Contents · DIPPR Database Expanded to 1,023 Pure Compounds ... International customers please add...

Volume 94, Number 3, May-June 1989

Journal of Research of the National Institute of Standards and Technology

Contents

ArticlesA Brief Review of Recent Superconductivity D. R. Lundy, 147

Research at NIST L. J. Swartzendruber,and L. H. Bennett

Calibration of Voltage Transformers William E. Anderson 179and High-Voltage Capacitors at NIST

Consensus Values, Regressions, Robert C. Paule and 197and Weighting Factors John Mandel



News Briefs

GENERAL DEVELOPMENTSParticipants Wanted for OSI, ISDN Security ProgramVendors Provide Workstations for OSI Security WorkStandard for Interchanging Document ProposedMercury Ion Laser-Cooled to LimitNIST to Study New Polymer Resins for IndustryNew Building Criteria for PrisonsNIST Studying Unintentional EED FiringCD-ROM Speech Database AvailableNIST Report Summarizes Inventions ProgramDiamond Films Produce New GemsNew Way to Evaluate Protective Coatings on MetalsAbstracts of Recent Publications AvailableCollected Papers on Ion ResearchChemical Structure of DNA Damage UncoveredNIST, NSF Plan Joint Neutron Research FacilityMeasuring High-Temperature SuperconductorsSteel in Fracture Test Sets U.S. RecordNIST Invites Vendors for GOSIP Evaluation ProjectNational Earthquake Awareness WeekFiber-Matrix Interface Properties via an Instrumented Indenter TechniqueNIST Collaboration With Oak Ridge National Laboratory on Neutron StandardsMajor NIST Collaboration to Study Novel Magnetic SystemsAtomic Positions From x-Ray Standing WavesPatent Application on New x-Ray Diffraction DeviceMagnetic Thin Films With Large Perpendicular Moments

205

CALIBRATION SERVICES 210Industry Help Requested on Coaxial ConnectorsNew Calibration Services Users Guide AvailableNCSL Ad Hoc Committee 91.3 on the Change of the Temperature Scale

STANDARD REFERENCE MATERIALS 211Improving Lead-in-Fuel Analyses Is Aim of MaterialsNew Materials Can Help Gauge Coal Sulfur ContentNew Australian Bauxite Ore Standard Issued

STANDARD REFERENCE DATA 212DIPPR Database Expanded to 1,023 Pure Compounds

Calendar 213

The Secretary of Commerce has determined that the publication of the periodical is necessary in thetransaction of the public business required by law of the Department. Use of funds for printing thisperiodical has been approved by the Director of the Office of Management and Budget throughOctober 1, 1989.

United States Government Printing Office, Washington: 1989

Superintendent of Documents Subscriptions Order Form

Ordrer Processirg Code:

*6218

D YES,

Charge your order.it's easy! W

please send me - subscriptions to the JOURNAL OF RESEARCH OF THENATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY for $13 for oneyear so I can stay up to date on the latest developments in measurement science andtechnology.

1. The total cost of my order is $_ . All prices include regular domestic postage and handling and aresubject to change. International customers please add 25%.

Please Type or Print

(Company or personal name)

(Additional address/attention line)

3. Please choose method of payment:

D Check payable to the Superintendent of Documents

D GPO Deposit Account I I I ||||I -DD VISA or MasterCard Account

I I I I I I I I I I I I I I I I I I I I1-(Street address)

(Credit card expiration date)Thank you for your order!

(Signature) 2/89

4. Mail To: Superintendent of Documents, Government Printing Office, Washington, D.C. 20402-9391

2.

(City, State, ZIP Code)

( )(Daytime phone including area code)



A Brief Review of Recent SuperconductivityResearch at NIST

Volume 94 Number 3 May-June 1989

D. R. Lundy, A brief overview of recent supercon- determination of phase diagrams andL. J. Swartzendruber, ductivity research at NIST is presented. crystal structure. For the low-tempera-and L. H. Bennett Emphasis is placed on the new high- ture superconductors, research spans

temperature oxide superconductors, studying the effect of stress on current

National Institute of Standards though mention is made of important density to the fabrication of a newwork on low-temperature superconduc- Josephson junction voltage standard.

and Technology, tors, and a few historical notes are in-Gaithersburg, MD 20899 cluded. NIST research covers a wide Key words: ceramics; copper oxides;

range of interests. For the new high- cryords: eering; cr oxide;temperature superconductors, research cryogenic engineerng; crystal structure;activities include determination of physi- electronic structure; materials science;cal properties such as elastic constants measurement science; overview; per-and electronic structure, development of ovskites; superconductivity.new techniques such as magnetic-fieldmodulated microwave-absorption and Accepted: January 26, 1989

Contents 1. Introduction

1. Introduction .......................... 1472. Background .......................... 1483. Crystal Structure ...................... 1504. Impurity Effects ............ .......... 1505. Synthesis ............................. 1516. Processing-Property Relationships ..... .. 1577. Electrical Contacts ........... ......... 1618. Electronic Structure ......... .......... 1629. Physical Properties ........... ......... 164

10. Theory ............................. 16811. Applications .......................... 16912. Other High-Temperature Superconductors. 17013. Low-Temperature Superconductors ..... 17114. Conclusion ........................... 17315. Acknowledgements .................... 17416. References ........................... 174

Research in superconductivity at NIST has along history, in part because of its importantapplications to measurements and materialsscience. The recent discovery of materials withunexpectedly high superconducting transition tem-peratures has brought renewed interest in thescience and technology of superconductorsthroughout the world. The purpose of this paper isto briefly review recent activities at NIST in thisfield and to discuss current usage and futurepossibilities for both conventional and high temper-ature superconductors. Some references will bemade to historical superconductivity contribut-ions at NIST but a complete review is notattempted.

147



2. Background

The field of superconductivity began with thediscovery by H. Kamerlingh-Onnes in 1911 thatmercury wire at 4.2 K had zero electrical resis-tance. Zero resistance implied transmission of cur-rent at any distance with no losses, the productionof large magnetic fields, or-because a supercon-ducting loop could carry current indefinitely-storage of energy. These applications were notrealized because, as was quickly discovered, the su-perconductors reverted to normal conductors at-arelatively low current density, called the criticalcurrent density, Jc, or in a relatively low magneticfield, called the critical field, Hc. In 1916, Silsbee,at the National Bureau of Standards, hypothesized[1] that the critical current for a superconductingwire was equal to that current which gave the crit-ical field at the surface of the wire. The reason forthis behavior was not made clear until the discov-ery [2] of the Meissner effect in 1933.

The discovery and development, in the 1950sand 1960s, of superconductors which can remainsuperconducting at much higher fields and currentsmade practical the production of useful supercon-ducting magnets (see table 1). Such high-field su-perconductors, which exhibit two critical fieldsdesignated Hc: and HC2, are called type-II. In 1950,another NBS scientist, E. Maxwell, was the discov-erer [3] of the isotope effect, which was also inde-pendently discovered by Serin et al. [4]. Thisexperimental observation was an important key totheoretical explanations of the mechanism of super-conductivity. In the isotope effect, the critical tem-perature for many superconductors depends on theisotopic mass, indicating that lattice vibrations areinvolved in the superconductivity, and that the at-tractive coupling between electrons is through thelattice vibrations (i.e., phonon-mediated). The dis-covery [5] of the Josephson effect in 1962 openedup exciting potential for the use of superconductors

Table 1. Selected superconductor applications

Application Advantage Comments

Generators with superconducting Lifetime cost savings up to 40% U.S. built a 10 MW prototype; USSR iswires in rotors building a 300 MW prototype

Energy storage rings Efficient, site independent, can revert from Pilot programscharging to discharging mode in less thanI second. Can stabilize system

Power transmission lines Reduced resistive losses Prototypes tested

Magnets for magnetic resonance Superconducting magnets result in shorter Largest current commercial applicationimaging exposure times and sharper images compared of superconductivity; wide acceptance as

to conventional magnets medical diagnostic tool

Chip interconnects Lack of electrical resistance; reduces heat In research stagebuildup; permits dense packing...rapidtransmission of signal

SQUIDS (superconducting quantum Extremely sensitive to magnetic fields Used for mineral exploration; anti-sub-interference devices) marine warfare potential; development

underway for use in medical diagnosis

Josephson Junction switches Fast switching times, low power dissipation, Used in fast-sampling oscilloscope;dispersionless transmission potential computer logic elements

Josephson Junction voltage standards Reliable, stable. Absolute voltage based on In use by, and available from, NISTfundamental constants

Magnets for fusion devices Magnets to confine plasma Prototype systems constructed in Franceand the Soviet Union

High-energy physics High fields to guide beams, reduced energy Superconducting supercolliderconsumption

Ship propulsion (motors) Smaller, quieter motors; elimination of gearbox Navy has protoype

Magnetohydrodynamic (MHD) High fields interact with a plasma to generate Prototypes constructed by thepower generation electricity Soviet Union

Magnets for MHD ship propulsion Quiet, more efficient, higher potential speeds Demonstrated by JapanMagnetic casting Eliminates contamination

Magnetic separation Separates weakly magnetic materials

Magnetic bearings Eliminates friction

IR sensors Smaller packages

Magnets for magnetically levitated Rapid and efficient mode of transportation Demonstrated by Japantrains (MAGLEV)

148



in measurement science and in high-speed elec-tronic devices.

Until 1986, the highest critical temperature ob-tained for any superconductor was only 23.2 K.This meant that superconductors had to be cooledby liquid helium-an expensive and sometimes un-reliable process. Consequently, many potential ap-plications were not commercially viable. Inaddition, most scientists had come to regard super-conductivity as a mature field with little possibilityfor any significant increase in critical temperatures.All this suddenly changed with the discovery byK. A. Bednorz and J. G. Muller of high-tempera-ture superconductivity.

In April 1986, Muller and Bednorz published apaper [6] on the possible existence of superconduc-tivity in a ceramic material, La-Ba-Cu-O, with asuperconducting transition temperature, Tc, of30 K, the first increase since 1973. Their discoverywas the result of several years of extensive investi-gations on metal oxides, some of which had earlierbeen shown to be superconducting. It is note-worthy that superconductivity in oxides had beenknown for many years. In fact the first oxide super-conductor was discovered at NIST. In 1964,Cohen predicted [7] that, based on the Bardeen-Cooper-Schrieffer (BCS) theory [8], semiconduc-tors could become superconductors. At this time ametal-oxide semiconductor-SrTiO 3 -being inves-tigated at NIST seemed to have the characteristicspostulated by Cohen as necessary for superconduc-tivity. Kahn and Leyendecker [9] had determinedthe energy band structure of SrTiO3 andFrederikse et al. [10] had determined the density ofstates in the conduction band. Schooley et al. [11]found superconductivity below 0.3 K in reducedSrTiO3 in 1964; this oxide was demonstrated to be atype-II superconductor by Ambler et al. [12] in1966 (see fig. 1). Tc was only 0.3 K however.Though substituting Ca and Ba for Sr [13] raisedthe Tc to 0.5 K, the low critical temperatures lim-ited general interest in these materials.

According to Muller and Bednorz [6], their re-search was influenced by the French work on theLa-Ba-Cu-0 system [14]. However, the French sci-entists were not looking for superconductivity.When researchers from the University of Tokyo[151 confirmed the findings of Muller and Bednorz,the era of "High-Temperature Superconductivity"was ushered in.

The end of 1986 and the beginning of 1987 wasmarked by synthesis of rare-earth metal oxides ofincreasingly higher Tc, culminating with the dis-covery [16] of the Y-Ba-Cu-O (YBCO) supercon-ductor with a Tc of 93 K. This was a significant

breakthrough because the material was supercon-ducting in liquid nitrogen (boiling point=77 K).Nitrogen is much more abundant than helium,much less expensive, and liquid nitrogen cryogenicsystems are less complex than systems using heliumrefrigeration. One application which could benefitfrom nitrogen cooling is the development of hybridmicroelectronic technology (semiconductor-super-conductor devices)-both gallium arsenide and sili-con can be tailored to perform better at liquidnitrogen temperatures.

4

2-

0

-2 -

-4 -

-6 -

-820 40 go

Applied Field, Oe80 t00

Figure 1. Partial hysteresis loop of SrTiO3 obtained at 0.15 Kusing a ballistic galvanometer. The shape of the loop indicatesthat SrTiO 3 is a type-II superconductor. (Unpublished datacourtesy of J. F. Schooley.)

The ease of making YBCO permitted its investi-gation by many laboratories. In fact, a number ofhigh school students synthesized it for use in sci-ence fair projects. At various times researchers re-ported Tc's greater than 100 K-some reportedsuperconductivity at room temperature and above.These observations were not confirmed; many ofthe results were irreproducible or the samples werenot stable. At the end of 1987, the highest Tc stoodat 95 K. In February 1988, Japanese, Chinese, andU.S. researchers found superconductivity in cop-per-containing oxides without rare earths. Thesenew non-rare-earth containing superconductor ma-terials incorporate either bismuth or thallium.Compounds containing the latter have a confirmedTc of z 127 K. These new high-temperature super-conductors containing bismuth or thallium mayhave some advantages over the superconductorscontaining rare-earths. Since the critical currentdensity increases as T/Tc decreases, a Tc far abovethe operating temperature of liquid nitrogen (77 K)is advantageous. Furthermore, the new materialsare more stable than the rare earth superconduc-tors; they do not lose oxygen or react with water.

149

4



In addition to trying to develop new high-Tc ma-terials, researchers also were trying to fabricatematerials with improved critical current densities(Jc). Current densities as high as 105-106 A/cm 2

may be needed for applications such as magnets,motors, and electronic components. The high-tem-perature superconductors are ceramics and have allthe brittleness problems associated with non-super-conducting ceramics. In addition, Jc is not an in-trinsic property of superconductors but is afunction of the processing procedure. The rare-earth superconductors also have highly directionalproperties. Therefore, a crucial problem is to fabri-cate the material into a useful shape and still havesufficiently high 1 c and mechanical strength forpractical applications.

