Trace Gas Monitoring by Hollow-Core Anti-Resonant Fiber-Enhanced Raman Spectroscopy with Sub-ppm Sensitivity

Abstract

The demand for accurate and sensitive trace gas detection in environmental monitoring and industrial diagnostics has driven the development of compact, high-performance Raman-based sensing systems. In this study, a hollow-core anti-resonant fiber (HC-ARF)-enhanced Raman spectroscopy system was developed to improve detection sensitivity. A double-lens signal collection module coupled with a small-core multimode fiber (MMF) was designed to improve Raman signal collection efficiency while mitigating background interference. Together with the CCD row-selective integration strategy, this configuration effectively minimized spatially nonuniform noise and enhanced the overall signal-to-noise ratio of the system. The system performance was systematically evaluated under varying integration times, demonstrating linearity, repeatability, and long-term stability. Under optimized conditions, sub-ppm detection limits were achieved for CO₂ isotopic species (¹²CO₂: 5.13 ppm, ¹³CO₂: 0.82 ppm) and multiple hydrocarbons including CH₄ (2.5 ppm), C₂H₂ (2.7 ppm), C₂H₄ (2.84 ppm), and C₂H₆ (0.57 ppm). These results confirm the performance of the proposed configuration for multi-component gas detection. Overall, this work provides an effective strategy for noise suppression in HC-ARF-based fiber-enhanced Raman systems and demonstrates their potential for real-time, high-precision environmental and industrial gas analysis.

1. Introduction

Trace gas detection has become increasingly important in various fields, such as environmental monitoring, industrial process control, and medical diagnostics. With its expanding applications, higher accuracy and sensitivity requirements have been imposed on detection systems. Compared with conventional techniques such as semiconductor sensors, electrochemical sensors, and gas chromatography [1,2], optical spectroscopic methods offer distinct advantages including non-contact measurement, high selectivity, and high sensitivity. Various optical spectroscopic techniques have been extensively investigated and applied. Tunable diode laser absorption spectroscopy (TDLAS) provides excellent selectivity and sensitivity, but it typically relies on specific absorption lines and is therefore limited in simultaneous multi-component detection [2,3,4]. Photoacoustic spectroscopy (PAS) and photothermal spectroscopy (PTS) offer extremely high sensitivity, with detection limits reaching the ppb level, but are limited by their single-component detection capability and complex system configuration [5,6,7,8,9]. In contrast, Raman spectroscopy exhibits unique advantages in multi-component gas detection, as it enables the simultaneous identification of various gas species using a single excitation laser and can effectively detect infrared-inactive homonuclear diatomic molecules such as N₂ and O₂ [10]. Nevertheless, Raman-based gas sensing remains limited by the inherently weak scattering cross-section and the low molecular density of gases.

Various enhancement strategies have been developed to overcome the inherently low Raman scattering cross-section and weak signal intensity of gas molecules [11], thereby improving the detection sensitivity of Raman spectroscopy. Among these methods, cavity-enhanced Raman spectroscopy (CERS) [12], waveguide-enhanced Raman spectroscopy (WERS) [13], and fiber-enhanced Raman spectroscopy (FERS) [14] have demonstrated remarkable potential in signal enhancement. Compared with CERS and WERS, FERS offers superior compactness, optical stability, and integration capability, as it enables long-distance light–gas interactions within the optical fiber, effectively enhancing the Raman signal intensity [15,16]. In particular, hollow-core fiber (HCF) structures provide a highly promising platform for gas-phase Raman detection, as light predominantly propagates through the hollow core filled with the target gas, thereby maximizing the light–molecule interaction and significantly suppressing background noise originating from the silica material.

