atmospheric atomic mercury monitoring using differential absorption lidar techniques

Atmospheric atomic mercury monitoring using differentialabsorption lidar techniques

Hans Edner, Gregory W. Faris, Anders Sunesson, and Sune Svanberg

Three-dimensional mapping of atmospheric atomic mercury has been performed with lidar techniques, to ourknowledge, for the first time. Industrial pollution monitoring, as well as measurements of backgroundconcentrations, is reported. High-efficiency frequency doubling of narrowband pulsed dye laser radiation

was employed to generate intense radiation at the mercury UV resonance line. Field measurements were

supplemented with extensive laboratory investigations of absorption cross sections and interfering lines ofmolecular oxygen.

1. Introduction

Range-resolved monitoring of atmospheric atomicmercury pollution employing the differential absorp-tion lidar technique is reported, we believe, for the firsttime. A mobile lidar system that was equipped with anarrowband tunable laser transmitter able to generatepulses of adequate power at the mercury resonanceline at around 254 nm was employed. Interfering ab-sorption lines due to molecular oxygen were studied indetail to allow mercury measurements with a sensitiv-ity down to typical background levels, 2 ng/m 3 .

Atomic mercury is an atmospheric'pollutant that isdirectly generated from chlorine-alkali plants, coal-fired power plants and refuse-incineration plants.However, anthropogenic mercury enters the environ-ment also in the aquatic phase as water-soluble mercu-ry compounds, e.g., CH 3HgCl and HgCl2 . Again, in-dustrial activities as well as inadequate wastemanagement are responsible for such emissions. Acomplex and not fully understood interaction betweenthe water and atmospheric phases occurs in the envi-ronmental mercury cycle.1'2 Atomic mercury is alsoan interesting geophysical tracer gas associated withcertain ore deposits,3 -6 as well as geothermal, 7 8 seis-mic,9 and volcanic' 0 activities.

Typical background concentrations of atomic mer-cury are a few ng/m3 .11-13 Low concentrations of Hgare normally measured by point monitors employing

When this work was done all authors were with Lund Institute of

Technology, Physics Department, P.O. Box 118, S-221 00 Lund,Sweden; G. W. Faris is now with SRI International, Molecular Phys-

ics Laboratory, Menlo Park, California 94025, and A. Sunesson is

now with RIVM, Laboratory for Air Research, P.O. Box 1, 3720 BA

Bilthoven, The Netherlands.Received 17 March 1988.0003-6935/89/050921-10$02.00/0.© 1989 Optical Society of America.

gold amalgamation techniques combined with flame-less atomic absorption spectroscopy. 4"15 Direct opti-cal absorption or Zeeman absorption spectrometerscan also be utilized.16 Remote sensing techniques,such as lidar (light detection and ranging), and inparticular the differential absorption lidar (DIAL)17,18

provide important advantages over point monitoring.Typical Hg concentrations, which are exceedingly lowby DIAL standards, can still be measured by thattechnique only since the oscillator strength of the elec-tronic transition at 254 nm is concentrated in an atom-ic line rather than spread over the large number ofrotational-vibrational molecular transitions normallyencountered. Actually, mercury is the only pollutantthat is present in the troposphere in elemental form.Because of the sharpness of the mercury absorptionline a narrowband laser transmitter is necessary toattain maximum absorption when the laser is tuned toresonance with the line. For range-resolved lidar mea-surements that rely on atmospheric backscattering, asubstantial laser pulse energy is necessary to achieve auseful range of -1 km.

Early attempts to measure mercury with the DIALtechnique are reported in Ref. 19. Sensitivity andrange were severely limited by the large linewidth ofthe laser used (0.015 nm) and low output pulse energy(0.5 mJ) obtained by stimulated Raman scattering inH2 of frequency-doubled dye laser radiation (567 nm).To exploit a broad laser, the gas correlation lidar tech-nique20 was introduced and demonstrated for Hg. Byrecording on- and off-resonance signals simultaneous-ly, a greater immunity to atmospheric turbulence isobtained, but the sensitivity is reduced. In a paralleldevelopment we have also explored the potential ofpath-averaged measurements of mercury2l with theDOAS (differential optical absorption spectroscopy)method.22 23 Although this method is quite useful, it isvery difficult to achieve high spectral resolution to

1 March 1989 / Vol. 28, No. 5 / APPLIED OPTICS 921

obtain optimal sensitivity and freedom from the influ-ence of interfering lines.

Three-dimensional, highly sensitive mapping of Hg,reported in this paper, had to await the availability ofhigh-power narrowband pulsed laser sources at 254nm. The measurements were performed with our newmobile lidar system24 which, for this measurement,was equipped with a Quantel Datachrome system witha dual wavelength option and a linewidth of 0.001 nmin the UV. Using a betabarium borate (BBO) crystal,pulse energies up to 5 mJ could be generated by directfrequency doubling.

In the next section the mobile lidar system arrange-ments for the field work are described as well as thesetup for laboratory measurements of the Hg crosssection and interfering 02 lines. Laboratory measure-ments are described in Sec. III.A while the field work isreported in Sec. III.B. Results for plumes from achlorine-alkali plant are presented as well as measure-ments on background concentrations of Hg. In a sepa-rate section the lidar signal contributions due to Hgresonance fluorescence are discussed. Fluorescence isnormally of little importance in tropospheric lidarwork on molecules but is utilized in lidar monitoring ofstratospheric atomic layers.2526 For Hg, in atomicform, fluorescence occurs even at tropospheric pres-sures and can contribute substantially to the signalwhen polarization techniques are employed. Finally,conclusions are drawn in the last section.

