FT-IR metrology aspects for on-line monitoring of CO₂ and CH₄ in underground laboratory conditions

Abstract

The French National Radioactive Waste Management Agency (ANDRA) has recently developed a new experimental set-up which allows sampling water from marl rock formations, together with an in situ characterisation of the composition and migration mechanisms of the gases dissolved in the marl porewater. Gases and liquids are collected from vertical borehole drillings in underground laboratories. The analytical design, Fourier transformed infrared spectroscopy based, allows powerful and long term on-line monitoring of gases released by low-permeability media. The IR system is designed to cope with the unfavourable measurement conditions occurring in an experimental underground laboratory (moisture, dust, etc.). Because the working conditions in such an underground laboratory make complete purging of the IR spectrometer difficult, the IR spectra of geological gases are often perturbed by contributions from atmospheric CO₂ and water vapour. The metrology aspect is based on an IR low resolution sensor equipped with two measurement compartments. In the internal compartment linked to the borehole layout, gases are monitored on-line through a cell with a variable optical path, whereas in the external compartment, atmospheric CO₂ is measured through a short open path configuration. The experimental method and data processing procedure used to determine the real partial pressure of CO₂ arising from the marl rock formation are described in this paper. Results of the on-line gas (CO₂ and CH₄) monitoring conducted in the Mont Terri underground laboratory are presented and compared with punctual gas chromatography analyses.

---

Introduction

Sedimentary rocks, and among them, argillaceous rocks, contain various gases (CO₂, alkanes, H₂, O₂, N₂, H₂S, SO₂, COS, NH₃, noble gases) of organic and inorganic origin. They are present as vapour, or as supercritical phase and/or dissolved in porewater according to their concentrations and the conditions of total pressures and temperatures. Organic gases originate from the diagenesis of fossil organic matter trapped during sedimentation. Alkanes and CO₂ are the most common gases found in sedimentary rocks and are sometimes associated with H₂S and N₂.¹CO₂ is in chemical equilibrium with carbonate minerals. They result from the biotic–abiotic decay of organic matter and can also be derived from bacterial fermentation, whereas CH₄ mainly forms through the thermal or bacterial degradation of organic matter. CH₄ thermogenesis generates large amounts of heavier gas molecules (C2, C3, C4, C5), whereas bacterial methanogenesis does not produce any additional alkanes.^2,3 The presence of dissolved gases and their respective concentrations influence the geochemistry of porewater, and consequently alter the texture and mineralogy of the rock through chemical interactions. Carbonaceous clay formations are the main gas reservoir rocks in sedimentary basins.

Clay-based rocks are also considered as potential host formations for deep geological disposal of high-level and long-lived intermediate-level radioactive wastes.^4,5 Clay formations contain interstitial porewater but display a low permeability, thus limiting radionuclide mobility. However, the retention properties of clay formations can vary depending on the chemistry of interstitial water.^6,7 Characterising porewater is intrinsically difficult because its extraction from clayey media is complex (low content and low permeability).^8,9 Furthermore, the extraction techniques can alter the porewater chemistry through atmospheric exposure (i.e. pH variation and Eh increase). To avoid any atmospheric contamination of porewater during sampling, a new experimental concept of porewater extraction was implemented for the first time at the Mont Terri rock laboratory (Switzerland) through the PC-C (Porewater Chemistry-C) experiment.¹⁰ The experimental concept was duplicated in the underground research laboratory (URL) in Bure (North-East of France) built by the French National Radioactive Waste Management Agency (ANDRA). Early 2005, a borehole was drilled in the drift at a depth of 445 m (experiment PAC2002). At the end of 2005, another borehole was drilled at a depth of 490 m and was equipped (experiment PAC1002)¹¹ with a specific device called a completion. In each experiment, a borehole interval was filled with argon at a pressure close to 1.3 bar in order to follow the evolution of the composition of the gas mixture by exchange with the host rock and to sample porewater under controlled conditions. All conditions were designed to avoid rock oxidation and microbiological perturbation in either the borehole or the gas circuit (dry-drilling with nitrogen, aseptic tool and total pressure >1 bar). A water-sampling module pumped the water flowing from the rock due to the hydraulic pressure difference between the rock and the interval (40 bar to 1.3 bar). This module allowed online analysis of water and guaranteed that the main part of the interval remained free of liquid water. Gas in contact with rock and seepage water circulated in an external gas module. The gas mixture was then sampled at various dates, thus providing the evolution of its composition upon exchange with the rock. The gas circulation module also allowed on-line non-destructive gas monitoring.¹²

The main objective of this paper is to introduce a system, which allows powerful and long term on-line monitoring of gases released by low-permeability media. Knowledge of the nature, partial pressure and transfer mechanisms of various gases is very valuable for assessing porewater chemistry and consequently, for predicting the behaviour of radionuclides in deep geological waste storages.¹³

