1. Introduction

Sulfur dioxide (SO₂) as a well-known and major atmospheric pollution gas is strictly monitored since it has a strong stimulating effect on the mucosa of eyes and respiratory tract. SO₂ is colorless and has a strong pungent smell. Pulmonary edema, laryngeal edema, vocal cord spasm and suffocation can be caused when dangerous amounts of SO₂ are breathed in. Therefore, the SO₂ concentration needs to be monitored in industry, such as wastewater treatment, safety production, especially when operating underground. In the electric-power industry, SO₂ is produced inside the sulfur hexafluoride (SF₆) gas-insulated substation (GIS) when decomposition mechanisms occur during corona discharges and sparks. By online monitoring of the SO₂ gas concentration inside the GIS in situ, the inner condition of the insulation system can be obtained without stopping working and opening the devices. The different SO₂ concentration level indicates the different insulation defect occurring inside the GIS. The SO₂ content at levels ranging from 1 to 5 parts per million by volume (ppmv), latent fault defect may be involved within the GIS. The potential fault for overheating is high when SO₂ concentration is > 10 ppmv. The unavoidable impurity (H₂O vapor) in pure SF₆ will react with SO₂ to generate acidic decompositions, which may corrode the internal surface of the GIS and eventually cause a serious accident [1–4]. Numerous chemical-based detection methods have been used for the operating state diagnostic of SF₆ gas insulated equipment by detecting SO₂. In 2012 [5], a TiO₂ nanotube array sensor was reported for detecting the SF₆ decomposition product. The SO₂ concentration within 10 - 50 ppmv creates a pronounced resistance change in the nanotube array, but the poor detection sensitivity and high working temperature of 200 °C limits the sensor application. In 2021 [6], the adsorption of Sc-doped MoS₂ monolayer for SF₆ characteristic decomposition gases was reported by using a density functional theory (DFT) method. Besides, a Ru-doped MoS₂ monolayer was also investigated as a resistance-type sensor for detecting SO₂ and SOF₂ [7], the complicated sensor structure and poor long-term stability are not beneficial to practical application.

The photoacoustic spectroscopy (PAS) technique is widely investigated recently, due to its extremely high detection sensitivity and selectivity, fast response time, cost-effectiveness and compact absorption detection module [8–12]. The PAS technique is based on the target molecular absorption of the optical energy, which results in the excitation of molecular energy levels (electronic, rotational, and vibrational). The absorption lines are specific for each target molecular like a person’s fingerprint. The excited state loses its energy through radiation processes, in which acoustic or ultrasonic waves are generated instantaneously due to localized transient heating and expansion. A photoacoustic transducer, usually a microphone, a quartz tuning fork (QTF), or a fiber tip, will be used to detect the pressure fluctuations [13–16]. One superiority of the PAS is the wavelength-independent relationship between the excitation wavelength and the PAS detection module. For the detection of SO₂ molecules, the strongest infrared absorption spectroscopy is between 7.2 - 7.6 µm spectral range (ν₃ bands) according to the HITRAN database. A ppb-level detection limit was obtained in a N₂ buffer gas, which is sufficiently sensitive for decomposition monitoring in power equipment [17]. However, the SF₆ concentration in power system is typical > 99.8%. In this case, SF₆ molecules exist dense and strong absorption lines when the excitation wavelength > 6 µm despite these lines are not disclosed in the HITRAN database [18]. Furthermore, the expensive and hulky QCL laser (usually needs a water-cooling chiller) limits its application. For a non-interfering monitoring of SO₂ component, the ultraviolet (UV) spectral region is a superior choice, since the line strengths of electronic transitions are several orders of magnitude stronger than the ro-vibrational transitions in the near-infrared region. The UV lasers emitting at 225.7 nm and 302 nm were employed for the SO₂ detection in N₂ gas but not meeting the carrier gas requirement [19,20]. A ppbv-level SO₂ sensor in SF₆ was reported with a diode-pumped solid-state laser of 303.6 nm [21]. A UV excimer lamp as the excitation for the absorption spectroscopy was established with the light power fluctuating around 21.7 mW [22]. However, a random laser power jitter is inevitable since the immature laser manufacturing technology in UV spectral region. For the photoacoustic gas sensor, the signal amplitude S_a is related to the cell constant C, the incident optical power P₀ and the molecular absorption coefficient α, i.e., S_a= CαP₀ [11]. The excitation laser power instability will result in an unstable photoacoustic signal for a long-term practical application.

In this manuscript, a passive Q-switching laser diode (LD) pumped all-solid-state laser emitting at ∼ 266 nm was used to energetically excite SO₂ molecules. A Bragg acousto-optic modulator (AOM) together with an internal closed-loop feedback controller was adopted as a power stabilization system for a super-stable laser output power. A miniature differential photoacoustic cell (PAC) was designed for photoacoustic wave detection with a resonant frequency of 690 Hz and a high-Q factor of 85, which achieved a ppbv-level detection limit for SO₂ detection in SF₆ gas.

