1644 PROCEFDINGS OF THE IEEE, VOL. 59, NO ...web.stanford.edu/~rlbyer/PDF_AllPubs/Conference Papers...rotational relaxation time of the order of 10-9 s. The nonradiative vibrational

1644 PROCEFDINGS OF THE IEEE, VOL. 59, NO. 12, DECEMBER 1971

Comparison of Laser Methods for the Remote Detection of Atmospheric Pollutants

Abr t ra6Three m o t h o d s of remote air pollution detection-Raman bodncattering, resonance backscattering, and resonance absoWh+- a n discussed and compared. Theoretical expressions are derived for the minimum detectable pollutant concentration, and in each case the depth resolution and the problems of interference, pump depletion, and background noise are discussed. A brief discussion of possible l o u r sources is included, numerical examples of the dehdabilities based on prescnt technology are given. Tho atmospheric transparency limits the useful range to a few kilometers for the Raman and resonance backscattering schemes. For the resonance absorption technique the useful mnge can be as gnat CH 50 kilometers.

I. INTRODUC~ON

A T THE PRESENT time the National Air Pollution Control Administration is monitoring over 40 atmospheric pollutants across the United States [ I]. In general, 90 percent

of these pollutants are in a gaseous state in concentrations that vary between 0.01 to 10 ppm for molecules to 0.01 to 10 ppb for metal vapors. To detect and obtain quantitative measurements of these pollutants presently requires fixed monitoring stations using wet chemical techniques with integration times varying from one minute to a few hours. There is a need for a sensitive, instantaneous detection method which can quantitatively measure atmospheric gaseous pollutants over a wide range of concentration.

In this paper three optical methods for pollutant detection are compared and discussed: 1) Raman backscattering, 2) resonance backscattering, and 3) resonance absorption. The application of these methods is now possible due to the invention of the laser and to the recent development of tunable coherent light with the dye laser [2] and parametric oscillator [3].1

The Raman scheme has been discussed previously [4], and preliminary experiments have been performed [5]-[8]. Also, the resonance backscattering scheme has recently been used to detect sodium atoms in the upper atmosphere [g]. However, a complete discussion of the problem of pollutant detection based on these approaches has not been given, and questions Qf sensitivity, interference, atmospheric path limitations, detectors, and required laser powers have not been considered. In this paper we consider these questions in detail for the three detection schemes, and apply the results to the specific problem of measuring small concentrations of molecular and atomic pollutants.

In the next section, theoretical expressions are developed for the Raman backscattering intensity, the resonance backscattering intensity, and resonance absorption. The problems of saturation of the pollutant transition and of pump depletion by high pol-

Manuscript received February 18, 1971; revised July 29, 1971. This work was supported in part by the National Aeronautics and Space Administration, under Grant NGL-05-02Cb103, and the Of&e of Naval Research, under Contract N00014(67-A-0112)0044.

The authors are with the Microwave Laboratory, Stanford Uni- versity, Stanford, Calif. 94305.

1 This article [3] reviews recent progress in parametric oscillator development.

lutant concentrations are also considered. These results are then applied to specific examples in Section IV.

In Section 111 the practical problems associated with all three schemes are discussed. These include atmospheric transparency limitations due to molecular Rayleigh scattering, Mie scattering, and H a and COz absorption; detector capability in the visible and infrared; and background radiation limitations due to backscattered pump radiation and to solar radiation.

Section IV compares the three detection schemes. The discussion is aimed at showing the limitations of the schemes using present laser technology and assuming obtainable detector, spectrometer, and optical system parameters. The discussion includes the effect of atmospheric transmission and background radiation levels. Depth resolution is also discussed for the Raman and resonance backscattering schemes. Brief descriptions of present laser capabilities and of the important parameters for the dye laser and parametric oscillator are also given in order to place the use of these laser sources in proper perspective.

II. THEORY This section considers three potentially useful remote air

pollution monitoring schemes. The expressions for measured intensity for Raman backscattering, resonance backscattering, and resonance absorption are developed. Section IV uses the princi- pal results for a direct comparison of each detection scheme.

A . Raman Backscattering

The Raman scheme requires only a single wavelength laser, whereas the resonance backscattering and the atmospheric transmission schemes require a tunable optical source. The Raman scheme has previously been discussed by Inaba and Kobayasi 141. Cooney [6] has measured the Raman backscattering from nitrogen, Me@ [7] from water vapor in the atmosphere, and recently Inaba et al. have measured COz and SO2 concentrations using the Raman technique [8].

The Raman lines of a molecule are shifted from the laser pump frequency by the characteristic vibrational frequencies of the molecule. These vibrational frequencies are specific and allow interference-free detection of many important pollutant molecules. Fig. 1 shows the important pollutants and their re- spective Raman frequencies. The absolute concentration of each pollutant can be determined by comparing the backscattered intensity with the intensity of the Nz or 02 Raman lines.

The Raman backscattering scheme has the following advantages: use of a single wavelength laser in the transparent spectral region of the atmosphere; laser pump and detector optics at the same location; and the measurement of pollutant concentration with good depth resolution. These advantages make the Raman system very attractive for remote air pollution monitoring. The potential disadvantages are the lack of sensitivity over long distances and the need to use high-power lasers with potential eye safety hazards.

KILDAL AND BYER: LASER DETECnON OF ATMOSPHERIC POLLUTANl3 1645

LASER

0 1000 2000 3000 4Ooo

RAMAN SHIFT (cni')

Fig. 1. Raman shifts of important pollutant molecules in the atmosphere relative to the exciting laser frequency.

The Raman differential backscattering cross section for scattering from the fundamental vibrational mode of a molecule is given by

where w1 is pump frequency, w2 the Raman frequency, and ro,la is the Raman polarizability [lo]. The two frequencies differ by the vibrational mode frequency wv so that we have w e =wl -av. Fig. 2 shows a typical Raman spectrum at 300°K. The spectrum is for a (00'0) to (10'0) transition in Con. The Q branch has the same rotational fine structure as the 0 and S branches, but the dis- placement between the lines is very small. For example, for COz the width of the Q branch is only 0.3 cm-l, and the linestrength is about ten times larger than for the 0 and S branches.

The backscattered Raman intensity at time t after the laser pulse is given by

where W1 is the total energy of a single laser pulse, R is the distance to the polluted area contributing to the intensity at the time t , N d R ) is the concentration of a pollutant, and TI and TO are the atmospheric transmittances at the frequencies w1 and w2. The observed Raman intensity arises from a region of a depth one-half the laser pulse length ct0/2. For simplicity N d R ) and R' are assumed constant over this distance.

In the Raman scheme selective excitation is not possible, so the backscattered signal includes the Raman scattering from all the pollutants. Since the total width of a Raman line, including all three branches, is typically more than 100 cm-l, interference is a problem, especially from Oz'and NO. The interference problem can be reduced by limiting the detection to the narrow Q branch at a cost of a small reduction in the effective scattering cross section. For example, for COz Fig. 2 shows that the reduction in cross section is about 10 percent.

One of the principle advantages of the Raman scheme is the depth resolution. With the detector followed by a gating circuit of width to, the depth resolution A R R a m is

C ARRam = - ( t o + to) .

2

Since Raman scattering is an instantaneous process, only the laser pulse length to and the gatewidth limit the depth resolution.

B. Resonance Backscattering

With a tunable laser source it is possible to excite selectively various pollutants. We consider two possibilities: excitation of infrared vibrational transitions, and excitation of atomic and

(3)

0-BRANCH

Q-BRANCH

'S - BRANCH

WAVENUMBER (crn-')

Fig. 2. Calculated Raman spectrum for C O z for a (00'0) to (10'0) transition at 300°K.

P-BRANCH R- BRANCH

2050 I 00

WAVENUMBER ( c m ' : I

2200

Fig. 3. Calculated absorption spectrum for CO for a fundamental vibrational transition at 300°K.

molecular electronic transitions. After excitation, the excited pollutant emits spontaneous radiation into a solid angle of 47r steradians. Monitoring the backscattered radiation determines the pollutants and their relative concentrations. Absolute concentrations are not so easily determined as in the Raman scheme. This is particularly the case for infrared transitions, since the backscattered intensity cannot be compared to the infrared in- active Ne or 0 2 molecules. As in the Raman scheme, the transmitter and the detector are at the same location.

In the following discussion we consider resonance backscattering from molecular infrared transitions. The infrared absorption lines are generally narrow and specific for each pollutant molecule. The results also apply with minor changes to the case of atomic and molecular electronic transitions.

Several absorption lines are associated with each vibrational transition of a molecule. Fig. 3 shows a typical absorption spectrum for a simple molecule that consists of the P and R branches. Molecular rotation causes splitting into a line spectrum. The linewidth of a single vibrational-rotational line at atmospheric pressure is typically 0.1 5 cm-', while the total width of the spectrum can be several hundred wavenumbers at 300'K. In the cal-

1646 PROCEEDINOS OF THE IEFE, DECEMBER 1971

culations we assume that the linewidth of the tunable pump source is sufficiently narrow so that only a single vibrational- rotational transition is excited.

Several lifetimes are important for the resonance backscattering scheme. The excited molecules reach rotational equilibrium before they decay back to the ground state, because of the fast rotational relaxation time of the order of 10-9 s. The nonradiative vibrational relaxation timer is much longer, and is approximately lW4 to 1 P s at atmospheric pressure, depending on the molecule [ll]. The spontaneous radiation decay time is even longer, being about 10-l to 1 0 1 s. Since the nonradiative vibrational decay time is much faster than the radiative decay time, the nso- ~ n c e backscattering efficiency is reduced by about the ratio between the two decay times.

Since most of the molecules reach rotational equilibrium before they decay back to the ground state, the backscattered radiation is emitted over a wide spectral band consisting of the P and R branches. Use of a narrow-band light source, with a linewidth comparable to the linewidth of a single vibrational- rotational line utilizes the total energy to excite the desired vibrational transition. In addition, narrow-band excitation reduces possible interference from other pollutants.

The backscattered spontaneous radiation intensity is calculated in Appendix I. Equation (98) in this Appendix gives the total backscattered intensity from both the R and P branches. If the optical bandwidth of the detection system is less than the width of the two branches to reduce background radiation or interference from other molecules, ihe signal intensity is decreased. When a single vibrational-rotational line is detected, the reduction is approximately 2 0 .

Setting the transmittance of the atmosphere equal to unity and assuming uniform pollutant concentration beyond Rmin, the distance at which the pollutant starts, (98) reduces to

S P

I (t) = W u n . o N t o t - - - m wn,n-1 ~ n . n - 1 s n , o ( t > (4)

4~ Un.0 k i n

where r is the nonradiative relaxation time, to is the laser pulse length, and the other quantitites have been defined previously.

