Ctontents - NIST · Superconductor Abstracts Compiled in Single Volume Report Details Risks, Handling Care for 3480 Tape 313 NIST Issues "How-To" Manual on Smoke Control Better Solutions

Volume 97, Number 2, March-April 1992

Journal of Research of the National Institute of Standards and Technology

Ctontents

ArticlesThe Development of a Standard Reference Barbara C. Levin, Yves AMarie, 245

Material for Calibration of the University of Maryanne F. Stock,Pittsburgh Smoke Toxicity Method for and Susannah B. SchillerAssessing the Acute Inhalation Toxicity ofCombustion Products

Certification of NIST SRM 1962: Arie W. Hartman, Theordore D. 2533 jim Diameter Polystyrene Spheres Doiron, and Joseph Fu

Optical Calibration of a Submicrometer Jon Geist, Barbara Belzer, 267Magnification Standard Mary Lou Miller, and

Peter Roitman

Probe-Position Error Correction in Lorant A. Muth 273Planar Near Field Measurements at 60 GHz:Experimental Verification

Measuring the Electron's Charge and the E. R. Williams, Ruby N. Ghosh, 299Fine-Structure Constant by Counting and John M. MartinisElectrons on a Capacitor



News BriefsGENERAL DEVELOPMENTS 305Bioassay Reference Values Established for Three NIST SRMsAnalytical Method Developed to Investigate the Role of Vitamin C in Cancer

Prevention and Juvenile Arthritis

Intramolecular Vibrational Energy Redistribution Studied by PicosecondIR Spectroscopy 306

Frequency and Wavelength Standard for the Far InfraredIn Situ Characterization of the Pulsed Laser Deposition of Thin FilmsNew Computational Method for Pattern Formation During Alloy Solidification

Fiber Laser Offers Wavelength Standard 307NIST Research Leads to Significant Yield Improvement in the Manufacture

of Memory ChipsNIST Leads Industry to Accord in X-Ray Mask Specifications

Neutron Reflectometry Helps Characterize Polymer Brushes 308Fast Restoration of ImagesNeutron Focusing Using Capillary Optics

NIST Develops Methods for High-Precision Phase Noise Measurements 309Hubble Data Suggest No End for UniverseNew "Measures" Needed to Boost ProductivityMeasurement Service for CW Wattmeters ComingNew Glossary of Computer Security Terminology

New Proceedings Issued on High Integrity Software 310Automated Manufacturing Stars in New VHS ProgramNIST Reports on ISO 9000 Standard SeriesSoftware Package for Antenna Metrology Available

New Guide Designed to Help Federal GOSIP Users 311Book Reviews Supercritical Fluid ExtractionVideo Highlights 1991 Baldrige Award WinnersRadio Waves "Lower the Boom" on Buried MinesPhotonic Probes Developed to Measure EM Fields

Shaping New Designs for Wave Probes 312IC Device Offers Improved Speed, AccuracyAlaska Marine Mammal Pollution Data AvailableSuperconductor Abstracts Compiled in Single Volume

Report Details Risks, Handling Care for 3480 Tape 313NIST Issues "How-To" Manual on Smoke ControlBetter Solutions for Wave Equations OfferedNew Microscope Offers Unprecedented Range

New Test Improves Understanding of Smoke DangerNew Detector for Gas Chromatography PatentedGuide Makes SRMs Useful Tools in Chemistry LabsCARB Solves Second Sugar Transport Protein

314



Papers Requested for Fall Optics Symposium 315Conformity Assessment Program ProposedNew Atlas Improves Infrared Spectrometry CalibrationsComments Sought for Language C Test MethodReport Examines "Bridging" SGML/ODA Standards

Summer Technology Outreach Conferences Scheduled 316NIST Laboratory Metrology Seminar Held at the University of Puerto Rico (UPR), MayaguezSpherical-Wave Expansions of Piston-Radiator Fields, Algorithm Derived

NIST to Apply Probe Position-Error Correction in New Near-Field Spherical Scanning Range 317NIST Scientist Reports Findings of NASA's Airborne Arctic Stratospheric Expedition IIFluid Equilibration in Low Gravity

NIST Radio Broadcasts to Support INSPIRE Program for High School Science Students 318Ionospheric Receiver Technology Transferred to IndustryObservation of "Quantum Projection" Noise

Apparatus for Measuring Thermal Conductivity at Low Temperatures 319Proposed Consortium on Casting of Aerospace AlloysElectrodeposition of Compositionally Modulated AlloysTechnology of Glass-Ceramic Inserts Transferred to Dental Materials IndustryPredicting Performance of Adhesive Bonds

NIST Study of Staging Areas for Persons with Mobility Limitations 320NIST Collaborates with the National Archives and Records Administration

(NARA) on Optical Disk TechnologyNIST Proposes Federal Information Processing Standard (FIPS) for Secure Hash StandardSpatial Data Transfer Standard Implementation Workshop Attracts Large Turnout

International Workshop on Harmonizing Conformance Testing 321of Programming Language Standards Held

CALIBRATION SERVICES 321Udated Calibration Services Users Guide Available

STANDARD REFERENCE MATERIALS 321Easy Access to Materials Standards Offered

SRM Improves Lead Poisoning Diagnosis Accuracy 322Sample Available for Human Hair Drug AnalysisStandard Reference Material 893 and 1295-Stainless Steel (SAE 405)Standard Reference Material 2708- Zinc Sulfide Thin Film on PolycarbonateStandard Reference Material 2034-Holmium Oxide Solution

Wavelength Standard 240 to 650 nm

Reference Materials 8506 and 8507-Moisture in Transformer Oil and in Mineral Oil 323

STANDARD REFERENCE DATA 323New PC Package Speeds Chemical Microanalysis

Calendar 325



[J. Res. Natl. Inst. Stand. Technol. 97, 245 (1992)]

The Development of a Standard ReferenceMaterial for Calibration of the University of

Pittsburgh Smoke Toxicity Method forAssessing the Acute Inhalation Toxicity of

Combustion Products

Volume 97 Number 2 March-April 1992

Barbara C. Levin A standard reference material (SRM value would be expected to fall. Thus,1049) has been developed for the Uni- if an investigator were to test this SRM

National Institute of Standards versity of Pittsburgh smoke toxicity under their laboratory conditions ac-and Technology, method. SRM 1049 is a nylon 6/6 and cording to the specifications of the Uni-

Gaithersburg, MD 20899 has the molecular structure of versity of Pittsburgh test procedure and[-NH(CH 2 )6NHCO(CH 2)4CO-]1. This found the LC50 value fell within the

Yves Alarie and SRM is for calibrating the apparatus certified 95% prediction interval, theand providing confidence that the probability is good that the test is being

Maryanne F. Stock method is being conducted in a correct conducted correctly.manner and that the equipment is func-

University of Pittsburgh, tioning properly. The certified figure of Key words: combustion; combustionPittsburgh, PA 15261 merit is a LC 5 0 value plus its 95% pre- products; inhalation; nylon; nylon 6/6;

diction interval which were calculated SRM; standard reference material; toxi-and and found to be 4.4 + 3.4 g. The 95% city tests; University of Pittsburgh.

prediction interval indicates the rangeSusannah B. Schiller in which the next determined LC50 Accepted: February 10, 1992

National Institute of Standardsand Technology,Gaithersburg, MD 20899

1. Introduction

In 1973, The National Commission on Fire Pre-vention and Control issued the report "AmericaBurning" [1] which noted that most fire victims diefrom inhaling smoke and toxic gases. This informa-tion served as one of the motivating forces in thedevelopment and testing of many smoke toxicitytest procedures [2]. In 1983, 13 of these publishedmethods were evaluated by Arthur D. Little, Inc. toassess the feasibility of incorporating combustiontoxicity requirements for building materials and fin-ishes into the building codes of New York State [3].On the basis of seven different criteria, only twomethods were found acceptable. These two meth-ods were the flow-through smoke toxicity methoddeveloped at the University of Pittsburgh [4,5] andthe closed-system cup furnace smoke toxicity

method [6] developed at the National Institute ofStandards and Technology (NIST).

Based on the results of the A. D. Little report,the state of New York under Article 15, Part 1120of the New York State Fire Prevention and Build-ing Code decided that building materials and fin-ishes should be examined by the method developedat the University of Pittsburgh and that the resultsbe filed with the state [7]. It is important to note,however, that although the results are filed, thestate of New York does not regulate any materialsor products based on the results of toxicity testing.It is also important to note that, at the present time,no smoke toxicity method has been accepted as astandard test by ASTM or any other national or in-ternational scientific or technical society designed

245



to develop standard test procedures. Thus, thedevelopment of other smoke toxicity methods isstill being actively pursued.

Three methods currently under development arethe University of Pittsburgh II radiant furnacemethod [8,9,10], a radiant furnace smoke toxicityprotocol [11,12] which is being developed at NISTand the National Institute of Building Sciences(NIBS) toxic hazard test method [13,14]. Althoughthese methods differ significantly in numerouscharacteristics, all three use radiant heat to decom-pose materials. Documentation of the relevanceand accuracy of the radiant methodology may befound in Refs. [11] and [12].

Over the past decade, the number of smoke toxi-city test apparatus users has increased. A numberof Federal agencies, industrial laboratories, andtesting companies are capable of conducting boththe University of Pittsburgh and the cup furnacesmoke toxicity test procedures. Although there areno state or federal regulations, the results of thesesmoke toxic potency tests, along with the results ofother material flammability tests, are being used inthe decision making process regarding material se-lection and overall fire hazard. Therefore, it is nec-essary to assure that such testing devices areinstalled and employed properly both by those lab-oratories currently conducting these tests and bynew laboratories that enter the field. To help as-sure the reproducibility of results between labora-tories, NIST has developed two standard referencematerials (SRMs), one which can be used to cali-brate the University of Pittsburgh smoke toxicitymethod (SRM 1049) and another SRM (SRM1048) which can be used to calibrate the cup fur-nace smoke toxicity method [15]. It is important tonote that these SRMs were not selected to representthe toxic potency of the combustion products of an"average" material and are not designed to be usedfor the comparison of the relative toxic potency of thecombustion products of test materials. Therefore, toxicpotency of the smoke from a test material should notbe compared to the toxic potency of the smoke fromthese SRMs.

The following criteria were used in the selectionprocess of the University of Pittsburgh smoke toxic-ity SRM:

1. The material should have reproducible burningcharacteristics (i.e., the material must be ho-mogeneous),

2. The material should produce combustion prod-ucts whose toxic potency values are within therange where the values for some other materi-als are found,

3. Upon combustion, toxic gases in addition toCO should be generated and contribute to thelethal atmospheres, and

4. The selected material should generate combus-tion products which cause deaths during theanimal exposures. The University of Pittsburghmethod does not specify the post-exposure ob-servation of the test animals other than an im-mediate 10 min period following the exposure.

The polymer nylon 6/6, whose characteristics fitthe above criteria, was selected for the Universityof Pittsburgh smoke toxicity SRM. An intralabora-tory evaluation (performed at the University ofPittsburgh, Pittsburgh, PA) and an interlaboratoryevaluation (carried out by Anderson Laboratories,Inc., Dedham, MA, Southwest Research Institute,San Antonio, TX, University of Pittsburgh, U.S.Testing Co., Inc., Hoboken, NJ, and WeyerhaeuserCo., Longview, WA) were conducted to determinethe repeatability of results within a laboratory andreproducibility of results between laboratories, re-spectively. When the intra- and interlaboratoryevaluations showed good repeatability and repro-ducibility of results with nylon 6/6, additional mate-rial of a single lot number was ordered for certi-fication as an SRM. Further testing of the new lotwas conducted by University of Pittsburgh and An-derson Laboratories to provide the data necessaryfor the development of the final certified SRM.

This paper documents the research and develop-ment of SRM 1049 which will be used to calibratethe University of Pittsburgh smoke toxicity test pro-cedure and will help assure that the apparatus isperforming correctly. To use SRM 1049 in the cali-bration of the test procedure, a laboratory woulddetermine the LC5o value of the SRM according tothe published University of Pittsburgh test proce-dure [4,5,7] and compare it with the certified LC5ovalue and its 95% prediction interval.' If the exper-imental value obtained by the laboratory fallswithin the 95% prediction interval of the certifiedLC5o value of this SRM, the investigator can beconfident that the method is being conducted cor-rectly.

1 In the University of Pittsburgh smoke toxicity method, theLC50 is defined as the statistical estimate of the amount of mate-rial (in grams) which, when placed into the furnace, wouldcause 50% of the exposed mice to die within the 30 min expo-sure and the 10 min post-exposure observation period.

246



2. Materials and Methods2.1 Materials

Two separate lots of nylon 6/6 [poly(hexa-methylene adipamide)] with a molecular structureof [-NH(CH 2 )6NHCO(CH 2 )4 CO-]n were obtainedfrom Aldrich Chemical Co.2 The first sample ofnylon 6/6 (lot 80078; hereafter referred to as lot 1)was tested by the University of Pittsburgh and fourother laboratories. Based on the results of this in-terlaboratory evaluation of the University of Pitts-burgh smoke toxicity method, nylon 6/6 was foundto be a suitable candidate for an SRM. A secondbatch of nylon 6/6 was ordered for certificationpurposes. The second batch of nylon 6/6 was in thesame pellet form and had the same manufacturerspecifications as the first batch, but had a differentlot number (lot 08015; hereafter referred to as lot2). Each bottle containing 1 kg of nylon 6/6 (lot 2)was randomly numbered when received at NIST.Samples of four bottles, Nos. 10, 20, 30, and 40,from lot 2 were used for certification purposes.

2.2 Animals

Swiss Webster mice weighing between 22 and28 g were used for this study. They were allowed toacclimate to the laboratory conditions for approxi-mately 1 week. Animals were group housed withfree availability of food and water. Only animalsappearing healthy were used for the study.

2.3 Experimental Method

The University of Pittsburgh smoke toxicitymethod to evaluate the acute inhalation toxicity ofcombustion products was developed by Alarie andAnderson [4,5]. The experimental arrangement isillustrated in Fig. 1. In this method, the sample isplaced on a load sensor in the furnace at roomtemperature. The temperature is then increased atthe rate of 20 'C/min. The temperature at whichthe sample begins to decompose, the rate at whichthe material decomposes as the temperature in-creases, and the time of ignition (i.e., flaming) ofthe sample are recorded.

Animal (head only) exposure to the thermal de-composition products is started when 1.0% mass

loss of the material occurs. Four Swiss Webstermale mice between 22 and 28 g in weight are ex-posed in each experiment. The exposure is 30 minin duration. The toxicological endpoint is deathwhich occurs during the 30 min exposure periodand a 10 min post-exposure observation period.The amount of material which releases enoughsmoke to cause either 0, 25, 50, 75, or 100% of theanimals to die is used to calculate the LCso values.During the animal exposures, the major combus-tion products, carbon monoxide (CO), carbon diox-ide (CO 2), and reduced oxygen (02) arecontinuously monitored.

All dimensions in millimeters

Fig. 1. Schematic of the experimental test system used to de-compose the sample and expose the animals.

In the evaluation and development of SRM 1049(i.e., data presented in this paper), the LC5 o valuesand their 95% confidence limits were determinedby the statistical method of Weil [16]. At the Uni-versity of Pittsburgh, CO, C02, and 02 were mea-sured continuously by a Miran 1A infrared gasanalyzer', a Beckman LB-2 medical gas analyzer,and a Beckman OM-11 oxygen analyzer, respec-tively. At Anderson Laboratories, CO and C02were measured continuously by Horiba nondisper-sive infrared gas analyzers and 02 was measuredcontinuously by a Lynn electrochemical oxygen an-alyzer. Even though the material contained nitro-gen, concentrations of hydrogen cyanide generatedduring these experiments were not measured.

2 Certain commercial equipment, instruments, or materials areidentified in this paper to specify adequately the experimentalprocedure. Such identification does not imply recommendationor endorsement by the National Institute of Standards andTechnology, nor does it imply that the materials or equipmentidentified are necessarily the best available for the purpose.

I The absolute gas concentrations reported in this paper dependupon the analytical instruments used in measuring the specificgas species and the instruments' response times.

247



2.4 Comparison Factors in the Development ofthis SRM

2.4.1 Interlaboratory Evaluation To ascertainthe reproducibility across different laboratories,four laboratories (in addition to the University ofPittsburgh) were asked to participate in an inter-laboratory evaluation of nylon 6/6 (lot 1). Theselaboratories were Anderson Laboratories (Ded-ham, MA), Southwest Research Institute (San An-tonio, TX), U.S. Testing Co., Inc. (Hoboken, NJ),and Weyerhaeuser Company (Longview, WA).2.4.2 Intralaboratory Evaluation The Universityof Pittsburgh examined the repeatability of theLC5o for both lots of nylon 6/6. Three separate LC5ovalues were determined for nylon 6/6 (lot 1) andfour LC50 values for lot 2. In addition, two separatesamples from lot 2 were sent to Anderson Labora-tories which determined two LC5o values.

2.5 Statistical Analysis of Results

For this SRM, two types of statistical uncertain-ties, a 95% confidence interval and a 95% predictioninterval, were determined. The 95% confidence in-terval defines the precision with which the trueendpoint (the LC5o) is known; whereas, the 95%prediction interval provides the numerical bound inwhich the next LC5o should fall if the experimentsare conducted correctly. Unlike a 95% confidenceinterval, a 95% prediction interval does not get ap-preciably narrower if more laboratories participatein the study, since the 95% prediction interval willalways be larger than the interlaboratory standarddeviation. The difference between these two inter-vals is illustrated below for the simplified case inwhich each of "n" laboratories would determineone LC5o value.

The uncertainty based on a 95% confidence in-terval for the true LC50 is shown in Eq. (1).

tn..1(0.025) Vinterlab variance (1)

where t - i(O.025) is the appropriate cutoff from thestudent's t distribution for a two sided interval witha 95% confidence level [17].

The uncertainty based on a 95% prediction in-terval, using data from the same simple case asabove, would be determined by Eq. (2) [18]. Thiscalculation incorporates the variance of a singlenew determination of the LC50 value plus the vari-ance of the mean.

tn 1 (0.025) \/ (n + 1) interlab variancetn IO02 V n (2)

Since we had limited data on the SRM material(only two laboratories evaluated it), we estimatedthe variability in the certified LC5o using data fromthe interlaboratory study (on lot 1) as well as thedata on the SRM (lot 2). Analysis of these datashowed that the variability between laboratorieswas larger than the variability within the laborato-ries. These data could be pooled since the two lotsof nylon 6/6 were considered fairly similar based onthe material composition and the expectation thatthe measurement errors made by the laboratorieswould follow the same distribution for both lots. Inaddition, an Analysis of Variance indicated thatthere was no statistically significant difference be-tween the mean LCso of lot 1 and that of lot 2. Theinterlaboratory and intralaboratory variance com-ponents were estimated from an analysis of vari-ance on both materials simultaneously. Thevariance of the mean LC5o value for the SRM isgiven in Eq. (3):

1 interlab variance + 3 within-lab variance (3)2 16

and the effective degrees of freedom for this vari-ance are 4.9. Therefore, the uncertainty based on a95% confidence interval for the mean is shown inEq. (4):

t4.9(0.025) /2 interlab variance + 3 within-lab variance2 ~~~~16

(4)

The variance of a single new LC50 plus the varianceof the mean is shown in Eq. (5):

3 interlab variance + 19 within-lab variance2 16

(5)

and the effective degrees of freedom for this sumare 5.3. Therefore, the uncertainty based on a 95%prediction interval is seen in Eq. (6):

ts.3(0.025) 2 interlab variance + 19 within-lab variance2(16

(6)

248



3. Results3.1 Interlaboratory Evaluation

The toxicological data for nylon 6/6 (lot 1) pro-vided by the participating laboratories in the inter-laboratory evaluation are given in Table 1.

3.2 Intralaboratory Evaluation

The toxicological and chemical data for nylon 6/6(lot 2) obtained at the University of Pittsburgh andAnderson Laboratories are given in Tables 2 and 3,respectively. The LC5o values of four separatesamples (bottles 10, 20, 30, and 40) of lot 2 were

Table 1. Interlaboratory evaluation of nylon 6/6 (lot 1) toxico-logical data

Laboratory LC5 0 values Mean LC50 values

(g) (g)

Univ. of Pittsburgh 4.8 (4.2-5.3)- 5.25.2 (4.8-5.7)5.7 (5.3-6.2)

1 3.6 (3.6-3.6) 3.83.7 (3.1-4.6)4.1 (3.3-5.0)

2 6.1 (6.0-6.2) 6.67.1 (6.7-8.0)

3 5.4 (4.8-6.1)5.4 (4.8-6.1)

4 4.3 (3.6-5.4) 4.3

Overall mean +95% confidence interval" 5.1±1.2

a 95% confidence limits of the LC5 0 values.b Mean of all the laboratories' mean values. The 95% confi-dence interval incorporates both the within-laboratory and thebetween-laboratory variation.

determined by the University of Pittsburgh. Twoadditional LC5 o values were obtained for twosamples of lot 2 (bottles 20 and 30) by AndersonLaboratories. The material mass loss, evolution ofCO and C0 2 , and reduction of 02 found in testsconducted by the University of Pittsburgh at orclose to the LC5 o values are illustrated in Figs. 2 to5. These four figures also show the temperatures atwhich the material decomposition began and whenflaming ignition occurred, and the temperatures atwhich animal exposure was initiated as well as thetemperature at the time 50% of the animals died(LT50).

Table 2. Intralaboratory evaluation of nylon 6/6 (lot 2) toxico-logical data

Laboratory Bottle LC50 valuesNo. (95% confi- Mean LC5 0

dence limits) value

(g) (g)

U. Pitt. 10 3.6 (ND)20 3.9 (3.7-3.9) 3.730 3.6 (3.5-3.7)40 3.9 (3.9-4.0)

Anderson 20 5.1 (4.5-5.8) 5.130 5.15a (5.Pb-5.2c)

Overall mean +95% confidence interval 4.4+ 1. 9d

a Midpoint between values in parenthesis"No deaths at this mass loadingc 100% deaths at this mass loadingd The overall mean is based on the mean values from the twolaboratories; The 95% confidence interval incorporates both thewithin and between laboratory variation.ND-not determined.

Table 3. Intralaboratory evaluation of nylon 6/6 (lot 2) physical and chemical dataa

Laboratory Bottle Temp. Temp. Time of Max Time Max Time Minimum TimeNo. initial during flaming" CO max CO 2 max 02 minimum

exposure flaming (min) (%) CO (%) C0 2 (%) 02

(IC) (IC) (min) (min) (min)

Univ. of 10 410 431-513 1.5 0.64 3 7.5 3 13 3Pittsburgh 20 415 448-531 1.7 2.8 3.5 ND ND ND ND

30 410 447-536 2.0 0.90 3.5 ND ND ND ND40 415 441-531 1.3 1.1 4.5 ND ND ND ND

Anderson 20 409 434-612 NR 1.2 NR 7.0 NR 11.6 NR30 401 435-601 NR 1.2 NR 7.6 NR 11.8 NR

a Values in table are from experiments conducted at a concentration equivalent to the LC50 value."Time that material started flaming where the beginning of the animal exposure is 0 time.ND-not determined.NR-not reported.

249



8 _

To 4 -

0 _

100

CD 80zZ

L 60c:GO

E 40

zw20

0-

0

------ % carbondioxide

-E Oxygen

- 21

J 18

_ 16 -

- 12

30

27

24

21

18

15

12

9

6

3

0100 200 300 400 500 600 700 800 900

TEMPERATURE-0 C

Fig. 2. Gas concentrations and percent mass remaining as fur-nace temperatures increased during the decomposition of nylon6/6 (lot 2) from bottle No. 10. The initial mass of nylon 6/6 was3.6 g (the LCs5 value).

100 _

80 _0

[1] 60cc

x (n

X u2Co U,)H 400- wo cco

lH

zu r-

0

Decomposition I-and

Exposure Start

I I At

100 200 300 400 500

-+- Mass remaining--U--- CO (PPM)

,.- LT50

600 700 800 900

30

27

24

21

18 %

15 X

PL

12 °

9

6

3

0

TEMPERATURE-C

Fig. 4. Carbon monoxide concentrations and percent mass re-maining as furnace temperatures increased during the decom-position of nylon 616 (lot 2) from bottle No. 30. The initial massof nylon 616 was 3.6 g (the LC50 value).

30

100

80

zZ

W 60

ECU)Go

-40z0ccw

a-20

100 200 300 400 500 600 700 800 900

TEMPERATURE - IC

Fig. 3. Carbon monoxide concentrations and percent mass re-maining as furnace temperatures increased during the decom-position of nylon 6/6 (lot 2) from bottle No. 20. The initial massof nylon 6/6 was 3.9 g (the LC5 0 value).

100 F_

80 -

60

40 H

20 H

0

Decomposition I- and

Exposure Start2

I I i

0 Mass remaining---2--- CO (PPM)

p LT50

27

24

91

18 "' Z2

15 2 cco

12 8 2oz

9 0

CL6

3

0

100 200 300 400 500 600 700 800 900

TEMPERATURE-tC

30

27

24

21

18 ox

15 LIL

12 °0

Fig. S. Carbon monoxide concentrations and percent mass re-maining as furnace temperatures increased during the decom-position of nylon 6/6 (lot 2) from bottle No. 40. The initial massof nylon 6/6 was 3.9 g (the LC50 value).

250

I

0- Lk--r-0



4. Discussion

The development of a standard reference mate-rial requires the statistical determination of a certi-fied value (in this case, the LC5 o of the material)and an uncertainty which informs the user how wellthe certified value is known. For most standard ref-erence materials, this uncertainty is defined by the95% confidence interval, which, in the example ofSRM 1049, would tell the user how precisely thetrue LC5o value is known (with 95% confidence).However, for this SRM, we have determined the95% prediction interval rather than the 95% confi-dence interval. The 95% prediction interval tellsthe user the numerical bound in which the nextLC5o value should fall assuming the experimentsare conducted correctly and instruments are func-tioning properly. A confidence interval is narrowerthan a prediction interval and will get narrower asthe number of participating laboratories increases.Even if the user's system is operating correctly andhis precision is comparable to the precision foundin the research reported here, the next LC5 o valuecould reasonably fall in a much larger range thanthat given by the confidence interval. On the otherhand, the 95% prediction interval will not getsmaller than the precision of a single measurementand thus allows the user to judge if the next experi-mental value obtained in his laboratory is in theright range.

The mean LC5o value of SRM 1049 (nylon 6/6;lot 2) is 4.4 g; its 95% confidence interval is ± 1.9 g;whereas, its 95% prediction interval is ± 3.4 g. This95% prediction interval incorporates the variabilityfrom both the within-laboratory experiments andthe interlaboratory evaluation of nylon 6/6. There-fore, if the user's precision is comparable to theprecision found in the data in this report and thetest procedure is being conducted correctly, the95% prediction interval should include the user'snext LC5o measurement.

