(12) Brunelle, M. F.. Dickinson, J. E . , Hamming, W. J., “Effec-
tiveness of Organic Solvents in Photochemical Smog Formation,” APCD of the County of Los Angeles. Los Angeles, Calif., 1966. (13) Ralph. H . D., Walker, G. D., “Evaluation of Gaseous Fuels Supply for Motor Vehicle Usage in the Los Angeles Basin, PACE Co., Houston, Tex.. 1971. (14) Korth, M.W.. Rose, A . H.. Stahman. R. C.. J. Air Poilut. Contr. Ass., 14, 168 (1964). (15) Altshuller, A. P., Kopczynski, S. L., Lonneman, W. A., Sulterfield. F. D., IVilson, D. L., Enciron. Sei. Techno/.. 4, 503 (1970). (16) Larsen, Ralph I., J . Air Pollut. Contr. Ass., 11(2), 71-6 (1961).
(17) Glasson, W. A , , Tuesday, C. S., Enuiron. Sei. Technol., 4, 37-44 (1970). (18) Ulbrich, E . A , , Socio-Econ. Plan. Sei., 1,423-40 (1970). (19) HEW (Department of Health, Education, and Welfare), “Control Techniques for Carbon Monoxide, Nitrogen Oxide and Hydrocarbon Emissions from Mobile Source,” Washington, D.C., 1970. (20) National Academy of Sciences, “Semiannual Report by the Committee on Motor Vehicle Emissions,” Washington, D.C., 1972.
Receiced f o r recieu’ October 17, 1972. Accepted April 24, 1974.
Problems with Flame Ionization Detectors in Automotive Exhaust Hydrocarbon Measurements Keith Schofield’ Delco Electronics-Santa
~
~
Barbara Operations, Calif. 9301 7
~~~~
Three of the major problems encountered in measuring concentration and mass emissions of hydrocarbons from engines using flame ionization detectors (FID’s) have been studied in detail. The magnitude of possible correlation discrepancies for analyses run on various commercial instruments as a function of operating conditions and exhaust type has been estimated from measured relative molar response values. Large variations are found under certain circumstances. Results also imply t h a t the larger analyses values occasionally observed with heated analyzers may result from increased relative molar responses to the various individual hydrocarbons rather than decreased line absorption effects. The magnitude of the oxygen interference effect (synergism). important for real-time a u tomotive testing, has been measured on various FID’s a t different conditions. A simple modification to minimize the effect is recommended for the majority of analyzers that exhibit this feature. Finally, a n extensive study of instrument response times has been completed for a variety of sampling lines. The relative importance of such factors as t h e analyzer operating temperature, sample line temperature, instrument sample pressure, bypass flow rate, sample line length, diameter, and material has been established. Optimum response was realized with a heated FID using heated 1b-in. o.d. stainless steel or Teflon lines and a high bypass sample flow rate. The flame ionization detector (FID) now is generally accepted as being the best analyzer for integrating the total carbon content of a mixture of hydrocarbons. The instrument’s ability to respond in a n approximately uniform manner t o saturated, unsaturated or aromatic carbon atoms in a variety of organic structures has provided a means for realistically monitoring the total concentration or mass emission of unburned hydrocarbons from combustion engines. Little changed since its introduction about 15 years ago. flame ionization detectors currently are available in a variety of about 15 commercial total hydrocarbon analyzers. Of these, five are high-temperature 1 Present address. ChemData Research. 2800 \Villiams \Yay, Santa Barbara. Calif. 93105
826
Environmental Science & Technology
models for use where high molecular weight, high boiling point hydrocarbons may be encountered. Their analyzer train, the sample line, pump, filter, valving and burner assembly are all heated to temperatures up to about 400°F. Although all these units are basically very similar and involve measuring the extent of the chemi-ionization produced when a hydrocarbon sample is injected into a small hydrogen diffusion flame, certain concerns have been expressed that different models may not correlate exactly when measuring the complex mixture of hydrocarbons emitted from gasoline or diesel engines. This concern is quite understandable since