Continuous Infrared Analysis of N 2 0 in Combustion Products

Nitrous oxide (N20) levels In the atmosphere are Increasing, poten tially contributing to the greenhouse effect and depletion of strato spheric ozone. From a limited data base, combustion sources have been Identified as a major anthropogenic source of N20. However, the existing data base (obtained by tradltlonal grab sampling tech niques followed by gas chromatographic analysis) Is In question due to the discovery of a sampling artifact. A continuous on-line N 2 0 analyzer would enable and facllltate the accurate characterization of combustion sources over a range of operating conditions, and also aid In the development of an appropriate sampling technique. This paper addresses the development of a continuous measure ment technique, and the evaluation and Initial use of a fleld proto type continuous N20 analyzer developed at the UCI Combustion Laboratory In cooperation with a major Instrument manufacturer. The analyzer Is capable of measuring N20 levels down to a few ppm. The analyzer has been evaluated and used to study the N 2 0 emissions from a pulverized coal-fired boller. The N 2 0 levels found with the analyzer are substantlally lower than· levels previously attributed to such sources. Initial N20 measurements made with the analyzer suggest that N 2 0 levels are not a substantial fraction of the NO. levels, as previously suggested.

Attention has recently focused on the emission of nitrous oxide (N20) from fossil fuel combustion processes. Although the emissions and emission formation mechanisms of NO.
(NO + N02) are relatively well documented, the data base on nitrous oxide formation and control is small with relatively limited information available in the literature.
N20 is suspected to have a two-fold effect in the atmosphere. In the troposphere, the species is stable with a lifet ime of approximately 150 years. 1 Here N 2 0 is a likely contributor to the greenhouse effect2,3 with the global mean concentration increasing between 0.2 and 0.4 percent a year. 2 The stability of N 2 0 facilitates its transport to the stratosphere. In the stratosphere, N 2 0 is the largest source of nitric oxide, which subsequently results in the destruction of ozone (03). 4 Hence, an increase in N 2 0 emissions is expected to translate to an increase in 0 3 depletion.
. ~thropogenic sources, including the use of chemical fer-t1J1zers and the combustion of fossil fuels 3 are estimated to comprise one-third of the total N 2 0 prod~ced. l Usi~g the lim~ted data available, which are based on grab sampling techmques, NzO production was originally estimated to be between 20-25 percent of the NO. levels. [4][5][6] May 1989 Volume 39, No. 5 these levels, combustion sources have been considered to be t he largest source of anthropogenic nitrous oxide 4 •5 with flue gas concentrations normally varying between 1 and 200 ppm. 5 Measurements from pilot scale facilities have been reported to be as high as 400 ppm. 7 However, recent findings suggest that grab samples containing NO, S02, and condensed H 2 0 produce N 2 0 as an artifact. 8 • 9 As a result, the N 2 0 levels reported to date may be higher than the levels actually produced directly by pulverized coal combustion systems. Providing a valid grab sampling technique is available, the method is tedious, making the complete characterization of a combustion source difficult.
An accurate, on-line continuous N 2 0 measurement technique is needed to (1) obtain reliable measurements in the field; and (2) assess the extent to which combustion sources directly produce N20· The present paper reports an investigation of means by which N 2 0 can be continuously monitored in combustion effluents. The goal is to identify, evaluate, and demonstrate a measurement method which is susceptible to ~i nimal interferences and yet relatively simple, compact, reliable, and suitable for both field and laboratory use.

Approach
In addition to gas chromatography, various analytical techniques have the capability of measuring nitrous oxide concentrations. 5 These include: Tunable Diode Laser System, Fourier Transform Infrared Spectroscopy (FTIR), and Infrared (IR) Analyzers.
A Tunable Diode Laser System allows for continuous monitoring with the potential of avoiding interferences from other species in the combustion products due to the very fine spectral resolution of the technique. Major disadvantages (system cost and the requirement of cryogenic cooling) render the technique difficult for field application. An FTIR analyzer has the potential of providing sufficient accuracy. However, FTIR is moderately expensive, and while providing ~elatively rapid sampling and analysis, is not truly a contmuous analysis method.
In comparison, infrared (IR) analyzers are relatively simple and inexpensive, and Non-Dispersive InfraRed (NDIR) analyzers are ~outinely used for continuously measuring CO, C02, and S02 m combustion systems. Since N 2 0 is an active infrared species, an analyzer based on IR analysis is a logical starting point in developing a simple analysis system. Be-  cause N 2 0 has a number of clearly defined peaks in its absorption spectrum, the particular methodology selected as a point of departure in the present study was selective NDIR at a specific, single wavelength,(versus nonselective NDIR).

