Open Access
Add to BibSonomy
Abstract
Supercontinuum sources based on intrapulse difference frequency generation (IDFG) from mode-locked lasers open new opportunities in mid-infrared gas spectroscopy. These sources provide high power and ultra-broadband spectral coverage in the molecular fingerprint region with very low relative intensity noise. Here, we demonstrate the performance of such a light source in combination with a multipass cell and a custom-built Fourier transform spectrometer (FTS) for multispecies trace gas detection. The light source provides a low-noise, ultra-broad spectrum from 2–11.5 µm with ∼3 W output power, outperforming existing mid-infrared supercontinuum sources in terms of noise, spectral coverage, and output power. This translates to an excellent match for spectroscopic applications, establishing (sub-)ppb sensitivity for molecular hydrocarbons (e.g., CH4, C2H4), oxides (e.g., SO2, NOx), and small organic molecules (e.g., acetone, ethyl acetate) over the spectral range of the supercontinuum source with a measurement time varying from seconds to minutes. We demonstrate a practical application by measuring the off-gas composition of a bioreactor containing an acidic ammonia-oxidizing culture with the simultaneous detection of multiple nitrogen oxides (NO, NO2, N2O, etc.). As the different species absorb various parts of the spectrum, these results highlight the functionality of this spectroscopic system for biological and environmental applications.
© 2024 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Most molecules have their strongest vibrational absorption bands with unique spectral features in the mid-infrared wavelength range (2–12 µm). This region covers both parts of the molecular fingerprint region and the functional group region. Broadband, high-resolution absorption spectroscopy in this wavelength region provides the opportunity to detect a large number of species in the gas phase simultaneously and with high detection sensitivity. Traditionally, Fourier transform infrared spectroscopy (FTIR) setups based on thermal light sources have been used for this purpose. However, thermal sources have a low spectral brightness and an omnidirectional light beam. These properties make it difficult to produce a powerful and well-collimated beam to increase the interaction path length in the gas sample, e.g., in a multipass cell (MPC). In addition, the low spectral brightness of the thermal sources usually requires cryogenically cooled infrared detectors to reduce detector noise. Furthermore, achieving a spectrum with a high signal-to-noise ratio (SNR) and a high spectral resolution – for example 0.1 cm-1, which is needed for the detection of certain atmospheric gas species – usually translates into a long averaging time on the order of tens of minutes.
Recently, broadband and ultra-broadband supercontinuum (SC) sources based on highly nonlinear fibers have been developed whose spectra are extending deep into the mid-infrared region [1,2]. Their performance and usability, in combination with different types of spectrometers, have been demonstrated in infrared molecular spectroscopy on mixtures of gaseous species [3–5] with various applications [6,7]. These SC sources operate by pumping highly nonlinear fibers with long ps to ns light pulses, usually generated by semiconductor lasers. Typically, a cascade of fibers made of different soft glasses is used to extend deeper into the mid-infrared wavelength range. Although these SC sources are rather inexpensive and easy to use, they demonstrate high relative intensity noise (RIN) that arises during the noise amplification in the nonlinear processes [8,9]. More specifically, ultra-broad SC sources with an even greater number of cascaded nonlinear fibers demonstrate higher RIN levels that limit the SNR of the obtained spectra in standard spectroscopy systems. A major part of the RIN can be reduced by using alternative detection schemes, such as lock-in amplification [4], boxcar averaging [10], and balanced detection [5]. Note that short fs light pulses, e.g. from mode-locked lasers, can also be incorporated in these types of SC sources to generate a very low noise SC by coherent spectral broadening [11,12]; however, the low noise operation imposes several restrictions on the pulse peak power, pulse duration, and fiber lengths. These restrictions effectively limit the achievable bandwidth and power of the SC light generated by coherent spectral broadening from nonlinear fibers [13], making them less attractive for ultra-broadband spectroscopy.
In addition to fiber-based SC sources, high-power ultra-broadband or widely tunable broadband spectra can be generated in the mid-infrared region using mode-locked lasers and nonlinear conversion in a crystal by means of optical parametric oscillation (OPO) [14] and difference frequency generation (DFG) [15]. These sources demonstrate much less RIN compared to SC sources based on highly nonlinear fibers pumped by semiconductor lasers; however, they are bulky and more expensive. For spectroscopic applications, one can use these sources in combination with a broadband spectrometer [16–20] or use two of them in a dual-comb configuration providing very short single-shot measurement times without moving parts [21–23]. Traditionally, OPO-based sources demonstrate higher power than DFG-based sources, but they are more complex and need a resonance cavity.
Recently, high-power DFG- and intrapulse DFG-based (IDFG-based) systems have been developed that extend deep into the mid-infrared wavelength range [24,25] and have shown an excellent performance for molecular absorption spectroscopy. Both in traditional (linear) dual-comb spectroscopy (DCS) [26–28] and electric field sampling DCS [29,30], their potential in high resolution and precision ultra-broadband spectroscopy was demonstrated. Here, we use an IDFG-based SC source with a particularly interesting ultra-broad spectrum from 2–11.5 µm [31] in combination with our custom-built FTS [5] and a multipass cell that allows us to perform spectroscopic measurements over the full range of the source. We demonstrate the capabilities of the system for simultaneous multispecies trace gas detection and compare the performances of the IDFG-based source with the established fiber-based SC sources. We also demonstrate the possibilities for real-life applications by analyzing the gas composition in the off-gas of a bioreactor containing an acidic ammonia-oxidizing bacterial culture.
