WO2013011253A1

WO2013011253A1 - Method and apparatus for gas monitoring and detection

Info

Publication number: WO2013011253A1
Application number: PCT/GB2012/000588
Authority: WO
Inventors: Edward Dennison MCNAGHTEN; Katherine Anne GRANT; Philip Adrian MARTIN
Original assignee: The Secretary Of State For Defence
Priority date: 2011-07-15
Filing date: 2012-07-12
Publication date: 2013-01-24
Also published as: GB201112171D0; GB2492841A

Abstract

A laser photoacoustic spectroscopy apparatus for detecting gaseous analytes comprising a plurality of tunable lasers (26), such as four fibre-coupled DFB-TDLs (26), and a photoacoustic cell (12) wherein each of the tunable lasers (26) is tunable to excite one or more analytes of interest. Preferably the photoacoustic cell comprises at least one cantilever microphone (13) and a compact Michelson interferometer (16) for measuring an acoustic signal produced by the at least one cantilever microphone (13). Preferably the outputs from the tunable lasers (26) are fed through individual single mode optical fibres (27) and combined using a four-to-one fibre coupler (28). The present in¬ vention can monitor and measure gaseous species simultaneously or sequentially using photoacoustic spectroscopy.

Description

Method and Apparatus for Gas Monitoring and Detection

The invention relates to gas monitoring and detection and more specifically to simultaneous detection of multiple gases.

Accurate monitoring of gaseous species at low concentrations has been required in a wide range of academic and industrial fields for many years. The most common scenarios include atmospheric chemistry, pollution monitoring, industrial process monitoring and control, breath analysis, agricultural studies, and materials ageing and life prediction studies. Data obtained from the analysis of gas samples obtained from materials ageing studies can provide an indication of the manner in which materials undergo chemical and physical change when confined for long periods under various atmospheres and serve as an aid to life prediction. Part of the monitoring and life prediction of materials includes assessing ageing processes that may occur over time. When making an assessment or prediction of a change in a material with time it can be beneficial to monitor the physical and chemical changes which occur during the ageing process. Chemical changes which occur during materials ageing processes are often accompanied by the evolution or consumption of gaseous species. For example, polymeric materials can produce methane (CH4) and other hydrocarbons, oxides of carbon (CO and C0₂) or formaldehyde (HCHO). Degradation of energetic materials can produce oxides of nitrogen (NO_x, where x = 1 or 2). Metals can undergo corrosion or embrittlement when exposed to some gases. Determination of the concentration of water vapour, oxygen and hydrogen in gas samples acquired from environments in which such materials are stored can provide an indication of the condition in which the materials are likely to be found.

Monitoring the evolution of gases by materials can identify the kinetics and mechanisms of ageing processes and aid the prediction of material lifetimes. It is desirable to detect gases at concentrations of a few ppmv (parts per million by volume) or in some cases ppbv (parts per billion by volume) in order to identify the onset of chemical processes at an early stage. Analysis of the gaseous headspace in experiments designed to artificially enhance material ageing processes (normally carried out at elevated temperatures) can provide data to inform kinetic models of material degradation and provide assurance that a material will still perform against its design requirement. A system capable of detecting the products of the ageing process can be useful in determining how quickly a material is deteriorating.

For materials ageing studies which rely on trace gas detection it is necessary for an instrument to allow unambiguous detection of at least a few simple molecules such as CO, CO₂, C₂H₂ and CH4. Often, several analyte species will evolve during degradation processes, therefore the ability to measure different compounds with a single instrument is advantageous.

In material ageing studies, the timescale over which changes in the concentration of evolved gases occur can be considerable. It is often necessary to measure the concentration of the analytes of interest every few minutes in a non-invasive manner over extended periods of time. The technique used for this purpose must therefore enable in-situ measurements to be made and should also provide a wide concentration W dynamic range as the quantities of gaseous degradation products could range from trace amounts (ppbv or ppmv) to percentage levels.

Gas analysis may be performed by optical methods such as Fourier transform infrared spectroscopy (FTIR) and non-optical methods such as gas chromatography and mass spectrometry. FTIR is a Useful technique for analysing samples as spectra of unknown or unexpected species present in the sample are recorded. However, FTIR instruments are usually unable to resolve isotopologues and overlapping absorption features in the spectra from similar molecules can make the full speciation impractical in samples containing complex mixtures. FTIR can be used in applications which require non-invasive analysis.

Gas chromatography (GC) and mass spectrometry have been used to analyse gas samples for many years and offer high sensitivity. Whilst these technique perform well in most situations, both require extraction of a gas sample from the environment of interest. GC may also require sample preconcentration if analytes of interest are present at concentrations below the instrument detection limit. Gases eluted from GC columns can be analysed by a variety of methods. The GC technique is time consuming, requires gas sample extraction and cannot deliver results with the speed which may be needed in certain circumstances.

Optically-based techniques offer an attractive means of non-invasive monitoring of gas composition in materials ageing and compatibility experiments and have the potential to overcome most of the limitations associated with FTIR, GC and mass spectrometry. Unambiguous analyte speciation is a prerequisite of any trace gas detection system. For an analytical system to be effective interference between concomitant species must be kept to an absolute minimum or eliminated. This is sometimes problematic for classical analytical techniques such as FTIR, GC and mass spectrometry.

Tunable lasers which operate in the mid and near infra-red (mid-IR and near-IR) spectral regions have found numerous applications in high resolution vibrational spectroscopy and trace gas detection in recent years. Mid IR lasers such as lead-salt tunable diode lasers (TDLs), quantum cascade lasers (QCLs) and difference frequency generation (DFG) sources provide access to the strong fundamental vibrational bands of molecules which possess IR-active vibrational modes. Near-IR emitting distributed feedback (DFB) TDLs have also found widespread applications in these areas; such devices allow excitation of vibrational overtone and combination bands of molecules.

Near-IR TDLs have advantages over mid-IR sources for applications which demand remote detection or in situ monitoring as their output can be delivered to the sample via fibre optic cables. The high selectivity provided by the narrow linewidth radiation (typically <10 MHz) generated by these devices usually guarantees unambiguous analyte speciation, often with isotopic resolution. In addition, the high spectral power density of the laser radiation ensures high sensitivity. Both of these attributes make near-IR TDLs attractive sources for many applications, including situations in which it is vital to detect the onset of chemical changes which generate volatile products. The development of near-IR emitting TDLs has given a new aspect to laser-based trace gas sensing as they can be used in conjunction with a variety of spectroscopic techniques and signal detection methods. AH of these sources can by used in absorption and photoacoustic spectroscopy (PAS) systems. A wide range of analytical systems incorporating these devices have been developed, examples of which are described in recent reviews (e.g. P. Martin, Chem. Soc. Rev; 31., 201 (2002), P. Werle et al., Lasers in Chemistry 1 , 255 (2008).

The most common applications of TDLs in spectroscopic systems to date include atmospheric monitoring, detection of airborne pollutants such as combustion products and volatile organic compounds, industrial process monitoring and human breath analysis. In addition to their applications in these fields, TDL-based methods can also be used to detect gaseous species released during ageing of polymeric or energetic materials. Most of the species released during ageing of such materials exhibit IR absorption, examples being CO, C0₂, CH₄, C₂H₂, C₂H₄, NO, N0₂, N₂0, NH₃ and HCHO. Most of these species to be detected at concentrations in the parts-per-million or parts per billion ranges using absorption or PAS systems which incorporate one or more TDLs. As these lasers have limited tuning ranges and an individual laser is usually only suitable for monitoring a single species analysis.

