a new Photofragmentation-laser induced fluorescence: method for detecting atmospheric trace gases M. 0. Rodgers, K. Asai, and D. D.

Transcription

1 Photofragmentation-laser induced fluorescence: method for detecting atmospheric trace gases a new M. 0. Rodgers, K. Asai, and D. D. Davis A new method for the in situ detection of nonfluorescing molecular species is proposed: photofragmentation-laser induced fluorescence (PF-LIF). In this approach, the species to be detected is first laser photolyzed at a wavelength X 1, producing one or more vibrationally excited photofragments. Before vibrational relaxation occurs, one of these photofragments is pumped into a bonding excited state by a second laser pulse centered at wavelength X 2. Fluorescence is sampled at a wavelength X 3, where X 3 < \2 and X. This pumping configuration thus permits massive discrimination against Rayleigh and Raman scattering as well as white noise fluorescence from the laser wavelengths Xi and X 2. The technique should be both highly sensitive and selective for numerous atmospheric trace gases. Specific sampling schemes for detecting NO 2, NO 3, and HNO 2 are proposed. Various noise sources and chemical interferences are discussed. Specific calculations that estimate the sensitivity of the PF-LIF system for detecting NO 2, NO 3, and HNO 2 are given. 1. Laser Induced Fluorescence (LIF) Technique Laser induced fluorescence, as a highly sensitive and selective technique for the detection of gas phase molecules, dates back to the early work of Sukurai and Broida.1 Since that effort, numerous spectroscopic applications of the LIF technique, particularly using tunable dye lasers, have been reported in the literature. 2 The basic approach of the LIF technique is illustrated in Fig. 1 for the case of the OH radical. In this system, it is seen that the OH radical is initially excited into the v' = 1 manifold of the A22+ electronic state by absorption of laser radiation at nm-the Qj(1) transition. Due, however, to collisions with N 2, 02, and H 2 0, significant rotational relaxation and electronic quenching occur, which result in both attenuated fluorescence signals and fluorescence emissions from rotational states both in the v' = 1 and v' = 0 manifolds of the A 2 1+ state. Even so, the most intense fluorescence radiation is centered at nm. 3 Thus, optical pumping takes place at nm, and fluorescence sampling occurs at nm. 3 The latter approach The authors are with Georgia Institute of Technology, School of Geophysical Sciences, Atlanta, Georgia Received 26 December /80/ $00.50/ Optical Society of America. permits discrimination against Rayleigh and Raman scattered laser radiation without a significant attenuation in the signal photon flux. It does not, however, prevent photon noise resulting from aerosol and chamber wall fluorescence induced by the pumping frequency. The latter noise source, under the clean air conditions prevalent in the natural troposphere, typically defines the dominant noise source in the conventional LIF system. II. General Description of Photofragmentation-Laser Induced Fluorescence (PF-LIF) Technique A. Introduction Although the conventional LIEF technique can be expected to provide an effective means of detecting numerous trace gases at natural tropospheric concentration levels, a great many others are not detectable by this method due to the absence of bonding excited states that fluoresce. In the latter case, the absence of fluorescence in the parent molecule may be overcome if the species can be made to photodecompose. Halpern et al. 4 have used such an approach to detect NH 3 at high sensitivity under laboratory conditions. In their system, a high energy excimer laser, producing intense pulses at 194 nm, is used to photolyze NH3 in a multiphoton absorption sequence that results in the production of electronically excited NH radicals. The resulting fluorescence from the NH* radical is then detected at 336 nm. Since this technique depends on the square of the 1 November 1980 / Vol. 19, No. 21 / APPLIED OPTICS 3597

2 VI *2 vw an I -- J Fig. 1. OH energy diagram showing LIF pumping and sampling scheme. energy density, these investigators maximize their sensitivity by strongly focusing the 194-nm laser beam into the center of their detection chamber. This paper reports the details of a different photofragmentation detection scheme, the general approach of which was briefly described by Davis et al. 3 Of particular importance in the proposed new method is the fact that much of the already developed airborne hardware used in our previously reported LIF system 3 should also be compatible with the PF-LIF technique. Thus, the transition from a laboratory prototype system to an airborne compatible unit can be expected to occur without major engineering difficulties. B. PF-LIF Approach In general, five criteria must be met if the PF-LIF technique is to be successfully applied: (1) The molecular species to be detected must have an appreciable cross section at a wavelength Al that is accessible with existing lasers. (2) Absorption by this parent molecule at A must result in bond dissociation (or predissociation). (3) One of the photofragments formed in the bond dissociation process must have a bonding excited state that can be made to fluoresce at wavelength 3 by pumping at a wavelength 2 again accessible with existing lasers. (4) The fluorescence inducing pumping wavelength for the photofragment 2 should be spectrally