Typical HCF types include metal-coated capillary (MCC), hollow-core photonic bandgap fiber (HC-PBF), and hollow-core anti-resonant fibers (HC-ARF). William et al. [17] first proposed a MCC system for Raman gas detection, achieving a 30-fold signal enhancement for ambient N₂. The large core diameter enables fast gas exchange, but high transmission loss and fluorescence from the silver layer limit sensitivity. HC-PBF achieves efficient light confinement and high signal collection efficiency through the photonic bandgap effect. Khannanov et al. [18,19] developed portable Raman gas sensing systems coupled to hollow-core photonic crystal fibers (HC-PCFs) with 532 nm excitation, enabling rapid and accurate multi-component analysis of natural gas and gas mixtures, including infrared-inactive species such as H₂, O₂, and N₂. Although HC-PBF enables efficient light confinement and signal collection via the photonic bandgap effect, their small core size results in poor coupling stability, slow gas diffusion, and limited effective sensing length. In contrast, HC-ARF, based on the anti-resonant reflection mechanism, offers low-loss, broadband single-mode transmission and effectively suppress silica background interference, thereby significantly improving the signal-to-noise ratio and detection sensitivity. Knebl et al. [20,21,22] were the first to employ newly developed HC-ARF for Raman gas sensing, successfully demonstrating their applicability in environmental analysis. Overall, these advantages make HC-ARFs an ideal candidate for hollow-core fiber-enhanced Raman gas sensing and a promising direction for achieving high-sensitivity, low-detection-limit Raman spectroscopy.

In this work, a hollow-core anti-resonant fiber (HC-ARF)-enhanced Raman gas detection system incorporating a double-lens signal collection module, a small-core multimode fiber (MMF), and CCD row-selective integration was successfully developed, achieving high spectral fidelity and detection sensitivity. The optimized double-lens coupling configuration effectively enhanced Raman signal collection while suppressing background interference from the fiber and spectrometer. Combined with the CCD row-selective integration strategy, the system significantly improved the signal-to-noise ratio and overall detection stability. Under optimized operating conditions, ppm-level detection limits were achieved for CO₂ isotopic species (¹²CO₂: 5.13 ppm, ¹³CO₂: 0.82 ppm) and multiple hydrocarbons, including CH₄ (2.5 ppm), C₂H₂ (2.7 ppm), C₂H₄ (2.84 ppm), and C₂H₆ (0.57 ppm). These results demonstrate the sensitivity, stability, and multi-component detection capability of the proposed configuration, confirming its strong potential for real-time environmental monitoring, industrial process diagnostics, and compact field-deployable Raman sensing systems.

2. Experimental Setup

A compact forward-collection fiber-enhanced Raman spectroscopy system based on a hollow-core anti-resonant fiber (HC-ARF) was constructed, as illustrated in Figure 1. The system primarily consists of a laser coupling module, a compact gas chamber integrating both optical and gas flow paths, and a double-lens signal collection module. A continuous-wave (CW) all-solid-state 532 nm laser (Cobolt Samba 08-DPL, 0–150 mW, HÜBNER Photonics, Danderyd, Sweden) is first spectrally cleaned by a bandpass filter (BF, Semrock LL01-532E-12.5, Semrock (a part of IDEX Health & Science), Rochester, NY, USA) to remove stray light. The filtered beam is focused by an achromatic lens (AL1, JCOPTIX OLD2436-T2, f = 50 mm, Ningbo JCOPTIX Technology Co., Ltd., Ningbo, China) directly into the hollow-core anti-resonant fiber, where the excitation light interacts with the gas sample. A power meter (S120C, PM100USB, Thorlabs, 56 Sparta Avenue, Newton, NJ, USA) at the fiber end monitors coupling efficiency, reaching 80%. Forward-scattered Raman signal emerging from the HC-ARF is recollimated by a achromatic lens (AL2, JCOPTIX OLD2434-T2, f = 40 mm, Technology Park, Ningbo, China). Residual pump and Rayleigh light are rejected by a dichroic mirror (DM, Chorma RT532rdc, Chroma Technology Corp, Bellows Falls, VT, USA) and a long-pass filter (LF, Semrock LP03-532RE-2, Semrock (a part of IDEX Health & Science), Rochester, NY, USA) and the remaining Stokes signal is focused by achromatic lens L3 (AL3, JCOPTIX OLD2434-T2, f = 40 mm, Technology Park, Ningbo, China) into a multimode collection fiber (MMF, core 25 µm, NA = 0.10). The MMF terminates at the input of the spectrometer (Princeton Instruments HRS-300, Teledyne Princeton Instruments, Trenton, NJ, USA) equipped with a deep-cooled CCD (PIXIS-400BX, Teledyne Princeton Instruments, Trenton, NJ, USA) for spectral acquisition. This compact, direct-coupling arrangement minimizes alignment optics while maintaining efficient excitation and high-fidelity forward Raman collection.