Spectroscopic applications of tunable diode lasers usually focus on absorption and photoacoustic (PA)-based methods. Tunable diode laser absorption spectroscopy (TDLAS) has the potential to distinguish between IR absorbing species without the need for prior separation. The degree of absorption is determined by the

concentration of the gas present, the optical path length and the molecular absorption cross-section of the analyte. The sensitivity of TDLAS instrument is limited by the smallest detectable change in optical intensity that can be measured. Increasing the length of the light/gas interaction region enables improved detection limits to be achieved and as a result TDLAS systems often incorporate multipass absorption cells such as astigmatic Herriot or White cells. However, the quantity of gas required to fill such cells can be a disadvantage. Cavity-enhanced techniques can be used to increase the absorption path even further and improve detection limits beyond those possible with TDLAS. The TDL-based PAS approach is becoming well established and offers a

complementary approach to TDLAS, offering high sensitivities (parts-per-billion detection limits) and simple experimental systems. A typical TDL-PAS system for gas-phase studies comprises most of the hardware associated with conventional TDLAS systems except for the cell and photodiode, these components being replaced by a PA cell containing an embedded microphone which acts as the signal detector.

PAS involves measurement of a varying pressure wave which is generated when a molecule in an excited state undergoes non-radiative relaxation. A PAS system has an optical source which is modulated before its output is introduced into a resonator, preferably at its fundamental resonance frequency. The acoustic wave is generated and then measured by a pressure transducer whilst a photodetector measures the optical power exiting the cell for power normalisation purposes. The data acquisition system can be tailored to the research requirements, and records and averages PA signals. Optical sources for PAS applications often include TDLs and QCLs. The benefit of high resolution provided by these lasers is of particular importance in low pressure applications where the absorption lines are narrow as a result of the reduced pressure. Any buffer gas can be used as the diluent in PAS, providing it is not reactive with the analyte or cell components. In general, the PA process tends to be more efficient in argon gas than in nitrogen.

In gas-phase PAS radiant energy deposited in the vibrational modes of an analyte is partly or fully dissipated to thermal energy by non-radiative relaxation processes. The net effect is the conversion of radiant energy to acoustic energy and PAS is therefore an indirect method of determining the amount of optical radiation absorbed by an analyte. The pressure wave generated in the relaxation process can be detected using a variety of pressure transducers. The photoacoustic signal (S) is directly proportional to the molecular absorption coefficient at the laser wavelength (a, cm^"1), the incident optical power (P, W) and a system-specific cell constant (C) according to Equation 1 :

S = PaC (Equation 1)

The acoustic wave has a wavelength of the order of a few centimetres to metres, enabling it to propagate much further than a thermal wave. Therefore, acoustic and thermal waves are spatially separated. The intensity of these changes is directly proportional to the concentration of the absorbing species.

The detection limit for the pressure transducer corresponds to the minimum detectable change in pressure which can be measured by the transducer. As the change in pressure is related to the concentration of analyte species (c), the detection limit can be calculated using Equation 2: detection limit = c/SNR = (3S_dcv/S)c (Equation 2) where Sdcv is the standard deviation and S is the signal strength.

The experimental conditions in PAS can vary considerably from one PAS system to another, most notably in the optical power, P, and the signal averaging time, t.

Therefore, the normalised sensitivity (cm ¹ WHz^"l/2) is used to normalise

measurements in a PA regime for power and signal averaging time and allow PAS systems to be compared more easily.

The PAS technique is complementary to absorption spectroscopy and uses similar equipment. However, the TDL-PAS method has a number of distinct advantages over TDLAS. The techniques are fundamentally different. PAS is a zero background technique in which only the analyte in the sample cell contributes to the PA signal whereas absorption-based techniques rely on the measurement of transmitted optical radiation, this necessitating the measurement of very small changes in a large background signals. PAS provides a linear signal response with analyte concentration over several orders of magnitude and measurements of concentrations can be made over a wide concentration dynamic range (e.g. ppbv to percentage levels). High sensitivity (up to pptv (parts per trillion by volume)) can be achieved in PAS and single point calibration for each sample pressure is possible. As PAS systems incorporate a wavelength-independent detector no dispersive optics are required and optical alignment is simple. Small sample volumes (a few ml) can be analysed by PAS, thus enabling miniaturised systems to be developed. Single -point/multi-point analysis is possible with PAS and multipass cells are not required the sample can be passed through the cell reducing wall adsorption effects. Finally, the PAS technique is not subject to optical interferences which can have an impact on the detection limits achievable.

The most important advantages of PAS as compared to absorption techniques are its ability to detect the signal against "zero" background, provide a wide dynamic range of measurement and analyse small sample volumes. The last of these features provides scope for minimising the cell dimensions, which can be important in applications where space is limited.

The microphone is a key component of any PAS system and must be chosen with care in order to optimise the performance of the system. Capacitive microphones have been the traditional choice for PAS applications for many years. These devices contain a conducting, flexible, membrane which stretches when pressure variations occur in the surrounding atmosphere. These devices provide high response with minimal signal distortion. However, the pressure wave must contain sufficient energy to cause the membrane to stretch and induce a change in capacitance in order for a signal to be observed. An alternative detector based on micro-electrical-mechanical systems (MEMS) technologies have been developed in recent years by Kauppinnen et al. in Microchem. J. 76, 151 (2004). This uses a miniature silicon cantilever microphone, the functionality of which involves translational motion of the detecting element without the need for stretching. Detection of acoustic signals by cantilever microphones is achieved by measuring the deflection of the free end of the cantilever by a compact Michelson interferometer embedded in the gas cell. Translational processes are associated with cantilever microphones, in contrast to the stretching involved in capacitive microphones. As translational processes require less energy cantilever microphones offer higher sensitivity: the movement of the cantilever's free end can be at least two orders of magnitude greater than the movement of the midpoint of a strained membrane for an equivalent pressure change. Sensitivities of the order of 10 ¹⁰ cm~ ' WHz^'"² have consistently been demonstrated for single species detection with cantilever cells, such as by Kauppinen et al. in the abovementioned reference. The smallest reported sensitivity obtained with this type of cell is 1.7 x 10^" cm^•'WHz^""²; this was achieved for the detection of C0₂ at 1572.327 nm using a TDL source (V. Koskinen et al., Applied Physics B 86, 451 (2007)).

An acoustic resonator is less susceptible to Iff noise because the modulation frequency is usually in the kilohertz range. Noise will be amplified as well as the signal but its amplitude decreases with increasing frequency. Window noise can be reduced by including acoustic baffles between cell windows and the resonator.