2. Methods
The IDFG-based SC source is shown in detail in Fig. 1 and as part of the entire setup in Fig. 2. It consists of a mode-locked master oscillator (MO) and a single-pass power amplifier (PA), both based on polycrystalline Cr:ZnS gain elements and optically pumped by off-the-shelf Er-doped fiber lasers [28,30,31]. The output of the Cr:ZnS MOPA (∼19 fs, 50 nJ pulses at 2.4 µm, repetition rate fr = 80 MHz) is coupled to the IDFG stage based on a ZnGeP2 (ZGP) crystal. The resulting output of the source is a ∼3 W continuum spanning a wavelength range between 2–11.5 µm (see Fig. 3), with ∼300 mW in the 7–11 µm wavelength range. The main distinctive feature of the source in comparison to the established SC sources is the use of bulk polycrystalline Cr:ZnS laser medium providing the direct access to ∼2-cycle pulses with >2 MW peak power at the wavelength 2.4 µm. According to recent studies [32], these pulses are very well suited for IDFG in non-oxide nonlinear materials with broad IR transparency and high figures of merit, e.g. ZGP and GaSe. Another important feature of IDFG with the few-cycle 2.4 µm pulses is their significant augmentation and red-shift during the propagation through the nonlinear crystal. Therefore, the output SC spectrum consists of a long-wave infrared IDFG component that is merged with the augmented fundamental component. These two superimposed components have slightly different spatial distributions, beam divergences, and amplitude noises. Moreover, in their overlapping region between 4 and 7 µm, they beat at the offset frequency of the master oscillator (f0). The frequency f0 can in turn be set to any value 0 ≤ f0≤ fr/2 via the control of the oscillator’s pump power [31].
Fig. 1. Internal structure of the SC source, EDFL: Erbium-doped fiber laser, MO: master oscillator, PA: power amplifier, IDFG: intrapulse difference frequency generation, HR: high reflectors, OC: output coupler, OAP: off-axis parabolic mirror.
Fig. 2. Experimental setup, consisting of the SC source, MPC, and FTS. He-Ne: helium-neon reference laser, M: mirror, BS: beam splitter, RR: retroreflector, TS: motorized linear translation stage, PD: photodetector.
Fig. 3. Power spectral density of the IDFG-based SC source as measured by the FTS at 3 GHz spectral resolution. The dips shown in the spectrum are absorption lines from water vapor (∼5 –8 µm, and ∼3.3 µm) or carbon dioxide (∼4.3 µm) in the surrounding air.
The free-space output of the IDFG-based SC source is directed to a Herriott multipass cell (MPC, Thorlabs HC30 L/M-M02) with an optical path length of 31.2 m and then to a custom-built Fourier transform spectrometer (FTS), as illustrated in Fig. 2. The FTS design has been described in our previous work [5]. The CaF2 windows of the MPC were replaced by uncoated ZnSe windows to be able to transmit light over the whole spectral range of the source. The FTS is based on a Michelson interferometer, in which the incoming light is split into two arms by a beamsplitter (BS, Thorlabs BSW711). The beam in each arm is reflected by a cubic retroreflector that is mounted back-to-back on the same motorized linear translation stage (TS, Thorlabs DDSM100/M) to double the optical path difference (OPD). In this way, moving the stage over 2.5 cm results in an OPD of 10 cm between the two beams, which corresponds to a spectral resolution of 0.1 cm-1 (or 3 GHz). The reflected beams recombine on the beamsplitter and propagate to two photovoltaic detectors (PD 1 and PD 2 in Fig. 2, Vigo Photonics PVI-4TE-10.6), which have a spectral response between 3 and 11.5 µm. Two pinholes in front of the detectors were used to prevent saturation. Using balanced detection, the two signals are subtracted by a differential amplifier to decrease the RIN of the SC source. The obtained stream of data is then digitized with 16-bit resolution using a data acquisition card (National Instruments PCIe-6361). To track the OPD during the measurement, the output beam from a stable reference He-Ne laser is sent along the beam path of the SC source through the FTS and is detected by a separate photodetector (PD 3 in Fig. 2, Thorlabs PDA8A2). The interferogram of the reference He-Ne laser is used to calibrate the OPD. In general, the similarity of the spectrometer used in this work with the one from our previous work [5] allows an objective comparison between the performances of different SC sources and their potential for spectroscopic applications.
During the measurements, the gas flow towards the MPC was controlled using mass flow controllers (Bronkhorst EL-FLOW FG-201CV and F-201CV) and the pressure was regulated to 1 bar using a back-pressure regulator (Bronkhorst EL-PRESS P-702CV). The FTS was set to a spectral resolution of 0.1 cm-1 (3 GHz), which is sufficient for resolving spectral features of the small gaseous molecules at atmospheric pressure (1 bar) [33]. Background spectra to determine the baseline laser intensity (as presented in Fig. 3) were measured by flushing the MPC with nitrogen gas. Reference spectra are fitted to the acquired absorbance spectra using a least-square fit with simulated spectra from the HITRAN [34] or PNNL [35] databases. For the HITRAN spectra, we used a Voigt profile to account for Doppler and pressure broadening, and a sinc function convolution to account for the instrumental line shape due to the finite OPD window of the FTS (boxcar apodization). In the fit, baseline offsets were compensated using a constant offset (DC-offset), or up to a 3rd-order polynomial offset in the case of severe baseline drifts.
We demonstrated the capability of the system with the IDFG-based SC source by detecting various nitrogen-containing compounds including NOx species, as well as CO2, in a biological application. Gas samples were taken from the gas outflow of a bioreactor containing an enrichment culture of an acid-tolerant ammonia-oxidizing bacterium [36]. The bioreactor was supplied with a mineral medium at a constant flow rate of 1 L day-1, consisting of 70 mM NH4Cl, 1 mM NaCl, 0.8 mM MgSO4•7 H2O, 1 mM K2HPO4, 0.3 mM CaCl2•2 H2O, and 1 mL each of trace element solutions I and II [36]. The bioreactor was operated at room temperature and stirred at 850 rpm; pH 4 was maintained by dosing of 1 M KHCO3 solution via an automatically controlled pump. The working volume of the bioreactor was maintained at 2.2 L using a level sensor controller, with biomass retention via a membrane filter. In addition, biomass was removed continuously at a rate of 0.2 L day-1. The bioreactor was flushed with a mixture of 2% CO2 in air at a flow rate of 300 mL min-1. For gas analysis, the off-gas was captured in a 3 L Tedlar bag and transferred to the MPC.