Experimental Apparatus
To assess the viability of continuous N20 analysis in combustion effluents with a selective, single wavelength NDIR technique, and especially its sensitivity to interfering gases, a variable wavelength infrared analyzer (Foxboro Miran IA) . was used ( Figure 1). Assuming the suppression of excessive interferences from other gases, the instrument can be calibrated to detect and measure a specific, infrared-active gas over a fairly wide concentration range. A parametric variation was conducted with the Miran IA using simulated combustfon products. High Purity gases were blended to investigate (I) linearity of the analyzer with respect to N20 (as a function of absorption wavelength, pathlength, and slit width); and (2) interferences from NO, N02, CO, C02, S02, and H20. These interferences were evaluated over the gas concentration ranges listed in Table I.

Results
The infrared absorption spectrum of N20 (obtained with an FTIR) is shown in Figure 2 and exhibits IR absorption peaks at 4.5, 7 .8, and I 7 .0 microns, the primary N 2 0 absorption lines being 4.5 and 7.8 microns. 1 0 As a result, these two lines were selected to evaluate the linearity of the analyzer response to N20. Pathlengths were chosen at each wavelength to provide sensitivity to N 2 0 while maintaining linearity between N20 and IR absorption. Slit width (spectral bandwidth) was chosen to minimize interferences from the other IR gases expected to be present in the effluent. Once the linearity of the instrument response to N 2 0 was verified, interference curves for C02, CO, NO, N02, and S02 were generated by observing the response of the analyzer to the introduction of various gas mixtures.
Typically N 2 0 levels of 400 ppm resulted in absorption of 20-25 percent. Figure 3 compares the interferences found for the species expected in combustion products. At a wavelength of 4.5 microns, the principal interfering species are C0 2 and CO. At 7.8 microns, N0 2 and S02 are the primary

24
The effect of trace amounts of water, corresponding to a I -2 degree Celsius dew point, was also evaluated. An initially dry gas blend was passed through an ice bath prior to analysis, simulating the amount of water anticipated to be present in actual samples if the gases were dried in a refrigerated dryer. Water interferences can be accounted for by either subtracting the effect of water after the measurement is made or passing the zero and calibration gases through an ice bath prior to analysis. Drierite was also put in series between the ice bath and the analyzer to remove the excess water prior to analysis with no effect upon nitrous oxide levels.
To further reduce the error introduced from the interferences at the 7.8 micron wavelength, sodium sulfite and sodium carbonate solutions were used to condition the sample by scrubbing the N02 and 80 2 , respectively. As shown in Figure  4, a sodium sulfite solution effectively removes the N0 2 and a sodium carbonate solution effectively removes the S0 2 from the sample while having essentially no effect upon nitrous oxide levels. By scrubbing the N0 2 and 80 2 from the sample, the over prediction of N 2 0 levels can be reduced  from 6 to 1 ppm for a low sulfur, low NOx fuel and from 17 to 1 ppm for a high sulfur, high NOx fuel (assuming no corrections have been made for interferences). The results described above outline the requirements and criteria for a continuous analyzer utilizing infrared absorption for measuring NzO in combustion effluent. The key points from this study are as follows: • The 7.8 micron region in the infrared is most suitable for measuring N 2 0 in combustion products. • At 7.8 microns, the primary interfering species are S02 andN02. • 80 2 and N02 can be readily removed from the sample stream ahead of the analyzer without affecting NzO levels. • A precision of better than 3 percent can be achieved based on reported N 2 0 levels from coal combustion products. Based on these criteria, a field prototype instrument was designed and built in 'cooperation with a major instrument manufacturer. The evaluation of the field prototype instrument including both instrument performance and sensitivity to interferences is described below. In addition, recent field measurements from two coal-fired utility boilers are presented to establish (1) the instrument performance in the field; and (2) representative emission levels of NzO from practical coal-combustion sources.

Prototype Analyzer Description
The prototype analyzer was built by HORIBA Ltd. as an adaptation of the HORIBA Model VIA-500. The analyzer is a nondispersive infrared analyzer utilizing a 500 mm sample cell. A schematic of the analyzer is shown in Figure 5. Replaceable optical filters allow the instrument to measure NzO either in the optical region around 4.5 microns, or 7.8-8.5 microns. To date, the 7.8-8.5 micron region has been used. To minimize interferences, the analyzer uses two Luft-type detectors, containing NzO, in series. The first, or primary detector, senses NzO and any interfering gases that absorb in the 7.8-8.5 micron region. Since the first detector absorbs all of the radiation from the N 2 0 bands, the radiation reaching the second detector is only that due to interfering species . . The second detector then senses the interfering species and electronically compensates for their effect.
The initial specifications for the analyzer included two ranges: 0 to 250 ppm and 0 to 500 ppm. This specification was made at the time NzO concentrations up to 400 ppm were anticipated, prior to the identification of an artifact associated with the use of grab samples.8,9 However, the analyzer is capable of operating on a 0 to 25 ppm full scale range with sufficient signal to noise ratio to detect 0.5 ppm changes in N 2 0 level.