Several in situ gas analysers have been developed to monitor atmospheric trace species (CO₂, CH₄, N₂O, CO,…). The most frequently used ones for measuring CO₂ are based on non-dispersive infrared spectroscopy (NDIR). In standard air composition the accuracy can reach 0.01 µmol mol⁻¹ for punctual measurements and 0.003% for hourly measurements.¹⁴ Using Fourier transform spectroscopy with a resolution of 1 cm⁻¹, various gases can be analyzed simultaneously with an accuracy of 0.04% for 360 ppmv CO₂, 0.05% for 1700 ppmv CH₄, 0.5% for 50 ppmv CO and 0.1% for 310 ppmv N₂O.¹⁵ Such accuracy can be obtained by using an optical path length of 22 m and averaging 256 scan spectra in recording between 2300 and 3000 cm⁻¹ during 8 min. Non-optical techniques such as mass spectroscopy or gas chromatography¹⁶ can also be used for measuring CO₂, CH₄, N₂O, CO, etc. Both these methods require sample extraction and preconditioning, and the use of specific sensors. An accuracy of 0.02% for CO₂ (360 µmol mol⁻¹), 0.2% for CH₄ (1700 nmol mol⁻¹),¹⁷ 0.1% for N₂O (300 nmol mol⁻¹)¹⁸ and 1% for CO (100 nmol mol⁻¹) can be obtained from these experimental methods.¹⁹ Recent technological advances exhibit the potentialities of quantum cascade laser (QCL) for monitoring gases.^20,21 The advance of QCL technology allows the use of very promising tuneable, miniature, non-cryogenic mid-IR sources for online, in situ and continued monitoring of gases,^22–24 which cover down to 100 cm⁻¹spectrum range, with high resolution (<1 cm⁻¹) and high speed measurement (to 100 cm⁻¹ per second). Coupled to photoacoustic spectroscopy, QCLs could lower the detection limit to ppb or sub-ppb.²⁵

Choosing a sensor for on-line non-destructive and non-altering analysis of a quasi-unknown gas mixture in the context of an underground laboratory appears as a complex task that must take into account various key-parameters. The first stage is to choose the most appropriate technique for gas analysis, taking into account four critical constraints: (i) full spectrum analyses are needed because the gas composition is not sufficiently well-known at the beginning of the project, (ii) accurate analyses are needed because gas concentrations can be rather low, (iii) the technique must be insensitive to argon (carrier gas in the borehole), and (iv) the technique must be sensitive to CO₂ and to the main organic gases (light alkanes) present in the porewater of the rock formation. Infrared spectroscopy is a vibrational spectroscopy frequently used for analyzing natural gases such as CO₂ and light alkanes.^15,26,27 In the mid-IR range (5000–600 cm⁻¹), the fundamental vibration frequencies²⁸ of C–H, C=O and C=C are all active and precisions down to the ppb can be reached under favourable conditions.^25,29,30 The first constraint leads to the choice of a whole spectrum technique, like Fourier transform infrared spectroscopy (FT-IR) which can cover all the mid-IR range. QCL technology doesn't provide a full spectrum analysis (applicable only to ca. 100 cm⁻¹) and the composition of the gas mixture must be previously known in order to choose the type of sensor and select the appropriate line for each compound. Furthermore, Fourier transformed infrared spectrometers can be equipped with a variable long path White cell,^31,32 which presents two advantages: (i) it is versatile in terms of sensitivity since a variable optical path can be used, (ii) such a cell can be linked to a gas collection circuit and allows on-line and flux analyses.^28,33,34 The use of a White cell allows the second constraint to be addressed.

Working conditions in underground galleries require specific adaptations of the FT-IR spectrometer taking into account moisture, vibrations, dust, and the non-controlled atmosphere. Moreover, for security reasons, it was not allowed to continuously use pressurized gases in the URL. As a consequence, the FT-IR spectrometer must work without any atmospheric purge, in contrast to conventional laboratory equipment. In such conditions the infrared beam of the IR device crosses the gas inside the cell and the gas from the atmosphere, both containing CO₂. Therefore, the IR signature due to atmospheric CO₂ is collected together with the signature from the borehole gas mixture. Part of this paper will focus on the development of an original analytical procedure specific to the quantitative determination of CO₂ in the presence of a large contribution from atmospheric CO₂. An example of gas measurements in natural samples acquired in the Mont Terri rock laboratory will be presented and discussed. Because first results showed that CO₂ and CH₄ are the two major species identified by FT-IR, this paper focuses on their evolutions. The obtained spectra allow the determination of other gases, the study of which will be developed in future work.

Equipment and methods

IR system design: a double compartment system

One possible way to subtract the contribution of atmospheric CO₂ from measurements is to record a reference spectrum before each measurement. However, recording reference spectra requires either emptying the gas cell, or filling it with a neutral gas. Evacuating the gas cell induces a loss of gas from the circuit. Filling it with a neutral gas requires, after recording the reference spectra, that the neutral gas must be transferred into the circuit or into the atmosphere. Both solutions were, however, rejected because (i) gas transfer in the circuit dilutes the original gas mixture and impacts the equilibrium of the gas mixture/borehole system, (ii) neutral gas transfer in the atmosphere requires vacuum pumping that could lead to an important head loss when the cell is returned on-line. Consequently, the FT-IR spectrometer must work without any atmospheric purge and without recording new reference spectra before each measurement.

As the infrared beam crosses the gas cell and the atmosphere (internal compartment Fig. 1), IR signals due to atmospheric water vapour and carbon dioxide are collected together with the signals of water vapour and carbon dioxide from the borehole gas mixture. In order to decompose this superimposition of signals, a second compartment specifically dedicated to the monitoring of atmospheric gases must be added to the IR device. It uses the same source and interferometer (external compartment Fig. 1). In this second compartment, the IR beam crosses only the atmosphere from the source to the detector. Note that the geometry of this optical bench implies that the length of the optical path crossing the atmosphere is different for each compartment, and by consequence, must be determined for quantitative CO₂ measurements. Such a configuration has already been used for other purposes.³⁵

Fig. 1 IR spectrometer with a double compartment system. The modulated IR beam can be switched between the two compartments through the use of a dual-position rotational mirror. The internal compartment is in the left part and the external compartment in the right one.

