Towards traceability in CO2 line strength measurements by TDLAS at 2.7 µm

https://doi.org/10.1016/j.jqsrt.2013.07.011Get rights and content

Highlights

  • Line strengths of the P36e and P34e lines in the 2.7 µm ν13 CO2 band were measured.

  • GUM-compliant uncertainties of 1–1.3% have been determined.

  • Line strengths uncertainties reduced by a factor of two compared to literature.

  • Traceability of the results has been addressed.

Abstract

Direct tunable diode laser absorption spectroscopy (TDLAS) was combined in this study with metrological principles on the determination of uncertainties to measure the line strengths of the P36e and P34e line of 12C16O2 in the ν13 band at 2.7 µm. Special emphasis was put on traceability and a concise, well-documented uncertainty assessment. We have quantitatively analyzed the uncertainty contributions of different experimental parameters to the uncertainty of the line strength. Establishment of the wavenumber axis and the gas handling procedure proved to be the two major contributors to the final uncertainty. The obtained line strengths at 296 K are 1.593×10–20 cm/molecule for the P36e and 1.981×10–20 cm/molecule for the P34e line, with relative expanded uncertainties of 1.1% and 1.3%, respectively (k=2, corresponding to a 95% confidence level). The measured line strength values are in agreement with literature data (line strengths listed in the HITRAN and GEISA databases), but show an uncertainty, which is at least a factor of 2 lower.

Introduction

Carbon dioxide (CO2) plays an important role in biological and combustion processes, and due to its strong opacity in the infrared, it has a major role in radiation transfer in planetary atmospheres. With an average amount fraction of 390 µmol/mol carbon dioxide is the third most abundant trace gas, as well as the most important anthropogenic greenhouse gas in the Earth's atmosphere.

Carbon dioxide amount fractions in the gas phase are mainly measured by spectroscopic techniques. The standard technique used in environmental monitoring is absorption spectroscopy based on a broadband infrared light source (also called the non-dispersive infrared (NDIR) method) [1]. For specific applications, including measurements of biosphere–atmosphere exchange fluxes of CO2, determining the ratio of abundances of different CO2 isotopes, or measuring at unusual locations, e.g. in the upper troposphere of Earth or in the atmosphere of other planets, several other spectroscopic techniques are available. Current state-of-the-art measurement techniques include cavity ring-down [2] and cavity enhanced spectroscopy [3], photoacoustic spectroscopy [4], Fourier-transform infrared spectroscopy [5] and tunable diode laser absorption spectroscopy (TDLAS) [6], [7].

Instruments based on the methods listed above are usually calibrated with gas mixtures before use. Typical commercial calibration gas standards usually have a relative uncertainty of 1%, while agreement between primary mixtures is 0.3% [8], which limits the uncertainty of amount fraction measurements. However, some of these spectroscopic techniques are also suitable for absolute measurements, i.e. calculating the amount fraction of CO2 directly from the measured spectrum, without the need for calibration with gas standards. In this case the above mentioned methods are used as an absolute method; they use spectroscopic parameters of the probed absorption lines and determine amount fractions using the Beer–Lambert law. Provided that highly accurate spectral data are available (with uncertainty smaller than that of calibration gas standards in the required amount fraction range), absolute measurements might be able to provide lower uncertainties than methods based on calibration using gas standards. Moreover, for remote sensing, where no direct calibration is possible, accurate knowledge of line parameters is inevitable. One example is the Orbiting Carbon Observatory (OCO), which aims to perform global satellite-based CO2 measurements with high precision, spatial resolution and coverage. The aimed precision of the measurements is <1 µmol/mol, i.e. 0.3% [9], which requires also spectral line parameters with sub-percent uncertainty. Global radiation transfer and climate models also rely on line parameters; inaccuracies in the used spectroscopic data might lead to incorrect model outputs.

Spectral line parameters from literature are collected in spectroscopic databases. The two most extensive databases for atmospheric applications are the HITRAN and the GEISA databases [10], [11]. These databases have been developed for several decades (the first edition of HITRAN was published in 1973, while the development of the GEISA database started in 1976), and contain valuable information on line data of a considerably high number of molecules in a wide spectral range. CO2 has been an important molecule in both databases from the beginning. However, line parameters in these spectroscopic databases originate from various sources: measurements performed by different methods in different research groups, as well as calculations using different models. Lack of homogeneity between these sources regarding how they report their data makes data comparability and in particular standardized uncertainty assessment difficult. A concise description on how uncertainties of the measured or calculated data have been estimated is often not given by the authors; therefore a uniform uncertainty assessment cannot be applied for the whole database [12]. As a result, line data in these databases are often presented with a relatively high uncertainty [13]. Most databases give uncertainty classes instead of exact numbers, and in some cases even data without uncertainty figures.

