-
MEASUREMENTS AND MODELING OF LWIR SPECTRAL EMISSIVITY OF
CONTAMINATED QUARTZ SAND
John Kerekes, Michael Gartley, Christopher De Angelis, Carl
Salvaggio
Digital Imaging and Remote Sensing Laboratory, Chester F.
Carlson Center for Imaging Science Rochester Institute of
Technology, Rochester, New York USA
Christopher Gittins1, Michael Costolo, Bogdan Cosofret Physical
Sciences, Inc., Andover, Massachusetts USA
1 C. Gittins is now with UTC Aerospace Systems in Westford,
Massachusetts.
ABSTRACT The fundamental understanding of effects of liquid
contaminants on the longwave infrared spectral emissivity of
surfaces contaminated is desirable. This research describes
modeling and longwave infrared spectral emissivity measurements for
samples of SiO2 (sand) with and without 0.3% (by weight) of SF96
(poly dimethyl siloxane) oil. Two different sand particle size
ranges were considered. The modeling was performed using a
micro-scattering code and the empirical emissivity measurements
were made outdoors using a D&P Instruments Model 102F MicroFTIR
non-imaging spectrometer. The data were calibrated and processed to
retrieve the spectral emissivity. General observations included a
significant increase in emissivity in the 8 to 9 and 12.5 to 13
micron regions due to the presence of the SF96. The comparison
between the modeled and measured emissivities shows a consistent
trend and significant separability between the spectral emissivity
of sand with and without the SF96 present.
Index Terms— infrared spectral emissivity, contaminated
surfaces, longwave hyperspectral imaging
1. INTRODUCTION The fundamental understanding of the effects of
liquid contaminants on the longwave infrared (LWIR) spectral
emissivity of surfaces contaminated is desirable. Measurements of
the LWIR emissivity of materials have most often been done in the
pristine environment of a laboratory. While there have been efforts
at field measurements of materials contaminated with different
liquids and solids (e.g., [1]), these have just scratched the
surface of the range of situations of interest. This research was
undertaken to expand the understanding of how liquid contaminants
affect the LWIR emissivity of sand in particular.
The paper is organized as follows. First we describe the
materials considered and the sample preparation. Then we describe
the empirical emissivity measurements. This is followed by a
discussion of the modeling approach, and a comparison between the
modeling and measurement results. We conclude with a summary and a
discussion of future work.
2. SAMPLE PREPARATION Four types of samples were prepared for
measurement of their LWIR spectral emissivity. Figure 1 shows a
microphotograph of the two sizes of pristine samples.
• Plain (pristine) sand with particles ranging in size from 425
– 1000 microns
• Plain sand with particles ranging in size from 1000 – 1400
microns
• Sand with particles ranging in size from 425 – 1000 microns
mixed with 0.3% (by weight) of SF96 (poly dimethyl siloxane)
oil
• Sand with particles ranging in size from 1000 – 1400 microns
mixed with 0.3% (by weight) of SF96 (poly dimethyl siloxane)
oil
Figure 1. Microphotograph of plain (pristine) sand
particles.
-
3. EMPIRICAL MEASUREMENTS
The instrument used by RIT for the measurements was a D&P
Instruments Model 102F MicroFTIR. While the instrument collects
data from 2 to 16 microns, the data in the midwave infrared (MWIR)
are typically of low quality in a passive collection mode, and for
this experiment, the emphasis was on the longwave infrared (7 to 13
microns).
The samples were poured into low profile cardboard containers
and formed into a thin layer (1/8” to 1/4” deep). These sample
containers were then placed on a small hot plate to raise their
temperature to approximately 308 K and enhance the contrast with
the atmospheric background. A total of eight emissivity
measurements were made comprised of the four sample mixtures
collected at nadir and 45º off-nadir viewing angles. The instrument
aperture was approximately 26” above the samples during the nadir
measurements. Given the 4.8º field of view of the instrument, this
resulted in a spot size of approximately 2” at nadir, which
expanded to 3” for the 45º off-nadir configuration, well within the
extent of the sand in the cardboard containers. Figure 2 shows the
measurement configuration on the roof of the Chester F. Carlson
Center for Imaging Science building.