The chosen FT-IR device is a double compartment low resolution Fourier transform infrared spectrometer (TENSOR 27 from BRUKER), as presented in Fig. 2. To minimise the impact of an open system, it is equipped with a “water resisting” ZnSe beamsplitter and windows. The first compartment, called the internal compartment, is equipped with a multipass gas cell (variable long path gas cell A 136/2-L BRUKER) connected to the gas collection circuit. The optical path length of this gas cell can vary from 0.8 m to 8 m and its volume is 2 L. The gas cell is equipped with CaF₂ windows. A drawback associated with such a choice is that the infrared range accessible to CaF₂ is limited to wavenumbers at least higher than 1000 cm⁻¹. The second compartment, called the external compartment, is dedicated to the monitoring of the atmosphere and acts as an open path system working in active mode. The modulated beam is switched between the two compartments through a software controlled rotational mirror. Each compartment is equipped with a DTGS sensor.

Fig. 2 View of the IR spectrometer with a double compartment system in use at the Mont Terri laboratory.

Global experimental setup: borehole design, completion, gas circulation module and IR system. Fig. 3 presents the global layout of the experimental device used in the URL. The equipment consists of three elements: the borehole and its equipment (element A), the module for gas circulation and gas sampling (element B) and the previously described IR system (element C). The borehole equipment, designed by Solexperts AG,¹² consists of a hollow cylinder 15 m in length through which gas and water lines, together with temperature and pressure transducers, cross the borehole up to the module for gas circulation and gas sampling (element B). Plugging of the cylinder is obtained by using two packers at both ends and resin in contact with the clay formation along the first 10 m. The last 5 m constitute the test interval for gas collection. At a distance of 10 m from the gallery, the test interval is considered to be located in rock unperturbed by the excavation of the gallery. The lines crossing the completion are two gas circulation lines (injection and extraction), several pressure and temperature measurement lines and an independent line for pumping water linked to the water sampling module especially designed by Metro-Mesures company.^10,11 The module for gas circulation and gas sampling (element B), specifically designed by Solexperts AG, is located in the gallery next to the borehole.¹² A low flow pump (20 mL min⁻¹) allows gas circulation without any contact with the atmosphere. Using several cells from 75 to 150 mL, gas sampling can be carried out without any pressure change in the gas lines.

Fig. 3 Layout of the experimental device used in the Mont Terri and the ANDRA experimental underground laboratories A: sectional drawing of the borehole and completion, B and C: layout of the device for sampling and online analyses.

Measurement parameters. The temperature is constant in the clay formation (about 20 °C). Therefore, the temperature of the gas mixture collected from the borehole is considered to be constant except when ventilation stops. In contrast, the pressure varies in the gas circuit because of leaks in the circuit (about 1 mbar per day). The pressure in the gas cell and bulk atmospheric pressure are then recorded during each experiment. All spectra are measured with a spectral resolution of 1 cm⁻¹, which guarantees a good signal to noise ratio and provides sufficient time resolution for the measurements carried out with a Blackman–Harris apodization function and a Mertz-type phase correction. The spectra are scanned ten times, resulting in a measurement time of 30 s. In each measurement, two single-beam spectra are recorded sequentially, the first one from the internal compartment and the second one from the external compartment. Absorbance conversions are carried out using background correction. In both cases, the background correction is obtained through a single beam spectrum modified by spectral treatment. In the case of the internal compartment, the single beam spectrum is recorded with the gas cell under vacuum, i.e. the IR beam crosses the atmosphere on either side of the gas cell. In the case of the external compartment, the single beam spectrum is acquired without changing the configuration, i.e. the IR beam crosses the atmosphere. In both cases, the spectral treatment used for the modification of the single beam spectra used for background correction consists in deleting the residual atmospheric CO₂ contribution by the generation of a straight-line between 2200 and 2450 cm⁻¹.

Quantitative aspects. The initial spectra acquired in the clay formation show intense CO₂ and CH₄ contributions. The procedure detailed in the companion paper of this work³⁶ was performed in laboratory conditions (i.e. purged spectrometer) at a constant temperature of 22 °C, i.e. 2 °C higher than the temperature of the gas mixture in the borehole. In this study, this difference of temperature was neglected. The results show significant deviations from Beer–Lambert law, mainly due to the low spectral resolution of 1 cm⁻¹.³⁷ Our study developed the quantification of CO₂ and CH₄ partial pressures, analysed by our low resolution FT-IR system taking into account bulk pressure dependence to cope with the bulk pressure variations observed during the field experiments. Evolution of the integrated intensity of the CO₂ν₃ band as a function of pCO₂ can be reproduced using a third degree polynomial relation at constant bulk pressure. This relation f can be expressed as:

f(A) = k₁A + k₂A² + k₃A³ = lp_CO2 (1)

Our study³⁶ confirmed that coefficients k₁, k₂, and k₃ are dependent on the bulk pressure according to polynomial relations. Consequently, eqn (1) becomes:

f_Pbulk(A) = k₁(P_bulk)A + k₂(P_bulk)A² + k₃(P_bulk)A³ = lp_CO2 (2)

The same type of result was obtained for CH₄ quantification.

