United States
Environmental Protection
Agency
<&ERA Research and
Development
EPA-600/R-93-088
May 1993
DEVELOPMENT OF SAMPLING
AND ANALYTICAL METHODS FOR THE
MEASUREMENT OF NITROUS OXIDE
FROM FOSSIL FUEL COMBUSTION SOURCES
Prepared for
Office of Policy, Planning and Evaluation
and
Office of Air and Radiation
Prepared by
Air and Energy Engineering Research
Laboratory
Research Triangle Park NC 27711
-------
EPA-600/R-33-088
May 1S93
DEVELOPMENT OF SAMPLING AND ANALYTICAL METHODS
FOR THE MEASUREMENT OF NITROUS OXIDE
FROM FOSSIL FUEL COMBUSTION SOURCES
PROJECT REPORT
Prepared by:
Jeffrey V. Ryan and Shawn A- Karris
Acurex Environmental Corporation
4915 Prospectus Drive
P.O. Box 13109
Research Triangle Park, NC 27709
EPA Contract No. 68-DO-0141
TDs 91-021, 92-066, 93-133
EPA Project Officer William P. Linak
Air and Energy Engineering Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
Prepared for
U.S. Environmental Protection Agency
Office of Research and Development
Washington, D.C. 20460
-------
EPA REVIEW NOTICE
This report has been reviewed by the U.S. Environmental Protection Agency, and
approved for publication. Approval does not signify that the contents necessarily
reflect the views and policy of the Agency, nor does mention of trade names or
commercial products constitute endorsement or recommendation for use.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
ABSTRACT
The combustion of fossil fuels is suspected to contribute to measured increases in ambient
concentrations of nitrous oxide (N,0). Accurate and reliable measurement techniques are needed to
assess the relative contribution of fossil fuel combustion N20 emissions to the increase in ambient
concentrations. The characterization of N,0 emissions from fossil fuel combustion sources has been
hindered by the lack of suitable and acceptable grab sampling and on-line monitoring methodologies.
Grab samples have been shown to be compromised by a sampling artifact where N;0 is actually
generated in the sample container in the presence of sulfur dioxide (SO2), nitrogen oxides (NO,), and
moisture. On-line monitoring techniques are limited and of those available, instrument costs are often
prohibitive, detection levels are often insufficient, and the techniques are often susceptible to
interferences present in combustion process effluents. The report documents the technical approach
and results achieved while developing a grab sampling method and an automated, on-line gas
chromatography method suitable to characterize NzO emissions from fossil fuel combustion sources.
The two methods developed were ultimately documented in the form of the U.S. Environmental
Protection Agency (EPA) Air and Energy Engineering Research Laboratory (AEERL) Recommended
Operating Procedures (ROPs).
ii
-------
ACKNOWLEDGEMENTS
The work described herein was performed by Acurex Environmental Corporation through
EPA/AEERL Contract No. 68-DO-0141. William P. Linak of AEERL's Combustion Research Branch,
was the EPA Project Officer. The authors would like to acknowledge the assistance of Mike Messner,
Constance Wall, and Shirley Wasson of the Research Triangle Institute as well as Judith Ford of
EPA/AEERL, for their quality assurance/quality control assistance during the development of the
sampling and analytical procedures. The authors would also like to acknowledge the contributions of
Richard Rinehart during development of the grab sampling procedure.
iii
-------
TABLE OF CONTENTS
Section Paee
ABSTRACT ii
ACKNOWLEDGEMENTS iii
LIST OF FIGURES vii
LIST OF TABLES viii
1.0 INTRODUCTION 1
1.1 BACKGROUND 1
1.2 AEERL RESEARCH EFFORTS 3
13 OBJECTIVES 4
2.0 DEVELOPMENT OF ANALYTICAL PROCEDURES 6
2.1 DEVELOPMENT OF INITIAL ANALYTICAL METHOD 6
22 INITIAL ON-LINE ANALYTICAL EFFORTS 8
2.3 ANALYTICAL METHOD IMPROVEMENT REQUIREMENTS 9
2.4 CONFIGURATION OF THE ANALYTICAL SYSTEM 10
2.4.1 Precolumn Selection 10
2.4.2 Description of tlie Backflush Method 11
25 ANALYTICAL METHOD PERFORMANCE 16
25.1 Method Quantitative Capabilities 16
2.5.2 On-line Monitoring Performance 21
2.6 ANALYTICAL METHOD SUMMARY 24
3.0 GRAB SAMPLE METHOD DEVELOPMENT 27
3.1 BACKGROUND 27
3.2 GRAB SAMPLING EQUIPMENT CONSIDERATIONS 29
3.4 EXPERIMENTAL APPARATUS 31
3.4.1 Flue Gas Simulation System (FGSS) 31
3.4.2 Flue Gas Generation System 32
3.4.2.1 Moisture Generator 32
3.4.2.2 Mixing Chamber 34
3.43 Sampling System 34
3.43.1 Bypass Loop 36
3.43.2 Sorbent/Sample Bomb Loop 36
3.4.33 Analyzers 36
(continued)
iv
-------
TABLE OF CONTENTS (continued)
Section Paee
3.4.4 Flue Gas Measurements 37
3.4.4.1 Continuous Emission Monitors 37
3.4.4.2 GC/ECD and N,0 Measurements 37
3.4.5 Sampling Procedure 38
3.5 INITIAL SORBENT TESTS 39
3.5.1 Introduction 39
3.5.2 Water Removal 40
3.53 FGSS Shakedown Tests 40
3.5.4 Sorbent Cartridge Design 41
3.5.5 Dispersion Tubes 42
3.5.6 S02 Color Indicator 46
3.5.7 Chemical Sorbent Screening 46
3.6 SOz SORBENT OPTIMIZATION 47
3.6.1 Introduction 47
3.6.2 Sand Particle Size 49
3.63 Sand-to-Reactant Ratio 50
3.6.4 Sand/Sorbent Preparation 54
3.6.5 Sorbent Volume 54
3.7 SAMPLE CONTAINER OPTIMIZATION 57
3.7.1 Introduction 57
3.7.2 Sample Bomb Preparation 57
3.73 Teflon Coated Sample Bombs 57
3.8 COMBUSTION SOURCE GRAB SAMPLE METHOD EVALUATIONS 59
3.8.1 Introduction S9
3.8.2 Moisture Removing Devices 59
3.83 Evaluation or Source Sampling Configuration 62
3.8.4 EPA's Innovative Furnace Reactor 65
3.9 WORST CASE SCENARIO TESTS 67
3.9.1 Introduction 67
3.9.2 Nominal Inlet Concentrations 71
3.93 Worst Case Conditions 71
3.9.4 EPA's Innovative Furnace Reactor Worst Case Conditions 76
3.10 FURTHER METHOD EVALUATION: SNCR TESTS 79
3.11 GRAB SAMPLING METHOD SUMMARY 81
(continued)
v
-------
TABLE OF CONTENTS (concluded)
Section Page
4.0 SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS 85
5.0 QUALITY CONTROL EVALUATION REPORT 87
6.0 REFERENCES 88
APPENDICES:
A: Non-continuous sampling and analysis of nitrous oxide from combustion sources
ROP No. 43 A-l
B: Standard operating procedure for determining nitrous oxide concentrations in combustion
flue gas B-l
C: Recommended operating procedure for analysis of nitrous oxide in combustion Que gases,
AEERL/ROP No. 45 C-l
-------
LIST OF FIGURES
Figure Page
2-1. Schematic diagram of 10-port valve system 12
2-2. Automated, on-line GC/ECD N20 monitoring system 17
2-3. ECD response to N20 calibration gases (linear regression) 19
2-4. ECD response to N20 calibration gases (linear regression of log transformed variables) ... 20
2-5. Automated N20 analysis results of span checks over course of a test day 23
2-6. Automated N20 analysis results of span checks over the course of a 2-week period 25
3-1. Flue gas generation system 33
3-2. Flue gas sampling system 35
3-3. S02 removal with 20:1 sand/Ca(OH)2 sorbent 43
3-4. SO, sorbent cartridge assembly 44
3-5. Effect of dispersion tubes on SOz removal efficiency 45
3-6. Comparison of candidate sorbents and S02 removal capability 48
3-7. Comparison of sand-to-rcactant ratios on N20 generation 53
3-8. Effect of sorbent preparation process on N20 generation 55
3-9. Effect of sorbent volume on N20 generation 56
3-10. Effect of Teflon coating on N,0 generation 58
3-11. Sample container schematic 60
3-12. Effect of moisture removal on S02 sorbent performance 61
3-13. Location of sampling system 63
3-14. Sorbent/sample container schematic 64
3-15. Comparison of N20 sample container generation with and without use of sorbents 66
3-16. Performance of sampling method on actual combustion source 68
3-17. Comparison of effects of sample location on N20 generation. 69
3-18. N20 generation in sample containers; repeat tests of nominal conditions 73
3-19. N->0 generation in the refrigeration condenser 74
3-20. Effect of addition of refrigeration condenser on N20 generation within the sample delivery
system 75
3-21. N20 generation under worst case conditions 78
-------
LIST OF TABLES
Table Page
2-1. Comparison of Relative Bias Using Differing Mathematical Approaches 21
3-1. Predetermined Limits 37
3-2. N20 Generation in Grab Samples While Optimizing SO-> Sorbents 52
3-3. Repeat of Nominal Conditions 72
3-4. N20 Generation Under Worst Case Conditions 77
3-5. Evaluation of Sampling Method on Actual Combustion Facility Under Worst Case
Conditions 80
3-6. N20 Generation in Samples Collected During NOx Control Tests 82
-------
SECnON 1
TNlRODUCnON
1.1 BACKGROUND
Nitrous oxide (N,0) has been of concern to the combustion community largely because fossil
fuel combustion has been proposed as a potential contributor to the measured increases in ambient
N-,0 concentrations.1"2*3 Cunently, atmospheric N20 concentrations are increasing at nearly 1 ppbv
annually from a present level of 303 ppbv.2,3,4 This increase is of concern because N,0 is considered
a "greenhouse" gas owing to its infrared (IR) radiation absorptive properties as well as a contributor to
stratospheric ozone depletion.5 To further substantiate the supposition that increases in atmospheric
N20 concentrations are associated with the combustion of fossil fuels, studies tracking atmospheric
increases of carbon dioxide (CO-,) over time reveal that the increase of N20 and C02 occur similarly.6
The increase of both anthropogenic pollutants correlate well with increases in industrial activity.
Early efforts to characterize N^O emissions from fossil fuel combustion sources focused on
identifying a relationship between nitrogen oxides (NO^) and N,0 emissions. Data were nominally
collected in a "piggy back" manner, where N20 grab samples were collected during NOx performance
tests. Considerable data exist comparing NOx emissions to N20 emissions from diverse combustion
sources and techniques firing on various fossil fuels.2*7^9,10 As a result of increasing concern over
rising atmospheric N;0 concentrations, the first of a series of workshops specifically designed to
address this issue was conducted in 1986. The U.S. Environmental Protection Agency's (EPA's) Air
and Energy Engineering Research Laboratory (AEERL) sponsored this workshop designed to assist
1
-------
EPA in identifying critical issues related to fossil fuel combustion emission of N20 that would guide
EPA in developing an N20 research program plan.8 Additional EPA/AEERL-sponsored workshops
were conducted that continued to evaluate the role of fossil fuel combustion and N20 emissions. At
the 1988 workshop, the N-,0 grab sampling artifact was presented.10
The grab sampling artifact is a situation in which the presence of NOx, sulfur dioxide (SO-j),
and moisture, N-,0 is actually generated in grab sample containers through a chemical reaction/series
of chemical reactions.11,12 N-,0 generation approaching 200 ppm in grab sample containers has been
observed.13 Much of the data reported on N,0 measurements from fossil fuel combustion sources
were obtained using grab sampling methods conducive to the sampling artifact10-13 For EPA/AEERL
to continue conducting research characterizing N20 emissions from fossil fuel combustion sources,
sampling and monitoring methods that provided representative measurements were required.
N-,0 measurement from combustion sources has been performed using a variety of
methodologies including grab sampling and on-line monitoring techniques. Grab samples collected are
normally analyzed using gas chromatography (GQ methods. On-line monitoring techniques include
GC, nondispcrsive infrared (NDIR), Fourier-transform infrared (FTIR), and tuneable diode laser
infrared (TDUR) real-time analyzers-9,14-15'16,17 Each method has its own advantages and more often
than not, disadvantages. Grab sampling methods are appealing from a cost and convenience stand
point; however, the sample integrity has been demonstrated to be compromised under most common
sampling conditions.10,11,12,13 On-line, real-time analyzers are desirable for obvious reasons although
instrument costs are often prohibitive, detection levels are often insufficient, elaborate conditioning
systems are routinely required, and overall operation is often complex. Realizing that accurate and
reliable N20 measurements were essential to emissions characterization research, the Combustion
Research Branch (CRB) of EPA's AEERL initiated a program to concurrently develop grab sampling
and on-line monitoring methodologies suitable for characterizing N20 emissions from various
combustion sources and processes. As a result of this program, two AEERL Recommended Operating
-------
Procedures (ROPs) were generated. ROP No. 45, "Analysis of Nitrous Oxide from Combustion
Sources," details a gas chroma tography/electron capture detector (GC/ECD) method suitable for grab
sample analysis as well as on-line monitoring purposes.14 ROP No. 56, 'Collection of Gaseous Grab
Samples from Combustion Sources for Nitrous Oxide Measurement," details a grab sampling method
suitable for collection of gaseous grab samples bom combustion sources for the screening of N20
emissions.18 This report documents the approach and results obtained by Acurex Environmental while
developing these procedures.
1.2 AEERL RESEARCH EFFORTS
The CRB of EPA's AEERL has been active in evaluating N-,0 emissions from a variety of
fossil fuel combustion sources and equipment Early research efforts used grab sampling techniques
where the sampling artifact was later confirmed to be present. Following sampling artifact
identification, research efforts focused on developing reliable sampling and monitoring techniques to
re-evaluate these same combustion processes. Direct comparisons of on-line measurements to grab
sampling measurements were performed on in-house combustion facilities firing on varied fossil
fuels.13 These tests demonstrated the vast difference between the on-line and grab sampling
measurements. On-line N20 concentrations less than 2 ppm were common, whereas measurements
from the grab samples often yielded concentrations approaching 200 ppm.13 Actual NzO generation
within the sample container was found to vary with respect to initial (stack) SO,, NOx, and moisture
concentration. With this in mind, several tests were performed evaluating methods of moisture
removal and the subsequent artifact. Similarly, tests were also performed in which crude attempts at
SO-, removal were evaluated. To further understand the reactions occurring within the sample
container, measurements over time of N20, SO,, and NO/NO, were made using GC methods.
Having demonstrated that N20 measurements from pilot-scale fossil fuel combustion sources
in which grab sampling techniques were used could bias reported emissions by as much as several
orders of magnitude,13 the AEERL/CRB conducted a field study to evaluate the emissions from full-
3
-------
scale utility fossil fuel combustion equipment.19 This study also demonstrated the large disparity
between on-line and aged grab sample measured concentrations. The low on-line or actual
concentrations supported the tenet that the direct emission of N20 from fossil fuel combustion was not
a significant contributor to the increase in observed atmospheric N,0 concentrations. On-line and grab
sample measurements were performed on 12 coal-fired utility commercial boilers of varied firing
configurations and thermal load.19 On-line measurements revealed direct emission concentrations
nominally less than 5 ppm, whereas grab sample measurements often yielded N20 concentrations in
excess of 100 ppm.
During the course of the pilot-scale and full-scale field fossil fuel combustion emission
evaluations, the problem areas of AEERL/CRB's N20 measurement methodologies were identified.
Hie on-line GC method was susceptible to interferences present in flue gases measured. Memory
effects from moisture and S02 resulted in detector baseline instability as well as chromatography
difficulties.19 These effects had a direct impact on detector sensitivity, often reducing detection levels
to values above actual N,0 concentrations present in measured gas streams. Identical problems were
encountered when analyzing a large number of grab samples.
13 OBJECTIVES
Having taken the position that the direct emission of N,0 from fossil fuel combustion was not
a significant contributor to measured increases in atmospheric N.,0 concentrations, the AEERL/CRB
was interested in developing an economical method for screening various fossil fuel combustion
sources to further support this tenet as well as identify potentially high NzO emitting sources. The
most cost-effective method for meeting this objective was to develop a grab sampling method suitable
for this purpose. AEERL/CRB researchers also realized that developing a grab sampling technique
that completely eliminated the generation artifact would be difficult and that developing a grab
sampling technique that consistently minimized the artifact to acceptable levels for screening purposes
would be more practical. If the grab sampling method were to minimize the N-,0 generation artifact
4
-------
to less than 10 ppm over a 1-2 week period following sample collection, the resulting measurements
would be acceptable for direct comparison of previously reported data as well as identifying potentially
large emission sources. The intent was that if a potentially large emission source was identified
through grab sample screening, then that source could be further evaluated using on-line monitoring
techniques; on-line measurements provide the most accurate means for assessing actual direct N,0
emissions.
This would necessitate the development of an on-line monitoring method suitable for
combustion source application. The interfering effects of SO-, and moisture would need to be
eliminated if a gas chromatographic method were to be used. In addition, an automated monitoring
system would make the monitoring process more efficient, allowing for unattended operation. Using
this approach, AEERL/CRB implemented a series of tasks to develop sampling and analytical
capabilities to meet these objectives. Specifically, these objectives were to:
* Improve the existing GC/ECD instrumentation so that potential interferences present in
combustion process emissions do not effect continuous N-,0 measurements
* Develop a method to automate the GC/ECD system for near continuous on-line
monitoring purposes
* Configure the GC/ECD system so that it could be used for grab sample analyses as well
as on-line monitoring purposes
* Develop a grab sampling method that minimizes N20 generation in grab sample
containers to less than 10 ppm over a 1-2 week period
This project was performed under an AEERL-approved Category IV Quality Assurance Project
Plan (QTRAK No. 89014). This report documents the approach and results obtained while meeting
these objectives.
5
-------
SECTION 2
DEVELOPMENT OF ANALYTICAL PROCEDURES
2.1 DEVELOPMENT OF INITIAL ANALYTICAL METHOD
Acurex Environmental^ initial work for AEERL/CRB investigated suitable methods of grab
sampling analysis by GC/ECD. Because of the relatively recent need for combustion source grab
sample measurements, the number of analytical methods available were limited. In 1986, Dr. R.
Weiss proposed an analytical configuration at the first EPA workshop on NzO emissions from fossil
fuel combustion.8 This configuration is similar to the system used by Dr. Weiss for ambient
applications.20 The Energy and Environmental Research Corporation further adapted Weiss' proposed
method (see Appendix A). Another analytical approach was developed by Radian (see Appendix B).
The initial procedure, developed by Acurex Environmental, incorporated elements from both of these
procedures. Hie initial procedure was ultimately adopted as an AEERL ROP and presented in 1988 at
the third N20 workshop held in France.10 The original ROP, since modified, is contained in
Appendix C The appended ROP details the analytical apparatus and methodology and will not be
reiterated here.
During the initial development of the analytical method, a number of experimental concerns
were investigated. Hie non-linearity of the ECD was characterized over varied N-,0 concentrations as
well as carrier gases. The detector was found to have a more pronounced non-linearity at
concentrations less than 20 ppmv. For quantitative purposes, the linearity problem was accommodated
by increasing the number of calibration points and breaking up the overall analytical quantitative range
-------
into smaller, calibrated ranges. It was also found that the ECD-suitable carrier gases evaluated bad no
significant, effect on detector linearity. Hie detector sensitivity was slightly enhanced by the 5 percent
methane/95 percent argon (?5) carrier relative to the nitrogen carrier. C02 was evaluated as a
potential analytical interference. Reportedly, C02 could positively bias the ECD response to N-,0 if
the two ana Ivies were to coelute.9 To evaluate this possibility, a test was performed that compared the
detector response to N-,0 from an N20 and C02-containing calibration standard where the C02
component was eluted both before and after N20 using different chromatographic columns. No
significant difference in N20 ECD response was observed.13 Lastly, concern over detector
desensitizing from repetitive oxygen exposure was evaluated. A standard gas mix containing nominal
combustion effluent oxygen and C02 concentrations was analyzed continually over a 7-h period with
no disccmable loss in detector sensitivity.
Under the original Acurex Environmental analytical configuration, grab samples were
introduced to the analytical system via a vacuum evacuation apparatus (refer to Appendix C). The GC
gas sample loop was brought down to near absolute vacuum (-5 mm Hg), and a valve located between
the sample loop and grab sample container was opened allowing the gaseous sample to fill the sample
loop. An absolute manometer was used to determine the absolute pressure within the sample loop to
correct the sample volume. This system had a number of limitations, the majority of which were leak-
related. In addition, the grab sample containers (bombs) provided were of insufficient volume to
perform reliable replicate analyses. Ultimately, the vacuum evacuation apparatus was abandoned and
the analytical system and sample containers were configured for syringe injections.
Shortly before the European N-O workshop in June 1988, AEERL researchers became aware
of the N-,0 sampling artifact Based on this information, AEERL initiated efforts to characterize, by
on-line means, the direct N-,0 emissions from fossil combustion.
