Three-wavelength light transmission technique to measuresmoke particle size and concentration

Kenneth L. Cashdollar, Calvin K. Lee, and Joseph M. Singer

This paper describes an optical technique and instrumentation for measuring the average particle size andmass concentration of smoke. Transmission through the smoke at three wavelengths (0.45 Atm, 0.63 Atm, and

1.00 ,um) is measured using a white light source and a compact, three-wavelength detector assembly. Sizes

and concentrations are then calculated from the transmission data using Mie theory. Graphs of the calcu-lated Mie extinction coefficients are presented for several complex refractive indices. This three-wave-

length optical technique was used to study smoke from a wood-tunnel fire, and the results are compared tothose of other sizing techniques.

Introduction

Recent interest in the characteristics and hazardsof smoke particulates generated in fires has promptedthe development of various techniques for determiningthe physical and chemical properties of smoke. Twocommonly used methods are sample withdrawal (col-lection of particles on a filter or removal of smokethrough tubing for external analysis) and in situ opticalmeasurements (light transmission or angular scatter-ing). The optical techniques have the advantage ofproviding continuous measurements without disturbingthe system. Some investigators have measured lighttransmission at a single wavelength or a broad wave-length band to obtain the optical density of a smokecloud. Powell et al. and Zinn et al.1, 2 used forwardscattering at two angles to calculate average smokeparticle size, but this method is limited to low concen-trations of smoke where single scattering can be as-sumed. Chippett and Gray3 used a monochromator tomeasure transmission of light through smoke for the0.4-0.8-Am wavelength range. They compared the datawith size distributions of collected particles to deter-mine the complex refractive index of the smoke parti-cles.

The present paper describes simple, portable in-strumentation consisting of a white light source and asmall, self-contained, three-wavelength detector as-sembly which has been developed to measure thetransmission of light through a smoke cloud. Thisoptical technique provides continuous, in situ trans-

The authors are with Pittsburgh Mining & Safety Research Center,Bureau of Mines, U.S. Department of the Interior, Pittsburgh,Pennsylvania 15213.

Received 8 August 1978.

mission measurements at three wavelengths (0.45 um,0.63 ,um, and 1.00 gim) without interfering with thesmoke cloud. Average smoke particle sizes and massconcentrations are then calculated using Mie theory.As will be described later in this paper, this method canbe used even for dense concentrations where multiplescattering is in effect. This three-wavelength opticaltechnique was used to study the smoke generated in twofires set in a wood-lined tunnel; the results were com-pared with data obtained using other sizing methods.This paper is an expanded version of an earlier pre-sentation.4

Instrument Design

The measurement of optical transmission at threewavelengths is achieved by use of a tungsten-filamentwhite light source and a three-channel detector as-sembly. A schematic of the detector is shown in Fig. 1.Light from the tungsten source travels through thesmoke cloud to the detector. When the attenuatedbeam enters the detector assembly, it is split into threeparts, each of which passes through an interference filterto a corresponding silicon photodiode. The filters arecentered at wavelengths of 0.45 ,im, 0.63 Aim, and 1.00gm with nominal bandwidths of 0.01 Am each. Theoutput of each photodiode is fed directly into an oper-ational amplifier, thereby providing a linear output overseveral orders of magnitude of input light. The entirethree-channel detector assembly shown in Fig. 1 ismounted in an aluminum housing 8 cm by 10 cm by 5cm.

Theory

The transmission of light through a cloud of uniformparticles is given by Bouguer's law:

T = exp(-QAnL) = exp[-(3QCmL/2pd)b, (1)

1 June 1979 / Vol. 18, No. 11 / APPLIED OPTICS 1763

where T is the fraction of light transmitted, Q is thedimensionless extinction coefficient, A is the cross-sectional area of a particle, n is the number concentra-tion of particles, L is the path length, Cm is the massconcentration of particles, p is the density of an indi-vidual particle, and d is the particle diameter. Theextinction coefficient Q includes loss of light due to bothabsorption and scattering. Using Mie scatteringtheory 5 6 for single spherical particles, Q can be calcu-lated as a function of particle size, wavelength of light,and complex refractive index of the particle. Eventhough the Q values are calculated on the basis of singlescattering, it is valid to use these values in Bouguer'stransmission law for the multiple-scattering case at highconcentrations of particles. The law breaks down onlywhen particles are spaced closer than a few particle di-ameters and interact electromagnetically, thus changingthe Q values. 7 However, Bouguer's law is only valid ifthe transmission measurements are made properly; thefield of view of the detector must be narrow enough sothat the amount of scattered light received by the de-tector is insignificant. Hodkinson7 suggested that thedetector half-angle field of view be limited to less thanone-tenth the angle of the first angular minimum in theFraunhofer diffraction pattern:

01/2 < 7.OX/d degrees, (2)

where 01/2 is the half-angle field of view, X is the wave-length, and d is the particle diameter. As describedlater in the paper, this condition is easily satisfied by thethree-wavelength detector.

