1. Introduction

Trace-gas detection of ethylene (C₂H₄) has numerous applications in atmospheric chemistry [1], medical diagnostics [2] and plant biology [3]. In atmospheric chemistry, C₂H₄, produced mainly from automobile exhaust and industrial emissions, is a primary pollutant in promoting the formation of tropospheric ozone in urban areas [4]. The average concentration of C₂H₄ in air is ~20 parts per billion (ppb), requiring a ppb-level sensitivity of a detection system [5]. In biomedical research, compared with the normal level of 6.5 ppb in human exhaled breath, an immediate increase of C₂H₄ concentration from 0.15 ppm to 0.78 ppm has been observed in elderly patients with renal failure after dialysis treatment [2]. The increased C₂H₄ level may intensify the oxidative stress that contributes to morbidity. The sensitive and reliable detection of C₂H₄ provides a non-invasive way of identifying renal failure by analyzing patient’s exhaled breath [2]. Additionally, C₂H₄ is a plant hormone significantly affecting the growth, ripening, and decay of fruits [3]. It is thus essential to control the C₂H₄ concentration level during fruit transportation. For example, an ethylene sensor with 1 ppm sensitivity is utilized for this purpose during banana transportation [5].

Laser absorption spectroscopy is an effective and promising technique for trace gas detection because of its high sensitivity, fast time-response, and cost effectiveness. Of various absorption detection schemes, cavity ring down spectroscopy (CRDS) and multipass absorption cell were two major techniques implemented previously for C₂H₄ detection of trace amount [5–11]. A CRDS C₂H₄ sensor using a distributed feedback (DFB) diode laser (1.62 µm, <10 mW) was developed by Aziz et al. [5] to detect C₂H₄ in ambient air. A minimum detection limit (MDL) of 280 ppt at 30 min integration time was achieved by employing a 30 cm optical cavity composed of two high-reflectivity mirrors (reflectivity R > 99.99%) [5]. A similar work was reported by Wahl et al. [6] to obtain 2 ppb detection limit of C₂H₄ at 4.4 min integration time using CRDS at 1.618 µm. The emission spectrum of carbon dioxide has a strong overlap with the v₇ fundamental absorption band of C₂H₄ at 10 µm spectral region, making carbon dioxide lasers possible light sources for C₂H₄ detection. Mürtz and co-authors reported the development of a C₂H₄ sensor using cavity leak-out spectroscopy (R = 99.5%) combined with a continuous-wave (CW) carbon dioxide laser to achieve a MDL of 1 ppb at 100 s integration time [7]. The carbon dioxide laser was also recently employed for combustion diagnostics of C₂H₄ near 10.5 µm using direct absorption spectroscopy [8].

Compared with the carbon dioxide laser with a large system size, the commercial quantum cascade lasers (QCLs) with a compact HHL package can readily access the mid-IR spectral range from 4 μm to 13 μm. Weidmann et al. [9] utilized a pulsed QCL (1 mW peak power and 10 kHz pulse repetition rate) emitting at 10 µm to detect C₂H₄ with a MDL of 30 ppb at 80 s integration time using an astigmatic Herriott multipass cell (182 passes, 100 m optical path length). Similar C₂H₄ sensors employing a Herriott multipass cell and a pulsed QCL at 10.3 µm were developed later by several other groups to achieve ppb-ppm detection limit [10,11]. It is noted that high detection sensitivity was achieved for these C₂H₄ sensors using CRDS or multipass cell techniques. However, the adoption of an optical cavity or multipass cell significantly increases the optical complexity, gas sampling volume and system cost. In particular, an expensive photodetector working near the long-wavlength infrared of 10 µm was required to obtain accurate measurement for relatively weak laser absorption.

Photoacoustic spectroscopy (PAS) is another common technique, which uses a microphone for acoustic signal measurement instead of a photodetector for laser absorption measurement [12]. Regarding the C₂H₄ detection using PAS, the carbon dioxide laser emitting at ~10.5 µm is mostly employed because of the high power output of commercial carbon dioxide lasers [13–16]. A typical PAS sensor using a CW carbon dioxide laser (1.9 W, 2.4 kHz amplitude modulation) was developed to monitor C₂H₄ concentration with a MDL of 16 ppb at 0.3 s integration time [1]. Such PAS sensors with a similar system configuration were applied in breath analysis [13], flooding research [14], and identifying industrial sources [15], ripening of fruits [16] and plant emissions in a stress condition [17]. However, the further applications of these PAS-based C₂H₄ sensors are significantly limited by the large system size of the laser source and the photoacoustic cell.

