68
Anal. Chem.
of the film optical properties. The measured values of the copper optical constants were used rather than literature values because the use of measured values will tend to cancel the effect of any systematic errors in the film-covered copper spectra. It is apparent that a whole range of possible optical constant spectra can be generated by estimating the film thickness. One can see from the optical calculations of Figure 6 that the derived k(n) band position shifts and the band shape changes as the estimated thickness varies. Given a single independent measurement of one band position, it is possible to select the correct thickness from a plot of the band positions and estimated thicknesses of Figure 7. Once the thickness has been determined then the optical constants can be calculated. This technique relies on the fact that peak positions as measured by external reflection are shifted to frequencies higher than their true positions. This effect is strictly an optical artifact and its magnitude depends on the strength of the absorption band. For transmission measurements peak position artifacts are minimal and so the peak position measured from a transmission spectrum is very close to the true value. An implicit assumption is that the film on the substrate will have the same peak position as the independently measured value. The optical constants derived by ellipsometry shown in Figure 8 compare quite well with those derived from transmission with respect to peak position, shape, and relative intensity, but the absolute intensity does not compare well for the k(v) spectra and there is a constant offset in the n(v) spectra. The thickness determined by surface profilometry was about 27% thinner than that determined by infrared ellipsometry. With improved optical components such as better polarizers and goniometers, the accuracy of the mea-
surement will be imporved. Such refinement study is currently under way in our laboratory. Registry No. PVAc (homopolymer),9003-20-7. LITERATURE CITED Ishida, H., Case Western Reserve University, 1980, unpublished work. Aspnes, D. E. Surf. Sci. 1980, 701, 84. Stobie, R. W.; Rao, 8.; Dignam, M. J. Appl. Opt. 1975, 74, 999. Dignam, M. J. Polym. Prepr. (Am. Chem. SOC.Div. Polym. Chem.) 1884, 2 5 , 149. Allen, T. H.; Sunderland, R. J. Thin Solid Films 1877, 4 5 , 169. Adams, J. R.; Zeidler, J. R.; Bashara, N. M. Opt. Commun. 1975, 75, 115. Schaefer, R. R. J. Phys. Colloq. 1983. C70, 87. Stobie, R. W.; Rao, B.; Dlgnam, M. J. J. Opt. SOC.Am. 1975, 6 5 , 25. Roseler, A. Infrared Phys. 1981, 21, 349. Roseier, A.; Molgedey, W. Infrared Phys. 1984, 2 4 , 1. Dlgnam, M. J.; Baker, M. D. Appl. Spectrosc. 1981, 3 5 , 186. Azzam, R. M. A.; Bashara, N. M. "Ellipsometry and Polarized Light"; North-Holland: New York, 1977. Muller, R. H. I n "Advances In Electrochemistry and Electrochemical Engineering"; Wlley: New York, 1973; Vol. 9, p 167. Hauge, P. S. Surf. Sci. 1980, 96, 108. Born, M.; Wolf, E. "Principles of Optics", 6th ed.; Pergamon: Oxford, 1980. Dlgnam, M. J.; Rao, B.; Moskovits, M.; Stobie, R. W. Can. J. Chem. 1971, 49, 1115. Dignam, M. J.; Fedyk, J. D. Appl. Spectrosc. Rev. 1878, 74, 249. Fedyk, J. D.; Mahaffy, P.; Dlgnam, M. J. Surf. Sci. 1978, 89, 404. Dlgnam, M. J. "Vibrations at Surfaces" [Proceedings of the International Conference on Vlbratlons at Surfaces]; 2nd ed.; Plenum: New York, 1982; pp 265-88. Molectron Corp. Infrared Grid Polarizers"; Technical Bulletin, 1981. Heavens, 0. S. "Optical Propertles of Thin Solid Films"; Dover: New York, 1965; Chapter 4. Brown, K. M. Dennis, J. E. Numer. Math. 1972, 78, 289. Cameron, D. G.; Escolar, D.;Goplen, T. G.; Jones, R. N. "Computer Programs for Infrared Spectrophotometry"; National Research Council Canada: Ottawa, 1977; Bulletin 17, p 67. Allara, D. L.; Baca, A.; Pryde, C. A. Macromolecules 1878, 7 7 , 1215.
RECEIVED for review May 15,1985. Accepted August 16, 1985.