The detailed structure of the HC-ARF gas cell is illustrated in Figure 2. The gas chamber is designed to simultaneously provide optical alignment and controlled gas flow. Both ends of the chamber are terminated with customized fiber adapters (FA₁ and FA₂) that provide mechanical stability and optical sealing. The two adapters are connected by a tubular gas conduit, inside which the HC-ARF is mounted to serve as the optical and gas interaction channel. The HC-ARF is fixed in a standard V-groove fiber holder (LBTEK OFF32) using double-sided adhesive tape and precisely mounted within FA₁ and FA₂ to ensure alignment stability. Each adapter includes a 3-way valve and pressure gauge to control gas inlet/outlet flow and monitor internal pressure. Optical windows, fixtures (1–3) and a seal ring ensure airtight sealing.

The gas-chamber module employs a 4 m-long hollow-core anti-resonant fiber (HC-ARF, STF-VIS 500–650, China Linfiber Tech. Ltd., Beijing, China), featuring a core diameter of 29 μm. The microscopic image of the fiber structure is shown in Figure 3a. Furthermore, the measured transmission spectrum, shown in the inset of Figure 3c, reveals a low-loss window spanning 500–650 nm, which fully covers the excitation wavelength of 532 nm and the corresponding Raman-shifted spectral range. The 4m-long HC-ARF is coiled on a 600 mm diameter spool to minimize bending loss. Finite element simulations using COMSOL 6.2 Multiphysics were conducted to obtain the mode field distribution of the HC-ARF at 532 nm, as shown in Figure 3b. The simulation was performed under typical structural parameters (cladding diameter: 200 μm, wall thickness: 400 nm). Mesh refinement followed standard criteria: λ/5.8 in silica and λ/4 in air, with a perfectly matched layer (PML) applied at the outer boundary to suppress reflection. The simulated fundamental mode exhibits strong confinement within the hollow core, indicating efficient light guidance and minimal overlap with the silica structure, which is beneficial for reducing background Raman scattering.

3. Experimental Details

3.1. Apparatus for Background Noise Suppression

Enhancing the signal-to-noise ratio (SNR) of the HC-ARF-based Raman spectroscopy system requires effective suppression of the spatially nonuniform background fluorescence noise originating from scattering at the fiber coupling interface and within the cladding region [23]. This background, typically concentrated in the outer cladding and surrounding the central Raman signal, severely restricts detection sensitivity and stability.

Accordingly, a double-lens signal collection module was developed to improve Raman signal collection efficiency while mitigating such background noise, as illustrated in Figure 4a. In this configuration, the Raman light emerging from the HC-ARF is relayed by two achromatic lenses (AL₁ and AL₂, focal length ratio of 1:1) to form a~30 μm image at the entrance of a multimode fiber (MMF, 25 μm core diameter) that is directly connected to the spectrometer. Experimental results, presented in Figure 4b, confirm that the double-lens configuration substantially enhances Raman signal intensity while suppressing silica background, demonstrating its superiority in both signal collection efficiency and noise reduction.

To further suppress the spatially nonuniform background noise after optimizing the collection module, additional optical filtering was implemented to remove the residual excitation light. In forward Raman configurations, the transmitted laser and Raman signal co-propagate, and a single long-pass filter is often insufficient to completely reject the strong 532 nm excitation component. This residual leakage can raise the baseline and mask weak gas Raman features. To address this issue, a dichroic mirror was added in series with the long-pass filter to enhance laser suppression without significantly attenuating the Raman signal.

