PLEASE SCROLL DOWN FOR ARTICLE
Access details: Access Details: [subscription number 908165593] Publisher Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
Journal of Environmental Science and Health, Part A Publication details, including instructions for authors and subscription information: http://www.informaworld.com/smpp/title~content=t713597268
Applying open-path Fourier transform infrared spectroscopy for measuring aerosols
Chang-Fu Wu ab; Yen-Ling Chen b; Chih-Chieh Chen b; Tzu-Ting Yang c; Pao-Erh Chang d
a Department of Public Health, National Taiwan University, Taipei, Taiwan b Institute of Occupational Medicine and Industrial Hygiene, National Taiwan University, Taipei, Taiwan c Department of Environmental
Engineering and Health, Yuanpei University, Hsinchu, Taiwan d Industrial Technology Research Institute, Hsinchu, Taiwan
Online Publication Date: 01 January 2007
To cite this Article Wu, Chang-Fu, Chen, Yen-Ling, Chen, Chih-Chieh, Yang, Tzu-Ting and Chang, Pao-Erh(2007)'Applying open-path Fourier transform infrared spectroscopy for measuring aerosols',Journal of Environmental Science and Health, Part A,42:8,1131 — 1140
To link to this Article: DOI: 10.1080/10934520701418631 URL: http://dx.doi.org/10.1080/10934520701418631
Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf
This article may be used for research, teaching and private study purposes. Any substantial or systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or distribution in any form to anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.
Journal of Environmental Science and Health Part A (2007) 42, 1131–1140 CopyrightCTaylor & Francis Group, LLC
ISSN: 1093-4529 (Print); 1532-4117 (Online) DOI: 10.1080/10934520701418631
Applying open-path Fourier transform infrared spectroscopy
for measuring aerosols
CHANG-FU WU1,2, YEN-LING CHEN2, CHIH-CHIEH CHEN2, TZU-TING YANG3 and PAO-ERH CHANG4
1Department of Public Health, National Taiwan University, Taipei, Taiwan
2Institute of Occupational Medicine and Industrial Hygiene, National Taiwan University, Taipei, Taiwan 3Department of Environmental Engineering and Health, Yuanpei University, Hsinchu, Taiwan
4Industrial Technology Research Institute, Hsinchu, Taiwan
This paper examines the feasibility of using Open-Path Fourier Transform Infrared Spectroscopy (OP-FTIR) to measure aerosols. The extinction spectra of water, ammonium nitrate, and ammonium sulfate aerosols were first simulated with various particle size distributions (geometric mean ranged from 2 to 10µm; geometric standard deviation ranged from 1.1 to 2.5) based on the Mie theory. An optimization procedure was developed to retrieve the geometric mean and standard deviation of the aerosols size distributions from the spectra, assuming that the complex refractive index is known. To test sensitivity, we also added 4%, 7%, and 10% noise levels to the spectra and compared the reconstruction results. In the experimental study, water aerosols were generated by a two-fluid nozzle inside a cylindrical chamber (3325 cm3). The extinction spectrum was collected with a modified FTIR and the size distribution information was
retrieved following the same optimization procedure as the one used in the simulation study. The optimization procedure developed in this study reconstructed the size distribution reasonably well for particles with known refractive index (i.e. homogeneous or internally mixed aerosols). The results were robust with the added noise levels up to 10%, after removing inaccurate estimates with the use of the censoring criteria for reconstructed GSD<1.3, reconstructed GM<2.5 µm and GSD<1.5, and reconstructed GM>10 µm. With regard to externally mixed aerosols, the reconstructed results were sensitive to the noise within the measuring systems, although most ambient aerosols were internally mixed. The reconstructed size distribution in the chamber experiment had a GM of 3.85µm and GSD of 1.70. The simulation results were applied to support this reconstruction result. We conclude that OP-FTIR can be used to measure aerosols and screen for the right region for a more detailed aerosol measurement campaign.
Keywords: Particulate matter, OP-FTIR, extinction spectrum, Mie theory, Smooth Basis Function Minimization.