Single crystal films of YBCO have current densi-ties above a million A/cm2 . However, results forbulk polycrystalline materials are orders of magni-tude less. Recently, researchers have grown non-oriented polycrystalline thallium films with Jc inthe millions [17]. Novel processing techniques suchas explosive compaction, rapid solidification andlaser ablation are currently being explored.

NIST personnel have been actively engaged infabricating and characterizing high-Tc supercon-ducting materials. This paper is a brief synopsis oftheir wide-ranging activities. Some mention of sig-nificant work in other superconductors will also bemade.

3. Crystal Structure

As was true for the previously mentionedSrTiO3 , the superconducting YBCO phase is a dis-torted perovskite [18]. Ideal perovskites have theform ABX3 , where A and B are metallic cationsand the X atoms are non-metallic anions. In super-conducting yttrium barium copper oxide, the struc-ture (fig. 2) is a defect perovskite of the formYBa2Cu 30 7 -x (YBCO). Oxygen and oxygen vacan-cies are the key to the superconductivity. Onewidely-used method for refining the structure ofYBCO is neutron diffraction because x-rays are notsensitive to oxygen atoms.

YBa2Cu3O7-, with x =0, 0.2, 0.5, and 1 was stud-ied [19-22] using the neutron diffraction facilitiesof the NIST reactor. YBa2Cu3O7 and YBa2 Cu3 068are orthorhombic and superconducting, and arecharacterized by Cu-O chains along the b axis. Inthe 07 material, oxygen atoms occupy 4 differentsites with 0(4) forming chains along the b-axis di-rection of the orthorhombic cell. The environ-

ments of the barium atoms and the copper atomslocated at (000), (0(z) change significantly with theamount of oxygen in the cell. YBa2Cu3O6 , which istetragonal and a semiconductor, is derived fromYBa2 Cu3 07 by removing oxygen along the b axis.While all the oxygen sites are occupied in the 07material, in the 06.8, there are oxygen vacancies lo-cated in the chains. The superconducting Tc is re-duced for deviations from O7, indicating thatoxygen vacancies disrupt conduction pathways.

* Cu(I)o Cu (2)

0f BaO Y

0 0(1)(1 0(2)

@ 0(3)

e 0(4)

c

b

a

Figure 2. Crystal structure of YBCO as determined from neu-tron diffraction showing location of the four 0 sites, two Cusites and single Ba and Y sites 1211.

The crystal structure of a strontium analog ofLa-Ba-Cu-O, Lals 5Sr0 15CuO4, was also examinedby neutron diffraction [23]. This highly two-dimen-sional structure, shown in figure 3, was found to betetragonal at ambient temperature, but became or-thorhombic at 200 K, resulting in the buckling ofthe Cu-O planes.

4. Impurity Effects

One of the outstanding questions in high-TcYBa2Cu3O7 superconductors is the relative impor-tance of the Cu-0 2 planes [Cu(2)-site] and the Cu-Ochains [Cu(l)-site]. Zn and Ga [24] have been used

150



to selectively substitute the Cu(2) and Cu(l) siteswhile maintaining the oxygen content near 7, asdetermined using neutron diffraction. These re-sults, plus other recent work on Al, Co, and Fesubstitutions, have shown that, in general, 3+ ions(e.g., Ga, Co, etc.) substitute predominantly forCu"+ on the chain sites, suppress the orthorhombiccrystal distortion, and enhance the overall oxygencontent above 7 due to valency effects. On theother hand, di-positive Zn++, which substitutesonly on the "plane" sites, retains the orthorhombicstructure, but rapidly destroys superconductivitywith only a few percent Zn substitution. Theseneutron results (combined with bulk data) havedemonstrated that the integrity of the planes ismuch more important in sustaining high-Tc super-conductivity than the chains, and have also shownthat the orthorhombic cell distortion is not essen-tial for high transition temperatures.

CT tI

OT

Figure 3. Generalized view of the highly two-dimensional te-tragonal Lal. 55 r0. 5CuO4 structure. On the scale of the figure,the orthorhombic and the tetragonal structures are not distin-guishable except that the orthorhombic unit cell is twice thearea in the plane perpendicular to c. The large shaded areas areLa and Sr atoms; the small circles, Cu. Oxygen atoms are at thevertices of polyhedra [23].

Much of the prevailing theoretical work pertain-ing to the origin of the superconducting pairing inthe high-Tc superconductors has focussed on amagnetic coupling of spins. YBa2Cu307.., with to-tal oxygen content below z6 .5 (i.e., x>0.5) hasbeen shown to exhibit strong antiferromagneticcorrelations. To examine these matters in more de-tail, neutron diffraction studies [25] were carriedout on materials in which Co has been substitutedfor the Cu on the "chain" site in order to enhancethe magnetic interactions. The results showed mag-netic ordering temperatures near 400 K, represent-ing more than enough energy to account for the=95 K superconducting transition temperatures.For Co concentrations of 20%, distinct orderingtemperatures were found for plane (z 400 K) andfor chain (z 2 0 0 K) site antiferromagnetic order-ings, while for 80% substitutions, both sites or-dered at the same temperature (z 4 35 K). The formof the temperature dependence of the observedmagnetization revealed strong couplings, bothwithin the planes and also between chain and planesites.

By using the Mbssbauer effect in Fe-doped rareearth-barium-copper-oxygen samples (with the rareearth being Y, Pr, or Er) the antiferromagneticcoupling correlations could be directly observed[26, 27]. Since the Fe atoms substitute for the Cuon both chain and plane sites, the antiferromag-netism present on the plane sites and the paramag-netism on the chain sites could be simultaneouslyobserved in the Mdssbauer patterns. It was alsodemonstrated that asymmetries in the M6ssbauerspectra are the result of a preferential alignment ofthe crystallites that arise during the normal samplepreparation process.

5. Synthesis

Phase equilibria diagrams provide phase compo-sitions and relationships under specific conditions.Such information is needed to characterize materi-als and develop synthesis procedures. Roth and hiscolleagues have been active in determining phaserelationships for the Y-Ba-Cu-O system. Prelimi-nary phase diagrams were constructed [28] for thebinary systems BaO-1/2Y2 03 ; BaO-CuO; andl/2Y 203-CuOx, the bounding oxide systems of theternary, and Y-Ba-Cu-O. Nine compounds arefound in the BaO-Y20 3-CuO system [29, 30] in thetemperature range 950-10000C. Three of the com-pounds, Ba2 YCu307_x (the superconducting phase),BaY2 CuO5 (the "green" phase which is found to

151



exist with other phases), and Ba3YCu2O2 werecharacterized by x-ray diffraction (XRD) [31-33].Substitutions of lanthanides for yttrium in BaY2 05were also characterized by XRD [34]. Fourteenstandard reference patterns for six high-Tc super-conducting and related phases have been reported[35]. Roth [28] found that all compositions in theternary system containing 50% BaO alwaysshowed a small amount of the green phase. A tenthphase, Ba2 CuO3 , which has a melting point below950 'C was characterized by XRD [36]. Studies ofphase equilibria in air [37] showed that the super-conductor phase melts through a four-phase region(fig. 4) from about 950-1002 'C. This region is dueto the presence of C02, probably mainly in the liq-uid phase. The presence of CO2 in the supercon-ducting phase was inferred. The transition fromtetragonal to orthorhombic structure was con-cluded to be metastable and no large primary phasefield consisting only of the superconducting phaseand liquid was identified.

The substitution of SrO for BaO in theBaO:Y2 03 :CuO system was studied to determinethe extent of solid solution of Sr in YBCO and toidentify any new phases. It was found that Sr couldbe substituted for Ba up to about 60%. There wereno ternary compounds in the Sr-Y-Cu-O equivalentto the three ternary phases in the Ba system, but anew binary phase Sr14Cu2404 1 was found. TheSrO-CaO-CuO system was also studied as part ofan investigation of the SrO-CaO-Bi2 -03 -CuO sys-tem. At 950 'C, there were three extensive solidsolutions at (Sr, Ca):Cu ratios of 2:1, 1:1, and 24:41.A new ternary SrxCa1_xCuO2 (xO.15) was foundand a new phase, probably CaCu2 , stable only be-low -740 'C, was identified [38, 39].

Since the oxygen content in YBa2Cu30,.xstrongly affects the superconducting and structuralproperties, the effects of variations in annealing(oxygenating) were studied. Single-phase sampleswere annealed at temperatures from 400'C to1000'C [40, 41], then quenched in a liquid-

BaO (BaCO3)

01/2(Y 203 )

20 40 Y2Cu2 05 60 80Mol %

100CuO

Figure 4. Ternary phase diagram of Y-Ba-Cu-O constructed from figures I and 6 of reference [37].(Unpublished figure courtesty of R. Roth.)

152



nitrogen-cooled copper cold well, through whichliquid-nitrogen-cooled helium gas was passed at arapid rate. The goal was to quench in the high-temperature structures and stoichiometries. Sam-ples were initially examined by x-ray diffraction. Itwas observed that, as the annealing temperaturedecreased, the ceramic became more orthorhom-bic-going from fully tetragonal at 1000 'C to fullyorthorhombic at 400 'C (fig. 5). The phase transi-tion occurred at 708-719 'C. The dependence ofcell volume on temperature was not linear, becom-ing substantial only at 400-650'C, the orthorhom-bic region. The limiting volumes were the volumeof YBa2 Cu3 07 annealed in air and YBa2Cu3O6 an-nealed in argon. Two possible orthorhombic re-gions were indicated-a <b =c/3, and a <b <c/3.ac susceptibility measurements were made for sam-ples annealed up to 708 'C (samples an-

nealed above 750 'C showed no Meissner effect).As shown in figure 6, a plot of Tc0 (Tc onset) ver-sus annealing temperature showed two plateaus-91 and 58 K. While these might indicate twodifferent orthorhombic phases, lack of corroborat-ing x-ray data prevented a firm identification.YBa2 Cu30 7 was also examined by thermogravi-metric analysis (TGA) and differential scanningscanning calorimetry (DSC). TGA results alsogave indications of two regions. DSC/TGA analy-sis showed a thermal event when YBa2 Cu30 7 .,_ washeated to 900 K which might be the result of mi-crohomogeneities or of discontinuities in the oxy-gen vacancy ordering [42]. DSC studies of thismaterial [43] showed that the phase transition doesnot associate with an enthalpy change, a character-istic of second-order transitions. Based on this andthe x-ray data, the phase transition appears to be a

CT

Co

bo a

aT

a0

Orthorhombic Tetragonal

I I I. I I400 500 600 700 800 900 1000

Temperature ('C)

Figure 5. A plot of the cell dimensions of YBCO as a function of annealing temperature.As the temperature increases, the cell dimension bo decreases while ao increases [40].

153

11.84

11.82

11.80 _-

11.78

11.76

11.74 _-

11.72 _

11.70 _

11.68 _

3.88 W

0

(n0co

E

a,C)

3.86 H

3.84 -

3.82 F-

3.80

.,



second order, order-disorder type. TGA was alsoused to determine the oxygen diffusion coefficientof YBCO [44]. The oxygen was found to diffusefaster than in some insulating oxides, such as A1203 ,but slower than in oxides which have been classedas oxygen conductors. The diffusion constant plot-ted against l/T is linear from 400-600 K. From 600-750 K, it is not linear and only weakly dependenton temperature, which may be due to structuralphase changes.

Tco

400 500 600 700 800 900

Annealing Temperature

1000

Figure 6. A plot of Tc0 versus annealing temperature. Note thetwo plateaus at 91 K and 58 K [41].

Structural phase transitions of Ba2RCu306+,(where R = Sm, Gd, or Er) were studied [45] todetermine the effect of the span of radii and themagnetic properties of Gd and Er. Samples wereannealed at 400-1000'C and quenched in a liquid-nitrogen-cooled copper well. The x-ray spectrawere similar to that of YBCO. The orthorhombic-tetragonal transition always occurred between 625-770'C. The Gd compounds showed an increase inthe c axis due to oxygen vacancies, as in YBCO.The rare-earth elements with smaller radii stabi-lized the orthorhombic phase to a higher tempera-

ture. The phase transformations are apparently sec-ond order and may involve two orthorhombic re-gions which correspond to two Tc plateaus-oneat 92 K and one between 52 and 60 K. These re-gions have the same general structure, but differentoxygen distributions corresponding to ordered anddisordered modifications within the orthorhombicstructure. No obvious plateau was detected for theGd compounds however. TGA also identifiedplateaus-three apparently single-phase regions forSm, Y, and Er and two for Gd. The fact that theliterature Tc values are about the same for all thecompounds suggests that the superconducting elec-trons are not strongly associated with the rare-earth elements [46].

The effect of annealing atmosphere (fig. 7) wasalso studied [47, 48]. Samples annealed and cooledin oxygen were found to have sharper supercon-ducting transitions than those annealed and cooledin air. They also had sharper diffraction peaks. Thebroadening is due to crystallite size differences andmicrostrain/chemical inhomogeneity which canoriginate from twinning, anisotropic thermal ex-pansion, and oxygen vacancies. Thermal analysisdetermined that the maximum oxygen content isobtained by annealing at 450 'C, or slightly above,and that the oxygen loss is reversible [43].

It is clear from these results that the processingparameters must be carefully controlled to yieldthe desired material. There is an additional con-cern. While it is known that exposure to water candestroy the superconducting ability of YBa2Cu3O7it also has been found that acetone can be deleteri-ous [49]. To obtain dense, strong ceramics, thepowders are often milled to a small size before sin-tering. When an acetone slurry is used, a non-superconducting tetragonal phase can be formed ifthe slurry is dried at 200 'C. The superconductingorthorhombic phase can be restored by annealingat 950 'C in Oz.