Figure 5a compares the Raman spectra obtained using a single long-pass filter and the combined long-pass + dichroic mirror configuration under identical integration time. The enlarged view of the CO₂ Raman region in Figure 5b (1300–1380 cm⁻¹) provides a clearer comparison of the background noise level. The baseline fluctuations are reduced when employing the dual-filter configuration, and the CO₂ Raman peaks appear more distinct. To quantitatively evaluate the noise performance, the standard deviation of the background signal in the 1298–1350 cm⁻¹ region was used as a measure of noise. The noise level decreased from 248.64 to 33.15 counts when employing the dual-filter configuration, corresponding to an improvement factor of approximately 7.5. This demonstrates that the dual-filter design effectively eliminates unwanted background light, thereby improving the signal-to-noise ratio (SNR) and overall spectral fidelity of the system.

3.2. Spatial Filtering Method

Following optical filtering, spatial nonuniformity originating from the fiber and spectrometer was further reduced by employing a row-selective integration strategy on the CCD detector. This technique enables flexible control of the effective detection area by selectively integrating only the pixel rows containing clean Raman signals, thereby eliminating residual lateral noise. To evaluate the effect of the integration range, Raman spectra were extracted by integrating 3, 5, and 7 CCD pixel rows, respectively, under the following conditions: excitation power of 150 mW, slit width of 20 μm and CCD exposure time of 450 s, and measurements conducted at standard atmospheric pressure (1 atm) and room temperature, as shown in Figure 6a. Increasing the number of accumulated rows enhances the overall signal intensity; however, excessive integration introduces additional background noise from adjacent regions.

As summarized in Figure 6b, the optimal signal-to-noise ratio (SNR) was obtained when integrating 5-pixel rows, which provided a balance between signal intensity and noise suppression. This result demonstrates that selecting an appropriate CCD readout region is crucial for achieving stable and high-SNR Raman detection in hollow-core fiber systems.

4. Experimental Results

To evaluate the temporal performance and detection stability of the proposed hollow-core fiber Raman sensing system, the influence of CCD integration time on the Raman signal quality was systematically investigated. Unlike the spatial filtering optimization discussed previously, this section focuses on the temporal accumulation process during spectral acquisition. Raman spectra of CO₂ were recorded under constant experimental conditions (excitation power of 150 mW, slit width of 20 μm and CCD integration over the central 5 pixel rows) at standard atmospheric pressure (1 atm) and room temperature, while varying the CCD integration time from 10 s to 300 s, as illustrated in Figure 7a.The Raman peaks corresponding to ¹²C¹⁶O₂ (1286 cm⁻¹ and 1388 cm⁻¹, around 396 ppm) and ¹³C¹⁶O₂ (1369 cm⁻¹, around 4 ppm) become progressively more pronounced with longer integration times, indicating an enhanced photon accumulation efficiency. As shown in Figure 7b,c, both signal intensity and noise increased with integration duration, but the growth rate of signal surpassed that of noise, leading to a steady rise in signal-to-noise ratio (SNR). The monotonic improvement of SNR demonstrates that appropriate extension of the integration time effectively enhances detection sensitivity and measurement repeatability in the HC-ARF Raman system.

Under a 300 s integration condition, the ν⁺ vibrational peaks of ¹³CO₂ (1369 cm⁻¹) and ¹²CO₂ (1388 cm⁻¹) were selected for limit-of-detection (LOD) analysis. The corresponding signal intensities were 1856 and 29,161 counts, respectively, with a baseline noise level of 126.09. The calculated signal-to-noise ratios (SNRs) were 14.72 for ¹³CO₂ and 231.41 for ¹²CO₂. Based on these values, the LODs were estimated to be 0.82 ppm for ¹³CO₂ and 5.13 ppm for ¹²CO₂.