Introduction
Particles and gases need to be measured simultaneously to assess air quality accurately. Most air monitoring in-struments measure either gaseous (e.g., using organic va-por monitor) or particulate (e.g., using dust monitor) air pollutants, not both. One technique potentially capable of monitoring both at the same time is Open-Path Fourier Transform Infrared (OP-FTIR) spectroscopy, an optical re-mote sensing (ORS) technique that is conventionally used for monitoring gaseous air pollutants.[1−−7]When there are aerosols in measurement beam path, baseline features of the spectra change.[8]This change is usually considered un-desirable, but ”shearing” of the baseline of the spectra ac-Address correspondence to Chang-Fu Wu National Taiwan Uni-versity, Room 717, No.17, Xu-Zhou Rd., Taipei 100, Taiwan; E-mail: [email protected]
Received January 11, 2007.
tually contains important information about the physical and chemical characteristics of aerosols.
With regard to retrieving aerosol size distribution infor-mation from OP-FTIR measurements, Arnott et al.[9] gen-erated water clouds in a 1 m3chamber and used FTIR to collect the extinction spectra of the water aerosols. In that same study, droplet size information was retrieved using an iterative algorithm developed based on the Mie theory.[10] Hashmonay and Yost,[8]conducting a similar experiment, generated water droplets in a standard shower chamber (0.85 m width, 1.18 m length, 1.53 m height). Assuming a Gaussian size distribution for the aerosols, they calculated the extinction spectrum of water aerosols based on the Mie theory. They then manually matched the calculated extinc-tion spectrum with the measured spectrum by changing the parameters of the size distribution.
The aim of this study was to develop an optimization pro-cedure for reconstructing aerosol size distributions from the OP-FTIR measurements, extending the methodology
developed by Hashmonay and Yost.[8]In this study, we de-veloped an optimization procedure to find a best-fit solu-tion, and then conducted a series of simulations to eval-uate how this procedure performed for different aerosols with various size distributions. We also performed an ini-tial chamber experiment to demonstrate the feasibility of this approach.
Theory development
The theory of retrieving aerosol size distribution from the OP-FTIR spectra was discussed in Hashmonay and Yost[8] and summarized briefly as follows. The light extinction phe-nomenon due to aerosols is described by the Mie theory.[10] The associated extinction spectrum can be computed if the extinction efficiency (Qe) is known. The Qe is a func-tion of the particle size, light wavelength, and the complex refractive index m at the corresponding wavelengths, and can be calculated using the recurrence procedure described in Wickramasinghe.[11] The refractive index is usually ex-pressed as a complex number:[12]
m= n − ik (1)
where n is the real refractive index (real part), i is√−1, and
k is related to the absorption coefficient (imaginary part)
of the aerosols.
The extinction coefficient (σe) is calculated from the Qe and the size distribution of the aerosols according to the relationship of:[12] σei = π 4 j QeijNjd 2 j (2)
where Nj is the number concentration at the jt h particle size class; dj is the mean diameter for the jt h particle size class; i is the index number for the wavelength; and j is the index number for the particle size class.[8] The extinction spectrum, i.e., the OP-FTIR spectrum, is then calculated from multiplying extinction coefficient by OP-FTIR beam path length.
During the retrieval process, the spectra for various size distributions of certain compounds are calculated, assum-ing the complex refractive index is known. When a specific size distribution results in a match between the calculated spectrum and target spectrum (e.g., spectrum collected ex-perimentally), that size distribution is regarded as the solu-tion.
Method
Simulation investigation
In the first part of this study, we obtained the complex re-fractive index information from the published literature and simulated infrared spectra on computer, assuming different
aerosol size distributions. We then applied the Mie theory to develop an optimization procedure to reconstruct the in-frared spectra and obtain the underlying size distribution information. The total particle number concentration and distribution parameters predicted from the reconstruction procedure were compared to the input parameters that were used to generate those simulated spectra.
Single compound scenario
The complex refractive indices of three compounds were obtained from published literature: water,[13] ammonium sulfate,[14] and ammonium nitrate.[15]The FTIR spectrum for each compound was simulated independently accord-ing to Equation 2. The latter two compounds were chosen originally because they are two important anthropogeni-cally produced components of ambient particulate matters. Assuming a log-normal distribution for the aerosols, vari-ous distribution parameters (i.e., geometric mean, or GM, ranged from 2 to 10 µm; geometric standard deviation, or GSD, ranged from 1.1 to 2.5) were applied to obtain the corresponding extinction spectra. All the distributions were assumed to have a total particle number concentration of 10,000 #/cm3. Although it is rare to have only aerosols with one pure chemical species in the ambient environment, testing on the single compound scenario helped identify the limitation of our optimization procedure since fewer pa-rameters were involved in the inversion process, compared to testing on mixtures of compounds.