differences in flame composition, burner and electrode design, and a lack of specific recommended operating parameters provide many variations which could conceivably affect the relative responses to the different hydrocarbons and so alter the resulting integral measure. In a period when the magnitude of exhaust analyses are of prime importance to various concerned groups, a lack of instrument correlation is of particular interest. Consequently some effort has been directed toward establishing the extent of such variation and conveying how closely the integral measure approximates the true exhaust hydrocarbon content. As the need for accurate real time and integral concentration and mass measurements of vehicular emissions has developed, several operational problems have become evident. One concerns the different FID response observed for the same hydrocarbon sample in either air or a n inert background gas. Since raw exhaust is neither one nor the other. the analyst is confronted with the dilemma of how to calibrate the instrument. Another centers upon the response time of the sampling system and analyzer. Since the monitoring instrumentation must be somewhat removed from the vicinity of the automobile dynamometer test area. it is important that source concentration changes be not degraded by the sampling system. Such problems form the basis for this paper a n d will be considered individually in detail. I t is surprising for a n instrument t h a t performs so well, combining the desired features of linear response. high sensitivity, good stability, and a n insensitivity to minor changes in operating parameters, that our basic understanding of the flame processes is still quite poor (I).
Suitability of FID for Vehicular Exhaust Analyses
It may be appropriate initially to illustrate why t h e FID became so widely accepted as the total hydrocarbon analyzer for automotive applications. With this detector the hydrocarbon concentration of a vehicular exhaust generally is expressed in terms of t h e equivalent concentration of propane. I t is quoted in terms of parts per million carbon equivalent (i.e., propane equivalent x3). T h e implicit assumption is t h a t all exhaust hydrocarbons have equal FID responses per unit carbon, equal in fact to t h a t of t h e carbon atom in propane. This not being strictly t h e case, the measured concentration will only approximate the true hydrocarbon cencentration. T o illustrate the extent of possible divergencies a n d the ability of the instrument to sum such complex mixtures realistically, consider t h e case of the two different type hydrocarbon emissions illustrated in Table I. These represent gas chromatographic analyses of hydrocarbon emissions from reciprocating and rotary type engines. The former contains hydrocarbons resulting mainly from incomplete combustion, whereas the latter is more representative of the gasoline fuel. The hydrocarbons present only in trace amounts have been summed together a t the bottom of t h e table and listed as “Other Paraffins,” and so forth. T h e number of contributing species t o each such sum is shown in parentheses. Sternberg et al. (2) were instrumental from the start in establishing tables of the effective instrument relative sensitivities to various hydrocarbon types. Their values have changed little a n d relative to methane taken a s 1.0, t h e unit carbon responses for hydrocarbons a r e generally considered as aliphatic
Table I. Representative Volume Compositions of Hydrocarbon Component of Vehicular Emissions ( 4 ) Reciprocating engine,
%
Methane Acetyl en e Ethylene Propyl e ne To1uene 1-Butene,1,3-butadiene Benzene Ethane 3 Ethyl pentane, 2,2,4-trirnethyl pentane I sopenta ne n-Butane, 2,2-dimethyl propane 2,3- and 3,3-Dimethyl hexane, 2,3,3and 2,3,4.trimethyl pentane p,m-Xylene Propad ien e 0-Xylene, phenyl ethylene cis.l-Phenyl-l.propene, t-butyl benzene, 1,2,4-trimethyl benzene 2,3-Dimethyl pentane, 2-methyl hexane 2-Methyl pentane 1-Methyl 3- or 4-ethyl benzene Ethyl benzene cis-2- Butene 1-Methyl-2-ethylbenzene, 2. phenyl.1-propene 2,4-Dimethyl pentane, 2,2,3trimethyl butane Other paraffins Other aromatics Other olefins
24.27 17.51
14.12 7.34 5.97 4.07 2.15
1.97
