1. Introduction

Photoacoustic spectroscopy (PAS) is a sensitive method in trace gas analysis. The photoacoustic signal is directly proportional to the optical power absorbed by the sample. The proportionality factor depends on the sensitivity of the detecting element (e.g., a microphone) and the geometry of the photoacoustic cell. PAS is basically a zero background technique, since absence of absorption should not generate any signal. Recently, tunable diode lasers have been combined with PAS (TDLPAS) [1–4]. This technique has the potential for compact, selective, highly sensitive, and relatively low-cost trace gas analysis.

Several ways of improving the sensitivity of PAS have been explored, for example acoustic cell resonance [5], wavelength modulation of the laser beam [1–3], and intracavity measurement [6]. Lately, attention has also been paid to the development and improvement of the sensing element, and thus novel sensing methods for PAS have been proposed [7,8]. Kosterev et. al. [7] introduced quartz-enhanced photoacoustic spectroscopy (QEPAS), where a small quartz tuning fork is used as a resonant element to detect the acoustic signal. The best sensitivity achieved to date with QEPAS is 7.2×10^-9 cm^-1WHz^-1/2 as measured for ammonia [9].

In this work we have applied tunable diode laser spectroscopy to a novel interferometrically-enhanced cantilever detection scheme proposed to be used in photoacoustic spectroscopy [8,10]. Prior to this work only a black body radiation source has been used with the cantilever-type microphone to generate the photoacoustic signal. The purpose of the present work was to demonstrate that the micromechanical interferometric cantilever device can be employed as a detector for TDLPAS, and that it has the potential for substantial improvement in the sensitivity of trace gas detection. In the experiments a noise equivalent sensitivity of 2.8×10^-10 cm^-1WHz^-1/2 was achieved.

2. Experimental

In the experiment a distributed feedback (DFB) telecom diode laser operating near 1572 nm (Furukawa Electric FOL 15DCWD-A81-19060) was used to excite the R(18) rotational line of the [00⁰0]_I→[30⁰1]_II vibrational CO₂ transition. This CO₂ band is fairly weak, but it was selected for the experiment on the basis of laser availability. Wavelength modulation at half of the detection frequency was used to carry out the intensity modulation for TDLPAS. This was performed by modulating the output of the laser current driver (ILX Lightwave LDX-3525). The maximum cw output power of the laser was about 50 mW. A cylindrical photoacoustic cell, originally designed for a black body excitation source [10], was used in the experiment. The diameter of the cell is 1 cm and the volume is 8 cm³.

The TDLPAS measurement setup is shown in Fig. 1. The cylinder is sealed with a calcium fluoride window at the front end and the micromechanical cantilever is located between two chambers. The interferometer resides in the chamber behind the cantilever.

 

Fig. 1. Experimental setup: SG, signal generator; ITD, current and temperature driver of the laser; DFBL, distributed feedback diode laser; FC, fiber collimator; W, window; C, micromechanical cantilever; MI, Michelson interferometer; PC, personal computer.

Download Full Size | PDF

The laser beam was collimated with a fiber collimator (Thorlabs F260FC-1550). The collimated beam was reflected back from the silver coated cantilever, thus traversing the measurement cell twice. The line of the DFB laser was tuned to the peak of the absorption line by adjustment of the laser temperature and the DC driving current. The wavelength of the laser line was measured with a wavelength meter (Burleigh WA-1500). At constant temperature (27 °C) the DFB laser was tunable with the current across three CO₂ absorption lines (the rotational lines R(16), R(18) and R(20) at 1572.018 nm, 1572.335 nm, and 1572.66 nm, respectively). These CO₂ lines are the strongest ones in the [00⁰0]_I→[30⁰1]_II vibrational band. The R(18) line was chosen for the demonstration measurements. The tuning characteristics of the laser and the spectral parameters of the R(18) line are summarized in Table 1.

Table 1. Characteristics of the measured CO₂R(18) rovibrational line and of the DFB laser used. The spectral parameters are taken from Ref. [11].

