PAPER

www.rsc.org/analyst | The Analyst

Measuring acetylene concentrations using a frequency chirped continuous wave diode laser operating in the near infrared Ruth E. Lindley, Manik Pradhan and Andrew J. Orr-Ewing* Received 12th January 2006, Accepted 10th April 2006 First published as an Advance Article on the web 24th April 2006 DOI: 10.1039/b600506c Two frequency chirped continuous wave diode lasers operating in the near infrared (IR) at wavelengths of l y 1.535 mm and l y 1.520 mm have been used to measure acetylene concentrations using the P(17) and R(9) rotational lines of the (n1 + n3) vibrational combination band. The diode lasers were frequency chirped by applying an electrical current pulse to the laser driver at a repetition rate of greater than 1 kHz. As the laser is operated at high repetition rates, more than 1000 spectra per second can, in principle, be acquired and summed, allowing fast accumulation of data, rapid averaging and consequent improvement of the signal to noise ratio and detection limit. Experiments were performed using a single-pass cell with a path length of 16.4 cm, and also an astigmatic multi-pass absorption cell aligned to give a path length of 56 m. Detection limits corresponding to minimum detectable absorption coefficients, amin, of 5.6 6 1025 and 7.8 6 1028 cm21, respectively, were obtained over a 4 s detection bandwidth. These detection limits would correspond to mixing ratios of 21 parts per million by volume (ppmv) and 59 parts per billion by volume (ppbv) of acetylene at 1 atm in air, with the deleterious effects of pressure broadening accounted for. The single-pass cell was used to perform breakthrough volume (BTV) experiments for the low volume adsorbent traps used to pre-concentrate organic compounds in air, taking advantage of the capability of the system to measure concentrations in real time.

1. Introduction Semiconductor diode lasers have provided an important source of laser radiation for several decades, with pioneering research into their development carried out in the early 1960s.1 Since that time, they have developed to become among the most efficient of lasers, with an internal quantum efficiency of y70%, operating over a wavelength range of 0.7–30 mm with output powers as high as 50 mW at room temperature.2 Owing to their compact size, low power requirements and long lifetimes (>106 hours), they have played an important role in the telecommunications industry, which has led to further development in lasers as well as their associated optics and electronic components, reducing costs in recent years. Individual diode lasers can be tuned over a fairly limited wavelength range, typically of the order of a few nanometres. This wavelength tuning can be achieved in one of two ways; by changing either the temperature or the current passed through the semiconductor. It is often necessary to scan the diode laser wavelength (or frequency) for spectroscopic applications, to resolve an absorption feature. A standard method involves applying a ramp voltage to the current driver of the laser that slowly increases the current and, in turn, the output wavelength. This method has the advantage of continuous wave operation of the diode, allowing the wavelength to be monitored and calibrated by an external wavemeter as the ramp voltage is applied. School of Chemistry, University of Bristol, Cantock’s Close, Bristol, BS8 1TS. E-mail: [email protected]; Fax: 0117 925 0612; Tel: 0117 928 7672

This journal is ß The Royal Society of Chemistry 2006

Diode lasers are now used as light sources for a range of spectroscopic techniques, the most popular of which is tunable diode laser absorption spectroscopy (TDLAS). TDLAS is commonly combined with long path length astigmatic Herriott-type cells in order to measure low concentrations of analytes in gaseous samples. Applications include monitoring of atmospheric constituents,3 breath diagnostics4 and process monitoring in plasmas and flames.5,6 Diode lasers are also increasingly used in cavity ring-down spectroscopy7 (CRDS) and photoacoustic spectroscopy8 (PAS), which can achieve lower detection limits than standard absorption methods but at the expense of a more demanding experimental set up. Quantum cascade (QC) lasers are a new type of semiconductor device in which the wavelength of the emitted light is dependent on the properties of a set of square well potentials.9 They can be reliably operated either continuously (at cryogenic temperatures), or pulsed (at room temperature). Tuning the frequency of the emitted laser radiation can be achieved by varying the temperature, either slowly, using a Peltier device, or very rapidly by applying an electrical current pulse.10 The second method causes a frequency chirp as the temperature of the QC laser increases. QC lasers produce radiation in the mid-infrared (IR), covering a wavelength range of 3.5–11.5 mm. This range allows access to the fundamental vibrational bands of many molecules, and these bands may have cross-sections that are one or more orders of magnitude larger than the overtone and combination bands that occur in the near IR region. QC lasers are increasing being used in applications similar to those outlined above, and can be combined with long path length astigmatic cells to measure low concentration gaseous samples.11 Analyst, 2006, 131, 731–738 | 731

In this paper, we describe the application of the rapid tuning method established for QC lasers to a distributed feedback (DFB) diode laser by applying short current pulses to the laser driver. This gives a frequency chirp that can be used to scan over a small spectral region of interest at high repetition rates, with consequent advantages for rapid accumulation of spectra, data acquisition and sensitivity. This chirping of a DFB laser has been exploited previously to lock the frequency of a diode laser transiently to modes of a high finesse optical cavity,12 to study molecular de-phasing times by measuring free induction decays,13–15 and to record rapid absorption spectra of Rb atoms.16 Here we demonstrate for the first time the advantages of such frequency chirping in simpler approaches to rapid analytical absorption spectroscopy.