shifted such that the condition X2 > 3 is satisfied. This condition dictates that the photofragment be formed in an excited vibrational state and that electronic pumping take place before significant vibrational relaxation occurs. A fifth criterion that must be applied to many systems (see later discussion for exceptions) is that the photofragmentation wavelength A also satisfy the condition X > 3. This preferred wavelength arrangement is in diagram form in Fig. 2. Both criteria (4) and (5) are concerned with reducing the white fluorescence noise in the system. This noise source is produced from atmospheric aerosols and from organic contamination on the walls of the fluorescence chamber upon absorbing laser radiation at XA and X2. If criteria (4) and (5) are satisfied, all white fluorescence noise must occur at wavelengths significantly longer than that of the sampled signal fluorescence. Under these conditions, long wavelength rejection filters, in combination with solar blind photomultiplier tubes (PMT), can be used to reduce the white fluorescence noise and Rayleigh and Raman scattered radiation to the level of the PMT dark count. This type of massive discrimination is not possible with conventional LIF systems since some laser induced white fluorescence noise is always present at the central wavelength of the bandpass interference filter. Results from our laboratory, for example, show that a rejection factor of is near maximum for the LIF system, whereas, now appear to be practical in the case of the PF-LIF technique. It was mentioned earlier that, in some cases, the requirement that X > 3 may not be a limiting factor in the application of the PF-LIF technique. This is true for a system in which the vibrationally excited photofragment formed by photolysis at A, does not undergo rapid vibrational relaxation. 5 In this case, the fluorescence pumping laser, 2 could be fired -1 sec after the photolysis laser. Under these conditions, all the white fluorescence noise generated by the photolysis laser beam is decayed away before the fluorescence V 2 V'- I V,* o V * 3 V"- 2 V V,,* X\1 2 > 3 * I o AB2. AB t B fi i ABE _ Fig. 2. Wavelength detection scheme of PF-LIF technique APPLIED OPTICS / Vol. 19, No. 21 / 1 November 1980

3 detecting PMT is gated on at the time of the second laser firing (provided the laser pulse width is not much > nsec). 6 This delayed firing technique may result in reduced sensitivity if, during the delay time, the initial rotational quantum state distribution undergoes further relaxation, thereby reducing the maximum population available for pumping from a single quantum state. Nevertheless, for many molecules that have significant absorption cross sections only in the far UV, the delayed firing sequence may be essential. Ill. Description of PF-LIF Experimental Hardware A laboratory PF-LIF system consists of three major components: (1) one or two high energy pulsed dye laser systems; (2) a fluorescence chamber with associated detection optics; and (3) sampling electronics and data processing hardware. The use of either one or two dye lasers to provide the photolysis and fluorescence pumping wavelengths, X and X 2, depends both on the type of molecular species being detected as well as the type of high energy driver laser used to excite the dye lasers. For example, for some chemical systems either the fundamental wavelength from the driver laser or some harmonic of the fundamental could be used to provide the photolysis pulse at X,. In the latter case, only a single dye laser would be required. The selection of the best driver laser to excite the one or more dye lasers must be based on two criteria: (a) the availability of short pulse widths (i.e., <15 nsec), and (b) the availability of high energies at one or more UV and/or visible wavelengths. The choice in this case, therefore, is limited to either solid state Yag or excimer gas lasers. Because of their rugged durability and very high energies at multiple wavelengths, we believe that a Nd:YAG system is the preferred driver laser. It should be noted also that, even in the case where two dye lasers are to be used to generate the X and X 2 laser pulses, a single Nd:YAG driver laser can suffice to drive both systems if a simple optical delay line is interfaced with the input to the 2 dye laser. For a chemical species requiring that XI be less than X 3, a second Nd: YAG laser is needed. In the latter system, an electronic delay is required to adjust the firing sequence of the XI and X2 lasers. The fluorescence chamber required for the PF-LIF system can be of rather simple design, having the same basic characteristics as outlined for conventional LIF systems. The detection optics, on the other hand, must be somewhat unique if the full potential of the PF-LIF is to be realized. The most important features of this proposed system are (a) the use of chemical optical filters rather than standard solid state interference bandpass filters; (b) the use of nonimaging collection optics rather than conventional imaging spherical lenses; and (c) the use of solar blind PMTs. The use of chemical filters in this system is preferred due both to the higher transmission of a tuned multiliquid filter system and to the relative insensitivity of this type filter to the angle of incidence of the incoming radiation. The latter point becomes quite important when trying to assess the ultimate effectiveness of nonimaging collection optics. Although this