3. Results
The unique combination of a high power, ultra-broadband IDFG-based SC source with low RIN makes it very well suited for sensitive multispecies trace gas detection. Firstly, we measured individual gas species, such as CO2 around 4.8–5.0 µm and 10.1–11.0 µm, and SO2 around 7–7.5 µm, next to a complex gas mixture of acetone, methanol, ethyl acetate, and ethylene between 7–11.5 µm, to demonstrate the ability of the system to retrieve the concentrations of species with both narrow and broadband absorption features in different spectral regions. Secondly, the linear response curve and detection limits have been determined for ethylene at 10–11 µm. Thirdly, to characterize the setup and quantify its performance compared to other SC-based FTS setups, we have assessed the sensitivity of the setup for different gases in various parts of the spectrum using Allan-Werle deviation analysis. The noise equivalent absorption sensitivity (NEAS) of the system was characterized. Finally, to demonstrate the applicability of the system, a gas sample from a bioreactor containing highly enriched ammonia-oxidizing bacteria was analyzed.
3.1 Single and multispecies gas detection
To demonstrate the performance of the system, we measured the spectrum of CO2, consisting of individual absorption lines, and that of SO2, which has a broadband absorption spectrum with many overlapping lines. The absorbance spectrum of a certified mixture of 4.9 ± 0.1% CO2 in synthetic air (Linde Gas) was measured simultaneously in two wavelength regions, from 4.8–5.0 µm and 10.1–11.0 µm, and averaged for 100 seconds. The result is shown in Fig. 4 along with the HITRAN-based fitted spectrum of CO2. The retrieved concentration of the fit is 5.04 ± 0.07%, where the uncertainty is determined by the standard deviation of 5 consecutive measurements. Points with saturated absorption were not included in the fit. Apart from some remnants of very strong lines of atmospheric water vapor removed during the fitting procedure, no residual CO2 features can be detected in the residual of the fit, indicating an accurate fit. It should be noted that the absorbance strength of CO2-lines at 10.4 µm is an order of magnitude weaker compared to those around 4.8 µm. This poses no challenge to the system, which can detect both CO2 bands simultaneously.
Fig. 4. Absorbance spectrum of 4.9 ± 0.1% CO2 in N2 (in black) measured in 100 seconds, with the simulated spectrum using HITRAN of CO2 (red, shown inverted). The residual of the fit is shown in the bottom.
The spectrum of a mixture of 32.6 ± 0.7 ppm SO2 in N2 (Linde Gas) was measured at 7–7.5 µm in 28 seconds. To optimize the spectral power in these wavelengths, a 6.7 µm optical longpass filter was inserted in the optical beam path. The obtained absorbance spectrum is shown in Fig. 5 along with the HITRAN-simulated SO2 spectrum. A concentration of 32.1 ± 0.12 ppm was obtained, with the uncertainty based on the standard deviation of 10 consecutive measurements. The small features remaining in the residual can be attributed to atmospheric water absorbance lines which are close to 100% absorption in both the measurement and background spectrum (i.e. the spectrum without the SO2 mixture). Saturated absorption line profiles are difficult to model and fit properly in the spectrum. Since the remainder of the residual is largely flat, it follows that SO2 can be accurately fitted to the simulated spectrum. The spectrum of SO2 in this spectral region has been investigated previously by a state-of-the-art chalcogenide-based SC source [5].
Fig. 5. Measured absorbance spectrum of 32.6 ± 0.7 ppm SO2 in N2 (in black) with the simulated spectrum containing SO2 (in red, inverted). Water lines have been removed for clarity. Marked in grey is the noise dominated part of the residual used for sensitivity calculations.
The sensitivity of a system was assessed by considering the standard deviation (σ) of a noise-dominated (thus flat) part of the residual (marked grey in Fig. 5) and normalizing it to the measurement time (T): Sensitivity = $\sigma \sqrt {2T} $.
It results that the sensitivity of the system using the IDFG-based SC source for SO2 detection is 0.032 Hz-1/2. As the sensitivity of the chalcogenide-based SC source in an identical setup with MPC and FTS was 7.8 Hz-1/2 [5], a comparison between the two systems indicates an improvement in sensitivity for this spectral region of over 240x in favor of the IDFG-based SC source.
The strength of covering such a broad wavelength range between 7–11 µm is the ability to detect multiple gases simultaneously, especially when these gases exhibit broad and overlapping absorbance features. To demonstrate this, we used mass flow controllers (Bronkhorst EL-FLOW FG-201CV and F-201CV) to generate a dynamic mixture of acetone (20.0 ± 1.0 ppm), ethyl acetate (19.0 ± 0.4 ppm), ethylene (19.6 ± 0.4 ppm), and methanol (19.6 ± 0.4 ppm) in N2. To retrieve the concentration from the spectrum, reference spectra were taken from the HITRAN (for ethylene) and PNNL database (for the other gases). The measured and calculated spectra are shown in Fig. 6. The retrieved concentrations from the fit are 20.1 ± 0.11 ppm ethyl acetate, 20.2 ± 0.6 ppm ethylene, 20.1 ± 0.8 ppm methanol, and 17.2 ± 0.3 ppm acetone, in which the uncertainty is given by the standard deviation of 10 consecutive measurements (each spectrum averaged for 190 seconds). In Fig. 6, it can be clearly noticed that the very broad absorption bands of ethyl acetate overlap significantly with the absorption band of methanol and those of acetone, but thanks to the wide spectral range it is still possible to determine the concentration of all gas species, even though the measured acetone concentration is lower than expected. Since acetone is known to have strong wall adsorption properties (being a so-called “sticky gas molecule”) and the fit of acetone leaves no visible remaining features in the residual, we are confident that the retrieved concentration of acetone is accurate and the lower concentration measured is due to the partial condensation of acetone on the tubing or walls of the MPC.