Multiple compounds scenario
Spectra for mixtures of the three compounds were also sim-ulated and tested. The mixed compounds were classified into two categories: external and internal mixtures.[16] An external mixture is defined as the condition in which each aerosol is chemically pure and aerosols are mixed in space without coagulation. The extinction coefficient of external mixture (σem) is the sum of the contributions from each pure species (σemk): σem = N k=1 σemk (3)
where N (=3 in this study) is the total number of species. On the contrary, an internal mixture is defined as the condition in which different chemical species are mixed in each aerosol in a fixed proportion. To calculate its extinction coefficient, the effective refractive index of the mixture is computed us-ing the refractive indices of the individual species based on various mixing rules (e.g., the volume weighted method[16]).
Optimization procedure
Theoretically, we should be able to reconstruct the N and d information from the measured OP-FTIR spec-trum, assuming the complex refractive index is known
Applying open-path fourier transform infrared spectroscopy for measuring aerosols
1133
(Equation 2). The process of searching for a solution de-scribed in the Theory Development section can be done manually or solved by the nonlinear least-squares optimiza-tion methods. In this study, we developed an optimizaoptimiza-tion procedure similar to the Smooth Basis Function Minimiza-tion (SBFM) method used in radial plume mapping anal-yses with the OP-FTIR measurements.[17−19]With this ap-proach, a known smooth basis function (e.g., log-normal distribution) with unknown parameters is assumed to de-scribe the size distribution of the aerosols. The SBFM ap-proach requires a ‘predicted’ extinction spectrum and an error function for minimization. The predicted extinction spectrum can be calculated from Equation 2 with the as-sumed distribution.
The error function is defined as the sum of square dif-ferences between the test and the predicted spectra. It is minimized by optimization functions provided in the Mat-lab computer software (MathWorks, Inc., Natick, MA) to yield a set of parameters that produce the best-fit between the test and predicted spectra. For single compound and in-ternal mixture scenarios, only three parameters need to be retrieved (i.e., GM, GSD, and particle number concentra-tion). However, for external mixture scenario, nine param-eters need to be retrieved (i.e., three paramparam-eters for each of the three compounds). Mathematically, the reconstruction performance in the internal mixture scenarios should be similar to the one in the single compound scenarios, since the internally mixed aerosols were also assumed to have one set of known complex refractive index (compared to the three sets of complex refractive indices for externally mixed aerosols). This was also confirmed by our prelimi-nary study results. Thus, for the multiple compound sce-narios, only the results from the externally mixed aerosols are presented.
Data analysis
The Pearson’s correlation coefficient (R) values between the test and the reconstructed spectra (Rspec) were used to repre-sent the fitness of the reconstructed spectrum. The R values
between the test and reconstructed size distributions (Rsize) were used to represent the fitness of the reconstructed size distribution. For each compound and size distribution, we also calculated the percent difference between the test and reconstructed total particle number concentrations (N%). In addition, we conducted a series of sensitivity analysis by adding 4%, 7%, and 10% noise levels to the simulated spectra and observed how they affected the reconstruction results.
Experimental investigation
An initial chamber experiment was conducted using a FTIR system (Illuminator, MIDAC, California, USA) with a modified White’s cell. The outer glass shell of the White’s cell was removed so that aerosols generated outside the chamber could enter the chamber. The cylindrical cham-ber was 3325 cm3in size (radius= 5.5 cm; length = 35 cm) and was constructed with 0.5 cm thick plexiglass (Figure 1). One 165 cm2 opening was located in the center of the chamber floor for aerosols access. A two-fluid nozzle (SU2, Spraying Systems Co., Wheaton, USA) with a sprayed an-gle of 15◦is set up to generate aerosols with the air pressure at 25 lb/in2. The spectrum was collected with the sampling regime using 32 co-averaged scans at 2 cm−1resolution. The total beam path length was 440 cm.
The crude spectra contained very fine absorbance fea-tures that could not be accounted for using the available complex refractive index information, so we smoothed the measured extinction spectra before subjecting to analysis. The smoothing process, which was performed by calculat-ing the 50-point movcalculat-ing average of the spectra, also reduced the noise levels for better reconstruction results.