II. Experimental Arrangement

The system that was used has been described exten-sively in Ref. 24. It is housed in a mobile truck with alaboratory area of 6.0 X 2.3 M2 . Power is supplied by a20-kVA diesel generator installed in a trailer towed bythe truck.

A Nd:YAG-pumped dye laser is tuned alternately totwo close-lying wavelengths, one that is on the mercuryresonance line and one that is off the line. The laserbeam passes a 6X beam expander and is transmittedinto the atmosphere using quartz prisms and a largeplane mirror housed in a dome construction that ishoisted up through a trapdoor in the roof of the truckduring measurements. The large mirror can be rotat-ed around both the horizontal and the vertical axes,thus allowing the beam to be aimed in the desireddirection. Backscattered radiation is collected withthe same mirror and directed into a Newtonian tele-scope with a 400-mm diameter. After the telescope,an interference filter selects the appropriate wave-length range for the detection. Detection is per-formed with an EMI 9816QA photomultiplier tube.The photomultiplier is gain-modulated to reduce thedynamic range of the signal and the mean power dissi-pation in the tube.

The electronic signal is A-D converted in one of thetwo channels of a LeCroy transient recorder with 10-nstime resolution. A GPIB interface transfers the digi-tized signal to an IBM AT-compatible computer wheredata are averaged and stored on floppy disks. Duringa measurement the system is controlled by the com-puter. It handles laser wavelength switching, mea-

surement direction setting, and control of data sam-pling and data averaging. Up to three plume scanswith fifteen directions each can be performed withoutinterference from the operator. Both horizontal andvertical sweeps are possible.

For this measurement the system was equipped witha new Quantel YG 581C Nd:YAG laser and a TDL 50dye laser. The pump laser delivers pulse energies of500 and 200 mJ at 532 and 355 nm, respectively, with10-Hz pulse repetition rate. The dye laser can deliverup to 200 mJ at 560 nm with a linewidth of 0.08 cm-'.To generate the 254-nm radiation the dye laser wasoperated with coumarin 500 at 507 nm and the outputwas frequency doubled with a betabarium borate(BBO) crystal, supplied by CSK Co., Los Angeles.Betabarium borate is a relatively new material fornonlinear optical processes. It has a high damagethreshold, good thermal stability, and it allows effi-cient doubling down to -200 nm. It was possible togenerate pulse energies of up to -5 mJ at 254 nm froma dye laser pulse energy of 25 mJ at 507 nm. Thelinewidth of the frequency-doubled laser beam was0.001 nm. To tune the dye laser on and off resonance adual-wavelength feature supplied by Quantel was uti-lized. In the dye laser oscillator the rear end of thecavity is split into two parts. The two parallel beamsare, after reflection from the single grating, tuned bytwo individual mirrors. The two cavities are tunedtogether by the laser wavelength tuning mechanismand the wavelength of the second cavity can be offsetrelative to that of the first one. A small chopper thatblocks the two cavities alternately as it rotates ismounted in the intracavity space and a computer-controlled stepper motor rotates the chopper, thusalternately allowing the laser to oscillate on one of thetwo wavelengths. This is advantageous comparedwith moving the tuning mechanism of the laser at 5 Hz,which can cause undesirable vibrations and wave-length instability.

To calibrate the dye laser wavelength 10% of theoutgoing beam is split off with a beam splitter anddirected into a calibration unit. After passagethrough neutral-density filters it is split by a 50% beamsplitter, and one part of it is passed through a thinquartz cell containing a drop of mercury in air whilethe other part is used as a reference beam. The inten-sities of the two beams are measured using photodi-odes. The signals are A-D converted in the transientrecorder, which is run as a boxcar averager by thecomputer. Measurements of the mercury differentialabsorption cross section were performed with the sameunit.

The setup shown in Fig. 1 was used for the laboratorymeasurements, where the oxygen absorption spectrumaround the mercury line was studied. The laser sys-tem and the data acquisition system in the mobile lidarsystem were used. The laser beam was directed fromthe truck, which is docked to the laboratory when it isnot used in a field campaign, to an optical table where amirror directed it through a lens and a diaphragm to,the input window of a White multipass cell that wasconstructed in our laboratory. 2 7 The cell length is 2 m

922 APPLIED OPTICS / Vol. 28, No. 5 / 1 March 1989

and its body is a Pyrex tube with end flanges of stain-less steel that, together with Invar rods, make up theframe of the device. The mirrors are Al and MgF2coated for maximum UV reflectance. Path lengths upto 200 m with lamps and 500 m with lasers have beenachieved. A He-Ne laser was used to align the cell.After the cell, the beam passed through one more dia-phragm and a lens and reached the detector, and EMI9558 QA photomultiplier tube. The signal could beviewed directly on an oscilloscope, while to collect thedata the transient recorder and the computer wereutilized. To provide a marker of the mercury lineposition the part of the beam that was split off forcalibration was passed through a thick mercury cell inthe calibration unit and was detected by a photodiode.This signal was also digitized in the transient recorderand stored together with the spectra.