Measured Relative Molar Responses
Rotary engine,
%
4.88
3.30 8.09 5.34 16.34
2.99 1.31 1.32
1.59
2.89 8.64 4.51
1.54
2.73
1.30 1.00
5.57
0.76
2.67
0.76
2.45
0.64
2.86
0.55
1.36 2.31 1.67
1.81 1.73
0.95
0.52 0.51 0.51
0.25
0.45
1.07
0.38
1.32 6.26 (29) 3.78 (20) 5.11(76)
4.07 (26) 1.11(20) 3.37 (62)
0.94, olefinic 0.90, aromatic 0.89, and acetylenic 1.30. These approximate values will be used initially in this preliminary illustration. Their variations with operating conditions will be described in detail later. When we integrate over t h e spectrum of exhaust hydrocarbons listed in Table I, it can be concluded t h a t for these mixtures. a n FID calibrated with propane will record a parts per million carbon equivalent t h a t is about 1.7% higher than t h e true value for t h e reciprocating engine and about 2.8% lower for t h e rotary engine case. By means of a dynamometer and a constant volume sampler, concentration measurements can be converted into hydrocarbon mass emissions per vehicular mile. However, in so doing, a n additional assumption invoked is that a mean density (16.33 g/ft3 per carbon atom at 68°F and 1 a t m pressure) may be taken for the hydrocarbon mixture. T h e mean density of the two exhausts considered above can readily be calculated from the d a t a in Table I and are 16.33 and 16.15, respectively. Consequently. mass measurements in these cases would be 1.7% too high and 1.770 too low. respectively. This illustrates how remarkably well suited the FID is for realistically integrating the concentration and mass of the complex hydrocarbon mixture emitted from internal combustion engines. True variations may be slightly more or less than these figures. dependent upon the actual relative molar responses for the particular instrument used. However, analyses are not expected to be seriously in error a t present. This results from the larger relative response to acetylene being counterbalanced by t h e lower responses of olefinic and aromatic hydrocarbons. Nevertheless. with the advent of catalytic converters and emission controls, which profoundly simplify the nature of the hydrocarbons present in the exhaust, this may not continue to be strictly the case.
T o better illustrate the extent of possible intermodel variations with various type exhausts, t h e relative carbon sensitivities have been measured for several hydrocarbons on three FID models operating at a variety of conditions. The individual hydrocarbons investigated were methane (CHI), ethylene ( C Z H ~ ) ,acetylene ( C Z H Z ) . toluene (C7Hs). and propane (C3H8). With the exception of propane, used solely as a reference species. these constitute the single most important group of organic compounds emitted from vehicles and together can represent typically from 30-907~by volume of t h e total, depending on the engine type and emission control features. The analyzers available to this program were manufactured by Beckman Instruments, Inc.. and consist of a Model 402 (analyzer train heated in t h e range 200-400”F), a Model 400 (thermostatically controlled a t 120”F), and a Model 108A ( a m bient temperature operation). Although from the same manufacturer. these are quite differently designed instruments, varying in burner and electrode design, and operating temperatures. They should typify generally available commercial instruments. More recent work appears to confirm this conclusion ( 3 ) .The Model 400 was operated in both its normal and a modified mode. The latter, to be described in more detail, consisted of a modification to minimize the response difference generally noted between equivalent samples of the hydrocarbon in either air or a n inert background gas. Individual gas cylinders of 2000 ppm by volume of methane (CH4), 980 ppm ethylene ( C Z H ~ ) 900 , ppm acetylene (CzHz), 310 ppm toluene (C7Hg). and 500 ppm propane (CsHs), all in air were obtained as primary gas standards. Their relative responses were measured on the three analyzers. The heated model was operated at either 200” or 400°F and all models Volume 8. Number 9 , September 1974 827
~~~
~~
~~
~~
Table II. Flame Ionization Detector Molar Responses for Several Model 40P
Model 400
196OF
391'F H./He
HIIN?
