Radiocarbon integrity of dissolved inorganic carbon (DIC) samples stored in plastic and glass bottles: implications for reliable groundwater age dating

ABSTRACT

Various approaches based on the natural variations of carbon isotopes (¹⁴C and ¹³C) in dissolved inorganic carbon (DIC) are routinely used to study groundwater dynamics and to estimate recharge rates by deriving groundwater ages. However, differences in ¹⁴C activities in groundwater samples collected repeatedly from the same wells and discordantly young ¹⁴C groundwater ages compared to noble gases led some authors to question the validity of radiocarbon dating. Poor sampling protocols and storage effects (¹⁴C contamination) for radiocarbon analysis are a critical factor in explaining age determination discrepancies. We evaluated the impact of storage protocols on carbon isotope exchange with atmospheric carbon dioxide by comparing glass versus standard plastic field sampling bottles for various storage times before radiocarbon and ¹³C analyses. The ¹⁴C bias after 12 months in pre-evacuated glass vials was minimal and within analytical precision. However, storage of DIC samples in plastic sampling bottles led to marked changes in ¹⁴C and ¹³C contents (up to ∼15 pmC and ∼ 5 ‰, respectively, after 12 months), meaning contamination led to younger groundwater age estimations than it should have been. Protocols for sampling and storing DIC samples for radiocarbon using pre-evacuated glass bottles help avoid atmospheric ¹⁴CO₂ contamination and microbial activity.

KEYWORDS:

• Carbon-13
• carbon-14
• dating
• DIC
• groundwater
• isotope measurements, methods, and equipment
• storage
• water sampling

1. Introduction

Environmental radioisotopes (e.g. ¹⁴C, ³H, ³⁶Cl) have been used for decades to study groundwater recharge and to evaluate the time scales of groundwater flow and transport. Reliable estimates of groundwater ages are critical for consolidating our understanding of aquifer flow dynamics and their exploitation sustainability as well as for developing and calibrating numerical groundwater flow and transport models (e.g. [1–4]). A widely used natural radiotracer available to study groundwater transport is the radiocarbon (¹⁴C) content of dissolved inorganic carbon (DIC), which has been used effectively over the past 60 years [5–7]. The prevalence of DIC (>100 mg/L C) in most groundwater systems, along with a long radiocarbon half-life (5730 ± 40 a), make ¹⁴C an attractive tracer for medium-sized and deep aquifers, spanning groundwater ages from several hundred to ca. 40,000 years. To derive ‘corrected’ groundwater ages by using radiocarbon, carbon stable isotopes (δ¹³C) data are included in geochemical reaction and transport models to correct for additional ¹⁴C-free inorganic carbon input derived from carbonate dissolution and other geochemical reaction complexities such as those occurring between soil gas CO₂ and HCO₃^– in recharge water [8–11].

However, sampling and applying ¹⁴C as a tracer in hydrological studies is not straightforward. The field sampling protocols require expertise in appropriate collection methods (e.g. [12,13]). Moreover, ultra-trace levels of ¹⁴C in very old (or ¹⁴C-free) groundwater samples are easily contaminated with minimal exposure of the water samples to atmospheric CO₂ that contains modern radiocarbon with higher than 100 percent of modern carbon (pmC) [12]. Additional physicochemical parameters such as water chemistry, temperature, pH, and field alkalinity are needed when using the various geochemical age correction models (e.g. [9,14]). From an analytical perspective, ¹⁴C analysis of the DIC samples involves complex laboratory methods for quantitatively extracting and converting DIC into pure CO₂ gas without contamination, both for low-level scintillation counting and accelerator mass-spectrometry methods [15,16]. Relating the ¹⁴C contents to accurate groundwater ‘ages’ requires technical skills in geochemical correction methods that need additional chemical and stable isotope data to account for various carbon sources (DIC, DOC, CH₄, POC) and complex geochemical reactions (water–rock) in the aquifer (e.g. [17]).

