Spectrochimica Acta Part B 68 (2012) 65–70
Contents lists available at SciVerse ScienceDirect
Spectrochimica Acta Part B journal homepage: www.elsevier.com/locate/sab
Technical note
Laser induced breakdown spectroscopy: A potential tool for atmospheric carbon dioxide measurement Vivek Dikshit a, Fang-Yu Yueh a, Jagdish P. Singh a,⁎, Dustin L. McIntyre b, Jinesh C. Jain b, Nouredine Melikechi c a
Institute for Clean Energy Technology, Mississippi State University, 205 Research Boulevard, Starkville, MS, USA National Energy Technology Laboratory, United States Department of Energy, 3610 Collins Ferry Road, Morgantown, WV, USA Center for Research and Education in Optical Sciences and Applications (CREOSA), Department of Physics and Pre-Engineering, Delaware State University, 1200 North DuPont Highway, Dover, DE, USA
b c
a r t i c l e
i n f o
Article history: Received 14 June 2011 Accepted 16 January 2012 Available online 26 January 2012 Keywords: LIBS Sensor Carbon dioxide Detection Ambient air
a b s t r a c t Carbon dioxide (CO2) is a main contributor to global warming, making up approximately 80% [1] of the greenhouse gases in the atmosphere. Therefore, a precise measurement of the atmospheric CO2 concentration is essential. Although a number of analytical techniques are available for measuring CO2 in air samples, laser induced breakdown spectroscopy (LIBS) offers a relatively simple and straightforward analysis which is why it was utilized in this study. LIBS requires a simple experimental setup and offers real-time carbon dioxide measurement. The strong C(I) emission line at 247.85 nm was selected for CO2 measurement, which yielded a detection limit of 36 ppm with a pulse energy of 145 mJ. Real-time measurement has been demonstrated: a single measurement can be made in 40 s with a relative standard deviation (RSD) of 3.6%. © 2012 Elsevier B.V. All rights reserved.
1. Introduction Carbon dioxide (CO2) is the predominant anthropogenic greenhouse gas responsible for continuous rise in global surface temperature and ocean acidity: relevant impacts being coastal flooding, eradication of wet lands, diminished glacial water sources, and damage to ocean ecosystems. A number of governing bodies recognize these threats and are beginning to mitigate CO2 emissions; therefore, better measurement techniques are crucial for producing rapid and accurate results. Currently, CO2 is measured by various methods [2]. The two commonly used methods are GC–MS and infrared spectroscopy. The working principle of GC–MS is based on selective adsorption and elution of the gas molecules of interest in the GC column and the subsequent analysis by the mass-spectrometer. Within the sample gas is bombarded with a high-energy electron beam which causes causing fragmentation of the molecules in the sample. These fragmented molecules are accelerated by a magnetic field and separated by virtue of their different mass to charge ratios. Infrared (IR) absorption spectroscopy is an optical technique, which is the most commonly used technique for CO2 detection. Carbon dioxide has a non-zero dipole moment which causes it to absorb specific wavelengths of light in the infrared region. Thus, analysis of absorption spectral bands gives the CO2
⁎ Corresponding author. E-mail address:
[email protected] (J.P. Singh). 0584-8547/$ – see front matter © 2012 Elsevier B.V. All rights reserved. doi:10.1016/j.sab.2012.01.009
concentration in the sample. This technique has a distinct disadvantage in the form of interference from water vapor and carbon monoxide. In the present work we have applied laser induced breakdown spectroscopy (LIBS) for measurement of CO2 in air. LIBS is a promising detection tool for solid, liquid and gaseous samples [3–5]. A LIBS set up essentially comprises of a high intensity laser as an excitation source and a time-gated spectrometer to collect the signal. Generally, a laser beam is focused onto a solid sample or within a gaseous or liquid sample to create high temperatures, dissociating the sample to form plasma. Radiation from the plasma is then collected by the spectrometer. Subsequent analysis of the radiation gives qualitative and quantitative information of chemical species present in the sample. Various authors have used LIBS for detection of gaseous samples [6–11]. While Hahn et al. [6] have used LIBS for hydrogen leak detection, Winefordner et al. [7] and McNaghten [8] have utilized it for detection of gaseous/particulate fluoride and helium, and argon in binary and ternary gas mixtures with nitrogen, respectively. Recently, Eseller et al. [9] used the technique for monitoring Ar, He and oxygen impurities in hydrogen. Ferioli et al. [10] have applied LIBS to measure the equivalence ratio of a spark-ignited engine where they showed that the ratio of either C(711.3 nm) or CN(707–734 nm) peaks and any of the N(746.3 and 743.8 nm) or O(776.6 nm) spectral lines can be used to estimate the equivalent ratio. In fact, the LIBS calibration curve of gaseous samples shows a better linearity and reproducibility than solids and liquids. This may be attributed to weak interactions relevant to the matrix effect in low density gas sample at atmospheric pressure and to the homogeneity of the gaseous sample.
