1. Introduction

After the introduction of cavity ring-down spectroscopy (CRDS) about two decades ago [1] a large number of applications in various field of molecular spectroscopy have been reported [2-4], some of which exploit multiplexing features to cover a large spectral range simultaneously. These broad-band cavity-enhanced absorption methods have developed, either as broad-band CRDS [5-9] or as broad-band cavity-enhanced spectroscopy (CEAS) employing pulsed lasers [10-12], or finally also as incoherent broad-band cavity-enhanced absorption spectroscopy (IBB-CEAS) [13].

The latter approach is experimentally very simple and robust. It thus has great potential for atmospheric trace gas detection and laboratory studies and was significantly developed in the past few years [14-23]. The combination of optical cavities with Fourier-transform spectrometers (FTS) has been mainly based on intra-cavity laser absorption spectroscopy (ICLAS) [24-28]. The first experimental implementation of FT spectroscopy in combination with cavity ring-down spectroscopy was reported by Engeln and Meijer over 10 years ago [29]. More recently a cavity-enhanced phase-shift setup was demonstrated using an incoherent light source [30]. The application of incoherent broad-band cavity-enhanced absorption spectroscopy (IBB-CEAS) in conjunction with a high-resolution FT spectrometer as the detector was demonstrated only recently [31] in the visible region of the spectrum. An advantageous aspect of high resolution methods is their species selectivity. In order to fully exploit the selectivity feature of FT spectroscopy, the near IR region is of interest since many overtone spectra of atmospherically relevant trace gases are located in this part of the spectrum. Many weak molecular overtone bands are yet unexplored and the higher energy part of the ground state potential of numerous species are still being studied theoretically, requiring more data based on highly sensitive experimental techniques.

In the present paper we report the first application of incoherent broad-band cavity-enhanced absorption spectroscopy in combination with a conventional (i.e. unmodified commercial) high-resolution Fourier-transform spectrometer in the near-infrared (IR) region, covering over 250 nm (~1000 cm⁻¹) between 1450 and 1700 nm (~5900-6900 cm⁻¹). The sensitivity and resolution of FT-IBB-CEAS is compared with long-path absorption spectroscopy using multi-pass reflection cells. The good agreement of the spectra obtained makes the FT-IBB-CEAS approach interesting for investigations of unstable species or radicals because of the much smaller sampling volume required for optical cavities as opposed to multi-pass reflection cells. It is expected that in the near future, the use of supercontinuum sources [22, 23, 32] will also lead to an additional, significant increase in signal-to-noise ratios, which is expected to make the combination of IBB-CEAS and FTS a very promising tool for both laboratory and field studies.

2. Experimental

The experimental set-up, schematically shown in Fig. 1, was mainly based on the one reported in [31].

 

Fig. 1. Sketch of the experimental set-up. Mirrors M₁ and M₂ are high-reflectivity mirrors forming an optical cavity. Lenses L₁ and L₂ are used to focus the light into the cavity and into an optical fiber, respectively.

Download Full Size | PDF

The Fourier-transform spectrometer was a Bruker IFS 125 HR with a dielectrically coated CaF₂ beamsplitter and a maximum spectral resolution of about 0.001 cm⁻¹. The spectra shown here were recorded at a resolution (unapodised) of 0.015 cm⁻¹ (note that the Doppler-width of CO₂ in the near-infrared region is about 0.013 cm⁻¹). The detector was an InGaAs photodiode at room temperature with high sensitivity in the near-infrared. The broad-band light source was a stabilized Xenon-arc lamp (Hamamatsu L2194, 75 W). The optical cavity was formed by two dielectric mirrors with a radius of curvature of 2 m and a specified reflectivity of R=0.997 (Laser Optics, Garbsen). The mirrors were attached to a stainless steel tube of 4 cm diameter and were separated by a distance d=90 cm; the resulting sample volume was about 1100 cm³. An external cavity diode laser (Toptica DL100) was used to facilitate aligning the optically stable cavity. The accurate determination of the mirror reflectivities by introducing a known quantity of a gaseous absorber (CO₂) into the cavity [15, 17, 18] will be discussed in the next section. To limit the spectral bandwidth to the region of high mirror reflectivity, optical bandwidth filters centered at ca. 1600 nm with a Full-Width at Half-Maximum (FWHM) of ca. 100 nm (BK Interferenzoptik Elektronik GmbH) were placed into the light beam in front of the cavity. The filters transmitted more than 1% of the light over a range of ca. 300 nm.

