Mid-infrared laser absorption spectroscopy utilizing a high-finesse optical cavity enables high precision trace analysis of gas molecules. In particular, optical detection of radiocarbon (14C) based on cavity ringdown spectroscopy using a quantum cascade laser (QCL) is gaining attention as an alternative to accelerator mass spectrometry. This paper reports a compact-packaged narrow-linewidth QCL system utilizing resonant optical feedback from an external V-shaped cavity. Based on frequency noise analysis, the derived laser linewidth is 44 kHz for 100 μs integration time with the capability to perform seamless frequency scanning around 10 GHz. We installed this laser system within a table-top cavity ringdown spectrometer for 14CO2. A single-shot detection limit of 1.2 × 10−9 cm−1 Hz−1/2 leading to a detectable abundance evaluated from a noise analysis of 0.2 in fraction modern 14C for a 10-s averaging time was achieved. This capability of rapid analysis for 14CO2 is suitable for various applications requiring trace 14C analysis.
I. INTRODUCTION
Quantum cascade lasers (QCLs) have played an important role in the progress of mid-infrared laser spectroscopy and its applications.1 Their wide wavelength tunability, compact size, and ease of handling are highly appropriate for in situ analysis of gas-phase samples based on mid-infrared laser absorption spectroscopy. Especially, for trace detection of molecules or their isotopologues, sensitivity enhancement by a high-finesse optical cavity is widely established as cavity-enhanced absorption spectroscopy (CEAS) and cavity ringdown spectroscopy (CRDS).2,3 For these applications, a single-mode continuous wave (CW) narrow-linewidth laser is generally used as a probe laser, such as a distributed feedback (DFB)-QCL. Another candidate with a wide tuning range and high output power is an external cavity (EC)-QCL, generally utilizing optical feedback from a diffraction grating4,5 or a partial reflector6 though it is usually more sensitive to environmental disturbances and can experience mode-hops. These narrow-linewidth CW QCLs are commercially widely available with typical linewidth from 500 kHz to 50 MHz, depending on observation time.
The performance of mid-infrared mirrors has advanced as well. Recently, monocrystalline mirror coatings7 have been developed and optical loss below 10 parts per million (ppm) at a wavelength of 4.54 μm was reported.8 Implementing a super-high-finesse optical cavity consisting of these ultra-low loss mirrors significantly improves the sensitivity of trace gas analysis by CRDS. In addition, it facilitates higher intracavity optical power, which is demanded in new CRDS techniques based on nonlinear absorption effects in the optical cavity such as saturated cavity absorption spectroscopy9 or two-photon/two-color CRDS.10–12 However, as the finesse grows, it imposes stricter constraints on the linewidth and stability of the mid-infrared laser. In the case of our CRDS system designed for radiocarbon (14C) analysis,13 newly obtained CRD mirrors with higher reflectivity (ZnSe substrate, CRD optics Inc.) were recently installed. The reflectivity of the mirrors was determined to be 99.996% at 4.526 μm with a corresponding cavity finesse of ∼80 000, which is higher than the reported value of 52 000 for the natural abundance 14C measurement;14 incidentally, these mirrors seem to have a larger optical (absorption/scatter) loss compared to the crystalline mirrors, as we observed significantly reduced transmission compared to our previously installed mirrors. Assuming a free spectral range (FSR) of 340 MHz, the width of a resonance in our new optical cavity reaches about 4 kHz. In this context, linewidth reduction of the QCL is now strongly desired to allow efficient optical coupling.