Over distances where the pump depletion can be neglected, the resonance backscamring scheme has a depth resolution MIe8 given by

a r e a Ej: C - ( t o + 7 + 1,) (6) 2

where to is the gatewidth. Combining (6) and (3) yields

The Raman depth resolution can be improved by decreasing the gatewidth and shortening the laser pulse length. In the resonance backscattering scheme the depth resolution is limited to

From our previous estimates for T we see that ARre” varies between 0.15 and 15 km for infrared transitions. This leads to the consideration of resonance backscattering for special detection geom- etrih, such as layered polluted regions with clear air between the detector and the polluted layer. Under this circumstance the lack of depth resolution may not be a severe handipp.

The fluorescence decay time for electronic transitions is as short as a nanosecond. For this case the depth resolution is comparable to the Raman scheme.

Present levels of pollution in the atmosphere are at times so higb that pump depletion is important. Under these conditions radiation trapping may present a problem particularly for transitions terminated in the ground state. The trapping both reduces and delays the backscattered radiation.

It is easily shown that radiation trapping is important only for distances greater than the pumpdepletion distance R-. In general, R- is given by

In other words, the pump beam is depleted before it reaches a depth from which trapping of the backscattered radiation is important.

C. Resonance Absorption Resonance absorption, similar to resonance backscattering,

may be used for the remote detection of pollution [12]. How- ever, it has the disadvantage of needing a remote detector or re- flective target to receive the transmitted beam. The resonance absorption technique measures the total amount of pollutants in the light path without depth resolution. The advantages of this scheme are its relative simplicity, its good sensitivity, and the use of low-power light sources.

The transmitted intensity at the detector follows from (94) in Appendix I. We have

where Z(0) is the intensity of the emitted pump beam, Tn,0 the atmospheric transmittance, and R the path length from the pump source to the detector. Saturation is neglected. For the conditions under which this is true we refer to the discussion in section IV.

By detuning the pump source so that it is outside the absorption line, the transmitted intensity is

The integrated pollutant concentration is determined from the ratio between the on and off line intensities and is given by

KEDAL AND BYER: LASER DETECTION OF ATMOSPHERIC PoLLUTANTs 1647

TABLE I RAYLEIGH AND MIE SCATI'@RPK~ con?ncmm

I < " I [micron]

0.2

0 3

0 . h

0.6

0.0

1 .o

5 .O 10.0

3.1

2.0

1.1

0.P

Assuming that it is possible to detec :t a

2.2 1.b

1.6 0.84

1.1 0.58

0.72 0.35

0.9 0 . a

0.45 0.11)

0.086 0 . a

0.&3 0.010

L 5-percent change in transmission on tuning to the absorption line, the minimum integrated density which can be measured in a given path length is

The value chosen for Z,,Otrsns/Zotftrans is conservative and can be improved by proper signal processing techniques.

Before considering each system and its parameters it is necessary to discuss general factors that limit the sensitivity and usefulness of all three schemes. These factors include atmospheric transmittance and scattering, background radiation, and detector parameters. The next section considers these factors as they apply to each of the detection schemes.

ITI. ATMCSPHERIC T R A N ~ A N C E , DETECTION S Y ~ PARAMETERS, AND BACKGROUND RADIATION

A. Atmospheric Transmittance Atmospheric transmittance is limited by two mechanisms:

elastic scattering loss due to Rayleigh and Mie scattering, and absorption. The transmission loss limits the range of the remote detection systems by pump depletion. In addition, the elastic scattering increases the background level at the detector. This section discusses the Rayleigh and Mie scattering cross sections and the atmospheric absorption for both the ultraviolet and the infrared spectral regions. The results are stated so that the practical limitations set by the atmospheric properties can be determined for each of the three detection schemes.

Elastic scattering is classified as Rayleigh or Mie scattering, depending on whether the particle size is much smaller or comparable to the wavelength. Rayleigh scattering is inversely proportional to the fourth power of the wavelength, so it is most important in the ultraviolet region of the spectrum. Mie scattering is much less wavelength dependent and is usually much larger than the Rayleigh scattering in the visible and near infrared. The magnitude of Mie scattering depends on the aerosol concentration, and varies with changing atmospheric conditions in con- trast to molecular Rayleigh scattering, which remains fairly constant. A simple empiricial relation relating the Mie scattering coe5cient to the visual range is often used [ 131 and is given by

3.91 0.55 o.5ssv* aXYis = - __

V [ x 1 km-l. (14)

Here X is the wavelength in microns and V is the visual range in kilometers. Raman scattering also contributes to scattering loss. However, the cross section is typically three orders of magnitude smaller than the Rayleigh scattering cross section, and Raman scattering is therefore neghgible. For reference we list the Ray-

m -

1800 2000 2200 2400 2600 2800 3000 3200 WAVELENGTH ( % I

Fig. 4. Atmospheric absorption coefficients for ozone and oxygen in the ultraviolet [15].

leigh [14] and Mie scattering coefficients for various wavelengths and visibility ranges in Table I.

In the Raman scheme a short pump wavelength is preferred due to the larger scattering cross section. For wavelengths longer than about 2500 A no problems are normally encountered with atmospheric absorption, and the range of the pump beam is limited by the elastic scattering loss. At wavelengths shorter than 2500 A, oxygen absorption becomes important for longer pathlengths (> 100 m), and at wavelengths shorter than 1850 A the atmosphere is totally opaque. At heavy ozone concentrations, additional absorption occurs due to the wide ozone absorption band between 2OOO and 3000 A. At an ozone concentration of 0.1 ppm the absorption compares with the oxygen absorption at 2500 A. Fig. 4 illustrates how the oxygen and ozone absorption vary with wavelength.

In the infrared the atmosphere has several absorption bands because of water vapor and C02 absorption which limit the useful spectral region to the atmospheric windows. Fig. 5 shows the atmospheric absorption bands together with the vibrational band centers of the most important pollutants. For a more complete discussion of the infrared atmospheric windows see [13]. Within the atmospheric windows the absorption loss over a 1-km pathlength is significant and exceeds the scattering loss in medium- clear-day conditions. As an example, for a visibility of 5 km, a relative humidity of 15 percent, and an air temperature of 60"F, the absorption constants in the two infrared windows 2.70 to 4.30 pm and 4.30 to 6 pm are approximately 0.4 and 0.8 kml, respectively [ 131. The corresponding Mie scattering coefficients are only 0.14 and 0.08 km-1.

From the previous discussion we conclude that the total atmospheric transmission losses, including both scattering and absorption, are approximated by an absorption constant of 1 k m 1 for the ultraviolet spectral region, and 0.5 k m - 1 for the infrared windows.

B. Detection System Parameters This section discusses limiting factors on the detection capa-

bility of the receiving system. Two types of detectors are considered: a photomultiplier tube for the ultraviolet and visible, and a photovoltaic InSb detector, cooled to liquid nitrogen temperature, for the infrared. The cooled InSb detector is useful between 1 and 5.5 pm.

1648 PROCEEDINOS OF THE IEEE, DECEMBER 1971

Fig. 5. Atmospheric transmission in the 1- to 14p region of the infrared together with the important absorption bands of molecular pollutants (visual range is 14 km and water content is 17 mm/nmi which corresponds to a relative humidity of 70 percent at W(F)) [16)-[18].

For shot-noise-limited detectors the signal-to-noise ratio is given by [19]

where i,, iB, and & are the average photodetector currents due to signal radiation, background radiation, and dark current. The electrical bandwidth of the detection system is B, and q is the electronic charge. Depending on which noise source gives the largest contribution, the detectors are said to be photon, background, or darkcurrent limited.

Equation (15) gives the signal-to-noise ratio for detection of a single pulse. The signal-to-noise ratio improves by averaging over a number of pulses. The improvement is directly proportional to the number of averaged pulses n, since the effective bandwidth is reduced to B/n . The maximum number of averaged pulses is &t by practical limitations, such as signal processing electronics and averaging time.

Introducing a detector sensitivity So, we have

i, = SDP, and i g = SDPB (16)

when the signal and the background radiation powers are P, and Pg. In order to detect a signal, we require that the signal-to-noise ratio be larger than (SI&&. Therefore, for photon-limited detection the minimum detectable signal power is

S D

and similarly for dark-current limited operation

Neglecting the background radiation noise, the detector is photon limited when

and darkcurrent limited when the reverse is true. The cathode dark current for the photomultiplier tube is approximately leu A, which means the photomultiplier is photon limited for bandwidths B larger than 3 kHz when (S/&in is set equal to one.

The photovoltaic InSb detector is always dark-current limited

TABLE II DET~c~~RP-

- 1

3.0 X 10"

1

1

3.5 x 10-5

3.0 x 10-9

for practical bandwidths. Introducing the detectivity P for the detector, given by

(1 8) for the darkcurrent limited InSb detector becomes

where AD is the detector area. Together with the sensitivity and detectivity, Table II lists the

minimum detectable signal (MDS) powers for (S/&b equal to one. Table II also gives the maximum background radiation powers allowed for the two detectors in order that they are not background limited. The electrical bandwidth of the photomultiplier is assumed to be 17 MHz, which corresponds to a minimum gatewidth of 30 ns. For the InSb detector the bandwidth is 1 MHz, and the detector area is 0.01 cm2.

For a dark-current limited detector, the constant background intnseity can be much larger than the signal intensity, provided that the detector does not saturate. For the photon-limited detector, however, it is necessary that Pd- is less than the signal power. In the table we have used the stricter condition that P B ~ ~ is less than the minimum detectable signal power.

C. Background Radiation

It is important to notice that, besides the background radiation from natural sources, such as the sky and the terrain, additional background radiation is generated by the elastic backscattering of the pump beam in the backscattering schemes. First we will discuss the background radiatioo from natural sources.

The background power at the receiving mirror, due to an extended source filling the mirror field of view, is [19]

KILDAL AND BYER: LASER DETECXION OF ATMOSPHERIC POUUTANrS 1649

lb’, , , 1

WAVELENGTH x (p) Fig. 6. Spectral radiance of the sky under clear daytime

conditions [19].

P B = TIAXQ,,,AJV(X) (22)

where Ta is the atmospheric transmittance at X, M the optical bandwidth of the detection system, Q,,, the receiving mirror field of view, and N(x) the spectral radiance of the background source. The spectral radiance of the sky under clear daytime conditions is shown in Fig. 6. Equation (22) shows that the background intensity can be reduced by decreasing the field of view and the optical bandwidth of the detection system.

The ozone absorption in the upper atmosphere provides a virtual screen against solar radiation below MOO A. This “night” background condition and the oxygen absorption starting at shorter wavelengths determine the optimum pump laser spectral range for the Raman scheme as between 2500 and 3000 A.