Over the past decades, accelerator mass spectrometry (AMS) is the preferred analytical method for ¹⁴C analyses used for groundwater dating. AMS is preferable to low-level liquid scintillation counting (LSC) methods due to the comparatively minimal sample size requirements (micrograms vs grams of C) and lower cost. AMS eliminated the need to precipitate large amounts of dissolved DIC as a solid barium or strontium carbonate in the field using up to 60 L of groundwater to obtain the 2–3 g of carbon needed for analysis by LSC. Only 50–300 mL of groundwater is typically required to obtain an amount of 1–2 mg carbon for ¹⁴C analysis by AMS and for ¹³C/¹²C analysis by isotope ratio mass spectrometry (IRMS) (see references within the section Carbon isotope analyses). Groundwater sampling protocols for radiocarbon recommend field-filtering (0.2 µm filters to remove carbonate particles and bacteria) into immediately tightly capped plastic bottles (low-density polyethene – LDPE, high-density polyethene – HDPE, and polypropylene – PP) for storage and transport to the ¹⁴C laboratory. The use of bactericides, such as saturated HgCl₂ or sodium azide solution, for poisoning water samples along with glass bottles is recommended to prevent microbial activity and also minimize sample handling.

Sampling protocols for ¹⁴C analysis must preserve the integrity of the DIC isotope composition (¹⁴C and ¹³C) from the field collection to the laboratory's carbon isotope analysis. Depending on the groundwater chemistry, DIC can partially precipitate as a solid carbonate during CO₂ degassing or undergo carbon isotope exchange reactions with atmospheric CO₂ after sampling, potentially altering the groundwater’s original carbon isotopic composition from the aquifer. Special care is also required to avoid uptake or exchange with atmospheric ¹⁴CO₂ during sampling (e.g. by groundwater sample aeration with pumping and handling), which poses a significant source of error when using ¹⁴C for ancient groundwater [12].

Additional factors revolve around sample integrity with extended storage and waiting times before the radiocarbon analysis is completed. Using inappropriate or low-quality plastic sampling bottles can lead to unwanted carbon isotope fractionation effects by allowing for the isotope exchange of carbon in the form of DIC and atmospheric CO₂ through the walls of the sample bottle. The magnitude of contaminating carbon-isotope exchange processes can significantly alter the derived radiometric groundwater ages. Synthetic organic polymers used in plastic bottles are semi-permeable barriers to CO₂ between the plastic wall; hence, diffusion of CO₂ through (in or out) the plastic bottle walls could alter the DIC carbon isotope composition. The magnitude of carbon isotope ‘storage effects’ in various sampling bottle types has not been fully considered for the application of radiocarbon dating of groundwater.

The negative impact of sampling and storage artifacts for groundwater dating using ¹⁴C is particularly pronounced for old ground waters, which present ¹⁴C activities lower than 1–10 pmC. For such samples, analytical artifacts derived from poor sampling methods or sample storage in low-quality bottles are critical because it requires only a few pmC gained by external contamination to cause a modern shift in apparent groundwater age of several thousand years and, unfortunately, in the absence of proper controls, could go unnoticed.

The too-frequent discordance between ¹⁴C activities in samples repeatedly collected from the same wells and the estimated groundwater ages using ¹⁴C versus noble gases [18,19] prompted this test regarding potential effects of storage (several types of plastic containers and long storage times) for ¹⁴C and ¹³C from DIC in groundwater. A laboratory experiment evaluated the magnitude of carbon isotope exchange by testing commonly used plastic field sampling bottles held for various storage periods before radiocarbon analyses were conducted, and these results were compared with a pre-evacuated sealed glass bottle technique. Both types of bottles, plastic and glass, are used with standard protocols for ¹⁴C-DIC sampling (e.g. [20]). HDPE is often recommended because CO₂ more easily diffuses through LDPE containers.