7
-------
22. INITIAL ON-LINE ANALYTICAL EFFORTS
As a result of the grab sampling container N-,0 generation artifact discovery, the need for the
development of on-line measurement/monitoring techniques became moie imperative. Real-time
monitoring capabilities were essential to establishing "true" N,0 emissions from fossil fuel combustion
and developing reliable grab sampling methodologies. Realizing this, AEERL/CRB initiated a series
of in-bouse tests to compare on-line measurements from pilot-scale fossil combustors to aged grab
samples collected at the lime of on-line measurement. NzO measurements were made from the grab
samples over progressive, elapsed periods of time to illustrate the extent of tbe sampling artifact.
These data are reported in detail elsewhere.13 Similarly, AEERL/CRB initiated a field study, also
conducted by Acurex Environmental, that performed similar tests on full-scale, coal-fired utility
boilers. These data are also described in detail elsewhere.13,19
These studies were conducted using GC/ECD systems configured as described in the
associated references. In summary, the GCs used were equipped with 63Ni ECDs nominally
maintained at 330 °C. N-,0 was chromatographically separated from flue gas components with a
0.125-in (0.32-cm) OD by 12-ft (3.66-m) stainless steel column packed with Porapak Super Q, 80/100
mesh (Alltech Associates Inc), using PS as the carrier at 20 cc/min. The analyses were performed
isothermally at 35 °C A 0.25-in (0.64-cm) OD by 1.5-in (3.8-cm) section of Teflon tubing filled with
indicating P2Os (AquaSorb, Mallinckrodt Inc.) was used as a precolumn for moisture removal.
Gaseous samples were introduced on column via a 6-port switching valve with a 1-cc sample loop.
Flue gas samples were obtained from a sample delivery system configured for use with continuous
emission monitors (CEMs). A portion of the sample stream, conditioned for moisture (refrigeration
condenser only) and particulate removal, was diverted under positive pressure to the gas sampling
valve.
During these studies, various fossil fuels and combustion configurations were evaluated. As a
result, a fairly representative cross section of combustion process effluents was encountered, both in
S
-------
composition and concentration. Also during these tests, ihe presence of high SO-, concentrations in
post-combustion gas streams was discovered to present analytical problems. Once on the analytical
column, the S02 component ultimately eluted from the column. Under the analytical conditions
employed, SO-, eluted from the column nearly 1 h after sample injection. In addition, because of the
low-column temperature, the peak shape was very broad (several minutes), resulting in an upset of the
baseline conditions and chromatographic difficulties. Similarly, coclution of the NzO and SO-,
components caused a reduction in detector sensitivity to N20. During the field study, this problem
was minimized through the use of dual detector GCs.19
23 ANALYTICAL METHOD IMPROVEMENT REQUIREMENTS
Realizing that the majority of reported fossil fuel combustion N20 emissions data were suspect
because of the discovery of the grab sampling artifact, AEERL/CRB researchers felt that although the
relative direct emissions of N20 emissions from fossil fuel combustion were probably much less than
previously reported, it was still necessary to characterize the actual direct N-,0 emissions from fossil
fuel combustion. AEERL/CRB believed this could be accomplished through a combination of grab
sampling and on-line monitoring campaigns where the grab sampling approach could be a mechanism
for screening potentially large N-,0-emiiting sources, which subsequently could be characterized in
detail through on-line monitoring efforts. However, substantial improvements would be required to the
analytical procedure to make it suitable for efficient, reliable, on-line monitoring applications. In
addition, an on-line monitoring method would be essential to the development of a grab sampling
method. Initial N20 concentrations would have to be established to evaluate the performance of the
grab sampling method.
Based on past results, the required analytical method improvements were fairly well defined.
The chromatographic interferences present in combustion process emissions would need to isolated
from the analytical system. The GCECD system would need to be automated to increase method
efficiency. An improved quantitative approach, compensating for the non-linearity of the detector,
9
-------
would be desirable. Lastly, the system should also be suitable for grab sample analyses. Based on
these requirements, improvements to the analytical method/system were initialed.
2.4 CONFIGURATION OF THE ANALYTICAL SYSTEM
2.4.1 Precolumn Selection
The interfering flue gas components, believed to be the cause of chromatographic/analytkal
difficulties, were isolated through a chromatographic backfiushing procedure. This technique uses a
precolumn to isolate the analyte of interest from slower eluting, undesirable constituents. Once the
analyte of interest has eluted from the precolumn to the secondary analytical column, the carrier gas
flow through the precolumn is reversed, flushing the undesirable components from the precolumn.
The primary combustion process flue gas components of concern were rao;>ture (H20) and
SO,, both of which have moderate response to the ECD. The relative retention times of these
components as well as N;0, CO,, and 02 were compared, and an elution order was determined for a
variety of potential chromatographic columns. The ideal precolumn would have adequate separation of
analytes at greater than ambient temperature, and the interfeiants (SO, and H20) would both elute
after N20. In addition, the length of the precolumn should be minimized to avoid excessive back
pressure of the carrier gas within the chromatographic system. Based on these criteria, precolumn
candidates were evaluated. Realizing that any change in elution order of the C02 and NzO
components would probably complicate the analyses, columns where the elution did not change
narrowed the selection. Of the remaining candidate preoolumns, relative separation of N20 and H20
was used to further isolate precolumn suitability. Using this selection technique, the precolumn
packing materials were narrowed to basically two choices: Pcrcpak Q, is sasst packing material
contained in the analytical column, and HayeSep D 100/120 mesh (AllLech Associates Inc), a packing
similar in properties to the Poropak Q but apparently more efficient at separating identical compounds
at comparable temperatures. Several HayeSep D columns of varied length were obtained for
evaluation.
10
-------
2.4.2 Description of the Backflush Method
The backflushing method uses a single, 10-port valve to divert/direct the flow of carrier and
sample gas streams through the chromatographic system. A schematic diagram of the 10-port valve
system is presented in Figure 2-1. The 10-port valve can be operated in two positions or modes. In
the off or backflush position (diagram 2-la), the precolumn is backflnshed by carrier 2 to a vent (ports
10, 9, 6, and 8, consecutively). The analytical column, supplied by carrier 1 (ports 5 and 7,
consecutively), is interfaced to the detector. A 1-cc sample loop, bridged by ports 3 and 4, can be
charged with the sample stream (ports 1 and 2, consecutively). In the on or analyze position (diagram
2-lb), the valve is switched to align tbc carrier gas flow so that the sample loop, precolumn, and
analytical column are routed in series (consecutively) to the detector. Once the valve is switched,
carrier 1 purges the sample loop onto the precolumn (pons 5, 3, 4, and 6, consecutively). The effluent
of the precolumn is routed to the analytical column and on to the detector (ports 9 and 7,
consecutively). Carrier 2 is vented via ports 10 and 8. The sample stream is vented via ports 1 and 2.
Once the analyte of interest has eluted from the precolumn onto the analytical column, the valve is
returned to the backflush position, the flow through the precolumn reversed, and the undesirable
sample components is purged from the precolumn. The NzO GC/ECD analytical system was
configured using this approach.
All previous analytical work was performed using either rented or borrowed Varian GCs or
CRB's Shimadzu GC. The Shimadzu GC had a number of hardware limitations that made changes in
plumbing more complicated than necessary. Similarly, the addition of a 10-port valve, required for
precolumn backflushing, could not be easily incoiporatcd into the Shimadzu system. As a result, the
Shimadzu GC was not considered for backflushing configuration. A Hewlett-Packard (HP) 5890 GC
was made available for the backflushing configuration. An ECD was installed on the GC, and a 10-
port valve was incorporated into the analytical system. This required installation of separate carrier
mass flow controllers.
11
-------
FROM CARRIER!
LOOP
PRECOLUMN
OUT
SAMHLB
VENT
COLUMN
DCsrECrOR
FROM CARRIER!
LOOP
PRECOLUMN
VENT
SAMPLE
OUT
IN
COLUMN
GO
DETLCTOR
FROM CARRIER 2
FROM CARRIER 2
2-la. Off position.
2-lb. On position.
Figure 2-1.
Schematic diagram of 10-porl valve system.
-------
To eliminate the need for manual valve switching, an air actuator, controlled by the GC
system, was used to change valve position. The automation of the valving system was accomplished
by interfacing the GC and integrator to a limed event control module that converted digital commands
from the integrator to time-controlled electrical switches. The integrator could be programmed to turn
the solenoid valves on or off at specific times. The solenoid valves, when energized, allowed
compressed air to pressurize the air actuator. When pressurized, the air actuator rotates the 10-port
valve to the desired position.
The backflush system pre column operating parameters were determined by characterizing the
retention limes of N20 for the HaycSep D precolumns at varied isothermal oven temperatures. Both a
3-ft (0.91-m) and 6-ft (1.83-m) precolumn were characterized. These retention times were used to
determine when the 10-port 'alve should be switched and backflushing initiated. The precolumns
were then evaluated individually when incorporated into the entire analytical system. A decision was
made to retain use of the 12-ft Porapak Q column as the analytical column. The 3-ft HayeSep D
column displayed acceptable chromatographic resolution when coupled with the analytical column;
however, baseline upset, resulting from pressure changes within the system during valve switching,
interfered with integration of the X20 peak. The isothermal oven temperature was varied in an
attempt to eliminate the condition, but unsuccessfully. The carrier gas head pressures, required to
obtain the targeted flow rates (20-30 cc/min), varied greatly between the 3-ft and 12-ft columns (-15
psig vs. -40 psig). This pressure disparity was the likely source of the baseline upset
The 6-ft HayeSep D column was evaluated with much more success. Baseline upsets were
much less severe and ultimately disappeared altogether. The disparity between column-head pressures
was also much less (-30 psig vs. 40 psig). Acceptable chromatographic resolution of N20 was
observed. Because of the encouraging results obtained with the 6-ft HayeSep D precolumn, this
column was selected as the backflush method precolumn. All future tests were performed with this
precolumn.
13
-------
To confirm the backflush activation time, the automated program was modified in small-time
increments, to decrease the elapsed time into the run when backflushing was initiated. The elapsed
time was reduced in 0.1-min increments until the N-,0 component no longer ciuted from the Porapak
Q analytical column. Adding 0.2 rain (12 sec) to this elapsed time into the run was felt to be
sufficient to backflush the S02 and H20 interferants.
At this time, the analytical system was ready for more rigorous evaluation. The GC/ECD
backflush system was incorporated into the Flue Gas Simulation System (FGSS), described in detail in
Section 3, to evaluate the method under more realistic conditions. A simulated Que gas, containing
realistic concentrations of SO-, (-1,200 ppm), NO (-600 ppm), and moisture (~5 percent by volume)
was routed to the system. An ice bath moisture condenser and a P2Os desiccant cartridge was located
upstream of the 10-port valve sample loop to remove moisture. No difficulties were encountered
during continuous analysis of the simulated flue gas sample. The system was subjected to varied
sample moisture concentrations by varying the moisture removal devices. Tests were performed where
only the ice bath was used for moisture removal. No discernable difference in system performance
was observed. Similarly, no moisture removal was attempted; the unconditioned, simulated flue gas
was routed straight to the sample loop. A long-term baseline upset and loss of detector sensitivity was
observed under this condition.
At this point, two options were evident. A different precolumn, suitable for high moisture
content use, could be identified and evaluated or, the system would require the moisture conditioning
of the sample stream before sample loop delivery. The latter option was not compromising to
analytical requirements, primarily because the analytical system would be used as an on-line
monitoring device and moisture removal by refrigeration condensation was commonly used by
continuous emission monitoring (CEM) sample conditioning systems. To verify this approach, the
GC/ECD analytical system was incorporated into FGSS CEM system for long-term evaluation.
14
-------
To further aid in analytical system automation, a solenoid valve -vas installed upstream of the
10-port valve sample loop. The purpose of this valve was to allow continuous purging of the sample
loop until the actual time of analysis. By interfacing the solenoid valve to the timed event control
system, the valve could be automatically controlled to open and close in coordination with analytical
sequence. The valve was controlled so that the sample loop was continuously purged with the sample
stream up to the lime of analysis, at which time the solenoid valve was closed, sample flow was
stopped, and the sample loop was equilibrated to atmospheric pressure. At the time of backflushing,
the 10-port valve was returned to the off position, and the solenoid valve opened, restoring flow to the
sample loop. This was essentially the last modification to the backflushing method.
At this point, the backflushing method analytical para mete is were clearly defined. These
parameters are as follows:
• Precolumn — 6-ft (1.8-m) by 0.125-in (0.32-cm) OD stainless steel, packed with HayeSep
D - 100/120 mesh support; carrier flow of 30 cc/min (head pressure at -30 psig)
• Analytical Column — 12-ft (3.7-m) by 0.125-in (0.32-cm) OD stainless steel, packed with
Porapak Super Q - 80/100 mesh support; carrier flow of 30 cc/min (head pressure at -40
psig)
• Carrier Gas — 5 or 10 percent methane in argon (P5, P10)
• Detector — 63Ni constant current cell ECD maintained at 300 °C
• GC Oven Temperature — Isothermal, 50 °C
The sequence of timed events were programmed as follows (times denote elapsed time into
run):
• 0.0 min — Close, solenoid valve (stop sample flow to sample loop)
• 0.1 min — Actuate 10-port valve, move to analyze position
• 3.6 min — Actuate 10-port valve, move to backflush position
• 3.7 min — Open solenoid valve (restore sample flow to sample loop)
15
-------
• 8.0 min — Stop run, integrate peak areas
Figure 2-2 depicts a schematic diagram of the automated system. The system was also capable
of unattended, continuous operation, by incorporating the programmed timed events into a separate
BASIC program capable of loop functions. At the end of the analytical run, the system was capable
of automatically re-initiating the sequence of timed events.
2.5 ANALYTICAL METHOD PERFORMANCE
2.5.1 Method Quantitative Capabilities
The quantitation of N20 is accomplished by relating integrated peak area to the linear
relationship between calibration variables (N-,0 concentration and peak area). A least squares linear
regression of the calibration variables is a commonly used calibration approach. The linear
relationship can be expressed by the equation:
y ~ mx + b
where: y = integrated peak area
m = the slope of the calibration curve
b = Jie intercept of the calibration curve
x s concentration
To determine unknown concentration, the following equation is used:
* = Lil
m
However, this quantitative approach has limitations. The non-linear response of the detector to
N-,0 concentrations nominally less than 20 ppm had been demonstrated early in the N->0 measurement
program. As described in the original ROP 45 (Appendix Q, this situation was compensated for by
narrowing the quantitative concentration ranges. To improve quantitative accuracy as well as to
expand the linear range of quantitation, the linear properties of the ECD were evaluated further. With
16
-------
ToCEMi
Rcfngcntloa
Dryer
Vacuum Pump
rion SUcm
Staple Um
115 VAC
N20 Spaa
ConprtiKd Air
FSCarricf
6 n
Pr*-co!ajna
Analytical
Cotumo
N20
Air
PJ
Spta
or
Carrier
Om
Howe
Om
Air
Sample Event
_ Control module
Integrator
Figure 2-2. Automated, on-line (3C/ECD N2P monitoring system.
-------
the assistance of the Research Triangle Institute (RTI), alternative mathematical approaches were
considered.
The linear regression approach enables the determination of quantitative bias on an absolute
basis. With this approach, error can be reported as less than a certain concentration, often reported as
percentage of full scale or as deviation from the true or known value. A problem arises in that the
estimated bias for low concentrations will be very large relative to the measured or true
concentration.14 By performing a linear regression of natural log (ln) transformed calibration
variables, error is capable of being reported on a relative basis. The equation for the curve is of the
form:
In (y) = m[ln(x)] + ln(fc)
where: ln(y) = the natural log of integrated peak area
m = the slope of the calibration curve
ln(b) = the natural log of the intercept of the calibration curve
ln(x) = the natural log of the concentration
The unknown concentration is determined using the formula:
- InQQ - lnfl)
m
A comparison of these two quantitative approaches are presented Figures 2-3 and 2-4 and
Table 2-1. Figures 2-3 and 2-4 compare ECD response to the mathematical linearizing approach while
Table 2-1 demonstrates the relative bias of calculated concentrations (relative to the true concentration)
18
-------
J3
*a
I
1,000,000
800,000 -
600,000
&
g 400,000
8 3
200,000 -
40 60 80 100
N20 Concentration (ppm)
R-square = 0.9935379
Figure 2-3. ECD response lo NjO calibration gases (linear regression).
-------
I I I I
"(1) 0 12 3
In (N20 Concentration)
R-squarc = 0.9992821
Figure 2-4. ECD response lo N20 calibration gases (linear regression of log transformed variables).
-------
using both quantitative approaches. The linear regression of the transformed calibration variables was
TABLE 2-1. COMPARISON OF RELATIVE BIAS USING DIFFERING
MATHEMATICAL APPROACHES
Linear Regression Linear Regression
(untransfonned variables) (transformed variables)
N20 Known
PPm
N20 Calc.
ppm
% Bias
N20 Calc.
ppm
% Bias
0.51
-3.11
-705.1
0.47
-8.6
0.97
-2.13
-319.6
0.99
2.1
1.99
-0.40
-120.1
2.02
1.5
5.03
4.58
-8.9
5.36
6.6
9.85
11.35
15.2
10.41
5.7
19.4
23.18
19.5
20.11
3.7
40.4
45.74
13.2
40.45
0.1
80.1
83.36
4.1
77.74
-2.9
128
123.68
-3.4
120.79
-5.6
effective in minimizing the relative error of calculated concentrations. Less than 10 percent bias was
observed over the entire quantitative range as opposed to as much as 700 percent relative bias for the
non-transformed quantitative approach.
2.5.2 On-line Monitoring Performance
The automated, on-line GC/ECD system was evaluated extensively on a number of diverse
EPA/AEERL fossil fuel combustion test facilities. Initially, the analytical system was used exclusively
during the development of the N,0 grab sampling method. On-line and grab sample measurements
were performed on gases generated by the Flue Gas Simulation System (FGSS). The on-line
concentrations measured were compared to grab sample measured concentrations to assess artifact
generation. These tests are described in detail in Section 3. Once the reliability of the analytical
system had been demonstrated, the on-line monitoring device was evaluated on actual combustion test
equipment.
21
-------
For AEERL's Gas Cleaning Technology Branch (GCTB), the N,0 monitoring system was
used to measure N-,0 emissions resulting from the combustion of various coals during parametric SO-,
removal testing. These tests, performed on EPA's Innovative Furnace Reactor (1FR), are described in
further detail in Section 3. The N-,0 concentrations measured ranged from 0_5 to 10 ppm.
During these tests, quality control (QC) span checks were performed nearly every hour over
the course of the S-h test period. The QC checks were used to assess method analytical bias and
precision over the course of the entire test period. The reliability of the analytical system was without
question. All span checks performed were within method QC objective limits. The results of these
QC checks in the form of a control chart, are graphically presented in Figure 2-5. The average bias
observed (Z9 percent) was well within the targeted level of less than 15 percent. Similarly, the
precision observed (2.7 percent), expressed as percent relative standard deviation (RSD) was well
within the targeted level of less than 10 percent.
The on-line GC/ECD system was loaned to GCTB for a scries of selective non-catalytic NOx
reduction (SNCR) tests. During these tests, additives such as ammonia and urea were injected into the
IFR to reduce NOx emissions. The on-line measurements were used to compare N20 emissions with
and without NOx control. The N-,0 concentrations measured ranged from 0.5 to 35 ppm. No
difficulties were encountered during analysis. All QC checks were within method requirements. The
analyzer was loaned to GCTB because their primary method of N,0 measurement, a tunable diode
laser, was experiencing operating difficulties. During the development of the TDLIR system, the oil-
line GC/ECD system was relied on to establish the actual flue gas N-,0 concentrations for performance
evaluation purposes.
The automated, on-line GC/ECD system was also used by GCTB to characterize the N20
emissions from a selective catalytic NOx reduction (SCR) pilot-scale test facility. N;0 concentrations
were measured both before and after the catalyst was evaluated. Measured concentrations ranged from
0.5 to 3 ppm. Again, the GOECD system performed reliably.
22
-------
K
a
Ǥ
a
ffl
$
20
15
10
5'
0
-5
-10
-15
-20
1
Precision = 2.7 % RSD
Span b S.03 ppm N20 In N2
_l I I L
4 5 6 7
Span Check Number
8
10
Figure 2-5. Automated tifi analysis results of span checks over course of a test day.
-------
The GC/ECD system was also evaluated under ambient conditions. For the Radon Mitigation
Branch (RMB), the system was used to assess the N->0 mass emissions resulting from the open-hearth
combustion of coal. In China, the open-hearth combustion of coal comprises a significant portion of
all coal burned. These ambient measurements were used to assess the magnitude of the mass
contribution of N20 to the environment from this combustion source. The N20 concentrations
measured were only slightly above ambient concentrations. However, the GC/ECD analytical system
was sensitive enough to resolve this 100-200 ppb relative increase.
These ambient measurements were performed over the course of several weeks. At the
beginning and end of each test period, QC span checks were performed. The results of these QC
checks are graphically presented in Figure 2-6 in the form of a control chart. The results demonstrate
that the analytical system is capable of long-term, reliable performance. The average bias over the 2-
weck period was only 3.4 percent, whereas the average precision was 2.9 percent.