The background radiation due to elastic backscattering is important for both the Raman and resonance backscattering schemes. For the Raman scheme both Rayleigh and Mie backscattering contribute. However, for clear air conditions and an ultraviolet exciting wavelength, Mie scattering is neghgible. The intensity ratio between Rayleigh backscattering from air and Raman backscattering from a single pollutant is approximately given by

where 7 is the relative density of the pollutant with respect to the density at one atmosphere. In (23) it is assumed that Ray- leigh scattering is from a “gas” at one atmosphere, and that the scattering cross section is about three orders of magnitude larger than the Raman cross section. With 11 equal to 1W6, which corresponds to a pollution concentration of 1 ppm, the ratio is 109. Therefore, the receiver must reject the Rayleigh light by at least lo9 against the Raman line in order to detect a 1-ppm pollutant concentration for the photon-limited photomultiplier detector. Present day spectrometers [20] have a rejection ratio of about 106 within 250 cm-’ of the exciting line, and a dual-beam spectrometer [21] has a rejection ratio of about 1Ol2. So for a single-beam spectrometer additional filtering is necessary if the desired sensitivity level is to be reached.

In the infrared Mie scattering is most important. For a Mie scattering constant mYi*, the elastic backscattered intensity is

”_ Yie ,.

assuming that Mie scattering is approximately isotropic. From (4) we have for the excitation of a fundamental vibrational transition that

IMie axark c Rmin -----. (25) Ire’ u 1 , d V t o t WYOR 2RS1,o

Typical values for the absorption cross section and the spontaneous transition probability are

ul.dV,t - 5 X 10% m-l and w:r0 - 30 sl. For a visual range of 5 km and a wavelength of about 5 pm, the Mie scattering coekient from Table I is about 0.086 k n - 1 . For a range R of 100 m we obtain

I Y i a

Ire’ - - lO-”lr] (26)

with R m ~ / 2 R S l , o ~ 1 . For resonance backscattering from vibrational transitions, the detected signal frequencies are closer to the exciting frequency than in the Raman scheme, so the rejection ratio offered by a grating spectrometer is smaller. It is of the order of 104 within 25 cm-1 of the exciting line for a single-beam spectrometer [ 2 0 1 . For a darkcurrent limited detection the background intensity can be somewhat larger than the signal intensity without affecting the signal-to-noise ratio. Therefore, Mie backscattering should. not be a problem for resonance backscattering detection of molecules in the infrared. For the case of atomic pollutants, where the fluorescence radiation is at the transmitted frequency, rejection of the backscattered Mie light is a more serious problem.

IV. NUMERICAL EXAMPLES OF D-ON ~A~ABII ,IIY

In this section we apply the results of the previous analysis and consider in turn the Raman backscattering, the resonance backscattering, and the absorption schemes for air pollution detection. The emphasis is to illustrate the potential of each system assuming the use of the optimum pump laser source which is now available or is soon to be available. The advantages and disadvantages of each system as it relates to sensitivity, interference problems, depth resolution, and range are discussed. The discussion therefore includes a brief review of present laser systems: the ruby laser and its second harmonic output; the Nd:glass laser and its quadrupled output; the Nd:YAG laser and its use as a pump for tunable parametric output; and the dye laser for tunable radiation in the violet and visible part of the spectrum.

Research in the field of nonlinear optics and dye lasers is pro- ceeding at a very rapid pace. Therefore, recent progress is discussed in order to fully evaluate the potential of the remote sensing schemes.

A . Raman Backscattering For the Raman detection scheme we consider the geometry

shown schematically in Fig. 7. The transmitted laser beam is passed through a beamexpanding telescope which is collinear with a large collecting-mirror receiving telescope. The use of the telescope is followed by a spectrometer and necessary filters to provide rejection against backscattered Rayleigh and Mie radiation. Reference [22] describes one of a number of Raman transmitter and receiver optical systems that have been built.

Equation (2) shows that the Raman signal intensity is proportional to the cross section, the atmospheric transmittance, and the laser energy per pulse. Since the cross section is inversely proportional to the fourth power of the wavelength, it is significantly increased by using a pump laser in the ultraviolet. The

1650 PROCEEDINGS OF THE IEEE, DECEMBER 1971

and (2), we obtain a condition for the minimum detectable con- R A M A N SCATTERING centration of the pollutant q given by SCHEME

RANSMITTED v where n is the integrated number of pulses. It is apparent from (28) that if the detector is photon-noise limited, the sensitivity is determined by the average transmitted power. In other words, if we compare two lasers with different repetition rates and pulse energies, but with the same average output powers, we find that the detection sensitivity is the same when integration times as well as other system parameters are equal. However, for darkcurrent or background-limited detection, where q is inversely proportional to &, the sensitivity is improved for larger pulse energies at the given average power level [cf (1 8)].

In order to evaluate the Raman scheme for pollution detection, we assume a l -J transmitted laser pulse in a 30-11s pulsewidth. At the present time this is available for a doubled Q- switched ruby laser. We also chose an electrical bandwidth of 17 MHz so that the gatewidth is equal to the laser pulsewidth. In ” this case, the depth resolution is 9m. Assuming a receiving mirror

Fig. 7. Schematic of the Raman scattering scheme showing the tram- area of 0.01 mz, a receiving optics efficiency of 0.1, and (s/ain equal to one, we find from (28) for Q branch detection that mitting laser, receiving optics, and spectrometer.

shortest useful ultraviolet wavelength is determined by the de- crease in the atmospheric transparency due to Rayleigh and Mie scattering and to oxygen absorption below 2500 A.

In the following discussion we consider a Q-switched ruby laser with output at the second harmonic at 3471 A as the pump laser. The ruby laser has a short pulse length and large energy per pulse. Another possible pump is the quadrupled Nd:glass laser at 2660 A. This laser has the advantage of the virtual “night” conditions existing at wavelengths shorter than 3000 A due to the ozone absorption screen. By proper design, however, natural background radiation problems can be avoided atthe doubled ruby wavelength. The pulsed NS laser at 3371 A and the second harmonic output of a visible dye laser are also possible pump sources, but they do not offer any power advantage over the ruby laser. As will be discussed, the tunable dye laser may offer some advantage due to possible resonant enhancement of the Raman cross section.

Equation (1) gives an expression for the Raman cross section. The direct Raman backscattering cross section, including all three branches for COS with a 3471-A pump, is calculated to be [IO1

COS is not usually considered an air pollutant, but its cross section is characteristic for other gaseous air pollutant molecules [lo], [23], [24]. Fig. 1 shows the Raman [ Q ( O ) line] shift for various pollutant molecules and the full spectral width of the Raman spectra for COZY O,, and Nz. If the detection is restricted to just the Q branch to reduce interference and background radiation problems, the cross section is somewhat smaller. For COS it is reduced by 10 percent.

Multiplication of the Raman intensity by the area of the receiving mirror A, and the efficiency of the receiving optics K yields the Raman power at the detector. This must be larger than the MDS power for the photomultiplier detector. At large electrical bandwidths the MDS for a photomultiplier tube is given by (17). Combining this expression with the previous cross section

R2 TIT2

q > - 7.6 X le9 (m-’).

The atmospheric transmittance in (29) is determined by the combined Rayleigh and Mie scattering. The total attenuation coefficient in the 3000 to 4OOO-A region is ax-5 km-l and 0.8 km-1 for visibilities of 1 km and 10 km, respectively [cf. Table I]. With Tl =T2 we can rewrite the restriction on the minimum detectable concentration q as

Fig. 8 shows the minimum detectable concentration of pollutants as a function of distance as determined by the previous equation. The detectability can be improved by increasing the pump pulse energy above 1 J, by increasing the receiving mirror area, by decreasing the bandwidth at the cost of depth resolution, and by integrating over a number of pulses.

As stated in Sectioq 111, for the Raman scheme the necessary rejection ratio against the Rayleigh line is of the order of l @ / ~ . Fig. 8 shows the necessary Rayleigh rejection .ratio for various 7. Since both the Raman and the Rayleigh scattering are inversely proportional to the fourth power of the wavelength, this restriction is independent of the excitation wavelength. Filtering methods include spectrometers, dielectric reflection filters, cutoff absorption filters, and possibly absorbing gases.

For photon-limited detection the natural sky background radiation power must be less than the signal power. Refemng to (22), this condition is

MDS > KTxAAQ,A,,,N(A) (31)

where the MDS is given by (1 7). From Fig. 6 the spectral radiance at 3700 A is No) =5 X lea We cm-1 pm-1 ST^. For an optical bandwidth of 1.2 A (10 cm-l), TX equal to one, and the same area and optical e5ciency as for the Raman example, (31) leads to a maximum receiver field of view given by

8, < 1.5 X les sr. (32)

For an optical receiving system matched into a spectrometer we have A,,,&= A,Q,, where a, and A. are the solid-angle field of

KILDAL AND BYER : LASER DPIZCIlON OF ATMOSPHERIC POLLUTANTS

Fig. 8. Minimum detectable pollutant concentration versus range for the Raman backscattering scheme. Also shown is the necessary rejection ratio against the Rayleigh backscattered intensity and the &ect of atmospheric attenuation.

view of the spectrometer and the area of the entrance slit. For an f-9 spectrometer, Q. is equal to 0.01 sr. With this value the restriction on the slit area is A,<0.15 cm2. When this condition is satisfied, background is not a serious noise source even at the doubled ruby wavelength. According to [19], the sky background is reduced by about seven orders of magnitude for the quadrupled Nd: glass laser or for nighttime operation and is completely negllgible for these cases.

It is possible to improve the detectability by increasing the receiving mirror area. With a 1-m2 mirror and a signal-to-noise ratio of 10, Fig. 9 shows that the detectability is about 10 ppm at a 100-m range with 10-m depth resolution. This is assuming opti- mistic parameters for the detection system. For detection beyond 1 km the scheme is severely limited due to both the small Raman signal and the pump depletion caused by atmospheric scattering. In the major cities of the world visibilities less than 10 km are common. The practical range of the Raman scheme is therefore limited to less than 1 km. The scheme is not sufficiently sensitive to measure general air-pollution levels, which typically are in the range of 0.01 to 0.1 ppm. It can be used, however, to monitor emissions from smokestacks or other sources where the pollutant concentrations are generally much higher.

Recently, resonance Raman scattering has been suggested as a means of improving the Raman backscattering efficiency. By choosing the pump frequency close to an allowed electronic transition of the molecule, an enhancement in the scattering cross section occurs due to a resonance effect in the polarizability ten- sor. The enhancement is approximately

1651

where w1 is the laser pump frequency, 00 the electronic transition frequency, and AU the linewidth of the electronic transition. In practice, resonance Raman scattering requires at least a dis- cretely tunable laser to match the electronic transition frequencies for different molecules. Except for a few molecules, the electronic transitions are in the ultraviolet where high-power tunable lasers are not yet available and the atmospheric transmission is limited by elastic scattering loss and absorption.