Among several techniques for laser linewidth reduction, optical feedback has been widely utilized, commonly known as self-injection locking in semiconductor lasers. In this technique, a fraction of the photons is reinjected to the gain chip by an external cavity construction. Our previous reports15,16 showed QCL linewidth reduction obtained by utilizing a cat-eye partial reflector. Recently, strong optical feedback in a QCL by a partial reflector was reported from the other group as well.17 A more advanced approach is resonant optical feedback using on-resonance transmission from a high-finesse cavity.18 Optical feedback (OF)-CEAS19 successfully improved sensitivity by self-frequency locking of the probe laser to a V-shaped three-mirror cavity (referred to as V-cavity here). Trace gas detection in mid-infrared regions with QCL was performed,20–22 and frequency stabilized (FS)-CRDS combining a reference V-cavity optical feedback was proposed with a DFB-diode laser at 1.6 μm.23 The V-shape cavity geometry is beneficial as it avoids direct non-resonant reflection so that feedback is only active near a cavity resonance. A similar approach conducted with a QCL employing a mode-mismatched linear Fabry–Pérot cavity was proposed as well.24 In addition, the linewidth reduction of the QCL utilizing “weak” resonant optical feedback from a highly mode-matched linear cavity was demonstrated recently,25 used for two-photon cavity-enhanced absorption spectroscopy. These resonant optical feedback techniques are a great alternative to Pound–Drever–Hall (PDH)26 locking, which requires high-speed electronics and an electro-optical modulator.
For long-term operational stability with optical feedback, precise control of the phase of the feedback is necessary: generally, the phase control is done by path length adjustment with a piezo actuator. Simultaneous laser current control is combined with phase tracking to enable wider tuning capability. This complexity still prevents the technique from the widespread application.
In this paper, we report an external V-shaped cavity optical feedback (EVC)-QCL system for mid-infrared CRDS applications, particularly toward 14C optical detection, which is a “compact-packaged” narrow-linewidth mid-infrared laser system with close to 10 GHz mode-hop-free tuning range. Although the phase and the current control are still necessary, a seamless and simple operation has been realized due to robust locking combined with a thermally stabilized enclosure. The optical performance of the EVC-QCL was evaluated, and sensitivity improvement of our mid-infrared CRDS system owing to the implementation of the EVC-QCL was demonstrated.
II. EXPERIMENTAL DETAILS
Our mid-infrared cavity ringdown spectrometer was designed to detect the P(20) line in the ν3 band of 14CO2 lying at 2209.108 cm−1 (4.526 μm). For a general description of our CRDS spectrometer including sample preparation, we refer to our previous report.13 A combustion tube was utilized for conversion of the carbon content of a sample into CO2 gas and a temperature-controlled column selectively trapped and released the CO2. Consequently, CO2 gas partial pressure in the CRDS gas cell above 60% of the total pressure was achieved.
Our first studies of optical feedback with the simple partial reflector are explained in Refs. 15 and 16. Here, we focus on the newly developed EVC-QCL with a schematic of the experimental layout depicted in Fig. 1. The EVC-QCL system was enclosed in a 300 × 450 mm2 box and placed on a 600 × 600 mm2 optical breadboard containing the remaining optical components such as an acoustic optical modulator, optical lenses, and parabolic mirrors for beam shaping.
Schematic of the EVC-QCL system. The details of the CRDS system for 14C analysis are given in our previous work.13 The components with no indications are either gold mirrors or optical lenses. Note: BS, beam splitter/sampler; PZT, piezo actuator; HV, high voltage amplifier; DAQ, data acquisition system with analog to digital converters; and PD, photodiode.
Schematic of the EVC-QCL system. The details of the CRDS system for 14C analysis are given in our previous work.13 The components with no indications are either gold mirrors or optical lenses. Note: BS, beam splitter/sampler; PZT, piezo actuator; HV, high voltage amplifier; DAQ, data acquisition system with analog to digital converters; and PD, photodiode.
A DFB-QCL (Hamamatsu Photonics, LE0833QCL) with an integrated collimation lens was fixed to a side plate of an outer enclosure for partial environmental insulation of the whole laser system. A wire-grid polarizer combined with a λ/2 waveplate was used to control the feedback strength. The external V-cavity consisted of three ZnSe CRD mirrors (reflectivity: R ∼ 99.98%), with a finesse of 7850. An arm-length of the V-cavity (Lvc) was designed to be 55 mm with two 150 V ring-piezo actuators included for frequency scanning, leading to a calculated FSR of 1.36 GHz. However, due to a defect in one of the actuators, only a single-side piezo scan was demonstrated in this paper. Gold-coated mirrors transported the beam to the V-cavity. A 150 V piezo stack actuator attached to the rear side of a light-weight gold mirror was employed as a feedback phase adjuster and a hollow roof prism mirror was installed for adjustment of the total path length between QCL and V-cavity.