2.6 ANALYTICAL METHOD SUMMARY
The GC/ECD backOush method developed was found to be suitable for the measurement of
N,0 from a variety of combustion sources and applications. In addition, the method was found to be
equally suitable for on-line monitoring or grab sample analysis purposes. Analytical interferences,
present in combustion process effluents, were negated through the use of a backflushing technique.
Method accuracy, expressed as percent bias, and precision, expressed as percent relative standard
deviation, were determined to be s r 15 percent and s 10 percent, respectively. The method was
found to be suitable for the quantitation of N20 concentrations ranging from 0.100 to 200 ppm.
Using this method for on-line monitoring purposes allows a semicontinuous measurement
approximately every 8 min. The system can be easily incorporated into most continuous emission
monitoring sample delivery/conditioning systems. The only requirement is the removal of particulate
and moisture from the sample stream by a refrigeration condenser. The sample stream should be
diverted to the analytical system before further moisture conditioning by a desiccant.
-------
£
K
I
1
m
£
20
15
10
5
0
-5
-10
-15
-20
0 PreTest Span Checks
A Post Test Span Checks
_L
10/30/91
PkUH Precision »4.11 % RSD
Posl Tul Precision = 1.77 % RSD
_L
A
0
A
O
_L
11/06/91 11/06/91 11/12/91
Test Date
11/13/91 11/14/91
Span b 0.97 ppm N20 Id N2
Figure 2-6. Automated N^O analysis results of span checks over Ihc course of a 2-week period.
-------
Through use of the backflushing technique, known interferences such as S02 and moisture are
isolated by a precolumn and purged from the system by backflushing the precolumn. Other common
~uc gas components such as O,, CO, CO-* NOx, unburn ed hydrocarbons (THC), and ammonia (NH3)
were found not to interfere with the analytical procedure.
The non-linear response of the detector to N,0 at low concentrations was minimized through
use of a logarithmic transformation of the calibration variables. The transformed data are used to
derive a least-squares linecr regression.
26
-------
SECTION 3
GRAB SAMPLE METHOD DEVELOPMENT
3.1 BACKGROUND
The discovery of the N-,0 sampling artifact attenuated the need for a standardized, reliable
sample method to accurately assess the N,0 emissions from fossil fuel combustion sources. As
previously mentioned, much of the reported N20 combustion emissions data was collected under
conditions conducive to the N,0 sampling artifact.10,12,21 Grab samples were collected in a variety of
sampling containers including glass flasks, stainless steel canisters and Tedlar bags.
Muzio and Kramlich were among the first researchers to identify the sampling artifact
reactants as well as potential formation mechanisms.11,12 The group identified the key artifact
reactants as SO-* NOx, and water, components present in most fossil fuel combustion process
emissions. The sampling artifact was also independently confirmed by a number of other
researchers.12-22,23
Solution-phase reactions between NOz and S02 with N%0 as a product have been documented.
Martin et al., identified N20 as a product in the reactions of NOx with SO., in the aqueous phase of
atmospheric aerosols.24 Chang et al., studying the chemistry of flue gas desulfurization, identified a
mechanism in which hyponitric acid decomposed into N2O.25 Lyon and Cole have performed kinetic
modeling on the proposed reactions occurring within aged grab sample containers.26 DeSoete also
conducted a detailed examination on the kinetics of solution-phase reactions leading to the formation
of NzO in grab sample containers.22
27
-------
The selective removal of any or all of these reactants was targeted as ao approach to
eliminating the sampling artifact Tests performed by Muzio et al., evaluated the effect of drying the
gas stream sampled as well as neutralizing S02 with NaOH scrubbing solutions.12 Results of the
performed tests indicated that N,0 generation within aged grab sample containers could be drastically
reduced, possibly even eliminated.
AEERL/CRB also performed work that investigated the use of methods to minimize the
sampling artifact13 Efforts focused on methods for removing moisture from the sample gas stream
only. The use of a desiccant, phosphorus pentoxidc (P205), was effective to drastically reduce the
artifact generation but unable to eliminate it completely.
Realizing that it would be extremely difficult, if not impossible, to consistently eliminate the
sampling artifact entirely, the AEERL/CRB believed that if NzO generation within aged sample
containers could be minimized to consistent levels, this would be suitable to screen for high K20-
emitting combustion sources. Specifically, it was felt by AEERL researchers that if N,0 generation
within grab sample containers could be consistently minimized to less than 10 ppm over a 1-2 week
period, this would be more than acceptable to screen for high I^O-emitting fossil fuel combustion
sources. The screening technique could then be used to direct on-line monitoring efforts.
The screening of intended fossil fuel combustion sources would require the voluntary
cooperation of commercial and research combustion facilities. Therefore, the grab sampling equipment
and technique must be easy to use and pose minimal imposition to those participating in screening
surveys. Specifically, the grab sampling method should not require a great degree of sampling
expertise. In addition, the grab sample should be capable of being obtained in a manner compatible
with commonly employed CEM sample delivery systems.
Because die screening of numerous fossil fuel combustion sources was intended, great
consideration into the preparing, shipping, and receiving of the grab sampling equipment was essential.
28
-------
The sampling equipment must be durable and compact Similarly, the restrictions of shipping
chemical reagents must be considered.
With these factors in mind, development of a N20 grab sampling method was initiated.
Specifically, the objective of this study was to determine, if possible, the conditions under which a
grab sample could be collected that minimized the N-,0 generation artifact while still allowing reliable,
representative N,0 measurements so that major N;0 eraitleis could be isolated. Primary objectives of
the study were as follows:
• Identify and evaluate materials that effectively remove the key flue gas constituents of
SO-, and moisture
• Incorporate and optimize these materials into an apparatus that can be easily adapted for
use on existing on-line, continuous emission monitor systems
• Minimize the N20 generation sampling artifact to less than 10 ppra
• Identify the NOx, SO-,, and H,0 concentration ranges where the method is applicable
• Validate the sorbent system on an actual combustion systems
• Determine the methods suitability through field evaluations
The following information demonstrates the approach taken and the tests conducted to meet
these objectives. The majority of the work was performed between January 1990 and August 1991.
Ultimately, the sampling procedures developed were documented in the form of an EPA/AEERL
ROP.18
3.2 GRAB SAMPLING EQUIPMENT CONSIDERATIONS
Ultimately, the grab sampling method, once developed, would be used to conduct a rigorous,
comprehensive field survey of the emissions from various fossil fuel combustion sources including
industrial boilers and power plants, fluidized bed combustors, and various pilot- and full-scale test
facilities. Voluntary cooperation of solicited participants would be critical to the success of the
29
-------
screening campaign. Therefore, it was important that the collection of grab samples for screening
purposes be as unobtrusive as possible.
The logistics of transporting the sample equipment would also be extremely important. Hie
complexities of shipping the equipment both to and from prospective screening candidates must be
considered. For example, the scaling and secondary containment of liquid samples would create an
added burden to screening participants. In addition, if liquid chemical reagents were used, precautions
could be necessary to ensure their safe shipment by screening participants. Similarly, chemical
shipping restrictions could have an adverse impact on the screening efforts if liquid chemical reagents
were used.
How the sampling system would be used was the most important factor when considering the
ideal characteristics of the field grab sampling method. It was felt that to increase survey
participation, the sampling method and equipment would need to be very simple to use. Potential N-,0
screening survey participants may possess little, if any, stack or source sampling experience. In
addition, it was believed that if the grab sampling equipment could be adapted or incorporated into
existing gaseous sample delivery systems, then participation in the screening survey could be
increased. It was expected that the vast majority, if not all, candidate combustion sources would
possess some type of continuous emission monitoring system. If the NzO grab sampling equipment
could be incorporated into this system, the need for a stand-alone sampling system could be
eliminated.
With these considerations, the actual components of the grab sampling equipment and method
were further identified. The use of dry sorbents for the neutralization of SO, was chosen as a starting
point Dry, calcium-based sorbents are commonly used for flue gas desulfurization processes and
could easily be used in a stack sampling configuration. The use of dry sorbents could eliminate the
neci for impingeis and other glassware associated with the use of liquid scrubbers and, therefore,
minimize glassware breakage problems. By using dry sorbents and thereby eliminating the use of
-------
liquid-filled impingers, the concern over N20 solubility in water could be avoided. DeSoete has
shown however, that N,0 solubility does not appear to be a problem.22 The use of dry sorbents could
also minimize sampling equipment shipping concerns, and thus, eliminating the risk of chemical spills
or leakage.
The elimination of potential scrubbing solutions such as sodium hydroxide, which also
removes CO-,, could possibly relieve potential quantitative concerns. If CO, were to be removed from
the sample gas, the NzO concentrations measured could be biased as a result The contribution of CO-,
to the entire sample volume is significant (-8-15 percent), and the loss in volume would require a
correction to measured N-,0 concentrations. In addition, the remaining CO, concentration would have
to be measured to complete the volume correction, requiring a separate analytical method for CO-,.
3.4 EXPERIMENTAL APPARATUS
Specially designed test equipment was used during the development of the grab sampling
methodology. This equipment allowed independent evaluation and control of parameters effecting the
integrity of aged grab samples. A test facility was designed and built that simulated typical
combustion process effluents, both in composition and concentration. A separate emission monitoring
system was used to determine generated gas concentrations. The individual components of these
systems are described in following sections.
3.4.1 Flue Gas Simulation System fFGSS')
The objective when designing the Flue Gas Simulation System (FGSS) was to simulate a flue
gas in the laboratory with the capability to vary the concentration of NO, SO,, and moisture,
independently (NO, 0-1,000 ppm; SO,, 0-2^00 ppm; and moisture, 0-20 percent by volume). This
system could then be used to conduct studies of the absorption of H,0 and SO, from a flue gas
stream by solid sorbents. The system was engineered and assembled with the capacity to accomplish
the following:
• Vary the concentrations of NO, SO,, and moisture, independently
-------
* Incorporate other gases into the system
* Continuously monitor for NO and SO, at varied locations
* Measure NzO on-line
¦ Maintain stable readings
The FGSS is a combination of two independent systems: the flue gas generation system and
the sampling system. This design allows for flexibility in sampling positions while continuously
monitoring NO and SO^
3.4.2 Flue Gas Generation Svstem
The FGSS requires three supply gases: nominally 2,000 ppm NO in nitrogen; 5,000 ppm SO-,
in air, and pure N2. N-,0 is introduced into the system from the NO cylinder which inherently
contains between 1-5 ppm N^O.
The three supply gases flow directly to four calibrated rotameters (Figure 3-1).
The supply gases are then fed into the mixing system in two gas streams. Gas stream No. 1 contains
NO and N2- Gas stream No. 2 contains S02 and N,. Both streams are balanced to the same flow
rate, 9 L/min. The N2 is a makeup gas in both gas streams (e.g., if the NO flow is decreased, the N-,
Cow is increased to maintain the 9 L/min, flow rate). Both gas streams are equipped with pressure
gauges to make rotameter flow corrections. These calculations and corrections are used to roughly set
the rotameters. The actual S02 and NOx concentrations are measured at the exit of the gas stream.
Gas stream No. 1 is directed upstream of the moisture generator because of the relative insolubility of
NO and N20 in water. Gas stream No. 2 is introduced downstream of the moisture generator because
of the greater solubility of S02 in water.
3.4.2.1 Moisture Generator
Moisture is generated from a 2-L, insulated, and temperature-controlled glass impinger and is
filled with 1 L of deionized water. The flow rate of gas stream No. 1 into the impinger is held
32
-------
NO.
Row
Figure 3-1. FIuo gas generation system.
-------
constant at 9 L/min while the tempera aire can be varied to change the percent moisture. A multipoint
calibration ranging from 60-100 °C (5-15 percent moisture) was performed on the moisture generator.
3.4.2.2 Mixing Chamber
After the impinger, the two gas streams combine to make gas stream No. 3, which has a
combined total flow rate of 18 L/min. At close to atmospheric pressure, stream No. 3 flows directly
into a mixing chamber and (ben into a vacuum pump. The mixing chamber is an insulated and
temperature controlled 6-in ID by 12-in stainless steel pipe (volume: 0.2 ft3 or 5.6 L). "The chamber
temperature is held at 105 °C The humidity of the gas stream during sampling is roughly monitored
by wet and dry bulb temperatures. The wet and dry bulb temperatures are monitored separately. A
regulating valve is located between the mixing chamber and the vacuum pump to balance the sample
pressure in the system and to ensure a constant flow rate of 18 L/min. The chamber is also equipped
with a pressure/vacuum gauge to monitor the chamber pressure.
For sampling purposes, an atmospheric dump is located on the outlet side of the chamber
pump. The atmospheric dump allows a sample to be withdrawn without affecting the total flow of the
flue gas generation system. This is achieved by enlarging the 1/4-in tubing to a 1/2-in tee. The
majority of the simulated flue gas vents through the 1/2-in tee. The 1/2 in-tee also connects the flue
gas generation system with the sampling system.
3.43 Sampling System
Through the addition of a smaller sample pump, the flue gas generation system and the
sampling system can operate independently. The smaller sample pump pulls a fraction of the
simulated flue gas into the sampling system from the 1/2-in tee (Figure 3-2). A 1/4-in tee is located at
the outlet of the sample pump. Part of the gas is directed under positive pressure at a regulated flow
through a 1/4-in heated Teflon sample line to a S02 "high" analyzer, 0-5,000 ppm (Teledyne UV).
The remainder of the sample is directed to the common port of a three-way valve. The valve allows
the sample to flow through either a "bypass" loop or a sorbent/sample loop.
-------
MAONEJIEDC
S03 Safest
SO] ID
O-MWM
ANALYZER
IORBENT/BOMB IjOGP
TtfAiuiMn
IbitedSWclJ
S021DOII
0 ¦ 3000 MM
ANALYZE*
(GC/EjCu)
Figure 3-2. Flue gas sampling system.
-------
3.4.3.1 Bypass Loop
The bypass loop (do sorbents) allows measurement of the initial flue gas concentrations (SO-,
high, NO, H20) untouched by sorbents. This allows the monitoring of any effects the sorbents may
have on the initial flue gas concentrations once the sorbents are placed in-line. The sample flows into
the fiist three-way valve through a water knockout device and then to a second three-way valve that
directs the simulated flue gas to the analyzers.
3.4.3.2 Sorbent/Sample Bomb Loop
The sorbent/sample bomb loop allows measurement of the flue gas concentrations (SO, low,
NO, and N20) after the gas flows through the sorbent system. The sorbent/sample bomb loop is
equipped with a rotameter to measure the sample flow rate through the sample bombs. Another
bypass loop between the sorbents and the bombs allows continuous flow through the sorbents when
samples are not being collected. The exit of the sorbent/sample loop is connected to the second three-
way valve.
3.4.3.3 Analyzers
The sample stream leaves the second three-way valve and is diverted four ways:
1. Through a rotameter to an NO analyzer, 0-1,000 ppm (Thermo Electron, Model 10,
chemiluminescent N0-N02-N0x analyzer).
Z To a SO, low analyzer, 0-50 ppm (Thermo Electron, Series 40, pulsed fluorescent
analyzer with the Perma Pure Dryer reissvsd). The SO, analyzer is used only when
sorbents are placed in the sorbent/sample bomb loop. The analyzer requires a dry sample.
3. Through a rotameter to a GC/ECD with a 1-mL sample loop for N-,0 measurements.
4. To a differential pressure gauge (0-10" HzO) with a needle valve to regulate pressure on
the system and then to vent.
36
-------
3.4.4 Flue Gas Measurements
3.4.4.1 Continuous Emission Monitors
The SO, and NO analyzers were calibrated Co verify linearity before initial testing. The
analyzers were then calibrated every 2 weeks or when the bias exceeded predetermined limits. These
limits are shown in Table 3-1.
The calibration consisted of at least three points (zero, span, and mid-point). All span gases
were delivered at a constant pressure and flow rates identical to those used during sampling.
The analog output from each CEM instrument was interfaced with a computer data
acquisition system. Since the instrument was based on linear measurement properties, the slope or
range was used to calculate concentration in ppm or percent. Data were collected over a timed
average and were automatically stored on disk. A hard copy was also produced for permanent record.
The daily OC checks conducted before and after each test period were used to validate data and
monitor system performance.
3.4.4.2 GC/ECD and N,0 Measurements
N20 measurements were performed on a Hewlett Packard 5890 GC/ECD configured for
automated, on-line N-,0 measurements. The analytical system has been previously described in
TABLE 3-1. PREDETERMINED LIMITS
Analyzer
Accuracy (% bias)
Precision (% RSD)
NO
± 20
10
so2
± 20
10
SO, lo
± 3 ppm
10
Section 2. EPA/AEERL ROP No. 45 was used as the procedural guidelines.14 N-,0 measurements
were either taken on-line or through direct injection with a 10-mL glass syringe. A multipoint
calibration was performed using the on-line method. The method of direct injection was verified using
three different span gases. There was no bias between the two methods. The sample loop required at
37
-------
least seven volume exchanges to ensure a representative sample. This required two 10-mL syringe
flushes before every N20 measurement when using the direct injection method.
3.4.5 Sampling Procedure
A standardized test plan for the evaluation of sorbents was followed. The plan called for
holding the FGSS conditions constant while varying the sand/sorbent mixtures, flow rates, and
sampling positions.
Throughout the first part of the study the FGSS conditions were held constant (10 percent
moisture, 1,200 ppm SO-,, 600 ppm NO, and about 0.5 ppm N^O). These were referred to as the
nominal inlet conditions.
The procedure for evaluating the sorbent cartridges (unless otherwise specified) was as
follows:
1. Shoot standard on GC.
2. Span CEMs.
3. Fill impinger with 1 L of deionized HjO.
4. Set temperatures: impinger, 60 °C; wet/dry bulb, 105 "C
5. Insert sorbent cartridge system into FGSS.
6. Switch FGSS to bypass loop.
7. Turn on SOP NO, and N2 supply gases.
8. Set rotameters at calculated values.
9. Let system equilibrate (- 5 min).
10. Take on-line bypass N20 measurement
11. Switch to sorbent/sample loop.
12. Let flue gas run through sorbent cartridge (-5 min).
13. Take a 2-min bomb sample (-4 L/min).
14. Take on-line N20 measurement from bomb exit
38
-------
15. Let Qoe gas run through soibent cartridge (-10 min).
16. Repeat steps 13-15 for two additional bombs.
17. Switch to bypass loop.
18. Take on-line bypass N20 measurement.
19. Turn off supply gases.
20. Span CEMs.
21. Shoot GC standard.
22. Shut down system.
This procedure is for one sorbent cartridge system with the collection of three sample bombs.
A test usually evaluated three sorbent cartridge systems at the same conditions with the collection of a
total of nine bombs. Each of the two remaining cartridges were inserted after step IS and steps 10-18
were repeated for each cartridge set At least seven volume exchanges were passed through the 600 cc
sample bombs to ensure a representative sample.
The bombs were nominally aged 5-7 days. The nine bombs from one test were aged the same
amount of time. The aged bombs were then analyzed for N20 by direct injection into a GC/ECD.
Duplicate 10-mL samples were withdrawn through the septa on the sample bomb with a 10-mL glass
syringe.
3.5 INITIAL SORBENT TESTS
3.5.1 Introduction
The initial series of tests necessitated a qualitative screening approach. Many of the tests were
of the yes/no or go/no-go nature. These types of tests were required to identify candidate materials
early in the study and then optimize their performance. The tests were conducted under the FGSS
conditions of 10 percent moisture, 1,200 ppm S02, 600 ppm NO, and about 0.5 ppm N20 and wQl be
referred to as the nominal inlet conditions. These concentrations are representative of actual emissions
39
-------
from typical coal combustion facilities. Initial N,0 concentrations were inherent to the NO supply gas
and varied with each individual cylinder.
The initial tests were concerned with SO-, removal efficiencies of various dry sorbents and the
best cartridge design. The SO, removal efficiency was defined, at this point in the study, as the
measured SOz exit concentration of the flue gas after the gas had passed through the sorbent system.
These SO, concentrations measured by the SO, low CEM in the sorbent/sample bomb loop, were
referred to as the SO, breakthrough data. The initial efforts also included the use of SO, and acid
color indicator. The various tests and their results are discussed in respective subsections.
3.5.2 Water Removal
As mentioned previously, P,0, was the selected desiccant based on its greater moisture
removing ability and its color indicating properties. The 120-cc refutable traps were filled with -50 g
of P,Oj and held in place by glass wool plugs. The water removal cartridge was placed, in scries,
after the S02 sorbent cartridge and before the sample bomb. The placement of the SO, cartridge was
critical because S02 must dissociate in water to form an acid and then react with the sorbent,
therefore, water was necessary to enhance the SOz neutralization process in the sorbent cartridge.
3.5.3 FGSS Shakedown Tests
The first set of tests took place during the design and construction of the FGSS. These first
screening tests were performed to ascertain if a two-cartridge solid-sorbent design was feasible.
The first screening test was performed to evaluate SO, removal by a 40:1 (by weight)
sand/Ca(OH)2 mix and to evaluate the effect of the Ca(OH), on the initial N,0 concentration.