The resonance Raman process is very similar to laserexcited fluorescence from molecular electronic states., The fluorescence frequency shift for molecules is of the same order as the Raman shift so that the discrimination problems are similar. However, the backscattered fluorescence intensity may be larger by orders of magnitude than the resonance Raman intensity at the same detuning from the resonance absorption frequency. The relation between the absorption cross section and resonance Raman scattering cross section is given by

fibs AO duh

4~ wap( dil )- where AU is the linewidth and fl is radiative decay rate. As an example, we estimate that the absorption cross section for benzene is 7 X 106 times larger than the resonance Raman cross section for a linewidth, and a decay time of 60 cm-' and 6.2XlW' s, respectively.

Quenching reduces the backscattered fluorescence efficiency. At 1 atm, the quenching time can be shorter than a nanosecond for electronic transitions. For the case of rapid fluorescence quenching times the expression for the backscattered fluorescence intensity simplifies [cf. Appendix II], and the ratio of fluorescence to Raman intensity is

--- -

Ires(t) (flb8"/7r)FQ = (33) Ir€S""(t> (5)-

where F is the fraction of monitored fluorescence. Q is the Stern- Volmer [25] quenching factor given by

1 Q =

Tcol

T8*

1 + y ) zz-z (34)

T c o 1

where the gas kinetic coltision time T ~ I at atmospheric pressure is approximately 10-l0 s. The probability of quenching per collision is 1/Z. For oxygen and nitrogen quenching Z is typically between 1 and 10 [26]. For benzene with ~~*=6.2X10-' s, and assumingZ= 1 and F=0.1, the fluorescence-to-resonance Raman intensity ratio is Zre8(t)/Zr,.R"(t) = For quenched-resonance fluorescence the depth resolution is essentially the same as for the Raman scheme. For molecules with rapid spontaneous fluorescence decay times the fluorescence scheme offers increased sensitivity compared to resonance Raman scattering.

B. Resonance Backscattering For both the resonance backscattering and the resonance ab-

sorption scheme, radiation sources are needed at frequencies matching the absorption lines of the pollutants of interest. With the recent development of the dye laser and the optical parametric oscillator, tunable coherent radiation is now available in the near ultraviolet, visible, and infrared spectral regions. Before the resonance detection methods are considered in detail, we will give a brief description of these new tunable laser sources.

Reference [2] gives a review of dye lasers. A single dye laser is tunable over an 800-A bandwidth in the visibk. With different dyes, tunable output is available from 3600 A in the ultraviolet to about 1 p in the infrared [27], [28]. Laser pumped dye lasers operate with efficiencies from 1 perant up to 50 percent, and reproduce the peak power, energy per pulse, and repetition rate of the pumping laser. Peak powers of up to 100 kW with a IO-ns pulse length at repetition rates of up to 100 pps are possible [29]. For a flashlamp pumped dye laser, the efficiency is decreased and the pulse length is longer, but the energy available per pulse is significantly improved from the l-mJ level in the previous example to oyer 1 J [MI.

Tuning of dye lasers by a diffraction grating [31] results in a spectral-width of about 1 cm-l. A birehgent (Lyot) filter or etalon further reduces the bandwidth so that it is narrower than the absorption line of a gas. Walther and Hall [32] recently obtained a spectral width of about 0.03 cm-’ with a birefringent element, and similar results have been obtained with etalons [33].

Optical parametric oscillators provide tunable radiation in the infrared at the vibrational frequencies of the pollutant molecules. In addition they tune over most of the visible and may be doubled to provide tuning over the near ultraviolet spectrum. In a recent review article, Harris [3] describes the operation characteristics to date. The oscillator requires a laser pump source and, like the dye laser, it reproduces the power, pulse length, and repetition rate of the pumping laser with efficiencies varying between 1 and 50 percent. For example, a commercial oscillator tunes from 0.55 pm to 3.75 pm, with a bandwidth of 1 cm-’ in the visible and less than 0.25 cm-l in the infrared. The single pulse energy is 1 mJ, and the repetition rate is up to loo0 pps [34], [35]. With an intracavity etalon the oscillator linewidth has been reduced to less than 0.01 cm-l [36].

New nonlinear materials are under study for the purpose of extending the tunable region further out in the infrared [ 371. The spinflip Raman laser also provides tunable radiation in this region [38], [39].

One other source of tunable coherent radiation might be particularly useful for absorption measurements in the infrared. This is the diode laser which generates coherent light directly from the applied current. Individual diodes tune over a narrow spectral region of about 1 cm-’ and the tuning is achieved by changing the current [40]. For diodes constructed of mixed crystals, it is possible to achieve oscillation at a particular frequency by adjust- ing the material composition. As an example, GaJn1,As has been “tuned” between 0.85 and 3.5 pm [41] and Pbl,Sn;Te between 6.5 and 32 pm [42], [43]. The spectral bandwidths are less than a few kilohertz and individual vibrational-rotational lines are resolved [43]. Diode lasers require cryogenic cooling when operating in the infrared spectral region, and the output is at best only diffraction limited in the plane of the junction [44]. In the plane perpendicular to the junction the beam divergence can be more than ten times the diffraction limit. Therefore, over a long pathlength not all of the diode power may be useful.

In the discussion of the resonance backscattering from vibrational transitions we consider CO as an exwple. It is a well- known air pollutant and should provide results that are typical, since the vibrational absorption cross sections and the spontaneous lifetimes are about the same for most simple molecules. For resonance detection of metal-vapor pollutants we take sodium and mercury as examples. The cross sections and spontaneous lifetimes for these atoms are known and should also be representative for other atomic pollutants. We also include a brief discussion of resonance backscattering from electronic transitions of molecules with benzene, NOZ, and SO2 as examples.

I ) Molecular Pollutants-Infrared Excitation: The spontaneous transition probabilities for CO are [45]

~ 1 . 0 = 30.3 5-l .P

W:pI = 59.33 5-1 SP

W Z , ~ = 0.55 S-’. (35)

Using (93) with Z ~ 0 . 1 and Av=O.15 cm-l, we obtain the absorption constants for the fundamental and the first overtone vibritional-rotational transition. We have

ul,ONtot = 5 X 10% (m-l)

and

at.dvtot = 231 (m-3 (36)

where 71 is the relative density. The following discussion considers only absorption at the fundamental transition. We have, however, explicitly stated the absorption cross section for the first overtone transition, because it may be useful in situations where pump depletion is important.

Assuming that the pollution begins at %in, and is of uniform concentration, (4) yields

?WSl,O(t) kin

~ ~ . ~ * q t ) = 12.1 ‘x 103 (MKS) (37)

for the total backscattered fluorescence where W is the transmitted pulse energy. If the atmospheric transmittance is different from unity, this will be included in the integral SI,&). In obtain- ing (37), we have implicitly assumed that the tunable laser used for the excitation of the vibrational transitions has a spectral linewidth much narrower than the vibrational-rotational linewidth and that the laser frequency is tuned to the peak of the absorption line. Otherwise, we must use an effective absorption cross section. To reduce background or interference from other molecules only a fraction of the total fluorescence at frequencies shifted from the exciting frequency is monitored. Detection of only a single vibrational-rotational line reduces the backscattered fluorescence intensity by about 20.

Multiplying the intensity at the receiving mirror by the area of the mirror A, and the efficiency of the receiving optics K yields the resonant backscattered power at the detector. For the photovoltaic InSb detector we find from (21) and (37) that the minimum detectable concentration is

d o - 1)

D*& 12.1 x 103S1,0(t)WAmK

From Table I1 d/ADB/D* = 3 X 10-0 W. Referring to the previous discussion of tunable sources, we assume a single pulse energy of 0.001 J in a 100-ns pulsewidth. With a receiving mirror area of 0.01 m2, a receiving optics efficiency K=0.1, and a signal-to- noise ratio of one, (38) reduces to

The resonance backscattering intensity as a function of time is proportional to the integral Sn.O(t). Fig. 9 shows the integral S,,,dt) for various relaxation times, absorption constants, and pump pulse lengths, assuming a uniform pollution concentration beginning at Rmin equal to 10 m. The rapid increase in S,,O after

KlLDAL AND BYER: LASER DETECllON OF ATMOSPHERIC POLLWTANl3 1653

i b )

'n

4 8 I2

0.13pes

I3puc L3pnc OJ3psw 0.013pyc

O L Q 6 7 W

the exciting radiation and assuming that the thickness of the polluted layer L is smaller than the distance out to the layer, we have for Sn,o that

where t>2R,,/c. For no pump depletion and 2L/m<<l this reduces to

L S*,O(O = exp [ - +(t - F)]. (42)

For large pump depletion where Lan,oNt.,&>l, then

1 S n , o ( t ) exp[ - ' ( t - -)I. (43)

%,O In this case the concentration dependence cancels since, according to (41,

Iru(O a cn,&totSn,o(t) . (44) I 2 3 4 Therefore, the backscattered intensity is independent of concen-

2 R m i n

Elmincn,&tot 7 C ( 5 1 0.4

02

t ( p r e s ) tration for pump-depletion lengths shorter than the layer thick-

absorption constants uNbt, and pulse lengths ta assuming a uniform The information most readily available from resonance back- pollutant concentration beginning at R,,,i,= 10 m. scattering from vibrational transitions is the distance to the

beginning of the polluted region. Since there is no straightforward

Fig. 9. The integral S,,o(t) evaluated for various relaxation times 7 , ness.

a time delay equivalent to Rm, gives the distance to the polluted region even though the long upper-state decay time limits depth resolution to ARre"=m/2. For typical vibration decay times of T = 10-4 to 1 0 - 6 s, the depth resolution is lo2 to lo4 m.

Due to pump depletion most of the resonance backscattering is from molecules that lie at a distance less than the pump depletion length into the polluted region. From (9), the pump depletion length is

1

a n , d V t o t Rdepl = Rmax - R m i n N- (40)

and the corresponding depletion time[cf. Fig. 1O]is tdepl= 2/c%,l

For CO the depletion length as a function of concentration is

Rdepl = 2 X lO-'/v (m), (0 - 1) excitation

and

Rdepl = 4.4 X (m), (0 - 2) excitation.

As an example, for v equal to 1 ppm the depletion length is 200 m for the fundamental excitation and 44 km for the first overtone excitation. The problem of pump depletion can be avoided for large pollutant concentrations by working at overtone excita- tions. The cost is the loss of sensitivity, but at the higher pollutant concentration the relative sensitivity is not reduced significantly.

The integral Sn,o(t) simplifies when the pollutant is concen- trated within a s m a l l region, for example, in a thin layer or in the plume from a smokestack. Neglecting the k i t e pulse length of

relationship between the intensity of the backscattered radiation and the concentration of the pollutant, quantitative measurements are complicated. The uncertainty in the nonradiative vibrational relaxation time r further complicates absolute measurements. Since it is different for different molecules, the pulse shapes of the backscattered radiation are not the same for two different pollutants, even if the absorption constants and the pollutant distributions are the same.