Compared to our first FT-IBB-CEAS experiments in the visible part of the spectrum [31], the optical interface between the cavity and the entrance aperture of the FTS was improved. The photons exiting the cavity were focused into an optical multimode fiber (Ocean Optics, 0.6 mm core diameter, 5 m length, numerical aperture 0.22) that was used to couple the light into the source compartment of the Bruker FTS. A lens made of CaF₂ was used to image the end of the fiber onto the entrance aperture of the Michelson interferometer inside the FTS. Optics were mounted on a fully adjustable optical holder and care was taken to homogeneously illuminate the entrance aperture of the FTS. The homogeneous illumination of the entrance aperture of the FTS caused line shapes to be fully symmetric and also improved the general signal-to-noise ratio. No residual etalon fringes due to potential interferences in optical components (e.g. in the cavity mirrors) were observed. The most critical factors that lead to an improvement in the signal to noise ratio in comparison to [31] are the use of a current stabilized (“super-quiet”) light source, a the better light coupling efficiency between the cavity and the FTS and the better sensitivity of the detector.

All gases were provided by Sigma Aldrich France and had the following purities: CO₂ (99.7%), OCS (99.5%), D₂O (98%), and H₂ ¹⁸O (98%). The pressure in the optical cavity was monitored using two calibrated MKS Baratron pressure transducers (with 1330 mbar and 13.3 mbar maximum readouts) with an absolute accuracy of 0.5%.

 

Fig. 2. High-resolution spectrum of CO₂ in the near IR obtained using an optical cavity in conjunction with a Fourier-transform spectrometer (a.u. = arbitrary units of intensity). The upper trace (shifted upwards for clarity) shows the spectrum of the empty cavity, the lower trace shows the spectrum obtained with 26.7 mbar of CO₂ in the cavity. The acquisition time is 90 min, the spectral resolution is 0.02 cm⁻¹. The inset in the left upper corner shows a weak overtone band (30011-00001) of CO₂ centered at 6503.08 cm⁻¹, and the inset in the right upper corner shows the P(18) line in this band to illustrate the signal-to-noise ratio, the spectral resolution, and the symmetry of the instrumental line shape.

Download Full Size | PDF

3. Results and discussion

High-resolution spectra of the evacuated optical cavity and of the cavity filled with 26.7 mbar of pure CO₂ are shown in Fig. 2 in the region between 5800 and 7000 cm⁻¹. The spectral resolution of the CO₂ spectrum is about 0.02 cm⁻¹ (apodized using the Norton-Beer “weak” apodization function); the spectrum was obtained from the average of 200 interferometer scans corresponding to an acquisition time of ca. 90 min. The spectrum of the empty cavity (recorded at a reduced resolution of 4 cm⁻¹) is shifted upwards by 0.2 divisions for clarity. Fig. 2 demonstrates the combined advantages of broad-band coverage, high spectral resolution, and high sensitivity, when using FTS together with IBB-CEAS. The quality of the spectrum in comparison to our previous proof-of-principle measurements of absorption lines of H₂O and O₂ (B-Band at 688 nm) in the visible [31] is improved due to factors outlined in the experimental section. The distortion of absorption features was prevented and an example of a fully symmetric line shape is shown in the upper right inset in Fig. 2. While the sensitivity is modest in comparison with CRDS experiments (a value of 7.5×10⁻⁶ cm⁻¹ Hz-^1/2 was determined for the noise equivalent absorption from the 1σ rms value of the noise in a region without absorption feature at 6665 cm⁻¹), the advantage of this approach lies in the broad coverage at high resolution.