type of optical system is capable of collecting many times more light than possible with simple spherical lenses, the resulting higher optical collection efficiency can only be fully realized if subsequent optical filtering elements do not strongly discriminate against off-axis rays. As noted, this is true of simple chemical filters but not of interference bandpass filters. To further complement the long wavelength discrimination characteristics of the proposed chemical optical filtering system, solar blind PMTs are also recommended. Tubes are now available with quantum efficiencies that fall off 11 orders of magnitude in shifting from the near UV to the visible region of the spectrum. The sampling electronics to support a PF-LIF system can be rather simple in design, consisting primarily of photon counting hardware. This system requires gating circuitry, which permits the counting electronics to be activated in a few nanoseconds and permits variable sampling gate widths of nsec. Data processing requirements for the overall system should be minimal due to the low photon count rates expected and could certainly be handled by a small microcomputer. Conversion of a laboratory PF-LIF system into one suitable for aircraft field sampling should be reasonably straightforward if much of the basic airborne LIF hardware 3 already in existence can be used. The PF- LIF system results in an increase in weight of 35% and an increase in electrical power requirements of -100% over those of an airborne LIF system. A ground base PF-LIF sampling system is obviously even more straightforward than that needed for an aircraft platform. IV. PF-LIF Signal Calculations For purposes of illustration, we consider a PF-LIF system that detects an atmospheric species ABC by photolyzing the molecule at a wavelength X, to form a vibrationally excited product AB. It is assumed that a small fraction of this photogenerated AB species is produced in quantum state i. AB in quantum state i is excited by a second laser, centered at wavelength X 2, resulting in the formation of an electronically excited AB species in the j quantum state. Fluorescence from electronically excited AB occurs at wavelength X 3. Thus, assuming that no additional AB is produced in quantum state i from other sources, the general equation for the number of signal X3 photons detected per laser shot can be expressed as DX 3 = (total number of X 3 photons emitted P 3 ) X (optical detection efficiency for X 3 photons Ed) X (electronic detection efficiency Ee). In abbreviated form DX3 = PX 3 X Ed X Ee. (1) The simplest term in Eq. (1) Ee represents the fraction of the total PMT signal detected by the counting electronics: 1 November 1980 / Vol. 19, No. 21 / APPLIED OPTICS 3599

4 E = (PMT signal pulses counted Pd) (2) (PMT signal pulses emitted P.) In most cases, it is expected that the value for this term is unity. 7 The second term in Eq. (1), the optical detection efficiency, may be defined as or E (fraction of total fluorescence at samplin wavelength X3,y m ) X (collection optics efficiency factor at X 3, YX 3 ) X (filter transmission factor at X 3, ZXA) X (quantum efficiency of PMT, X, Ed = YX3 X YX3 X Z3 X OX 3 (3) The first term in Eq. (1) is more complex than those given above and is presented here in several steps. The most general expression for PX 3 is given here in the form of Eq. (4): PX 3 = (total number of X 2 photons absorbed Nx 2 ) (4) X (fluorescence efficiency Ef). The number of X 2 photons absorbed by photofragment AB molecules within the sampling volume may be expressed as NX =(total number of AB molecules in quantum (state i within sampling region C {fraction of molecules in quantum (5) state i absorbing photons Ex 2 The total number of AB molecules in quantum state i within the sampling region can be related to the photolysis pumping scheme and the number of precursor molecules by Eq. (6): C (number of ABC molecules) in sampling region Cr X fraction of ABC molecules undergoing photolysis and producing photofragments in quantum state i The number of ABC molecules within the sampling region C is given by the product of the sampling volume V and the concentration of the precursor species [ABC]: C, = (volume of sampling region V) X concentration of the precursor) species [ABC] P The volume V is given by J I (6) V = a, X 1, (8) where I is the laser path length, and ax, is the laser beam cross section. The photolysis efficiency Ex, from Eq. (6) is defined as Ex = 1 - exp - PA1ux\) x Q x fx F. (9) The exponential term in Eq. (9) accounts for saturation effects that are sometimes important in the photolysis step. The equation presented here is based on an absorbing sphere model, which is valid for low concentrations of absorbers. In this equation, Px 1, X 1, and ax identify the laser photon flux, the absorption cross section of ABC molecules at X, and the X laser beam cross-sectional area, respectively. The Q. term in Eq. (9) is the primary quantum yield for production of AB by photolysis of ABC, whereas, fi represents that fraction of the newly formed AB species that is in the desired quantum state i. The variable F is a correction term designed to take into consideration any change in the population of the ith quantum state that might occur in the time interval between the firing of the X and 2 lasers. This term, therefore, corrects for the effects of vibrational and rotational relaxation. For short (<20-nsec) time delays, the value of this term should be near unity. 