It may be possible to avoid the grinding stepsaltogether by employing chemical synthesis. Foursystems-the coprecipitation of yttrium, barium,and copper hydroxy-carbonates; the hydrolysis ofyttrium, barium, and copper alkoxides in ethanol/toluene; the reaction of barium and yttriumalkoxides with Cu(OH)2; and the hydrolysis ofyttrium, barium, and copper alkoxides inmethoxyethanol/ethanol were studied [50]. All thesystems showed BaCO3 , CuO, and Y203 whenheated at 400-600 'C. The samples must then beheated to 800-950 'C to obtain YBa2 Cu3 O,_^ andsubsequently annealed in oxygen at 450-600 'C toobtain superconductivity.

154



0 25 30 35 40 45 50 55 60

20

25 30 35 40 45 50 55 60

28

Figure 7. Effect of oxygen on the structure of YBCO. a) x-ray diffraction pattern ofYBCO annealed in air (orthorhombic). b) x-ray diffraction pattern of YBCO annealed inargon (tetragonal). The most striking features are the intensity reversal of the two sets ofdoublets at around 32-33° and 57-60°, and the shifting of positions of corresponding peakswhich indicates different cell dimensions [47].

The making of YBCO films was also investigatedusing two different techniques. In the first ap-proach, films were made from a bulk superconduc-tor by laser ablation [51], as shown in fig. 8. Apulsed-laser source was used to vaporize the sur-face of a disk made from superconducting YBCOand deposit a film on a fused silica substrate. Filmswere made by irradiating a spot or raster scanning.The resulting films were 1 cm2 in area and thickerin the center, 2 ltm, than on the edges. The as-de-posited films had superconducting regions withproperties comparable to the bulk material. Similarresults were obtained for La-Sr-Cu-O (LSCO). Thepotential advantage of this method is that the filmdoes not need a high temperature anneal to incor-porate oxygen. This is important in hybrid elec-

tronic (superconductor-semiconductor) applica-tions where a heat treatment could destroy thesemiconductor.

A second technique investigated for makingYBCO films [52] utilized co-evaporation of Y, Cu,and BaF2 . These materials were deposited simulta-neously onto a room-temperature substrate. Oxy-gen was introduced into the vacuum system duringdeposition. At this stage, films containing BaF2 aretolerant to moisture, air, positive photoresist, de-veloper, and common solvents. Annealing in oxy-gen and water vapor incorporated additionaloxygen into the film and reacted away the fluorine.The choice of substrate is critical for many applica-tions. The best films have been fabricated on Sr-TiO3 . A resistive transition, about 0.5 K wide, isshown in fig. 9.

155

1000

c.

500

c

0 2

1000

C.,

500

021

Ba2 YCu3 0 7x Annealed in Argon

_ o



VACUUM C$IAM&lEII

WINDOW

Figure 8. Schematic of laser-ablation process [141].

600

500

E 400

C

If>. 0

Q) 200a)~

100-

PAIIENT nULKSPECIMEN

ABLATEDMATERIAL

FI rLMSUBSTRArTE

Temperature (K)

Figure 9. Resistive transition of YBCO film made by coevaporation of Y, Cu, and BaF2 on aSrTiO3 substrate [52].

Films deposited using the second method arepatterned using conventional photolithographicprocesses [52]. Prior to deposition of the YBCO,the substrate is coated with photoresist, exposedwith a pattern using a projection printer, and de-veloped. The exposed and developed resist exposes

bare substrate where the patterned YBCO is to re-main. After the room temperature deposition of the Y,Cu, and BaF2 , but before the oxygen anneal, the re-maining photoresist is dissolved, removing the un-wanted portions of the film. Annealing in oxygenas before creates a patterned superconducting film.

156



Superconducting strips with dimensions as smallas 1.5 ,tm have been successfully fabricated. A pat-terned film having a constriction of 5 X 5 ,tm had acritical current density of 5.6x 106 A/cm2 at 4 K.As mentioned earlier, such high critical currentdensities are not yet achievable in bulk samples.Patterned films of this sort are being used in funda-mental studies of noise in small constrictions and astransition edge bolometers. Efforts to make high-Tc Josephson junctions are in progress.

6. Processing-Property Relationships

The understanding of the relationship betweenmicrostructure, processing, and properties is par-ticularly important for the high-temperature super-conductors. These are oxygen-sensitive, brittlematerials whose processing parameters need to becontrolled to produce optimum properties. The un-derstanding of this relationship requires, in additionto the measurement of electrical conductivity,techniques such as x-ray diffraction, magnetic sus-ceptibility, and ultrasonics. Neutron activationanalysis can be used to determine the stoichiometry

1

C_

U)

.I

YJ

* -

0cz_

X

0.5

0

-0.5

-175 80

[53]. ac susceptibility measurements can be used tocharacterize the superconducting properties. Theac data consist of a real and an imaginary compo-nent. The real part can be used to determine Tc andto estimate the percentage of superconducting sam-ple. There have been questions about the interpre-tation of the imaginary part. Goldfarb et al. [54]have provided evidence that the imaginary part isalmost totally due to hysteresis losses and haveshown how the temperature at which the slope ofthe imaginary component becomes positive uponwarming can be used to estimate Hc 1. To observe asharp magnetic transition and complete bulk dia-magnetism, the applied measuring field must bevery small. As shown in figure 10, two distinctsuperconducting components in a single-phasespecimen were identified [55]-one a relativelyhigh Tc, Hcl superconductor and the other a rela-tively low Tc, HCl superconductor (see fig. 11).These two components were found in all sinteredhigh-temperature superconductors that were exam-ined. The results of subsequent experiments on sin-tered and powdered samples suggested that thefirst component was intrinsic to the material, whilethe second arose from inter-granular coupling [56].

85Temperature (K)

90

0.06

.U

0.03 7=

00

0.00 z

CO

-0.03 z

- .-0.06

95

Figure 10. ac susceptibility vs temperature for YBCO. In the imaginary part, two peaks are apparent. Notethat the susceptibility is almost independent of frequency [55].

157

Y1 Ba2Cu 3 08-6Oxygen Annealed 5X

700 0C, 40 h ° 5

,d ¢t 91.3 K

86.8 K

0510~~~~~~~~~~~~~~~~~~~~~~~1

x = 10 Hz X

0= 100 Hz+ =1000 Hz X :

_X

Field Amplitude

225 A/m (2.8 Oe) , P

~ p I I



1.2

C)C-)

10

._,G

.Z

.z

5U

To4

1.0

0.8

0.6

0.4

0.2

0.020 40 60 80

15

Q)

10 '0

a)0.

5 "s

Q)

*j 0100

Temperature (K)

Figure 11. Lower critical fields vs temperature for the two components observed in figure 8. The grefers to the higher T. component [55].

Since the superconducting properties of the newsuperconductors are strongly dependent on mi-crostructure and composition, techniques availablefor elemental and moleculaf-microanalysis, princi-pally, electron-probe compositional mapping andmicro-Raman spectroscopy were employed toinvestigate a variety of samples in the YBCOsystem. Electron-probe compositional mapping iscomputer-aided x-ray microanalysis furnishingspatially-resolved digital images in which the dis-played grey scale is related to the true compositionof the specimen and not merely to x-ray intensity ofany given element. In studying YBCO, three wave-length dispersive spectrometers were employed;one each for the detection of yttrium, barium, andcopper. A representative result is shown in figure12. Compositional mapping is most useful, there-fore, in the identification of compositional hetero-geneities on the micrometer-scale, and thedetermination of dissimilar phases in a high-temperature superconductor [57-59].

Raman and infrared spectroscopy are widelyused tools for investigating and characterizinghigh-Tc superconductors. The Raman spectra ex-hibit vibrational modes mostly related to Cu-Obonds and to vibrations of other atoms in the lat-tice. The spectra are sensitive to differences incrystal structure, bonding, and phase relationshipand, furthermore, show a variation with oxygencontent, thus providing information on oxygen sto-ichiometry. Micro-Raman spectroscopy extendsthese capabilities into the microscopic domain witha spatial resolution comparable to that of electron-probe microanalysis. Preliminary work [60] hasshown that this technique can provide molecularinformation not revealed by macro or averagestructure methods. A micro-Raman spectrum ofthe YBCO superconducting ceramic in the or-thorhombic phase is shown in figure 13. Any varia-tions in the frequency positions and relativeintensities of the bands observed in these spectraare indicative of compositional and structural dif-ferences attesting to sample heterogeneities.

158



I 'mT7 -;0;A00 4 X0 0 -

4

B

Figure 12. Electron-microprobe compositional maps for Y, Ba, and Cu and corresponding SEMimage of YBCO sample. Region A shows a decrease in yttrium concentration, but no copper orbarium enhancement; B shows an yttrium-poor region corresponding to a barium-rich but un-changed copper region; C shows an enhancement of barium with no changes in copper oryttrium concentrations [58].

Lattice defects in YBCO can be identified byfield-ion microscopy (FIM). This technique per-mits the qualitative determination of surface andbulk atomic configurations and microstructural fea-tures. Atomic striations observed in FIM imagesare possibly due to preferentially conducting layersin the material. Thus the superconductivity is pos-sibly localized to specific layers, which are tenta-tively identified as the Cu-O planes of theorthorhombic unit cell. FIM identified various lat-tice defects such as dislocations and grainboundaries in the superconductors YbBa2Cu3 O,7,SmBa 2Cu 30 7 , GdBa 2(Cu. 96Fe.04)307_x, andGdBa2(Cu92Fe.s)O,_X (0<X<.5), in addition toYBCO [61-64].

A new technique has been developed to observethe superconducting transitions-magnetic-field-modulated microwave absorption (MAMMA) de-tection [51, 65]-which differs from conventionalmicrowave techniques in that it observes only mag-netic-field induced changes in the sample's mi-crowave loss as a function of temperature. Thetechnique is accomplished in a conventional ESRspectrometer by applying a small ac magnetic fieldto the sample and phase detecting the microwavepower reflected from the cavity at the ac modula-tion frequency. It has the advantages of ease of im-plementation using commercial ESR apparatus:high sensitivity due to noise reduction by narrow-band amplification and phase-sensitive detection,

159

a �51,- W , � Z

�, ANZ I

B IW I



800 600 400RAMAN SHIFT (Acm-,)

Figure 13. Micro-Raman spectrum of an arbitrary isolated particle of the superconduc-tor YBa2 Cu30 7_ (X WO) identified to be in the orthorhombic phase. The spectrum isexcited with the 514.5 nm line of an argon/krypton ion laser at low irradiance, employ-ing 5 mW in a - 12 ,pm beam spot. The microparticle is supported by a lithium fluoridesubstrate. The resolution is 7 cm-' [60].

and selectivity since only changes in sensitivitywhich are magnetic field dependent will be ob-served. The latter is characteristic of a supercon-ductive transition, as illustrated in figure 14a. ThisMAMMA technique has been used to study bulkand film specimens of lanthanum, yttrium, and bis-muth high-temperature superconductors (fig. 14b).These films were prepared from oxide targets bylaser ablation [51].

The composition and microstructure of YBCOwas studied as a function of processing [45]. Stron-tium was found to be the major contaminant. Thestarting compositions were barium rich relative tothe Y:B:Cu ratio, which remained constant duringprocessing. Electron-probe microanalysis revealedthree types of inhomogeneities that are within re-gions which correspond to the YBCO composi-tion-(i) Ba-rich; Y, Cu-poor, (ii) Y-rich, Ba-poor,and (iii) Cu-rich with lesser amounts of Ba and Y.These phases have been identified as (i) BaCu2 04 ,(ii) BaY2 CuO3 , and (iii) the remnants of a liquidphase that is present at the sintering temperature.The liquid phase limits Jc since the intergranularphases are not superconducting. Another source ofinsulating grain boundary film is carbon, which

may arise from atmospheric CO2 and from solventsused during grinding. During low-temperature oxi-dation of the sintered material, residual carbon mayreact with oxygen to form gas-filled pores alongthe grain boundary and a high concentration of de-fects adjacent to the grain boundary. Hence, it isapparent that a large degree of compositionalcontrol is needed to control the properties ofYBCO.

Processing-property relationships for YBCOhave been studied as a function of annealing tem-perature and environment. It was found [57, 58, 66]that annealing at low temperature in oxygen is nec-essary to obtain the highest Tc, sharpest transition,and the largest superconducting fraction. Samplescontained a small amount of inhomogeneous sec-ond phase liquid, insufficient to prevent currentflow. Sintering was more rapid and to a higherdensity in air than in oxygen. Segregation occurredduring sintering and pores and micro-cracks wereobserved. The thermal expansion was very high fora ceramic-indicating that thermal shock may be aproblem for these materials. Fracture toughnesswas quite low and the material was susceptible tomoisture-enhanced cracking.

160

I-zwI-z07IL

0U)



toTEMPERATURE WIC)

90 100

TEMPERATURE (K)

apparatus is shown in figure 15. The c axis tendedto be aligned parallel to the applied stress direction.The transport Jc was less than 100 A/cm 2 , how-ever, and in some cases even zero. This was due toweak linking. Transmission electron microscopyshowed second phases at grain boundaries, formingS-N-S junctions [67].

i5

Figure 14. a) Microwave (MAMMA) signal vs temperature fora bulk sample of niobium. The T. obtained was in good agree-ment with known values of T. for niobium. b) Microwave(MAMMA) signal vs temperature for bulk YBCO (above) andthin film (below) made by laser ablation from the bulk. Thevalue of T 0=95 K is in good agreement with resistivity andMeissner data. The double peak for the thin film is indicative oftwo phases with slightly different T.'s [51].