The CO₂ bands in the spectral region shown in Fig. 2 have been studied extensively before (see [33] and references therein). The signal-to-noise ratio and spectral resolution of the data shown are best compared to those of FT spectra reported recently by Boudjaadar et al. [34]. Note, however, that in [34] a multi-pass reflection cell was used with a substantially larger sampling volume (factor of ~100) than in the present study. The same is true for the sampling volume in the FTS study by Miller and Brown [35], which is referred to in the HITRAN database [36] for the description of the 30011-00001 absorption band, shown in the left inset in Fig. 2. The 30011-00001 band has also been studied by laser based methods. Chou et al. [37] also used a multi-pass cell (100 m path length) and tuned an external cavity diode laser over each CO₂ line separately (lasting ca. one minute per line) at a CO₂ pressure between 21.3 and 26.7 mbar. He and Orr [38] used a swept-frequency variant of continuous wave CRDS to measure this band (over more than 30 cm⁻¹) in less than 5 min at ~659 mbar CO₂. Even though the latter study may have some advantages (in particular the shorter acquisition time) compared to the present approach (e.g. for environmental sensing or process control), it is more difficult to adapt over a substantially wider spectral range.

Broadband setups with a high spectral resolution have an inherent advantage over low resolution instrumentation. (A) Non Lambert-Beer absorption behavior, resulting in a broadening of absorption line widths due to an insufficient spectral resolution of the detection system, does essentially not occur. This problem has indeed been encountered before with broadband cavity-enhanced absorption instruments [7, 39, 40]. (B) If the expected absorption features are narrow and the overall absorption spectrum is not too congested so that the “zero-absorption” baseline (I₀) can be observed, an effective reflectivity loss (R_eff) can be easily determined by a single measurement with a calibration gas, as shown below. Moreover, if a mixture of a calibration gas and a target species is used for experiments, and the partial pressures of these gases are known, absolute absorption cross-sections (σ[cm² molecule⁻¹]) of the target species can also be determined in a single measurement. The following evaluation method, which is expected to be widely applicable in the near IR for high-resolution methods, has significantly smaller systematic errors concerning the measurement of R_eff (and thus of gas concentrations or absorption cross-sections) than low-resolution broadband cavity-enhanced absorption applications.

 

Fig. 3. (a) Intensity, I₀, transmitted by the empty cavity; the broad spectral structure is a result of the wavelength dependent lamp spectrum, filter function, reflectivity and detector response (a.u. = arbitrary units of intensity). (b) Effective reflectivity of the mirrors (see text below for the definition of R_eff) determined using 75 of the CO₂ lines in the panels below. (c) The integrated line intensities of the calibration gas, S_cal, from the HITRAN database. (d) Measured absorption spectrum plotted as fractional intensity change. Note that with increasing R_eff (i.e. towards higher wavenumbers) the CO₂ bands in panel (d) appear to be stronger than in panel (c). This is due to the increased effective optical path length in the cavity at higher effective R_eff. The effective absorption pathlength is ca. 33 m at 6200 cm⁻¹, 58 m at 6350 cm⁻¹, 180 m at ~6500 cm⁻¹, and 370 m at ~6680 cm⁻¹.

Download Full Size | PDF

The mirror reflectivity is generally not well defined in cavity-enhanced applications where the full mirror surface area is illuminated, due to the dependence of the cavity’s finesse on the cavity modes. The reflectivity found in an experiment where a multitude of transverse modes is excited is commonly lower than the reflectivity measured on the optical axis of the cavity only [41]. In addition to these genuine mirror reflectivity properties, other loss mechanisms that are weakly dependent on the wavelength, notably Rayleigh scattering, always increase the overall cavity loss, which is hence best described by an effective reflectivity (R_eff). This reflectivity must be calibrated for quantitative applications. Using a calibration gas at known partial pressure (P _cal [Pa]), R_eff can be determined from the known integrated line intensity of the calibration gas (S _cal [cm molecule⁻¹]), if the measurements are performed at high spectral resolution. Let d [cm] be the cavity length, T [K] the temperature, and k the Boltzmann constant, then the effective reflectivity as a function of maximal absorption wavenumbers (ν̃_max [cm⁻¹]) of the calibration gas is given by:

In Eq. (1) I _cal ($\tilde{\nu }$ ) [a.u.] is the intensity transmitted by the cavity on an absorption line of the calibration gas at wavenumber ν̃, and Ī₀(Δν̃) [a.u.] is the zero-absorption baseline averaged over a small wavenumber interval Δ $\tilde{\nu }$ in the vicinity of corresponding absorption line of the calibration gas. ν̃₀ is a spectral position off the absorption line (ν̃₀<ν̃_max) that defines the integral boundaries in Eq. (1). The choice of the integration limits depends on the line profile and line width. For example, in the case of a Lorentzian line shape (with broad wings) integration over 6 half widths at half maximum (HWHM) yields only 91% of the line area. In such cases, fitting of the measured data with an analytical expression for the line profile might be helpful, but requires knowledge of the line shape and – if possible – a negligible (or well known) contribution of the instrument function.

The evaluation approach outlined here is also valid for applications using coherent light sources or experiments with dispersive spectrometers as long as the measurements are performed at sufficiently high spectral resolution. An example of an effective reflectivity spectrum calculated with Eq. (1) is shown in Fig. 3 for the case of the CO₂ data shown in Fig. 2. Figure 3 also shows the spectra used in the calculation of R_eff(ν̃).

Knowing the effective mirror reflectivity the absorption coefficient, α, of the gas can be calculated [13]: α=[(I₀/I)-1] (1-R_eff)/d. An example is shown in Fig. 4, where a spectrum of the weak hot band (11121–00001) of CO₂ is shown, recorded at a pressure of P _cal=26.7 mbar. The upper and lower traces correspond to the observed spectrum, and the spectrum modeled with data from the HITRAN database [36] using Gaussian profiles with FWHM of ~0.02 cm⁻¹, respectively. Using pressure broadening coefficients of the HITRAN database Lorentzian FWHM of 0.008–0.013 cm⁻¹ were obtained. The Gaussian FWHM of the lines is of the same order of magnitude. Thus the resulting Voigt linewidths are about 0.011–0.018 cm⁻¹. Together with the instrumental line shape of the FTS (which was not studied in detail here) the measured lines have FWHM of about 0.015-0.020 cm⁻¹ (a value of 0.02 cm⁻¹ was used for the modeled spectrum based on the HITRAN data – lower trace in Fig. 4).

The inset in Fig. 4 shows the ro-vibrational line Q(18) at 6677.97 cm⁻¹ (solid circles) and the corresponding Gaussian fit (solid line). The area of 0.00236±0.00005 cm⁻¹ under the solid line in the inset corresponds to the value of the integral in the denominator of Eq. (1); since it is determined by a non-linear least-squares fit based on the analytical expression for a Lorentzian line shape (including the baseline offset as a fit parameter) it is insensitive regarding the exact choices of ν̃₀ and Δν̃, which were taken to be 6677.90 cm⁻¹ and 0.15 cm⁻¹, respectively (see inset in Fig. 4). Using the HITRAN line intensity of Q(18), S _cal=1.024×10⁻²⁵ cm molecule⁻¹ [36], P _cal/(kT)=6.54×10⁺¹⁷ molecule cm⁻³ at 300 K, and d=90 cm, yields an effective reflectivity of R_eff=0.9974 (corresponding to an effective optical path length of ~350 m). Note that using a Gaussian line shape in the fit yields a lower value of R_eff indicating the importance of the line shape used for the determination of the absorption area. Therefore, if the instrumental function (line shape) is not well characterized but contributes to the measured line profiles (as in our case), or if collisional narrowing is present (see [42, 43]), it is clearly preferable to integrate the measured spectrum to obtain the absorption line area, as described above. In this context it is important to minimize collisional broadening (i.e. the broad wings of the Lorentzian part of the line profile) so that the integration limits can be chosen as narrow as possible. For our measurements, the difference between the line area obtained by integration and by fitting with a Lorentzian line shape is only 3%, hence has negligible impact on the R_eff values obtained.

