High-resolution spectroscopy temperature measurements inlaminar channel flows
Anouar Soufiani and Jean Taine
A gas temperature measurement technique, based on the inversion of the spectral shape of a vibration-rotation absorption line, is presented. A tunable diode laser system is used to record the fully resolvedspectral shape. A precise calibration of the method is given in the 300-900 K temperature range. Themethod is used in channel laminar air flows in which temperature is the only unknown quantity. Theintrinsic uncertainties of this technique are estimated to 2% at 300 K and +3% at 900 K. Good agreementbetween the diode laser measured profiles and theoretical predictions is obtained.
I. Introduction
Measurements of temperature and species concen-tration are often required in combustion investigationsand heat transfer studies. Several nonintrusive- tech-niques have been recently developed; they includeCARS spectroscopy,' laser-induced fluorescence,2
Rayleigh scattering,3 two-line absorption,4 etc.High-resolution spectroscopy techniques have been
previously used to detect trace gases,5 or for concentra-tion6 7 and temperature7 8 measurements in combus-tion media. Temperature was deduced from a two-absorption line intensity ratio; this method ispracticable at high temperature. The temperaturemeasurement technique used here is based on the tem-perature dependence of the Lorentz halfwidth of avibration-rotation absorption line. A diode laser sys-tem is used to fully resolve the spectral line shape.The method is applied in this study to homogeneouschannel flows in which temperature is the only un-known quantity. This application enables an evalua-tion of the intrinsic precision of the method.
The basis of the method and the choice of the ab-sorbing species are explained in Sec. II. A precisecalibration in thermally stabilized cells is given in Sec.III. The apparatus and some results are then shown inSecs. IV and V, respectively.
The authors are with CNRS/ECP Laboratoire d'Energetique Mo-leculaire et Macroscopique, Combustion, Ecole Centrale des Arts etManufactures, Grande Voie des Vignes, 92295 Chatenay-MalabryCEDEX, France.
The temperature measurement technique devel-oped here is based on the inversion of the spectralshape of an infrared absorption line of a molecularspecies. The parameters describing the absorptionline are the integrated intensity So (T) and the half-width (HWHM) y. Measurements of these two pa-rameters enable the determination of temperature andmolar fraction xi of the absorbing species i. The spec-tral transmissivity -r, of homogeneous and isothermalgaseous column of length is
T, = exp(-K, 1)(1
where K, is the spectral absorption coefficient. For anisolated line of species i centered at the wavenumbervo, K, is given by
K = xPSo(T)F(v - ). (2)
The normalized line shape F(v - vo) depends on totalpressure P, molar fractions xj of all the species, andabsolute temperature T. For combined collisionaland Doppler broadening effects, the line shape is aVoigt one9 :
F(v - o)= = J: I P((Y2 ) dy, (3)
with a = JIi(YL/YD) and = n2(v-vo)/YD, where theDoppler halfwidth YD is a known function of tempera-ture and the Lorentz halfwidth L depends on tem-perature and partial pressures xjP:
YL = Z xj(P/Ps)YLij(T)(TT)flpi. (4)species j
Here YLij(Ts) is the Lorentz halfwidth of the absorbingline of species i diluted in species j under reference
temperature and total pressure conditions (T8 = 300 K,P8 = 1 atm). ij is the line temperature dependencecoefficient, generally included in the 0.5-1 range. Forknown molar fractions xj ( = i included) and parame-ters YLi](T8) and Al, measurement of YL enables a tem-perature determination from Eq. (4).
The absorbing species i may be one of the majorflowing gases or just present as a trace gas in themixture (xi < 0.01). In the latter case, the term xi(P/Ps)ALii(Ts)(T/T)lii in Eq. (4) is negligible in compari-sion with the other terms and a precise knowledge ofthe molar fraction xi is not required for temperaturedeterminations. This procedure has been used in ourexperiments. The flowing mixture to be studied isseeded with a suitably chosen absorbing species at avery low molar fraction. This results in a negligiblechange of the thermophysical properties of the mix-ture studied.