2. Experimental For each of the diode lasers, the frequency was chirped by applying a square voltage pulse to the laser driver at a repetition rate of greater than 1 kHz. During the current pulse generated by the driver, the diode emits laser radiation, and its output wavelength increases as the temperature of the diode rises. The down-chirp in frequency does not occur linearly with time, and so the changing output wavelength must be monitored. When the diode first begins to emit laser radiation, the wavelength increases most rapidly as resistance to the current passed through the diode heats it. External temperature control of the diode works against this current induced heating and the increase in wavelength thus slows in the later stages of the current pulse. An etalon can be used to monitor the changing wavelength with time and convert spectra to a linear wavelength or frequency scale. 2.1. Single-pass cell A schematic diagram of the experimental set up is shown in Fig. 1. The output from a distributed feedback (DFB) diode laser operating at wavelengths close to l y 1.535 mm (Marconi Optical Components, LD 6204) or l y 1.520 mm (NEL, NLK1S5GAAA), with an output power of ,10 mW and a bandwidth of y1 MHz, was initially passed through an optical isolator (OFR, IO-D-1550-SS) to prevent feedback to the

diode arising from back reflected light. The laser beam was then intercepted by a beam splitter which reflects a small portion (,10%) through 90u to an etalon. The etalon was comprised of two dielectric coated mirrors with high reflectivity in the near IR, separated by a distance of y12 cm. Light that passed through the etalon was collected by an optical fibre and transferred to a fibre coupled (FC) detector (New Focus, IR DC-125 MHz 1811). The light that passed through the beam splitter was directed by a prism through a single-pass cell fitted with quartz windows, and was then focused by a lens onto a fast response photodiode (Hamamatsu, G8370-01). The cell had a path length of 16.4 cm and was attached to a pressure gauge, gas sample line and pump. The gas samples used were acetylene (BOC Edwards, 97%) and a 1% dilution of acetylene in argon (BOC Edwards, 99.999%). A function generator (TTi, TG230) produced the square pulses input to the laser diode driver (Newport, 505B). These pulses, and the signals from the two photodiode detectors, were captured by a 14-bit, 100 MHz bandwidth digitiser capable of collecting data at a sampling rate of 100 MS s21 (National Instruments, PCI-5122). All signals were analysed on the PC housing the data acquisition card using LabVIEW software, either in real time or by co-adding spectra to improve the detection limit. The rate of acquisition was limited by the LabVIEW software, which typically accumulated spectra at a rate of 500 Hz. The frequency scales of such spectra were calibrated using the etalon transmission signal. 2.2. Astigmatic multi-pass absorption cell The set up for the experiments employing an astigmatic multipass absorption cell was similar to that described in the previous section, utilising the same light source and electronics. It differed only in the type of gas cell and the need for more careful laser beam alignment. The near IR light emitted from the diode laser was directed by two prisms into the multi-pass cell (Aerodyne, AMAC-76). The laser beam path was originally aligned using a visible diode laser (Maplin) operating at l y 650 nm with an output power of ,3 mW, and the near IR beam then overlapped with the optimum path with the aid of two irises. On exiting the cell, the near IR laser beam was coupled into an optical fibre which was attached to the FC photoreceiver.

3. Results and discussion

Fig. 1 A schematic diagram of the experimental apparatus. OI, optical isolator; B, beam splitter; FC, fibre coupled. The single-pass cell has a path length of 16.4 cm.