View Table

Wavelength modulation was accomplished by adding a sinusoidal small-amplitude current, of frequency f, to the DC driving current of the laser. The modulation current was adjusted to make the laser line sweep back and forth across the absorption line of the gas. Thus, the photoacoustic signal generated by line absorption in the gas appears at the frequency 2f. In addition to excitation of the gas molecules, the modulated laser beam may also generate photoacoustic background signal in the window, cell walls, and the cantilever. However, the background effects that may appear are generated at f and, therefore, are not present in the signal at 2f. This is a particular advantage of wavelength modulation over amplitude modulation. However, analysis of the signal formation shows that the line absorption signal contributes, in general, also to the peak at f, and that this contribution is minimized on the condition that the wavelength modulation is adjusted to be symmetric with respect to the absorption line maximum.

The measurements were made in static conditions with a total pressure of 250 mbar. CO₂ was mixed with dry nitrogen using calibrated mass flow controllers. The CO₂ concentration was varied between 0 and 100 %. The photoacoustic cell was placed on a support frame to reduce building vibrations. In the current setup the photoacoustic cell had to be removed from the frame each time the sample gas was changed. To change the gas sample the cell was first evacuated with a vacuum pump and then new sample gas was allowed to flow trough the cell for several minutes, after which the cell was closed and the gas sample reached quiescence.

The dimensions of the cantilever and the interferometer setup are shown in Fig. 2. The frequency response and the resonance frequency of the cantilever inside the cell depend on the cantilever geometry and mass; the surrounding gas pressure, temperature, and composition; as well as the total cell volume. For dilute CO₂ mixtures in N₂ at 250 mbar total pressure the resonance frequency of the cantilever used in this work was near 400 Hz. Above the resonance frequency of the cantilever the response is proportional to f ^-3 [8]. For this reason, it is not advantageous to operate at a detection frequency above the cantilever resonance. In the present experiment the modulation frequency was 163 Hz and the detection frequency, correspondingly, 326 Hz.

For interferometric measurement of the cantilever deflection the reference mirror in the interferometer was adjusted to direct one-half of an interference fringe to each of the photodiodes D1 and D2. The signals from D1 and D2 had a nominal 90° phase difference. The phase difference between photodiodes D1 and D3 was exactly 180°. A 16-bit data acquisition board and a personal computer were used to calculate the Fourier transform (FT) of the photodiode signals. Using the three photodiodes, the phase of the interferometer can be continuously tracked over a 2π phase change, corresponding to a cantilever deflection of one-half of the wavelength of the interferometer laser. In the current setup the FT processing was not phase selective with respect to the wavelength modulation.

 

Fig. 2. Dimensions of the cantilever pressure sensor and the interferometer setup.

Download Full Size | PDF

3. Results and discussion

A typical Fourier transform spectrum of the cantilever deflection signal in the range 0–500 Hz is shown in Fig. 3 for two different CO₂ concentrations (500 ppm and 5000 ppm) at 250 mbar cell pressure. The FT spectrum is an average of 100 individual FT signal spectra corresponding to a 110 s measurement time. The peaks below 50 Hz are due to building vibrations. The broad peak around 400 Hz is the resonance peak of the cantilever. The peak at f is practically absent from the FT signal spectrum for 500 ppm CO₂ concentration, whereas an f peak is clearly present at 5000 ppm concentration. The fact that the f peak could be made to vanish indicates that in this particular case the background effects were actually negligible and that the 2f peak represents all of the line absorption in this case. In contrast, the FT signal spectrum at 5000 ppm exhibits an f peak, as do, in fact, most of the measurements at other concentrations (not shown in the figure). As explained above, the f peaks may include contributions from background effects as well as line absorption. Furthermore, the resolution of the laser current driver (0.1 mA) was not sufficient for repeatably achieving fully symmetric wavelength modulation in each measurement. Thus, in most of the measurements there was a line absorption contribution to the f peak. This was further confirmed by the fact that the height of the f peak increased slightly with CO₂ concentration. It cannot be fully excluded that the procedure employed in changing the gas sample, described above, induced some misalignment of the laser beam, which could also lead to background effects. Therefore, there is a small uncertainty in the results caused by these effects, which will be overcome in further work.