Ca(OH)2 was initially chosen because of its proven ability to scrub SO,. This test mix was added to
an empty, gas-tight air purifier tube and then evaluated on the FGSS. A gas stream containing -1,500
ppm SO,, 600 ppm NO, and 8 percent moisture was passed through the sorbent cartridge. The SO,
concentration exiting the cartridge was measured to be less than 10 ppm. The N,0 concentrations
were measured by on-line GC/ECD upstream and downstream of the Ca(OH), cartridge. A 0.8
-------
percent difference existed between the upstream (initial) N^O concentration and the downstream N-,0
concentration. This difference was considered negligible.
The second screening test evaluated the use of an acid color indicator, methyl red, in the
sorbent mix. The methyl red was added to the sorbent mix in an attempt to possibly indicate the
expenditure of Ca(OH)2 during sampling. A 40:1 sand/Ca(OH)2 (80 g, 2 g, respectively) mix was
made and 82 mg of methyl red was then added. This mix was evaluated under the same conditions as
(he previous test. The sand mixture changed color as the test progressed, but the color change did not
occur evenly throughout the cartridge. The SO-, exit concentrations and N-,0 results were similar to
those of the previous screening test.
These first screening tests indicated that a two-cartridge solid-sorbent system may indeed be
feasible. The initial results also indicated that Ca(OH)2 had no effect on the initial N-,0 concentration.
Initial N.,0 concentration would become a critical measurement later on in the study. These initial
tests also indicated a need for a bypass loop within the FGSS to allow measurement of the flue gas
concentrations untouched by sorbents. A continuous data acquisition system was also necessary to
monitor and archive the initial and exit concentrations of the Que gas.
3.5.4 Sorbent Cartridge Design
After completing the many FGSS modifications, the next study objective was to find
commercially available gas-tight, swage-compatible cartridges to contain the S02 sorbent and the
desiccant The most desirable option was to purchase gas purifier tubes available through a variety of
vendors. Vendors were contacted and asked whether the clear traps, empty of any purifying agent,
could be obtained. One vendor was able to provide empty, 120-cc, refillable, and gas-tight cartridges.
The cartridge end caps were compatible with stainless steel O-ring sealed straight thread to
compression fitting connectors. These were added to the cartridge to ensure a leak-tight fitting and a
1/4-in compression fitting compatibility. These traps were used to contain both the SO, sorbent and
the desiccant
-------
A test was performed to evaluate the SO, removal efficiency of a 20:1 sand to Ca(OH)2 mix
using a slurry method of mixing then drying. The test mix was added to the 120-cc traps and was
held in place by glass wool plugs. The S02 sorbent and the desiccant cartridges were placed in the
sorbent/sample bomb loop and evaluated under nominal inlet conditions. Figure 3-3 is a plot of the
nominal inlet conditions. A comparison of the initial SO, concentration and the SO, exit
concentrations showed that the test mix was still removing 98 percent of the S02 after 20 min. This
test verified the suitable operation of the FGSS and the sorbent cartridge sampling system. The
cartridges were selected to be used in further testing.
3.5.5 Dispersion Tubes
The first few scoping tests gave insight into the sorbent system operation and enhancement
During these tests, there was concern that the contact between the flue gas and the reactant was less
than optimal. To alleviate this concern, dispersion tubes were added to the inlet and outlet of the SO,
sorbent cartridge. The dispersion tubes are designed to "spray" the flue gas through the solid sorbent
thus maximizing the contact between gas and sorbent
Dispersion tubes ate made from 1/4-in Teflon tubing, 6 in long, with a compression fitting
stainless steel nut, ferrules and cap on the end (Figure 3-4). About 70 0.6-mm holes are drilled in a 2-
in section behind the nnt and cap. The collective area of the 70 holes is greater than the inner annular
area of the Teflon tubing. These tubes are used at both the inlet and outlet of the SO-, sorbent
cartridge.
A test was performed to evaluate the effect of the dispersion tubes on the SO, removal
efficiency of the sorbent cartridge. The S02 breakthrough data were compared to a previous test
where the tubes were not used (Figure 3-5). After 20 min, the dispersion tube cartridge bad minimal
S02 breakthrough (2 ppm) compared to the cartridge without tubes (25 ppm). Both tests used a 20:1
sand:Ca(OH)2 slurry mix. Because of SO., removal efficiency enhancement, the dispersion tubes
became a permanent part of the SOz sorbent cartridge design.
-------
Nomina] Inlet Concentrations
_ 120 cc cartridge
20:1 sand/Cj(OH)2
A
' X
«.
X
V
\
\
\
\
/\ s
X \
"i •
• , t ¦ • 1 * i * ¦ ¦ . . * ^ ¦ i ' ? -1. >. » r *¦*_.<—L^yi——1—9—L ~
T . . . ri 1 W . 1 j>T r FM 1 M ¦l"!"!' 1 f ¥ -» - * • (T
" / l+I
: r—x / \ ?
• / \ ¦
\
1114 : • »< < > m »V
^ Bomb Sample Taken
» s •.
„ \
1 1 1_ ! 1 I 1 1 ! 1 1 1 I I 1 1
12
16
20
24
28
S02INIT (XI00)
Time (miowes)
+ NOx (*100)
S02 EXIT CONG
Figure 3-3. S02 removal with 20:1 sand/CaCOH^soibent.
43
-------
1/4-in. ceramic-filled Teflon ferrule
120cc refillable trap
1/4-in. nut
1/4-in. compression fitting
7/16-20-in.
straight
thread
mrfjmdgj
£
1/4-in. ceramlc-fillcd
Teflon feirule
stainless steel
O-ring seal connector
S02 sorbent cartridge components. cntl "P
dispersion tube (w/0.0225-in. holes) ,/4.in slain|css s)ec| fcfru|cs
mk
!!!!!!!!!!!!!!!!
¦g j/4-in.cap
^^^1/4-in. nut
l-ring seal end cap (O-ring inside end cap)
Dispersion tube hardware.
&
e
Complete S02 sorbent cartridge.
Figure 3-4. S02sorbont cartridge assembly.
-------
¦U
1/1
Valve turned lo bypass loop
and ihen sorbent loop
1.5 3.0 4.5 6.0 7.5 9.0 10.5 12.0 13.5 15.0 16.5 18.0 19.5 21.0 22.5 24.0 25.5 27.0 28.5 30.0
Time (min)
Without Tubes
With Tubes
Figure 3-5. Effect of dispeision tubes on S02 removal efficiency.
-------
3.5.6 SO-, Color Indicator
To investigate alternative SO-, sorbents, considerable effort was made contacting vendors of
commercial SO2 air sampling tubes to inquire about the chemicals used in the indicating exposure
tubes. The contents of these sampling tubes were appealing both from a standpoint of reactive
specificity as well as quantitative indicating properties. The possibility of purchasing indicating
sorbents in bulk form was explored. Unfortunately, the identity of the indicating reagents was
proprietary. In addition, none of the materials were available in bulk form, some due to their toxicity.
Commercially available SO-, color indicating and sampling tubes were investigated with the
possibility that they would indicate the expenditure of the Ca(OH); and/or divulge any major SO,
breakthrough during field sampling. These tubes indicate the concentration of S02 in air through a
linear color change along the length of cartridge proportional to volume and concentration. The tubes
were tested along with a 10:1 sand/Ca(OH), slurry mix The air sampling tube was placed after the
P,Os cartridge to ensure dry gas was entering the air sampling tube. The results of the screening test
indicated a large pressure drop, approximately 8 psig, across the sorbent system due to the relative
small size (-6 mm diameter) of the air sampling tubes.
Also during testing, the color indicator along the tube did not change in a linear manner. The
color change started in the middle of the cartridge with the beginning of the cartridge never changing
color during the test. No further work was conducted on the SO-, color indicators.
3.5.7 Chemical Sorbent Screening
As previously mentioned, Ca(OH)2 was initially used in the study because of its proven ability
to remove S02. To validate the use of Ca(OH)2, scoping tests were performed to compare the length
of time that three selected dry sorbent materials, Ca(OH)2, NaOH, and Na,C03, were effective in^
removing S02. NaOH and Na2C03 were selected because of their similar basic natures. Each
chemical was mixed with sand at a 20:1 sand-to-reactant ratio using a 20-30 mesh (0.8S-.60 mm)
Ottawa sand. The chemicals and sand were mixed using a slurry method with dcionized water.
-------
*
The appropriate amounts of sand and sorbent were weighed and added to a large pan along
with 50 mL of deionized water. The reagents were then mixed by hand with a putty knife for about
15 min. The pan and sorbent mix was then placed in an oven at 105 °C to dry overnight. A sluny
method was chosen on the assumption that the reactants would coat the sand particles thus creating a
larger reactive surface area and greater scrubbing efficiency.
The Na,C03 mix exhibited difficulties during sorbent mix preparation. The Na2C03 dried in
clumps and did not disperse through the sand. The Na,C03 was not evaluated owing to these
problems in the sorbent preparation. The NaOH mixture dried to form a brick, which subsequently
bad to be broken up before addition to the cartridges. There were no problems encountered during the
sand/Ca(OH)2 preparation.
The NaOH and Ca(OH)., mixtures were added to the sorbent cartridges (with dispersion tubes)
and placed in the sorbent position of the FGSS. They were then evaluated at nominal inlet conditions.
Hie S02 exit concentrations were measured vs. time using a 10 ppm S02 breakthrough as the
threshold.
Figure 3-6 shows that the sand/Ca(OH)2 mix lasted much longer (-45 min) than the Na(OH)
mix (—13 min). Also, the reaction of NaOH with moisture/SO, during testing was found to be very
exothermic thus causing a safety concern. The sand/Ca(OH), mix was chosen for use in additional
studies because of its longer S02 removing capabilities and its ease in preparation.
3.6 S02 SORBENT OPTIMIZATION
3.6.1 Introduction
Once the S02 sorbent cartridge design and chemical sorbent had been selected, the next
priority of the study was to enhance the S02 removal capabilities of the sandZCa(OH)-, mix. Tests
were designed to determine an optimum sand/Ca(OH)2 mix that would consistently minimize the N;0
generation to less than 10 ppm. This determination was accomplished by varying the sand particle
47
-------
J I I I I L-
0 10 20 30 40 50
Time (min)
¦ 20:1 NaOH + 20:1 Ca(OH)2
Figure 3-6. Comparison of candidate sorbents and SO^ removal capability.
-------
size, the sand-to-reactant ratio, sorbeni volume, and the sorbent mix preparation. All tests were
conducted under the nominal inlet conditions.
The previous screening tests defined the SO, removal efficiency of a sorbent by the length of
time it was efficient in removing SO^ This was achieved by measuring the SO-, exit concentrations of
the sorbent cartridges and plotting these concentrations vs time. The next tests were designed 10
further evaluate the SO, removal efficiency by measuring the effect of the sorbents on the
minimization of the NjO generation artifact in sample containers.
The N,0 generation measurement was defined as the difference between the initial N20
concentration obtained from the bypass loop during testing and the actual N-,0 concentration found in
the aged sample containers. The SO-, breakthrough data were used to quickly determine the SO-,
sorbent and the cartridge design. The study now focused on N-,0 generation in sample containers.
The SO, breakthrough data were still used as a variable in the decision making process. Each test and
its results are discussed in the subsequent subsections.
3.6.2 Sand Particle Size
As previously mentioned, the Ca(OH)2 was dispersed through sand to increase the reactant's
usable surface area. It was then theorized that with the same cartridge volume, a decrease in the sand
particle size would allow an increase in reactive surface area. Along with the increase of reactive
surface area, there was also the possibility that the flow rate and pressure drop through the sorbent
system could be compromised. The objective was to find a sand particle size that would increase the
chemical's usable reactive surface area but not effect the flow rate and pressure drop through the
sorbent system. Tests evaluated and compared the effect of three different sand particle sizes on the
flow rate and pressure drop through the sorbent cartridges and also on the mix uniformity. Each
particle size was mixed 2t a 20:1 ratio with CaCOH)^ using the slurry method. Each mix was then
added to a sorbent cartridge and tested on the FGSS.
49
-------
The first sand evaluated was a commercially available "play" sand. The play sand was sifted
to more than 18 mesh particle size (particle diameter > 1 mm). To collect the > 1 mm particle size
sand, an 18-mesh sieve was placed on a shaker. The play sand was added and sifted. The sand that
did not go through the sieve was collected for testing.
A slurried 20:1 mix was prepared with the > 1 mm sand. The sand and Ca(OH), did not
"mix" well. The larger sand particles caused the Ca(OH)-> to form in clumps or pockets when the
mixture was added to the cartridge. The > 1 mm sand/Ca(OH)2 allowed a flow rate of > 7 L/min and
a minimal pressure drop across the sorbent cartridges. Although this sand did not compromise the
flow rate and pressure drop, the larger sand panicles did compromise the mix homogeneity.
The second sand evaluated was a commercially available Ottawa sand (20-30 mesh). The sand
mixed well with the Ca(OH)-,. The fine particle sand allowed a flow rate of 4 LVmin, and caused a
large pressure drop (7 psi) across the cartridges. Sand particles were also found in the dispersion tube
holes.
The third sand evaluated was the commercially available "play" sand sifted to 18-20 mesh
(particle diameter = 0.85-1.00 mm). To collect the 0.85-1.00 mm sand, an 18- and 20-mesh sieve
were placed on a shaker with the 18-nicsh sieve on top. The play sand was added and sifted. The
sand that went through the 18-mesh sieve but not through the 20-mesh sieve was collected for testing.
There were no difficulties encountered with the preparation of a 20:1 mix using the 0.85-1.00
mm particle-size sand. This mix compared to the 20:1 Ottawa sand mix, gave a greater flow rate and
a reduced pressure drop (6 L/min, 4 psi, respectively). Because of the enhanced flow rate, the 18-20
mesh sand was the choice for further studies.
3.63 Sand-to-Reactant Ratio
With a goal to further optimize the Ca(OH)2 cartridge, the Ca(OH)2 concentration in the
sorbent cartridge was doubled from a 20:1 sand:Ca(OH)-, mix to a 10:1 sand:Ca(OH)-, mix. Tests
SO
-------
were performed evaluating the effect of each mix on the SO-, breakthrough of each cartridge and the
effect on N-,0 generation in bomb samples.
The 20:1 mix and the 10:1 mix were prepared using the slurry method. Each mix was added
to the sorbent cartridges and evaluated under nominal inlet conditions. Each test evaluated a single
sorbent cartridge with the collection of one bomb at 7, 19, and 31 min after the sorbent was placed in
the sample loop. A total of three bombs were collected for each cartridge tested. Each test mix was
performed in triplicate with a total of nine bomb samples collected per mix. During two of the 10:1
mix tests, a fourth bomb sample was collected at the 45-min interval.
The initial N20 concentration was established from the bypass loop before the placement of
the sorbent system in the sorbent/bomb loop. N20 measurements were also taken at the exit of the
sorbent/sample bomb loop during the collection of each bomb sample. This measurement was taken to
ascertain the effect of the cartridge system on initial N,0 concentration. Each bomb was then sealed
and stored at room temperature.
The bombs collected with the 20:1 mix were aged 3 days and then analyzed for N-,0. The
bombs collected with the 10:1 mix were aged 4 days and then analyzed for N20. Table 3-2 lists the
conditions and results of the individual test cartridges and its replicate bomb samples. The SO, exit
concentrations were not available because of an analyzer failure. Figure 3-7 graphically presents the
N20 generation of the bomb samples at each sample interval. N20 generat ion is defined as the
measured bomb concentration minus the initial (on-line) N-,0 concentration. A comparison of these
results revealed that the 10:1 sand:Ca(OH)2 mix was consistent in minimizing the N-,0 generation to <
5 ppm even after 45 min, whereas the 20:1 mix showed significant N20 generation in the bomb
samples after 31 min. The 10:1 sand-to-reactant ratio was chosen for use in further studies owing to
its consistency in minimizing the generation artifact
51
-------
TABLE 3-2. Np GENERATION IN GRAB SAMPLES WHILE OPTIMIZING SO2 SORBENTS
Sorbent Conditions and Tests Results
Sorbent
Conditions
No. Days
Aged
Initial N20
Concentration
Bomb 1
Generation
Bomb 2
Generation
Bomb 3
Generation
Bomb 4
Generation
Avg Cart
Generation
1A
20:1
3
1.4
2.0
5.1
22.4
-
9.8
IB
20:1
3
1.4
1.8
3.1
9.9
—
4.9
1C
20:1
3
1.4
5.5
0.1
2.4
—
2.7
2A
10:1 Sluny
4
0.4
1.5
2.8
1.9
13
1.9
2B
10:1 Sluny
4
0.5
23
3 2
23
IS
23
2C
10:1 Sluny
4
OS
4.1
3.8
3.4
—
3.8
3A
10:1 Dry
5
0.4
0.4
1.2
65
142,
5.6
3B
10:1 Dry
5
0.4
22.
9.0
0.9
1.7
33
3C
10:1 Dry
5
0.4
0.7
0.8
1.0
12.
0.9
4A
10:1 (200 CC)
5
0.5
2.1
8.1
-
-
5.1
4B
10:1 (200 CC)
5
0.5
2.7
2.0
-
-
2.4
Concentrations in ppm
52
-------
TEST NO. 1; 20:1 SAND:Ca(OH)2
N20 Generation per Bomb Sample
25
7 mill 19 min 31 min Carl Avg
Tine since sorbent placed in-line
¦ Cartridge A ~ Cartridge B ~ Cartridge C
TEST NO. 2; 10:1 SAND:Ca(OH)2
N20 Generation per Bomb Sample
7 sua
19 mm
45 min
31 min
Time since sorbent placed in-line
Nominil Inlet ConStiaQa
I Cartridge A ~ Cartridge B ~ Cartridge C
Figure 3-7. Comparison sand-to-reactant ratios on N20 generation.
53
-------
3.6.4 Sand/Sorbent Preparation
The slurry method of mixing the sand and Ca(OH)2 was chosen or the assumption that the
water would cause a more uniform dispersion of the Ca(OH), through the sand. This method was
found to be time consuming because of the overnight drying process. A dry mix of 10:1
sand:Ca(OH)^ was prepared and evaluated to ascertain whether the slurry process was necessary when
dispersing the Ca(OH)2 through sand. The mix was prepared by weighing out the appropriate amounts
of sand and Ca(OH)2 and adding them to a pan. The two solids were then mixed by hand using a
puny knife for approximately 15 min. The mix was then added to the sorbent cartridges.
Test 3 evaluated one sorbent cartridge with the collection of one bomb at 7, 19, 31, and 45
min after the sorbent system had been placed in the sorbent loop for a total of four bombs collected
per cartridge. The test is performed in triplicate. During testing, the SO-, exit concentrations and
initial NzO concentrations were monitored. These results were compared to Test 2 where a 10:1 mix
using the si airy method was evaluated.
Table 3-2 lists the results and conditions for each test Figure 3-8 graphically compares the
N20 generation of the bomb samples in the replicate tests. From these results, the slurry process
demonstrated a slight performance advantage over the dry mix. The resulting N20 generation was still
much less than 10 ppm. Although the dry mix exhibited a higher generation in the bombs, the dry
method of preparation was chosen because of the short preparatory time. The excessive preparatory
time for the slurried sorbent material was also hindering the progress of scoping tests.
3.6.5 Sorbent Volume
The sorbent volume was the last variable examined during the optimization of the SO, sorbent
cartridge. The 120-cc cartridges contained about 160 g of sand and sorbent. To approximately double
the sorbent volume, 200-cc refillable traps were filled with the 10:1 sand/Ca(OH)-, dry mix and
evaluated (Test 4). Table 3-2 shows that the SO, exit concentrations were similar to those observed
using shorter cartridges (< 15 ppm after 40 min). Figure 3-9 shows that the NzO generation results
-------
TEST NO. 2; 10:1 SAND'.Ca(OH)2 SLURRY
N20 Generation per Bomb Sample
25
E 20
5
Cart Avg
0
31 min
45 min
7 mm
19 min
Time since sorbent placed in-line
¦ Cartridge A ~ Cartridge B £3 Cartridge C
Nominal tain Coedirieni
7 min
TEST NO. 3; 10:1 SAND:Ca(OH)2 DRY
N20 Generation per Bomb Sample
ESi |o]et CsuUgi
19 GQiii 31 min 45 min
Time since sorbenz placed in-line
¦ Cartridge A ~ Cartridge B Q Cartridge C
Cart Avg
Figure 3-8. Effect of sorbent preparation process on N-,0 generation.
55
-------
TEST NO. 3; 10:1 SAND:Ca(OH)2 (120 CC)
N20 Generation per Bomb Sample
n
M , M
1
Normal! InlaCflDdiliou
7 min 19 mio 31 min 45 mill
lime since sorbeni placed in-line
¦ Cartridge A ~ Cartridge B ~ Cartridge C
Cart Avg
TEST NO. 4; 10:1 SAND:Ca(OH)2 (200 CC)
N20 Generation per Bomb Sample
25
1* 20 -
45 min 75 min Cart Avg
Tune since sorbeut placed in-line
¦ Cartridge A Q Cartridge B
Ndannii IalerCoodfrtast
Figure 3-9. Effect of sorbeat volume on N20 generation.