Fig. 6. Measured absorbance spectrum of a mixture of acetone, ethyl acetate, ethylene, and methanol (each around 20 ppm) in N2 (in black) and corresponding reference spectra from HITRAN (ethylene) or PNNL (others) (in color, shown inverted).
Furthermore, the small visible peaks in the residual between 1350 and 1400 cm-1 can be attributed to water vapor, whose concentration seemingly varied in the atmosphere during the measurements. The increased noise visible in the residual at the peak intensity of ethyl acetate (1230–1250 cm-1) and the Q-branch of ethylene (950 cm-1) is caused by saturation of these strong absorbances. However, this does not affect the ability to retrieve accurate concentrations of the species, thanks to the broadband nature of the source that allows to measure the weaker absorption bands of these species as well.
3.2 System linearity, detection sensitivity and long-term stability
The dynamic range of the system can be demonstrated by characterizing the linear response of the system with dilution series of a gas. Ethylene (C2H4) is chosen because of its specific absorption spectrum at the end of the spectrum of the SC source at 10 µm. The concentration was varied between 25 ppb and 1 ppm using dilutions from a certified and calibrated mixture of 1.04 ± 0.05 ppm ethylene in N2 (Linde Gas). The measured concentrations are shown for each of the applied concentrations in Fig. 7(a), with the uncertainty in dilution resulting from the MFCs (which have ±0.5% reading and ±0.1% full scale uncertainty), and the measurement uncertainty resulting from the standard deviation of 8 consecutive measurements (each averaged for 190 seconds). In Fig. 7(b), the absorbance spectrum of 1 ppm C2H4 is shown, with the corresponding HITRAN spectrum of C2H4 in red (inverted). The excellent match between measurement and simulation is shown by the flat and low noise in the residual. A linear fit yields an R2-value of 0.9994. To demonstrate that the linearity and dynamic response of the system extends even further to almost 5 orders of magnitude, two dilutions of 49 and 98 ppm ethylene were measured and averaged for 190 seconds from a certified, calibrated mixture of 98 ± 2 ppm ethylene in N2 (Linde Gas) and shown in Fig. 7(a).
Fig. 7. a) Linear response of the system to a dilution series of C2H4 (ethylene). The retrieved linear fit on the data points between 25 and 1000 ppb (shown in red) yields an R2-value of 0.9994, b) The measured absorbance spectrum of 1 ppm C2H4 (in black) with corresponding reference spectrum from HITRAN (shown in red, inverted) and residual of the fit.
To determine the long-term stability of the system, the optimal averaging time, and the detection sensitivity, we have assessed the Allan-Werle deviation in two different spectral regions of the source. The first region of interest is around 3–3.5 µm, where most hydrocarbons have strong absorbance bands related to their C-H stretching vibrational modes. Moreover, it is a wavelength range covered by commercially available ZBLAN fiber-based SC sources [10]. Therefore, we made a comparison between the detection sensitivity of the system equipped with the IDFG-based SC source, and with a commercial (ZBLAN) fiber-based SC source [37]. The second region of interest is at 10–11 µm, as it is an interesting part of the molecular fingerprint region with strong distinctive absorption bands of many molecules. Moreover, to this date, this spectral region has been unexplored for spectroscopic research with SC sources.
The Allan-Werle plots in Fig. 8 were obtained by measuring spectra of pure nitrogen gas for several hours. Noise equivalent absorbance spectra were obtained by using the first measured spectrum of nitrogen as background for the rest of the measurements. In the case of the obtained Allan-Werle deviation curve near 3 µm (Fig. 8(a)), a 3–3.5 µm bandpass filter (Thorlabs FB3250-500) was placed in the beam path to enhance the spectral brightness for this wavelength region (similar to [37]), and methane was fitted to these spectra in order to acquire noise equivalent concentrations (NECs) for methane in this spectral window. For the long-wavelength range of the IDFG-based SC source, we repeated a similar experiment with an 8–12 µm bandpass filter (Edmund Optics 11-999) and fitted ammonia in the spectral window 900–1250 cm-1 (8–11 µm) to determine the NEC of ammonia, resulting in the Allan-Werle plot shown in Fig. 8(b).
Fig. 8. Allan-Werle plot of the NEC of a) methane around 3–3.5 µm for this work (in blue) and previous work (in black) [37], and b) ammonia around 8–11 µm, with the τ-1/2-dependency of white noise (dashed line).
The white noise can be reduced over 103 seconds as it follows a τ-1/2-dependency, yielding an optimum averaging time of ∼1000 seconds with a detection sensitivity of ∼200 ppt for methane. This is a 60-fold improvement over commercial ZBLAN fiber-based SC sources [37], which is a remarkable result, as that the main strength of the IDFG-based SC source is its much broader spectrum extending further in the mid-infrared wavelength region. The Allan-Werle deviation curve of ammonia starts deviating from the τ-1/2-dependency after about 100 seconds, indicating that the stability of the source at the extreme end of its long-wavelength emission is suffering slightly more from long-term drifts. However, a detection limit of 6.2 ppb ammonia in 500 seconds can be achieved.
