EPA/600/A-92/215
Gn-Line Measurement of Nitrous Oxide from Combustion Sources
by Automated Gas Chromatography
Jeffrey V. Ryan
Aeurex Environmental Corporation
Environmental Systems Division
P.O. Box 13109
Research Triangle Park, NC 27709
William P. Linak
U.S. Environmental Protection Agency
Air and Energy Engineering Research Laboratory
Combustion Research Branch, MD-65
Research Triangle Park, NC 27711
Abstract
The combustion of fossil fuels is suspected to contribute to the measured increases in the
ambient concentrations of nitrous oxide (N2O). Characterization of N2O emissions from fossil
fuel combustion and associated pollution control systems has been hindered by a sampling art fact
whereby N2O may be generated from nitrogen oxides, sulfur dioxide, and moisture present in the
sample vessel while these samples await analysis. To truly assess the N2O emissions from fossil
fuel combustion, a real-time or near real-time measurement technique is required. To accomplish
this, a gas chromatograph equipped with an electron capture detector was configured and
automated. This system is capable of detection levels below ambient concentrations and a practical
quantifying range of 0.1 to 200 ppm. A pre-column backflushing system negates the effects of
interferants present in fossil fuel combustion emissions. The automated system is capable of one
on-line measurement every 8 minutes and has been used to evaluate N2O emissions from a variety
of combustion sources, fuels, and post-combustion pollution control techniques.
Introduction
Nitrous oxide (N2O) has been of concern to the combustion community largely because the
combustion of fossil fuels has been proposed to be a contributor to the measured increases in the
ambient concentrations of N2O (Pierotti and Rasmussen, 1976; Weiss and Craig, 1976; Hao et al„
1987). This increase is significant because N2O is considered to be a "greenhouse" gas due to its
infrared radiation absorptive properties as well as an active participant in stratospheric ozone
depletion kinetic mechanisms (Ramanathan et al, 1985).
-------
N2O from combustion sources has been measured using a variety of methodologies
including both grab (container) sampling and on-line monitoring techniques. Once collected, grab
samples are often analyzed using gas chromatography (GC). On-line monitoring methods include
non-dispersive infrared (NDIR), tunable diode infrared laser (TDIR), Fourier transform inflated
(FTIR), and GC techniques. Each of these methods presents its own advantages and
disadvantages.
Grab sampling methods are appealing from cost and convenience considerations.
However, the integrity of samples taken in this manner has been demonstrated to be compromised
under most common combustion sampling conditions (Muzio and Kramlich, 19B8; Muzio et al.,
1989; Linak et al., 1990). This sample artifact has been observed when nitrogen oxides (NO*),
sulfur dioxide (SO2), and moisture, present in most combustion samples, react in the sampling
containers to produce N2O. N2O generation in grab sample containers approaching 200 ppm has
been observed (Linak et al., 1990). However, current research is evaluating the use of modified
sample conditioning techniques that minimizes or negates N2O formation in contained samples.
On-line, real-time N2O analyzers are desirable; however, the commercial availability of
these monitors is limited. Of those available, detection levels may be insufficient, and elaborate
sample conditioning systems are routinely required. Additionally, these systems are largely
research-oriented (Lanier and Robinson, 1986; Kramlich et al., 1988), requiring extensive operator
attention and may not be designed for field use. Several continuous, real-time monitoring
techniques based on infrared radiation absorption have been developed for N2O combustion source
monitoring applications. A NDIR system, developed at the University of California's Irvine
Combustion Laboratory, has been used to characterize the N2O emissions from several pulverized
coal utility boilers (Montgomery et al., 1989). The NDIR system provides real-time measurement
capabilities and is capable of monitoring N2O levels as low as several parts per million. However,
the system is susceptible to interferences from other compounds present in combustion gases that
absorb IR radiation at similar wavelengths as N2O. These interferences can often be minimized
through the use of elaborate conditioning systems as well as electronic background correction.
TDIR spectrometry is another potential real-time monitoring technique. A TDIR analyzer
has been developed by EPA's Air and Energy Engineering Research Laboratory (AEERL) to
monitor N2O emissions from its fossil fuel combustion test facilities (Briden et al., 1991). TDIR
systems offer excellent sensitivity, and are capable of detection levels in the pasts per billion range.