We considered the extent of atmospheric CO₂ exchange and diffusion using three groundwater samples with contrasting original alkalinity and isotopic composition (250–1700 mg/L and 2–40 pmC). This technical note aims to provide practical protocols for consideration of sample bottles used for storing DIC samples for radiocarbon dating, and to evaluate the magnitude of the carbon isotope shifts as a function of storage time and the type of sampling bottle.

2. Materials and methods

2.1. Groundwater samples

Three groundwater samples were used in a ¹⁴C exchange and contamination experiment covering a 12-month sample storage timeframe. The two groundwater samples were commercial spring waters provided in polyethylene terephthalate (PET) from Austrian deep groundwater systems. The mineral waters were purchased in July 2020 (3 bottles of 1.5 L each); one groundwater was from 660 m depth (VOS; Voslauer Natural; https://www.voeslauer.com/produkte/mineralwasser/ohne) and the other from 90 m depth (JUV; Juvina; https://www.juvina.at/de/produkte/juvina-still/). The third water sample was from a deep artesian well at Grafendorf, Austria, sampled in July 2020 (GRA). The JUV water is carbonic acid rich with an alkalinity of ∼ 1.7 g/L, while the alkalinity of VOS is about 0.25 g/L. We hypothesized that carbon isotope exchange at room temperature during long-term sample storage could progressively alter and allow for the differentiation in the DIC isotopes over time.

2.2. Plastic sample bottles

The test bottles included three of the most widely used plastic sampling bottles: a 500-mL LDPE bottle manufactured by Kautex (www.kautex.com), a 500-mL HDPE wide-mouth plastic bottle (type Nalgene®), and a standard 500-mL HDPE bottle provided by Flakado (www.flakado.de). All experimental waters were done in triplicate based on the type of plastic bottle.

2.3. Glass sample bottles

We tested pre-evacuated 125-mL glass serum bottles (Wheaton, Millville, NJ, USA) sealed with a thick butyl blue stopper. The pre-evacuated glass bottle method followed the approach for dissolved oxygen isotopes and carbon-13 that was proven to prevent atmospheric contamination for >1 year [21]. The samples were poisoned to prevent microbial activity during storage by adding a few drops of sodium azide solution prior to bottle evacuation. Other bactericides that do not contain ¹⁴C, such as saturated HgCl₂ salt solution, are also possible but subjected to toxin handling guidelines.

2.4. Bottle sampling and storage

For each bottle type, we first gently transferred ∼ 3 L of each groundwater sample into an open bucket. We then filled and closed the plastic bottles while fully submerged underwater to avoid bubbles and contact with atmospheric CO₂. The plastic bottle types were filled in duplicate. The pre-evacuated (10⁻³ Torr) glass bottles were filled by immersing under water and using a 21G needle. All air bubbles were removed from the needle by purging underwater. We then punctured the bottle septum with the needle underwater and allowed the bottle to fill by vacuum, after which the needle was pulled from the septum and the filled sample bottle was taken out of the water. After sampling, all DIC sample bottles were stored in a dark cabinet at room temperature for one week, 6 and 12 months before being analyzed by AMS for ¹⁴C and ¹³C. The samples in evacuated glass bottles were analyzed only twice, after one week and 12 months.