732 | Analyst, 2006, 131, 731–738

Representative examples of the 6.0 V peak to peak (Vp-p) square pulse sent to the laser driver, the driver output pulse received by the laser diode and the resultant signal from a photoreceiver placed behind the etalon, are shown in Fig. 2. The signal input to the laser driver was partially distorted by the driver electronics before it was input to the diode, with an y17 ms delay, and a slower increase to the maximum voltage. At 35 ms, the signal received by the laser reached a constant level and this continued until the square pulse input to the laser driver ended at 84 ms. The signal received by the laser diode began to decay at this point. The etalon signal in panel (c) illustrates the non-linearity of the laser frequency chirp, with the wavelength increasing rapidly at the start of the chirp and slowing down before the This journal is ß The Royal Society of Chemistry 2006

3.1. Single-pass cell

Fig. 2 (a) The Vp-p = 6.0 V square pulse sent to the laser driver at a repetition rate of 1.3 kHz. (b) The signal output from the laser driver and input to the laser diode. (c) The signal measured by the photodiode placed behind the etalon.

square pulse ends and the laser ceases to emit radiation. There was a short delay between the start of the pulse in panel (b) and the emission of radiation, which is due to the threshold current that must be attained. The increase in wavelength between 35 ms and the end of the pulse arises from heating of the diode, as the applied current is constant throughout this time period. For the experiments presented here, the square wave function sent to the laser driver was maintained at Vp-p = 6.0 V with a repetition rate of 1.3 kHz. The shape of the square pulse was adjusted so that the pulse lasted for 84 ms and the laser frequency chirp covered a wavenumber range of 0.35 cm21. This wavenumber scan corresponded to an increase in temperature of 0.9 uC. There was an off period of y700 ms between each chirp, which allowed the diode to cool to the externally maintained temperature before the next current pulse. The minimum off period required to allow this cooling was determined to be y400 ms at 1.3 kHz, which imposed a maximum chirp range of 0.875 cm21 at this repetition rate. The requirement for a 400 ms off period limits the rate at which the pulse train can be input to the laser driver, ultimately limiting the rate of data acquisition (assuming the data collection software can be made to run faster than our current version). Maintaining a 400 ms off period with a 0.35 cm21 chirp gives a maximum repetition rate of y2 kHz. QC laser chirp rates, in contrast, typically range from 10 s of kHz up to a maximum of 100 kHz. The frequency of the input pulse train is again limited by the temperature control of the device. The spectra acquired by the chirped DFB laser method can be analysed using the Beer–Lambert law. If I0 is the initial intensity of light passing through a sample with a concentration [X] and a wavelength dependent absorption cross-section s, contained within a path length l, the light exiting the sample will have intensity, I, given by: I = I0 exp(2s[X]l) This journal is ß The Royal Society of Chemistry 2006

(1)

The spectra shown in Fig. 3 are of the R(9) rotational line of the (n1 + n3) vibrational band of acetylene, taken using the chirp conditions described previously. Each spectrum consists of 2000 individual spectra accumulated and summed at 500 Hz. The detection limit calculated for this experiment is 6.62 6 1013 molecule cm23 for a 4 s optimum integration time, which would correspond to a mixing ratio of 2.7 parts per million by volume (ppmv) in a bulk gas such as air at 1 atm. Note, however, that the line shapes in Fig. 3 are determined by Doppler broadening, and the presence of a bath gas would induce pressure broadening which reduces the absorption cross-section at the peak of the absorption line. The pressure broadening coefficient of the R(9) line of acetylene in air was measured using the fast-chirp technique and was found to be 0.0779 ¡ 0.001 cm21 per atm. This is in reasonably good agreement with the HITRAN database,17 which lists a pressure broadening coefficient of the R(9) line of acetylene in air of 0.0808 cm21 per atm. In 1 atm of air, the peak absorption cross-section was reduced by a factor of y8, meaning a detection limit in ambient air of 21 ppmv for the single-pass absorption cell. The detection limit was evaluated from the signal to noise ratio, with the noise taken to be 26 the standard deviation of a section of the baseline. The optimum integration time was determined by Allan variance analysis of this experimental set up, a more detailed description of which is given in section 3.3. The demonstrated sensitivity corresponds to a minimum detectable absorption coefficient, amin = 5.6 6 1025 cm21. The inset in Fig. 3 shows the integrated area of the absorption line as a function of path length multiplied by concentration. The linear fit to the data set has a gradient equal to the line-integrated absorption cross-section and is

Fig. 3 The main figure shows the R(9) rotational line of the (n1 + n3) vibrational combination band of acetylene measured at four concentrations using the single-pass cell. The concentrations were achieved by varying the pressure of a 97% pure acetylene sample in the cell. The pressures were 1.79, 1.04, 0.71 and 0.40 Torr, giving acetylene concentrations of 5.65, 3.30, 2.26 and 1.27 6 1016 molecule cm23. The inset shows the integrated area as a function of path length multiplied by concentration.