Quartz-enhanced photoacoustic spectroscopy (QEPAS) invented by Tittel’s group at Rice University [18], is one of the most sensitive spectroscopic techniques for trace-gas detection applications using different laser sources [19–26]. It is a spectroscopic method that works by combining PAS and a quartz tuning fork (QTF). The QTF acts as a resonant acoustic transducer that converts the acoustic wave generated by gas absorption of the modulated laser intensity into an electrical signal via piezoelectric effect. The commercially available QTF with a sharply resonant frequency of 32.768 kHz makes the sensor low-cost and less sensitive to environmental noise compared with the broadband electric microphone used in traditional PAS. Additionally, the tiny QTF (4 mm × 1.5 mm × 0.35 mm) utilized for acoustic detection allows the realization of a compact sensor with a small sampling volume. Only a couple of QEPAS-based C₂H₄ sensors were reported previously using a 1.62 µm (15 mW) diode laser to obtain a MDL of 4 ppm at 0.7 s data acquisition time [25], or a 3.32 µm DFB laser (1.5 mW) to achieve a MDL of 63 ppm at 25 s averaging time [26]. However, as illustrated in Fig. 1 of the C₂H₄ absorption spectra (PNNL database [27]), C₂H₄ shows the strongest absorption band near 10.5 µm compared with that at 1.6 µm and 3.3 µm. The QEPAS sensor detecting the strongest absorption band near 10.5 µm enables the possibility of more sensitive detection of C₂H₄.

 

Fig. 1 (a) Absorption spectra of C₂H₄ between 1 and 13 µm at 296 K for 1 ppm C₂H₄ and 1 m optical path length (PNNL [27] database). (b) Simulated absorption coefficients of 1 ppm C₂H₄ and standard air near 10.5 µm at 296 K and 500 Torr using HITRAN [28] database (the QCL tuning range with a single current scan is highlighted).

Download Full Size | PPT Slide | PDF

In this work, we report the development of a novel QEPAS-based C₂H₄ sensor by exploiting the C₂H₄ spectra near 10.5 µm. Recent availability of the compact mid-IR QCLs allows the access of the strongest absorption band. Sensor development is discussed to overcome the new challenges at this long-wavelength infrared compared with previous QEPAS sensors. The sensor calibration and optimization, long-term stability, and a practical application in analyzing the C₂H₄ efflux of real biological samples, are presented in this paper.

2. Spectral analysis and laser characterizations

2.1 C₂H₄ spectral analysis

The C₂H₄ absorption spectra between 1 and 13 µm adopted from PNNL database [27] are depicted in Fig. 1(a). The previous QEPAS detection of C₂H₄ exploited the absorption bands near 1.62 µm and 3.32 µm [25,26], corresponding to the (ν₅ + ν₉) combination band and the ν₁₁ fundamental band, respectively. Figure 1(a) also reveals that the peak C₂H₄ absorbance at ~10.5 µm (ν₇ fundamental band) is about 300 times and 6 times stronger compared to that at 1.62 µm and 3.32 µm, respectively.

The simulated absorption coefficient of the targeted C₂H₄ absorption feature located at 949.34 cm⁻¹ (10.534 µm) is illustrated in Fig. 1(b) using the HITRAN database [28]. C₂H₄ shows unresolved congested spectra compared with those simpler molecules. The simulation was performed for 1 ppm C₂H₄ in standard air (H₂O 2.5%, CO₂ 0.04%, CO 200 ppb, CH₄ 1.76 ppm, N₂O 317 ppb, N₂ 75.5% and O₂ 21%) at a pressure of 500 Torr to assess potential interferences from other atmospheric species. It is found that the strong H₂O line centered at 948.26 cm⁻¹ is more than 1 cm⁻¹ away from the C₂H₄ absorption peak, showing no overlap with the C₂H₄ peak even at atmospheric pressure. When employing wavelength modulation spectroscopy (WMS) for the 2f peak detection to infer the gas concentration, the spectral interference from H₂O is thus negligible. A slight interference from the CO₂ line centered at 949.48 cm⁻¹ is also negligible due to its much smaller absorption compared with that of C₂H₄. Recent commercially available QCLs are able to access this strong C₂H₄ feature near 10.5 µm. For example, at a fixed laser temperature, the DFB-QCL used in the present work can be tuned across the C₂H₄ absorption feature with a single current scan as highlighted in Fig. 1(b).