Knowing R_eff, the cross-section of a potential target species (σ) can be determined at wavenumbers different from the ones used in Eq. (1), if a mixture of calibration gas and target species is used:

In Eq. (2) I(ν̃) is the intensity transmitted by the cavity at wavenumber ν̃ on an absorption line of the target gas, Ī₀(Δν̃) is the zero-absorption baseline averaged over a small wavelength interval Δν̃ in the vicinity of an absorption line of the target gas, and P is the known corresponding partial pressure of the target gas. Alternatively also the target species’ line intensities, S [cm molecule⁻¹] or concentrations can be determined accurately.

 

Fig. 4. Small section of the high-resolution spectrum shown in Fig. 2 and Fig. 3. The upper trace shows the observed spectrum of the CO₂ (11121–00001) band [33]. The lower trace is a calculated spectrum using CO₂ line parameters from the HITRAN database (0.02 cm⁻¹ FWHM). The strong lines that are not reproduced in the modeled spectrum are due to residual water vapor in the interferometer. The effective optical path length is ca. 350 m corresponding to an effective reflectivity of about 0.9974. The inset shows a Lorentzian fit (solid line) of absorption line Q(18) (solid circles) in this hot band at 6677.97 cm⁻¹ (dotted vertical line) used in the calculation of R_eff. The limits of the integral in Eq. (1) are given by the first tick (ν̃0=6677.90 cm⁻¹) and last tick (ν̃₀+Δν̃=6678.05 cm⁻¹) of the inset’s wavenumber axis.

Download Full Size | PDF

Figure 4 also illustrates what a two component spectrum might look like, since the measured spectrum contains absorption features of a second species (compare upper and lower traces in Fig. 4). If the additional species had been deliberately filled into the cavity at a known partial pressure, P, its cross-sections or line intensities could have been evaluated with the approach outlined here. However, in the present case the additional lines in Fig. 4 are merely features due to ambient water vapor in the FT spectrometer (negligible vestiges of H₂O may also have been left on the walls of the cavity, but were verified to be insignificant for the evaluation of R_eff based on their well-known line strengths [36]).

The advantage of a high spectral resolution in the near IR with FT-IBB-CEAS is that baseline, R_eff-calibration, and target species measurements can be performed simultaneously. This reduces the influence of systematic errors on the result. The systematic error on σ is due to the (generally small) uncertainties in P, P _cal, T, d, S _cal and (1-R_eff). In the approach presented here, the systematic uncertainty of (1-R_eff) is reduced due to I₀ and I _cal being measured simultaneously. Hence the evaluation of R_eff is largely insensitive to light source fluctuations. Using integrated calibration cross-sections, i.e. line intensities S _cal rather than resolution-dependent cross-sections (σ _cal), also improves the accuracy. The systematic error on S _cal is limited by the uncertainty of available line parameters of potential calibration gases in spectroscopic databases (for CO₂ for instance the uncertainty of line strengths is of the order of 1% [34] in the region shown in Fig. 4) and the purity of the calibration gas. The errors in I₀, I _cal and I are of a statistical nature and can in principle be minimized through detector cooling and longer acquisition times, hence the error in R_eff and σ can also be minimized. Using more than several lines over a small wavelength region in the calibration procedure will further lower the contributions of I ₀, I _cal and I to the uncertainty of (1-R_eff), which can then be determined to within a few percent or even better. In this case the uncertainties of P, P _cal, and T start to become important. In the example above the error of (1-R_eff) equals approximately 3.3% (i.e. R_eff=0.9974±0.0001) assuming conservative relative uncertainties of Δd=1.0%, ΔP _cal=0.5%, Δ(kT)=0.5%, ΔS _cal=1.0% (systematic), and Δ(integral in Eq. (1))=2.7% as obtained after the fit (inset Fig. 4). If the error of the integral in Eq. (1) is neglected by using many absorption lines in a small spectral region, the lower limit of the error Δ(1-R_eff)=2.2%. This is significantly better than in previous IBBCEAS studies [17, 18] where NO2 had to be used for calibration purposes in the visible. Note that if a line fit is used in the evaluation procedure, the choice of the line profile may also contribute a systematic error (which was left out in the estimate above).