Results and discussion
Simulation—single compound
The complex refractive indices and the spectra simulated by assuming that the GM was 5 or 10µm and GSD was
Fig. 1. The experimental setup of using a two-fluid nozzle to generate water aerosols.
Fig. 2. The complex refractive index of water, ammonium nitrate, and ammonium sulfate and the corresponding extinction spectra
with size distributions of (GM= 5 µm, GSD = 2) and (GM = 10 µm and GSD = 2).
2 are shown in Figure 2. The locations of the minima of the spectra (e.g., 950, 1650, and 3500 cm−1in Figs. 2d and 2g) were approximately the same as the locations of the minima of the real part of the complex refractive index (e.g., 950, 1650, and 3500 cm−1in Fig. 2a). These locations were independent of the assumed size distributions, suggesting that these minima features are the signature characteristics of these aerosols. The effects of size can be identified from the overall shape of the baseline features. For all the three compounds, the baseline of the spectra of larger particles (Fig. 2, the third row) had sharper slopes in the 2000 to 3500 cm−1spectral range than the baseline of the spectra of smaller particles (Fig. 2, the second row).
Figures 3a and 3c are examples of the test and recon-structed size distributions for water aerosols, assuming that the test distribution had a GM of 2 and 5µm, respectively, and both had GSD of 1.5. Their corresponding OP-FTIR spectra are presented in the right column in Figure 3. The Rspec and Rsize levels in both examples were larger than 0.99. The more systematic testing results are shown in Fig-ure 4, in which the x- and y-axes represent the test GM and
GSD of the water aerosols, respectively. The grayscale bar in Figure 4a represents the Rspecwhile the grayscale bar in Figure 4b represents the Rsize. In ideal conditions (i.e., noise level= 0%), the optimization procedure was able to recon-struct the spectra with good Rspec for almost all the test distributions. The Rsize was generally good for most test distributions except for the distributions with GSD= 1.1, in part because the Rsizeparameter for a narrow distribution is relatively sensitive to the shift of the particle size distribu-tion. For example, in Figure 3e (GM= 9 µm, GSD = 1.1), the Rsizeis only 0.36 but the reconstructed distribution was quite similar to the test distribution when inspected visually. Another region worthy of notice is the upper-right cor-ner of Figure 4b. There the Rsizevalues were also low. This is because that the Mie theory is generally applied to con-ditions in which the particle size and the wavelength of the light source are of the same order of magnitude.[10,12] In our case, the assumed wavelength of the FTIR spectrome-ter ranged from 2.5 to 25µm. Therefore, when there was a greater portion of larger particles, the reconstructed size distribution was less than ideal, as can be seen in Figure 3g
Applying open-path fourier transform infrared spectroscopy for measuring aerosols
1135
Fig. 3. Examples of the test and reconstructed size distributions (left column) and their corresponding spectra (right column) for
water aerosols. (a) & (b) test GM= 2 µm, test GSD = 1.5; (c) & (d): test GM = 5 µm, test GSD = 1.5; (e) & (f): test GM = 9 µm, test GSD= 1.1; (g) & (h): test GM = 8.5 µm, test GSD = 2.3.
(test GM = 8.5 µm, GSD = 2.3, Rsize= 0.86). With re-gard to total particle number concentrations (Fig. 4c), the N%was less than 10% in all regions, except in cases in which distributions had GSD= 1.1 or in which distributions had larger particles (i.e. lower-left and upper-right regions).
Sensitivity analysis was performed by adding different levels of noise to the test spectra of water aerosols (Figs. 4d– 4l). As expected, the more noise, the greater the decrease in
Rspec (Fig. 4, first column). The 5th∼95th percentiles of Rspecwere 0.96∼1, 0.90∼1, and 0.81∼1, respectively, when 4%, 7%, and 10% noises were added to the test spectra. This phenomenon was more apparent for distributions with larger particles (upper-right region), because larger parti-cles have stronger extinctions and the noise was added in proportion to the strength of extinction signals. Regardless of noise levels, the Rsizeand the N%(Fig. 4, second and third
Fig. 4. The sensitivity analysis results by adding different levels of noises to the test spectra for water aerosols. The grayscale bars in
the 1st to the 3rd columns represent the Rspec, Rsizeand N%, respectively. The 1st to the 4th rows represent the results from adding
0%, 4%, 7%, 10% noises, respectively.
columns, respectively) did not vary significantly, indicating that the reconstruction algorithm is robust enough even for spectra with a noise level of 10%.
