1. Introduction

Photonic crystal fibers (PCFs) and photonic bandgap fibers (PBFs) offer new and practical strategies for gas sensing (see [1, 2] and references cited within). In particular, PBFs provide practically free-space beam guidance at path lengths of many meters – much longer than conventional gas cells – and are flexible, robust and allow remote measurements. The fundamental molecular vibration-rotation transitions of a wide range of common gases lie in the mid-infrared (mid-IR) wavelength range from 3–10 µm, therefore detection based on measuring changes in the optical transmission through a gas cell is potentially most efficient in this spectral region. Despite the fact that silica-based PBFs have already been shown to guide mid-IR light at wavelengths as long as 3.25 µm [3], their application in high-resolution spectrometry in the mid-IR has, until now, not been demonstrated. Previous work has implemented high-resolution gas sensing by exploiting the overtone molecular vibrations of methane in the 1–2 µm region and using a PBF that guided in the near-infrared [4]. The absorption strengths of molecular overtone resonances are much weaker than the corresponding fundamental absorptions therefore overtone spectroscopy is inferior in sensitivity to fundamental band spectroscopy. In this paper, we present methane sensing using a femtosecond optical parametric oscillator (OPO) and a PBF fabricated to efficiently guide light in the 3.0–3.5 µm wavelength region which corresponds to the fundamental absorption band of methane.

In earlier work we reported an all-silica PBF for mid-IR transmission beyond 3 µm [3]. More than 98% of the optical field is guided through the air core in such a fiber [5], therefore the overlap between the gas and the optical mode volumes is close to optimum. A cross-section of a silica PBF designed by us for gas-sensing applications is shown in Fig. 1: the diameter of the hollow core was typically 40 µm. The fiber can be filled with any gas, allowing the core to be used as an arbitrarily long gas cell.

 

Fig. 1. Cross-sectional image of a photonic bandgap fiber. The fiber shown exhibits the transmission pictured in Fig. 4. and was used to obtain spectral data shown in Fig. 5a.

Download Full Size | PDF

2. Optical parametric oscillator

The application of PBFs to mid-IR gas sensing requires a means of making high-resolution spectrally-resolved transmission measurements. This requirement is normally met by using narrow-line tunable laser diodes and a point detector [4] or by using a broadband source and a spectrometer. Previously, we demonstrated an implementation of methane sensing using a femtosecond optical parametric oscillator (OPO) and a Fourier transform infrared (FTIR) spectrometer which achieved around 2.6 cm^-1 resolution, sufficient to resolve individual rotational lines in the methane P and R branches [6, 7]. The broad spectral bandwidth, high spatial coherence, high peak power and wide mid-IR tunability of a femtosecond OPO make it a very attractive light source for FTIR fiber-coupled gas sensing and we adopted the same approach in the present work.

We used an OPO based on a 1 mm thick periodically-poled lithium niobate (PPLN) crystal containing gratings with periods from 20.5 to 22 µm. A Ti:sapphire oscillator running at around 800 nm and with a spectral width of 10 nm was used as a pump. The OPO cavity mirror coatings had high reflectivity in the range 900–1100 nm and a 20.5 µm period was chosen to phasematch the OPO for signal pulses in this wavelength range, typically around 1050 nm. Idler pulses were then produced with center wavelengths in the 3.0–3.4 µm region which spanned the fundamental methane absorptions under investigation. Other details are similar to those described in [6] and [7]. Signal and idler tuning was possible by adjusting the end mirror position and thus modifying the OPO cavity length. Very accurate control was obtained using a piezo-electric actuator capable of moving the end-mirror position and changing the cavity length with a 10 nm precision. The measured idler tuning range is shown in Fig. 2 and spectra were recorded for different cavity lengths with a 1.2 µm step size. The maximum extracted average power of the idler output across the wavelength range from 3.0 to 3.4 µm varied from 10 to 2 mW. After the interferometer the average power delivered to the fiber input-coupling lens was typically 500 µW and the interference signal was just below the detector saturation level at this value of power.