Since fluctuations of the V-cavity length in turn cause frequency instability of the QCL locked onto the V-cavity resonance, a cordierite ceramic (from Nishimura Advanced Ceramics) was adopted as a spacer fixing the V-cavity mirrors. The cordierite has a low thermal-expansion coefficient of ∼10−7 K−1 at room temperature. To further minimize drifts, an aluminum box enclosed the V-cavity. Thermo-electrical cooling (TEC) fixed to a thermal link kept the temperature of the box stabilized and the internal pressure was kept below 100 Pa using a pumping port. The sidewall of the enclosure was angled to eliminate unwanted feedback from the BaF2 entrance windows. A water-cooled aluminum breadboard was placed on the bottom of the outer enclosure to stabilize the whole laser layout for the optical feedback, simultaneously acting as a heatsink for the TEC of the inner enclosure through the thermal link.
The feedback path length is a crucial parameter as well, defined by the distance between the QCL output facet and the first V-cavity mirror, from here on denoted as Lphase. To maintain the phase-matching condition for all resonant cavity modes, Lphase is required to be an odd multiple of Lvc.19 This would enable performing systematic mode-by-mode jumps of the laser frequency, which could be used to effectively increase the tuning range without reaching the limits of the piezo-adjusters. While the even-multiple length is also acceptable, two adjacent resonant modes would have opposite phase conditions, so larger jumps would need to be performed. However, so far, we only focused on a single-mode and followed it without jumps for frequency scanning. In this case, the integer-multiple of the Lvc maybe not required. Nevertheless, for future exploitation, Lphase was adjusted to roughly 495 mm (∼9 × Lvc).
The transmitted beam after the polarizer was passed through a 30 dB optical isolator and transported to the outside via periscope and a wedged output window. A small portion of the output laser beam was directed to an N2O reference cell (wavelength references, pure N2O with 0.5 Torr) by a beam sampler with the remainder guided further to our CRDS cavity. A portion of light before the output window of the enclosure was picked up by a combination of a λ/2 waveplate and a wedged CaF2 window for wavelength calibration. In the previous setup, we used a solid silicon etalon (from LightMachinery). While the large refractive index of silicon realizes an FSR of 527 MHz with only 83 mm length, its large thermal-expansion coefficient was found to be detrimental even with frequent calibration using the reference of the N2O absorption lines. Instead, a new self-made quadrature interferometer was constructed, which consisted of 50:50 BaF2 beam splitters and square-shaped gold mirrors, all epoxied directly onto another cordierite spacer. It was installed within the same temperature-stabilized inner enclosure. In front of the interferometer, another 50/50 beam splitter divided the laser path into two, both beams aimed toward the etalon for wavelength calibration using a quadrature interferometer technique,27 with details beyond the scope of this article. Its performance will be reported elsewhere in the near future. In this paper, only a single beam was used so that the quadrature interferometer effectively works as a normal low-finesse etalon with an FSR of 202 MHz. Since the etalon fringes require power normalization, a reflection from a rear surface of the CaF2 window picking up the light for wavelength calibration was detected as well. Room-temperature InAsSb photodiodes (Hamamatsu Photonics, P13243-011MA) were used to monitor each signal.
Due to disturbances from room-temperature drift and pressure change as well as vibrations, the V-cavity resonant optical feedback is difficult to maintain without any stabilizing mechanism. In addition to the dual enclosures, a bench-top laser stabilization device (TEM Messtechnik, Laselock) performed two-channel electrical feedback control of the optical feedback phase and the QCL current, respectively. This device is a key element to realize robust resonant optical feedback capable of rapid continuous frequency scanning over more than 10 GHz. The Laselock generated two independent error signals from the V-cavity transmission signal using first harmonic phase-sensitive detection (or lock-in detection), applying dither modulation to the phase piezo and the QCL current. The phase piezo mirror was dithered at a frequency close to a mechanical resonance of the piezo assembly (∼31 kHz) to obtain a sufficient error signal for stable locking. The QCL current was dithered at a frequency of 57 kHz. A voltage divider was installed before the modulation input of the QCL current driver to suppress digitization noise.