However, care must be taken to avoid introducing pressure fluctuations due to turbulence with the gas in flowing systems. Other drawbacks of the use of resonant cells include sensitivity to parameters that affect the cell resonance frequency (e.g. pressure, temperature and buffer gas composition). Modifying a cell to obtain resonance characteristics may instigate an increase in its volume, which in turn increases the probability of gas adsorption onto internal surfaces. As adsorbed gas does not contribute to the photoacoustic signal thus slower cell response and longer residence times are observed. The performance of a diode laser-based photoacoustic system depends crucially on the PA signal, which is proportional to the incident laser power (Equation 1 ). This has a direct bearing on the detection limit. To improve the detection limit of PAS, lasers of higher power may be used. Much of the early laser-based PAS gas sensing work involved the use of high power (-10 W) mid-IR lasers. More compact near-IR tunable diode lasers have output powers in the milliwatts range and therefore have limited scope in PAS applications due to the reduction in detection sensitivity incurred using these sources. Despite this, the small size, low cost, reliability and room temperature operation of these devices compensate for their limited powers. The disadvantages arising from the limited power obtainable from TDLs can be overcome using fibre amplifiers in conjunction with TDLs, providing that the TDL operates in the gain region of the amplifier. This allows the power available to be increased from milliwatts to typically 0.5-2W.

PA signals embedded in noise can be recovered using wavelength modulation spectroscopy (WMS) combined with a phase-sensitive detector such as a lock-in amplifier (LI A). Of course, this introduces an added layer of complexity and additional cost to the experimental arrangement. However, modulation techniques are key to achieving high sensitivity and noise reduction and are used in many

applications of the TDLAS and PAS techniques.^

Several TDLs can be used in a diode laser spectrometer to extend the analyte range and facilitate multispecies detection. This approach makes use of multiplexing techniques to facilitate simultaneous/quasi-simultaneous detection of a number of analytes and involves coupling the outputs from several TDLs into a single optical fibre, for example a silica glass fibre for delivery to the sample of interest. Silica glass transmits near-IR and visible wavelength radiation efficiently, but does not transmit in the mid-IR region. Near-IR TDLs are therefore the lasers of choice for multiplexed systems.

The measurement of multiple components in a gas mixture using PAS and TDL sources can be achieved in a variety of ways. The simplest method involves use of a single laser which has a tuning range covering the absorption lines of the analytes of interest. Coincidental overlap of the spectral features of different gases can be used provided the wavelength of the diode laser can be tuned sufficiently far in a single sweep. This method relies on well separated lines with no strong overlapping.

Measures taken to resolve spectral lines usually involve reducing the sample pressure to 100 mbar or less. The output wavelength range of the laser diode must coincide with lines from two or more species for this approach to be effective as multispecies detection method. In theory a single laser can by used, providing its tuning range covers spectral features of the different analytes. In Appl. Phys. B 82, 495 (2006) Scotoni et al. demonstrated this by simultaneously detecting NH₃, CH₄ and C₂H₆ using a single TDL which produced output in the 1.63 μη region. However, scope for using this approach is rather limited in view of the narrow tuning range afforded by the lasers and the rarity of spectral near-coincidences in the near-IR region. In Appl. Spectrosc. 65, 108 (201 1 ) Cai et al. demonstrated simultaneous detection of CO and C0₂ using a single DFB diode laser operating near 1.57 μπι It is more common to employ several lasers in a system, each of which is dedicated to the detection of a single analyte. This can be achieved using a separate detector for each laser, as demonstrated by Ebert et al. in Proc. Combust. Inst. 28, 423 (2000). If there is no convenient overlap of several spectral features within the wavelength tuning range of a single laser more complex techniques such time division

multiplexing (TDM) and modulation frequency division multiplexing (MFDM) can be used. TDM and MFDM involve using a number of lasers and provide the most promising solution for multispecies PAS detection. These techniques involve combining the laser outputs and provide sequential or simultaneous detection of the analyte signals respectively. Both methods only require a single detector.

TDM is the simplest multiplexing method to implement and requires only one detector. In this regime the TDLs are operated in sequential mode and the sample experiences only one excitation wavelength at a time. TDM is relatively

straightforward to implement and is usually the method of choice for multispecies detection in PA cells which contain capacitive microphones.

MFDM provides an alternative approach and enables simultaneous detection of different species. This technique involves modulating each laser at a different frequency and detecting the signals simultaneously using a single detector. In conventional MFDM ac and dc components are applied to the laser driver to scan across the transition of interest and the lasers are modulated at different ac

frequencies. The ac component applied to each laser is then used to demodulate each response from the combined signal. The composite signal is demodulated into separate frequency components using a lock-in amplifier. MFDM can suffer from "cross-talk" between concomitant species because signals from two channels can interfere with each other. The potential for "cross-talk" occurring is dependent on the differences between the modulation frequencies applied to the lasers. This is more likely to occur if the frequencies are not sufficiently different. Simultaneous detection of multiple species using TDLs in conjunction with the MFDM technique has been demonstrated in absorption experiments reported by Gerard et al. in Appl. Opt. 46, 3937 (2007).

The sensitivity of trace gas detectors is often determined by measuring the signal to noise ratio (SNR) observed in spectra of a gas. High responsivity to noise will lower the usefulness of the spectrometer when measuring small gas concentrations. The noise sources in PAS systems are a composite of ambient acoustic noise, acceleration noise, Brownian noise and electrical noise; although in certain circumstances one type of noise contributes above all others.

Ambient acoustic noise is generated from external acoustic waves which leak into the cell. High frequency mechanical vibrations produce acoustic noise when external structures connected to the PA cell vibrate. This type of noise can be reduced by isolating the PA cell from noise sources by placing it into a sound proofed container and by using flexible tubing in the gas transfer line instead of stainless steel tubing. Acceleration noise is generated from movements of the transducer because of external disturbances acting on the system. These usually arise from low frequency vibrations which couple directly into the support structure of the cell and therefore directly to the pressure transducer. Brownian noise is random and due to thermal fluctuations arising from molecules colliding with the transducer. Acceleration noise is the main noise associated with the cantilever technique.

Vibrational components that are perpendicular to the surface of the cantilever will cause the cantilever to bend and produce noise. This is minimised by ensuring that that plane of the cantilever pressure transducer is parallel to the direction of vibration components. Acceleration noise is also generated from the acceleration of gases inside the cell. This latter effect can be compensated for by using proper geometrical designs.

In the cantilever design the cell is always operated below the fundamental resonance frequency of the resonator because operating at resonance will increase the noise by the same factor as the PA signal. This method for compensating the acceleration noise in the cantilever design is crucial because the benefits of operating at the resonance frequency of the cantilever cell are redundant.

The performance of a TDL-PAS system can be enhanced significantly by using a fibre amplifier to increase the optical power available across the tuning range of the laser Although multi-species detection using TDLAS regimes is already commonplace there is no commercially available analytical instrument which exploits TDL technology, multispecies detection and PAS together with or without fibre- amplification. The present invention overcomes these problems described above by providing a system incorporating these methodologies while using low and high optical power sources and which can be used to deliver results in much shorter time than in the prior art. To date, investigations using TDLs in conjunction with cantilever-based PA cells have focused on single species detection exclusively although detection of multiple species with high sensitivity is desirable for many applications. Accordingly, the present invention provides a tunable laser photoacoustic spectroscopy apparatus for detecting gaseous analytes comprising a plurality of tunable lasers and a photoacoustic cell wherein each of the tunable lasers is tuned to excite one or more analytes of interest. The present invention also provides a method for detecting multiple trace gases by using a tunable laser photoacoustic spectroscopy apparatus.

