FTIR Technology Development
EPA Contract 68D90055
Work Assignment 2.107
Thomas J. Geyer, Ph.D.
Grant M. Plummer, Ph.D.
Thomas A. Dunder, Ph.D.
Entropy Environmentalists, Inc.
P. O. Box 12291
Research Triangle Park, North Carolina 27709
Lori Tussey Lay
K. William Grimley
U. S. Environmental Protection Agency
Emission Measurement Branch
Research Triangle Park, North Carolina 27711
This document was prepared by Entropy Environmentalists, Inc. under Contract No. 68D90055, Work Assignment No. 2.107. This document has not been reviewed by the U. S. Environmental Protection Agency. The opinions, conclusions, and recommendations expressed herein are those of the authors, and do not necessarily represent those of the United States Environmental Protection Agency. Mention of specific trade names or products within this report does not constitute endorsement by the EPA or Entropy Environmentalists, Inc.
TABLE OF CONTENTS
Entropy Environmentalists Inc., under U.S. EPA contract, is developing techniques for quantitative identification of air toxic compounds in a sample stream extracted from an emissions source. The Clean Air Act amendments of 1990 have increased the number of hazardous air pollutants (HAP's) for which emission rates must be measured. This has prompted research activity into the development of Fourier transform infrared (FTIR) spectrometry as a measurement method. To that end Entropy has performed a number of tasks for the Emissions Measurement Branch (EMB).
FTIR has the advantage that a number of compounds can be detected simultaneously and unambiguously. The experimental difficulties are connected with accurately quantifying any compounds that are detected. In order to do this it, is necessary to posses an extensive library of reference spectra. This library contains multiple spectra for each of a number of compounds over a range of concentrations. Once a compound is detected in a sample stream, its concentration can be determined only if spectra for that same compound are included in the reference library.
During the past year, Entropy has obtained quantitative reference spectra for over 100 HAP compounds. It is the purpose of this report to describe the experimental procedures under which these spectra were collected. The reference library is stored on 3.5 inch diskettes included with this report.
Entropy has been developing a sample concentration system to achieve lower detection limits for some compounds. Laboratory experiments and field tests have been conducted and some preliminary results are presented below.
The spectral files presented here are for concentrations at 500, 100, and 20 ppm. The concentration-path length products covered a range from 45 to 5625 ppm meters. At least two spectra were collected at each concentration. These were required to meet the condition that maximum band intensities agree to within five percent. In some cases, additional spectra were taken until this condition could be satisfied. For some neat samples, the time required to fill the mixing chamber was relatively long so it was difficult to obtain reproducible band intensities at 20 ppm. For these, the lowest concentration has been omitted.
Over the course of the project, two different sampling systems were utilized. The first gas handling system is depicted in Figure 1. Details of this system are given in a draft report submitted to EMB in September 1991. Cylinder gases could be introduced directly into the absorption cell, with a fixed path of three meters, or to the mixing chamber for dilution with UPC grade nitrogen. The mixing chamber was maintained at approximately 50oC. Pressure measurements were made using MKS Baratronr capacitance barometers models 128 and 221. Ambient pressure was recorded twice daily. The mixing chamber, cell, and sample loop were filled above ambient pressure. The sample was then vented through the GC loop to lower the pressure to ambient levels. The GC provided a check on the purity of the sample. For comparison, concentrations were calculated using GC peak areas.
Many of the HAP compounds have low vapor pressure at 20oC. Therefore, a heated gas handling system was developed for obtaining reference spectra from neat samples. The following changes were made to the original configuration. A stainless steel manifold was used and there was no cold trap between the manifold and the pump. Typically, the absolute pressure was 50 to 60 Hg. The leak rate, checked before each sampling, was 4 to 5 Hg per minute. The mixing chamber was 22 liters. The absorption cell was a Wilkes/Foxboro model 9020 variable path cell. The teflonr transfer lines, mixing chamber, GC loop, manifold, and absorption cell were all maintained at a temperature of approximately 100oC. The cell path length was 2.25 meters or 11.25 meters for weakly absorbing compounds. The temperature control bath was eliminated. The path lengths were verified by comparing calculated areas for the ethylene band centered at 949 cm-1. The fixed path length cell was assumed to be three meters.
Ultra pure laboratory grade chemicals were purchased from Aldrich, Pfaultz and Bauer, or Kodak. After pumping away any dissolved gases, vapor from the liquid or solid compound was introduced to the manifold and mixing chamber from a pyrexr sample tube. The partial pressure of the sample varied from 20 to 510 Hg depending on the desired concentration. First, the leak rate was measured. Then a target total pressure was determined for a the sampling period. The maximum sampling time was about 5 to 15 minutes. Once an acceptable pressure was reached, UPC grade nitrogen was introduced through a pre-heated line. The mixing chamber was filled to 1025 mm Hg or higher to achieve the target concentration. The mixture was allowed to thoroughly combine for 15 minutes before introduction to the GC and the infrared cell.
