Particulate Matter Measurements Using Open-Path Fourier Transform
Infrared Spectroscopy
Ram A. Hashmonay
ARCADIS Gcraghty & Miller, Inc., P.O. Box 13109, Research Triangle Park, NC 27709
D. Bruce Harris
U.S. EPA, ORD, NRMRL, Mail Drop 61, Research Triangle Park, NC 27711
Abstract
Open-path Fourier transform infrared (OP-FTIR) spectroscopy is a well-accepted technology
designed for measuring gaseous air contaminants. OP-FTIR absorbance spectra acquired during
changing aerosol conditions reveal related changes in very broad baseline features. Usually this
"shearing" of a spectrum's baseline is viewed as undesirable because it interferes with
quantifying gases. However, this paper demonstrates that this wavelength-dependent absorbance
can be used to measure particulate matter (PM). A developed inversion algorithm for retrieving
the size distribution of the PM in the beam path is described. This paper presents preliminary
results from OP-FTIR measurements conducted downwind from an active dirt road. Using the
developed inversion algorithm and measured optical properties, the calculated extinction spectra
were fitted to the measured extinction by using the Mie theory for spherical particles. Time series
data reveal significant rapid shifts in size distribution of the PM in the OP-FTIR. Results indicate
that size distribution parameters may be retrieved from OP-FTIR spectra acquired over an open
optical path. The suggested method may provide real-time concentrations of gaseous and PM
contaminants simultaneously. Emission flux estimates of PM may be generated for near-ground-
level line (roads) and area sources when several beam paths are deployed downwind from the
source.
Introduction
OP-FTIR spectroscopy is an accepted technique to measure gaseous air toxics and volatile
organic compounds.1 OP-FTIR instruments pass an IR light along an open-beam optical path up
to 1 km total length to measure and identify chemical contaminants directly in the field. This
yields real-time data (<1 minute/sample) for multiple chemical species typically with part-per-
billion detection levels. However, in many field studies it was observed that OP-FTIR
absorbance spectra acquired during changing aerosol conditions (dust and fog) revealed related
changes in very broad features of the baseline. Usually, these changes in the spectrum's baseline
were viewed as interfering with quantifying gases. We have published that these wavelength-
Abstract No. 550 Session No. AO-1b
Page:
1/1
-------
dependent baseline features ean be quantitatively related to changes in the aerosol size
distribution and type (-). We applied Mie theory to an assumed size distribution function and
known optical properties of water to match the measured extinction spectrum of condensation
water aerosol collected in a controlled shower chamber. In this paper an inversion algorithm is
presented and tested for the same water condensation aerosol spectra collected in the shower
chamber. In order to quantify these broad spectral features for the size distribution, one must
know the optical properties of the PM. Extending the analysis to non-water particles requires
collecting particles on filters concurrently with OP-FTIR measurements. Analyzing for the
optical properties of the sampled material allows the subsequent measurement of the changes in
size distribution and concentration of target particles in high temporal resolution.
Optical Properties
Light extinction (absorption + scattering) due to PM as a function of wavelength can be
computed by Mie theory if the optical properties as a function of wavelength and the size
distribution of the PM are known. The optical properties are expressed in terms of the complex
refractive index m, and are defined by (1):
111 = n - in K (1)
Where the real part, n, is the refractive index defined as the ratio between the wavelength in free
space, and the wavelength in matter, A.; K" is the absorption index. The imaginary part of the
complex refractive index is the absorption factor, and the absorption index, k, is related to the
absorption coefficient of Beer's law, a, by:
aX
K = (2)
An
Several methods have been developed to determine the real and imaginary parts of the complex
refractive index of PM in the IR spectral region. Most of them rely on the analysis of PM
samples and can be classified into three groups: 1) Modeling methods; 2) Diffused reflectance
methods; and 3) Transmission methods. (4) Our research program focuses on the second group
using a bench-top FTIR spectrometer with a suitable diffuse reflectance attachment. In both the
reflectance <5) and transmission (6) approaches, a Kramers-Kronig analysis is performed to retrieve
the complex refractive index spectrum. This analysis needs only a reflection or transmission
spectrum, which is easily measured with an ordinary bench-top FTIR spectrometer with the
commercially available attachment. There are some advantages in employing the Kramers-
Kronig analysis for an external reflection spectrum near the normal angle of incidence. The key
advantage is that it does not require a least squares solution that could diverge, and further, no
prior information on the sample material is needed. Therefore, this approach will be explored for
calculating the complex refractive of the target PM in this research program.