The novel tunable laser photoacoustic spectroscopy system comprises a plurality of lasers and a photoacoustic cell containing a cantilever microphone.

In an embodiment of the present invention the apparatus comprises at least one cantilever microphone. The displacement of the cantilever can be measured by a compact Michelson interferometer. The pressure wave produced from absorption of the modulated radiation in the photoacoustic cell results in deflection of the cantilever which can be processed by (i) a Fourier transform of the signal or (ii) demodulating the signal using a phase-sensitive detector. The Fourier transform method

deconvolutes the signal into all of its frequency components, thus retaining all frequency information and as such the entire frequency spectrum can be monitored simultaneously. Demodulation of the signal using a phase-sensitive detector (lock-in amplifier (LIA)) deconvolutes the signal and ensures that only certain frequency components are retained, these being determined by the modulation frequency and bandwidth of the LIA. This method filters all frequencies which do not fall within the bandwidth of the low pass filter and as such only a small portion of the frequency spectrum is retained.

The number of tunable lasers can be equal to the number of analytes to be detected although a laser can be used to detect more than one analyte. The power of the lasers can be increased using a fibre amplifier.

Careful adjustment of the thermoelectric cooler temperatures specific to each laser allows their wavelengths to be tuned to the centre wavelengths of the analyte absorption lines and a Burleigh WA 1000 wavemeter can be used to monitor the wavelength of the radiation emitted by each laser. The wavelengths of the tunable lasers are tuned to the analyte absorption lines of the analytes by adjusting the laser temperatures, after which their outputs were coupled into a single fibre using a four- to-one fibre coupler. The combined collimated output is directed towards the sample chamber of the cantilever ceil by two gold plated planar mirrors and focused into the cell by an anti-reflection coated calcium fluoride lens of 30 cm focal length. The cell is located in a position which ensured that the focal point occurred just below the free end of the cantilever microphone.

MFDM experiments are performed by irradiating the gas sample with the outputs of four DFB-TDLs simultaneously. In each case the amplitude of the modulation applied to a laser was optimised to ensure that the scanned wavelength range exceeds the full width of the absorption feature of the analyte of interest The outputs from the lasers can be coupled into a combined beam. Further, the laser outputs can be coupled into a single mode optical fibre. Preferably, the optical fibre is made from silica glass. The combined beam is used to excite the analytes of interest either simultaneously or sequentially.

Frequency division multiplexing is used to acquire the measurements simultaneously.

Sequential measurements are taken when each of the lasers are operated at a different time to another laser by time division multiplexing.

The acoustic signals produced are measured by a lock-in amplifier. A lock- in amplifier can be used to demodulate the acoustic signal. These acoustic signals can be processed by a computer program.

The PA signals registered by the cantilever are Fourier transformed to allow the full frequency spectrum to be recorded. Modulation frequencies are limited by the cantilever response with frequency and are therefore tens of Hertz rather than in the kilohertz range.

The modulation frequency is not matched to the resonance frequency of the sample chamber as hysteresis effects are significant for the cantilever at high modulation frequencies (kHz) and deteriorated the performance.

The tunable lasers are preferably tunable diode lasers or are wavelength tunable. The invention also relates to a method of using the apparatus. The invention will now be further described with reference to the accompanying figures.

Figures 1(a) to (d) show the transitions targeted with each laser in a four laser system. The band spectra of CO, CO2, C₂H₂, and CH₄ are shown. The intensities of each line reflect the populations of the rotational levels at 298 . The transition lines chosen are presented in Table 1.

Figure 2 is a schematic diagram of the cantilever PAS cell showing the optical path of the laser beam through the sample chamber. The movement of the cantilever microphone is monitored by a Michelson interferometer.

Figure 3 is the experimental arrangement for multispecies detection using four fibre coupled DFBs and the cantilever photoacoustic cell. FC = fibre combiner, PM - powermeter, f( = modulation frequency of laser i, λί = wavelength of laser i.

Figure 4 shows a typical C₂H₂ photoacoustic signal obtained when using the cantilever cell.

Figure 5 comprises a main graph and an inset. The main graph shows the variation in magnitude of the second harmonic C₂H₂ photoacoustic signal at constant analyte concentration (0.5 %) and 1000 mbar with modulation frequency. The inset shows the response curve for C0₂ under these conditions. P T/GB2012/000588

Figure 6 (a) shows a cantilever signal obtained when 4 TDLs were used to probe

C2H2, CH₄, CO and C0₂ simultaneously. The largest peak observed at 340 Hz was attributed to C2H2. The insert graph shows the section of the transform in which photoacoustic signals attributed to C02, CO and CH₄ occurred (this is expanded by a factor of 30 for clarity).

Figure 6 (b) - the main figure shows the acoustic signals generated when the

1534.099 nm C2H2 absorption peak was probed. The inset graph is an expanded portion of the main graph (highlighted by the oval). The main peak was flanked by a series of sidelobes which decreased in intensity. FW and FWHM denote the full peak width and the peak width at half maximum respectively. The intensity of the largest sidelobe was 0.75 % of the main excursion.

Figure 7 shows photoacoustic signals of C0₂, CH₄ and CO recorded sequentially (B) and simultaneously (A). The plots are offset for clarity.

Figure 8 shows a photoacoustic signal response as a function of sample pressure for the cantilever cell. Figures 1(a) to (d) show the transitions targeted with each laser in a four laser system. The band spectra of CO, C0₂, C₂H , and CH₄ are shown.

range/nm ^" Table 1

Figure 2 shows a schematic of the configuration of the acoustic chamber (12), cantilever microphone (14) and Michelson interferometer (16). The cantilever photoacoustic cell used was manufactured by Gasera Ltd. of Finland. The cell was of modular design and comprised units containing the measurement and balancing acoustic chambers (15), gas inlet (18) and gas outlet (20) valves and a Michelson interferometer ( 16). The resonator which constituted the sample chamber comprised a 1 15 mm cylindrical copper tube of 3 mm diameter which was sealed at both ends with 1° wedged BK7 windows (22). A novel silicon cantilever microphone of dimensions 6 mm x 1.5 mm x 10 μπι (length x width x thickness) was positioned between the sample and balancing chambers (15) in an orientation parallel to the longitudinal axes of the chambers and the direction in which the laser radiation (24) propagated through the sample chamber. The cantilever was silver coated to ensure high reflectivity and was located on the top of the tube which connected the chambers. In order to minimise the effect of volume changes behind the cantilever on the acoustic signal the volume of the balancing chamber exceeded that of the sample chamber.

Figure 3 shows the experimental arrangement in which the outputs from four fibre- coupled DFB-TDLs (26) were fed through individual single mode optical fibres (27) and combined using a four-to-one fibre coupler (28). Detection of trace quantities of four gaseous analytes (CO, C0₂, C₂H₂ and CH₄ in nitrogen) has been achieved using this instrument. All of these species are relevant to degradation of polymeric / organic materials and atmospheric studies. The analytes were detected sequentially using the TDM method and simultaneously using MFDM. Four fibre-coupled DFB TDLs were used in the system, each of which was dedicated to the detection of a single analyte. The central wavelengths of the lasers were approximately 1534 nm (C₂H₂), 1567 nm (CO), 1568 nm (C0₂) and 1620 nm (CH₄). The lasers generated 20 - 30 mW output at source, although the powers available for the PAS studies were reduced to 1 - 3 mW after fibre coupling.

