Letter pubs.acs.org/ac
A Wide Spectral Range Photoacoustic Aerosol Absorption Spectrometer C. Haisch,* P. Menzenbach, H. Bladt, and R. Niessner Chair for Analytical Chemistry, Technische Universität München, Marchioninistr. 17, D-81377 Munich, Germany S Supporting Information *
ABSTRACT: A photoacoustic spectrometer for the measurement of aerosol absorption spectra, based on the excitation of a pulsed nanosecond optical parametrical oscillator (OPO), will be introduced. This spectrometer is working at ambient pressure and can be used to detect and characterize different classes of aerosols. The spectrometer features a spectral range of 410 to 2500 nm and a sensitivity of 2.5 × 10−7 m−1 at 550 nm. A full characterization of the system in the visible spectral range is demonstrated, and the potential of the system for near IR measurement is discussed. In the example of different kinds of soot particles, the performance of the spectrometer was assessed. As we demonstrate, it is possible to determine a specific optical absorption per particle by a combination of the new spectrometer with an aerosol particle counter.
A
photon.23,24 The most common OPOs are based on beta barium borate (BBO) crystals, which combine high conversion efficiencies with high damage thresholds. The combination of PA spectroscopy with wavelength-tunable OPO or CO2 laser sources is described already by several authors.13,25−31 Pulsed as well as modulated sources were employed in the near infrared (NIR) to mid infrared (MIR) spectral range for a highly sensitive detection of different trace gases. However, none of these developments was devoted to aerosol characterization. We present here a PA aerosol absorption spectrometer suitable for absorption measurements of aerosols in the spectral range from 410 to 2500 nm. We employ a pulsed OPO system, which is less cost intensive and more robust than a continuous wave (cw) OPO, while similar sensitivity has to be expected.32
erosol particles are known to significantly influence the global radiation balance.1,2 While light scattering by aerosol particles can be measured by a number of commercial instruments,3,4 there are hardly any routine analytical instruments for aerosol absorption analyses. For a concise overview on these techniques, we refer to the review articles by Moosmüller et al.5,6 The common approaches for aerosol absorption measurements are based on filter sampling prior to the optical characterization, which is known to be prone to artifact formation.7−9 Online analysis of absorption can be based on combined transmission and scattering measurements10,11 or directly by the photoacoustic (PA) effect. Photoacoustics is based on a local warming of an absorbing medium due to light absorption, which again leads to local expansion of the gas. Modulated or pulsed illumination leads to modulated gas expansion, which can be detected by a microphone.12−14 In the special case of PA analysis of aerosols, light gets absorbed by the aerosol particles, and the heat is transferred to the surrounding air, which in consequence expands periodically. Several PA instruments dedicated to aerosol analysis are described in the literature.15−20 All these instruments have in common that they measure the absorption at a single wavelength, defined by the selected laser source. A special instrument was developed by Szabo’s group, which allows for simultaneous measurements at four wavelengths (1064, 532, 355, and 266 nm) generated by a single laser source using different nonlinear optical crystals.21 In a very early publication, Killinger et al. presented a PA system for aerosol characterization equipped with a wavelength-tunable dye laser, though the tuning range of the system was limited to 20 nm for practical reasons.22 We are not aware of any other instrument which allows for a direct measurement either of complete aerosol absorption spectra or at arbitrary single wavelengths. Optical parametrical oscillators (OPOs) are using a nonlinear process to split pump photons into a signal and an idler © 2012 American Chemical Society
■
EXPERIMENTAL SETUP Optics. The layout of the system is depicted in Figure 1. The light source of the spectrometer is a BBO type II-based OPO (Premiscan-ULD, GWU, Germany), pumped by the third harmonic (λ = 355 nm, pulse energy 60 mJ, pulse width 5 ns) of a Q-switched Nd:YAG laser (SpitLight-200, InnoLas GmbH, Germany). With its circular beam and ultralow divergence (ULD), this type of OPO guarantees a good spatial filling of the acoustic resonator where the PA effect takes place. The signal (410−710 nm) and the idler (710−2500 nm) beams, which are generated simultaneously by the OPO, are separated by dichroic mirrors. The signal and the idler beams are 180°folded by prisms, respectively, by mirrors and sent through independent PA cells. As the diameter of the cylindrical resonators in the PA cells (Di = 6 mm) was adapted to the Received: July 31, 2012 Accepted: October 4, 2012 Published: October 4, 2012 8941
dx.doi.org/10.1021/ac302194u | Anal. Chem. 2012, 84, 8941−8945
Analytical Chemistry
Letter