Abstract No. 550 Session No. AO-1b
Pago:
2/2
-------
Size Distribution Inversion
The extinction efficiency is a function of particle size, light wavelength, and the complex
refractive index, at a given wavelength for a monochromatic incident light source. The relevant
size parameter x of Mie theory (7) is defined as the ratio between the particle diameter d and the
wavelength X multiplied by K. However, when a broad spectral band of incident light is detected,
one has to calculate the extinction efficiency for each combination of particle size and
wavelength, due to changes in the complex refractive index over the spectral domain.
The extinction efficiency Qe, for a given m and x, is a function of the Mie coefficients, an and
bn.(1) Our method for calculating the Mie coefficients followed the recurrence procedure in
Wickramasinghe (7). We summarize the extinction efficiency values in a matrix Qe ,j where each
row (/) is for a different wavelength (or wavenumber) and each column (j) is for a different
particle size class. We calculate the extinction efficiency matrix Qe ,j, using the previously
measured complex refractive index spectrum, for 48 wavenumber values over the range of 500 -
5000 cm"1 (20 - 2 Jim, respectively), and for a variable user-defined number of diameter size
values between 0.2 to 32 |im. We expressed the extinction coefficient
-------
/;n)=ri(l+Ğ';<",a,y) (5)
i
Where a is an empirically determined relaxation factor. The next step vector of concentrations,
Cj"+,>, is then calculated as:
c
(H + l) _ r () ()
=/rcr ^
This iterative procedure proceeds until the difference of the criteria parameter between
consequent steps, reaches a very small threshold value.
We applied our developed inversion algorithm to the spectra collected in the controlled shower
chamber. Prior to the inversion, we deresolved the spectrum from 2 cm"1 down to 90 cm"1, to
comply with the algorithm requirements. This means that an OP-FTIR spectrum in very low
resolution may be collected, avoiding most of the gaseous absorption interference.
In Figure 1 we summarize the inversion results for the two spectra presented in our earlier study H
Graphs 1A and 1C overlay the fitted spectrum on the measured spectrum. The corresponding size
distributions are shown in Graphs IB and ID, respectively. The results of the inversion procedure
are consistent with the monitored processes in the shower chamber(2). The locations of the three
minima of the spectra baseline are due to the complex refractive index function, and are
independent of the size distribution. These minima behave like a signature of water aerosol in this
spectral range. The mismatch between the measured and fitted spectra at about 1500 and 3500 cm"1
are due to water vapor absorption.
The Concordance Correlation Factor (CCF)9 was chosen as a measure to represent the quality of
the reconstructed spectrum relative to the measured spectrum. The CCF is similar to the Pearson
correlation coefficient, but is adjusted to account for shifts in location and scale. Like the
Pearson correlation, CCF values are bounded between -1 and 1, yet the CCF can never exceed the
absolute value of the Pearson correlation factor. The CCF will be equal to the Pearson
correlation when the linear regression line intersects the ordinate at 0 and its slope equals 1. This
measure was calculated from a linear regression of paired point values taken from the fitted and
measured spectra. The CCF results for both inversions are quite high, indicating that the
underlying physical process of the inversion algorithm provides an appropriate prediction of the
measured spectrum.
Abstract No. 550 Session No. AO-1b
Paqe:
4/4
-------
D
<
-------
A
0.15
3
<
<1)
O
c
0.1 Ĥ
JD
<
0.05
700 1400 2100 2800 3500 4200 4900
Vfeuemrrber [cm1]
0.15
^ 0.1
v) 0.05
£>
<
700 1400 2100 2800 3500 4200 4900
V\faveruTber[cm1]
B
D
0.15
w 0.05 -
0.15
D
<
c
a
XI
0.1 -
3 0.05
XI
<
700 1400 2100 2800 3500 4200 4900
V\favwxrrter [cm1]
700 1400 2100 2800 3500 4200 4900
Wfewemrrtw [cm1]
Figure 2. Examples of absorbance spectra where different PM contents exist in the OP-FTIR beam. A
significant <10 um; B significant <7 um; C significant <5 um; and D few particles.