56
-------
were also similar (5 days, < 5 ppm). The smaller volume was chosen because of the ease of handling
and concerns over waste generation.
3.7 SAMPLE CONTAINER OPTIMIZATION
3.7.1 Introduction
Concern over N20 generation inconsistencies between bomb samples led to improving
methods to clean and condition the sample bombs. At this point, there was not a clear understanding
of the reaction mechanisms between SO,, NO^ and H,0 that generated N20. Most of the proposed
mechanisms hypothesized a liquid-phase wall reaction involving these three reactants. If this was
indeed true the inconsistencies between replicate bomb samples could in part be caused by the inside
walls of the sample bombs which were not consistent from container to container. The inconsistencies
could also be caused by residual S02 and moisture in the bomb itself. Efforts then concentrated on
creating greater uniformity between sample containers.
3.7.2 Sample Bomb Preparation
Initially, the 600 cc stainless steel sample containers went through a S02 neutralization
process. The bombs were then washed, dried, and stored under vacuum. A method developed ensured
a clean, dry, and pressurized bomb. The conditioning consisted of a hot, soapy water soak, a
deionized water rinse, and a methanol rinse. The rinsed bombs were oven dried at 105 °C for 12 h.
The hot bombs were sealed and cooled. Dry nitrogen was used to purge and pressurize the cooled
bombs.
3.73 Teflon Coated Sample Bombs
To create greater uniformity between sample containers, the inner surfaces of several sample
bombs were Teflon coated. A test using the Teflon-coated bombs with the sorbentfoomb system
revealed little effect on the N,0 generation artifact (Figure 3-10). Both the stainless steel bombs and
the Teflon-coated bombs had less than 5 ppm N-,0 generation.
57
-------
TEFLON COATED BOMBS
N20 Gcootoob per Bomb Sample
?mia
Nominal Inlet Coaditioas
10:1 ood:Ci(OH)2 dry
24 bos <1 can
TIME (race jorbeci pUoed in-liae)
Cirt Avg
RUN NO. 7A TEFLON BOMBS
10:1 »ad.~0(OH)2 NOT SLURRY
14 -
13 -
"¦ ! —
30
TIME (min)
¦ SO2HI(X10a) SQ2LO O NO (XI DO)
Figure 3-10. Effect of Teflon coating on N,0 generation.
58
-------
Although this test showed no visible performance enhancement from Teflon coating, a
decision was made to Teflon coat the inner surfaces of all sample bombs. It was believed the Teflon
coating would prolong the life of the stainless steel container, lessen the chance of S02 residue build-
up along the container walls, and create a more inert sample contact surface.
3.8 COMBUSTION SOURCE GRAB SAMPLE METHOD EVALUATIONS
3.8.1 Introduction
At this point in the study, the sorbent/bomb system consisted of a two-cartridge, solid-sorbent
system (-160 g 10:1 sand:Ca(OH)>, dry; -50 g P205) and a clean, dry, and pressurized 600-cc
stainless steel sample container equipped with toggle valves and a side port for syringe injections (see
Figure 3-11). The combination of reagents and equipment had demonstrated acceptable performance
wbSe incorporated into the FGSS. The focus now shifted to evaluating the sampling method as it
would be used on actual combustion sources. This required the consideration of appropriate sampling
equipment and sampling configuration. Emphasis was placed on the ability to incorporate the
sampling equipment into conventional CEM sample delivery/conditioning systems.
3.8.2 Moisture Removing Devices
Realizing that many sampling systems use moisture removing devices, particularly refrigeration
condensers, a test was conducted to monitor the effect of such a device on the performance of the
sampling method. A Hankason refrigeration dryer was placed in the FGSS sampling system between
the FGSS atmospheric dump and the sample pump of the nominal grab sampling system. The test
conducted under the nominal inlet concentrations, monitored the effects of the dryer on the SO,
removal efficiency of the calcium hydroxide. Two bomb samples were also collected. After 20 min,
the S02 exit concentrations approached inlet conditions (1,200 ppm). Analysis of the bomb samples
after 7 days exhibited N,0 generation of 10 ppm (Figure 3-12). The high SO-, exit concentrations
verified the importance of moisture (in the sample gas) on the neutralization of SO, in the sorbeni
59
-------
s
Stainless sleel
toggle valve
600 cc sample container
with septum port
1/4-in. stainless steel nut
—
8/32-in. washer
d
3/8-in. OD septum
O
_]/4-in. compression
fitting
fl
1/4-in. NPT 1/4-in. NPT
(female) (male)
Figure 3-11. Sample container schematic.
-------
1,400
1,200
o 1,000
0
«£»
d 800
0
1
€ 600
8
a
U 400
200
0
REFRIGERATION CONDENSER
CEM daU and N20 generation data
/ *
/ *
5.0
Nominal Inlet Conditions grn ui qq2 i0
10:1 sand:Ca(OH)2
10.0
Time (min)
NOx
\ /
w-
/ :
/
/
i
15.0
N20 Generation (xlO)
20.0
Figure 3-12. Effect of moisture removal on SQ^ sorbcnl performance.
-------
cartridge. Because of the relatively high N,0 generation, sampling downstream of any moisture
conditioning devices was no longer considered.
3.83 Evaluation of Source Sampling Configuration
Hie grab sampling configuration of the FGSS was designed such that a large number of
method performance parameters could be monitored or measured concurrently. This tended to make
the sampling system and equipment more complicated than necessary. To make the sampling system
more compatible with conventional CEM sample delivery/conditioning systems, the system needed to
be simplified. Essentially, all that was required was a means to extract a representative portion of a
flue gas from the CEM sample system and push it through the grab sampling system. This could be
accomplished with a small vacuum pump. Hie grab sample could be obtained in parallel to the CEM
sample at a location between the stack and any CEM mo is tu ring conditioning devices (see Figure
3-13).
To accomplish this, a sampling system separate from the normal FCSS sampling system was
installed. A representative portion of the Due gas mix was pulled from the atmospheric dump section
of the FGSS system and subsequently pushed through the S02 sorbent, the H20 sorbent, and then
directly into the sample container (see Figure 3-14). The gaseous sample was still obtained from the
atmospheric dump (vent) of the FGSS. A toggle valve was located between the atmospheric dump
and sample pump to isolate the sample delivery systems during sampling. The toggle valve was
connected to the inlet of the sample pump by a 2-ft section of 1/4-in OD Teflon tubing. The sorbent
cartridges and sample bomb were located at the outlet (positive pressure) side of the pump.
A rotameter was placed at the outlet of the sample container to measure the flow rate through
the sorbent cartridge/sample bomb system. The gas stream exiting the rotameter was occasionally
used to measure initial N20 concentrations. Inlet NO, SO-,, and N,0 concentrations were monitored
from nominal locations in the FGSS (bypass loop position). Outlet (exit) SO-, concentrations were no
longer measured.
-------
Suck
To CEMs, dryeis, etc.
Flow
Sample
pomp
Figure 3-13. Location of sampling system.
63
-------
Inlet
Oullcl>
From slack
8
S02 sorbcnl
*
]> *c
>
H20 sorbent
Flow
Sample container meter
Sample pump
Figure 3-14. Sorbent/sample container schematic.
-------
Following design and installation of the sampling system, a test was designed to evaluate its
performance relative to tests conducted to date. The nominal test condition simulated flue gas
concentrations were used for this test. Several "wet" grab samples were collected without the use of
any of the gas conditioning sorbents to demonstrate the full extent of N-,0 generation.
Once an on-line N20 measurement was taken in the bypass loop of the sampling system, the
toggle valve between the atmospheric dump and the grab sample pump was opened. The flue gas was
allowed to run through the sorbents cartridges. Three separate bomb samples were collected. For
each sample collected, the conditioned gas stream flowed through the sample bomb for at least 2 min.
At the flow rate measured through the bomb, 9 L/min, a 2-min purge time ensured that a
representative sample was collected. At least seven volume exchanges of the sample container took
place. Also during the 2-min sample period, an on-line N,0 measurement was taken at the bomb exit
to again verify that the sorbents had no effect on initial N20 concentration. Three samples were also
collected without the use of any sorbent cartridges at 11 L/min for 2 min each.
The samples collected using the sorbent cartridges demonstrated N-,0 generation less than 1
ppm when analyzed after 6 days (Figure 3-15). The "wet" samples revealed N,0 generation of more
than ISO ppm when analyzed after the same period. Because of the success of this test, the sampling
configuration used was deemed acceptable for further testing.
3.8.4 EPA's Innovative Furnace Reactor
The next logical step in the sampling method evaluation process was the application of the
sorbent cartridge sampling system to actual combustion situations containing similar flue gas
constituents and concentrations. As a result, the sampling method was evaluated further on the
EPA/AEERL-GCTB 15 kW (50,000 Btu/h) Innovative Furnace Reactor (IFR). The grab sample
method was evaluated while the furnace fired on Illinois No. 2 coaL Samples were obtained by
tapping into the existing CEM sampling system in parallel so that continuous CEM data could also be
65
-------
FLUE GAS SIMULATION SYSTEM
N20 ARTIFACT GENERATION OF SAMPLES COLLECTED
3
165 ppm
0.8 ppm
0.5 ppm
SAMPLE BOMBS AGED 6 DAYS
Sj WITHOUT SORBENTS
SAMPLE FLOW = 11 LPM
~
WITH SORBENTS (X10)
SAMPLE FLOW = 9 LPM
ON-LINE N20 CONC.
(X10)
SORBENT(10:1 SAND, Ca(OH)2 AND P203)
1,200 ppm S02,600 ppm NO, 10% MOISTURB
N20 INITIAL = 0,5 ppm
Figure 3-15. Comparison of tyO sample container generation
with and without use of sorbcnls.
-------
collected (see Figure 3-13). On-liae N20 measurements were also obtained using the automated, on-
line GC/ECD system.
Figure 3-16 represents data from tests designed to evaluate the sampling method configuration
on an actual fossil fuel combustion source. The (list test used four sets of sorbent cartridges with a
collection of three sample bombs per cartridge for a total of 12 bombs collected. The sample flow
through the cartridge system was 2 L/min. Three bombs without sorbent cartridges were also
collected. The average flue gas concentrations were 416 ppm NO and 1,900 ppm S02. The moisture
content was not measured but estimated lo be roughly 8 percent by volume. The grab samples when
analyzed after 7 days showed excellent agreement with the on-line N-,0 measurements. The "wet"
samples were an order-of-magnitude larger, illustrating the N,0 generation artifact.
An additional test was conducted to further evaluate the effect of moisture removal on the
collection of samples. Six samples were collected using two sorbent systems upstream of the
condenser. Three samples were collected using one sorbent system downstream of the condenser.
Three "wet" samples were taken both upstream and downstream. The sorbent sampling flow rates
were also varied during upstream sampling. Figure 3-17 shows that minimal N20 generation from
within the sample containers when analyzed after 5 days. The flow rates and refrigeration condensers
appear to have had negligible effect on the N20 generation. However, given the results of the
moisture conditioning tests conducted on the FGSS, the decision was made to remain sampling
upstream of all moisture conditioning devices.
3.9 WORST CASE SCENARIO TESTS
3.9.1 Introduction
The tests conducted up to this point in the study confirmed that the development of a grab
sampling method that minimized the NzO generation artifact to the extent that meaningful N20
measurements could be obtained was indeed possible. Under the nominal simulated flue gas
concentrations tested (-600 ppm NO; -1,200 ppm SO,; and -10 percent moisture) an unslunied 10:1
-------
ILLINOIS NO. 2 COAL—INNOVATIVE FURNACE REACTOR
N20 ARTIFACT GENERATION OF SAMPLES COLLECTED
WITH AND WITH OUT SORBENTS
NO SORBENTS
-
v* , * * / * 3 v
.
y* " ,
WITH AUTOMATED
J /
* r
SORBENTS ON-LINE GC/ECD
V*,, »
-
***? 1 v ~
BOMBS AGED FOR 7 D AYS
12 BOMB AVERAGE
WITH SORBENTS
7 INJECTION AVERAGE
ON-LINE (GC/ECD)
3 BOMB AVERAGE
WITHOUT SORBENTS
SORBENTS (10:1 SAND, Ca(OH)2 AND P205)
AVO NO = 416 ppm, CO = 166 ppm, C02 n 14.3ft
AVO 02 = 4.9%, S02 = 1900 ppm, MOISTURE UNKNOWN
LOAD - 15 kW (50,000 Btu/h)
Figure 3-16. Performance of sampling method on actual combustion source.
-------
ON-LINE N20 CONC. vs. AGED GRAB SAMPLE N20 CONC.
ILLINOIS NO. 2 COAL
INNOVATIVE FURNACE REACTOR LOAD = 15 kW (50,000 Btu/h)
UPSTREAM (before llankison dryer) DOWNSTREAM (after Hankison dryer)
SAMPLE njOW RA11! I
SAMPLE PLOW KATC
SAMPLE FLOW RATE
ILPM
jtfMlVlbVltW
1LPH
NOSORBENTS SORBBNTS SORBBNTS NOSORBENTS SORBF.NTS
¦ ON-LINE N20 ANALYSIS [H§1 AGED BOMB N20
FROM BX1T OP BOMB HO ANALYSIS (5 DAYS)
SORBBNTS (10:1 SAND, Oi(OM)2 and P205)
AVO NO - 425 ppm, CO • 91 ppo, C02 - 12.7%
AVO 02 • 7.8*, S02 - 1,620 ppm, MOISTURE UNKNOWN
Figure 3-17. Comparison of effects of sample location on NjO generation.
-------
mix of sand/calcium hydroxide coupled with P205 was shown to be effective in minimizing the N20
generation artifact in sample containers to less than 10 ppm when samples were analyzed within 1-2
weeks of collection. In addition, the sampling equipment and configuration developed was relatively
simple to use and easily incorporated into conventional CEM sample delivery/conditioning systems.
Although the sorfoent cartridge system had been evaluated under representative, controlled, and
realistic conditions, the full extreme of potential Que gas concentrations that may be encountered had
not been evaluated. The minimum moisture content critical to quantitative SO-, neutralization had not
been determined. Similarly, higher SO, and NO concentrations had yet to be investigated.
It was then determined that the next major step in the method development process was to
determine the range of NO, SO,, and H,0 concentrations under which the sorbent system can operate
effectively. An emphasis was put on testing the worst case conditions to determine if high
concentrations of SO-, and NO or low moisture content would increase the potential for N-,0
generation within the sample container. On the other extreme, low flue gas concentrations of S02 and
NO were considered inconsequential and were therefore not considered for testing. Similarly, since
the samples were intended to be collected upstream of any pollution control equipment, including wet
scrubbers, high moisture concentrations were considered favorable and were also not considered for
testing.
The worst case scenario was defined to be the maximum range of the key flue gas
concentrations that would be found in the field. The worst case conditions were defined as 2,500 ppm
SO,, 900 ppm NO, and 5 percent H,0.
70
-------
3.9.2 Nominal Inlet Concentrations
Two tests were run at nominal inlet concentrations in order to repeat baseline conditions.
Each test evaluated three sorbent cartridge sets with the collection of three bombs per cartridge set for
a total of nine sample bombs. The initial N20 concentration was measured from (he bypass loop of
the FGSS. The GC/ECD was programmed to automatically sample for N-,0 every 8 min throughout
the test No N20 measurements were made at the bomb exits.
The results revealed inordinately low N-,0 generation when the bombs were analyzed after 5
days (Table 3-3). A comparison of Run 100 and Run 101 revealed the same low N,0 generation for
both tests (Figure 3-18). Another review of the test data revealed a linear increase in the on-line initial
N-,0 concentrations (Figure 3-19). It was then suspected that N20 was being generated within the
FGSS. Since the initial N20 concentration is dependent on the concentration of N.,0 in the NO
supply gas and the FGSS operates at "steady-state" conditions, an increase in initial N20 concentration
should not occur without changes in the measured concentrations of the other supply gases. Efforts
were focused on finding the source of this generation since a correct initial N-,0 concentration was
crucial to the study. Knowing that the formation of N20 occurs in the liquid phase, the investigation
was centered on a source of "standing" water. The source was found to be a cyclone type water
knock-out device that had been added to the system to remove water before the flue gas entered the
CEMs and GC/ECD. The device's drain had plugged and the cyclone had filled with water.
This water removal device was replaced by a small refrigeration condenser. Tests were
performed to confirm that the condenser removed water without effecting the initial flue gas N.,0
concentrations. Figure 3-20 compares the on-line N20 concentrations vs. time before and after the
addition of the condenser.
3.9.3 Woist Case Conditions
With the N20 problem solved and the verification that the FGSS was in acceptable operating
order, two tests were performed to evaluate the sorbent cartridge system at worst case conditions
71
-------
TABLE 3-3. REPEAT OF NOMINAL CONDITIONS
Sorbcnt Conditions and Tests Results
Sorbent
Conditions
No. Days
Aged
Initial N20
Concentration
Bomb 1
Generation
Bomb 2
Generation
Bomb 3
Generation
AvgCait
Generation
100A
10:1 Dry
5
0.540
0.235
0.968
0.196
0.466
100B
10:1 Dry
5
0.610
3302
1.710
2.874
2.630
100C
10:1 Dry
5
0.750
0.651
0.566
0.157
0.458
101A
10:1 Dry
5
0.822
0.147
0.276
0.137
0.190
101B
10:1 Dry
5
0.966
0.734
0.000
0.845
0-526
101C
10:1 Dry
5
1.046
0.214
0.014
0.067
0.098
Concentrations in ppm unless otherwise noted
72
-------
35
NEW SAMPLING POSITION
Rub No. 100; Ncxsinai Inlet CaxStiaa»
^15
i
8 2
•Jf
e
115
8
S 1
z
05
0
10:1 nodO(OH)2
I
Bccob 1
Bomb 2 Bomb 3
Bomb Sample* per CtrTrid^e
¦ C&rvidgeA OCUnridgeB ~ CsrtridgeC
CartAvg
NEW SAMPLING POSITION
Ram Nol 101; Nanmil klet Caidiiau
¦
:'
¦ ¦
¦
HI
0.5
ra6
iw
02
10:1 sudbOt(OH)2
Bomb 2 Bc*ab3
BGabSmpUi pv (^nidge
BGnfeidge A Q Cartridge B QCmndgcC
CmAvg
Figure 3-18. NzO generation in sample containers; repeat tests of nominal conditions.
73
-------
1.1
ON-LINE INITIAL N20 CONCENTRATIONS
Runs 100 & 101
& 0.9
¦= 0.8
O 0.7
OJ
0.6
0.5
J.
X
525
1405
Tc«5 ran comecnrjvtly
Nominal Inlet CoocfifiocB
1935 2430 2825 3355
Time (seconds)
3800
4150
4670
Figure 3-19. NzO generation in the refrigeration condenser.
74
-------
ON-LINE N20 CONCENTRATION
2980
860
4400
TIME (SECS)
N20W/0 CONDENSER N20 WITH CONDENSER
Figure 1-20. Effect of addition of refrigeration condenser on NzO generation
within the sample delivery system.
75
-------
(2,500 ppm SO-v, 900 ppm NO, and 5 percent H-,0). The first test, Run 103, evaluated three sorbent-
cartridge systems using the original design of earlier tests. This design placed a dispersion tube at the
inlet and outlet of the SO, sorbent cartridge. The second test. Run 104, evaluated three modified
sorbent-cartridge systems. The modification was performed on the S02 sorbent cartridge only. The
dispersion tube was removed at the outlet of the S02 sorbent cartridge. This modification was
performed to ascertain if the removal of the dispersion tube would generate a higher flow rate and less
of a pressure drop without compromising the effectiveness of the Ca(OH)2 to remove SO-,. Also,
insertion of the second dispersion tube into the dry sorbent was extremely difficult Elimination of the
second dispersion tube would make S02 sorbent cartridge assembly much easier.
Three bombs were collected with each cartridge system with a total of nine bombs per run.
The bombs were aged 7 days and then analyzed for N20. The average artifact N20 generation for
both runs appeared to be minimal (Table 3-4). The average artifact N,0 generation was observed to
be greater with the modified cartridge and higher flow rate (Figure 3-21). The higher average N-,0
generation for Run 104 may be due to the SO-, sorbent packing. During testing, the sorbent appeareo
"loose;" thus, the actual contact between the flue gas and S02 sorbent may have been less than
optimum. Run 10j, with a lower N20 generation and lower flow rate, appears to operate consistently
from cartridge to cartridge. A decision was made to continue using two dispersion tubes in the SO-,
sorbent cartridge.
The results of the worst case scenario tests, performed under controlled conditions, indicate
that the sampling configuration used is capable of controlling N20 generation in the sample container
to acceptable levels. The final performance evaluation would be to duplicate the worst case scenario
tests under actual combustion process conditions.