8 The second term in Eq. (5), the fluorescence pumping efficiency Ex 2, must take into account both the fraction of molecules absorbing photons within the X2 beam and the imperfect overlap of the and A 2 laser beams. Applying the Beer-Lambert law, we find PX 2 X OA2 ax 2 EX 2 = X - ax 2 ax, (10) where PX 2 is the A 2 laser photon flux, AX 2 is the absorption cross section for the photofragment at X2, ax, and ax 2 are the cross-sectional areas of the photolysis and the fluorescence pumping beams. This form of the equation assumes that the A 2 beam is completely overlapped by the X beam and that the fraction of photofragments absorbing photons is small enough that a single term Taylor series approximation to the Beer- Lambert law yields acceptable results. Returning finally to the second term in Eq. (4), Ef, it is recognized that this quantity can be expanded according to Eq. (11), i.e., Ef = fraction of excited kf molecules producing} fluorescence [kf + kd + kq [MI] where kf is the reciprocal of the natural radiative lifetime r of AB in the ith excited state, kd is the first-order rate constant for dissociation, kq is the bimolecular electronic quenching rate constant, and [M] is the corcentration of the quenching species. Based on detailed Eqs. (5)-(11), Eq. (4) can now be rewritten in the following form: (11) P, 3 = Ex, X Ex 2 X El X V X [ABC]. (12) And, upon substituting Eq. (12) into Eq. (1), the final signal expression becomes Dx 3 = Ex, X Ex 2 X Ef X Ed X E X V X [ABC]. (13) Equation (13) is a very useful form of the signal equation since each term within it has an easily identified physical significance, for example, 3600 APPLIED OPTICS / Vol. 19, No. 21 / 1 November 1980

5 photolysis\ Ex, = efficiency = - exp - ) X fi X F; optical pumping PX 2 X X 2 l EX 2 = efficiency = at 2 ax2 ax, E I-fluorescence [kf I efficiency I [kf + kd + kq[m]] optical Ed = detection efficiency/ (electronic E= detection = Pd efficiency volume of P = 'YX 2 X Yxs X Zx 2 X S V sampling =ax, X 1. region From an examination of Eq. (13) in terms of the detailed expressions for Ex,, EX 2, Ef, Ed, Ee, and V, two important characteristics can be defined regarding the relationship between signal photons detected and the concentration of ABC. (1) To a first approximation, the signal depends upon the average value of the product of the laser photon fluxes. Since the average value of the product of these two quantities is generally not equal to the product of the individual average values, that is, (PxPx 2 ) D (PX,) X (PX 2 ), either a measurement of both laser energies for each laser firing is required or a reliable description of the energy distribution statistics for each laser is needed. (2) In the case of a completely overlapped X 2 beam, the signal is seen to be independent of the X 2 beam cross section. V. Evaluation of Photon Noise Sources and Chemical Interferences In a PF-LIF system, noise sources may be divided into two general categories. First, there are those sources that are independent of the transmission of either laser beam through the fluorescence sampling chamber. Sources in this category include light leaks in the PMT housing, the dark count of the PMT, and natural background radiation entering the fluorescence chamber through the laser entrance port. With the use of short time gates and careful mechanical design, we now estimate that the collective count rate from these sources is negligibly small (in the counts/ laser shot range). x,; The second category of noise involves sources that are directly related to the transmission of one or both laser beams through the fluorescence sampling chamber. For purposes of clarity, we have further divided this second category into two subcategories: (1) photon noise occurring at wavelengths 2X 2 ; and (2) photon noise occurring at wavelengths <X 2 - Sources in the first subcategory include white fluorescence noise from gas phase molecules, aerosols, and wall contaminants; Stokes Raman scatter from gas phase molecules; and Rayleigh scatter from both gas phase molecules and aerosols. For the proposed PF-LIF system, the latter noise source, in principle, could be reduced to levels equal to or less than that of the PMT dark count via a combination of long wavelength chemical filters and solar blind PM tubes. The second subcategory of laser induced noise (i.e., sources that generate noise wavelengths <X 2 ) is obviously more difficult to eliminate. At this time, two possibilities have presented themselves. The first could involve multiphoton short wavelength fluorescence. This type of fluorescence could result from any one of five basic processes, i.e., hv2 (a) AB 2 + hv 2 -[AB 2 ] -- )AB2 AB* - AB 2 + hv 3 hv 2 (b) AC 2 + hv 2 -A AC*2 AC- (V3 > V2), AC* AC 2 + hv 4 (V4 > V2), M (c) AC 2 + hv 2 - ACACI AC + hv 2 - ACI* - AC 2 + hv 5 (V5 > V2), (d) AC + hv 2 - AC2 - AC + hv6 (v6 < V2) ACI + hv 2 AC* - AC 2 + hv 7 (7 > V2), (e) AD 2 + hv 2 - AD2 hv' 2 AD* + D AD* - AD + hvs (v8 > V2) In the above schemes V 2 is the pumping laser frequency, [AB 2 ] represents a virtual excited state, AB* is electronically excited AB 2, AC** is a bonding upper electronic excited state of AC 2, AC is vibrationally excited ground electronic state AC 2, AC* is a highly vibrationally excited state of the first excited electronic state of AC 2. The above two-photon noise sources are by far the most difficult to evaluate due to a lack of quantitative data on photochemical quantum yields, quenching cross sections, absorption cross sections, Franck-Condon factors, and atmospheric concentration levels. However, for purposes of obtaining order or magnitude estimates of this noise source, we examined here the minor atmospheric constituents N 2 0, C0 2, H 2 0, CO, and CH 4 plus the fluorescing and nonfluorescing simple polyatomic trace gases SO 2, CS 2, CH 2 0, H 2 0 2, CH 3 00H, NH 3, HNO 3, H 2 S, and COS. Finally, several representative complex species were investigated; these include acetaldehyde, butylaldehyde, acetone, benzene, 1 November 1980 / Vol. 19, No. 21 / APPLIED OPTICS 3601