5. Conclusions

A standard reference material (SRM 1049) hasbeen developed for calibration of the University ofPittsburgh smoke toxicity method for assessing theacute inhalation toxicity of combustion products.The certified material is a nylon 6/6 and the certi-fied LC5o value was based on four series of testsconducted at the University of Pittsburgh and twoseries of tests conducted at Anderson Laboratories.The 95% prediction interval is based on the vari-ability of results found in both interlaboratory andintralaboratory evaluations. The certified LC5o

value and 95% prediction interval is 4.4 ± 3.4 g. If alaboratory were to test this SRM under their condi-tions in their apparatus and found the LCso valueto fall within the certified 95% prediction interval,the probability is good that their equipment isfunctioning appropriately and that the test is beingconducted correctly.

Acknowledgment

The participation of Anderson Laboratories, Inc.(Dedham, MA), Southwest Research Institute(San Antonio, TX), University of Pittsburgh (Pitts-burgh, PA), U.S. Testing Co., Inc. (Hoboken, NJ),and Weyerhaeuser Co. (Longview, WA) in the in-terlaboratory evaluation is greatly appreciated.

6. References

[1] America Burning, Report of the National Commission onFire Prevention and Control, U.S. Government PrintingOffice, Washington, DC, 1973.

[2] H. L. Kaplan, A. F. Grand, and G. E. Hartzell, Combus-tion Toxicology, Principles and Test Methods, TechnomicPublishing Co., Inc., Lancaster, PA (1983).

[3] R. C. Anderson, P. A. Croce, F. G. Feeley, III, and J. D.Sakura, Study to Assess the Feasibility of IncorporatingCombustion Toxicity Requirements into Building Mate-rial and Furnishing Codes of New York State. Final re-port to Department of State, Office of Fire Preventionand Control, Albany, New York, From Arthur D. Little,Inc., Cambridge, MA, Ref. No. 88712, (May 1983).

[4] Y. C. Alarie and R. C. Anderson, Toxicologic and AcuteLethal Hazard Evaluation of Thermal DecompositionProducts of Synthetic and Natural Polymers, Tox. Appl.Pharm. 51, 341-362 (1979).

[5] Y. C. Alarie and R. C. Anderson, Toxicologic Classifica-tion of Thermal Decomposition Products of Synthetic andNatural Polymers, Tox. Appl. Pharm. 57, 181-188 (1981).

[6] B. C. Levin, A. J. Fowell, M. M. Birky, M. Paabo, A.Stolte, and D. Malek, Further Development of a TestMethod for the Assessment of the Acute Inhalation Toxi-city of Combustion Products, NBSIR 82-2532, Natd. Bur.Stand. (U.S.) (1982).

[7] New York State Uniform Fire Prevention and BuildingCode, Article 15, Part 1120, Combustion Toxicity Testingand Regulations for Implementing Building Materials andFinishes; Fire Gas Toxicity Data File. New York State,Department of State, Office of Fire Prevention and Con-trol, Albany, NY 12231 (December 16, 1986).

[8] D. J. Caldwell and Y. C. Alarie, A Method to Determinethe Potential Toxicity of Smoke from Burning Polymers: I.Experiments with Douglas Fir, J. Fire Sci. 8, 23-62(1990).

[9] D. J Caldwell and Y. C. Alarie, A Method to Determinethe Potential Toxicity of Smoke from Burning Polymers:II. The Toxicity of Smoke from Douglas Fir, J. Fire Sci. 8,275-309 (1990).

[10] D. J. Caldwell and Y. C. Alarie, A Method to Determinethe Potential Toxicity of Smoke from Burning Polymers:III. Comparison of Synthetic Polymers to Douglas Fir

251



Using the UPitt II Flaming Combustion/Toxicity ofSmoke Apparatus, J. Fire Sci. 9, 470-518 (1991).

[11] V. Babrauskas, R. H. Harris, Jr., E. Braun, B. C. Levin,M. Paabo, and R. G. Gann, The Role of Bench-Scale TestData in Assessing Full-Scale Fire Toxicity, NIST Tech.Note 1284, Natl. Inst. Stand. Technol. (1991).

[12] V. Babrauskas, B. C. Levin, R. G. Gann, M. Paabo, R. H.Harris, Jr., R. D. Peacock, and S. Yusa, Toxic PotencyMeasurement for Fire Hazard Analysis, NIST SpecialPublication 827, Natd. Inst. Stand. Technol. (1991).

[13] J. C. Norris, National Institute of Building Sciences Com-bustion Toxicity Hazard Test, in Proceedings of the JointMeeting of the Fire Retardant Chemicals Association andthe Society of Plastics Engineers, Grenelefe, FL (March1988) pp. 146-155.

[14] H. J. Roux, The NIBS SMOTOX WG Program, Proceed-ings of the 37th International Wire and Cable Sympo-sium, Reno, NV (November 1988) pp. 543-545.

[15] B. C. Levin, M. Paabo, and S. B. Schiller, A StandardReference Material for Calibration of the Cup FurnaceToxicity Method for Assessing the Acute Inhalation Toxi-city of Combustion Products, J. Res. Natl. Inst. Stand.Technol. 96, 741-755 (1991).

[16] C. S Weil, Tables for Convenient Calculation of Median-Effective Dose (LD50 or EDso) and Instructions in TheirUse, Biometrics 8, 249-262 (1952).

[17] M. Natrella, Experimental Statistics. NBS Handbook 91,Natl. Bur. Stand. (U.S.) (1966).

[18] G. A.Whitmore, Prediction Limits for a Univariate Nor-mal Observation, Am. Statistician 40, 143-145 (1986).

About the authors: Barbara C Levin is a ProjectLeader and Toxicologist in the Fire Hazard AnalysisGroup in the Building and Fire Research Laboratoryat the National Institute of Standards and Technol-ogy. She is also an Adjunct Professor in the Fire Pro-tection Engineering Department at the University ofMaryland, College Park, Maryland. Yves Alarie is aFull Professor of Respiratory Physiology and Toxicol-ogy in the Department of Environmental and Occupa-tional Health, University of Pittsburgh GraduateSchool of Public Health. Maryanne F. Stock is aSenior Research Specialist in Combustion and In-halation Toxicology in the Department of Environ-mental and Occupational Health at the University ofPittsburgh Graduate School of Public Health. Susan-nah B. Schiller is a statistician in the Computing andApplied Mathematics Laboratory at the National In-stitute of Standards and Technology. The NationalInstitute of Standards and Technology is an agency ofthe Technology Administration, U.S. Department ofCommerce.

252




Certification of NIS T SRM 1962:3 ,Am Diameter Polystyrene Spheres


Arie W. Hartman, Theodore D. This report describes the certification of b =2.990± 0.009 tlm, based on mea-Doiron, and Joseph Fu SRM 1962, a NIST Standard Reference surement of 120 spheres. The reported

Material for particle diameter. It con- value covering the two results isNational Institute of Standards sists of an aqueous suspension of mono- D = 2.983 plm with a maximum uncer-and Technology, size 3 plm polystyrene spheres. Two tainty of 0.016 pym, with EYD = 0.020 p1m.Gaithersburg, MD 20899 calibration techniques were used: optical

microscopy and electron microscopy. Key words: electron microscopy; micro-The first one gave a mean diameter of spheres; optical microscopy; particle siz-D =2.977±0.011 pm and a standard de- ing; polystyrene; standard referenceviation of the size distribution TD = 0.020 materials.pIm, based on measurement of 4600spheres. The second technique gave Accepted: September 16, 1991

1. Introduction and Summary

This report contains the procedures, measure-ment results, and error analysis for the certificationof SRM 1962, a Standard Reference Material(SRM) for particle diameter. The SRM consistsof a 0.5% aqueous suspension of monosizepolystyrene microspheres with a mean diameter ofnominal 3 pm.

The calibration was carried out by two indepen-dent methods: specialized forms of optical andelectron microscopy. The two methods are de-scribed in Sec. 2, the measurement results areshown in Sec. 3, and the error analysis is given inSec. 4.

The results of the calibration are as follows:Optical microscopy:

Mean diameter: D = 2.977 ± 0.011 am

Diameter distribution: Gaussian from 10 to90%Standard deviation oD

= 0.020 am

Number of outliers (defined here as ID -b I >4OrD):

<0.5% for oversizenil for undersize

Seven samples totaling over 4600 spheres weremeasured.

Electron microscopy:Mean diameter: D = 2.989 ± 0.009 jm

Reported mean diameter value: D = 2.982+0.016 ltm.

2. Methods

The two methods used in the calibration of SRM1962 are described in this section.

2.1 Optical Microscopy (CDF)

A drop of microsphere suspension is placed on amicroscope slide, and allowed to flow out and dry.

253



During drying the drop breaks up into numeroussmaller droplets that dry individually. The spheresthat these droplets contain are pulled together bysurface tension forces, resulting in strands andsmall clusters of contacting spheres (Fig. 1). Thecontacting spheres are illuminated in the micro-scope by near-parallel light (condenser stoppeddown), and a number of small and circular "focalspots" form in the common back focal plane, asshown in Fig. 2. When a photomicrograph is takenof this back-focal plane, each recorded spot marksa sphere center. The distances C between adjacentspots represent the sum of two sphere radii. If thesphere diameters D are distributed normally(Gaussian), the C-values will also be distributednormally. The mean value C then equals D and thestandard deviation oc of the C-distribution equalsaDV/2. In this way D and aD are found. This tech-nique is called Center Distance Finding, or CDF[1].

The photographed focal spots are small (about0.5-0.6 tm in the object plane), uniform and circu-lar, permitting center distances C to be measuredwith high precision: a few hundredths of a microm-eter in the object plane. It thus allows a measure-ment of the diameter distribution, which would be

Fig. 1. Strands and clusters of 3 Am spheres.

difficult to do from measurements of the sphere im-ages themselves.

To make the CDF measurements a number ofmicrosphere slides are prepared and photographed.A large number of photographs are measured un-der computer control (see Appendix A). The filmscale (image magnification) is measured, as out-lined in Appendix B. The image distortion, whichfor high-quality optics is a function of off-axis dis-tance only, is measured also (see Appendix B). Thecomputer then applies a radial correction to eachmeasured focal spot position. The corrected centerdistances C are determined, which leads to 1D andOb.

2.2 Electron Microscopy (MEM)

With this method, called Metrology Electron Mi-croscopy or MEM, the focused beam of a scanningelectron microscope (SEM) is held stationary whilea single-axis scanning stage with interferometric po-sition readout moves the specimen such that the fo-cused electron beam traverses the distance betweenleading and trailing edge of the sample to be mea-sured. In our case the sample consists of a straightrow of equal-size microspheres.

An interferometer system measures stage travelversus time during a constant-speed scan, and thesecondary electron detection system measures theelectron output varying with time, all under com-puter control. The two data streams are combinedresulting in a value for the length of the measuredmicrosphere row, from which an average sphere di-ameter is found. The operation resembles that of anoptical measuring microscope, where a set ofcrosshairs defines a stationary reference point inthe field of view and a micrometer screw measuresstage travel. See also Refs. [2] and [3].

3. Measurements

In this section details are given of the specimenpreparation, data collection and reduction, and themeasurement results. Section 3.1 covers optical mi-croscopy, Sec. 3.2 treats electron microscopy.

3.1 Optical Microscopy (CDF)

Four samples were taken from one vial of SRM1962 microsphere suspension, and one sample fromeach of three other vials. The vial contents werehomogenized by rolling and shaking for 2 min, priorto dispensing a drop of suspension for analysis.

254



Fig. 2. The CDF microsphere sizing scheme.

The microscope used was an Olympus ModelBH-21 with a 100x/0.90NA objective, producingimages with 1000 x magnification on 4 x 5 in Polar-oid sheet film.

Focal-spot patterns from the contacting micro-sphere structures were photographed on Type 57positive film. This high-speed material (3000 ASA)has adequate dimensional stability [1] and lowgranularity, permitting its use for this SRM calibra-tion. Seventy-six photographs were measured, con-taining over 4600 measured focal spots. Themeasurement path through each microspheregrouping was selected such that each sphere wasmeasured only once. The groupings of contactingspheres were examined first for overdetermined-ness, to indicate where small air gaps between ap-parently contacting spheres could have formedduring the drying process. Such gaps have mini-mum widths ranging from zero to typically one on.

' Certain commercial equipment, instruments, or materials areidentified in this paper to specify adequately the experimentalprocedure. Such identification does not imply recommendationor endorsement by the National Institute of Standards andTechnology, nor does it imply that the materials or equipmentidentified are necessarily the best available for the purpose.

Air-gap formation can occur in contacting micro-sphere groupings, such as hexagonal arrays wheresix neighboring spheres surround a center spherewhile the spheres have slightly different diameters[4,5]. Such overdetermined sphere groupings wereavoided in the measurement phase. An example ofa selected measurement path is given in Fig. 3.

The measured photographs had a print magnifi-cation of nominal 1000 x . The measured focal spotshad 0.5-0.6 mm diameters, their 3 mm center spac-ings were measured with 0.01 mm resolution. Themicroscope image calibration for magnification andimage distortion is detailed in sec. 4.1.1 and in Ap-pendix B.

Measurement results are given in Table 1 and inFig. 4. The data were originally plotted with centerdistances as the horizontal axis. This was then con-verted into a diameter scale by compressing thehorizontal scale by V2 to reflect the fact that fornormal distributions ob = ac /2, and by centeringthe D-scale such that the mean diameter D coin-cides with the mean center distance C. The result-ing "diameter distribution" of Fig. 4a alreadyimplies that this distribution is considered a normalone. The information extracted from Fig. 4 is:

255



H 14 RB

030

.010

£010_ ; 3 A

030

.040

lii [ll lI IdI l L lL II U.LLLULLI Liilit.i.LLtILLUiltSilLUlU I

-. 40 -. 030 -. 020 -. 010 .000 .010 .020 .030 .040

mm

Fig. 3. Measurement path for a sphere grouping.

a) the median diameter (which corresponds withthe average diameter D if the distribution is nor-mal), b) the diameter range over which it actually isnormal, and c) the value for the standard deviationob associated with that diameter range.

3.2 Electron Microscopy (MEM)

The contents of an SRM 1962 vial were homoge-nized by rolling and shaking for 2 min. Then a dropwas taken from the vial, diluted in 50 ml of 18 Mflcm deionized water, and washed three times to re-duce the amount of dissolved material remaining(biocide). Each washing cycle involved low-powerultrasonication, settling and decanting four-fifths ofthe clear liquid. A small drop of the final suspen-sion was placed on three microscope slides and air

dried, causing formation of single-layer arrays withhexagonal ordering. The slides were then coatedwith about 30 nm of amorphous carbon to minimizecharging in the electron beam. With this techniquethe formation of sphere arrays was sought, as op-posed to the case of CDF (see Sec. 4.2.1).

The electron microscope used for the micro-sphere diameter measurements is a Vacuum Gen-erators VG HB-50A scanning electron microscope.It has in the secondary electron imaging mode anedge resolution of 0.03 jim at 30 keV and a 25 mmworking distance. The interferometer is a single-pass polarization Michelson type, mounted in theSEM vacuum on the fixed and moving parts of apiezo-electric scanning stage. The two reflectors arecorner cube prisms to accommodate the relativelylong distance (some 80 cm) from the stage insidethe SEM column to the interferometer readout sys-tem outside. The scanning stage is placed on top ofthe X-Y stage in the SEM. The X-Y stage is usedfor searching.

The interferometer readout is a Hewlett-PackardModel 5526A, utilizing a two-frequency stabilizedHe-Ne laser and a heterodyne scheme for measur-ing optical path differences. The two reflectors aremounted in the SEM vacuum on the fixed and mov-ing parts of the piezo-electric one-axis scanningstage. The reflectors are corner cube prisms, to ac-commodate any misalignment over the relativelylong distance (some 80 cm) from the stage insidethe SEM column to the interferometer readout sys-tem outside. A block diagram of the MEM systemis given in Fig. 5.

Forty microspheres were measured on each ofthree slides, selecting array rows that were essen-tially parallel to the scanning stage axis. The rowswere at least 12 spheres long allowing measurementof the spacing between two contact planes sepa-rated by 10 spheres, from which an average sphere

Table 1. Measurement results with optical microscopy'

Vial # Sample # Sphere diameter # of measurements # of photographs(plm)

1 1 2.975 1092 151 2 2.972 571 91 3 2.974 475 111 4 2.975 501 171 to4 2.974 2639 522 5 2.978 720 113 6 2.975 214 44 7 2.975 1061 9

all all 2.975 4634 76

a Diameter distribution is approximately normal for 10 to 90%. Standard deviation over this interval: 0.021 pm.

256



Diameter Distribution of SRM1962 (3lgm Microspheres)

z

z3

C-)

CO)

zD0

m_JM

M0

3L

D

0

DIAMETER (,U m)

a)

Fig. 4. SRM 1962: a) diametertribution.

99

9590705030105

…I

I iI i 2 D

I

2.9 3.0DIAMETER (at m)

3.1

b)

distribution; b) cumulative dis-

data point spacing. Measurement results are givenin Table 2.

4. Error Analysis

In this section sources of uncertainty (called "er-rors" for short) are identified and evaluated for thetwo microsphere size measurement techniques.They are expressed as "3 or" or "maximum" errorsas indicated, the individual independent contribu-tions are summed in quadrature, and the total sys-tematic and random errors are added linearly toform "the uncertainty" of the measurement pro-cess (see also Tables 3 and 4).

Output

Fig. S. Diagram of the Metrology Electron Microscope (MEM).

diameter was calculated. Visibly obvious outlierswere excluded from the measurements. After eachcomputer-controlled scan along a microsphere rowthe microscope was reverted to scan mode (SEMmode) and the next sphere row positioned manu-ally for a line scan (spot mode). The scans were 60jim long, with the secondary electron intensity pro-file being sampled at 2000 equispaced points. Eachminimum in SEM output current signals the pass-ing of a contact plane between two touchingspheres (see Fig. 6). Each minimum fell within one

4.1 Errors in Optical Microscopy

The errors in measuring the average diametercan be arranged in three groups: errors associatedwith finding the image magnification of the mea-sured photographs, errors associated with measur-ing photographed focal spot spacings (centerdistances between contacting spheres), and errorsassociated with the diameter distribution. To findestimates for the first two errors, five repeat photo-graphs were taken. Averaging of the repeat datawas done to find the magnification at lower uncer-tainty, while comparison between the photographswas used to find scatter in measured focal spotspacings from which uncertainties in the magnifica-tion and in a single measurement of center dis-tance can be derived. The three groups of errorsare discussed separately.

4.1.1 Errors Associated with Image Magnifica-tion The print magnification was found by pho-tographing an interferometrically calibratedchrome-on-glass stage micrometer (NIST No.5525). The line center spacings were measured ona SGIP Universal Measuring Machine, Model MU-214B. The measured and averaged lengths are cor-rected for image distortion which had beenmeasured separately (Appendix B). The result wasan image magnification value valid over the wholefield of view, this value is equal to the on-axis valueprior to image distortion removal. A number of er-ror sources affected the result, as detailed below.a) The object micrometer.

A length 2.10-2.20 mm of the micrometer men-tioned before was used. The length of segment 0-2.11 mm is 2100.58 jim with a maximum error of0.17 jim, that of segment 0-2.20 mm was2199.51 ± 0.11 jim, giving for 2.11-2.20 mm a

257



FIg. 6. Measuring microspheres with MEM.

Table 2. Measurement results with electron microscopy

Microsphere row # Average diameter (>m)

1 3.0082 3.0003 3.0004 2.9985 3.0036 2.9937 2.9968 2.9979 2.997

10 2.98111 2.99412 2.996

Average 2.997

length 88.93 ± 0.20 plm. This corresponds to± 0.227% or 0.0068 psm for a 3 plm length in theobject plane, a systematic error.b) The SGIP film measuring machine.

The scale error amounted to approximately 1.3pm maximum per setting, or about 2 pm for a dif-

ference between two settings. For the measuredfilm distances (89.271 mm mean value) thisamounts to about 0.002% or 0.0001 pm in the ob-ject plane, a systematic error.c) Film emulsion shifts, image magnification scat-ter, and film readout errors.

Polaroid Type 57 positive film exhibits local ran-dom emulsion shifts, like most photographic emul-sions. These lateral shifts are caused bynon-uniform film processing and drying.

Magnification scatter is caused by slight changesin film position in the cassette (measured along theoptical axis) when exposed film is replaced by anew film sheet.

Film readout errors reflect the precision withwhich one can visually pinpoint the center positionof the scale division lines of the photographed ob-ject micrometer.

The combined contribution by these three errorsources was found as follows. The utilized 0.09 mmsection of the calibrated object micrometer wasphotographed five times at 1000 x, giving imagelengths (in mm): 89.807, 89.924, 89.863, 89.903, and

258



Table 3. Error budget for a single 3 Am center distance measurement, using CDFP

Category Error source Error contribution (Am)Systematic Random

On-axis Stage micrometer calibration 0.0068magnification

Film measuring machine 0.0001calibration

Film readout, emulsion 0.0053shifts, and magnificationscatter (5 exposures)

Image distortion uncertainty 0.0020

0.0089

Center distance Film readout and emulsion 0.017measurement shifts

Magnification scatter 0.008

Image distortion-worst case 0.025

Sphere flattening on contact 0.002

Non-sphericity 0.030

Subtotals 0.0089 0.043

Finite sample Diameter distribution width 0.095

size (N = 4600)

Totals 0.0089 0.104

a Uncertainty in D/0.0089+0.104/V4600=0.011 Am. Measured D (after correction, see Sec. 4.1.2): 2.977 Am.

Table 4. Error budget for an average diameter measurement of 3 Am spheres, using MEM,

Category Error source Error contribution (glm)Systematic Random

Microsphere Imperfect scan conditions:sensing scan direction 0.0011

unequal-size spheres 0.0002E-beam exposure 0.0009SEM resolution and 0.0008E-beam wander

Row length Stage travel sampling 0.0009measurements interferometer least count 0.0005

Microsphere Sphere flattening on contact 0.0020grouping air gaps between spheres 0.0020

Sample size Small sample with non-zero °b 0.0055(N= 120)

Totals 0.0030 0.0057

a Uncertainty in D: 0.0030 + 0.0057 = 0.009 Am. Measured 1D (after corrections, see Sec. 4.2.5): 2.990 Am.

259



89.607. The mean was 89.837, with a 3 o scatter of0.142 mm or 0.159%. This total scatter contributesa 0.0048 pm error to a 3 pm center distance mea-surement. After correction for image distortion,see d) below, the mean becomes 89.271 mm, givingan on-axis image magnification M.= 89.271 mm/88.93 pm= 1004 x.d) Image distortion.

The microscope used exhibits radial image dis-tortion: each off-axis image point is shifted radiallyby a small amount from its intended position. Inour case each end point of the 89 mm measuredlength was shifted outwards by 0.30 mm typically,with an estimated maximum uncertainty of 0.030mm (Appendix B). This amounts to a combined0.060 mm uncertainty in the measured length (thetwo error contributions are correlated), or 0.067%,corresponding to 0.0020 ,Lm in the object plane.

The error contributions a) through d) combineto a total systematic error of 0.0089 pm (see Table3).

4.1.2 Errors Associated With the Determina-tion of Microsphere Center Distances The accu-racy of center distance measurements is affected byvarious uncertainties: those associated with (a) pin-pointing the positions of focal spot centers in thefilm, (b) with the correction of measured focal spotpositions due to image distortion, (c) with the fluc-tuations in print magnification when new film isinserted in the cassette, (d) with possible distortionof the spheres at the contact areas, and (e) with thepossibility that the individual spheres might beslightly deformed (showing a non-circular crosssection when measured perpendicular to the line ofsight).a) The combined effect of film readout (pinpoint-ing sphere centers) and emulsion shifts was foundby taking three repeat exposures of a hexagonal ar-ray of the 3 pm spheres, and measuring each timethe same 17 distances between adjacent spherecenters in a selected microsphere row, under com-puter control as described in Appendix A. All 17sets of three center distance readings each werescaled down to the same average value (nominally3.0 mm, as a result of 1000 x magnification of the 3pm sphere objects). The 51 values were thenpooled, resulting in a 3 o- scatter of 17.3 pm whichamounts to 0.017 iim in a 3 ,um object distance, arandom error. As can be seen, this procedure re-duced the effects of magnification scatter andavoided the effects of off-axis magnificationchanges due to image distortion and of unequal-size spheres.

When pinpointing the center positions of the fo-cal spot recordings in the film the utilized coordi-nate measuring machine with TV-microscopeprobe exhibited a reproducibility of better than 0.5jim at 1 C (see Appendix A). This translates to amaximum error of 2 jim in film distances betweentwo focal spots, or 0.002 jim in distances betweenmicrosphere centers. This random error is includedin the above error discussion, and does not notice-ably increase the 0.017 jim random error calcu-lated.b) The effect of image distortion in our case (seeAppendix B) is maximum for a sphere pair at theedge of the measured field of view. At 40 mm off-axis distance the maximum error in the measuredimage distortion is about + 20 jim, at 37 mm it is± 16 j. Assuming as a worst case that these errorsare uncorrelated, the resultant maximum error for3 mm center distances near the edge of the 80 mmfield of view will be ± 25 jim or 0.025 jim in theobject plane, a random error. For center distancescloser to the optical axis this error will be consider-ably less, and for those on the axis the error will bezero.c) Magnification scatter, occurring when replacingsheet film in the cassette, was measured as 0.27%at 3 o for the central area of a single exposure (seeAppendix B). This value is considerably larger thancan be expected from the data in c) of Sec. 4.1.1. Areason is that c) relates to measurements near theedges of the film sheet (the imaged object micro-meter segment filling the field of view), where it isclamped by the cassette mechanism and conse-quently flexes much less. The correspondingmaximum error for a 3 jim center distance mea-surement is 0.008 jim, an essentially random error.It has been applied to all areas in the film, as aworst case.d) One can adapt the model that two polystyrenespheres approaching each other during the dryingprocess will finally be in intimate contact over acircular area, the extent of which is controlled by abalance between van der Waals attraction and elas-tic deformation. This model has been analyzed byDerjaguin et al.; they have derived an expressionfor the resultant sphere flattening [6]. For thepresent case the two-sided flattening wouldamount to a shortening AC of the measured centerdistance C given by

AC = [6(1 _7) 2 DA 2 ]l 3 in which

260



7 =Poisson constant, 0.3 for polystyreneD = sphere diameter, 3 x 10' cmA = Hamaker constant, 1 x 10-12 erg for polystyreneE =Young's Modulus, 3 x 1010 dyne/cm2 for

polystyrenee = distance of closest approach, 3 x 108 cm.

This gives AC= 1.3 nm = 0.0013 jim, lowering themeasured diameter. If the selected values for Aand E are each uncertain by a conceivable factor 2,then AC could change by a factor whose maximumvalue is 3V16 = 2.5. The AC estimate then rangesfrom 0.0005 to 0.003 jim.