For a polydisperse system of particles, Dobbins8 re-vised Bouguer's transmission law:

T = exp [(3QCmL/2pd 3 2)], (3)

where Q is an average extinction coefficient and d3 2 isthe volume-to-surface mean particle diameter. Theyare defined as

Q= _Q(d)N(d)d2

dd/_N(d)d2 Ad, (4)

d3 2 = E N(d)d 3Ad/ZN(d)d 2 Ad, (5)

where N(d) is the number of particles with diametersbetween d and d + Ad. Dobbins8 found that Q de-pends primarily on the mean diameter d32 and not very

D f Photodetector

XI tSSlnterference filter

Beam from flight sourc_ X

through smoke

Window Cubebearnsplitter

C7Fig. 1. Schematic diagram showing optical configuration of the

three-wavelength smoke detector.

much on the exact shape of the size distribution func-tion N(d). The calculations of Q for this paper weremade for several log-normal size distributions withdifferent geometric standard deviations g based onprevious studies3 9 which showed that smoke particlesapproximate a log-normal size distribution.

Since the complex refractive index m of wood smokeparticles is not known exactly, calculations of the ex-tinction coefficients Q were made for several values ofm. The first, ml = 1.95-0.66i, is the complex refractiveindex of carbon measured by Senftleben and Benedict.10

The second, m2 = 1.8-0.6i, is an average of the index forcarbon and the index (1.56-0.48i) for propane andacetylene soots measured by Dalzell and Sarofim.11The third, m3 = 1.8-0.3i, is similar to the value(1.87-0.19i) derived by Kunitomo and Sato12 for varioussoots and to the value (1.9-0.35i) measured by Chippettand Gray3 for propane and acetylene soots. Since thesemeasured refractive indices showed little variation withwavelength in the visible and near infrared, they wereassumed to be constant at the three wavelengths (0.45gim, 0.63 gim, and 1.00 gim) for the Mie calculations.

The calculated mean extinction coefficients for thethree wavelengths and for the three refractive indicesand two size distributions are shown in Figs. 2, 4, 6, 8,and 10. The size distributions are log-normal withgeometric standard deviations ug = 1.5 or 2.0. Sincethe three-wavelength detector measures the transmis-sion for all three wavelengths over an identical pathlength through the smoke cloud, the ratio of the loga-rithms of the transmissions [see Eq. (3)] at any twowavelengths is equal to the ratio of the calculated ex-tinction coefficients for the same wavelengths:

lnT(X1 )/lnT(X 2 ) = Q(Xi,d32)/Q(X2,d32). (6)

In Figs. 3, 5, 7, 9, and 11, the calculated extinction ratiosare graphed as a function of mean particle diameter d32.Using these graphs and the measured log-transmissionratios, the mean particle diameter (d32) of the observedsmoke can be determined. Due to the nonlinear natureof the Q-ratio versus d32 curves, it is better to use allthree log-transmission ratios even though only two ofthem are independent. If the refractive index (m) andsize distribution width (g) are correct, the three valuesof d32 (from the corresponding three log-transmissionratios) will be consistent. If the three values of d32 areinconsistent, either the size distribution width or therefractive index or both must be varied. However, sincethere are only two independent log-transmission ratios,an exact solution can be found for only two of the fol-lowing values: particle size, size distribution width, andreal and imaginary parts of the refractive index. Thereis general agreement on a value of 1.8 to 1.9 for the realpart of the refractive index. Therefore, the particle sizeand either the size distribution width or the imaginaryindex can be determined. In the present paper severalreasonable values of the distribution width and theimaginary index are considered in the graphs in Figs.3, 5, 7, 9, and 11, and a determination is made as towhich values are consistent with the data from the woodsmoke. A more detailed computer analysis of the data

1764 APPLIED OPTICS / Vol. 18, No. 11 / 1 June 1979

150

01,_

I.1LI

.1111

I;r

I

AVERAGE PARTICLE DIAMETER d,,, pm

Fig. 2. Average extinction coefficients for three wavelengths as afunction of average particle diameter for a log-normal size distribution

with org = 1.5 and for ml = 1.95-0.66i.

