FTIR Monitoring in High Water and Carbon Dioxide Environments

Document Sample
FTIR Monitoring in High Water and Carbon Dioxide Environments Powered By Docstoc
					FTIR Monitoring in High Water and Carbon Dioxide Environments
                                    Robert L. Spellicy
                           Industrial Monitor & Control Crop.
                                   Round Rock, Texas

                                     Scott Swiggard
                               Golden Specialty Consulting
                                    Deer Park, Texas

There is a widespread opinion that FTIR is not applicable to monitoring in sources with
high H2O and/or CO2 content. In this paper we present a measurement series in two
sources one with >95% water and one with >70% CO2. In these sources the compounds
of concern included: water, Carbon monoxide, Carbon Dioxide, methane, Formaldehyde,
acetaldehyde, propane, propylene, ethylene, ethanol, and heavy hydrocarbons. Some
special sampling handling was required to avoid condensation and with this sample
handling all compounds were successfully detected at low ppm levels.


There was a recent requirement to monitor a series of compounds (water, Carbon
monoxide, carbon dioxide, methane, formaldehyde, acetaldehyde, propane, propylene,
ethylene, ethanol, and heavy hydrocarbons) in an industrial de-aerator stack and a CO2
vent. Because of the very high water levels (>95%) in the first source and the high CO2
levels in the second, no monitoring method was ideal. It was decided that FTIR had as
good a chance as other methods, so an extractive FTIR was used. Typically, FTIR has a
difficult time monitoring in these environments because the absorption of the water and
CO2 obscures the bands of the compounds of interest. However, the concentrations of
interest in this case were in the low ppm-range so it was felt that FTIR could successfully
monitor them provided careful analysis procedures were developed. The second
challenge to the program was sample handling. In the de-aerator stack water
concentrations were >95% so liquid water was formed as soon as the stream left the stack
or cooled at all. As a result, not only heated lines were necessary but low pressure
sampling was required.

The Measurement System

The FTIR used was an Imacc extractive system consisting of a Base Unit with a high
resolution (0.125 cm-1) interferometer coupled to a multi-pass heated-cell accessory. The
Base Unit, shown in Figure 1, contains the interferometer, the power supply, the laser
control systems, and the on-board-computer (OBC) which controls the FTIR itself. The
FTIR has a “dash-pot” moving mirror which is essentially a graphite piston moving in a
precision glass tube. This provides great stability of the instrument because the mirror
has only one degree of freedom, that being along the desired direction of travel.
However, to correct for any possible variation in the mirror motion caused by vibration or
other factors, the fixed mirror is dynamic aligned using the HeNe laser to keep the two
mirrors in “perfect” alignment throughout the mirror motion.

                 Figure 1 Imacc FTIR Base Unit with IR source,
                 interferometer, power supply, laser control and FTIR

The cell accessory, shown in Figure 2, contains the multi-pass cell, the infrared detector,
the cell heater controller, and the cell pressure monitor. In the current test an un-cooled
DTGS detector was used because the sensitivity of the Liquid nitrogen detector was not
necessary. The cell in the accessory can be either a 1 m to 10m or a 3 m to 30 m path
adjustable cell with an internal cell volume of 2.3 liters. In this case a 10 m cell was used
and it was held at 185 C.

The gas handling system is shown in Figure 3. A heated probe extracted the sample from
the duct passed it through a heated line, held at 185 C, and passed a portion to the FTIR
for analysis. Two pumps were used, one providing continuous bypass flow and one
extracting a portion of the overall flow to the FTIR cell. These were to provide sufficient
flow to minimize line losses. Both pumps were also at the output of the line and cell so
the sample did not contact the pumps until after analysis. Calibration gases could also be
injected directly into the FTIR or to the probe. A mass flow controller was used to meter
the gases into the line and a rotometer was used to monitor total flow through the FTIR
cell. This allows for spike ratios to be determined.
              Figure 2 The Imacc 10 m multi-pass accessory with long
              path cell, temperature controller, pressure monitor, and IR

                                                                                       Ball Valve
                               Spike System Manifold

                                                                                                     Mass Flow
                                                                                                      0 - 4 lpm
                                  Analyte 4

                                              Analyte 3

                                                          Analyte 2

                                                                      Analyte 1
                                                                      Analyte 1