From the above it is obvious that measurements of spectra with an evacuated cavity (also shown in Figs. 2, 5 and 6) are not necessary for reflectivity and cross-section (or line strength) measurements. They are merely included in the Figs. (and arbitrarily shifted by two division) to illustrate the shape of the zero-absorption baseline for each measurement. If the pure baselines were in fact used for the evaluation of R_eff they would introduce a (small) additional systematic error to the measurement, since they were not recorded in the same measurement.

Having a way to measure R_eff easily due to the high resolution of the FTS approach is particularly important for broadband cavity-enhanced absorption measurements, since the zero-absorption baseline, which depends on the overall losses of the cavity (i.e. the wavelength-dependent mirror reflectivity and other wavelength-dependent loss mechanisms, e.g. Rayleigh scattering) is generally time-dependent and can change slightly from measurement to measurement. From R_eff the effective absorption path length, L_eff=d/(1-R_eff), of ca. 350 m at 6678 cm⁻¹ was calculated and used for the simulation of the synthetic spectrum in Fig. 4. Note that the reflectivity of the mirrors is lower in other regions of the spectrum, for example L _eff in the region at 6500 cm⁻¹ (inset in Fig. 2) is “only” about 173 m corresponding to a mirror reflectivity of about R _eff=0.9948. It is therefore important to use a calibration gas with broad spectral coverage; in this regard CO₂ is well suited for this purpose in the near infrared. Using mirrors with higher reflectivities and slightly longer cavities one can thus expect to achieve optical path lengths of several thousand meters (provided more light can be effectively coupled into the cavity, for instance by means of a supercontinuum source [22, 23, 31]). For example, Majcherova et al. [33] used mirrors with 0.99995>R>0.99984 in a laser application to measure the (11121–00001) band and achieved a minimal detectable absorption coefficient of 3×10⁻¹⁰ cm⁻¹ at 25.66 mbar of CO₂; however, measurements using ultimately high mirror reflectivities commonly cover a smaller spectral region. Therefore in general there is a trade-off between the achievable sensitivity, spectral coverage, choice of the light source, and required spectral and temporal resolution. One interesting approach to optimize these experimental aspects may be by using prism based cavities where applicable [23]. For broad coverage and high mirror reflectivities more powerful broadband light sources, such as supercontinuum sources [22, 23, 31], will provide a greater photon flux into the optical cavity and further improve the signal-to-noise ratio and thus sensitivity. A supercontinuum source can typically deliver an average power of ~1-2 mW cm⁻² nm⁻¹ in the near IR. Practically all of this light can be delivered to the entrance mirror of the cavity. In comparison, the Xe-lamp used in this experiment delivered roughly 15 µW cm⁻² nm⁻¹ at a distance of 10 cm, where the light was imaged by a lens. Not all of the collected light can be successfully imaged onto the cavity’s entrance aperture. Therefore a significant increase in the signal (> factor of 100) can be expected with a directional supercontinuum source.

Figure 5 shows a high-resolution overtone spectrum of pure OCS in the optical cavity recorded at a sample pressure of 26.9 mbar. The spectrum was measured 6 times with 200 scans and then averaged resulting in an overall acquisition time of 9 hours. Compared to spectra obtained with conventional multi-pass absorption cells as shown in Fig. 7 of [44], the spectral resolution and signal-to-noise ratio are indeed at least equivalent (note that the spectral position of the band shown in the inset of Fig. 5 is approximately 500 cm⁻¹ higher than that of the one shown in Fig. 7 of [44]). In [44] the authors studied nearly the same spectral region (5920-7140 cm⁻¹) with an optical absorption path of 42 m in a White-cell at an OCS pressure of 30.7 mbar. They used a spectral resolution of 0.009 cm⁻¹ (unapodized) and averaged a total of 1200 scans, hence the recording time must have been at least several hours. While working with a smaller sampling volume, similar results in terms of sensitivity and spectral resolution (some 0.01 cm⁻¹) were achieved with the FT-IBB-CEAS setup.

