Tunable diode laser svstems for measuring trace gases in tropospheric air A discussion of their use and the sampling and calibration procedures for NO, N 0 2 , and "03 H. I. Schiff'J D. R. Hastiel G . I. MackayZ T. Iguchi'J B. A. Ridley'J York University Downsview, Ontario M3J l P 3 Canada Unisearch Associates Inc. Concord, Ontario L4K l B 5 Canada
Photochemical smog and acid deposition are two environmental problems in which the nitrogen oxides, NO, N02, and HN03 play a central role. Measurements of these constituents in the troposphere are therefore essential for understanding the chemistry of these problems. NO, emitted mainly from automobile exhausts and thermal power plants, is the major anthropogenic source of the nitrogen oxides. It is oxidized in tropospheric air to NO2 by reactions with 03,H02, and organic peroxy radicals. NO2 is an essential ingredient in the formation of photooxidants. It is rapidly photolyzed by solar radiation at wavelengths less than 420 nm to produce atomic oxygen, which combines rapidly with oxygen to form ozone. NO2 is removed mainly by reaction with H O radicals to form "03. It also reacts with organic radicals to form strong oxidants and irritants such as peroxy-acetyl nitrate (PAN). HN03,the major sink for the nitrogen oxides is, in turn, removed mainly by wet or dry deposition, often in the form of aerosol nitrate. The oxides of nitrogen, therefore, not only contribute directly to the production of photooxidants, but also control the concentrations of HO radicals, which 352A
Environ. Sci. Technol., Vol. 17, No. 8,1983
are the major initiators of chain reactions involving hydrocarbons. Wet chemical and filter methods have been used for many years to measure these compounds in the atmosphere. The reliability of these methods has not been high, and long collection times are required. For these reasons they have been largely supplanted by optical methods for measurement of NO and NO2. Only indirect chemiluminescence methods have been reported for measuring HN03 with the sensitivity and the short response time required for tropospheric modeling ( 1 - 4 ) . Infrared absorption in the 2-15-pm region offers a number of advantages for atmospheric trace gas measurements; it is a passive technique, well adapted to in situ, real-time measurements. While virtually every minor or trace constituent absorbs in this spectral region the major gases, N2 and 02, do not. In fact, the absorption spectra for the trace gases are so rich that high resolution comparable to the line width, typically 2 X cm-' (60 MHz), is usually required to avoid mutual interferences. Such resolution can be obtained with a Fourier transform infrared (FTIR) spectrometer using a broadband light source. But because of its low sensitivity, path lengths of several kilometers are needed to detect constituents at the ppbv level. Recently, tunable diode lasers have been developed that provide a wavelength-tunable source of extremely narrow line width. Techniques have been developed for measuring very small absorbances with these diodes. Reid and co-workers (5-8) have described optical systems for detecting
tropospheric gases, but not the calibration methods or techniques for sampling real air into the system. This paper describes the use of tunable diode laser absorption spectrometers (TDLAS) for measuring nitrogen oxides in clean and polluted tropospheric air. We have used this spectrometer in a mobile system capable of in situ measurements of the oxides of nitrogen and other trace atmospheric gases. Principles of TDL spectroscopy Currently available tunable diode lasers (TDL) that operate in the midinfrared are lead salt semiconductors of general composition Pbl-,Sn,Te, PbS1-,Sex, Pbl-,SnxSe and Pbl-,Cd,S. A p-n junction is formed by diffusion through an appropriate mask. The crystal is cut or cleaved to produce a diode of approximately 400 Fm in length and 200 pm in width and height, with the crystal faces defining the cavity. Electrical contacts are evaporated onto the laser, which is then mounted on a rigid block. Application of an electric current (typically 0.5-2.0 A) causes lasing action by producing a population inversion between the almost filled valence band and the slightly populated conduction band in the active region. The frequency of the output laser radiation is given, to a first approximation, by the band gap of the semiconductor. Since this band gap depends on the metals that comprise the crystal and the composition parameter x, lasers can be manufactured to operate at a chosen frequency. Lasers are available for any frequency in the range of 500 to 3500 cm-' (2.8-20 pm);frequencies outside this range are available on special order.