 

Fig. 2. Measured OPO idler tuning range in 1.2 µm cavity length steps. Every other spectrum is shown using a dashed line for clarity.

Download Full Size | PDF

3. Measurement procedure

The spectral measurement and processing was performed using a FTIR spectroscopy technique which is a commonly used method known for its potential high resolution and sensitivity. Compared with grating spectrometry, which was used previously with a thermal broadband light source to investigate methane absorption in PBF [8], the FTIR technique can provide sufficient spectral resolution to clearly resolve individual rotational lines in the methane absorption spectrum. As indicated in Fig. 3, the FTIR spectrometer consisted of two main parts: a Michelson interferometer and a point detector. Scanning the length of one interferometer arm allowed us to obtain a first order autocorrelation trace of the idler pulses by using a PbSe detector. The idler pulse spectrum is the Fourier transform of the trace acquired through a single scan and the spectral resolution of the idler measurement is proportional to the physical length of the interferometer scan. Light from a He-Ne laser was coupled into the same interferometer (see Fig. 3), and the fringes detected by a silicon photodiode at the output of this auxiliary interferometer were used to calibrate the delay axis of the idler autocorrelation profile. We used a loudspeaker-based linear translation stage driven by a sinusoidal signal. The He-Ne calibration method allowed us to use data collected even during the nonlinear parts of the scan range. After the interferometer, wavelengths shorter than 1.7 µm were blocked by a germanium filter and the idler beam was either steered directly onto the PbSe detector to measure the idler spectrum, or coupled into a 80 cm-long photonic bandgap fiber using a ZnSe lens and detected immediately afterwards. The beam line used to measure the idler spectrum before the fiber is indicated by the dashed elements in Fig. 3.

 

Fig. 3. Experimental configuration. OPO idler pulses (3.0–3.4 µm) leave the cavity through mirror M1 and are then collimated with a CaF₂ lens before entering a Michelson interferometer. After the interferometer the pulses are coupled into the fiber and detected with a PbSe photodiode or steered directly onto the detector, omitting the fiber (dashed beam path). Mirrors M1–M6 have high reflectivity from 900–1100 nm and high transmission at other wavelengths. Mirror M1 is coated on a CaF₂ substrate. PZT, piezoelectric translator; BS, 50:50 mid-IR beamsplitter; Ge, uncoated germanium filter; ZnSe, zinc selenide lens; PBF, photonic bandgap fiber.

Download Full Size | PDF

Femtosecond OPOs are synchronously-pumped systems and so are subject to fluctuations in the center wavelength of the resonant pulses which are caused by small changes in the cavity lengths of the OPO or the pump laser. As we did not have the facility to simultaneously record the spectrum of the idler pulses traveling through the core when it was both empty and filled with methane, an auxiliary spectral measurement of the stray second harmonic of the signal radiation at around 530 nm was used to monitor the idler wavelength. In this way we ensured the same idler conditions for the empty and filled fiber by storing a reference spectrum of the signal second harmonic pulses for one measurement and adjusting the piezo-actuator on the OPO end mirror if necessary to obtain the same idler spectrum during the measurements.

First the transmission of six fibers with a 40 µm-core diameter and a cladding pitch (distance between two neighboring holes) between 6 and 7 µm, was measured to determine which one was the most suitable for methane sensing. The OPO tunability was essential to span the transmission windows of the different fibers. The measurement procedure comprised an acquisition of idler spectrum (using the detector configuration shown by a dashed line in a Fig. 3) and a subsequent acquisition of spectrum transmitted through the fiber. Such pairs of measurements were recorded at two or three tuning positions to span entire window and a fiber transmission was inferred from each pair. The individual transmission results were then concatenated to form the final transmission profiles shown in Fig. 4.

 

Fig. 4. Spectral transmissions of six different fibers having core diameters and cladding pitches similar to the example pictured in Fig. 1.