Ill. Measurements

A. Laboratory Measurements

Laboratory measurements were performed to estab-lish the possible interference from oxygen absorptionin a measurement of mercury vapor in the atmosphere.The setup illustrated in Fig. 1 was utilized, where theWhite multipass cell was filled with 1 atm of pureoxygen. The path length through the cell was adjust-ed by observing the signal from the PMT at the output,either directly on an oscilloscope or after digitizationon the transient recorder. A read out from the latter isshown in Fig. 2. The sharp peak originates from thetransmitted laser light, while the signal at shorter de-lay times/distances is due to scattered light in sequen-tial reflections within the cell. The absorption wasmeasured by integrating the channels containing thesignal from the transmitted laser beam and normaliz-ing to a few channels for the first reflections. Thisconstituted an easy way of compensating for fluctuat-ing laser power during a wavelength scan, withoutinserting a beam splitter and an additional photode-tector. Thus, possible problems with fringes and dif-ferent characteristics of two photodetectors wereavoided. During an absorption measurement thenoise was lowered by using a small capacitor on thePMT, which distributed the signal due to the trans-mitted light over more channels in the transient re-corder. The path length used during the experimentswas 340 m.

Figure 3(a) shows the measured oxygen absorptionin a region close to the mercury resonance line. Arecorded Hg spectrum is inserted as a dotted line. Anenlargement of the Hg cell spectrum is shown in Fig.3(b). The comparatively broad structure of the Hgline is due to the different isotopic and hyperfine struc-ture lines present, as indicated in the figure.28 Thetypical linewidth of these lines at atmospheric pres-sure, together with the laser linewidth used, is alsoinserted. As can be seen from Fig. 3(a) two weakoxygen lines are very close to the mercury line. Thesetwo lines lie on the long-wavelength side of the v = 0 -v = 7 band of the Herzberg I system but can neither beidentified with Herzberg's listings nor with later list-

MOBILE LIDAR SYSTEM… _ m

LABORATORY

Osilloscop

Fig. 1. Experimental arrangement.

z

0 200 400PATHLENGTH ()

Fig. 2. Temporally resolved White cell signal.

ings of the Herzberg I, II, and III systems, to ourknowledge.29-33 The interference from these lines in aDIAL measurement of low background values of Hgover long atmospheric paths can be minimized if thetwo wavelengths are chosen carefully. With the laserlinewidth of 0.001 nm used here, one wavelength can beplaced on the Hg absorption peak, while the referencewavelength must avoid the oxygen lines. In thepresent measurement we used a wavelength differenceof 0.008 nm with the reference on the long-wavelengthside.

The relevant Hg absorption cross section for thepresent system was established by measuring the ab-sorption in several small Hg cells of known sizes andtemperature. The results are plotted in Fig. 4, yield-ing a cross section of 3.3 X 10-18 m2 or 9.8 X 10-6 m2/ng.

B. Field Measurements

During a one-week period the mobile lidar systemwas used in a field measurement of Hg emission from achlorine-alkali plant, which uses mercury in the pro-cess. The system was placed -550 m from the largestHg emission source, which was the outlet from a 60-mlong clerestory 13 m above the ground. Measurements


0.75

z

0 .0

1 .03

0.70

0.50

O.m I , I

253.645 253.650 253.655

WAVELENGTH (nm)

Fig. 3. Cell spectra: (a) overview with oxygen interference and (b)enlarged Hg signal with isotopic and hyperfine structure lines indi-

cated.

0.6

0.5

02 O.4

. 0.3

0. 2

0. I1

10 20 30 40 50 60OPTICAL DEPTH (<m')

Fig. 4. Diagram of data used for cross-sectional determination.

could be performed directly on the outlet or downwindon the spreading plume with a free atmospheric path of-900 m before the laser beam hit a hillside. Thepresence of Hg vapor in the clerestory outlet could beexplicitly shown by measuring the decrease in the idarsignal over a 300-m path containing the plume whilescanning the laser slowly over the Hg line. The ab-sorption spectrum thus obtained is shown in Fig. 5.The on- and off-resonance wavelengths in the quanti-tative DIAL measurements are indicated in the figure.

0.40

2 0.30InU)

0.10ON

0.03

253.64 253.65 253.66

WAVELENGTH (n)

Fig. 5. Recording of the absorption line from mercury releasedfrom a 500-m distant chlorine-alkali plant. The transmission onthe vertical scale has not been compensated for natural extinction

and 1/R2 dependence.

ON/OFF

Zo O

o1x s aW mW *ii MD SW0 7e SW

DISTANCE m

Fig. 6. (a) On/off-resonance lidar curves and (b) resulting DIALcurve.

Figure 6 shows an example from a DIAL measure-ment on the Hg plume. The curves in Fig. 6(a) showthe detected signal intensity due to backscattered laserlight for the on- and off-resonance wavelengths. Thesignal here is averaged over 1600 shots at each wave-length. The shape of the lidar curves is due to twofactors: geometrical dependence and gain modulationof the photomultiplier. Figure 6(b) shows the result-ing ratio of DIAL curve, where the significant effect ofthe plume is easily seen. By performing several se-


25.423.5236253.W

02

.. ....

II

. I. I

. __ - I

253.64 253.95

Hg PLUME

408 s88

> 280 ng'm3240 - 288288 - 240160 - 200120 - 160

se - 12040 - 80

0 - 40

Height

13 m

31~~~~~~~ 5a

g D1l

baa z~~~~~~

Distance

/m.6 E.0 8'5 i

by- ~ K, t , s s

Fig. 7. Vertical scan of the Hg plume downwind from a chlorine-alkali plant.

quential measurements of this kind in different direc-tions in a vertical section, the Hg concentration in thespreading plume could be charted. An example of thisis shown in Fig. 7, where the Hg concentrations wereevaluated from the DIAL curves with a range resolu-tion of 45 m. The behavior of the fast-descendingplume was studied in this way. The flux of Hg vaporthrough the vertical section was also measured by inte-grating the concentration over the area of the plumecross section and multiplying the integrated contentby the wind velocity normal to the vertical section. Inthe example given here, the flux was found to be 30 g/hof atomic Hg vapor.