2. Excitation laser source

According to the MPI-Mainz database, SO₂ molecule has three absorption bands in the UV spectral region (170-400 nm). As shown in Fig. 1, A. the second allowed band of 170-240 nm (¹B₂←¹A₁), which is the strongest, B. the first allowed band of 240-340 nm (¹A₂, ¹B₁←¹A₁), and C. a weak forbidden band of 340-400 nm (³B₁←¹A₁), which is the weakest. However, another dominating SF₆ decomposition gas (H₂S) exhibits strong electronic absorption between 160 nm-250 nm. The absorption cross interference of different decomposition components results in a gas signal overlapping, therefore the UV laser wavelength located in the spectral region B is the optimum solution [23].

 

Fig. 1. Three absorption bands of H₂S and SO₂ molecules.

Download Full Size | PDF

3. Photoacoustic sensor system

The photoacoustic signal amplitude is proportional to the excitation power of the laser excitation source. With the development of the frequency-doubling technology, the 266 nm laser output power can be reaching several hundreds of mW even watt level [24]. The stability of the power affects the detection accuracy of the sensing system. However, the output power of commercial UV lasers will fluctuate due to these factors such as operating temperature, mechanical vibration, electromagnetic interference, discharge current, and activation medium gain coefficient.

A frequency conversion of Q-switched laser emitting at 266.4 nm (Changchun new industries, model MOL-F-266) with a max average output power of 30 mW was employed as the excitation source. The pulse repetition rate of the laser current was set as the PAC resonant frequency. The transverse mode of the laser beam is near TEM₀₀, which has a beam quality factor (M²) of < 2. The elliptical beam (4:1) diameter is < 1.2 mm with a full divergence angle of < 1.5 mrad. The average power stability over 4 hours is < 5%, which is a significant disincentive to acoustic signal detection for long-time monitoring. As shown in Fig. 2(a), a closed-loop feedback power stabilization system was designed for a super-stable photoacoustic SO₂ detection in SF₆ buffer gas. The UV laser beam passes through an AOM in the Bragg diffraction configuration with the linear polarization. The AOM can be operated in 240-400 nm, which has a diffraction efficiency of >90%. The AOM driver is commercially available (China Electronics Technology Group Corporation 26 Research Institute), which has a rise and fall time of < 10 ns. The output light was divided into the first-order beam (which was blocked further away) and the zero-order beam. For this experiment, only the zero-order beam is the useful beam, the transmitted beam (∼ 95% power ratio) was employed to detect SO₂ gas by a differential photoacoustic cell (as shown in Fig. 2 (b)), and the reflected beam (∼ 5%) was collected by a silicon avalanche photodiode (Thorlabs, APD440) to act as the feedback control beam. The photodiode voltage signal was converted into a 24-bit digital signal by a custom-made A/D converter (ADS1255). The sample frequency was set as 30 kHz. A power deviation value can be obtained by the actual photodiode digital signal and the setting power digital signal, which was employed to obtain a real-time AOM control voltage signal by utilizing a Field Programmable Gate Array (FPGA) with an incremental PID control algorithm. The AOM control voltage signal was then transmitted to an electrical controller to carry out the servo controls of the power balance between AOM diffraction orders. The zero-order power fluctuation can be alleviated by compensating for the first-order beam power. The laser output power fluctuation was transferred to the first-order power variation by the Bragg diffraction, and the goal of stabilizing the zero-order laser output power of the system was achieved. The used digit circuit exists a distinct advantage over the analog circuit published before [25–28], because it is easy to adjust equipment by editing program. Analog circuit processing analog volume is completed through the circuit structure, to solve the interference, distortion and other problems, the structure is complex and debugging is difficult. Particularly when optical wavelength of the excitation laser source changes, the feedback resistance of the analog circuit should be optimized. The actual photo and the 3D schematic diagram of the developed feedback laser power controller were shown in Fig. 2(c) and (d).

 

Fig. 2. Schematic diagram of photoacoustic SO₂ gas detection system with a closed-loop feedback power controller (a), a differential photoacoustic cell (b) and a lock-in amplifier. (c) and (d) are the actual photo and 3D mode of the laser power stabilization system.

Download Full Size | PDF

The stable zero-order laser beam passed through a Helmholtz two-channel differential PAC to detect the desired SO₂ photoacoustic signal. The length and diameter of each resonator were 11 cm and 0.8 cm. Two buffer volumes with a Brewster angle were used on either side of the resonator to minimize laser beam multiple reflections and to reduce gas flow noise. Two electret condenser microphones were selected to achieve an identical signal response of around 1 kHz and mounted on the middle of each resonator. A custom-made differential pre-amplifier was used to collect the microphone current signals with a gain factor of 13. The voltage signal was fed into a lock-in amplifier (FEMTO, LIA-BVD-150) in the 1f mode. The decomposition photoacoustic signal was collected by the FPGA and then transferred to a computer. The filter slope of the lock-in amplifier was set as 12 dB/oct. The time constant was 1 s, which corresponded to a detection bandwidth of 0.25 Hz.