Analyst, 2006, 131, 731–738 | 733

1.347 ¡ 0.0024 6 10220 cm2 molecule21 cm21. This compares well (to within 1%) with the HITRAN database17 value of 1.34 6 10220 cm2 molecule21 cm21, demonstrating that the method is suitable for precise, quantitative absorption measurements. The usual method of scanning the frequency of a diode laser over a molecular absorption feature is to apply a relatively slowly varying ramp voltage to the laser driver (fast sweeping of a ramp voltage introduces an additional nonlinear frequency chirp as we show below). This linear ramp scanning method was tested as a comparison with square-wave pulse chirping the frequency of the diode using the same rotational line of acetylene, and was carried out using an input to the laser driver of a sawtooth wave with Vp-p = 1.0 V at a repetition rate of 1.3 kHz (for fair comparison with the rapid chirp scheme). This input waveform gave a scanned wavenumber range of 0.37 cm21, which is comparable to that of the frequency chirped experiment using the 84 ms duration squarewave pulse (henceforth referred to simply as the fast chirped experiment). The sweep progressed over a 680 ms time period and spectra were acquired at a rate of approximately 500 Hz. Allan variance analysis was used to determine the optimum integration time, and was found to be the same as that for the fast chirped experiment. 2000 wavelength sweeps were thus accumulated and summed for each spectrum to mimic the conditions of the fast chirp data. The laser output wavelength sweep did not occur on a linear wavelength scale at the fast repetition rate used, because of an increase in temperature of the laser diode as the ramp voltage was applied, thus the frequency scales of acquired spectra were calibrated using the etalon transmission signal. The detection limit calculated for this varying ramp voltage method was 8.64 6 1013 molecule cm23, which gives amin = 7.3 6 1025 cm21. The detection limit for the fast frequency chirped experiment is a small improvement on that obtained by varying the ramp voltage over a similar frequency range at an equivalent repetition rate. The magnitude of the additional chirp induced by the y500 Hz ramp voltage was found to contribute y0.06 cm21 of the 0.37 cm21 wavenumber scan. This additional chirp is sufficient to warrant the implementation of an etalon to convert spectra acquired by this method to a linear wavenumber scale, as required for the square-wave chirped frequency method. The raw data obtained with the fast ramp voltage showed absorption lines on a steadily rising background because of changes in the laser output power with drive current. In contrast, the fast chirp with a square wave pulse generated raw spectra with a flat background level, a property used to advantage in measurements described in section 3.4. To avoid the additional non-linear chirp, the laser wavelength can be swept slowly by applying a ramp voltage at a much lower frequency (e.g., y10 Hz). This method of wavelength scanning was also tested, primarily to verify the purity of acetylene samples, and gave amin = 9.5 6 1025 cm21 for an optimum integration time of 500 s. The optimum integration time was determined through Allan variance analysis (described in detail in section 3.3). This method of frequency tuning gave comparable sensitivity to the fast chirped technique; however, data acquisition times in excess 734 | Analyst, 2006, 131, 731–738

of 8 min make it unsuitable for measuring rapid changes in concentration. The reproducibility and stability of the frequency chirp induced by square-wave pulses was assessed by co-adding increasing numbers of spectra of a single rotational line of acetylene and measuring the width of the resultant profile by fitting to a Gaussian function. If the swept frequency range is not constant, the absorption spectrum should broaden as increasing numbers of spectra are accumulated. The expected Doppler width (FWHM) of the absorption line investigated, the P(17) rotational line of the (n1 + n3) vibrational band of acetylene, is 0.015 67 cm21 at 21 uC, with negligible contributions to the linewidth from the y1 MHz laser bandwidth. Deviations of measured spectra from the expected Doppler width for integration periods between 0.2 and 4 s were a maximum of ¡ 0.000 05 cm21 (0.3% of the Doppler width) and thus negligible. For longer integration periods of up to 20 s the deviation increased to ¡ 0.000 15 cm21 (1.0% of the Doppler width). The stability of the frequency range swept by the fast chirped diode laser is thus excellent, and more than sufficient to allow rapid measurement of pressure broadening coefficients. 3.2. Astigmatic multi-pass absorption cell An astigmatic off-axis resonator cell is similar in design to a Herriott cell,18 but differs in that the latter employs two matching astigmatic mirrors separated by their common radius of curvature, whereas the former uses two mirrors with different radii of curvature.19 Light enters the cell through a hole in one of the mirrors and is reflected many times before exiting through the same hole. The pattern of spots where the laser beam is reflected fills the mirror surface, allowing a long path length while maintaining a small cell volume, which is advantageous in applications where there is a limited sample size. The modification to the astigmatic off-axis resonator offers the advantage of a more evenly distributed spot pattern on the cell mirrors. The spacing of spots on the mirrors is of great importance in a multi-pass cell, as this spacing determines the overall pattern size and thus the minimum cell volume for a given number of passes. Both Herriott and astigmatic cells require only simple alignment procedures, have a robust design that will maintain long optical path length alignment for long time periods and under mechanical vibrations, and efficiently suppress interference fringes. The astigmatic multi-pass absorption cell used in the current study had an optimum path length of 76.16 m and an internal volume of 0.5 l. The distance between the two mirrors was 0.32 m and so the optimum alignment required 238 passes through the cell. There was a second possible alignment of 55.68 m arising from 174 passes. Fig. 4 shows spectra of the P(17) rotational line of the (n1 + n3) vibrational band taken at increasing acetylene concentrations. The laser wavelength was rapidly scanned in the same way as described previously by applying an 84 ms, 6.0 Vp-p, 1.3 kHz repetition rate square pulse to the laser driver, and 2000 wavelength sweeps were accumulated and summed to form each individual spectrum, requiring a 4 s measurement time. The insert in Fig. 4 shows the variation in This journal is ß The Royal Society of Chemistry 2006