1.25

.25 I_

0 0.2 0.4 0.6 08AVERAGE PARTICLE DIAMETER d, 2,pm

1.0 1.2

Fig. 3. Average extinction coefficient ratios for three wavelengthsas a function of average particle diameter for a log-normal size dis-

tribution with o-g = 1.5 and for ml = 1.95-0.66i.

04 06 08AVERAGE PARTICLE DIAMETER d, 2,pm

1.25

1.00

.75

50

25

0 2 0.4 06 0.8

AVERAGE PARTICLE DIAMETER d 2, m10 12

Fig. 5. Average extinction coefficient ratios for three wavelengthsas a function of average particle diameter for a log-normal size dis-

tribution with og = 1.5 and for m2 = 1.8-0.6i.

0.4 06 08AVERAGE PARTICLE DIAMETER d2 2, pm

Fig. 6. Average extinction coefficients for three wavelengths as afunction of average particle diameter for a log-normal size distribution

with ag = 2.0 and for m2 = 1.8-0.6i.

1.50

1.251

I1.00

I-

. .75

50o

. .50

02 0.4 0.6 0.8

AVERAGE PARTICLE DIAMETER d 32 ,pm1.0 1.2

Fig. 4. Average extinction coefficients for three wavelengths as a Fig. 7. Average extinction coefficient ratios for three wavelengthsfunction of average particle diameter for a log-normal size distribution as a function of average particle diameter for a log-normal size dis-

with tg = 1.5 and for m2 = 1.8-0.6i. tribution with og = 2.0 and for m2 = 1.8-0.6i.

1 June 1979 / Vol. 18, No. 11 / APPLIED OPTICS 1765

I

0

) - z ~~~ ~ ~~~~~KEY_s _ _, ~~~~~~~m 80 -60 i

. --- {:j~~~~~~~~~~(X63)/5(X.45)> - {~~~~~~~~~~~~~~~5(X .00)/ a X 063) -

1,( 00) /O( X 0 45)