Although this model for sphere flattening oncontact is not the only one [7] available, experi-mental data (comparison with other calibrationtechniques for various monosize microsphere Stan-dard Reference Materials) support the Derjaguinmodel. Therefore the measured diameter values inTable 1 are corrected afterwards by a somewhatarbitrary increase of 0.002 jim, and a systematicerror 0.002 jim is entered in the Error Analysis.e) If a microsphere is elongated perpendicular tothe line of sight, its focal spot will be elongated bythe same amount [1]. The photographed focalspots are almost all very uniform and circular, witha diameter of 0.4-0.5 mm in the film plane corre-sponding to 0.5 jim in the object plane. A non-cir-cularity of 0.03 mm is visually detectable, and anyresidual non-sphericity will then not exceed 0.030jim-a random error.

The random contributions combine to a maxi-mum random error of 0.043 jim.

4.1.3 Errors Associated With the MicrosphereDiameter Distribution Figure 4b shows that thediameter distribution is not quite normaL Of themeasured population 98% covers the size range2.89 to 3.08 jim. The maximum error contributionto a single center distance measurement can be setat ± 0.095 Gum-a random error.

4.1.4 Combining the Various Error Contribu-tions for the D Measurement From Table 3 thetotal random error amounts to 0.104/A/4600=0.0015 jim, the total systematic error is 0.0091 jim,therefore the total error in D is 0.011 jim, givingD = 2.977 ± 0.011 jim.

4.1.5 Finding the Standard Deviation AD of theSize Distribution Figure 4 shows that the diame-ter distribution is. normal from 10 to 90% (3700spheres), and the calculated value of eD for thatpopulation is 0.021 jim with a statistical uncer-tainty in cm of ± 14% at 3 ai, or 0.003 jim.

Subtracting in quadrature the 1 or random uncer-tainty in a single measurement of center distance

(equal to 0.043/3 jim or 0.014 jim, see Table 3)would lower oD to 0.017 jim, but the uncertainty inthis value will increase correspondingly.

The reported value for OLD has been set at 0.020jim.

4.2 Errors in Measuring Microsphere AverageDiameter by Electron Microscopy

These error contributions can be divided intofour groups: errors associated with the microspheresensing process, errors associated with the mea-surement, of scanning stage travel, errors associatedwith the kind of grouping the measured micro-spheres are in (single-layer microsphere arrays withhexagonal ordering), and errors associated with themicrosphere diameter distribution width.

4.2.1 Errors Associated With MicrosphereSensinga) Imperfect scan conditions.

Referring to Fig. 7, path 1 represents a possiblescan path, while path 2 produces deeper minima ofSEM output at the contact plane position midwayof the line segments AB. When measuring contactplane spacings scan 2 gives no additional errorsover scan 1.

When the scan path makes a small angle a (radi-ans) with the center line, a cosine error occurs andthe measured spacing is too large by a fraction 1/2a2 . A 10-sphere scan beginning at 1/2 R above thecenter line and ending at 1/2 R below it (a worstcase), has a = 0.05 and will have an error in thecalculated scan length of 0.125% or 0.0375 jim. For12 such scans, covering all 120 measured spheres,the random error in the calculated average diame-ter D is then (0.0375A/12)/120 jim, or 0.0011 jim.

If as a worst case two unequal-size contactingspheres have diameters 2% smaller than D and 2%larger than D respectively, the midpoint of the seg-ment AB in Fig. 7 shifts a calculated 0.0014D or0.0042 jim. If this happens at both ends of the mea-sured contact plane spacing containing N spheres,the maximum error in the calculated value for D is0.0028 DIN. In the present case all 12 measuredrows contained 10 spheres each. When all data arepooled the resulting random maximum error in Dwill be 0.0028 D/(10V12) = 0.0002 jim.

Path 2-Path 1-

Fig. 7. Row length errors (see text).

261

1=1�1. 4=�%AV B AV B I I



b) E-beam exposure.If E-beam exposure causes the polystyrene

spheres to swell or shrink uniformly, the spherescould conceivably move with respect to each otherand with respect to the substrate. A uniformswelling causes in principle increased contact area(sphere flattening) without changing sphere centerdistances, giving no measurement errors. A uni-form shrinking causes in principle no errors if eachsphere remains anchored to the substrate, while er-rors will occur if a sphere can move laterally (eitherby rolling over the substrate and sliding against itsneighbor, or by sliding over the substrate while be-ing in fixed contact with its neighbor).

Lateral sphere motions were not observed, andin a few cases development of a tiny "neck" or"bridge" was found between adjacent spheres, in-dicating that a uniform shrinking might be takingplace there. Tentatively an upper limit of 0.005Dhas been placed on row end point shifts caused bysphere motions, leading to a maximum error of0.01D in the measurement of a 10-sphere contactplane spacing and a systematic error of 0.01D/(10V12) or 0.0009 jim in the calculated D-valuecovering 120 spheres.c) SEM-resolution and E-beam wander.

The combined effect is estimated at 0.02 jimwhen pinpointing an individual contact plane or0.02V2 for a 10-sphere contact plane spacing. Thisgives for 120 pheres a random error in D, equal to0.02V2 x V12/120 = 0.0008 jim.

4.2.2 Errors Associated With Stage Travel(Row Length) Measurementsa) Stage travel sampling.

Each scan is 60 jim long, and covers 2000 equallyspaced data points for a least count of 0.033 jim.This gives a sampling error of 0.033 jim when mea-suring contact plane spacings. For 10-sphere rowsand a total of 12 rows the resultant error in D be-comes 0.0009 jim.b) Interferometer digitizing.

The interferometer has a least count of A/40, giv-ing an error in contact plane spacing of A/40 and aresultant error in D (using A = 0.6328 jim, 10-sphere rows, and 12 rows total) equal to 0.0005jim.

The errors associated with the electronics ofSEM, interferometer, and computer, are consid-ered negligible.

4.2.3 Errors Associated With Contacting Mi-crosphere Groupingsa) Sphere deformation at the contact areas.

At the contact areas sphere flattening can occuras discussed in Sec. 4.1.2. It is in effect a change in

scale, leading to a correction in the measured b setat + 0.002 jim, and a systematic error of 0.002 jim.b) Air gaps between spheres.

Hexagonal arrays of slightly unequal-size spheresexhibit small air gaps throughout the array; the av-erage gap width for normal size distributions isabout 0.46 times the standard deviation of the di-ameter distribution [2,3]. For this microsphereSRM oD = 0.020 jim as found in Sec. 3.1, leading toa diameter correction of -0.46 aD = - 0.0090 Aimwith an estimated systematic uncertainty of 0.0020jim.

4.2.4 Errors Associated With Sample Size Themaximum error in the measurement of average di-ameter of a microsphere material with eD = 0.020jim based on a sample of 120 spheres, will be equalto 3 ab/V120=0.00 55 jim-a random error.

4.2.5 Combining the Various Error Contribu-tions to b The total random error is found as alinear sum of the sampling and digitizing errors,plus an rms sum of the other random contributions,for a total of 0.0057 jim. The total systematic erroris an rms sum of the individual contributions,amounting to 0.0030 jim. The reported total uncer-tainty in D is then 0.009 jim. The mean diameteritself, after applying the corrections for sphere flat-tening (+ 0.0020 jim) and air gaps (-0.0090 jim),is 2.990 jim.

5. Diameter Calibration Final Results

Optical Microscopy

Average diameter of SRM 1962 microspheres:b = 2.977 + 0.011 jim (optical microscopy)qD = 0.020 jim (central peak)

Electron Microscopy

b= 2.990 ±0.009 jim (electron microscopy)

Because the two methods gave noticeably differ-ent b-values and have almost the same total uncer-tainty, the reported value for the average diameterwas set at b = 2.983 ± 0.016 jim.

The quoted uncertainties are maximum values.

6. Sample Uniformity

From Table 1 an impression of sample unifor-mity can be obtained: the within-vial variation inthe measured mean diameter is less than ± 0.1%,

262



and the between-vial variation amounts to a slightlylarger (± 0.10%) amount. Therefore, we take as anupper limit for the SRM non-uniformity: ±0.1%for within-vial and between-vial sampling.

7. Outliers

As with the sample uniformity, only upper limitscould be set to the percent oversize and undersizeof the measured 2000 spheres. When outliers aredefined as spheres with sizes more than 4qD differ-ent from the average diameters, there are about1% oversize and about 1% undersize (Fig. 4b).Spheres that were outsize by some 10% or morewould be detectable by visual inspection of thephotomicrographs. No such spheres were found.

8. Appendix A. Measuring MicrosphereCenter Distances From Photomicro-graphs

The relative positions of the photographed mi-crosphere focal spot positions are read out with acoordinate measuring machine (CMM). Its probeis a microscope containing a high-resolution imagesensor (vidicon) and circuitry to pinpoint the cen-ter of each focal spot, which produces steering sig-nals to the X-Y CMM drives to bring the focal spotcenter to the boresight axis (center of the micro-scope's field of view). The CMM's X-Y coordinatesare then stored.

Next, for each photograph being measured a de-cision is made as to what measurement path will beselected from one focal spot to the next, in order tomake sure that each sphere is measured only onceand that each sphere in the grouping was, free toassume its contacting position with its neighborsduring drying (no mechanically overdeterminedsphere arrangement). The path selection is done byan operator using an interactive graphics routine.The CRT display shows all focal spots in their rela-tive positions with circles drawn around each one,to simulate the actual microsphere scene ratherthan the focal spot representation of it. It alsoshows the keyed-in measurement path (see Fig. 3).The result is a string of X-Y coordinates in whicheach increment represents a center distance (CD)between two adjacent (contacting) spheres.

Before the CDs are computed each focal spotposition is first corrected for image distortion ofthe microscope (this distortion can be determinedas in Appendix B). Because the distortion is a ra-

dial function these position corrections take theform of small radial shifts. Their magnitudes de-pend on the initial off-axis distances, and are foundby means of a stored function or lookup table.

An impression of the repeatability and accuracyof the film measuring system can be obtained fromthe following information:1. The CMM is normally used for the calibration

of precision grid plates. It has been extensivelystudied, and its positional accuracy is betterthan 0.1 jim over the area of interest [8].

2. Each focal spot was centered to the same pointof the camera field, and all the spots were aboutthe same size. Therefore any geometrical errorsin the camera scan can have only secondary ef-fects on the measurement.

3. A photograph selected at random was run anumber of times without being moved. The cen-ter distances (3 mm nominally, the print magni-fication being 1000x) showed a repeatabilitywith standard deviation less than 0.5 jim.

4. One other photograph was measured repeatedlywhile the microscope focus was varied in bothdirections. Since the focal spot brightness wasnot completely isotropic, the focus settingschanged the apparent size and shape of the fo-cal spots by small amounts. The resultantchanges in focal spot positions were very smalland uncorrelated, with an estimated standarddeviation of less than 0.5 jim.

5. A few photographs were reexamined after allhad been measured in order to check for anyprocess changes during the measurements. Nochanges larger than the measured repeatabilitywere found.

The CMM used was a five-axis Moore V fromMoore Special Tools Company. The TV camerasystem consisted of a Dage-MTI Inc. high-resolu-tion vidicon camera Model 65, with a Bausch &Lomb lens type Mono Zoom 7 and a vision systemfrom Videometrix VPU, Model 101110-501-14. Thescene illumination was a diffuse one, using a fiber-optic ring illuminator from Titan Tool Company.

9. Appendix B. Measuring the PrintMagnification (Film Scale) and ImageDistortion

To set the scale of the photographs a section of acalibrated object micrometer is photographed andmeasured. To maximize resolution the pho-tographed length is made to span the whole field ofview. The measured length is then corrected for

263



the effects of image distortion which causes off-axisimage points to be radially shifted from their in-tended positions. The corrected length now showsthe image magnification in the absence of imagedistortion. In other words: one has found the on-axis image magnification value.

For the determination of image distortion themicroscope magnification itself is not needed. Inthe following is shown how these properties weremeasured.a) Image distortion.

Following a scheme outlined in Ref. [1] this goesas follows. A row of microspheres is placed suchthat it crosses the center of the field of view. Itsrow of focal spots is photographed. Then the row isshifted in-line by four sphere diameter (4D) and itsfocal spots are recorded again. All distances be-tween adjacent sphere centers are measured inboth photographs; it will be seen that some centerdistances will have decreased while others in-creased. These changes are so small that an in-lineshift of 4D rather than D was chosen in order tomake the center distance changes stand out frommeasurement noise (see Fig. 8a).

Assuming a 4D shift from left to right, and start-ing with the far left (first) sphere pair, one findsthe accumulated length changes when the centerdistance length D is shifted in-line by 4D at thevarious points in the field of view. A second dataset is found by repeating the length change calcula-tions starting at the next left sphere pair. A thirddata set is obtained from the third left sphere pair,and likewise for the fourth set. The four data setsbelong to a common curve, and a best fit of all foursets is obtained as shown in Fig. 8b.

Figure 8b therefore shows by what percentage acenter distance D will vary as it is shifted all acrossthe field of view (this is equal to the percentchange in magnification for radial objects as afunction of off-axis distance). A graphic integrationthen yields the radial shifts of off-axis image pointsas a function of their off-axis distance, that is, theimage distortion as shown in Fig. 8c.b) Film scale.

With the image distortion known, the film scaleor on-axis magnification M. is found next as shownin Fig. 8b. If M. is measured a number of times theresults will show data scatter, typically 0.3% at 3 oa.This is caused by small changes in axial positionwhen fresh sheet film is inserted in the cassette.The corresponding scale changes can be reducedby averaging over a number of repeat exposures,for instance five, of the object micrometer.

ZO-zuJ<

< 0

zUi0

+50

0

-50

0

S AM_~~~~ * o12 mm

DIAL POSITION

a)

zo-z <0

ZZz <cl: DC) UJ

w<0CC

-40 -20 0 20 40OFF-AXIS DISTANCE (mm)

b)

IMAGE DISTORTION (gum)

200-_

100-_

IAn o II

-,ou U zfU

OFF-AXIS DISTANCE (mm)

1030z0

1020 <0

1010 0

1000

40

c)

Fig. 8. Finding microscope image distortion and magnification.

The effect of inserting fresh sheet film in the cas-sette can be measured by placing a row of micro-spheres such that it crosses the center of themicroscope's field of view, taking five repeat photo-micrographs, and measuring a number of spherecenter distances in all photographs. The scatterfound in these lengths contains the combined ef-fects of film readout, local emulsion shifts andchanges in film scale. As shown in Fig. 9 these dataapproximate a straight line through the origin; theslope of that line represents the scatter in filmscale or magnification (0.09% at 1 oa in our case),for areas near the center of the photographs. Forthis experiment a 10 jm microsphere array wascentered in the field of view, and the variouslengths in Fig. 9 were realized by summing thelengths of a number of adjacent microsphere rows.The found magnification scatter was indicative offilm flexure at the central area of the film frame.

264

.....

-m+uon



'-S

O/

/1

- /

Io

0 50 1 00 mm

MEASURED LENGTH(5 exposures)

Fig. 9. Scatter in image magnification.

10. References

[1] A. W. Hartman, Powder Technol. 46, 109 (1986).[2] G. G. Hembree, Proc. 44th Ann. Meeting, G. W. Bailey, ed.,

San Francisco Press, San Francisco (1986) p. 644.[3] D. A. Swyt and S. W. Jensen, Prec. Eng. 3, 11 (1981).[4] H. E. Kubitschek, Nature 192, 48 (1961).[5] A. W. Hartman, Powder Technology 42, 269 (1985).[6] B. V. Derjaguin, V. M. Muller, and Yu P. Toporov, J. Coll.

Interf. Sci. 53, 414 (1975).[7] K. L. Johnson, K. Kendall, and A. S. Roberts, Proc. Roy.

Soc. Lond. A324, 301 (1971).[8] R. J. Hocken et. al., Ann. C.I.R.P., Vol. 26 (1977).

About the authors: A. Hartman, T. Doiron, and J. Fuare physicists in the Precision Engineering Division atNIST. The National Institute of Standards and Tech-nology is an agency of the Technology Administration,U.S. Department of Commerce.

265

0.2

0.1

E-

I-

C/)

0Cn

Cn

cs

I



[J. Res. Nati. Inst. Stand. Technol. 97, 267 (1992)]

Optical Calibration of a SubmicrometerMagnification Standard


Jon Geist, Barbara Belzer, The calibration of a new submicrometer oxide. The uncertainty in the derivationMary Lou Miller, and magnification standard for electron mi- of a thickness for the layer from the el-Peter Roitman croscopes is described. The new stan- lipsometric parameters is also derived.

dard is based on the width of a thinNational Institute of Standards thermal-oxide film sandwiched between Key words: calibration; electron micro-and Technology, a silicon single-crystal substrate and a scope; magnification standard; scanningpolysilicon capping layer. The calibra- electron microscope; transmission elec-Gaithersburg, M 20899 tion is based on an ellipsometric mea- tron microscope.

surement of the oxide thickness beforethe polysilicon layer is deposited on the Accepted: January 13, 1992

1. Introduction

It was recently proposed [1] that the same cali-bration technique [2] used to produce the NIST2530 Standard Reference Material (SRM) series ofoxide thickness standards could be used as the basisfor a new submicrometer magnification standard.The key ideas for the new standard are the follow-ing: 1) A thin thermal oxide is grown on a 76.2 mm(3 in diameter) silicon wafer, 2) the thickness of theoxide is measured at the center of the wafer withthe NIST High-Accuracy Ellipsometer [3], 3) thevariation in oxide thickness is measured at ninepoints over the wafer using a high-precision reflec-tometer, and 4) a polysilicon cap is deposited overthe oxide layer subsequent to certification of the ox-ide thickness. Sections of the resulting wafer can beused as submicrometer magnification standardswhen viewed edge on. For use in a Scanning Elec-tron Microscope (SEM), the cleaved edge must bedipped in a weak solution of HF for a few secondsbefore viewing in order to create topographic fea-tures. For use in a Transmission Electron Micro-scope (TEM), the standard can be sectioned in theusual way.

This paper describes the calibration of this newstandard and the associated uncertainty analysis.There are two very different types of uncertainty as-sociated with this standard. The first is the uncer-tainty in the thickness of the oxide at any point onthe wafer. This uncertainty, which is reported in thecalibration certificate, is described here. The sec-ond is the uncertainty associated with the resolutionof the oxide-silicon interfaces in an electron micro-scope. This uncertainty will vary depending uponthe machine and the operating conditions, and mustbe estimated by the user. With some SEMs, this un-certainty will be much larger than the calibrationuncertainty, and with some TEMs, it may be smallerthan the calibration uncertainty. In either case,these two uncertainties are added in quadrature be-cause they are uncorrelated. This paper is con-cerned only with the calibration of the oxidethickness, and does not address the uncertainty as-sociated with the actual use of the standard in anelectron microscope.

A model of the optical properties of the magnifi-cation standard is needed to extract the thickness

267



of the oxide layer from the measured ellipsometricparameters. The model used for the silicon/silicondioxide system consists of three layers as shown inFig. 1: 1) a homogeneous, isotropic top layer char-acterized by a film thickness tf and a real index ofrefraction nf; 2) a homogeneous, isotropic inter-layer characterized by an interlayer thickness t; anda real index of refraction ni; and 3) a homoge-neous, isotropic substrate characterized by a com-plex index of refraction n.. The three-layer modelis used because Taft and Cordes showed that atwo-layer model produces an oxide index of refrac-tion that depends upon oxide thickness, which isclearly unphysical [4]. The bottom layer is inter-preted as single-crystal silicon and the top layer asamorphous silicon-dioxide. The interpretation ofthe interlayer in terms of a physical structure is lessstraightforward, and is discussed in detail later inthis paper.

a)

tf

b)

SiO2

ti

,T _

. 1~-- - - -- - -CT6

._------- -------- _

Si

Fig. 1. Optical model of a thermal oxide osilicon substrate that was used to derive an oximeasurements of ellipsometric A and q/ data for the principalangle at 633 nm.

It is not possible to determine tf, ti, nf, ni, and nsfrom measurements of the ellipsometric parame-ters [2] A and qi on a single wafer. The derivationof the values reported in the certificate of calibra-tion for this new standard is carried out on a lot ofwafers as described in Ref. [2]. The remainder ofthis paper describes in more detail how a lot ofwafers is measured, how the oxide thickness is cal-culated, and how the uncertainty in the oxide thick-ness is estimated. The details provided herecomplement those presented in Ref. [2], with theexception that tf+ti is used as the oxide thicknessin Ref. [2], whereas tf+tj/

2 is used for the oxidethickness for this magnification standard. The main

reason for this change is that the ellipsometricstandard is concerned with the thicknesses of lay-ers defined by an optical model, whereas this stan-dard is concerned with the thickness of a physicallayer, and ti/2 is more consistent with alternativephysical models of the optically determined inter-layer. This point is explained in more detail later inthis paper.

2. Definition of a Batch and a Lot ofWafers

For the purposes of this magnification standard,a batch of wafers is defined as a set of at least fivewafers that were put into an oxide-growth furnaceat the same time, so all of the wafers in the batchhave the same growth conditions and same nominaloxide thickness. A batch of wafers is processed andmeasured in the following steps:

1) The wafers are cleaned and dried immedi-c) ately before loading into an oxide-growth furnace.

-k A 2) A nominal thickness oxide is grown on the

tf 3) The relative variation of the oxide thicknesstF is measured at nine points on the surface of each

-f--- wafer as shown in Fig. 2.ti 4) A simple model of the variation in oxide

thickness with position on the wafer is fit to theS c---- data measured in Step 3 above, as discussed later

in this paper.5) The ellipsometric parameters A and + at the

center of each wafer for the principal angle of inci-dence are measured at 633 nm with the NISTHigh-Accuracy Ellipsometer [3].

,n a single-crystal 6) A layer of polysilicon is deposited over thede thickness from oxide.

A lot of wafers is defined to be a set of at leastthree batches of wafers. To qualify as a lot, 1) atleast one batch must consist of wafers with nominal50 nm oxides, another with nominal 100 nm oxides,and a third with nominal 200 nm oxides, and 2) alloxides must have been grown at the same tempera-ture with the same gas-flow conditions in the samefurnace. Batches of wafers with nominal 12 and 25nm oxides may also be included in a lot on a fairlyroutine basis, and batches of other thicknesses maybe included on occasion. The thicknesses of 50,100, and 200 nm were chosen for consistency withRef. [2]; the restrictions on lots and batches ofwafers are required because nf and either ni or t,have been determined to depend upon growth con-ditions [4].

268



Fig. 2. Order in which the thickness of the oxide is measuredrelative to the center of the wafer at the points denoted bycrosses. The crosses are on 2 cm centers. The center is mea-sured as points 1, 6, and 11.

When all of the wafers in all of the batches consti-tuting a lot have been measured, the thicknesses tfof the oxide films on the individual wafers in the lotare adjusted in a least-squares fit [2,5] to the exper-imental A and 4i data based on the model shown inFig. 1. The extinction coefficient k, of the siliconsubstrate at 633 nm is assumed to be the same forall of the wafers in the lot, and is fixed at0.0156 ± 0.0003 [6]. The indices of refraction nS, nff,and ni of the silicon substrate, the silicon-dioxidefilm, and the interlayer, respectively, are assumedto be the same for all wafers in the lot, and areadjusted to produce the best fit. The thickness t1 ofthe interlayers is also assumed to be the same forall wafers, and is also adjusted in the fit [4].

3. Oxide Thickness and Uncertainty

The thickness of the oxide at any point (xy) onthe wafer for which -2 cm < x < 2 cm, and- 2 cm < y < 2 cm can be calculated from the val-ues of tf, ti, a, and 13 reported on the certificate ofcalibration for the new magnification standard byusing

t (xy) =tf +ti/2 +a x +.8 y. (1)

The point (0,0) is the center of the circular exten-sion of the circumference of the wafer as shown in

Fig. 2. Equation (1) assigns tf+tj/ 2 as the oxidethickness at the center of the wafer. The reason forthis choice is associated with the interpretation ofthe interlayer, and is described in the sections ofthis paper devoted to that topic.

The estimated standard deviation of t(xsy) isgiven by

At = (Atf2 + At 2+ ti 2 + ti 2 + Atmap2 + AtPoly2)1'. (2)

The quantity At is the uncertainty in the oxidethickness, and it is reported in the calibration cer-tificate. The uncertainties in a and f3 are so smallthat they do not contribute significantly to theoverall uncertainty, and they are not reported. Thenature of the uncertainties contributing to At andhow they were calculated is described below.

The quantities Atf and at; in Eq. (2) are the un-certainties associated with the adjustment of tf andti in the fit referred to in the previous section. Inprinciple, Atf and Ati vary from wafer to wafer, butin fact they are dominated by systematic errors thatare common to all of the ellipsometric measure-ments, so Atf and At; are the same for all wafers,even from different lots and batches. The first ti inEq. (2) is an uncertainty associated with the use ofthe model of Fig. 1. The second t; is an uncertaintythat accounts for any microscopic roughness of theoxide-silicon interface and any microscopic varia-tions in the oxide thickness. Microscopes are sensi-tive to these variations, but the High-AccuracyEllipsometer is not, due to its macroscopic beamsize, which is measured in millimeters. The uncer-tainty Atmap is the residual standard deviation ofthe fit of the model of Eq. (1) to the measured dataon the variation of oxide thickness over the surfaceof the wafer, and Atpoiy is an uncertainty associatedwith assigning a thickness to the oxide under thepolysilicon on the basis of measurements carriedout before the polysilicon was deposited.

4. Film and Interlayer Thickness

The uncertainties Atf and Ati are determined asdescribed in Ref. [2]. They are one-sigma estimatesfor the sum in quadrature of the uncertainties asso-ciated with both the random errors and the system-atic errors in the measured values of A and qi forthe individual wafers in the lot under the assump-tion that the model in Fig. 1 accurately describesthe wafers. The problem with this assumption isthat different interpretations of the interlayer re-quire that it be apportioned differently betweenthe top oxide layer and the silicon substrate.

269



5. Interpretation of Interlayer

The physical interpretation of the interlayershown in Fig. 1 is not straightforward because dif-ferent experimental results reported in the litera-ture do not appear to be consistent. The interlayerhas been interpreted as a graded layer of SiO.where the x value varies from 0 on the silicon sideof the interlayer to two on the oxide side of theinterlayer [4]. This interpretation is consistent withthe fact that the optically derived thickness for theinterlayer varies from 0.5 to 2.0 nm, dependingupon the particular samples constituting a lot ofwafers. This interpretation is also supported byx-ray photoemission spectroscopy (XPS) measure-ments that detect silicon in every allowed oxidationstate at the interface [7]. According to this inter-pretation, most of the interlayer (if not all of it)should be considered part of the oxide.

On the other hand, TEM [8] and grazing inci-dence x-ray scattering measurements [9] on ther-mal oxides grown between room temperature and900 0C on (100) silicon surfaces detect only a singleatomic layer of silicon that is not bonded either likesingle-crystal silicon or like silicon dioxide. This re-sult is not necessarily inconsistent with the XPS re-sults. Some of the dangling bonds associated withthe specially bonded layer of silicon atoms could bedimerized or tied up by bridging oxygen atoms tocreate the various XPS peaks associated with theremaining oxidation states. (It does not seem possi-ble to determine the concentrations of differentspecies from the heights of the XPS peaks.)

