PHOTOACOUSTIC SPECTROSCOPY Light and sound—photoacoustic spectroscopy C. Haisch and R. Niessner Institute of Hydrochemistry,Technical University of Munich, Marchioninistrasse 17, D-81377 Munich, Germany. E-mail: Christoph.Haisch@ch.tum.de Introduction Like many other spectroscopic tech- niques, which are today carried out with laser light sources, photoacoustic spectroscopy is much older than the development of the first laser in 1960. In 1881 A.G. Bell proposed a spectro- phone (Figure 1) “for the purpose of examination of the absorption spectra of bodies in those portions of the spec- trum that are invisible”.1 This instru- Figure 1. The spectrophone, the first proposed application of photothermal spectrometry for absorption analysis, taken from a publication from the year 1881. ment was based on his experiments for the transmission of sound without a cable connection. The field then lay largely dormant until the mid-1970s, when Allan Rosencwaig and Allen Gersho laid the theoretical basis for the photoacoustic effect in solids, the so- called R-G Theory.2 The basic principle of all photother- mal (PT) techniques is the absorption of light in a sample with a subsequent change of its thermal state. This may be either a change of the temperature or other thermodynamical parameter of the sample related to the temperature. Measurement of either the tempera- ture, pressure or density change that occurs due to optical absorption is ulti- mately the basis for all PT spectroscop- ic methods. PT analysis can be consid- ered as an indirect absorption measure- ment as the measured quantity is not an optical signal. (It should be noted here that the classical absorption measure- ment is not a direct measurement either. Though the measurement value 10 © Spectroscopy Europe 2002 in this case is an optical one, namely the transmitted light, the absorbed light quantity is derived from the difference of the incident and the transmitted energy.) The sample heating which produces the PT signal is correlated directly to the absorbed electromagnet- ic energy. In contrast to conventional transmission spectroscopy, neither scat- tered nor reflected light contributes to the signal. This makes PT spectroscopy particularly attractive for absorption measurements in gaseous, liquid and solid media containing scattering parti- cles, and on solid boundaries. The most straightforward detection scheme for a photoacoustic signal is the observation of the temperature change at the irradiated sample surface. If this observation is carried out by tempera- ture transducers on the sample surface the technique is called thermometric detection. The more common tech- nique is the detection of the emitted thermal radiation from the sample sur- face, which represents the temperature distribution within the sample. If the temperature rise in the absorbing sam- ple volume occurs faster than this vol- ume can expand a local pressure increase (wave) is the consequence. This pressure wave can be considered as a sound signal. PT absorption mea- surements based on the detection of this sound wave are named photo- acoustic (PA) techniques. The basic steps of PA analysis are summarised in Figure 2. The means for the detection of a PA signal are multifarious: in a gaseous matrix a microphone may be employed, whereas pressure fluctua- tions in a solid or liquid sample can be probed by pressure sensitive elements like piezo transducers. Alternatively, the pressure fluctuations can be observed by optical methods. Although a PT effect can be induced by any light source, lasers are the pre- ferred excitation source nowadays for two reasons: (i) the PT signal, to a first approximation, is proportional to the temperature rise in the sample and thus proportional to the absorbed energy, i.e. the pulse energy, (ii) for many Figure 2. Principle of a photo- acoustic experiment. applications the selectivity of a PT analysis, as with any other absorption analysis, depends on the band-width of the excitation wavelength. Selected applications Quantification of soot particles in diesel engine exhaust gas For analytical purposes, gas analysis is the oldest application of PA spec- troscopy (PAS). The components of A.G. Bell’s set-up can all also be found in today’s PAS gas analysers. A modu- lated laser source, tuned to an absorp- tion band of the analyte gas, replaces the sun as the light source (see Figure 3). Instead of the ear, a sensitive micro- phone detects the sound signal. The microphone signals are electronically amplified and detected by a lock-in amplifier. A lock-in amplifier is a fre- quency and phase-sensitive data recorder that picks up only signals with the same frequency as a reference sig- nal. This type of data recorder can be used as an extremely narrow band-pass Spectroscopy Europe 14/5 (2002) 