A range of mixtures containing 0.5% of each species in nitrogen diluent was supplied by BOC and used throughout the experiments. Samples containing lower concentrations of the analytes were prepared by diluting the gas mixture with pure nitrogen using a vacuum system in conjunction with calibrated mass flow controllers (MKS 722). These samples were prepared by varying the ratio of the flow rates between the analyte mixture and the diluent and used to determine the detection limits. The controllers were originally calibrated by the manufacturer using nitrogen gas streams and correction factors of 1.0 (N₂), 0.63 (C₂H₂), 0.7 (C0₂), 1.0 (CO) and 0.72 (CH ) had to be applied.

The outputs of the four DFB-TDLs, were fibre coupled into a single beam which was used to detect the four analytes either simultaneously (MFD ) or sequentially (TDM). The gas cell was located on an optical breadboard mounted on vibration- isolated laser table throughout the experiments. Pre-prepared gas samples were transferred to the cell via a manifold equipped with a baratron-type pressure transducer (MKS, 1000 mbar). Care was taken to avoid rapid fluctuations of the pressure within the cell due to the fragile nature of the silicon cantilever. Following transfer of gas, the cell was isolated using two electronically actuated shut-off valves to allow measurements to be made under static pressure conditions. It was recognised that operating in this regime was likely to present problems when the analysis B2012/000588 involves detecting molecules which tend to become adsorbed on the interior surface of the cell which leads to "wall loss" effects.

At atmospheric pressure the resonance frequency of the sample chamber was approximately 1460 Hz for a sample containing 1 % C₂H₂ in nitrogen diluent at 298K. The cantilever itself had a resonance frequency of 300Hz. Although the cantilever can vibrate in several modes, only the region below the first resonance frequency is considered. It is preferable to operate the cantilever cell below the resonance frequency as this avoids having to correct for slight drifts in temperature during the analysis. The displacement of the cantilever was monitored by the compact Michelson interferometer which measured the interference between two internal laser beams with a phase shift of 90°. The two internal laser beams were collected on four photodiodes, the signals from which were transferred to a computer and Fourier transformed using the Gasera software to generate a frequency spectrum. Each Fourier transformed spectrum contained 4096 averaged data points. The extent of averaging was controlled by the number of samples collected (131 ,072) and the data acquisition rate (50 kHz). The total time taken to acquire an averaged frequency spectrum was approximately 2.62 seconds.

The laser powers ranged from 0.8 mW to 3.06 mW. . The current tuning rates (nm mA^"1) applied to the DFBs were 0.0065 (CO), 0.0080 (C0₂), 0.0070 (C₂H₂₎ and

0.0084 (CH ) respectively. The lasers were operated using a Profile PRO 8000 laser diode controller and connecting cables. The drivers enabled the current and

temperanire of the diodes to be varied, thus enabling the output of each laser to be tuned to the absorption iine chosen for the analyte of interest. Data were acquired W using a LabVIEW program developed by Gasera. Ltd. The Gasera program was modified to incorporate a virtual function generator which provided low frequency ( 10 - 500 Hz) sinusoidal voltage waveforms to modulate the laser current drivers. As designed, this program only facilitated single species detection and it was therefore necessary to modify it to enable multispecies detection to be performed. The modified program incorporated a virtual function generator which provided low frequency ( 10 - 500 Hz) sinusoidal voltage waveforms to modulate the laser current drivers. The virtual waveform was processed through a high speed 16-bit data acquisition card (PCI-6259), sampling at a rate of 1.25 MS/s. The photodiode voltage signals were amplified and then subtracted from each other before the signals were acquired with the same 16-bit data acquisition card. The modulation frequency was not matched to the resonance frequency of the sample chamber as hysteresis effects were significant for the cantilever at high modulation frequencies (kHz) and deteriorated the performance. Resonant amplification of the acoustic signal was of no benefit.

A LabVIEW based program was used to process the photodiode signals from the interferometer. This data contained information covering the frequency response of the cantilever between 0 and 5000 Hz. This program allowed the implementation of four function generators, outputting sinusoidal waveforms to four DFB-TDLs.

Additional modification of this program allowed the magnitude of four FT-PA signals to be logged and raw interferometer signals registered for further post-processing and as a gauge of interferometer stability. An Ophir Nova II power meter and a PD300 IR photodiode were used to measure the optical power incident on the cell. A suite of pigtailed DFB-TDLs were used to detect CO, C0₂, C2H2 and CH₄ respectively. These lasers were selected to match the absorption features of interest. The output from each laser was collimated (Thorlabs) and then steered with gold coated mirrors through a 30 cm focal length lens such that the focal point coincided directly below the free end of the cantilever. The beam entered the PAS cell via a BK7 glass window and exited through a second BK7 window. The laser output was attenuated with a fibre-based attenuator (Oz Optics) when necessary.

Gas samples were transferred to the cantilever cell in a controlled manner. First of all, the cell was carefully evacuated using a rotary vacuum pump (Edwards) after which the pre-prepared gas samples were transferred to the cell via a manifold equipped with a baratron-type pressure transducer (MKS, 1000 mbar). Care was taken to avoid rapid fluctuations of the pressure within the cell due to the fragile nature of the silicon cantilever. Following transfer of gas, the cell was isolated using two electronically actuated shut-off valves to allow the sample to reach quiescence.

CO, C0₂, C₂H₂ and CH₄ were detected simultaneously in the cantilever PA cell giving the first demonstration of MFDM using a cantilever based PA spectrometer. Careful adjustment of the thermoelectric cooler temperatures was necessary to realign the output wavelengths with the centre of the absorption peaks. Each laser was modulated below the resonance frequency of the cell and PA signals were detected at twice the modulation frequency. Wavelength modulation of four DFB-TDLs across the absorption features of interest resulted in the formation of multiple pressure waves in the PA spectrometer. These led to displacements of the cantilever microphone ( 13) which were measured using an interferometer ( 16). The composite PA signal could be decoupled into its constituent components by applying a Fourier transform to the interferometer signal. The magnitude of each peak in the spectrum was directly proportional to the cantilever displacement at the frequencies corresponding to the peak. Each component in the spectrum corresponded to a single analyte and occurred at a frequency that was a multiple of the laser modulation frequency.

Resonant amplification of the acoustic signal was of no benefit with the cantilever cell as noise resulting from mechanical vibrations would be amplified with the signal. The virtual waveform was processed through a high speed 16-bit I/O card (PCI-6259), sampling at a rate of 1.25MS/s.

Data obtained from measurements of single and multiple species are presented in Table 2.