In addition, interferences from other IR radiation absorbing gases can be minimized by the TDIR's
2
-------
ability to isolate appropriate N2O absorbing wavelengths with greater resolution. The main
disadvantages of TDIR spectrometry are its complexity, high cost, short life span of laser diodes
(1-2 years), and the necessity for cryogenic cooling.
FTIR spectrometry is yet another viable real-time N2O monitoring method. The FOR
offers excellent wavelength resolution and sensitivity. In addition, FTIR is capable of monitoring
multiple wavelengths simultaneously, making the technique suitable for real-time, multi-component
monitoring of combustion gases. The main disadvantages of FTIR systems are their complexity
and high cost.
N2O emissions in combustion gases are most commonly monitored using GC. GC
techniques, coupled with electron capture detection (BCD), offer excellent sensitivity with detection
levels less than ambient concentrations (approximately 300 ppb). Other detection methods such as
thermal conductivity detection (TCD) and mass spectrometry (MS) have also been used
successfully. An important limitation of GC methods is that they do not allow a continuous real-
time measurement. Other disadvantages include analytical difficulties and detector de-sensitivity
caused by other compounds present in the combustion gases. Conversely, in addition to excellent
sensitivity, GC methods are typically easy to construct and operate, and relatively inexpensive. A
GC analytical procedure suitable for N2O measurement from combustion gases has been
documented (Ford, 1990).
Realizing that much of the N2O emissions data reported prior to the discovery of the
sampling artifact were at best suspect, the EPA conducted a series of tests characterizing the N2O
emissions from numerous pilot- and full-scale fossil fuel combustion facilities (Clayton et al.,
1989; Linak et al., 1990). Due to the lack of available on-line analytical techniques, the EPA
research group chose to develop a near real-time method based on GC/ECD. During the pilot- and
full-scale tests, the limitations of the on-line GC/ECD system used were identified. These
limitations included susceptibility to interferences present in the flue gases measured and memory
effects from moisture and SO2 resulting in detector baseline instability as well as operational
difficulties. These effects had a direct impact on detector sensitivity, often raising detection levels
to values above actual N2O concentrations present in the gas streams measured. However,
regardless of these limitations and the fact that this technique would never yield real-time
information, the GC/ECD system offered the advantages of being highly sensitive, relatively
inexpensive, widely available, and fairly simple to operate. Therefore, since the development of a
reliable on-line technique was important to the continued characterization of N2O emissions from
fossil fuel combustion sources, EPA/AEERL set out to develop a GC/ECD analytical system and
3
-------
procedure suitable for the on-line measurement of N2O from fossil fuel combustion sources. The
development of the system required negating the effects of interferences present in combustion
process emissions, configuring the instrument for automated operation, and improving the linear
working range of N2O emission quantification.
Experimental
The GC/ECD analytical system developed uses a precolumn backflush method to isola\e the
interfering flue gas components. The system is automated by using the timed event commands
associated with the GC operation/data acquisition system to control and activate the
sampling/valving hardware. N2O is quantified by relating integrated peak areas to a least squares
linear regression of logarithm transformed calibration variables (peak area and N2O concentration).
The system requires that a particulate free, moisture conditioned, sample stream be delivered to the
system under slight positive pressure. Figure 1 is a schematic diagram of the analytical system.
The analytical conditions of the GC/ECD system are presented in Table 1.
Table I. GC/ECD analytical system equipment and conditions.
Gas Chromatograph
Hewlett-Packard Model 5890A
Integrator
Hewlett-Packard Model 3392A
Timed Sample Event Controller
Hewlett-Packard Model 19505A
Detector
63nj constant current cell electron capture detector maintained at 3003C
GC Oven Temperature
Isothermal, 50°C
Carrier Gas
Precolumn
j 5 or 10% methane in argon (PS,.PI0)
j 6 ft 0 8 m) x 0.125 in. (0.32 cm) O D stainless steel, packed with"
| HayeSep D - 100/120 mesh support.