6 napthalene, and anthracene. Our general findings are that process (a) was always found to be unimportant as a noise source because of the low cross-section values inherently involved in two-photon absorption processes. Calculated noise signals from process (b) are significantly greater than those derived from (a) but still are typically lower than the PMT dark count. The low efficiency of process (b) reflects the high probability for bond dissociation when molecules are energized by two photons. Process (e) has been estimated to be more important than (b) but usually only for that case where V8 < 2. A study of the latter type of system was reported by Halpern et al. 4 in which excited NH was formed from the photolysis of NH 3 (see Sec. II. for details). Potentially, the largest source of two-photon noise comes in the form of processes (c) and (d). For these processes, the most important molecules are believed to be the strong UV absorbing species napthalene and anthracene. If present in the gas phase at the pptv concentration level, it is estimated that photon noise levels as high as 10-6 /laser shot result. Although we presently believe that this is a worst case estimate, the possible uncertainties in the values used for the internal conversion quantum yields and for quenching seem to justify a very careful examination of these processes under controlled laboratory conditions. The second low wavelength noise source (i.e., Xnoise < 2 ) is anti-stokes Raman scattering. The molecules N 2, 02, and CO 2 are, in general, the dominant anti- Stokes Raman scattering sources in atmospheric sampling due primarily to their high ambient concentrations. Calculations indicate that the minimization of this noise source requires the selection of a laser pumping scheme that maximizes the separation between the pumping and sampling wavelengths, for purposes of discrimination, but does not result in major losses in signal counts due to poor Franck-Condon factors. In general, if the separation between 2 and X 3 is at least 20 nm, chemical filters can be designed that strongly discriminate against first-order anti-stokes lines and, to some degree, second-order anti-stokes noise. Even so, present calculations indicate that anti-stokes Raman scattering probably define the lower limit on detection sensitivity for a PF-LIF system. In addition to the real noise sources outlined, one must also consider interference signal photons derived from chemical species other than the one specifically being monitored, in our example, ABC. If, for instance, the photofragment AB also occurs naturally in the atmosphere, a potential problem could develop if the Boltzmann distribution of this naturally occurring species contained a significant number of molecules in the same quantum state i used for fluorescence pumping. This interference, however, can be eliminated by the proper selection of the AB quantum state i. If the i state, for example, consists of AB molecules energized by two or more vibrational quanta, the naturally occurring population at this energy is expected to be negligible. The second type of chemical interference that can occur in the PF-LIF system is that involving the production of the photofragment AB molecule from the photolysis of an atmospheric trace gas other than ABC. In this case, some fraction of the AB formed could be produced in quantum state i, and these species would be indistinguishable from those produced from ABC. Even so, the selection of XI and X 2 provides a great deal of flexibility in eliminating interferences of this type. For example, by varying the photolysis wavelength, the absorption cross section for molecule ABC, relative to that for one or more possible interfering species, could very likely be changed by factors ranging from 2 to 10. Thus, the magnitude of the interference signal could be quantitatively defined and then eliminated, if necessary, by selecting a new photolysis wavelength. Equally important is the potential selectivity afforded by varying X 2, the fluorescence pumping wavelength. Since it is highly probable that each atmospheric trace gas photodissociated at wavelength XI produces fragment molecules having a unique distribution of rotational and vibrational quantum states, varying the pumping wavelength X 2 should provide a clear identification of the molecular species being sampled.' 