Content uploaded by Steven Joseph Gluck
Author content
All content in this area was uploaded by Steven Joseph Gluck on Jan 02, 2014
Content may be subject to copyright.
Evaluation of NO
x
Flue Gas Analyzers for Accuracy and Their
Applicability for Low-Concentration Measurements
Steven Gluck, Chuck Glenn, Tim Logan, Bac Vu, Mike Walsh, and Pat Williams
Dow Chemical Co., Freeport, Texas
ABSTRACT
The requirements of the Texas State Implementation Plan
of the U.S. Clean Air Act for the Houston-Galveston
Ozone Nonattainment Area stipulate large reductions in
oxides of nitrogen (NO
x
) emissions. A large number of
sources at Dow Chemical Co. sites within the nonattain-
ment area may require the addition of continuous emis-
sion monitoring systems (CEMS) for online analysis of
NO
x
, CO, and O
2
. At the outset of this work, it was not
known whether the analyzers could accurately measure
NO
x
as low as 2 ppm. Therefore, NO
x
CEMS analyzers
from five different companies were evaluated for their
ability to reliably measure NO
x
in the 2–20 ppm range.
Testing was performed with a laboratory apparatus that
accurately simulated different mixtures of flue gas and, on
a limited basis, simulated a dual-train sampling system on
a gas turbine. The results indicate that this method is a
reasonable approach for analyzer testing and reveal im-
portant technical performance aspects for accurate NO
x
measurements. Several commercial analyzers, if installed
in a CEMS application with sampling conditioning com-
ponents similar to those used in this study, can meet the
U.S. Environmental Protection Agency’s measurement
data quality requirements for accuracy.
INTRODUCTION
At the time of this writing, the Texas State Implementa-
tion Plan of the U.S. Clean Air Act required 90% reduction
of oxides of nitrogen (NO
x
) from industrial sources in the
eight-county area surrounding Houston.
1
The Dow
Chemical Co.’s plan for compliance includes the applica-
tion of continuous emission monitoring systems (CEMS)
or predictive emission monitoring systems (PEMS) to a
large number of NO
x
sources. At the outset of this study,
it was not known whether NO
x
could be accurately mea-
sured at concentrations less than 10 ppm. The specifica-
tions for accuracy of NO
x
measurements are defined in
U.S. Environmental Protection Agency (EPA) Performance
Specification 2.
2
In short, this specification requires that
the NO
x
CEMS analyzer results and all system variability
be within 20% of the reference method during a relative
accuracy test audit (RATA). It also requires that the NO
x
CEMS analyzer not drift more than 2.5% of its span value
in a single day for seven consecutive days. Because a large
number of new CEMS were to be specified, an informal
review of the current industrial practices was undertaken
to determine the seriousness of the measurement prob-
lem. No recent peer-reviewed publications or case studies
were available to assist in a simple assessment of whether
low levels of NO
x
could be measured in flue gas or which
suppliers made technically acceptable equipment. The
purpose of this study was to confidently select the ana-
lyzer technology that would meet the needs of Dow
Chemical.
HISTORICAL BACKGROUND AND THEORY
Much of the current information available in textbooks
regarding NO
x
analyzers is based on older studies and
references. For the purposes of historical perspective and
to pull together information from different sources im-
pacting this study, this knowledge is included here with
appropriate referencing. In discussions with practitioners
over the last three years, we realized that some of the
historical information has been forgotten. The informa-
tion presented in this section is not new; it provides good
background and a basis for the current problems in low
NO
x
measurement. The use of critical orifices in instru-
mentation, the combustion kinetics for NO
2
formation,
and the impact of converter tube temperatures on NH
3
measurements cannot be found in a single source any-
where. These are all significant considerations for low
NO
x
measurements.
Chemiluminescence is the most common type of
NO
x
analyzer. In fact, EPA reference method 7E
3
currently
requires a chemiluminescence analyzer. The technology
IMPLICATIONS
NO
x
analyzers stable and sensitive enough to measure
levels of NO
x
between 2 and 20 ppm are currently available.
Loss of NO
2
in the NO
x
analyzer sampling system is unlikely
if the system is properly designed. Acceptable accuracy
performance can be achieved through technically sound
overall system design and analyzer selection.
TECHNICAL PAPER ISSN 1047-3289 J. Air & Waste Manage. Assoc. 53:749–758
Copyright 2003 Air & Waste Management Association
Volume 53 June 2003 Journal of the Air & Waste Management Association 749
is based upon the reaction of O
3
with NO to form NO
2
in
an excited state.
4,5
From reference 4, the important reac-
tions in a chemiluminescent detector are
NO ⫹O3
O
¡
k1
NO2*⫹O2(1)
NO2*
O
¡
k2
NO2⫹hv (2)
NO2*⫹M
O
¡
k3
NO2⫹M(3)
In the first reaction, O
3
reacts with NO to form excited
nitrogen dioxide (NO
2
*). This NO
2
* reaches equilibrium
either through chemiluminescence (reaction 2) or
through collisional energy transfer with any third body,
M(quenching, reaction 3). The species and concentration
of Mdetermines the amount of quench. Therefore, the
overall intensity, I, can be described as
I⫽k1k2关NO兴关O3兴
k2⫹
冘
M
k3M关M兴(4)
where k
3M
is a function of the specific third body. It can
be seen in eq 4 that if k
3M
or Mare small and [O
3
] is much
larger than [NO], then the intensity of the chemilumines-
cence is only a function of the [NO]. This is achieved in
modern chemiluminescence analyzers through decreas-
ing the collisional probability of the third body by oper-
ating the reaction chamber at a reduced pressure and
using critical orifices to maintain a constant flow ratio of
sample to O
3
. A critical orifice operates at the velocity of
sound such that any changes in downstream pressure do
not change the flow rate of gas through the orifice. Both
H
2
O and CO
2
are significant quench bodies that reduce
the NO response by 4.75–3.43% per mol % H
2
O and
1.05–0.896% per mol % CO
2
relative to N
2
.