                                                                                              0 - 20 lpm

                                              IMACC FTIR
              System                                                              Heated
              Computer                                                            FTIR cell
                                  10.3 meter path length
              750 Mhz
              20 gigHD                  DTGS Detector, Kbr
                                 Beamsplitter, IR Source

Exhaust to

Figure 3 The Gas Handling system used with the FTIR. A heated probe was used along with
two pumps to maintain maximum flow through the long extraction line.
Analysis Procedures

Figure 4 shows the full infrared spectrum of water, CO2, and the analytes of primary
interest. Two regions are in general available for analysis one around 3000 cm-1 and one
around 1000 cm-1. Because of the encroachment of water on the 3000 cm-1 region only
formaldehyde and acetaldehyde were analyzed in this region while ammonia, methanol
and ethanol were monitored in the 1000 cm-1 region.

 Figure 4 Background and primary analyte spectra. Top 71% H2O 10% CO2, second panel
 downward references for: formaldehyde , acetaldehyde, ammonia, methanol, and ethanol.

The full CLS-analysis1 matrix for all compounds is shown in Figure 5. This matrix
shows the compounds along the top. An “S” in a column of the matrix indicates that the
compound above that column is analyzed in the “region” shown at the left of the
particular row. An “I” means the compound is an interferant in that region. For example,
the fourth and fifth columns are for formaldehyde and acetaldehyde. These are analyzed
in the region from 2550 cm-1 to 2795 cm-1. In this region the possible interferants are:
water, carbon dioxide, methane, ethylene, propane, propylene, and methanol. These are
shown as “I” in the same row. The software treats both analytes and interferants on an
equal footing in each analysis region but only reports the values for the analytes. Some
very similar regions or even overlapping regions are seen in the matrix. This is because
the software also allows for use of “windows” within any given region to minimize
interference and bias. As an example, Figure 6 shows the region from 867.6 cm-1 to
1057.1 cm-1. On the right is shown the windows used in this region. This in effect
produces a piecewise continuous region that can “jump over” areas with severe
interference or little information. Addition of windows to the analysis regions provides
for substantially reduced errors (spectral residuals) in the data.

    Figure 5 Analysis matrix for the compounds listed along the top of the

             Figure 6 The analysis region from 867.6 cm-1 to 1057.1
             cm-1 showing the windows defined for it.
Collection Procedures and Representative Data

The gas sampling system shown in Figure 3 above was used to continuously flow stack
sample through the FTIR cell. Initially very erratic results were observed in the FTIR
particularly with the high water content source. After some experimentation, it was
found that when running the FTIR cell at a pressure 100 Torr or more below atmospheric
the data became much more stable. This implied that there was still condensation
problems in the flow system because the lower pressure would lower the boiling point
and consequently reduce the possibility of condensation. Therefore, for all data
collection the flow system was run with two pumps one providing a bypass of the FTIR
and one pulling a slip stream through the FTIR cell. Under these conditions, a higher
flow and lower pressure could be maintained throughout.

Figures 7 shows spectra collected on both the CO2 vent and the high water content de-
aerator vent. It is clear that the de-aerator spectrum with nearly 98% water has severe
interference in the 3000 cm-1 region and the 1000 cm-1 region making analysis much
more difficult. The CO2 vent with 60% to 70% CO2 produces some augmented
absorption near 1000 cm-1 but overall its effect is much less troublesome. Use of
measured high H2O and CO2 spectra in the analysis method as well as windows in the
analysis regions to avoid strongly interfering lines substantially reduced error and bias in
the analysis results.

Figure 7 Comparison of high CO2 (top) and high water (bottom) spectra from the CO2 vent
and de-aerator stacks.
Typical data from the test series is shown in Figures 8 and 9 for the two sources.
Compounds seen regularly in the CO2 vent included: water, carbon dioxide, carbon
monoxide, methane, formaldehyde, propylene, methanol and ethanol. In the de-aerator
stack, the composition was similar but here ammonia was also seen at significant levels.