Even though the TEM and grazing-incidencex-ray scattering results are not necessarily inconsis-tent with the XPS results, the former strongly sug-gest that the interlayer should be 0.5 nm thick, andshould not vary from lot to lot. These results alsopredict that the interlayer should be consideredpart of the silicon in some experiments such asTEM measurements, but part of the oxide in otherexperiments such as those in which etches removethe atypically bonded atomic layer of silicon atomsas well as the oxide.

A possible resolution of this paradox is thatsome portion of the optically derived width of theinterlayer is an artifact that produces a better fit tothe experimental data because it involves an extrafree parameter, but does not correspond exactly toany real structure at the interface. For example,the interlayer in the optical model might be com-pensating for the polarization properties intro-duced by the roughness of the silicon surface andthe microscopic variations in thickness of the oxidelayer [10]. This situation is illustrated in Fig. 3.

SiO 2 nf If

Interlayer ni ti

Si ns

Fig. 3. Schematic illustration of plausible worst-case limits a)and c) between the film and interlayer thicknesses tf and t i in theoptical model and the average oxide thickness tF if the interlayeris an artifact that describes the change in polarization inducedby the roughness at the oxide-silicon interface as shown in b). Ifthis is the case, tf and ti are adjusted to produce the same polar-ization of the radiation leaving the oxide surface as is producedby the roughness and IF, and t

F need not equal if.

6. Microscale Roughness

Figure 3b represents a rough interface betweenthe oxide and the silicon substrate for spatial fre-quencies great enough so that the top surface ofthe oxide is not conformal with the interface underthe assumption that such spatial frequencies exist.The average oxide thickness is denoted by tF andthe root-mean-square (rms) roughness is denotedby a, and it is assumed that the ellipsometricparameters A(a) and if(a) depend upon a. There-fore, tf and t;, which are obtained by fitting the op-tical model to the measured A and V4 data for awafer, also vary with a. In fact, there is no reasonthat tf should be identical to tF even though thisresult would be intuitively satisfying.

The only way to definitively determine the de-pendence of tf and t; on a is to solve Maxwell'sequations for oblique incidence on a rough surface.This is not within the scope of this paper. However,there is some evidence that ti = a for a < 5 nm[11], and Figs. 3a and 3c shows worst-case esti-mates for the limits of variation of tf with a. Forone limit, the average film thickness is given by

tF=tf+ tj+ a, (3)

and for the other limit it is given by

tF=tf O.

270

(4)

1



This range is covered by

tF =tf + ti/2 ± (ti/2 + a). (5)

This is the justification for the first two terms inEq. (1).

The range + (ti/2 + o-) represents a limit of error.What is needed for Eq. (2) is something more likea one-standard deviation estimate. This can be ob-tained by multiplying the range by 2/3. If we nowuse ti as an approximation to ci as suggested in Ref.[11], the uncertainty becomes ±ti. This is the firstti appearing in Eq. (2).

The uncertainty just described accounts only forthe model uncertainty. An additional ± o- must beadded to allow for the possibility that the micro-scope might sample the oxide thickness at a crest, atrough, or or any other part of the micro-roughoxide-silicon interface illustrated in Fig. 3. Onceagain, we approximate or by ti, and obtain thesecond t, appearing in Eq. (2). Interpretation of theentire interlayer as an artifact of roughness asshown in Fig. 3 is a worst-case scenario. The otherinterpretations of the interlayer that were dis-cussed above are all described by Eq. (5), and thestated uncertainties should be sufficiently conser-vative.

(Xxn tn)

2 4 mm2 (7)

(Xy. 0 8)24 mm (8)

and

A Atmap / 9.4 mm,

where A is the uncertainty in the coefficients a and/3.

Table 1. Relation of position coordinates x andy to measure-ment number in uniformity map of oxide thickness

Measurement x yNo. (cm) (cm)

1 0.0 0.02 -2.0 2.03 0.0 2.04 2.0 2.05 2.0 0.06 0.0 0.07 -2.0 0.08 -2.0 - 2.09 0.0 -2.0

10 2.0 -2.011 0.0 0.0

7. Variations in Oxide ThicknessBefore the polysilicon is deposited on the wafer,

the thickness t. of the oxide on each wafer ismeasured with a reflectometer for the points n = 1,2, 3, ... , 11 indicated in Fig. 2. The parameters a,/3, and t in the equation

tn = t + ax,, +3By" (6)

are adjusted in a least-squares fit to the measureddata points, where xn and y. are given in Table 1.The parameters a and /3 describe the variation ofoxide thickness about the thickness at the center ofthe wafer, and are reported in the calibration cer-tificate. The residual standard deviation of the fit isused as Atmap in the calculation of the uncertaintyin Eq. (2), and if any single residual is larger than0.6 nm, the wafer is rejected for use as a standard.The value t is not reported, since tf+ti/2, which isderived from the measurements made with theHigh-Accuracy Ellipsometer, is a more accurate es-timate of the thickness of the oxide at the center ofthe wafer.

The symmetric location of the points in Fig. 2about the wafer center greatly simplifies the least-squares analysis [12], so

8. ThicknessPolysilicon

of the Oxide Under the

The following experiment was conducted to setan upper limit on the change in thickness of theoxide caused by deposition of the polysilicon layer.Seven wafers with oxides with nominal 100 nm ox-ides were selected, and the oxide thicknesses weremeasured at the center of each wafer with theHigh-Accuracy Ellipsometer. A layer of polysiliconsimilar to that used on the new submicrometermagnification standard was deposited over the ox-ide on each wafer, and was removed after thewafer cooled to room temperature by etching in 14M KOH for 30 min at room temperature [13]. Thewafers were then rinsed for 15 min in deionizedH20 at room temperature. The thicknesses of theoxides were then remeasured. The average value ofthe difference between the oxide thickness beforeand after the deposition and etch was 0.6 nm with astandard deviation of the mean of seven measure-ments of 0.1 nm. The difference is used for Atp.,y inEq. (2). This is a conservative estimate because thedecrease in thickness of the oxide layer is morelikely associated with the removal of the polysiliconthan with its deposition.

271



9. Conclusion

Typical values of At and of the terms in Eq. (2)that contribute to it are listed in Table 2. It is clearthat the uncertainties of magnitude t1 completelydominate the other uncertainties listed in Table 2in their contribution to At through addition inquadrature. As far as this standard is concerned,there is no reason to attempt to decrease any of theother uncertainties listed in Eq. (2) until the physi-cal meaning of the interlayer derived from the opti-cal model is clearly understood, so that theseuncertainties can be replaced by smaller values.Nevertheless, the uncertainty At in the calibrationof the oxide thickness is quite satisfactory for manyapplications, and will, in fact, be much smaller thanthe errors associated with the use of the standardin many electron microscopes.

[11] J. R. Blanco and P. J. McMarr, Appl. Opt. 30,3210 (1991).[12] N. R. Draper and H. Smith, Applied Regression Analysis,

Wiley, New York (1966) p. 58.[13] D. L. Kendall, Ann. Rev. Mater. Sci. 9, 373 (1979).

About the authors: Jon Geist and Peter Roitman arephysicists, Barbara Belzer is a physical science techni-cian, and Mary Lou Miller is an electronics technicianin the Semiconductor Electronics Division of theNIST Electronics and Electrical Engineering Labora-tory. The National Institute of Standards and Tech-nology is an agency of the Technology Administration,U.S. Department of Commerce.

Table 2. Typical values for the uncertainties in the oxide thickness

Error term Value Source of error(nm)

Atf 0.5 tf from fit4ti 0.4 ti from fit4 1.4 meaning of interlayerI, 1.4 possible roughness of interfaceA1.3p 0.3 variation in oxide thicknessAtp-ly 0.6 deposition of polysiliconAl 2.2 sum in quadrature

10. References

[1] P. Roitman, unpublished.[21 G. A. Candela, D. Chandler-Horowitz, J. F. Marchiando,

D. B. Novotny, B. J. Belzer, and M. C. Croarkin, Natd. Inst.Stand. Technol., SP 260-109 (1988).

[3] G. A. Candela and D. Chandler-Horowitz, SPIE 480, 2(1984).

[41 E. Taft and L. Cordes, J. Electrochem. Soc. 126, 131(1979).

[5] J. F. Marchiando, NatI. Inst. Stand. Technol., SP 400-83(1989).

[6] J.Geist, A. R. Schaefer, J-F. Song, Y. H. Wang, and E. F.Zalewski, J. Res. Natl. Inst. Stand. Technol. 95, 549(1990).

[71 M. Niwano, H. Katakura, Y. Takeda, Y. Takakuwa, andN. Miyamoto, J. Vac. Sci. Technol. A9, 195 (1991).

[8] A. Ourmazd, D. W. Taylor, J. A. Rentschler, and J. Bevk,Phys. Rev. Lett. 59, 213 (1987).

[9] G. Renaud, P. H. Fuoss, A. Ourmazd, J. Besk, B. S.Freer, and P. 0. Hahn, Appl. Phys. Lett. 58, 1044 (1991).

[10] T. Yamnaguchi, J. Lafait, A. Bichri, and K Driss-Kodja,

Appl. Opt. 30, 489 (1991).

272




Probe-Position Error Correction inPlanar Near Field Measurements at 60 GHz.

Experimental Verification


Lorant A. Muth This study was conducted to verify that field data as successfully as from simu-the probe-position error correction can lated data; some residual errors, which

National Institute of Standards be successfully applied to real data ob- are thought to be due to multiple re-and Technology, tained on a planar near-field range flections, residual drift in the measure-Boulder, CO 80303 where probe position errors are known. ment system, and residual probe

Since probe position-error correction is position errors in all three coordinates,most important at high frequencies, are observed.measurements were made at 60 GHz.Six planar scans at z positions sepa-rated by 0.03A were obtained. The cor- Key words: error correction; experi-rection technique was applied to an mental verification; planar near fields;error-contaminated near field con- ve-position errars.structed out of the six scans according pto a discretized periodic error function.The results indicate that probe positionerrors can be removed from real near- Accepted: December 24, 1991

1. Introduction

In this study I establish the correctness and effec-tiveness of the probe-position error correction usingreal rather than simulated data. All of the requiredtheory has been thoroughly discussed in [1] and [2]for theplanar and spherical error correction, respec-tively. In these publications we demonstrated thatcomputer simulations using exact near fields andcomputer-generated error-contaminated near fieldssuccessfully produced error-corrected near fieldsthat agree with the original error-free near fieldswithin fractions of a decibel in amplitude and towithin a fraction of a degree in phase. The methodhas been shown to work for errors as large as 0.2 Aat 3.3 GHz. Similar results were obtained with the-oretical computer simulations at 60 GHz in the con-text of this study. Such results indicate that thetheoretical formalisms appearing in [1] and [2] arecorrect, independent of the frequency or the near-field pattern.

However, an important aspect of the correctionproblem has not yet been addressed. We must ex-amine the error correction procedure in the pres-ence of

(a) multiple reflections in measured near fields(b) residual drift in measured near fields, and(c) residual errors in the probe's position,

which are experimental factors not taken into ac-count in the theory nor in the numerical simula-tions. In addition, any probe position error functionused with real data is necessarily discretized ratherthan continuous, and the extent to which this effectsthe success of the correction procedure is not im-mediately apparent. Therefore, the technique mustbe robust and stable enough that the introductionof these additional uncertainties (experimental ef-fects) into the procedure does not destroy its useful-ness. For example, the uncertainties due to multiplereflections are of the order of or larger than the

273



uncertainties due to probe position errors [3]. Wecan consider these two effects to be independentand, therefore, expect that the error correctiontechnique will not fail with real data.

This work completes a long series of studies. Itestablishes the theoretical correctness of the error-correction formalism, even when used with datacontaminated by experimental effects, and demon-strates the practical usefulness of the error-correc-tion procedure.

We first present a brief overview of the theoreti-cal concepts needed to understand probe-positionerror correction. Next, we describe the experimen-tal design and measurements we obtained to cor-rectly implement the error correction, and, finally,we observe what effects the experimental contami-nations, as described in (a), (b) and (c) above, haveon the results.

2. Theoretical Review

In planar near-field measurements, data are, inreality, taken on an irregular plane on an irregulargrid. We denote an irregular grid on which mea-surements are taken byx + Ax, where x =x(x,y,z) isan exact set of grid points separated by constantincrements in x and y, z =zO is a constant, and6x(x,y,z) is the deviation of the position of theprobe from the exact grid position at (x,y,zo). Weassume that the probe-position error function3x(xy, z) is known at every point of measurement.If the near field that exists at the exact grid posi-tions is denoted by b(x) and the measured nearfield is denoted by b (x + 6x), then we can write

b (x + Ax) = (1 + T)b (x), (1)

where (1 + T) is the infinite Taylor series operatortaking a continuous function from a point to aneighboring point. An error-contaminated near-field function can now be defined as

b(x;8x) 3 b (x + Ax), (2)

which means that the error-contaminated nearfield exists on a regular grid x, and computationsusing fast Fourier transforms can be performed onit. Equations (1) and (2) yield

b(x;6x) = (1 + T)b (x). (3)

The error-correction technique can be stated math-ematically as

b (x) = (1 + T) -' L (x; 3ix), (4)

which implies that the error-free near field on a reg-ular grid can be obtained from the error-contami-

nated (measured) near field by obtaining theinverse of the operator 1 + T, and applying this in-verse to the measurements. This leads to the well-ordered small parameter expansion (to fourthorder),

b(x) = (1 -t 1 -t 2 -t 3 -t 4 +t 1t 1+tlt 2 +tIt 3 +t 2 tl +t 2 t2

+ t3th - tititl -t112t1 - tltlt2 - t2tltl + tltltlt1) (x;8x), (5)

where t,, = 1/n !(xk)"(d/dxk)', the n th-order termof the Taylor series for the coordinate Xk. For fur-ther details and in-depth discussions see [1,2],wherein the method of computing the requiredderivatives and the structure of the individualterms in Eq. (5) are discussed further. Also, for athorough documentation of the the software pack-age developed to implement the error-correctiontechnique see [4].

3. Experimental Design andments

Measure-

To enhance the high-frequency calibration capa-bility at NIST, we decided to test the probe correc-tion procedure at 60 GHz using a center-fedCassegrain parabolic reflector antenna of 0.5 m indiameter. Since the wavelength A= 5 mm at thisfrequency, a probe-position error as small as 1 mm=0.2 A can lead to significant errors in the mea-sured near fields, and, consequently, in the farfields obtained from such error-contaminated nearfields. For example, the phase error would be anunacceptable 72° for a plane wave.

A schematic of the sequence of measurementsand the experimental design parameters is shownin Fig. 1. Given planar near-field scans labelledwith indexes 0... N, and separated by small dis-tances 8z, we can construct any number of error-contaminated near fields by selecting a near-fieldvalue at each (xy) point in the scan area accordingto the scan plane index. The index can be specifiedusing any criteria that uniquely assigns the integers0 to N at each point in the scan plane. When thecorresponding near-field values are thought of aspart of a single scan, an error-contaminated nearfield is obtained. This field can be arbitrarily as-signed to exist at zo, without loss of generality.

For the error-correction study we have used theindex function

j = int [Ncos2(3 ax) cos2(3 ay) + 0.5],

where a= 2rr/L, L is the length of the scan planein centimeters, and int is the fortran truncate-to-

274

Journal of Research ofVolume 97, Number 2, March-April 1992

the National Institute of Standards and Technology

C 0 n p(o 90ToTo TanTh0 xO G

6zL 0 (9(9

r-,\0U 0 G1_n U 0o Qo G

z5

z4

z3

Z2

ZI

ZOo

Q residual probe position errors

z= 0.03 A, which represents a phase error ofAO = 10.80 in a plane wave, was chosen as the dis-tance between the individual scans (see Fig. 1). Thisdistance is also greater than the residual positionerrors of -0.04 mm known to exist on the accu-rately aligned planar near-field range at NIST [6].

a)

a)

0)in3

aQFig. 1. The experimental design used to verify the probe posi-tion error-correction technique at 60 GHz. The six scan planes zoto Z 5 were separated by 8z = 0.03 A, so that the maximum possibleprobe position error is Az = 0.15 A. The residual errors are shownschematically.

40

32

24

16

40

32

24

16

21:46:10 29 JUL 1991

++++++++ + + + 4- + + + + +

+ + + +. .- + A+. + +. . +

+ + +9 + 4- + + =+ + +

01:08:50 30 JUL 1991.+ .-f + + - . = - + + + o + =

+ + + + 4- l l + l l

'+_ + + .+ 4- .+ + + .+ A +

U 5 1 U 1 5 20 25 30 35Time (min)

40 45 50

integer function. The discretized probe position-error function is then [z] =j6z, and the maximumprobe position error Az=N8z. Previous simula-tions using the continuous probe position-errorfunction z = Azcos2(3ax)cos 2 (3ay) have demon-strated the success of the error-correction tech-nique at 3.3 GHz [1] for Az =0.2A. The choice ofperiodic error functions was motivated by the factthat periodic position errors induce high sidelobeerrors in the far field.

The magnitudes of the experimental parameters6z and Az =Naz, where N + 1 is the number of scan

planes, were chosen to ensure that system imperfec-tions do not overwhelm the experimental resultssought. Residual system drift during a tie-scan [5]and repeatability of the laser positioning systemused to locate the scan planes are the relevant ex-perimental factors here. If 8o denotes the systemphase drift during At, the time needed to completea tie scan, and AO is a typical phase difference be-tween the near fields at successive scan planes, thenwe must have AO >30 to be able to isolate positionerror effects. Similarly, phase errors due to re-peatability must be much less than AO.

3.1 System Drift

Figure 2 shows the phase as a function of time atthe normalization point in a scan plane over a pe-riod of a few hours. The crucial experimentalparameter here is the maximum expected phaseshift So during St 2 min, which was the time re-quired to obtain a single tie-scan. Less stringently,an average or most probable phase shift can beused. From Fig. 2 we estimate that S- 4°. Hence,

Fig. 2. The phase drift in degrees as a function of time as ob-served at the normalization point. Each horizontal box repre-sents 20 min. The residual drift is estimated to be -4°.

3.2 Repeatability of the Laser Positioning Sys-tem

The scan planes were located using a laser posi-tioning system that could measure translations inthe z direction accurately at the normalization pointof the scan. The initial point was located in the zOscan plane, and the distance between the probe andthe antenna was increased manually by 3z (read ona digital display) seven times. This whole cycle wasrepeated eight times. As shown in Fig. 3 the relativenormalization values in each scan plane wererecorded at the time the scan plane was located. Weobserve that as the probe is moved from the zo scanplane, the variability in the amplitude of the signalincreases, until at z5 the variability begins to be largeenough to compete with the variations as we stepfrom scan plane to scan plane. Hence, the scanplane at zs was chosen to delineate the maximumacceptable probe position error Az under theprevailing experimental conditions, givingAz =53z =0.15A. An examination of the re-peatability of the phase reveals no additional prob-lems with this choice. The variability observed herecan be attributed to system drift and to human errorin positioning the probe. These two effects cannotbe separated without further study, and no attemptwas made to do this. Figure 3 also shows theoreticalnormalization values in each scan plane. This is fur-ther discussed in Sec. 6 when multiple reflectionsare considered.

275

Az

V

n



Figure 4 shows the amplitude of the near fieldmeasured at zo. The scan area was 1 x 1 i

2 , anddata were taken at 2 mm intervals in both the x andy directions. The resulting 501x501 data matrixwas zero-filled to get a 512x512 dataset, whichwas then used in all subsequent analysis. The near-field amplitudes obtained in the scan planes z1 to Z5

.5

IS

E

0

0.14

0.10

0.06

0.02

-0.02

-0.06 -

are almost indistinguishable from the one shown;hence they are not included here.

Figure 5 shows composite plots of the far fieldsobtained from the six near fields. By definition,error-free near-field measurements should yieldthe same far field; hence, the vertical spread amongthe lines in the plots is a measure of the

50

t70_D

,50 //

C,0

O 0

E 30

C 2 theoretical values20

C 10_

0O - - - -s l 5 in d 0 1 2 a i

z scan-plane indlex z scan-plane indlex

Fig. 3. The repeatability of measurements. The set of measurements show the relative amplitudes (left) and phases (right) at thenormalization point as a function of z scan-plane index. Errors increase as z increases. Theoretical normalization points, partially elim-inating multiple reflection effects, are also shown.

co CN

(D

N

Fig. 4. The near field of the center-fed Cassegrain parabolic reflector antenna used in the experiment. The antenna was 0.5 min diameter and operates at 60 GHz.

276

o.Ila



experimental factors present, as discussed in Sec. 1.Of these, multiple reflections are thought to be themost significant, but an independent verification ofthis interpretation at 60 GHz has not been pur-sued. We also observe greater vertical spread inFig. 5a than in Fig. Sc, indicating that errors in thex coordinates might be greater than in they coordi-nates.

4. Error-Contaminated Near Fields

We want to establish to what extent error correc-tion is effected when it is applied to real rather thansimulated data. To accomplish this we compare twoerror-contaminated fields created computationally(which contain no experimental contamination) andthe error-contaminated fields obtained experimen-tally.

Figures 6 and 7 show the continuous and discreteprobe position-error functions used to obtain simu-lated error-contaminated near fields. We per-formed simulations using the discrete errorfunction to observe any possible significant conse-quence to the error-correction technique due to dis-cretization, although no serious differences wereexpected.

Figure 8 shows the amplitude and phase of theratio of the error-contaminated and error-free nearfields when the continuous error function was usedto generate the error-contaminated field; Fig. 9shows the corresponding fields when the discreteerror function was used. The nine-lobe structure ofthe error function is clearly discernible in each ofthese plots. In the case of discrete errors the error-contaminated region is more sharply delineatedfrom the background, where the error in the ampli-tude ratio is 0 dB, and the phase errors show up indiscrete steps rather than as a continuous surface.These features were expected, as they mirror thestructure of the error function. As is shown in thenext section, these differences in the error-contam-inated fields do not effect the success of the errorcorrection to any significant degree.

We constructed an error-contaminated near fieldfrom the measurements according to the techniquedescribed in Sec. 3. The amplitude and phase of theratio of this field to the error-free near field ob-tained at zo are shown in Fig. 10. There seems to bemore rapid and pronounced oscillations in the am-plitude errors when compared to the errors shownin Fig. 9. We cannot tell at this point to what extentsuch a difference will degrade the effectiveness ofthe error correction.

To further observe the differences between simu-lated and measured error-contaminated near fields

discussed in the preceding paragraphs, we enlargedthe center (main beam) region of the plots shown inFigs. 8-10. Some small-scale features in these plotsare worthy of observation. We observe that the sim-ulated error amplitude ratios (Figs. 11a and 12a)differ little, but that the measured error amplituderatio (Fig. 13a) is significantly different when com-pared to simulations. Further the steps in the dis-crete phase plots are a lot smoother for thesimulated field (Fig. 12b) than for the field con-structed from data (Fig. 13b). Each of the effectslisted in the introduction, multiple reflections, resid-ual dnft, and residual probe position errors, has prob-ably contributed to this difference. We anticipatethat this difference in the phase will effect the accu-racy of the error correction, since near-field phaseerrors are the most significant source of errors inthe far field. We cannot, however, predict to whatextent the error correction will be effected, sincethe whole procedure is a nonlinear function of theerror fields and of the position errors [see Eq. (5)].

5. Error-Corrected FieldsWe applied the error-correction technique

[Eq. (5)] to the error-contaminated near fields dis-cussed in the previous section. Again, a comparisonof the results of the error correction applied to sim-ulated and to measured data will reveal the extentto which experimental effects in the data confirmthe success of the technique. We also transformedall near fields to obtain the corresponding far fields.As will be seen in the subsection below, comparisonof the far fields demonstrates that the error-correc-tion technique is successful for real data both in themain beam and sidelobe directions. In the sidelobedirections, however, the simulated data yieldslightly better results.

5.1 Error-Corrected Near Fields

In Figs. 14 and 15 we show the main beam por-tion of the ratios of simulated error-corrected nearfields to the error-free near field for the continuousand the discrete probe position-error functions, re-spectively. Both the amplitude and phase surfacesseem to be well defined with a few minor irregularfeatures, and the magnitudes are very close to 0 (acase which would indicate no residual errors). Thestructure of the residual error surfaces clearly re-flect the continuous or discrete nature of the posi-tion error function. The phase surface in Fig. 15bclearly brings out the nonlinearity of the correction,since each step in the error-contaminated field is ofequal height originally (see Fig. 13b).

277



Figure 16 shows the result of the error correctionwhen measured data are used. The well-definedstructures observed in Figs. 14 and 15 are no longerpresent. When the error-contaminated amplituderatio (Fig. 13a) is compared with the error-cor-rected amplitude ratio (Fig. 16a), we observe someimprovements, but the success of the error correc-tion is not obvious. The well-defined residual errorsurface observed with the simulated data is not ap-parent. Comparing the error-contaminated phasedifferences (Fig. 13b) with the error-correctedphase differences (Fig. 16b), we see that the dis-crete phase structure has been altered to what isessentially a random surface made up of singlepoint ridges and spikes. The residual error surfaceobtained with simulated data is not discernible;most likely it is buried beneath the spikes. Never-theless, the lack of a discrete structure, togetherwith a decrease in the maxima of the phase plots,indicates that phase correction has occurred.

5.2 Error-Corrected Far Fields

In Figs. 17-19 we show the results of near-fieldto far-field transformations. In these center-cutplots we superimposed the error-contaminated,error-corrected and error-free far fields. (Thesedatasets are drawn with solid circles connectedwith dotted line, open circles connected with asolid line, and a solid line, respectively.) Figure 17shows these results obtained from measured datafor the full range of kx and ky; because of the den-sity of points, essential details of the results are notdiscernible, except at maxima and minima of thecuts, where the success of the error correction isapparent. To observe more detail, we enlargedthese center cuts between kx, = y +0.3 both forsimulated and measured data

In Fig. 18 we compare the results of the errorcorrection using simulated error-contaminatedfields. Here only single solid lines connecting theopen circles are observed. This is readily inter-preted to show that the error-corrected and error-free values overlap on the scale of this plot; that is,the error correction fully succeeds. Indeed, themaximum error in the spectrum [7] that appears atkx =kky = (127r/L )k = 0.03k induced by the periodicerror function used in this simulation is fully re-moved.

In Fig. 19 we compare the results of the errorcorrection using error-contaminated fields ob-tained from real measured data. We now see twosolid lines, one with open circles, that do not over-lap everywhere, and solid circles connected withdotted lines. We observe the difference between

the error-free and error-corrected lines to be muchless than the difference between the error-free anderror-contaminated data at most of the data points.We thus conclude that we can almost recover theerror-free far field from measured error-contami-nated near fields by applying the probe-position er-ror correction, but residual errors will be present. Weagain observe that the maximum error in the spec-trum at kx =ky =(12n7/L)k = 0.03k is removed, asin Fig. 18, but a very small residual error is visiblein this region. The magnitudes of these residual er-rors, and those elsewhere, are at acceptable levels.