From Fig. 9 and (42) we have that the approximate maximum values for Sn.o(t) are about 0.5 and L/R,in for the cases of uniform pollutant concentration in a thick and a thin layer. Assuming integration over lo00 pulses, (39) reduces to

~ ( 0 - 1) > 1.6 X 10-*Rmin (45)

for a thick layer and

v(o - 1) > 7.9 x 10-9~m~,2 /~ (46)

for a thin layer where Rmin and L is in meters. Fig. 10 shows the minimum detectable concentration 40 - 1) for resonance backscattering for the two cases. As in the Raman scheme, it is possible to improve the sensitivity by increasing the receiving mirror area from the assumed 0.01 mf to 1 mf and enhance the sensitivity by a factor of one hundred. The atmospheric attenuation limits the range to a few kilometers, since within the infrared atmospheric windows the attenuation constant is approximately 0.5 km-1.

Referring to Table II, we see that the background radiation power can theoretically be more than four orders of magnitude larger than the minimum detectable signal power for the dark-

1654 PROCEEDINGS OF THE IEE, DECFMBER 1971

Id W - t

- Id- a - a * 0 Id- 5

8 10 K I- - 2 2

z 0 0 W

I-

A w K

L IO a

-

I -

Id -

RANGE (meters)

Fig. 10. Minimum detectable pollutant concentration 40-1) versus range for resonance backscattering from CO.

current-limited InSb detector. In practice the background radiation fluctuates, and, therefore, we conservatively require the background power to be less than the minimum detectable signal power [cf. (31)]. With a spectral bandwidth of 0.25 pm, TA equal to one, and the MDS given in Table 11, this leads to a limitation on the receiving mirror field of view given by

!A < 1.2 X sr (47)

for a spectral sky radiance at 5 pm of NO) = 10-4 [ W. cm- pm-* ST']. As discussed for the Raman scheme, this restriction on can be satisfied so that the natural background radiation is not a problem.

Elastic Mie scattering is also a source of background radiation. According to (26), the necessary rejection ratio against the Mie line is of the order of l(r3/r] for the Mie intensity to be less than the signal intensity. For a pollutant concentration of 0.1 ppm, the necessary rejection is therefore 104, which is possible to achieve.

Saturation of the excited vibrational transition is possible at high transmitted pump energies. For tO/r<<l we have from (87) in Appendix I that

Ito << 100 [ J m-2]. (48)

With a pulse energy of 0.001 J and a beam cross section of 0.01 m2 this restriction is satisfied.

2) Metal Vapor Pollutants: The pollutant metal vapors include arsenic, cadmium, zinc, sodium, and mercury. For our discussion we consider sodium and mercury vapor as examples. The detection sensitivity and depth resolution is good for metal vapors because of large absorption cross sections and short fluorescence decay times. We assume that the available pulse energy is 1 mJ at the sodium frequency (58% A) and 0.1 mJ at the mercury frequency (2537 A).

For reference Table 111 lists the relevant parameters for sodium and mercury atoms, including the collision time, the probability of quenching per collision 1/Z, and the quenching factor Q [a].

At atmospheric pressure, collision broadening determines the linewidth Av of the transitions. As used previously, r] in Table I11 is the relative concentration of the pollutant.

Pump depletion is a potential problem because of large atomic absorption cross sections. As a function of detuning for a Lorent- zian line with a linewidth Au, the effective cross section is given by

where fibs is the absorption cross section at the center frequency w0. By detuning, the pump-depletion length increases so that more of the polluted region is probed until it reaches a limit determined by the decreased backscattered intensity. Referring to (a), the depletion length as a function of pollutant concentration for sodium and mercury is

9.1 x 10-10

w ( 4 Na R d e p l = ( 4

The range of metal vapor concentration presently measured in the atmosphere is between 10 and 0.01 parts per billion (ppb). The expected pump-depletion length therefore ranges between 0.1 and lo2 m for sodium and 1 to 1@ m for mercury when the exciting frequency is at line center.

From Table I11 we see that quenched fluorescence decay time T = T ~ ~ Z is approximately s. With this short fluorescence decay time we can use the simplified expression for the fluorescence backscattering intensity given by (101) of Appendix I1

where Q is the quenching factor defined in (34) and pump depletion is neglected.

Applying the condition that the backscattered power is greater than the minimum detectable power for photon-limited detection, the minimum detectable concentration for sodium is

2qB(S/N)min RZ ? > (52) SDn 1.4 X lO'd/(w)WA,K

with the quantities expressed in MKS units. This expression compares directly with that derived for the Raman case given by (28), except for the increased sensitivity due to the much larger absorption cross section. Using the parameters

KILDAL AND BYER: LASER DETECTION OF ATMOSPHERIC POLLUTANTS 1655

K = 0.1 For mercury the Mie scattering limikd detection sensitivity is

A , = 0.01 m2

SD = 0.06A * IT-' ( S / N ) m i n = 1

W = 1W3 J for Na B = 10 MHz

(52) reduces to

R2 tl > 3.8 x 10-19 - (MKS). (53)

nY (4

The combination of increased cross section, fast radiative lifetime, and use of a photomultiplier detector greatly increases the sensitivity for metal vapor detection. This result is valid if the detector is not background limited. However, the background radiation caused by elastic Mie and Rayleigh backscattering is not negligible for resonance backscattering from atoms. Except under extremely clear daytime conditions, the largest contribution is from Mie backscattering. From (24) and (51) the approximate ratio between the Mie and the resonance backscattering intensity is

I ?die ax Y ie -- - . I'"B ( T y W ) N t o t &

(54)

For a visibility of 5 km the Mie scattering coefficient at the sodium wavelength is 0.7 km-l. Using this value and the parameters for sodium we find

A major problem associated with the resonance backscattering from atoms is the lack of effective discrimination against the elastic backscattering. Equation (55) determines the detection limit set by Mie backscattering. For tuning on and off the resonance line and assuming a minimum detectable change in intensity of 10 percent, we have that

I Yie

pa - < 10

if no special techniques are used to discriminate against the Mie intensity. This condition with (55) gives

Comparing this result with (53), we see that the elastic backscattering severely limits the sensitivity. The detectivity is independent of distance since both the fluorescence and elastic backscattering intensities have the same l /Rz dependence.

To improve the detection sensitivity it is necessary to discriminate against elastic backscattering. For example, off line center excitation of the resonance transition allows spectral filtering. For an increased signal-toelastic backscattering intensity ratio, the detection sensitivity improves even for a reduction in total signal intensity. According to (53) and (56), the signal intensity may be reduced several orders of magnitude before the detector noise limit is reached. Therefore, filtering techniques should improve the detection sensitivity given in (56).

(57)

for a visibility of 5 km with the corresponding Mie scattering coefficient of 1.7 km+. The detection sensitivity for mercury is lower than for sodium due to the smaller mercury cross section and longer spontaneous lifetime.

The depletion length determines the depth over which the pollutant can be probed. In (51) we neglected pump depletion. If we include the exponential depletion term [cf. (lol)], we can solve for the maximum probing depth R-&h consistent with the detection sensitivity given by (56) and (57). At a 100-m range we find that

R - Rmin = 390 m

for sodium and

R - Rmin = 11 m

for mercury. Due to the short quenched fluorescence decay time the trans-

mitted laser pulse length and gatewidth determine the depth resolution. Therefore, the depth resolution is the same as for the Raman scheme.

When the laser pulse length is longer than the fluorescence decay time 7, saturation is negligible when

WBt7 << 1 (58)

where +t is the stimulated transition rate [cf. (W)]. For sodium this condition requires

I << 4 kW/cm2

and for mercury

I << 400 kW/cm2.

Therefore, for reasonable transmitted pump energies and beam areas, saturation is negligible.

The discussion of natural background radiation and atmospheric attenuation for the Raman scheme in Section IV-A applies for the electronic resonance backscattering as well since the transition frequencies are in the near ultraviolet. We have already concluded that the natural background radiation can be neglected for a narrow field of view. At wavelengths shorter than 3000 A, natural background radiation is not a problem because of the ozone absorption in the upper atmosphere.

As in the Raman scheme, the atmospheric attenuation limits the range of detection. Except for possible pump depletion due to the resonance absorption, the maximum range is a few kilometers, depending on the visibility at the time of the measurements.

3) Molecular Poilutants-Electronic Excitation: As a final example we consider resonance backscattering from benzene. This example illustrates the potential sensitivity associated with resonance backscattering from excited molecular electronic states. Most molecules absorb in the middle ultraviolet, and the absorption is generally over a wide spectral range. As a consequence, there is significant overlap between individual absorption bands. Some of the molecules show distinct absorption spectra consisting of individual lines, while others have diffuse absorption bands with very little structure. The fluorescence is normally at a different frequency than the absorbed radiation. The frequency is shifted toward the visible region, and the difference is due to one or more vibrational frequencies. This frequency shift allows good

1656 PROCEEDINGS OF THE IEE, DECPMBW 1971

TABLE IV ELECIRONIC EXCITATION PARAMFIFRS FOR

SOME MOLECULAR POLLUTANIS

RO2 902 I '6'6

WAVENUMBER (~6')

Fig. 11. Absorption spectrum of benzene vapor [a].

WAVENUMBER ( c d ' )

Fig. 12. High-presure fluorescence spectrum of benzene excited by a mercury lamp [Xi].

rejection of the Rayleigh and Mie backscattering. Since the fluorescence consists of several lines, the interference problem may be reduced by correlation spectroscopy techniques [47].

The absorption spectrum for benzene is shown in Fig. 11. Benzene absorbs between 2300 and 2700 A, and the various lines reflect the vibrational structure. Individual rotational lines are not resolved, and at atmospheric pressure they overlap so that the linewidth factor used in the absorption constant is the total linewidth of an electronic vibrational transition including all rotational fine structure. This linewidth is approximately 60 cm-1.

Table IV lists the absorption cross sections and quenching factors for NOs, SOr, and benzene. For benzene we have assumed 1 / Z is approximately 1 and ~~1 is approximately s. For a review of fluorescence quenching in small molecules see [ 5 5 ] .

Fig. 12 shows the fluorescence spectrum of benzene which consists of several lines over a 30OOcm-1 spectral range. Equation (51) gives the backscattered fluorescence intensity for benzene. For the values in Table IV we have

where F is the monitored fraction of the total fluorescence spectrum. By combining (59) and (17) the minimum detectable pollutant concentration is

R2

for a laser pulse energy of lo-' J. Here we have assumed the same parameters for the receiving system as in our previous discussion for the metal vapor example. For F=0.1 and integration over loo0 pulses the minimum detectable concentration for benzene is

similarly, we obtain for SOr and NOs that

and

NOu 7 > 2.3 X 10-"R2. (6.3)

Fig. 13 shows these results as a function of range including atmospheric transmission loss.