Similar results were observed for ammonium nitrate and ammonium sulfate aerosols. The optimization procedure reconstructed the spectra well for most distributions for noise level of 0% (5th∼95th percentile of Rspec= 0.98∼1). In addition, the Rsize was larger than 0.98 and N% was smaller than 10% for most distributions, except in cases in which the distribution was narrow or the distribution had larger particles. The sensitivity analysis confirmed the robustness of the optimization algorithm. That is, the Rsize and N% did not vary as fast as the Rspec when noise was added to the spectra.
The above analyses also suggest that the Rspecalone is not a sufficient indicator for evaluating the quality of the recon-structed size distribution. Combining the results from the three types of aerosols with all noise levels, we found that po-tentially inaccurate results can be identified using the three censoring criteria: (i) reconstructed GSD<1.3 for narrow distribution; (ii) reconstructed GM<2.5 µm and GSD<1.5 for distribution with a significant portion of small particles; and (iii) reconstructed GM>10 µm for distribution with a
significant portion of large particles. After applying these censoring criteria to the reconstructed distributions, the me-dian values (with 5th∼95th percentiles) of the Rsizeand N% were 0.99 (0.95∼1) and 3.3% (0.5%∼17.4%), respectively. More stringent criteria could be applied to improve the quality of fit, though this would also limit the applicable aerosol size range.
Simulation—multiple compounds
In Figure 5a can been seen one example of the test and re-constructed size distributions for external mixture. The size distributions of each compound were the same (GM= 5
µm and GSD = 1.5). The corresponding OP-FTIR
spec-trum is presented in Figure 5b. As in the single compound scenario, the optimization procedure retrieved the spectrum accurately and produced a good estimate of the aerosol size distribution for all the three compounds (Rsize> 0.99 and N%< 2.1%). We also tested the situation in which each compound had different distributions. As can be seen in Figures 5c and 5d, the searching algorithm produced good results (Rsize> 0.99 and N%< 2%).
Applying open-path fourier transform infrared spectroscopy for measuring aerosols
1137
Fig. 5. Examples of the test and reconstructed size distributions (left column) and their corresponding spectra (right column) for
externally mixed aerosols. (a) & (b): test GM= 5 µm, test GSD = 1.5 for all compounds; (c) & (d): test GM = 5, 6, 7 µm for water, ammonium nitrate, ammonium sulfate aerosols, respectively. The test GSD is 1.5 for all compounds.
The results from a more systematic testing approach are shown in Figures 6a–6g. The size distributions of each com-pound were assumed to be the same but the optimization procedure was designed to retrieve three sets of size pa-rameters (one set for each type of aerosols) in each of the simulated spectrum of the externally mixed aerosols. In an ideal situation with a noise level of 0%, the algorithm re-constructed well for all the spectra. The Rsize (5th∼95th percentiles: 0.98∼1) was higher than 0.98 and the N% (5th∼95th percentiles: 2%∼12%) was smaller than 12% for most size distributions, except for distributions with small GSDs or small particles. Unlike in the single compound scenarios, our sensitivity analysis showed that adding only 4% noises affected the reconstruction quality (Figs. 6h–6n). The 5th∼95th percentiles of Rsizewere 0.87∼1, 0.73∼1, and 0.57∼1, respectively, when 4%, 7%, and 10% noises were added to the spectra. This is because the noise was added in the composite spectrum (Equation 3) and the algorithms were not be able to smooth out the noise in the presumed individual compound’s spectrum.
In this study, in addition to water aerosols, we also chose to test ammonium sulfate and ammonium nitrate. Al-though these two aerosols usually exist in the ambient
envi-ronment as small particles and thus may not be suitable for FTIR measurement, they represent some of the few com-pounds having complete complex refractive indices. Most other compounds (e.g., mineral aerosol particles)[20] have the complex refractive index only at a few wavelengths. The similarity in results obtained from all the three test com-pounds provide adequate evidence that the FTIR can be used for aerosol monitoring when the complex refractive index of the aerosols in interest are available at multiple wavelengths. One possible extension of this technique may be its potential use for remote sensing instruments with light sources of shorter wavelength (such as the Ultraviolet Dif-ferential Optical Absorption Spectroscopy, or UV-DOAS) to detect submicrometer particles.