chemistry, reacting with the OH radical, the principal oxidising species in the atmosphere, as well as halogen atoms and the nitrate radical.21 The sensitivities presented here are as yet not sufficient for direct atmospheric acetylene measurements, but the rapid data accumulation rates may provide a useful tool when measuring fast concentration changes in other environments such as plasmas and flames. An application of the fast frequency chirped technique involving C2H2 detection is presented in Section 3.4. 3.3. Allan variance

Fig. 4 The main figure shows the P(17) rotational line of the (n1 + n3) vibrational band of acetylene measured at four concentrations using the astigmatic multi-pass absorption cell. The concentrations were achieved by varying the pressure of a y1% sample of acetylene in argon. The pressures were 1.17, 0.88, 0.65 and 0.34 Torr, giving acetylene concentrations of 3.75, 2.81, 2.09 and 1.08 6 1014 molecule cm23. The blown up section of the baseline is enlarged by a factor of 20 and shows interference fringes arising from collecting the transmitted laser light via a FC photoreceiver. The inset shows the integrated area plotted as a function of concentration. The path length can be derived from the gradient and is equal to 53.4 ¡ 2.7 m.

the integrated area of the spectral line as a function of concentration. Using the HITRAN database17 value of the absorption cross-section, the path length derived from the gradient of this plot is 53.4 ¡ 2.7 m. The error associated with this incorporates a 5% uncertainty in the concentration of the 1% acetylene in argon standard mixture used in the measurements. The detection limit calculated for the spectra shown in Fig. 4 is 1.74 6 1012 molecule cm23, with amin = 7.4 6 1027 cm21, again using 26 the standard deviation of baseline points as a reference noise level. In 1 atm of air (and without the deleterious effects of pressure broadening), this number density would correspond to a mixing ratio of 70 parts per billion by volume (ppbv), but, as has been discussed previously, pressure broadening will degrade the sensitivity by a factor of y8. Interference fringes are visible across the whole spectrum and arise from coupling the light exiting the cell into an optical fibre. The fringe pattern can be seen in the enlarged section of baseline in Fig. 4. Using a free space detector would reduce the effect of this noise source and improve the detection limit. To simulate the effect of a free space detector, the baseline oscillation was fitted to a sinusoidal function which was subtracted from the spectrum. The resultant detection limit was then estimated to be 1.83 6 1011 molecule cm23, which gives amin = 7.8 6 1028 cm21 for a 4 s accumulation time. Acetylene mixing ratios in the troposphere typically range from 0.8 to 2.5 ppbv in rural environments and have been measured up to 20 ppbv in urban environments.20 Acetylene has purely anthropogenic sources, and may be used as a chemical marker when determining the source of air packets in atmospheric measurements. It plays a role in atmospheric This journal is ß The Royal Society of Chemistry 2006

The Allan variance, s2A , provides a method to determine the optimum integration period for a particular system.3 It is evaluated as the time average of the sample variance of two adjacent averages of time series data, and can be determined for a range of integration periods. It is calculated using the formula given in eqn 2, where A1 and A2 are the first and second time series averages. 1 (2) s2A ~ ½A1 {A2
2 As the integration period is increased, the Allan variance should decrease as long as increasing signal averaging leads to a reduction in the noise level. Enhancement in sensitivity resulting from longer signal averaging will occur only up to a point where long-term instabilities (e.g., drift in alignment, temperature, laser power, etc.) come into effect and cause the Allan variance either to remain constant or increase. The Allan variance for the rapid frequency chirped method was measured using the P(17) line of the (n1 + n3) vibrational band with the astigmatic multi-pass cell for integration periods ranging from 0.008 to 72 s. The quantity investigated was the integrated area of an absorption line, which is directly related to the concentration of acetylene within the cell. Wall adsorption loss of acetylene was shown to be negligible. The calculated Allan variance as a function of integration time is

Fig. 5 A graph of the Allan variance, s2A , plotted against the integration time. The straight line is a linear best fit to the first section of the data set.