66
V. Dikshit et al. / Spectrochimica Acta Part B 68 (2012) 65–70
Fig. 1. Schematic of experimental set up for CO2 measurement.
The purpose of this study is to investigate whether LIBS can be used as an analytical technique to detect carbon dioxide in air. Our working hypothesis is carbon dioxide concentration in air can be measured by creating a spark in the air and subsequently measuring the carbon signal at C-247.85 nm; a line used by other researchers [12,13] for carbon measurement in soil. We assume that the main source of the carbon signal in air is from carbon dioxide, considering the approximate concentrations of CO2, methane, and carbon monoxide in air are 390 ppm, 1.79 ppm, and 0.1 ppm respectively (current values are available online at http://gaw.kishou.go.jp/cgi-bin/wdcgg/ catalogue.cgi). Though in highly polluted cities some interference might come from the soot (black carbon) present in the air. Soot concentration in the atmosphere is highly inhomogeneous. Typical soot concentration in big cities is 0.46 μg/m 3 [14]. Generally, soot concentration may be neglected. However for those cases where this contribution cannot be neglected, this interference can be removed through the use of a commercial particulate filter (Diesel Particulate Filter has a removal efficiency of 85% to 100%). Alternatively, soot concentration can be measured separately and its contribution can be subtracted from the LIBS signal to get the contribution due to CO2 only. To our knowledge, the research reported here constitutes the first study utilizing LIBS for atmospheric carbon dioxide measurement. 2. Experimental A schematic of the experimental set up is shown in Fig. 1. A twoway cylindrical cell was used to hold the gaseous sample during the measurement. The cell was fitted with two quartz windows at both ends to allow the laser beam to enter and exit the cell. The sample cell contained ancillaries to allow inflow and venting of gases, and measurement of the cell pressure. A rotary vane pump (Edwards E2M2) and a Baratron absolute pressure gauge (MKS 122A) were connected to the cell. The Nd:YAG laser (Big Sky Inc. CFR400) with a wavelength of 532 nm was used in this investigation. It was operated in Q-switched mode with a pulse repetition rate of 10 Hz. The pulse width (FWHM) and the maximum pulse energy were 8 ns and 180 mJ respectively. The beam diameter was 6.5 mm and had a Gaussian beam profile. A spherical plano-convex fused silica lens of focal length 10 cm was used for focusing the laser beam into the center of the sample cell to create a plasma spark within the gas mixture. The emission from the laser-induced plasma was focused into a fiber optic cable by a fused silica lens of focal length 10 cm. The other end of the optical fiber was coupled to a UV–visible Echelle optical spectrograph (LLA Instruments, GmbH, ESA 3000 EV/I, Berlin, Germany). The spectrograph had the spectral range of 200–780 nm (though it had some spectral gaps). It had a linear dispersion of approximately
5–19 pm/pixel. A 1024 × 1024 element intensified charge-coupled device (Kodak KAF-1001) with a pixel width of 24 μm was cooled by Peltier elements and attached to the exit of the spectrograph which was used to detect the light from the laser spark. The detector was operated in gated mode using a dedicated high-voltage fast-pulse generator (Stanford DG 535) that was synchronized with the laser pulse. The data acquisition and analysis were performed on a personal computer using ESAWIN software. Two certified gas mixtures supplied by Nexair were used in the experiments. These comprised a cylinder of compressed air and another of carbon dioxide (5% CO2 balanced with air). The certified gas mixtures were used to prepare the gas sample for the system calibration. 