The particular absorption line used is the P(4) line ofthe fundamental band of carbon monoxide, centeredat v = 2127.693 cm-'. The choice of CO as the absorb-ing species is made for several reasons: (i) CO is asimple molecule and its high-resolution infrared spec-tra have been extensively studied theoretically1 0-16and experimentally.14-24 (ii) As the CO rotational con-stant is sufficiently large (B, - 1.92 cm-'), the absorp-tion lines are well resolved; there are no overlappingeffects to be accounted for in the data reduction proce-dures. (iii) The intensity of the CO fundamental bandis sufficiently small (260 cm-2 atm-1 at STP25); heattransfer in transparent gases can be studied withoutadding radiation effects. (iv) Infrared active mole-cules such as CO2 and H2 0 are transparent for thechosen central wavenumber r'o; the method can beeasily used in the presence of these molecules.
III. Calibration
A precise determination of temperature dependencecoefficients sBCoj and room temperature halfwidths'YLCO-j(TS) is required before proceeding to tempera-ture measurements. As these two parameters dependon the rotational quantum number J, the calibrationmust be carried out for the particular absorption lineto be used in the temperature measurements. To ourknowledge, there were no systematic measurements ofcollisional broadening of a CO line at high temperaturewith C02, H20, N2, and 02 as collision partners. Lo-rentz halfwidths of the P(4) line of the CO fundamen-tal band have been measured in the 293-800 K tem-perature range for CO2
16 and N2 partners and in the446-852 K range in the case of H2 0.15 The opticalsetup and data reduction procedures used for the ex-perimental determinations are the same as in Refs. 14-16 and will not be detailed here. The thermally stabi-lized cell and its environment are described in Ref. 26.
Experimental Lorentz halfwidths for CO-N2, CO-CO2, and CO-H20 broadening of the P(4) line areshown in Fig. 1 vs temperature. The error bars indi-cate standard deviations observed in the measuredvalues when the total pressure of the gas sample wasvaried over the 0.2-1.2-atm range (an experimental
0.08 -
F'
0.06 F
.04
1
r,
CO-N2 1<ss
------ T(K)
300 400 500 oo 700 800 900
Fig. 1. Broadening of the 1-0 P(4) line of CO diluted in N2 , CO2 ,and H20: I, experimental data; , best least-squares fit of the
measured halfwidths by a power law [Eq. (5)].
point is typically deduced from ten experiments atdifferent pressures). The uncertainties are estimatedto +1% at 300 K and +2% at 850K. The best fit of theexperimental results by the temperature dependencepower law,
yco-j(T) = yco-j(Ts)(Ts/T)fCOi, (5)
is also shown in Fig. 1. The parameters yCoj(Ts) andflco-< deduced from this best least-squares fit are givenin Table I.
For temperature measurements in air, the air-broadened halfwidth is calculated from nitrogen- andoxygen-broadened halfwidths assuming the relation
'YCO-air = 797CO-N + 0.2 1-YCO-0 (6)
Here co-o2 is approximated in the 300-1000 K tem-perature range by
ycto -o = 0.0606(300/T) 0 " cm-' atm 1 (7)
according to the theoretical study of Bouanich12 andthe low or room temperature measurements of Na-kazawa and Tanaka.2 2 27
IV. Experimental Facility and Procedures
The diode laser temperature measurement tech-nique is used for simple air flows with well-knownboundary conditions to validate the apparatus anddata reduction procedures. A horizontal wind tunnelwith a rectangular cross section is used (200 cm long, 15cm wide, and 3 cm high). The upper and lower walls ofthe channel are maintained at constant but different
Table . Temperature Dependence Coefficients for the 1-0 P(4) CO LineBroadened by N2, H20, and CO2 [to be Used with Eq. (5)]
temperatures. The first is heated electrically and sta-bilized at a temperature between 400 and 900 K; thesecond is held to -373 K using a liquid-vapor waterphase change. The channel end is open to the atmo-sphere, all the studied flows are then at the atmospher-ic pressure which is measured during each experiment.Four sapphire windows enable optical measurementsat two cross sections of the channel 120 cm away fromeach other. More details on the wind tunnel and ther-mal stabilization of wall temperatures are given in Ref.28.
