 4 He value of the sample and R A is the 3 He/ 4 He value of air (1.39 × 10 −6 − Mamyrin et al., 1970).Data are conventionally represented graphically as isotope ratio variation diagrams, as concentrations normalised to air and as noble gas concentrations in the sample normalised to 36 Ar and air i.e.:

Sample noble gas Sample Ar
Air noble gas Air Ar (2) Noble gas concentrations in the captured CO 2 are typically two to three orders of magnitude lower than in air, while isotope ratios are similar to air.The natural gas fuel sample (SMG) has a helium content higher than air by two orders of magnitude but the other noble gases are of low concentration.Noble gas concentrations and isotope ratios other than 3 He/ 4 He in the analysed flue gas are similar to air.The postinjection subsurface gases have variable noble gas concentrations and isotopic compositions.The AQ1 sample is notably similar to air.

Discussion
4.1.Composition and controls on the tracer fingerprint of captured CO 2 4.1.1.Carbon isotopes Fig. 2 shows the C-isotope composition of the analysed captured CO 2 samples, along with C-isotope data for fuel feedstocks, where appropriate.Flude et al. (2016) predicted that the C-isotope composition of CO 2 captured from combustion power plants would be mostly controlled by the fuel feedstock with possible fractionation during both combustion and amine capture.They predicted δ 13 C Combustion_CO2 values of −31 to −21‰ for coal, less than −21‰ for natural gas, and −24 to −30‰ for C3 biomass combustion (Flude et al., 2016), with subsequent amine capture potentially imparting an extra fractionation of between −20 and +2.5‰.CO 2 captured from gasification plants was predicted to show a possible small negative fractionation, producing captured CO 2 with δ 13 C < -20‰.All of the coal and gas combustion-derived captured CO 2 falls within the range of values expected for combustion CO 2 , based on worldwide ranges of fuel δ 13 C. CO 2 captured from biomass combustion falls outside of the values expected for combustion, but within the range of values expected due to fractionation during amine capture.
Specific feedstock C-isotope data are available to compare with amine captured CO 2 from combustion of biomass (c.f.PACT10), lignite (c.f.Nied#3), coal (c.f.BD1) and natural gas (c.f.PACT7).Fig. 3 shows the difference in δ 13 C (Δ) between captured CO 2 and the fuel, and for intermediate stages of the combustion and capture process, where available, ranked in terms of their hydrocarbon maturity.Δ Captured-Fuel range from −5.0‰ (biomass) to +2.6‰ (natural gas) with an apparent increase in Δ value with increasing hydrocarbon maturity.Coal shows the least fractionation between fuel and captured CO 2 .CO 2 in the pre-capture combustion flue gas was also analysed for biomass (PACT2) and natural gas combustion (PACT8).Δ values were calculated for Flue gas-Fuel and for Captured CO 2 -Flue gas to investigate the influence of these different processes (combustion and amine capture) on the Cisotope value.In both cases Δ Flue-fuel is positive (+1.4‰ for biomass, +13.3‰ for gas), opposite to the expected −1.3‰ fractionation during combustion (Flude et al., 2016;Widory, 2006).Δ Captured-Flue is negative in both cases (-6.4‰ for biomass, −10.7‰ for gas).This is consistent with the predicted fractionation of between −20 and +2.5‰ (Flude et al., 2016).The greatest fractionations observed during combustion and capture are for the gas-combustion sample.As previously outlined, the flue sample (PACT8) may be contaminated with biomass flue gas and this value should be interpreted with caution.
The effect of amine capture on the δ 13 C CO2 was investigated by comparing δ 13 C values for flue gas, absorber outlet and captured CO 2 samples derived from biomass (samples PACT2, 9 and 10).Absorption of CO 2 by amine solutions involves formation of bicarbonate and carbamate species in the solution (Lee et al., 2013).Carbamate-CO 2 enrichment factors of +11‰ have been reported (Usdowski and Hoefs, 1988) while bicarbonate − CO 2 enrichment factors for fresh water at 35 to 45 °C range from 6.9 to 5.9‰ (calculated from Clark and Fritz, 1997).Assuming that these enrichment factors are appropriate for amine capture and that CO 2 desorption is more efficient than CO 2 absorption, it should be possible to determine the efficiency of the CO 2 capture process by comparing the δ 13 C values of the flue, absorber Table 5 Noble gas concentration data.Concentrations are volume fractions and given in cm 3 (STP)/cm 3 .Standard temperature and pressure is as reported in Ozima and Podosek (2002).Air composition is after Ozima and Podosek (2002) and references therein, and air saturated water (ASW) is calculated after Kipfer et al. (2002).Uncertainties are one standard deviation.outlet, and captured CO 2 .In this case, the δ 13 C of the flue gas CO 2 would be the initial reacting CO 2 , the absorber outlet reflects the residual CO 2 remaining after absorption, and the captured CO 2 represents the amine dissolved inorganic carbon (DIC) (assuming that significant fractionation does not take place during desorption).A Rayleigh fractionation model was created to try and reproduce the absorber outlet and captured CO 2 δ 13 C values, using Eq. ( 3) (Clark and Fritz, 1997).