3.9.4 EPA's Innovative Furnace Reactor: Worst Case Conditions
A final grab sampling method performance evaluation test was conducted on the EPA's IFR
under similar worst case scenario conditions. The furnace was fired on natural gas. No SO-> or NOx
76
-------
TABLE 3-4-. Np GENERATION UNDER WORST CASE CONDITIONS
Socbent Conditions and Tests Results
Soibent
Conditions
No. Days
Aged
Initial N20
Concentration
Bomb 1
Generation
Bomb 2
Generation
Bomb 3
Generation
AvgCart
Generation
103A
2 Disp Tubes
7
0.75
0.80
0.91
0.61
0.77
103B
2 Disp Tubes
7
0.96
0.63
0.94
0.96
0.84
103C
2 Disp Tubes
7
Z14
0.17
0.42
0.80
0.46
104A
1 Disp Tube
7
1.2Q
0.00
0.00
0.75
0.25
104B
1 Disp Tube
7
0.80
1.19
1.18
1.18
1.18
KMC
1 Disp Tube
7
0.84
0.90
0.99
234
1.47
Concentrations ill ppm unless noted S02 = 2^00 ppm, NO = 900 ppm, ICO = 5% vfv
77
-------
Run No. 103; 2 Dispersion tubes
3
25
Bomb 1 Bomb 2 Bomb 3 Cart Avg
Boob Samples per Cartridge
¦ Cartridge A ~ Cartridge B ~ Cartridge C
Run Ho. 104; 1 Dispersion tubes
Bomb 1 Bomb 2 Bomb 3
Bomb Samples per Cartridge
¦ Cartridge A ~ Cartridge B 0 Cartridge C
Cart Avg
Figure 3-21. N20 generation under worst case conditions.
78
-------
pollution control devices or techniques were used during these tests. The concentrations of the key
flue gas constituents measured during the worst case testing were approximately 3,200 ppm SO-, and
1,500 ppm NO. Moisture was not measured but estimated to be approximately 10 percent by volume.
The high levels of SO-, and NO were obtained by doping the combustor with S02 and ammonia,
respectively.
The grab samples were collected from the same location as the earlier performance evaluation
tests. Specifically, the samples were obtained at a location in the CEM sample delivery system
upstream of the moisture conditioning unit Three sorbent cartridge systems were evaluated with a
collection of three bombs per cartridge. The initial N20 concentration was determined by the on-line
GC/ECD which also sampled from the same location in the CEM system. N20 measurements were
made as closely to the time of grab sampling as possible.
Hie bomb samples were aged for 8 days and then analyzed for N20. Table 3-5 lists the
conditions and results of the three tests. All three tests resulted in the generation of N20 within the
aged bomb sample. The average N20 generation ranged from 3-7 ppm. One of the nine bomb
samples did exhibit generation greater than 10 ppm (11.2 ppm).
Because the data were acceptable for this final performance evaluation test, the grab sampling
method was deemed suitable for the screening of high N20-emitting fossil fuel combustion sources.
3.10 FURTHER METHOD EVALUATION: SNCR TESTS
Although this procedure was designed and tested for use with flue gases from conventional
combustion sources without the application of any pollution control techniques or devices, an
opportunity was presented to evaluate the grab sampling method on a NO, control technique. The
control technology employed, SNCR, uses additive reactants such as urea and cyanuric acid, injected
in the post-combustion zone, to control NOx emissions. This technology, however, has the potential
to increase NzO emissions. During this particular test, a proprietary reagent, hereafter referred to as
NOx-OUT, was evaluated.
-------
TABLE 3-5. EVALUATION OF SAMPLING METHOD ON ACTUAL
COMBUSTION FACILITY UNDER WORST CASE CONDITIONS
Performed on tbe C-Wing IbK
Conditions Results
Test
CRUN5
CRUN6
CRUN7
Date
06/24/91
06/24/91
06/24/91
Fuel
Nat Gas
Nat- Gas
Nat Gas
S02ppm
3175
3200
3220
NO* ppm
1562
1537
1537
CO ppm
25
20
18
CQ2%
10.02
9.94
9.78
02 %
2.875
285
2.875
NOxOUT
no
no
no
Aged Days
8
8
8
Init N20
-52
.53
1.04
N20 Gen Bomb A
2.46
1.81
2.66
N20 Geo Bomb B
2.94
7.04
4.92
N20 Gen Bomb C
3.41
11-23
536
Avg Cart Gen
2.94
6.69
431
Avg Bomb N20 = 3 bomb avg
Concentrations in ppm unless noted
-------
Because the automated, on-line GC/ECD N20 monitoring system was on loan to GCTB to
measure N,0 emissions, it was an opportune time to evaluate the grab sampling method- Bomb
samples were collected while the NOx-OUT reagents were injected into the Down-fired Tunnel
Furnace. Four tests were performed, each on separate days. Each test consisted of the collection of
three bomb samples, which were then aged for more than 7 days and were then analyzed for N20.
Table 3-6 gives the conditions and results for the four tests.
The results of these analyses demonstrated that no N-,0 generation was observed. In fact, the
N-,0 concentration decreased from the initial, on-line N,0 concentration that was taken before the flue
gas was sampled. The average reduction from initial N20 to aged N20 concentrations was 31 percent
In an attempt to explain the N-,0 concentration decrease, the bombs were checked for leaks and found
to be under pressure. The daily GC/ECD QC checks exhibited analytical bias of less than 3 percent,
so an analytical error was also ruled out
No explanation for the decrease in N-,0 concentration in the aged bomb samples is apparent
A potential explanation may be linked to high stack concentrations of ammonia (NH3), a byproduct of
the NOx-OUT additive. It may be possible that the ammonia participates in a reversible N-,0 reaction
within the bomb because of the basic nature of NH3. The basic property alone is not enough to
explain the reduction, as sodium hydroxide, a strong base, has been used in impirger solutions to
acrob S02 from combustion flue gas samples for subsequent sampling for N20 measurement No
negative bias on grab sample analyses have been isolated. However, an important conclusion can be
drawn from this test; the grab sample method should be used on conventional combustion sources
without the application of any pollution control equipment or technique only.
3.11 GRAB SAMPLING METHOD SUMMARY
The method developed was designed so that it could be used compatibly with continuous
emission monitoring sample delivery/conditioning systems or as a stand alone procedure. Specifically,
the method developed employs the use of reactant-specific dry sorbents to remove the gaseous
-------
TABLE 3-6. Np GENERATION IN SAMPLES COLLECTED DURING NOx CONTROL TESTS
Selective Non-catalytic Reduction
Performed on the C-WIng LFR
Conditions and Results
Test
CRUN1
CRUN2
CRUN3
CRUN4
Date
06/10/91
06/11/91
06/12/91
06/17/91
Fuel
Nat. Gas
Nat Gas
Nat Gas
Fitt No. 8 Coal
SOZ ppm
725
1625
1570
100
NOx ppm
80
90
388
280
CO ppm
30
80
33
10
C02 %
7.6
7.46
7.6
8.6
02 %
7.475
7.85
7.45
10.625
NOxOUT
yes
yes
yes
yes
Aged Days
9
8
7
9
Ink.N20
38.6
52.9
33.6
8.79
N20 Gen Bomb A
(5-06)
(19.73)
(9-5)
(2.00)
N20 Gen Bomb B
(6.98)
(8.05)
(6.58)
(2.95)
N20 Gen Bomb C
(932)
(2438)
(12.65)
(2.90)
N20 Artifact
(7.12)
(1739)
(9-58)
(2.62)
Avg bomb N2Q = 3 bomb avg
Concentrations in ppm unless noted
( ) indicates negative values
82
-------
components of SO-, and H-,0 to the degree that N-,0 generation in stored (1-2 weeks) sample
containers is consistently minimized to < 10 ppm. Sequentially, SO, is neutralized, and HzO is
removed from a fossil fuel combustion process flue gas sample stream before entering a Teflon-lined
stainless steel container. The neutralization of S02 requires the presence of H20 in the flue gas
stream. Therefore, the flue gas sample must be collected upstream of any moisture conditioning
devices such as condensers and/or desiccants that may be present in CEM sample delivery/conditioning
systems.
The flue gas sample is extracted from the combustion source using a vacuum pump which
pushes the gaseous sample through the two-cartridge, solid sorbent system and ultimately through the
grab sample container or "sample bomb." The sample container is sealed and stored for up to two
weeks at room temperature. The sample containeis can be analyzed for N20 at any point during the
2-week holding period.
This procedure was developed for use with Que gases from conventional fossQ fuel combustion
sources and processes. Samples were designed to be collected upstream of any pollution control
equipment or on combustion facilities where pollution control equipment did not exist This grab
sampling method may not be suitable for use where sampling is performed downstream of pollution
control devices or processes.
During the development of this sampling method, tests were conducted to determine the fossil
fuel combustion process Que gas NO, S02, and HzO concentration ranges where NT20 generation in
aged (1-2 week) samples would be consistently minimized to less than 10 ppm. This method was
found suitable for use on combustion systems with Que gas concentrations in the following ranges:
• S02 — 0-2,500 ppm
• NO — 0-1,000 ppm
• H-,0 — 5-25 percent (by volume)
83
-------
These flue gas concentration ranges were verified under actual combustion conditions as well.
During these tests, the flue gas components of CO, CO-,, and unburned hydrocarbons, typically present
in fossil fuel combustion process streams, were found not to interfere with sampling method
performance. Other common flue gas components such as hydrogen chloride (HQ) and ammonia
(NH3) were not evaluated and may act as interferences.
84
-------
SECTION 4
SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS
The analytical and sampling methods developed during this project were found to be effective
tools for the characterization of N20 emissions from fossil fuel combustion processes and equipment.
The automated, on-line monitoring system has proved to be a particularly effective means to acquire
accurate, near real-time N-,0 measurements from diverse combustion sources. The prototype
instrument developed has since been duplicated and field tested at a commercial power utility.
Application of the grab sampling methodology has not been as aggressive, primarily because of the
need for absolute measurement of N20.
Because of the N-,0 generation sampling artifact, the reliability of accurate measurements
becomes even more critical. The procedures developed through this task are by no means the ultimate
answers to N20 sampling and analysis needs. The procedures developed were meant to rapidly enable
the characterization of fossil fuel combustion source emissions. Although the automated, on-line
GC/ECD monitoring system has proved to be accurate and reliable, it is not a real-time analyzer and is
therefore not capable of continuous monitoring or measurement Unfortunately, the commercial
availability of dedicated, state-of-the-art combustion process N-,0 monitoring equipment is extremely
limited. Of those available, detection levels may be insufficient. In addition, these NDIR systems are
susceptible to interferences from other combustion process gases that absorb IR radiation at
wavelengths close to those that are absorbed by N-,0. These interferences are often minimized
through the use of elaborate sample gas conditioning systems.
-------
Realizing that the continued development of continuous on-line monitoring instrumentation is
likely, consideration should also be given to the sample delivery systems. The residence lime in these
systems may be long enough that generation of N-,0 within the system may be possible. The volume
and length of sample tubing along with sample flow rate should be considered. During tests
performed by AEERL where a TDLTR continuous monitoring system was used, N20 concentration
spikes were observed following flow stoppage in sample delivery lines. These spikes were observed
even when sample Oow was stopped only for a period of several minutes. It is possible that sample
delivery systems exist where the residence time between the source and analyzer can approach several
minutes. Further examination of sample delivery systems are warranted.
The grab sampling method developed, although suitable for the screening of high N-,0-
emitting fossil fuel combustion sources, is not suitable for the collection of grab samples for the
determination of absolute N20 measurements. The authors are unaware of a grab sampling method(s)
that ensures the collection of uncompromised grab samples where generation does not take place, even
after long periods of storage. Is addition, although most researchers measuring N->0 emissions from
combustion using grab sampling techniques take means to collect samples where the sampling artifact
is drastically minimized, the variety of procedures to do so are quite diverse. It seems logical that
some type of standardized grab sampling approach be developed.
In summary, the procedures developed during the course of this project were sufficient to meet
AEERL's fossil fuel combustion source characterization needs. These procedures are documented in
the form of EPA-AEERL ROPs. The ROPs contain detailed descriptions of the respective
methodologies.
86
-------
SECTION 5
QUALITY CONTROL EVALUATION REPORT
This study was conducted following quality assurance/quality control guidelines set by
EPA/AF.F.RL. This study was performed under an AEERL Category IV Quality Assurance Project
Plan (QTRAK No. 89014) reviewed and approved by EPA. The goal of the project was to develop
sampling and analytical methodologies suitable for the characterization of N,0 emissions from fossil
fuel combustion sources.
The approach taken during this study was predominantly qualitative in nature. Many of the
tests conducted were based on a go/no go or yes/no approach in order to effectively screen candidate
sorbents for the development of the grab sampling method.
The measurements made by this project were of sufficient quality to more than adequately
accomplish project goals. Essentially, the only measurements made were for NO, SO;, and N-,0. The
accuracy requirements, expressed as percent bias, were 20, 20, and IS percent, respectively. Unless
stated otherwise in the report, all data validating quality control checks performed before, during
and/or after each test were within these limits.
87
-------
SECTION 6
REFERENCES
1. D. Pierotti, R. A. Rasmussen, "Combustion as a source of nitrous oxide in the atmosphere,"
Geophysical Research Letters. 3(5): 265-267 (1976)
2 W. M. Hao, S. C. Wofey, M. B. McElroy, "Sources of atmospheric nitrous oxide from
combustion," Journal of Geophysical Research. 92(D3): 3098-3104 (1987)
3 R. F. Weiss, H. Craig, "Production of atmospheric nitrous oxide by combustion," Geophysical
Research Letters. 3(12): 751-753 (1976)
4 M. A. Kahlil, R. A. Rasmussen, "Increase in seasonal cycles of nitrous oxide in the earth's
atmosphere," Tellus. 35B: 161-169 (1983)
5 V. Ramanaihan, R. J. Cicerone, H. B. Singh, J. T. Kiehl,. "Trace gas trends and their potential
role in climate change," Journal of Geophysical Research. 90(D3): 5547-5566 (1985)
6 G. L Pearman, D. Ethcridge, F. de Silva, P. J. Fraser, "Evidence of changing concentrations of
atmospheric CO->, N->0, and CH. from air bubbles in Antarctic ice," Nature. 320: 248-250
(1986)
7 C Castaldini, R. DeRosier, L. R. Waterland, H. B. Mason, "Environmental assessment of
industrial process combustion equipment modified for Iow-NOx operation," Proceedings of the
1982 Joint Symposium on Stationary Combustion NO^ Control. Vol. II, EPA-600-9-85-022b
(NTIS PB85-235612), U.S. Environmental Protection Agency, Research Triangle Park, 1985
8 W. S. Lanier, S. B. Robinson, EPA Workshop on N,Q Emission from Combustion. EPA-600-
8-86-035 (NTIS PB87-113742), U.S. Environmental Protection Agency, Research Triangle
Park, 1986
9 J. C Kramlich, R. K. Lyon, W. S. Lanier, EPA/NOAA/NASAAJSDA N,Q Workshop. Vol. I:
Measurement Studies and Combustion Sources. EPA-600-8-88-079 (NTIS PB88-214911), U.S.
Environmental Protection Agency, Research Triangle Park, 1988
10 J. V. Ryan, R_ K. Srivastava, EPAyihb* European Workshop on the Emission of Nitrous Oxide
from Fossil Fuel Combustion. EPA-600-9-89-089 (NTIS PB90-126038), U.S. Environmental
Protection Agency, Research Triangle Park, 1989
88
-------
11 L. J. Muzio, J. C Kramlich, "An artifact in the measurement of N20 from combustion
sources," Geophysical Research Letters. 15(12): 1369-1372 (1988)
12 L. J. Muzio, M. E. Tcague, J. C Kramlich, J. A. Cole, J. M. McCarthy, R. K. Lyon, "Errors in
grab sample measurements of N,0 from combustion sources," JAPCA. 39: 287-293 (1989)
13 W. P. Linak, J. A. McSorley, R. E. Hall, J. V. Ryan, R. K. Srivastavn, J. O. L. Wendt, J. B.
Mereb, "Nitrous oxide emissions from fossil fuel combustion," Journal of Geophysical
Research. 95(D6): 7533-7541 (1990)
14 J. S. Ford, Recommended Operating Procedure No. 45: Analysis of Nitrous Oxide from
Combustion Sources. EPA-600-8-90-053 (NTIS PB90-238502), U.S. Environmental Protection
Agency, Research Triangle Park, 1990
15 T. A. Montgomery, G. S. Samuelson, L. J. Muzio, "Continuous infrared analysis of N,0 in
combustion products," JAPCA. 39: 721-726 (1989)
16 F. E. Briden, D. F. Natschke, R. B. Snoddy, "The practical application of tunable diode laser
infrared spectroscopy to the monitoring of nitrous oxide and other combustion process stream
gases," presented at 1991 Joint Symposium on Stationary Combustion NOn Control.
Washington, DC (March 1991)
17 J. V. Ryan, S. A. Karris, "On-line monitoring of nitrous oxide from combustion sources using
an automated gas chromaiograph system," Presented at the Air and Waste Management
Association Conference on Measurement of Toxic and Related Air Pollutants. Durham, NC,
M-W 1992
18 J. V. Ryan, S. A. Karris, Recommended Operating Procedure No. 56: Collection of Gaseous
Grab Samples from Combustion Sources for Nitrous Oxide Measurement. EPA-600-R -92-141
(NTIS PB92-216928), U.S. Environmental Protection Agency, Research Triangle Park, 1992
19 R. Clayton, A. S. Sykes, R. Machflek, K. Krebs, J. Ryan, N..O Field Study. EPA-600-2-89-006
(NTIS PB89-166623), U.S. Environmental Protection Agency, Research Triangle Park, 1989
20 R- F. Weiss, "Determination of carbon dioxide and methane by dual catalyst flame ionization
chromatography and nitrous oxide by electron capture chromatography," Journal of
Chromatographic Science. 19: 611-616 (1981)
21 R. K. Lyon, J. C Kramlich, J. A. Cole, "Nitrous oxide: Sources, sampling, and science
policy," Environmental Science and Technology. 23(4): 392-393 (1989)
22 G. G. deSoete, Parametric Study of N-O Formation from Sulfur Oxides and Nitric Oxide
During Storage of Rue Gas Samples. Report 36 732, Institute Franca is du Petrol, Ruefl-
Malmaison, France (1988)
23 S.A. Sloan, C. K. Laird, "Measurements of nitrous oxide emissions from P. F. fired power
stations," Atmospheric Environment. 5: 1199-1206 (1990)
89
-------
24
L. R. Martin, D. E. Damschen, H. S. Judeikis, "The reactions of nitrogen oxides with SO-, in
aqueous aerosols," Atmospheric Environment, 15: 191-195(1981)
S. G. Chang, D. Littlcjohn, N. H. Lin, In Flue Desulfurization; J. L. Hudson, G. T. Rochelle,
Ed.; ACS Symposium Scries 168; American Chemical Society: Washington, DC, 1982; 127-
152
R. K. Lyon, J. A- Cole, "Kinetic modeling of artifacts in the measurement of N,0 from
combustion sources," Combustion and Flame, 77: 139-143 (1989)
90
-------
APPENDIX A
NON-CONTINUOUS SAMPLING AND ANALYSIS OF NITROUS OXIDE
FROM COMBUSTION SOURCES
ROP NO. 43
PREFACE
This appended method is included for historical perspective only. Aspects of this method were
evaluated to develop the sampling and analytical methods described in this report. Specifically, the
vacuum evacuation method for transferring the gaseous sample from the sample container to the
analytical system was evaluated. The vacuum evacuation method was found to be susceptible to
system leaks and was found not to be suitable for small volume samples. The sample introduction
method ultimately selected used syringe injection.
A-l
-------
NgO Analysis
Date: April 1988
Revision No. 1
Page 1 of 20
S1
NON-CONTINUOUS SAMPLING AND ANALYSIS OF i&)US OXmkJP
FROM COMBUSTION SOURCEst"^^ —
ROP NO. 43 4^ ^f/l
T.C. Grogan
J.M. McCa
J.C. Kram
sT & ¦*&>
Energy and Enviri
Irvine,
cv
onmen&w«5earch$|5jraorat1on
fjp*™ J&P &
C»Kom1a ^rTw?798
$> o® .©*
Si - ^
1.0 Introduction ^
X? N. Ci
The method desag^ro bel owMjj^es' the pr^^ple of gas chromatographic
separation and eleq|g0h captu^g^ytection for the measurement of nitrous
oxide (NgO) in c(^^^ion g^s M$es. Pertfer^hce of this method should not
«yj
jfco^-e unfy»raer wit
th sQMjg^^ampl knc
o
Pnnci^e^^pplic^fejy
be attempt^^mthose ayr^ye unfqptjfc^ with the operation of a gas
chromatogrfpite^Rrith sqyj^^^ampll4jj^^oe£aijse knowledge beyond the scope of
this preslntatj
is used^w
ox id
e is extracted from a combustion source and
lilted sample container. The sample is analyzed
20) by gas chromatographic (GC} separation and
ion CECD).
A-2
-------
N2O Analysis
Date: April 1988
Revision No- 1
Page 2 of 20
2.2 Applicabil ity. The approach is used for measuring the NgO
of combustion gases, both within the flame and in flue gases.