PHOTOACOUSTIC SPECTROSCOPY Acoustical filter volumes Sample gas in Gas out Resonator Window Microphone Flow controller Fig. 3 Gas flow Diode laser <<< >>> Designed by Haisch Digital Delay Generator PASS cell Pump Beam trap Pre-amplifier Made by Main Aux. W IDTH W IDTH 100 ns 100 ns ^ < > 1 2 3 4 5 6 7 8 9 . 0 # Func. C E Powe 5350 2390 5350 2390 1 1 1 1 1 1 1 1 1 1 ENTER 1 1 1 1 1 1 1 1 1 1 1 ST Designed by Lock-in amplifier Notebook Figure 3. Experimental set of the PASS (photoacoustic soot sensor) system. filter. For PA spectroscopy with a modulated light source, the modulation frequency is taken as a reference signal, since the PA signal has exactly the same frequency as the excitation source. The PA system presented here, named PASS (photoacoustic soot sen- sor), is optimised for the quantitative detection of soot particle mass in the exhaust gas of diesel engines, and serves as an example for a variety of similar PA gas sensors. The specific require- ment for the soot detection was a detection limit below 50 µg m–3 with a time resolution of 1 Hz. The insensi- tivity of PA spectroscopy to light-scat- tering particles in the probe volume compared to other techniques is espe- cially favourable, as the exhaust gas stream may contain a high amount of liquid droplets, which can act as strong light-scatterers. For the excitation a diode laser with a wavelength of about 800 nm and an output power of 1 W was chosen. Soot, or black carbon, shows a wide absorption continuum over the whole visible and near infrared spectrum. To avoid cross sen- sitivity with any potential trace gases in the exhaust gas stream, e.g. NO or water vapour, a limited wavelength range around 800 nm is used. To increase the PA signal, resonance amplification of the acoustical wave is applied. The absorption of the modu- lated laser beam takes place in an acoustical resonator tube, through which the sample gas is sucked. Since the intensity of the generated pressure wave follows the modulation of the exciting laser, they both have the same frequency. Hence, the sound wave can be tuned to match the resonance fre- quency of the resonator by tuning the 12 modulation frequency of the laser. The resulting resonance amplification can reach values from 10 to more than 1000, depending on the design of the resonator. The optimum resonance amplification for a certain analytical application is not necessarily the high- est, because higher resonance amplifi- cation, i.e. a higher resonance profile always corresponds to a more narrow profile shape. The resonance frequency depends both on the shape of the res- onator and on the velocity of sound of the gas within the resonator. If this velocity changes, e.g. due to changing gas temperatures or humidity, a narrow profile leads to significantly reduced amplification, and, as a consequence, to signal fluctuations. As the conditions of the exhaust gas being investigated can change drastically over on experiment, a relatively low resonance amplification of around 20 was chosen. At each end of the resonator tube is a cylindrical buffer volume, which acts as an acoustical filter. Their length measures about half the length of the resonator. The destructive interferences in these buffers filter external noise with the resonance frequency of the laser. Noise sources with other frequencies are less relevant as they are not amplified, nei- ther by the resonator nor by the lock- in amplifier. A gas handling system completes the PASS system. It consists of an intake tube which fits to standard exhaust gas probes and leads directly into the PA cell. After the cell the gas is filtered to protect the mass flow controller. The gas is sucked through this system by a pump. After this pump the sampled gas can either be released or fed back into the exhaust gas stream, which is Figure 4. Photograph of the PASS (photoacoustic soot sen- sor) system. required for some exhaust gas probe systems. In order to prevent condensa- tion of water vapour in the PA cell, it can be heated to 50°C. A picture of the complete PASS sensor, without the lock-in amplifier, is shown in Figure 4. In Figure 5, the calibration of the PASS system is illustrated. The system is calibrated in soot mass per cubic metre. Reference analysis is carried out by sampling the soot on a filter fol- lowed by thermochemical carbon analysis. The calibration points labelled in red were measured with strongly varying NOx concentrations, which are given on the graph. Clearly, no cross- sensitivity of the soot measurement to this gas, which is present in any exhaust gas, can be observed. A typical applica- tion of the PASS system is displayed in Figure 6. Here, the PASS was employed to analyse the soot mass con- centration over a 30 min transient dri- ving cycle. These test cycles at an engine test stand simulate a truck dri- ven through a city, on a highway and on a motorway. They are used for engine development and type approval of new engines. PA analysis of highly concentrated textile dyes The analysis of a highly concentrated textile dyestuff is used as a second example for demonstrating a specific advantage of PA absorption spectrome- try over conventional transmission spectroscopy. The concentrations of these dyes are in the range of more than 5 g L–1, resulting in absorption coefficients of 103 cm–1. Changes in the concentrations are required to be mon- itored during the dying process. An additional problem is the continuously increasing contamination of the dye solution by fuzz from the textile. This combination of extremely high absorp- tion and scattering particles in the dye solution makes a classical transmission spectroscopic analysis impossible. PA Spectroscopy Europe 14/5 (2002) 