Figure 2 shows the optical arrangement used fortemperature measurements. The diode laser system(Spectra-Physics) uses a lead-salt diode workingaround 2128 cm-'. The spectral width of the diodeline is lower than 10-3 cm-', and a deconvolution pro-cedure is not necessary for the Lorentz lines consid-ered here. The laser beam is mechanically choppedand enters a grating monochromator, with typically0.5-cm-1 spectral resolution, to allow selection of asingle emission mode. The parallel beam from the exitlens of the monochromator is split in three parts byCaF2 beam splitters (BS2 and BS3). The first beam(B1) is directly focused on the infrared detector D1; itpermits recording of variations of the laser radiationintensity. The second beam (B2) goes through a refer-ence cell containing a CO-N2 etalon mixture at roomtemperature. The signal delivered by detector D2leads to the determination of radiation frequency vari-ations from the diode current. Two diaphragms (01and 02) of 1-mm diameter are then used to limit thespatial resolution of the measurements. The laserbeam from 02 is split into two equal parts (beam split-ter BS4, CaF2) to sound the flow at the two crosssections of the channel. Measurements at the up-stream section (B3) are used as inlet conditions in flowcalculations; results of these calculations are then com-pared to measurements at the downstream section(B4). Two motorized units (MU1 and MU2), includ-ing step-by-step motors (microcontrol) and focusinglenses, allow precise vertical displacements of thebeams, with 1-Atm resolution, and permit temperaturemeasurements at any vertical position between thehorizontal channel walls. Figure 3 shows the details ofone of these units. After crossing the flow, the beamsare collected by similar motorized units (MU3 andMU4) where the focusing lenses are replaced by dia-phragms to cut the radiation emitted by the hot chan-nel. Optical paths around the flow are in vacuum-tight enclosures to avoid beam perturbations due tonatural convection outside the studied flow. BeamsB3 and B4 are finally focused on detectors D3 and D4.Optical alignments are performed using two He-Nelasers (632.8 nm) as shown in Fig. 2. D and D2 arenitrogen-cooled GeAu detectors while D3 and D4 arenitrogen-cooled InSb detectors.
A parallel input/output card of a PDP 11-23 com-puter (Fig. 4) monitors the displacements of the mo-torized units MU1 to MU4 and, occasionally, the rota-tion of the monochromator grating. The four choppedsignals delivered by detectors D1,. . .,D4 are first fil-
Fig. 2. Optical setup for diode laser temperature measurements:M, Au mirror; L, CaF2 lens; BS, CaF2 beam splitter; 0, diaphragm;
D, detector; MU, motorized unit for vertical displacements.
Fig. 3. Motorized unit MU1.
tered in lock-in amplifiers (PAR 5204 or StanfordSR510) before being digitalized together with the di-ode laser current by a 12-bit A-D conversion card ofthe computer. Numerical averaging of the convertedvalues is made over time intervals 1 to 10 times greaterthan the lock-in time constant, depending on the sig-nal-to-noise ratio. The diode current is tuned overtypically 100 mA (0.5 cm-') during each experiment.Some 50 to 100 averaged values of the four signals andof the diode current are synchronously stored by thecomputer.
The spectral absorption coefficient of the consid-ered line is deduced from two experiments. The firstis performed with the flowing mixture containing CO
Fig. 5. Signals S3 and S4 vs the diode current: , with CO;- -- - -- -, without CO.
lock-in preampifiers detectorsamplifiers
Fig. 4. Data acquisition system and motor monitoring.
as a trace gas and the reference cell filled with theetalon CO-N2 mixture. It leads to the values S(C),SE(C), S4(C), 84RC), and C of the signals related to thefour detectors Dl,...,D4 and to the diode currentintensity C. The second experiment is performedwith an empty reference cell and without CO in theflow. The recorded values are then called 8R(C),82NC), SR(C), SR(C), and C. Figure 5 shows typical
records of the signals S3 and S4 with and without CO.The spectral transmissivity r(C) of optical path k oflength 1k (k = 2,3,4) is calculated from the ratio
Frequency variations as functions of diode currenthave been measured using a 3.48-cm germaniumFabry-Perot etalon. Figure 6 shows examples of thesevariations for four measurements in the used range ofthe current C. A linear form of this function is satis-fied within a maximal deviation of 0.002 cm-' in a totalspectral range of 0.6 cm-'. We assume afterward thatthe linear law,
V = ac + 0, (10)
is checked. The loss of accuracy resulting from thisassumption is <0.5% for the measured temperature.But as the coefficient a may depend on the thermody-namic conditions of the lead-salt diode, it is deter-mined in each experiment from linewidth measure-ments in the etalon CO-N2 mixture at roomtemperature (beam B2).