Sample
Where δ is the δ 13 C of the residual reactant (in this case the absorber outlet gas), δ 0 is the initial δ 13 C of the reactant gas (in this case the flue gas), ε is the enrichment factor, and f is the fraction of reactant remaining.It was not possible to reproduce the residual reactant (absorber outlet) and dissolved carbon (captured CO 2 as a proxy) δ 13 C values using the expected enrichment factors, which are both positive.It is possible that DIC-CO 2 enrichment factors are different for amine solutions and fresh water, and that enrichment factors for amine capture can be estimated by iterative calculation of Rayleigh fractionation models with different enrichment factors until there is convergence between measured and modelled absorber outlet and captured CO 2 (DIC) values.Results of this modelling are shown in Fig. 4, which suggests a DIC-CO 2 enrichment factor during absorption of −7‰ and that only ∼20% of the CO 2 was being captured at the time of sampling.This may indicate that the PACT absorber column was not operating at full capacity during sampling, an interpretation backed up by the discrepancy between expected (0.9%) and measured (up to 2.4%) CO 2 concentrations in the absorber outlet samples (Section 3.1).

Oxygen isotopes
Fig. 5 shows the O-isotope composition of captured CO 2 , compared to natural sources of oxygen.Combustion of C-rich fuels should produce CO 2 with δ 18 O similar to that of atmospheric oxygen, with a possible fractionation of up to −21‰ for combustion of low ignition temperature fuel and/or combustion in an oxygen-rich environment (Serno et al., 2017).Boundary Dam and Niederaussem captured CO 2 are consistent with this hypothesis, but three amine-captured and one oxyfuel combustion CO 2 samples have δ 18 O higher than that of atmospheric oxygen (Ferrybridge, PACT, Callide).δ 18 O values of flue gases collected at PACT are also higher than that of atmospheric oxygen, suggesting either that oxygen isotope behaviour during combustion is not straightforward, or that passing of flue gas through a candle filter causes isotope fractionation.
Sample COSPL is CO 2 captured from the Callide Oxyfuel Plant, Australia, which uses a cryogenic air separation unit to supply high purity O 2 for combustion, and a cryogenic CO 2 separation unit to purify the end produced CO 2 (Uchida et al., 2013).It is possible that largescale cryogenic separation causes O-isotope fractionation, with the heavier isotopes preferentially partitioning into the cryogenically trapped phase, resulting in δ 18 O enrichment of the purified O 2 and CO 2 .
The δ 18 O of CO 2 produced by gasification processes is expected to be dominated by that of the water (steam) used during steam reforming and the shift reaction (Flude et al., 2016;Serno et al., 2017).The δ 18 O of Tanker#1 is particularly surprising as this value (+19‰) is significantly higher than the global range of meteoric water and is similar to atmospheric O 2 , as would be expected for a gasification plant that used partial oxidation instead of steam reforming for the gasification process (Serno et al., 2017).Alternatively it is possible that the fertilizer plant uses an isotopically unusual source of water to generate steam, or that significant O-isotope fractionation occurs during steam reforming, with 16 O being preferentially retained in the water phase and 18 O preferentially combining with natural gas derived carbon to form CO 2 .(Serno et al., 2017) noted that the O-isotope composition of aminecaptured CO 2 should be influenced by any water in the CO 2 generation system, and thus it is difficult to predict due to the wide range of δ 18 O of meteoric waters and a lack of known amine solution-CO 2 enrichment or fractionation factors.At the current time, enrichment factors between the water in amine solutions and captured CO 2 have not been published, perhaps due to difficulties in measuring δ 18 O aminesolution .We analysed δ 18 O in a sample of demineralised water from Boundary Dam's water treatment plant, which is used to dilute amine solution for the CO 2 capture process.This water has a notably light O-isotope composition (δ 18 O = −28‰) and the difference between the water and CO 2 captured from Boundary Dam is ∼+46‰.This degree of enrichment can only be produced between fresh water and CO 2 at temperatures of ∼ 0 °C (calculated via Clark and Fritz, 1997), much lower than the temperatures involved during amine capture (35-50 °C and 90-110 °C for the absorber and desorber, respectively).This suggests that either the amine solution-CO 2 enrichment factors are much greater than for fresh water-CO 2 systems, or that isotopic equilibrium is not reached during the capture process.Alternatively, contrary to predictions, the O-isotope composition of flue gas CO 2 dominates the O-isotope budget of the system.In hindsight, this result is not surprising as amine absorber columns are designed to maximise gas-water contact areas (Fitzgerald et al., 2014) and thus the gas to water ratio will be large, Table 6 Noble gas isotope data.Air composition is after Ozima and Podosek (2002) and references therein and Lee et al. (2006), and air saturated water (ASW) is calculated after Kipfer et al. (2002) resulting in the isotopic composition of the gas dominating the system.This hypothesis is consistent with the very small difference in δ 18 O values observed between flue gas CO 2 and captured CO 2 for the PACT samples (Fig. 5).