Another example of a Hg concentration map isshown in Fig. 8, where a horizontal scan at low heightwas performed. Several scans were averaged to give amore representative picture. The result here is copieddirectly onto a map of the plant. This is often a usefulrepresentation, e.g., to locate different sources. Forclarity, isoconcentration lines are also drawn by handbeside the grey scale presented by the computer.

Some studies of the ability of the present system tomeasure the low background concentration of atomicmercury have also been performed. The typical rangefor concentration evaluation was found to be -1 kmwith normal visibility. This limited range is due to thenatural atmospheric extinction at these short wave-lengths, which was measured to 1.2-1.6 km-', in goodagreement with existing theory for Rayleigh scatteringand empirical formulas for the Mie scattering contri-bution. The absorption by ozone gives a contributionof 0.3 km-1 for a typical background concentration of10 ppb. A homogeneously distributed low concentra-tion can be evaluated by fitting the DIAL curve with astraight line. The slope of this line then gives theaverage concentration over the path. Figure 9 showsan example of this kind. Several similar measure-ments were performed, also with the laser tuned off theHg line, to estimate the noise in the concentrationevaluation. The noise was found to be between 0.5and 2 ng/m3, depending on the atmospheric condi-tions. Unfortunately, during the period of the mea-surement only values close to or slightly over the noise

of Hg distribution over a chlorine-alkaliplant.

Fig. 8. Horizontal scan

1.25

1.03

0

0 7

, .

tow

H 0.00

0.25

0.0

Fig. 9. Recording of Hg background concentration.

limit were detected. The small slope in Fig. 9 gives aconcentration of 1.5 t 1 ng/m 3 .

IV. Hg Resonance Fluorescence

Lidar detection of atomic mercury is complicated byresonance fluorescence, as previously reported in Ref.19. Because it occurs at the same wavelength as thedetected backscattered light, resonance fluorescencecannot be separated spectrally from the signal. Wehave examined the effects of fluorescence on DIALmeasurements both experimentally and theoretically,and present a technique for correcting for the effects offluorescence.

For DIAL measurements, resonance fluorescencewill usually only affect the on-resonance measure-ment. This can result in a signal for the on-resonancewavelength which is greater than that of the off-reso-nance wavelength. This effect is seen in the lidarcurves shown in the lower part of Fig. 10(a). Thesewere taken near the outlet of the clerestory at thechlorine-alkali plant and the sharp gradient in themercury concentration results in a fluorescence peakat the on-resonance wavelength at the onset of absorp-tion. The peak is also visible in the DIAL (upper)


It 35,,1, V n Ao 01 an 200 4W all DA (i)

DISTANCE (m

, / e-

Mu :u ILW law *swn

0

, .03

0

0.00

z.^o

,2..

DISTANCE (m)

Fig. 10. Lidar/DIAL curves recorded for (a) both polarizations and(b) parallel and (c) perpendicular to the laser polarization.

curve which shows the ratio of the on- to off-resonancewavelengths. If evaluated over a suitably narrow in-terval, this would result in a seemingly negative mercu-ry concentration. The qualitative effects of uncor-rected resonance fluorescence are demonstrated in amodel study described in the Appendix. These in-clude a displacement of the calculated concentrationcurve to a position further from the lidar system, and adistortion of the signal. For small absorptions, thedisplacement is independent of the concentrationpresent and depends solely on the ratio of the fluores-cence backscattering coefficient to the nonresonantbackscattering coefficient.

A simple argument demonstrates that the total con-centration measured is not affected by resonance fluo-rescence. The natural fluorescence lifetime of the6s6p 3P level in mercury is -100 ns.3 4 The quenchingfactor for Hg atoms in air is -2 X 10-3, correspondingto a quenching time of 0.2 ns.35 Thus, finite relaxationeffects36 are not significant, and the fluorescence signalis range resolved. The fluorescence from a given vol-ume will be detected simultaneously with the back-scattered signal, making the effective scattering at the

resonance wavelength greater than that at the nonres-onance (off) wavelength. Thus, errors in the DIALsignal due to fluorescence will be limited to the regionswhere mercury is present. Thus, the absorption of thereturning signal from beyond a mercury plume is unaf-fected by fluorescence, and the total integrated con-centration obtained is correct, even though the exactshape of the concentration calculated from the DIALmeasurements may be incorrect.

To correct for the effects of resonance fluorescence,the strength of the fluorescence backscattering rela-tive to the nonresonant Rayleigh and Mie backscatter-ing must be known. While this ratio is treated asvariable in the theory presented in the Appendix, it isdifficult to determine this ratio for all points in a lidarcurve. The ratio may vary due to particles in a plumeor local variations in the quenching rate. Variation inthe backscattering due to particles may be seen asincreased scattering at the off-resonance signal abovethe level for clean air. Variation in the quenching rateis more difficult to determine. For the measurementat the chlorine-alkali plant, little scattering due toparticles was observed, and the fluorescence-to-non-resonant backscattering ratio was treated as a con-stant. It is noted that the effects of fluorescence inthis case of relatively clean air were small and that anincrease in either backscattering or quenching wouldmake the effects still smaller.