6. Suggestions for Further Study

The residual errors observed in Fig. 19 are dueto the experimental factors enumerated in the in-troduction. Further reduction of these errors couldbe obtained by altering the measurement and/ordata analysis process with such sources of experi-mental errors in mind. Some steps that need orcould be taken to improve accuracy of measurednear fields will be briefly examined here.

6.1 Multiple Reflections

Standard near-field measurement seeks to mini-mize multiple reflections by appropriately choosingthe distance between the probe and the antennaunder test. Residual multiple reflections, stillpresent in the data, could possibly be removed bysome filtering and/or averaging technique. In Fig. 3we have shown theoretical normalization values,which were derived by transforming the near fieldat zo to obtain the far field, filtering the evanescentmodes, and then transforming back to the variousnear-field scan planes. This procedure providesnormalization values for a fictitious near field thatis a composite of the actual (filtered) near field ofthe antenna and the real modes of the multiplereflections present in the data. When such near-field data are transformed from scan plane to scanplane, the transformations occur in a reflectionlessenvironment. Alternatively, the far fields obtainedfrom the various near fields can be averaged andtransformed back to a near-field plane. In thismanner some of the multiple reflections can be re-moved from the data, and the effect of multiplereflections on the error correction technique canbe studied. Neither of these approaches, however,removes all of the multiple reflections, since multi-ple reflections are not treated as a realistic func-tion of x and y. Currently, no manageable analytictechnique that can accomplish this is known.

278

Journal of

1 0

0

-1I0

-o

-0(0-003

E0

-20


Research of the National Institute of Standards and Technology

30 k

40 -

50 I

-60

- 1.2 -0.8 -0.4 -0.0

Kx/k0.4 0.8 1.2

Fig. Sa. Nonoverlapping far fields as functions of kg derived from the near fields measured on six scanplanes zo to z5. Theoretically, the far fields should be exactly the same, so the differences can be safelyattributed to experimental effects as discussed in the text.

- 1.2 -0.8 -0.4 -0.0

Kx /k

Fig. Sb. The phases of the nonoverlapping far fields

279

0.4 0.8 1.2

shown in Fig. 5a.

-70

200

1 60

1 20

Q)

CD)(F)_0

80

40

0

-40

-80

- 1 20

- 1 60

- 200



10

O

1 0

1--

Qn

0)

E

-20

30 K

40 o

50 K

60 I

-70 I I I I I l I

1 .2 -0.8 -0.4 -0.0 0.4 0.8 1.2

Ky/ k

Fig. 5c. The amplitudes of the nonoverlapping far fields as functions of k, (see Fig. 5a).

200 r

1 60

1 20

80

40

0

-40

-80

- 1 20

- 1 60

-200- 1.2 -0.8 -0.4 -0.0

Ky/ k

0.4 0.8

Fig. 5d. The phases of the nonoverlapping far fields as functions of ky (see Fig. 5a).

280

0)0)

0)

0-)0)7)

0-cCx

1.2



, 0

0%

:00)

00

C)

Fig. 6. The theoretical probe position-error function used in the simulations. The error surface is given byz = Azcos2 (3ax) cos2 (3ay),where a= 2i7r/L.

0 "I

. _

C)N00) (9

Fig. 7. The discrete probe position error function. The discrete error surface is given byz = ez int[Ncos 2 (3ax) cos2 (3ay) + 0.5], whereN = 5 is the maximum scan plane index, int is the fortran truncate-to-integer function, and a = 27r/L .

281

so



I...I I l, II.,,1 I I1 11~lL ,III I

I, 1 ll

I!

I .,1

7-~' 11l1 1 1 1 j 1 ,1

Fig. 8a. The ratio of the simulated error-contaminated and error-free near-field amplitudes when the continuous error function in

Fig. 6 is used.

OD

co

N

U)

(I

Qc)

1~O

ci)-~0

ci)

(0(3'IQ-r

(N'

I

00)

Fig. 8b. The phase difference between the simulated error-contaminated and error-free near fields when the contin-

uous error function in Fig. 6 is used.

282

0

0

1---\ 01

\ 01

ci)

-0 zc3

0 C

I1I

I 111' I

-

I I



01

,'I-

o

n

'--- 01

QCcN

\_- 01

Q)

_T01

0_101

E I

rI

I I

1 liIII II I.11 .

I . ,I Ii II i ,Il

.~ F ,1I ' I

Fig. 9a. The ratio of the simulated error-contaminated and error-free near-field amplitudes when the discrete error function inFig. 7 is used.

0D0o

01(N

W

ci)

Qi)

(FjQ_

-Cm_ o_

(N

0)

I1 II1

Fig. 9b. The phase difference between the simulated error-contaminated and error-free near fields when the discreteerror function in Fig. 7 is used.

283

III1

1!11 I.II11 1� P �1 1�



-o

4i)

Fig. 10a. The ratio of the error-contaminated and the error-free near-fieldused to construct the error-contaminated near field from measurements.

000

,C_

Ci)

cm

Qi)

Ci)

Q)0

co

I

1 II

.II I I ,l. Ii'a

PO.

amplitudes when the discrete error function in Fig. 7 is

.1

Fig. 10b. The phase difference between the error-contaminated and the error-free near-fields when the discrete errorfunction in Fig. 7 is used to construct the error-contaminated near field from measurements.

284

I ,1 11

,==�

"71,11,

I



cc

Q)

-00

Fig. 11a. The center (main beam) portion of the the ratio of the simulated error-contaminated andamplitudes when the continuous error function in Fig. 6 is used (see Fig. 8a).

0o F

0cc

'-N

Q)

ci)

ci)

Q.)U)0

I

0cm

error-free near-field

Fig. 1ib. The center (main beam) portion of the phase difference between the simulated error-contaminated and error-free nearfields when the continuous error function in Fig. 6 is used (see Fig. 8b).

285

P.,

I _11';_�

I

O:1-

I- -



p.-

II v�4./

'V

Fig. 12a. The center (main beam) portion of the ratio of the simulated error-contaminatedamplitudes when the discrete error function in Fig. 7 is used (see Fig. 9a).

0)N_

Q-)

Q)O

Z0I

I

0cmF_

and error-free near-field

Fig. 12b. The center (main beam) portion of the phase difference between the simulated error-contaminated and error-free nearfields when the discrete error function in Fig. 7 is used (see Fig. 9b).

286

Q9

19II,

I

'I- <_-1-

Journal of ResearchVolume 97, Number 2, March-April 1992

of the National Institute of Standards and Technology

I

1r1' u 11I!f )

II'

�- -�-

Fig. 13a. The center (main beam) portion of the ratio of the error-contaminated and the error-free near-field amplitudes when thediscrete error function in Fig. 7 is used to construct the error-contaminated near field from measurements (see Fig. 10a).

0cm

O-0

ci)0

ci)

ci)UZ)0

cc

0

cOI

to0)

F-

Fig. 13b. The center (main beam) portion of the phase difference between the error-contaminated and the error-free near-fieldswhen the discrete error function in Fig. 7 is used to construct the error-contaminated near field from measurements (see Fig. 10b).

287

co

Q0

W iA 1~I

nA A . I I

I'

Or

A

I



00

1-1co

aci 0

0-Q

cc0�,

Fig. 14a. The center (main beam) portion of the ratio of the simulated error-corrected and error-free near-field amplitudes when

the continuous error function in Fig. 6 is used.

LCc

6

U)ci)ci)

ci)

i)0- O

U)

0

Q_

Fig. 14b. The center (main beam) portion of the phase difference between the simulated error-corrected and error-free near

fields when the continuous error function in Fig. 6 is used.

288

Ip



LF)

06

'--N

co

ci0

0-

0n %' W

Fig. isa. The center (main beam) portion of the ratio of the simulated error-corrected and error-free near-field amplitudes whenthe discrete error function in Fig. 7 is used.

O

I-

M

Q)ci)

ci)

ci)

½Q)0-Z0-

Fig. 15b. The center (main beam) portion of the phase difference between the simulated error-corrected and error-free near fieldswhen the discrete error function in Fig. 7 is used.

289

'I- �=- - -



-

ci)-0

O4-

O I

4 ''11

l I

Fig. 16a. The center (main beam) portion of the ratio of the error-corrected and the error-free near-field amplitudes when the discreteerror function in Fig. 7 is used to construct the error-contaminated near field from measurements.

U)ci)ma _)Og

_~ O\ -)

0 0(iI

QO

I

0Qc

Fig. 16b. The center (main beam) portion of the phase difference between the error-corrected and the error-free near-fieldswhen the discrete error function in Fig. 7 is used to construct the error-contaminated near field from measurements.

290

I I. 1.1 11I,,

II

II

. I

st

Y� f



0

- 1 0

- 20

-001_

-01

EC1

K

30 k

40 k

50 I.'

60 I

-70-1.2 -0.8 -0.4 -0.0 0.4 0.8 1.2

Kx/k

Fig. 17a. The amplitudes of the error-contaminated, error-corrected, and error-free far fields as func-tions of k, for the full range of kX derived from measured data. These fields are represented with solidcircles connected with dotted lines, open circles connected with solid lines, and a solid line, respectively.

-1.2 -0.8 -0.4 -0.0 0.4 0.8 1.2

Kx/k

Fig. 17b. The phases of the error-contaminated, error-corrected and error-free far fields as functionsof k 1 for the full range of k., derived from measured data. These fields are represented with solid circlesconnected with dotted lines, open circles connected with solid lines, and a solid line, respectively.

291

U)0)Q1

a)-01-c

Q-

200

1 60

1 20

80

40

0

-40

-80

- 1 20

- 1 60

-200



0

1 0

20 k

m

-)

-Q1

30 I

40 I

-50

-60 k

-70- 1.2 -0.8 -0.4 -0.0 0.4 0.8 1.2

Ky/ k

Fig. 17c. The amplitudes of the error-contaminated, error-corrected, and error-free far fields as func-tions of k, for the full range of k, derived from measured data (see Fig. 17a).

-1.2 -0.8 -0.4 -0.0

Ky/ k

0.4 0.8

Fig. 17d. The phases of the error-contaminated, error-corrected and error-free far fieldsof kl for the full range of kI derived from measured data (see Fig. 17b).

292

200 ,

1 60

120

0)01Q1

01

0)Q)

aD-

80

40

0

-40

-80

- 1 20

- 1 60

-2001.2

as functions



1O r

0

1 0

.01

_0

20 k

30 k

40 I

50 I

-0.3 -0.2 -0.1 -0.0

Kx/k0.1 0.2 0.3

Fig. 18a. The center (main beam) portion of the amplitudes of the error-contaminated, error-corrected,and error-free far fields as functions of k, derived from simulated data. These fields are represented withsolid circles connected with dotted lines, open circles connected with solid lines, and a solid line, respec-tively. The error-corrected and error-free lines overlap, hence, cannot be distinguished.

80 k

-0.3 -0.2 -0.1 -0.0

Kx/ k

0.1 0.2 0.3

Fig. 18b. The center (main beam) portion of the phases of the error-contaminated, error-corrected, anderror-free far fields as functions of k,, derived from simulated data. These fields are represented withsolid circles connected with dotted lines, open circles connected with solid lines, and a solid line, respec-tively. The error-corrected and error-free lines overlap, and, hence, cannot be distinguished.

293

60

200 ,

1 60

1 20

Cn0)0)

0)01

010)Cn

a-

40

0

-40

-80

- 1 20

- 1 60

-200



1 o

0

1 0k

m

0_

Q)D

a-

01

20 k

30 k

40 k

50 k

60 k

-0.3 -0.2 -0.1 -0.0

Ky/k0.1 0.2 0.3

Fig. 1&c. The center (main beam) portion of the amplitudes of the error-contaminated, error-corrected,and error-free far fields as functions of k, derived from simulated data (see Fig. 18a).

-0.3 -0.2 -0.1 -0.0

Ky/k0.1 0.2 0.3

Fig. 18d. The center (main beam) portion of the phases of the error-contaminated, error-corrected, anderror-free far fields as functions of kl derived from simulated data (see Fig. 18b).

294

-70

200 r

1 60

120

0)0101

01

0)

a-

80

40

0

-40

-80

-120

- 1 60

-200



0

-1 0

I4-'

_0

-20

30 I

-40 I

-50 k

-0.3 -0.2 -0.1 -0.0 0.1 0.2 0.3

Kx/kFig. 19a. The center (main beam) portion of the amplitudes of the error-contaminated, error-corrected,and error-free far fields as functions of k, derived from measured data. These fields are represented withsolid circles connected with dotted lines, open circles connected with solid lines, and a solid line, respec-tively. The error-corrected and error-free lines do not overlap, showing the presence of residual errors.

E

-0.3 -0.2 -0.1 -0.0

Kx/k0.1 0.2 0.3

Fig. 19b. The center (main beam) portion of the phases of the error-contaminated, error-corrected, anderror-free far fields as functions of k, derived from measured data. These fields are represented with solidcircles connected with dotted lines, open circles connected with solid lines, and a solid line, respectively.The error-corrected and error-free lines do not overlap, showing the presence of residual errors.

295

-60

200

1 60

1 20

u)0101U)

Q)

Q001-

0)01

a-

80

40

0

-40

-80

- 1 20

- 1 60

-200



0

10 k

20 k

m

0_

Q)-0

a,--

01

30 k

40 k

-50

-60 k

-0.3 -0.2 -0.1 -0.0Ky/ k

0.1 0.2 0.3

Fig. 19c. The center (main beam) portion of the amplitudes of the error-contaminated, error-corrected, and error-free far fields as functions of k, derived from measured data (see Fig. 19a).

200

1 60

1 20

0)01

cn01

Q)

Q)01

a-

80

40

0

-40

-80

- 1 20

- 1 60

E

-0.3 -0.2 -0.1 -0.0Ky/ k

0.1 0.2 0.3

Fig. 19d. The center (main beam) portion of the phases of the error-contaminated, error-corrected, anderror-free far fields as functions of kl derived from measured data (see Fig. 19b).

296

-70

-200



6.2 Residual Drift

In Fig. 2 we have shown the least amount ofresidual drift observed during a few days of moni-toring the system. In the initial stages of analysisfast tie-scans are used to analytically remove thedrift. But system drift during tie-scans will intro-duce both amplitude and phase errors that willshow up as part of the difference in the smoothnessin the discrete surfaces shown in Figs. 12b and 13b.We could remove such residual drifts by isolatingthe mechanism of interaction between the near-field system and the environment. It is thought, butnot proved, that temperature variations are themajor cause of such drifts. Hence, the temperaturesensitivity of each of the components of the scanerand its instrumentation should be studied, and themost sensitive components should be thermally iso-lated from the environment.

6.3 Residual Probe-Position Errors

In Fig. 1 we have represented residual probe po-sition errors schematically. In the analysis we haveassumed that probe position errors were only dueto data selection from the various scan planeszo.. .z5. Each scan plane was treated as perfectlyplanar. Since the interscan distance 3z is muchgreater than the residual errors in z, the error cor-rection was not effected to first order. We also ig-nored x and y position errors; accurate data onthese errors are not readily available. We can im-prove the final results of this study by initially cor-recting for known position errors on a single scanplane (residual errors). This was not done primar-ily because of practical limitations on time andfunds available for the project.

Finally, we note that the error-correction analy-sis in this study has been carried to fourth order, asexplicitly prescribed in Eq. (5). By including fifth -order terms in the analysis, we could improve con-vergence to the error-free values.

7. References[1] L. A. Muth and R. L. Lewis, IEEE Trans. Ant. Propagat.

38, 1925 (1990).[2] L. A. Muth, J. Res. Natd. Inst. Stand. Technol. 96, 391

(1991).[3] A. C. Newell, IEEE Trans. Ant. Propagat. 36, 754 (1988).[4] L. A. Muth and R. L. Lewis, Natl. Inst. Stand. Technol.,

Interagency Report, NISTIR 3970 (1991).[5] A. C Newell, NIST Planar Near-Field Measurements,

(1989) to be published.[61 L. A. Muth, A. C. Newell, R. L. Lewis, S. Canales, and

D. Kremer, 12th Annual Meeting and Symposium, AMTA,13-27 (1990).

[7] L. A. Muth, IEEE Trans. Ant. Propagat. 36, 581 (1988).

About the author: Lorant A. Muth is a physicist inthe Antenna Metrology Group of the ElectromagneticFields Division, which is part of NIST Electronics andElectrical Engineering Laboratory. The NationalInstitute of Standards and Technology is an agency ofthe Technology Administration, U.S. Department ofCommerce.

6.4 Conclusion

In this study we have shown that probe positionerrors can be removed from real near-field data bythe technique developed at NIST, and made sug-gestions to further improve the quality of near-fieldmeasurements at high frequencies by fine tuningboth the measurement system and the correctionprocedure.

297




Measuring the Electron's Charge and theFine-Structure Constant by Counting

Electrons on a Capacitor


E. R. Williams and The charge of the electron can be de- make an electrometer with sufficientRuby N. Ghosh, termined by simply placing a known sensitivity to measure the charge.

number of electrons on one electrodeNational Institute of Standards of a capacitor and measuring theand Technology, voltage, Vs, across the capacitor. If V. is Key words: calculable capacitor; cou-

measured in terms of the Josephson lomb blockade; electron charge; elec-Gaithersburg, MD 20899 volt and the capacitor is measured in tron counting; fine-structure constant;

SI units then the fine-structure constant single electron tunneling.and is the quantity determined. Recent de-

velopments involving single electronJohn M. Martinis tunneling, SET, have shown how to

count the electrons as well as how to Accepted: February 3, 1992National Institute of Standardsand Technology,Boulder, CO 80303

1. Introduction

The recent development of single-electrondevices [1,2,3,4,5] has made possible a new and veryprecise technique to measure the charge of an elec-tron. These devices are based on metal/insulator/metal tunnel junctions whose current-voltage (I-V)characteristics are determined by individual elec-tron tunneling events. If such a device is used toplace n electrons on a capacitor of capacitance C.and the resulting voltage V, is measured, the elec-tron charge e can be determined in terms of thisvoltage and capacitance, ne = VC. In practice, V, ismost precisely determined in terms of the voltagegenerated by a Josephson junction and thus the ex-periment will not be a determination of e in SIunits, but a measurement of e in terms of theJosephson volt [6]. In this paper we show that sucha measurement will soon be possible and that an ac-curate result would make a useful contribution tothe field of fundamental physical constants.

1.1 Relating the Fine-Structure Constant and theElectric Charge

First, we address how this measurement relatesto the fine-structure constant, a. By relating themeasured voltage to a Josephson voltage and thecapacitance to SI capacitance as measured in a cal-culable capacitor experiment [7], we can calculatethe fine-structure constant from the following sim-ple equations:

ne = hVs=C=jf 2 (1)

where fi is the frequency producing the Josephsonsteps and j is an integer associated with the Joseph-son effect. (Note that in a Josephson array j is thesum of all the integer steps of each junction.) Solv-ing Eq. (1) for e2Ih, a is given by

a= 2h 4n (2)

299



where go is the permeability of vacuum and C,must be measured in SI units. Specifically, the ca-pacitance must be related to the calculable capaci-tor experiment. In a calculable capacitorexperiment a change of capacitance of 0.5 pF canbe measured in terms of the meter with an accu-racy of 0.014 ppm [7] (1 ppm= 1 x 10-6; through-out, all uncertainties are one standard deviationestimates). This capacitance can then be relatedwith accuracies near the 0.01 ppm level to a 10 pFcapacitor that is stable and transportable.

If this experiment to measure e in terms of theJosephson volt and the calculable capacitor can berealized with high accuracy, it will provide a newpath to a. This new approach is similar to the mea-surement of a via the quantum Hall effect [8]. Al-though both experiments require a connection tothe calculable capacitor, this new method is muchmore direct. The 0.5 pF capacitance change used inmost calculable capacitor experiments is a goodmatch to the size of the capacitance that might beused to determine e; only one or two precision ra-tio transformer bridges need be involved in the cal-ibration of C,. By contrast, in the quantum Hallcase a chain of calibrations might involve these in-termediate standards: 0.5 pF, 10 pF, 100 pF,1000 pF, 100 k Q, 10 k f, 1 k f, and 6453.2 f. Themost accurate value of a determined from the cal-culable capacitor and quantum Hall effect has anuncertainty of 0.024 ppm [7,9] with 0.014 ppm com-ing from the calculable capacitor and the rest fromthe chain of intermediate standards. An alternatedetermination of a using the proton gyromagneticratio in H2 0 and the quantum Hall effect but notthe calculable capacitor, has an uncertainty of0.037 ppm [9]. However, these two results differfrom each other by (0.10 ± 0.043) ppm. A value of adetermined from the anomalous magnetic momentof the electron and quantum electrodynamic theory(QED) has an uncertainty of 0.007 ppm [10] and itsvalue lies between the two non-QED values.Therefore, the unexplained 0.10 ppm difference inthe non-QED values limits the accuracy to whichQED theory is tested. Any measurement of a bythis new SET method at the 0.1 ppm level or betterwould be helpful in the field of fundamental con-stants [8].

1.2 SET Devices

K. K. Likharev and his colleagues [1] have longbeen proponents of the application of Coulombblockade effects arising from the discreet charge ofelectrons to realize a precise current source. Their

pioneering ideas and experiments paved the wayfor much of the recent progress. In the past fewyears, new devices employing metal/insulator/metaltunnel junctions have been demonstrated whichcould allow a precise measurement of the electroncharge e and thus a. These include an electrometer[2], observation of single electron tunneling oscilla-tions [3], a "turnstile" current source [4], and a"pump" current source [5]. A brief sketch of sin-gle-electron tunneling is presented to describe theoperation of the electrometer and pump.

Consider a normal-metal tunnel junction biasedby a current source and having a capacitance C.The change in energy of the junction after anelectron tunnels through the barrier is AE =eQ/C-e 2/2C, where Q is the charge across thejunction and e2/2C is the Coulomb energy cost ofthe tunneling event. If the tunneling resistance ofthe junction is greater than the quantum unit ofresistance, R,>h/e2 , and thermal fluctuations donot mask the charging energy, kT<e2/2C, then aCoulomb blockade appears in the junction I-Vcharacteristic where the tunneling probability isgreatly reduced for IVI <e/2C. At 50 mK, and using1 fF for the junction capacitance, one finds (kT)/(e2 /2C)-1/20, hence temperature effects are smallbut non-negligible. The single-electron devices ex-ploit this Coulomb blockade of the tunneling cur-rent.

The Fulton-Dolan (SET) electrometer [2] isshown schematically in Fig. la. Aside from tunnel-ing events, the electrode a between the two junc-tions and the gate capacitor Co is an "island"electrically isolated from the circuit. The electrom-eter provides a very high impedance technique tomeasure the potential U. At a constant bias V, thedevice conductance as a function of gate voltage isperiodic with the period AU =e/Co. The biasvoltages V and U are chosen to maximize the sensi-tivity of the electrometer so that the device currentis linearly proportional to small changes of the po-tential U. Such electrometers have demonstrated

I I

_Co°C VU

Fig. la. This circuit shows an electrometer formed by two tun-nel junctions each having a capacitance C. The input is potentialU coupled to the isolated island a through CQ.

300



charge sensitivity of 1.5 x 10- e/Hz" 2 at a fre-quency between 2 and 200 Hz [11].

Convincing experimental evidence of the SEToscillations producing the controlled transfer ofelectrons with the relation I =ef was reported byDelsing et al. [3]. Geerligs et al. [4] were the first todemonstrate flat plateaus and observe a currentwith accuracies of better than 1% using the turn-stile device. However, certain properties of thepump device suggest that it will be more precise incounting electrons and we therefore sketch itsoperation.

An electron pump [5] (see Fig. lb) is a devicesimilar to the electrometer, in which a single elec-tron tunnels sequentially through the "islands" la-beled b and c. For a given voltage V, a controllednumber of electrons can be transferred in eitherdirection by appropriately cycling the gate voltagesU1 and U2 at the frequencyf. The pump also deliv-ers an electron current at the rate I = ef. The direc-tion of the current can be reversed simply byreversing the phase of U1 and U2. A pump based onfour or five tunnel junctions operating at 1-10MHz has been theoretically predicated to be ableto deliver a known current with errors below the 1ppm level [12,13].

I

calibrated in a range that will be useful in the fol-lowing experiment.

2. Experimental Configurations [14]

2.1 Measurements of the Electric Charge e

Figure 2a shows one possible configuration thatincorporates these new SET devices to measure theelectronic charge. A standard capacitor C. is con-nected to a current source (pump) and also to theinput of a Fulton-Dolan electrometer via a couplingcapacitor Cc, forming an isolated island a. Using theelectrometer as a null detector a voltage V, is ap-plied across C. to maintain the island a at a fixedpotential (near ground), while pumping electronsonto or off of the island. The stray capacitance toground, Cg, limits the sensitivity of the electrometerto V.. The voltage, Vs, on the capacitor is adjusted sothat the electrometer input is always at the samepotential (near ground). Thus the current from the

VP

--- . .. I

C C Cm m mI

V U1 U 2 V

2 2

US U2Time Time

Fig. lb. This circuit shows a three junction pump that pumpsone electron per cycle of the wave-forms shown. The electron isfirst pumped to island b by potential U, then to c by U 2 .

The currents produced by these pumps are onlyabout 10-12 A and thus precision measurements willbe a challenge. Direct measurement by sensing thevoltage across a high resistance is a problem be-cause it is difficult for metrologists to accuratelycalibrate the required high resistances. Capacitors,on the other hand, have already been accurately

Calibtioninput

VS

Fig. 2a. Proposed circuit for measuring a. The voltage V, re-quired to keep the potential of the island a constant when n elec-trons are pumped onto the island can be used to measure theelectron charge in terms of C, and V,. The electrometer monitorsthe potential at a.

301

T



pump serves as the charging current of the capaci-tor as the voltage is ramped up. A specific exampleis illustrated in Fig. 2b. With the pump operating at6.2 MHz for 10 s, the voltage across a 1 pF capaci-tor standard will charge to 10 V at a linear rate inorder to keep the electrometer input constant. Westop the pump and measure the 10 V signal as wellas any deviation of the electrometer from null.Reversing the pump for 20 s then brings thevoltage to -10 V.

s-' (1592 Hz) using a room temperature detectorwith a noise figure of 20 el Hz"X. A SET electrome-ter should have similar noise performance at thesame frequency and capacitance. Further improve-ments in the sensitivity of the SET electrometerare expected when measuring smaller capacitances.

C2-- --

a) 0_0U) IE

I U I C

I I

>Pa)

U,I mC I

a) I

Ca

a)

C.)

I UI Oa)

-A 20 s

Fig. 2b. The measurement sequence for measuring a, showingthe potential V, as a function of time.