5,6
A typical
chemiluminescence analyzer is shown in Figure 1.
NO
x
is measured because NO
2
is in equilibrium with
NO in the presence of O
2
. Directly out of the combustion
zone where the temperature is the highest, NO is the
predominant species unless the gas is quickly cooled, in
which case NO
2
will predominate.
5
At cooler tempera-
tures, NO
2
is favored. To manage these issues, NO
x
, the
sum of NO plus NO
2
, is measured. In reporting mass, NO
x
is calculated as NO
2
. To measure NO
2
and NO, a converter
tube in the analyzer converts the NO
2
to NO. There are
several catalysts available for this conversion; stainless
steel, Mo, vitreous C, and Mo coated on vitreous C are
some of the more common choices. At higher converter
tube temperatures, NH
3
will be converted to NO and can
cause a positive bias in the NO
x
measurement. This is an
important consideration when NH
3
is expected in the flue
gas, such as when nonselective catalytic reduction or se-
lective catalytic reduction (SCR) units are employed to
chemically minimize NO
x
. At lower converter tube tem-
peratures, the potential for conversion efficiency to de-
crease exists, resulting in an erroneously low NO
x
mea-
surement. Reaction temperature, however, should not be
the sole consideration in selecting a NO
x
converter; side
reactions, efficiency, and durability should also be con-
sidered. A properly selected and designed NO
x
converter
should produce acceptable conversion efficiency.
Each of the analyzers evaluated in this paper measure
NO rather than total NO
x
.TogetaNO
x
reading, the NO
2
in the sample must be converted to NO by passing the
sample over a heated catalyst. Different types of catalysts
are available, each with certain advantages. None of the
suppliers used high-temperature stainless steel, appar-
ently because it converts NH
3
to NO and would give false
high readings in an SCR-based system. Horiba uses a com-
bination Mo/vitreous C catalyst. The California Analytical
analyzer uses a proprietary low-temperature (200 °C) vit-
reous C catalyst, which has high efficiency and low re-
placement cost. Rosemount uses vitreous C. The ABB and
Siemens analyzers use Mo catalyst.
NO
x
is measured on a dry basis for low concentration
applications. Condensed water in the sample lines de-
composes the NO
2
to form HNO
2
and HNO
3
.
NO2⫹H2O⫽HNO2⫹HNO3(5)
NO
2
is very water-soluble; it decomposes as described in
the previous reaction. However, NO has limited solubil-
ity: 7% at 0 °C and insoluble at 100 °C. An additional
measurement problem with water was the previously
Figure 1. Typical chemiluminescence analyzer.
Gluck, Glenn, Logan, Vu, Walsh, and Williams
750 Journal of the Air & Waste Management Association Volume 53 June 2003
mentioned third body effect. The water problem was par-
tially solved by selectively removing it via a “cold-finger”
or impinger or, alternatively, with a Perma Pure dryer. The
Perma Pure dryer is a tube and shell arrangement of acid-
treated Nafion polymer. It is a highly water-permeable,
dessicant-like material. On the tube side, flue gas has a
high water concentration. On the shell side, the same gas,
or another gas, is at a much lower partial water pressure,
thus providing the necessary chemical potential to drive
the water across the polymer wall. The Perma Pure dryer
was shown to absorb and release NO
2
and may, therefore,
not be useful for accurate low NO
x
measurements.
7
There-
fore, this kind of dryer was not used in this investigation
even though recent claims suggest that it is a suitable
material.
Many CEMS today use Peltier-type devices to cool the
sample stream in a short zone to reduce the contact time
of NO
2
with the water. The condensed water is continu-
ally pumped out of the bottom of this device. Normally,
the temperature of the Peltier cooler is maintained at 2–4
°C, which corresponds to a maximum concentration of
less than 0.8% water, thus minimizing the quench effect,
extending the lifetime of the NO converter, and protect-
ing the analyzer from water condensation.
The CO
2
quench is less significant than the H
2
O
quench and is managed differently. Chemiluminescence
technology is extremely sensitive and able to detect part
per trillion concentrations of NO. A commercially avail-
able chemiluminescence NO
x
analyzer is essentially “de-
tuned”by using less gain on the optics and less sample in
the analyzer, along with other appropriate engineering
modifications. As a result, commercially available chemi-
luminescence NO
x
analyzers currently offer a minimum
measuring range of 0–2 ppm. The more sensitive analyz-
ers may operate with a higher ratio of sample to O
3
than
the less sensitive analyzers. However, this increase in sen-
sitivity comes at the expense of an increase in the third
body quench. It turns out that a modern analyzer de-
signed with a detection limit of 0.1 ppm will have a CO
2
quench that is so small that it does not significantly
impact the accuracy of the NO
x
measurement, whereas an
analyzer designed with a detection limit of 0.02 ppm will
be impacted. The EPA methods permit calculation of this
effect in the results reporting, but it adds an uncertain
step in data processing that most users wish to avoid.
An entirely different approach to overcoming the
NO
x
quench problem has been taken by some users, par-
ticularly those with high particulate loading and SO
2
.