H?/He
~~~~~
"
Normal
._
HI,"?
HIIN.'
-~
~
HI/Hed
Hydrocarbon
5'
3
5
3
5
3
5
3
5
3
5
Methane, CH, Ethylene, C2H, Acetylene, C2H. Toluene, C;H. Propane, C H,
1.096 1.781 2.852 6.339 3.000
1.057 1.903 2.788 6.382 3.000
1.051 1.963 2.801 6.940 3.000
0.996 2.031 2.677 6.899 3.000
1.055 1.987 3.156 6.865 3.000
1.004 2.037 3.022 6.935 3.000
0.994 2.056 2.938 7.395 3.000
0.949 2.058 2.974 7.403 3.000
1.169 1.718 2.954 6.398 3.000
1.072 1.837 2.623 6.386 3.000
1.063 1.915 2.690 6.909 3.000
Fuel 30 psi, Air 20 psi.
'' Fuel 30
psi, Air 30 psi.
Fuel 30 psi, Air 13 psi (Normal), 10 psi (Modified).
burned 40% hydrogen/60%nitrogen or 40% hydrogen/6017o helium standard FID mixed fuels. Fuel/air flows were typically of the order of about 100 and 250-400 cm3 m i n - l , respectively. Sample addition rates were either a t their maximum setting, 5 psi gauge pressure. which corresponds to a flow of about 18 cm3 m i n - l or a t 3 psi (10 cm3 m i n ~ - l ) The . instruments fue1,'air settings were those for which a maximum signal could be produced. This criterion for locating the optimum flow rates produces a n instrument with maximum response and has been shown to be least sensitive to slight instabilities in the fuel or air flows a t such settings. The measured responses to the five hydrocarbons. suitably corrected so that each refers to the same molar concentration of hydrocarbon are listed in Table I1 for the various cases considered. The values represent. relative to propane taken as three, the effective number of carbon atoms in the hydrocarbons structure as monitored by the FID. The values of these effective carbon number,'molecule fall in the generally accepted range; that is. methane (1.1). ethylene (1.91, acetylene (2.81, toluene (6.61, relative to propane (3.0) but are seen to vary extensively between instruments and for different conditions. Sone of the d a t a was degraded by instrumental nonlinearity problems. The accuracy with which the cylinder concentrations Table 111. Selected Hydrocarbon Exhaust Compositions -Percentage by Volume of the Total Hydrocarbon Emission 350 C!D V8 engine EPA cycles Hydrocarbon
13-18
Wankel engine EPA 1st cycle
Methane Ethylene Acetyl en e To1 ue ne % Representation of total hydrocarbon emission
10.7 22.3 14.5 7.0 54.5
Prototype engine EPA cycles
EPA cycles
3-10
1-18
4.9 8.1 3.3 16.3
64.2 24.4 0.9 1.9
47.1 22.6 8.1 2.0
32.6
91.4
79.8
3
0.999 1.994 2.532 6.921 3.000
Fuel 30 p s i , Air 16 p s i (Normal), 1 4 p s i (Modified).
are known only affects the results insofar as their absolute molar response scale. Any such error would affect all the data for a particular hydrocarbon by the same factor. Since it has already been noted that the effective molar responses do fall within the usual prescribed ranges, the cylinder concentrations appear to be sufficiently accurate for this program. The range of variations for a specific hydrocarbon, which is of prime interest here, will be unaffected by such considerations. The spread of the data in Table I1 is extensive and shows a maximum scatter in values of 25% for methane, 2 1 7 ~for ethylene, 24% for acetylene, and 17% for toluene between the highest and lowest recorded sensitivities for each. General trends can be observed which may be important in explaining quantitative correlation differences that have been observed in exhaust analyses. For example, whereas the heated Model 402 operating a t 400°F has the largest responses of all to ethylene, acetylene, and toluene. it is the least sensitive to methane. At 200°F it responds not too differently from t h e Model 400. On hydrogen-nitrogen mixed fuel, the relative responses of methane and acetylene are higher than on hydrogen-helium fuel. However, the opposite is the case for ethylene and toluene. Reduction of the sample flow rate decreases the relative responses to methane and acetylene but increases that of the ethylene. Such interesting features can be invoked in explaining exhaust analyses correlation differences.