In our experimental conditions (P 1 atm, 300 K <T < 1000 K), the broadening of the absorption line ismainly due to collisions. A Lorentz spectral shape isthen adopted. The theoretical absorption coefficientKh can be written as a function of the current C:
where the Lorentz halfwidth -Yk is expressed in units ofcurrent intensity (mA). The parameters 7Yk and xcoSo(T)/a are deduced from a nonlinear least-squares fitof the measured absorption spectra [Eq. (9)] by thetheoretical Lorentz shape [Eq. (11)] for the three pathsk = 2,3, and 4. The coefficient a is then obtained fromthe measured value 72 (mA) and its theoretical value(cm-') given by Eq. (5) for j = N2 . Figure 7 shows anexample of the measured products Kk1k and the besttheoretical fits for the three signals S2, S3, and S4. Theagreement between theoretical and measured profilesis excellent.
The above procedure used for frequency calibrationresults in a small number of points to be acquired; thesystematic use of a Fabry-Perot etalon6 "14 requires -6times more points to precisely resolve the positions ofthe interference fringes. The required time for a tem-perature measurement is then smaller with our proce-dure.
V. Results and Discussion
To assess the accuracy of the diode laser techniqueand data reduction procedures, measurements havebeen made in an isothermal air flow at 290 K. The
Fig. 9. Spectral line profiles at 1 mm from the channel walls:, experimental profile; - -- -, best fit by a Lorentz profile.
k 3
600
500k 4
T (K)
'I
Q- down-stream. It ;
, '- ,
up-stream
400 -
C(mA)
-0.2 -0.1 0 0.1 v-vo (c m4)
Fig. 7. Typical absorption spectra recorded for the three paths k =2,3, and 4: +, experimental points; ,theoretical Lorentz shape
obtained from a least-squares fit.
310k T (K)
30d-
290
I -* -2 % - .--- -- -- -- -_ ~~~~~~~ I , . * . 2%T
280)-
270 r I . . I ..I y (mm)I . * I
0 6 12 18 24 30
Fig. 8. Measured temperatures at the * upstream and * down-stream sections of the channel. Isothermal air flow at 290 K.
measured temperatures at different vertical positionsin the two optical cross sections are shown in Fig. 8 for asingle experiment. Approximately 90% of the experi-mental points are inside the confidence interval 290 ±6K which represents an uncertainty of ±2%. This factis consistent with the observed dispersion on the mea-sured linewidths at different pressures when the meth-od was calibrated (Sec. III). In fact, a relative uncer-tainty Ay/-y results in a temperature uncertainty AT/T= (1/0Coj)Ay/-y for a flowing species j. This leads tothe uncertainties ±2% at 300 K and ±3% at 900 K.
300Y(mm)
I I . I . I
0 6 12 18 24 30
Fig. 10. Measured and theoretical temperature profiles in the caseof a laminar air flow at a Reynolds number Re = 1950: *, thermo-couple measurements at the downstream section; , diode lasermeasurements at the downstream section; - - - - -, calculated tem-perature profile at the downstream section; A, constant wall tem-peratures; ....... , diode laser measurements at the upstream
section.
Averaging over two or three measurements generallyimproves the accuracy of the method.
It is worth noting that measurements at 1 mm fromthe walls can be easily performed. Figure 9 showstypical spectra recorded near the upper and lowerwalls for a heated laminar flow. The agreement be-tween experimental data and Lorentz profiles remainsexcellent. The spatial resolution in the direction per-pendicular to the propagation direction is reduced bythe use of the diaphragms 01 and 02 of 1-mm diameterand the focusing lenses (Fig. 2). This spatial resolu-tion is -0.3 mm in our experiments.
Measurements have been carried out in laminar bi-dimensional air flows with wall temperatures equal to618 i 4 K and 363 i 4 K for the upper and lower walls,respectively. Thermocouples (Chromel-Alumel, 1-mm diameter) with radiation shields have also beenused to make comparisons with diode laser measure-ments. Figure 10 shows temperature profiles mea-
sured at the downstream section of the channel withthe two techniques. Bar errors for diode laser mea-surements indicate a 2.5% confidence intervalaround averaged values obtained from two experi-ments. Thermocouple measurements lead to up to 35K higher temperatures than diode laser measure-ments. This overestimation is attributed to the upperwall radiation which affects the equilibrium tempera-ture of the probe. Numerical simulations of this ef-fect, including the radiation shields, lead to compara-ble overestimations. 2 8 This is due to the smallconvective heat transfer between the probe and the gasin the case of laminar flows.