Detection regime CO CiHi C A

Laser parameters Laser power at line centre ) m W 2.93 2.35 1 .7 1 0.81

Detection frequency / Hz 92 42 340 74

(a) Single species Detection limit / ppmv 177.0 170.6 1 .72 2 14.3 detection Normalized sensitivity / x 10^"* 2.4 0.8 4.1 0.6

cm-' Hz"⁷

(b) Multispecies Detection limit / ppm 249.6 181.3 1 .5 293.7 detection Normalized sensitivity I K J O^'9 3.4 0.9 3.6 0.9

cm^"' W Hz ^ia

Tabic 2 Row (a) presents detection limits and sensitivities obtained when the analytes were detected individually and Row (b) presents data obtained simultaneously using the MFDM method. All sensitivities listed in table 2 have been normalised to the laser power and date acquisition time (2.62 s). Although the sensitivities presented in Table 2 were expected to be identical (being an assessment of the instrument's capability) marked differences were evident.

The data presented in Table 2 indicate that power and bandwidth normalised sensitivities were different for each analyte; this was consistent with observations in other PAS measurements. Although the inventors do not wish to be bound by theory, this is thought to result from differences between the rates of the vibration- rotation/translation (V→ R7T) energy transfer process associated with the analytes, although other contributing factors have arisen from recorded signals in different noise regions. Slight deviations of analyte concentrations from the theoretical amounts and the dependence of the microphone response on detection frequency may also be contributing factors. The detection limits and sensitivities for CO, C0₂ and CH₄ were lower when these analytes were monitored individually than when all four analytes were detected simultaneously. The sensitivity determined for C0₂ absorption is in excellent agreement with previously published data. However, power and bandwidth normalised sensitivities were different for each analyte. This behaviour was consistent with observations in other PAS measurements. The Fourier transforms of the cantilever response contained the fundamental, or first harmonic, and a range of higher harmonics which decreased in magnitude with increasing order due to a decrease in the response of the cantilever with increasing frequency. Figure 4 presents a typical photoacoustic signal of C2H2 using the cantilever cell and was obtained by repeatedly scanning the laser frequency back and forth across the maximum of the 1534.099 nm absorption line at a rate of 37 Hz. As shown, the scanning resulted in the appearance of first, second and higher order harmonic peaks. Slight scanning asymmetry, attributed to laser power variations across the modulation cycle, resulted in the appearance of the weaker odd harmonic features. It was therefore decided to record second harmonic peaks in preference to the fourth or sixth. This was achieved by adjusting the modulation amplitude applied to each laser driver before acquiring data to ensure that the even harmonic components were predominant in the spectrum. Another advantage of measuring the second harmonic signal is that the background signal arising from window and wall absorption occurred at the fundamental frequency.

The sensitivity achieved for the multispecies detection depended significantly on the modulation frequency for each laser. As shown in Figure 5, the photoacoustic signal for C2H2 varied across the possible cantilever operating frequencies. However, for multispecies detection it is not always possible to modulate each laser in the most sensitive frequency range for cantilever sensitivity and a range of frequencies must be selected. Modulation frequencies of 21 , 37, 46 and 170 Hz were selected for C0₂, CH₄, CO and C₂H₂ respectively. In principle the sensitivities determined for each species could be further normalised or scaled for the variation in cantilever sensitivity across the selected frequency range. 0588

The effect of gradually increasing the number of DFBs used in the system was assessed in a series of MFDM experiments. Photoacoustic spectra were acquired in simultaneous mode for each combination of lasers. This study was designed to determine if additional noise from window heating was produced when the number of lasers deployed in the system was gradually increased.

As the response of cantilever microphones was known to be dependent on the sample pressure and modulation frequency it was essential to characterise these dependencies in order to define the optimum measurement conditions. The response of the cantilever as a function of pressure was determined for each analyte in the 30 to 1000 mbar range by recording photoacoustic signals from each analyte simultaneously.

The frequency response was investigated by modulating the four lasers at different frequencies, all of which were less than the resonance frequency of the sample chamber ( 1460 Hz). In each case, the analyte absorption line was accessed twice during the modulation cycle and the photoacoustic signal was detected at twice the modulation frequency. This allowed the magnitude of sensitivity variations arising from non-uniform response of the microphone to be assessed. A series of sequential measurements in which each laser was modulated at 37 Hz were then performed for comparison; under these conditions a uniform microphone response was expected.

Analytes with small absorption cross sections (i.e. CH₄, CO and C0₂) were monitored using modulation frequencies below 100 Hz as the cantilever gave optimum response in the 0 - 00 Hz region. Detection of C₂H₂ was carried out close to the cantilever resonance frequency (300 Hz). Working at the cell resonance frequency was not necessary, as ambient noise and photoacoustic signals were amplified by equal amounts. Prior to recording spectra, background scans were obtained to identify the frequencies at which ambient noise was low.

Signal to noise ratios, detection limits and normalised sensitivities were derived for each analyte in the gas mixture. The extent to which interference between

concomitant species occurred in the cantilever cell was evaluated by comparing data acquired from measurement of single analytes with data obtained by simultaneous detection of the four analytes using the MFDM method. These measurements were performed under identical experimental conditions. The wavelength of each DFB was tuned to the absorption maximum of the selected line pertaining to its target species. As the DFBs produced stable outputs (wavelength and power) stabilisation was not required.

As the experimental conditions in PAS vary considerably it is common practice to derive the noise-equivalent absorption coefficient (NNEA, cm^"' WHz ^"l/2) for a particular system. The NNEA is defined by Equation 3 where a_mi_n is the minimum detectable absorption coefficient (cm^"1), P is the laser power (W) and t is the signal averaging time (s). NNEAs allow the sensitivities of different PAS systems to be directly compared.

NNEA = a_minPVt (Equation 3)

The detection limit is the equivalent concentration of analytes under these conditions. Microphone response curves were obtained for detection of the second harmonics of the C₂H₂ and C0₂ acoustic signals in the nitrogen-based gas mixture and are presented in Figure 5. These show the large variation in microphone response with frequency, [n the case of C₂H₂ the signals were recorded at twice the modulation frequency at 1000 mbar sample pressure using the certified gas mixture.

The cantilever resonance frequency occurred at 300 Hz for C₂H₂. The inset in Figure 5 shows the response curve obtained when detecting C0₂: in this case the resonance frequency was 290 Hz. The 10 Hz difference between the resonances was attributed to slight changes in ambient temperature between recording the two data sets. Both graphs indicate that the response of the cantilever increased with decreasing frequency.

Figure 6(a) shows a Fourier transform of the cantilever movement when four DFB- TDLs were used to detect analytes simultaneously. The modulation frequencies used were 21 Hz (C0₂), 37 Hz (CH ), 46 Hz (CO) and 170 Hz (C₂H₂) respectively and the second harmonic signals from the cantilever occurred at 42, 74, 92 and 340 Hz. The inset in Figure 6(a) a shows a portion of the spectrum between 0 and 100 Hz, magnified by a factor of 30 for clarity.

The modulation amplitude applied to each laser driver was optimised before data collection to ensure only even harmonic components were present in the spectrum. Data were collected at a rate of 50,000 kHz for 2.62 seconds. The use of an interferometer to probe the cantilever movement and subsequent Fourier transform of the signal allowed the separation of different frequency components. Knowledge of the cantilever response allowed careful selection of each modulation frequency, with analytes that had the smallest absorption cross-section probed at frequencies where cantilever response was greatest. Background spectra were also recorded to ascertain frequencies where ambient noise was low. Although working at the resonant frequency was not necessary as ambient noise and PA signals were amplified by equal amounts, C₂H₂ detection was carried out close to the cell resonance frequency (340 Hz). Despite the large signal amplitude at high frequencies, it was still possible to measure precise changes in small signals at low frequencies and the insert on the graph shows.