Noble gases
Fig. 6 shows normalised concentrations of noble gases in the captured CO 2 , while Fig. 7 shows selected noble gas isotope ratios.
Flude et al. ( 2016) made predictions regarding the noble gas content of captured CO 2 .Conventional fuel combustion is expected to incorporate air-like noble gases into the flue gas, while oxyfuel combustion will introduce noble gases via the source of O 2 .Use of fossil fuel feedstocks is expected to produce a relative enrichment of He, and possibly Ar, due to the accumulation of radiogenic 4 He and 40 Ar in the fossil fuel.Use of coal as a feedstock may give an enrichment in heavy noble gases, which have a high adsorption affinity for fine-grained organic material and coals (Flude et al., 2016;Podosek et al., 1981).
Noble gases in amine-captured CO 2 were predicted to be of especially low concentration (much lower than air saturated water −ASW), with a possible small relative enrichment in heavy over light noble gases (upward slope from left to right on Fig. 6A and B), due to massdependent differences in noble gas solubility (Flude et al., 2016).Our new data mostly agree with this prediction; Boundary Dam (BD2), Ferrybridge (Ferrybridge-1) and Niederaussem (Nied#3) CO 2 streams all have noble gas concentrations up to two orders of magnitude lower     (Ciais et al., 1997;Clark and Fritz, 1997;Drimmie et al., 1991;Ettayfi et al., 2012;Kroopnick and Craig, 1972;Newell et al., 2014;Nisi et al., 2013;Serno et al., 2017).PACT Biomass Flue = Sample PACT2.PACT Gas Flue = Sample PACT8.See than ASW (Fig. 6A).The relative concentrations of noble gases in the sample from Niederaussem also fit this prediction well, with an elevated 4 He content and an increase in abundance relative to air from Ne to Xe (Fig. 6A and B).
Noble gas concentrations in oxyfuel produced CO 2 were expected to be high − possibly of higher concentration than in air, especially for Kr and Xe which may be concentrated in cryogenically separated oxygen (Flude et al., 2016).This was corroborated by noble gas data for oxyfuel CO 2 produced at Laq, France, which showed enrichment of all noble gases above the natural gas fuel, and krypton concentrations an order of magnitude higher than found in air (Garcia et al., 2012) (Fig. 6).Our oxyfuel sample (COSPL) has the highest noble gas concentrations of all our captured CO 2 samples, but concentrations remain one to two orders of magnitude below air concentrations (Fig. 6A).Helium concentrations were expected to be dependent on the radiogenic He content of the fuel, while the remaining noble gases were expected to show a relative enrichment in heavier noble gases (an upward slope from Ne to Xe on Fig. 6A and B).However, an overall decrease in slope from Ne to Kr is observed, with a relative enrichment in Xe.The difference between our measured oxyfuel noble gas values, and those of the Laq oxyfuel may represent differences in post-capture purification of the CO 2 , or differences in oxygen purification techniques used to generate O 2 for the oxyfuel combustion process resulting in different trace noble gas concentrations in the purified O 2 .
Noble gases in CO 2 produced by gasification (steam reforming) were expected to be of low concentrations, similar to that of ASW, and with relative elemental abundances similar to ASW.The measured concentrations are within an order of magnitude of ASW, but the relative elemental abundances show the opposite trend to ASW, with enrichment of light and depletion of heavy noble gases, other than Xe, which is also enriched.The feedstock for the Tanker#1 CO 2 was natural gas.Enrichment of Xe is often observed for natural gases (Torgersen and Kennedy, 1999) and so this observed Xe enrichment likely derives from the organic content of the feedstock.
Noble gas isotope ratios were expected to be air like but with enrichment of radiogenic 4 He and possibly 40 Ar in CO 2 derived from fossil fuels (Flude et al., 2016).Fig. 7 plots He, Ar and Ne isotope ratios relative to air and air saturated water.Enrichments in 40 Ar/ 36 Ar over air, due to addition of radiogenic 40 Ar were not observed.This is likely due to the low K-content of fossil fuels combined with the long half life of 40 K (1.25 × 10 9 years − McDougall and Harrison, 1999) leading to a slow build up of 40 Ar in the fuel (c.f.coal age of 6.6 × 10 7 years), combined with the relatively high concentration of Ar in the atmosphere (0.934% − Ozima, M. and Podosek, 2002) and the high level of atmospheric input in most CO 2 production processes.The majority of samples Exhibit 40 Ar/ 36 Ar ratios significantly below the atmospheric value (298.56 − Lee et al., 2006) indicating that isotopic mass fractionation takes place at some point in the CO 2 generation and purification process.
Helium isotope ratios (presented as R/R A − Fig. 7) in samples Nied#3, COSPL and Tanker#1 show a small enrichment in 4 He (R/ R A < 1), as expected.Samples Ferrybridge-1 and AS-P-2 both have 3 He/ 4 He ratios greater than the atmospheric value.Ferrybridge-1 plots close to the mass fractionation line, calculated using Graham's law (Holland and Gilfillan, 2013).Ne isotope data (Fig. 7) also indicate that most samples have noble gas isotope ratios consistent with fractionated air.
Unfortunately, it was not possible to analyse noble gases on a full suite of samples including fuel, flue gas, and end product captured CO 2 , to assess how different stages of the CO 2 production and capture process control the noble gas fingerprint of the captured CO 2 stream.Samples SMG and PACT1 represent natural gas fuel and the flue gas derived from combustion of this gas in air; the associated sample of captured CO 2 (PACT7) was, unfortunately, too contaminated with air to provide meaningful noble gas data, having been sampled via the FTIR exhaust (see Sections 2.3.5 and 3.1).However, comparison of the noble gas concentrations and ratios between SMG and PACT1 may provide some constraints (Fig. 8).
Sample SMG has low concentrations (less than ASW) for all noble gases other than He, which shows an enrichment in 4 He, as would be expected for a gas that has resided in the geological subsurface.Argon and helium isotopes also show radiogenic isotope enrichment ( 40 Ar/ 36 Ar > 298.56; 3 He/ 4 He < < 1 R/R A − see Fig. 8 and Table 6).The flue gas sample has a higher 3 He/ 4 He ratio, but this is still less than the atmospheric 3 He/ 4 He ratio, suggesting that the radiogenic 4 He signature has been retained, albeit diluted.The flue gas 40 Ar/ 36 Ar value is atmospheric, indicating that the argon fingerprint of the natural gas has been diluted beyond recognition due to mixing with air during combustion.A simple two component mixing model between SMG and air (Fig. 8) was calculated to assess whether the He and Ar isotopes in sample PACT1 could be explained by mixing between SMG and air during combustion.The flue gas plots on the mixing line at ∼3% SMG and 97% air.However, the noble gas concentrations cannot be reproduced using this mixing model.
A mixing model invoking 60% SMG and 40% air best fits the concentrations of Ne and Ar, but this model over-estimates He and underestimates Kr and Xe.A potential source for the excess Kr and Xe is the candle filter between the combustion rig and the flue gas sampling port.Kr and Xe have high adsorption affinities (Podosek et al., 1981) and it is possible that Kr and Xe released from fuel in previous combustion Fig. 6.Noble gas concentrations in capture CO 2 normalised to concentrations in air (A) and normalised to 36 Ar and air (B).Rousse data from Garcia et al. (2012).The range of air saturated water (ASW) noble gas concentrations for water equilibrated with air at between 10 and 60 °C (calculated according to Kipfer et al., 2002) is also shown.Ferrybridge = Sample Ferrybridge-2.All other samples are as described in Fig. 2. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).experiments were trapped in the candle filter via adsorption onto the high surface area porous interior, or adsorption onto the fine grained particles trapped by the filter.These adsorbed noble gases would then be released into the gas stream when encountering a hotter or lower Kr and Xe concentration gas.