*Ss
3.0
Range and Sensitivity
&
This method is designed for the 1 to
upper limit can be extended by proper caH|
sample. The lower limit can be extended by^
loop beyond the 0.6 milliliters employed
4
range.%?Is
^^^e of
ution
4.0 Interferences
4.1 Particulate Matter,
interference by clogging of
be eliminated by use of
fm
JS
tn ga^gawiles may cause
This interference «ust
4.2 Moisture.
unreproducible
disturbances
trap described
e con^^^ation jSS se interferences or
ytH&?
$r wit*|nt
th¥Thpany fW
proc^.4#
system due to the flow
Water 1s removed by the
4.3
^kted Comsi^jrfos. H
fl u o r^^ft^Ty droca,
basel%ie Jlsets
backflush systei
4.4 Carborre&jf@3Hde.
ted compounds, in particular
can esttfi&iriate the ECD and lead to long-term
hterf^encp 1s avoided by use of the precolunm/
bed in. the method.
K
dioxide can cause a negative interference
on the dS^Ktor, p^yyj^arly if it co-elutes with H2O under the high
concentrations presera^ln combustion sources. The column used in the
presertfLj^fchod avoj^^thls problem by eluting N2O before C02* Also, the
us# ifHi Arier gas, as described in the method, prevents the
erence i
A-3
-------
NgO Analysis
Date: Apri 1 1988
Revision No. 1
Page 3 of 20
4.5 Oxygen. The detector becomes desensitized by exposure to o:
Because of the oxygen present in the samples, a progre
desensltlzation of the detector cannot be avoided during repet^Tt
analysis. As outlined in the procedure, a rep&tltive cat^^rfeion is
performed throughout the sample run and the d<
drift. Overnight conditioning returns the
sensitivity.
Apparatus
5.1 Sampling Probe. The probe si
EPA Method 7, with the followin
stainless steel, quartz, or
above 190°C. Probe cooli
above 200°C. This is prf^Jpally
surface and to quench
cooling media. ^
5.2 Sample Ligg.f^'^flon^r st
connection betw^^toe end or^he.
be as short as &a£&nable.
5.3
r the
e baseli
N
d for
may be
robe ma
flon sh
-------
MgO Analysis
Date: April iggg
Revision No. 1
Page 4 of 20
through the valves. Sample containers with free septa should n
used.
5-5 Sample Train. The sample train consists jjf. the prois, slpple
filter, and sample container. The outlet of Jsm iaaple coriiyplr is
connected through a flow metering device to a
flow meter should be si led to allow
condensate knockout may be used betwee
flow meter, if desired.
jttzrxx
» a smgAfi puap.^e pisap andlfTi
=f s
.^^ampV^a^i ner aw^m
0
-------
N.jQ Analysis
Date: Feb. 1988
Revision No. 0
Page 5 of 20
POLYETHYLENE TU3IHS
STAIMLES^TIEL TU8I
TYGQN
So
Ar/CH,
CYLINDER
10 PORT
JU.VE
tR-LME
REGULATOR
GAS PACIFIER
COLUW
HIXE-UP
RON
COmUOUJER
I ie 1 I r 1 t i
O
»• 1 t/«-
VALVf ISOLATION
COIL
« PSKTi r—mi
WWIJ» i ' 3
jV
LAB-LINE OVEN
70*C
CONTAINS
6" x 1/8" S.S.
PRE COLUMN
PACKS) WITH
PflRAPAK Q
diagram of the N^O analysis system.
A-6
-------
N2O Analysis
Date: April 1988
Revision No. l
Page 5 of 20
6.0
Calibration
Two levels of calibration are used. The most detail<
performed after the initial setup, whenever the has
upset, and whenever the system is repowered
cool off. The second level is perform
variations in system response, and to p
drift during the analytical day.
6.1 Detailed Calibration. N1 trou#^))£T3e em is si
pulverized coal flames general!j^a^Tietweel
the recommended initial callbr^an&j^ntervaff^flTh the
interval be extended to^^jfc^&Ude anv^iaffifes
encountered during analvsiglgyilote that, sources
different N2O emissions; reference
best guidance on selectf&y#) ini
ne an
ut-down an
r a full
check
for de
once
d 150
The instrument re
concentrations i
response. Ea&h
triplicate.
from
is is
viso that the
this range
eld markedly
ie literature will provide the
1 bra t1 oiQIfeerval.
should B®Wetenphiew|£t five evenly separated
er to dfe|§ct any fjgn-ljnearities in the detector
leparate^swfei gas shotffia be measured at least in
kO
Cy y of the normal means used for
kknown dffrul|lons ^fe|es may be employed. Including the
precalfcrlfed stananHK. The normal procedure 1s to prepare
the sample J n^fes HaffiLof known volumes which are fitted with
septla. Th^J^^is eva^^raTto less than 1 mn Hg using a vacuum pump
and a knotfgfnpunt d^^ews injected through the septum port. The
pressure is^Basuretf^ll^S^ means capable of resolving 1 mm Hg pressure.
The cylinder is then Erackfilled with nitrogen to atmospheric pressure.
The $^|taoie fracfio^is calculated as follows:
'ati on
'inc
Cn2o] =
CI]
A-7
-------
N20 Analysis
Date: April 1988
Revision No. I
Page 7 of 20
where VN20 and vcy1 are the volume of the injected N2O and the
volume, respectively. Mote that if the HgO and the dilution niti%|wi 4?
are all at ambient pressure and temperature, then no correc^^mWs
required for these variables.
o
6.1.2 Procedure. Once the standards are
procedure follow exactly the analysis
is are ^Jseffred, the, cal ibratl0*^9*t
is .g3tcr1.Jty.
ft%»^S^ cylinftTi^h1"ts preaS®M
samples. As samples are withdrawn cyyi£Ja^lts Pre
" sampi^i
develop a standard calibration curveg~ |lTs proce^gpfoll
decreases. Since a fixed volume, samnfc Sod
aliquot, 1t is necessary to correct ^je^respon$ap39
*ocemyr
£
to raea
The initial calibration AffLUp'esul t e t^t will contain
three entries for each ca&|»^.tion paiJHEy Theseagaa^^C 1) the NgO
^ -rr
concentration in the stand^|^(2) the 1^%#^ the N2W$^Rc, as obtained
from the gas chromatograph7*and (3p^jie sample pressure. The NjO
concentration corrected 0 atauroreSre is c m ated as follows:
[f^Olcurve ^ ClKHilx Pstapdar^l>®Q C23
Pstarute^S^60
ISs cl^
1 the pr$|g|re in the^fSiaple loop at the time of
g (th1#%MSuremeat described in the analytical
procedure s»|afe*aon). of peaJMraa against £N20]curve the
true ca^N|r^ion cuHgLfor a sample loop pressure. Figure 2
showMQfeical ca|rfS$|tion cur^gWerated over a 10 to 150 ppst range.
The cwttmlon ^gwwrally T%nyexcept for a slight non-linearity at
low sensl tlvgQfep. Su^La non-11 near!ty 1s consistent with ECD
cnaracter^^tcg^ror theJitejMtical reasons discussed by Weiss [1981).
Figure sin&ama&t that covers a more sensitive range: 0.5
to 15 ppn?
d Leve
Ibration
The ECO sensitivity is subject to drift
nd within a single day. For N2O analysis from
A-8
-------
N2O Analysis
Date: Feb. 1988
Revision No. 0
Page 8 of 20
m
m
£
(mdd) ON
A-9
-------
^0 Analysis
Date: Feb. 1988
Revision No. 0
Page 9 of 20
A—10
-------
N20 Analysis
Date: April 1988
Revision No. 1
Page 10 of 20
combustion gas, this drift is mainly due to the exposure of the de
to the oxygen in the sample. Since this is unavoidable, the da
corrected by repetitive single point calibrations during the
day. A calibration is performed at least once^jpr each l|b saiples
(i.e., the sequence is duplicate determi natiaj^|p>' cal ibrat&isf' gas,
duplicate determination of sample A, duplica^^j^erminat^^^of sampjel
B, and return to calibration gas).
6.2.1 Calibration Standards. This
concentration is selected to be negt^
concentration of the samples. For Itw0 lieasurei
a good value is 100 ppm. A co
pressure cylinder is the mostJ|bm^fe'nent a$ftro
of N2O in such cylinders ens^^ OTisistan^
6.2.2 Procedure. The
'sCiaHy pn
m co
pressure regulator
connector. When th.
relax to ambie
performed as de
concentration is'
Ibration
se
ed
tion
span.jjas Tn a high
h. The^^h stability
day performance.
e looptaffs been
n&M^sure^y stftppin^bu*
SiaSNted in g&^alyticall|g$jfi
.0
Praced
where^lL»tnlfis
proportional ae«lchoice depends on the nature of the source and the
;e of cld&actbrization desired. Systems in which the sample gases
A-ll
I
-------
N2O Analysis
Date: April 1988
Revision No. 1
Page 11 of 20
are well homogenized and time steady may be well characterizi
single measurement at a single location. Character!zation 0
highly stratified flows will necessarily require detailed trayelggiif'of
the duct or stack. Procedures for both approache&^are desc*""
7.2 Single Point Characterization. The
described in Section 5.0. The sample coir
or a similar insulating material. The o
and the probe is inserted into the flo
the flow is adjusted to a point in thera^ to
sample container is continuously p^g^| with
reaches thermal equilibrium w^HtShe prob
however, should the container tomdess th
through it (e.g., no less,#8ip%>5 Ute
ml. container). At the^M&l usion,
first closed, and a nomfipt is all
equilibrate with the MsUTiM stai Satire
closed, and the con^^gr is4 red fri
required, the cmei^gaave are,
ready for anal vsi^pi^.
Full Dj
involve
that
re qui
e emi
1zfction.
es on
until
train Ja&assenbled
eK&is wrapqM&$ glass
flow^,i£*a$grted (11
samp«t^^pp is sta^tecTand
The
tainer
erature. no case,
tiroes ^^volume purged
be through a 250
iwnstream Wftainer valve is
or thg^Ojg^ainer pressure to
tream valve is then
sampling system. If
and the containers are
;haract
usM duct characterization will
~
ciently fine grid across a duct
be obtained by integrating N2O
Hon wltftfmw ve&ci^fe. Due to the specialized hardware
%thlsUl^pd is appTOHble only to streams below 200°C. Hie
procedure <£1|fiftthat d^fej^ibed in Section 7.2, with the following
exceptions ^
The pnSBfc is moiMfcied^to Include the pi tot probe, thermal sensor,
a*(Lprobe tijyjspecified for EPA Method 5. The EPA Method 5 probe
suitableLprlbe for this application.
A-12
-------
(3)
NgO Analysis
Date: April 1988
Revision No. 1
Page 12 of 20
{2) The pi tot manometer, pump, and flow measurement system spec
for EPA Method 5 are used.
(4)
Operation procedures specified in Section l^^are f ol%^wetffhere,
with the exception that the flow is se^tP^frovide an isokinetljc^f^
sampling rate by use of the procedures in ||>5R;§tethod
EPA Method 1 may be used as a gu1
traverse points required, and
an overly large number of samn»s?'W!.g.1
approach is to reduce the auabgr of S|
multiple of the Method 1 sparfff:ation.
(5) The average N2O c
calculated in accordance
ss the^sUiot or stack is
edures outlined in EPA Method
8.0
Refer to Figu
Comparison of Fi
relationship be
The
flows thri
the detector
flow is wlthdra
0.6 0111111*
ejsto s
ocation
this c
an even
tration
th the
Analysis Proce
unctio
r* gas
Fed schemaWf of the analytical system,
facilitate the identification of the
DfllHlters/nlnute. This initially
f, the precolunn, the main column, and
Is connected into the system, and a small
set
ample
e fli
the jfeBer^the 10-port sample valve, (Including the
isolation coll, and into the vacuum pump,
ed to4S&F8BaS? the flow indirectly by measuring the rate at
The manometer
which the ffiask is evacuaWd. When the sample loop has been filled with
sample, t&qjgrabuum pun^glve is closed and the system is allowed to come to
presstf^^es^jptibr^K^Ji^manometer reading 1s recorded to allow calculation
of ifw^otal corwfcctJ" volume of sample within the sample loop. The direct
A-13
-------
N?0 Analysis
Date: Feb. 1988
Revision No. 0
Page 13 of 20
ISOLATION
COIL
Ar/CH,
CARRIER
GAS
VACUUM
£
MANOMETER
SAMPL186
VALfC
WITH
SAI*»U
LOOP
BAOCFLuSh gas
rarr
AQUASORB
DRYER
(P205)
1Ifled flow schematic of the analysis system.
A-14
-------
N2O Analysis
Date: April 1988
Revision No. 1
Page 14 of 20
analysis is started by switching tne 10-port valve so the sample loc
placed into the carrier gas flow path. The carrier flow is directed
the precolumn for 2.5 minutes. This time is sufficient for the
elute onto the main column, Put for any fluorinated ctffifounds telle reiained
sJlf -S?
on the precolumn. As the 6-port valve is switched^tW^precolunn is removed^-
from the carrier gas stream. A 20-mil 111 Iter
backflushes the precolumn during this
chronatogram in Figure 5, the NjO elutes
injection. The sample run is not terraij
elapsed, however. This is to allow the.
step ensures that the subsequent sarapl
from the preceding analysis. The respgjisjl from
by a cnart recorder or a digital ^icxfeLrator.
9.0
Calculations
The following cal
concentration. The rei
obtain an uncorrecte
loop pressure and
[NgQldry =
determine the N2O
'e calibration curve to
corrected for the sample
AS S
imatef
until
~2 peaCtd| ful 1 y,
not ejeJerjence CO 2^
5 min
amplifier is recorded
X
ORS
peak are
concentration.
ibratioflrSt&Jt as fol
cess
eM76Q/Pioop) W
atlon 1n the sample on a dry
where: if^O^jry = ^O^cc
value form the peak area and the calibration
:ted N2O concentration In the conmercial second-
'vel standards (from equation 3). This is the
Pverage of the calibration determinations made
immediately before and after the sample analysis.
A—15
-------
fUO Analysis
Date: Feb. 1988
Revision No. 0
Page 15 of 20
3%
Example chromatogram for ^0 analysis.
a—16
-------
[NjOlcurve
N20 Analysis
Date: April 1988
Revision Mo. 1
Page 16 of 20
N2O concentration obtained from the calibratior
using the peak area generated by the seconc !
standard.
absolute pressure in the^i^le loop
injection, ran Hg.
The dry N20 value can be corrected to any
use of the following:
[N20]
-------
MgO Analysis
Date: April 1988
Revision No. 1
Page 17 of 20
calibration is outside the bounds described below. In theorj^TWltt
calibration should be linear with the exception of a curving
concentrations, as shown in Figure 2. The recommendation approajnfois to
manually fit a line through the curved portion of the^jirve, and ttse a|31near
least squares fit to derive a straight line through £l|gj|jP near poi^^rof the^
curve. The two curves are constrained to meejjB&h in terfflfrpf absolute, ,
value and slope.
The goodness of fit 1s estimated by
deviation for all the points:
I C(yi - f(*i))/f(*i)]2
fii
ratfl^ for ead&jacTrii
lative
where:
yi
fUl>
^turve at the sane
NjO concentrate^ for ead^^
)J20 concetfftlion cal^fete
peak *sr ^
ro'"2s O
Although the seljffltarfT of an«cS^>tance/reiect1on criteria is somewhat
arbitrary, past%Stt»riene^fiwi^fchown t&SWlT s is greater than 0.1 the
calibration
10.2
done.
If the second level calibration 1s
the primary calibration, the primary
consistantly grea
claibration shou
10.3 C^i stencMMU^s. As described above, the second level
cal ibratioi|Vis performed^fr least once between every two samples. All
samples aasjCrorformefl^Wfeduplicate. and in all cases additional aliquots are
analyzedf4£jfche fference between replicates is greater than 5%. A
blawfes^ylinder -coB,ta1liing nitrogen is analyzed once each day. The response
A-18
-------
N2O Analysis
Date: April 1988
Revision Mo. 1
Page 18 of 20
of the second level calibration should be maintained as a quality contro^
chart-
11.0
Bibliography
The following references discuss vario us ^a«ico aches to^9fe N2O
measurement problem. The approach described abovejfl^ nodifigfpfon of ttjat^
reported by Weiss (1931). The modifications wege^tequired^^awnsforn
ambient measurement technique into one suitabllpta" combiffue gas
wish to acknowledge the advice Dr. Weiss proviflltois. JC
1. Bowman, C. T. and R. J. Roby: CombuaT iPane 7i
Jmuiu —
2. Hao, h. M.: Industrial Sources
Ph.D. dissertation, Divisj
Cambridge, MA (1986).
3. Kramlich, J. C., R.
Heap: Combust. Race
4. Pierottl, D. and
and CHjBr,
University,
rshing, and N. P.
Lett. 3, 255 (1976).
ospheeJc^?0,
pllea^a^B^s.
5. Weiss, R. F.: |$jj^hrom.^Sca7 jfe9, 511
6. Weiss, H. Craim» Geop^9*?p. Lett. 3 , 751 (1976).
-------
NjO Analysis
Date: April 1988
Revision No. 1
Page 19 of 20
EXAMPLE CALCULATION
The following example provides a guide to the dal^r^duction
^ &
The input conditions are as follows:
Methane/air combustion with 41 O2 (1
gas.
Peak area obtained from tJie sai
of 450 mm Hg. From Figure yields
ppo.
The secondary standar^l^^ 100 ppa^K^^iich the ^igtrured peak area
is 1.5 at an atmosphericpressurejjL750 mn Hg.
ures.
e past
pie ffoopjKessure
Hue of Cl^H^aw = 54
Since the calibration cu
r^™^75u mn Hg.
step is to correct the s^^gary standan^-to lAufrl^y equation 3
pressure, the first
[NgOlstd.latm
BO ppra)
r(760} =
.7 ppm.
Because of the^|%gntly rej^cfej^pressuraifjl^the sample loop, the loop would
contain PP® its were adjusted to 760 mn Hg.- by
isotherm# d^ution. #R«t, thfc^B^gpratlon curve is entered with the
measured «tatfWdary Bfr^grd peaic^ifS, 1.5, and the value recovered is:
- - - ^0 value in the sample is calculated by
CK20lcurve ' |8
equation 4.
[NgQldry =
im/88ppm) {760nmHg/450mmHg) = 102.3 ppm.
In essenciF^pns equag!
dri f aci
corrects the raw reading for the downward detector
sample pressure was reduced.
A-20
-------
NgO Analysis
Date: April 1988
Revision No. 1
Page 20 of 20
To correct the NjO value to 0" 02» equation 5 is used:
CWdry.OiO = H02.3) (213/(21-4) = 126.4 ppm.
To correct to a wet basis requires a calculatii
combustion for CH4/air at 45 (dry) 02- The tot
produced per mole of CH4 is 10.52, while the
Thus, the wet basis N2O is calculated from equatagh|^s:
JS& %
EWwet = (102-3ppo) (10.52/12.52) =
This calculation basic!y represents
larger basis.
the pro
noles^al product
wet raang^s 12.52
ct that %ObO is sprefff^over a
£
A-21
-------
APPENDIX B
STANDARD OPERATING PROCEDURE FOR DETERMINING
NITROUS OXIDE CONCENTRATIONS IN COMBUSTION FLUE GAS
PREFACE
This appended method is included for historical peispective only. The chromatographic conditions
contained in this method were selected for evaluation during development of sampling and analysis
methodologies. The chromatographic conditions were found to be susceptible to baseline upsets and
chromatographic difficulties resulting from interferences present is combustion samples.
B-l
-------
tj
<»
't /
Document Control No.:
Date of Issue: Julv 20
Revision Nuacer: -
Page 1 of 4
STANDARD OPERATING PROCEDURE FOR !
NITROUS OXIDE CONCENTRATIONS IN COMBI
Alston
Radian C
P.O.
Research Trian^^ft
Office Director:
Effective Date:
N FLUE
3000
S-
Signed:
ate Signed:
clai
has be
Indust
of
an
Envir
Signature Approvals and
Immediate Supervisor:
_ ird Operating Procedure
>for the sole use of the
jtal Research Laboratory
Triangle Park, North Carolina
rifically applicable to the
Sr organizations.
-------
Document; Control No.:
Dare of Issue: -Tnlv 20.
Revision Numoer: Q_
Pace 2 of 4
1. APPLTCABTLTTV AND
PRINCIPLE
1.1 Scope; This method is applicable
nitrous oxide (N^O) in combustion flue gas
on the separation of nitrous oxide from £
using a gas chromatographic column and
vrith an electron caoture detector (ECD) .
1.2 Range of Applicability:
is limited by the electron cap'
useful upper limit concentration
The minimum detectable
les.
ANALYSIS
1.3
operator can be obtains
from the previous va^u4
1.4 Interfpr
demonstrated to
calibration
line shoul
additional
injections tf
2.
uanti
e methods is based
A
pper apraic
detect^p^saturatr
P pai^^oer mi, taion.