In order to use a capacitor as a standard it musthave a well defined capacitance; it thus is necessaryto understand how precision capacitors are mea-sured and under what circumstances they are welldefined. Figure 3 shows the essential features ofone side of a standard bridge that is used to com-pare capacitors CQ and C2. The dashed lines repre-sent the grounded shield that prevents thepotentials V, and V2 from affecting the detector ex-cept through their respective capacitances. Thisshielding is critical to making precise capacitancecomparisons. The bridge is balanced by adjustingthe potentials V1 /V2 =CJC 1. C., and C.2 representthe stray capacitance to ground. The bridge bal-ance is not affected by Cg2 because there is novoltage across Cg1 when the bridge is balanced.However, Cg1 does affect the detector sensitivity soit should be kept small. The capacitance Cg2 doesnot affect the balance as long as the impedancefrom the source to C, is small compared to theimpedance of C52. Present day precision 10 pFcapacitance standards with Cgl=200 pF are com-pared at the 0.01 ppm level at a frequency ow = 104

I - - -

I I

I I

I -

I gI

I

L- _ - V

I -I V2_ - - - \

I V

_l Cg2I IC 1 Cg2

Fig. 3. Circuit showing half of a bridge used to compare capaci-tors C, and C2 .The potentials V, and V2 are usually supplied bya precision transformer. The effect of stray capacitances Cg, andC42 are discussed in the text.

The key to this experiment to measure e using acapacitor is to replace the detector in Fig. 3 with aFulton-Dolan electrometer as shown in Fig. 2a. Byplacing the electrometer at 20 mK near the sourceof charge, we anticipate fewer problems from leak-age and less stray capacitance to ground. In addi-tion, we expect to gain from the impressivesensitivity of this device, reduce the leakage dra-matically, and still make a well defined capacitorthat can be calibrated in situ.

302

I



Because of the requirement to shield thestandard capacitor, it is easier initially to have thiscapacitor on a different chip from the electrometerand pump. As multilayer technology becomesincorporated into the junction fabrication processthis capacitor could be placed on the same chip.Having it on a separate chip leads to some practicallimits in the optimum choice for Cs. The sensitivityof the electrometer is approximately given byCJ(Cs + C), where C5 represents all other capaci-tances to ground. This two chip arrangement willincrease Cg, but we expect to keep it near 1 pF.Choosing Cs to be small both increases V, and max-imizes the electrometer sensitivity. A 1 pF capaci-tance for C. constitutes a practical compromise.

Calibration of the capacitance requires making aconnection to room temperature. However, theFulton-Dolan electrometer could still be used bymeans of the calibration input switch shown inFig. 2. Note that care must be taken to keep thecapacitance from the island to ground low.

2.2 Dual Pump Bridge

Another interesting experiment (see Fig. 4) is tohave two current sources (drawn as two, four-junc-tion "pumps" in Fig. 4) feeding into an island awith a total capacitance of about 10 fF. The chargestate of island a is then measured with an electrom-eter. This geometry will not measure e but will al-low the very precise comparison of two currentsources. Any difference in current from the twosources will show up as an accumulation of the is-land charge. The island capacitance is chosen to belarge enough to greatly reduce the interaction ofthe two pumps but small enough to still have singlecharge resolution. Because the error is detected asan integrated charge on the island, only a shorttime is required to obtain extremely high precision.It would also be interesting to compare the errorrate of a "pump" and a "turnstile."

3. Future Prospects

A new precision technique to measure the fine-structure constant via Eqs. (1) and (2) has beendescribed. At present SET experiments have beenused to determine currents only at the 0.1 to 1.0%level, hence it is risky to predict the ultimate accu-racy of the technique. Based on theoretical predic-tions [12,13] we expect that the number ofelectrons could be measured with 1 ppm accuracyin a four junction pump. A five junction pump

a

| Pump 1

ElectrometerI.

C I

Ic

II Pump I

I I I

I Pump 2

I

IL,

I~~~ I~~~~~~~~~~~~~~~~~~~~~~~~

I V

-1

Fig. 4. A bridge circuit used to compare two pump circuits. Theelectrometer detects any accumulation of charge on island a.

should be more accurate still. At present the elec-trical standards needed to support this measure-ment do not appear to limit its accuracy. Forexample, capacitance standards of 10 pF are com-pared between national laboratories near the 0.02ppm level and 10 V Zener voltage references arealso compared with 0.02 ppm accuracy. In the firstexperiments these accuracies will likely not bereached, but much physics will be learned aboutSET devices as we make another accurate determi-nation of a.

Acknowledgment

We thank M. H. Devoret, D. Esteve, C. Urbina,H. Pothier, and P. Lafarge of Service de Physiquedu Solide et de Resonance Magnetique, Centred'Etudes Nucleaires des Saclay, for many informa-tive discussions about SET devices. We thank J. Q.Shields for crucial discussions about the techniquesused to measure standard capacitors. Discussionswith B. N. Taylor, Hans Jensen, and R. L. Kautzwere also very useful.

303



4. References[1] D. V. Averin and K. K Likharev, in Mesoscopic Phenom-

ena in Solids, B. Al'tshuler, P. A. Lee, and R. A. Webb,eds., Elsevier, Amsterdam (1991) Chapter 6.

[2] T. A. Fulton and G. J. Dolan, Observation of Single-Electron Charging Effects in Small Tunnel Junctions,Phys. Rev. Lett. 59, 109-112 (1987).

[3] P. Delsing, K. K Likharev, L. S. Kuzmin, and T. Claeson,Time-Correlated Single-Electron Tunneling in One-Dimensional Arrays of Ultrasmall Tunnel Junctions, Phys.Rev. Lett. 63, 1861-1864 (1989) and Phys. Rev. B 42,7439-7449 (1990).

[4] L. J. Geerligs, V. F. Anderegg, P. A. M. Holweg, J. E.Mooji, H. Pothier, D. Esteve, C. Urbina, and M. H.Devoret, Frequency-Locked Turnstile Device for SingleElectrons, Phys.Rev. Lett. 64, 2691-2694 (1990).

[5] H. Pothier, P. Lafarge, C. Urbina, D. Esteve, and M. H.Devoret, Single-Electron Pump Based on ChargingEffects, Europhys. Lett. 17, 249-254 (1992).

[6] Resolution 6 of the 18th Conference Generale des Poidset Mesures. See B. N. Taylor, ed., The International Sys-tem of Units (SI), Natl. Inst. Stand. Technol. Spec. Pub.330 (1991) p. 45.

[7] J. Q. Shields, R. F. Dziuba, H. P. Layer, New Realizationof the Ohm and Farad Using the NBS Calculable Capaci-tor, IEEE Trans. Instrum. Meas. 38, 249-251 (1989).

[8] B. N. Taylor and E. R. Cohen, How accurate are theJosephson and Quantum Hall Effects and QED?, Phys.Lett. A 153, 308 (1991).

[9] M. E. Cage, R. F. Dziuba, R. E. Elmquist, B. F. Field, G.R. Jones, Jr., P. T. Olsen, W. D. Phillips, J. Q. Shields, R.L. Steiner, B. N. Taylor, and E. R. Williams, NBSDetermination of the Fine-Structure Constant and theQuantized Hall Resistance and Josephson Frequency-to-Voltage Quotient in SI Units, IEEE Trans. Instrum. Meas.38, 284-289 (1989).

[10] T. Kinoshita and W. B. Lindquist, Eighth-order MagneticMoment of the Electron. V. Diagrams Containing No Vac-uum-Polarization Loop, Phys. Rev. D 42, 636-655 (1990).

[11] L. J. Geerligs, V. F. Anderegg, and J. E. Mooij, TunnelingTime and Offset Charging in Small Tunnel Junctions,Physica B165/166, 973-974 (1990).

[12] H. Pothier, 1991, Blocage de Coulomb et Transfertd'Electons Un Par Un, PhD Thesis, de l'Universite Paris6.

[13] Hans Jansen and J. Martinis, private communications.[14] The configurations shown here were first presented by the

authors as a poster paper at the NATO Advanced StudyInstitute on Single Charge Tunneling, Les Houches,France, March 1991. J. Martinis will present a paper onthis subject at the Conference on Precision Electromag-netic Measurements, June 1992, in Paris France.

About the authors: Edwin R. Williams is a physicistin the Electricity Division at NIST. Ruby N. Ghosh isan American Society of Engineering Education Post-doctoral Fellow in the Electricity Division at NIST.John M. Martinis is a physicist in the ElectromagneticTechnology Division at NIST. The National Instituteof Standards and Technology is an agency of theTechnologyAdministration, U.S. Department of Com-merce.

304



News Briefs

General DevelopmentsInquiries about News Briefs, where no contact personis identified, should be referred to the Managing Editor,Journal of Research, National Institute of Standardsand Technology, Administration Building, A635,Gaithersburg, MD 20899; telephone: 301/975-3572.

BIOASSAY REFERENCE VALUESESTABLISHED FOR THREE NIST SRMsShort-term bioassays are being used in many coun-tries to evaluate the exposure of humans to complexmixtures of mutagens and potential carcinogens inair, water, soil and from various emission sources.The salmonella mutagenicity assay developed byProfessor Bruce Ames and coworkers in 1925 is themutagenesis bioassay most frequently used in suchstudies.

In 1987, the International Programme on Chemi-cal Safety, in collaboration with the EnvironmentalProtection Agency and NIST, initiated an interna-tional collaborative study to determine the variabil-ity of data being generated by labs using the AmesBioassay. The study involved 20 laboratories fromNorth America, Europe, and Japan that used theAmes test on a regular basis. Three NIST SRMsthat are certified for chemical composition wereused in the study: 1597, Complex Mixture ofPolycyclic Aromatic Hydrocarbons from Coal Tar;1649, Urban Particulate Matter; and 1650, DieselExhaust Particulate Matter.

The results from this study revealed an interlab-oratory variability (approaching 100 percent coeffi-cient of variation) among the twenty study parti-cipants. Representatives from these expert labora-tories agreed that the variation found representedthe current state-of-the-art practice for theseassays. While a true value exists for chemical andphysical properties that can be measured with arelatively high degree of accuracy and precision,

bioassays are inherently more variable in partbecause the living organisms that serve as detectiondevices are constantly undergoing dynamic pro-cesses of growth, replication, metabolism, andinteraction.

The data from the collaborative study has beenused to establish two types of reference values.One, expressed as a 95 percent confidence interval,defines the actual mutagenic activity as measuredby the Ames bioassay test. The other, expressed asan 80 percent tolerance interval, characterizes thedifferences in reported mutagenic activity andestablishes a realistic target for comparing individ-ual laboratory results with the state-of-the-practice.With these Ames Mutagenicity reference values,SRMs 1597, 1649, and 1650 can now be used by theinternational bioassay measurement community aspowerful tools for reducing the variability of Amesbioassay data.

ANALYTICAL METHOD DEVELOPED TOINVESTIGATE THE ROLE OF VITAMIN CIN CANCER PREVENTION ANDJUVENILE ARTHRITISThe role of selected vitamins in the prevention ofcertain diseases, such as cancer, has been the sub-ject of great interest in recent years. Scientists atNIST and the National Cancer Institute recentlydeveloped a method for the accurate and precisemeasurement of vitamin C in human plasma.Procedures also were developed for stabilizingcalibration standards and for preparing stable andhomogeneous quality control materials (70 'C).The results of studies conducted to assess the long-term stability of vitamin C in preserved frozenplasma and the short-term stability in whole blood(4 'C) indicated that plasma samples containingvitamin C can be stored for at least 1 yr withoutmeasurable analyte degradation. In addition, vita-min C remains intact for at least 6 h in wholeblood. These studies also were used to assess the

305



performance of laboratories participating in NCIstudies designed to examine the role of vitamin Cin carcinogenesis.

This accurate method for the measurement ofvitamin C has proven to be useful in a collaborativestudy with the Children's Hospital National Medi-cal Center and George Washington University,where the vitamin C content of the serum ofpatients with juvenile arthritis was measured. Theresults indicate that the serum vitamin C levelscorrelate well with the pathological types ofarthritis exhibited by the patients.

Each of these studies, the multilaboratory studyon relationship of vitamin C to cancer incidenceand the grouping of pathological states as a func-tion of serum vitamin C levels, requires the use ofaccurate and precise trace organic measurementtechnology.

INTRAMOLECULAR VIBRATIONAL ENERGYREDISTRIBUTION STUDIED BYPICOSECOND IR SPECTROSCOPYMeasuring molecular vibrational energy redistribu-tion rates and extracting transfer mechanisms arekey steps toward understanding energy involve-ment in all chemical transformations. Obtainingthis dynamical information is especially difficultfor polyatomic molecules in high temperature con-densed media since such processes occur on thepicosecond and femtosecond timescales. NISTscientists have been able to examine these events inreal time by developing novel ultrashort pulsedinfrared techniques which monitor transient energypopulation in specific vibrational modes. Carefulselection of the molecular species and surroundingenvironment has enabled them to directly probefundamental transfer events after energy has beenselectively deposited in a vibrational mode. Thescientists have recently measured for the first timevibrational energy transfer rates for a meta-carbonyl-nitrosyl molecule, Co(CO) 3NO, in the gasphase and as dissolved in CFC13 solution. The ratesincreased by factors of two to five for the moleculein solution and the measurements indicate thatpopulated lower energy states readily relax bydepositing energy into solvent motions. Theseenergy transfer mechanisms are now being studiedby selecting buffer gases with specific acceptorvibrational energies (e.g., CO, NO, CFCl3) andmeasuring vibrational energy transfer processes asa function of pressure and density.

FREQUENCY AND WAVELENGTH STANDARDFOR THE FAR INFRAREDVisiting scientists at NIST recently completed ahigh-resolution study of the far-infrared spectrumof gaseous carbon monoxide (CO). The frequen-cies of pure rotational transitions in the groundstate of the molecule were measured using theirlaser-synthesized, tunable far infrared spectrome-ter. A complete set of frequencies was calculatedfrom these data with an accuracy of about 5 partsin 109. This is 10 times more accurate than theprevious best measurements.

Accurate measurements of the spectrum ofcarbon monoxide are important for two reasons.First, CO is an abundant interstellar molecule,and many astrophysical measurements and tests oftheory depend on the accuracy of CO rotationalfrequencies. Second, CO is often used as a refer-ence standard to calibrate spectrometers (espe-cially Fourier transform spectrometers) in the farinfrared. With the completion of this work, the COtransitions are now the most accurate far-infraredfrequencies available in the frequency range from100 to 4300 GHz (wavelength of 300 to 7 pum).

IN SITU CHARACTERIZATION OF THEPULSED LASER DEPOSITION OF THIN FILMSPulsed laser deposition (PLD) has been shown tobe an effective means of depositing thin films fromrefractory targets. Excimer and NdIYAG lasers arebeing used by NIST to deposit thin films of high Tcsuperconductors, high dielectric constant BaTiO3 ,ferroelectric PbZr.53Ti.470 3 (PZT), and magneticthin films of Fe3O 4 and Ag-Fe3O4 . During the PLDof these thin films, optical emission spectroscopy isused to determine the identity and energy (temper-ature) of the excited state species present in thelaser induced plumes. These results, combined withthe results of in situ molecular beam mass spectro-metric analysis of the plumes, permits determina-tion of both the minor and major species presentduring the PLD process, including nonexcitedneutral species and complex molecular species.Correlations are being made with the film'sstoichiometry, morphology and electric or magneticproperties.

NEW COMPUTATIONAL METHOD FORPATTERN FORMATION DURING ALLOYSOLIDIFICATIONA new model has been developed that predicts thegrowth and resulting segregation patterns of solid

306



particles forming from a melt during alloy solidifi-cation. These segregation patterns are responsiblefor defects that limit the mechanical and corrosionproperties of many alloys. This new approach waspreviously developed for pure materials but,through work at NIST has now been applied to themore useful case of binary alloys. The system istreated in its entirety with a continuous fieldvariable describing the state (liquid or solid) ofthe various regions of the system. Classical meth-ods previously used to calculate these growthmorphologies have required numerical tracking ofthe free boundary that separates the liquid fromthe solid and individual treatment of each phase.By contrast, the present approach greatly simplifiesthe calculations, making predictions for practicalalloys feasible.

FIBER LASER OFFERS WAVELENGTHSTANDARDNIST has successfully stabilized the wavelength ofan erbium-doped fiber laser to a narrow atomicabsorption line in the 1.5 pm region, offeringindustry a highly accurate optical wavelength stan-dard. The work is critical to the development ofadvanced optical communication systems utilizingcables smaller in size and cheaper to operate thanconventional wire lines. Narrow wavelength opera-tion permits many optical channels to be packedinto a single fiber. To create a system like this, anadequate laser wavelength standard is necessary.NIST's proposed standard uses both a diode laseroperating near 0.78 pm and a fiber laser near1.53 pm to drive transitions in rubidium. Addi-tionally, the fiber laser already has been usedsuccessfully in high-resolution spectroscopy, aidingresearchers in their investigations of atomic andmolecular behavior. For technical information,contact Sarah Gilbert, Div. 814.02, NIST, Boulder,CO 80303, 303/497-3120.

NIST RESEARCH LEADS TO SIGNIFICANTYIELD IMPROVEMENT IN THEMANUFACTURE OF MEMORY CHIPSResearch by a NIST scientist has directly con-tributed to a significant improvement in yield in themanufacture of integrated circuit memory chips inSIMOX material. SIMOX material is made byimplanting oxygen in silicon wafers to form asilicon dioxide layer beneath the silicon surface.Because of the smaller volume in which hole-

electron pairs can be generated, electronic circuitsfabricated in SIMOX material are much moreresistant to radiation effects than bulk silicon.

If there are particles on the wafer surface duringimplanting, conducting channels or pipes formbetween the silicon layer on the top and thesubstrate. Devices made on such a SIMOX waferdo not work properly. The scientist developed apotassium hydroxide etching procedure whichmade the pipes visible using optical microscopy.Further investigation showed that the number ofsurface particles could be reduced from 10/cm2 to0.1/cm2 by taking certain steps to reduce contami-nation within the implanter. Two private compa-nies adopted these procedures for producingSIMOX material. One company used its materialto manufacture SIMOX-based static random-access memories (SRAMs) and reports a signifi-cantly increased yield in 64 kilobit SRAMs. Thecompany also reports that use of the new SIMOXmaterial made manufacture of a 1 megabit SRAMpossible.

NIST LEADS INDUSTRY TO ACCORD IN X-RAYMASK SPECIFICATIONSA NIST scientist in his capacity as chairman of theX-Ray Program Mask Standards Committee estab-lished by the Defense Advanced Research ProjectsAgency, has led industry to adopt a standard limit-ing set of dimensions (the envelope) of the masksused for manufacturing integrated circuits definedthrough x-ray lithography. This accomplishmentsatisfies the charge given to the committee on itsformation. An x-ray mask consists of a thin siliconmembrane supported by a heavier frame; x-raylithography extends the capabilities of opticallithography to finer dimensions, permitting theproduction of smaller integrated-circuit elements.The agreement constitutes an important steptoward bringing x-ray lithography from the labora-tory to the production line.

The committee also concluded that there isinsufficient motivation for developing commonalitybetween masks used with point (lower flux) andcollimated (higher flux) sources, that there is notechnical advantage anticipated from the conceptof employing differing standards for primarymaster patterns and working x-ray mask copies,and that it would be neither necessary nor advanta-geous to impose any further dimensional con-straints on x-ray masks until the technologybecomes further advanced.

307



NEUTRON REFLECTOMETRY HELPSCHARACTERIZE POLYMER BRUSHESPolymer brushes are central to many importantproblems in polymer science and biophysics, in-cluding colloid stabilization, polymeric surfactants,and polymer compatibalizers. Understanding thebrush structure of the polymer chains is importantin predicting the behavior of each of these systems.

Recently NIST scientists, in collaboration withresearchers from two universities, have made asignificant advance in determining the polymersegment density profile of these brushes usingneutron reflectivity. The polymers used in thisstudy are so-called "sticky foot" polystyrene inwhich one end of the polystyrene molecule has afunctional group which adsorbs strongly onto thesurface from the solution. This gives a very largedensity of adsorption of polymer near the surface,thus making neutron reflectivity measurementspossible. NIST scientists use a technique unique toneutrons wherein the neutron beam traverses alarge single crystal of silicon and reflects from thesilicon-polymer solution interface, with the re-flected beam again traversing the silicon crystal.For the first time, the neutron data show verystrong minima and maxima in reflectivity, therebysignificantly increasing the precision with which thestructure of the polymer brushes adsorbed on thesilicon-solution interface can be determined.

FAST RESTORATION OF IMAGESTwo NIST scientists have demonstrated the speedand flexibility which modern parallel computingcan provide in restoring blurred images from anoisy background. Their restoration procedure, ona massively parallel machine (the ConnectionMachine at the University of Maryland Institutefor Advanced Computer Studies), takes advantageof what the operator may know in advance and canrecognize in the image as it is recovered.

Starting from a 512 x 512 image sample, thescientist first artificially blurred it, by convolutionwith a Gaussian-like point spread kernel, and thenadded random noise. This produced a blurredimage, which resembled typical raw image data ona medical diagnosis system.

The recovery was undertaken by the methodof "Tikhonov regularization"-using a low-passspatial filter with an interactively set regularizationparameter. In the presence of noise, the recoveryof the image is a mathematically ill-posed problem.The computer by itself, not knowing the features ofthe image that are most important, may wanderfrom the desired solution. The operator can

sharpen important details in the recoveredimage by resetting the regularization parameterinteractively.

In the recovered image, most of the significantdetail from the original is present, plus a back-ground pattern which originated with the noisecomponent of the blurred data. These extraneousdetails can be easily ignored because the operatorcan distinguish them visually.

Massively parallel processing is fast, in this casemaking it possible to run the restoration processwith many different settings of the regularizationparameter.

NEUTRON FOCUSING USING CAPILLARYOPTICSNIST researchers have been collaborating withresearchers from the Kurchatov Institute forAtomic Energy Institute in Moscow and from theState University of New York at Albany on theuse of multicapillary fibers for the focusing ofneutrons. Recently, the researchers at NISTdemonstrated the focusing of cold neutrons fromthe Cold Neutron Research Facility using a simplelens comprising a dozen of these capillary fibers.These neutron measurements were performed inconjunction with a newly developed charge transferdevice neutron detector having a few tens ofmicrometers spatial resolution. A cooperativeresearch and development agreement to developmulticapillary optical fibers for the transmissionand focusing of cold neutrons has been signed witha private company.

These multicapillary fibers, which were devel-oped at the Institute for Roentgen Optical Systemsin Moscow, consist of thousands of parallelchannels within a glass fiber of sub-millimeterdiameter. Neutrons are guided down numerousopen channels that are only 6 prm in diameter, aconcept that has proved successful with x rays.Low-energy neutrons have optical propertiesanalogous to those of photons and may be focusedto produce significantly higher fluxes using theprinciple of mirror reflection from smooth surfacesat small grazing incident angles. The lens is formedby bending and orienting the fibers in such a waythat the beams that emerge from each fiber pointto the same spot, and so increase the neutroncurrent density. Narrowly focused beams allowgreater spatial resolution for neutron absorptionmeasurements than previously available andsuggest microprobe capabilities to complementother radiation techniques for multidimensionalanalysis in materials science.

308



NIST DEVELOPS METHODS FOR HIGH-PRECISION PHASE NOISE MEASUREMENTSIn response to new requirements of the U.S.aerospace industry for tighter phase noise specifi-cations, NIST has developed the first measurementsystems that yield high-precision phase noisemeasurements over a broad frequency range. Phasenoise has been rising in importance in radars,telecommunications, and other systems where highspectral purity of signals is critical. Its presence insignal channels degrades the spectral purity andthus the performance of the system. The NISTphase noise measurement systems operate atcarrier frequencies from 5 MHz to 26 GHz and 33to 50 GHz with bandwidths up to 10 percent of thecarrier frequency. The measurement precisionranges from about 1 to 2 dB depending on theoperating conditions.

Commercial equipment for phase noise measure-ments have been available for several years, butthere has been no satisfactory method for assuringthat such systems were indeed giving accuratereadings. The commercial systems are substantiallymore limited in dynamic range and precision. TheNIST measurement systems have uncovered anumber of measurement errors, some ranging from5 to 20 dB.

HUBBLE DATA SUGGEST NO END FORUNIVERSEA team of scientists using NASA's Hubble SpaceTelescope (HST) has made the most precise mea-surement to date of the percent of deuterium(heavy hydrogen) in space. Led by a NISTastronomer and fellow of the Joint Institute forLaboratory Astrophysics in Boulder, CO, the re-searchers used HST's Goddard High ResolutionSpectrograph to determine that the ratio of deu-terium to hydrogen in space is 15 parts per million(15 x 10-6). If the assumptions of the Big Bangtheory for the universe's creation are correct, thisratio accounts for the maximum possible density ofordinary matter that can be present in the universe.The Hubble data also suggest that unless largeamounts of some exotic form of matter exist inspace, there is not enough matter present to haltthe expansion of the universe. If true, the universehad a brilliant beginning but will have no end.

NEW "MEASURES" NEEDED TO BOOSTPRODUCTIVITYAs American electronics firms strive to producetop-notch products, one factor emerges as a key toboosting domestic semiconductor productivity:

measurement technology. This is the theme of anew NIST report, Metrology for the Semiconduc-tor Industry. The 46-page document explains theimportance of measurement in integrated circuitfabrication and how it enables manufacturers tobetter understand and control processes. Thereport says reliable measurement factors, such ascircuit component dimensions and properties atvarious steps during the production process, arecrucial. This ensures that processing errors areavoided and that faulty products are eliminatedearly in the production process. Besides presentingits measurement message, the report also details,in lay language, the semiconductor manufacturingprocess. For a free copy, contact Jane Walters,B344 Technology Building, NIST, Gaithersburg,MD 20899, 301/975-2050.

MEASUREMENT SERVICE FORCW WATTMETERS COMINGNIST expects to establish a measurement servicefor high power continuous wave (cw) wattmeters inthe near future. Measurements will be available atseveral points from 2 to 30 MHz (1 to 1,000 W) and30 to 400 MHz (1 to 500 W). Advance notice isbeing provided because use of this service willrequire prearrangement with NIST. Wattmetersmust be controllable via an IEEE-488 bus, have atype N male input connector, and either have atype N female output connector or be suppliedwith a load. At each measurement point, thecustomer will receive a calibration factor for thewattmeter, defined as the ratio of the wattmeterreading to the power incident on it. The measure-ment uncertainty for the new service is expected tobe less than plus or minus 2 percent, depending onthe frequency, power level, and electrical charac-teristics of the wattmeter/load combination. Formore information, contact Gregorio Rebuldela orJeffrey Jargon, Div. 813.01, NIST, Boulder, CO80303, 303/497-3561.

NEW GLOSSARY OF COMPUTER SECURITYTERMINOLOGYComputer security practitioners will be interestedin The Glossary of Computer Security Terminology(NISTIR 4659), a collection of terms and defini-tions used by federal departments and agencies intheir policies and standards. The 176-page docu-ment was developed under the auspices of theNational Security Telecommunications and Infor-mation Systems Security Committee and publishedby NIST as part of its effort to distribute federallysponsored work. This glossary provides more than

309



one definition for many computer and communica-tions security terms, reflecting the various uses bydifferent federal agencies. Technical informationon the publication is available from EdwardRoback at 301/975-3696. The publication is for saleby the National Technical Information Service,Springfield, VA 22161 for $26 (hard copy) and$12.50 (microfiche) prepaid. Order by PB 92-112259/AS.