Comparison of Fourier Transform Infrared Spectrometry and 2,4=Dinitrophenylhydrazine Impinger Techniques for the Measurement of Formaldehyde in Vehicle Exhaust Larry P. Haack,* Diane L. Lacourse, and Thomas J. Korniski Ford Motor Company, Scientific Research Laboratory, Room S-3061, Dearborn, Michigan 48121- 2053
Experiments were conducted to valldate a Fourier transform Infrared (FT-IR) sampling and analysis system for measurement of trace gases In vehicle exhaust utlllzlng gasoline-, gasohol-, diesel-, and methanol-fueled vehicles as the emisslon source and formaldehyde (HCHO) as the test molecule. The 2,4-dlnltrophenylhydrazlne lmplnger method was chosen as the reference method. Dlluted exhaust was drawn continuously through the FT-IR cell and measured every 3 s. The FT-IR signals were averaged over a complete drlvlng-test cycle and compared to the concentrailon determined from concurrent lmplnger sampling. By lmplnger measurements It was shown that HCHO losses between the tallplpe and the FT-IR cell were on the order of only 5 % , Independent of vehicle type or HCHO concentration (0.02-8.5 ppm). Comparisons between FT-IR and lmplnger measurements on 43 tests of methanol-fueled vehicles under transient condRlons (diluted-exhaust HCHO 0.28-8.5 ppm) showed FT-IR/lmplnger = 1.055 f 0.095.
Formaldehyde has been identified in the combustion ex0003-2700/86/0358-0068$0 1.50/0
haust of various alternate-fueled vehicles (1-3). Formaldehyde is a reactive molecule that will readily oxidize, hydrate, or self-polymerize. It has been shown to combine with ambient species to contribute to photochemical smog (4). It was therefore chosen as a challenging model compound to aid in the evaluation of a Fourier transform infrared (FT-IR) sampling and analysis design for multicomponent exhaust measurements. Past work has documented the use of long-path IR and FT-IR for the measurement of ambient formaldehyde (5-9). However, the published literature is scarce concerning the use of FT-IR spectrometry for the measurement of components in combustion exhaust (10-1.2). There is a continuing effort to advance this technique, since it can be used for multicomponent exhaust analysis and is adaptable to continuous on-line sampling so that time-resolved transient information can be obtained. In the present vehicle tests the FT-IR spectrometer measured concentrations at a 3-s time interval. The individual 3-s concentration values were averaged over an entire test cycle and compared to the integrated value measured by using impingers. The generally accepted 2,4-dinitrophenylhydrazine (DNPH) impinger method (1,13)was used as the standard 0 1985 Amerlcan Chemical Society
ANALYTICAL CHEMISTRY, VOL. 58, NO. 1, JANUARY 1986 INTINE D I L U E N T A I R
69
8.5pprn HCHO in Diluted Exhaust
1 VENT
I * '
&HEATER
W
u z a
25.5ppm HCHO in Nitrogen
II
m
a
0 v)
m
a
.--
-I
2840
2790
2815
2765
I
2740
WAVE NUMBER (ern-')
Flgure 3. FT-IR formaldehyde (HCHO) spectra in nitrogen and in diluted methanol-fueledvehicle exhaust.
II
I
% t T ON-OFF
REG VALVE
FILTER
1-0-
/DILUTION
ROTARY VANE PUMP EXHAUST
I..
1
46 7 m TEFLON SAMPLE LINE
I1
\fhHr
CHARCOAL TRAP
ME PUMP
WETTEST METER
DILUTE EXHAUST
Flgure 4. Impinger sampling system.