The photoacoustic signal generated by probing the C₂H₂ absorption at 1534.099 nm registered some sidelobes ("ringing") which had greater amplitude than the noise. These comprised a series of extra peaks flanking the main peak and reducing in magnitude, as shown in Figure 6(b), the largest of which was 0.75% that of the main peak. Whilst the inventors do not wish to be bound by theory, the sidelobes are thought to arise from frequency instabilities in the system. The main plot in Figure 6(b) shows a transform of the acoustic signals which were generated when the

1534.099 nm C₂H₂ absorption was probed. The inset graph shows an expanded portion of the main graph (highlighted by the oval). Although the contribution from sidelobes was much smaller than that of the main peak it could be reduced using a variety of functions (known filters or apodization functions) to help remove them.

Detection of C₂H₂ was carried out at 340 Hz to prevent the sidelobes associated with the strong response from this species masking weaker signals produced by the other analytes. The minimum separation required between C₂H₂ peak and peaks from the other analytes was governed by the width of the ringing, which in turn had implications on the frequency resolution of the spectrometer. The sidelobes in the proximity of the strong C₂H₂ peak prevented other species from being detected within 15 Hz of the main peak and therefore had the potential to cause interference during multiple species detection. The photoacoustic signals from the other analytes did not show any sidelobes above the level of the background noise. Despite the large signal amplitude of the C2H2 peak and the presence of its sidelobes it was still possible to measure the precise changes in the small signal produced by C0₂, CH₄ and CO at the low frequencies, as the inset graph in Figure 6(a) shows.

In the MFDM experiments each analyte was detected at a different frequency. This provided opportunity to detect the four analytes simultaneously and also to obtain normalised sensitivities for each species. It was recognised that use of different frequencies could result in variation in sensitivity. In view of this, an investigation was conducted to assess the influence of the cantilever response on the measurement sensitivities. There were pronounced hysteresis effects at the higher frequencies which reduced the responsivity of the transducer, therefore depending on the frequency at which the analyte was detected, the sensitivity changed. To determine whether the frequency response of the cantilever transducer could account for the variation in sensitivity, each laser was sequentially modulated at 37 Hz. Modulating each laser at 37 Hz ensures that the response of the transducer was the same for all the species. The noise was taken as 3 standard deviations of a 50 point section of spectrum at the base of the second harmonic peak. The noise level was calculated in this way for each spectrum to account for any noise variations from experiment to experiment and give a fair evaluation of the noise. The detection limits were calculated using this noise measure and the maximum peak height. The detection limits and sensitivities were normalised for the microphone response yielding the results in Table 3. Simultaneous multispecies detection using a cantilever microphone also requires sensitivities to be normalised against the microphone response.

Table 3

Figure 7 shows photoacoustic signals of C0₂, C¾ and CO recorded sequentially (B) and simultaneously (A). The plots are offset for clarity. The variations in cantilever response with pressure for the four analytes are presented in Figure 8(a) and (b). in each case the photoacoustic signals increased in the 0- 100 mbar pressure range and then decreased as the pressure in the cell was increased. Although all of the 0588 photoacoustic signals were enhanced at lower pressures, the degree of enhancement was not uniform. The optimum pressure for photoacoustic signal detection ranged from 55 to 200 mbar, depending on the analyte. The lowest optimum pressure (55 mbar) was obtained for C₂H₂ with values of 103, 1 18 and 200 mbar being obtained for CO, CH₄ and C0₂ respectively. Enhancement factors achieved by operating at the optimum pressure instead of 1000 mbar ranged from 3.9 (C0₂) to 25.6 (C₂H₂).

The peak attributed to C₂H₂ is not shown, but displayed similar characteristics to the other analytes. Second and fourth harmonic peaks were observed for C0₂.

Normalised sensitivities for each analyte were calculated and compared to determine whether an increase in multiplexing complexity had deleterious effects on the sensitivity of the spectrometer. No calculable difference was observed between data obtained by measuring one analyte at a time and all four analytes simultaneously and there was no evidence of analyte interference. Variations of the total pressure in the cantilever cell are known to have an effect on sensitivity. The optimal pressure inside the cantilever cell has been shown to be approx 100 mbar. At 100 mbar the dominating source of noise is Brownian motion of the cantilever. Detection limits for CH₄ were quoted as 3 ppb at 100 mbar, whereas those at 1000 mbar were approximately 25 ppb (an 8.3 fold increase). In order to assess this samples of the nitrogen based mixture containing 0.5% CH₄, 0.5% C0₂, 0.5% CO and 0.5% C₂H₂ mixture were introduced into the PA cell at various pressures between 30 and 1000 mbar and the PA signals of the four analytes recorded simultaneously. It was noted that the PA signal decreased when the pressure in the PA cell was increased. All the PA signals were enhanced at lower pressures, but some more than others. The optimum operating pressure varied with analyte: for C2H2 this was approximately 55 mbar, CO approximately 103 mbar, CH₄ approximately 1 18 mbar and C0₂ approximately 200 mbar. At the optimum operating pressure for each analyte the detection limits have been summarised in Table 4.

Analylc Detection Optimum Potential Potential Anticipated

limit at (000 pressure / enhancement detection sensitivity at mbar / ppm mbar factor at optimum limit / ppmv optimum pressure / pressure lO^ Wcm-' Hz-"²

CO 249.6 103.0 5.9 42.2 5.7

C⁽¾ 181.3 200.0 3.9 46.5 0.5

C2H2 1.5 55.0 25.6 0.1 1.0

CH₄ 293.7 1 18 0 5.8 50.6 1.5

Table 4

Data are presented in Table 4 along with the potential detection limit and sensitivity for each analyte at its optimum pressure. Under these conditions the dominant source of noise was attributed to Brownian motion of the cantilever.

When the interferometer was performing properly, noise registered in the signal could be attributed to ambient acoustic noise. The effect of instability on the spectra could not be removed by taking the ratio of the second harmonic signal to the baseline noise. As the noise in the system registered itself in the baseline, any increase by the same factor as the noise (i.e. doubling the noise did not double the second harmonic signal), but instead contributed to the magnitude of the signal. Therefore, taking a ratio of the two did not remove noise. The degree of interference between analyte signals during simultaneous

measurements was assessed by comparing the performance achieved by this method with that obtained by detecting the analytes sequentially. In these studies the gas mixture was flushed through the cell for several minutes after which the valves were closed to retain a sample at 1000 mbar. Photoacoustic spectra were acquired by operating the four DFBs sequentially at the optimised modulation index pertaining to each analyte. During sequential measurements one laser was operated at a time and the injection currents applied to the other three lasers were reduced to zero. Fourier transforms of each signal were recorded and the signal-to-noise ratios derived from the peak height and the baseline noise, taken as three standard deviations (3S_dcv) of a 50-point section in the vicinity of the main peak. Due to known adsorption and desorption effects associated with C₂H₂, the concentration of this gas was verified independently by single pass absorption measurements at 1534.099 nm. This was achieved by directing the laser radiation through the cell and scanning the laser wavelength across the spectral region of interest by applying a 5 Hz triangular waveform (2 V peak-to-peak) to the laser driver. The intensity of the transmitted radiation was recorded by a fast response detector and captured by a 14-bit digitiser. The 100 MHz bandwidth of the digitiser enabled data to be collected at a sampling rate of 100 MS/s. The sample pressure was varied between 50 and 1 150 mbar during this investigation and normal absorption spectra were recorded.