Background radiation is not a problem for fluorescence wavelengths less than 3000 A, as is the case for benzene. For SO* and NO2, however, with their wide fluorescence band at wavelengths longer than 3000 A, these results are valid for nighttime conditions. Under daytime conditions the sensitivity is reduced approximately 200 for NO2 for a bandwidth such that F=0.1 and a telescope acceptance angle of %,= lW5 ST.

In practice interference from other molecules may reduce the detection sensitivity. As discussed earlier, the range is limited to a few kilometers due to the atmospheric attenuation.

The depletion length and sensitivity determine the maximum probing depth. From (40) and (61) we have for benzene that

Rdepl = 3.6 X 108/R2 (m). (64

KILDAL AND BYER: LASER DETECTlON OF ATMOSPHERIC POLLUTANTS 1657

TABLE V MINIMUM DETECTABLE CONCENTRATION FOR AN

ABSORPTION LENGTH OF 100 M ~ R S

TRANSMITTER

TRANSMITTER AND

RECEIVER

Fig. 14. Schematic of two possible long pathlength absorption schemes for monitoring pollutants.

Cornpared to the metal vapor case, the probing depth for benzene is considerably longer.

Since the backscatter fluorescence is at a shifted frequency relative to the pump, the discrimination problem against the Mie backscattering is not as severe as for atomic fluoresoace. Fig. 13 shows the necessary rejection ratio for NOz and SOo. This can be obtained ,yith standard filtering techniques. Resonance backscattering from molecular electronic transitions has sufficient sensitivity to be useful and has good depth resolution due to rapid quenching times. However, interference from other molecules due to the wide fluorescence bands and background radiation are potential problems that require further investigation.

C. Resonance Absorption Detection of air pollutants by absorption has the advantage of

both high sensitivity and the requirement for minimal beam powers. However, it is necesary to use a remote detector or retro- reflector. The absorption scheme measures the total integrated pollution density in the beampath.

Fig. 14 illustrates two possible geometries for the long pathlength absorption measurement technique. The advantage of a laser compared to conventional light sources is its high intensity per bandwidth and small beam divergence [19]. For example, for a transmitting mirror area of 10 cm2 at a 1-pm wavelength the beam area only doubles over a 1-km pathlength.

When the detection limit is taken as a necessary 5-percent change in intensity for frequency switching on and off the absorption line, (13) gives the minimum detectable integrated density. For a uniform pollutant concentration starting at Rmh, (13) becomes

> 1.9 x 10-27 u"b"(W)(B - Emin)

rl= (m3). (65)

This expression is generally applicable if the transmitted laser spectral-width is less than the absorption linewidth of the transition. Otherwise, it is necessary to use an effective absorption cross section.

The condition used to estimate the detection sensitivity is conservative. When the integration time is sufficiently long to average out laser or atmospheric fluctuations, the ultimate detection limit is set by the detector noise. From (109) of Appendix III the minimum detectable change in relative signal power is

where SDP, = i,. For a signal-tcmoise ratio of 10, a 1-MHz bandwidth, integration over lo00 pulses, and other detector parameters as in Table II, we find

or a photomultiplier, and

for an InSb detector. In these expressions the first term results from the signal shot noise and the second term from dark-current shot noise. To detect a change of 0.1 percent in transmission requires a signal power of only W. This is a theoretical limit disregarding atmospheric and laser fluctuations. Independent of fluctuations, a minimum limit is also set on APmin/P, by the dy- namic range of the detector. Commercial spectrometers measure approximately 0.1-percent change in transmission. For a 0.1- percent detection limit the detection sensitivity given by (65) is improved by a factor of 50.

Table V gives the minimum detectable concentration for various pollutants for a 100-m absorption path. These results neglect interference problems. For example, NO2 and SO2 have wide electronic absorption bands with little characteristic structure. In this case, absorption measurements at several characteristic wavelengths can help to reduce the interference.

Fig. 15 shows the minimum detectable concentration versus range for CO absorption at the fundamental vibrational transition. The figure applies to other pollutants if the relative concentration scale is divided by the cross section ratio with respect to the CO absorption cross section. The absorption technique allows sensitive measurements of pollutant levels over short and long pathlengths. Also, unlike the Raman and rescpance backscattering schemes, the absorption scheme sensitivity improves with pathlength.

Referring to Table V, we see that for a pathlength of 1 m at 0.1-percent detection limit the detection sensitivity for CO is 0.2 ppm. A folded beampath improves the sensitivity without increasing the size of the detection device. Such an instrument is similar to the more familiar nondispersive infrared analyzers (NDIR) [57], except that the sensitivity is increased to the level

1658

' 0 2 r -

I I I I

10 100 loo0 1oooo RANGE (malerrl

Fig. 15. Minimum detectable pollutant concentration do-1) versus range for CO resonance absorption.

necessary to detect dispersed pollutants. A possible use for a portable path absorption spectrometer could be to monitor street level, subway, or even closed-work-area pollutant levels. The folded path cell makes use of the collimation properties of the laser beam properly matched into a spherical cavity formed by two curved high reflecting mirrors, such as described by Herriott and Schulte [ 5 8 ] in their laser delay line work. By using proper cavities of only moderate volume, it is possible to achieve over 100 passes between the mirrors. Thus, unlike the NDIR or other single-pass absorption instruments, the sensitivity is greatly increased at little cost in overall absorption volume. If gold coated mirrors are used with a reflectivity of 98.3 percent in the infrared, the beam intensity is reduced by a factor of 0.2 after one hundred bounces. Thus a low-power laser is sufficient to provide enough intensity for a good signal-to-noise ratio over an effective pathlength of between 10 to 30 m.

The useful range for the absorption scheme is much longer than for the backscattering scheme, and for a given range the required transmitted power is much less. For optimum pollutant absorption the necessary transmitted laser power P(0) to give a 30-percent accuracy in the integrated pollutant concentration for a darkcurrent limited detector is [cf. (113)]

where is the atmospheric absorption constant and K is the receiving optics efficiency. The amount of transmitted power to maintain the same meaurement accuracy increases for non- optimum values of integrated absorption. From (69) the maximum range for a given transmitted power is

For an MDS of 3X lW9 W, an optics efficiency of 0.1, and a transmitted peak pulse power of 1 kW, the maximum range is

R,, 5 44 km (7 1)

for equal to 0.5 k m l . In most situations near urban areas,

PROCEEDINGS OF THE IEEE, DECEMBER 1971

pathlengths of the order of a few kilometers are of interest. From (69) the necessary power for a 1-km pathlength is only

P(O) - 4.5 x 10-7 w. (72)

This result is important when consideration is given to laser sources that are alternatives to the relatively expensive dye and parametric oscillator systems. For example, the tunable diode lasers [41], [42], [43] presently available can provide nearly diffraction-limited light output over most of the infrared spectral region. The diode laser is small and relatively inexpensive. Present day infrared diodes, however, require cryogenic cooling.

The sensitivity of the absorption scheme improves with increased pathlength. The necessary laser power for sensitive detection is much smaller than for the Raman and the resonance backscattering schemes, and in addition there is no need for a large receiving mirror. So in situations where depth resolution is not a necessity, the absorption scheme offers the most direct and least expensive solution to the pollution detection problem.

V. CONCLUSION

The Raman and the resonance backscattering schemes require sophisticated optical systems. To obtain adequate sensitivity it is necessary to use high-power lasers, large receiving mirrors, and expensive signal processing instrumentation. Resonance absorption has more modest requirements, particularly if semiconductor lasers are used as sources in the infrared region. The resonance schemes require tunable lasers, and moreover, the linewidth should be narrower than the width of the resonance transitions. With the addition of a remote reflecting mirror, a system for resonance backscattering is also useful for resonance absorption. The two detection schemes can provide complementary information. For example, resonance backscattering may locate the pollution, while the more sensitive absorption scheme may determine the integrated concentration.

All molecules in the atmosphere contribute to the Raman scattering. To select the Raman signal for the pollutant of interest, good filtering techniques are required. Interference problems may arise for lines close to the Nz or 02 Raman lines because of the high Raman intensity for these molecules. In order to detect dispersed pollutants at a 1 ppm level, the Nz and O2 intensities must be rejected by at least a lo6 ratio. Elastic Rayleigh and Mie backscattering present a background problem since the recessary rejection ratio is larger than 109 at the 1 ppm pollutant concentration level.

We have considered resonance backscattering and absorption for both vibrational and electronic transitions. For vibrational transitions, it is necessary to work within one of the infrared atmospheric windows. Interference can be avoided by exciting vibrational rotational lines not overlapped by other pollutants. For electronic transitions, atmospheric transmission losses limit the resonance backscattering and absorption techniques to atoms and molecules with absorption at wavelengths longer than 2500 A. The potential sensitivity is high for the electronic fluorescence scheme, but interference is a problem since the electronic absorption and fluorescence bands of the molecules are wide and over- lapping. The total bandwidth may be more than 5 0 0 0 cm-', compared to a total width of only a few hundred wavenumbers for Raman and vibrational transitions. For molecules the fluorescence is frequency shifted from the absorption frequency, so discrimination against Mie backscattering is possible. For backscattering from atoms, however, the Mie backscattering limits the detection sensitivity.

Natural background radiation may also limit the sensitivity. Small acceptance angles and optical bandwidths reduce the back-

KILDAL AND BYER: LASER DFTECnON OF AfMOSPHERIC POLLUTANIX 1659

ground radiation, so it is not a problem except for the resonance backscattering from molecular electronic transitions at wavelengths longer than 3000 A.

Atmospheric transmittance limits the maximum range of the Raman and resonance backscattering schemes to a few kilometers. To achieve this range for metal vapors it may be necessary to tune off line center because of pump depletion due to the large absorption cross sections. The transmission scheme has the greatest range of up to 50 k m .

Electronic transitions usually have fast fluorescence decay times. Therefore, the depth resolution depends only on gatewidth and laser pulse length, as for the Raman scheme. There is a trade- off between depth resolution and sensitivity since the large electronic bandwidths necessary for good depth resolution also increase the detector noise. Backscattering from vibrational transitions provides only partial depth resolution.

The interpretation of measured data is fairly straightforward for the Raman and transmission schemes. For the resonance backscattering from vibrational transitions a complicated integral relates the pollutant concentration to the signal intensity. The information most readily available is the location of the polluted region. For electronic fluorescence the integral simplifies, due to the short fluorescence decay times, to the same form as the Raman expression.