We have attempted to measure the fluorescence-to-nonresonant backscattering ratio through lidar mea-surements at orthogonal polarizations. An Ealing UVpolarizer was inserted between the collecting telescopeand the photomultiplier. The returning signal wasfound to be strongly polarized. Measurements weretaken with the polarizer oriented to give maximumtransmission and then at 90°. The resulting lidarcurves are shown at the bottom of Figs. 10(b) and 10(c).All three data sets shown in Fig. 10 were taken during aspan of 23 min. The relative scales for curves (b) and(c) differ by a factor of 10, so there is a difference in thebackscattering signal of a factor of 15. This indicatesthat the scattering at this short wavelength is predomi-nantly polarized. Rayleigh scattering, which pre-serves polarization, varies as 1/X4. Mie scattering hasa slower wavelength dependence and it maintains po-larization if the scatterers are spherical. Nonsphericalscatterers or multiple scattering may induce depolar-ization. The fluorescence signal, however, appears tobe unpolarized. That is, the relative magnitudes ofthe two polarizations are about the same. This wouldindicate that the rate of elastic depolarizing colli-sions37 is significantly greater than the quenching rate.Because the two curves were not recorded simulta-neously, it is impossible to determine the exact ratiobetween the two polarizations. Because of the differ-ence in polarization between the main backscatteringsignal and the fluorescence signal, a polarizer may beused to suppress the fluorescence effects. This couldyield a factor of 2 reduction in the fluorescence effectsfor a completely polarized backscattered signal, and anunpolarized fluorescence signal. The factor F/B de-scribed in the Appendix was found to be 2 X 10-4 m3/ng


l .3

1.00

0.75

CM

*.20

OMC

0.0

from the height of the fluorescence peak shown in Fig.10(c), the magnitudes of the background scattering inFigs. 10(b) and (c), and the calculated concentrationfrom Fig. 10(a). With this value, the method de-scribed in the Appendix may be used to correct for thefluorescence. This was done for the curves shown inFig. 10(a) and the results are shown as the solid curvein Fig. 11. The dashed curve is the result calculatedwithout any correction for fluorescence. This curve isdisplaced from the position of the corrected curve by10 m, in agreement with the expected displacementdescribed in the Appendix. Additionally, the uncor-rected curve is distorted, lying too low on the near side,and too high on the far side, as expected from themodel study presented in the Appendix.V. Conclusions

Lidar measurements of atomic mercury in the atmo-sphere have been demonstrated, yielding 3-D maps ofthe gas distribution. High laser pulse energies and anarrow laser linewidth were found to be essential forthis development. We found that the effects of fluo-rescence led to a displacement of the calculated con-centration curve and a distortion of the concentrationprofile. We have presented a theoretical technique forcorrecting these effects, and an experimental tech-nique for measuring the fluorescence to backgroundfactor necessary to implement this correction. TheHg lidar technique shows high potential for pollutionmonitoring and also for studies of geophysical phe-nomena. A search for mercury in the atmosphere atIcelandic geothermal fields has already been per-formed38 and further studies of atmospheric mercuryof geophysical origin seem to be of considerable inter-est.

The authors are grateful to B. Galle for helping withthe practical arrangements for the field work. Thiswork was supported by the Swedish Board for SpaceActivities, The Swedish Environmental ProtectionBoard, and the Swedish Natural Science ResearchCouncil. One of us (G.W.F.) would like to thank theU.S. National Science Foundation for a grant enablinghim to spend a year at the Lund Institute of Technol-ogy.

Appendix: Correction for Resonance Fluorescence

The lidar equation for the detected power P(X,R) ata wavelength X, resulting from atmospheric backscat-tering at a distance R, may be written as

PAS(X,R) = C(X,R)3(X,R) exp{-2 J [Nj(R')o(X)

+ Nr(R')ar(X)]dR} - (Al)

C(XR) is a factor which includes such effects as theemitted laser energy and temporal profile, the laser'seffective beam area at range R, the overlap of thetransmitted beam and the receiving optics field ofview, the solid angle of the backscattered flux capturedby the receiving optics, and the optical efficiency of thecollection optics, filters, and the detector. 3(X,R) is

11121.03

14.00

0

6 03.0 E.

z.

a20.0 Im X al 300 400

DISTANCE ()

MO G 700 O

Fig. 11. Mercury concentration curves calculated with (solid line)and without (dashed line) correction for fluorescence.

the volume atmospheric backscattering coefficient.N(R) and as(X), and Nr(R) and aor(X) are the concentra-tions and absorption cross sections of an atmosphericconstituent of interest and the remaining constituents,respectively. A similar equation may be written forthe detected power due to resonance fluorescence,PF(A,R), for the constituent of interest:

PF(X,R) = C(X,R)Nj(R)F(X,R) exa{-2 J [Nj(R')aj(X)

+ Nr(R'),rQ()]dR} (A2)

where F(X,R) represents the normalized fluorescencescattering coefficient, which includes the effects of theabsorption cross section, the quantum efficiency forresonance fluorescence, and quenching. It is assumedhere, as discussed in Sec. IV, that the lifetime of theupper level is much shorter than the laser pulse length.These two equations may be combined to yield thetotal detected power due to atmospheric and fluores-cence scattering:

P(X,R) = C(X,R) [O(X,R) + Ni(R)F(XR)]

X exp{+2 J [Nj(R')a(X) + Nr(R')ar(X)]dR'}* (A3)

DIAL measurements are taken at two closely spacedwavelengths, that on resonance Xon-, and that off reso-nance Xoff. It is then assumed that the factors fl(XR)and ,r(X) are the same for the two wavelengths, whileC(Xon,R) and C(XoffR) are related by a constant. Theratio of the powers for the on- and off-resonance wave-lengths, P(R) = P(XonR)/P(XoffR), may be written as