Analytical development

This section describes how the spectra recorded in the two compartments are treated to acquire the CO₂ partial pressures in the borehole. CH₄ partial pressures are obtained with the calibration procedure developed in the companion paper of this work.³⁶

Virtual representation of the double compartment system: combined gas cells approach. In order to accurately model the interaction of the IR beam, from the source to the detector, with all the gas phases, the experimental device presented in Fig. 1 can be described in terms of a series of gas cells. This approach was first developed by Plass^38,39 in order to approximate the extremely variable conditions along an atmospheric slant path. The virtual representation of the double compartment IR system, according to such an approach, is described in Fig. 4. The internal compartment can be simulated as a set of double gas cells in series (α and β in Fig. 4, left). The α gas cell is a virtual gas cell filled with atmospheric gases at atmospheric pressure (P_atm). The optical path length l_α of the α gas cell represents the optical path length of the internal compartment that crosses the atmosphere. The β gas cell represents the real multipass gas cell (internal compartment, Fig. 1) filled with the borehole gas mixture at gas cell bulk pressure (P_cell). The optical path length l_β is the optical path length fixed in the multipass cell. In this case, the atmospheric gas mixture can be considered as a sample filling a gas cell (α) installed “in series” with the real gas cell β. The IR signal reaching the detector results in the combination of the IR response from both α and β gas cells.

Fig. 4 Representation of the double compartment system. The full line represents the IR beam crossing the gas cell filled with borehole gas mixture. The dotted line represents the IR beam crossing the atmosphere. α, β and α′ refer to the corresponding simulated gas cells.

The second compartment can be simulated as a simple gas cell noted α′ (Fig. 4, right), filled with atmospheric gases at atmospheric pressure (P_atm). The optical path length l_α′ of the α′ cell represents the optical path length of the external compartment which crosses the atmosphere from the IR source to the DTGS sensor of the external compartment (Fig. 1). The difference between the two virtual α and α′ gas cells is their optical path length l_α and l_α′, which is unknown and must be determined.

Use of the combined gas cell approach to determine the optical path ratio (κ) between the external and internal compartments. When the β gas cell (Fig. 4) is empty, the part of the IR beam crossing the β gas cell does not contribute to the acquired spectra. The IR contributions to the spectrum of the virtual internal compartment therefore originate from atmospheric gases present in the α gas cell. In the external compartment, the IR contribution is due to atmospheric gases present in the α′ gas cell. The upper part of Fig. 5 shows the IR spectra of the CO₂ν₃ band recorded in this configuration for each compartment. A_α and A_α′ are the integrated areas (absorbance) of CO₂ contribution in the internal and external compartments, respectively. Bulk pressure, temperature and p_CO2 are the same in both α and α′ gas cells. In these conditions, we can write:

p^α_CO2 = p^α′_CO2 (3)

where p_CO2^α and p_CO2^α′ correspond to the atmospheric CO₂ partial pressure in the α and α′ gas cells, respectively. Combining eqn (2) and (3), yields eqn (4):Eqn (4) links IR signals from the two compartments. Data interpretation requires knowing the optical atmospheric path length ratio κ. To that end, complementary experiments were developed in the ANDRA underground research laboratory in Bure (experiment PAC1002 in −490 m experimental drift). During 16 days the multi-pass gas cell (β) of the double-compartment system was maintained in vacuum conditions (about 0.1 mbar) with a vacuum pump linked to the gas cell. The atmospheric pressure and CO₂ spectra, allowing one to obtain the A_α and A_α′ values, were regularly recorded (every hour). Fig. 6 shows the evolution of the atmospheric pressure of f_Patm(A_α), of f_Patm(A_α′), and of the k ratio calculated using eqn (4), except during the fifth day of measurement because of a power cut. The bulk atmospheric pressure varied between 1 and 1.05 bar (Fig. 6a). The values of f_Patm(A_α) and f_Patm(A_α′) represent the partial pressure evolution of CO₂ expressed in mbar m. They follow the same evolution in both compartments, but their values are significantly different (Fig. 6b and c) due to the difference between l_α and l_α′. As expected, κ determined through eqn (4) is constant. Its average value is 0.716 with a standard deviation of 0.005 (Fig. 6d).

Fig. 5 Upper part: internal (a) and external (a′) compartment spectra with β gas cell emptied. Lower part: internal (b) and external (b′) compartment spectra with β gas cell filled with a standard gas mixture containing argon and 750 ppmv of CO₂ (p_CO2^β = 0.825 mbar) with a bulk pressure P_cell of 1.1 bar.

Fig. 6 Evolution of (a) bulk atmospheric pressure, (b) f_Patm(A_α), (c) f_Patm(A_α′) and (d) κ ratio between 28 November 2006 and 28 December 2006. Measurements were obtained in the ANDRA underground laboratory in Bure (−490 m). f_Patm(A_α) and f_Patm(A_α′) represent p_CO2^α and p_CO2^α′ values balanced by l_α and l_α′ optical path lengths, respectively.

Determination of absolute CO₂ content in the borehole. Once the optical path length ratio k is determined, the spectra measured in the internal and external compartments can be used to determine the absolute partial pressure of CO₂ coming from the borehole.

Experiments were first performed in a surface laboratory with the same equipment in experimental conditions replicating those encountered in the URL. The gas cell (β) can then be either empty or filled with standard gas mixtures (from Messer France SAS). Bulk pressures in the cell (β) and in the atmosphere are recorded (α and α′ gas cell). The equipment design is that of Fig. 4. The lower part of Fig. 5 shows an example of spectra obtained when the β gas cell is filled with a standard gas mixture containing argon and 750 ppmv of CO₂ (p_CO2^β = 0.825 mbar) with a bulk pressure P_cell of 1.1 bar. The optical path length l_β in the cell is fixed at 1.6 m. The bulk pressure involved in α and α′ gas cells is the atmospheric bulk pressure P_atm (983 mbar). A_αβ and A_α′ are the integrated areas of this band measured in the internal (both α and β gas cell) and external compartment (α′ gas cell), respectively.