Precolumn Carrier How
1 30 cm^/min (head pressure -30 psig. 308.2 kPa)
Analytical Column
I 12 ft (3.7 m) x 0.125 in. (0.32 cm) O.D. stainless steel, packed with
Porapak Super Q - 80/100 mesh support
Analytical Column Carrier Flow
| 30 cm^/min (head pressure -40 psig, 377.1 kPa)
Backflush Method
The backflushing method uses a single 10-port valve to divert/direct the flow of earner gas
and sample gas streams through the chromatograph system. A schematic diagram of a 10-port
valve is presented in Figure 2. The 10-port valve can be operated in two positions or modes. In
the "off or backflush position (Figure 2a), the precolumn is backflushed by carrier 2 to a vent
4
-------
(ports 10, 9, 6, and 8 consecutively). The analytical column, supplied by carrier 1 (ports 5 and 7
consecutively), is connected to the ECD. A 1 cm3 sample loop, bridging ports 3 and 4, can be
charged with the sample stream (ports 1 and 2 consecutively). In the "on" or analyze position
(Figure 2b), the valve is aligned so that carrier 1 purges the sample loop onto the precolumn (ports
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 pons 10 and 8. The
sample stream is vented via ports 1 and 2. Once the analyte of interest (N2O) has eluted from the
precolumn onto the analytical column, the valve is returned to the backflush position: the flow
through the precolumn is reversed and the interfering sample components are purged from the
precolumn.
An electronically controlled air actuator was used to automate valve switching. The valving
system was controlled by interfacing the GC and integrator to a timed event control module that
converted digital commands from the integrator to time controlled electrical switches. To further
aid in analytical system automation, a solenoid valve, installed upstream of the 10-port valve
sample loop, allows continuous purging of the sample loop with sample gas until the time of
analysis. The valve was controlled so that, just prior to analysis, 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. The system was also configured for unattended,
continuous operation by incorporating the programmed timed events into a separate program
capable of automatically re-initiating the sequence of timed events.
Calibration and Linearity
The linearity of the GC/ECD system was evaluated with varied concentration N2O in
nitrogen span gases ranging from 0.514 to 128 ppmv (Figure 3a). The detector response to N2O
(area counts/ppm N2O) exhibited decreasing sensitivity with increasing concentration. This effect
tends to limit the linear working range of quantitation. The linearity of the detector was evaluated
using two mathematical approaches: a least squares linear regression of the calibration variables,
concentrations, and peak areas; and a least squares linear regression based on transformed (natural
logarithmic) calibration variables. A comparison of these two approaches is presented in Table 2
and Figure 3. The approaches are compared by back-calculating the concentration of each
calibration standard and determining the percent bias from the known value.
5
-------
Table 2. Two mathematical approaches to evaluate detector linearity."
Linear Regression
N2O (known) j Peak area 1 N2O (calc.) Bias %
0.514
In Transformed Linear Regression
[ [ !
ln(N2Q) known ln(Peakarea) | N2O (calc.) | Bias%
-0.66553 8.833171
-8.6
-705.1
0.47
0.97
-319.6
-0.03045
9.484405
0.99
0.688134
10.09406 s 2.02
1.99
-0.40
-120.1
5.03
56075 4.58
8.9
1.615419
10.92444
5.36
9.85
2.287471
1.50567
99278
10.41
5.7
19.4
174880
23.1
19.5
2.965273
3.7
318970 ! 45.74
83.36
40.4
3.698829
80.1
4.383275
559344
13.23451
77.74
816984 ! 123.68
-3.4
4.852030
3.61337
120.79
-5.6
128 I 816984 j 123.68
aN20 concentrations in ppmv.
'Hie lineal" regression of the transformed calibration variables was effective in minimizing
the relative error of calculated concentrations. Less than 10 percent bias was observed over the
entire range. By calibrating in a narrower concentration range, more specific to anticipated
emission concentrations, the relative error can be further reduced.
System Performance
The automated, on-line GC/ECD system was evaluated extensively on a number of diverse
fossil fuel combustion test facilities. The system was used to monitor N2O emissions from the
combustion of various coals during parametric SO2 removal testing. N2O concentrations measured
ranged from 0.5 to 10 ppm.