0013-936X/83/0916-0352A$01.50/0
@ 1983 American Chemical Society
Tunability can be accomplished by subjecting the laser to a variable magnetic field ( 9 ) . a variable hydrostatic pressure (10). or a variable temperature. Since these devices usually must be operated at a stable temperature below 50 K, temperature variation is the most convenient method for tuning the diode. Small temperature changes can be made by varying the current through the diode. Early work was carried out with the diodes held in liquid helium Dewars. But this work was limited by problems with the supply, handling, and cost of liquid helium. More recently, reliable, closed-cycle refrigerators capable of achieving 12 K under fractional watt heat loads have become available. They have permitted lasers to be held at stable temperatures between 15 and
60 K, providing a wide frequency range for such lasers. In principle, output frequencies corresponding to a number of welldefined longitudinal modes separated in frequency by AU = c/2nl are possible, where c is the speed of light, I is the cavity length, and n is the refractive index of the laser material. The modes are not usually well defined and longitudinal, however, nor are the crystal faces exactly planar and parallel. Consequently, multimode operation, with variable frequency separation between modes, usually occurs. Since the intermodal separation is generally one wavenumber or greater, a laboratory monochromator is more than adequate to resolve the mode structure and is used, in some applications, to ensure passage of only a single mode through the system. Single
mode performance often can be achieved by operating the laser at a suitable temperature and drive current. Under these conditions the frequency generally can be varied by 0.5 cm-I for a current variation of 100 mA. The line width of a TDL is extremely narrow. If a Dewar system is used to cool the diode the laser halfwidth can be as small as 27 kHz ( / I ) .The use of a refrigerator increases the noise because of piston vibration, and linewidths are typically between 1.6 and 50 MHz (12). Since gas-phase Doppler half widths are on the order of 60 MHz, TDL systems operating in single mode can obtain absorption spectra limited by Doppler broadening. The narrow line width and fast response of these diodes permit their frequency to be altered and
x
Air moniloringsystem. The runable-diode-larer-basrd Tropocpherir Air Moniroring S,srem (TAMS) (IS usedjor rhe LOTAngele.5 measuremenr campaign. The oprical sysrem and gas control hardware are mounred on rhe bench wirh rhe Whire cell and dererror i n rhe foreground. The laser conrro1,jlow control, gas calibrarion, and sipnal processing and signal recording e1ecrronic.r are housed i n rhe racks below rhe benrh. Environ. Sci. Technol.. VoI. 17. NO. 8. 1983
353A
FIGURE 1
A schematic of the laboratory tunable diode laser system'
Gas inlet
modulated. The laser can therefore be scanned rapidly over the spectral range of interest. A TDLAS takes advantage of the narrow line width, tunability, and frequency modulation characteristics of these diodes to measure accurately small absorbances (10-3-10-5) due to a single rotational line in the vibrational band spectrum of a molecule. To measure this small absorbance, the absorption band must have a resolved rotational fine structure. Pressures less than 35 torr are used to substantially lower collisional broadening, but even at these pressures large molecules such as organic aromatics, PAN, NzO5,and CINO3 have so many rotational lines that the spectrum resembles a continuum. Most molecules with 5 atoms or less, however, as well as some larger molecules such as SFg and C2H4, have resolved fine structures. All heteronuclear diatomics and some polyatomics have regions where the spectrum is resolved, even at atmospheric pressure; NO, NOz, and HNO, fall into this category. Typical line strengths are in the range of 6X to 2 X cm-I molec-1 cm2, and therefore a path length of 20-50 m is sufficient to measure sub354A
Envlrm. Scl. Techml.. Vol. 17. No. 8. 1983
ppbv levels of these molecules with a TDLAS capable of detecting absorbThe long path length ances of can be provided by directing the beam through the atmosphere and returning it to the detector by a retroreflector. For molecules that do not have resolved absorption spectra at atmospheric pressure, or to avoid interferences by other species, reduced pressure is required, and the long path can be achieved with a multiplereflection absorption cell.
Absorption reletionsbips The fraction of the power transmitted through an absorbing medium of length I cm can be related to the concentration of absorbing molecules, N molecules/cm3, by the Beer Lambert law I = exp(-u(v).N.I)
Io
(1)
where I and IOare the transmitted and incident powers and u(u) the frequency-dependent absorption coefficient in cm2. For a single isolated absorption line the absorption coefficient can be expressed as the product of an absorption
coefficient at line center uo and a line shape function L(u) U ( Y ) = UOL(V) (2) At line center and for an optically thin system the Beer Lambert law reduces to
Thus the power absorbed is always proportional to the concentration of the absorbing molecule. The functions u(u), L(P) have three different forms, depending on the pressure. It can be shown, however, that for each of these cases the absorption signal remains linear with concentration for optically thin conditions.