Numerical predictions of temperature profile at thedownstream section of the channel are also shown inFig. 10. The diode laser measured profile at the up-stream section is used as an inlet condition in a para-bolic calculation of the flow. Coupled bidimensionalmomentum and energy equations are then solved si-multaneously using the real temperature-dependentthermophysical properties of the fluid. A numericalscheme based on Patankar recommendations 2 9 is used.Details on these theoretical predictions and numericalprocedures are given in Ref. 28. The agreement be-tween high resolution spectroscopic measurementsand the numerical simulation is excellent if we accountfor measurement uncertainties. Other experiments atvarious Reynolds numbers (<2500) lead to similaragreements. 2 8
VI. Concluding Remarks
A diode laser temperature measurement techniquehas been applied to simple laminar channel flows inwhich temperature is uniform along the optical path.A precise calibration of the method has been carriedout. As temperature was the only unknown parameterin this application, the intrinsic accuracy of the meth-od has been studied. Uncertainties are estimated to±2% at 300 K and ±3% at 900 K. The principallimitation of this technique is that it yields an integrat-ed line-of-sight measurement. It can be applied in thecase of isothermal paths or in the case of cylindricalgeometries using mathematical inversion proceduressuch as Abel inversion.20 On the other hand, the timeresolution of the method, which is the time required fora diode laser scan of the line shape, can be reduced to0.1 ms. This enables the study of rapid transientphenomena. Application of the method to laminarflows leads to excellent agreement with theoreticalpredictions. The use of this spectroscopic techniquein the case of turbulent flows must be carefully under-taken owing to spatial temperature fluctuations alongthe path length.
References1. S. A. J. Druet and J. P. E. Taran, "CARS Spectroscopy," Prog.
Quantum Electron. 7, 1 (1981).2. J. D. Bradshaw, N. Omenetto, G. Zizak, J. N. Bower, and J. D.
3. R. K. Cheng, R. G. Bill, and F. Robben, "Experimental Study ofCombustion in a Turbulent Boundary Layer," in Proceedings,Eighteenth International Symposium on Combustion (TheCombustion Institute, Pittsburgh, PA, 1981), p. 1021.
4. A. Y. Chang, E. C. Rea, and R. K. Hanson, "Temperature Mea-surements in Shock Tubes Using a Laser-Based AbsorptionTechnique," Appl. Opt. 26, 885 (1987).
5. D. T. Cassidy and J. Reid, "High-Sensitivity Detection of TraceGases Using Sweep Integration and Tunable Diode Lasers,"Appl. Opt. 21, 2527 (1982).
6. R. K. Hanson, P. A. Kuntz, and C. H. Kruger, "High-ResolutionSpectroscopy of Combustion Gases Using a Tunable IR DiodeLaser," Appl. Opt. 16, 2045 (1977).
7. S. M. Schoenung and R. K. Hanson, "CO and TemperatureMeasurements in a Flat Flame by Laser Absorption Spectrosco-py and Probe Techniques," Combust. Sci. Technol. 24, 227(1981).
8. R. K. Hanson and P. K. Falcone, "Temperature MeasurementTechnique for High-Temperature Gases Using a Tunable DiodeLaser," Appl. Opt. 17, 2477 (1978).
9. S. S. Penner, Quantitative Molecular Spectroscopy and GasEmissivities (Addison-Wesley, Reading, MA, 1959).
10. G. D. T. Tejwani, "Half-Width Computations for Air-Broad-ened CO Lines," J. Quant. Spectrosc. Radiat. Transfer 12, 123(1972).
11. P. Varanasi and G. D. T. Tejwani, "Half-Width Calculations forCO Lines Broadened by C0 2 ," J. Quant. Spectrosc. Radiat.Transfer 11, 255 (1971).
12. J. P. Bouanich, "Largeurs de raies spectrales de CO autoper-turbe et perturbk par N2, 02, H2, HCl, NO et C02," J. Quant.Spectrosc. Radiat. Transfer 13, 953 (1973).