According to the mathematical development of Plass,^38,39 the IR beam crosses the cells in series, with independent bulk pressures, temperatures and concentrations of absorbing gas. Plass modelled a multi-cell system with jcells in series by a virtual homogeneous (h) one-layer system associated with a bulk pressure P_h, a temperature T_h, an amount of absorbing gas u_h and an optical path length l_h. The signal received by the detector corresponds to the signal produced by a virtual homogeneous one-layer system (i.e. from all the jcells in series). In the case of the integration of a non-well defined and non-opaque band and if the temperature in the cells is constant, Plass gives the following simple relation linking the virtual homogeneous one-layer system with the model of jcells in series:

with u_j = ρ_jl_j (5b)

In our case, the mass of absorbing gas (CO₂) per unit area u and f_P(A) are expressed by the following relation, based on the ideal gas law and eqn (2):Eqn (6) shows that f_P(A) and u are proportional.

As our internal compartment has been simulated as a two-cells in series system (α and β gas cell), u_h becoming noted u_αβ, eqn (5a) is as follows:

u_αβ = u_α + u_β (7)

The combination of eqn (6) and (7) gives eqn (8):

f_Pαβ(A_αβ) = f_Patm(A_α) + f_pcell(A_β) (8)

where P_αβ is the “average” pressure of the bulk system with two cells in series (α and β gas cells).

For the rest of this work, it will be assumed that the bulk pressure P_αβ characteristic of the bulk signal received by a detector from the internal compartment is the average of P_cell and P_atm:

The final objective of the procedure is to determine p_CO2^β, the partial pressure of CO₂ in the β gas cell representing the multipass gas cell of the internal compartment. Determination of p_CO2^β means determining f_pcell(A_β) (eqn (2)) or u_β (eqn (6)). Combining eqn (4) and (8) we arrive at:

f_pcell(A_β) = f_Pαβ(A_αβ) − κf_Patm(A_α′) (10)

The partial pressure of CO₂ (p_CO2^β) in the cell is calculated by dividing f_pcell(A_β) values by the value of the optical path length in the cell (l_β):

Results and discussion

Validation of the method for the absolute determination of CO₂

Gas cells filled with standard gas mixtures, with known bulk pressure and the same optical path length as used in the URL experimentation (i.e. 0.80 m), were analyzed to validate both the experimental and data treatment aspects. Six standard gas mixtures were used; they were different to those used for the calibration step.³⁶ They contained 152, 344, 490, 1480, 1970 and 2500 ppm of CO₂, respectively, the rest being argon. The atmospheric bulk pressure during the experiments was around 1000 ± 5 mbar. Triplicate spectra for each sample were acquired with five different bulk pressures, between 900 and 1300 mbar. Gas mixture concentrations were certified with a relative uncertainty of 3% and bulk pressures were measured with an accuracy of ±4 mbar. In each measurement, the CO₂ spectra were recorded from both the internal and external compartments, which allowed us to calculate A_αβ and A_α′, coupled with bulk pressure in the cell and atmospheric pressure (P_cell and P_atm) measurements. A_αβ values higher than 100 U.A. were not taken into account because repeatability of the measurements is not sufficient.³⁶

Results are presented in Table 1: the mean obtained value for all gas cell bulk pressures is given with its relative standard deviation (RSD) and the relative root mean squared error of prediction (RMSEP) of the determination method of absolute CO₂. Taking into account that the RMSEP of the calibration model is included between 1 and 3%, the relative RMSEP obtained for absolute CO₂ determination is reasonable. The calculations are considered as truly valid for values higher than 0.3 mbar m, as the relative RMSEP in the calibration model increases for values lower than 0.3 mbar m.^36,40 It can be concluded that the performance of the proposed calculation algorithm depends on the amount of CO₂ in the cell. In the case of low CO₂ amounts, close to 0.3 mbar m, it is proposed to increase the optical path length in the cell in order to increase accuracy.

Table 1 Comparison of the true concentration of CO₂ and the estimation calculated from FT-IR spectra taking into account atmospheric pressure and bulk pressure in the gas cell

Declared (ppm)	Mean obtained value (ppm)	RSD (%)	RMSEP (%)
152	141	19	18
344	339	5	5
490	480	2	3
1480	1484	2	2
1970	2016	2	3
2590	2623	1	1