TDL absorption spectrometers The first TDLAS to be used as a trace gas detector was that of Hinkley et al. (13-16). Initially, it was used to measure ambient carbon monoxide in cities (13). A total path length of 610 m was achieved using a retroreflector. Since the P(1) to P(8) absorption lines of CO at around 2100 cm-I are all resolved at atmospheric pressure it was possible to use the frequency modulation (FM) technique,
with detection taking place at the fundamental frequency. They reported C O measurements of 400-1000 ppbv with an estimated noise equivalent of 5 ppbv. They later used these results to estimate sensitivities for C2H4, ”3, O,, Freon I I , Freon 12, and vinyl chloride based on their ability to measure absorbances of over a I-km path (14). Several improvements to this instrument have been proposed including line locking, 2f detection ( 1 7 ) , and frequency stabilization using a tunable etalon (18). Sachse et al. ( 1 9 , 20) have used a TDLAS of this type on an aircraft using 10 m on the outside of the fuselage as the absorption path. They measured carbon monoxide with an absorbance limit of 8 X 10-4-about a fourfold-lower sensitivity than that achieved in the laboratory. To overcome the problems of atmospheric turbulence and spectroscopic interferences, Reid et al. modified the TDLAS by using a multiplereflection White cell in place of the atmospheric long path and ran the cell at reduced pressure ( 5 ) . They also detected the absorption at twice the modulation frequency (2f), which had the advantage over detection at f of a ‘‘zero’’ baseline and a maximum at line center. By extrapolating observed absorbances in a 10-cm cell to those expected in a 300-111 White cell they estimated their detection limit for SO2 to be 3 ppbv in the v I band at 1140 cm-I. They estimated the absorbance detection limit to be 10-5, corresponding to limits of 0.5 ppbv for 03, 0.3 ppbv for SO2 at 1370 cm-I, 2 ppbv for NzO, 0.01 ppbv for CO, 0.05 ppbv for NH3,0.03 ppbv for CH4.0.02 ppbv for N02, and 0.03 ppbv for NO. Later work by theseauthors showed that this limit could be improved by locking the laser frequency ( 6 ) .They also showed the ability to measure SO2 in a White cell but were unsuccessful in their attempts to measure NO2 because of adsorption of the calibration gas on the cell walls (8). Based on the results of Reid and cc-workers, a number of TDLAS trace gas detectors have been built using their design and techniques. Connell et al. (21) have measured N2O in air with an absorbance detection limit of 3 X Mankin et al. ( 2 2 ) have measured C O from an aircraft, and Pokrowski et al. ( 2 3 ) have measured HCI from on board a ship at sea. A laboratory TDLAS The laboratory TDLAS used in this work is based on the design of Reid et al. and is shown schematically in Fig-
alignment of the optical path, it is convenient to pass a helium-neon laser beam through the center of the infrared beam. The red beam can be introduced into the optical train by inserting the folding mirror, PM5. The White cell can be replaced by a ‘I4-m monochromator for approximate frequency determination or for examining the mode structure of the laser output. The spectrometer can be operated in either theamplitudemodulation (AM) or the FM mode. In the AM mode the beam is chopped by the mechanical chopper and detected at this frequency by the lock-in amplifier to give a direct measure of the power incident on the detector. The absorption spectra are obtained as the laser frequency is varied. Figure 2 shows an example of the absorption spectrum of HNO, obtained in this way. The concentration of the absorbing species can be calculated directly from the spectrum, provided that the absorption coefficient in Equation 1 is known. This method of determining concentration is limited to the 1% absorbance level by the difficulty in measuring small differences between I and Io. This limitation can be circumvented by operating the system in the FM mode. The multiple-function generator
ure 1. It routinely achieved absorbance or better, detection limits of 2 X which-for the oxides of nitrogencorresuonds to mixine. ratios below the I-ppb; level. The diode laser is held in a source assemblv maintained to fO.OO1 K at the operating temperature, which is in the 10-60-K range for the diodes used. Temperature control is provided by the combination of a closed-cycle, helium cryocooler, a heater, and a servo temperature control system. The current through the laser is supplied by a highly stabilized dc current control system. These items were supplied by Spectra Physics. The current can be changed or modulated by inputs received from a multiple-function generator. The laser radiation, which is generally emitted as a multilobe pattern within an f / l cone, is collected and focused by an f / l lens, LI. The plane fold mirrors, PMI and PM2, permit maneuvering the beam for accurate entry into the White cell. The exiting beam is directed by a plane mirror, PM3, through a sample cell and another f / l lens, L2, onto a liquid-nitrogen-cooled, mercury-cadmium-telluride detector, the output of which is fed into an oscilloscope for visual representation, then to a lock-in amplifier for processing. To aid in
-
FIGURE 2
The HNOo absorption spectrum taken in the region of 1722 cm-’ using the 11-cm cell’
spectrum
0.2
1
,.,..~ ..,,. ::;::< ji
1721.6605
1721.7926
1722.0000
1
8
Frequency (cm-‘) ‘Total Pressure = 1.0 tor. “0% p“aI
pmrsure
~
0.01 fori, and laser power = 1 (IW.
