Carbon monoxide (CO) and carbon dioxide (CO₂) affect the heat balance of the atmosphere and participate in the atmospheric chemistry.^[1] CO is an important atmospheric pollutant in the troposphere.^[2] CO also serves as an indicator of urban pollution, and its anthropogenic source is mainly the incomplete combustion of hydrocarbon fuels.^[3] Moreover, CO can react with O₂ to produce CO₂, therefore it is an indirect greenhouse gas (GHG). CO₂ is the most important anthropogenic GHG. It is the main product of combustion processes and human activities. Its annual emissions had grown by about 80% between 1970 and 2004, from 21 gigatonnes (Gt) to 38 Gt, and accounted for 77% of the total anthropogenic GHG emissions in 2004.^[4] Reliable and sensitive detection of atmospheric CO and CO₂ simultaneously is necessary,^[5,6] as these measurements can be used to determine their sources, leading to an improved understanding of the carbon cycle and climate change.

Tunable diode absorption spectroscopy (TDLAS) technology is ideally suited for trace gas detection in environment monitoring, industry control, and combustion diagnosis due to its advantages of in situ, high sensitivity, non-intrusive, and high selectivity.^[7–11] Simultaneous detection of atmospheric CO and CO₂ based on TDLAS has been reported by several research groups.^[12–15] Gabrysch et al.^[12] used a near-infrared laser (1.578 μm) for CO and CO₂ simultaneous detection and the sensitivity was increased by a factor of 15 compared to direct absorption, the minimum detection limit for CO and CO₂ was about 10 ppm (100 torr, 1 m) and 53 ppm (8 torr, 1 m). Engelbrecht^[13] reported a compact fiber-optic diode laser spectrometer for the two species measurement around 1580 nm, achieving a detection limit of 5.1 ppm CO and 9.1 ppm CO₂ with a 1 m absorption path length at 80 hpa. Chen^[14] developed a system for the detection of CO and CO₂ with two near-infrared lasers (1563.62 nm and 1580.99 nm) and a white cell. A remote laser-based sensor for vehicle CO and CO₂ emission detection was designed with a single laser (1.57 μm).^[15] Those researches were all based on the near-infrared spectrum, however, the fundamental absorption lines of most trace gases are in the mid-infrared region.^[16] Quantum cascade laser (QCL) is now a perfect mid-infrared light source^[17] and it is more suitable for high-sensitive trace gas detection in combination with TDLAS and the wavelength modulation spectroscopy (WMS) technique.^[18–20] Recent developments of QCL for simultaneous atmospheric trace gases detection have been reported, such as simultaneous detection of atmospheric CH₄ and N₂O emission fluxes by QCLS (4.5 μm, 7.5 μm, and 7.9 μm),^[21] and simultaneous atmospheric N₂O and CO detection by a 4.5 μm QCL.^[22] A QCL (7.73 μm) based sensor for simultaneous detection of atmospheric N₂O, CH₄, and H₂O was developed with a Herriot cell at 100 torr.^[23] However, using a single QCL for simultaneous measurement of atmospheric CO and CO₂ has rarely been reported.

In this paper, we develop a mid-infrared laser sensor for simultaneous measurements of atmospheric CO and CO₂ based on a single 4.85 μm continuous wave QCL (CW-QCL). The appropriate wavelength enables simultaneous multi-species measurements using a single laser, simplifying the sensor system and reducing its cost greatly. Two neighboring CO (2055.40 cm⁻¹) and CO₂ (2055.16 cm⁻¹) absorption lines can be covered in a single QCL wavelength scan. A multiple pass absorption cell with an effective path length of 74 m and the WMS technique are employed to achieve high sensitivity.

The experimental single-QCL-based system is presented in Fig. 4. A continuous wave distributed feedback QCL (Alpes Laser, Switzerland) around 2055 cm⁻¹ in a high heat load (HHL) package was used as the light source, with a beam-divergence angle less than 3 mrad. The temperature and the current were simultaneously controlled by a commercial laser controller (LDC 3724B, ILX Lightwave, USA). The mid-infrared light was combined with an aligned visible beam (λ = 650 nm) by a beam splitter (T > 90% at 4.85 μm, R > 85% at 650 nm) for aiding the optical alignment. The combined light beam went through two reflect mirrors, then passed into a white multi-pass cell. The cell consists of three spherical mirrors, with a base length of 1 m, which can provide an optical path length of 74 m. The leakage rate of the multi-pass white cell is 0.01 Pa·L/S. The output mid-infrared beam was focused onto an AC thermoelectric cooled mercury cadmium telluride detector (MCT) (PVI-2TE-10.6, Vigo Systems, Poland) using a parabolic mirror (f = 50.4 mm). The detected signal was acquired by a data acquisition (DAQ) card (M2i.4210 PCI 14 bit 20 Ms/s, SPECTRUM, Germany), and then input to a Labview-based lock-in amplifier to extract the 1 f-normalized 2 f signal.

Fig. 4.

	Figure Option ViewDownloadNew Window
	Fig. 4. Schematic of the experimental apparatus.

The optical path outside the gas cell was about 50 cm, which was 148 times smaller than the inside optical path of 74 m, so the influence of CO and CO₂ in ambient air can be neglected. It is better to make the outside part closed or purge with N₂ particularly for CO₂ detection, because it is easily influenced by human breath in the laboratory.