The spectra were collected using an RFX-65 spectrophotometer capable of a maximum resolution of 0.12 cm-1 and equipped with an MCT liquid nitrogen cooled detector. The interferometer, detector and computer were leased from KVB/Analect, INC. The nominal resolution of the data presented here is 0.25 cm-1. Two hundred fifty scans were co-added and transformed into a single beam spectrum. This was divided point by point by a 500 scan background of the empty cell. The resulting transmission spectrum was converted to absorbance and saved along with the original interferograms. At the longest path length twice as many scans were used.
Table 1 presents the list of spectral files included on the diskettes. The diskettes are sequentially numbered and that number appears in the table. The compound name is given along with the file names and corresponding concentrations. Each file name begins with a three digit Entropy number followed by a letter identifying the cell. For some files water was spectrally subtracted. In these the third letter from the end of the name is an "s." All files are Analect absorbance spectra and have the ending .asf. The path length for each spectrum is given in meters. The spectra were run at room temperature or in a heated line system. The temperature was recorded each day and actually varied between 20 and 25oC for the cold system or between 100 and 105oC for the heated system. For simplicity, a value of 25 or 100oC is given.
Two different concentration values are reported for each spectrum. The accepted standard concentration (ASC) is the barometrically measured value. The indicated standard concentration (ISC) is the value calculated from a regression analysis of absorbance versus concentration. Two linear regressions of the FTIR band areas versus concentration (as determined by the barometric and GC concentration values) were performed. The R2 values for these regressions are typically in the 0.998 range. The fractional calibration uncertainty (FCU) represents the average fractional difference (calculated minus regression values divided by the regression values) between the ISC and the ASC. An FCU was obtained from the both the GC and FTIR band areas. Those using the FTIR band areas versus barometric pressure are given in the table.
Calculations were performed on up to six bands for each compound, but the FCU is reported for only one band. This was done to limit the size of the table. The band reported is often, but not always that which gives the lowest FCU. An attempt was made to choose a band free from spectral interferants, one that had maximum absorbance less than 1.0 for the highest concentration, and was also relatively uncomplicated. If these three criteria could be met for more than one band, then the one with the lowest FCU value was chosen. The mid point of the chosen band is given in cm-1. This frequency actually represents the center of the analytical region used in the calculation and may include a number of superimposed bands. In some cases only a portion of an absorption band was contained within the analytical region. This was done to avoid inclusion of an interferant band. An example is methyl chloride, whose absorption at 732 cm-1 lies near the CO2 bending mode.
All compounds for which quantitative data were obtained are included in the library. For some compounds only qualitative spectra were generated and no FCU values are reported. No attempt was made to obtain quantitative data for some common interferants, but spectra for these are included in the library. Interferant spectra will be useful for spectral subtraction. If possible, we obtained results from both a neat sample and a cylinder gas. This was done for allyl chloride, where also two different path lengths were used.
Hydrazine, methyl hydrazine, and 1,1-dimethyl hydrazine all exhibited impurity peaks in the GC that occurred due to decomposition in the heated system. This could be diminished by taking spectra immediately after dilution instead of waiting 15 minutes. This worked best for the dimethyl hydrazine.
Some spectra of the calibration transfer standard (CTS), 100 ppm of ethylene in nitrogen, are also included. Three spectra for each path length were randomly chosen. Within each group there is a variation of about 3 percent in the maximum intensity of the 949 cm-1 Q-branch. The variation in pressure for all of the CTS spectra was about 2 per cent. A complete analysis of the CTS spectra will include computing band areas that are corrected for pressure. These can then be compared over a given path length to determine the fractional reproducible uncertainty (FRU) as defined in the FTIR protocol.
Features due to residual water are apparent in most of the spectra. This normally had little effect on the FCU values. For improved appearance, residual water features were removed by spectral subtraction. This process was not completed for all of the compounds listed, but for some it was not necessary. The water spectra used for subtraction contained CO2 so there may be negative peaks near 2350 cm-1 and 667 cm-1. The quality of the subtraction can easily be improved, if necessary, since all of the original spectra are stored.
It should not be inferred that this library, as extensive as it may be, is complete. With design changes it should be possible to obtain spectra for some of the compounds that presented difficulties. Use of absorption cell and sampling handling materials capable of withstanding higher temperature could lead to the addition of compounds to the library.