 

Fig. 5. High-resolution spectrum of OCS obtained using an optical cavity together with a Fourier-transform spectrometer. The upper trace (shifted upwards for clarity) shows the spectrum of the empty cavity, the lower trace shows the spectrum obtained with 26.7 mbar of OCS in the cavity. The recording time is 540 min, the spectral resolution is 0.02 cm⁻¹. The inset in the left upper corner shows the weak 3020-0000 overtone band of OCS with band edge at about 6649.83 cm⁻¹ [44].

Download Full Size | PDF

Finally, Fig. 6 shows a high-resolution Fourier-transform spectrum recorded with a mixture of 8.0 mbar of D₂O and 12.4 mbar of pure H₂ ¹⁸O in the optical cavity. The spectral resolution was 0.02 cm⁻¹. Even though H₂ ¹⁸O has been studied before in this spectral region by different techniques with typical absorption path lengths of up to 100 m [45], the superior absorption sensitivity of FT-IBB-CEAS in combination with D₂O being present in the cavity results in the observation of a large number of new absorption features in the spectrum in Fig. 6 notably due to the rare HD¹⁸O isotope of water. Over 4700 lines could be identified between 6000 and 7000 cm⁻¹, the vast majority of them are not saturated. A detailed analysis of this spectrum is currently in progress in our group [46]. The spectrum demonstrates the advantage of FT-IBB-CEAS in applications concerning rare (and often expensive) isotopic samples. Due to the much smaller sample volume of the optical cavity compared to long-path absorption cells, substantially smaller concentrations of sample molecules are required to achieve detectable optical densities. This is of course also an important advantage for other types of samples, such as toxic species, radicals, ions, flames and combustion plasmas, or liquids with very low vapor pressure. In this context FT-IBB-CEAS will also be an advantageous technique for applications with continuous molecular jets.

4. Conclusion

We report the first application of a combination of Fourier-transform spectroscopy and incoherent broad-band cavity-enhanced absorption spectroscopy in the near IR between 5800 and 7000 cm⁻¹. The spectra shown demonstrate the potential of combining FTS with IBB-CEAS, in particular when small sample volumes are required. The measurements demonstrate the usefulness of this approach for spectroscopic investigations of isotopic or dangerous samples, electric or plasma discharges, flames, or chemical sources like flow tubes in a steady state. However due to its modest time resolution the approach is less suited for kinetic studies. Furthermore it was shown that high-resolution spectra of molecular lines provide an improved way to calibrate effective mirror reflectivities, or absolute absorption cross-sections and line strengths, without the need of additional measurements using an empty cavity. An extension of this method into the mid-IR region using a synchrotron light source [47, 48] is presently in preparation.

 

Fig. 6. High-resolution spectrum of a mixture of 8.0 mbar of D₂O and 12.4 mbar of pure H₂ ¹⁸O in the optical cavity recorded with a spectral resolution of 0.02 cm⁻¹. The upper trace (shifted upwards for clarity) shows the spectrum of the empty cavity, the lower trace shows the spectrum obtained with the isotopic water mixture in the cavity. The recording time was 540 min, the spectral resolution was 0.02 cm⁻¹. The inset in the left upper corner shows the region from 6747 to 6757 cm⁻¹. More than 4700 absorption lines are observed in the 6000-7000 cm⁻¹ region, many of them are due to the rare water isotope HD¹⁸O. The horizontal dotted ‘zero’-line demonstrates that below ca. 6800 cm⁻¹ the absorption lines are not saturated.

Download Full Size | PDF