2.2 QCL characterization

A 10.5 µm CW DFB-QCL mounted in an HHL housing (Alpes Lasers, Switzerland) was used as the excitation source. The laser tuning characteristics of the emission wavelength and power were characterized by directing the laser beam into a power meter (Ophir Optronics) and a spectral analyzer (Bristol Instruments, accuracy of ± 0.75 ppm), respectively. Figure 2 depicts the laser power and wavelength as a function of injection current at different temperatures (0-30°C). As shown in Fig. 2(a), the maximum output power of the QCL varies between 5 and 27 mW over the entire temperature range at the maximum injection current. This DFB-QCL can be tuned over the frequency range between 946.5 and 950.4 cm⁻¹ (10.522-10.565 µm) by adjusting both the laser temperature and injection current. The absorption feature of C₂H₄ centered at ~949.3 cm⁻¹, corresponding to the grey line in Fig. 2(b), can be well covered by the QCL. Considering that a higher excitation power is preferred for photoacoustic detection, the laser temperature is set at 0 °C to obtain an output power of ~23 mW. At this fixed temperature, the laser frequency can be tuned between 949.2 cm⁻¹ and 950.4 cm⁻¹ using a single current scan with a current tuning parameter of −0.00651 cm⁻¹/mA.

 

Fig. 2 The (a) power and (b) wavelength tuning characteristics of the CW DFB-QCL at different operating temperatures.

Download Full Size | PPT Slide | PDF

Besides the ramp voltage applied to scan the laser wavelength across the absorption feature, a sinusoidal dither at half of the resonant frequency of QTF (f₀/2) is superimposed onto the ramp voltage for WMS detection. The frequency response of the sinusoidal modulation, or modulation depth, is another important parameter that needs to be characterized. Here a Germanium etalon with a designed length of 76.2 mm (free spectral range, FSR = 0.0164 cm⁻¹) was placed on the optical path between the QCL and the photodetector (see the optical setup below) to track the frequency response. The etalon curve of the interference fringe when the laser current was modulated at 16.3 kHz (f₀/2) with a modulation amplitude of 0.36 V is plotted in Fig. 3(a). Considering that the two adjacent peaks correspond to one FSR of the etalon, the frequency response curve can be obtained using the method as described in [29] and the results are shown in Fig. 3(b). The laser frequency modulation (FM) is well described by a sinusoidal fitting with an amplitude measured to be 0.0998 cm⁻¹ (modulation depth). Along with the etalon curve, the intensity curve is also plotted in Fig. 3(b) for comparison. Similarly, the intensity modulation (IM) was best-fitted to a sine function with the residual shown at the bottom panel of Fig. 3(b). A phase shift of 1.07 π between the FM and IM is found for this particular DFB-QCL. The amplitude of the residual curve, namely the nonlinear term of IM, is measured to be within ± 2% of the intensity modulation.

 

Fig. 3 (a) Interference signal of etalon (two adjacent etalon peaks correspond to one FSR). (b) QCL frequency modulation (FM) depth and the phase shift relative to the intensity modulation (IM); bottom panel, residual of the sine curve fit to the IM.