3.3 Noise-equivalent absorption sensitivity (NEAS)
An elegant way to compare the performance of the system with the IDFG-based SC source to systems with other SC sources is the noise-equivalent absorption sensitivity (NEAS). The NEAS is defined as
in which σ is the standard deviation of the noise in a normalized absorption spectrum, T the measurement time of this spectrum (e.g., 2 seconds for the sample measurement plus 2 seconds for the background measurement), L the effective interaction path length (or optical path length, 31.2 m) and M the number of spectral elements (given by the width of the obtained spectrum divided by the resolution of the spectrometer). The noise behavior of the IDFG-based SC source is not identical over its full spectral range. As mentioned in the Methods section, the full spectral coverage of the source is a combination of two co-propagating, superimposed beams. The first beam covers a spectral range between 2–7 µm, while the second beam covers a range between 4–11.5 µm, with the two beams having a spectral overlap in the 4–7 µm region. As the two beams exhibit different RIN levels, two different figures for the NEAS can be calculated: one covering the entire spectrum with the higher associated RIN, and another one in which only the spectrum in the longer wavelengths is considered with the lower associated RIN. For the entire spectrum, which has 27500 spectral elements (2750 cm-1 / 0.1 cm-1) and a σ of 5.1 ${\times} $ 10−2, the NEAS was found to be 2.7 ${\times} $ 10−7 cm-1 Hz-1/2 per spectral element. For the long wavelength range (using a 6.7 µm longpass filter, providing 6500 spectral elements and a σ of 9.1 ${\times} $ 10−3), the NEAS was calculated to be 9.8 ${\times} $ 10−8 cm-1 Hz-1/2 per spectral element.In comparison to other systems [5], the current system with IDFG-based SC source outperforms the other SC-based FTS systems (NEAS of 9.9${\times} $ 10−7 and 5.3${\times} $ 10−6 cm-1 Hz-1/2 per spectral element for short and long wave mid-infrared wavelength range, respectively) by one to two orders of magnitudes. Moreover, the performance aligns with other fs-laser-based systems, with the clear advantage that the IDFG-based system presented in this study can perform over a much wider spectral range.
3.4 Application: bioreactor off-gas samples
To demonstrate the applicability of the system, we analyzed the off-gas from a bioreactor containing an enrichment culture of acid-tolerant ammonia-oxidizing bacteria [36]. The reactor content was constantly stirred and purged with a mixture of 2% CO2 in air at a flow rate of 300 mL min-1. In this study, the off-gas of the bioreactor was collected and measured to test for nitrogen oxides.
During canonical ammonia oxidation under neutral to slightly alkaline conditions, NH3 is oxidized to $\textrm{NO}_2^ - $via the intermediates NH2OH and NO [38]. The exact physiology of ammonia-oxidizing microorganisms at acidic pH remains uncertain, partially due to the chemical instability of $\textrm{NO}_2^ - $ and NO, especially at low pH. Consequently, biotic and abiotic reactions occur simultaneously in the bioreactor, and a wide variety of nitrogen species are expected to be present. This bioreactor setup is therefore of particular interest to test the IDFG-based system. At acidic conditions, $\textrm{NO}_2^ - $ protonates partially to HNO2, which then chemically degrades mostly into volatile compounds such as NO, NO2, and N2O3 [39]. In addition, N2O can be produced both biotically and abiotically in this system [40]. Simultaneous observations of multiple nitrogen oxides such as NOx and N2O are challenging, given their complexity to be measured by mass spectrometric methods, as well as that their absorption spectra are in different parts of the mid-infrared spectrum. For example, the main absorption band of NO2 is between 6–6.4 µm, NO between 5.1–5.6 µm, N2O between 4.4–4.6 µm, and HNO2 between 7.7–8.3 µm and 5.7–6.3 µm (both bands being about equally strong). Given the strong water absorption between 6–7 µm, the NOx bands in this spectral window are not useful, and consequently, the weaker NO2 absorption band around 3.4–3.5 µm had to be used. The detection of these nitrogen oxides (including NOx, N2O, and HNO2) would thus either require several laser sources or a source with at least a spectral bandwidth of 3.4–8.3 µm. Therefore, the broad spectral coverage of the IDFG-based SC source opens new possibilities to detect and quantify the major NOx species simultaneously.
As shown in Fig. 9, in a single measurement (averaged for 190 seconds), absorption bands of NO, NO2, N2O, HNO2, and CO2, could be detected. The top panel contains the full coverage of the power spectral density obtained from the measurement, of which only the narrow parts with a white background were used to retrieve the concentration of the six species of interest. Even though the full absorbance bands of the species of interest can be observed, the high spectral resolution allows us to zoom into specific parts of the spectrum with maximum absorbance strength and minimal interference with, e.g., strongly absorbing water lines. Consequently, the regions shown in the absorbance spectrum (middle panel) were used for retrieving the concentrations of the species of interest (Fig. 9). Fitting these spectra to simulations using the HITRAN database (or PNNL in the case of HNO2) yields concentrations of 89 ± 9 ppm NO, 270 ± 40 ppm NO2, 11.2 ± 0.10 ppm N2O, 357 ± 2 ppm HNO2, and 1.34 ± 0.02% CO2, in which the uncertainty is given by the standard deviation of 4 consecutive measurements.
Fig. 9. Measured power spectral density (top) and absorbance spectrum (middle) of a bioreactor off-gas sample from an acidic ammonia-oxidizing enrichment culture (in black). The absorbance spectrum is zoomed to the regions used for fitting gas species concentrations and fitted reference spectra from PNNL (HNO2) and HITRAN (others) are shown in color and inverted. The residual of the fit is shown in the bottom.
The ability to simultaneously observe and quantify all these nitrogen-containing species offers the possibility of constructing nitrogen mass balances of the pathways converting ammonia under acidic conditions. In contrast to previous methods [41–43] targeting only a few of these molecules mainly by colorimetric assays, gas and ion chromatography, the method described here is a major step towards establishing a completely closed mass balance.
4. Discussion
The IDFG-based SC source in combination with FTS shows great possibilities for sensitive, multispecies gas detection. However, there are some technical considerations which need further discussion. As mentioned before, the output of the source consists of a combination of two co-propagating, superimposed beams that each cover a different part of the spectrum with an overlap in the middle (4–7 µm). Previous works [31,32] describe how those two beams are resulting from the interaction within the ZGP crystal, where the broadened and red-shifted fundamental component stretches up to 7 µm and the IDFG-generated component spans from 4–11.5 µm. Performing dual-comb spectroscopy with two IDFG-based SC sources over the entire spectral coverage of the source is rather complicated since the carrier-envelope offset frequencies of these two components of the beam are different. Therefore, the demonstrations of dual-comb spectroscopy using two of these sources have been limited to a spectral coverage between ∼6.6 and 11.5 µm, which is the spectral range that is solely generated by the IDFG process [28,30], limiting the capability of the detection system. Here, we have shown that it is possible to use both components in the beam simultaneously for spectroscopic applications using an FTS, yielding – to the best of our knowledge – an unprecedented spectral coverage from 3–11.5 µm for sensitive and simultaneous multi-species trace gas detection based on SC sources.