3. Results and discussion The initial efforts were concentrated on investigating the effect of experimental parameters (laser energy, gate delay, and choice of focusing lens) on the intensity of the carbon 247.85 nm line (Fig. 2). In addition, the use of a signal enhancement device that employed a metal substrate was also evaluated. The objective of the metal substrate was to improve signal-to-noise ratio of the analyte carbon line for better CO2 detection. Finally, the optimized experimental conditions were used to record LIBS spectra for instrument calibration and CO2 concentration measurement of ambient air. 3.1. Optimization of experimental parameter Optimization using the available instruments and experimental setup is necessary to get the best signal. The most persistent carbon line that could be read with the spectrometer (LLA ESA 3000) was found to be C-247.85 nm. Intensity of the selected line was read directly from accompanying software (ESAWIN), which calculated the intensity by subtracting the background from the peak maximum. First, the effect of focusing lenses on the signal strength was investigated. The laser beam was focused with a 10 cm focal length lens and a spark was created in air. On-chip accumulation was found to be 40 to get the best signal-to-noise ratio without saturation. 1 The signal intensity was recorded with different pulse energies and gate delays. The whole procedure was repeated with a 20 cm focusing lens. The use of a 10 cm lens gave a stronger signal than the use of the 20 cm lens (Figs. 3 and 4). The plasma spark size observed in the experiment was smaller in the case of the 10 cm lens. The tight focusing of the laser beam in a small volume by a 10 cm lens had 1 On-chip accumulation refers to adding of multiple exposures right on the interline ICCD before a single readout.
V. Dikshit et al. / Spectrochimica Acta Part B 68 (2012) 65–70
Fig. 2. Spectral line of carbon at 247.85 nm at various CO2 concentrations in the cell. Smooth lines connecting the points are to guide the eye.
resulted in greater irradiance and thus a better signal. The maximum signal-to-noise ratio (S/N) was found to be approximately 60 using a 10 cm lens and 27 using a 20 cm lens. Fig. 4 shows that maximum signal strength can be obtained by setting the gate delay to 500 ns. Next, to study the nature of the plasma in air and to select the appropriate pulse energy, the intensity ratio of the C 247.85 nm line with various nitrogen lines against the pulse energy were plotted (Fig. 5). The ratio plots show similar trend using three different nitrogen lines; this suggests the uniformity of plasma formed. The intensity ratio increases with increase in laser pulse energy. When the laser energy reaches above ~145 mJ, the ratio becomes constant. This serves as a secondary indication of saturation both for carbon and nitrogen lines. Based on this study, laser pulse energy of 145 mJ was selected for the instrument calibration. 3.2. Signal enhancement with metal substrate It has been demonstrated [6,15] that LIBS signal can be enhanced by placing a metal plate close to the plasma spark. This technique was evaluated for CO2 detection in order to improve the measurement sensitivity. For this part of the experiment, the sample cell shown in Fig. 1 was replaced by a platform to mount a metal substrate. LIBS measurements were taken by slowly translating the
Fig. 3. The scatter plot shows the effect of using focusing lenses of different focal lengths where discrete values of carbon spectral line intensity (C-247.85 nm) at different laser pulse energies are shown. Gate delay and gate width were fixed at 0.5 μs and 5 μs respectively. Each data point is an average of 10 measurements; on-chip accumulation was 40.