Sintering at 950 'C gave the best density butpoorest superconductivity due to a lack of porosityrequired for oxygen diffusion. Sinter-forging wasinvestigated [66] reasoning that it should be possi-ble to increase the current density by aligning thegrains. The grains had a high density center withthe edges cracked and not very dense. Large yt-trium-rich bands were formed perpendicular to thestress direction as a result of local segregation. The

Silicon Carbide Ram

- Alumina Platen

Sample

Figure 15. Sinter-forging apparatus. The load was applied in thevertical direction with no die wall constraints. The sample wasseparated from the ram by alumina plates [67].

Another possible method to align the grainswould be to cast the samples in a magnetic field.The alignment is due to anisotropy of the paramag-netic susceptibility of the grains. Ostertag et al. [68]studied the magnetic casting of YBCO and HBCO(H=holmium). A slurry of the superconductingpowder and isopropanol was placed in a homoge-neous magnetic field of 2 T for 30 minutes (see fig.16). The samples, which were then pressed and sin-tered, tended to align with their c axes parallel tothe applied field ([001] alignment). However, thisalignment is not sufficient for high Jc. Alignment inthe [010] and [100] or [010] is also needed since tiltdecreases Jc. Clean grain boundaries are also a re-quirement. Current densities of samples aligned inthe oxygen-rich state were up to five times greaterthan samples aligned in an oxygen-deficient stateand then oxygenated, due to the presence of non-superconducting junctions. Bulk Jc calculated frommagnetic measurements were 103-104 A/cm2 .

7. Electrical Contacts

One problem that existed in the study of high-temperature superconductors was too high a resis-tivity in the electrical contacts. Contacts made ofindium solder, silver paint or epoxy, direct wire

161

(a)

z

a.:

kIn

tnC

>2

0.-<.a

S

(b)

zo -

v. -

,-,

c:0

80I



MntFilter = Vacuum

Pump

Mge Man

Figure 16. Configuration for casting in a magnetic field. Theslurry of isoproponal and powder was placed in a homogeneous2 T magnetic field [68].

bonds, and pressure contacts have contact surfaceresistivities in the range of 10-2 to 10 fl-cm2 whichis several orders of magnitude too high for mea-surement and applications. Contact resistivities oflo- 4 to 1o-' fl-cm2 or lower are needed. Ekin andcoworkers [69-71] developed a method consistingof sputter etching the surface of the superconduc-tor to remove the degraded surface area immedi-ately before depositing noble metal (Ag or Au)pads, followed by annealing the noble metal/super-conductor interface in oxygen. Contact resistancefor the silver pads showed metallic behavior, de-creasing by a factor of 3 to 12 as the temperaturedecreased from 295 to 76 K. Contact surface resis-tivities less than 10 pLfl cm2 at 76 K were achievedwithout oxygen annealing. After annealing in oxy-gen at 500 'C for 1 hour, contact resistivities werereduced to as low as 0.1 nil cm2 [70]. The low oxy-gen affinity of the noble metals may play an impor-tant role in passivating the contact interface. Onthe other hand, oxygen and indium formed a semi-conducting oxide with resistivities greater than thatof pure indium [70, 72]. Room-temperature diffu-sion of oxygen is limited in the noble metals,

REACTIONFORCE

GLUE JCINTa

thus protecting the YBCO. This may explain whythere can be low contact resistance despite expo-sure of the YBCO to air.

Moreland and Goodrich [73] have developed sil-ver screen contacts for rapid characterization ofYBCO. The screens can be used for makingvoltage contacts and voltage taps. Silver wirescreens are interleaved between calcined powdersections and fixed to form a composite pellet. Sil-ver diffuses in the powder during sintering to formproximity contacts permeable to oxygen.

8. Electronic Structure

One method of obtaining information on theelectronic structure of superconductors is by tun-neling measurements. A technique used for suchmeasurements was developed by Moreland andEkin [74]-the break-junction technique (fig. 17).In a break junction, tunneling occurs across thefracture of a bullk sample. A small piece of a bulkmaterial is mechanically fractured under liquid he-lium and the freshly fractured surfaces are adjustedto form a tunneling barrier with helium as the insu-lator. The sample can be a single crystal, polycrys-tal, or sintered pellet. Unlike other tunnelingtechniques, break junctions give information on theinterior of bulk samples. Break junctions have beenused to study both the lanthanum and yttrium su-perconductors. Tunneling junctions for LSCO(fig. 18) exhibited a variety of tunneling behavior[75-77]. Scanning-electron microscopy showed arough surface with numerous voids and scattered

"FREE' COVER BENDING GISLIP FOR BEAM

CONCENTRATINGSTRAIN

Figure 17. Fabricating a break junction. A superconducting filament is mounted on abeam which is bent using an electromagnetic force. Once the filament is fractured, thebeam is relaxed to form a tunneling contact within the fracture of the filament. Contactmay be either through a thin insulating medium (vacuum, gas or liquid) or by closingthe fracture to form a point contact. An electromagnetic assembly affords precisecontrol of the tunneling gap [84].

162



inclusions. This variability may be due to tunnelingbetween different phases in the material. Large en-ergy gaps and deep structure in the conductancederivatives are evidence for a strong couplingmechanism.

C

C

-60 -,0 -20 0VOLTAGE

20 'd 60(mV)

Figure 18. Voltage vs current curves for a LSCO electron tun-neling break junction immersed in liquid helium at 4 K for threedifferent barrier settings. The bottom curve was the most com-mon V-1 characteristic found 175].

Break junctions for YBCO gave results indica-tive of strongly coupled superconductors [78] buthad the same variability as LSCO. Variable resultsin perovskites can be explained as being due to thestructure which consists of alternating layers of in-sulating and conducting platelets which can be su-perconducting, semiconducting, or both [77].While evidence for the usual pairing state associ-ated with the BCS theory was found, so was alower IcR product which is indicative of a lowerenergy gap than that expected from BCS theory. Inaddition, a Josephson junction effect was found[79]. Evidence of an intrinsic energy gap was foundin both LSCO and YBCO [80]. The gap scales withTc and decreases and vanishes when approachingTc from a lower temperature. This points to theenergy gap being quasiparticle in nature.

Break junctions in single crystals should permit adirect measurement of gap anisotropy if the sam-ples are fractured along cleavage planes. To thisend, break junctions of single crystal HoBa2Cu3Ob(HBCO) were compared with polycrystallineYBCO [81]. Both had junction conductance in-creasing linearly with junction bias. Gap structureof YBCO occurred more often during adjustmentof the junctions than with HBCO. This may havebeen due to a lack of oxygen penetration in thesingle crystal. The results may have been affectedby the fact that the HBCO fracture surfaces werenot ideal. The V-I curves showed the square-lawdependence of current seen in many tunneling mea-surements of polycrystalline YBCO.

Based on the fact that the anomalies in the breakjunction results may be microstructural in originand not due to the electron coupling mechanisms,several models have been proposed. In the granularmodel [82], the superconductor is divided intograins isolated from each other by insulating tun-neling junctions. A second model, the multiparticlemodel [83], assumes that the grains are oriented toform a series array of junctions near a primary tun-neling contact. Moreland et al. have developed athird model to explain these results in perovskites[84]. In the laminar model, the microstructure con-sists of a complex tunneling matrix with parallelsuperconducting laminae connected to each other,the point contact, and the surrounding grains bytunneling junctions. This structure may be mani-fested in a layered perovskite single grain with su-perconducting layers separated by high dielectricinsulating barriers. The individual laminae form aseries-parallel network of superconducting junc-tions within a single grain of the material. Al-though there is some evidence that casts doubtupon the granular model, the exact model is still inquestion.

Tunneling measurements were also made onYBCO thin films [85] using the method of squeez-able electron tunneling (SET) junctions developedby Moreland et al. [86]. In contrast to the breakjunction measurements of bulk samples where thespectra are often without energy gap features, SETspectra invariably contain such features. This im-plies that the film is superconducting near the sur-face, in contrast to results on bulk materials whichindicate that only parts of the interior are super-conducting. Improvement of surfaces by the addi-tion of very thin noble metal films, which becomesuperconducting by the proximity effect, is underinvestigation [87].

163

, I I I I ' I ' a ' I 'Ti ] I I I

no ~~~~~~~~~~~~~~I

I I

7

1iOj …--------'

2

L.-Sr -CU-0ere.k Jfl~CtCof

T :01K

. I . I . I . I I I I I I I I



Measurement of the electronic structures of thehigh-temperature superconductors are important inproviding supporting evidence for theoretical mod-els of superconductivity. Kurtz [88] has written areview of the experimental measurements of thevalence electronic structure of LSCO and YBCO.NIST's Synchrotron Ultraviolet Radiation Facility(SURF-II) was used to study these features. Theelectron structure of YBCO was measured usingresonant photoemission, which is associated withthe enhancement of valence photoelectron featuresresulting from the coupling of excitation and decaymechanisms at the core-electron photoabsorptiononsets. Radiation in the 60-160 eV range was used[89], (fig. 19). The upper edge of the valence bandwas found to nearly coincide with the Fermi leveland the density of states was small. There was nodistinctive edge. The valence band did not resonatewith the photon energy. Furthermore, there wasno evidence of valence band structural changes asthe temperature was lowered below the criticaltemperature. The copper oxide in the material wasfound to give spectra similar to CuO.

25 20 15 10Binding Energy (eV)

Figure 19. Ultraviolet photoemission spectra of fracturedYBCO. The top curve is at h= 60 eV, the bottom 106 eV, andeach curve is separated by 2 eV. At 60 eV there are two valenceband features-at binding energies of 5 and 9.4 eV. Increasingthe photon cncrgy, features become apparent at 12.4, 15.0, and28.8 eV. The features at 9.4 eV are due to Y/Cu; at 12.4 eV toCu, and at 15.0 and 28.8 eV, Ba [89].

Another study using SURF-II but at an energyrange of 20-600 eV [90], confirmed the 2+ valencyof copper in YBCO. The National SynchrotronLight Source at Brookhaven was used to provideinformation on oxygen, barium, and yttrium. It wasfound that the p-type partial density of states isvery small at the Fermi energy. The electronicstructure observed in the photoemission measure-ments is associated with the oxygen 2p orbitals.This study also observed no change in the spectraas the temperature was lowered below the criticaltemperature.

Additional studies carried out by Kurtz,Stockbauer, and coworkers included photoemis-sion of YBCO and LSCO, which revealed a reso-nance in the peak located at a binding energy of-9.5 eV for photon energies spanning the onset ofO-2s excitations. This feature is associated withoxygen excitations. The satellite is suppressed onsurfaces that are superconducting within the probedepth of the spectroscopy [91]. Photoelectronspectroscopy of high-Tc superconductors, includ-ing the newer bismuth and thallium superconduc-tors revealed that the materials have a highlyhybridized Cu-O valence band and resonant satel-lites which imply that the materials are highly con-nected. No substantial changes were observed inthe electronic structure as the materials werecooled from room temperature to below Tc. Thematerials reacted strongly with H2 0 and CO2,forming hydroxides and carbonates, but reactedmore weakly with 02 and CO [92, 93].

Photoemission measurements of YBCO revealedtwo constraints on any theoretical treatments of itselectronic structure based on the observation of a2.3 eV feature [94]. First, YBCO has a highercharge carrier concentration at the Fermi levelthan in related lower-Tc and non-superconductingcompounds. Secondly, there is a large contributionfrom oxygen to the density of states near the Fermilevel, mainly derived from oxygen in the Cu-Ochains. The 2.3 eV feature is intense in the or-thorhombic phase, but weak in the tetragonal.

9. Physical Properties

Current densities are a critical parameter for thesuccessful application of high-temperature super-conductors. A cryogenic bathysphere developedby Moreland et al. [95,96] for resistance measure-ments of high-Tc superconductors is shown infigure 20. This device thermally isolates an envi-ronmental chamber from surrounding cryogenic

164



fluids. The bathysphere has the advantages of (i)being compact enough to fit in the base of a high-field superconducting solenoid without the use of are-entrant dewar; (ii) the sample remaining dry; (iii)being inexpensive; (iv) having no moving parts;and (v) having sufficient thermal contact betweensample and thermometer provided by the ambientpressure exchange gas to maintain thermal equi-librium within +0.1 K while the temperaturechanges as fast as 3 K/min. It may also be possibleto adapt this device to susceptibility, critical cur-rent, and electron tunneling measurements. Thebathysphere has been successfully tested with NbTiin liquid helium and YBCO in liquid nitrogen.

As previously mentioned, current densities of theorder of one million A/cm2 will be required formost applications. While films with these currentdensities have been produced, bulk materials havehad much lower current densities. Ekin et al. [97,98] studied bulk sintered YBCO samples from sev-eral different laboratories. Using V-I characteris-tics, they found that while a field of over 30 T wasneeded to suppress all superconductivity, a field ofonly a few tesla could suppress the transport cur-rent (fig. 21). The measured transport current wassignificantly lower than that measured by magne-tization. The superconducting transition in poly-crystalline YBCO is very broad. This is consistentwith a model of a weak-link region between high-current-density grains. At least part of the behavioris due to intrinsic conduction anisotropy. This an-isotropy has been observed in YBCO single crys-tals with the weakest conduction along the c axis[99]. The low current density could be due to: (i)impurities or low-Tc phases at the grain boundariesor (ii) misalignment of the grains. Electron mi-croscopy gives no evidence of the former [100].The location of the weak links in YBCO could beat the grain boundaries, within the grains or be-tween the Cu-O planes. Transport critical currentdensities have been measured at low magneticfields in several kinds of high-Tc superconductorsfabricated in many different laboratories, and fittedwith a model which assumes that the barriers tocurrent flow are Josephson weak links which havea statistical distribution of sizes and orientations[101-103]. The data were shown to follow the Airycurrent-field pattern. The fits of the data to theoryare good for all the samples. The fitting parameteressentially gives the average dimension of the junc-tions, which in all instances is about equal to thegrain size, thus furnishing convincing evidence that

the barriers at low magnetic fields are at the grainboundaries. This finding indicates that a possiblemethod for increasing the current density would beby processing in such a manner that the grainswould be aligned.