Download Full Size | PPT Slide | PDF

3. Experimental

3.1 Sensor setup

The schematic of the QEPAS-based C₂H₄ sensor setup is depicted in Fig. 4. The CW DFB-QCL with ~23 mW power output at the target wavelength near 10.5 µm was current and temperature controlled by a commercial combination laser driver (ILX Lightwave, LDC 3736). Two ZnSe anti-reflection (AR) coated plano-convex lenses, L₁ (f = 20 mm) and L₂ (f = 40 mm), were used to focus the laser beam between the two prongs of QTF and through two micro-resonator (mR) tubes inside an acoustic detection module (ADM). These two mR tubes were installed adjacent to the QTF to enhance the acoustic signal by a factor of ~30 [30]. A pinhole with a diameter of 400 µm placed between L₁ and L₂ was used as a spatial filter to improve the beam quality. A typical QCL beam profile after the ADM was captured by a pyroelectric camera (Ophir Optronics) as illustrated in the inset graph of Fig. 4. The custom-designed ADM mounted on a xyz micro-positioning system was equipped with two ZnSe windows wedged at 8° to avoid unwanted etalon effect. A photo of the C₂H₄ sensor is also shown in Fig. 4.

 

Fig. 4 Schematic and photo of the QEPAS sensor setup for C₂H₄ detection. L₁, L₂: plano-convex lens; ADM, acoustic detection module; QTF, quartz tuning fork; mR, micro-resonator; CEU, control electronics unit; DAQ, data acquisition.

Download Full Size | PPT Slide | PDF

Besides the sensor system, Fig. 4 also presents the additional optical set up for laser characterization. The flip mirror placed before L₁ was used to direct the laser beam into the wavelength analyzer for measuring the absolute wavelength or through the Germanium etalon for characterizing the wavelength modulation depth. The output laser beam from the etalon was focused onto a room-temperature MCT detector (VIGO System S.A.) and further processed by a data acquisition (DAQ) module (PCI 6110, National Instruments).

The piezo-electrical signal generated by the QTF was amplified by a transimpedance amplifier with a 10 MΩ feedback resistor and then demodulated by a lock-in amplifier to obtain the 2f signal. These operations were carried out by a custom-designed control electronics unit (CEU) controlled by LabVIEW programs. The CEU was also programmed to measure the QTF parameters (i.e., quality factor, dynamic resistance and resonant frequency) and modulate the laser injection current.

3.2 Optimization of the micro-resonator (mR)

The previous research concluded that the mR tubes with a length of 4.4 mm and an inner diameter (ID) of 0.6 mm yield the best signal-to-noise ratio (SNR) at gas pressures between 400 and 800 Torr [30]. However, we observed a relatively large background noise with this configuration, which was analyzed and found to be mainly due to the optical interference between the 10.5 μm laser beam and the mR tubes. Note that a power loss of (5-8%) was still found after the laser beam was carefully aligned through the QTF and mR tubes. This is because a longer wavelength of 10.5 µm was implemented in the present work, leading to a larger spot size at the focal point compared to the previous study that used a 1.53 µm near-IR diode laser [30].

Instead of using the mR configuration proposed in reference [30], here we designed and tested new types of mR tubes suitable for the long mid-IR wavelength at 10.5 µm. Two micro-tubes with a length of 4.4 mm and ID of 0.9 mm were selected to reduce the laser transmission loss below 0.2%. Due to the acoustic coupling between the QTF and the mR of new geometry, the gaps between the QTF and mR tubes need to be increased from 50 µm (the optimized value in reference [30]) to 70 µm to obtain the highest photoacoustic signal. Figure 5 compares the measured QEPAS 2f spectra of C₂H₄ (100 ppm in N₂) and the background noise (pure N₂) using two different mR dimensions. As shown in Fig. 5(a), the 2f peak amplitude using the mR tubes of 0.6 mm ID was measured to be 21.4 mV, slightly higher than in the case of 0.9 mm ID (15.1 mV). However, the new mR geometry (4.4 mm length, 0.9 mm ID) significantly reduced the optical noise by a factor of 6 as indicated in Fig. 5(b). Hence, our new QTF-mR configuration improves the SNR by a factor of 4.2 in general and was adopted in the sensor development.

 

Fig. 5 QEPAS (a) 2f signal and (b) background noise of two different mR dimensions. Red line, ID 0.9 mm; black line, ID 0.6 mm.