4. Laser power stability assessment

In order to verify the effect of the system control on the laser power fluctuation, the system output power needs to be quantitatively analyzed. Another silicon avalanche photodiode detector was used to detect the transmitted light beam power. The detector voltage was recorded by a six and a half digit multimeter (Agilent, 34410 A). The acquisition time interval was 1.2 s. Since the direct output data of the detector was voltage value, the voltage value is used to analyze the fluctuation of the laser output power in the following experiments. An evaluation coefficient S was used to evaluate the power stability:

The stability performance of the photodiode detector was first tested for long-term work. The light-sensitive surface of the detector was covered with a black light shield. And the enlargement factor was set as 30 dB. The voltage of the used photodiode detector was continually recorded for about two hours. As shown in Fig. 3(a), a maximal voltage changing value was < 30 µV, which means that the system accuracy was not affected by the detector dark current. Afterward, the photodiode voltage was sampled for 5,000 points in a row before and after starting the 266 nm laser power stabilization system. The comparative data were depicted in Fig. 3(b). A fluctuation coefficient ${S_1}$ of 0.41% was obtained using Eq. 1 for the UV laser power. The maximum voltage fluctuation was 38.3 mV. Commendably, a stabilized coefficient ${S_2}$ of 0.04% was achieved after starting the power stability control system. In this case, the maximum voltage fluctuation was 8.1 mV. The laser power fluctuation was significantly suppressed, and the degree of power instability was down by an order of magnitude.

 

Fig. 3. (a) Photodiode voltage of the detector dark current. (b) The original (without feedback) and stable laser output power voltage.

Download Full Size | PDF

5. Sensor performance assessment

For the sensor performance evaluation, 50 ppmv SO₂/SF₆ gas mixtures and pure SF₆ gas were used to prepare different desired SO₂/SF₆ gas concentrations by utilizing a gas dilution system (Environics Inc., EN 4000). The sensor gas flow and pressure were maintained with a diaphragm pump and a pressure controller (MKS Instruments, 649B). The experimental environment was controlled at room temperature and under standard atmospheric pressure. To obtain an optimal photoacoustic signal, the PAC resonant frequency was measured first, which was proportional to the ratio of the sound velocity υ and the resonator length L (f = υ/2 L). A PAC fundamental longitudinal vibration frequency of 690.4 Hz was obtained with a high Q-factor of 85. The detailed description of the SF₆ buffer gas-induced high Q-factor (> 50) PAC was demonstrated previously [11], since its particular physical property relative to N₂. Therefore, the modulation frequency of the laser current was set as the PAC vibration frequency of 690.4 Hz. The duty cycle of the modulation square signal was 50%.

As shown in Fig. 4, continuous photoacoustic signal values were recorded under different concentrations of SO₂/SF₆ gas mixtures from 5 ppmv to 20 ppmv for 100 s (1 point per 1 s). The standard deviations (1σ) of pure SF₆ gas signals were defined as the sensor noise levels, which was 1.8 µV. For the sensor detection limit analysis, the average signal amplitude of 64.34 µV of 5 ppmv SO₂/SF₆ gas mixture was used. A signal-to-noise ratio (SNR) of 35.7 was obtained, which resulted in a detection limit of 140 ppbv and a normalized noise equivalent absorption (NNEA) coefficient of 3.2 × 10⁻⁹cm^-1WHz^-1/2 by taking into account the laser power and the lock-in detection bandwidth.

 

Fig. 4. Photoacoustic signal amplitudes as a function of five different SO₂/SF₆ gas mixtures.

Download Full Size | PDF

To verify the developed sensor linear concentration response, the 50 ppmv SO₂/SF₆ gas mixtures were diluted with pure SF₆ to generate different concentration levels. The measured average photoacoustic signal amplitudes for 100 points were plotted in Fig. 5 as a function of the SO₂ concentrations. A calculated R-square value of 0.9997 was calculated for a linear fitting analysis, which indicated that the sensor system exhibited a super linear response under different target gas molecules after adopting a power stabilization system.

 

Fig. 5. Response linearity of SO₂ photoacoustic signal amplitudes for different SO₂/SF₆ gas mixtures.

Download Full Size | PDF

6. Conclusions

A ppb-level photoacoustic SO₂ gas sensor in a SF₆ buffer gas was demonstrated by employing a dual-resonator differential photoacoustic cell. A super-stable laser output power at 266 nm was developed using a Bragg acousto-optic modulator (AOM) together with an internal closed-loop feedback controller. A power stabilization factor of ∼ 10 times was achieved with respect to the raw laser power. Such stable laser power was vital for practical application in the long operation. In addition, one of the advantages of the photoacoustic gas sensor is the wavelength-independent relationship between the PAS spectrophone and excitation light source. The power stabilization system can also be used for the different excitation light at another wavelength by adjusting the wavelength dependent devices, such as the AOM, reflector and photoelectric detector. Furthermore, the AOM can provide a power modulation function for a laser source without the modulation capability instead of using an external chopper [29]. The sensor performance with a detection limit of 140 ppbv (< 1 ppmv) and an NNEA coefficient of 3.2 × 10⁻⁹cm^-1WHz^-1/2 were obtained, which was sensitive enough for the safety monitoring in the electrical power system.