S upuport Tube

. < .Vacuum CanSample

Thermometer

> Radiation Shield

C 1 ~ Circuit Board

Thermal Anchor

S- Heater Coil

f l vTest Leads

a U | / End Plug

D 1 Stainless

ZL SCopper

-- F E Fiberglass-Y§§ Epoxy

fr-41 cmFigure 20. Cryogenic bathysphere for resistance measurementsof high-T, superconductors [95, 96].

165



0.8

0.7

0.6

E

0)cU

0.5

0.4

0.3

0.2

0.1

0

Current, A

Figure 21. Voltage vs current characteristics for a YBCO sample in transverse magnetic fields in liquidnitrogen at 77 K. High magnetic fields were required to increase the slope to the normal resistance valueat Tc, and the transport critical current is suppressed by very low fields. The curves are nearly linear atcurrents well above the critical currents indicated by the arrows [100].

Other physical properties of interest include theelastic constants, which have practical significance,such as in stress and fracture toughness, and arerelated to physical properties such as specific heatand hardness. Elastic constants relate strongly tointeratomic potentials and force constants and canbe used to calculate the Debye temperature whichis used in the BCS calculation of the critical tem-perature. They also relate strongly to any phonon-mediated superconductivity mechanism. Values ofelastic constants can be determined by ultrasonicmethods and one of them, the bulk modulus, byx-ray diffraction. The elastic properties of metal-oxide superconductors have been reviewed byLedbetter [104].

Ledbetter and coworkers have measured theelastic constants of YBCO using ultrasonic tech-niques. YBCO was compared to BaTiO3 [105] andwas found to have a lower elastic stiffness whichcould arise from oxygen vacancies or microcracks.The latter have a larger effect than a comparablefraction of spherical voids. Study of six YBCOspecimens [106] showed that some specimens maybe free from softening defects and their propertiesmay reflect intrinsic behavior. A check on this is tocompare elastic and thermal Debye characteristictemperatures. Elastic constants were measured asthe specimens were cooled through the transition

temperature [107-110]. Samples run in helium gavereproducible results suggesting that these measure-ments represented intrinsic material properties.Elastic constants showed irregularities above andbelow, but not at, the critical temperature. Theshear-modulus results (fig. 22) departed from thoseexpected for a simple second-order normal/super-conducting transition, in agreement with the resultsfor the dilation [111]. The value of the Poisson ra-tio behaved irregularly below the transition tem-perature indicating a change in interatomic forcessupporting Geballe's view [87, 112] that a largefraction of electrons enter into Cooper pairs, thegap is approximately equal to the Fermi energy,and coupling is strong. During cooling from 160-70 K, YBCO behaved as if it underwent a sluggishphase transition. Two YBCO materials with differ-ent oxygen contents, x = 6.70 and 6.92, showedsimilar ambient-temperature elastic-constant val-ues, and similar temperature behavior [113], but thex = 6.92 YBCO demonstrated a higher elastic stiff-ening during cooling to 4 K.

The behavior of the elastic constants can be de-scribed by a "reentrant softening" model [114,115]. Softening occurs just above the critical tem-perature suggesting growing lattice instability withdecreasing temperature. Premonitory behavior ofthis type is known to be associated with martensitic

166



or displacive structural transformations in variousmaterials including the Al5 superconductors [116].The increased stiffness below the transition temper-ature is the result of the softening being offset bythe increased stiffness associated with the develop-ing superconducting phase. The calculation of theDebye temperature based on this model is in agree-ment with other experimental measurements. Themodel also predicts that the elastic constants willhave a higher value in the normal state than in thesuperconducting state. The results of measure-ments on LSCO are also in agreement with thismodel. Strong thermal hysteresis, especially in thedilational modes, were found in subsequent studiesby Ledbetter and Kim [117].

1.100

trN-r-CMI IO!

1.075

1.0501

1025

0 50 100 150

T (K)

YBCO was found to be different on warming thanon cooling, with the greatest difference occurringin the first cycle. Three attenuation peaks werefound on warming: I at 65-75 K, II at 134 K andIII at 183 K. The hysteretic velocity changes andpeak I appear related to a first-order phase transi-tion involving magnetic superstructure in non-superconducting portions of the sample. Peak IIIappears to be consistent with a defect relaxationprocess. The origin of peak II, which was depen-dent on thermal history, could not be identified.

175

4r-)

0<170 -0)

>1 65-

1600

200 250 300

Figure 22. Relative shear modulus G =pv,2 between 275 and4 K for YBCO. Above T. (65 K for this material), behavior isnormal. Below T., contrary to expectation, G apparently in-creases. However, a reentrant-softening model reconciles thisapparent anomaly [107].

The bulk modulus of YBCO was determined bymeasurements in a diamond-anvil cell using an en-ergy dispersive x-ray diffraction technique [1181.The least compression was observed within theperovskite layers because of the oxygen packing,and the largest was observed perpendicular tothese layers. As seen in figure 23, the decrease involume was essentially linear with applied pres-sure. The value of the bulk modulus was largerthan that determined by ultrasonic techniques.Ledbetter and Lei [119] focused on this differenceand its implication for the related Grilneisenparameter, supporting their measurements byionic-bonding calculations.

Ultrasonics can also be used to provide addi-tional information [120]. The ultrasonic velocity in

2 4 6 8 10

Pressure, GPa12

Figure 23. The pressure dependence of the volume of YBCO.The lattice parameters are determined from x-ray diffractiondata. These data are used to calculate the isothermal bulk mod-ulus [1 18].

Neutron inelastic scattering was used to measurethe phonon density of states in an attempt to ascer-tain if any significant perturbations occurred in thephonon modes at the superconducting transitiontemperature [21]. A "softening" of such modes is akey aspect of conventional phonon-driven super-conductivity. The normalized density of states at120 K for YBa2 Cu3 07 is shown in figure 24. Thespectra consist of a strong double peak near 20meV and a second principal maximum at approxi-mately 70 meV. Measurements below the transitiontemperature gave only negligible changes in theobserved spectra which could be accounted for byanharmomic effects, and did not indicate any majorchanges in the overall phonon modes accom-panying the superconducting transition. Oxygen-deficient YBa2 Cu30 7 showed a pronouncedweakening of the 70 meV features in the density ofstates and a filling in and broadening of the lower

167

I L

Y' Ba2 Cu3 07X

I I I I I

Ba2Cu 3Y07

V = 172.2-0.83 P-

I



b

0

cis

0C o;E'S

Io

45 60E(me V)

105

Figure 24. Vibrational density of states as measured with inelastic neutron scattering at 120 K. Thelargest spectral weight is contained in peaks involving oxygen vibrations [21].

energy features, reflecting a change in the har-monic modes associated with the absence of oxy-gen in the Cu-O "chain" structure. These featuresare equivalent to modes observed by Raman scat-tering [60].

The magnetic hysteresis loops of YBCO werealso studied [121]. The shape of the loops well be-low Tc (fig. 25) brought to mind the constrictedhysteresis loops observed in certain ferromagneticmaterials which are usually associated with mag-netic aftereffects. Similar dynamic effects with timeconstants on the order of 10 seconds at 40 K werefound to be present in YBCO. This is in addition toflux creep (due to thermally activated jumpingover flux-pinning sites) observed for longer timeperiods. When the measurement time is fast com-pared to both time constants, the hysteresis loopscan be approximated by a critical-state (i.e., Bean-Kim) model [122-126]. The experimental hysteresisloops at higher temperatures are more pinchedthan the critical-state model because of the move-ment of fluxoids.

The critical-state model, which provides amethod for calculating the energy losses in type IIsuperconductors, has been extended by Peterson[127] to include the train of magnetization jumpsoften seen at low temperatures in moderate-to-highmagnetic fields. Chen and Goldfarb [128] have de-veloped an analytic method for using the critical-

state model to determine critical currents frommagnetization measurements on the sample shapesmost often encountered in developmental studies.

10. Theory

The interaction between two test charges in asolid can be described in terms of a total dielectricfunction that includes electronic and lattice polar-ization. Stability requirements place restrictions onthe dielectric function. Allen et al. [129] show thatthe eigenvalues of the inverse dielectric matrix, Xi,satisfy Xi<l. As a result, the electron-electron in-teraction (as determined by test charges) which en-ters BCS theory is not restricted to positive valuesby general stability requirements. Casella [130]considered other intermediate bosons, besidesphonons, mediating the superconducting interac-tion and carried out a semiphenomenological anal-ysis of the effects of certain band-gap features onthe gap ratios of high-temperature superconduc-tors. Comparison with experiment suggests that theintermediate boson is not a phonon.

Melamud et al. [131] studied the near-neighborenvironments and the bonding of atoms in lan-thanum and yttrium based copper-oxide super-conductors using Wigner-Seitz cell construction.Wigner-Seitz cells can identify the nearest neigh-

168



300

0_

0-p

NO0)

b-0

-300

Field. kOeFigure 25. Experimental (solid line) and calculated (dotted line) hysteresis loops for YBCO at 38 K. Theinsert is an expanded view of the virgin curve [121].

bors, the site symmetry arising from the presenceof these neighbors, and the number of nearestneighbors common to a near-neighbor pair. Differ-ent results were obtained depending on whetherionic or covalent/metallic bonding is assumed. Co-valent/metallic bonding gave more reasonablechemical results and was consistent with knownproperties of these materials. The barium, lan-thanum, and yttrium atoms all had large coordina-tion numbers (see fig. 26) implying a three-dimensional chemical bonding scheme. The resultsare in agreement with the conclusion of Pauling[132] that the bonding at the important copper sitesis not limited to oxygen but involves substantial in-teractions with large atoms such as lanthanum andbarium.

11. Applications

Problems with current density and fabricationhave hindered many applications of high-tempera-ture superconductors. However, a successful pro-totype transition edge bolometer and a SQUID(superconducting quantum interference device)made from YBCO have been developed [133, 134].

The breaking fixture used to form the Josephsoncontact for the SQUID is shown in figure 27. Vari-ations in performance were found with differentYBCO batches, and the first devices constructedshowed considerable noise above 61 K, althoughquantum interference effects persisted up to 81 K.However, SQUIDs made from well-characterized,high-quality YBCO, operated in liquid nitrogen,with only a modest increase in noise over thatfound at 4 K. This provided the first demonstrationthat sensitive high-TC SQUIDs operating at liquidnitrogen temperature are possible.

Many applications for high-temperature super-conductivity depend on understanding and im-proving the critical current. To this end, a YBCOmacrobridge (bridge dimensions are much greaterthan the coherence length) was fabricated [135] tounderstand not only Jc, but intra-film Josephsoneffects. Extremely noisy sections of the V-I curvewere observed, always well below Tc. This behav-ior could have ramifications for potential low-noiseapplications of high-Tc superconductors. The noisedepends on temperature, bias current, and the mag-netic field. A very rapid change of switching ratewith very small fields and small changes in biascurrent was observed, which suggests that the

169



noise may be due to the motion of vortices in andout of pinning sites.

The ability of a superconductor to levitate amagnet above its surface is well known, and forhigh-TC superconductors it is often demonstrated.Recently, it has been realized [136, 137] that spe-cially processed samples of a high-Tc superconduc-tor can be levitated below a magnet. This unusualtype of levitation involves "attraction" of the su-perconductor by a magnet rather than the Meissnereffect "repulsion" seen for a levitated magnet. Animportant application for this effect would be inmagnetic bearings (see table 1).

Figure 26. The (O 12 0 8 0 6) Wigner-Seitz polyhedron (coordi-nation number=26) of the Ba, La or Y atom in the ideal ABO3perovskite structure, obtained with the use of metallic radii. Inthe high-temperature superconductors with distorted perovskitestructure, the Wigner-Seitz cells for these sites are derivationsfrom the ideal polyhedron [131].

spring

Al holder

YBCO / lpellet I I

68 mm I

taper pin

saw cut inpellet

holedia.

Figure 27. A SQUID made from a YBCO break junction. AYBCO pellet was secured with epoxy in an aluminum breakingfixture. A hole was drilled through the pellet and a saw cutmade part way through so that the pellet would break along adiameter when the two arms of the fixture were spread apart bya thin tapered pin. Springs close the break as the taper is with-drawn [1331.

12. Other High-temperatureSuperconductors

Following the discovery of high-temperature su-perconductivity in Bi-Sr-Ca-Cu-O ceramics [138],Bi2Sr2CaCu2O was synthesized both chemicallyand by a solid state reaction [72]. ac susceptibilitymeasurements showed transitions at 80 K and110 K and a low Hci. The appearance and amountof the 110 K superconductor was sensitive to theannealing procedure. Magnetic hysteresis loopsconstructed at 80 K were narrow, signifying asmall amount of trapped flux. The loops were con-stricted in the center, indicating the probable exis-tence of time effects similar to those seen in YBCO[121]. The bismuth superconductor was also stud-ied by the magnetic-field-modulated-microwave-absorption (MAMMA) technique [65, 139]. Super-conducting transitions were observed at 72, 100,and 110 K. An applied magnetic field broadenedthe microwave response peak much more than inthe case of YBCO. Thin films of the bismuth super-conductor were made [140] by laser ablation onZrO2 and characterized by MAMMA. The filmquality was affected by substrate temperature andan annealing process. Unlike previous work [141],the films were not superconducting as deposited.