Two sets of the PID (proportional, integral, and differential) gains were tuned to minimize the error signals. Once locked, the coupled laser frequency can be easily scanned by continuously changing the V-cavity length with the piezo actuators while maintaining the resonant optical feedback (as shown in Fig. 4). In addition, re-optimization was not necessary unless we had intentionally changed the operational settings such as frequency tuning rate, dithering parameters, or the QCL chip temperature.
III. RESULTS AND DISCUSSION
A. The optical properties of the EVC-QCL
Resonant feedback effects from the V-cavity were experimentally observed by monitoring the V-cavity transmission intensity as shown in Fig. 2(a), with a setting of 20% reflection at the polarizer, denoted as Rpol = 20%. During this experiment, a fixed DC voltage was applied to the piezo actuators in the V-cavity, while the applied voltage to the phase controlling piezo was changed in several steps to acquire the characteristic transmission patterns. The QCL current was scanned linearly and converted into the free-running QCL frequency: ωfree.
Phase offset dependence of experimentally acquired V-cavity transmission signals (a) and simulated patterns (b). The bottom panels [(c)–(e)] indicate the simulated coupled QCL frequency: ω as a function of the relative free-running frequency; ωfree of the QCL to the resonance frequency center of the external V-cavity; ωres with different phase offsets of 0, −π/4, −π/2, respectively. The red dashed lines show the coupled frequency jump when ωfree is tuned from the direction indicated by the gray arrow.
Phase offset dependence of experimentally acquired V-cavity transmission signals (a) and simulated patterns (b). The bottom panels [(c)–(e)] indicate the simulated coupled QCL frequency: ω as a function of the relative free-running frequency; ωfree of the QCL to the resonance frequency center of the external V-cavity; ωres with different phase offsets of 0, −π/4, −π/2, respectively. The red dashed lines show the coupled frequency jump when ωfree is tuned from the direction indicated by the gray arrow.
The coupled laser frequency of a QCL laser with weak resonant optical feedback can be represented as a function depending on external laser cavity parameters and V-cavity frequency filtering effects. A detailed discussion is shown in Sec. 3.3 in Ref. 19: we calculated the coupled laser frequency in python using Eq. (3). Figure 2(b) shows simulated V-cavity transmission patterns for various feedback phase conditions. Since the parameters of our DFB-QCL required for the simulation were not provided by the manufacturer, these calculations were conducted with values used in similar studies with a different QCL.17,25 The calculation was started for Lphase = 9 × Lvc and the phase offset was adjusted to qualitatively match the shapes observed in the experimental results. It must be noted that the Henry factor28 of the QCL affects the phase as well, although QCL typically has a small value compared to other diode lasers. The feedback power ratio k is a crucial factor describing the strength of the feedback and is defined as the intensity ratio of the feedback injection power to the total output power. For the simulation shown in Fig. 2(b), k = 1% was used so that the simulated patterns matched the experimental results. In our setup using the polarizer and the λ/2 waveplate, a polarization mismatch between the QCL output and the returning feedback light occurs. Assuming that the wrong polarization state does not contribute to the optical feedback, this effectively induces a combined optical loss, predicted to be proportional to . Furthermore, in the V-cavity layout, only one of the four transmission directions returned to the QCL. Considering these loss factors, was expected. A value of k = 1% for Rpol = 20% was found to be consistent with this consideration. Even with a moderate feedback strength of k = 1%, the QCL frequency was self-injection locked to the V-cavity and captured for a range of ωfree over more than 100 MHz.
We performed similar experiments and simulations with different k-values. As shown in Fig. 3, the experimental data corresponded with the simulation result within a few percent errors even without considering other losses. Larger k-values provided a wider capture range; however, there is a trade-off between the k-value and the available laser power for spectroscopy. Here, we chose a 20% power reflection as a reasonable balance between ease of locking and high power availability for the CRDS system. The good agreement between theory and experiment should help to predict the performance of different conditions and to implement better resonant optical feedback.
Dependence of feedback capture range on feedback ratio: (a) experimental result and (b) simulation result. The simulation was conducted with the phase offset of −π/4 so that the transmission patterns matched to the experiment.