These users prefer to use a dilution system, which splits
the flow of the flue gas at a controlled ratio of, for exam-
ple, 100:1. This approach has been developed to lower the
concentration of interfering and potentially corrosive
substances in the analyzer. However, dilution presents a
challenge to accurate measurements. This was outlined in
the EPRI CEMS Analyzer Bias and Linearity Effects study.
8
Because there are so many variables involved with making
this approach work, the decision was made to not use it.
Dilution requires more hardware and, in Dow flue gas
applications in Texas, would require more maintenance,
including dilution gas usage. In addition, employing di-
lution requires more complicated software control and
higher user training and expertise.
Another early approach that deserves mention but
did not turn out to be historically successful was a chemi-
luminescence method that does not require an NO con-
verter. It instead relies on the reaction of NO and NO
2
with H
2
that has been generated through electrical dis-
charge.
9
It is likely that interference and mechanical com-
plexity caused this technique to fail in commercial appli-
cations. Yet another approach involves IR absorbance.
This approach has been used with reasonable success, but
currently the lowest range of the technology is 0–50 ppm,
too high for our application requirements. Several com-
panies market IR analyzers.
An alternate to the chemiluminescence method is
differential UV. ABB offers a differential UV absorbance
analyzer that determines NO.
10
A converter tube is used to
convert NO
2
to NO, similar to the chemiluminesence type
analyzers. The light source is an electrodeless discharge
lamp (EDL), which uses high-frequency induction in an
atmosphere of N
2
and O
2
to create a plasma. The plasma
gives some select emission lines for NO; 226 nm is se-
lected by an interference filter. The spectrum is shown in
Figure 2. The optical bench of this analyzer is shown in
Figure 3.
Measurement stability is obtained with a four-beam
optical method that uses either a clear or an NO-filled
gas cuvet to block NO specific absorption and a beam
splitter with a reference detector. This method allows for
Figure 2. Plasma emission from N
2
and O
2
EDL.
Gluck, Glenn, Logan, Vu, Walsh, and Williams
Volume 53 June 2003 Journal of the Air & Waste Management Association 751
compensation of source and detector aging and also for
contamination or fouling of the sample cell. There are four
intensity measurements, R1, R2, M1, and M2, as shown in
Figure 3. R1 is the measurement of the source intensity
when the filter wheel is open. R2 is the measurement of
source intensity when the filter wheel contains a cuvette
with NO, thus blocking specific NO wavelengths. M1 is the
intensity of the source less that absorbed by gas in the
sample cell. M2 is the intensity of the source less that ab-
sorbed by the NO gas filter and nonspecific absorbance in
the sample cell. M1 and M2 are separated in time, as are R1
and R2. The absorbance caused only by the analyte, NO, in
the sample cell is thus
A⫽
R1
M1⫺R2
M2
R1
M1
(6)
Another feature of this analyzer is the ability to calibrate
without a flowing gas. The calibration wheel contains
gas-filled cuvets, which, when rotated into the optical
path, are equivalent to an absorbing gas in the sample cell
and are thus useful for calibration. The lowest measure-
ment range available is 0–5 ppm, which is likely to be
more sensitive than is necessary. Differences between dif-
ferential UV and chemiluminescence measurement of
NO
x
are summarized in Table 1.
One promising technology is in situ analyzers. These
units have no sample extraction or sample conditioning
systems. Currently, Procal and Sick Optic manufacture
such units. They consist of a porous metal tube, which
acts as a filter. This metal tube is installed inside the
process pipe or vessel. The NO
x
that penetrates the tube is
measured in the IR (high-level measurements; e.g., 50–
200 ppm) or UV (low-level measurements; e.g., 0–20
ppm) range. The products differ in their multivariate sta-
tistical algorithms used to selectively determine the spe-
cific concentration of NO and NO
2
. The overall system is
greatly simplified because the analyzer and electronics
mount directly on the stack; there is, therefore, no sample
conditioning system or analyzer house required. Al-
though this approach has the potential to save consider-
able cost, it was not evaluated here because the companies
are not well represented in the United States (service
could be a problem), the technology is too new for a large
number of installations, and maintenance personnel do
not favor having to climb a stack to maintain an analyzer.
The future of NO
x
measurement technology will
likely be advanced through sensor development for the
automotive industry. Every new car may have a NO
x
sensor in the coming years. These sensors are currently
under development at various institutions and are based
on electrochemical, surface plasmon resonance, surface
acoustic wave technology, and other mechanisms.
CHALLENGES FOR LOW NO
x
MEASUREMENTS
During the last 10 years, analyzer manufacturers have
significantly increased the maintenance reliability of
CEMS NO
x
analyzers. At the same time, regulations have
required much lower measurement concentrations to the
point that measurements less than 15 ppm are considered
to be low NO
x
measurements. There are three economic
drivers for having accurate measurement technologies: (1)
process control, (2) NO
x
trading credits, and (3) the ability
to certify accuracy to effectively give the operating unit
the “license to operate.”Accurate NO
x
measurements al-
low the combustion sources to be optimized for power
Table 1. Differences between differential UV and chemiluminescence measurement
of NO
x
.
Chemiluminescence Differential UV
Method Chemical Spectroscopic
Ozone generator Yes No
Quenching effects Yes No
Auxiliary gas (air or O
2
)Yes No
Deozonizer Yes No
Vacuum pump Yes No
Converter Yes Yes
Calibration Flowing gas Internal cuvettes or flowing gas
Dynamic range 1000 (e.g., 10–10,000 ppm) 500
Figure 3. Optical bench and measurement modes for ABB LIMAS
differential UV analyzers.