---

Measurements of gases in natural samples

The PC-C experiment borehole located in the underground rock laboratory of Mont-Terri (St Ursanne, Switzerland) was studied for gas phase exchange with the rock from March to December 2004.¹⁰ During this period of 301 days, the test interval, which was filled with pure argon the first day, was purged twice with argon after 44 and 232 days for technical reasons. Before the second argon filling, experimental disturbances took place, impacting mainly on the homogeneity of the total pressure in the gas circuit. Only the last 70 days are considered to have yielded significant results. These results are presented in Fig. 7 for CO₂ (a) and CH₄ (b), respectively. During this period, 838 pairs of IR spectra were acquired (with an optical path length of 0.8 m) and 7 gas sampling cells were extracted from the circuit and analyzed by gas chromatography. For CO₂ concentration evolution, measurements by the IR system were stopped on the 254^th day while the quantitative measurement of CH₄ continued until the end of the experimental period. After the 254^th day, the intensity of CO₂ band (sum of atmospheric and borehole CO₂ signals, i.e.A_αβ) exceeded the limit of quantification (100 U.A.).³⁶ The evolution of the CO₂ concentration showed a sharp increase in the first 5 days of the experiment, followed by a slight regular increase during the rest of the time. These results are in good agreement with the 6 gas chromatography results acquired on sampling days 231, 253, 286, 294, 296, 299 and 301, except for the datum acquired on day 273 (Table 2), that is not in accordance with the other five gas sampling GC data.¹⁰CH₄ concentration evolution shows a quasi-monotonic increase with time during the same period, that is confirmed by seven gas samples analysed by gas chromatography.¹⁰ The continuous measurements obtained through the IR system show a weak dispersion compared with those obtained through the gas sampling method. Dispersion of gas sampling data can be due to several reasons: (i) the gas cells are not located exactly at the same place in the gas circuit and data dispersion could represent heterogeneities in CO₂ amount in the circuit, (ii) manipulation or storage of gas cells can induce slight leakages by transfer operations from one container to another, (iii) local adsorption of CO₂ on the wall of containers, (iv) low reproducibility of GC data with time because analyses were not acquired together on the same day. The advantage of the IR technique is the low dispersion of data giving direct access to the gas transfer curves. Gas transfer models in porous media can be numerically resolved taking into account the experimental parameters. The curves resulting from this modelisation can then be adjusted with the experimental gas transfer curves in order to quantify the respective contributions of the diffusion, advection and/or adsorption involved in gas transfer in the porous media. Such curves can be used to determine the gas concentration in the source-rock responsible for gas migration towards the borehole. It can also be used to discuss the reliability of some geological data such as effective diffusion coefficient and petro-physical parameters (porosity, permeability, adsorption surfaces).

Fig. 7 Evolution of CO₂ (a) and CH₄ (b) gas concentration between day 230 and day 310 of the PC-C experiment in the Mont Terri laboratory obtained by gas chromatography on gas samplings (circles) and by IR device monitoring (dots).

Table 2 Gas chromatography results of CO₂ and CH₄ between 230^th and 310^th day of PC-C experiment in the Mont Terri laboratory

Sampling days/d	CO2 (ppmv)	CH4 (ppmv)
231	1800 ± 75	125 ± 3
253	3900 ± 162	1410 ± 33
273	3400 ± 141	2700 ± 64
286	—	2800 ± 66
294	4300 ± 178	3400 ± 81
296	4400 ± 183	3800 ± 90
299	4400 ± 183	3800 ± 90
301	4000 ± 166	4900 ± 116

---

The high sensitivity of this method allows us to detect slight variations in the partial pressure of gas. Fig. 8 represents a period of two days and a few hours of the PAC1002 experiment carried out in the Bure URL.¹¹ This period occurred 573 days after drilling of the borehole and after argon refilling of the gas circulation circuit and the gas cell, while the borehole was isolated from the rest of the circuit. The gas mixture inside the borehole contained a given quantity of CO₂, CH₄ and other gases (mainly argon) resulting from the 573 days of gas equilibration with the rock formation (marl). The pressure in the gas circuit was slightly higher than in the borehole. When we reconnected the borehole with the gas circulation circuit filled with argon, we observed very specific behaviours for CO₂ and CH₄. The evolution of CH₄ concentration during the days after the connection initially followed a sine cardinal (aka sampling function (sinc)) profile, and then stabilised. Whereas, the concentration of CO₂ reached a maximum value after 2 hours, then decreased slightly to tend towards stabilization after 30 hours. The same behaviours could be observed for both gases after starting and stopping the gas circulation pump, but with lesser amplitudes.

Fig. 8 Evolution of CO₂ (green circles) and CH₄ (orange circles) concentrations between the day 573 and the day 576 of the PAC1002 experiment in the Bure laboratory monitored by IR device after a dilution of the borehole gas mixture by addition of argon.

Concerning the behaviour of CH₄, the duration of the sine cardinal period was 10 hours, which corresponds to the time necessary for the circulation pump to pump out the whole volume of the gas circuit (about 11.14 L at 20 mL min⁻¹). The transfer mechanism of CH₄ in clayey rock, probably diffusion–advection, is not rapid enough to compensate for the CH₄ decrease induced by argon input. The behaviour of CH₄ concentration seen by the IR device could thus be the result of three phenomena acting in concert: gas flow induced by the pump, diffusion of the mixture of the two gases (gas in the borehole and argon in the circulation circuit) and pressure re-equilibrium between the borehole and the circulation circuit.

Concerning the behaviour of CO₂, its concentration reaches a maximum and decreases slightly to tend towards stabilization. The fast increase at the beginning of the cycle could have the same origin as for CH₄: the action of the pump coupled to pressure re-equilibrium allows the CO₂ present in the borehole to reach the infrared cell. In the borehole, the arrival of argon (i.e. decrease of CO₂ concentration) is quickly compensated by CO₂ production from the rock. The equilibrium of the gas mixture in the borehole with fluid in the rock system is governed here by the dissolution of the carbonate minerals within the porewater, with regard to the pH.

Thus, the difference of behaviour between CO₂ and CH₄ is attributed to the difference of chemical activity of the two components. The behaviour of CH₄ is representative of a purely mechanical mixing of two gases in a circuit while the concentration in CO₂ is influenced mainly by the chemical reaction between CO₂, porewater and the rock. These details of the evolution of the two gases are only accessible with continuous, online, and non-destructive measurement.