Fingerprint retention during transport
Transport of captured CO 2 from the capture plant to the storage site is most likely to be via pipeline or tanker.In either case, the CO 2 will be pressurised and transported as either a liquid or supercritical fluid.Sample BD1 was collected from the Boundary Dam capture plant compressor building, immediately before the final stages of compression for pipeline transport.Transport of CO 2 from the compressor building to the Aquistore injection site is via ∼5 km long underground pipeline (Whittaker and Worth, 2011) maintained at ∼2200 psi (15 MPa − SaskPower, pers.comm.).Sample AS_P_2 was collected 11 days later from the Aquistore injection wellhead.The coal used as fuel at Boundary Dam is sourced locally from large open cast coal mines, and so variation in fuel type or quality is not expected to influence the CO 2 fingerprint.
Recent work has identified differences in solubility of noble gases between low and high density CO 2 (Warr et al., 2015), and so opensystem fluctuations in CO 2 density may result in significant elemental fractionation of the noble gases.As noted in Section 2.3.2,dry ice formed during sampling of AS_P_2 and the sample was thus not collected under equilibrium conditions meaning that any fractionation effects will likely be enhanced.Carbon and oxygen isotopes (Figs. 2 and  4) show negligible variation between the two samples, suggesting the C and O isotope fingerprint of the captured CO 2 will be preserved during transport.
Figs. 6 and 7 compare the noble gas content of the samples.Concentrations of Ar and Kr show good agreement but He, Ne and Xe are all higher in concentration in sample AS_P_2.Noble gas isotopes also show differences between the two samples with AS_P_2 having lower 40 Ar/ 36 Ar, and higher 3 He/ 4 He and 20 Ne/ 22 Ne.AS_P_2 lies on the mass fractionation line on both Ar and Ne isotope variation diagrams (Fig. 7), suggesting that kinetic isotope fractionation of noble gases may take place during pipeline transport and/or CO 2 stream phase changes as the noble gases diffuse between the different phases.However, it is likely that these modifications to the noble gas fingerprint occurred during the sampling procedure, rather than transport.The pressurised pipeline is a relatively closed and simple system compared to the sampling port and equipment, where the CO 2 experienced a pressure drop from 1608 psig (11 MPa) to atmospheric pressure and a phase change from supercritical phase to a mixture of dry ice and gas.Nevertheless, the possibility of elemental and isotopic fractionation should be taken into account, especially when sampling injected CO 2 from pressurised Fig. 7. Isotope ratio variation diagrams for He, Ne and Ar in the captured CO 2 streams.MFL = mass fractionation line.R/R A represents the 3 He/ 4 He value of the sample normalised to that of the atmosphere.The grey box shows the field that noble gases in fossil-fuel derived CO 2 streams were expected to plot in (Flude et al., 2016).Samples are as described in Fig. 6. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).sources, such as injection wellheads.
4.3.Is the inherent fingerprint retained during subsurface migration in the storage reservoir?Three of the captured CO 2 streams that we sampled have been injected into geological storage pilot sites.Pressurisation of the Aquistore pipeline (Saskatchewan, Canada) was undertaken using 37 m 3 of Tanker#1 CO 2 (Agrium Fertilizer Plant), followed by injection of pipeline transported Boundary Dam CO 2 .At the Otway pilot injection site, Australia, Callide CO 2 (COSPL) was injected as part of the Otway Phase 2B extension experiment.Both pilot storage sites are equipped downhole fluid recovery systems to facilitate sampling of geological fluids and the CO 2 in the storage reservoir.