The detection sensitivities given below are based on realistic values for system parameters. For the Raman scheme we assume a laser pulse of 1 J at 3471 A and the detection of only one pulse. The minimum detectable concentration is

where R is in meters, and the maximum useful range is limited to about 1 km due to atmospheric transmittance loss TIT2. Reso- nance backscattering from CO in the infrared gives

vco(0 - 1) > 1.6 X 10-'Rmi, (45)

for a thick layer and

vco(o - 1) > 7.9 x 1 0 - 9 - RminZ

L

for a thin layer. For this case we assume that the tunable laser has a pulse energy of 0.001 J and that lo00 pulses are integrated. Also, the linewidth of the tunable laser is assumed to be less than the absorption linewidth of the pollutant. For resonance backscattering from atomic pollutants the minimum detectable concentration is limited by Mie scattering to

6 X 7Na > (56)

Y(4 and

(57)

which is independent of distance out to a range set by the detector sensitivity. Resonance backscattering from molecular electronic levels gives

grated pulses. The detectable concentrations for SOz and NOt assume nighttime conditions. For daytime operation the background reduces the sensitivity by approximately 200 for NOz.

The minimum detectable concentration for the absorption scheme is (cf. Table V)

1 0 - 5 mo(0 - 1) >

R - Rmin

for CO molecules in the infrared with an assumed intensity change of greater than 5 percent. For atomic pollutants the sensitivity is k f . Table V)

5 x 1 0 - 1 1

R - Rmin I]N8 >

In this case pump depletion is an important consideration. For molecular electronic transitions the sensitivity is (cf. Table V)

7 x 10-5 '?NO: >

R - Rmin

The Raman scheme and the resonance backscattering scheme for infrared molecular transitions are not sensitive enough to detect dispersed pollutants and simultaneously provide depth resolution. However, they can be used to detect highly concen- trated pollutant sources such as smokestacks. The resonance backscattering scheme for the detection of atomic pollutants and for scattering from molecular electronic transitions has better sensitivity, but it is still marginal for the detection of dispersed pollutants.

In evaluating the usefulness of these schemes it is important to consider complexity and cost. Both the Raman and resonance backscattering schemes require complex optical and data reduction systems at a relatively high cost. The atmospheric absorption scheme is less complex and costly and requires minimal laser power. It lacks depth resolution, but it is the only scheme sensitive enough to detect dispersed pollutants.

APPENDIX I RESONANCE BACKSCATTERING I m s r r y FROM

VIBRATIONAL TRANSTIONS In this Appendix we derive an expression for the spontaneous

backscattered radiation intensity from vibrational transitions. Be- cause of the short rotational relaxation time, we will consider each vibrational level to be in rotational equilibrium characterized by an equilibrium temperature T. With N, representing the total density of molecules in the Nth vibrational state, the Boltzmann distribution gives for the density of molecules N,,, J in the different rotational states that

where Be is the rotational constant of the molecule. When AB,/kT<<l (73) reduces to

' I C ~ H ~ > 8 X 10-13R2 vso, > 4 X 10-"R2 NO* > 2.3 X 10-11R2

(61) N,J = N , - (2J + 1) exp [ - - J(J + l)] (74) AB, fiBe

(62) kT kT (63) which we write as

where we have assumed a laser energy of lo-' J and lo00 inte- N,,J = XJNn (75)


a typical pollutant molecule is [ s e e (78)] we proceed to calculate the number of molecules excited by one pump pulse. When the excitation corresponds to an R transition, the rate of excitation of the nth vibrational state is

rnL

Wl.0 - 0.11.

We therefore must have

8P N n - W n . n - l N n - - (77)

7

where T is the nonradiative vibrational relaxation time and wiTn-l is the spontaneous transition probability between the vibrational states n and n- 1. The stimulated transition probability is

It sp Wn,0 = Wn.0 ___ I (78)

4r2hcAv

where Z and are the intensity and the wavelength of the exciting radiation, and Av is the linewidth of a single vibrational-rotational line.

We solve (77) with the initial conditions

No = Ntot and N, = 0 a t t = 0

in order to avoid saturation. In the resonance backscattering scheme a short pump pulse is

preferred to achieve depth resolution. In the resonance absorption scheme which lacks depth resolution, there is no restriction on the pump pulse length. It is still important, however, to avoid saturation of the excited vibrational level since this would complicate quantitative measurements. For a CW pump laser the number of excited molecules is

c&tot Nn = -

cz

and saturation can therefore be neglected when

- << 1. c 1

C2

where N,, is the total density of the pollutant. Assuming a short pump pulse of length t o , we make the approximation that This leads to the restriction

No(t) + N n ( t ) N t o t (79) E t J SP 1 Wn,O ~ XJ-I << W n , n - I + - . (90)

7 for t < to. Using (75) and (79) we rewrite (77) as

dN. s t J J

2 J - 1

When we neglect the spontaneous decay term, assume XJ-14.1 , and make use of (86), the restriction on the laser pump intensity

IT << 200 [J m-2]. (91)

-- - Wn,O- X J - 1 N t o t - { w::O [ 2J-1 X J - I dt 2 J - 1 for excitation of the first vibrational level is

+--J ] ap XJ + Wn.n-1 + 2 J + 1 Emation (88) is strictly valid only for excitation of the first vibra-

Equation (80) is of the form . .

tional level since, otherwise, the population density of the inter- coupled levels would have to be considered.

d N n For resonance detection schemes, pump depletion may be -= C l N t o t - c z N n

dt (81) important. With no saturation the attenuation of the pump laser per unit length due to the vibrational excitation of the pollutant is

where c1 and c2 are constants for an assumed square pump pulse. With the initial conditions given above, the solution is

dI st J

dR _ - W - 1 _ - - - ftWn.OWn.0 ~ X J - I N t o t . (92)

~ , ( t ) = - [I - exp ( - c z t ) ] . tot

(82) Substituting for the absorption cross section

For a short laser pulse where st L o 3 J un.0 = fiWn,OWn,O- ___ X J - 1 (93)

C2tO << 1 (83) 4r2hCAv 2J - 1

is satisfied, the total number of excited molecules is (93) yields

L t J Wn,O ___ X J - l N t o t t O .

2J - 1 (84) Pump depletion is important when the integral in (94) is larger

Equation (83) implies that Nn(to)<<Ntot, which means that the We next calculate the backscattered spontaneous radiation, laser pulse does not saturate the transition. For X J d . 1 and including pump depletion and assuming no saturation. The radia- neglecting the spontaneous decay term in (80), (83) gives tion intensity at the detector at a time t after the pump pulse is

than unity.

KILDAL AND BYER: LASER DETECllON OF ATMOSPHERIC POLLUTANTS 1661

APPENDIX 11

RESONANCE BACKSCATTERING INIwm FROH

ELECTRONIC TRANSITIONS

. N n ( - $ ' ') AdR (95) At atmospheric pressure the fluorescence decay time of excited electronic transitions can be as smaU as 1Wlo s due to rapid

where Rmi, is the distance to the start of the polluted area, Tn,-l is quenching. The expression for the resonance backscattering the atmospheric transmittance over the distance R at the fre- intensity [cf. (4)] simplifies for rapid fluorescence decay times. quency w ~ , ~ - ~ , and A is the cross section of the pump beam. From Equation (5) uses the approximation ( t 0 / 7 ) < 1 [cf. (83)] which (77) the number density of excited pollutant molecules is is valid for vibrational transitions. Without making the approxi-

2R I t - -

Nn[t-(2R/c), R ] is the density of the pollutant at R contributing mation the integral S n , 0 is to the backscattered intensity at the detector at time t . The first expression in (%) is the density at R during the passage of the f k o ( t ) pump pulse and the second is the density after the pump pulse has passed. Combining (84) with (77), (93), and (94), Nn(to, R ) is = given by t o J e t ' ' c(t-t0)/2 g(1- exp [ - +(t - f>l>

a n , o N t o t .exp [ - u n , o N t o t ( R - R m i n ) ] Nn(t0, R) = ~ T , , d d o

frwn ,O c( t -b) /2 dR + [ 1 - (- 3 1 SRmh

.exp { - JR:h g n , o N t o t ( E ) W } . (97) .exp[ - +(t - to - -

C

where T,,,,, is the atmospheric transmittance. Finally, substitution of (%) into (95) yields the following de-

sired result : - a n , o N t o t ( R - Rmin)]} . (99)

For electronic transitions where f0>>7 and the depletion length is larger than the laser pulse length

where W = Z d t o is the total energy per pump pulse.

For R>>cto/2, the expression for the backscattered intensity is of the same form as for the Raman scheme. Combining (100) with (4) we obtain

I"(t) % ~ - - u n , O N t o t (J2 ?- c 4rR2 w1 rsP 2

-w

(98) aexp [ - a n , o N t o t ( R - R m i n ) ] (101)

where w1 is the absorption frequency and w 2 the fluorescence frequency.


APPENDIX III DEIFCTOR NOISE CALCULATIONS FOR THE

AFSORPTION S a m m

From (12) arid (16)

where il and il are the average signal currents at the detector for the laser tuned on and off the absorption line. The integral a is the integrated pollutant concentration so that

il

22 a = I n T . (103)

The ratio Aa/a is the relative uncertainty in the measured pollutant concentration. For Aa we have that

where Ail and Ais are the uncertainty in signal currents.

We have from (15) that the shot noise is The shot noise in the detector sets a theoretical detection limit.

B n

= 2q - (il i d ) (105)

and B n

= 2q - ( i 2 + id) ( 106)

neglecting background generated noise. Combining the previous equations, we find

For smaU absorption a&, (107) reduces to

The right-hand side of (108) is recognized as the signal-to-noise expression (S/N)i,+ for the difference current. From (108) the minimum detectable change in signal current at a given signal-to- noise ratio is

For example, for a signal-to-noise ratio of 10 for the difference current, we see from (108) that the integrated pollutant concentration is measured with an accuracy of 33 percent.

For sufiicient pollutant absorption there is a optimum value for the absorption integral a which maximizes the CY/ACY ratio. We determine the optimum value for CY by differentiating (107). For the cases of photon-limited and dark-current limited detection we obtain respectively = 2.22 and h, = 1 . 1 1. Substitution of these

values into (107) gives for photon-limited detection il>>id that

n

Requiring that (a/Aa)2mar>lO we rewrite (110) and (111) to obtain a restriction for the necessary power P1=il/SD outside the absorption-line incident on the detector. We have for a photon- limited detector that

2qB nSD

P i > 21 - = 21 MDS

and for a darkcurrent limited detector that

P1 > 9 . 1 d A ~ : / D * = 9.1 MDS (113)

where the detector parameters have been defined in Section 111.

ACKNOWLDGMENT The authors wish to thank S . E. Harris and H. Inaba for their

participation in useful discussions; and, R. T. H. Collis and T. Kobayasi for similar talks during the concluding phase of this work. They also wish to thank Mrs. Valerie Lipinski for the preparation of the manuscript.