P(R) = [ l+ N (R)R] exp[-2adiff J Ni(R')dR'] (A4)

where 0Jdiff = c(Xon) - c(Xoff) represents the differencebetween the cross sections for the on- and off-reso-nance wavelengths, the fluorescence backscatteringfor the off-resonance wavelength is assumed to be neg-ligible compared to the atmospheric backscattering,and the power ratio has been normalized by the factorC(X0nR)/C(XoffR). F(R) and d(R) are understood to


I

I'

apply to the wavelength Xo, and N now refers to theconcentration of the constituent of interest. Thisequation may be written in an (exact) differentialform:

P(R + AR) = IP() [1+ F(R + A) N(R + AR)+ F(R) NR L fl(R +AR~)I

fl(R)

X expi-2adiff JR N(R')dR .

Equation (AS) may alternatively be written as

P(R + AR) = PNF(R) 1 + gF(R + AR) N(R + AR)]

X exl{GR+ARd

X exp-20diff f, N(R')dR' I

(A5)

1.0

..20

0

Z 0.00

0.30

(A6)

where PNF(R) represents the on- to off-resonance pow-er ratio that would be present if there were no fluores-cence. This equation shows explicitly that, for negligi-ble upper level relaxation times, the power ratioreturning from a point R + AR is not dependent on thefluorescence contributions from other points. If theabsorption over the length AR is small, this equationmay be rewritten as

P(R + AR) PNF(R)[I + F(R + AR) N(R + AR)-2adifAR

(A7)

where N represents a value averaged between R and R+ AR. For small AR such that N, F, and are aboutconstant, then

P(R + AR) PNF(R)[l + N - 2odiffAR)]. (A8)

Note that, for a displacement AR given by

AR-: F (A9)2adiffo A9

the displaced fluorescence power ratio, (R + AR), isequal to the undisplaced power ratio in the absence offluorescence. As this relation holds for every point R(so long as the concentration does not vary greatly overthe width AR), the dominant effect of fluorescence isseen as a displacement of the DIAL signal away fromthe observer by a distance of -F/2ofl. Note that thisdisplacement is independent of the concentration N.Thus, to first order, an equally large displacement willbe found for small concentrations as for large concen-trations. For the mercury differential cross sectiongiven in Sec. III, 9.8 X 10-6 m2 /ng, and the ratio F/lgiven in Sec. IV, 2 X 10-4 m3 /ng, the displacement is'10 m. Simulation studies of fluorescence lidar39

have found a similar type of shift, but in that case,absorption causes the measured fluorescence curve tobe shifted toward the observer.

This displacement and other effects of resonancefluorescence may be illustrated with a simulation.Figure 12(a) shows the P ratio, or DIAL curves for asimulated rectangular mercury cloud, of width 75 mand a uniform concentration of 1000 ng/m3 . The solid

0.03

'403.03

I

O.X

Z (3.0E.

z

-218.03

DISTANCE ()

Fig. 12. DIAL curves for a simulated Hg cloud: (a) displaying thefluorescence effect (dashed line) and (b) calculated concentrationprofiles obtained with (dashed line) and without (dotted line) cor-

rection for fluorescence.

line shows the DIAL curve if there were no fluores-cence and the dashed line shows the DIAL curve whenthe normalized fluorescence ratio, F/, is constant,equal to 2 X 10-4 m3/ng. While a Gaussian or othersmoothly varying cloud is more typical of real clouds, arectangular cloud was chosen to better demonstratethe effects of resonance fluorescence and a method tocorrect for these effects. The displacement betweenthe two curves agrees with the expression in Eq. (A9),but at the sharp concentration gradients at the edges,other effects appear. There the mercury concentra-tion changes considerably in the displacement dis-tance given by Eq. (A9), and discontinuities arepresent at the boundaries of the cloud.

The effects of fluorescence may be corrected usingEq. (A5). This is similar in form to the differentialequation normally used to evaluate concentrationsfrom a DIAL curve based on absorption alone. How-ever, an additional fluorescence term contains the con-centration to be evaluated, N(R + AR), and the absorp-tion and fluorescence terms effectively have oppositesigns [see Eq. (A7), for example]. This makes theconcentration calculation more sensitive to noise thanfor absorption alone, and, in the worst case, for anevaluation interval AR given by Eq. (A9) there is nodependence on the concentration at all. However,when integrating in the opposite direction, that is,from larger R to smaller R, both the absorption and


fluorescence terms have the same sign, and the integra-tion is less sensitive to noise. This is the approach thathas been used to solve for the fluorescence-correctedconcentration.

Using subscript 1 to denote the position R and 2 todenote R + AR, and substituting the average concen-tration value (N + N2)/2 in the exponential, Eq. (A5)may be written as

Pi 1P + [1+ F2 N 2] exp[-diff(Nl + N 2)AR]. (A10)1 +-N,

This can be rewritten as a transcendental equation forthe quantity N in terms of N2 and the known powerratios at points 1 and 2, PI and P2 :

F1+ -- N11 exp(oAdiffJN1AR) - 1+ F2 N2] exp(-adjffN2AR).[ T3 ] 2 [ /2

(All)

