|OFR 01-0429: Laboratory Reflectance Spectroscopy (RS) Studies of WTC Samples|
|About USGS / Science Topics / Maps, Products & Publications / Education / FAQ|
Spectroscopy is a tool that detects chemical bonds in molecules (solid, liquid or gas) through absorption (or emission) features in the spectrum of the material. While crystallinity affects the shape of spectral features, non-crystalline materials (such as glass or plastic) still display absorption features in their spectra. Results of the spectral analysis of the WTC samples is given in Results Table 1 (in the Integration of Results section, below). In general, ultraviolet to near-infrared reflectance spectroscopy used for this analysis (0.35 to 2.5 microns) has different sensitivities to materials than XRD. Reflectance spectroscopy in this spectral range is particularly sensitive to hydroxyl and water-bearing materials (OH and H2O), organic compounds (materials containing C-H), carbonates (e.g. found in marble), water-bearing sulfates (like gypsum used in wallboard), and iron-containing compounds (like hematite: iron rust used to color bricks red).
Laboratory spectroscopy differs from airborne imaging spectroscopy in 3 respects: 1) Single spots are analyzed, not images, 2) There is little interfering atmosphere to block portions of the spectrum, and 3) Signal to noise in the laboratory is much higher. Consequently, laboratory analyses have much higher sensitivity to the individual materials in a sample.
To see plots of spectra for each sample, and to get the digital spectral data for each sample (ascii list), CLICK HERE.
Reflectance spectra of 33 dust samples collected on Sept. 17, 18, and 19th from areas within a 1 km radius circle of the WTC collapse site (RS Fig. 1) were measured in a laboratory High Efficiency Particulate Air Filter (HEPA) fumehood with an Analytical Spectral Devices Full Range Spectrometer over the range from 0.35 - 2.5 microns using a halogen lamp for illumination and spectralon panel for reference. (Use of trade names is for descriptive purposes only and does not constitute endorsement but the USGS.) The entire unsplit sample was first poured from the plastic sample bag onto a clean sheet of paper, then the sample was mixed with a spatula leaving a relatively flat broad pile up to a few centimeters thick for spectral measurement. Mixing allowed us to avoid possible inadvertent effects of particle sorting that may have occurred during transport. Given that reflectance spectroscopy does not detect materials any deeper than can be seen with the eye, we tried to compensate by first measuring ten spectra of the pile and then remixing the pile before collecting an additional ten spectra. Our intent was to expose previously unmeasured material at the surface for the last set of measurements. The spectrometer optical fiber was held a few centimeters above the pile and moved constantly in an elliptical manner to spatially average the surface of all but the edges of the pile. This method allowed us to spectrally characterize about two thirds of the entire sample. Samples were then split by quartering and portions sent out for various other analyses. We used one quarter of the sample as a research split supplementing evenly from the other three quarters as needed to build the bulk of the split. After separating the research split any remaining sample was then recombined and set aside as a chain of custody archive sample.
Spectra of each dust sample were averaged and corrected to absolute reflectance using a National Institute for Standards and Technology traceable spectral correction. Overall spectra of the dust samples look nearly identical with reflectance levels that vary between 20 - 45 % with a strong absorption edge between 0.35 - 0.8 microns (sharp downturn in RS Fig. 2). There are only weak spectral absorptions, apart from the absorption edge, in the electronic region of the spectra (0.35 - 1.35 microns) indicating that the dust contains only low abundances of materials with Fe absorptions or absorptions from other transition elements. However, there are numerous spectral absorptions in the vibrational spectral region from 1.35 - 2.5 microns principally at 1.45, 1.75, 1.94, 1.97,and from 2.2 - 2.4 microns. The strongest spectral feature is at 1.94 microns and is due to structural and adsorbed water, with the next weaker feature at 1.45 microns (due to water and/or OH) which varies in depth from 2 - 6 % and up to 13 % in WTC01-17. Spectral features at 1.75 microns are next weakest with depths of only a few %. Spectral features in the 2 micron region are weakest varying from a fraction of a percent up to 1 % depth.