Chamber experiment
The spectrum collected using the two-fluid nozzle to gener-ate wgener-ater aerosols and the reconstructed spectrum is shown in Figure 7b. There were three minima at about 950, 1650, and 3500 cm−1corresponding to the minimum location in the real part of the complex refractive index of water (Fig. 2a). The Rspec value was 0.96 and the reconstructed
Fig. 6. The sensitivity analysis results by adding different levels of noises (a–g: 0 %; h–n: 4 %) to the test spectra for externally mixed
aerosols. The grayscale bars in the 1st to the 3rd columns represent the Rspec, Rsizeand N%, respectively. The 1st & 4th, 2nd & 5th, and
3rd and 6th rows represent the results for the components of water, ammonium nitrate, and ammonium sulfate aerosols, respectively.
parameters were GM= 3.85 µm, GSD = 1.70, and N = 2833 #/cm3. As stated earlier, this initial experiment was conducted to demonstrate the feasibility of the optimiza-tion procedure. Thus, no non-FTIR measurements were collected to verify the reconstruction results experimentally. The fact that the reconstructed size distribution did not fit the censoring criteria, as proposed in the simulation study, gives evidence of its validity.
We originally attempted to use the Aerodynamic Parti-cle Sizer (APS, TSI, Inc., Shoreview, MN, USA) to measure aerosol size distribution simultaneously. However, APS and most conventional aerosol measuring instruments (such as the impactors or time-of-flight instruments) were found to be unsuitable for measuring volatile aerosols (e.g., the wa-ter aerosols), since these instruments usually require the aerosols be sampled into the sensing chamber and the
Applying open-path fourier transform infrared spectroscopy for measuring aerosols
1139
Fig. 7. (a) The reconstructed aerosol size distribution (GM= 3.85 µm, GSD = 1.7) in the chamber experiment. (b) The spectrum
collected with a FTIR spectrometer and the reconstructed spectrum.
aerosol property might change during the sampling pro-cess. This contrasts the in-situ types of measurements from the OP-FTIR for which no sampling of air is required.
To apply this technique in the field, two issues should be addressed. First, availability of known refractive index at multiple wavelengths is crucial for this inversion process. Where such information is unavailable, one could conduct certain calibration experiments (e.g., calibrate the extinc-tion signals against filter samples) to obtain a suitable cor-rection factor. This is done routinely for many conventional aerosol monitoring devices such as the nephelometer or other light scattering instruments. Second, judging from the inherent uncertainty and the N% error, this system might be better used as a screening tool. For example, an OP-FTIR or UV-DOAS, with multiple scanning beam paths, could scan a wide area first to identify the regions with high aerosol concentrations. Then, a more established aerosol measuring system (e.g. impactors or tapered element oscil-lating microbalance) can be used at the identified region to obtain more accurate results.
Conclusions
The simulation results showed that the optimization proce-dure developed in this study reconstructed reasonably well for particles with known refractive index (i.e., homogeneous or internally mixed aerosols). The results were robust with added noise levels up to 10%. We also found that fitness between the test and reconstructed spectra (i.e. the Rspec) alone is not a sufficient indicator of the quality of the re-constructed size distribution. The result can be improved, however, by implementing appropriate censoring criteria based on the reconstructed GM and GSD. With regard to the externally mixed aerosols, the results of our recon-structions were sensitive to noise within the measuring sys-tems. Note that in the ambient environments, most aerosols were internally mixed. In the initial chamber experiment, we demonstrated that the optimization procedure could re-construct the size distribution of water aerosols generated in a controlled environment. Future studies should consider
generating other types of aerosols to test the performance of this system.
Acknowledgments
The authors thank the National Science Council of Tai-wan for financially supporting this research under contract NSC94-2320-B002-040.
References
[1] Grant, W.; Kagann, R.; Mcclenny, W. Optical remote measurement of toxic gases. J. Air Waste Manag. Assoc. 1992, 42(1), 18–30. [2] Childers, J.W.; Thompson, E.L.; Harris, D.B.; Kirchgessner, D.A.;
Clayton, M.; Natschke, D.F.; Phillips, W.J. Multi-pollutant concen-tration measurements around a concentrated swine production fa-cility using open-path FTIR spectrometry. Atmos. Environ. 2001, 35(11), 1923–1936.