A classical test to determine the contribution ofan electron-phonon interaction to the superconduc-tivity is to measure the isotope shift [3, 4] in Tc.Substitution of 150 for 160 in the Bi-Sr-Ca-Cu-Osystem [142], the La-Sr-Cu-O system [143, 144],and the Y-Ba-Cu-O system [145, 146] has demon-strated a measurable, albeit small, isotope shift inTc. Although this small effect indicates that theelectron-phonon interaction contributes to the su-perconductivity, it is probably too small to accountfor the high values of Tc, and other mechanisms,

170



e.g., spin fluctuations must be operative. The possi-ble role of various magnetic interactions have re-cently been addressed at a workshop held at NIST,Gaithersburg [147].

Magnetic measurements were made [148, 149] onchemically synthesized Bi-Pb-Sr-Ca-Cu-O. Thelead substitution appears to encourage or stabilizethe high-Tc Bi phase. The superconductor dis-played extremely narrow hysteresis loops aboveliquid nitrogen temperatures, indicating a smallnumber of effective flux pinning sites. Below 40 K,a dimpling was observed but only when the samplewas a loosely packed powder. A flux depinningwas observed, as illustrated in figure 28, for twotemperatures. A plot of the flux-depinning field vstemperature appears to be linear (fig. 29).

0.4

0.2- -403

-O.I

S-0.3 ji

-0.4-I -0.8 -0.6 -0.4 -0.2 D 0.2 0.4 0. 6 0.8 I

Applied Field, kcOe

Figure 28. Hysteresis loops at two temperatures for a Bi-Pb-Sr-Ca-Cu-0 superconductor, illustrating flux depinning [148].

a.5

_ tHp = £ '100 (It-t) < ,

-- 7.5 -

- 0.1

-0.

5--2 -0.8 -1.6 -0.4 -0.2 -I .0.6

Ln( I - O

Figure 29. The frux-depinning field of the sample of figure 28 asa function of temperature [149].

Ultrasonic elastic-constant studies were carriedout for Bi-Pb-Sr-Ca-Cu-0 [150], with results simi-lar to YBCO. There was stiffening during cooling,

no measurable change at Tc, and hysteresis. How-ever, the Bi-Cu-O is much softer than YBCO, withan elastic Debye temperature of 312 K vs 437 K.

Resonant photoemission has been used to study[151] the electronic states and electron-electron in-teractions in a bulk sample of Tl-Ba-Ca-Cu-O. Theelectron structure is similar to that of YBCO indi-cating that the electron states and interactions aresimilar. The surface of the TI superconductor is notas reactive toward atmospheric gases as YBCO.

13. Low-Temperature Superconductors

Despite all the current interest in high-tempera-ture superconductors, low-temperature supercon-ductors will still be required in many applications.For example, the cost savings realized by switchingfrom helium to nitrogen for cooling large magnetsmay be only a small part of the total operating cost.Additionally, much experience has been gained inlearning how to fabricate these materials into prac-tical conductors-in the shape of tapes or wires-that can support high current densities under realis-tic operating conditions. Therefore research onthese materials is continuing, with the goal of opti-mizing their properties-e.g., current-carrying ca-pabilities stability, ac losses, etc.

NIST is developing facilities and standards forthe definition and measurement of superconductiv-ity parameters [152, 153]. The facilities developedfor this project enable critical currents up to 3000A to be measured in fields up to 12 T in the pres-ence of longitudinal or transverse stress. NIST isalso involved in round robins on critical currentmeasurements of Nb-Ti and Nb3 Sn with both do-mestic and foreign participants. Calibration tech-niques developed for the Nb-Ti study were used inthe Nb3 Sn study. It was found that a small changein the mounting technique could result in a 40%change in the critical current density at 12 T. Man-drel material and geometry were also a source oferror [154, 155].

The problem of current ripple on critical currentmeasurement was studied [156, 1571. Ripple (theperiodic departure from a dc output level) reducesthe measured dc critical current, Ic, and causesnoise at the input to the voltmeter used for mea-surements. A theoretical model of rippling was de-veloped which was in good agreement with theexperimental data and can be used to estimate theeffects of current ripple on the measured dc 4c. Itwas also found that the effect of ripple should scalewith its fraction of the Ic, and will depend upon theshape of the V-I curve.

171



At present, the material of choice in the wind-ings for magnets is Nb-Ti. The effect of stress oncurrent degradation has been studied by Ekin et al.Current degradation as a function of strand loca-tion and field angle on cable compacted into a key-stone shape was evaluated [1581. It was found thatcabling can lead to localized reductions in 4cwithin a single strand. The widest spread in local Icalong the cable strands was with the field perpen-dicular to the cable edge. Unfortunately, in thedipole magnet orientation, this orientation is nearthe critical orientation. The relevant Ic criteriamay be a spatial average (the strand Ic). Therefore,both magnetic-field orientations, perpendicular andparallel to the cable width, need to be tested for Ic[159]. In addition, a large difference in current car-rying capacity can exist between thick and thin ca-ble edges, and thus, changing the direction of thetest current can affect the measured Ic.

As illustrated in figure 30, the effects of varioustypes of stress on 'c at 4 K also were studied [160].It was found that Ic degradation from transversecompression was much less than from axial tensionin terms of overall conductor stress but comparablein terms of stress on NbTi filaments. More stresscan be developed in axial tension than in transversecompression because of the matrix. Ic is 95% re-versible for both stresses indicating that the effectsof stress will be seen only when the conductor isunder stress. The primary source of degradation isa stress-induced reversible decrease in HC2. It wasfound [161] that the effect on the critical current isindependent of the temperature at which the stressis applied. Existing data obtained at 4 K can there-fore be used to determine the degradation of Tcarising from room-temperature fabrication stress,cool-down stress, and 4 K stress due to the Lorentzforce when the magnet is energized. Couplinglosses in multifilamentary NbTi wire were studied[156] by vibrating sample magnetometry. Lossesfor wires with long twist lengths were up to twicethe hysteresis losses. Using short twist lengths re-duced these losses.

Non-uniformity of sample diameter (sausaging)of the filaments also degrades performance [152].Sausaging causes a change in E-I response resultingin a significant electric field below 'c leading toheating and decreased stability. In the relationshipE a In, the value of n is related to the degree ofsausaging with smaller values of n implying morenecking. Therefore, the value of n can be used toestimate filament regularity (fig. 31).

TRANSVERSE FORCE PER UNrr LENGTH, FILkN/m)

& OmI

-4

TRANSVERSE COMPRESSION, aj(MP)

Figure 30. Effect of transverse compression force on the criticalcurrent of a NbTi conductor [160].

1 0

0.8 I _

I / lC

Figure 31. Logarithmic plot of electric field vs current for Nb-Ti samples with different n [152].

Another low-temperature superconductor whichcan be used for magnet applications is Nb3 Sn. Astudy of the effect of transverse stress on Ic degra-dation showed that the intrinsic effect on the uppercritical field is about 10 times that of axial stress[162, 163]. This effect scales with conductor thick-ness and as a result places limits on conductor di-mensions and the spacing between distributedreinforcements in large magnets. This is importantin applications calling for larger conductors neededto limit inductance and keep induced quenchvoltages low in large magnet applications. Stressconcentration at strand crossover points can signif-icantly enhance the effects. This effect is re-versible, but not totally. Hysteresis losses were

172



measured [156] on a series of fine filament Nb3 Snsuperconductors made by the internal tin process.Hysteresis was measured as a function of filamentdiameter and interfilament separation. Losses weregreater than predicted. This was due to interfila-ment bridging across the wires. The critical inter-filament separation, for which the critical-statemodel would be accurate, was determined.

The cable matrix can also play a role in improv-ing performance. The addition of manganese to acopper matrix of fine filament Nb-Ti wire was in-vestigated by Goldfarb et al. [164]. Manganese ad-ditions had been shown to reduce proximity-effectcoupling between closely-spaced filaments [165,166]. The investigation found that as long as themanganese content was less than 4%, there wereno adverse effects.

NIST has developed a wide variety of applica-tions of superconductor electronics (which will bethe subject of a future review). The most successfuldevices that NIST researchers have produced arearray voltage standards (see fig. 32) containing asmany as 19,000 Josephson junctions [167]. Such in-tegrated circuits made at NIST using VLSI tech-niques are already in use in most national standardslaboratories around the world and in two U.S.companies. Other devices made at NIST are ultra-

high-speed analog-to-digital convertors, supercon-ductor-insulator-superconductor mixers for radioastronomy at frequencies up to 300 GHz, SQUIDswith sensitivities approaching the uncertainty prin-ciple limit, samplers with response times of lessthan 10 ps, counters with rates above 100 GHz andsensitivity to pulses of 10'" J, and an ultra-sensi-tive microwave and infrared detector based on thekinetic inductance of very thin superconductingfilms.

14. Conclusion

This review paper has attempted to show thebreadth of NIST's work in superconductivity. Ma-jor contributions to the materials science, standard-ization, and engineering applications of super-conductors are evident. To maintain a reasonablelength, many topics have not been covered in thedepth they deserve. Some of these, e.g., supercon-ductive electronics, will be the subjects of futurereview articles. With all the world-wide attentionon the new high-temperature superconductors andtheir potential economic impact, we can anticipatethat NIST personnel will continue to make newand important contributions to this exciting field.

Figure 32. The layout for a Josephson junction array voltage standard chip fabricatedusing a seven-level photolithographic process [167].

173



15. Acknowledgements

We thank R. Powell for his aid in providingtimely information on the NIST publications. R. A.Kamper, H. P. R. Frederikse, and K. Moorjani fur-nished extensive comments which have been incor-porated in the manuscript. We are grateful to manyof the referenced NIST authors for their aid in re-ducing our errors and omissions.

About the authors: All three authors are with theMagnetic Materials Group of the Metallurgy Divi-sion, Institute for Materials Science and Engineering,NIST, Gaithersburg. Donald R. Lundy is a GuestScientist. Lydon J. Swartzendruber is a ResearchMetallurgist. Lawrence H. Bennett is a Physicist andthe Group Leader.

16. References

* Papers authored or co-authored by NIST scientists.

*[I] Silsbee, F. B., J. Wash. Acad. Sci. 6, 597 (1916).[2] Meissner, W., and Oschenfeld, R., Naturwissenschaften

21, 787 (1933).*[3] Maxwell, E., Phys. Rev. 78, 477 (1950).[4] Reynolds, C. A., Serin, B., Wright, W. H., and Nesbitt,

L. B., Phys. Rev. 78, 487 (1950).[5] Josephson, B. D., Phys Lett. 1, 251 (1962); Rev. Mod.

Phys. 36, 216 (1964); Adv. Phys. 14, 419 (1965).[6] Bednorz, J. G., and Muller, K. A., Z. Phys. B-

Condensed Matter 64, 189 (1986); Miller, K. A., andBednorz, J. G., Science 237, 1133 (1987).

[7] Cohen, M. L., Phys. Rev. 134, A442 (1964).[8] Bardeen, J., Cooper, L. N., and Schrieffer, J. R., Phys.

Rev., 108, 1175 (1975).*[91 Kahn, A. H., and Leyendecker, A. J., Phys. Rev., 135,

A1321 (1964).*[10] Frederikse, H. P. R., Thurber, W. R., and Hosler, W. R.,

Phys. Rev. 134, A442 (1964).*[1I] Schooley, J. F., Hosler, W. R., and Cohen, M. L., Phys.

Rev. Lett. 12, 474 (1964).*[12] Ambler, E., Colwell, J. H., Hosler, W. R., and Schooley,

J. F., Phys. Rev. 148, 280 (1966).*[13] Frederikse, H. P. R., Schooley, J. F., Thurber, W. R.,

Pfeiffer, E., and Hosler, W. R., Phys. Rev. 16, 579(1966).

[14] Michel, C., Er-Rakho, L., and Raveau, B., Mater. Res.Bull. 20, 667 (1985).

[15] Tagaki, H., Uchida, S., Kitazawa, K., and Tanaka, T.,Japn. J. App]. Phys. 26, L123 (1987).

[16] Wu, M. K., Ashburn, J. R., Torng, C. J., Hor, P. H.,Meng, R. L., Gao, L., Huang, Z. J., Wang, Y. Q., andChu, C. W., Phys. Rev. Lett. 58, 908 (1987).

[17] Ginley, D. S., Kwak, J. F., Hellmer, R. P., Baughman,R. J., Venturini, E. L., and Morosin, B., Appl. Phys.Lett. 53, 406 (1988).

[18] Cava, R. J., Batlogg, B., vanDover, R. B., Murphy, D.W., Sunshine, S., Siegrist, T., Remeika, J. P., Reitman,

E. A., Zahurak, S., and Espinosa, G. P., Phys. Rev. Lett.58, 1676 (1987).

*[19] Beech, F., Miraglia, S., Santoro, A., and Roth, R. S.,Phys. Rev. B 35, 8778 (1987).

*[20] Santoro, A., Miraglia, S., and Beech, F., Sunshine, S. A.,Murphy, D. W., Schneemeyer, L. F., and Waszczak, J.W., Mater. Res. Bull. 22, 1007 (1987).

¶[21] Rhyne, J, J., Neumann, D. A., Gotaas, J. A., Beech, F.,Toth, L., Lawrence, S., Wolf, S., Osofsky, M., andGubser, D. U., Phys. Rev. B 36, 2294 (1987).

*[22] Miraglia, S., Beech, F., Santoro, A., Tran Qui, D.,Sunshine, S. A., and Murphy, D. W., Mater. Res. Bull.22, 1733 (1987).

*[23] Cava, R. J., Santoro, A., Johnson, D. W., Jr., andRhodes, W. W., Phys. Rev. B 35, 6716 (1987).