An FT-IR spectrometer differs from conventional grating or prism
instruments in that wavelength determination is accomplished by modulating each wavelength at its own unique audio-range frequency via a scanning Michelson interferometer. The digital interferogram was frequency analyzed (Fourier transformed) to produce the IR spectrum. The sample cell was a custom goldcoated Wilks 20-m variable-path cell, used in 14th order with an effective path length of 21.75 m. A HgCdTe detector from Infrared Associates, Inc., exhibiting negligible nonlinearity, was used. The FI'-IR data were collected by acquiring an interferogram over a 3-s time period and writing the intereferogram onto the magnetic tape. The spectral resolution was 0.25 cm-l corresponding to a 2-cm stroke of the interferometer moving mirror. The interferograms were processed with a VAX 780 computer. The transmission spectra were generated by applying the Fourier transformation to the interferograms. The transmission spectra were corrected for the contribution of room-temperature stray radiation by subtracting a previously determined background. An absorbance spectrum was generated by calculating the logarithm of the ratio of the corrected spectrum of the tunnel dilution air to the corrected sample spectrum. The formaldehyde IR signal was measured by integrating the absorbance of selected vibration-rotation lines in the C-H stretch region from 2739 to 2827 cm-l. Figure 3 displays the FT-IR formaldehyde spectra of a standard and a methanol-fueled vehicle exhaust sample. The exhaust spectrum was generated from one interferogram. By use of this standard the formaldehyde diluted exhaust concentration was determined to be 8.5 ppm. D N P H / A c e t o n i t r i l e I m p i n g e r Technique. The impinger technique used for formaldehyde analysis was a modification of a procedure developed by Lipari and Swarin (1). This technique involves the impinger trapping of formaldehyde by forming the DNPH derivative in acidified acetonitrile solution followed by separation and quantification of the derivative by high-pressure liquid chromatography (HPLC). Impinger Sampling System. The sampling system used for the impingers is shown in Figure 4. Sample lines made of Teflon (0.64 cm o.d., 4 mm id.), with stainless-steel Swagelok fittings, were used to connect all components. The diluted exhaust was drawn through ice-cooled impingers at approximately 1.5 L/min using a Metal Bellows Model MB-158 pump. A microregulating valve before the pump was used to meter the sample flow. Total sample flow was measured by using wet test meters. Charcoal (coconut, activated, 8-12 mesh) traps were used to inhibit ace-
70
ANALYTICAL CHEMISTRY, VOL. 58, NO. 1, JANUARY 1986
tonitrile vapors from reaching the wet test meters. Reagents. The DNPH and DNPH-formaldehyde derivatives were purchased from Chemical Samples Co., Columbus, OH. The DNPH was further purified by recrystallization in HPLC grade acetonitrile. The formaldehyde-DNPH derivative was recrystallized from HPLC grade ethanol. A stock solution of the formaldehyde-DNPH derivative was prepared at a concentration of 423 pg/mL. HPLC standards were prepared by dilution of the stock solution to the proper concentration range as determined by the formaldehyde levels encountered during each vehicle run. A stock solution of DNPH/acetonitrile was prepared at a concentration of 1.25 mg/mL. Impinger solutions were prepared by adding 10 mL of the stock solution to 30 mL of acetonitrile. The solutions were acidified with 10 drops of 5.9 N perchloric acid just prior to sample collection. The samples were neutralized with 10 drops of 5.9 N ammonium hydroxide immediately after sample collection. After the solutions were warmed to room temperature, they were diluted with acetonitrile to 50 mL in volumetric flasks and then refrigerated until analysis by HPLC. HPLC Analysis. Analysis of the formaldehyde-DNPH derivative was performed with HPLC instrumentation consisting of a Du Pont Zorbax ODS 4.6 mm X 25 cm column, two Waters Model M-6000A chromatography pumps, and a Varian Model UV-50 variable-wavelength detector operated at 365 nm. Samples were injected by use of a Waters WISP 710B automatic sampler. The injection volume for all standards and samples was 15 pL. Separations were performed isocraticallyusing a 40/60 blend of HPLC grade water/acetonitrile. FT-IR/Impinger Calibration. A modified permeation tube calibration system (Tracor Instruments, Model 420/PERM) was used to generate a continuous standard stream of formaldehyde in nitrogen. A stream of nitrogen was passed over paraformaldehyde in a Pyrex capillary diffusion tube heated to a constant elevated temperature. A mass flow meter was used to deliver a constant flow of nitrogen. Closely matched periodic FT-IR spectra obtained from the gas stream ensured the generation of a consistent dynamically blended standard. The mass of formaldehyde delivered per unit time was determined by successive weighings of the diffusion tube. Preparation of Formaldehyde/Water Aerosol Standard. A stock solution of 10% formaldehyde in water was prepared by dissolving paraformaldehyde (95% formaldehyde, Matheson Coleman and Bell, Norwood, OH) in boiling purified (glass distilled, deionized) water. This solution was standardized gravimetrically by the dimedone method (19). Formaldehyde Aerosol Injections. Recovery studies were performed by aspirating, at the upstream end of the dilution tunnel (site B, Figure l),diluted mixtures of the stock formaldehyde/water solution to conditioned room air. Impinger samples were drawn from the downstream end of the dilution tunnel (site C) to determine the amount of formaldehyde transferred. Tunnel flow and temperature conditions were chosen to ensure complete vaporization of the aerosol. The tunnel flow rate used was approximately 17 m3/min. The diluent air temperature ranged from 35 to 40 "C. The standard samples were delivered through an aerosol aspirator nozzle. Argon gas was used to propel the aerosol into the tunnel at a liquid flow rate of approximately 1 cm3/min. An analytical top-pan balance was used to gravimetrically determine the total amount of liquid delivered during each test. These studies were performed without a vehicle. Standard additions of formaldehyde to vehicle exhaust were performed by using the same injection apparatus. The aerosol was injected directly into the hot (200-350 O C ) tailpipe exhaust (site A, Figure 1). A background measurement was first obtained from the vehicle. The standard injection then was made with the vehicle driven under the same conditions. Recovery was determined by comparing the difference between the impinger measurements with and without injection against the actual amount injected. One test was made with a methanol-fueled vehicle operated continuously at a fast idle. Another was made with a diesel-fueledvehicle driven under the HWFET test cycle. Repetitive HWFET cycles were separated by a 10-min soak. As with the tunnel injections,formaldehyde recovery was determined by using downstream impinger samples (site C, Figure 1).