The on-line GC/ECD system was also used extensively during a series of selective non-
catalytic NOx reduction (SNCR) tests. During these tests, additives such as ammonia and urea
were injected into the combustion test facility to reduce NOx emissions. The on-line measurements
were used to compare N2O emissions with and without NOx control. The NoO concentrations
measured ranged from 0.5 to 35 ppm, and compared well with measurements taken by a tunable
diode laser. Similarly, the 011-line GC/ECD system was used to characterize the N2O emissions
from a selective catalytic NOx reduction (SCR) pilot-scale test facility. NoO concentrations were
6
-------
measured both before and after the catalyst evaluated. Measured concentrations ranged from 0.5 to
3 ppm.
The GC/ECD system was also evaluated under ambient conditions. The system was used
to assess the N2O mass emissions resulting from the open hearth combustion of Chinese coal. I11
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 N2O
to the environment from this combustion source. The NoO 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.
During on-line analyses, span or performance checks were conducted on a routine basis.
These checks, used to evaluate method accuracy and precision, were conducted at various times
during the measurement process. Figures 4 and 5 present results of span checks conducted during
representative ambient and source monitoring activities. Overall, the accuracy and precision levels
achieved during various on-line monitoring requirements were consistent. The type of combustion
source monitored did not appear to affect method performance. Accuracy of span checks,
expressed as percent bias, was consistently less than 15 percent. The precision of replicate span
checks, expressed as percent relative standard deviation (RSD), was consistently less than 10
percent.
Summary
The GC/ECD backflush method developed was found to be suitable for the measurement of
N2O 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. Analytical interferences,
present in combustion process effluents, were negated through the use of a backflushing
technique. Common flue gas components such as NOx, SO2, oxygen, carbon monoxide, carbon
dioxide, moisture, unburned hydrocarbons, and ammonia were not found to interfere with the
analytical procedure. Method accuracy (percent bias) and precision (percent RSD) were
determined to be less than 15 and 10 percent, respectively. The method was found to be suitable
for the quantitation of N2O concentrations ranging from 0.100 to 200 ppm. Ultimately, the
procedure was approved as an AEERL Recommended Operating Procedure (ROP) (Ford, 1990).
Using this method for on-line monitoring purposes allows a semi-continuous measurement
approximately once eveiy 8 minutes. The system can be easily incorporated into most continuous
7
-------
emission monitoring sample delivery/conditioning systems, the only requirements being the
removal of particulate and moisture from the sample stream by a refrigeration condenser. The
sample stream should be diverted to the analytical system prior to further moisture conditioning by
a desiccant.
The non-linear response of the detector to N2O at low concentrations was minimized
through use of a logarithmic transformation of the calibration variables. The transformed data ire
used to derive a least-squares linear regression.
Acknowledgments/Disclaimer
Portions of this work were conducted under EPA contract 68-DO-0141 with Acurex
Environmental Coiporation. The authors would also like to thank R.E, Hall, Chief, Combustion
Research Branch, AEERL, for his support. This article has been reviewed by the EPA's AEERL,
and approved for publication. The contents of this article should not be construed to represent
Agency policy nor does mention of trade names or commercial products constitute endorsement or
recommendation for use.
References
Briden, F.E., Natschke, D.F., and Snoddy, R.B., The practical application of tunable diode laser
infrared spectroscopy to the monitoring of nitrous oxide and other combustion process stream
gases, 1991 Joint Symposium on Stationary Combustion NOx Control, Washington, DC, March
1991.
Clayton, R.. Sykes, A., Maehilek, R„ Krebs, K„ and Ryan, J., N20 field study, EPA-600/2-89-
006 (NTIS PB89-166623), February 1989.
Ford, J., Recommended operating procedure no. 45: analysis of nitrous oxide from combustion
sources, EPA-600/8-90-053 (NTIS PB90-238502), June 1990.
Hao, W.M., Wolfsy, S.C., McElroy, M.B., Beer, J.M., and Toqan, M.A., Sources of
atmospheric nitrous oxide from combustion, J. Geophys. Res., 92, 3098-3104, 1987.
Kramlich, J.C., Lyon, R.K., and Lanier, W.S., EPA/NOA A/NASA/USD A N2O workshop,
volume I: measurement studies and combustion sources (September 15-16, 1987, Boulder, CO),
EPA-600/8-88-079 (NTIS PB88-214911), May 1988.
