https://doi.org/10.3390/chemosensors10080321
“A TDLAS-based H2 sensor is proposed in this paper. A continuous-wave (CW), distributed feedback (DFB) diode laser was employed to cover the strongest absorption line of H2 at 2121.83 nm. Because the H2 absorption line has a weak line strength, a 10.13 m Herriott multipass gas cell (HMPC) was used to increase the optical absorption path of H2 in the gas chamber to laser energy, thereby enhancing the signal amplitude. The wavelength modulation spectroscopy (WMS) and second harmonic (2f) demodulation techniques were utilized to reduce the noise level and improve the sensing performance. Daubechies (DB) wavelet was applied for further signal denoising. The system’s stability was investigated using the Allan variance.”
”
2.1. Principle of TDLAS Technique
As an absorption spectroscopy technology, TDLAS is based on the Beer–Lambert law:
I=I0exp[−α(ν)L]�=�0���−���
When α(ν)L≪1���≪1, this formula can be rewritten as:
I=I0(1−α(ν)L)�=�01−���
where I0 is the intensity of laser output, I is the intensity of the laser received by the photodetector (PD), L is the absorption optical path, α(ν)�� is the absorption coefficient, and can be expressed as follows:
α(ν)=PS(T)φ(ν)C��=������
where P is the gas pressure, C is the gas concentration, S(T) is the line strength of the gas at temperature T, and φ(ν)� is the line shape function.
When the emission laser makes contact with gas molecules, the gas molecules will absorb the laser energy if the wavelength of the laser coincides with the absorption line of the gas molecule. In combination with Equations (2) and (3), the attenuation of the light intensity is proportional to the gas concentration C, absorption optical path L, and line strength S(T) [42]. A larger attenuation of laser intensity tends to produce a stronger TDLAS sensor signal, so there are many methods to improve the detection limit of the TDLAS technique, such as selecting the appropriate absorption spectroscopy line, thereby increasing the absorption optical path [43,44,45,46,47].
2.2. The Selection of Laser Source
In terms of gas spectroscopy detection technology, the performance of the laser has a decisive influence. The main requirements of the laser source are as follows:
-
A laser with a suitable output power. A moderate output power is advantageous in obtaining a high signal-to-noise ratio (SNR).
-
A single-longitudinal-mode and narrow line width output can improve the selectivity and sensitivity of system detection.
-
A stable mode and no mode-hopping output, which can reduce the noise caused by mode competition and improve the stability of system detection.
-
The working wavelength can be quickly tuned, and the tuning range is wide enough to ensure that the spectral absorption range of the target gas can be covered. The requirements of fast response and high-speed data acquisition can be met. The output wavelength range of the laser determines the species and quantity of detectable gases.
-
High beam quality, that is, a small divergence angle and ideal spot shape, which facilitates beam coupling in the optical system and reduces optical noise.
-
A source that is insensitive to changes in environmental conditions. External environment temperature, humidity, and pressure changes will not significantly influence the output performance of the laser.
-
The volume, weight, and power dissipation of the laser should be small, and the lifetime should be long so as to meet the requirements of engineering application.
According to the above, the development of TDLAS technology, which is a spectroscopy technique, is strongly restricted by laser sources [48]. Laser sources that can match H2 absorption lines have just emerged in recent years, which has resulted in rare research on H2 sensing using TDLAS. A distributed feedback laser with a built-in Bragg grating can result in a single mode laser. Due to the excellent monochromaticity and side-mode suppression ratio, the DFB laser has wide gas detection. Compared with other laser sources, it possesses the characteristics of small size, low power consumption, and a low cost. Therefore, a DFB diode laser was adopted in this research.
2.3. The Selection of H2 Absorption Lines
It is worth noting that each gas has a different line strength at different wavelengths, which avoids the problem of cross-sensitivity between different gases by selecting the appropriate absorption lines. A strong line strength often results in a large attenuation of the light intensity. Therefore, the selection of absorption lines is important and usually obeys the following rules:
-
The line strength of the selected absorption line should be as large as possible.
-
The selected gas absorption lines minimize the degree of overlap with other gas absorption lines.