0 The latter type of selectivity, however, necessitates a rather complete advance knowledge of the rotational distributions within one or more vibrational levels of any molecular species believed to be a potential problem. In most cases, it is expected that serious chemical interferences could be avoided by simply comparing the relative concentration levels of the species involved, and the relative absorption cross sections, and the energetics for bond rupture and photofragment excitation. This simple approach was used in formulating the pumping schemes outlined for several molecules in the next section. VI. PF-LIF Sampling Schemes Within the NO. family there are at least three species that appear amenable to detection via the PF-LI F approach. When photolyzed in the visible and/or near UV, the gases NO 2, NO 3, and HNO 2 all produce photofragments capable of fluorescence. Here we explore possible PF-LIF photolysis and pumping wavelength schemes for each of these gases. A. NO 2 (a) NO 2 + hv(x = 300 nm) - NO(X 2 11,V" = 2) + 0(3P), (b) NO(v" = 2) + h 2 (X 2 = 248 nm) - NO(A 2 F+,v' = 0), (c) NO(A 2 j+,v' = 0) - NO(X 2 11,v" = 0) + hv 3 (X3 = 226 nm). The selection of 300 nm for X, in process (a) involves a compromise between the NO 2 absorption cross section," the achievable photon flux, and the fraction of the NO photofragment population that can be produced in the v" = 2 vibrational manifold. The wavelength assignment of 300 nm, therefore, represents a near maximum in the product of the achievable photon flux from a frequency doubled dye laser and the absorption cross section for NO 2 in a wavelength region believed 3602 APPLIED OPTICS / Vol. 19, No. 21 / 1 November 1980

7 optimal for populating the v" = 2 state of NO. Wavelength selections for X 2 and X 3 are based on consideration of the values for 0 X 2, P 2 and the Franck-Condon factors for the several possible electronic transitions for NO. Those selected here tend to maximize the value of Dx 3 without resulting in proportionate increases in the anti-stokes Raman noise level. As seen in scheme (a)-(c), the assignment of 248-nm radiation for X 2 and 226 nm for the sampling wavelength also makes possible a 22-nm separation between X 2 and X 3. Thus, the use of the v" = 2 vibrational manifold as the origin for process (b) allows for large scale discrimination against the first anti-stokes Raman lines from N 2, 02, and CO 2 and even limited discrimination against second-order anti-stokes radiation from 02 and CO 2. The second-order lines from N 2 are sufficiently weak as to be of negligible importance as a noise source. B. NO 3 The wavelength selection of 248 and 226 nm for X 2 and X 3 in the NO 2 system can also be used in the detection of NO from NO 3. In the latter system, however, due to the very low natural concentration levels of NO 3, a photolysis wavelength X, must be selected that energetically can produce v" = 2 NO photofragments from NO 3 but not from the far more populous NO, species such as NO 2 or HNO 3. Based on the known photochemistries for these three species, the desired energy configuration can be achieved by selecting photolysis wavelengths that are 580 nm. Neither NO 2 nor HNO 3 undergoes bond rupture at these wavelengths. For NO 3, the process (d) NO 3 + hv(xi = 589 nm) - NO(X 2 11,V" = 2) (0A) should occur with high probability. The exact assignment of 589 nm as the photolysis wavelength was determined from available information on the absorption cross section for NO 3, given by Johnston and Graham,12 and from the quantum yield values recommended by Magnotta.1 3 C. HNO 2 The detection scheme for HNO 2, like that for NO 3, can involve X 2 and X 3 assignments of 248 and 226 nm, respectively. Once again, however, NO in the v" = 2 level needs to be formed from the photolysis of HNO 2 at a wavelength that energetically could not produce this same quantum state of NO from NO 2 or HNO 3. In this case, the selection of the third harmonic Nd:YAG (i.e., 355 nm) appears to be a unique assignment since (1) HNO 2 has an absorption maximum near this wavelength; (2) very high photon fluxes can be generated at 355 nm; (3) the HNO 3 absorption cross section is -3 orders of magnitude lower than that for HNO 2, and available photochemical evidence also suggests that NO is not a primary product in the photolysis of HNO 3 at long wavelengths; (4) absorption by NO 2 can only produce NO in the v" = 2 level provided all excess energy were to go into vibrational excitation, a highly improbable event; and (5) the absorption cross section for NO 3 is very low at 355 nm, and photolysis at that wavelength results in a negligibly small quantum yield for production of NO. The proposed method for detecting HNO 2 may therefore be summarized as (e) HNO 2 + hv(xi = 355 nm) NO(X 2 11,v" = 2) + OH(X 2 11) followed by processes (b) and (c) given earlier. In the HNO 2 system, the possibility also exists that OH could be used as the monitored photofragment; however, the data reported by Cox et al.' 4 "1 5 suggest that in the photolysis of HNO 2 at 355 nm, very little excess energy is partitioned into internal energy involving the OH species. If true, this results in a decrease in the separation between X 2 and X 3 and thus significantly reduces the sensitivity of the PF-LIF technique. VIl. Other Species At the present time, the number of trace gases that potentially could be detected via the PF-LIF technique appears to be quite large. Among the possible candidates are H 2 0 2, HNO 3, CH 3 00H, CH 3 I, CH 3 Br, CH 3 0NO, CS 2, CH 3 SH, NH 3, and INO 3. Detectable photofragments consist of OH, I, Br, NO, CS, SH, NH, or NH 2 and NO or 10. Future advances in the spectroscopy and photochemistry of each of the above parent molecular species should thus provide a basis for evaluating PF-LIF detection schemes. Vil. Sensitivity Calculations for Three NO, Gases Calculations are presented that estimate the detection sensitivity of the PF-LIF system