Absorption features can be diagnostic of the materials contained in the dust. Gypsum, probably from crushed wallboard, has the strongest spectral features accounting for most of the bands observed in the vibrational portion of the spectra. This mineral has three diagnostic bands that form a stair-step like triplet between 1.42 - 1.54 microns recognizable in nearly all dust samples. It also has bands at 1.75, 1.95, 1.975, 2.17, 2.217, and 2.268 microns closely matching the positions and geometry of the bands observed in the dust. In addition to those absorptions attributable to gypsum, an absorption at 1.413 microns may be due to portlandite or muscovite, both mineral components of the WTC concrete (see XRD results in Table 1 for samples WTC01- 37A&B). Spectral evidence for muscovite is seen in the 2 micron region where the overall shape of the absorptions match gypsum but the expected band position of gypsum at 2.217 microns is shifted to 2.209 microns as would be expected in a sample containing both gypsum and muscovite. Additional absorptions located at 2.307 and 2.343 microns are due to C-H stretches from organic materials.
SEM energy dispersive analysis indicates fibrous glass with elemental composition closely matched by slag wool in all dust samples analyzed by this method. Slag wool has only weak spectral absorptions and is virtually spectrally transparent over the 0.35- 2.5 micron spectral range. However, other materials associated with slag wool may contribute to some of the spectral features observed in the dust. An asbestos-free coating (WTC01-9) taken from a steel girder from the debris pile west of WTC tower 1 has relatively strong 2.307 and 2.343 micron absorptions that match the positions of similar but weaker bands in all of the dust samples. Although the positions of these bands are compatible with C-H absorptions in most organic materials (e.g. plastics, paper, fabric) washing of sample WTC01-9 with methanol was sufficient to significantly reduce the strength of these absorptions indicating they are caused by a soluble organic material, possibly an oil used as a dust suppressant in insulation containing slag wool. Spectra of both girder coating samples (WTC01-8 and WTC01-9) have an absorption at 1.725 microns, perhaps too weak to be seen in the dust samples, that closely matches the position of a C-H absorption found in oil. Use of slag wool, a type of mineral wool insulation, was widespread in the WTC towers (Hyman Brown, personal communication) as part of fireproof coatings on steel girders and the undersides of floors, thermal insulation around glass windows, and possibly in ceiling tiles. The apparent widespread use of slag wool and its brittle nature may explain its presence in all of the dust samples as a volumetrically significant component.
Notably absent from the reflectance spectra of the dust samples is evidence for paper in the form of a cellulose absorption expected at 2.1 micron. Visual examination of the dust samples reveals numerous strips of paper. A possible explanation is that microscopic gypsum particles form an optically opaque coating on the paper and other coarser fragments thereby concealing them from spectral detection. When pieces of dust coated paper are cleaned and then re-measured their spectra have distinct 2.1 micron absorptions. Gypsum from pulverized wallboard, in the dust samples, has a strong wide absorption at 1.95 microns that may obscure the 2.1 micron cellulose absorption from paper.
There are several possible sources of asbestos in the collapsed buildings at the WTC. One source is the fireproof coatings sprayed onto steel girders and the undersides of floors. Asbestos was reportedly used in a fireproof coating in Tower 1 up to the 20th floor (Hyman Brown, personal communication). Sample WTC01-8 was taken from a coating on a steel girder removed from the debris pile west of WTC tower 1. Most notable in its spectrum are absorptions at 1.385 and 2.323 microns which match in position and overall shape those of chrysotile (RS Figure 3). Chrysotile is one of three serpentine minerals and the only one to form with an asbestiform crystal habit. XRD analysis and the depth of these spectral absorptions is consistent with chrysotile forming up to 20 wt% of this coating material. The position of the 2.323 micron absorption in WTC01-8 is consistent with that of chrysotile but the band width is somewhat wider possibly indicating the presence of overlapping 2.307 and 2.343 micron absorptions from dust suppressant oil.
Other potential asbestos sources are ceiling tiles and to a more limited extent, vinyl flooring (Hyman Brown, personal communication). Apparently the cement used to construct the WTC towers probably did not contain asbestos (Leslie Robertson, written communication). Spectroscopy and XRD analysis of our two concrete samples did not reveal any chrysotile asbestos. Vermiculite has the potential to contain amphibole asbestos, but its possible use in fireproof coatings at the WTC is unclear at this time. Vermiculite was not present in the one sample of an asbestos-free coating (WTC01-9) we collected at the site. Vermiculite was not detected in the dust samples during SEM, XRD, and spectral analysis. No amphibole asbestos was identified by these analytical methods either. However, a few flakes of a vermiculite-like mica were noted during visual examinations of the dust.