The developed inversion algorithm was applied to the three spectra shown in Figure 2, which
contained particle information. As for the water aerosol experiment, the spectrum was deresolved
from 2 down to 90 cm"1 prior to the inversion to comply with the algorithm requirements. Since the
complex refractive index for the sampled dust has not been determined yet, assumed similar6
optical properties of dust from the Sahara Desert were applied. Figure 3 illustrates the optical
properties of the Sahara dust as measured by Volz10
Page:
Abstract No. 550 Session No. AO-1b 6/6
-------
3.5
x 3
ĤS 2.5
a>
>
*->
o
(0
M
0)
DC
2
1.5
1
0.5
0
Ref. Index Abs. Index
V
V *
1
X
-------
0.12
A. Spectral Fit 1 (CCF=0.71)
C. Spectral Fit 3 (CCF=0.76)
Reconstructed Spectrum Measured Spectrum
Reconstructed Spectrum
Measured Spectrum]
2 0.06
5 0.06
1500 2500 3500
Wavenumber [cm-1]
1500 2500 3500
Wavenumber [cm-1]
0.12
B. Spectral Fit 2 (CCF=0.88)
Reconstructed Spectrum Measured Spectrum1
0.09
5 0.06
c
o
£>
§ S E
Ĥ= o
c
o
o
D. Reconstructed Size Distribution
0.3
0.1
Fit 1 Fit 2 Fit 3
i
\
1
,4
A
500 1500 2500 3500
Wavenumber [cm-1]
4500
5 10 15
Size Distribution [p.m]
20:
Figure 4. Summary of the inversion results for three spectra presented in Figure 2 A-C.
Conclusions
The inverted extinction functions and measured baselines for IR spectra in the presence of water
aerosol and dust particles matched quite well. Some mismatch or error is present in the calculated
spectra, but overall there is substantial agreement with the baseline features of the measured
spectra. The errors are probably due to error in the complex refractive index functions, especially
for the dust samples. However, it was found in the numerical simulation that these errors have little
effect on the reconstructed size distribution.
Our observations indicate that the general complex refractive index function of the aerosol is the
key information needed to account for the broad absorbance features in the baseline of the OP-
FTIR spectrum in the presence of aerosols. With some knowledge of the complex refractive index
function for the specific aerosol material and the measured infrared spectrum, it is possible to
estimate the aerosol size distribution parameters in near real-time. This raises the interesting
possibility of using OP-FTIR to simultaneously measure both pollutant gases and aerosols.
The preliminary results of this study provide a physical explanation for the observed baseline
absorbance features in an open-path spectrum in the presence of aerosols. Application of this work
could broaden the scope of OP-FTIR applications to include real-time determination of path-
averaged fugitive dust concentrations and size distribution parameters.
Abstract No. 550 Session No. AO-1b
Page:
8/8
-------
References
1. Xiao, H.K., S.P. Levine, W. F. Hcrget, J. B. D'Arcy, R. C. Spear, and T. A. Pritchett. Am. Ind.
Hyg. Assoc. J. 1991, 52, 449.
2. Hashmonay, R.A., and M.G. Yost. "On the Application of OP-FTIR Spectroscopy to Measure
Aerosols: Observations of Water Droplets," Environmental Science & Technology, 33(7),
1141-1144, April 1999.
3. Kerker, M. "The scattering of light," Academic Press, New York, NY (1969).
4. Yamamoto, K., and A. Masui. "Complex Refractive Index Determination of Bulk Materials
from Infrared Reflection Spectra," Applied Spectroscopy, 49(5), 639-644 (1995).