for NO 2, NO 3, and HNO 2. The choice of these three species was based both on there being available reasonable estimates of the absorption cross sections and the fact that, for all three systems, NO could be selected as the photofragment from which fluorescence is monitored. In these calculations, we also made the simplifying assumption that in each case quantum state i is the NO state from which fluorescence pumping occurs. This simplification allowed us to establish several common parameters for each system investigated. For example, since the terms Ef, Ed, and Ee do not depend upon the characteristics of the X, or X 2 laser beams, their values can be taken to be the same for each NO. system. The product E 2 X V is also independent of the X, laser; however, to establish specific values of EX 2 and V, we made the further assumption that ax is the same for all three detection systems. Estimated values for these common terms are in Table I. Ex 1 has not been included in Table I since this quantity must be estimated independently for each NO. species being considered. In Table II we summarized the estimated values of Ex 1 for each NO. system, assuming negligible vibrational and rotational relaxation. Finally, Tables I and II were combined to yield mathematical expressions that give the detected signal as a function of the ambient concentration (in molecules/cc) of the parent NO. species. The appropriate equation for each of these systems can be expressed as 1 November 1980 / Vol. 19, No. 21 / APPLIED OPTICS 3603

8 Table 1. Estimation of E,,, E,, Ed, E., and V for Detection of NO 2, NO 3, and HNO 2 Term Variables Ref. E,\ 2 = 3.4 X 10-2 Px 2 = 2.1 X 1015 photons ax, = cm 2 18 a,, = 0.5 cm 2 Ef = 7.1 x 10-3 kf = 6.3 X 106 sec kd = 0 20 k = 1.7 x 1010 cm 3 /sec 21 [Ml = 5.2 X 1018 cm Ed = 1.4 X YX 3 = YX3 = 0.09 ZX3 = 0.05 OX 3 = 0.20 Ee = 1.0 V = cm 3 ax = 0.5 cm 2 I = 1.5 cm Table II. Estimation of ER for Detection of Three NO, Species Px, I x, fi Species (X101 6 ) (cm 2 ) 1 1 (X10-2 ) Q F EXI NO X X 10-5 HNO x X 10-4 NO X 10-' X 10-3 D, = 1.1 X cm 3 X [NO 2 ], Dx 3 = 2.5 X cm 3 X [HNO 2 ], Dx3 = cm 3 X [NO 3 ]. In Table III the signal strength expressions given above were combined with anti-stokes Raman noise estimates to permit an evaluation of the SNR for each NO, species. Several different ambient concentration levels and integration times are included. These results suggest that for integration times of -5 min., 3 pptv (at sea level) of NO 2 could be detected with a SNR of 2:1. For NO 3 and HNO 2, the concentrations required to produce SNRs of 2:1 for the same integration time correspond to 0.3 and 2 pptv, respectively. These estimated detection sensitivities indicate that NO 2 and NO 3 could be detected at their natural levels in the unpolluted troposphere with reasonably good time resolution. On the other hand, the natural concentration level of HNO 2 appears sufficiently low that detection of this species is doubtful. The detection of HNO 2 at the few pptv level, however, could prove useful for other applications. For instance, the free radical species HO 2 is estimated present in the atmosphere at concentration levels of 2-40 pptv. This important atmospheric free radical is also known to react very rapidly with NO according to the reaction HO 2 + NO - NO 2 + OH. Thus, if high concentrations of NO were injected into a sampled airstream, the HO 2 is first converted into OH and NO 2 followed by the production M of HNO 2 via the very fast reaction OH + NO HNO 2. Since HO 2 is present in the atmosphere at concentrations >1 order of magnitude higher than natural OH, the HNO 2 formed in the above reaction presents itself as a unique diagnostic molecule, which potentially could be used to determine the atmospheric concentration level of HO 2. Even the NO 2 formed in reaction NO + HO 2 - NO 2 + OH could be used as an effective monitor of HO 2 provided the natural tropospheric NO 2 level was not much different than that of HO 2. Potential complications resulting from other peroxy radicals (i.e., CH ) could be resolved by measuring the concentrations of both titration products NO 2 and HNO 2. IX. Summary and Conclusions In this paper we outlined a new method for detecting nonfluorescing atmospheric trace gases. This technique appears both highly sensitive and specific. The general equations for calculating the signal strength were given. Possible sources of noise as well as chemical interferences were discussed. Based on available information, the dominant noise source in the PF-LIF system was identified as anti-stokes Raman scatter involving the atmospheric species N 2, 02, and CO 2. It is also noted, however, that multiphoton fluorescence noise might later prove significant in some systems. Extensive experimental testing is required to resolve this question. Detection schemes for employing the PF-LIF system in the monitoring of several NO. species were presented. Using the signal equations developed here, Table. PF-LIF Signals for Detection of Three NO, Species Number of laser shots Estimated Estimated 2 4 Signal Signal 2 5 (min at signal noise real statistical Species Concentration 20 Hz) counts counts noise noise NO 2 5 X (5 min) :1 4.5:1 2.5 X (10 min) :1 4.5:1 7.5 X (20 min) :1 3.5:1 HNO 2 5 X (5 min) :1 6.8: X (10 min) :1 4.8:1 2.5 X (10 min) :1 5.5:1 NO X (10 min) :1 5.5:1 7.5 X (20 min) :1 4.5:1 2.5 X (20 min) :1 2.6: APPLIED OPTICS / Vol. 19, No. 21 / 1 November 1980