67
Fig. 4. Variation of C247.85 nm line intensity with gate-delay using two different focusing lenses (focal 10 cm and 20 cm) is shown. The laser pulse energy and the gate width were respectively fixed at 145 mJ and 5 μs. Each data point is an average of 10 measurements and on-chip accumulation was 40. A linear least square fit is shown for data.
substrate with a stepping motor to ensure that the laser beam fell on different spots during the measurement. The data was recorded using copper and steel substrates. For comparison, data was also recorded by simply creating spark in air without the use of metal substrate. In each case, laser pulse energy was varied and the signal was recorded. At low pulse energies, the signal intensity from the spark with metal substrate was comparable to the signal intensity without metal substrate (Fig. 6). At higher pulse energies, the signal with metal substrate was more intense than the signal without metal substrate. However, there are disadvantages associated with the use of a metal substrate as a signal enhancement device: (i) the formation of a metal oxide layer and minute pitting, which necessitates replacing the substrate, and (ii) increased fluctuation in data, which necessitates averaging over a larger number of spectra. Table 1 quantitatively compares the signal fluctuation using metal substrates and without using any metal substrate, %RSD being the criterion. From the table we notice that for each laser pulse energy fluctuation is lowest when the substrate was not used. Thus, it can be concluded that a better precision can be achieved by making measurements without the use of metal substrate. Various parameters set for the experiment have been given in the table.
3.3. Calibration The schematic of the experimental set up used for calibration is shown in Fig. 1. The pressure inside the sample cell was set at atmospheric pressure. The laser spark was created in the sample cell without using any metal substrate. The experimental parameters used for the instrument calibration are as follows: laser pulse energy of 145 mJ, gate delay of 0.5 μs, and gate width of 5 μs. ICCD on-chip accumulation was set at 40, therefore 40 shots were needed for a single measurement, which took 8 s. Ten such measurements (time taken: 8 × 10 s) were averaged to remove shot-to-shot variation. Carbon 247.85 nm line was selected as the analyte line for CO2 measurement. The observed analyte line at three different CO2 concentrations is shown in Fig. 2. The desired gas samples for the instrument calibration were prepared by diluting the 5% (50,000 ppm) carbon dioxide mixture with air. The calibration curve for the entire concentration region was found to be non-linear, but when the whole concentration region was divided into three sub-regions, the calibration curve showed linearity in each separate region (Fig. 7). A high R 2 value ensures that CO2 concentration is directly proportional to the intensity of C-247.85 nm spectral line. The concentration can be estimated from the intensity if the slope and intercept of the calibration line is known. The non-zero y-intercept in Figs. 7(c) and 8(a) might be
68
V. Dikshit et al. / Spectrochimica Acta Part B 68 (2012) 65–70
Fig. 5. Variation of intensity ratio of different spectral lines with laser pulse energy is shown. The gate delay and gate width were set at 0.5 μs and 5 μs. Each data point is an average of 10 measurements and on-chip accumulation was 40.
Fig. 6. Variation of C 247.85 nm line intensity with and without metal substrates at different laser pulse energy is shown. The figure clearly shows that the use of metal substrate gives a better signal at high laser pulse energy, but at low energy their responses are comparable. The gate delay and gate width were set at 0.5 μs and 5 μs. Each data point is an average of 10 measurements and on-chip accumulation was 40. Smooth lines connecting the points are to guide the eye.
due to deviation in shape of the calibration curve from linear in the very low concentration region. The limit of detection (LOD) of carbon dioxide was evaluated using the standard method: LOD ¼
3σ m
where σ is the standard deviation of a blank measurement and m is the slope of the calibration line. In this experiment mean σ was 10 and slope was 0.83 (from Fig. 7c). Using these numbers LOD was estimated to be ~36 ppm. To ensure the repeatability of the measurement, the calibration data was collected on five different days and calibration curves were made. We found that these calibration curves overlap well. To present this fact we divided these data sets into two groups: training
data and test data. The training data set was comprised of data from the first four days and the test dataset was comprised of data obtained on the fifth day. The training data set was used to plot the calibration curve shown in Fig. 8a, which was further used to predict the concentrations of the test data set. ‘Actual’ concentrations from the test data were plotted against ‘predicted’ concentrations (Fig. 8b). The data presented in the both figures show measurements are repeatable. The precision and rapid real-time detection was demonstrated by CO2 measurement of the laboratory ambient air. First, a calibration curve was made, then the sample cell was filled with the ambient air and measurements were made continuously for 30 min. Fig. 9 shows CO2 concentration measurements obtained at various time instants. Each point in Fig. 9 is an average of 5 measurements. RSD found was 3.6%. The average CO2 concentration over a period of 30 min was 298.82 ppm.