Otway 2BX
Samples L283/2 and L284/2 were collected from the Paaratte Formation, via the U-tube system (Freifeld, 2005), 26 days after the second injection of Callide CO 2 (COSPL) as part of the Otway 2BX-1 experiment.Hence, these potentially provide a measure of how the inherent fingerprint of the Callide CO 2 sample evolves on injection and subsequent production from a subsurface storage reservoir.
Carbon isotope values are slightly higher (by 2.4-3.0‰) in the reservoir samples (-24.4 to −23.7‰) than the injected Callide CO 2 (-26.7‰).Subsurface processes that can cause a shift in δ 13 C CO2 are mixing with a source of CO 2 that has a different isotopic signature, and isotopic fractionation during partial dissolution or exsolution.Fractionation during dissolution of the CO 2 is expected to be negligible as the experiment was designed to promote full dissolution of the CO 2 upon injection (see Section 2.3.4).C-isotope fractionation may have occurred during collection of the gas sample if incomplete degassing of the pressurised fluid occurred and a significant proportion of DIC remained in the fluid phase.Paterson et al. (2011) concluded that all Paaratte Formation water DIC was bicarbonate.Enrichment factors between bicarbonate and CO 2 gas (ε HCO2-CO2 ) will be positive at temperatures less than ∼100 °C, meaning that CO 2 will be isotopically lighter, and thus have a lower δ 13 C, than co-existing bicarbonate.Our reservoir samples are isotopically heavier (higher δ 13 C) than the presumed DIC and so the difference in isotope values cannot be explained by incomplete degassing during sample collection.Baseline δ 13 C data have not been published for the Otway 2 experiment series, but naturally occurring CO 2 from a range of formations in the Otway basin have δ 13 C between −16 and 0‰ (Watson et al., 2004).It is thus most likely that the observe difference in δ 13 C between the injected and produced CO 2 is due to mixing in the reservoir with a naturally occurring, heavier CO 2 source.
δ 18 O values in the reservoir samples are higher than in the Callide sample (Fig. 9).Oxygen readily exchanges between CO 2 and water and the resulting isotopic shifts can be used to calculate freephase CO 2 pore space saturation (Johnson et al., 2011a;Serno et al., 2016).δ 18 O values for Paaratte formation waters during Otway 2BX1 phase 1 showed little variation over the duration of the experiment (−6.2 to −5.7‰, Serno et al., 2016).Serno et al. (2016) calculated a δ 18 O CO2-H2O enrichment factor for the Paaratte formation of ε = 36.84‰.Assuming isotopic equilibrium is established between the formation water and the injected CO 2 , the maximum expected δ 18 O CO2 value is that of the water plus the calculated enrichment factor.This gives a maximum expected δ 18 O CO2 of 31‰, but the measured δ 18 O CO2 is ∼35‰.This 4‰ difference may reflect undocumented natural sample variability, isotopic fractionation during diffusion of the CO 2 through the formation waters, as hypothesised and observed for C-isotopes (Freundt et al., 2013;Mayer et al., 2015;Myrttinen et al., 2010), or a separate source of oxygen.The second CO 2 injection of the Otway 2BX-1 experiment included added impurities of 9 ppm NO 2 , 67 ppm SO 2 and 6150 ppm O 2 (Haese et al., 2016).The O-isotope composition of these added oxygen sources, and the extent to which they will isotopically exchange with the water and CO 2 is unknown.
Noble gas data for the injected CO 2 (COSPL) and the post-injection reservoir gas (L283/2) are compared in Fig. 10.The noble gas fingerprint of injected CO 2 is expected to mix with the reservoir noble gases with an anticipated increase in radiogenic isotopes ( 4 He, 21 Ne and 40 Ar).Baseline noble gas data are unfortunately not available.Post injection, the reservoir CO 2 has lower He and Ne and higher Ar, Kr and Xe than the injected CO 2 , which is unexpected.The xenon concentration (0.1 ppm) is notably high and an order of magnitude greater than in air.The source of this excess Kr and Xe is likely residual gas from the 2011 Otway 2B experiment, where Kr and Xe were added as tracers to measure CO 2 residual saturation.
During the 2B experiment, the noble gases were measured via gas chromatography (GC) mass spectrometry, with Kr and Xe detection limits of ∼1 ppm (Paterson et al., 2011).Our measured Kr and Xe concentrations in sample L283/2 are very high for geological fluids and compared to air, but remain below the GC detection limits, meaning that post-experiment alteration to the reservoir noble gas baseline by the 2B experiment would not be detectable without noble gas mass spectrometry analysis.Some of the impurities added in the second part of the Otway 2BX-1 (Haese et al., 2016) experiment may also contribute additional Kr and Xe.For example, oxygen and nitrogen are often cryogenically separated from air and may preferentially concentrate the heavy noble gases.
The lack of expected 4 He enrichment in the subsurface gases may be due to gas stripping of the subsurface fluids during the Otway 2B injection experiment.The Paaratte formation is 1075-1472 m deep and so is expected to have accumulated radiogenic 4 He.Noble gas solubility in water and brine increases with increasing mass, and it is likely that the lighter noble gases (He and Ne) preferentially partitioned into the newly injected gas phase during the 2011 Otway 2 B experiment and were effectively stripped from the reservoir during back-production of the injected CO 2 .This would result in a He and Ne depleted (and Kr and Xe enriched) baseline for the 2BX experiments and emphasises the need for accurate baseline determination when using noble gases as tracers, especially when working in geological formations that have been disturbed by human activity.