This equation may be solved using Newton's method40

for all values of N and converges typically in one tothree iterations. In evaluating an experimentallymeasured lidar curve, noise in the data combined witha limited number of samples necessitates spatial aver-aging. This was done with a sliding average techniqueas is done in the evaluation of lidar curves withoutfluorescence. Because of the large separation betweenpoints 1 and 2 for a sliding average technique, thesolution of Eq. (All) is sensitive to overshoot andundershoot. This is partly due to the dependence ofthe solution for the concentration N on previouslydetermined concentrations N2. This is not a problemin the solution of absorption alone. In evaluating Eq.(All) for longer integration ranges AR, it was foundthat keeping the resulting value of N and storing it atposition 1 (appropriate for fluorescence alone, this issomewhat like a forward difference approximation 41 ),results in underdamped overshoot when there is a largeconcentration gradient and absorption is predomi-nant. When the calculated value is stored as a value ofN, or (N + N2)/2, at a position midway between 1 and2 (appropriate for absorption alone, this is similar to acentral difference solution41 ), undershoot occurs forlarge concentration gradients when fluorescence ispredominant. Very good results were found wheneach evaluation was stored as the interpreted valuebetween N and N2 at a distance

AR L diffAR 1 (A12)

, + tldiffAR

toward point 2 from point 1. Thus, when fluorescenceeffects are dominant (large F/fl), the evaluation is morelike a forward difference approximation, but whenabsorption is dominant (small F/fl), it becomes morelike a central difference solution.

The performance of this technique was evaluatedusing the DIAL curves in Fig. 12(a), a demanding testcase because of the infinitesimally sharp concentrationgradients. The resulting concentration curves areshown in Fig. 12(b). The curves were calculated using

a sliding average technique with a path length of 30 mand an average length of 11 points (15 m). Theselengths are typical for an experimental curve with goodsignal-to-noise ratio. Shown as the solid line in Fig.12(b) is the concentration curve evaluated from thesolid line in Fig.12(a), that without fluorescence. Thedotted line in Fig. 12(b) shows the concentration calcu-lated from the DIAL curve containing fluorescence[the dashed line in Fig. 12(a)] evaluated in the normalmanner. This curve is shifted by the expected dis-placement, and there is additionally a dip on the nearside of the concentration cloud and an overshoot on thefar side. The dashed line in the figure shows theconcentration calculated from the same DIAL curvebut evaluated using Eq. (All) together with Eq. (A12).This curve follows closely the proper, solid line, withsome small transients at the corners.

References1. J. 0. Nriagu, Ed., The Biogeochemistry of Mercury in the Envi-

ronment (Elsevier/North-Holland, Amsterdam, 1979).2. 0. Lindqvist and H. Rodhe, "Atmospheric Mercury-a Re-

view," Tellus 37B, 136 (1985).3. Q. Bristow and I. R. Jonasson, "Vapour Sensing for Mineral

Exploration," Can. Min. J. 93, 39 (1972).4. J. H. McCarthy, Jr., "Mercury Vapor and Other Volatile Com-

ponents in the Air as Guides to Ore Deposits," J. Geochem.Explor. 1, 143 (1972).

5. E. Kromer, G. Friedrich and P. Wallner, "Mercury and MercuryCompounds in Surface Air, Soil Gas, Soils and Rocks," J. Geo-chem. Explor. 15, 51 (1980).

6. R. W. Klusman and J. D. Webster, "Meteorological Noise inCrustal Gas Emission and Relevance to Geochemical Explora-tion," J. Geochem. Explor. 15, 63 (1981).

7. J. C. Varekamp and P. R. Buceck, "Hg Anomalies in Soils: aGeochemical Exploration Method for Geothermal Areas,"Geothermics 12, 29 (1983).

8. D. E. Robertson, E. A. Crecelius, J. S. Fruchter, and J. D.Ludwick, "Mercury Emissions from Geothermal Power Plants,"Science 196, 1094 (1977).

9. V. Z. Furzov, N. B. Volfson, and A. G. Khvalovskiy, "Results of aStudy of Mercury Vapour in the Tashkent Earthquake Zone,"Dokl. Acad. Nauk SSSR 179, 208 (1968).

10. J. C. Varekamp and P. R. Buseck, "Mercury Emissions fromMount St. Helens During September 1980," Nature London293, 555 (1981).

11. F. Slemr, W. Seiler, and G. Schuster, "Latitudinal Distributionof Mercury over the Atlantic Ocean," J. Geophys. Res. 86, 1159(1981).

12. S. H. Williston, "Mercury in the Atmosphere," J. Geophys. Res.73, 7051 (1968).

13. C. Brosset, "Total Airborne Mercury and Its Possible Origin,"Water Air Soil Pollut. 17, 37 (1982).

14. W. F. Fitzgerald and G. A. Gill, "Subnanogram Determination ofMercury by Two-Stage Gold Amalgamation and Gas PhaseDetection Applied to Atmospheric Analysis," Anal. Chem. 51,1714 (1979).

15. F. Slemr, W. Seiler, C. Eberling, and P. Roggendorf, "The Deter-mination of Total Gaseous Mercury in Air at BackgroundLevels," Anal. Chim. Acta 110, 35 (1979).

16. J. C. Robbins, "Zeeman Spectrometer for Measurements ofAtmospheric Mercury Vapour," in Geochemical Exploration,M. J. Jones, Ed. (Institute of Mining & Metallurgy, London,1973), p. 315.


17. D. K. Killinger and A. Mooradian, Eds., Optical and LaserRemote Sensing (Springer-Verlag, Berlin, 1983).

18. R. M. Measures, Laser Remote Sensing (Wiley-Interscience,New York, 1984).

19. M. Alden, H. Edner, and S. Svanberg, "Remote Measurement ofAtmospheric Mercury Using Differential Absorption Lidar,"Opt. Lett. 7, 221 (1982).