The metrological term traceability refers to the “property of a measurement result whereby the result can be related to a reference through a documented unbroken chain of calibrations, each contributing to the measurement uncertainty” [14]. The general rules of evaluating and expressing uncertainty are summarized in an ISO standard (Evaluation of Measurement Data – Guide to the Expression of Uncertainty in Measurement (GUM) [15], [16]). This means that following the GUM would ensure reliability and comparability of the measurement results, based on transparently and reproducibly derived measurement uncertainties.

The GUM differentiates between standard and expanded uncertainty. Standard uncertainty (u) is the “uncertainty of the result of a measurement expressed as a standard deviation”, while expanded uncertainty (U) is a “quantity defining an interval about the result of the measurement that may be expected to encompass a large fraction of the distribution of the values that could reasonably be attributed to the measurand”. Expanded uncertainty is calculated from standard uncertainty by multiplying it with the so-called coverage factor (k), typically in the range of 2–3, depending on the type of distribution and the confidence interval. In this paper we give expanded uncertainties, which refer to a confidence interval of 95%, and k=2.

Difficulties in achieving traceability arise from the fact that even in the simplest experiment several factors influence the measurement result and the experimentalist has to be quite prudent and rigorous to include all significant sources of uncertainty. A common mistake is the underestimation of measurement uncertainties by neglecting significant sources of uncertainty. Several publications report highly accurate line strength measurements and give impressive numbers for the measurement uncertainty, but a clear and transparent description of the uncertainty assessment is not given, and results obtained by different research groups often do not agree with each other, simply because uncertainties have been underestimated or neglected.

For example, Toth et al. [17] report line strengths measured by FTIR in the 4550–7000 cm−1 range with an accuracy of 1% (we interpret this number as an expanded uncertainty). Results of these measurements are listed in the HITRAN 2008 database for a number of lines in this wavenumber range, where an uncertainty range of 1–2% is given. Casa et al. [18], [19] report CO2 line strengths of the R(2)–R(18) lines in the ν1+2ν23 band between 4980 and 4991 cm−1 obtained by TDLAS measurements using an intensity stabilized diode laser. They report 1σ accuracies of 0.1–0.15% (which we interpret as standard uncertainty and apply a coverage factor of k=2 to calculate an expanded uncertainty). From our research group, line strength studies on the R(12) line of CO2 at 4987.3 cm−1 were reported [20], [21]. Referring to the GUM a relative expanded uncertainty of 1% was estimated [20]. To analyze agreement between results of these three studies, we took uncertainties as described above, and calculated the normalized error defined asEn=|(S1S2)/U12+U22|.Therein S1 and S2 are line strengths reported on two different measurements, with the corresponding expanded uncertainties U1 and U2.

Measurement results for the R(12) line, which is included in all the above-mentioned studies are given in Table 1 together with normalized errors. Normalized errors were found to be close to or even exceeding 1, which indicates that measurement uncertainties in some of these studies are probably underestimated.

The reason for unclear uncertainty figures in databases or discrepancies between different measurements might stem from the fact that most authors aim to determine a wide range of spectral line parameters. For such large data sets accurate determination of uncertainties for each measured quantity would be too laborious. However, if one concentrates on single absorption lines and single line parameters, a more comprehensive and accurate uncertainty assessment can be given. In this work we measure line strengths of only two lines with the aim to give a full uncertainty budget including every input quantity that contributes to the final line strength uncertainty. This also allows identifying the most important sources of uncertainty, thus pointing out where the experiments should be improved to reach lower uncertainty in the line strength. Measured line strengths of individual lines with reliable uncertainties can later be used as anchor points for the validation of larger data sets containing a high number of lines in a broad spectral range, which is typically measured by FTIR spectroscopy.

Our present work was carried out within the framework of the European Metrology Research Programme [22] in a project on metrology for chemical pollutants in air [23]. Our task in the project has been the certification of zero gas standards by means of laser spectroscopic measurements. For measuring CO2 impurities in zero gas samples we have chosen two absorption lines from the strongest NIR CO2 band (ν13): the P36e line at 2.7153 µm, which is a spectrally well isolated line; and the P34e line at 2.7139 µm, which is close to a relatively strong water line at 2.7141 µm. This allows simultaneous detection of CO2 and H2O in zero gas samples by recording absorption lines of both analytes on a single wavelength scan. The project aims at absolute measurement of amount fractions of zero gas impurities, thereby avoiding the use of calibration gas standards in the measurement process. For this, accurate spectral line parameters, most importantly line strengths are inevitable. To achieve traceability in the measured amount fractions [24], the used line strength values need to be traceable as well.

We have measured line strength of two CO2 lines at 2.71 µm by TDLAS aiming at traceability of the results. Although these lines belong to the strongest near infrared (NIR) CO2 band, only very few studies have been reported so far for this spectral range. Line strengths listed in spectroscopic databases are based on a few FTIR measurements or calculations [25]. This limited interest in the past was most probably due to the lack of available light sources in this wavelength range, which is now expected to change due to new developments of commercial DFB diode lasers [26].