*[24] Xiao, G., Cieplak, M. Z., Musser, D., Gavrin, A., Streitz,F. H., Chien, C. L., Rhyne, J. J., and Gotaas, J. A.,Nature 332, 238 (1988).

*[25] Miceli, P. F., Tarascon, J. M., Greene, L. H., Barboux,P., Griroud, M., Neumann, D. A., Rhyne, J. J.,Schneemeyer, L. F., and Waszczak, J. V., Phys. Rev. B38, 9209 (1988).

*[26] Rubinstein, M., Harford, M. Z., Swartzendruber, L. J.,and Bennett, L. H., J. Phys., Suppl., in press (1988).

*[27] Swartzendruber, L. J., Bennett, L. H., Harford, M. Z.,and Rubinstein, M., J. Superconductivity 1, 219 (1988).

*[28] Roth, R. S., Davis, K. L., and Dennis, J. R., AdvancedCeramic Mater.: Ceramic Superconductors, Vol. 2, 303(1987).

[29] Frase, K. G., and Clarke, D. R., Adv. Cer. Mat. 8, (313)295 (1987).

[301 Wang, G., Hwu, S. J., Song, S. N., Ketteson, J. B.,Marks, L. D., Poeppelmeir, K. R., and Mason, T. O.,Adv. Cer. Mat 2 (3B), 313 (1987).

*[31] Wong-Ng, W., McMurdie, H. F., Paretzkin, B.,Hubbard, C. R., Dragoo, A. L., and Stewart, J. M.,Powder Diffraction 2, 106 (1987).

*132] Wong-Ng, W., McMurdie, H. F., Paretzkin, B., Zhang,Y., Davis, K. L., Hubbard, C. R., Dragoo, A. L., andStewart, J. M., Powder Diffraction 3, (1987).

*[33] Wong-Ng, W., Roth, R. S., Beech, F., and Davis, K. L.,submitted to Advances in X-Ray Analysis.

*[34] Wong-Ng, W., Kuchinski, M. A., McMurdie, H. F., andParetzkin, B., submitted to Powder Diffraction.

*[35] Wong-Ng, W., McMurdie, H. F., Paretzkin, B.,Kuchinski, M. A., and Dragoo, A. L., Powder Diffrac-tion, to be published.

*[36] Wong-Ng, W., Davis, K. L., and Roth, R. S., J. Amer.Ceramic Soc. 71, C64 (1987).

*[37] Roth, R. S., Rawn, C. J., Beech, F., Whitler, J. D., andAnderson, J. O., in Ceramic Superconductors II, Yan,M. F., Ed. (Am. Ceramic Soc., Westerville, OH, 1988)p. 13.

*[38] Roth, R. S., Rawn, C. J., Ritter, J. J., and Burton, B.,submitted to the Journal of the American Ceramic Soci-ety, Communications.

*[39] Roth, R. S., Rawn, C. J., Whitler, J. D., Chiang, C. K.,and Wong-Ng, W., submitted to the Journal of theAmerican Ceramic Society.

*[40] Wong-Ng, W., and Cook, L. P., Advanced CeramicMater.: Ceramic Superconductors 2, 624 (1987).

*[41] Wong-Ng, W., Cook, L. P., Chiang, C. K.,Swartzendruber, L. J., Bennett, L. H., Blendell, J., andMinor, D., J. Mater. Res. 3, 832 (1988).

174



*[42] Cook, L. P., Chiang, C. K., Wong-Ng, W., and Blendell,J., Advanced Ceramic Mater.: Ceramic Superconduc-tors 2, 656 (1987).

*[43] Chiang, C. K., Cook, L. P., Chang, S. S., Blendell, 1. E.,and Roth, R. S., Advanced Ceramic Mater.: CeramicSuperconductors 2, 530 (1987).

*[44] Chiang, C. K., Cook, L. P., and Chang, S. S., submittedto the Proceedings of the Materials Research SocietySymposium with the Japanese Materials ResearchSociety.

*[45] Blendell, J. E., Handwerker, C., Vaudin, M. D., andFuller, E. R., Jr., J. Crystal Growth 89, 93 (1988).

*[46] Wong-Ng, W., Cook, L. P., Chiang, C. K.,Swartzendruber, L. J., and Bennett, L. H., Adv. Cer.Materials, in press (1988).

*[47] Wong-Ng, W., Roth, R. S., Swartzendruber, L. J., Ben-nett, L. H., Chiang, C. K., Beech, F., and Hubbard, C.R., Advanced Ceramic Mater.: Ceramic Superconduc-tors 2, 565 (1987).

*[48] Zhang, Y., Wong-Ng, W., Morosin, B., Hubbard, C. R.,Stewart, J. M., and Freiman, S. W., Physica C 152, 130(1988).

*[49] Ostertag, C. P., Beech, F., and Fuller, E. R., Jr., in Ce-ramic Powder Science II, Messing, G. L., Fuller, E. R.,Jr. and Hausner, H., Eds. (Amer. Ceramic Soc., Wester-ville, OH, 1988) p.501.

*[50] Ritter, J. J., in Ceramic Powder Science II, Messing, G.L., Fuller, E. R., Jr. and Hausner, H., Eds. (Amer. Ce-ramic Soc., Westerville, OH, 1988) p.7 9 .

*[51] Moorjani, K., Bohandy, J., Adrian, F. J., Kim, B. F.,Shull, R. D., Chiang, C. K., Swartzendruber, L. J., andBennett, L. H., Phys. Rev. B 36, 4036 (1987).

*[521 Ono, R. H., Beall, J. A., Cromar, M. W., Mankiewich,P. M., and Howard, R. W., Proc. of 1988 Applied Super-conductivity Conf., to be published in IEEE Trans.Magnetics.

*[53] Lindstrom, R. M., Trans. Amer. Nuclear Soc. 56, 231(1988).

*[54] Goldfarb, Ri B., Clark, A. F., Panson, A. J., andBraginski, A. I., Materials Research Society, Vol. EA-11, 261 (1987).

*[55] Goldfarb, R. B., Clark, A. F., Braginski, A. I., andPanson, A. J., Cryogenics 27, 475 (1987).

*[56] Chen, D.-X., Goldfarb, R. B., Nogues, J., and Rao, K.V., J. Appl. Phys. 63, 980 (1988).

*[57] Blendell, J. E., Chiang, C. K., Cranmer, D. C., Freiman,S. W., Fuller, E. R., Jr., Drescher-Krasicka, E., Johnson,W. L., Ledbetter, H. M., Bennett, L. H., Swartzendru-ber, L. J., Marinenko, R. B., Myklebust, R. L., Bright,D. S., and Newbury, D. E., Advanced Ceramic Mater.:Ceramic Superconductors 2, 512 (1987).

*[58] Blendell, J. E., Chiang, C. K., Cranmer, D. C., Freiman,S. W., Fuller, E. R., Jr., Drescher-Krasicka, E., Johnson,W. L., Ledbetter, H. M., Bennett, L. H., Swartzendru-her, L. J., Marinenko, R. B., Myklebust, R. L., Bright,D. S., and Newbury, D. E., Chemistry of High Temper-ature Superconductors, Amer. Chem. Soc. Symp. 351,240 (1987).

*[59] Marinenko, R. B., Newbury, D. E., Bright, D. S.,Myklebust, R. L., and Blendell, J. E., Microbeam Analy-sis-1988, Newbury, D. E., Ed. (San Francisco Press,1988) p. 37.

*[601 Etz, E. S., Wong-Ng, W., Blendefl, J. E., and Chiang,C. K., in Microbeam Analysis 1988, Newbury, D. E.,Ed. (San Francisco Press, San Francisco, CA 1988)p. 187.

*[61] Melmed, A. J., Shull, R. D., Chiang, C. K., and Fowler,H. A., Science 239, 176 (1988).

*[62] Melmed, A. J., Shull, R. D., Chiang, C. K., and Fowler,H. A., Mater. Sci. Eng. 1tN, L27 (1988).

*[63] Melmed, A. J., Shull, R. D., Chiang, C. K., and Fowler,H. A., J. Appl. Phys. 63, 4232 (1988).

*[64] Melmed, A. J., Shull, R. D., and Chiang, C. K., J. Phys.,in press (1988).

*[65] Bohandy, J., Adrian, F. J., Kim, B. F., Moorjani, K.,Shull, R. D., Swartzendruber, L. J., Bennett, L. H., andWallace, J. S., J. Superconductivity 1, 191 (1988).

*[66] Blendell, J. E., and Stearns, L. C., in Ceramic PowderScience II, Messing, G. L., Fuller, E. R., Jr. andHausner, H., Eds. (Amer. Ceramic Soc., Westerville,OH, 1988) p. 1146.

*[67] StearnsL. C., Vaudin, M. D., Ostertag, C. P., Blendell,J. L., and Fuller, E. R., Jr., to be published in the Pro-ceedings of the Superconductor Symposium, 90th An-nual Meeting of the American Ceramic Society.

*[68] Ostertag, C. P., Shull, R. D., Vaudin, M. D., Blendell, J.E., Steams, L. C., and Fuller, E. R., Jr., in Ceramic Su-perconductors 11, Yan, M. F., Ed. (Amer. Ceramic Soc.,Westerville, OH, 1988) p. 332.

*[69] Ekin, J. W., Panson, A. J., Blankenship, B. A., Appl.Phys. Lett. 52, 331 (1988).

*[70] Ekin, J. W., Larson, T. M., Bergren, N. F., Nelson, A. J.,Swartzlander, A. B., Kazmerski, L. L., Panson, A. J.,and Blankenship, B. A., Appl. Phys. Lett. 52, 1819(1988).

*[71] Ekin, J. W., Panson, A. J., and Blankenship, B. A., inHigh Temperature Superconductors, Brodsky, M. B.,Tuller, H. L., Dynes, R. C., and Kitazawa, K., Eds. (Ma-terials Research Society Symposium Proceedings, Pitts-burgh, Pa., 1988) Vol. 99.

*[72] Wallace, J. S., Ritter, J. J., Fuller, E. R., Jr., Bennett,L. H., Shull, R. D., and Swartzendruber, L. J., Phys.Rev. B. in press.

*[73] Moreland, J., and Goodrich, L. F., Proceedings of theApplied Superconductivity Conference, IEEE Transac-tions on Magnetics, in press.

*[74] Moreland, J., and Ekin, W. J., J. Appl. Phys. 58, 3888(1985).

*[75] Moreland, J., Clark, A. F., Ku, H. C., and Shelton, R.N., Cryogenics 27, 227 (1987).

*176] Moreland, J., Clark, A. F., Goodrich, L. F., Ku, H. C.,and Shelton, R. N., Phys. Rev. B 35, 8711 (1987).

*[77] Moreland, J., Ekin, J. W., Goodrich, L. F., Capobianco,T. E., and Clark, A. F., Materials Research Society, Vol.EA-I1, 73 (1987).

*[78] Moreland, J., Ekin, J. W., Goodrich, L. F., Capobianco,T. E., Clark, A. F., Kwo, J., Hong, M., and Liou, S. H.,Phys. Rev. B 35, 8856 (1987).

*[79] Moreland, J., Goodrich, L. F., Ekin, J. W., Capobianco,T. E., Clark, A. F., Braginski, A. I., and Panson, A. J.,Appl. Phys. Lett. 51, 540 (1987).

*[80] Moreland, J., Goodrich, L. F., Ekin, J. W., Capobianco,T. E., and Clark, A. F., Japanese J. Appl. Phys. 26, Sup-plement No. 26-3, 999 (1987).

*[81] Moreland, J., Clark, A. F., Damento, M. A., andGschneider, K. A., Jr., Proc. Intern. Conf. on HighTemperature Superconductivity Interlaken, Switzer-land, (1988) p. 1383.

[821 Zeller, H. R., and Giaver, 1., Phys. Rev. 181, 789 (1969).[83] Kapailtulnik, A., APS March Meeting 1987.

175



*[84] Moreland, J., Goodrich, L. F., Ekin, J. W., Capobianco,T. E., and Clark, A. F., Adv. Cryogenic Eng. 34, 625(1988).

*[85] Moreland, J., Beall, J. A., Ono, R. H., and Clark, A. F.,Materials Research Society, EA-14, 351 (1987).

*[86] Moreland, J., Alexander, S., Cox, M., Sonnenfeld, R.,and Hansen, P. K., Appl. Phys. Lett. 43, 387 (1983).

[87] Geballe, T. H., in Novel Superconductivity (Plenum,N.Y. 1987) p. 1108.

*[88] Kurtz, R. L., AIP Proc. 165, Amer. Vac. Soc. Series 3:Thin Film Processing and Characteristics of High-TcSuperconductors, Harper, J. M. E., Colton, R. J., Feld-man, L. C., Eds., NY (1988) p. 223.

*[89] Kurtz, R. L., Stockbauer, R. L., Mueller, D., Shih, A.,Toth, L. E., Osofsky, M., and Wolf, S. A., Phys. Rev. B35, 8818 (1987).

*[90] Tsang, K.-L., Zhang, C. H., Callcott, T. A., Canfield, L.R., Ederer, D. L., Blendell, J. E., Clark, C. W.,Wassdahl, N., Rubensson, J. E., Bray, G., Mortensson,N., Nordgren, J., Nyholm, R., and Cramm, S., Phys.Rev. B 37, 2293 (1988).

*[91] Kurtz, R. L., Robey, S. W., Stockbauer, R. L., Mueller,D., Shih, A., and Toth, L., submitted to Phys. Rev. B.

*[92] Kurtz, R. L., Robey, S. W., Stockbauer, R. L., Mueller,D., Shih, A., Toth, L., Singh, A. K., and Osofsky, M., tobe published in Vacuum as part of the Proceedings of 1stIberian Vacuum Meeting.