Table I. Formaldehyde Recovery through the Dilution Tunnel Using the Aerosol Injector tunnel formaldehyde concn, ppm 0.143 0.467 1.320
mass injected, mg
63.1 103
291
mass collected, mg 63.4 106 266
%
recovered 100.5
102.9 91.4 98.3
a
Mean (standard deviation).
FT-IR Sample Line Formaldehyde Transfer Efficiency Studies. Gasoline-,gasohol-,diesel-, and methanol-fueledvehicles were used as the formaldehyde emission source for these experiments. Separate impingers sampled from both ends of the FT-IR sample line (at sites C, impinger I, and D, impinger 11, Figure 1). A comparison of the formaldehyde collected at each end determined the transfer loss through both the filter and sample line. RESULTS AND DISCUSSION FT-IR/Impinger Calibration, A FT-IR/impinger formaldehyde calibration was performed by using the modified permeation tube calibration system (PTCS). Two sets of impingers sampling in tandem drew the entire PTCS-generated standard gas stream over a period of 10.0 min. The PTCS delivered 28.6 pg of formaldehyde/min (286 pg/10 min). HPLC analysis of the impinger samples yielded a combined collection of 285 pg of formaldehyde. This experiment showed that the standard generated by means of the modified PTCS correlated precisely with the DNPH/impinger analysis technique. The FT-IR standard spectrum for formaldehyde, against which samples were ratioed, was generated dynamically from the same PTCS-produced standard gas stream. Thus, the FT-IR and DNPH/impinger techniques were matched closely in calibration. Determination of Dilution Tunnel Formaldehyde T r a n s f e r Efficiency. A major effort of this study was to determine the integrity of the FT-IR sampling system for transfer of formaldehyde. An integral part of the sampling system is the 7.6-m dilution tunnel. (Refer to Figure 1 for the following discussion.) Mass-balance experiments were conducted to determine the transfer loss of formaldehyde through the dilution tunnel (from site B to C) and from the vehicle tailpipe to the downstream end of the dilution tunnel (from site A to C). Table I shows recoveries for single injections (at site B) at three different concentration levels. For these injections, the average recovery was 98.3 f 6.1%. A higher percentage mass transfer loss, due to adsorption or reaction with the inner tunnel surface, would be expected at low concentration levels. These results show no measurable transfer loss of formaldehyde even at levels as low as 0.14 ppm. Injections (at site A) into the hot tailpipe exhaust were performed to ascertain whether exhaust components interfere with the impinger measurement of formaldehyde. Methanoland diesel-fueled vehicles were chosen as representative emission sources of formaldehyde in a simple and a complex matrix, respectively. Table I1 includes the results of three standard additions to vehicle exhaust. Formaldehyde recoveries (at site C) for the methanol vehicle additions were 95.0 and 94.8% for standard injections of 1.00 and 2.72 ppm, respectively. With the methanol vehicle a high recovery of formaldehyde was somewhat predictable, since the low exhaust emissions of various species at idle condition should not interfere with the measurement of a relatively high addition of formaldehyde. Recovery of formaldehyde from the more complex exhaust mixture of the diesel-fueled vehicle would be more challenging. The diesel vehicle was driven by using
ANALYTICAL CHEMISTRY, VOL. 58, NO. 1, JANUARY 1986
71
~
Table 11. Formaldehyde Recovery through the Dilution Tunnel from Vehicle Standard Additions
mass injected,
mass collected,
%
mg
mg
recovered
+ i (239)
+ i (618)
18 227 586
95.0 94.8
HWFET (h) + h (94)
54 90
95.7
vehicle methanol
idle (i) 221 600
diesel
40
Table IV. Formaldehyde Transfer Efficiency of FT-IR Sample Line Using Diesel-Fueled Vehicles
impinger I concn, ppm
impinger I1 concn, ppm
0.173 0.197 0.216 0.225 0.235 0.254 0.260 0.435 0.449 0.513 0.610 0.631 0.635 0.650 0.994
0.206 0.190 0.207 0.212 0.199 0.224 0.254 0.347 0.388 0.541 0.501 0.542 0.754 0.616 0.864
95.2 (0.5)"
" Mean (standard deviation). Table 111. Formaldehyde Transfer Efficiency of FT-IR Sample Line Using Gasoline- and Gasohol-Fueled Vehicles