NEW PROCEEDINGS ISSUED ON HIGHINTEGRITY SOFTWAREThe Proceedings of the Workshop on HighIntegrity Software (NIST Special Publication500-190) presents the results of a recent Assuranceof High Integrity Software workshop sponsored byNIST. Participants addressed techniques, costs andbenefits of assurance, controlled and encouragedpractices, and hazard analysis. The workshopprepared a preliminary set of recommendationsand proposed future directions for NIST to coordi-nate an effort to produce a comprehensive set ofstandards and guidelines for the assurance of highintegrity software. Technical information on theworkshop is available from Dolores R. Wallace at301/975-3340. The publication is for sale by theNational Technical Information Service, Spring-field, VA 22161 for $17 (hard copy) and $9(microfiche) prepaid. Order by PB 92-109040/AS.

AUTOMATED MANUFACTURING STARS INNEW VHS PROGRAMA new videotape describing information manage-ment strategies developed for NIST's AutomatedManufacturing Research Facility (AMRF) wasrecently released. The AMRF is a prototype small-batch manufacturing facility used for integrationand measurement-related standards research."Information Management in the AMRF, Novem-ber 1991," a 12 1/2 min tape describes five majormanufacturing functions: design, process planning,off-line programming, shop floor control, and ma-terials processing. Special emphasis is placed onhow the AMRF manages information flow betweenthese functions. The video is available on free loanby requesting title number 24798 from ModernTalking Picture Service Inc., 5000 Park St. North,St. Petersburg, FL 33709, 800/237-4599. VHScopies are $12 (shipping included) and may bepurchased from Video Transfer Inc., 5709-BArundel Ave., Rockville, MD 20852, 301/881-0270.

NIST REPORTS ON ISO 9000 STANDARDSERIESAnswers to many commonly asked questions aboutquality, quality systems, and the use and applica-tion of such systems are found in Questions andAnswers on Quality, The ISO 9000 StandardSeries, Quality System Registration, and RelatedIssues (NISTIR 4721). The report briefly describesthe contents of the ISO 9000 standards that werepublished in 1987 by the International Organiza-tion for Standardization (ISO) in Geneva, Switzer-land. The five international standards, along withthe terminology and definitions contained inISO Standard 8402, Quality-Vocabulary, provideguidance for industry decision makers on the selec-tion of an appropriate quality managementprogram or system. Among other topics, the reportdescribes the conformity assessment scheme of theEuropean Community (EC) for EC-regulatedproducts. To obtain a copy of NISTIR 4721, send aself-addressed mailing label to the StandardsCode and Information Program, A629 Administra-tion Building, NIST, Gaithersburg, MD 20899,301/975-4031.

SOFTWARE PACKAGE FOR ANTENNAMETROLOGY AVAILABLEA demanding task in antenna metrology is develop-ing computer programs to process the massiveamounts of measurement data needed to deter-mine antenna performance characteristics. NISTscientists are currently creating a comprehensiveand easy-to-use personal computer softwarepackage for conducting advanced research relatedto near-field antenna metrology. Ultimately, it willsimplify methods for precisely determining thefar-field properties of state-of-the-art antennasbased on laboratory measurements. The softwarewill isolate common computational themes anddevelop dedicated modules for use in performingcomputational tasks (involving real or simulateddata for research and error analysis in antennametrology). A recent NIST publication describes asoftware system that achieves a high degree ofpower, modularity, and flexibility for studyingplanar near-field measurements. The packageconsists of many FORTRAN modules that can becalled from DOS to achieve a variety of computa-tional tasks. For a copy of Personal ComputerCodes for Analysis of Planar Near Fields (NISTIR3970), contact the National Technical InformationService, Springfield, VA 22161. Order by PB 92-112283; the price is $19 prepaid.

310



NEW GUIDE DESIGNED TO HELP FEDERALGOSIP USERSFederal managers, procurement personnel, andtechnical specialists seeking to buy or evaluateGovernment Open Systems Interconnection Profile(GOSIP) products should find a new users guidehelpful. The guide is a valuable reference designedto help federal users interpret GOSIP technicalinformation. It also can aid users in setting upGOSIP-compliant products in the workplace.Persons with little or no experience in OpenSystems Interconnection (OSI)-which allowsunrelated computer systems to communicate in anopen network environment-should be able tounderstand all portions of the guide. The guidealso contains worthwhile information for theexperienced OSI user. Government Open SystemsInterconnection Profile Users' Guide, Version 2 isavailable for $26 prepaid from the NationalTechnical Information Service, Springfield, VA22161. Cite order number PB 92-119676/AS. (Note:This users guide serves as a companion documentto Version 2 of GOSIP, issued as Federal Informa-tion Processing Standard (FIPS) 146-1. It is bestused in conjunction with GOSIP and/or OSIdocuments.)

BOOK REVIEWS SUPERCRITICAL FLUIDEXTRACTIONTwo NIST researchers have authored and edited adefinitive book that reviews the theory and applica-tions of supercritical fluid technology. Over thepast 5 years, there has been renewed interest insupercritical fluid extraction (SFE) methods. SFEhas many industrial applications, but the systemsmust be carefully chosen. The best candidates forextraction are high-cost, low-volume commoditiesthat require the use of a non-toxic solvent in theirprocessing. The researchers have collected a seriesof reviews covering theoretical and experimentalwork geared toward the more exact requirementsof current SFE applications. Supercritical FluidTechnology: Reviews in Modern Theory andApplications contains 16 chapters written by someof the foremost experts in this technology. Formore information, contact Tom Bruno at Div. 838,NIST, Boulder, CO 80303, 303/497-5158.

VIDEO HIGHLIGHTS 1991 BALDRIGE AWARDWINNERSWhat is quality management? Why is it importantto U.S. competitiveness? How is it achieved?In a new 20 min video, representatives from thethree 1991 winners of the Malcolm Baldrige

National Quality Award answer these questionsand others on quality and quality managementstrategies. The 1991 winners are Solectron Corp.,San Jose, CA; Zytec Corp., Eden Prairie, MN; andMarlow Industries, Dallas, TX. The award, namedfor the late secretary of commerce, was establishedby federal legislation in August 1987. It promotesnational awareness about the importance ofimproving quality management and recognizesquality achievements of U.S. companies. However,the award is not for specific products or services.The award program is managed by NIST, with theactive involvement of the private sector, includingthe American Society for Quality Control (ASQC).The 1991 winners video is available from ASQC for$15. Item TA996 in ASQC's inventory, it can beordered by calling 1-800/952-6587.

RADIO WAVES "LOWER THE BOOM" ONBURIED MINESNIST has completed a 6 year effort to help theU.S. Army detect buried land mines (either plasticor metal) using electromagnetic (EM) technology.The method involves sending radio waves into thesoil and analyzing the reflections received. A newreport, Qualifying Standard Performance of Elec-tromagnetic-Based Mine Detectors (NISTIR3982), discusses the best ways of conducting con-trolled tests of portable detectors under varyingsoil conditions and recommends means for estab-lishing a stable test range. Topics covered include:theoretical relationships between EM fields andmatter; critical EM performance factors forportable mine detectors; measuring EM propertiesof materials relevant to EM mine detection; EMproperties of soils and recommendations for simu-lated soil standards; desired properties and recom-mendations for simulated mine standards (bothplastic and metal); recommended mine detectortesting strategy; maintenance of test records and atest range; and recommendations for future work.Copies of NISTIR 3982 are available from theNational Technical Information Service, Spring-field, VA 22161. Order by PB 92-116292; the priceis $26 ($9 microfiche).

PHOTONIC PROBES DEVELOPED TOMEASURE EM FIELDSPhotonic probes using lasers and fiber optics offermany advantages in electromagnetic field measure-ments, especially when measuring electromagneticpulses (EMP) and high-power microwaves. Theseprobes provide the wide bandwidth and low disper-sion needed to measure EMP signals. They are free

311



from EM interference and provide minimal distur-bance of the field being measured. NIST's interesthas been in passive probes, where the probe headcontains no active electronics or power supplies. Arecent paper discusses the characteristics ofphotonic probes, presents an overview of systemdesign (with emphasis on the transfer functions ofappropriate antennas and electro-optic modula-tors), and details three photonic systems NIST hasdeveloped. For a copy of paper 61-91, contact JoEmery, Div. 104, NIST, Boulder, CO 80303,303/497-3237, fax: 303/497-5222.

SHAPING NEW DESIGNS FOR WAVE PROBESThere is a growing need for basic measurementsupport at millimeter wave and upper microwavefrequencies. One requirement is for transferstandard probes -electric or magnetic field probesused for comparing standard electromagnetic fieldsgenerated in calibration laboratories. Anotherrequirement is for radiation hazard meters. NISTscientists are using an optical sensor to measurethe temperature rise of a resistive element in anelectromagnetic field. In a recent paper, theyreport on the analyses and evaluations of numer-ous configurations for the resistive element. Asphere was chosen as the most promising shape.Several probe tips of this design were made andtested for sensitivity, frequency response, andlinearity. Test results were very encouraging, withone design providing adequate performance for atransfer standard for millimeter waves. Report54-91 is available from Jo Emery, Div. 104, NIST,Boulder, CO 80303, 303/497-3237.

IC DEVICE OFFERS IMPROVED SPEED,ACCURACYNIST researchers have created a novel integratedcircuit "comparator" that samples electrical signalsfaster and with less error than existing devices.Potential uses for the circuit include measuringperformance of video display integrated circuits,where output voltage must change and settlerapidly to assure a crisp image. The NIST deviceeliminates thermal effects that can cause errorsand affect voltage settling time. More than twice asfast as most other available comparators, it makesdecisions with 0.1 percent accuracy within 2 nsfollowing an abrupt input change. Unlike othercomparators, the NIST device is intended for useonly in a "strobed" or sampling mode (in which anunknown signal and a reference signal are

compared on command). The circuit's inventorshave applied to patent the device. For more

information, contactMetrology Building,20899, 301/975-2406.

Michael Souders, B162NIST, Gaithersburg, MD

ALASKA MARINE MAMMAL POLLUTIONDATA AVAILABLETissue samples from seals and whales in Alaskacontain varying levels of pesticides, PCBs and traceelements, according to a new report by scientistsfrom NIST and the National Oceanic and Atmo-spheric Administration (NOAA). The documentpresents baseline data from analyses of selectedtissue samples in the Alaska Marine MammalTissue Archival Project (AMMTAP). The publicationalso describes the current inventory of tissuesamples in the AMMTAP and methods used toanalyze the samples. Since 1987, NIST and NOAAscientists have collected marine mammal tissuesamples on subsistence hunts with Native Alaskans.The samples are stored at NIST's NationalBiomonitoring Specimen Bank in Gaithersburg,MD. The publication, AMMTAP: Sample Inven-tory and Results of Selected Samples for Organicand Trace Elements (NISTIR 4731), is availablefrom the National Technical Information Service,Springfield, VA 22161 for $26 prepaid. Order byPB 92-143718.

SUPERCONDUCTOR ABSTRACTS COMPILEDIN SINGLE VOLUMEThe abstracts of 243 papers published by NISTauthors in the field of high critical temperaturesuperconductors have been collected and reprintedin High-Temperature Superconductivity: Abstractsof NIST Publications, 1987-1991 (NIST SP 826).These papers-which appear in proceedings ofmeetings, scientific and technical journals, and asNIST publications-represent work from nine ofthe institute's divisions. Many are the result ofcollaborations with researchers in industry, acade-mia, and other government agencies. Abstracts aregrouped in the following categories: compositionalanalysis; critical current and critical temperature;crystal structure; elastic constants and phononspectra; electrical contacts; electronic structure;energy gap and tunneling spectra; Josephson effectand devices; magnetic measurements; phaseequilibrium; processing-bulk material; andprocessing -thin films. SP 826 is available from theNational Technical Information Service, Spring-field, VA 22161, for $19 (paper) or $9 (microfiche)prepaid. Order by PB 92-126564/AS.

312



REPORT DETAILS RISKS, HANDLING CAREFOR 3480 TAPESince the mid-1980s, data centers have used 3480-type magnetic tape to archive information. Someusers have saved critical data on literally thousandsof 3480 cartridges. But what kind of shelf life canusers expect from these magnetic media? Are 3480tapes a dependable way to save crucial material?NIST researchers recently examined these issuesand published their results in a new report, The3480 Type Tape Cartridge: Potential Data StorageRisks, and Care and Handling Procedures to Mini-mize Risks (NIST SP 500-199). Under sponsorshipof the National Oceanic and Atmospheric Admin-istration, NIST appraised potential risks in using3480 materials, studying factors such as chemical,mechanical, and magnetic failure mechanisms. Thereport summarizes reasonable procedures for thecare and handling of the tapes to minimize risk.Also included are recommendations from applica-ble scientific literature and opinions from othertape experts. Available for $3.50 prepaid from theSuperintendent of Documents, U.S. GovernmentPrinting Office, Washington, DC 20402-9325. Citestock number SN 003-003-03127-5.

NIST ISSUES "HOW-TO" MANUAL ONSMOKE CONTROLSmoke is the killer in most fires. Flowing silentlythrough corridors, elevator shafts, and stairwells, itcan be deadly even to those far removed from thefire. For example, most of the 85 people who diedin the 1980 fire in the Las Vegas MGM GrandHotel were far above the first floor where the firestarted and was contained. Researchers at NISThave worked with private industry for several yearsto find ways to prevent such tragedies. The result isa new "how-to" manual for designing systems tocontrol or contain smoke. The systems described inthe manual use pressurization produced bymechanical fans. Such techniques have been usedfor at least 50 years to control the flow of airborneproducts such as dangerous gases or bacteria. How-ever, only recently have they been considered forcontrolling smoke. The project was funded by theAmerican Society of Heating, Refrigerating, andAir-Conditioning Engineers. Technical assistancecame from the National Research Council ofCanada and numerous industry experts. TheDesign Manual for Smoke Control Systems(NISTIR 4551) is available from the NationalTechnical Information Service, Springfield, VA22161 for $43 prepaid. Order by PB 92-144104.

BETTER SOLUTIONS FOR WAVE EQUATIONSOFFEREDNIST has published a tutorial monograph thatfeatures an improved method for deriving solutionsto one of the basic equations of mathematicalphysics, the wave equation. Modified AiryFunction and WKB Solutions to the Wave Equa-tion (Monograph 176), defines the Modified AiryFunction (MAF) method. The MAF method ismore useful than the WKB method developed inthe 1920s and long used for approximate solutionsto problems in quantum mechanics and opticalwaveguides with a non-uniform refractive index.The monograph discusses and compares the WKBand MAF methods, drawing examples from theoptical waveguide community. It also deals witheigenvalue and initial value problems. Single copiesof Monograph 176 are available at no charge fromR.L. Gallawa, Div. 814.02, NIST, Boulder, CO80303, 303/497-3761. Multiple copies may bepurchased from the Superintendent of Documents,U.S. Government Printing Office, Washington DC20402-9325, for $5.50 each. Order by stockno. 003-003-03123-2.

NEW MICROSCOPE OFFERSUNPRECEDENTED RANGEImagine a camera that can image individual flowersin a Philadelphia tulip bed and then expand tosuccessively larger frames until the 170 km expansebetween Philadelphia and Baltimore is in view. Byanalogy, a novel scanning tunneling microscope(STM) designed by NIST researchers accommo-dates a comparable range for material surfaces.Sixty times better than the span of ranges achievedwith standard STMs, the 600 000 times differencein the field permits greater viewing flexibility. Onecan attain a resolution of 1 nm-fine enough toscrutinize a cluster of molecules-or enlarge thefield to survey a square area measuring 600 pm ona side. Researchers at NIST report the magnifica-tion range and zooming capabilities of theirlong-range STM rival that of a scanning electronmicroscope (SEM). Also, high-magnificationviewing does not require placing samples in avacuum chamber like an SEM. For moreinformation, contact Theodore Vorburger, A117Metrology Building, NIST, Gaithersburg, MD20899, 301/975-3493.

313



NEW TEST IMPROVES UNDERSTANDING OFSMOKE DANGERNIST researchers have developed the first compre-hensive method for determining the toxic potencyof smoke from burning materials. The techniqueprovides a more realistic way of analyzing thepotential hazard to humans of a burning product,such as a chair or wall covering. As manufacturers,code officials, and others gain experience and con-fidence with this method, say NIST researchers,the toxic potency of some products, such as thosewith similar chemical makeups, could be assessedwithout testing. This could speed the developmentof new products without adding risk to the con-sumer and will reduce the need for animal testing.The test includes a unique small-scale instrumentfor gathering accurate smoke toxicity data in con-junction with a method of predicting smoke toxicitythat is faster and less costly than current toxicitytests. NIST is working with standards groups tohave the NIST method accepted both in theUnited States and abroad as a standard smoke tox-icity measurement. A report, Toxic Potency Mea-surement for Fire Hazard Analysis (NIST SpecialPublication 827), is available from the NationalTechnical Information Service, Springfield, VA22161 for $26 prepaid. Order by PB 92-137546.

NEW DETECTOR FOR GASCHROMATOGRAPHY PATENTEDA NIST researcher has been granted a patent(U.S. Patent No. 5,070,024) for a hydrocarbondetector designed for use during gas chromato-graphy in harsh industrial process environments.The research was sponsored by, and the patentassigned to, the Gas Research Institute, Chicago.The device indicates the presence and quantityof hydrocarbons in the exit stream of a gaschromatography column by sensing catalytic crack-ing of the hydrocarbons. It is much simpler inconstruction and operation than commonly avail-able commercial detectors. By recording thetemperature difference between gas flowing past acatalyst-coated thermopile (a linked group ofthermocouples) and a similar uncoated thermopile,the device can determine the proportions of chemi-cal components, identify unknowns, and distinguishbetween straight chain and branched hydrocar-bons. The new device uses simple voltage-measuring instrumentation and is estimated to costabout one-tenth as much to fabricate as existingdetectors. The detection cavity is very small, sominute amounts of hydrocarbons can be detected.For licensing information, contact Jeffrey Savidge,

Gas Research Institute, 8600 W. Bryn Mawr Ave.,Chicago, IL 60631, 312/399-8100. For technicalinformation, contact Thomas Bruno, Div. 838.02,NIST, Boulder, CO 80303-3328, 303/497-5158.

GUIDE MAKES SRMs USEFUL TOOLS INCHEMISTRY LABSNIST chemists describe how analytical laboratoriescan assess the accuracy of their chemical analyses ina new publication describing the use of StandardReference Materials (SRMs). The publication, Useof NIST Standard Reference Materials forDecisions on Performance of Analytical ChemicalMethods and Laboratories (NIST Special Publica-tion 829), guides laboratory managers on the mosteffective ways to use SRMs. SRMs serve as qualitycontrol standard samples for laboratories analyzingorganic and inorganic compounds in a wide varietyof materials. The NIST publication gives qualitycontrol managers specific guidelines for usingSRMs to design measurement methods and inter-pret the accuracy of data. Copies are available fromthe Chemical Science and Technology Laboratory,A317 Chemistry Building, NIST, Gaithersburg, MD20899, 301/975-3146.

CARB SOLVES SECOND SUGAR TRANSPORTPROTEINResearchers at the Center for Advanced Researchin Biotechnology (CARB) have defined thethree-dimensional structure of a second bacterialsugar-transport protein. The protein, known ashistidine-containing phosphocarrier protein HPr, isessential for transporting energy from the environ-ment into bacterial cells. The same CARB teamsolved the structure of a related bacterial sugar-transport protein last fall. Both proteins areinvolved in the same chain reaction that bringsglucose molecules into bacterial cells. With proteinstructures such as these in hand, pharmaceuticalcompanies may soon be designing drugs that pre-vent sugar transport and, in turn, starve infectiousbacteria by cutting off their energy supply.Researchers from NIST, the University of Maryland,and the University of California at San Diegocollaborated on the protein studies at CARB. Apaper describing the new structure appears in theMarch 15, 1992, issue of Proceedings of theNational Academy of Sciences. CARB was estab-lished in 1984 by NIST, the University of Maryland,and Montgomery County, MD, as a unique centerfor government, academic, and industry scientists.

314



PAPERS REQUESTED FOR FALL OPTICSSYMPOSIUMNIST, in cooperation with the Institute of Electri-cal and Electronics Engineers (IEEE) Lasers andElectro-Optics Society and the Optical Society ofAmerica, will sponsor the 7th Biennial Symposiumon Optical Fiber Measurements on Sept. 15-17,1992, in Boulder, CO. Symposium sessions willcover measurements for things such as telecommu-nications fibers, fiber lasers and amplifiers, fibersfor sensors, couplers, connectors, multiplexers,integrated optics, sources, detectors, modulators,switches, long-haul systems, LANs, subscriberloops, field and laboratory instrumentation, andstandards. Experimental and analytical paperscurrently are being solicited on all aspects of themeasurement of guided-lightwave technology.Summaries of proposed papers must be submittedby May 29, 1992, to Gordon W. Day, ProgramChairman, Symposium on Optical Fiber Measure-ments, NIST, Div. 814.02, Boulder, CO80303-3328. For more information on the sympo-sium, contact the general chairman, Douglas L.Franzen, at the same address, or call 303/497-3346.

CONFORMITY ASSESSMENT PROGRAMPROPOSEDNIST is seeking public comments on a proposal toestablish a voluntary conformity assessmentsystems evaluation (CASE) program. CASE willenable the Department of Commerce, actingthrough NIST, to assure foreign entities that U.S.conformity assessment activities related to labora-tory testing, product certification, and quality sys-tems registration satisfy international guidelines.NIST proposes to prescribe criteria and a system toevaluate and, when requested or directed,recognize specified conformity assessment activi-ties. Program operation will be fully fee supported.Deadline for remarks is May 26, 1992. Commentsor requests for information should be sent toStanley I. Warshaw, Director, Office of StandardServices, A603 Administration Building, NIST,Gaithersburg, MD 20899, 301/975-4000, fax:301/963-2371.

NEW ATLAS IMPROVES INFRAREDSPECTROMETRY CALIBRATIONSNIST recently published the most accurate atlas todate for the calibration of infrared spectrometers.Several hundred spectral maps and tables (withdata based on frequency rather than wavelengthtechniques for absolute references) are presentedfor various calibration molecules. The text portion

of the atlas contains a description of heterodynefrequency measurement techniques, details of theanalysis, and calculation procedures. Intensitiesand lineshape parameters also are included. Toorder multiple copies of Wavenumber CalibrationTables from Heterodyne Frequency Measurements(NIST Special Publication 821), contact theSuperintendent of Documents, U.S. GovernmentPrinting Office, Washington, DC 20402. Order bynumber 003-003-03136-4 for $31 prepaid.

COMMENTS SOUGHT FOR LANGUAGEC TEST METHODNIST is soliciting the views of industrial, public,and federal computer users about its choice of amethod for testing C compilers for conformance toFederal Information Processing Standard (FIPS)160. This FIPS, which adopts ANSI StandardX3.159-1989 Programming Language C, applies toC compilers acquired for federal use after Sept. 30,1991. NIST has chosen the Perennial ANSI C Vali-dation Suite as its test method and needs feedbackto assess the method's suitability for appraisingFIPS compliance. NIST also is using a trial valida-tion service to verify the accuracy and complete-ness of its C validation procedures. The C standardis designed to permit portability of C applicationsacross a variety of hardware configurations. NIST'stest method and the results from its validationservice will be especially helpful to federal agenciesthat require FIPS 160 conformance. Writtencomments on the NIST method should be sent tothe Computer Systems Laboratory, Attn: C TestService, A266, Technology Building, NIST,Gaithersburg, MD 20899.

REPORT EXAMINES "BRIDGING" SGML/ODASTANDARDSStandard Generalized Markup Language (SGML)and Office Document Architecture (ODA) are thetwo most prominent international standards formarkup and interchange of electronic documents.However, the two are incompatible. Documentsencoded in SGML cannot be used directly in anODA-based system, and vice versa. Users of onestandard must learn the other before importing orexporting material encoded in the other standard.In a new report, NIST researchers explore creationof a "bridge" between the two standards. Thereport evaluates Office Document Language(ODL), an SGML application specifically designedfor ODA documents. Also described is a transla-tion program that converts SGML documents toODA and back. The report concedes that both

315



standards have loyal users and will likely be aroundfor some time, so it suggests ways to minimizecostly duplication efforts and allow users of eachstandard to take advantage of features offered bythe other. On the Interchangeability of SGML andODA (NISTIR 4681) is available for $17 (print) or$9 (microfiche) prepaid from the National Techni-cal Information Service, Springfield, VA 22161.Order by number PB 92-149830/AS.

SUMMER TECHNOLOGY OUTREACHCONFERENCES SCHEDULEDAn updated schedule has been announced for theeight remaining National Technology Initiative(NTI) meetings. The new schedule includes a date,July 9, for the 10th meeting in Gaithersburg, MD,and plans for an 11th meeting on July 22 in Prince-ton, NJ. Co-sponsored by the Departments ofCommerce, Energy, and Transportation, andNASA, NTI is designed to promote industry/government collaborations and commercializationof federal R&D technologies. Meetings cover twotopics of regional interest (listed below) andinclude core sections on manufacturing partner-ships, financing, and cooperative research anddevelopment relationships. The remaining sched-ule for NTI meetings is: May 14, Seattle, WA(transportation, environment); May 28, Pasadena,CA (aerospace, biotechnology); June 11, Denver,CO (natural resources, communications); June 25,Kansas City, MO (agricultural technology, ad-vanced manufacturing); July 9, Gaithersburg, MD(life sciences, systems integration); and July 22,Princeton, NJ (transportation, electronics).

NIST LABORATORY METROLOGY SEMINARHELD AT THE UNIVERSITY OF PUERTO RICO(UPR), MAYAGUEZThe NIST Office of Weights and Measures (OWM)sponsored a laboratory metrology seminar at UPRin Mayaguez Feb. 10-14, in cooperation with thePuerto Rican Weights and Measures Division,Department of Consumer Affairs. The seminar wascoordinated through the Center for Cooperation inEducation and Research in Engineering andApplied Science (Co-Hemis), also representingnational agencies in charge of technologicalresearch in Argentina, Canada, Costa Rica,Cuba, Chile, Guatemala, Mexico, Peru, DominicanRepublic, Trinidad-Tobago, Uruguay, andVenezuela.

The seminar was taught in English by a memberof NIST OWM and in Spanish by a representativeof the P.R. Department of Consumer Affairs. Theycovered basic measurement concepts and proce-dures in mass, volume, statistics, and measurementcontrol programs as part of laboratory qualityassurance. Most of the participants representedlegal metrology or legal weights and measuresjurisdictions (U.S. Virgin Islands, St. Lucia Bureauof Standards, Direccion General de Normas ofMexico, Jamaica Bureau of Standards, GrenadaBureau of Standards, and Barbados NationalStandards Institution); private industry also wasrepresented (National Standards of Puerto Rico,Inc., and Advanced Instruments, also of PuertoRico). Participants were excited about this oppor-tunity for training; plans are being made for inter-laboratory measurements in mass calibrationduring the next year. Another meeting is beingplanned for February 1993, with the objective ofcreating an ongoing Regional MeasurementAssurance Program Group.