Aquistore
The Aquistore observation well is situated 150 m northeast of the injection well (Rostron et al., 2014) and is fitted with a downhole fluid recovery system (FRS) (Worth et al., 2014).A sample of CO 2 and methane, degassed from fluid collected at the FRS in April 2016, one year after injection commenced at Aquistore, was provided by the University of Alberta for noble gas analyses.
Noble gases in the injected CO 2 (Tanker #1/Agrium, AS_P_2/ Aquistore Pipeline and BD1/Boundary Dam) are compared to the subsurface sample (AQ1) in Fig. 10.The Aquistore storage reservoir is within sandstones of the Cambro-Ordovician Deadwood and Winnipeg formations (Jensen and Rostron, 2014), which directly overlie the Precambrian basement (Worth et al., 2014).It was not possible to collect a baseline sample for noble gas analysis before CO 2 injection took place and we have thus used published noble gas data from Canadian Shield brines (Greene et al., 2008) as an estimate of the Aquistore noble gas baseline data.
In Fig. 10, the relative proportions of Ne through to Xe in AQ1 are comparable to air (i.e. the slope of the AQ1 line is parallel to that of air, despite lower absolute concentrations) and concentrations of most noble gases are higher than the injected CO 2 , estimated reservoir baseline, or ASW.Ne isotope ratios and 84 Kr/ 132 Xe are atmospheric, but 40 Ar/ 36 Ar is sub-atmospheric, suggesting isotope fractionation of atmospheric Ar has taken place.These observations suggest that this sample has been heavily contaminated with atmospheric noble gases.This air contamination may have been introduced during the sampling procedure, if the copper tube was not sufficiently vented prior to sealing, or post sampling, if the copper tube was not fully sealed.
Alternatively, atmospheric noble gases may have been introduced to the sample in the fluid recovery system, which uses nitrogen (which is often derived from air), as a lift gas.
The noble gas concentrations in AQ1 can be approximately reproduced by a hypothetical model involving mixing of 80% injected CO 2 (the relative proportions of BD1 and Tanker#1 make negligible difference to the model results), 15% air and 5% baseline noble gases, assuming that the baseline is similar to the Canadian Shield Brines measured by Greene et al. (2008).We note that the true baseline may be very different and that is not a unique model solution.
While the air contamination in this sample makes it difficult to use noble gases to assess subsurface mixing and migration, we note that sample AQ1 has a low 3 He/ 4 He ratio (0.44 R/R A ) and a high 4 He concentration (75 ppm, an order of magnitude higher than the concentration in air).This is consistent with our prediction that injected CO 2 will inherit the noble gas fingerprint of the reservoir.If this CO 2 were to subsequently leak to the surface, we would expect the elevated 4 He and below air 3 He/ 4 He ratio to be retained during this vertical migration and facilitate identification of the leaking gas as having a geologically deep source, as has been observed for natural leakage at Teapot Dome, Wyoming (Mackintosh and Ballentine, 2012), and St John's, Arizona/New Mexico (Gilfillan et al., 2011).