Extent of Exhaust Ana/?,.sisVariations Gas chromatographic analyses of the hydrocarbon components in a multitude of automobile exhausts for a variety of test cycles have been performed by Jackson ( 4 ) . These illustrate the particular predominance of methane, ethylene. acetylene, and toluene. Their relative importance in four representative exhausts. chosen because of the diverse nature of their compositions is illustrated in Table 111. Together. the four components constitute from 32-9170 by volume of the total hydrocarbons emitted. Considering solely this part of the total hydrocarbon exhaust, it is possible. utilizing the measured relative molar responses of
Table IV. Calculated Variability of Analyses for Various Mixtures Model 400
Model 402 ~~~~
196OF
~-
~~
H,/N,
-~
Exhaust type mixtures
350 CID V8 engine
Wankel engine Prototype, cycles 3-10 Prototype, cycles 1-18 "
5;.
97.1 95.1 98.9 99.0
H:/He
~~
3
98.3 96.0 99.3 99.3
HJN~
_____ 5
Environmental Science & Technology
Normal
5
3
105.8 103.0 101.8 103.8
105.1 103.6 100.2 102.1
3
1 0 2 . 1 101.3 102.9 102.3 101.0 99.4 101.1 99.4
_ ~ _ _ _
H./He .
Scaled t o a m e a n value of 100 p p m p r o p a n e equivalent for each mixture.
828
____~
~~~~~
391°F H.lN.
~,/Iie
~
~~
5
5
3
106.8 106.9 108.9 108.9 98.5 100.7 102.2 100.8
98.0 95.9 101.5 101.2
Sample operating pressure, psi.
3
5
3
95.7 100.1 99.3 95.3 102.0 102.0 98.7 100.5 98.7 97.7 99.9 98.0
Hydrocarbons Measured Relative to Propane M o d e l 400 (contd.) Modified
n!/NIC
__.__
3
5
1.216 1.660 3.161 6.437 3.000
1.134 1.769 2.758 6.368 3.000
Model 108A"
n:/N?
n./ned
3
5
1.111 1.839 2.893 6.922 3.000
1.051 1.924 2.631 6.866 3.000
3
5
1.143 1.724 2.810 6.231 3.000
1.054 1.851 2.578 6.283 3.000
H?/He
5
1.039 1.930 2.518 6.777 3.000
3
0.995 1.989 2.488 6.829 3.000
.
3 ,
1c
S a m p l e o p e r a t i n g pressure, psi.
I
ips ike
i
338
CpEr,l
Table I1 to calculate the expected variability of analyses of mixtures of CH4, C2H4, C2H2, and C7H8 formulated in the proportions typifying such exhaust types. The results for such hypothetical analyses are listed in Table IV. For convenience they have been scaled in each case to a mean value of 100 ppm propane equivalent to better illustrate the range of variation. It is seen, for example, that for one of the mixtures representing a prototype engine exhaust, 91.4% of the total hydrocarbons are in the form of CH4, C2H4, CzH2, and C7Hs. Analysis of such a mixture in the same proportions would produce values on this set of instruments differing a t the most by s%,dependent on the FID and mode of operation. Since the four hydrocarbons represent just about the entire hydrocarbon emission in this particular example, this total variation would be expected for the actual exhaust analyses. For the other mixtures. calculated variation are larger, 12.17~for the 350 CID (in.3 displacement) V8 engine, 15.4% for the Wankel, and 7.0% for the second set of prototype engine data. The particular extent of the scatter is controlled by the relative contributions of each hydrocarbon and the indivdual variations of the response sensitivity. For example, the 15.4% figure noted for the Wankel engine reflects almost entirely the variation of the response to toluene. The 12.1% scatter for the 350 CID V8 mixture arises from almost equal contributions from C2H4, C2Hz. and C7H8, with a negligible contribution from CHI. whereas for the prototype mixtures CHI and C2H4, contributions are predominant. For these cases where the four hydrocarbons constitute only a fraction of the hydrocarbon emission, it is not possible to assess without additional data whether the variation for the total hydrocarbon exhaust analyses will change to quite the same extent. The variation of the relative sensitivities of the additional species may modify this to a greater or lesser extent. Actual exhaust analyses. however. confirm t h a t such possible modification appears slight. For example. diluted exhaust samples from a 350 CID V8 engine were collected in a Tedlar bag and analyzed on several occasions with all the instruments under their various possible operating modes. The data showed iden-
of Methane, Ethylene, Acetylene, and Toluene,( Model 400 (contd.) Modified
Model lO8A HI/He
H.lN.