To better constrain the abundance of chrysotile asbestos in the dust, we constructed mixtures of dust and chrysotile. We chose WTC01-6 as representative of an asbestos-free dust sample based on XRD and spectral evaluations. We added known amounts of a well characterized, very finely ground, National Institute of Occupational Safety and Health (NIOSH) Chrysotile standard (CH-29) to splits of WTC01-6 dust to create mixtures with 0.25, 0.5, 1.0, 2.5 and 5.0 weight % chrysotile. These samples were then measured spectrally and characterized by XRD, to provide a set of spectra and diffractograms that could be directly compared to those of the dust samples. Chrysotile (RS Fig. 3) has, as mentioned earlier, two strong absorptions at 1.385 and 2.323 microns due to the 1st OH stretch overtone and an Mg-OH combination band respectively. Even though the 2.323 micron chrysotile absorption is the stronger of the two bands, its wavelength position is wedged between stronger C-H absorptions at 2.307 and 2.343 microns, and overlaps with a 2.317 micron absorption from concrete (e.g. muscovite?), all of which tend to mask the presence of chrysotile in the dust samples at levels below 1 wt% in laboratory spectra. The weaker 1.385 micron chrysotile absorption occurs in a relatively less cluttered spectral region on the short wavelength side of the 1st OH stretch overtone gypsum absorption at 1.446 microns, thus providing a means of estimating chrysotile contents at even lower concentrations.
RS Figure 4A shows spectra of dust + chrysotile mixtures with contina removed from the 1.4 micron region (Clark and Roush, 1984). The presence of chrysotile is indicated by a distinct absorption in the 5, 2.5, and 1 wt% mixtures and as an inflection in the 0.5, and 0.25 wt% mixtures. RS Figure 4B shows these same samples with a curved continuum removed from the region centered on the 1.385-micron chrysotile absorption. Non-chrysotile bearing dust samples have convex upward curves in the 1.385 micron region. Band depths of the 1.385 absorption do not increase linearly with increasing chrysotile abundance due to the effects of multiple scattering (e.g. Clark and Roush, 1984). Sensitivity to changes in concentration is lowest for mixtures with trace amounts of chrysotile, however, the lack of convexity at the wavelength corresponding to the 1.385 micron chrysotile band, is a reliable indicator of the presence of chrysotile for the WTC dust samples. Because the chrysotile absorptions are relatively weak, even for the 5 wt% mixture (about 0.25 % depth), diagnostic spectral measurements must be done with high signal-to-noise ratio.
RS Figure 5A shows examples of the detection of chrysotile in three of the WTC dust samples at the 0.25, 0.5, and 0.75 wt% level. Arrows mark the inflection points (or shoulders) of the 1.385 micron chrysotile absorption. Note how the inflection points move outward as chrysotile content increases. Sequentially overlaying all of the dust spectra on top of the asbestos-free WTC01-6 spectrum provided a means of detecting these inflection points and thereby estimating the amount of chrysotile present in each sample. Figure 5B shows the same spectra as in RS Figure 5A with a convex continuum (provided by WTC01-6) removed from the 1.385 micron region. This process reveals absorptions due to chrysotile even in the 0.25 and 0.5 wt% samples.
The lower limit of detection for chrysotile in the WTC dusts depends to a degree on the grain size and albedo (brightness) of the dust. RS Figure 6 shows a series of dust + chrysotile mixtures constructed using dust sample WTC01-15, a coarser gained sample than WTC01-6. The dotted curves are spectra of the mixtures after hand grinding to fine powders. Because diffuse reflectance is a non-linear process, decreasing the grain size of the dust results in more scattering causing a decrease in the strength of all the absorptions including those of chrysotile. The size of individual grains in the 33 WTC dust samples is quite variable ranging from centimeter-sized fragments to submicron diameter fibers. It is likely that the grain size of the hand ground mixtures is smaller than the average grain size of even the finest of the dust samples, therefore the hand ground samples may represent the worst case detection scenarios. Even though grinding does significantly decrease the depths of the chrysotile absorption, the lack of convexity over the 1.385 micron chrysotile band is still evident in spectra of the hand ground mixtures down to the 0.5 wt% level. As an empirical rule of thumb, a given estimate of chrysotile content decreases by about half the wt% value over this extreme grain size range.