                          H2O    = 94 +/- 0.04 %
                          CO2    = 5.2 +/- 1%
                          CH4    = 290.7 +/- 28 ppm
                          H2CO = 2.1 +/- 0.9 ppm
                          CO     = 94.3 +/- 9 ppm
                          NH3    = 493.5 +/- 10 ppm
                          CH3OH = 689.2 +/- 11ppm

       Figure 9 Example of observed compounds in the de-aerator vent and
       their concentrations

To estimate minimum detectable levels, three approaches were used: 1) Noise limited
detection, 2) Detection levels defined by the spectral residuals, and 3) The standard error
defined by the CLS procedures. The first two were computed as recommended in EPA
Method 3202 with Noise limited detection given by:

                                 RMS noise * [ν u − ν l ]
                       MAU =                                                (1)
                                      Ci * L i

where RMSnoise is the measured absorbance noise in the analysis region, νu and νl the
wavenumber limits of the region, Ai the area of the analyte reference spectrum in the
region, Ci the concentration of the reference spectrum, and Li the path length of this
spectrum. This, in essence, ratios the “area” of the noise to the area of the reference
spectrum normalizing to unit ppm*m. Dividing this by the path length of the cell will
give the noise limited concentration in ppm. The second approach is similar but instead
of using the noise spectrum of the instrument one uses the residual spectrum of the CLS
analysis. In this case the corresponding ppm*m detection limit is:
                                 RMS resid * (ν u − ν l )
                      FMU * =                                              (2)
                                      (C i * L i )

All terms here as the same as in equation (1) except the RMS is now evaluated on the
residual spectrum not the noise spectrum. Because both of these equations use the full
analysis region (νu - νl ), these may over predict the minimum detection for CLS routines
(like the Imacc routines3 used here) that use sub-windows in analysis regions and that do
dynamic selection of references to minimize residual error.

Table 2 below shows a comparison of the three estimated detection limits. Because an
un-cooled DTGS detector was used, the noise limited values are roughly ten-times what
one would expect with a cooled HgCdTe detector. As is usually the case, the noise
limited values are the smallest. These limits can be reached in practice only if
interference is not significant so the instrument noise becomes the limiting factor. The
residual MDC value and that given by the CLS standard error should be comparable.
However, the residual error may be larger due to the regions and windows issues
discussed above.

                                      Table 2
                Average Detection Limits (ppm) Computed from Noise,
                     Spectral Residual, and CLS standard errors
             Compound                       MDC                       CLS Low
                                 Noise Limited  Residual                2*σ
       Acetaldehyde                   2.4         4.9                   4-6
       Formaldehyde                   3.8         7.9                    1.0
       Ammonia                        2.4         16.3                  3 - 10
       Propylene                      2.9         19.7                   8.0
       Ethylene                       4.6         31.0                  18.0

       Carbon Dioxide                3405.8           22988.8          3400.0
       Methanol                        1.6              10.5           5 - 10
       Ethanol                         1.4               9.6            7 - 10
       Carbon Monoxide                 1.8              3.8             5-9
       Water                         55918.4          114900.9         ~8000
       Propane                         0.6               1.3            2-5
       C5+                             0.4              0.8             1-3
       Methane                         4.1               8.3            8 -10

  * This equation differs slightly from that in Method 320; M320 measures a relative
  uncertainty dividing by the gas concentration, here this term is not included.
The one area that was unsuccessful in the tests was the dynamic spiking. During all
spiking tests, the FTIR reported large fluctuations in the spike + background
concentration. Because of the observed sensitivity of the sampling system to pressure
and temperature, this was believed to be a result of increased pressure (and possibly
reduced temperature) at the point of calibration gas injection. In subsequent tests, the cal
gas will be pre-heated and a different flow system will be used to better equalize
pressures before injection.


Preliminary tests have shown that many organics can be detected to the single digit ppm
range in sources with very high water or carbon dioxide concentrations. To accomplish
this a sampling system allowing for high temperature and low pressure flow was needed
and analysis routines had to be used which provided for piecewise selection of the
analysis regions to avoid strong interfering lines.


   1. D.M. Haaland and R.G. Easterling, Appl. Spectrosc., 34, 539 (1980)

   2. US Code of Federal Regulations 40 CFR Part 63, Appendix A, Test Method 320
      “Measurement of vapor phase organic and inorganic emissions by extractive
      Fourier Transform Infrared (FTIR) spectroscopy”, US Government Printing
      Office , Public Documents Distribution Center, Laurel, MD, USA.

   3. “Challenges to CLS Analysis In The Real World”, Robert L. Spellicy, Industrial
      Monitor & Control Corp., Austin, Texas, Presented at the National Air & Waste
      Management Association (AWMA) conference Salt Lake City, Utah June 18-22,