E n v i r ~Sci. . Technol., Vol. 17, No. 8. 1983
355A
applies a I-8-kHz sine wave to the laser current to give a current amplitude modulation that produces a frequency-modulated laser output. Measurement of the detector output at this frequency gives a signal related to thevariation in power with frequency rather than the power incident on the detector itself. The signal obtained by scanning the modulated radiation over a single absorption line has a shape that is similar in appearance to the derivative of the absorption feature. This signal is a true derivative only if the modulation amplitude is much smaller than the absorption line width, but under these conditions the signal is relatively small. The signals are substantially larger for larger modulation amplitudes and it can be shown theoretically and experimentally that the maximum signal occurs when the peak-to-peak modulation amplitude is 2.2 times the half width (24). This determines the magnitude of the current modulation to be 1%) of NO in the 11-cm cell. PreBIYre = 10 torr.
0
1 Time (min)
2
‘Measured 8n 21 defecttan for a NOalr rnlxlng ratio Of 11 ppbv in the While cell Pressure = 10 IOR. MDC = 0.5 ppbv
Environ. Sci. Technol.. Val. 17. NO.8. 1983
357A
FiGURE 5
The observed variation in the measured NO 2f signal vs. the expected mixing ratio.
1.5 6-
z
> 1.0
--a m
c
-
P II)
0.5
N
0 0
25
Mixing ratio (ppbv)
the residence time (7-27 s) in the White cell. Apparently, losses due to adsorption or reaction of NO at the walls are immeasurably small under these conditions. NO*. Since NO2 is more polar than NO, problems were anticipated in maintaining sampling integrity at low concentrations. A cylinder containing a 10-ppmv mixture of NO2 in oxygen was used as an NO2 source for many of the experiments. But concern over the long-term stability of this source and possible surface effects on the regulators, valves, and flow meters used to take the mixture from the cylinder prompted cross calibration with an independent source. Passing known flows of nitrogen over a permeation tube was selected. Since the permeation rate is independent of the pressure outside the tube, mixtures can be made at low pressure and fed into the system without the gas contacting anything but transfer tubing. Permeation tubes are calibrated by the manufacturer from the weight loss as a function of time. Four tubes were used, all of which had permeation rates in the 50-150ng/min range. For N2 flows of I O SCCM, mixing ratios in the 2.5-7.5ppmv range of NO2 were produced. The mixtures could be further diluted to the tens of a ppbv range. Flow lines of Teflon or glass were found to be unreactive to NO2. A titration system was used to check the NO2 sources. A known flow of NO mixture was oxidized to NO2 by ozone: 358A
Environ.
Sci. TeChnOl.. Vol. 17. No. 8. 1983
+
-
+
+
-
+
NO 0 3 NO2 0 2 (R1) Reaction RI provides quantitative amounts of NO2 when carried out in an excess of N O or near its end point. Under these conditions the rate of the competing reaction: NO2 O3 NO3 O2 (R2) is negligible since k2 = 3.2 X IO-” cm3 s-* (26) and the O3concentration is small. The decrease in NO mixing ratio is then equal to the NO2 mixing ratio produced. A series of experiments was performed in which this titration was used to calibrate the permeation tubes and to determine the MDC of N02. In some of these experiments the NO mixture was titrated to the end point and the signal due to the NO2 generated was measured. The output of the permeation tube was then introduced to the flow and the measured signal compared with the signal from the titration source. For experiments involving an excess of NO, the residual NO was measured with a chemiluminescent instrument. The permeation rates determined by these experiments were compared with those given by the manufacturer and with values obtained from independent weighings by the Atmospheric Environment Services (Canada). These comparisons are shown in Table I. In all cases, the manufacturer’s permeation rates were smaller than those obtained by the other methods, which suggests changes since shipment. The permeation rates appear to remain stable within f5% provided the tubes
are not subjected to large changes in pressure or temperature. However, care is needed in transportation since some changes were observed when we shipped tubes by air. One tube was found to have a white region in the plug indicating some damage, which perhaps explains its high permeation rate. Permeation tubes, therefore, provide satisfactory NO2 calibration sources provided adequate care is taken in their handling. As anticipated, problems were encountered in the handling of trace concentrations of NO2 from the cylinder source. The measured concentration of 10 ppmv was found to agree within 5% of the manufacturer’s figure only after the gas was run through the system for several minutes at high flows. The major source of the adsorption indicated by this observation appears to be the regulator, although it contained only stainless steel and Teflon. Low flow rates (