The processing of the data proceeded as follows. First the raw
spectra were calibrated to spectral radiance using the blackbody
measurements. Then the emissivity of the sample εsamp was derived
using the following equation.
€
εsamp =Lmeas − Ldwr
LBB (Tsamp ) − Ldwr
Figure 2. Nadir measurement configuration showing samples and
the D&P instrument.
Here, Lmeas is the measured spectral radiance for the sample,
Ldwr is the measured reflected downwelling spectral radiance, and
LBB(Tsamp) is the blackbody radiance for the retrieved temperature
of the sample. The retrieved temperature was found by stepping
through a probable range of sample temperatures and finding the
temperature that maximized the smoothness of the retrieved
emissivity over a small spectral range. For these results the
spectral range 8.2 to 8.5 microns was used. The temperature
emissivity separation (TES) algorithm used in this processing was
developed from references [2] and [3]. The calibration and TES
processing was performed using in-house IDL codes.
Figure 3 shows the spectral emissivity curves measured in the
nadir-viewing configuration for the different size particles with
and without the SF96 present. Results for the 45º off-nadir
configuration were very similar.
Figure 3. Spectral emissivity measurements using the nadir
configuration for the small (a) and large (b) particles.
4. MODELING APPROACH The RIT code used to predict the
emissivities is a version of the first-principles physics-based ray
tracing image simulation tool DIRSIG [4]. This version is known as
microDIRSIG. It uses microscale geometric structure and material
scattering characteristics to predict spectral bidirectional
reflectance distribution functions (BRDFs).
A variety of sand surface geometries were constructed and the
surfaces were attributed with a spectral complex index of
refraction taken from the literature [5]. It should be noted that
the spectral shape of the complex index of refraction varies
between authors due to silica material property variability
(optical clarity, grain sizes, etc.) and the one assumed for this
modeling may not precisely reflect that for the actual sand used in
the empirical data collection.
Similarly, the spectral complex index of refraction for the
single contaminant considered, SF96, was also leveraged from the
literature [6] and utilized as an input to the microDIRSIG model.
For the SF96 dosed modeling runs, we calculated the average surface
film thickness required on each silica particle to achieve the
target dosing level of 0.3% by weight. We determined a SF96 film
thickness of
-
0.96 microns for the smaller grain size (750 microns) and 1.65
microns for the large grain size (1,200 microns) sand to cover the
surface of each silica sphere and achieve the target total dosing
of 0.3% (by weight).
Each microDIRSIG modeling run was conducted for a single
wavelength from 7 to 13 microns in 50 nm increments for a total of
121 individual modeling runs per (1) view geometry, (2) silica
particle size, (3) contaminant present or not present. The silica
particle sizes chosen were a single value for each sand type for
the modeling. The resulting output was a total of eight spectral
cubes of bi-directional hemispherical reflectance maps.
Hemispherical integration of the eight output spectral cubes
permitted calculation of the total directional hemispherical
reflectance (DHR) and also the directional hemispherical emissivity
(DHE).
The nature of the current RIT modeling capability permits only
Geometric Optics (GO) regime radiative transfer, leaving out
diffraction effects when the particle size approaches the
wavelength of light of interest. However, these modeling runs
utilized particles that were two orders of magnitude above the
wavelengths of interest keeping them safely in the GO regime.
5. MEASUREMENT VS. MODELING COMPARISON
This section presents plots to compare the empirical
measurements with the model predictions. Figures 4 and 5 compare
the spectral emissivity of the sand with and without SF96 for the
two particle size ranges investigated. While there are differences
between the measurements and modeling, a consistently clear trend
is observed in that the addition of the SF96 to the sand leads to a
higher emissivity in the Restrahlen band between 8 and 9 microns,
and near 12.5 microns.