Conclusions

A new FTIR procedure for on-line monitoring of molecular gases (mainly CO₂ and CH₄) released by low-permeability geological media has been presented. In an underground laboratory context, gas monitoring by FT-IR is specific. In contrast to conventional surface laboratory equipment, the FT-IR spectrometer must work without any atmospheric purge. Such conditions are highly problematic for CO₂ monitoring, because this gas is present in the rock formation (e.g. the atmosphere of the URL) at high concentrations. To deal with such specific conditions, an original procedure based on FT-IR analysis using a spectrometer with two compartments sharing the same source and interferometer has been developed. The first internal compartment is connected to the borehole through a variable optical path gas cell connected to the gas circuit. The external compartment is open to the atmosphere. In the case of the internal compartment, the bulk signal recorded by the IR system comes from two contributions: atmospheric and borehole CO₂. Deconvolution allows extraction of the borehole CO₂ concentration expressed in terms of absolute partial pressure or concentration. The results of this procedure are limited by the accuracy of calibration and the quantification limit of the device: explored partial pressure ranges are between 0.3 and 4 mbar m for CO₂ and 0.3 and 12 mbar m for CH₄, and relative RMSEP of the models is included between 1 and 3% in the explored pressure ranges for each gas. Quantification limits can be lowered by using gas cells with higher path lengths, which increases the signal/background ratio.

Automatic spectra recording, coupled with an appropriate calculation procedure, permits on-line measurements of partial pressures of CO₂ and associated gases such as CH₄ over short to long periods of time. The periodicity of measurement can be adapted or modified by the operator as needed. Quantitative calculations take into account bulk pressure variations in the gas circuit because the experiment is not carried out in isobaric conditions.³⁶ Comparison of on-line measurements acquired by the IR device and measurements acquired by gas sampling coupled to GC analysis favour the FT-IR technique. It is not affected by gas manipulation and allows the determination of accurate physical laws to describe on-site gas transfer from porous geological media. Gas content determination is a very useful tool to complete porewater geochemistry as well as the measurement of accurate pH deduced from CO₂ partial pressure at equilibrium with water. Such data are of great interest for the description of rock formations acting as host rocks for nuclear waste disposal. It is also useful to monitor CO₂ or CH₄ geological storage: survey during the injection or post-injection period requires comparison with the baseline acquired before injection.⁴¹ The development of the FT-IR analysis of other natural gases (alkanes, H₂O) by statistical spectral analyses (classical least square method)³³ and new advances in ¹³C isotopes determination of CO₂ and CH₄ offer new challenges for the immediate future. This novel approach used to quantify gas emissions in a rock formation over time is an important contribution to the knowledge of gas transfer in the lithosphere. It is also a comprehensive way to monitor gas exchanges between the geosphere and the atmosphere.

Nomenclature

A	intensity of the CO2ν3 band integrated between 2220 and 2400 cm^−1
A α, Aα′, Aβ, Aαβ	intensity of the CO2ν3 band integrated between 2220 and 2400 cm^−1 considering the α, α′, β or αβ gas cell
f	polynomial calibration function
f P	polynomial calibration function for a given constant pressure P
k 1, k2, k3	polynomial coefficients of function f
k 1(P), k2(P), k3(P)	polynomial coefficients of the function fP
l	optical path length
l α, lα′, lβ, lαβ	optical path length in the α, α′, β or αβ cell
M m	molar mass of the considered gas
P α, Pα′, Pβ, Pαβ	bulk pressure of the sample in the α, α′, β or αβ cell
P atm	atmospheric pressure around the system
P bulk	bulk pressure of the considered sample
P cell	bulk pressure in the cell
p CO 2 α, pCO2α′, pCO2β	partial pressure of CO2 in the α, α′, or β cell
R	Rydberg constant
T	temperature of the gas sample (K)
u α, uβ, uαβ	mass absorbing gas per unit area, nomenclature derived from Plass' nomenclature, considering the α, β or αβ cell
κ	optical path ratio crossing the atmosphere between the two compartments of the IR system
α, β, αβ, α′	indices referring to the virtual gas cell

Plass' Nomenclature³⁸

h	index referring to a single-layer homogeneous system
j	index referring to a single cell in series
u h	mass absorbing gas per unit area for all the jcells in series
u j	mass absorbing gas per unit area for the jcell
r j	density of absorbing gas for the jcell
l j	optical path length of j^th cell in series

Acknowledgements

This work was financially supported by the French National Agency for Radioactive Waste Management (ANDRA) to whom we express our gratitude. Experiments in the surface laboratory were performed at Nancy University (LEM laboratory). We gratefully acknowledge Gordon Filby and Anne de Henning for improving the English.