l l l l l l l l l l~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

1.50

l-

I 1.00

u 75

l- .50

U-~

I I I I II I I I

\ II/ I/

~~~~~~~~~ ,

//'/ KEY

= 1.5I- I I m= 1.95 -. 66 i

'--S'~~~~~~~~~~~~ - A,, sxn .1 , 7, xr 4

O(X l.00)/Q IX 0.631- ---- Q(X 1.00)/ Q(X 0.45)

I , I I , I I

10

l-

I

.10

'm

x

I I I I I I I

_ ~ ~ ~~ _

-,' KEY _

- m (g.?u

-- -5X063)/5(X0.45)R5 (X 0.)/00I/X 0.63)

- ---- (X .00I/(X0.45)

I , I , I

:: _-

3.0

L2.0

0/

C 1.5

_ 1.0 KEY

/ / m ~~~~~~~~~~~~~~1.80-0.30)5 5 // 0-xO4

5pfim,5 / -063,ml

I,// x I

0 02 04 06 0.8 1.0 1.2AVERAGE PARTICLE DIAMETER d,, pm

Fig. 8. Average extinction coefficients for three wavelengths as afunction of average particle diameter for a log-normal size distribution

with ag = 1.5 and for m3 = 1.8-0.3i.

1.50

1.25_

1.00

U .75

50 ~~~~~~~~~~~KEYMxI.80-Q301

-a ~XC63/6IX0.45))l-I, 1.00/ 0.61

--- OIUXI.00I/0(Xo45)

0 02 0.4 0.6 0.8 1.0 12AVERAGE PARTICLE DIAMETER d32,pm

Fig. 9. Average extinction coefficient ratios for three wavelengthsas a function of average particle diameter for a log-normal size dis-

tribution with rg = 1.5 and for m3 = 1.8-0.3i.

,2.0 //_10-°I-5 // /K

Z 1.0 K//EY

LU2.0 /~~~~~~~~~~~~~~~~,-.

- /63,m

0 012 04 06 8 1.0 1.2AVERAGE PARTICLE DIAMETER d, Hm

Fig. 10. Average extinction coefficients for three wavelengths as afunction of average particle diameter for a log-normal size distribution

with Ig = 2.0 and for m3 = 1.8-0.3i.

I.:

I

I 5

Iwil-lw0Z

I11

I

11�

02 0.4 06 0.8AVERAGE PARTICLE DIAMETER d,,sm

1.0 12

Fig. 11. Average extinction coefficient ratios for three wavelengthsas a function of average particle diameter for a log-normal size dis-

tribution with ag = 2.0 and for m3 = 1.8-0.3i.

To exhaustfan

Damper control

Exit gate Water

0.3 in out

03 41

0.5 1.2I IL I

I f Cooling sectionJ(water-cooledheat exchanger)

3x detector-position

61 m loroof

ZFirebrickinsulation

Entrance gate for 0.15m diametertest section stock

ng wood lining(3.2cm thick) Exhaust gate forand two side walls n =, Ignition section

/Propagating fire O91m long 4-wallwood lining0.2m diameter

duct28mx0.28m, Air

:r U Vane7.0 1.8 | 1.8 _ anemometer

Test section / I Ignition section1 Intake air flowZMarinite lining control section(1.9cm thick)on floor

Fig. 12. Vertical-section view ofthe fire tunnel (Note: dimensions

are in meters).

1766 APPLIED OPTICS / Vol. 18, No. 11 / 1 June 1979

50 -I KEY -

'g 2.0" - m1.80-0.30i

' -' -- -a(X0.63)/I(X0.45)25 U O(X .00)/0a(X 0.63) -

(X l0 0) (X 0 45)

I , I , I , I , I

/ I I .. - *1

.!

.1

I

I

using numerical inversion techniques could also bedone, but transmission data at several additionalwavelengths would be required for an exact solution.The attempt in the present paper is only to determineapproximate smoke particle sizes by using relativelysimple instrumentation.

Knowledge of the exact path length of the lightthrough the smoke is not necessary in the above deter-mination of particle size. However, if the path lengthis known and the density of a smoke particle can be es-timated, Eq. (3) can be used to calculate the mass con-centration of the smoke.

Fire Tunnel Experimental Apparatus

To test its applicability and reliability, the three-wavelength optical transmission technique was used tostudy the smoke from two wood-lined tunnel fire ex-periments. The present paper will discuss only theresults of the three-wavelength particle-sizing techniqueand comparisons with two other particle-sizing meth-ods. Details on the significance of the smoke mea-surements in relation to other fire data from thewood-tunnel fires are reported elsewhere. 13 Details ofthe tunnel facility and instrumentation were reportedpreviously. 14 The tunnel, shown in Fig. 12, was 0.3 msquare in cross section and 10.0 m long with a 1.8-mignition section and a 7.0-m test section. Wood (3 cmthick) lined half the ignition section and the first 6.1 mof the test section. Figure 13 is a detailed horizontal-section view of part of the fire tunnel showing the whitelight source and the three-wavelength detector at the6.3-m position in the test section, 0.2 m past the end ofthe wood lining. Both the light source and detectorwere at the ends of 33-cm-long by 1.1-cm-diametertubes which extended into the tunnel. The tubes wereblackened on the inside to prevent reflected light fromreaching the detector. The half-angle field of view ofthe three-wavelength detector was about 0.9 degree,thus eliminating any significant contribution of lightscattered from the smoke particles [see Eq. (2)]. Thework of Deepak and Box15 also confirms the fact thatthe correction for forward-scattered light is negligiblefor our detector. The outer ends of the tubes weresealed at the light source and detector; the resultingdead air spaces in the tubes, combined with the gas flowin the tunnel, prevented any appreciable amount ofsmoke from entering the tubes. The path lengththrough the smoke was therefore assumed to be the 10cm between the open ends of the tubes. The lightsource was a 100-W, halogen-filled, tungsten-filamentlamp with a stabilized power supply. The light fromthe lamp could be blocked off during the experiment toverify that radiation from the hot smoke particles orinterior walls of the tunnel was insignificant.