Gluck, Glenn, Logan, Vu, Walsh, and Williams
752 Journal of the Air & Waste Management Association Volume 53 June 2003
output versus NO
x
concentration or to minimize NH
3
reagent usage in SCR units. NO
x
credits are based on “Cap
and Trade”agreements. An industrial source will be per-
mitted to emit no more NO
x
than a previously defined
emission limit. If the source emits under this cap, it has
NO
x
credit that it can sell through a brokerage to another
source that is over its cap. Being under the limit could
produce substantial revenues. The “license to operate”
implies that the operating unit’s CEMS (or PEMS) is cer-
tified to produce accurate data during a RATA and, thus,
the unit is allowed to continue operation.
The main technical challenge in this work was to
establish whether a CEMS NO
x
analyzer could reliably
make accurate measurements in the 2–20 ppm range. For
a CEMS NO
x
analyzer to achieve this (1) there should be
no measurement bias caused by the sampling system or
by quench effects, (2) the drift should be within EPA
Performance Specification 2 guidelines, and (3) the con-
centration resolution for measuring NO
x
should be signif-
icantly less than the expected NO
x
levels in the sources
(2.5 ppm), to allow for adequate measurement confi-
dence. A 5% error in a CEMS with a range of 0–10 ppm
would mean a 0.5-ppm error in the NO
x
concentration,
effectively challenging the 20% accuracy requirement of a
RATA for a low NO
x
source. Potential errors investigated
included converter efficiency, loss of NO
2
in the sample
conditioning system, CO
2
quench, and third body effects
related to the calibration of the analyzer in a dry N
2
matrix while running the analyzer in an entirely different
matrix stream.
EXPERIMENTAL PROCEDURES
NO
x
Analyzers Selected for Evaluation
Two previously completed studies on low-level CEMS
NO
x
analyzers served as screens for selecting such analyz-
ers for this evaluation; one study was completed by Gen-
eral Electric Energy and Environmental Research,
11
the
other by the University of California–Riverside.
12
Based
on our review of these studies, NO
x
analyzers from five
suppliers were selected: Horiba Instruments, Inc.; Califor-
nia Analytical Instruments (CAI), Inc.; Rosemount Ana-
lytical, Inc.; ABB Automation, Inc.; and Siemens Applied
Automation, Inc.
Testing Apparatus
The testing apparatus used to evaluate the NO
x
CEMS
analyzers is shown in Figure 4. It was designed to meet the
three major objectives of simulating an actual NO
x
source
with precise control of components and compositions,
providing sample conditioning similar to a typical CEMS
installation, and simultaneously testing all the analyzers.
NO
x
Source. The NO
x
source synthesized in the laboratory
contained components typically found in flue gas. NO
and NO
2
were diluted with blends of N
2
,CO
2
,O
2
, air,
and, at times, He. This gas was preheated and then mixed
with water vapor and NH
3
. To precisely control the com-
ponents, an Environics Series 2040 computerized multi-
component gas blending/dilution system was used to
blend the diluent (dilution) gases with the NO and NO
2
.
The gas blending/dilution system uses multiple mass flow
controllers to generate a gas blend of a specified concen-
tration and flow rate.
Various gases and gas blends were used as the diluent
gas. Zero-grade N
2
cylinder gas was used for zero and span
calibration of the analyzers. A Scott Master Class Ac-
cublend gas cylinder of 14.95% CO
2
, 3.03% O
2
, and N
2
balance was used as a matrix-blend diluent gas. Zero-grade
air was used to dilute instrument-grade CO
2
during the
chemiluminesence quenching test.
NO and NO
2
gas standards were chosen in the 2000-
ppm range so that their addition was small relative to the
diluent gas flow rate. This allowed for changes in the NO
or NO
2
concentrations that did not significantly change
the matrix. The NO
2
standard used was a Scott Certified
Master Class gas cylinder at 2020 ppm in N
2
. The NO
standard used was a Scott Certified Master Class gas cyl-
inder at 1995 ppm in N
2
. This standard was diluted to
17.3 ppm with N
2
by the Environics system to calibrate
the analyzers. A Scott RATA Class Calibration Standard
Figure 4. NO
x
analyzer testing apparatus.
Gluck, Glenn, Logan, Vu, Walsh, and Williams
Volume 53 June 2003 Journal of the Air & Waste Management Association 753
NO gas cylinder at 17.3 ppm was chosen to verify the
Master Class standard and the dilution system.
The output from the Environics was valved for flow
either directly to the analyzers, via the flow drawer, or
through the sampling system. Flow was valved directly to
the analyzers for zero and span calibration. For the sam-
pling system, the blended gas was preheated to 120 °C
through 25 ft of heated tetrafluoroethylene (TFE) tubing.
At this point, water vapor was injected. The water vapor
was generated by pumping water through heat-traced 1/8-
in. copper tubing, through a heat-traced 500-mL type 316
stainless-steel (316 SS) cylinder, and then through heat-
traced 1/16-in. tubing. The cylinder helped dampen the
flow fluctuations caused as the water flashed within the
hot tubing. The water addition was calibrated gravimetri-
cally with a Mettler PC 2000 analytical balance. For NH
3
addition, NH
4
OH was added to the water. A Kenics static
mixer (0.38-in. tube 316 SS with 21 elements) immedi-
ately following the water vapor injection provided the
final mixing of this NO
x
source. The flow from the Envi-
ronics dilution system was controlled at two rates. Flow
was set at 8000 mL/min when the output was valved
directly to the Baldwin flow drawer and at 9000 mL/min
when the output was valved to the sample system. The
higher rate provided excess sample to the sample system.