Conclusions and implications
We find that the δ 13 C values in captured CO 2 are mostly controlled by the feedstock composition, as previously predicted (Flude et al., 2016).However, we also find that a small fractionation between fuel and CO 2 (−5 to +2.6‰) may occur.This appears to be distinct from the previously observed fractionation of −1.3‰ during combustion   (Serno et al., 2016).CO-SPL = Callide CO 2 (injected).L283/2 and L284/2 are reservoir samples collected via utube.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).(Widory, 2006).The majority of the CO 2 samples exhibit δ 18 O similar to atmospheric O 2 , although captured CO 2 samples from biomass and gas feedstocks at one location in the UK are significantly higher.
The δ 18 O of combustion derived CO 2 in flue gas was expected to be similar to or lower than the δ 18 O of atmospheric O 2. with amine capture being predicted to control δ 18 O of captured CO 2 via O-isotope exchange with the water in the amine solution (Serno et al., 2017).However we find that the difference in δ 18 O between the water used to dilute the Boundary Dam amine solution, and the Boundary Dam captured CO 2 is larger than can be explained by equilibrium O-isotope exchange between water and CO 2 .Hence, it appears that amine capture exerts little influence on the O-isotope composition of captured CO 2 , and that combustion flue gas CO 2 may have δ 18 O higher than that of atmospheric O 2 , contrary to the expected fractionation (Serno et al., 2017).The reason for this is unknown but may relate to fractionation during filtering/cleaning of the flue gas.CO 2 derived from a fertiliser plant via steam reforming has a δ 18 O surprisingly similar to that of atmospheric O 2 , suggesting that the O-isotope composition of CO 2 produced by gasification is not, as predicted (Serno et al., 2017), controlled by the water used during steam reforming.
Our measured noble gas concentrations in captured CO 2 are generally as expected from previous work (Flude et al., 2016), apart from an oxyfuel sample which exhibited lower concentrations than expected.Relative noble gas elemental abundances are variable and often show an opposite trend to that of ASW.Expected enrichments in radiogenic noble gases ( 4 He and 40 Ar) for fossil fuel derived CO 2 (Flude et al., 2016) are not always observed due to dilution with atmospheric noble gases during the CO 2 generation and capture process.In the case of combustion, the majority of this dilution takes place during fuel ignition.Noble gas isotope ratios indicate that isotopic fractionation takes place during the CO 2 generation and capture process, resulting in isotope ratios similar to fractionated air.Phase changes associated with CO 2 transport and sampling may induce noble gas elemental and isotopic fractionation, due to different noble gas solubilities between high (liquid or supercritical) and low (gaseous) density CO 2 .
Samples from two CCS injection sites were collected to assess how effectively the inherent CO 2 fingerprint is retained after injection into the storage formation, but full interpretation of the data was hampered by a lack of reliable baseline data (Otway 2BX) and air contamination of the sample (Aquistore).Nevertheless, data from Otway 2BX indicate that δ 13 C values of the CO 2 changed slightly once injected due to mixing with pre-existing CO 2 .Furthermore, noble gas data from the post-injection Otway storage reservoir identify noble gas stripping of the Paaratte formation and contamination with Kr and Xe related to an earlier injection experiment.Finally He data from Aquistore illustrate that injected CO 2 will inherit distinctive crustal radiogenic noble gas fingerprints from the subsurface once injected into a geological storage reservoir.
Inherent tracers thus have the potential to be valuable geochemical monitoring tools for tracking the fate and migration of CO 2 injected into the subsurface for storage but further research from test injection sites is needed to resolve how inherent tracers can be most effectively used to monitor commercial scale CO 2 storage.We conclude that the majority of captured CO 2 sources will have carbon and oxygen isotope compositions that fall within a relatively restricted range (δ 13 C:−40 to −20‰; δ 18 O: +15 to +30‰), despite generation from a range of feedstocks and processes.These ranges show some overlap with, but are at least partially distinct from expected natural baselines and so can be expected to act as useful geochemical tracers in the majority of cases.Noble gas elemental and isotopic compositions will be distinct from reservoir baseline values, and so may prove to be valuable inherent tracers, but more work is needed to assess the sensitivity of this technique due to the low noble gas concentrations in captured CO 2 .We also recommend further work is carried out into the baseline noble gas conditions of CO 2 storage reservoirs in depleted oil and gas reservoirs where hydrocarbon production may have stripped out initial reservoir noble gases.In such situations, the distinctive radiogenic signature associated with deep crustal fluids, and which is a reliable tracer of CO 2 origin for naturally occurring CO 2 seeps, may no longer be present.Finally, we note that carbon and oxygen isotopes and noble gases may also act as tracers for industrial and capture processes.Whilst the goal of our data analysis was to understand what controls the geochemical and isotopic fingerprint of captured CO 2 streams to improve prediction of this fingerprint, in doing so we illustrate that these geochemical tracers may have potential for assessing process efficiency in industrial applications relevant to CO 2 capture.