2.5. Carbon isotope analyses (¹⁴C and δ¹³C)

All DIC samples were analysed for carbon isotope analyses (¹⁴C and ¹³C/¹²C) at the Centre for Isotope Research (CIO) at the University of Groningen, Netherlands. In the Groningen laboratory, 20–50 mL of water in the glass bottles were displaced with helium gas and acidified with phosphoric acid to release DIC as CO₂ gas to the headspace. The CO₂ in the headspace was cryogenically transferred to a Pyrex break seal tube and the ¹⁴C content was analysed by AMS and expressed as pmC [22]. Radiocarbon results were normalized using the reference material NIST oxalic acid (SRM 4990C), and a ‘zero’ point was obtained from CO₂ derived from ¹⁴C-free marble. Analytical uncertainty, reported as one standard deviation of replicate samples, was better than ± 0.2 pmC (https://www.rug.nl/research/centre-for-isotope-research/research/radio-carbon-dating/?lang=en). Stable carbon isotope measurements in dissolved inorganic carbon (δ¹³C-DIC) were obtained from an aliquot of headspace CO₂ after treatment with phosphoric acid of water samples, and by isotope-ratio mass spectrometry (IRMS). The uncorrected δ¹³C values were calculated relative to a high-purity research-grade CO₂ pulse and then normalized to the VPDB scale using laboratory reference materials. δ¹³C values were expressed in the delta notation as parts per thousand (‰) deviations from the international standard (Vienna Pee Dee-Belemnite, VPDB). Analytical precision of the IRMS method for DIC was better than ± 0.2 ‰ for δ¹³C.

2.6. Effect of storage and bottle type on radiocarbon ages

Using the carbonate-reaction and isotope mass balance method, we evaluated changes in apparent radiocarbon ages [23,24] due to storage by bottle type and over sample storage time. To illustrate the importance of improper sampling and storage, ‘uncorrected’ ages were calculated based on the initial ¹⁴C amount without any radioactive decay plus a correction for geochemical processes, i.e. ¹⁴C₀. The value of ¹⁴C₀ of dissolved CO₂ in equilibrium with modern atmospheric CO₂ was assumed to be 100 pmC. The following equation represents the calculation: (1) $t=-{\displaystyle \frac{5730}{\ln 2}\ln \left({\displaystyle \frac{{}_{}^{14}{\mathrm{C}}_{\mathrm{DIC}}}{{}_{}^{14}{\mathrm{C}}_0}}\right),}$(1)

where t is the ¹⁴C age, and ¹⁴C_DIC is the measured value of DIC. We also determined ‘corrected’ radiocarbon ages using δ¹³C and a two-source isotope mixing approach [25,26]. This simple model is a two-step reaction process: the DIC in equilibrium with CO₂ from soil (100 pmC) and the DIC in equilibrium with solid (¹⁴C-free) carbonates. The model calculated the initial ¹⁴C content (¹⁴C₀) following equation (2): (2) ${}_{}^{14}{\mathrm{C}}_0={\displaystyle \frac{\left(100\left(\delta {}_{}^{13}{\mathrm{C}}_{\mathrm{DIC}}-\delta {}_{}^{13}{\mathrm{C}}_{\mathrm{s}}\right)\right)}{\left(\delta {}_{}^{13}{\mathrm{C}}_{\mathrm{g}}-\delta {}_{}^{13}{\mathrm{C}}_{\mathrm{s}}+{ϵ}_{\mathrm{b}/\mathrm{g}}\right)}\times \left(1+{\displaystyle \frac{2{ϵ}_{\mathrm{b}/\mathrm{g}}}{1000}}\right)={\displaystyle \frac{{}_{}^{14}{\mathrm{C}}_{\mathrm{g}}-0.2{ϵ}_{\mathrm{g}/\mathrm{b}}}{\delta {}_{}^{13}{\mathrm{C}}_{\mathrm{g}}-{ϵ}_{\mathrm{g}/\mathrm{b}}}\delta {}_{}^{13}{\mathrm{C}}_{\mathrm{DIC}},}}$(2)

where δ¹³C_DIC is the isotopic composition of dissolved carbonate species (DIC), δ¹³C_g is the isotopic composition of soil gas CO₂ that was assumed to be −25 ± 0.5 ‰ (1σ); δ¹³C_s is the isotopic composition of the carbonate in the aquifer matrix that was assumed to be +1 ± 0.5 ‰, and all C comes from the rock CaCO₃; and ϵ_g/b is the ¹³C fractionation factor of CO₂ (gas) with respect to HCO₃⁻ (−7 ± 0.25 ‰, [8,27,28]). It should be noted that the test samples were not exposed to soil or atmospheric CO₂ during sampling and storage. Still, these calculations illustrate the potential effects of sample storage on age calculations.