Analyst, 2006, 131, 731–738 | 735

shown in Fig. 5. Fitting a linear function to the first section of the data set, where the Allan variance is decreasing, allows an estimation of the optimum integration time. For this experimental set up the optimum integration time was found to be y4 s, which corresponds to 2000 accumulated spectra. After this point there is no further improvement in the Allan variance and there is no benefit in additional signal averaging. The detection limit and minimum detectable absorption coefficient for this integration period were given in the previous section. In the initial part of the graph, where the Allan variance is decreasing with increasing integration time, the gradient of the line of best fit is indicative of the noise source. A gradient of 21 corresponds to a white noise source and the Allan variance improves by the inverse of the integration period. The gradient of the line of best fit is 21.02 ¡ 0.04 s21, indicating that the improvement in Allan variance is indeed due to reduction of white noise with increasing averaging. After the 4 s optimum integration time, there is a small increase in Allan variance at longer integration times. This is caused by long term drifts in the stability of the system, as demonstrated by Werle et al.,3 and is most likely to be due to fluctuations in the laser temperature. If the temperature of the laser fluctuates, the position of the absorption line will shift within the swept spectral region. With increased integration time, this instability would lead to an apparent broadening of the absorption line accompanied by a decrease in peak height. 3.4. Breakthrough volume experiments To illustrate a potential application of the fast-chirping technique, absorption spectroscopy was used to measure the retention and breakthrough of acetylene by a range of adsorbent traps. Pre-concentration of very dilute analytes is a technique that is commonly used prior to gas chromatographic (GC) separation and subsequent detection for monitoring atmospheric constituents. Flame ionisation detection (FID), electron capture devices (ECD) and mass spectrometry (MS) are commonly used to measure the separated components eluting from a GC column.22 Pre-concentration of atmospheric gas samples was recently also demonstrated to be viable for enhancing detection limits for analysis by CRDS.23 An air sample is typically passed through a trap containing an adsorbent material; the adsorbent retains the trace molecules of interest while the rest of the sample is flowed to a vent. The trap can then be heated to desorb the retained molecules and a small volume of carrier gas used to transfer them either to the GC for separation and subsequent detection or, in the case of detection by CRDS, directly to the ring-down cavity. When performing a pre-concentration step it is essential to know the capacity of the adsorbent trap for the species to be retained. If more molecules are passed through the trap than it is capable of retaining, saturation will occur and analyte molecules will be lost to the vent, resulting in an underestimation in the concentration of the original sample. A standard procedure for assessing the capacity of the trap is to perform breakthrough volume (BTV) experiments.24 For the evaluation of adsorbent traps, the BTV is usually defined as the volume of gas that has passed through the trap before the 736 | Analyst, 2006, 131, 731–738

concentration of the target species in the effluent stream reaches between 5 and 50% of the concentration in the inlet stream, with the latter case described as the 50% BTV. The adsorbent used in all BTV experiments was molecular ˚ pore size of this zeolite material sieve 5A (MS 5A). The 5 A makes it a highly selective adsorber of certain small molecules such as ethylene and acetylene. MS 5A traps were made from 1/80 stainless steel tubing packed with zeolite, with glass wool and glass beads at either end to contain the adsorbent within the trap. The BTV experiments were performed using the single-pass cell with the apparatus shown in Fig. 1, but with the modified sample line shown in Fig. 6. The DFB laser operating at l y 1.535 mm was used to measure the absorption via the P(17) rotational line of the (n1 + n3) vibrational band of acetylene. Acetylene (97% pure) was passed through a mass flow controller (MFC) (MKS, 10 sccm N2), through a trap containing MS 5A and then to the cell. The flow rate varied between 2 and 4 sccm, but with a constant flow rate maintained throughout each experiment. The cell was held under vacuum, and pumped throughout with the pump line throttled in order to regulate the rate of gas flow through the cell. The 50% BTV was measured for 3 traps containing 50, 200 and 250 mg of adsorbent. The traps were cleaned between experiments by heating to 220 uC using a heat gun, while passing N2 through the trap at a flow rate of 10 sccm. The absorption of diode laser radiation by acetylene within the single-pass cell was analysed using a custom written LabVIEW program that obtained I and I0 values and thus calculated the concentration. I and I0 were taken as the maximum and minimum values of a data array spanning the absorption line. This method of analysis relied on a flat baseline level of accumulated spectra, which is achieved using the fast chirped technique as the output intensity of the laser is constant for the duration of the chirp. The accuracy of the programme was tested and found to calculate the concentration inside the cell to within 3% of the value determined by the increase in pressure. 300 fast chirp frequency sweeps were accumulated for each spectrum that was analysed to improve the detection limit, giving an actual sampling rate of 1.7 Hz.