The LOD was determined using the standard definition [24]:

$$\mathit{LOD} = {\displaystyle \frac{3C}{\mathit{SNR} }},$$

(1)

where $C$ is the actual concentration of the target gas. Here, the LOD corresponds to the minimum detectable concentration at which the Raman signal equals three times the baseline noise. Notably, the proposed configuration enables direct quantitative Raman measurements without the need for gas pre-concentration or background subtraction. This advantage stems from the strong light–gas interaction within the HC-ARF, which effectively suppresses silica-induced Raman signals, and from the optimized optical and CCD integration design that minimizes system background. Consequently, the obtained spectra feature a clean baseline and high SNR, allowing real-time and straightforward trace gas analysis under ambient conditions.

Building upon the quantitative evaluation of CO₂ detection performance, the developed HC-ARF Raman sensing system was further examined for its capability to detect trace hydrocarbon gases. Prior to gas injection, the gas chamber was evacuated for approximately 15 min to remove residual air and moisture, ensuring a clean optical path and stable baseline. Subsequently, hydrocarbon gases at a concentration of approximately 5 ppm were introduced into the hollow-core fiber for trace-level detection sensitivity assessment. All test gases (CH₄, C₂H₂, C₂H₄, and C₂H₆) were provided as certified mixtures in an argon (Ar) balance by Shanghai Ujali Special Gases Co., Ltd.; The corresponding Raman spectrum, presented in Figure 8, was recorded under an excitation power of 150 mW, a slit width of 20 μm, and a CCD integration time of 300 s, with signal accumulation over the central 5 pixel rows, at standard atmospheric pressure (1 atm) and room temperature. Distinct Raman peaks associated with CH₄ (2917 cm⁻¹), C₂H₆ (2954 cm⁻¹), C₂H₄ (3020 cm⁻¹), and C₂H₂ (1981 cm⁻¹) are clearly identifiable, and a quantitative summary of their spectral parameters—including peak positions, molecular assignments, measured intensities, signal-to-noise ratios (SNRs), and limits of detection (LODs)—is presented in

Table 1. Quantitative Raman detection results of trace hydrocarbon gases (~5 ppm) obtained using the HC-ARF-based Raman sensing system. (The SNR values were calculated based on the average signal intensity and average baseline noise from three independent measurements under identical experimental conditions).

Gas	Gas Concentration/ppm	Raman Shift/cm−1	Signal-To-Noise Ratio	LOD/ppm
CH4	5.1	2917	6.1	2.5
C2H2	5.2	1981	5.7	2.7
C2H4	5.4	3020	5.7	2.84
C2H6	5.2	2954	27.2	0.57

, demonstrating the system’s strong potential for multi-component hydrocarbon detection at trace concentrations.

5. Conclusions

In summary, we have demonstrated a hollow-core anti-resonant fiber (HC-ARF)-enhanced Raman gas detection system integrating a double-lens signal collection module, a small-core multimode fiber (MMF), and CCD row-selective integration, achieving high spectral fidelity, sub-ppm detection limits, and exceptional detection sensitivity. The optimized optical collection design significantly improves Raman signal coupling efficiency while mitigating background interference, and the adaptive CCD integration effectively suppresses spatially nonuniform noise, enhancing detection stability. This system has reliably detected CO₂ isotopic species (¹²CO₂: 5.13 ppm, ¹³CO₂: 0.82 ppm) and several hydrocarbons, including CH₄ (2.5 ppm), C₂H₂ (2.7 ppm), C₂H₄ (2.84 ppm), and C₂H₆ (0.57 ppm). These results highlight the system’s strong capability for multi-component gas detection and its potential for real-time environmental monitoring and industrial diagnostics. Future work will focus on optimizing system sensitivity by extending the fiber length, improving optical coupling, and exploring pressurized conditions for further enhancement of detection limits, along with evaluating different core-diameter MMFs to achieve a better balance between coupling efficiency and background suppression.