However, as explained in the Methods section, these two superimposed components of the beam have slightly different spatial distributions and beam divergences although the two beams are co-propagating. As a result, only one of these two components can be properly mode matched with the correct entrance angle to the MPC and achieve the nominal effective path length, while the other component will be lost (cut off in the cell). During the experiments, where we used a bandpass or a longpass filter and only transmitted one of the two beam components, it was possible to achieve the maximum effective path length of the MPC (32.1 m). However, by removing the filter the maximum achievable path length for both beam components, without one of them being cut off, was shorter. We measured and calibrated this path length by using an independently certified CO2/N2 mixture in the MPC that absorbs in both shorter and longer wavelengths, where the two beam components cover separately. This calibration was performed prior to the reported experiments in this manuscript. Despite being shorter than the nominal interaction length, the path lengths for both beams agreed over the entire spectrum to be 25.6 m. This is the interaction length that we used to retrieve concentration of CO2 in the two spectral ranges shown in Fig. 4. The featureless residual of the fit and the unique retrieved concentration shows that the interaction length was properly calibrated.
Replacing the MPC with a resonant cavity would be an option to further enhance the sensitivity of the system, considering the recent developments in the ultralow loss mid-infrared crystalline mirrors [44]. However, it is not possible to use enhancement cavities for the entire wavelength range simultaneously, because the HR-coatings of the cavity mirrors have a limited bandwidth. Moreover, the IDFG process removes the carrier-envelope offset frequency, while this is a necessary tuning parameter when optimizing the coupling to the cavity (in the tight-locking scheme), given that the dispersion in the cavity leads to changes in the regularity of the free spectral range in the cavity over the spectrum. In addition, the high spectral power of the IDFG-based SC source makes it possible to use MPCs with a higher number of reflections and still have enough power on the detectors to maintain a good SNR. This can increase the detection sensitivity without sacrificing the spectral bandwidth of the source.
5. Conclusion
The development of IDFG-based SC sources opens new opportunities in mid-infrared spectroscopy. In combination with an MPC and a custom-built FTS, the IDFG-based SC source presented here proves to be an excellent tool for multispecies gas detection. In comparison to existing mid-infrared SC sources, the current system provides one to two orders of magnitude better NEAS. Moreover, the IDFG-based SC source extends much further into the IR spectrum, with a major part of its spectral power in the 7–11.5 µm wavelength region, thereby accessing the highly interesting molecular fingerprint region. Although there are existing DFG- and OPO-based systems that also access (a part of) this spectral window [37], the immense wavelength range of 2–11.5 µm proves to be an interesting advantage, which to the best of our knowledge, has not been demonstrated yet for SC-based spectroscopy.
Characterizing the performance of the source for various gaseous molecules, (sub-)ppb levels of sensitivity are achieved for hydrocarbons (e.g., CH4, C2H4), oxides (e.g., SO2, NOx), and small organic molecules (e.g., acetone, ethyl acetate) in measurement times of the order of seconds to a few minutes. In comparison to the fiber-based SC sources, the system equipped with an IDFG-based SC source proved to be 60 times more sensitive than using a ZBLAN-based SC source (for CH4 around 3 µm) and 240 times more sensitive than using a chalcogenide-based SC source (for SO2 around 7 µm).
In a real-life application, various nitrogen species including nitrogen oxides were measured simultaneously in the off-gas of a bioreactor demonstrating the advantage of a wide spectral range of the system. Further applications of such a system include the analysis of notoriously complex gas mixtures, such as gas phase chemical conversions in electrical discharges or plasmas [7], or human breath, which contains different compounds in ppb- to ppm-level concentrations while the spectrum is significantly obscured by highly absorbing spectral lines of CO2 and H2O [45–48]. Finally, an interesting application for this source is open-path spectroscopy for environmental monitoring, where the combination of the broad spectral coverage and spatial coherence of the source could be an advantage over the current state-of-the-art dual-comb systems used to monitor multiple greenhouse gases or pollutants in larger outdoor areas [49–55].
Funding
H2020 Industrial Leadership (101015825, TRIAGE Project); Nederlandse Organisatie voor Wetenschappelijk Onderzoek (016.Vidi.189.050).
Disclosures
The authors declare no financial relationship between the company manufacturing the IDFG-based SC sources (IPG Photonics) and the academic institute developing the FTS and performing the measurements (Radboud University). The company did not fund this study/manuscript. The development of the FTS instrument and reporting of the results were carried out by the authors at Radboud University, independent of the company that manufactures the SC sources.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon request.