3. Results and discussion

3.1. The integrity of ¹⁴C and δ¹³C -DIC by bottle type

The ¹⁴C and δ¹³C values of the DIC samples measured at the start of sampling (VOS, JUV, GRA) ranged between 2 and 41 pmC and between −10 and −5 ‰, respectively, revealing a wide range of initial ¹⁴C activities at the beginning of the sample storage experiment (Table 1). The ¹⁴C and δ¹³C values after 6–12 months of storage were evaluated for bias and z-scores (z). The z-score was used to assess the accuracy of a ¹⁴C assay measurement after various months of laboratory storage by considering analytical uncertainty. We used a performance standard deviation (σ) of 0.4 ‰ for comparison purposes to include the uncertainties related to the analytical measurement, normalization and reference materials used, which was estimated as two times the analytical uncertainty of AMS measurements. (3) $z={\displaystyle \frac{X-\mathrm{Xa}}{\sigma },}$(3) where the numerator is the bias of measurement, X and X_a are the isotope values for the sample to be evaluated, and the reference sample value, respectively.

Table 1. Carbon isotope (¹⁴C and δ¹³C) values of the DIC samples stored in pre-evacuated glass bottles at the start of the experiment (C_o), at the end of the experiment after 12-month storage (C_f), and differences between both experimental times (Δ).

	δ^13Co (‰)	δ^13Cf (‰) t = 12 m	Δ^13C (‰)	^14Co (pmC)	^14Cf (pmC) t = 12 m	Δ^14C (pmC)
VOS	–5.95	–6.71	–0.76	7.63	5.01	–2.62
JUV	–0.79	–1.83	–1.04	2.20	1.69	–0.51
GRA	–10.38	–10.26	0.12	41.0	40.78	–0.22
			–0.56 ± 0.61			–1.12 ± 1.31

3.1.1. Pre-evacuated glass bottles

The storage bias for ¹⁴C after 12 months in the pre-evacuated glass vials in absolute terms was a 1.1 ± 1.3 pmC for ¹⁴C and 0.6 ± 0.6 ‰ for δ¹³C (average, SD). These radio- and stable-carbon isotope changes were minimal, within the AMS analytical uncertainty when z-scores were <2.0. The ¹⁴C data obtained from the glass bottles at t = 0 were used as our ‘reference’ to assess the plastic bottle results after 6–12 months of storage. In the case of ¹⁴C, the measured values of experimental waters were within analytical uncertainty after one year of storage except for VOS (|z| ∼ 6.5). For δ¹³C, two samples (VOS and GRA) were acceptably reproducible after one year. Still, the carbonic-rich water (JUV) had a |z| between 2.0 and 3.0. This significant but relatively small isotope change in the carbonic-rich sample was likely related to its exceptionally high alkalinity (ca. 2 g/L) and dissolved CO₂ gas content compared to the other samples, where possibly some release of CO₂ occurred during later handling (preparative analysis).

3.1.2. Plastic bottle storage effects

Figure 1 shows the ¹⁴C and δ¹³C values of the samples stored in various plastic bottles for 6–12 months compared to the data from the glass bottles, as well as potential contaminating sources of CO₂. These data for the plastic bottles show marked changes that fall along three distinctive curves (dashed lines) for each type of water. Each curve shows two potential processes for carbon gain or loss: (1) loss of CO₂ from the plastic bottle as the dominant carbon isotope fractionation process during the first six months of storage; (2) carbon isotope exchange which becomes the dominant process after six months in storage. These results indicated that regardless of storage time, all the water samples stored in plastic bottles would eventually approach the same ¹⁴C values, whose final values depend on the (modern) radiocarbon content of the lab air CO₂. The initial data points of GRA plotted close to the zero-age line, indicating that this water was relatively young (in terms of ¹⁴C age) compared to the other two waters that fell below the zero-age line.