Sample Collection. From the output of the static mixer, the
sample was collected as though it was from an actual NO
x
emissions source. A 100-ft heated TFE sample line, previ-
ously used for stack gas sampling, was connected to a tee
downstream of the static mixer tube. The tee was con-
nected to a valve to provide a vent for excess sample. The
heated sample line was temperature controlled by a Clean
Air Engineering controller at 120 °C. The output of the
sample line was connected to a sample conditioner. A
Vaisala relative humidity sensor (HMP230) was inserted
before the conditioner. The sample conditioner consisted
of a Universal Analyzers Model 3080 SS sample cooler
(maintained at 4 °C) and an Air Dimensions DIA-VAC
Model 19310T vapor pump. Conditioned sample was
pumped to a Baldwin Environmental BEI Model 3300
flow-control drawer.
Analyzers. The Baldwin flow-control drawer distributed
conditioned sample to the analyzers in parallel. This ar-
rangement allowed for simulta-
neous testing of all the analyzers.
Recommended flow rates were con-
trolled for each individual analyzer.
The sample pressure at the flow
drawer was approximately 11 psig.
This pressure was maintained
whether the sample arrived directly
from the Environics dilution system or through the sample
system. The flows to the analyzers were
(1) Horiba CLA-510SS 2 L/min
(2) CAI 400 HCLD CE 1.5 L/min
(3) Rosemount CLD 194006 1.5 L/min
(4) ABB LIMAS 11 CEM 2 L/min
(5) Siemens CLD 70SE 0.3 L/min
All of the analyzers used NO
x
converters. The range
was set at 0–20 ppm for all of the analyzers except the
CAI, which was set at 0–30. The CAI was an analog device,
with preset selectable measuring ranges; 0–30 was the
lowest selectable range on the analyzer. The output ana-
log signals were transmitted to a LabVIEW program on a
desktop computer for data collection and logging into an
Excel spreadsheet.
RESULTS AND DISCUSSION
Experimental System Validation
Gas blends created with the blending system had a spec-
ified accuracy of ⫾2%. To confirm the accuracy of the
blended gases, a 17.3-ppm gas cylinder standard was com-
pared with a blended standard of 17.3 ppm that was
created from a 1995-ppm calibration gas. The data in
Table 2, averaged over a 4–5-min period of measurement,
confirmed the accuracy of the blending system. All di-
luted measurements were less than 1% different than the
undiluted reference gas.
Check for Measurement Interferences
Figures 5 and 6 show the effect of CO
2
on the analyzer
readings at constant NO and NO
2
levels. With the excep-
tion of the ABB analyzer, the presence of CO
2
caused
some negative interference with the NO
x
analysis. The
ABB UV analyzer was not expected to show CO
2
interfer-
ence because CO
2
does not have UV absorbance at 226
nm. The test data, shown in Figure 5, confirm this expec-
tation. Most of the chemiluminescence analyzers showed
acceptable performance with respect to CO
2
interference,
indicating that the analyzers had been designed to mini-
mize this effect. The one exception was the Siemens an-
alyzer. This analyzer was reportedly designed to achieve
lower detection limits than the other analyzers by intro-
ducing a larger mixing ratio of sample to O
3
in the reac-
tion chamber. The result was a higher sensitivity to CO
2
quenching. As shown in Figures 5 and 6, the NO and NO
2
Table 2. Blended standard gas vs. gas cylinder standard gas.
Horiba CAI Rosemount ABB Siemens
Average reading on blended 17.3 ppm standard (ppm) 17.5 17.6 17.5 17.5 17.3
Average reading on gas cylinder 17.3 ppm standard (ppm) 17.4 17.4 17.4 17.4 17.4
Difference (%) 0.68 0.88 0.49 0.64 ⫺0.13
Gluck, Glenn, Logan, Vu, Walsh, and Williams
754 Journal of the Air & Waste Management Association Volume 53 June 2003
analyses for the Siemens analyzer indicate a downward
trend with increasing CO
2
in the diluent gas (water com-
position was constant, nullifying NO
2
absorption effects).
Interference levels or quenching effect from CO
2
were
calculated ona%NO
x
quench per % CO
2
basis; the
results are summarized in Table 3.
Interference Particular to the ABB Analyzer
Because the ABB UV source produces O
2
emission lines,
there is a small sensitivity to O
2
in the sample. Figure 7
shows this effect. This effect applies only to the UV ana-
lyzer and not to the analyzers that use chemilumines-
cence detection. As long as the NO
x
concentration is
greater than 4 ppm, the error of this effect is less than 2%
of measurement and does not need to be corrected in the
data collection system. In a later version of this analyzer,
a new interference filter had been selected that the manu-
facturer claimed reduced this O
2
response by a factor of 5.
Analytical Precision, Accuracy, and Linearity
Linearity and estimated detection limits for the analyzers
are summarized in Table 4. The detection limit was esti-
mated by multiplying the SD (or noise level) of the zero
gas readings during five different experiments by 3.
13
Table 4 and Figures 8 and 9 show the results of the
linearity testing for NO and NO
2
.