Dependence of feedback capture range on feedback ratio: (a) experimental result and (b) simulation result. The simulation was conducted with the phase offset of −π/4 so that the transmission patterns matched to the experiment.
Figure 4 shows a typical frequency scan obtained by modulating the V-cavity length with a triangle voltage ramp applied to one of the V-cavity piezo actuators. Due to the electrical control from the Laselock, the phase piezo voltage (regulator A), as well as the current (regulator B), automatically followed the change of the resonance condition of the V-cavity; the phase piezo voltage was inversely correlated with the current as they had an opposite sign with respect to laser frequency changes. The etalon fringes as well as the N2O spectra prove a seamless 4.65 GHz (∼23 × FSR of the etalon) frequency scan per second. The advantages of this dual modulation and dual feedback technique are the capability to perform wide-range scans without mode-hops and long-term lock stability, which are usually not feasible if relying on simple feed-forward current control, which is often applied in EC lasers.29 Considering the ring-piezo stroke of 10.5 μm/150 V specified for the actuator, the observed range of 5 GHz (equivalent to 3.7 × FSR of the V-cavity) is close to the maximum achievable with a single-side piezo scan. As the other tuning parameters still have enough headroom, a dual-side piezo scan can enable a 10 GHz mode-hop-free frequency tuning range. We attempted faster scanning at more than 1 Hz/scan, which required a higher D gain parameter, which in turn caused larger fluctuations in the coupled frequency.
The nonlinear dependence of the laser frequency seen in Fig. 4 is due to hysteresis and nonlinearity of the ring-piezo actuators. However, this is no problem in the EVC-QCL system with wavelength calibration combining the etalon and the N2O reference cell.
The frequency noise power spectral density (PSD) as shown in Fig. 5 was calculated from the N2O signal acquired by fixing the laser frequency at the side of a strong Doppler-broadened absorption peak [at the red point shown in Fig. 4(c)]. Since the concentration of N2O in the cell as well as the total pressure were specified, the acquired N2O spectrum was fitted by the HAPI package from HITRAN30 to convert transmitted intensity fluctuation to frequency fluctuation. A liquid nitrogen-cooled InSb detector (Teledyne Judson Technologies, J10D-M204) in conjunction with a low-noise transimpedance amplifier (FEMTO, DLPCA-200) was used to monitor the N2O cell transmission with 400 kHz bandwidth. The signals were acquired by a high-resolution PXI oscilloscope (National Instrument, PXI-5922) operated in the 22-bit, 1MS/s mode, and 106 data length. To estimate the laser intensity noise, the laser power was measured as well by removing the N2O cell. As clearly shown in Fig. 5, the EVC-QCL decreases the noise of the laser frequency in a wide frequency range except for the spikes between 200 Hz and 10 kHz. The origin of these spikes was not clearly identified; they may be related to ground-loop noise and a mechanical resonance of one of the piezo assemblies. Intensity fluctuations were found to slightly increase by a factor of two because of the optical feedback.
Typical frequency scanning by the 0.5 Hz triangle ramp voltage applied to the V-cavity piezo actuator (a) and the passive response of the regulator (reg.) A and B (b) in the Laselock. (c) and (d) show the N2O absorption spectra and etalon fringes, respectively. The times when the scan direction switched is shown (gray dashed lines). The red dot in (c) indicates the point where the frequency of the QCL was fixed for the frequency noise measurement shown in Fig. 5.
Typical frequency scanning by the 0.5 Hz triangle ramp voltage applied to the V-cavity piezo actuator (a) and the passive response of the regulator (reg.) A and B (b) in the Laselock. (c) and (d) show the N2O absorption spectra and etalon fringes, respectively. The times when the scan direction switched is shown (gray dashed lines). The red dot in (c) indicates the point where the frequency of the QCL was fixed for the frequency noise measurement shown in Fig. 5.
Frequency noise analysis. (a) The plot of power spectral density (PSD) with the β-separation line (green dotted line) and (b) the Gaussian linewidth depending on integration time, which equals the inverse of Fourier frequency (same color code).