frequency. The input and output windows of the cells are wideband antireflective coated for the respective wavelength range. A gas flow rate of 1 L min−1 through each cell is maintained by a diaphragm gas pump (NMP 850 KNDC, KNF Neuberger, Germany). Since the system was not intended for the analysis of highly transient gas compositions, the gas flow was guided sequentially through the three cells. By a specially designed switching valve, a HEPA-filter (high efficiency particulate air filter, DIF-BN-30, datatronic GmbH, Germany) can be implemented into the gas flow before the PA cells in order to determine the gas absorption background (“gas background”). As opposed to a normal solenoid valve, this valve opens a straight tube of an inner diameter of 4 mm without any edges or bends, thus avoiding particle deposition. Data Acquisition and Handling. The microphones are connected to a Digital Analog Converter (PCI-6221, National Instruments, Germany) which is triggered by the pockels cell trigger of the Nd:YAG laser. In a time window of 10 ms, the amplified PA signals are taken with a time resolution of 5 μs. The signal at each single wavelength is the mean value of 50 laser pulses, which is a compromise between a low standard deviation and an acceptable measuring time. The PA signal we use for further calculations is the intensity of the Fourier transformation of the PA signal at 4.0 kHz (see Figures S-2 and S-3 of the Supporting Information). Without normalization to the corresponding laser pulse energy, the relative pulse to pulse standard deviation of the pulse energies, caused by the flash lamp-pumped Nd:YAG laser itself and the following three nonlinear processes (second and third harmonic generation, OPO-process), are in the range of 20−25%, depending on the wavelength region. From each raw signal, first a constant background, caused by the electronic noise from the laser power supply (“electronic background”), is subtracted. Following, energy normalization is carried out, reducing the pulse to pulse standard deviation to about 3%. The PA absorption spectra of the aerosol particles, as shown for instance in Figure 2, are generated by calculating the difference between the aerosol measurement and gas background measurements, both after electronic background correction and energy normalization.
Figure 1. Layout of the OPO-based PA spectrometer.
beam diameter, no focusing optics were required, thus simplifying the setup and avoiding problems due to the chromatic aberration caused by the broad tuning range especially of the idler beam. A fraction of the 355 nm pump beam back-reflected from the OPO crystal, which is usually sent to a beam dump, is coupled out and guided through a third PA cell, thus expanding the accessible spectral range further into the UV. This fixed-wavelength beam has a pulse energy of up to 5 mJ. However, being a back-reflection from the OPO crystal, the pulse energy depends on the signal/idler wavelength setting of the OPO, which is tuned by rotating the crystal in the beam. Behind each PA cell, each of the beams is guided onto a pyroelectric energy detector (Ophir PE-series with USB-Pulsar II, by ACAL technology, Germany) for monitoring the pulse energy of each individual laser pulse. Output energies of the signal beam are in the range from 0.8 to 8 mJ, depending on the wavelength, while the idler pulse energies range from 0.1 to 1.8 mJ (see Figure S-1 of the Supporting Information). The line width of the OPO system is about 5 cm−1 (150 GHz). This bandwidth in combination with a spectral substructure of the emission makes the laser hardly suitable for gas phase analysis. Calibration of the system in the IR spectral range by a welldefined gaseous absorber is not possible. Photoacoustic Cell. Keeping in mind a future integration in a robust, mobile system, we decided for a rather simple and straightforward PA cell design with a longitudinal resonator, similar to a cell we described earlier.33 The gas in and outlets were modified to circular slits of 1 mm width around the nodes of the longitudinal resonances in order to reduce flow noise (see Figure 1). The resonator itself is a glass tube with an outer diameter of 8 mm and an inner diameter of 6 mm. The microphone (EK 3029, Knowles, GB) is placed in the center of the resonator. The resonance frequency of the resonator at room temperature is 4.0 kHz with a Q-factor of 13.3 (see Figure S-3 of Supporting Information). This rather low Q value ensures stable resonance enhancement for largely varying gas compositions and temperatures, which is crucial for practical field application. Opposite to the microphone, but outside the resonator tube, a speaker is installed which serves to test the microphone performance and to find the precise resonance
Figure 2. Absorption spectra of fresh and aged spark discharge soot, measured with the PA absorption spectrometer. 8942
dx.doi.org/10.1021/ac302194u | Anal. Chem. 2012, 84, 8941−8945
Analytical Chemistry
Letter
Table 1. Temporal Behavior of the Spectral Properties of Pure Propane Soot aging time 35 min particle number conc./cm quality of fit (n = 20) Angstrom exponent 450 nm 550 nm 700 nm
−3
4
65 min
95 min
4
6.23·10 R2 = 0.9481 α = 1.5
4.38·10 R2 = 0.9528 α = 1.60
1.78·10−10 1.23·10−10 7.36·10−11
2.04·10−10 1.42·10−10 8.69·10−11
The complete measurements were handled by a Windows PC and LabView 7.0 (National Instruments, Germany) controlling the OPO scans, data acquisition, and storage of the PA signals as well as of the pulse energy monitors and the output of a condensation particle counter (CPC model TSI 3720, Germany), which serves as reference for some measurements. The pulse energy and particle number density corrected magnitude of the first longitudinal cavity resonance out of the fast Fourier transform (FFT) spectrum was taken as a reference for the absorption. A single wavelength measurement takes about 30 s, including the wavelength tuning of the OPO to any desired value and a 10 s interval we allow for energy stabilization at the chosen wavelength.