Research toward extractive FTIR detection of pollutants in stack gas effluent has revealed several limitations to the technique. Firstly, even when path lengths up to tens of meters are used the detection limits only extend to the low parts per million (ppm) range for some compounds. This results from the magnitude of their infrared absorption cross sections. Secondly, the typically high concentration of water vapor and CO2 in gaseous effluent streams may mask the strongest absorbance bands of some species because water vapor gives a strong infrared spectrum. Removal of water by condensation or permeation treatments can be performed, but has the potential to affect the concentrations of other species. For these reasons,it is desirable to develop a technique that would improve detection limits while at the same time reducing sensitivity to water vapor and CO2 in the waste stream. One approach to this problem is sample concentration on solid phase adsorbents.
There are techniques currently available, such as the VOST and Semi-VOST methods, for sample concentration to enhance the detection of pollutants in gas streams. The VOST method, for example, concentrates pollutants on Tenax and charcoal adsorbents, followed by thermal desorption to evolve the collected species into a Gas Chromatograph/Mass Spectrometer (GC/MS) for detection and quantification. This requires shipping the sample to a specialized laboratory for evaluation, at considerable cost as well as delay before the results are known. There would be significant advantages in developing a similar technique that permitted the analysis to be performed at the test site so that same day results could be obtained.
A research program was initiated to determine the viability of sample concentration on solid phase adsorbents followed by thermal desorption and direct measurement using FTIR spectroscopy. The first phase of the research was a laboratory evaluation of the technique and of several adsorbents. Scale-up to field testing at a coal-fired electric utility was performed in the second phase.
The laboratory evaluation of thermal desorption FTIR (TD-FTIR) was initiated in June 1992 to determine the viability of direct FTIR detection of thermally desorbed species from solid phase adsorbents. To this end an EGA-500 (Evolved Gas Analysis) FTIR cell was obtained from Aabspec Instrumentation Ltd. of Ireland. This cell consists of two independently heated stages with separate temperature controllers. The adsorbent is packed into a stainless steel sample tube that fits into the first stage. The second stage is a heated light pipe of 1 millimeter diameter and 5 centimeter length. Windows are affixed to the light pipe with compression O-rings and Teflon spacers. Due to the narrow diameter of the light pipe, alignment within the FTIR sample compartment is critical to obtain sufficient throughput, but the alignment is very reproducible since the cell mounts onto a standard mount in the sample compartment.
The adsorbents were weighed into the sample tubes with glass wool plugs on either end. Two modified charcoal adsorbents (Ambersorb and Anasorb from SKC, Inc.) and one organic polymer adsorbent (Tenax GC from Alltech, Inc.) were evaluated. These adsorbents could be doped with various gases and desorbed without being removed from the cell. Certified gas mixtures (Scott Specialty Gas) were used to flow a measured volume through the adsorbent. Following deposition on the room temperature adsorbent, the cell was evacuated to remove residual gas, isolated from the pump, and the adsorbate was thermally desorbed into the light pipe where its spectrum was measured.
Three different materials - carbon tetrachloride, methylene chloride, and ethyl benzene - were deposited on the adsorbent at a typical loading of 1 milligram per gram of adsorbent. For all three species, the Tenax adsorbent gave a readily detectable infrared absorbance. Conversely, the charcoal-based adsorbents yielded no measurable infrared signal for any of these species. The magnitude of the infrared absorbance from the Tenax samples indicated that sample loadings as low as 1 microgram per gram of absorbent could be detected using this arrangement.
Though these experiments demonstrated that the approach was viable, it was obvious that the EGA-500 cell was impractical for field measurements. This stemmed from the limited amount of absorbent that could be packed into the sample cells (1-2 grams) and the restricted throughput and path length of the light pipe. It was recognized that the detection limit of this technique could be enhanced by the combination of a larger quantity of absorbent, a longer infrared path length, and flowing a larger volume of stack gas through the absorbent. Entropy designed a new thermal desorption apparatus that was interfaced with a long-path White cell. Absorbent tubes were fabricated out of stainless steel tubing and were able to hold ten grams of Tenax GC. These tubes could be heated to 250 C to desorb materials directly into the long path cell; a maximum temperature of 350 C could be achieved to purify the Tenax for re-use.
A variant of the VOST method for flowing stack gases through the absorbent was employed. The absorbent tube was immersed in an ice bath and connected to the heated sample line on one end and a dry gas meter on the other end. A water sparge tube and a silica tube protected the dry gas meter from the water in the gas stream. Up to 10 cubic feet of extracted stack gas was passed through the absorbent. Following deposition, dry nitrogen gas was passed through the ice temperature tube to carry off residual water.
This experimental set-up was tested at a coal-burning electric utility. The gas stream had a high water content, limiting the utility of direct extractive FTIR measurements. Diluted or conditioned (dried) gas samples did not reveal the presence of hazardous air pollutants. The absorbent tubes, by contrast, showed the presence of several compounds regulated by the Clean Air Act amendments at less than part per million concentrations. Ambient air samples and field blanks did not show evidence of these compounds.