Wood-Tunnel-Fire Smoke Data and Analysis

Figure 14 shows the optical transmission measuredby the three-wavelength detector for the first wood-tunnel fire. Note that the transmission dropped rap-idly as the fire built up in the tunnel and then remainedfairly constant after the fire was fully developed. Fig-

Fig. 13. Horizontal-section view of part of the fire tunnel showingposition of light source and three-wavelength (3X) detector.

1.0

X LOOP~

t- .8

aU, )~~~~~~~~~~X.063pjm.6

.2

0 2 4 6 8 10 12 14

TIMEt, mI

Fig. 14. Optical transmission measured at three wavelengths duringa wood-tunnel fire.

8

6

4L

2

0 2 4 6 8 10 12 14l

TIME t, mi

Fig. 15. Optical density per unit length for three wavelengths duringa wood-tunnel fire.

ure 15 shows the optical density per unit length DIL =

-(log1 OT)/L calculated from the transmission valuesof Fig. 14. For visible light (X = 0.55 gm), the opticaldensity per unit length ranged from 3 m-1 to 5 m- 1 afterthe fire was fully developed, indicating a very densesmoke cloud.

During the period of relatively constant transmissionfrom 3 min to 7 min after the start of the fire, the mea-sured log-transmission ratios were nT(X1.00)/lnT(X0.45) = 0.25, lnT(X1.00)/lnT(X0.63) = 0.47, andlnT(X0.63)/lnT(X0.45) = 0.54. Using Eq. (4) to relatethese values to the extinction coefficient ratios, themean particle size can be determined from the curvesof Figs. 3, 5, 7, 9, or 11. For most of the graphs, themeasured points fall below the lowest values of thecalculated curves. Figure 9, for m3 = 1.8-0.3i and a

1 June 1979 / Vol. 18, No. 1 1 / APPLIED OPTICS 1767

0 0.2. 1 1

Scale,,Am

Fig. 16. Electron-microscope photograph of smoke particles froma wood-tunnel fire.

log-normal distribution with 0g = 1.5, is the only graphthat gives a self-consistent result of d3 2 = 0.12 + 0.02 gmfrom all three ratios. During the later stages of the fire,the particles were about the same size. During the earlystage of the fire, the rapid change in transmission as thefire built up prevented an accurate determination ofparticle size, but the best estimate is that the averageparticle size was about one-half as large as that duringthe time of steady-state burning.

During the same wood-tunnel fire, smoke particleswere collected on grids at the 4.0-m position in the testsection for examination by transmission electron mi-croscopy (TEM). A representative TEM photographof smoke particles collected 9 min after the start of thefire is shown in Fig. 16. The individual particles werereasonably circular, although some agglomeration oc-curred owing to particle accumulation on the grid.These spherical particles were in contrast to the irreg-ular soot agglomerates observed by Chippett and Gray.3

The difference may be related to the difference in fuels(wood versus propane) or to the high temperature(8000C) in the fire tunnel compared to the low tem-perature at which they collected the soot agglomerates.Other TEM photographs of particles collected at dif-ferent times in the fire showed particles similar to thosein Fig. 16. These observed, spherical particles offersome justification for the assumption of spherical par-ticles in the Mie calculations. In addition, Hodkinson 7

reports that the extinction coefficients () for non-spherical particles differ only slightly from the valuescalculated for spherical particles. The TEM particleshad diameters from 0.04 gim to 0.10 ,um. There wereinsufficient particles on the grid to determine an accu-rate mean size, but most are somewhat smaller than thevolume-to-surface mean diameter (d32 = 0.12 gm) de-duced from the three-wavelength optical technique.This may have been due to agglomeration between theTEM sampling point and the three-wavelength opticalmeasurement point farther down the tunnel. In sum-mary, there appeared to be reasonably good agreementbetween these two methods.

Once the mean particle size d32 had been determined,

the mass concentration Cm could be calculated from Eq.(3), assuming a p-value of 1.5 g/cM3 for the density ofthe individual smoke particles. The range of the cal-culated concentrations was 1-2 mg/liter duringsteady-state burning. The uncertainty would be largerfor Cm than for d32 because of the uncertainties in thevalues of p and Q in addition to the error in d32.