5. Sokolnik, I., A. Andronova, and T.C. Johnson. "Complex Refractive Index of Atmospheric
Dust Aerosols," Atmospheric Environment, 27A(16), 2495-2502 (1993).
6. Volz, F.E. "Infrared Optical Constants of Aerosols at Some Locations," Appl. Opt., 22, 3690-
3700(1983).
7. Wickramasinghe, N.C. Light Scattering Functions for Small Particles with Applications in
Astronomy, Adam Hilger, London, U. K. 1973, 25-27.
8. Chahine, M. T. "Inverse Problems in Radiative Transfer: Determination of Atmospheric
Parameters," J. Atmos. Sci., 17, 960-967, 1970.
9. Hashmonay, R.A., M.G. Yost, and Chang-Fu Wu, "Computed Tomography of Air Pollutants
Using Radial Scanning Path-Integrated Optical Remote Sensing," Atmospheric Environment,
33(2), 267-274 (1999).
10. Volz, F.E. "Infrared Optical Constants of Ammonium Sulfate, Sahara Dust, Volcanic Pumice,
and Flyash," Applied Optics, 12(3), 564-568 (1973).
Abstract No. 550 Session No. AO-1b
Page:
9/9
-------
T\J T?l\/I "RT _ dtp- "P- TECHNICAL REPORT DATA
in ruvi xvi-' Al ~ -l o O o (Please read Instructions on (he reverse before completing)
1. REPORT NO. 2.
EPA/600/A-01/060
3. RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE
Particulate Matter Measurements Using Cpen-path
Fourier Transform Infrared Spectroscopy
5. REPORT DATE
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
R. A. Hashmonay (ARCADIS) and D. B. Harris (EPA)
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
ARCADIS Geraghty and Miller, Inc.
P. C. Box 13109
Research Triangle Park, North Carolina 27709
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
68-099-201
12. SPONSORING AGENCY NAME AND ADDRESS
EPA, Office of Research and Development
Air Pollution Prevention and Control Division
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
Published paper; 1/00-1/01
14. SPONSORING AGENCY CODE
EPA/600/13
15.supplementary notes ^PPCD project officer is D. Bruce Harris, Mail Drop 61, 919/
541-7807. Presented at AWMA Annual Meeting, Orlando, FL, 6/24-28/01.
I6" abstract Qpen_path Fourier transform infrared (OP-FTIR) spectroscopy is an ac-
cepted technology for measuring gaseous air contaminants. OP-FTIR absorbance
spectra acquired during changing aerosols conditions reveal related changes in very
broad baseline features. Usually, this shearing of a spectrum's baseline is viewed
as undesirable because it interferes with quantifying gases. However, the paper
shows that this wavelength-dependent absorbance can be used to measure particulate
matter (PM). It describes a developed inversion algorithm for retrieving the size
distribution of the PM in the beam path. It gives preliminary results from CP-FTIF
measurements conducted downwind from an active dirt road. Using the developed
inversion algorithm and measured optical properties, the calculated extinction spec-
tra were fitted to the measured extinction, using the Mie theory for spherical par-
ticles. Time series data reveal significant rapid shifts in size distribution of the PM
in the OP-FTIR. Results indicate that size distribution parameters may be retrie-
ved from CP-FTIR spectra acquired over an open optical path. The suggested me-
thod may provide real-time concentrations of gaseous and PM contaminants simul-
taneously. Emission flux estimates of PM may be generated for near-ground-level
(roads) and area sources when several beam paths are deployed downwind.
17. KEY WORDS AND DOCUMENT ANALYSIS
a. DESCRIPTORS
b. IDENTIF IERS/OPEN ENDED TERMS
c. COSATI Field/Group
Pollution Gases
Particles
Measurement
Infrared Spectroscopy
Fourier Analysis
Aerosols
Pollution Control
Stationary Sources
Particulate
13 B
14 G
14 B
12 A
07D
18. DISTRIBUTION STATEMENT
Release to Public
ia. SECURITY CLASS (This Report)
Unclassified
21. NO. OF PAGES
9
20. SECURITY CLASS (This pagej
Unclassified
22. PRICE
EPA Form 2220-1 (9-73)
------- |