Table 1 Relative standard deviation (RSD) comparison of the C(247.85) spectral line intensity with the use of steel and copper metal substrate and without the use of metal substrate is shown. Energy (mJ)
%RSD with steel substrate, gate delay=1500 ns, on-chip accumulations=2
%RSD with copper substrate, gate delay=1500 ns, on-chip accumulations=1
%RSD without metal substrate, Gate delay=500 ns, On-chip accumulations=40
51.0 79.0 110.5 141.0 170.1 197.5
18.1 19.0 4.7 3.1 6.8 9.8
46.6 13.0 15.9 10.6 15.4 19.0
10.9 5.4 3.2 2.5 2.1 1.3
V. Dikshit et al. / Spectrochimica Acta Part B 68 (2012) 65–70
69
Fig. 8. The experiment was repeated five times; first four datasets were used to make the calibration plot shown in panel a, and this calibration curve was used to predict concentrations of the 5th dataset. Actual concentrations and predicted concentrations are plotted in panel b.
Fig. 7. Calibration curves for CO2 detection in different concentration ranges are shown. Each data point represents an average of 10 measurements. The error bars represent plus or minus one full standard deviation. The experiment was performed with a gate delay of 0.5 μs and gate width of 5 μs with an on-chip accumulation of 40.
mixture with air. The calibration curves showed good linearity (R 2 ~ 0.99) and were reproducible when experiment was repeated on different days. The precision of the instrument is ensured by low RSD which was found to be comparable to commercially available IR CO2 sensors (RSD 1% to 3%). The LOD was estimated to be 36 ppm. Thus, this technique was investigated for rapid CO2 measurement in concentration range 36 ppm to 50,000 ppm. The accuracy and the LOD can be further improved by employing more accurate sample preparation methods for calibration and by modifying the experimental set up. Significant errors in the IR based method for atmospheric CO2 measurement are introduced during sample collection, while drying the samples, and also due to minute amounts of leakage within the system [16]. LIBS based method requires no sample collection. It uses a simple experimental setup which eliminates such errors. Therefore, this technique can be used as a complementary data collection technique with other techniques for atmospheric CO2 measurement to provide more reliable, rapid results.
Typically, LIBS signal intensity shows shot-to-shot fluctuation. This fluctuation is minimized by averaging over ‘n’ number of measurements. It was observed that by increasing number of measurements, ‘n’, fluctuation can be reduced. This behavior is quantitatively presented in Table 2. While RSD corresponding to a single measurement was 9.2% and the time required was 8 s, averaging 40 such measurements reduced the RSD to 1.24% and the time required rose to 320 s.
4. Conclusion In summary, analytical results found in this work support the hypothesis that LIBS can be used to measure CO2 in air. A good signal intensity (C-247.85 nm) was achieved by optimizing different parameters, the most important of them were found to be choice of focusing lens and ICCD on-chip accumulation. The desired gas mixtures for calibration were prepared by diluting the 50,000 ppm carbon dioxide
Fig. 9. Variation of CO2 concentration of ambient air with time is shown. Each point is an average of five measurements. Lines connecting the points are to guide the eye. The laser pulse energy, the gate delay and the gate width were set at 145 mJ, 0.5 μs, and 5 μs respectively.