Download Full Size | PDF

Based on the transmission measurements shown in Fig. 4 we concluded that the bandgap of fiber (b) was the best match to the methane absorption spectrum. We therefore terminated this fiber at both ends with specially designed valves in order to allow gas to be introduced into the fiber in a controlled way. The valves were equipped with CaF2 windows to allow the coupling of the idler beam into the fiber. Any remaining air in the core was pumped out before filling the fiber with a 5:95 methane:nitrogen mixture under a relative pressure of about 1 bar for spectroscopic measurements. A second measurement was performed when the fiber was filled with pure nitrogen to provide a reference spectrum in the absence of methane. Each complete idler measurement consisted of two data sets: an idler interferogram, providing the idler spectrum after a Fourier-transformation, and a He-Ne laser interferogram, recorded simultaneously with the idler interferogram and used to determine the exact delay-axis of the idler data. The full-width half-maximum bandwidth of the idler pulses from the OPO was typically 100 nm, so it was possible to cover the appropriate part of the fiber transmission window with one spectrum only, and no further concatenation of data was necessary.

4. Results

Figure 5(a) shows two spectra recorded for the core containing only nitrogen (thick line) and the methane:nitrogen mixture (thin line). The complicated structure of the reference spectrum originates from the transmission properties of the fiber used. In Fig. 5(b) we present the experimental transmission profile obtained by taking the ratio of these two spectra (thick line) and we compare this with the corresponding calculated transmission spectrum for a 5:95 methane:nitrogen mixture (thin line). The calculated spectrum was based on data from the Hitran database [9] and simulated a measurement with a resolution of 3.3 nm which is close to our estimated experimental resolution of 3.1 cm^-1.

There are certain differences between the observed and calculated spectra, and some of these may arise because we were operating in a spectral region in which the fiber transmission profile contained steep edges, and modulations at frequencies close to those in the methane spectrum. Under these conditions, small OPO fluctuations occurring between the acquisitions of the measurement and reference spectra can account for such inaccuracies. Modifying the system to include twin fibers of equal lengths and two detectors would allow the measurement and reference spectra to be obtained simultaneously. In this arrangement one fiber would be filled with the methane:nitrogen mixture and the other with nitrogen only. The idler beam would be shared between both fibers, and by simultaneously recording interferograms after both fibers one would obtain common-mode noise rejection, ensuring that the artifacts from the OPO noise and fiber structure would be automatically removed.

 

Fig. 5. Experimental results. (a) Idler spectra recorded after the fiber. Thick line, fiber filled with pure nitrogen; thin line, fiber filled with 5:95 methane:nitrogen mixture. (b) Experimental (thick line) and calculated (thin line) methane transmission profiles.

Download Full Size | PDF

Apart from improving the quality of the gas absorption profile with the use of a pair of identical fibers and detectors, we would like to highlight some further enhancements which could be introduced. Certainly, the resolution can be easily improved with commercially available FTIR sets to a level below 1 cm^-1. Furthermore, recent results still in preparation indicate sensing at methane concentrations as low as 0.1%, implying a 50-fold improvement in potential sensitivity compared to the results described in this paper. Other enhancements such as using a longer fiber, using perforated fibers, or multiple short lengths of fibers could improve the sensitivity by allowing the light to interact with a bigger volume of gas.

5. Conclusions

In this proof-of-principle experiment, by combining FTIR spectrometry and photonic bandgap fibers, we have shown that mid-IR gas sensing can be carried out in a convenient spectral region containing fundamental molecular absorption lines. This can be done at spectral resolutions of the order of single nanometers while also benefiting from the versatility of using a PBF as a flexible and arbitrarily long gas cell. Commercially available Fourier-transform spectrometers could be applied to significantly improve the resolution by providing a longer scan range than was used in our apparatus. A common-mode noise rejection technique based on the simultaneous measurement of both the reference and absorption spectra would make the system significantly less vulnerable to OPO instability. Longer fibers should increase sensitivity, however further work is still needed to identify the best way of introducing the gas into the fiber, ideally allowing it to passively diffuse into the core itself, so avoiding the active gas handling used in this work.