REFERENCES [l ] G. B. Morgan, G. Ozolins, and E. C. Tabor, “Air pollution surveil-

[2] B. B. Snavely, “Flashlampexcited organic dye lasers,” Proc.

[3] S. E. Harris, “Tunable optical parametric oscillators,” Proc. IEEE, vol. 57, Dec. 1969, pp. 2096-2113.

[4] H. Inaba and T. Kobayasi, “Laser-Raman radar for chemical analysis of polluted air,” Nature, vol. 224, 1969, p. 170. -, “Laser-Raman radar for air pollution probe,” Proc. IEEE (Special Issue on Optical Commmication), vol. 58, Oct. 1970, pp. 1568-1571; also, T. Hirschfeld, S. Klaimer, and R. Burton, “New fields for laser Raman spectroscopy,” presented at the Electro- Optical Systems Conf., New York, Sept. 16-18, 1969.

[5] D. A. Leonard, “Observation of Raman scattering from the atmosphere using a pulsed nitrogen ultraviolet laser,” Nature, vol. 216, 1967, p. 142.

[a] J. A. Cooney, “Measurements on the Raman component of Laser atmospheric backscatter,” Appl. Phys. Lerr., vol. 12, 1968, p. 40.

[7] S. H. Melfi, J. D. Lawrence, Jr., and M. P. McCormick, “Observa- tion of Raman scattering by water vapor in the atmosphere,” Appl. Phys. Lett., vol. 15, 1969, p. 295; also, J. A. Cooney, “Com- parisons of water vapor profiles obtained by radiosonde and laser backscatter,”J. Appl. Meteorol., vol. 10, 1971, 301.

[8] T. Kobayasi and H. Inaba, “Spectroscopic detection of SO2 and C02 molecules in polluted atmosphere by laser-Raman radar technique,” Appl. Phys. Lett., vol. 17, 1970, p. 139.

[9] M. R. Bowman, A. J. Gibson, and M. C. W. Sandford, “Atmo- spheric sodium measured by a tuned laser radar,” Nature, vol. 221, 1969, p. 456.

[lo] E. J. Stansbury, M. F. Crawford, and H. L. Welsh, “Determina- tion of rates of change of polarizability from Raman and Rayleigh intensities,” Can. J . Phys., vol. 31, 1953, p. 954.

lance systems,” Science, vol. 170, 1970, p. 289.

IEEE, VOI. 57, Aug. 1969, pp. 1374-1390.

K L D A L AND BYER: LASER DEI’ECTION OF A‘IMOSPHERIC POJLUTANlX 1663

T. L. Cottrell and J. C. McCoubrey, Molecular Energy Transfer in Gases. London, England:Butterworth, 1961. P. L. Hanst and J. A. Morreal, “Detection and measurement of air pollutants by absorptions of infrared radiation,” J. Air Pollut. Contr. Ass., vol. 18, 1968, p. 754; also, S. Zaromb, “Remote sensing of invisible air pollutants by Lidar absorption spectroscopy,” presented at the Electro-optical Systems Design Conf., New York, Sept. 1618, 1969. P. W. Kruse, L. D. McGlauchlin, and R. B. McQuistan, Elements of Infrared Technology. New York: Wiley, 1963. K. Ya. Kondratyev, Radiation in the Atmosphere. New York: Academic Press, 1969. A. E. S. Green, The Middle Ultraviolet. New York: Wiley, 1966. R. A. Smith, F. E. Jones, and R. P. Chasmar, The Detection and Measurement of Infra-Red Radiation. London, England: Oxford, 1 OAR

[17] G. Herzberg, Spectra of Diatomic Molecules. New York: Van Nostrand, 1959.

[I81 -, Infrared and Raman Spectra. New York: Van Nostrand, 1968.

[19] W. K. Pratt, Laser Communication Systems. New York: Wiley, 1969.

[ a ] Reference Manual for the 1700 Series Spectrometers. Spex Indus- tries Inc., 1969.

[21] Reference Manual for the I401 Double Spectrometer. Spex Indus- tries Inc., 1969.

[22] T. Hirschfeld and S. Klaimer, “Remote Raman spectroscopy as a pollution radar,” Opt. Spectra, vol. 4, July-Aug. 1970, p. 63.

[23] D. A. Leonard, “Measurement of NO and SOz Raman-scattering cross sections,” J. Appl. Phys., vol. 41,1970, p. 4238.

[24] V. E. Derr and C. G. Little, “A comparison of remote sensing of the clear atmosphere by optical, radio, and acoustic radar techniques,” Appl. Opt., vol. 9, 1970, p. 1976.

[25] 0. Stern and M. Volmer, “ U b e r die Abklingungszeit der Fluores- zenz,” Phys. Z. , vol. 2 0 , 1919, p. 183.

[%I H. D. Mettee, “Foreign gas quenching of sulfer dioxide vapor emission,” J. Phys. Chem., vol. 73, 1969, p. 1071.

[27] J. A. Myer, I. Itzkan, and E. Kierstead, “Dye lasers in the ultraviolet,” Nature, vol. 225, 1970, p. 544.

[28] Y. Miyazoe and M. Maeda, “Stimulated emission from 19 poly- methine dyes-Laser action over the continuous range 710-1060 mp,” Appl. Phys. Lett., vol. 12, 1968, p. 206.

[29] D. A. Leonard, “Saturation of the molecular nitrogen second posi- tive laser transition,” Appl. Phys. Lett., vol. 7, 1965, p. 4.

[30] H. W. Furumoto and H. L. Ceccon, “Optical pumps for organic dye lasers,” Appl. Opt., vol. 8, 1969, p. 1613.

[31] B. H. Soffer and B. B. McFarland, “Continuously tunable, narrow-band organic dye Iziers,” Appl. Phys. Lett., vol. 10, 1967, p. 226.

[32] H. Walther and J. L. Hall, “Tunable dye laser with narrow spectral output,” Appl. Phys. Lett., vol. 17, 1970, p. 239.

[33] T. W. Hansch, 1. S. Shahin, and A. L. Schawlow, “High rwlution saturation spectroscopy of the sodium D lines with a pulsed tunable dye laser,” to be published.

[34] R. W. Wallace and S. E. Harris, “Extending the tunability spectrum,” Laser Focus, vol. 6, 1970, p. 42.

[35] J. F. Young, R. B. Miles, S. E. Harris, andR. W. Wallace, “Pump

A<”...

linewidth requirement for optical parametric oscillators,” J. Appl. Phys., vol. 42, 1971, p. 497.

[36] L. B. Kreuzer, “Single mode oscillation of a pulsed singly resonant optical parametric oscillator,” Appl. Phys. Lett., vol. 15, 1969, p. 263.

[37] R. L. Byer, H. Kildal, and R. S. Feigelson, “CdGeAsz-A new nonlinear crystal phasematchable at 10.6 pm,” to be published.

[38] E. D. Shaw and C. K. N. Patel, “Stimukted anti-Stokes spin-flip Raman scattering in InSb,” Appl. Phys. Lett., vol. 18, 1971, p. 215.

[39] S. R. J. Brueck and A. Mooridian, “Efiient, single-mode, CW, tunable spin-flip Raman laser,” Appl. Phys. Lett., vol. 18, 1971, p. 229.

[40] E. D. Hinkley, T. C. Harman, and C. Freed, “Optical heterodyne detection at 10.6 pm of the beat frequency between a tunable Pbo.s&no.ltTe diode laser and a COZ gas laser,” Appl. Phys. Lett., vol. 13, 1968, p. 49.

[41] 1. Melngailis, A. J. Straw, and R. H. Rediker, “Semiconductor diode masers of I&Gal-Js,” Proc. ZEEE (Corresp.), vol. 51,

[42] J. F. Butler and T. C. Harman, “Long-wavelength infrared Pbl-.SnzTe diode lasers,” Appl. Phys. Lett., vol. 12, 1968, p. 347.

[43] E. D. Hinkley, “High-resolution infrared spectroscopy with a tunable diode laser,” Appl. Phys. Lett., vol. 16, 1970, p. 351.

[44] J. C. Dyment, “Hermite-Gaussian mode patterns in GaAs junction lasers,” Appl. Phys. Lett., vol. 10, 1967, p. 84.

[45] C. K. N. Patel, “Vibrational-rotational laser action in carbon monoxide,” Phys. Reo., vol. 141, 1966, p. 71.

[46] A. C. G. Mitchell and M. W. Zemansky, Resonance Radiation and Excited Atoms. New York: Cambridge University Press, 1961.

[47] D. T. Williams and B. L. Kolitz, “Molecular correlation spec- .trometry,” Appl. Opt., vol. 7, 1968, p. 607.

[48] J. H. Callomon: T. M. DUM, and I. M. Mills, “Rotational analysis of the 2600 A absorption system of benzene,” Phil. Trans. Roy. Soc. London, vol. 259A, 1966, p. 499.

[49] J. B. Birks, Photophysics of Aromatic Molecules. New York: Wiley-Interscience, 1970.

[50] J. G. Calvert and J. N. Pitts, Photochemistry. New York: Wiley, 1966.

[51] K. F. Greenough and A. B. Duncan, “The fluorescence of sulfer dioxide,” J. Amer. Chem. Soc., vol. 83, 1961, p. 555.

[52] G. H. Myers, D. M. Silver, and K. Kaufman, “Quenching of NO, fluorescence,” J . Chem. Phys., vol. 44,1966, p. 718.

[53] T. C. Hall, Jr., and F. E. Blacet, “Separation of thc absorption spectra of NOr and NrOl in the range of 2400-5000 A,” J. Chem. Phys., vol. 20, 1952, p. 1745.

[54] P. Warneck, F. F. Marmo, and J. 0. Sullivan, “Ultraviolet absorption of SO*: Dissociation energies of S O 2 and S O , ” J. Chem. Phys., vol. 40, 1964, p. 11 32.

[55] J. I. Steinfeld, “Quenching of fluorescence in small molecules,” Accounts Chem. Res., vol. 3, 1970, p. 31 3.

[56] C . K. Ingold, “The Structure of benzene,” Proc. Roy. Soc. (Lon- don), vol. 169A, 1939, p. 149.

[57] B. M. Sturgis, J. W. Bozek, W. F. Biller, and S. B. Smith, “The application of continuous infrared instruments to the analysis of exhaust gas,” SAE Tech. Prog., Ser. 6, 1964, p. 81.

[58] D. R. Herriott and H. J. Schulte, “Folded optical delay lines,” Appl. Opt., vol. 4, 1965, p. 883.

Aug. 1963, pp. 1154-1155.

Documents

1644 PROCEFDINGS OF THE IEEE, VOL. 59, NO ...web.stanford.edu/~rlbyer/PDF_AllPubs/Conference Papers...rotational relaxation time of the order of 10-9 s. The nonradiative vibrational