[3] Ross, K.R.; Todd, L.A. Field evaluation of a transportable open-path FTIR spectrometer for real-time air monitoring, Appl. Occup. Environ. Hyg. 2002, 17(2), 131–143.
[4] Yan, L.; Wang, J.D.; Huang, Z.H.; Xu, H.Q.; Zhou, X.T. Monitoring leaking gases by OP-FTIR remote sensing, J. Environ. Sci. Health A 2002, 37(8), 1453–1462.
[5] Bacsik, Z.; Mink, J.; Keresztury, G. FTIR spectroscopy of the at-mosphere. I. Principles and methods. Appl. Spectrosc. Rev. 2004, 39(3), 295–363.
[6] Shores, R.C.; Harris, D.B.; Thompson, E.L.; Vogel, C.A.; Natschke, D.; Hashmonay, R.A.; Wagoner, K.R.; Modrak, M. Plane-integrated open-path Fourier transform infrared spectrometry methodology for anaerobic swine lagoon emission measurements. Appl. Eng. Agric. 2005, 21(3), 487–492.
[7] Bacsik, Z.; Komlosi, V.; Ollar, T.; Mink, J. Comparison of open path and extractive long-path FTIR techniques in detection of air pollutants. Appl. Spectrosc. Rev. 2006, 41(1), 77–97.
[8] Hashmonay, R.; Yost, M. On the application of open-path Fourier transform infra-red spectroscopy to measure aerosols: observa-tions of water droplets. Environ. Sci. Technol. 1999, 33(7), 1141– 1144.
[9] Arnott, W.P.; Schmitt, C.; Liu, Y.G.; Hallett, J. Droplet size spectra and water-vapor concentration of laboratory water clouds: Inver-sion of Fourier transform infrared (500–5000 cm−1) optical-depth measurement. Appl. Opt. 1997, 36(21), 5205–5216.
[10] Bohren, C.F.; Huffman, D.R. Absorption and Scattering of Light by Small Particles. Wiley, New York, 1983.
[11] Wickramasinghe, N.C. Light Scattering Functions for Small Parti-cles, with Applications in Astronomy. Wiley, New York, 1973. [12] Hinds, W.C. Aerosol Technology: Properties, Behavior, and
Measure-ment of Airborne Particles. 2nd ed. Wiley, New York, 1999. [13] Downing, H.D.; Williams, D. Optical-constants of water in infrared.
J. Geophys. Res. 1975, 80(12), 1656–1661.
[14] Toon, O.B.; Pollack, J.B.; Khare, B.N. Optical-constants of several atmospheric aerosol species - ammonium-sulfate, aluminum-oxide, and sodium-chloride. J. Geophys. Res. Oceans Atmos. 1976, 81(33), 5733–5748.
[15] Jarzembski, M.A.; Norman, M.L.; Fuller, K.A.; Srivastava, V.; Cut-ten, D.R. Complex refractive index of ammonium nitrate in the 2-20-mu m spectral range. Appl. Opt. 2003, 42(6), 922–930.
[16] Ouimette, J.R.; Flagan, R.C. The extinction coefficient of mul-ticomponent aerosols. Atmos. Environ. 1982, 16 (10), 2405– 2419.
[17] Drescher, A.C.; Gadgil, A.J.; Price, P.N.; Nazaroff, W.W. Novel ap-proach for tomographic reconstruction of gas concentration distri-butions in air: use of smooth basis functions and simulated anneal-ing. Atmos. Environ. 1996, 30(6), 929–940.
[18] Hashmonay, R.A.; Yost, M.G.;Wu, C.F. Computed tomography of air pollutants using radial scanning path-integrated optical remote sensing. Atmos. Environ. 1999, 33(2), 267–274.
[19] Wu, C.F.; Yost, M.G.; Hashmonay, R.A.; Park, D.Y. Ex-perimental evaluation of a radial beam geometry for map-ping air pollutants using optical remote sensing and com-puted tomography. Atmos. Environ. 1999, 33(28), 4709– 4716.
[20] Volten, H.; Munoz, O.; Rol, E.; De Haan, J.F.; Vassen, W.; Hovenier, J.W.; Muinonen, K.; Nousiainen, T. Scattering matrices of mineral aerosol particles at 441.6 nm and 632.8 nm. J. Geophys. Res. Atmos. 2001, 106(D15), 17375–17401.