13. J. Bonamy, D. Robert, and C. Boulet, "Simplified Models for theTemperature Dependence of Linewidths at Elevated Tempera-tures and Applications to CO Broadened by Ar and N2," J.Quant. Spectrosc. Radiat. Transfer 31, 23 (1984).
14. J. M. Hartmann, M. Y. Perrin, J. Taine, and L. Rosenmann,"Diode-Laser Measurements and Calculations of CO 1-0 P(4)Line Broadening in the 294- to 765-K Temperature Range," J.Quant. Spectrose. Radiat. Transfer 35, 357 (1986).
15. A. Soufiani and J. M. Hartmann, "Measurements and Calcula-tions of CO-H20 Line-Widths at High Temperature," J. Quant.Spectrosc. Radiat. Transfer 37, 205 (1987).
16. A. Soufiani and L. Rosenmann, "Diode-Laser Measurementsand Calculations of CO-C02 Line-Widths at High Temperature,"J. Quant. Spectrosc. Radiat. Transfer in press.
17. D. A. Draegert and D. Williams, "Collisional Broadening of COAbsorption Lines by Foreign Gases," J. Opt. Soc. Am. 58, 1399(1968).
18. J. N. P. Sun and P. R. Griffiths, "Temperature Dependence ofthe Self-Broadening Coefficients for the Fundamental Band ofCarbon Monoxide," Appl. Opt. 20, 1691 (1981).
19. J. P. Bouanich, R. Farreno, and C. Brodbeck, "Direct Measure-ments of N2 Broadened Line-Widths in the CO Fundamental atLow Temperatures," Can. J. Phys. 61, 192 (1983).
20. P. L. Varghese and R. K. Hanson, "Collision Width Measure-ments of CO in Combustion Gases Using a Tunable Diode La-ser," J. Quant. Spectrosc. Radiat. Trangfer,26, 339 (1981).
21. R. E. Willis, H. C. Walker, and H. S. Lowry II, "CollisionWidths of CO Lines Broadened by Water Vapor at ElevatedTemperatures," J. Quant Spectrosc. Radiat. Transfer 31, 373(1984).
22. T. Nakazawa and M. Tanaka, "Intensities, Half-Widths andShapes of Spectral Lines in the Fundamental Band of CO at LowTemperatures," J. Quant. Spectrosc. Radiat. Transfer 28, 471(1982).
23. H. S. Lowry III and C. J. Fisher, "Line Parameters Measure-ments and Calculations of CO Broadened by H20 and CO2 at
Elevated Temperatures," J. Quant. Spectrosc. Radiat. Transfer31, 575 (1984).
24. R. A. Toth and L. A. Darnton, "Linewiths of HCl Broadened byCO 2 and N 2 and CO Broadened by C0 2," J. Mol. Spectrosc. 49,100 (1974).
25. C. B. Ludwig, W. Malkmus, J. E. Reardon, and A. L. Thompson,Handbook of Infrared Radiation from Combustion Gases,NASA SP-3080 (U.S. GPO, Washington, DC, 1973).
26. R. Levi Di Leon and J. Taine, "Infrared Absorption by GasMixtures in the 300-850 K Temperature Range 1-4.3 pm and 2.7pm C02 Spectra," J. Quant. Spectrosc. Radiat. Transfer 35, 337(1986).
27. T. Nakazawa and M. Tanaka, "Measurements of Intensities andSelf- and Foreign-Gas-Broadened Half-Widths of SpectralLines in the CO Fundamental Band," J. Quant. Spectrosc. Ra-diat. Transfer 28, 409 (1982).
28. A. Soufiani, "Etudes thboriques et expbrimentales des transfertscoupl6s par convection laminaire ou turbulente et rayonnementdans un milieu gazeux A temperature 6levbe," These d'Etat, U.Paris XI, No. 3384 (1987).
29. S. V. Patankar, Numerical Heat Transfer and Fluid Flow(Hemisphere, New York, 1980).
30. V. Ditkine and A. Proudnikov, Transformations integrales etcalcul operationnel (Mir Ed., Moscow, 1978).
Life on the edge continued from page 3722
lectures and hands-on use of the SRM. Recognizing theimportance of the new system, photomask instrument mak-ers loaned commercial measuring equipment to the seminarsso attendees could practice on the same microscopes theywould use at work. Almost every seminar was booked upweeks in advance. Information at these seminars flowed twoways, according to Nyyssonen, "We would have feedbacksessions at the seminars on what measurement problems theparticipants were seeing in their work. Other than the visitsto companies, I think the linewidth measurement seminarswere our most important source of information. Peoplewere very open with us, taking us into their facilities to seetheir equipment and discuss their problems."