H./N
n./He
CPS
-
H2 V ? PC
'13
>I8 P-
P
Figure 1 . Analysis of a diluted exhaust sample taken from a 350 CID V8 engine on a variety of analyzers under different operating modes
tical trends and one such set of values has been plotted in Figure 1. A line representing the magnitude of a 2% change on the same scale illustrates the extensive spread. The difference between minimum and maximum values is of a 14% magnitude. However, when scaled to the data of Table IV: the distribution can be superimposed almost exactly on that representing a 350 CID V8 type mixture of only the four basic hydrocarbons, agreement being to better than about 2%. The analyses (Figure 1) confirm t h a t substantially higher values are obtained in this instance with the heated FID. Also, values are seen to be larger in general when the instruments operate on the hydrogen-helium fuel. Variations for a particular instrument as a function of operating mode are much smaller and are typically & 53%. It appears safe to conclude that although the FID does a creditable job in integrating the carbon content of an exhaust mixture and is undoubtedly accurate to within 1070, large variations between commercial instruments are possible under certain circumstances. These will tend to be less with exhaust emissions rich in the light hydrocarbon components and more pronounced for example with rotary engines not fitted with control features to reduce the contribution from toluene. The results also illustrate that it is quite possible that the higher analysis values generally noted with heated FID's may not result altogether from reduced absorption/condensation of high-molecularweight-boiling point hydrocarbons but be a consequence of larger relative molar sensitivities. R e s p o n s e D i f f e r e n c e s B e t w e e n S a m p l e s of S a m e Concentration i n Nitrogen a n d in Air-Oxygen Interference
Flame ionization detectors invariably burn a diffusion flame of hydrogen fuel on a small diameter jet in a surrounding flow of air. The sample generally is added to the fuel flow just prior to combustion. It has long been realized (2) that the background gas of otherwise similar samples can affect the resulting flame response. The terms oxygen effect, synergism, or interference have been used interchangeably to describe this. The extent of the interference depends on various factors including the particular analyzer model and its operating parameters, the burner and electrode design, the fuel type. the organic species. and the oxygen content of the sample. Given the same concentration of propane in nitrogen or in air, the latter will produce a lower flame ionization. However, dependent upon their nature, some hydrocarbons can show an enhanced response. This effect is particularly important in cases where it is not possible to use calibration
__~~, ~~
5
3
5
3
99.8 96.4 101.4 99.4 96.7 95.3 102.4 101.4 102.9 100.6 101.6 100.0 103.3 99.6 101.7 99.2
5
95.6 93.5 100.0 99.2
3
5
3
94.8 97.8 98.3 94.0 100.1 100.8 97.9 99.3 98.3 96.8 98.1 97.4
Volume 8. Number 9, September 1974
829
Propane l p p r ~ l
Figure 2. Response differences between samples of propane in nitrogen and propane in air gases having a n identical gas medium to t h a t of the sample. Such is the case for real-time analyses of undiluted automotive exhausts in which a variable oxygen content exists. For these it is necessary to minimize this interference to a n acceptable level. The extent of the oxygen synergism has been measured for propane on the Beckman Model 402. and Models 400 and 108A with both hydrogen-nitrogen and hydrogen-helium mixed fuels a t various sample pressures and analyzer temperatures. Instrument operational settings, as mentioned before. always were those which would give a n optimum response to a n organic sample. Four cylinders of propane in nitrogen (50.5, 100, 214, and 410 ppm) and four propane in air (49.6, 96.3. 