Fig. 1 .
Fig. 1.Schematic of the PACT facility with sampling locations indicated by stars.

Fig. 2 .
Fig. 2. C-isotope composition of captured CO 2 and feedstock compared to natural CO 2 sources.Baseline and predicted values from Flude et al. (2016).Dark grey predicted values represent the anticipated C-isotope composition of CO 2 produced during combustion.Light grey areas represent additional fractionation that may take place during amine capture.Boundary Dam CO 2 = Sample BD2.Aquistore Pipeline = Sample AS_P_2.Niederaussem CO 2 = Sample Nied#3.Ferrbyridge = Sample Ferrybridge-1.PACT Biomass CO 2 = Sample PACT10.PACT Gas CO 2 = Sample PACT7.Callide Oxyfuel = -Sample COSPL#1.Agrium Fertiliser Plant = Sample Tanker#1.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).

Fig. 4 .
Fig. 4. Rayleigh fractionation model to calculate an estimation of the DIC-CO 2 enrichment factor for amine capture of CO 2 .Dotted lines are measured values.Solid lines are calculated Rayleigh fractionation curves for a starting flue gas CO 2 with δ 13 C of −25.3‰ dissolving into amine solution with an iteratively calculated DIC-CO 2 enrichment factor of −7‰.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).

Fig. 5 .
Fig. 5. Oxygen-isotope values for the captured CO 2 along with combustion flue gases where available, and amine solution values.Baseline ranges from(Ciais et al., 1997;Clark and Fritz, 1997;Drimmie et al., 1991;Ettayfi et al., 2012;Kroopnick and Craig, 1972;Newell et al., 2014;Nisi et al., 2013;Serno et al., 2017).PACT Biomass Flue = Sample PACT2.PACT Gas Flue = Sample PACT8.See Fig. 2 caption for other sample designations.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).
Fig. 5. Oxygen-isotope values for the captured CO 2 along with combustion flue gases where available, and amine solution values.Baseline ranges from(Ciais et al., 1997;Clark and Fritz, 1997;Drimmie et al., 1991;Ettayfi et al., 2012;Kroopnick and Craig, 1972;Newell et al., 2014;Nisi et al., 2013;Serno et al., 2017).PACT Biomass Flue = Sample PACT2.PACT Gas Flue = Sample PACT8.See Fig. 2 caption for other sample designations.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).

Fig. 8 .
Fig.8.Noble gas concentrations (A) and isotope ratios (B) for the Sheffield mains gas (SMG) burnt in the gas rig at PACT to generate flue gas sample PACT 1 (Flue).The dotted grey line in the left hand panel represents a two end member mass-balance mixing model between 60% SMG and 40% air that provides the best fit to sample PACT1.The grey line in the right hand panel is a model mixing line between noble gases in sample SMG and air, with tick marks representing the proportion of SMG.The He and Ar isotopes of sample PACT 1 are consistent with mixing between 3% mains gas and 97% air.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).

Fig. 10 .
Fig. 10.Comparison of noble gas concentrations in injected and post-injection reservoir CO 2 for Otway 2BX-1 and Aquistore.Grey line represents a mixing model between 80% injected CO 2 (proportions of Tanker#1 and BD1 make little difference), 5% baseline and 15% air.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).

Fig. 9 .
Fig. 9. Oxygen isotope data for injected and post-injection reservoir CO 2 in the Otway 2BX-1 experiment.Paaratte Formation water values from (Serno et al., 2016).CO-SPL = Callide CO 2 (injected).L283/2 and L284/2 are reservoir samples collected via utube.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).

Table 1
Sample name, source, feedstock and capture process context.

Table 2
Stable isotope composition of captured CO 2 .Data with uncertainties (1σ) are averages of multiple analyses.Data without uncertainties represent single data points.Full data are provided in Supplementary Table 3.

Table 3
Stable isotope data from the fuel to capture process.AO = Absorber Outlet and refers to the gas released from the amine absorber column exhaust.Data with uncertainties (1σ) are averages of multiple analyses.Data without uncertainties represent single data points.Full data are provided in Supplementary Table3.

Table 4
Stable isotope composition of injected and subsurface fluids from Otway.Data are averages of multiple analyses and uncertainties are 1σ.Full data are provided in Supplementary Table3. name.
. Uncertainties are one standard deviation.