References
1. C. R. Petersen, P. M. Moselund, L. Huot, et al., “Towards a table-top synchrotron based on supercontinuum generation,” Infrared Phys. Technol. 91, 182–186 (2018). [CrossRef]
2. C. R. Petersen, U. Møller, I. Kubat, et al., “Mid-infrared supercontinuum covering the 1.4-13.3 µm molecular fingerprint region using ultra-high NA chalcogenide step-index fibre,” Nat. Photonics 8(11), 830–834 (2014). [CrossRef]
3. K. E. Jahromi, Q. Pan, L. Høgstedt, et al., “Mid-infrared supercontinuum-based upconversion detection for trace gas sensing,” Opt. Express 27(17), 24469 (2019). [CrossRef]
4. K. E. Jahromi, M. Nematollahi, Q. Pan, et al., “Sensitive multi-species trace gas sensor based on a high repetition rate mid-infrared supercontinuum source,” Opt. Express 28(18), 26091 (2020). [CrossRef]
5. M. A. Abbas, K. E. Jahromi, M. Nematollahi, et al., “Fourier transform spectrometer based on high-repetition-rate mid-infrared supercontinuum sources for trace gas detection,” Opt. Express 29(14), 22315 (2021). [CrossRef]
6. K. E. Jahromi, Q. Pan, A. Khodabakhsh, et al., “A broadband mid-infrared trace gas sensor using supercontinuum light source: applications for real-time quality control for fruit storage,” Sensors 19(10), 2334 (2019). [CrossRef]
7. R. Krebbers, N. Liu, K. E. Jahromi, et al., “Mid-infrared supercontinuum-based Fourier transform spectroscopy for plasma analysis,” Sci. Rep. 12(1), 9642 (2022). [CrossRef]
8. N. R. Newbury, B. R. Washburn, K. L. Corwin, et al., “Noise amplification during supercontinuum generation in microstructure fiber,” Opt. Lett. 28(11), 944 (2003). [CrossRef]
9. U. Møller, S. T. Sørensen, C. Jakobsen, et al., “Power dependence of supercontinuum noise in uniform and tapered PCFs,” Opt. Express 20(3), 2851 (2012). [CrossRef]
10. I. Zorin, P. Gattinger, A. Ebner, et al., “Advances in mid-infrared spectroscopy enabled by supercontinuum laser sources,” Opt. Express 30(4), 5222 (2022). [CrossRef]
11. A. M. Heidt, A. Hartung, G. W. Bosman, et al., “Coherent octave spanning near-infrared and visible supercontinuum generation in all-normal dispersion photonic crystal fibers,” Opt. Express 19(4), 3775 (2011). [CrossRef]
12. M. Klimczak, G. Soboń, R. Kasztelanic, et al., “Direct comparison of shot-to-shot noise performance of all normal dispersion and anomalous dispersion supercontinuum pumped with sub-picosecond pulse fiber-based laser,” Sci. Rep. 6(1), 19284 (2016). [CrossRef]
13. S. Rao D. S., M. Jensen, L. Grüner-Nielsen, et al., “Shot-noise limited, supercontinuum-based optical coherence tomography,” Light: Sci. Appl. 10(1), 133 (2021). [CrossRef]
14. K. L. Vodopyanov, “Ultra-broadband mid-infrared frequency combs produced by optical subharmonic generation,” Quantum Electron. 52(4), 307–312 (2022). [CrossRef]
15. K. Krzempek, D. Tomaszewska, A. Głuszek, et al., “Stabilized all-fiber source for generation of tunable broadband fCEO-free mid-IR frequency comb in the 7–9 µm range,” Opt. Express 27(26), 37435 (2019). [CrossRef]
16. F. Adler, P. Masłowski, A. Foltynowicz, et al., “Mid-infrared Fourier transform spectroscopy with a broadband frequency comb,” Opt. Express 18(21), 21861 (2010). [CrossRef]
17. A. Khodabakhsh, V. Ramaiah-Badarla, L. Rutkowski, et al., “Fourier transform and Vernier spectroscopy using an optical frequency comb at 3–54 µm,” Opt. Lett. 41(11), 2541 (2016). [CrossRef]
18. A. J. Fleisher, B. J. Bjork, T. Q. Bui, et al., “Mid-infrared time-resolved frequency comb spectroscopy of transient free radicals,” J. Phys. Chem. Lett. 5(13), 2241–2246 (2014). [CrossRef]
19. A. Hjältén, M. Germann, K. Krzempek, et al., “Optical frequency comb Fourier transform spectroscopy of 14N216O at 7.8 µm,” J. Quant. Spectrosc. Radiat. Transfer 271, 107734 (2021). [CrossRef]
20. K. Iwakuni, T. Q. Bui, J. F. Niedermeyer, et al., “Comb-resolved spectroscopy with immersion grating in long-wave infrared,” Opt. Express 27(3), 1911 (2019). [CrossRef]
21. A. V. Muraviev, V. O. Smolski, Z. E. Loparo, et al., “Massively parallel sensing of trace molecules and their isotopologues with broadband subharmonic mid-infrared frequency combs,” Nat. Photonics 12(4), 209–214 (2018). [CrossRef]
22. G. Ycas, F. R. Giorgetta, E. Baumann, et al., “High-coherence mid-infrared dual-comb spectroscopy spanning 2.6 to 5.2 µm,” Nat. Photonics 12(4), 202–208 (2018). [CrossRef]
23. M. A. Abbas, Q. Pan, J. Mandon, et al., “Time-resolved mid-infrared dual-comb spectroscopy,” Sci. Rep. 9(1), 17247 (2019). [CrossRef]
24. C. Gaida, M. Gebhardt, T. Heuermann, et al., “Watt-scale super-octave mid-infrared intrapulse difference frequency generation,” Light: Sci. Appl. 7(1), 94 (2018). [CrossRef]
25. S. Vasilyev, I. Moskalev, V. Smolski, et al., “Multi-octave visible to long-wave IR femtosecond continuum generated in Cr:ZnS-GaSe tandem,” Opt. Express 27(11), 16405 (2019). [CrossRef]
26. H. Timmers, A. Kowligy, A. Lind, et al., “Molecular fingerprinting with bright, broadband infrared frequency combs,” Optica 5(6), 727 (2018). [CrossRef]
27. A. J. Lind, A. Kowligy, H. Timmers, et al., “Mid-infrared frequency comb generation and spectroscopy with few-cycle pulses and χ (2) nonlinear optics,” Phys. Rev. Lett. 124(13), 133904 (2020). [CrossRef]
28. S. Vasilyev, A. Muraviev, D. Konnov, et al., “Longwave infrared (6.6–11.4 µm) dual-comb spectroscopy with 240,000 comb-mode-resolved data points at video rate,” Opt. Lett. 48(9), 2273 (2023). [CrossRef]
29. A. S. Kowligy, H. Timmers, A. J. Lind, et al., “Infrared electric field sampled frequency comb spectroscopy,” Sci. Adv. 5(6), eaaw8794 (2019). [CrossRef]
30. D. Konnov, A. Muraviev, S. Vasilyev, et al., “High-resolution frequency-comb spectroscopy with electro-optic sampling and instantaneous octave-wide coverage across mid-IR to THz at a video rate,” APL Photonics 8(11), 110801 (2023). [CrossRef]
31. S. Vasilyev, V. Smolski, J. Peppers, et al., “Middle-IR frequency comb based on Cr:ZnS laser,” Opt. Express 27(24), 35079 (2019). [CrossRef]
32. S. Vasilyev, V. Smolski, I. Moskalev, et al., “Multi-octave infrared femtosecond continuum generation in Cr:ZnS-GaSe and Cr:ZnS-ZGP tandems,” in Nonlinear Frequency Generation and Conversion: Materials and Devices XIX, P. G. SchunemannK. L. Schepler, eds., (SPIE, 2020), pp. 6.