During the second wood-tunnel fire, the three-wavelength optical size determination technique wascompared to size analyses from an ionization-typeparticulate detector developed by Litton et al.16,17 Thisionization-type detector measured the size distributionof smoke particles based on the mobility of chargedparticles. Both detectors observed the smoke at the6.3-m position in the test section of the fire tunnel.During the early stage of the fire, both methods gave avolume-to-surface mean particle size between 0.15 gmand 0.20 gm with an uncertainty of about 0.05 gm. Atthis time the transmissions at all three wavelengths werebetween 90% and 97%. The ionization-type detectoralso gave a size distribution that was roughly log-normalwith a geometric standard deviation g of 1.9. This waslarger than the best fit for the three-wavelength detectordata, which was a log-normal distribution with og = 1.5and m3 = 1.8-0.3i. The differences may be related tothe fact that the measured distribution was only roughlylog-normal.

After the fire became fully developed, the transmis-sion dropped to 13% at X = 0.45 gim, and the meanparticle size was d32 = 0.35 0.10 gim, based on thethree-wavelength detector data. Again the best fit wasfor m3 = 1.8-0.3i and g = 1.5, although the calculatedmean diameters based on the other refractive indiceswere mostly within the listed uncertainty in d 32. Themeasured data points did not fit well to the broader sizedistributions with g = 2.0 for any of the refractive in-dices. The ionization-type detector measured a meanparticle size d32 = 0.4 0.1 m and a distribution withrg = 2.4 during the same time as the three-wavelength

detector measurements. The mean particle sizes ob-tained by both methods agree to within the uncertain-ties, but the ionization-type detector measured abroader distribution. Again, this may be due to the factthat the distribution was only approximately log-normalor to inherent errors in both measurements. In general,however, there was good agreement between thethree-wavelength detector and the ionization-typeparticulate detector for the smoke particle size mea-surements. Based on the three-wavelength detectordata, the maximum concentration during this secondfire was about 3 mg/liter.

Conclusion

The three-wavelength optical transmission techniquedescribed in this paper appears to be a feasible methodfor in situ monitoring of the sizes and concentrationsof smoke particles generated in fires. Reasonably goodagreement was found between the three-wavelengthtechnique and the two other sizing methods used duringsimultaneous observations of the smoke from two ex-periments with wood-lined-tunnel fires. In addition,

1768 APPLIED OPTICS / Vol. 18, No. 11 / 1 June 1979

the measured smoke particle diameters were compa-rable to those observed by other investigators studyingthe flaming combustion of wood.9 "18

The advantages of the present three-wavelengthoptical technique are the compact portable instru-mentation and the fact that very dense smoke cloudscan be monitored. In addition to the determination ofmean particle size, some information on the approxi-mate refractive index and width of the size distributioncan also be obtained. This technique may also be ap-plicable to other aerosols in addition to the wood smokeobserved in the present experiments.

The authors wish to thank C. D. Litton and T. P.Weldon of the Pittsburgh Mining and Safety ResearchCenter (PMSRC) and Z. J. Fink, formerly of PMSRC,for their assistance in the development of the computerprograms for the calculation of the Mie extinctioncoefficients.

References1. E. A. Powell, R. A. Cassanova, C. P. Bankston, and B. T. Zinn,

AIAA Paper No. 76-67, presented at the 14th AIAA AerospaceSciences Meeting, Washington, D.C. (January 1976).

2. B. T. Zinn, E. A. Powell, R. A. Cassanova, and C. P. Bankston, FireResearch 1, 23 (1977).

3. S. Chippett and W. A. Gray, Combust. Flame 31, 149 (1978).4. K. L. Cashdollar, C. K. Lee, and J. M. Singer, "Smoke Particle

Size and Concentration Measurements by a Three-WavelengthTransmission Technique," presented at the Fall TechnicalMeeting of the Eastern Section of the Combustion Institute, heldin Hartford, Connecticut, 10-11 Nov. 1977.

5. H. C. Van de Hulst, Light Scattering by Small Particles (Wiley,New York, 1957), Chap. 9.

6. M. Kerker, The Scattering of Light and Other ElectromagneticRadiation (Academic, New York, 1969), Chaps. 3 and 4.

7. J. R. Hodkinson, in Aerosol Science, C. N. Davies, Ed. (Academic,New York, 1966), Chap. 10, pp. 290-297.

8. R. A. Dobbins and G. S. Jizmagian, J. Opt. Soc. Am. 56, 1345(1966).