Frequency noise analysis. (a) The plot of power spectral density (PSD) with the β-separation line (green dotted line) and (b) the Gaussian linewidth depending on integration time, which equals the inverse of Fourier frequency (same color code).
To evaluate the linewidth of the EVC-QCL, a method utilizing the so-called β-separation line31 was adopted. In this method, only noise above the β-separation line is focused on for the linewidth estimation, which contributes to the Gaussian line profile as high modulation index components. Noise below this line is assumed to contribute to the Lorentzian tails of the line profile which do not affect the laser linewidth. The integral of the PSD in the high modulation index region gives the Gaussian laser linewidth for an arbitrary integration time: T0, which is defined as an inverse of the start frequency of the integration (=1/T0). After subtracting intensity noise, we estimate the QCL linewidth depending on integration time [Fig. 5(b)]. For long integration times, the linewidth plateaus to values of 183 kHz and 1.1 MHz with and without feedback, respectively. This is equivalent to a factor-five linewidth reduction. The characteristic noise peaks observed in the EVC-QCL within the 200 Hz–10 kHz range mentioned above limit the reduction of the linewidth. In the future, we will attempt to further investigate the origin of this noise and attempt countermeasures. For timescales of the order of the photon lifetime of the ringdown cavity (∼100 μs), the linewidth without feedback slightly drops to a value of 980 kHz, in contrast to the case with feedback showing a significant improvement to 44 kHz. Due to the limited bandwidth of the transimpedance amplifier, evaluation of higher Fourier frequency was not feasible.
In addition, we checked the medium-term stability by monitoring the N2O signal for 2 min (not shown here). The root mean square of the frequency change of EVC-QCL was only 300 kHz, compared to the 1.1 MHz of the free-running QCL.
B. Cavity ringdown spectroscopy with the EVC-QCL
In this section, we demonstrate CRDS measurements with the EVC-QCL and investigate its performance for the 14C detection. Figure 6 shows a comparison of 100 typical ringdown signals acquired in identical experimental conditions except for the activation of the feedback lock with a 106 V/A gain current amplifier and similar incident laser power. Due to the high finesse of around 80 000 of our present CRDS cavity, the free-running QCL did not couple to the cavity efficiently nor with stable amplitude as revealed by distinct intensity fluctuations in the raw transmitted signals [gray lines in Fig. 6(a)]. After 100 signal averaging (black curve), the average ringdown amplitude is 51 mV with a 40% standard deviation. The signals with the EVC-QCL have higher repeatability and higher average intensity: the fluctuation is only 3.7% for an average amplitude of 264 mV.
Acquired ringdown signals with (a) the free-running QCL and (b) the EVC-QCL. The gray lines indicate the raw trace of 100 signals. The solid black line shows the averaged signal of each 100 samples and only the first point of the decay is plotted with the standard deviation. The exponential decay fitting result (red dashed line) is also shown.
Acquired ringdown signals with (a) the free-running QCL and (b) the EVC-QCL. The gray lines indicate the raw trace of 100 signals. The solid black line shows the averaged signal of each 100 samples and only the first point of the decay is plotted with the standard deviation. The exponential decay fitting result (red dashed line) is also shown.
The great benefit provided by the resonant optical feedback can finally be summarized in the sensitivity or detection limit of the cavity ringdown spectrometer. Commonly, the CRDS sensitivity is discussed using the Allan–Werle deviation.32 Figure 7 shows the calculated Allan–Werle deviations in terms of absorption coefficient or a fraction modern: F14C which is the ratio of 14C content in the sample to the one in the standard modern carbon sample. Here, the ringdown signals were acquired for 5 min at the fixed laser frequency at the wavelength without gas absorption. The CRDS gas cell was vacuumed below 100 Pa. The acquisition rate was approximately 55 ringdowns/s. An absorption coefficient of 2 × 10−10 cm−1 is equivalent to natural 14C abundance (F14C = 1) for our measurement conditions.13 The EVC-QCL improves the single-shot detection limit by almost an order of magnitude from 1.1 × 10−8 cm−1 Hz−1/2 with the free-running QCL to 1.2 × 10−9 cm−1 Hz−1/2. With the EVC-QCL, the detection limit of F14C = 1 is attained within a short averaging time of only 0.5 s. In addition, our system is stable for more than 10 s and the reachable detection limit in 10 s is F14C = 0.2. At longer timescales, a linear drift of the ringdown rate prevents the system from further averaging, which is mainly caused by an etaloning effect. This effect is usually the dominant noise source in CRDS, causing periodic oscillations in the background spectrum. The effect is changed by environmental drifts such as cavity temperature or pressure as well as the drift of the QCL frequency and appears as a systematic increase on the Allan plot for long timescales. We previously reported on techniques to suppress these effects13,33 and recent progress will be published elsewhere. However, in this experiment, these techniques were not implemented.