-
It should be ensured that a suitable laser source is available to excite the selected gas absorption lines.
Usually, the high-resolution transmission molecular absorption database (HITRAN) is used to predict and simulate the transmission and emission of light in the atmosphere, and we used the database in this study.
2.4. The Selection of Multipass Gas Cell
The line strength of the strongest absorption line of H2 (10−26) is still much weaker than other gases (~10−21) [38,39,40]. Therefore, it is necessary to increase the amount of light intensity attenuation by using a long optical path gas cell.
Presently, the commonly used method to increase the optical path is through the use of a multipass gas cell, which includes two main kinds of multipass cells: White cells and Herriott cells. The most major difference between a Herriott cell and White cell is the reflection point. Every reflection of a White cell is at the center of the mirror, meaning that there are multiple reflections at the center of each small mirror at the same time, and reflection points overlap with each other. On the other hand, the reflection points of the Herriott cell are distributed around the mirror, forming a ring; each reflection point will form a separate spot of light without overlapping each other. Therefore, if an excitation light source with a narrow line width (such as a laser) is used for the White cell, the overlapping of the reflection points causes the mutual interference of the laser, resulting in interference noise. The Herriott cell can solve this problem as each reflection point of the Herriott cell is independent and does not overlap with another point, so no interference fringes will be generated. A White cell is usually formed by using three mirrors and a Herriott is created by using two mirrors. The Herriott cell is widely adopted in TDLAS technology.
2.5. WMS and Harmonic Demodulation Technique
The detection system based on TDLAS technology usually has low frequency background noise, such as the 1/f noise, and most of them have frequencies of less than 1 kHz. Therefore, if WMS technique is employed and the laser wavelength is modulated with a high frequency (>1 kHz), it can effectively reduce the influence of low frequency noise on the signal. Generally, there are two existing WMS methods: fixed-point wavelength modulation and scanning wavelength modulation. In TDLAS, the former method only uses a sine wave to modulate the wavelength corresponding to the center of the absorption line; the latter not only adopts a low frequency sawtooth wave to scan the selected whole absorption line, but applies a high frequency sine wave to modulate the wavelength of the whole absorption line at the same time. The former is only used for single wavelength point measurements, while the latter can scan the entire absorption line and realize multi-point and multi-wavelength measurements. Therefore, the second method is widely used in TDLAS and other laser spectroscopy-based gas sensing.
Harmonic demodulation technique is often used for the demodulation of modulated signals. In TDLAS, the magnitude of the harmonic signal is related to the strength of the gas absorption line corresponding to that wavelength [48]. Hence, the shape of the harmonic signal obtained by demodulation is theoretically a straight line for the first method, while for the second method, the shape of the harmonic signal obtained by demodulation is related to the harmonic order number. In terms of each order of harmonic, the even harmonic peak is located at the center of the absorption line, while the odd harmonic is zero at the absorption peak. With the increase in harmonic order, the harmonic peak decreases rapidly. Compared with other orders of harmonics, the peak value of the second harmonic (2f) is not only located at the absorption peak but also is the largest among even harmonics. Based on these features of harmonic signals, the 2f order is often used during the demodulation of modulated signals.
2.6. Allan Variance
System stability is the main factor that affects high sensitivity measurement. A perfectly stable gas detection system based on spectroscopy technology can theoretically obtain an extremely sensitive detection performance through infinite averaging. However, due to the fluctuations of laser emission wavelength, interference fringe, background spectroscopy, and other reasons, real TDLAS systems are only stable for a limited time, so that the system’s best detection sensitivity is only achieved in a limited time span.