All the instruments showed excellent linearity and
satisfactory detection limits. A satisfactory detection limit
is less than 2% of span. All of the analyzers had a 20-ppm
span with the exception of the CAI, which had a 30-ppm
span. Table 4 shows that all five analyzers demonstrated
detection limits that were less than 2% of span and that
the noise levels were within acceptable levels. The data
collection system logged data to the nearest 0.1 ppm, so
the detection limits reported here may be somewhat con-
servative. The average reading for the ABB analyzer on
zero gas was 0.2 ppm. This effect resulted because a poorly
designed prototype sample cell in the analyzer
required an extraordinarily long time (more than
10 min) to satisfactorily purge the cell; because of
inadequate purging, there were traces of NO
x
remaining in the cell when running the zero gas
samples.
As Figures 8 and 9 indicate, four of the five
analyzers exhibited excellent linearity. The Sie-
mens analyzer gave acceptable linearity, although
it showed the greatest deviation from unity on
Figure 5. CO
2
effect on NO
x
analysis (5 ppm NO).
Figure 6. CO
2
effect on NO
x
analysis (5 ppm NO
2
).
Table 3. CO
2
interference on NO
x
readings at 5 ppm NO.
Horiba CAI Rosemount ABB Siemens
%NO
x
quench per % CO
2
for 0% CO
2
00 0 0 0
%NO
x
quench per % CO
2
for 5% CO
2
⫺0.02 0.18 ⫺0.20 ⫺0.18 0.73
%NO
x
quench per % CO
2
for 10% CO
2
0.05 0.12 ⫺0.35 ⫺0.01 0.79
%NO
x
quench per % CO
2
for 14% CO
2
0.07 0.11 ⫺0.14 0.05 0.86
% average NO
x
quench per % CO
2
0.03 0.14 ⫺0.23 ⫺0.04 0.79
Figure 7. Effect of O
2
on ABB UV NO
x
analyzer.
Gluck, Glenn, Logan, Vu, Walsh, and Williams
Volume 53 June 2003 Journal of the Air & Waste Management Association 755
the slope, apparently because of CO
2
quenching. It is
interesting to note that the estimated detection limit for
the Siemens analyzer was not significantly better than
that of the other analyzers, even though the poor CO
2
interference rejection was stated to be caused by a design
that gave superior detection limits.
It is important to have a high converter efficiency so
analysis errors are minimized, particularly for low levels
of NO
2
(e.g., 5 ppm). Table 5 shows the converter
efficiency at 0 and 15% CO
2
for the various
analyzers. At low CO
2
concentrations, the
converter efficiencies ranged from approxi-
mately 90 to 95%. At 15% CO
2
, the measured
converter efficiencies were lower, possibly be-
cause of the reduction in sample O
2
concen-
trations that resulted from the blending of the
CO
2
with our air mixture. The effect of con-
verter efficiency versus O
2
concentration or
water content was not investigated. All the analyzers
tested came equipped with NO
x
converters. Following
span calibration, the NO
x
converter performances were
verified by measuring NO
2
diluted in N
2
. In the experi-
mental design, the analyzers were tested as a unit, con-
sisting of the analyzer with the converter. Therefore, the
effects of O
2
concentration or water content on NO and
NO
2
were attributed to the analyzer unit and not isolated
for the individual converters.
To determine the bias introduced as a result of NH
3
,
samples with 10-ppm NH
3
were prepared. Table 6 shows
the effects of the NH
3
on NO and NO
2
samples. Basically,
there was no significant effect from NH
3
. Ammonia was
detected in the condensate from the sample cooler, indi-
cating that the cooler was scrubbing NH
3
from the matrix.
Perma Pure manufactures an NH
3
-scrubbing sampling
system cartridge. The device was tested and was found to
effectively remove NH
3
from the sample stream, as indi-
cated by the fact that there was not a false high NO
x
reading. At 10-ppm NH
3
, the NO recovery averaged 101%
and the NO
2
recovery averaged 99%.
Another potential source of error in the measurement
of NO
x
is the loss of NO
2
in the water-removal stage. Gas
blends with and without water were generated to see if
there was any difference between the measured NO
x
val-
ues. Table 7 shows that the NO
2
losses were less than 2%,
which is within the error of the experimental system. The
large error in the Siemens analyzer results is likely caused
by a higher sensitivity to water quenching.
Drift
As specified by EPA Performance Specification 2, the drift
of the NO
x
CEMS analyzer must be less than 2.5% of full
span per day, measured daily for seven consecutive days
on both zero and span gas. Recalibration is permitted
Table 4. Estimated NO
x
detection limits and linearity.
Horiba CAI Rosemount ABB Siemens
Estimated detection limit (ppm) ⬍0.1 ⬍0.1 ⬍0.1 ⬍0.1 ⬍0.1
Slope 0.956 0.956 0.962 0.968 0.812
Intercept 0.0944 0.1306 0.1291 0.2460 0.0793
Regression coefficient 0.9999 0.9998 0.9999 0.9999 0.9998
Table 5. NO
2
converter efficiencies at 0 and 15% CO
2
.
%NO
2
Converted to NO
Horiba CAI Rosemount ABB Siemens
Dry sample with 0% CO
2
89.6 89.4 95.3 93.9 94.9
Dry sample with 15% CO
2
88.5 90.3 91.5 92.2 88.8
Figure 8. NO linearity (NO
x
reading vs. NO concentration).
Figure 9. NO linearity (NO
x
reading vs. NO
2
concentration).
Gluck, Glenn, Logan, Vu, Walsh, and Williams
756 Journal of the Air & Waste Management Association Volume 53 June 2003
every day. The span gas is typically at 80% of the span
range for the measured component. The zero drift for the
analyzers is listed in Table 8 for a period of five days in the
laboratory without calibration.