33. J. Kim, S. Lee, W. Lee, et al., “Design optimization and implementation of a Fourier transform spectrometer with rotating motion for 0.1 cm -1 resolution spectroscopy,” Opt. Express 31(20), 33041 (2023). [CrossRef]
34. I. E. Gordon, L. S. Rothman, R. J. Hargreaves, et al., “The HITRAN2020 molecular spectroscopic database,” J. Quant. Spectrosc. Radiat. Transfer 277, 107949 (2022). [CrossRef]
35. S. W. Sharpe, T. J. Johnson, R. L. Sams, et al., “Gas-phase databases for quantitative infrared spectroscopy,” Appl. Spectrosc. 58(12), 1452–1461 (2004). [CrossRef]
36. N. Picone, A. Pol, R. Mesman, et al., “Ammonia oxidation at pH 2.5 by a new gammaproteobacterial ammonia-oxidizing bacterium,” ISME J. 15(4), 1150–1164 (2021). [CrossRef]
37. K. E. Jahromi, M. Nematollahi, R. Krebbers, et al., “Fourier transform and grating-based spectroscopy with a mid-infrared supercontinuum source for trace gas detection in fruit quality monitoring,” Opt. Express 29(8), 12381–12397 (2021). [CrossRef]
38. M. M. M. Kuypers, H. K. Marchant, and B. Kartal, “The microbial nitrogen-cycling network,” Nat. Rev. Microbiol. 16(5), 263–276 (2018). [CrossRef]
39. K. M. Udert, T. A. Larsen, and W. Gujer, “Chemical Nitrite Oxidation in Acid Solutions as a Consequence of Microbial Ammonium Oxidation,” Environ. Sci. Technol. 39(11), 4066–4075 (2005). [CrossRef]
40. A. M. Quick, W. J. Reeder, T. B. Farrell, et al., “Nitrous oxide from streams and rivers: A review of primary biogeochemical pathways and environmental variables,” Earth-Sci. Rev. 191, 224–262 (2019). [CrossRef]
41. Marc. Roth, “Fluorescence reaction for amino acids,” Anal. Chem. 43(7), 880–882 (1971). [CrossRef]
42. S. A. Gordon, A. Fleck, and J. Bell, “Optimal Conditions for the Estimation of Ammonium by the Berthelot Reaction,” Ann. Clin. Biochem. 15(1-6), 270–275 (1978). [CrossRef]
43. S. Wang, K. Lin, N. Chen, et al., “Automated determination of nitrate plus nitrite in aqueous samples with flow injection analysis using vanadium (III) chloride as reductant,” Talanta 146, 744–748 (2016). [CrossRef]
44. G.-W. Truong, L. W. Perner, D. M. Bailey, et al., “Mid-infrared supermirrors with finesse exceeding 400 000,” Nat. Commun. 14(1), 7846 (2023). [CrossRef]
45. B. Henderson, A. Khodabakhsh, M. Metsälä, et al., “Laser spectroscopy for breath analysis: towards clinical implementation,” Appl. Phys. B 124(8), 161 (2018). [CrossRef]
46. J. D. Fenske and S. E. Paulson, “Human Breath Emissions of VOCs,” J. Air Waste Manage. Assoc. 49(5), 594–598 (1999). [CrossRef]
47. T. Mathew, P. Pownraj, S. Abdulla, et al., “Technologies for clinical diagnosis using expired human breath analysis,” Diagnostics 5(1), 27–60 (2015). [CrossRef]
48. C. Wang and P. Sahay, “Breath analysis using laser spectroscopic techniques: breath biomarkers, spectral fingerprints, and detection limits,” Sensors 9(10), 8230–8262 (2009). [CrossRef]
49. D. I. Herman, C. Weerasekara, L. C. Hutcherson, et al., “Precise multispecies agricultural gas flux determined using broadband open-path dual-comb spectroscopy,” Sci. Adv. 7(14), eabe9765 (2021). [CrossRef]
50. G. J. Mead, E. M. Waxman, D. Bon, et al., “Open-path dual-comb spectroscopy of methane and VOC emissions from an unconventional oil well development in Northern Colorado,” Front. Chem. 11, 1202255 (2023). [CrossRef]
51. F. R. Giorgetta, J. Peischl, D. I. Herman, et al., “Open-path dual-comb spectroscopy for multispecies trace gas detection in the 4.5–5 µm spectral region,” Laser Photonics Rev. 15(9), 2000583 (2021). [CrossRef]
52. K. C. Cossel, E. M. Waxman, F. R. Giorgetta, et al., “Open-path dual-comb spectroscopy to an airborne retroreflector,” Optica 4(7), 724 (2017). [CrossRef]
53. G. B. Rieker, F. R. Giorgetta, W. C. Swann, et al., “Frequency-comb-based remote sensing of greenhouse gases over kilometer air paths,” Optica 1(5), 290 (2014). [CrossRef]
54. C. B. Alden, S. C. Coburn, R. J. Wright, et al., “Single-blind quantification of natural gas leaks from 1 km distance using frequency combs,” Environ. Sci. Technol. 53(5), 2908–2917 (2019). [CrossRef]
55. S. Coburn, C. B. Alden, R. Wright, et al., “Regional trace-gas source attribution using a field-deployed dual frequency comb spectrometer,” Optica 5(4), 320 (2018). [CrossRef]