Figure 1. Carbon isotope composition of soil CO₂ (from soil organic carbon), atmospheric CO₂, HCO₃^– in water equilibrated with atmospheric CO₂, and groundwater samples in bottles (points). X: Initial samples shipped to the radiocarbon laboratory in glass bottles as the starting values. Solid triangles: samples stored in pre-evacuated glass bottles for 12 months before isotope analysis. Solid circles: Samples stored in plastic bottles. Red arrow denotes simulated changes in carbon isotope contents caused by CO₂ loss.

The evolution of carbon isotope contents of the three samples over time was considerable and well above expected level of uncertainties associated with isotope analysis and sample handling (Table 2, Figure 2). The ¹⁴C changes over time with storage in plastic containers were significant after just six months (+1–3 pmC for ¹⁴C and +0.2–4.5 ‰ for δ¹³C respect to the typical analytical uncertainty) and 12 months (+2–15 pmC for ¹⁴C and +0.5–5.5 ‰ for δ¹³C respect to the standard analytical uncertainty).

Figure 2. Temporal evolution of ¹⁴C and ¹³C contents in water samples after storage in three different types of plastic bottles for 6–12 months. Isotope values at t = 0 indicate the original contents determined in glass bottles at the start of the experiment.

Table 2. ¹⁴C activities (pmC) and ¹³C content (in per mil deviations vs VPDB) and their respective differences relative to the original value at different experimental times for waters stored in plastic bottles (Δ¹⁴C).

	^14C (pmC)	Δ^14C (pmC)	^14C (pmC)	Δ^14C (pmC)	δ^13C (‰)	Δ^13C (‰)	δ^13C (‰)	Δ^13C (‰)
Voslauer	At t = 0, ^14C = 7.63 pmC				At t = 0, ^13C = –5.95 ‰
	t = 6 months		t = 12 months		t = 6 months		t = 12 months
LDPE	11.1	3.47	22.70	15.07	–5.27	0.68	–3.63	2.32
Nalgene	9.92	2.29	13.04	5.41	–5.77	0.18	–4.93	1.02
HDPE	9.74	2.11	10.50	2.87	–5.81	0.14	–3.73	2.22
Juvina	At t = 0, ^14C = 2.20 pmC				At t = 0, ^13C = –0.79 ‰
	t = 6 months		t = 12 months		t = 6 months		t = 12 months
LDPE	3.54	1.34	4.42	2.22	3.69	4.48	4.22	5.01
Nalgene	3.21	1.01	12.50	10.3	2.85	3.64	4.59	5.38
HDPE	2.35	0.15	20.41	18.21	3.50	4.29	4.50	5.29
Grafendorf	At t = 0, ^14C = 41.0 pmC				At t = 0, ^13C = –10.38 ‰
	t = 6 months		t = 12 months		t = 6 months		t = 12 months
LDPE	42.46	1.46	51.83	10.83	–8.66	1.72	–6.25	4.13
Nalgene	41.19	0.19	50.61	9.61	–9.76	0.62	–6.96	3.42
HDPE	41.77	0.77	–	–	–9.59	0.79	–9.75	0.63

These data graphically show that the changes in ¹⁴C values were significant but relatively small during the first six months (up to 3 pmC), but after six additional months the changes in ¹⁴C became more pronounced (up to 15 pmC). In the case of δ¹³C, the observed isotopes shift reached 5 ‰ after one-year storage. One possible explanation is that during the initial stage of sample storage, irreversible escape of CO₂ from water was the predominant process; therefore, changes in ¹⁴C were mainly caused by carbon isotope fractionation, whose effects were minor relative to those found over storage periods longer than six months. However, after a longer storage times the loss of CO₂ from the diffusive ingress of atmospheric CO₂ became prevalent (i.e. carbon isotope exchange). Reversible carbon isotope exchange between water and the atmosphere becomes the dominant process at this later stage, and suggested the ¹⁴C content would continue to increase beyond 1 year of storage. This isotope effect was more significant than the loss of CO₂ from water because CO₂ in air has very high ¹⁴C values (currently 100–110 pmC; [29]), and the isotope separation from two compartments is higher. A similar pattern was observed for δ¹³C, except the CO₂ loss was higher for JUV (carbonic-rich water) at earlier storage times. Thus, it appears that δ¹³C can also be used as a proxy for modern air CO₂ contamination, which contradicts previous assertions [11].