Allan–Werle plots in terms of the absorption coefficient or the fraction modern of 14C (F14C). The gray dashed line indicates F14C = 1, equating to the 14CO2 signal level of a natural abundance sample in our system.
Allan–Werle plots in terms of the absorption coefficient or the fraction modern of 14C (F14C). The gray dashed line indicates F14C = 1, equating to the 14CO2 signal level of a natural abundance sample in our system.
The green dashed line in Fig. 7 indicates the shot-noise limit34 of the measurement condition with the EVC-QCL based on the averaged amplitude of the ringdown signals. The gap between the line of the EVC-QCL and the shot-noise limit is still one order of magnitude due to the detector noise component. Reduction of detector noise is a promising way to improve the single-shot detection limit. Higher laser power results in a lower shot-noise limit and leads to improving the level of the single-shot deviation as well. Then, single-shot measurement with a detection limit below F14C = 1 will come into reach.
Finally, we performed a CRDS measurement with the EVC-QCL using a 50 μl 14C-labeled glucose solution sample, whose amount of 14C was 85 mBq within a 2.5 mg stable carbon designed for F14C = 150. The total acquisition time was 10 min. While the gas cell was usually cooled to below 253 K in order to suppress 13CO2 interference absorption,13 this measurement was conducted at 293 K as the quantity of 14C was far above the detection limit. Figure 8 shows the acquired absorption spectrum with nonlinear least square fitting using Voigt absorption profiles calculated by the HAPI package. About 200 ringdown signals were averaged for each wavelength point, with standard deviations too small for display on the linear plot. The clear bump due to the 14CO2 absorption was observed near the peaks of stable CO2 and N2O, which were well distinguishable from the noise. The residuals of the fitting are in the range of 10−9–10−10 cm−1 in regions with weak absorptions, which is in good agreement with the sensitivity evaluation by the Allan–Werle plot. However, close to the absorption lines, larger distortions are visible as seen in the residual plot in Fig. 8(b). These characteristic structures in the fitting residual are generally caused due to the uncertainty of the wavelength calibration or absorption peak position values as well as their respective intensities in the HITRAN database. In addition, the etaloning effect causes periodic large oscillations in the fitting residuals, which is seen around 2009.10 cm−1. To realize a precise measurement below the level of the natural abundance, etaloning suppression is still required. The EVC-QCL will help in developing and implementing these techniques. Cooling down the sample gas is also crucial. In the case of a sample with a large concentration of nitrogen such as biomedical samples, interference from N2O absorption will be another concern that needs to be considered.
Room-temperature CRDS spectrum acquired from the 14C glucose sample: (a) during the full frequency range and (b) the 14CO2 absorption region with the fitting residual in the bottom panel. The total pressure in the cell was 15.5 mbar. The partial pressure of CO2 and N2O were derived from a spectrum fitting to be 10 and 1.27 × 10−5 mbar, respectively.
Room-temperature CRDS spectrum acquired from the 14C glucose sample: (a) during the full frequency range and (b) the 14CO2 absorption region with the fitting residual in the bottom panel. The total pressure in the cell was 15.5 mbar. The partial pressure of CO2 and N2O were derived from a spectrum fitting to be 10 and 1.27 × 10−5 mbar, respectively.