SPHERICAL-WAVE EXPANSIONS OFPISTON-RADIATOR FIELDS,ALGORITHM DERIVEDResearchers from NIST and the Air Force havederived simple spherical-wave expansions of thecontinuous-wave fields of a circular piston radiatorin a rigid baffle and presented their derivation in apaper recently published in the Journal of theAcoustical Society of America. The team also hasproduced an accurate and efficient algorithm fornumerical computation of the near field, basedupon the spherical-wave expansions. This workrepresents a key step toward the solution of animportant electromagnetic field problem and, at thesame time, constitutes a significant contribution tothe science of acoustics. The motivation for theresearch was the need for a model electromagneticfield radiator that is both totally understood andphysically realizable, for use in validating methodsfor measuring radiated fields. Through their calcu-lations, the researchers have shed new light on thelong-standing problem of determining the near fieldof a piston radiator transducer. This is the (scalar)acoustic analog of the (vector) electromagnetic fieldproblem of analytically determining the radiatedfield of a uniformly excited circular aperture. Theentire field, both near field and far field of the pis-ton radiator is, therefore, completely characterized.The work provides the scientific community with

316



the first exact spherical wave expansion that makespossible a complete understanding of the near-fieldstructure of the piston radiator from a theoreticalpoint of view. The expansions are of considerableinterest since there are relatively few non-trivialproblems for which a complete development inspherical-wave modes can be given.

NIST TO APPLY PROBE POSITION-ERRORCORRECTION IN NEW NEAR-FIELDSPHERICAL SCANNING RANGENIST scientists have developed methods forcorrecting for positional errors of the probes usedin near-field spherical scanning. These methods areboth theoretical and computational in nature andwill be used in the measurements to be made in theNIST near-field spherical scanning range nowunder development and have general applicabilityto other spherical ranges. In near-field scanning,amplitude and phase measurements are made closeto (i.e., in the near field) the antenna beingmeasured, and the resulting data are mathemati-cally transformed into plots of the antenna's farfield. In the NIST spherical range, position errors,defined as differences between the actual positionsof the probe and mathematically defined regulargrid points lying on the surface of a true sphere (orspherical segment), will be measured accurately bymeans of angle encoders and a state-of-the-art laserpositioning system. Position errors in any of thethree spherical coordinates introduce phase errorsin the near-field data, leading to significant errorsin the calculated far field. The removal of positionerrors from the near-field data therefore improvesthe accuracy of the determination of the far field.This process becomes increasingly importantand necessary at higher frequencies, whereposition errors are a significant fraction of thewavelength.

NIST SCIENTIST REPORTS FINDINGSOF NASA'S AIRBORNE ARCTICSTRATOSPHERIC EXPEDITION IIA NIST scientist reported the preliminary findingsof the NASA Airborne Arctic StratosphericExpedition II. He is the acting manager of theNASA Upper Atmosphere Research Program andserved as program manager/scientist for theairborne expedition.

The expedition took place between October andJanuary and involved sampling the upper atmo-sphere from aircraft flying out of Fairbanks,Alaska, and Bangor, Maine. The expedition teamsought information about the causes of widespreadstratospheric ozone decreases over mid-latitudesin the first part of the calendar year and about theeffects of volcanic debris on the chemical processeswhich govern stratospheric ozone depletion.

The researchers found elevated levels of reactiveintermediates known to be responsible for ozonedestruction over a wide area ranging from easternCanada to the southernmost latitudes examined inthe study, the central Caribbean. Evidence wasobtained that reactions on aerosol surfaces reducethe concentration of reactive forms of nitrogenoxides, which act as a brake on ozone destructionby CIO and BrO. Volcanic debris from the erup-tion of Mt. Pinatubo was shown to facilitate therate of ozone loss, leading to a conclusion that thevolcanic eruption may lead to enhanced ozonelayer reduction for several years.

FLUID EQUILIBRATION IN LOW GRAVITYTwo NIST scientists participated in an experimentflown on the space shuttle Discovery Jan. 22-30.Direct and interferometric images were used torecord the equilibration of a small sample of fluidsulfur hexafluoride (SF6) following changes in thetemperature. Since the density and temperature ofthis fluid are near the critical point, this process,which ordinarily takes place in seconds, can take aslong as 1 hour. It is known that on Earth, gravitygreatly affects the final density distribution; how-ever, the influence of gravity on the way the fluidachieves this equilibrium is unclear. Once the datafrom the microgravity experiment are analyzed, theresults should provide a better theoretical under-standing of this process and guide the design offuture low-gravity critical point experiments.

The experiment team was headed by a NASAscientist and included participants from theUniversity of Maryland and the Technical Univer-sity in Munich. This team was one of three usingthe Critical Point Facility developed through theEuropean Space Agency's microgravity program.Participation in the mission, named InternationalMicrogravity Laboratory-One, included researchersfrom 14 countries.

317



NIST RADIO BROADCASTS TO SUPPORTINSPIRE PROGRAM FOR HIGH SCHOOLSCIENCE STUDENTSNIST time and frequency broadcast station WWVwill provide short broadcasts of scheduling infor-mation to high school science students participat-ing in the Interactive NASA Space PhysicsIonosphere Radio Experiment (INSPIRE). Thisunique program will involve radio reception ofsignals produced by a beam of accelerated elec-trons modulated at various audio frequencies. Thebeam of electrons will be guided by the Earth'smagnetic field producing a large virtual an-tenna.The experiment requires reception of the signalson Earth with a geographically diverse network ofreceivers. This network will be provided by IN-SPIRE, which is a group made up of physicists,amateur radio enthusiasts, scientific experimenters,and teachers. One objective is to involve a largenumber of high school students in the program.

Equipment for the satellite-signal broadcasts wascarried on the space shuttle in March in the first of10 ATLAS (ATmospheric Laboratory for Applica-tions and Science) missions. Taped observationsmade by the INSPIRE network will be gathered todetermine the footprint of the arriving signals,providing information on the propagation of signalsthrough the ionosphere and magnetosphere. Sincethe schedule of broadcasts from space was subjectto changes, a means for communicating thesechanges to the observers was needed. With itsbroad geographic coverage, the NIST shortwavebroadcasts can meet this need. Furthermore, auto-mated methods for recording messages for broad-cast are already in place providing specialannouncements (e.g., marine storm warnings) forother agencies.

IONOSPHERIC RECEIVER TECHNOLOGYTRANSFERRED TO INDUSTRYNIST scientists have successfully transferred thetechnology for a codeless GPS ionospheric receiverto a local Colorado company that deals in atmo-spheric measurement technology. The general ideaof a codeless GPS receiver had been conceived byseveral organizations including the Jet PropulsionLaboratory, but no implementation to date hasbeen as simple and versatile as the NIST receiver.

NIST developed the receiver in order to providefor continuous measurement of the ionosphericdelay involved in time transfer using the common-view method (also developed by NIST). Thereceiver picks up signals from two different

frequencies broadcast by the GPS satellites and,from the phase difference of the signals, deter-mines the ionospheric delay and thus the freeelectron content along the path to the satellite. Thereceiver stores a catalogue of all GPS satellites andidentifies each from its Doppler signature. It cantrack approximately 10 satellites, simultaneouslyproviding ionospheric measurements for each.

While the market for receivers that measure theionospheric time-delay is significant, the systemscan potentially be used for improving the accuracyof geophysical GPS surveying and for studies of theionosphere. In fact, this latter application isimportant, because the current measurementmethod for studying the ionosphere involvesexpensive and cumbersome use of reflected radarsignals which then only provide a measure of thevertical ionospheric component.

OBSERVATION OF "QUANTUM PROJECTION"NOISENIST scientists have completed and publishedtheoretical and experimental work describing apreviously unrecognized noise source they havecalled "quantum projection" noise.

In spectroscopy, "technical noise," such as laseramplitude fluctuations caused by an unstablepower supply, often dominates the noise. Thesesources of noise can be eliminated by carefulengineering. Two examples of more fundamentalnoise sources are the detection shot noise in a laserabsorption spectroscopy experiment and the fluctu-ations in signal caused by the fluctuations in thenumber of atoms in an atomic beam experiment.These sources of noise also can be eliminated orsignificantly reduced. For example, laser shot noisecan, in principle, be reduced by use of squeezedlight. In a stored-ion experiment, the number ofions can be held constant, thereby reducing theatomic number fluctuations to zero. Also, whenabsorption is detected using electron shelving, thedetection noise approaches zero since 100 percentdetection efficiency is possible. Because of thisimmunity from some sources of noise, the NISTscientists have been able to observe this "quantumprojection" noise in a clear manner. Basically, thissource of noise is caused by the statistical fluctua-tions in the number of atomic absorbers (ions inthis case) that are observed to make the transitionin an absorption spectroscopy experiment.

318



APPARATUS FOR MEASURING THERMALCONDUCTIVITY AT LOW TEMPERATURESA guarded-hot-plate apparatus has been developedto measure the thermal conductivity of insulatingmaterials in the temperature range 10 to 400 K.NIST is using the new apparatus to performmeasurements for a private company on foaminsulations with environmentally acceptable fillgases. The specimen environment may be vacuumor non-condensing gas and the mechanical load onthe specimen may be varied from 222 N to 8896 Nduring a test. Specimen diameter is 30.5 cm andthickness up to 5 cm can be accommodated. Initialtest results suggest an uncertainty of 2 percent forpolymeric foam insulations. The apparatus wasdesigned to optimize operation at low tempera-tures, e.g., a variable heat switch is used to providerefrigeration from a static cryogenic bath. Dataacquisition and temperature control are completelyautomated so that extensive data sets can begenerated with minimal operator interaction.

PROPOSED CONSORTIUM ON CASTING OFAEROSPACE ALLOYSNIST scientists recently met with industrialrepresentatives to initiate the planning for a NIST-administered consortium on casting of aerospacealloys. The meeting, which was sponsored by theAerospace Industries Association, the NationalCenter for Advanced Technologies, and the NISTOffice of Intelligent Processing of Materials, tookplace on Jan. 16 at the Aerospace IndustriesAssociation in Washington, DC. There were 31attendees from a wide variety of aerospace compa-nies and universities. There were also representa-tives from NSF, DARPA, and the Bureau ofMines.

A consensus was reached at the meeting that aconsortium should be formed to carry out precom-petitive generic research to improve the processingof aerospace metal alloys by casting. Further plan-ning by the potential consortium members iscurrently under way in the areas of process model-ing, process sensing, and thermophysical propertiesdata. A tentative consortium initiation date of July1 has been set.

ELECTRODEPOSITION OFCOMPOSITIONALLY MODULATED ALLOYSNanostructural alloys are a new class of materialswhose unusual properties are a result of the largevolume fraction of atoms either in, or close to, aninterface within the structure. These materials maybe both compositionally and structurally modulated

and include both metals and ceramics. Studies atNIST show that it is possible to produce composi-tionally modulated alloys with layers as thin as twomonolayers using electrochemical techniques. Thevirtually atomic-scale control of composition allowsthe production of alloys with tailored (or designed)magnetic, optical, and mechanical properties. Ithas been demonstrated that microlayered alloysproduced electrochemically exhibit a much higherdegree of perfection than comparably producedsputtered alloys and that large-scale structures canbe electroformed with near atomic control overproperties. Microlayered alloys have clear applica-tions to electrical contacts, magnetic recording, andsemiconductor devices.

TECHNOLOGY OF GLASS-CERAMICINSERTS TRANSFERRED TO DENTALMATERIALS INDUSTRYGlass-ceramic inserts for composite dental restora-tions, a technology invented by an AmericanDental Association research associate, wererecently introduced to the dental materials market-place. Appropriate premolded sizes and shapes oftooth-colored beta-quartz solid-solution glass-ceramic are inserted into dental composite fillingsprior to polymerization. The inserts decreasepolymerization shrinkage, increase the modulus ofelasticity, and may improve the dimensional stabil-ity of occlusal and interproximal restorations. In-sert kits were introduced in fall 1991 to the UnitedStates and world markets.

The inventor is director of the PaffenbargerResearch Center in the dental and medical materi-als group at NIST. NIST scientists advised theinventor and his research team and provided glassmelts.

PREDICTING PERFORMANCE OF ADHESIVEBONDSA new study at NIST shows that the fractureenergy of adhesive bonds loaded in shear can bepredicted from a knowledge of the stress-strainbehavior of the adhesive. This is an importantresult because the inability to predict failure inadhesive joints is a major factor limiting theiruse, particularly in structural applications whereadhesives offer the significant advantages of lowerweight and design flexibility. The recent work suc-cessfully treats the critical role played by the bondthickness. Restricting the adhesive to thin layerssignificantly changes its mechanical properties,which are important in the fracture of bondedjoints. The study used video microscopy to record

319



the details of the deformation behavior in thecrack-tip region during fracture and comparedthese data with results from stress-strain experi-ments conducted on the same adhesive confined toa thin layer with a thickness equal to that in theadhesive bond. The strain to failure in thin-layerspecimens was found to be much greater than thatobtained from experiments on bulk samplesbecause of the inability for flaws to develop in verythin layers. The area under the stress-strain curveof a thin layer adhesive was found to correlate verywell with the adhesive fracture energy. This newunderstanding will help improve the design of newadhesive systems and increase the ability to predictthe lifetimes of bonded structures.

NIST STUDY OF STAGING AREAS FORPERSONS WITH MOBILITY LIMITATIONSNIST has completed a project for the GeneralServices Administration (GSA) to evaluate theconcept of a staging area as a means of fire protec-tion for persons with disabilities. GSA has modifiedsix buildings as test installations with staging areas.Staging areas are intended as spaces in which peo-ple with disabilities can safely wait during a fire.Spaces that were turned into staging areas includedpassenger elevator lobbies, service elevator lobbies,sections of corridor, and rooms. The NIST projectconsisted of field tests, human behavior studies, andthreat analysis. The threat analysis evaluated ten-ability in route to staging areas and within stagingareas due to transport of smoke from post flashoverfires, sprinklered fires, and smoldering fires. Thesprinklered fires were based on a model of post-sprinkler heat release recently developed by a NISTscientist.

It was concluded that staging areas can be eithera haven of safety or a death trap. The difference ishighly dependent on details of design, the type offire exposure, outside wind and temperature condi-tions, and the capability and reliability of the smokecontrol pressurization system. Without pressuriza-tion, all staging areas studied are subject to lethalfailure. In many cases, the persons most needingthe staging area protection may be unable to reachthat area before their pathway (corridor or aisleways) becomes lethal. Further, it was concludedthat the operation of a properly designed sprinklersystem eliminates the life threat to all occupants re-gardless of their individual abilities and can providesuperior protection for people with disabilities ascompared to staging areas.

NIST COLLABORATES WITH THE NATIONALARCHIVES AND RECORDS ADMINISTRATION(NARA) ON OPTICAL DISK TECHNOLOGYDevelopment of a Testing Methodology To Pre-dict Optical Disk Life Expectancy Values (NISTSP 500-200) reports the results of a cooperativeresearch effort to develop a methodology for deter-mining how long records may be safely stored onoptical disk media. Partially sponsored by NARA,NIST researchers set up an optical media measure-ment laboratory to test and analyze optical diskmedia characteristics and to produce a testingmethodology that can be used to derive lifeexpectancy values for optical disks. A standardizedtest methodology and standards for measuringmedia characteristics would assist federal managersin selecting the right media for the storage ofpermanent records.

NIST PROPOSES FEDERAL INFORMATIONPROCESSING STANDARD (FIPS) FOR SECUREHASH STANDARDA FIPS for Secure Hash Standard has beenproposed which specifies a Secure Hash Algorithm(SHA) for use with the proposed Digital SignatureStandard. For applications not requiring a digitalsignature, the SHA would be required whenever asecure hash algorithm is needed for federal appli-cations.

The hash standard provides a formula forproducing a numeric value (called a "messagedigest") of a message (or more generically, anydigital information). The formula is devised so thatvirtually any change to the message results in achange to the message digest. The message digestcan be recalculated at any time and compared tothe original. If they are not identical, the informa-tion has been changed. This mechanism verifies theintegrity of data.

SPATIAL DATA TRANSFER STANDARDIMPLEMENTATION WORKSHOP ATTRACTSLARGE TURNOUTNIST, in cooperation with the National MappingDivision, United States Geological Survey, and theStandards Working Group, Federal GeographicData Committee, held a Spatial Data TransferStandard (SDTS) Implementation Workshop Feb.18-21 at NIST. The workshop attracted about 150participants from government, industry, andacademia.

SDTS recently completed the Federal Registercomment and review period of the FIPS approval

320



process. SDTS is the result of major efforts byfederal and national organizations during the past10 years. Specifications are provided for theorganization and structure of digital spatial datatransfer, definition of spatial features andattributes, and data transfer encoding.

Following in-depth tutorials on SDTS implemen-tation, the workshop featured a demonstration ofencoding and decoding test data. Other sessionscovered the refinement of the Federal VectorProfile with Topology and an initiation in develop-ing a SDTS Raster Profile. This is the first of aseries of workshops designed to educate the vendorcommunity on the implementation of SDTS.

INTERNATIONAL WORKSHOP ON HARMO-NIZING CONFORMANCE TESTING OF PRO-GRAMMING LANGUAGE STANDARDS HELDNIST recently convened the sixth in a series ofinternational workshops on harmonizing validationprocedures for testing implementations for confor-mance to programming language standards. TheInstituto Italiano del Marchio di Qualita hosted theworkshop in Milano, Italy; countries representedincluded the United Kingdom, France, Italy, Japan,Germany, and the United States. The workshopbrought together conformance testing experts forthe programming languages of COBOL, Pascal,Ada, FORTRAN, MUMPS, C, and the graphicalstandards of Graphical Kernel System, ComputerGraphics Metafile, and Computer GraphicsInterface.

The European participants briefed participantson an application that has been made to theEuropean Committee for IT Testing and Certifica-tion to register a new agreement group, theGraphics and Languages-Agreement Group forTesting and Certification (GLATC). The GLATCgroup is using documents produced in previousworkshops as the basis for developing the GLATCrecognition arrangement. Expanding the format ofworkshops to include graphic standards validationactivities also was considered.

Attendees focused on updating guidelines forvendors to register equivalent hardware andsoftware environments and on developing a proce-dural model for coordinated maintenance andcontrol of test suites. A proposal was made to regis-ter test reports rather than certificates for productsfound to have areas of non-conformance. Finally,the workshop updated the Areas of Agreementdocument, which covers test laboratory accredita-tion, test reports, registration, certification, andmutual recognition.

Calibration Services

UPDATED CALIBRATION SERVICES USERSGUIDE AVAILABLEMore than 500 different calibration services,special test services, and measurement assuranceprograms (MAPs) are listed in the NIST Calibra-tion Services Users Guide 1991 (SP 250). Theservices referenced are the most accurate calibra-tions of their type in the United States. Thesecalibrations directly link a customer's precisionequipment or in-house transfer standards tonational standards. Services are listed in thefollowing seven areas: dimensional; mechanical(including flow, acoustic, and ultrasonic); thermo-dynamic quantities; optical radiation; ionizingradiation; electromagnetic (including directcurrent, alternating current, radio frequency, andmicrowave); and time and frequency measure-ments. Copies of the guide are available from theCalibration Program, A104 Building 411, NIST,Gaithersburg, MD 20899, 301/975-2002.

Standard Reference Materials

EASY ACCESS TO MATERIALS STANDARDSOFFEREDThanks to a group of new NIST brochures, produc-ers and manufacturers of plastic, rubber, glass,ceramic, and metal products now have rapid accessto special listings of Standard Reference Materials(SRMs). The brochures list more than 400 avail-able SRMs with certified chemical and physicalproperties for calibrating instruments and evaluat-ing measurement systems. These SRMs help indus-try achieve greater quality assurance of materialsand goods by improving measurement accuracy.The group of six brochures lists SRMs for poly-mers; glass and related materials; ferrous metals;non-ferrous metals; cast iron; and ores, minerals,and refractories. Copies of the brochures (part of aseries of 16 on select SRMs), as well as informationon more than 1,200 other certified calibration stan-dards on nutrition, clinical health, engineeringmaterials, and environmental standards, are avail-able from the Standard Reference MaterialsProgram, Rm. 204, Building 202, NIST, Gaithers-burg, MD 20899, 301/975-6776.

321



SRM IMPROVES LEAD POISONINGDIAGNOSIS ACCURACYPublic health officials warn that lead poisoning-from household paints, dust, and water pipes in57 million American homes-is a serious environ-mental threat facing U.S. children. The Centers forDisease Control (CDC) responded to this issue inOctober 1991 by lowering the tolerable thresholdfor lead in blood to 10 ,lg/dL (100 parts in 109), andrecommending lead screenings for all 12-montholds. In order to help labs more accurately measurelow lead levels (less than 50 parts in 109), NIST hasissued a new Standard Reference Material (SRM)for lead in blood. The SRM contains four vials offrozen cow blood, each with a certified leadconcentration between 5 and 55 pg/dL. NISTscientists collaborated with environmental healthspecialists at the CDC in preparing SRM 955a,available for $261. To order, contact the StandardReference Materials Program, Rm. 204, Building202, NIST, Gaithersburg, MD 20899, 301/975-6776.

SAMPLE AVAILABLE FOR HUMAN HAIRDRUG ANALYSISResearchers seeking an alternative to urinalysis foridentifying drug abusers are looking strongly atusing human hair. Studies show that after inges-tion, drugs are incorporated into hair and remaindetectable for long periods of time. This gives lawenforcers a method sensitive enough to reveal druguse from months past instead of just within a fewdays of the test date. In addition, only a smallamount of hair is required and samples can beobtained from the scalp without cosmetic effect.This could help avoid privacy issues often associ-ated with urinalysis. NIST researchers have createda reference material (Reference Material 8449) tohelp ensure hair test accuracy, making the analysismore reliable and, hopefully, more admissible aslegal evidence. The product is a bottled sample ofhuman hair in powdered form that contains knownamounts of each of four different drugs: cocaine,morphine, codeine, and benzoylecgonine. Becausethe drug quantities are fixed, investigators can usethe sample to check analytical instruments orensure measurement compatibility among labora-tories. One unit of RM 8449 costs $276 from theStandard Reference Materials Program, Rm. 205,Building 202, NIST, Gaithersburg, MD 20899, 301/975-6776.

STANDARD REFERENCE MATERIAL 893AND 1295-STAINLESS STEEL (SAE 405)The Standard Reference Materials Programannounces the availability of Stainless Steel (SAE405) Standard Reference Materials (SRMs) 893and 1295. SRM 893 is in the form of chips sizedbetween 0.5 and 1.18 mm, and is intended for usein chemical methods of analysis. SRM 1295 is inthe form of a disk, approximately 32 mm in diame-ter, and is intended for use in optical emission andx-ray spectrometric methods of analysis. TheseSRMs are certified for chromium at 13.5 weightpercent and for 10 other elements whichoccur at concentrations of less than one weightpercent.

They were developed and certified under theauspices of the ASTM-NIST Research AssociateProgram.

STANDARD REFERENCE MATERIAL 2708-ZINC SULFIDE THIN FILM ONPOLYCARBONATEThe Standard Reference Materials Programannounces the availability of a zinc sulfide thin filmstandard, Standard Reference Material (SRM)2708. It is intended primarily for use in thestandardization of x-ray fluorescence spectrome-ters for sulfur determinations in applications wheremeasurements are made of particulate mattercollected on filter media or on thin films. It consistsof a single layer of zinc sulfide, about 0.02 p.mthick, deposited on a polycarbonate filter andmounted on an aluminum ring to maintain uniformgeometry. Certification for sulfur is based onisotope dilution thermal ionization massspectrometry; an informational (non-certified)value is provided for zinc.

STANDARD REFERENCE MATERIAL 2034-HOLMIUM OXIDE SOLUTION WAVELENGTHSTANDARD 240 TO 650 nmThe Standard Reference Materials Programannounces the availability of a renewal lot ofStandard Reference Material (SRM) 2034. ThisSRM is a certified transfer standard intended forthe verification and calibration of the wavelengthscale of ultraviolet and visible absorption spectrom-eters having nominal bandwidths not exceeding3 nm. Certification is for wavelength location ofminimum transmittance of 14 bands at each of six

322



bandwidths, in the range 240-650 nm. SRM 2034 isan aqueous solution of four percent (by weight)holmium oxide in 10 percent (V/V) perchloric acid,flame sealed in a non-fluorescent fused-silicacuvette of optical quality. The square cuvette has anominal 10 mm pathlength and will fit into thesample compartments of most conventionalabsorption spectrometers.

REFERENCE MATERIALS 8506 AND8507-MOISTURE IN TRANSFORMER OILAND IN MINERAL OILThe Standard Reference Materials Programannounces the availability of Reference Materials(RMs) 8506 and 8507, primarily intended for use incalibration of equipment used to determine waterin transformer oils, mineral oils, and/or materialsof a similar matrix. RM 8506 is a petroleum electri-cal insulating oil containing 21 mg/kg water, whileRM 8507 is a petroleum lubricating oil containing47 mg/kg water. Coulometric Karl Fischer mea-surements from two round-robin analytical pro-grams were used to assign the referenceconcentration values; chromatographic and KarlFischer measurements made at NIST were used forconfirmation.

Standard Reference Data

NEW PC PACKAGE SPEEDS CHEMICALMICROANALYSISScientists using electron microscopy to analyzechemical composition can save hours of laboratorywork with a new software and database packagefrom NIST and the National Institutes of Health.The system, called the Desktop Spectrum Analyzerand X-Ray Database (DTSA), can simulate experi-ments and evaluate electron microscopy data. Itallows scientists to design effective experimentsquickly without using trial and error to find thebest experimental conditions. With a special acqui-sition board installed in the desktop computer, theDTSA also can collect data directly from an elec-tron microscope. The DTSA sells for $790 throughthe Standard Reference Data Program at NIST. Itis designed for Apple Macintosh II series comput-ers and requires a minimum of 4 megabytes ofmemory. For more information, contact theStandard Reference Data Program, A320 PhysicsBuilding, NIST, Gaithersburg, MD 20899,301/975-2208, fax: 301/926-0416.

323



Calendar

August 2-6, 1992NATIONAL CONFERENCE

OF STANDARDS LABORATORIESWORKSHOP AND SYMPOSIUM

"MANAGING WORLDWIDE MEASUREMENTS"

Location: Grand Hyatt WashingtonWashington, DC

Purpose: To provide a forum for exchange ofideas, techniques, and innovations among thoseengaged in measurement science and has evolvedas a major educational activity of the year for themeasurement community worldwide.Topics: Technical sessions include papers on me-chanical, dimensional, electrical, optical, and tem-perature. Managaement sessions on internationalstandards, develo;pment of a calibration lab, andlaboratoary accreditation. In addition there will beworkshops on subjects such as: Intrinsic andDerived Standards, Automatic Test Equipment,Measurement Assurance, National MeasurementRequirements, Biomedical and PhaarmaceuticalMetrology.Format: Symposium, exhibits and workshopsessions.Sponsors: NIST and the National Conference ofStandards Laboratories (NCSL).Contact: Wilbur Anson, 303/440-3339, NCSL,1880-30th St., Suite 305B, Boulder, CO 80301.

325