Therefore, storage of DIC samples in plastic bottles over more extended periods of time (e.g. several months) will eventually lead to increasing ¹⁴C-DIC contamination and, accordingly, progressive groundwater age underestimations (younger). The effects on groundwater with different radiocarbon content are also a factor, with a more significant impact on storage effects for old, low-¹⁴C ground water samples (Table 3), which would cause Type 2 errors [11]. These long-term errors on ¹⁴C-DIC due to inadequate storage were induced by carbon isotope exchange processes, and they appear to mimic geochemical processes found in natural systems. Based on this experimental evidence, it seems obvious that plastic bottles should never be used for ¹⁴C-DIC sample collection and storage. Plastic containers allow not only (irreversible) loss of CO₂ but also (reversible) carbon isotope exchange between atmospheric CO₂ with DIC in the water due to diffusion and permeability to CO₂. In summary, our study suggested that two processes affect the carbon isotope compositions of groundwater during long-term storage in plastic bottles: (1) carbon isotope fractionation during CO₂ loss, and (2) diffusive isotope mixing between the atmospheric CO₂ and the sample DIC. Our results also indicated that the storage bias for each type of plastic bottle or for each water sample of known isotopic composition was not systematic, which makes predictable mathematical corrections impossible.

Table 3. Uncorrected and corrected groundwater ages (see text for details) were calculated for the three mineral water samples after one week (t = 0), and after 6 and 12 months of storage in different plastic containers. For comparison purposes, the initial ¹⁴C value of DIC (¹⁴C₀) was corrected based on carbonate reactions in the bedrock.

	Uncorrected ^14C age (ka)			Δ age (ka)
	t = 0	t = 6 months	t = 12 months	after 12 months
VOS	23.0 ± 2.4	18.8 ± 0.6	15.9 ± 3.3	7.1
JUV	32.6 ± 1.5	29.0 ± 1.8	21.5 ± 6.1	11.1
GRA	7.4 ± 0.1	7.2 ± 0.1	5.5 ± 0.1	1.9
	Corrected ^14C age (ka)			Δ age (ka)
VOS	15.2 ± 3.1	10.29 ± 1.0	5.1 ± 3.8	10.1
JUV	15.1 ± 4.2	n/a	n/a	–
GRA	3.2 ± 0.1	2.3 ± 0.6	modern	3.2

3.2. Processes associated with carbon isotope fractionation

The CO₂ loss from groundwater samples during long-term storage in plastic bottles is an irreversible unidirectional process that causes carbon isotope fractionation. The lighter carbon isotopes preferentially go with CO₂ gaseous phase following ¹³C fractionation [30]. The carbon isotope fractionation caused by this process is independent of laboratory air CO₂ concentration because this process is irreversible [31]. The process of CO₂ loss will be more significant for carbonic-rich water samples high in HCO₃⁻ contents and pCO₂ (e.g. JUV). Similarly, the ¹⁴C isotope fractionation factor is about –0.2 × 9.6 pmC since isotope ¹⁴C fractionation is approximately proportional to the mass difference ([32], also see [24]). Put another way, a δ¹³C change of 1 ‰ leads to a change in ¹⁴C of ca. 2 ‰ ( =  0.2 pmC). This shift is represented graphically in Figure 1 (as red arrow), which shows changes in δ¹³C and ¹⁴C simulated using the isotope fractionation factors of carbon atoms between air CO₂ and HCO₃⁻ in water.