IV. CONCLUSIONS
A compact-packaged narrow-linewidth mid-infrared laser system based on resonant optical feedback from an external V-shaped cavity (called EVC-QCL here) was developed for application in mid-infrared laser spectroscopy, particularly for our CRDS radiocarbon analyzer. For higher long-term frequency stability, a cordierite ceramic with a low thermal-expansion coefficient was chosen as a spacer of the V-cavity. Additionally, a two-layer enclosure insulated the whole laser system from the environment. A feedback power ratio k = 1% was achieved with only a power loss of 20% resulting in a lock capture range of more than 100 MHz in the free-running QCL frequency. From the frequency noise power spectral density, the estimated linewidth of the EVC-QCL for the integration time of 100 μs was 44 kHz. Feedback controls applied to the QCL current and phase piezo actuator enabled efficient coupling to the V-cavity. Mode-hop-free continuous frequency scanning up to 10 GHz/s within a second was enabled by a two-side V-cavity piezo scan, while a scan over 4.65 GHz/second was demonstrated in this paper with only single-side piezo scan due to a defect in one of the actuators.
We installed the EVC-QCL in the table-top CRDS apparatus for the 14C analysis and performed CRDS measurements. The ringdown signals were stronger and with smaller fluctuations after installing the EVC-QCL even with our CRD cavity with a very high finesse of around 80 000. The Allan–Werle plot indicates that we achieved a single-shot detection limit of 1.2 × 10−9 cm−1 Hz−1/2 with the EVC-QCL and the minimum detection limit after 10 s averaging was F14C = 0.2. A rapid analysis for 14C below natural abundance is suitable for various applications requiring trace 14C analysis: recently, CRDS radiocarbon detection was applied to ADME studies in drug development,35 quantification of a biogenic fraction,36 and radiocarbon emission monitoring in nuclear facilities.37
Lastly, a low-noise absorption spectrum was observed. The fitting residuals in the regions without any absorption showed good agreement with the Allan–Werle plots, while characteristic deviations were observed close to the absorption peaks, which are caused by the uncertainties of the wavelength calibration or the line-profile parameters in the HITRAN database. In addition, etaloning effects were also visible in the residual plot. For demonstrating the 14C analysis below natural abundance, etaloning suppression becomes increasingly important in the CRDS system with the EVC-QCL. This will be implemented in our CRDS system in near future. Furthermore, the narrow linewidth of the EVC-QCL and the resulting improvement in coupling efficiency enable etaloning-free and doppler-free new CRDS techniques utilizing nonlinear absorption effects in the optical cavity, such as SCAR or two-photon/two-color CRDS. Finally, we believe that our one-box packaged mid-infrared narrow-linewidth laser system as well as the related techniques will contribute to the development and wider expansion of molecular spectroscopy and its numerous applications.
ACKNOWLEDGMENTS
This research was partially supported by JSPS Grant-in-Aid for Scientific Research (B) (No. 18H03469), Grant-in-Aid for Early-Career Scientists (No. 20K15205) and PRESTO of the Japan Science and Technology Agency (JST) (No. JPMJPR19G7).
AUTHOR DECLARATIONS
Conflict of Interest
The authors have no conflicts to disclose.
Author Contributions
Ryohei Terabayashi: Conceptualization (equal); Formal analysis (equal); Funding acquisition (equal); Investigation (equal); Methodology (equal); Project administration (equal); Software (equal); Supervision (equal); Visualization (equal); Writing – original draft (equal). Keisuke Saito: Formal analysis (equal); Investigation (equal); Visualization (equal); Writing – original draft (supporting). Volker Sonnenschein: Conceptualization (equal); Investigation (equal); Methodology (equal); Writing – review & editing (lead). Yuki Okuyama: Formal analysis (supporting); Software (supporting). Kazuki Iwamoto: Investigation (supporting); Software (supporting). Kazune Mano: Validation (equal); Writing – review & editing (supporting). Yuta Kawashima: Resources (equal); Validation (equal). Tetsuo Furumiya: Writing – review & editing (supporting). Koji Tojo: Writing – review & editing (supporting). Shinichi Ninomiya: Resources (equal); Writing – review & editing (supporting). Kenji Yoshida: Resources (equal). Hideki Tomita: Conceptualization (equal); Funding acquisition (equal); Project administration (equal); Supervision (equal); Writing – review & editing (equal).
DATA AVAILABILITY
The data that support the findings of this study are available from the corresponding authors upon reasonable request.