A comparison of data acquired from simultaneous detection of the four analytes using the MFDM method with data obtained for single species detection indicated that there was no evidence of cross-talk between the analyte signals obtained using the cantilever cell. The sensitivities achieved using the various laser combinations are summarised in Table 5.

Number of lasers CO C0₂ CH, C_H_

1 7.1 - - -

2 5.5 0.3 - ^■ -

3 6.4 0.5 0.8

4 8.7 0.6 0.7 5.8

Table 5

The difference between the largest and smallest sensitivity (approximately 20%) can be accounted for by changes in the baseline noise. The remaining differences were attributed to fluctuations in the photoacoustic signal. Gradual introduction of additional lasers did not have deleterious effects on the noise or the detection limits as the optical power of each laser was only 1 - 2 mW. Generally, the main source of noise in the cantilever cell arose from ambient acoustic effects.

The performance of cantilever PAS system was not influenced by the number of lasers deployed in the instrument. The stability of the interferometer proved critical in maintaining the sensitivity over an extended period time. It was found that the interferometer only remained stable for up to an hour before realignment was required. The best sensitivities were achieved when signals were averaged over a few minutes, indicating that continuous monitoring of analyte species is not suitable with this type of cell. Using a cantilever cell and no additional fibre amplification, it was possible to measure CO, C0₂ and C₂H₂ sequentially using TDM scheme; detection limits were 10.6, 16.0 and 0.86 ppmv respectively. The best normalised sensitivity registered was 6.27 x 10^"10 cra^''WHz^'"² for C₂H₂. This is the first demonstration of using the cantilever cell in a multiplexing regime and the first demonstration of using the cantilever cell with wavelength modulation regimes. Each spectrum took 130 seconds to acquire resulting in greater variations to the PA signal height when several spectra were recorded of the same analyte species and concentration. These sensitivities were achieved with very low powers (approximately 0.3 to 3 mW).

The cantilever microphone and the use of MFDM enabled high sensitivity, selectivity and multispecies detection. Sensitivities achieved for multispecies detection ranged between 0.95 x 10 ⁹ and 1 .69 x 10^"9 cm^'1 W Hz^"1/2 which is comparable to single species detection quoted in V. Koskinen et al, Vibrational Spectroscopy, 42 239 (2006). PA cells having volumes of a few cm³ can be used with low optical powers (0.81 and 2.93 mW) to measure ppmv levels of CO, C0₂, C₂H₂ and CH₄ with DFB- TDLs. The sensitivities were normalised with respect to power, measurement bandwidth, microphone response and assumed Lorentzian adsorption profile. The cantilever microphone demonstrated no measurable cross-talk between the signals.

Sequential fibre-amplification of the DFB-TDLs allowed the total optical power to be channelled to one wavelength at a time and gave the best option for maximising the optical power available for detection of each analyte. When sequential amplification was used the detection limit improved for all analyte species (CH₄, CO, C0₂ and C₂H₂). The TDM approach was coupled with a wavelength modulation regime. The movement of the cantilever was still monitored with the Michelson interferometer but the PA signals were detected with a lock-in amplifier. This helped to decrease the detection bandwidth to 1 MHz. This was the first demonstration of utilising the cantilever cell together with TDM and wavelength modulation schemes. Normalised sensitivities (without fibre amplification) varied between 0.61 x 10^"10 and 3.22 x 10 ¹⁰ cm ' WHz ¹'² for the group of species monitored. Again, fibre amplification of the DFB-TDLs was possible in this arrangement. With fibre-amplification the sensitivities ranged between 2.88 x 10^~8 and 13.4 x 10^"'° cm^"'WHz^{" l 2}.

The method and apparatus applies equally well to single species detection as it does to multispecies detection. The supporting data shows that it is possible to use four lasers it is possible to use any number of lasers. The cantilever PAS system could be used in the analysis of a wide range of gaseous species. It is preferable for the system to be portable and automated to minimise the user intervention along with room

temperature operation of the instrument. The system could be used to analyse gas samples obtained from a wide range of environments.

The high sensitivity of the cantilever pressure transducer coupled with the effective noise removal offered by phase-sensitive detection introduced a regime capable of detecting ppbv levels of C₂H₂.

Claims

1. A laser photoacoustic spectroscopy apparatus for detecting gaseous analytes comprising a plurality of tunable lasers and a photoacoustic cell wherein each of the tunable lasers is tuned to excite one or more analytes of interest.

2. An apparatus as claimed in claim 1 wherein the photoacoustic cell

comprises at least one cantilever microphone.

3. An apparatus as claimed in claim 2 wherein a compact Michelson

interferometer is used to measure an acoustic signal produced by the at least one cantilever microphone.

4. An apparatus according to any preceding claim wherein the number of tunable lasers is equal to the number of analytes to be detected.

5. An apparatus according to any preceding claim wherein optimised

wavelength modulation is applied to each of the lasers such that a scanned wavelength range exceeds a width of the absorption feature of each analyte of interest.

6. An apparatus according to any preceding claim wherein the outputs from the lasers are coupled into a combined beam.

7. An apparatus according to claim 6 wherein the outputs from the lasers are coupled into a single mode optical fibre

8. An apparatus according to claim 7 wherein the single mode optical fibre is silica glass.

9. An apparatus according to any of claims 6, 7 or 8 wherein the combined

beam is used to excite the analytes simultaneously.

10. An apparatus according to claim 7 wherein a modulation frequency division multiplexing technique is used to acquire measurements simultaneously.

1 1. An apparatus according to claim 6 wherein the combined beam is used to excite the analytes sequentially.

12. An apparatus according to claim 1 1 wherein each of the plurality of lasers are operated at a different time to any other laser by time division multiplexing to allow sequential measurements to be taken.

13. An apparatus according to claim 1 1 or 12 wherein a lock-in amplifier is used to detect the acoustic signals.

14. An apparatus according to any preceding claim wherein the apparatus further comprises a fibre amplifier to increase power output of the plurality of lasers.

15. An apparatus according to any preceding claim wherein a computer program processes the acoustic signals.

16. An apparatus according to claim 15 wherein the computer program Fourier transforms the signals to generate a frequency spectrum.

1 7. An apparatus according to claim 15 wherein a lock-in amplifier demodulates the acoustic signal.

1 8. An apparatus according to any preceding claim wherein the tunable lasers are tunable diode lasers.

1 9. An apparatus according to any preceding claim wherein the lasers are

wavelength tunable.

20. An apparatus substantially as hereinbefore described and with reference to the accompanying drawings.

21 . A method for detecting gaseous analytes using the apparatus according to any one of claims 1 to 20 wherein a gas sample is transferred to the photoacoustic cell and then the cell is closed to retain the sample for analysis.

22. A method substantially as hereinbefore described and with reference to the accompanying drawings.