The two absorption lines were covered by a sweep range of about 0.6 cm⁻¹ with a current scan signal of 200 Hz. The modulation depth was optimized, which affects the peak value of the 2 f signal. The modulation frequency was set at 20 kHz to reduce the low-frequency noise. Under this modulation frequency, the modulation depth varied with the modulation voltage directly, so an appropriate modulation voltage should be selected. The laboratory air flowed into the cell and the needle valves were adjusted to maintain the pressure of 250 torr. The peak value of the 1 f-normalized 2 f signal is plotted in Fig. 5. As shown in Fig. 5, both the 2 f peaks of CO and CO₂ reached the maximum with a modulation voltage of about 120 mV, thus this modulation voltage was selected for CO and CO₂ simultaneous detection in the following experiment. The detector signal, the 1 f-normalized 2 f signal, and the peak values of CO₂ and CO are also given in Fig. 6. The asymmetry of the 2 f signal in Fig. 6 resulted from the intensity modulation. In Fig. 6, the 1 f-normalized 2 f signal of CO is different from that of CO₂, this difference perhaps is due to the background effect. Also, the absorbance of CO in the laboratory air was much less than that of CO₂, which would result in different background effects to their 2 f signals.

Fig. 5.

	Figure Option ViewDownloadNew Window
	Fig. 5. The 2 f peaks of CO and CO2 in laboratory air at different modulation voltages.

Fig. 6.

	Figure Option ViewDownloadNew Window
	Fig. 6. An example of detector signal and related 2 f signal of laboratory air.

Several standard gases (Guangming Research & Design Institute of Chemical Industry, with 2% relative uncertainty) with different mole fractions of CO (200 ppb, 500 ppb, 950 ppb, 1.7 ppm, 3.7 ppm) and CO₂ (95 ppm, 257 ppm, 404 ppm, 702 ppm, 1001 ppm) were used for the sensor calibration, and the carrier gas was N₂. During the calibration, these standard gases flowed into the gas cell with the concentration increased. The 2 f amplitude increased with the concentration of both gases. When each 2 f peak became stable, the signal was recorded for 20–30 min. The calibration processes of CO and CO₂ are plotted in Figs. 7 and 8. After calibration, a continuous detection of relatively lower standard gas (950 ppb CO, 404 ppm CO₂) was conducted to test the sensor reproducibility, as illustrated in Figs. 7(a) and 8(a).

Fig. 7.

	Figure Option ViewDownloadNew Window
	Fig. 7. Calibration process (a) and calibration results (b) for CO.

Fig. 8.

	Figure Option ViewDownloadNew Window
	Fig. 8. Calibration process (a) and calibration results (b) for CO2.

For CO, the linear fit coefficient of CO is greater than 0.998. For the second 950 ppb CO detection, the average 2 f amplitude in 30 min is 2.14, and the corresponding value is 2.12 for the first detection in the calibration, so the amplitude difference is 0.02. The related concentration difference between the two stages is 10 ppb (1%). The two differences are small, showing the good reproducibility. Similarly, for CO₂, the linear fit coefficient is greater than 0.99. For two detections of the same concentration of the standard gases, the differences are 0.35 and 15 ppm (3.7%) for the amplitude and the concentration, respectively, which are both relatively small. All these results illustrate a good linear response and reproducibility of the sensor for the detection of CO and CO₂.

In order to test the detection limit and the long-term precision of the experimental system, the laboratory ambient air was closed in the gas cell and continuously measured at 250 torr. An example of the signals is shown in Fig. 9. The continuous detection results of 3000 s are shown in Fig. 10 with 1-Hz sample rate. According to the measurement results, the averaged concentrations of CO and CO₂ are about 1.07 ppm and 509.24 ppm, which are in the normal concentration ranges in atmosphere. During the detection period, the root mean square (RMS) noise is 8 ppb for CO (about 0.7% of the laboratory ambient air level) and 1.78 ppm for CO₂ (about 0.35% of the laboratory ambient air level).

Fig. 9.

	Figure Option ViewDownloadNew Window
	Fig. 9. An example of the related signals of ambient air: (a) detector absorption signal, (b) normalized 2 f and 1 f signals.

Fig. 10.

	Figure Option ViewDownloadNew Window
	Fig. 10. Continuous measured results of CO (a) and CO2 (b) in laboratory ambient air for 3000 s.

In general, the Allan variance plot is used to seek the optimum integration time and the corresponding minimum detection limit for the system. The Allan variance of the simultaneous detection results is plotted in Fig. 11, which demonstrates that the optimum integration time is about 85 s for the simultaneous detection of CO and CO₂. A minimum Allan deviation of 1.3 ppb for CO (∼ 1‰ of the ambient level) and 0.44 ppm for CO₂ (∼0.9‰ of the ambient level) is achieved at the optimum integration time. The high detection limit shows that the system is appropriate for atmospheric monitoring of these two species. With a longer integration time, there is an increase in the Allan Deviation due to the instrumental drifts, including laser wavelength drift, optical drift, and electronic drift.

Fig. 11.

	Figure Option ViewDownloadNew Window
	Fig. 11. Allan deviations of CO (a) and CO2 (b) as a function of integration time.

We have developed a single QCL based absorption spectroscopy system for simultaneous detection of atmospheric CO and CO₂ with a 74 m multi-pass cell. Utilization of a single QCL to realize dual-species detection reduced the cost and simplified the experimental configuration. The two neighboring CO (2055.40 cm⁻¹) and CO₂ (2055.16 cm⁻¹) absorption lines were covered within a single wavelength scan. The concentration calibration of the system was performed using standard gases with different concentrations, and the calibration results illustrated a rather good linear response for both CO (200 ppb–3.7 ppm) and CO₂ (95 ppm–1001 ppm). The detection limit and long-term precision of this CO/CO₂ sensor were examined by monitoring the laboratory room air for about 3000 s, and its Allan deviation plots showed that a detection limit of 1.3 ppb for CO and 0.44 ppm for CO₂ can be achieved at 85 s integration time.