The Allan variance was proposed in 1993 by Werle et al. for the TDLAS system to obtain the best detection sensitivity, which can be written in the following form as a time-domain approximation when neglecting the duty cycle effects [49]:
σ2A(τ)=cwhite noist(1/τ)+c1/f+∑αcdrift,ατα��2�=��ℎ��� �����1/�+�1/�+∑�������,���
where τ is the integration time, cwhite noist(1/τ)��ℎ��� �����1/�, c1/f�1/�, ∑αcdrift,ατα∑�������,��� represent the contributions of the frequency independent “white noise”, frequency dependent 1/f, and 1/fα(α > 1) noise to Allan variance, respectively, and α characterizes the type of system drift. It can be seen from Equation (4) that when the integration time τ is small, the Allan variance is mainly caused by white noise and decreases with the increase of integration time. However, as the integration time increases, the main factor affecting the Allan variance becomes the frequency dependent noise which increases with the increase in integration time. Therefore, the curve of the Allan variance as a function of integration time exists at a minimum. Additionally, the square root of the Allan variance in regions dominated by white noise is proportional to the detection limit [49], which means that the Allan variance can be employed to evaluate the system minimum detection limit.
3. Experimental Setup
According to the HITRAN 2020 database [50], all of the absorption lines of H2 are simulated and shown in Figure 1a. Concerning the interference of other gases in the air, the line strength of carbon dioxide (CO2), water vapor (H2O), and H2 located between 4712 cm−1 and 4713.5 cm−1 are depicted in Figure 1b. We observed that the absorption line of H2 located at 4712.90 cm−1 (2121.83 nm) not only had the maximum line strength (3.19 × 10−26 cm−1/cm−2 × molec) but also avoided the interference from H2O and had one order of magnitude stronger than the line strength of the adjacent CO2. H2 has narrow line widths due to its simple level structure. Therefore, even if the absorption line of H2 is located close to that of other gases, the detection of H2 by the TDLAS method is less likely to be disturbed. From the above analysis, the absorption line of H2 located at 4712.90 cm−1 was chosen and used in the subsequent studies.
Figure 1. Simulations of line strength based on the HITRAN 2020 database. (a) All absorption lines of H2. (b) The absorption lines of H2, CO2, and H2O between the spectral range of 4712.0 cm−1 and 4713.5 cm−1.
The experimental setup of the H2-TDLAS sensing system, based on the WMS technique and HMPC (model no.: PCI-HC10m; number of passes: 34; optical pathlength: 10.13 m), is shown in Figure 2. The cell volume of the used HMPC was 0.24 L, and the window was a wedge mirror of gallium fluoride (GaF2). First, a CW-DFB diode laser with a typical wavelength of 2121.83 nm (Nanosystems and Technologies GmbH, TO5 cube) was adopted as the excitation light source. A low-frequency sawtooth wave and a high-frequency sine wave were generated by a signal generator and lock-in amplifier (Zurich Instruments), respectively, which were superimposed and then sent into the LD controller (LD2TC5 LAB series combination laser diode and temperature control instrument). The LD controller adjusted the injection current and operating temperature of the laser to modulate the laser emission wavelength. The three parameters of the LD controller, proportion, integration, and differentiation, need to be set to appropriate values. The emission wavelength of the LD controller could match the chosen absorption line of H2 when the injection current and operating temperature of the laser were set to 80 mA and 39 °C, respectively. Under these conditions, the output power of the laser was 10.9 mW. The modulated laser beam was collimated by a convex lens with a 5.95 mm focal length and incident into the HMPC at a specific angle. Viewed from the perspective of facing each mirror, the schematic diagram spot patterns, generated by a correctly oriented input beam on the far and near mirrors of the HMPC, were simulated by mathematical software and are shown in Figure 3. The correctly oriented input laser beam came in between position 0 and position 1, and the corresponding exit laser beam came out between position 33 and position 34. Hence, the correct exit laser beam had the same angle in Y and opposite angles in X as the input laser beam. Finally, the exit light was received and converted into an electrical signal by a PD device (model No. DET10D2, Thorlabs). This electrical signal was demodulated at a particular frequency by the lock-in amplifier when the integration time was 200 ms and the low-pass filter order of the lock-in amplifier was set to 4. Considering that the peak value of the 2f is not only located at the absorption peak but also is the largest among even harmonics, the electrical signal received by the PD device was demodulated to 2f signals using a lock-in amplifier in this experiment. In addition, the signal generator also generated impulse waves to trigger the lock-in amplifier monitoring process, which ensured that the complete second harmonic could be obtained in each pulse period.