The model predictions show a more subtle difference for the two
particle sizes than observed in the empirical data. This may be due
to the assumed spherical shape for the modeled sand particles as
compared to the irregular shapes observed in the real sand, as well
as potential small errors in the emissivity data collection and
processing. Differences in the overall spectral shape are most
likely from the fact that the assumed silica index of refraction
used in the model did not precisely reflect that for the sand used
in the empirical measurements.
Figure 4. Comparison of the empirical (a) and model predicted
(b) spectral emissivity for with and without SF95 for the small
particles.
Figure 5. Comparison of the empirical (a) and model predicted
(b) spectral emissivity for with and without SF95 for the large
particles.
6. SUMMARY AND FUTURE WORK
This project has demonstrated the collection of spectral
longwave infrared emissivity measurements of sand contaminated with
liquid SF96 oil and generating comparable model predictions of
emissivity spectra using a physics-based radiative transfer code.
The measurements were challenging due to the heterogeneity of the
surface and the difficulties of measuring accurate skin temperature
for use in the temperature-emissivity separation algorithm.
The microDIRSIG model was seen to be capable of simulating the
appropriate relative trend in the spectral emissivity of sand when
contaminated with SF96 (an increase in the emissivity in the
Restrahlen band between 8 and 9 microns, and another feature near
12.5 microns). However, the exact prediction is complicated due to
the random geometric shape of real sand particles, which is
challenging to reproduce in the simulated world, and the lack of
perfect knowledge of the index of refraction for the real sand
material.
-
The observed differences in the longwave emissivity spectra of
sand with and without SF96 were of a significant level (up to 40%
change in emissivity), which suggests the presence of SF96 in
amounts considered here should be easily detectable by remote
spectral observations.
While this project demonstrated LWIR emissivity measurements of
contaminated sand, it was for a single material and contaminant
type. Additional research is recommended to expand the types of
substrates and contaminants and to collect data in more realistic
situations such as a dirt road or parking lot.
7. ACKNOWLEDGMENTS This material is based upon work supported by
U.S. Army Edgewood Contracting Division and the U.S. Army Edgewood
Chemical Biological Center (ECBC) under Contract Number
W911SR-12-C-0004. Any opinions, findings and conclusions or
recommendations expressed in this material are those of the authors
and do not necessarily reflect the views of U.S. Army. Dr. James
Jensen and Dr. Janet Jensen of ECBC are gratefully acknowledged for
their support of the project.
8. REFERENCES [1] J. Kerekes, K.-E. Strackerjan, and C.
Salvaggio, "Spectral Reflectance and Emissivity of Man-made
Surfaces Contaminated with Environmental Effects," Optical
Engineering, vol. 47, no. 10, October 2008.
[2] N. Bower, Knuteson, R., and Revercomb, H., “High spectral
resolution land surface temperature and emissivity measurement in
the thermal infrared. Proceedings of 10th Conference on Atmospheric
Radiation: A Symposium with tributes to the works of Verner E.
Suomi, Madison, WI, 28 June-2 July 1999, pp. 528-531, American
Meteorological Society, 1999. [3] K. Horton, J. Johnson, and P.
Lucey, “Infrared measurements of pristine and disturbed soils 2.
Environmental effects and field data reduction,” Remote Sensing
Environment, vol. 64, pp. 47-52, 1998. [4] DIRSIG - The Digital
Imaging and Remote Sensing Image Generation Model,
http://www.dirsig.org. [5] R. Kitamura, L. Pilon, M. Jonasz,
“Optical constants of silica glass from extreme ultraviolet to far
infrared at near room temperature,” Applied Optics, vol. 46, no.
33, pp. 8118- 8133, 2007. [6] M. Querry, “Optical constants of
minerals and other materials from the millimeter to the
ultraviolet,” Chemical Research, Development and Engineering
Center, University of Missouri-Kansas City, November 1987.