Lanier, W.S., and Robinson, S.B., EPA workshop on N2O emission from combustion (Durham,
NC, February 13-14, 1986), EPA-600/8-86-035 (NTIS PB87-113742), September 1986.
Linak, W.P., McSorley, J.A., Hall, R.E., Ryan, J.V., Srivastava, R.K., Wendt, J.O.L, and
Mereb, J.B., Nitrous oxide emissions from fossil fuel combustion, J. Geophys. Res., 95(D6),
7533-7541, 1990.
8
-------
Montgomery, T.A., Samuelsen, G.S., and Muzio, L.J., Continuous infrared analysis of N2C1 in
combustion products, J. Air Pollut, Control Assoc., 39(5). 721-726, 1989.
Muzio, L.J., and Kramlich, J.C., An artifact in the measurement of N2O from combustion
sources, Geophys. Res. Letters, 15, 1369-1372, 1988.
Muzio, L.J., Teague, M.E., Kramlich, J.C., Cole, J.A., McCarthy, J.M., and Lyon, R.K.,
Errors in grab sample measurements of N2O from combustion sources, J. Air Pollut. Control
Assoc., 39(3), 287-293, 1989.
Pierotti, D., and Rasmussen, R.A., Combustion as a source of nitrous oxide in the atmosphere,
Geophys. Res. Letters, 3, 265-267, 1976.
Ramanathan, V., Cicerone, R.J., Singh, H.B., and Kiehl, J.T., Trace gas trends and their
potential role in climate change, J. Geophys. Res., 90(D3), 5547-5566, 1985.
Weiss, R.F., and Craig, H., Production of atmospheric nitrous oxide by combustion, Geophys.
Res. Letters, 3, 751-753, 1976.
9
-------
To CEMs
From Stack
Refrigeration
Sample Line
115 VAC
N20 Span Gas
Compressed Air
115 VAC
10 Port Valve \*a
P5 Carrier Gas
QPD
P5
Carrier
Gas
, Precolumn..
6 ft (1.8 m) Analytical
Oolumn
12 ft (3.7 m
Sample Event
Control Module
Integrator
Figure 1. Automated on-line GC/ECD N2O monitoring system.
-------
(a) "off" or backflush position
(b) "on" or analyze position
FROM CARRIER 1
LOOP
PRECOLUMN
OUT
SAMPLE
VENT
COLUMN
TO
DETECTOR
FROM CARRIER 1
>
LOOP
PRECOLUMN
SAMPLE
OUT
IN
VENT
COLUMN
00
TO
DETECTOR
FROM CARRIER 2
FROM CARRIER 2
Figure 2. 10-port sampling valve.
-------
1000000
(a) Detector response to calibration gases
Peak area = 6389.233 (N20) + 26755.25
800000 J r2 = 0.993537
O Calibration standards (known)
600000
400000
200000
O,
0
14
13
12.
11 .
—i f i \ i I i I | 1 » f t | i I I—i j" i i I i"" |—i—I » i
0 25 50 75 100 125 150
N20 (ppm)
(b) Calibration transformation
In (Peak area) = 0.8597 In (N20) + 9.491897
r2 = 0.999238
O Calibration standards (known)
In (NpO)
Figure 3. ECD response to N20 calibration gases.
12
-------
7
Q.
a.
O
CM
(0
(0
c
a}
E
CD
Q_
Calibration checks (Illinois No. 2)
9.85 ppm full scale
ave. stack N20 conc. = 5.6 ppm
8
O n20 calibration (measured)
- N20 calibration (known)
8
0 50 100 150 200
Time (min)
O Calculated bias
250 300
Figure 4. Illinois No. 2 coal calibration checks, downfired furnace/stack sampling.
13
-------
1.1
1.0.
Calibration checks (open hearth China coal)
0.97 ppm full scale
ave. room N20 conc. = 0.45 ppm
O
CL
Q.
o
eg
0.9.
O n20 calibration (measured)
- N20 calibration (known)
0.8.
15
t 1 r
_1 1 1 , , , r
O Calculated bias
10
to
co
c
CD
2
------- |