20. H. Edner, S. Svanberg, L. Un6us, and W. Wendt, "Gas-Correla-tion Lidar," Opt. Lett. 9, 493 (1984).

21. H. Edner, A. Sunesson, S. Svanberg, L. Un6us, and S. Wallin,"Differential Optical Absorption Spectroscopy System Used forAtmospheric Mercury Monitoring," Appl. Opt. 25, 403 (1986).

22. U. Platt, D. Perner, and H. W. Pdtz, "Simultaneous Measure-ment of Atmospheric CH2 O, 03, and NO2 by Differential OpticalAbsorption," J. Geophys. Res. 84, 6329 (1979).

23. U. Platt and D. Perner, "Measurements of Atmospheric TraceGases by Long-Path Differential UV/Visible Absorption Spec-troscopy," in Ref. 17.

24. H. Edner, K. Fredriksson, A. Sunesson, S. Svanberg, L. Un6us,and W. Wendt, "Mobile Remote Sensing System for Atmo-spheric Monitoring," Appl. Opt. 26, 4330 (1987).

25. C. Granier, J. P. Jegou, and G. Megie, "Resonant Lidar Detec-tion of Ca and Ca+ in the Upper Atmosphere," Geophys. Res.Lett. 12, 10 (1985).

26. K. H. Fricke and U. von Zahn, "Mesopause Temperatures De-rived from Probing the Hyperfine Structure of the D2 ResonanceLine of Sodium by Lidar," J. Atmos. Terr. Phys. 47,499 (1985).

27. A. Sunesson, "Construction of a Multipass Absorption Cell,"Lund Reports on Atomic Physics LRAP-46, Lund Institute ofTechnology (1986).

28. Y. Nishimura and T. Fujimoto, " = 2537 A Line from a Low-Pressure Mercury Discharge Lamp. Emission Profile and LineAbsorption by a Gas Containing a Mercury Vapor," Appl. Phys.B 38, 91 (1985).

29. G. Herzberg, "Forbidden Transitions in Diatomic Molecules.II. The 3+-37 Absorption Bands of the Oxygen Molecule,"Can. J. Phys. 30, 185 (1952).

30. G. Herzberg, "Forbidden Transitions in Diatomic Molecules.III. New lZ- 3Z and 3A6 -32T Absorption Bands of the OxygenMolecule," Can. J. Phys. 31, 657 (1953).

31. D. A. Ramsay, "High-Resolution Studies of the Near-Ultravio-let Bands of Oxygen: I: The c'2]-X 3Z- System," Can. J.Phys. 64, 717 (1986).

32. P. M. Borrell, P. Borrell, and D. A. Ramsay, "High-ResolutionStudies of the Near-Ultraviolet Bands of Oxygen: II: TheA 32+-X 3 2T System," Can. J. Phys. 64, 721 (1986).

33. B. Coquart and D. A. Ramsay, "High-Resolution Studies of theNear-Ultraviolet Bands of Oxygen: III: The A' 3A6 3-X 31T Sys-tem," Can. J. Phys. 64, 726 (1986).

34. J. Brossel and F. Bitter, "A New "Double Resonance" Methodfor Investigating Atomic Energy Levels," Phys. Rev. 86, 308(1952).

35. H. Inaba, "Detection of Atoms and Molecules by Raman Scat-tering and Resonance Fluorescence," in Laser Monitoring of theAtmosphere, E. D. Hinkley, Ed. (Springer-Verlag, Berlin, 1976),p. 182.

36. H. Kildal and R. L. Dyer, "Comparison of Laser Methods for theRemote Detection of Atmospheric Pollutants," Proc. IEEE 59,1644 (1971).

37. P. M. Doherty and D. R. Crosley, "Polarization of Laser-In-duced Fluorescence in OH in an Atmospheric Pressure Flame,"Appl. Opt. 23, 713 (1984).

38. H. Edner, G. Faris, A. Sunesson, S. Svanberg, J. 0. Bjarnasson,H. Kristmannsdottir, and K. H. Sigurdsson, "Lidar Search forAtmospheric Atomic Mercury in Icelandic Geothermal Fields,"manuscript in preparation for J. Geophys. Res.

39. R. M. Measures and G. Pilon, "A Study of Tunable Laser Tech-niques for Remote Mapping of Specific Gaseous Constituents ofthe Atmosphere," Opto-Electronics 4, 141 (1972).

40. F. S. Acton, Numerical Methods that Work (Harper & Row,New York, 1970), p. 51.

41. G. P. Smith, Numerical Solution of Partial Differential Equa-tions: Finite Difference Methods (Clarendon, Oxford,1985), p.7.

Mercury Ion Laser-Cooled To LimitScientists at NIST's Time and Frequency Division, Boulder, Colo..have succeeded for the first time in loser-cooling a bound atomicIon to its fundamental limit. We pushed the atom Into the groundstate of its confining well. That's the end of cooling for a boundparticle,' says project leader David J. Wineland. Their finding Is Im-portant for spectroscopy. a study of the nature of matter throughvarious radiations it emits. One result may be the development of ahighly sensitive spectrum analyzer. A report on their work appearsIn the Jan. 23, 1989, issue of Physical Review Lefters. They shinedlaser light on a mercury ion sideband frequency generated by theDoppler effect associated with thermal motion. The result was toreduce the ion's kinetic energy, limit its movement, and sharpen Itsspectral features. The Ion was confined In a radio frequency'trap.'


atmospheric atomic mercury monitoring using differential absorption lidar techniques

Documents