Our previous measurements [27], [20] have proven that, due to its high spectral resolution, TDLAS is suitable for accurate line data measurements. In this work we put special emphasis on achieving traceability, identifying sources of uncertainty and giving a clear and concise explanation of our uncertainty assessment.

Section snippets

The studied spectral lines

Fig. 1 shows the absorption lines of CO2 listed in the HITRAN 2008 database [10] in the studied wavelength range. The selected absorption lines are the P36e and the P34e line of the ν13 band of CO2, at 3682.77 and 3684.75 cm−1, respectively, with line strengths in the range of 10–20 cm/molecule.

Experimental set-up

The set-up used in our experiments is depicted in Fig. 2.

We used a DFB diode laser (Nanoplus [28], DL in Fig. 2) as a light source, with a central wavelength of 2716 nm, maximum optical power of 2 mW,

Input parameters

As a first step, we discuss the individual input quantities, from the point of view of traceability, and estimate expanded uncertainties. All expanded uncertainties are given for a coverage factor of k=2.

Both pressure sensors used are traceable to the German national pressure standard. Calibration against the national pressure standard revealed an expanded uncertainty of p better than 0.2% for the pressure range of 10 to 50 Pa. In addition, long-term stability of the sensors and temperature

Conclusions and outlook

Employing tunable diode laser absorption spectroscopy based on a DFB diode laser with an emission wavelength around 2.7 µm, we have measured the intensity of the P36e and P34e lines of 12C16O2 in the ν13 band. Measurements were performed in a single pass gas cell on pure CO2 gas at room temperature, at pressures below 50 Pa, where pressure broadening of the absorption lines is significantly decreased, i.e. two orders of magnitude lower than the Doppler line width.

We have put special emphasis on

Acknowledgment

The authors acknowledge financial support and collaboration in the EMRP projects MACPoll and EUMETRISPEC (EMRP-ENV01 and EMRP-ENV06). The EMRP is jointly funded by the EMRP participating countries within EURAMET and the European Union. The authors are grateful to Steven Wagner for his support regarding the measurement software. The authors appreciate support from Steffen Scheppner and Karl Jouston regarding correctness of the pressure measurement and from Lars Nielsen regarding generalized

References (44)

  • H. Chen et al.

    High-accuracy continuous airborne measurements of greenhouse gases (CO2 and CH4) using the cavity ring-down spectroscopy (CRDS) technique

    Atmospheric Measurement Techniques

    (2010)
  • Baer DS, Paul JB, Gupta M, O'Keefe A. Sensitive absorption measurements in the near-infrared region using off-axis...
  • R. Lewicki et al.

    Carbon dioxide and ammonia detection using 2 μm diode laser based quartz-enhanced photoacoustic spectroscopy

    Applied Physics B

    (2007)
  • R. Haus et al.

    Mobile Fourier-transform infrared spectroscopy monitoring of air pollution

    Applied Optics

    (1994)
  • A. Farooq et al.

    CO2 concentration and temperature sensor for combustion gases using diode-laser absorption near 2.7 μm

    Applied Physics B

    (2008)
  • R.M. Wessel et al.

    International comparison CCQM-K52: carbon dioxide in synthetic air

    Metrologia

    (2008)
  • Volker E, Werhahn O, Brunzendorf J, Serdyukov A, Rausch A, Ott O, et al. EUMETRISPEC: traceability of spectral line...
  • International Organization for Standardization

    International vocabulary of basic and general terms in metrology

    (1993)
  • International Organization for Standardization. Guide to the Expression of uncertainty in measurement. 1. ISO Guide...
  • International Organization for Standardization. Supplement 1 to the guide to the expression of uncertainty in...
  • G. Casa et al.

    Highly accurate determinations of CO2 line strengths using intensity-stabilized diode laser absorption spectrometry

    Journal of Chemical Physics

    (2007)
  • G. Casa et al.

    The line shape problem in the near-infrared spectrum of self-colliding CO2 molecules: experimental investigation and test of semiclassical models

    Journal of Chemical Physics

    (2009)
  • Cited by (54)

    • Measuring techniques for concentration and stable isotopologues of CO<inf>2</inf> in a terrestrial ecosystem: A review

      2019, Earth-Science Reviews
      Citation Excerpt :

      There have been extensive efforts by several research groups (Gordon et al., 2017; Tashkun et al., 2019; Huang et al., 2017; Jacquinet-Husson et al., 2016) to expand the characterization of CO2 isotopologues, from both experimental and theoretical standpoints, while using/comparing different spectroscopic databases (e.g. HITRAN, CDSD, AMES and GEISA). However, the LAS analysis uncertainty resulting from inconsistency hidden among these spectroscopic databases for CO2 isotopologues, is still a question (Pogány et al., 2013). This question is well justified, since the intensity of most CO2 IR transitions in spectroscopic databases have >1% uncertainty.

    View all citing articles on Scopus
    View full text