(HDL)
~2 part
million.
tS by the same
curacy of + 15%
L -w
%
The ,V ytical system must be initially
^-*f^n^~rti^iy-iilv * i by running a
^ samp3j^f^the flue gas. The base-
pr wjjSnded time to determine if
interfere with consecutive
meters;
Poeajjak" Super Q, 12 ft. x 1/8 in. stainless
conditioned at 220°C
'O^n T^fcse^ture: 35 degrees C
[njectorf^fSO degrees C —
Detector (Electron Capture): Ni 63 at 350 degrees C
Carrier Gas: Nitrogen at 20 mi/min
-------
Gas SamDie Value:
2.2 Calibration;
Document Control No.: _
Date of Issue: July 2Q.
Revision Number:
Page 3 of £
Valco 6-port with 1 mi-loop, s
less steel
use of known concentrations of NjO
Calibration is acc
in nitr
Scott Environmental iP^pbstea&re£^s
be obtained from
concentrations in the range of 4.0 ppm
yo p^m4ka^-
seed t"*fTrimrTh the
Gasfixtures
The gas sample valve is placed
standard gas mixture is purged
The flow of standard i
allowed to equilibrate to atopic
The sample valve is then
injections should be made
£ach standard concentratig
curve.
15 sec.) .
/$ mt**wot
cm. . Snrrnr'ai
ZjS/Z.
re obtained.
the man li-
the 1
the 1
shut&KC, and
ric jffl&roxima
po
resul
cted to
t
uilib
Condi
mL/ntu>
andai
e.
ntiiXJi> ba
leasu
plot vs.
AnaJfttzei sanroles
and oven temperatures to
220WC with nitrogen flow of 20
e is stable, c
mixtures^ using 6-port gas sample
height (or area) of standards and
ncentration.
the same technique as
^u.wlBvec with standards.
^"^lion: A response factor can be calculated
from the standards by dividing the concentration (ppm) of a
standard by the peak height or the area counts. The
-------
Document: Control No.: L
Date of Issuer .Tilly 20. "I!
Revision Number: Q_
Page 4 of 4_
concentration of the unknown sample is computed by multiply ^
its peak height (or area) by the response factor.
3. oualtty ASEtJRANCE/onRT.rrv cowanr.
Each sample should be injected a
to establish repeatability. c<*£J>r*
/nT ***- a. yse* ~
After sample injection, suffici
allow later elutiag peaks to exi
injection is made.
As with all electron
current and sensitivity chec
niflnuf3 cture r s rBcoirnticnds
also required to check
f th
ust be
pez^S^c standing
eAT&tffrrmed according to the
¦jj^tets ofc-thn, flietector are
General period^^lfegnter
needed with approraiflSfc doc;
should be at lea9^g^9.S9J
installed betwee^J®^
All sjfl^lii'^E'ines
to verify
N
chromatocraph is
ition.^^fe carrier gas used
with trace oxygen filters
fand tjj®gj^ chroma tog raph.
ec and periodically checked
¦frh/JL
G.Z--(/r>) C * C6)
&r ~ £ct>
/?1- 5/»o*~ "f /"
-------
APPENDIX C
RECOMMENDED OPERATING PROCEDURE FOR ANALYSIS OF NITROUS OXIDE
IN COMBUSTION FLUE GASES
AEERL/ROP NO. 45
PREFACE
This appended method is included for historical perspective only. This was the original
AEERL/ROP No. 45 which has since been revised and published. This appended method was developed
combining sample introduction aspects of Appendix A with chromatographic conditions contained in
Appendix B. The vacuum evacuation method for transferring the gaseous sample from the sample
container to the analytical system was evaluated. The vacuum evacuation method was found to be
susceptible to system leaks and was found not to be suitable for small volume samples. The sample
introduction method ultimately selected used syringe injection. The chromatographic conditions were
found to be susceptible to baseline upsets and chromatographic difficulties resulting from interferences
present is combustion samples. Detector response to N->0 concentration was not linear over the desired
range of quantitation.
C-l
-------
RECOMMENDED OPERATING PROCEDURE FOR ^
ANALYSIS OF NITROUS OXIDE IN COMBUSTION FLUE^SASE
r v
AESrtURO? No. 45
by
J.V. Ryan*
RA Groie"
C.V. WallH?#*
COMBUSTION HESStaCH R
AIR AND ENERGY ENGU^g^ING RES
U.S. ENVIRON^^A?PnO"rEi
RESEARCHsM&NGL£ p.
H
_ BOF
^SENCC
27711
This draft reco;
Eaerav Ea
Triangle AdgS
"Acurex Corporation^.
Research Triande
torv, U
prepared for tfce sole use of tise Air and
'maeatal Protection Agency, Research
•cafly applicable to the activities of otter
"Combustion ResefBafa Branc
Air and Eneres-Engmeenne ReHircii Laboratory
U.S. Emirogge^J Protean Agency
Research "Zra^gl'c Park.*
,#
-.^^Environrra^a® uaiity Assurance
Researcnlpr riauele Instate
Research Tnaneic Parie NC
C-2
-------
Document No: AEERL/45
Sluus: Draft
Revision No. 0
Date: Mav 1988
Past i of 15
RECOMMENDED OPERATING PROCEDURE FOR
ANALYSIS OF NITROUS OXIDE IN COMBUSTION FLUE GAS^Sf
1.0 PROCEDURAL ELEMENTS
1.1 Scope and Application
Based on the principles of gas chromatographic see
this method is applicable to the measurement of nitrous
combustion flue gases and is designed for a 5 to 200 ppm
ed by changing the sampie loop pressure to reduce thi
of samples: the sensitivity limits shall be determined
standards.
This method requires that the analyst have ai^Wbf a
rwo (2) vears of excellence m gas caroinatoera
^hmiis can Mgasspha-
:e loop
-------
Docmfflf No: ASERL/45
Stanza: Draft
Revision No. 0
Dae Mzv 1988
Pace 2 of 15
MDL: minimum detection limit: the concentration corresponding 10 rxve (5x) times
background noise level of the measurement.
MQL. -mmirmm nnanririahie lirRtrr :fae concentration corresponding to ten (
the background cease level of the measurement, or the lowest gS&garian
1.4 interterences
The analytical system shall be demonstrated to be nee
a two-point calibration check prior to sample analysis and by
control (QC) samples with unknown samples.
Sample must be pulled through a son-reactive
however, the analytical system shall be protected from ^Eeifcence
moisture in the sample gas by placing a moisture tratffc3S!e •nrior »
vaive. Interference caused by organic wfll be
moisture trap.
1.5 Apparatus
1.5.1 Analysis
* Gas Chromatognph: Caf^l^^pera;
tron capture detector with Jm^&stant
* GC Column: 3.7m
mesh support (e.g- P®ap|ic5uper
* Carrier Gas: Ar
percent At
Moistun
percent
percent
amste con
to the
ova
g$L jj-C an
t cell, an'
inane
ntCH
tected to an elec-
of operating at 330* C
steel packed with 80/100
Inc~ Deenueld. IL).
;d for ECD analysis, containing 95
) O J). Teflon tube containing 50
Mallinckrodt. Sl Louis. MO) and 50
Vacuum Pump:
DASi Data
.«f.^3ftjP^eis>on.wfjj&tentxk capable of pulling 29 inches of mercury.
Compres Cylmaers^ix. containing various known concentrations of N-O/in
pure N-.
ents (N,
rs: Three. containing known concentrations cf mixed
'and O-}.
C-4
-------
Document No: ALERL-45
Stanrs: Draft
Revrnon No. 0
Daic Mavl988
Pane 5 of IS
isolation Coil: U ra x 0.032 cm (5 fi x 0.125 in.) stainiess steel.
Thermometer. Certified accaracv of — 0.5S&.
Sample: Dry grab sampie extracted from a combusdon
stainicss steel sample cylinder (no smaller than 250 ml_
aid canable of withstandine a full vacuum.
1.6 Analytical System
This method is intended for use with a gas chromai
capture deieoor (ECD) and capable of operating at
cm (12 ft x 0.125 in.) stainless steel column packed.
The ECD is a Ni63 constant current cell operate^
prepared for ECD analysis, is 95 percent aigoj
flows through a 5 A moiecuiar sieve and
the detector. The sampie is supplied
stainless steel tubing connected to a i-cc
tube containing a P.O, acidic absoi
organics. respectively. is placed
stainless steel tubing. The tub
prior to flowing through
stainless steei isolation coi
a with
GC use
d Z20°
ISuqer
enscruobe
e loon
that th
thajara&pd chamaL Vacuus
oiU^^d to supg^jH^sample t<
'on
032
support,
er gas. aspecially
e. Thcafarrier gas initially
reacnin^|^m.
-------
Document Nc AtfcRi -
Sums: Dran
Rrvsioii No. 0
Datr- Mzy 1988
Pase4of iS
ISOLATION m
COIL S
?
VACUUM
PUMP
-------
Document No. A' i R L 45
Stains: Draft
Revision No. 0
Daxe: Mav 1988
?aae 5 of 15
Remove the em
Set cetecto;
rare
(e». 330" C
ssure to 4 32.
t to the
~ allow.
acentratio
O in Dure
meed
cylinder
son o
a a muln
1.7.1 Pre-Calibration
• Refer to the analytical system's operations manual for SDecific steps of the :
analysis ana shutdown.
• Open the main valve of the carrier gas cylinder ana set the
ps'i;.
• Adjust the carrier gas flow to 20 ccnnin.
• Disconnect the GC column from the ECD. cap
column to vent into the oven.
• Turn power on.
• Condition the GC column overnight at 221
• Obtain six compressed gas cylinders co;
X. in tne range of 5 to 200 ppm.
• Verify the accuracy and stability
• Use verified concentration of
1.7.2 Multipoint Calibration
The following multipoint
samples. The integrity of the caBWetnBn will be^femtored
prior to analysis of sampics'&fctffiiMitli ilvsi^frpC saxnpi<
cylinder,
calibration.
rior to analysis of
''mid-point calibration check
lout the analvsis Dcriod.
iet and connect the detector to the GC.
xely tw^^cicgrccs beiow the maximum operating
turn tesjSggirure is 320" C).
coiugnD^ci
Access the data aggggrab:
* _
Establish th&&if%tfv of the
temperature.
Connect tfiSfeLQ/N, ufWffirjft! standard containing the lowest concentration of N.O so
th2t thegtandard flows fir^^he cylinder regulator into the connector for the sample
comaiheiwnd ilnailyo&jhe GC inieL
the t
earn in iroaica i
nse (in terms of integrated neak area i to each calibration
C-7
-------
Document No: AEERL'45
Siams: Draft
Revision No. 0
Date: May 1988
Paae 6 of i5
• Calculate the response m terms of average integrated peak area (PA) for each stanriau &
• ReDcat the analysis for the remainine catfbration eases.
Moid Figure 2 shows that tie ECD output is a nonlinear function
uonstap between the two variables exhibits a cozve at the
use two calibration curves to vertfv concentranons of QC
centrarion in combustion gas samnies. Fieures " —
generated for the uuper and lower rancesT The
the lowest point of the first curve and the hi gin
Correct verified N,0 concentration to
foDowine ea nation:
icon. Be
00 of the 1
bw caiibi
; oT the
the tela-,
method
N,0
^corrected = ^verified*
Linearize the response of thqjmafoical syste:
regression equation (see Noten&he least
regression parameters, slotatim) andinte:
the correaed N,0 conce^^^" '
The correlatii
Verify
conceh:
onn. v
), shall be
linear
b. The
determined by plotting
sSDwhere:
cient
& °
must be > 0.998.
&s*
iraQoi^Q^tre by back-calculating the
each caraflbon standard. Apply the average peak
||e appropriate least squares regression, where:
Calculate
corrected'
= 100
) between the verified N.O concentration ([N.O]
a N.O concentration ([N-O] where:
curve ~ corr.'/'^s-^ corr.
dard. the calibration is acceatable.
C-8
-------
ppm N|0
Figure 2. Detector response lo known concentrations of N20.
-------
Donrmcnt Nu AEERL45
Sums: Draft
Revision No. 0
Date Mav 1988
Page d of If
S
&
-------
ppm N20
Figure 4. N.,0 calibration curve for lower range.
-------
Document No: AEERL.'45
Sums: Draft
RsviaanNo.O
Pair- May 1988
Pace 10 of IS
The multipoint calibration shall be repeated if:
— I D I > 2-5% for any standard
— the analytical system is subjected to a full shutdown for any-jeason
— during analysis the result for any QC sample is great*
corrective action does not resoive the problem.
1.7.3 Calibration Checks
• Obtain three commercially prepared compresse
concentrations at o to 15 ppm and 100 ppm. Th<
similar to that ocdccxq in combustion flue
or pure nitrogen. TLsJie gases will be used
unicnowns.
• Verify the accuracy ana stability of
against the multiooint caiibrado:
ana other
1)
2)
Conduct triplicate measurements of the
a%i cement between each ZpAust be
Calculate the averagaiwepQted
standard as outii^dhfScric
to con
CS (CO
ined N,
Section 1.7
¦h cylinder
IfDJ-O]
unco:
tO^°' PA-bW
use t^ac^QC sample. The
percent previous run.
a and tlfK^^WDncentxation of each
vaiue must be used as the
unicnowns.
pressure ana temperature using the
293.16
Fill three 2
I) Plaq^rvo-siage. •„
Ltial " 760 * T"C - 273.16
"with a portion of each QC sample gas.
_ . puntv. flow regulator on the eas cylinder.
cm fO.25 inches) OX), sample line with Swageiock
ator outlet.
TM
C-12
-------
DocmneniNo: AEFRL.45
Slams: Dran
RcvbmbNo.O
Daie: Mav 1988
Paae 11 of If
3*i Open vaives located a: each end of ihc sampie container.
-) Attach me other end of the sampie line connected to the cylinder rq^safos^o me
sample container inlet. ~
5)
Open the cylinder ana regulator outlet valves. S
pressure to 0.35 kg/cnr (5 psi).
6) Purge the sample container with the QC
10 liters of eas flow throueh the container.
T) Qose the vaive on the sample contains xStfi- turn
outlet valve, ana allow the sampie to ea
ai
Qose the vaive on the samnie
Verify the N-O concentration
in the anaivncal nrocedures
(D) berwq$jS?he initial
he call
([N.O]qc for'^g^gsampj^^e: ^
e to
r ouiie
aes as outlined
Calculate the percent diffe
section 1.73 ([M-O] ^
If ID [ is <
samsie
initial
C sam
mousti
as o
of the
rane
Die co
e samo
xai
infection calculated in
;er the transfer
x 100
cceptabie for use as a caiiDranon cneac
samnies.
e anahmcal procedures section, at least once
1.8 Analysis Proee
• Verify the
that are '
,t calibration by aupiicaie analysis of two QC samples
ibranon curve.
er to the analytical svstem.
id to acuroxunateh-100 nunHe on the mercurv manometer.
c—13
-------
Document No: A£tiU_45
Stasis: Draii
Revision No. U
Date: Mav 1988
Paae 12 of 15
S» *
-------
Document No: AEERLZ45
Status: Draii
Revision No. 0
Date: Mav 1988
Pace 13 of 15
Open the sampie container ouilei vaivc ana fill the sanroie ioao by allowing the*^^
manometer to return to apuroxmiateiy 400 mmfie. Record the actual manomcicrreac- -';;
ing ior calculating the total coireaed volume of the sampie in the sample loon#
Direct the sampie flow from the sample loop into the carrier eas stream ;
column. ' " " %,
Three constituents will be ehiied in the following orde
O, at 21-- minutes. CO, at 4.5 minutes, and N.O at
be" terminated at 10 minutes. The response of we
snows a chromatoeiam for N,0 analysis.
Conduct the triplicate analysis or unknowns
The analysis sequence is QC samDle (duplii
(triplicate). and' QC sampie.
Determine die concentration of N-O ugeaeij uruc
analyzed immediately before or immcd&ieiv arte:
the calculations section.
eDAS.
d ab
ownJSitjiirate),
1.9 Calculations
The following calculations are
unknown gas sampie:
• Uncorrected N.O com
cc^noss
ted from calibration curve.
to auannfv
n ior
Die an
EJ^O] obtaSjap = f-1- %j/
wnere:
eem
m = sloce
N.O conce
oressure
€ .# nsfoi
the result oxme QC samDie
roie in Question, as described in
m eaco
neidsaaSfej corrected for instrument drift ana sample loop
lai >S*:01 ob7%£6Q/p^) x f293.i6/T8C - 273.16).
concentration in neld samoies. calculated from calibration curve.
c-15
-------
Document No: AEtKL-
Stains; Dran
Reman No. 0
Date: Mav 1988
Paae 14 of 15
= absolute pressure in sample loop as indicated by mercury maxtona
prior to injection.
T'C = ambient temperature in degrees Celcius.
* Accuracv expressed as oercent deviaricm.
100 • (X-T>t '
where:
X = average measured value = t X;
i«l 1
n
T =» true vaiue
• Precision expressed as relative standard gfUfepon (RSDj
100
where:
ucn fCV),
Gas cvunaers con
Calculations
ton- Not
S = Standard deviation
2.0 QUALITY CONTRO
2.1 QC Checks
• Toe inteenrv of
outlined i
from tin
pro"oieJranibr a full
louitored with verified QC samples as
n the results of the QC caeck deviate
steps will be taken to correct the
be conducted.
: (100 psi) witl not be used.
i on triclicaie analysis of each sample.
22 QC Controls
DocumeuiBtion
-Seep ail test results, calibration data, and quality control
atorv notebook. Sien ana date the notebook at time of data
C-16
I
-------
Docnxnent No: AEERL45
Sums: Diafi
Revision No. 0
Date: Mav 1988
Pace li 01 15
2) Instrument Notebook—Assign a logbook :o the GC to maintain a recor
calibrations. maintenance, ana reDairs.
5) Controi Charts—Use control charts to track daily response to
4-) Cenmcanan/Venfications—In a secure area. naxHiaaa^nle conramffl'jp
manufacturer's certmcanons and laboratory ven^ftrion of standatds.
5) Instrument Manuals—Keep operator ma
system available ana easily accessible.
• Raw Data
Maintain ail measurement data (storage
3.0 REFERENCES
The following were referenced ft^^u^on of
Cicerone. R-L et al_ Journal of Geopfysical Resist f^oL 83. No. C673042-3050, June 1978.
an. EPA-j®^8r8§-035.
:n Meast»di8fe|faSvstems. Vol. m. EPA-600/4-
sTotthe
ties.
EPA Workshop on N-O Emisoaryfrom
-
Qualirv Assurance Handboffl«iiar Air
77-027b.
Svkes. AX- Radian
Weiss. R-F- Joi
I coram n
?Science, VoL 19. 611-616. December 1981.
-------
TECHNICAL REPORT DATA
fPtecsc read Jnurvclions on the reverse before compter'
1. REPORT NO. 2.
EPA- 600/R-93-088
3.
U. TITLE and SUBTITLE
Development of Sampling and -Analytical Methods for
the Measurement of Nitrous Oxide from Fossil Fuel
Combustion Sources
5. REPORT DATE
May 1993
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
Jeffrey V. Ryan and Shawn A. Karns
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING OROANIZATION NAME AND ADDRESS
Acurex Environmental Corporation
P. C. Box 13109
Research Triangle Park, North Carolina 27709
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
68-DC-0141, Tasks 91-021.
92-066. and 93-133
12. SPONSORING AGENCY NAME AND AOORESS
EPA. Office of Research and Development
Air and Energy Engineering Research Laboratory
Research Triangle Park, NC 27711
13. TYPE OP REPORT ANO PERIOO COVERED
Task Final; 10/89 - 3/93
14. SPONSORING AGENCY CODE
EPA/600/13
^.supplementarynotes A£ERL project officer is William P. Linak. Mail Drop 65. 919/
541-5792.
i6. abstract rep0rt documents the technical approach ana results achieved while
developing a grab sampling method and an automated, on-line gas chromatography
method suitable to characterize nitrous oxide (N20) emissions from fossil fuel com-
bustion sources. The two methods developed have been documented in the form of
U.S. EPA/AEERL, Recommended Operating Procedures. The combustion of fossil
fuels is suspected to contribute to measured increases in ambient concentrations of
N2C. Accurate and reliable measurement techniques would help to assess the rela-
tive contribution of fossil fuel combustion N2C emissions to the increase in ambient
concentrations. The characterization of N2C emissions from fossil fuel combustion
sources has been hindered by the lack of suitable and acceptable grab sampling and
on-line monitoring methodologies. Grab samples have been shown to be compromised
by a sampling artifact in which N20 is actually generated in the sample container in
the presence of sulfur dioxide (S02), nitrogen oxides (NCx). and moisture. On-line
monitoring techniques are limited and. of those available, instrument costs are of-
ten prohibitive, detection levels are often insufficient, and the techniques are often
susceptible to interferences present in combustion process effluents.
17. KEY WORDS ANO DOCUMENT ANALYSIS
1. DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS
c. COS ATI Field/Croup
Pollution Analyzing
Nitrogen Oxide (N2C) Gas Chromato-
Fossil Fuels graphy
Combustion
Measurement
Sampling
Pollution Control
Stationary Sources
13 B
07B
21D 07D
21B
14G
14 B
18. DISTRIBUTION STATEMENT
Release to Public
19. SECURITY CLASS (ThisReport)
Unclassified
21. NO. OF PAGES
142
20. SECURITY CLASS (TWlpage)
Unclassified
22. PRICE
EPA Form 2220-1 <*-73)
------- |