194. and 357 p p m ) formed the basis for the data. The extent of the effect is illustrated for various cases in Figure 2. and its magnitude, expressed as a percentage effect, is listed in Table V for a variety of conditions. Since the instrument response is linear in these concentration ranges and refers to a common zero. the effect expressed in this way is independent of hydrocarbon concentration. The interference is significant in most cases and is particularly pronounced when the instruments are operated on hydrogen-nitrogen fuel. With the exception of the heated FID operating with its analyzer train maintained at its maximum temperature of 400”F, the effect is still quite large even on hydrogen-helium fuel which is generally recommended for minimizing the interference. Sample pressure effects are rather unpredictable. It might have been reasonable to expect that since a 5-2 psi change in sample pressure decreases the sample flow rate and, consequently, sample air by about a factor of three. that a much lower oxygen synergism would result. Although reduced effects are noted in most cases, the decrease is by less than a half; in fact, occasionally, a n increase is observed. Typically, in these instruments, about 125 cm3 min-l of 4 0 7 ~hydrogen-60% helium fuel burns in a flow of about 400 cm3 m i n - l of air. Sample flow a t 5 psi is about 20 cm3 m i n - l . This means that 4 cm3 m i n - l of oxygen are being premixed (1.4 cm3 m i n - l a t 2 psi) with about 50 cm3 m i n - l of hydrogen; 25 cm3 min-l would be required for a complete stoichiometric premixture. A trace addition of as low as 1.4 cm3 m i n - l of oxygen, that is about 1% of the gas flow through the burner tip, can under certain conditions produce these large differences observed. Consequently, the results of Table V illustrate t h a t as long as some minimum trace of oxygen is present in the sample, almost the full extent of the interference is realized. This observation formed the basis for the development and characterization of a modification which was initially in830
Environmental Science 8, Technology
stalled on the Beckman Model 400. However, the concept should apply equally to all FID’s and provide a means of minimizing the sample oxygen interference effect. If a trace of air can be continually added to the fuel and sample mixture just prior to combustion, little difference should be observed between a similar sample of hydrocarbon in air or in nitrogen. Moreover, such a small addition a s suggested should not otherwise change to any marked extent the nature of the burner flame or modify the general performance of the instrument. The plumbing of most FID’s is readily adaptable for such a modification which involves tapping the high pressure part of the air line and bleeding the required trace of air through a flow restrictor into the fuel line. This is illustrated in Figures 3a and 3b. In this way the burner air, fuel, and sample flow rates remain unchanged. No safety hazards are introduced with this new arrangement since the porous metal plug flow restrictors in each line prevent mixing of fuel and air upstream in the highpressure sections. Initially, calibrated amounts of air were bled into the fuel-sample line and the magnitude of the oxygen effect was measured as a function of this for propane samples. Small additions produced marked reductions and a minimum interference was obtained with a flow rate of about 7 cm3 m i n - l . Larger flows rates increased the magnitude of the effect again. Similar behavior was noted a t any sample flow rate (2-5 psi setting). Fitted with a restrictor through which the flow was 6.9 cm3 m i n - l for a n air pressure of 15 psi, the oxygen effect was reduced to those values quoted a t the bottom of Table V. The hydrogen-helium data, which are of prime in-
U
L
(hl
Figure 3 . Schematic of burner flow configuration ( a ) in typical flame ionization detector ( b ) modified as recommended for minimizing the sample oxygen interference
Table V. Response Differences Between Samples of Same Propane Concentration in Nitrogen and in Air