70
V. Dikshit et al. / Spectrochimica Acta Part B 68 (2012) 65–70
Table 2 Shot-to-shot intensity variation, which is inherent in LIBS, can be reduced by averaging an increased number of measurements, which is reflected by reduced standard deviation. However, the time taken to make measurement will increase. No. of measurements averaged, n
Relative standard deviation, RSD
Time taken (s)
1 5 10 20 40
9.2% 3.6% 2.5% 1.66% 1.24%
8 40 80 160 320
Acknowledgments Authors would like to thank Christopher Ramos and K.E. Eseller for useful discussion during the work. This work was supported by DESUNSF contract no. HRD-0630388. References [1] D.A. Lashof, D.R. Ahuja, Relative contributions of greenhouse gas emissions to global warming, Nature 344 (1990) 529–531. [2] M. Holzinger, J. Maier, W. Sitte, Potentiometric detection of complex gases: application to CO2, Solid State Ion. 94 (1997) 217–225. [3] D.A. Cremers, L.J. Radziemski, Handbook of Laser-Induced Breakdown Spectroscopy, Wiley, New York, USA, 2006. [4] A. Miziolek, V. Palleschi, I. Schechter, Laser-Induced Breakdown Spectroscopy (LIBS): Fundamentals and Applications, Cambridge University Press, 2006.
[5] J.P. Singh, S.N. Thakur, Laser-Induced Breakdown Spectroscopy, Elsevier Science B. V, Amsterdam, The Netherlands, 2007. [6] A.J. Ball, V. Hohreiter, D.W. Hahn, Hydrogen leak detection using laser-induced breakdown spectroscopy, Appl. Spectrosc. 59 (2005) 348–353. [7] M. Tran, B.W. Smith, D.W. Hahn, J.D. Winefordner, Detection of gaseous and particulate fluorides by laser-induced breakdown spectroscopy, Appl. Spectrosc. 55 (2001) 1455–1461. [8] E.D. McNaghten, A.M. Parkes, B.C. Griffiths, A.I. Whitehouse, S. Palanco, Detection of trace concentrations of helium and argon in gas mixtures by laser-induced breakdown spectroscopy, Spectrochim. Acta Part B 64 (2009) 1111–1118. [9] K.E. Eseller, F.Y. Yueh, J.P. Singh, Non-intrusive, on-line, simultaneous multispecies impurity monitoring in hydrogen using LIBS, Appl. Phys. B 102 (2011) 963–969. [10] F. Ferioli, P.V. Puzinauskas, S.G. Buckley, Laser-induced breakdown spectroscopy for on-line engine equivalence ratio measurements, Appl. Spectrosc. 57 (2003) 1183–1189. [11] V. Hohreiter, D.W. Hahn, Calibration effects for laser-induced breakdown spectroscopy of gaseous sample streams: analyte response of gas-phase species versus solid-phase species, Anal. Chem. 77 (4) (2005) 1118–1124. [12] D.A. Cremers, M.H. Ebinger, P.J. Unkefer, S.A. Kannerdiener, M.J. Ferris, K.M. Catlett, J.R. Brown, Measuring total soil carbon with laser-induced breakdown spectroscopy (LIBS), J. Environ. Qual. 30 (2001) 2202–2206. [13] M.Z. Martin, S.D. Wullschleger, C.T. Garten, V.A. Palumbo, Laser-induced breakdown spectroscopy for the environmental determination of total carbon and nitrogen in soils, Appl. Opt. 42 (2003) 2072–2077. [14] R. Banhadur, Y. Feng, L.M. Russell, V. Ramanathan, Impact of California's air pollution laws on black carbon and their implications for direct radiative forcing, Atmos. Environ. 45 (2011) 1162–1167. [15] V. Kumar, R.K. Thareja, Laser-induced breakdown of argon gas near a metal surface, Laser Part. Beams 10 (1992) 109–116. [16] K.W. Thoning, P.P. Trans, W.D. Komhyr, Atmospheric carbon dioxide at Mauna Loa Observatory 2. Analysis of the NOAA GMCC data, J. Geophys. Res. 94 (1989) 8549–8565.