transfer vehicle gasoline
gasohol
impinger I concn, ppm
impinger I1 concn, ppm
0.036 0.051 0.054 0.084 0.091 0.098
0.029 0.050 0.039 0.060 0.072 0.101
0.806 0.980 0.722 0.714 0.791 1.031
0.019 0.020 0.025 0.039 0.073 0.085
0.025 0.018 0.033 0.038 0.075 0.090
1.316 0.900 1.320 0.974 1.027 1.059
efficiency
transfer efficiency (II/I) 1.191 0.964 0.958 0.942 0.847 0.882 0.977 0.798 0.864 1.055 0.821 0.859 1.187 0.948 0.870
(II/I)
0.944 (0.120)"
" Mean (standard deviation). Table V. Formaldehyde Transfer Efficiency of FT-IR Sample Line Using Methanol-Fueled Vehicles
transfer impinger I concn, ppm
impinger I1 concn, ppm
methanol with catalyst
0.57 1.23 1.97 2.36 2.39 3.02
0.53 1.20 1.88 2.26 2.46 2.81
0.930 0.976 0.954 0.958 1.029 0.930
methanol without catalyst
2.22 3.39 3.52 3.88 6.64 7.58 8.48
1.80 3.13 3.28 3.41 7.29 7.48 8.44
0.811 0.923 0.932 0.879 1.098 0.987 0.995
vehicle
0.970 (0.202)'
" Mean (standard deviation). the HWFET cycle, which, due to cyclic engine operation, would generate a higher emission level of formaldehyde in the presence of a greater amount of potentially interfering species. The level of formaldehyde injected was chosen to be equivalent to the expected level for a HWFET cycle. Even under these conditions, there was a high recovery (95.7%)of formaldehyde. In these three tests, recovery from the two vehicles averaged 95.2 f 0.5%. Determination of FT-IR Sample Line Formaldehyde Transfer Efficiency. The second integral part of the FT-IR sampling system is the 45.7-m unheated sample line, made of Teflon, with filter. Experiments were codducted to determine formaldehyde transfer loss through the line and filter. The different alternate-fueled vehicles were used to produce formaldehyde concentrations ranging from 0.02 to 8.48 ppm. The gasoline- and gasohol-fueld vehicles yielded the lowest diluted exhaust formaldehyde concentrations (0.020-0.100 ppm); the diesel-fueled vehicle yielded intermediate levels (0.200-0.994 ppm); and the methanol-fueled vehicles-both with and without catalysts-yielded the highest levels (0.57-8.48 ppm). These concentration levels are averaged numbers in diluted exhaust for a complete test cycle. The transient test cycles of the 78 FTP and HWFET procedures caused maxima and minima that would deviate greatly from the average. By using the test vehicles as the formaldehyde source, it was possible to determine transfer efficiency as a function of both concentration and vehicle type. Transfer problems were anticipated with the diesel due to the high emission of particulate matter. The sample line filter therefore was changed daily for each diesel test. Visual inspection of the filter after a diesel test showed a significant collection of particulates, which may impede the transfer of formaldehyde. For this reason, the filter holder was heated to 55 OC during diesel tests.
efficiency
(II/I)
0.954 (0.070)'
'Mean (standard deviation). Tables 111,IV, and V show the results of transfer efficiency studies using the alternate-fueled vehicles. The transfer efficiency was 95.5 f 13.5%for all vehicle types. The standard deviations indicate that, within experimental error, essentially all the formaldehyde was transferred through the filter and sample line indepenent of concentration and vehicle type. Comparison of FT-IR vs. Impinger Formaldehyde Measurements. Results from the PTCS-generated formaldehyde standard comparison study indicated that the FT-IR and impingers should agree when measuring formaldehyde from the same sample stream. The transfer efficiency experiments showed that, for the formaldehyde concentrations encountered in these analyses, no loss of formaldehyde occurred in the FT-IR sampling system. For these reasons, direct comparison measurements could be made between the FT-IR and impinger I, which sampled directly from the dilution tunnel (site C, Figure 1). In its present operational state, including continuous sampling and 3-s time resolution, the detection limit of our FT-IR for formaldehyde is -0.1 ppm ( S I N (peak/peak) = 1). The FT-IR did not have sufficient sensitivity to detect the low levels (