 

Fig. 3. Fourier transform of the cantilever’s interferometer signal for two different concentrations of CO₂: 500 ppm (gray line) and 5000 ppm (black line).

Download Full Size | PDF

The background offset signal was determined separately for every concentration, since the broad resonance peak tends to vary slightly with different CO₂ concentrations [10], and, therefore, the background offset level at 2f also altered. In pure N₂ at 2 f there was no difference whether the laser beam was blocked or not, i.e. no background absorption at 2 f was observed in this case. The background offset level was determined at each concentration when the laser beam was blocked. The obtained background offset signals at 2f were subtracted from the photoacoustic 2f signals. In Fig. 4 the background-subtracted photoacoustic signal at 2f is presented as a function of the CO₂ concentration between 50 ppm and 5000 ppm at a total pressure of 250 mbar. The photoacoustic signal is linearly proportional to the CO₂ concentration. At higher CO₂ concentrations some nonlinearity appeared in the response. The increased slope of the signal curve at higher CO₂ concentrations suggests that there may exist alternative pathways, through which the absorbed energy is relaxed. The vibrationally excited ν₃ band of CO₂ is known to be in resonance with a metastable state of N₂ [12]. Due to this energy transfer N₂ decreases the photoacoustic signal from the ν₃ band. The nonlinear response at high CO₂ concentrations has also been reported with a higher overtone of the ν₃ band by Veres et. al. [13] who measured CO₂ with TDLPAS at 1.43 µm. Thus, the authors believe that the slowly relaxing metastable state of N₂ explains the observed nonlinearity since a part of the excited energy, although in a form of a combination band, is also stored in the ν₃ state. The expected linear behavior could be restored by using wetted nitrogen as a buffer gas, since water is known to promote the relaxation of vibrationally excited states into translational energy [13]. The division of relaxation pathways between rapidly relaxing states and metastable N₂ depends on the CO₂ concentration. However, at very low CO₂ concentrations the variation in CO₂ level is so small in relation to N₂ that the response is linear, as observed in our measurements. At intermediate CO₂ concentrations nonlinearity appears. On the other hand, with a phase selective detection setup the alternative relaxation paths could be studied in more detail.

Furthermore, the phase shifts between the three detector signals of the interferometer were, in fact, not exactly 90 degrees. This caused small periodical fluctuations of the signal (on a time scale of a few minutes). However, in a more recent design of the interferometer the phase adjustment is already substantially improved. The inset shows the photoacoustic signal at a constant 5000 ppm CO₂ concentration as a function of the laser power. As can be seen from the figure, the photoacoustic response to the laser power is indeed linear.

 

Fig. 4. Measured photoacoustic signal amplitude at 50, 100, 500, and 5000 ppm CO₂ concentration. Inset: Measured dependence of the photoacoustic signal on the DFB laser power at constant 5000 ppm CO₂ concentration.

Download Full Size | PDF

The noise level (σ=2.4×10^-5) of the present system was obtained as a standard deviation at 2f with the laser beam blocked. The noise level corresponds to a noise equivalent optical density (αl) of 1.60×10^-8. On the other hand, this corresponds to a minimum detectable CO₂ mole fraction of 7.2 ppm. As mentioned before, the present CO₂ absorption band is fairly weak. This explains why the minimum detectable CO₂ concentration is not very low in spite of the high sensitivity of the method. The noise equivalent (S/N=1) absorption coefficient normalized by the laser power (33 mW during the measurements) and the detection bandwidth is 2.8×10^-10 cm^-1WHz^-1/2.

In this work we have demonstrated that interferometrically-enhanced cantilever detection can be used with diode laser-based photoacoustic spectroscopy. The experimental results show that proposed system is very sensitive. The sensitivity of the cantilever detection method can be further improved with a photoacoustic cell that is more suitable for the laser source, e.g., a smaller cell having the cantilever placed at the side of the cell to remove any residual distortions arising from laser beam hitting the cantilever surface.