The drift of the ABB was biased high because of inad-
equate flushing of residual NO. (This was a design flaw
that was later fixed.) In all cases, the zero drift was low
enough to be considered acceptable. In a further explora-
tion of analyzer drift, the CAI NO
x
analyzer was subjected
to a field drift experiment in a mobile laboratory normally
used for process measurements and CEMS certification.
The analyzer was calibrated once and allowed to run for a
week on a flue gas NO
x
application. Periodically, the zero
and span drift were measured but the analyzer was never
recalibrated. The results are summarized in Table 9.
In all cases, the zero and span drift were well within
the required guidelines of EPA Performance Specification
2. It is not appropriate to suggest that if the span were
0–10 ppm, the drift would triple because the range was
reduced by a factor of 3 (specifically in the case of the CAI
NO
x
analyzer, which had a range of 0–30 ppm). Three
times this drift would not meet the requirements of EPA
Performance Specification 2. When the analyzer is
spanned at a low range appropriate for the application,
the noise decreases also. It is reasonable to believe that the
drift would also be less, especially because all of the in-
stallations are in temperature-controlled shelters. Because
testing was not completed with a lower range analyzer, it
cannot be absolutely concluded that there would not be
challenges in meeting the drift requirements of EPA Per-
formance Specification 2. However, the data in Tables 8
and 9, in conjunction with good installation design, in-
dicate that the challenges should be minimal.
Field Test
Two separate but similar sample trains were
placed on a gas turbine. One train had a single
CAI NO
x
analyzer. The other train had a CAI
NO
x
analyzer and an ABB NO
x
analyzer. The
purpose of the experiment was to determine if
these two different technologies gave equiva-
lent results and how much variation should be
expected between two similar analyzers.
When the data were plotted in terms of rela-
tive differences, the variability indicated a dif-
ference of approximately 1% for identical
manufacturers and 5% between manufacturers. Because
the relative accuracy criterion for EPA Performance Spec-
ification 2 is 20%, there was no effort to determine why
there was a 5% difference.
A “Worst-Case Scenario”
Understanding what was known from the laboratory and
field measurements can be applied to consider some ex-
treme conditions, specifically for the purpose of modeling
the impact of a worst-case scenario and determining if the
relative accuracy requirements would still be acceptable.
The conditions for this worst case assumed measurement
of 5 ppm NO
x
, with 50% of the NO
x
being NO
2
. It is also
assumed that 10% of the NO
2
is lost in the chiller (0.25
ppm) and there is a 0.08-ppm loss resulting from CO
2
quenching. If the converter efficiency is assumed to be
only 90%, the minimum required, then there is another
loss of 0.23 ppm NO
x
. Finally, a calibration error of 0.25
ppm is assumed. The measured NO
x
value would thus be
5⫺共0.25 ⫹0.08 ⫺0.23 ⫹0.25兲⫽4.19 ppm (7)
A value of 4.19 ppm corresponds to a relative accuracy of
⫺16%, still within the requirements of EPA Performance
Specification 2.
Summary of Measurement Capabilities
The evaluation data indicated that all of the analyzers
tested would be acceptable for flue gas applications. The
ABB analyzer that was originally supplied had an unac-
ceptably long response time because of an unsuitable
sample cell design. An improved sample cell was supplied
near the end of the evaluations, and the response time of
Table 6. Effect of NH
3
on NO
x
measurements.
% Recovery with 10 ppm NH
3
in Sample Calculated as
(measurement with 10 ppm NH
3
)/
(measurement without NH
3
)
Horiba CAI Rosemount ABB Siemens
10-ppm NO in matrix of 15% CO
2
,
3% O
2
, 15% H
2
O, balance N
2
100 100.9 94.9 100.3 97.5
10-ppm NO
2
in matrix of 15% CO
2
,
3% O
2
, 15% H
2
O, balance N
2
95.6 96.2 91.5 96.3 96.5
Table 7. NO
2
losses caused by solubility in water.
Horiba CAI Rosemount ABB Siemens
NO
2
lost from 10-ppm
standard (ppm) ⫺0.01 0.11 0.14 0.18 0.73
%NO
2
recovered 100.1 98.9 98.6 98.2 92.6
Table 8. Zero drift on NO
x
analyzers (5 days without calibration).
Horiba CAI Rosemount ABB Siemens
% average drift per day 0 ⫺0.02 ⫺0.04 ⫺0.24 ⫺0.05
Span (ppm) 20 30 20 20 20
Gluck, Glenn, Logan, Vu, Walsh, and Williams
Volume 53 June 2003 Journal of the Air & Waste Management Association 757
the analyzer was confirmed to be comparable to the other
analyzers tested.
CONCLUSIONS
NO
x
analyzers from five companies, Horiba, CAI, Rose-
mount, ABB, and Siemens, were selected for evaluation.
All of these analyzers, with the exception of the ABB, were
based on the O
3
-NO chemiluminescence reaction. The
ABB was based on a UV absorbance measurement using
the gas filter correlation principle. The measurement ca-
pabilities of the analyzers indicated that all but the Sie-
mens would give nearly equivalent data. The Siemens
analyzer had a significant quench effect from CO
2
, which
caused the reported NO
x
concentrations to be biased low.