9 along with estimated noise levels, we carried out detailed calculations to determine the detection sensitivity of the PF-LIF technique for the gases NO 2, NO 3, and HNO 2. While the detection sensitivity for HNO 2 (5 X 107 molecules/cc for 5 min of integration) does not appear sufficient to measure this species at its natural levels in the unpolluted troposphere, the sensitivity limits on both NO 2 and NO 3 now appear more than adequate. The calculations presented suggest that for a 5-min integration time period the detection limits for NO 2 and NO 3 at sea level should be 3 and 0.3 pptv, respectively. In addition to in situ atmospheric concentration measurements of trace gases, the PF-LIF technique may be amenable to other applications. For example, studies of molecular photodynamics, especially in the study of energy partitioning in photofragments, may be advanced with the application of this methodology. The authors would like to acknowledge NASA grant NAGL-50 which provided funding for this work. Appendix: Symbols axi cross-sectional area of X 1 beam; ax2 cross-sectional area of X 2 beam; [ABC] concentration of parent species; Ci number of AB molecules in quantum state i within sampling region; C, number of ABC molecules in sampling region; Dx, number of signal counts/laser shot; Ed optical detection efficiency for X 3 photons; Ee electronic detection efficiency; Ef fluorescence efficiency; Ex, photolysis efficiency; EX 2 fluorescence pumping efficiency; fi fraction of photofragment AB in desired quantum state i; F vibrational and rotational relaxation correction; lyx3 Franck-Condon factor of desired fluorescence transition; kd first-order rate constant for dissociation of AB in the jth excited state; kq bimolecular rate constant for electronic quenching of AB in the jth excited state by molecule M; kfi reciprocal of the natural radiative lifetime of AB in the jth excited state; 1 laser path length through sampling region; Al photolysis wavelength; X2 fluorescence pumping wavelength; X3 fluorescence wavelength; [M] concentration of quenching species; NX 2 number of photons absorbed by AB molecules within the sampling region; Pd number of signal pulses counted by detection electronics; Px 1 X1 photon flux; PX 2 X 2 photon flux; P, 3 number of fluorescence photons emitted; P., number of signal pulses outputted from photomultiplier tube; quantum efficiency of the photomultiplier tube at X 3 ; Q., primary quantum yield for production of AB by photolysis of ABC at X,; ax 1 absorption cross section for ABC at Xi; AX 2 absorption cross section for AB at X 2 ; V volume of sampling region; YX 3 optical collection efficiency for X 3 photons; ZX 3 filter transmission at X 3. References 1. K. Sukurai and H. P. Broida, J. Chem. Phys. 50, 2404 (1969). 2. See, for example, W. M. Jackson, J. Chem. Phys. 50,960 (1973); J. A. Silver, W. L. Dimpfl, J. H. Brophy, and J. L. Kinsey, J. Chem. Phys. 65, 1811 (1976); J. A. Gelbwachs, M. Birnbaum, A. W. Tucker, and C. L. Fincher, Opto-Electronics 4, 155 (1972). 3. D. D. Davis, W. S. Heaps, D. Philen, M. Rodgers, T. McGee, A. Nelson, and A. J. Moriarty, Rev. Sci. Instrum. 50, No. 12, 70 (1979). 4. J. B. Halpern, W. M. Jackson, and V. McCrary, Appl. Opt. 18,590 (1979). 5. That is, having Ile time constants for vibrational relaxation >1,usec at atmospheric pressure. 6. This also assumes that the RC time constant of the signal cable and related input-output detection electronics is <20 nsec. 7. This will be true if two conditions are satisfied: (1) the signal lost in output signal gating is negligible; (2) impedance mismatches that result in reflections in the signal cable terminations are small so that, after two reflections, pulses on the signal cable are undetectable. 8. Most diatomics have le relaxation times from 5 to 20 times longer than this time delay. 9. Interfering AB molecules might also be formed by secondary photochemistry involving the XA and/or X 2 laser beams. 10. This technique of using vibrational and rotational population distributions to "fingerprint" molecules may have widespread use as an analytical tool for identification of species. 11. E. R. Reiter, "The Natural Stratosphere of 1974," CIAP Monograph I, DOT-TST (1975). 12. H. S. Johnston and R. A. Graham, J. Phys. Chem. 77, 62 (1973). 13. F. Magnotta, unpublished results (1979). 14. R. A. Cox, J. Photochem. 3, 291 (1975). 15. R. A. Cox, J. Photochem. 3, 1975 (1975). 16. In the (2,2) transition, k' levels <15 want to be excited to prevent extensive predissociation of the OH. 17. A. B. Callear, Proc. R. Soc. London Ser. A: 276, 401 (1963). 18. Extrapolated from the work of T. Tajime, T. Saheki, and K. Ito, Appl. Opt. 17, 1290 (1978). 19. Corresponding to an assumed natural radiative lifetime of 160 nsec. 20. G. Herzberg, Spectra of Diatomic Molecules (Van Nostrand Reinhold, New York, 1950). 21. L. A. Melton and W. Klemperer, Planet. Space Sci. 20, 157 (1972). 22. U.S. Standard Atmosphere (U.S. GPO, Washington, D.C., 1966). 23. R. J. Spindler, L. Isaacson, and J. Wentink, J. Quant. Spectrosc. Radiat. Transfer 10, 621 (1970). 24. Estimated to be due principally to residual anti-stokes Raman scattering from the X 2 laser beam. 25. If the signal is assumed to follow a Poisson distribution, the standard deviation is given by a = an, where N represents the total number of counts. This column thus gives the ratio of the signal to the statistical counting uncertainty. 1 November 1980 / Vol. 19, No. 21 / APPLIED OPTICS 3605