Download Full Size | PPT Slide | PDF

3.3 QEPAS 2f signal vs. modulation depth & gas pressure

The QEPAS 2f signal is related to the modulation depth and gas pressure, which need to be optimized to obtain the largest 2f peak amplitude. The QEPAS signal for a mixture of 95 ppm C₂H₄ in nitrogen was thus investigated at varied gas pressures and modulation depths for this purpose. The experimental results are plotted in Fig. 6(a) with the representative 2f spectra at two different modulation depths shown in Fig. 6(b). At a fixed modulation depth, when the gas pressure is relatively low, the 2f peak amplitude increases significantly with the increasing pressure mainly due to the enhanced collision rate and vibration-translation (V-T) relaxation at higher pressures. After reaching the peak value at a specific pressure (determined by the modulation depth), the 2f amplitude starts to decrease with the increasing pressure mainly because of the reduced quality factor of the QTF at higher pressures as shown in Fig. 6(a).

 

Fig. 6 (a) Measured QEPAS 2f peak amplitude as a function of laser modulation depth and gas pressure. (b) Representative QEPAS 2f curves at two different modulation depths and a fixed pressure of 500 Torr. (c) Measured 2f peak amplitude as a function of laser modulation depth at two different pressures.

Download Full Size | PPT Slide | PDF

Figure 6(a) also indicates that the 2f peak amplitude increases with the modulation depth over a wide pressure range of 300-760 Torr. At the gas pressures of 500 and 760 Torr, the 2f amplitude is plotted in Fig. 6(c) as a function of modulation depth. The 2f amplitude increases almost linearly with the modulation depth from 0.0155 to 0.0939 cm⁻¹, but changes slightly when going to the highest modulation depth of 0.1211 cm⁻¹. Hence, the gas pressure of 500 Torr and modulation depth of 0.0939 cm⁻¹ were selected as the optimum parameters for C₂H₄ detection considering the largest 2f amplitude and QCL current limit. It should be noted that the 2f peak amplitude at 760 Torr is about 12% lower than that at the optimum pressure of 500 Torr. In practical applications, atmospheric pressure is preferred for sensor development due to the simpler system configuration involved; there is no need to implement a pumping system and a pressure monitor. The sensor performance at both pressure conditions (500 Torr and 760 Torr) were investigated in this work.

4. Results and discussion

4.1 Sensor performance

The sensor performance was examined using a series of C₂H₄ mixtures of known concentrations. The gas mixtures were generated by diluting the C₂H₄:N₂ with a certificated concentration of 100 ppm using the gas dilution system (Jinwei Electronics) and introduced into the ADM by a diaphragm pump (KNF, N86KTE). The gas pressure inside the ADM was monitored by a pressure transducer (MKS Instruments, 722B) and controlled by adjusting the two needle valves (Swagelok) installed before the inlet and after the outlet of the ADM, respectively.

The representative 2f signals at different C₂H₄ concentrations using the optimum modulation depth (0.0939 cm⁻¹) and pressure (500 Torr) are illustrated in Fig. 7(a). The corresponding peak amplitude of the QEPAS 2f signal is plotted in Fig. 7(b) as a function of C₂H₄ concentration (5-95 ppm). A linear fit to the experimental data yields an R-square value of 0.9994, indicating a good linear response of the sensor and a calibration coefficient of 41.6 ppb per 1 pA root-mean-square (RMS) of the QTF generated current.

 

Fig. 7 (a) Representative QEPAS 2f signals recorded with varied C₂H₄ concentrations at 500 Torr and modulation depth of 0.0939 cm⁻¹. (b) Measured 2f peak amplitude as a function of C₂H₄ concentration with the best linear fit.

Download Full Size | PPT Slide | PDF

As discussed previously, considering the advantage of applying the C₂H₄ sensor at atmospheric pressure, the sensor performance was also calibrate at 760 Torr using the same modulation depth of 0.0939 cm⁻¹. A better linear response (R-square value of 0.9999) at 760 Torr was found in the calibration curve as shown in Fig. 8(a), leading to a calibration coefficient of 42.8 ppb per 1 pA RMS of the QTF generated current. In order to evaluate the precision and stability of the sensor system, an Allan deviation analysis was conducted by measuring pure N₂ for one hour with the results illustrated in Fig. 8(b). The sensor achieves a precision of 560 ppb of C₂H₄ with 10 Hz sample rate. The turning point of the Allan deviation curve is approximately 70 s leading to an optimal C₂H₄ sensitivity of ~50 ppb, corresponding to a normalized noise equivalent absorption (NNEA) coefficient of 1.78 × 10⁻⁷ cm⁻¹W/Hz^1/2.