Over prolonged storage periods in plastic bottles, the carbon isotope values were mainly controlled by reversible exchange reactions with atmospheric CO₂. This process is comparable to a mixing process. After an infinite time of storage, the HCO₃⁻ in the sample water should have a ¹⁴C value of ca. 100 pmC, as shown in Figure 1, and a δ¹³C value that is ca. 9 ‰ higher than the surrounding lab air CO₂, considering the ¹³C fractionation factor that can be estimated by ϵ_CO2/HCO3 = −(9483/T) + 23.89 (−8.5 ‰) at 20 °C, and the δ¹³C value of atmospheric CO₂ is between −8.8 and −7.4 ‰ [28].

4. Conclusions and protocol recommendations

Proper protocols for sampling and storing DIC samples for radiocarbon should have (1) no atmospheric ¹⁴CO₂ contamination during collection, and (2) no influence of subsequent carbon isotope exchange and bacterial processes during sample storage. Both objectives can only be reliably achieved using glass bottles with additional precautions to eliminate microbial activity. Because groundwater samples are often stored for months before analysis in laboratories with AMS technologies, we highly recommend using pre-evacuated glass bottles (poisoned) sealed with butyl stoppers. Groundwater samples should be collected and preserved in the field without any exposure to the atmosphere (i.e. bottles submerged in a gently overflowing container).

An example of a pre-evacuated glass bottle DIC sampling kit is shown in Figure 3. This kit contains 12 acid-washed and oven-sterilized 125-mL glass serum bottles sealed with thick butyl-blue stoppers. The bottles are pre-evacuated to 10^–3 Torr using a rotary vacuum pump to allow for groundwater filling by suction. Based on our experience, these bottles hold vacuum for > 2 years. Adding bactericide, such as sodium azide, in the glass bottle prior to evacuation is optional, but highly recommended (toxin transportation restrictions may apply). Samples should have a minimum of 100 mg/L as HCO₃^– to obtain enough carbon for ¹⁴C and ¹³C. Samples should be collected in duplicate if the alkalinity is lower than 100 mg/L or use a larger glass sampling bottle. Groundwater samples are filled into the bottles by suction (or injecting filtered water through the septa by syringe). Samples collected and stored in this way with microbial preservation will retain the ¹⁴C and ¹³C values for at least one year and possibly longer. At the time of groundwater sampling, water temperature, pH and alkalinity (and additionally, dissolved oxygen (DO) and redox potential) are essential parameters to be measured at the field to interpret ¹⁴C- and ¹³C-DIC data and for use in the ¹⁴C geochemical age correction models.

Figure 3. Photo of a ¹⁴C DIC field sampling kit. It includes pre-evacuated (10^–3 Torr) glass containers sealed with thick butyl-blue stoppers, syringes and 0.45 µm barrel filters and 21G needles stored in a protective (air travel robust) field transportation case. For biologically active or nutrient-impacted groundwaters, the pre-evacuated glass bottles can be poisoned by adding HgCl₂ or Na-azide salt to prevent microbial growth that may alter the carbon isotope values during storage.

The accuracy of groundwater age determinations using ¹⁴C is critical to better understanding aquifer dynamics and characterizing climatic effects on water resources. The routine use of pre-evacuated glass bottles using the sampling protocols above is highly recommended. Inaccurate ¹⁴C and ¹³C results data lead to erroneous hydrogeological conclusions and recommendations regarding groundwater flow model calibration and resource management decisions. This technical note reveals that the kind of plastic bottles traditionally used for groundwater DIC sampling, even those made of HDPE, are inappropriate for ¹⁴C sample collection due to the loss and exchange of CO₂ through the plastic walls, and whose adverse effects increase markedly with longer storage times and cannot be fully predicted.

Acknowledgements

The International Atomic Energy Agency supported this study. Correspondence regarding the DIC sampling kit should be directed to [email protected].

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by the International Atomic Energy Agency.