Lowering the albedo of a mixture results in less scattering thereby decreasing the likelihood that a given photon will encounter a chrysotile fiber and be scattered to the sensor, resulting in a lower sensitivity to chrysotile content. The albedo of the dust samples varies from 0.2 to 0.4 in the 1.4 micron spectral region, with about 80 % of the samples varying between 0.29 and 0.35. The dust samples used for the mixtures have albedos around 0.32 in the 1.4 micron region. The effects of albedo on spectrally measuring chrysotile contents have not been quantitatively evaluated. However, because most of the dust samples have an albedo similar to that of the samples used to create the mixtures, errors in estimating chrysotile content will probably be smaller than those due to grain size variations. In samples brighter than the mixture samples, chrysotile contents may be systematically overestimated. The reverse is true for samples darker than the mixture samples.
Spectroscopy could not be used to estimate chrysotile content of the dust below the 1wt% level for 9 out of 33 samples because of interference from nearby spectral absorptions. RS Figure 7 shows a 1wt% chrysotile + dust mixture compared with dust sample WTC01-2 which contains a relatively high concentration of portlandite and/or muscovite. These minerals have strong OH overtone absorptions located at 1.41 - 1.42 microns, with shoulders that overlap and conceal the nearby chrysotile 1.385 micron band. None of the 9 samples contained more than 1wt% chrysotile as determined by spectroscopy. XRD did detect trace levels of chrysotile in 7 of these samples. The other known constituent of the dust samples that has a weak spectral absorption near the 1.385 micron chrysotile absorption is that of bassanite (CaSO4*1/2H20) at 1.391 microns. Potential interference from this bassanite absorption is negligible because in at least one of those samples where it was detected with XRD (WTC01-19), the spectral curve was convex upward in the 1.385 micron region showing neither the presence of chrysotile or bassanite.
Spectral and XRD estimates of chrysotile content were made for all 33 dust samples (RS Fig. 8). Circles represent dust samples from the surface that contain trace levels of chrysotile asbestos. Solid circles represent detection of chrysotile by both spectroscopy and XRD analysis. Circles with their upper halves filled represent detection of chrysotile by spectroscopy; circles with their lower halves filled represent detection of chrysotile by XRD analysis. The size of the circle represents the amount of chrysotile detected by spectroscopy. Circles with their right halves filled represent samples where spectral interference prevented an estimate of chrysotile below 1% but where XRD analysis did detect chrysotile. Triangles represent samples where spectral interference prevented an estimate of chrysotile content but where XRD analysis revealed no chrysotile at or above the 1 wt% level. Squares represent samples where neither spectroscopy or XRD analysis were able to detect chrysotile at or above the 0.25 wt% (spectroscopy) and 1.0 wt% (XRD) levels. Samples 20 and 36 were collected indoors and where not subjected to rain and wind during the September 14th thunderstorm. The large filled "I" represents the location of the chrysotile bearing coating collected from the steel girder; the smaller "I" represents the the asbestos-free coating collected from a different steel girder. If plotted in its true proportion, the "I" representing the chrysotile bearing coating (WTC01-8) would be so large that it would conceal information from the other samples on this diagram, therefore its size is not proportional to its estimated 10 - 20 wt% chrysotile content.
Chrysotile was detected in a little over two thirds of the dust samples. Neither spectroscopy or XRD analysis revealed a chrysotile concentration higher than about 0.75 wt% in any of the 33 dust samples. Chrysotile was detected in 24 of the dust samples (exclusive of the girder coatings and concrete samples) with spectroscopy detecting it in 15 of the samples and XRD analysis detecting it in 16 of the samples. Chrysotile was detected by both methods in 7 dust samples. One explanation for the apparent disagreement between methods is the heterogeneous nature of the dust samples. While XRD measurements were made on 1.5 gram splits of each dust sample, spectroscopic measurements were made on approximately two thirds of the entire sample (or between 50 - 300 grams of material). Initial XRD measurements on samples WTC01-14 and 28 showed no trace of chrysotile, while replicate XRD measurements on different splits of these samples revealed chrysotile in 28. The problems of detecting unevenly distributed chrysotile in the dust are likely to plague any analytical method.
From RS Figure 8 it is apparent that trace levels of chrysotile were distributed with the dust radially in west, north, and easterly directions perhaps at distances greater than 3/4 kilometer from ground zero. The lack of chrysotile at levels above the detection limits of both spectroscopy and XRD analysis in samples collected south of ground zero (except in WTC01-33) may indicate that chrysotile was not distributed uniformly during the collapse.
NEXT Section of Report: Scanning Electron Microscopy (SEM) and Energy Dispersive X-Ray Analysis (EDS)
Back to document Table of Contents
For more information, contact:
Dr. Gregg A. Swayze
Dr. Roger N. Clark
|AccessibilityFOIAPrivacyPolicies and Notices|