 

Fig. 8 (a) Linearity of the 2f peak amplitude to the C₂H₄ concentration at 760 Torr; inset graph, typical QEPAS 2f signals at different C₂H₄ concentrations. (b) Allan deviation plot of the C₂H₄ sensor system.

Download Full Size | PPT Slide | PDF

It is noted that a fast V-T relaxation occurs from the C₂H₄ excited state even in dry N₂ [25]. The relaxation effects that have been reported for QEPAS detection of CO₂ and HCN [31,32], due to the high modulation frequency, are not an issue in the case of C₂H₄. Hence, the possible H₂O interference for the current C₂H₄ sensor is negligible in terms of either the cross-talk absorption or the V-T relaxation effect.

4.2 Sensor application

C₂H₄ is a naturally occurring gas associated with the internal C₂H₄ concentration in the fruit core of plants under stress and fruit maturation in horticulture [33]. Our QEPAS sensor was applied to measure the C₂H₄ efflux of apples to investigate the sensor feasibility on the analysis of real biological samples.

The experiment was carried out in four steps as shown in Fig. 9. First, four apple samples (initially stored in a fridge and with a total weight of 0.67 kg) were put into a sample bag (E-Switch, 5 L volume) and then fully evacuumized. Second, nitrogen with 99.999% purity was introduced into the sample bag until at atmospheric pressure or slightly above. The first two steps were repeated several times to ensure the removal of the residual air. Third, after waiting for a certain time, the ADM was evacuated first and then connected to the sample bag via a needle valve. Finally, the gas sample was introduced into the ADM by adjusting the needle valve to 760 Torr. The C₂H₄ concentration was then measured by acquiring the QEPAS 2f signal.

 

Fig. 9 Experimental procedures of monitoring the C₂H₄ efflux of apples.

Download Full Size | PPT Slide | PDF

A typical QEPAS 2f signal of the C₂H₄ emitted from the apple samples is illustrated in Fig. 10(a), corresponding to the time of half an hour after starting the experiment. The C₂H₄ concentration was measured to be 8.17 ppm using the calibration curve described in the previous section. A calibrated C₂H₄/N₂ mixture of 8.0 ppm is also plotted in Fig. 10(a) for spectral comparison. This measured concentration corresponds to a C₂H₄ efflux of 12 ppb/g for the apple sample. These measurements were repeated in the first five hours with the measured C₂H₄ efflux (ppb/g) and efflux rate (ppb/g/min) summarized in Fig. 10(b). It is observed that the C₂H₄ efflux increases with time but at a decreasing rate from 0.32 ppb/g/min at 0.25 hour to 0.016 ppb/g/min at 5 hours.

 

Fig. 10 (a) Representative QEPAS 2f signal of C₂H₄ emission by the apples; the 2f signal of a known concentration of C₂H₄/N₂ mixture is also plotted for comparison. (b) Measured C₂H₄ efflux of the apples varied with time.

Download Full Size | PPT Slide | PDF

5. Conclusion

We have reported the development of a novel QEPAS-based C₂H₄ sensor using a CW DFB-QCL by detecting the strongest absorption band of C₂H₄ at ~10.5 µm. The mR-QTF configuration was optimized to significantly improve the detection SNR. Very good linear response to the C₂H₄ concentration was achieved for the sensor system. The long-term stability and precision of the sensor were evaluated to show a MDL of 560 ppb at 0.1 s integration time and an improved MDL of 50 ppb at 70 s integration time. The sensor was successfully applied in measuring the C₂H₄ efflux of apples under laboratory conditions, demonstrating its feasibility in future agricultural- and industrial-related applications. The present study provides a clear pathway for precise and stable C₂H₄ measurement using a QCL-based QEPAS sensor at the pressure of 500-760 Torr. Future work will also involve the improvement of the detection sensitivity considering the modest NNEA achieved in the present study, so that the sensor system is suitable for atmospheric field deployment requiring a ppb or sub-ppb sensitivity. It should be noted that the spectral interference of ambient CO₂ is non-negligible at ppb-level C₂H₄ concentration and must be quantitatively determined and subtracted from the measured C₂H₄ signal.