9. W. W. Foster, "The Size of Wood Smoke Particles," in Aerody-namic Capture of Particles, E. G. Richardson, Ed. (PergamonPress, New York, 1960).

10. H. Senftleben and E. Benedict, Ann. Phys. Leipzig 54, 65(1918).

11. W. H. Dalzell and A. F. Sarofim, J. Heat Transfer 91, 100

(1969).12. T. Kunitomo and T. Sato, Bull. Japan. Soc. Mech. Eng. 14, 58

(1971).13. C. K. Lee, J. M. Singer, and K. L. Cashdollar, Fire Materials 2,

110 (1978).14. R. F. Chaiken and J. M. Singer, "Experimental Coal Mine Fire

Research," Fourth International Symposium on CombustionProcesses held in Czestochawa, Poland, September 1975, pub-lished Polish in Arch. Termodynamiki Spalania 7, 529 (1976).

15. A. Deepak and M. A. Box, Appl. Opt. 17, 3169 (1978).16. C. D. Litton and M. Hertzberg, "Principles of Ionization Smoke

Detection. Development of a New Sensor for Combustion-Generated Submicrometer Particulates," U.S. Bureau of MinesReport of Investigations 8242 (1977).

17. M. Hertzberg, C. D. Litton, and R. Garloff (assigned to U.S. Dept.of Interior), "Sub-Micron Particle Detector," U.S. Patent4,053,776, 11 October 1977.

18. C. P. Bankston, E. A. Powell, R. A. Cassanova, and B. T. Zinn, J.Fire Flammability 8, 395 (1977).

Information about future meetings should be sent tothe Managing Editor, P. R. WAKELING, WINC,

1613 Nineteenth Street N. W., Washington, D. C. 20009

Feedback Control of Exposure Geometry in DentalRadiography Workshop, University of Connecticut,16 May 1978

Reported by Richard Gordon, University of Manitoba

Stereotactic methods are used in surgery when the precise locationof damaged or pathological tissue is known in three dimensions. Forexample, 3-D information is obtained by holding the patient's headin a rig, using rods stuck in the ears, while taking x-ray pictures. Infeedback-controlled stereotactic radiology a computer-controlledx-ray apparatus can compensate for changes in the orientation of thepatient. Feedback control in radiography thus achieves stereotacticimaging without bolting the patient to the apparatus. If instrumentswith such capabilities were built, stereotactic imaging could becomea more acceptable and more widely used diagnostic method.

Beyond facilitating precise surgery, stereotactic imaging allows oneto follow the time course of a disease or the motions of an organ in aquantitative fashion. Furthermore, diagnosis can be improved viaquantitative image processing which takes advantage of the time se-quence. The x-ray dose to the patient can also be greatly reduced.One can even conceive of stereotactic surgery or dentistry carried outunder direct computer control.

This workshop was convened in May last year at the University ofConnecticut School of Dental Medicine to explore the possibility ofusing feedback control to improve dental radiography, before anyapparatus are actually built. It was organized by Richard Webberof the National Institute of Dental Research (National Institutes ofHealth, Bethesda, Maryland) on behalf of the Technology andEquipment Committee of the American Academy of Dental Radi-ology.

A number of dose reduction schemes were suggested during theworkshop. If one can aim an x-ray beam, one can avoid tissues suchas bone marrow, which has proliferating cells, Such tissues have ahigher probability of becoming cancerous from radiation damage(Allen B. Reiskin, University of Connecticut). One can also avoiddental fillings, which over their area of the image contribute nothingto a diagnosis, since they are nearly x-ray opaque. A more refinedversion of such an approach uses a higher dose for better imagery onlyin those areas suspected of containing caries or periodontal disease(Roger Nagel, NIDR). Most of these methods are object dependent,require the use of a priori information, and thus involve highly non-linear computational algorithms.

Under computer control the x-ray beam may be confined to a vol-ume that strictly intersects the detector, so that no radiation is wasted(Richard Webber, NIDR). A simple computer search algorithmcould allow an intraoral detector to be precisely locatedin the mouth.If a detector array is used, the primary beam may be distinguishedfrom scattered radiation. A multilayered detector can perform arough energy discrimination.

An aimed beam allows one to attain a uniform signal-to-noise ratioby holding the beam at each point for an appropriate length of time.If the peak of the x-ray spectrum could also be dynamically tuned tothe optimum energy for each local optical thickness, further dosereduction could be achieved.

There are two roles for image processing in stereotactic radiologyin dentistry. The first is as an aid in the acquisition of an image. Thesecond permits one to follow the time course of a disease.

continued on page 1834

1 June 1979 / Vol. 18, No. 11 / APPLIED OPTICS 1769