PHOTOACOUSTIC SPECTROSCOPY 40000 35000 30000 25000 20000 15000 10000 5000 0 0 R2 = 0.987, n = 26 LOD (Limit of Detection) 10 µg m- 3 223 ppm NOx 656 ppm NOx Varying NOx Concentrations 1175 ppm NOx 0.2 0.4 0.6 0.8 1 1.2 Concentration / mg m-3 1200 1000 800 600 400 200 0 0 200 400 600 800 1000 1200 1400 Time / s Figure 5. Calibration curve of the PASS system vs thermochemical analysis. Figure 6. PASS signal of a driving cycle for diesel engine testing. Dye solution Entrance window Dye inlet Laser beam Focusing lens Dye outlet 0.5 0.4 0.3 0.2 0.1 0 0 BNC- connector (signal out) Piezo foil Coupling window R2 = 0.9994 n = 9 200 400 600 800 1000 Laser Pulse Energy / µJ Figure 7. Design of the flow cuvette for the PA analysis of concentrated dyestuff. Figure 8. Dependence of the PA signal on the laser pulse energy. spectroscopy is a viable approach to overcome the problems. Instead of a modulated light source, in this case a pulsed laser is employed. To get spectral information about the dye solution this laser source has to be tuneable over the whole visible spectral range. Suitable laser systems are dye lasers, for instance, which are optically pumped either by an excimer laser or by a nitrogen laser. The latter is the more cost-effective and manageable, since no gas refilling is required. Running costs and maintenance are key considerations for a system meant for on-line routine application in an industrial environment. The funda- mental experiments presented here were produced using a frequency dou- bled Nd:YAG laser, which was cou- pled into an optical fibre. This type of laser is not tuneable, but it is a real turn-key instrument and most conve- niently applied. The laser wavelength is 532 nm, the pulse energy behind the optical fibre is 1 mJ with a pulse dura- tion (FWHM) of 6 ns. The laser beam is directed through an entrance window into a cylindrical flow cell (see Figure 7). This cell has a 14 diameter of 0.8 cm and a length of 1.0 cm, resulting in a cell volume of c. 0.5 cm3. An inlet and outlet are installed at opposite sides of the cell. Thus, the sample flow passes through the cell perpendicular to the incident laser beam. The PA signal is detected by a hydrophone opposite to the laser input window. Hence, the laser impinges directly on the hydrophone, if there is no absorbing liquid in the cell. This layout was found to be advantageous for the case of strongly absorbing liquids, as the sound wave produced shows an approximately spherical propagation characteristic, while the sound wave produced in a low absorptive solution propagates in the form of a cylindrical wave. This kind of waveform with its axis of sym- metry collinear with the laser beam is detected best perpendicular to the laser beam. The hydrophone consists of a quartz front window (diameter 10 mm), which is coupled to the front contact of the piezo-active foil. Although there is no need to use an optically transpar- ent material for the front surface, quartz was found to result in an opti- mum coupling efficiency of the sound signal to the piezo foil. This polyvinyl- idene fluoride (PVDF) foil has a thick- ness of 25 µm. The second metallic contact is on the backside of the foil and directly connected to a BNC cou- pler. This whole hydrophone is encap- sulated in a metal housing. As mentioned above, the PA signal intensity depends on the laser pulse energy deposited in the sample. Figure 8 depicts this linear relation. A typical calibration graph of the described PA system for concentrated dyestuff is shown in Figure 9. This calibration is linear in a concentration range more than 10 times higher than achieved with conventional transmission spec- trometry. Further aspects of PA analysis The time-resolved PA signals as shown in Figure 10 contain additional information, which is not exploited in a bulk absorption analysis as described in the example above. The first signal in time is the signal generated by the Spectroscopy Europe 14/5 (2002) 

PHOTOACOUSTIC SPECTROSCOPY R2 = 0.995 n = 9 25 20 15 10 5 0 0.6 0.4 0.2 0 0 -0.2 -0.4 -0.6 -0.8 10 g/L 25 g/L 2 4 6 8 10 12 Transmitted light absorbed by the piezodetector Light absorbed in the cuvette, soundwave propagated through the cuvette 0 5 10 15 20 25 Concentration / g L-1 Time / µs Figure 9. Calibration of the PA sensor system for textile dyestuff. Figure 10. PA signal of two different concentrations of textile dyestuff in water. light that is transmitted through the sample and impinges directly on the hydrophone. The second signal stems from the light absorbed in the sample liquid. Due to the high absorption in the sample the electromagnetic ener- gy is deposited mostly directly on the front side of the cuvette. The pressure wave generated there propagates through the cuvette and is detected after a delay. This delay is the time (~7 µs) the sound needs to propagate the length of the cuvette (1 cm), (the speed of sound in water ~1500 m s–1). For a non-homogeneous sample a depth-resolved absorption measure- ment can be carried out in this way. Each delay corresponds to a certain depth in the sample. This simple example is included to give an idea of an interesting field of application or PA analysis. Of course, the design of such a sensor system would be differ- ent, for example, the hydrophone could be placed on the same side as the laser. However, the description of depth resolved analysis by PA spec- troscopy is beyond the scope of this short article. References 1. (a) A.G. Bell, Phil. Mag. 11, 510 (1881). 2. A. Rosencwaig and A. Gersho, “Theory of the Photoacoustic Effect in Solids”, J. Appl. Phys. 47, 64–69 (1976). Spectroscopy Europe 14/5 (2002) 15