References

1. J. M. Hunt, Petroleum Geochemistry and Geology, W.H. Freeman & Co, New York, 2nd edn, 1996.
2. A. Prinzhofer and A. Battani, Oil Gas Sci. Technol., 2003, 58, 299.
3. J.-P. Girard, C. Fléhoc and E. Gaucher, Selected Papers from the 5th International Symposium on Applied Isotope Geochemistry, Heron Island, Great Barrier Reef, Australia, 26–30 May 2003, 2005, 20, p. 713.
4. Andra, Dossier 2005 Argile: Evaluation of the Feasibility of a Geological Repository in an Argillaceous Formation, Dossier 2005 Argile, 2005.
5. M. Thury, C. R. Phys., 2002, 3, 923–933.
6. É. C. Gaucher, P. Blanc, F. Bardot, G. Braibant, S. Buschaert, C. Crouzet, A. Gautier, J.-P. Girard, E. Jacquot, A. Lassin, G. Negrel, C. Tournassat, A. Vinsot and S. Altmann, C. R. Geosci., 2006, 338, 917–930.
7. C. Degueldre, A. Scholtis, A. Laube, M. Turrero and B. Thomas, Appl. Geochem., 2003, 18, 55.
8. M. H. Bradbury and B. Baeyens, Geochim. Cosmochim. Acta, 1998, 62, 783–795.
9. P. Leroy, A. Revil, S. Altmann and C. Tournassat, Geochim. Cosmochim. Acta, 2007, 71, 1087–1097.
10. A. Vinsot, C. A. J. Appelo, C. Cailteau, S. Wechner, J. Pironon, P. De Donato, P. De Cannière, S. Mettler, P. Wersin and H. E. Gäbler, Phys. Chem. Earth, 2008, 33, S54–S60.
11. A. Vinsot, S. Mettler and S. Wechner, Phys. Chem. Earth, 2008, 33, S75–S86.
12. Solexperts, Rapport d'installation—Instrumentation hydrogéologique après creusement de la niche, rapport ANDRA D.RP.SOP.05.0007, 2005.
13. P. Marschall, S. Horseman and T. Gimmi, Oil Gas Sci. Technol., 2005, 60, 121–139.
14. W. D. Komhyr, T. B. Harris, L. S. Waterman, J. F. S. Chin and K. W. Thoning, J. Geophys. Res., 1989, 94, 8533–8547.
15. M. B. Esler, D. W. T. Griffith, S. R. Wilson and L. P. Steele, Anal. Chem., 2000, 72, 206–215.
16. D. Helmig, J. Chromatogr., A, 1999, 843, 129.
17. L. P. Steele, P. J. Fraser, R. A. Rasmussen, M. A. K. Khalil, T. J. Conway, A. J. Crawford, R. H. Gammon, K. A. Masarie and K. W. Thoning, J. Atmos. Chem., 1987, 5, 125.
18. R. Prinn, D. Cunnold, R. Rasmussen, P. Simmonds, F. Alyea, A. Crawford, P. Fraser and R. Rosen, J. Geophys. Res., 1990, 95, 18369–18385.
19. P. C. Novelli, J. W. Elkins and L. P. Steele, J. Geophys. Res., 1991, 96, 13109–13121.
20. M. Pushkarsky, M. Weida, T. Day, D. Arnone, R. Pritchett, D. Caffey and S. Crivello, Proc. SPIE-Int. Soc. Opt. Eng., 2008, 6871, 68711X.
21. S. Barbieri, J. P. Pellaux, E. Studemann and D. Rosset, Rev. Sci. Instrum., 2002, 73, 2458.
22. M. Taslakov, V. Simeonov and H. Van Den Bergh, Spectrochim. Acta, Part A, 2006, 63, 1002–1008.
23. E. Normand, I. Howieson, M. McCulloch and P. Black, Proc. SPIE-Int. Soc. Opt. Eng., 2006, 6402, 64020G.
24. Z. A. Allah, D. Sawtell, V. L. Kasyutich and P. A. Martin, J. Phys.: Conf. Ser., 2009, 157, 012001.
25. M. W. Sigrist, Rev. Sci. Instrum., 2003, 74, 486–490.
26. D. W. T. Griffith, R. Leuning, O. T. Denmead and I. M. Jamie, Atmos. Environ., 2002, 36, 1833.
27. M. Grutter, Atmosfera, 2003, 16, 1.
28. W. M. Doyle, Process Control Qual., 1992, 2, 11.
29. C. P. Rinsland, A. Goldman, F. J. Murcray, T. M. Stephen, N. S. Pougatchev, J. Fishman, S. J. David, R. D. Blatherwick, P. C. Novelli, N. B. Jones and B. J. Connor, J. Geophys. Res., 1999, 104, 18667.
30. D. W. T. Griffith, Appl. Spectrosc., 1996, 50, 59–70.
31. P. L. Hanst and S. T. Hanst, in Air Monitoring by Spectroscopic Techniques, ed. M. W. Sigrist, John Wiley & Sons Inc., New York, 1994, vol. 127, pp. 335–470.
32. J. U. White, J. Opt. Soc. Am., 1942, 32, 285.
33. A. Hakuli, A. Kytokivi, E. L. Lakomaa and O. Krause, Anal. Chem., 1995, 67, 1881–1886.
34. E. Lopez-Anreus, S. Garrigues and M. De La Guardia, Anal. Chim. Acta, 1995, 308, 28.
35. D. W. T. Griffith and B. Galle, Atmos. Environ., 2000, 34, 1087.
36. C. Cailteau, J. Pironon, P. de Donato, A. Vinsot, C. Garnier and O. Barrès, Anal. Methods, 2011 10.1039/C0AY00622J.
37. J. Bak and A. Larsen, Appl. Spectrosc., 1995, 49, 437–443.
38. G. N. Plass, Appl. Opt., 1965, 4, 69–78.
39. J. A. Jamieson, R. H. McFee and G. N. Plass, Infrared Physics and Engineering, McGraw Hill Publishing Company, New York, 1963.
40. C. Cailteau, PhD Thesis, Nancy Université, 2008.
41. J. Pironon, P. de Donato, O. Barrès, C. Garnier, C. Cailteau, A. Vinsot and G. Radilla, Int. J. Greenhouse Gas Control, 2010, 4, 217–224.

---