Figure 2. A schematic diagram of the H2-TDLAS sensing system.
Figure 3. A schematic diagram of the spot patterns on the far and near mirrors of the HMPC with a correctly oriented input beam, from the perspective of directly facing each mirror.
4. Results and Discussion
The intensity of the TDLAS 2f signal is related to the modulation amplitude of the sine wave [45]. Therefore, the relationship between the 2f signal amplitude and the modulation depth was measured when the concentration of H2 was 100%, which is depicted in Figure 4. With the increase in the modulation depth, the 2f signal amplitude increased at the beginning, flattened out, and finally slowly descended. The maximum 2f signal amplitude of 200.91 μV could be obtained when the modulation depth was 0.016 cm−1. This optimum modulation depth was kept for use in the subsequent experiments. In addition, the small value of the optimal modulation depth also proves that H2 has a narrow absorption line width compared to other gases [51,52,53,54,55,56,57].
Figure 4. The relationship between the 2f signal amplitude and the modulation depth when H2 concentration was 100% for the H2-TDLAS sensing system.
In order to evaluate the response performance of the WMS-HMPC-based H2-TDLAS sensing system, H2 with a concentration of 100% and pure nitrogen (N2) were sent into the gas dilution system, respectively, and mixed gases with H2 concentrations of 100%, 80%, 60%, 40%, 20%, 10%, and 5% were obtained by controlling the flow rate of two gas flowmeters in the dilution system. The mixed gas was sent into the gas cell at a constant flow rate of 300 mL/min. The 2f signal for the TDLAS sensor with respect to different concentrations of H2 was measured and is shown in Figure 5. The relationship between the 2f signal amplitude and the H2 concentrations is displayed in Figure 6. These data points were linearly fitted by using the equation y = 1.97x − 8.19, and the R-squared value was calculated as 0.995. Additionally, a calculation of 2f signal repeatability at 100% H2 concentration was performed and is shown in Figure 7. In the measurements that were taken 6 times, the mean value was 202.65 μV and the maximum volatility was less than 1.7%, which indicates a good performance in 2f signal’s repeatability in the H2-TDLAS sensing system.
Figure 5. The 2f signal for the TDLAS sensor with respect to different concentrations of H2 in the H2-TDLAS sensing system.
Figure 6. The linear fitting between the 2f signal amplitude and H2 concentration for the H2-TDLAS sensing system.
Figure 7. The 2f signal repeatability at 100% H2 concentration for the H2-TDLAS sensing system.
As shown in Figure 5, there was a presence of noise in the measurement of the H2-TDLAS 2f signal. Therefore, the wavelet denoising method was applied to further improve the sensor’s performance. A DB wavelet with 10 vanishing moments was employed for 2f signal denoising by 6 times. The 2f signal, with and without the DB wavelet denoising when H2 concentration was 100%, are shown in Figure 8. The data far from the absorption peak in the 2f signal were considered as the system noise. Compared to the original 2f signal, the 1σ noise deviation with the DB wavelet denoising was reduced by half while the 2f signal amplitude was almost identical. Furthermore, the SNR and minimum detection limit (MDL) improved from 248.02 and 0.40% to 509.55 and 0.20%, respectively.
Figure 8. H2-TDLAS 2f signal comparison with and without the DB wavelet denoising when H2 concentration is 100%.
The H2-TDLAS 2f signal was monitored continuously after filling the GMPC with pure N2 to investigate the optimum detection performance and the system stability of the WMS-HMPC-based H2-TDLAS sensing system. Based on these data, the Allan variance of the WMS-HMPC-based H2-TDLAS sensing system was calculated by the LabView software, and the results are shown in Figure 9. As the integration time increased, the MDL decreased. When the integration time of the system was 36 s, the MDL improved to 522.02 ppm. Finally, the properties of some H2 sensing technologies are listed in Table 1. Compared to these H2 sensors, the H2-TDLAS sensor achieves a better detection performance at a very weak H2 absorption line strength (~10−26).
Figure 9. Allan variance analysis for the WMS-HMPC-based H2-TDLAS sensing system.
“