Fig. 6 A schematic diagram of the sampling line and key elements of the optical set-up used for the breakthrough volume experiments. MFC—mass flow controller.

This journal is ß The Royal Society of Chemistry 2006

Although our demonstration of the principle of BTV measurements was made for a 97% acetylene sample, it is possible to measure 50% BTVs using diluted acetylene samples. Such measurements would be beneficial prior to a sample analysis requiring pre-concentration, in order to establish the volume of the dilute sample gas mixture that can be passed through the trap before the target molecules start to elute. Although the sensitivity is limited by the path length of the single-pass cell, the astigmatic multi-pass cell would allow detection of much lower concentrations of acetylene, enabling analysis with very dilute samples.

4. Conclusions

Fig. 7 Upper: The results from a BTV experiment for an adsorbent trap containing 250 mg of MS 5A. Acetylene was passed through the trap at a flow rate of 2.9 sccm. 50% breakthrough occurs at 171 s, which corresponds to a volume of acetylene of 8.3 ml. Lower: The 50% BTVs calculated for each of the traps as a function of the mass of adsorbent contained within the trap, and a best-fit straight line constrained to pass through the origin.

The results of a BTV experiment for the 250 mg MS 5A trap are shown in the upper panel of Fig. 7. In this experiment, acetylene was passed through the trap at a flow rate of 2.9 sccm. The flow was turned on at t = 0 s, and while t , 140 s no acetylene was measured in the detection cell as all molecules were adsorbed and retained by the trap. After 140 s, saturation of the trap occurred, and acetylene molecules passed through the trap and into the single-pass cell where they were detected. The concentration stabilised to a constant level after complete trap saturation, and this concentration was used to calculate a time of 171 s at which the 50% breakthrough occurred. The 50% BTV was thus 8.3 ml. This value may differ for mixtures of acetylene in, for example, air because other gases may compete for adsorption sites. The lower panel of Fig. 7 shows the results of BTV experiments performed on traps containing 50, 200 and 250 mg of MS 5A. As the mass of adsorbent increases, so too does the 50% BTV, with a clear linear dependence. The approximate number of acetylene molecules that have adsorbed to a particular trap during a BTV experiment can be calculated from the 50% BTV. For example, a trap containing 250 mg of MS 5A retained y1.9 6 1020 molecules of acetylene. MS 5A has a manufacturer specified (Jones Chromatography and Co.) surface area of 572 m2 g21, which can be used to estimate the surface area occupied by a molecule of acetylene as y0.8 nm2. This journal is ß The Royal Society of Chemistry 2006

Rapid (84 ms square-wave drive pulse duration, 1.3 kHz repetition rate) chirping of the frequency of a continuous wave diode laser has been shown to be an effective method of scanning the output frequency over a range of 0.35 cm21 for analytical spectroscopy applications. For a 4 s integration time, scanning the frequency this way gave a small improvement in sensitivity compared with applying a 10 Hz linearly increasing ramp voltage to the laser driver over a 500 s detection bandwidth. The optimum integration times for both systems were derived from Allan variance analysis. A comparison between the two methods, measuring the R(9) line of the (n1 + n3) band of acetylene in a 16.4 cm single pass absorption cell, yielded amin = 5.6 6 1025 cm21 for the rapid wavelength chirp compared with amin = 9.5 6 1025 cm21 for the linearly varying ramp voltage. Although the sensitivities are similar, the increased integration time for the linearly increased ramp voltage makes it unsuitable for rapid accumulation of data. The spectral range scanned by the frequency chirped diode laser is fully reproducible and the concentration or absorption cross-section can be accurately derived using Beer–Lambert law analysis. The Allan variance was measured for spectra of the P(17) line of the (n1 + n3) band of acetylene, using the fast frequency chirp method and an astigmatic multi-pass cell. Up to a 4 s integration period, the Allan variance decreased as increasing signal averaging improved the precision of the measured integrated area. After this time a small increase in Allan variance was observed as long-term instabilities, thought to be due to the temperature stability of the laser, affected the system’s performance. For a 4 s integration time, the detection limits were compared for the single-pass cell and an astigmatic multi-pass absorption cell with a measured path length of 53.4 ¡ 2.7 m. The latter gave amin = 7.4 6 1027 cm21 and was further improved to 7.8 6 1028 cm21 by removal of the interference fringes superimposed on acquired spectra by the light detection apparatus. The main advantage of rapid chirping of the frequency of a diode laser is the repetition rate at which spectra can be obtained. For the work presented here, a 1.3 kHz pulse train was used for successive frequency sweeps and complete spectral scans were accumulated at a rate of 500 Hz. This rapid data acquisition allows real time monitoring of fast changes in analyte concentration, as has been demonstrated by measuring the breakthrough volumes for a range of adsorbent traps. Analyst, 2006, 131, 731–738 | 737