Prototype Analyzer Characterization
Simulated combustion products were used for the initial evaluation of the prototype continuous N 2 0 analyzer. High purity gases were blended to (1) verify the linearity of the instrument; and (2) quantify the extent of interferences from C02, CO, NO, N02, and S02.
The linearity of the analyzer response to NzO was evaluated first. The analyzer responds linearly to NzO concentrations ranging between 0 and 250 ppm for the low range of the analyzer. The analyzer provides an excellent signal-to-noise level even at very low NzO levels ( < a few ppm).  Subsequently, the extent of the interferences produced by typical coal combustion products was evaluated. Interference curves were generated for CO levels between 6 and 500 ppm; C02, 0 and 20 percent; NO, 0 and 1250 ppm; 80 2 , 0 and 4000 ppm. In general, the results correspond to those obtained with the Miran IA; S0 2 is the primary interfering species. The analyzer utilized some construction materials that adsorb N02, which precluded generation of an interference curve for this species. Because of the N02 adsorption within the analyzer and the extent of 80 2 interferences, N0 2 and S02 removal from the sample stream is desirable.

Field Application
Uport completion of the initial laboratory evaluation and validation of the prototype analyzer, the analyzer was used to make measurements in utility boiler combustion effluents. This field effort was intended to (1) verify the suitability of the analyzer for field use; and (2) begin characterization of the N 2 0 emissions from full scale combustion field sources, particularly pulverized coal fired utility boilers.
Two combustion engineering units burning pulverized coal were tested. Both units are tangentially-fired and rated at 790 MW (Mega Watts). For these measurements, the NzO analyzer was integrated into a continuous gas analysis system being used to make continuous measurements of 02, C02, CO, NOx, and 80 2 at the site. Figure 6 shows a diagram of the sampling system and illustrates the way in which the N 2 0 analyzer was integrated into the existing system.
The gas samples were obtained from the ducts between the electrostatic precipitators and the stack. The gas sam- ples were drawn through stainless steel probes, passed through a heated filter, dried in an ice bath, and transported through a% inch unheated sample line (approximately 300 feet in length), further dried through a refrigerated dryer, and then distributed to the analzyer. As shown in Figure 6, the sample for continuous N20 analysis was obtained downstream of the 80 2 analyzer.
In addition to continuous analysis, grab samples for N 2 0 were obtained at the inlet to the N 2 0 analyzer and directly from the probes at the sample location (refer to Figure 6).
Continuous N20 measurements. The average values for the continuous gas measurements obtained at each unit are summarized in Table II. As seen in Table II, N 2 0 levels from the continuous analyzer were low at each unit, less than 1 ppm. The NOx levels at both units ranged between 275 and 419 ppm. These measurements provide evidence that, at least from these two utility furnaces, direct N 2 0 emissions from coal combustion are low and, in fact, not a large fraction of the NOx levels (e.g., N20/NOx < 0.4 percent), as previously reported.  testing of the two units. The grab samples were obtained two ways; using a "standard" flask sampling technique, and the procedure suggested by Muzio and Kramlich.8,9 The "standard" sampling technique simply captures flue gas in a clean, dry glass sample flask. The method suggested by Muzio and Kramlich8 and Muzio et al. 9 involves sampling into a flask containing 5 cc. of 10 N NaOH which effectively ties up the 802 and increases the sample pH. As shown by Muzio et . al., 9 this inhibits generation of N 2 0 in the sample flask.  Table III compares the results obtained with the continuous analyzer to those obtained from the sample flasks containing NaOH. The grab sample analyses showed N20 levels below 2 ppm, which is the detectability limit of the analytical technique. Agreement between the two different analysis May 1989 Volume 39, No. 5 methods is good. Table III also compares the two different grab sampling techniques. T he N20 value obtained from the flask without NaOH is higher, once again indicating N 2 0 generation in the sample flask.

Concluslons
The following conclusions can be drawn from the results d iscussed above: • The continuous analyzer system designed and built for the detection of N 2 0 is capable of N 2 0 analysis in combustion effluents at levels down to a few ppm. • Field tests of the continuous N20 analyzer demonstrated its suitability for continuous source testing. • N 2 0 measured with the continuous analyzer at two 790 MW pulverized coal fired boilers showed levels less than 1 ppm. These levels are substantially lower than the fraction of NOx levels suggested by data previously obtained using grab samples. • T he N20 levels were in agreement with grab samples using a grab sampling technique to inhibit N 2 0 generation in the sample flasks.