The ABB analyzer was found to have a small positive bias
from O
2
, which can be corrected in the data acquisition
system if necessary. A field test was performed at a gas
turbine for the ABB and CAI analyzers. The results of the
test indicated that at 6–8 ppm NO
x
, the maximum rela-
tive error was between 1 and 5%, well within the 20%
window allowed by EPA Performance Specification 2. In
addition, a worst-case scenario model for low NO
x
mea-
surements indicated that NO
x
can be measured with suf-
ficient accuracy.
REFERENCES
1. U.S. Environmental Protection Agency. The Plain English Guide to the
Clean Air Act; EPA-400-K-93–001; U.S. Government Printing Office:
Washington, DC, 1993. Available at: http://www.epa.gov/oar/oaqps/.
2. U.S. Environmental Protection Agency. Performance Specification 2,
Specifications and Test Procedures for SO
2
and NO
x
Continuous Emis-
sion Monitoring Systems in Stationary Sources. In Code of Federal
Regulations, Section 40, Part 60, Appendix B, Protection of Environment;
Office of the Federal Register, National Archives and Records Admin-
istration; U.S. Government Printing Office: Washington, DC, 2001.
3. U.S. Environmental Protection Agency. Determination of Nitrogen
Oxides Emissions from Stationary Sources (Instrumental Analyzer Pro-
cedure). In Code of Federal Regulations, Section 40, Part 60 Appendix A,
Method 7E;Office of the Federal Register, National Archives and
Records Administration; U.S. Government Printing Office: Washing-
ton, DC, 2001.
4. Sigsby, J.E.; Black, F.M.; Bellar, T.A.; Klosterman, D.L. Chemilumines-
cent Method for Analysis of Nitrogen Compounds in Mobile Source
Emissions (NO, NO
2
and NH
3
); Environ. Sci. Technol. 1973,7, 51.
5. Tidona, R.J.; Nizami, A.A.; Cernansky, N.P. Reducing Interference Ef-
fects in the Chemiluminescent Measurement of Nitric Oxides from
Combustion Systems; J. Air Pollut. Control Assoc. 1988,38, 806-811.
6. Mathews, R.D.; Sawyer, R.F.; Schefer, R.W. Interferences in the Chemi-
luminescent Measurement of NO and NO
2
Emissions from Combus-
tion Systems; Environ. Sci. Technol. 1977,11, 1092.
7. Campbell, N.T.; Beres, G.A.; Blasko, T.J.; Groth, R.H. Problems in the
Measurement of Oxides of Nitrogen in Gas Turbine Engine Exhaust.
Presented at the 74th Annual Meeting of the Air Pollution Control
Association, Philadelphia, PA, June 1981; Paper 81-38.3.
8. Berry, R.S.; Martin, J.C.; Dean, C.E. CEMS Analyzer Bias and Linearity
Effects Study; Electric Power Research Institute: Palo Alto, CA, 1998.
Available at: http://rmb-consulting.com/newpaper/cable/cable.htm.
9. Fontijn, A.; Volltrauer, H.N., Frenchu, W.R. Nitrogen Oxides (NO
x
)
Monitor Based on an Hydrogen-Atom Direct Chemiluminescence
Method; Environ. Sci. Technol. 1980,14 (3), 324-328.
10. Baronick, J.D.; Heller, B.; Lach, G.; Bogner, F.; Gerspach, U.; Schimpl,
H.; Gruber, D.; Fabinski, W.; Moede, M.; Zo¨chbauer, M. Evaluation of
an UV Analyzer for NO
x
Vehicle Emission Measurement; Soc. Automot.
Eng. 2001, Spec. Publ. 1588 (Emissions Measurement and Test Meth-
ods), 59–66.
11. Lanier, W.S. Assessment of Bias in Single Digit NO
x
Measurement.
Presented at the EPRI CEM User’s Group Conference, Charlotte, NC,
May 14–18, 2001.
12. Welch, W.A.; Fitz, D.A. Quantification of Uncertainties in Continuous
Measurement Systems for Low-NO
x
Emissions from Stationary Sources;
Revised Draft Final Report for California Energy Commission Contract
No. 700–99-019; University of California, Riverside, Bourns College of
Engineering—Center for Environmental Research and Technology:
Riverside, CA, 2001.
13. MacDougall, D., et al. Guidelines for Data Acquisition and Data Qual-
ity Evaluation in Environmental Chemistry; Anal. Chem. 1980,52,
2242-2249.
About the Authors
The authors work for corporate research and development
at Dow Chemical Co. Steven Gluck is the global technical
leader, Chuck Glenn is an environmental technician, and
Tim Logan is a senior analytical technologist for the envi-
ronmental analytical group. Bac Vu is an analytical special-
ist, Mike Walsh is a senior analytical specialist, and Pat
Williams is an analytical specialist for the process analytical
group. Address correspondence to: Steven Gluck, Dow
Chemical Co., 2301 North Brazosport Boulevard, Building
B-2009, Freeport, TX 77541; phone: (979) 238-7100; fax:
(979) 238-0906; e-mail: sgluck@dow.com.
Table 9. Zero and span drift deviation for CAI NO
x
analyzer (without calibration).
Date and Time Zero Drift (%) Span Drift (%)
10/17/01 02:10 1 1.3
10/17/01 08:15 1 0
10/17/01 14:20 1 1
10/17/01 20:26 1.3 1
10/18/01 02:31 1 1
10/18/01 08:36 1.3 0.7
10/19/01 09:17 0.3 2
10/24/01 09:31 0.7 1.3
Gluck, Glenn, Logan, Vu, Walsh, and Williams
758 Journal of the Air & Waste Management Association Volume 53 June 2003