Acknowledgements We thank Dr E. D. McNaghten and Dr A. M. Parkes from AWE for the loan of the astigmatic multi-pass absorption cell and for financial support through a CASE studentship for R.E.L. We are grateful to Professor G. Duxbury (Strathclyde) and Dr D. Martin (Bristol) for valuable discussion. Financial support from the EPSRC Portfolio Grant LASER, the EPSRC and Royal Society of Chemistry Analytical Trust Fund (R.E.L.) and the Dorothy Hodgkin Postgraduate Awards scheme (M.P.) is also gratefully acknowledged.

References 1 N. G. Basov, O. N. Krokhin and Y. M. Popov, Sov. Phys. Usp. (Engl. Transl.), 1961, 3, 702. 2 P. Werle, F. Slemr, K. Maurer, R. Kormann, R. Mu¨cke and B. Ja¨nker, Opt. Lasers Eng., 2002, 37, 101. 3 P. Werle, R. Mu¨cke and F. Slemr, Appl. Phys. B: Photophys. Laser Chem., 1993, 57, 131. 4 G. von Basum, H. Dahnke, D. Halmer, P. Hering and M. Mu¨rtz, J. Appl. Physiol., 2003, 95, 2583. 5 G. Lombardi, K. Hassouni, G. D. Stancu, L. Mechold, J. Ro¨pcke and A. Gicquel, Plasma Sources Sci. Technol., 2005, 14, 440. 6 R. G. Daniel, K. L. McNesby and A. W. Miziolek, Appl. Opt., 1996, 35, 4018. 7 M. Mazurenka, A. J. Orr-Ewing, R. Peverall and G. A. D. Ritchie, Annu. Rep. Prog. Chem., Sect. C: Phys. Chem., 2005, 101, 100.

738 | Analyst, 2006, 131, 731–738

8 J. Uotila, V. Koskinen and J. Kauppinen, Vib. Spectrosc., 2005, 38, 3. 9 G. Duxbury, N. Langford, M. T. McCulloch and S. Wright, Chem. Soc. Rev., 2005, 34, 921. 10 M. T. McCulloch, E. L. Normand, N. Langford, G. Duxbury and D. A. Newnham, J. Opt. Soc. Am. B: Opt. Phys., 2003, 20, 1761. 11 M. T. McCulloch, N. Langford and G. Duxbury, Appl. Opt., 2005, 44, 2887. 12 J. Morville, D. Romanini, A. A. Kachanov and M. Chenevier, Appl. Phys. B, 2004, 78, 465. 13 Y.-Q. Li, W. H. Burkett and M. Xiao, Opt. Lett., 1996, 21, 982. 14 P. Dube´, M. D. Levenson and J. L. Hall, Opt. Lett., 1997, 22, 184. 15 R. M. Macfarlane and M. Zhu, Opt. Lett., 1997, 22, 248. 16 Y.-Q. Li and M. Xiao, J. Opt. B: Quant. Semiclassical Opt., 1999, 1, 541. 17 The HITRAN database is available from www.hitran.com. 18 D. Herriott, H. Kogelnik and R. Kompfner, Appl. Opt., 1964, 3, 523. 19 J. B. McManus, P. L. Kebabian and M. S. Zahniser, Appl. Opt., 1995, 34, 3336. 20 ‘Volatile Organic Compounds in the Atmosphere’, ed. R. E. Hester and R. M. Harrison, RSC: Issues in Environmental Science and Technology, 1995. 21 Chemistry of Atmospheres, ed. R. P. Wayne, Oxford University Press, 2000. 22 Modern Practice of Gas Chromatography, ed. R. L. Grob and E. F. Barry, Wiley, 4th edn, 2004. 23 A. M. Parkes, R. E. Lindley and A. J. Orr-Ewing, Anal. Chem., 2004, 76, 7329. 24 G. Bertoni and R. Tappa, J. Chromatogr., A, 1997, 767, 153.

This journal is ß The Royal Society of Chemistry 2006