4
specific abs./m−1
130 min
3.01·10 R2 = 0.9691 α = 1.66
2.11·104 R2 = 0.9792 α = 1.80
2.20·10−10 1.48·10−10 9.48·10−11
2.37·10−10 1.51·10−10 9.37·10−11
and then be transferred to others, as long as a calibration standard with known absorption at this wavelength λC is available. For a given geometry, the PA signal only scales with the specific absorption and the applied laser energy, which is monitored by our system. As the specific absorption of the GFG soot particles is known from literature,11 a quantitative calibration of our system was possible based on three wavelengths (450, 550, and 700 nm) (see Figure S-5 of Supporting Information). Equally, we were able to establish the limit of detection (LOD), following the 3·sB criterion, on the basis of the same data. As reference for the quantification of the soot, we employed the SMPS system mentioned above. By measuring the particle size distribution and assuming a soot density of 2.0 g cm−3 according to Saathoff et al.,10 quantitative calibration could be carried out and an LOD of 2.5 × 10−7 m−1 for a measuring wavelength of 550 nm was found, which corresponds to about 90 ng m−3 of soot. Although the calibration is not wavelength-dependent, as stated above, the LODs are higher at other wavelengths due to lower pulse energies and thus lower signal-to-noise ratios. At measuring wavelengths of 450 and 700 nm, LODs of 4.8 × 10−7 m−1 and 5.7 × 10−7 m−1, respectively, were found. The lack of aerosol standards with well-defined optical properties in the spectral range beyond 800 nm makes the characterization of the system difficult for the idler spectral range (710−2500 nm), but it can be estimated from the laser pulse energies in this range that the LODs in the IR are at least by a factor of 50 higher than for the signal range (410−710 nm) and more than 100 times beyond 2000 nm. In order to demonstrate the potential of the system in this range, an absorption spectrum measured with the new instrument in comparison with the corresponding HITRAN data is presented in Figure S-4, Supporting Information.45 As mentioned above, the spectral bandwidth and substructure makes a quantitative calibration of the system by the H2O absorption impossible. After the calibration, the PA absorption spectrometer, in combination with the CPC, allows determination of specific particle absorption, i.e., the optical absorption per particle. This concept is demonstrated on the example of aging of propanegenerated soot containing different amounts of iron (see Table 1 and Supporting Information Figure S-6 for the spectra). These aerosol particles are generated by controlled burning of propane with iron pentacarbonyl sprayed into the flame for the simulation of the emissions of heavy-duty maritime diesel engines. Iron content was determined by filter sampling and weighing before and after burning of the filters at 650 °C. Details are described elsewhere.46 Aging was carried out in an in-house made polymer (DuPont material 7288400001 KIND 200A) bag with a content of about 100 L. For fresh, pure propane soot, an Angstrom exponent of 1.50 was found which increases within 2 h to 1.80. Over this period, agglomeration
■
RESULTS AND DISCUSSION Soot and soot-like combustion aerosol, such as black and brown carbon, are among the most scrutinized aerosols regarding their optical properties.34−39 The wavelength dependency of the absorption of this class of particles is described by the Angstrom exponent α, which describes the relation between the mass specific absorption cross section σ on the wavelength σ = γ·λ−α, with γ being a constant factor.40,41 For our first experiments, we compared the Angstrom exponents of fresh (