By 1981, Bureau researchers were considering improve-ments. SRM 474 was expensive largely because it was com-plicated and took a long time to certify. Discussions withindustrial customers revealed that most people really did notneed all the complexity. SRM 475 was designed, a simpli-fied standard that could be certified and sold for much less.(On 1 October 1987, SRM 475 will be priced at $3781.) TheNBS masks used a dark chrome to reduce reflections, butsome customers use bright chrome masks so a bright chromeSRM was developed and will be released as SRM 476.
Industrial response was gratifying. In 1982, a U.S. firm,one of the world's largest IC manufacturers, formally insti-tuted an in-house measurement system based on SRM 474."It has basically become our bible," noted the linewidthmeasurement project manager. "We've created a whole se-ries of secondary standards, sent them out to the front endsand our measurement squabbles have been cut by 90%. Ican't remember the last time anyone tried to challenge ourreadings on a linewidth based on our knowledge from yourmeasurement seminars and SRM 474, whereas that was aweekly if not daily occurrence before then." Another firmestimated in 1983 that it had invested about $400,000 in newmeasurement procedures built around the NBS linewidthstandard and "bought that back many times over" in re-duced production problems.
In 1982, NBS studied the impact of the linewidth standardon the semiconductor industry. A hypothetical companymaking $1000 photomasks would save about $179 per maskby using the NBS technology, of which at least half thesavings could be directly attributed to the NBS work. Im-plementation costs: $2 per mask. On worldwide mask salesof $375 million per year, that translates to a $30 millionsavings directly attributable to the NBS program, whichcosts less than $400,000 per year. The numbers are to betaken cautiously-too many estimates are involved to besure of the actual dollars-but almost every variation in thebasic assumptions resulted in significant economies.
Technology, of course, does not stand still and neither cana standards laboratory. Today NBS is looking to the 1990sand a whole new set of linewidth measurement problems.The current optical system used to calibrate the NBSlinewidth SRMs is useful for linewidths down to about 0.5pm, but manufacturers are beginning to think in terms of 0.1-pum linewidths. Accurate measurements at that level willrequire scanning electron microscopes, with characteristicsquite different from optical instruments. "What happenedin optics is what is happening now in scanning electronmicroscopy," notes Larrabee. "People have instrumentsand are making measurements that they believe are correct.When things go wrong, they don't question the measure-ment, just the results."
In addition, the present system relies on measuring lightcoming through a transparent mask. Industry needs a simi-lar standard for measurements made on the opaque semicon-ductor wafers themselves, and this introduces a whole newset of problems. Imaging in reflected light can be morecomplicated to model than imaging in transmitted light, andthe mathematical models developed for SRM 474 do notapply to features on integrated circuits. NBS researchersare looking for solutions to these problems, working as usualwith their private-industry counterparts. NBS currentlyhas formal agreements for joint research with several privatefirms, including:
CD Metrology, Inc., to develop mathematical models tointerpret the profiles of features as imaged by electron andoptical microscopes;
EDAX International, to develop linewidth calibrationtechniques for scanning electron microscopes;
E. Fjeld Company, to develop precision stages for SEMs;Hewlett-Packard, to develop state-of-the-art masks for
linewidth standards; andVLSI Standards, Inc., for basic research to model how light
reflects from thick lines on silicon wafers.The linewidth standard is just one of many measurement
problems studied at NBS, but it is an example of the way thesystem is supposed to work. Comments NBS Director Er-nest Ambler, "Behind each SRM, each individual calibra-tion, or each table of reference data is a substantial invest-ment in time and effort by our best people. In this case,industry thought that all they needed was an SRM, but infact they needed an entirely new measurement technology.We have the expertise to know the difference and to solve theproblem, and it is our mission at the National Bureau ofStandards to do so-it is what we are all about."
This report is based on a reprint from NBS ResearchReports of August 1987 by anNBSPublic Affairs Specialist.