*[93] Kurtz, R. L., Stockbauer, R., Madey, T. E., Mueller, D.,Shih, A., and Toth, L. E., Phys. Rev. B 37, 7936 (1988).

*[94] Mueller, D., Shih, A., Toth, L. E., Osofsky, M., Wolf, S.A., Kurtz, R. L., and Stockbauer, R. L., Phys. Rev. B 37,7936 (1988).

*f95] Moreland, J., Li, Y., Folsom, R., and Capobianco, T. E.,Rev. Sci. Instruments, in press.

*[96] Moreland, J., Li, Y., Folsom, R. M., and Capobianco,T. E., to be published in the Proceedings of the AppliedSuperconductivity Conference, IEEE Transactions onMagnetics.

*[97] Ekin, J. W., Panson, A. J., Braginski, A. I., Janocko, M.A., Hong, M., Kwo, J., Liou, S. H., Capone, D. W., II,and Flandermeyer, B., Materials Research Soc., Vol.EA-ll, 223 (1987).

*[98] Ekin, J. W., Advanced Ceramic Mater.: Ceramic Super-conductors 2, 586 (1987).

[99] Dinger, T. R., Wothington, T. K., Gallagher, W. J., andSandstrom, R. L., Phys. Rev. Lett. 58, 2687 (1987).

*[100] Ekin, J. W., Braginski, A. I., Panson, A. J., Janocko, M.A., Capone, D. W., II, Zaluzec, N. J., Flandermeyer, B.,de Lima, 0. F., Hong, M., Kwo, J., and Liou, S. H., J.Appl. Phys. 62, 4821 (1987).

*[101] Peterson, R. L., and Ekin, J. W., Phys. Rev. B 37, 9848(1988).

*[102] Peterson, R. L., and Ekin, J. W., Physica C (to be pub-lished).

*[103] Ekin, J. W., Peterson, R. L., and Bray, S. L., MaterialsResearch Society International Meeting on AdvancedMaterials, Tokyo, Japan, May 30-June 3, 1988.

*[104] Ledbetter, H., J. Metals 40, 24 (1988).*[1051 Ledbetter, H. M., Austin, M. W., Kim, S. A., and Lei,

M., J. Mater. Res. 2, 786 (1987).*[106] Ledbetter, H. M., Proc. MRS Int. Meeting on Advanced

Mater. (Tokyo, June 1988).*[107] Ledbetter, H. M., Austin, M. W., Kim, S. A., Datta, T.,

Violet, C. E., J. Mater. Res. 2, 790 (1987).

*[108] Ledbetter, H. M., Kim, S. A., Datta, T., Estrada, J., andViolet, C. E., submitted to Phys. Rev. B.

*1109] Ledbetter, H. M., Kim, S. A., Datta, T., Estrada, J., andViolet, C. E., submitted to Physica C.

*[110] Ledbetter, H. M., Kim, S. A., and Capone, D. W., inHigh-Temperature Superconductors II (Materials Re-search Society Symposium Series, 1988) p. 293.

[111] Bishop, B. J., Ramirez, A. P., Gammel, P. L., Batlogg,B., Reitman, E. A., Cava, R. J., and Millis, R. J., Phys.Rev. B 36, 2408 (1987).

[112] Geballe, T. H., and Hulm, J. K., Science 239, 367 (1988).*[113] Ledbetter, H. M., Kim, S. A., and Lei, M., in Cryogenic

Materials 1988 (Int. Cryog. Mater. Conf., Boulder, CO,1988).

*[114] Datta, T., Ledbetter, H. M., Violet, C. E., Almasan, C.,and Estrada, J., Phys. Rev. B 37, 7502 (1988).

*[I 15] Violet, C. E., Datta, T., Ledbetter, H. M., Almasan, C.,and Estrada, J., in High Temperature Superconductors(Materials Research Society Symposium Series) 99, 375(1988).

[116] Testardi, L. R., Rev. Mod. Phys. 47, 637 (1975).*[117] Ledbetter, H. M., and Kim, S. A., Phys. Rev. B 38,

11857 (1988).*[118] Block, S., Piermarini, G. J., Munro, R. G., and

Wong-Ng, W., Advanced Ceramic Mater.: Ceramic Su-perconductors, 2, 601 (1987).

*[119] Ledbetter, H. M., and Lei, M., to be published; Ledbet-ter, H. M., to be published.

*[120] Drescher-Krasicka, E., Johnson, W. L., Blendell, J. E.,Chiang, C. K., and Wadley, H. N. G., submitted to Phys-ical Review Letters.

*[121] Atzmony, U., Shull, R. D., Chiang, C. K., Swartzendru-ber, L. J., Bennett, L. H., and Watson, R. E., J. Appl.Phys. 63, 4179 (1988).

[122] Bean, C. P., Rev. Mod. Phys. 361, 39 (1964).[123] Bean, C. P., Phys. Rev. Lett. 8, 250 (1962).[124] Kim, Y. B., Hempstead, C. F., and Strnad, A. R., Phys.

Rev. Lett. 9, 306 (1962).[125] London, H., Phys. Lett. 6, 162 (1963).[126] Kamper, R. A., Phys. Lett. 2, 290 (1962).

*[127] Peterson, R. L., submitted to Physics Letters A, May 11,1988.

*[128] Chen, D,-X., and Goldfarb, R. B., to be published.*[129] Allen, P. B., Cohen, M. L., and Penn, D. R., Phys. Rev.

B. 38, 2513 (1988).*[130] Casella, R. C., Nuovo Cimento 1OD, 1439 (1988).*[131] Melamud, M., Bennett, L. H., and Watson, R. E., Phys.

Rev. B 38, 4624 (1988).[1321 Pauling, L., Phys. Rev. Lett. 59, 225 (1987).

*[1331 Zimmerman, J. E., Beall, J. A., Cromar, M. W., andOno, R. H., Appl. Phys. Lett. 51, 617 (1987).

*1134] Zimmerman, J. E., Beall, J. A., Cromar, M. W., andOno, R. H., Japn. J. Appl. Phys. 26, Suppl. 26-3, 2125(1987).

*1135] Ono, R. H., Beall, J. A., Cromar, M. W., Mankiewich, P.M., Howard, R. W., and Skocpol, W., IEEE Trans. onMagnetics, MAG-25 (1989).

*[136] Huang, C. Y., Shapira, Y., McNiff, E. J., Jr., Peters, P.N., Schwartz, B. B., Wu, M. K., Shull, R. D., andChiang, C. K., Mod. Phys. Lett. B2, 869 (1988).

*[137] Shull, R. D., Swartzendruber, L. J., Chiang, C. K., Wu,M. K., Peters, P. N., and Huang, C. Y., in High-T, Su-perconductors: Magnetic Interactions, World ScientificPublishing Co. (1988).

176



[138] Cava, R. J., Batlogg, B., van Dover, R. B., Murphy, D.W., Sunshine, S., Siegrist, T., Remeika, J. P., Reitman,E. A., Zahurak, S., and Espinosa, G. P., Phys. Rev. Lett.58, 1676 (1987).

*[139] Adrian, F. J., Bohandy, J., Kim, B. F., Moorjani, K.,Shull, R. D., Swartzendruber, L. J., and Bennett, L. H.,Physica C 156, 184 (1988).

*[140] Kim, B. F., Phillips, T. E., Green, W. J., Agostinelli, E.,Adrian, F. J., Moorjani, K., Swartzendruber, L. J.,Shull, R. D., Bennett, L. H., and Wallace, J. S., AppI.Phys. Lett., 53, 321 (1988).

*[141] Moorjani, K., Bohandy, J., Adrian, F. J., Kim, B. F.,Atzmony, U., Shull, R. D., Chiang, C. K.,Swartzendruber, L. J., and Bennett, L. H., J. Appl. Phys.63, 4199 (1988).

*[142] Katayama-Yoshida, H., Hirooka, T., Oyamada, A., Ok-abe, Y., Takahashi, T., Sasaki, T., Ochiai, A., Suzuki, T.,Mascarenhas, A. J., Pankove, J. I., Ciszek, T. F., Deb, S.K., Goldfarb, R. B., and Li, Y., Physica C 156, 481(1988).

[143] Batlogg, B., Kourouklis, G., Weber, W., Cava, R. J.,Jayaraman, A., White, A. E., Short, K. T., Rupp, L. W.,and Rietman, E. A., Phys. Rev. Lett. 59, 912 (1987).

[144] Faltens, T. A., Ham, W. K., Keller, S. W., Leary, K. J.,Michaels, J. N., Stacy, A. M., zur Loye, H. C., Barbee,T. W., III, Bourne, L. C., Cohen, M. L., Hoen, S., andZettI, A., Phys. Rev. Lett. 59, 915 (1987).

[145] Leary, K. 5., zur Loye, H. C., Keller, S. W., Faltens, T.A., Ham, W. K., Michaels, J. N., and Stacy, A. M., Phys.Rev. Lett. 59, 1236 (1987).

[146] Morris, D. E., Kuroda, R. M., Marketz, A. G., Nickel, J.H., and Wei, J. Y. T., Phys. Rev. B 37, 5936 (1988).

*[147] High-Tc Superconductors: Magnetic Interactions, Ben-nett, L. H., Vezzoli, G., and Flom, Y., Eds. (World Sci-entific Publishing Co., Teaneck, NJ, 1989).

*[1481 Lundy, D. R., Ritter, J., Swartzendruber, L. J., Shull, R.D., and Bennett, L. H., J. Superconductivity in press,1989.

*[149] Lundy, D. R., Ritter, J., Swartzendruber, L. J., Shull, R.D., and Bennett, L. H., in High-Tc Superconductors:Magnetic Interactions, (World Scientific Publishing Co.,Teaneck, NJ, 1989).

*[150] Ledbetter, H. M., Kim, S. A., Goldfarb, R. B., andTogano, K., to be published.

*1151] Stockbauer, R. L., Robey, S. W., Kurtz, R. L., Mueller,D., Shih, A., Singh, A. K., Toth, L., and Osofsky, M.,AIP Conference Proceedings of AVS Topical Confer-ence on Thin Film Processing and Characterization ofHigh-Temperature Superconductors, Atlanta, Georgia(October, 1988).

*1152] Ekin, J. W., Cryogenics 1987 27, 603 (1987).*[153] Bray, S. L., Goodrich, L. F., Dub6, W. P., submitted to

RSI, May 12, 1988.*[154] Goodrich, L. F., and Bray, S. L., submitted to Cryogen-

ics, July 7, 1988.*[155] Goodrich, L. F., Bray, S. L., and Stauffer, T. C., Pro-

ceedings of the Applied Superconductivity Conference,IEEE Transactions on Magnetics.

*[156] NBSIR 88-3088, Goodrich, L. F., Editor, published bythe National Bureau of Standards, 75 pages (1988).

*[157] Goodrich, L. F., Bray, S. L., and Clark, A. F., Advancesin Cryogenic Engineering Mater. 34, 1019 (1988).

*[158] Goodrich, L. F., Pittman, E. S., Ekin, J. W., and Scan-lan, R. M., IEEE Trans. on Magnetics, MAG-23, 1642(1987).

*[159] Goodrich, L. F., and Bray, L. S., to be published in theProceedings of the Applied Superconductivity Confer-ence, IEEE Transactions on Magnetics.

*[160] Ekin, J. W., IEEE Trans. on Magnetics, MAG-23, 1634(1987).

*[161] Bray, S. L., and Ekin, J. W., Journal of Applied Physics,in press.

*[162] Ekin, J. W., Advances in Cryogenic Engineering Mater.34, 547 (1988).

*[163] Ekin, J. W., J. Appl. Phys. 62, 4829 (1987).*[164] Goldfarb, R. B., Ried, D. L., Kreilick, T. S., and

Gregory, E., to be submitted to Applied Superconduc-tivity Conference-1988 for publication in IEEE Trans-actions on Magnetics.

[165] Collings, E. W., Adv. Cryoeng. (Materials) 34, 817(1988)

[166] Kreilick, T. S., Gregory, E., Wong, J., Scanlan, R. M.,Ghosh, A. K., Sampson, W. B., and Collings, E. W.,Adv. Cryoeng. (Materials) 34, 895 (1988).

*[167] Lloyd, F. L., Hamilton, C. A., Beall, J. A., Go, D., Ono,R. H., and Harris, R. E., IEEE Electron Device Letters,EDL-8, 449 (1987).

The following additional papers authored or co-authored byNIST personnel were not received in time for inclusion in thereview:

Moreland, J., Ono, R. H., Beall, J. A., Madden, M., and Nelson,A. J., submitted to Applied Physics Letters.

Chaillout, C., Santoro, A., Remeika, J. P., Cooper, S. A.,Espinosa, G. P., and Marezio, M., Solid State Communica-tions 65, 1363 (1988).

Bordet, P., Hodeau, J. L., Strobel, P., Marezio, M., and Santoro,A., Solid State Communications 66, 435 (1988).

Chang, K. J., Cohen, M. L., and Penn, D. R., Phys. Rev. B 38,8691 (1988).

Goldfarb, R. B., Cizek, T. F., and Evans, C. D., J. Appl. Phys.64, 5914 (1988).

Nikolo, M., and Goldfarb, R. B., submitted for publication.Moreland, J., Goodrich, L. F., Ekin, J. W., Capabianco, T. E.,

and Clark, A. F., National Institute of Standards and Tech-nology, NISTIR 88-3090 (May 1988).

Zhang, C. H., Callcott, T. A., Tsang, K.-L., Ederer, D. L.,Blendell, J. E., Clark, C. W., Scimeca, T., and Liu,Y.-W., Phys. Rev. B (March 1, 1989).

Callcott, T. A., Tsang, K.-L., Zhang, C. H., Ederer, D. L.,Clark, C. W., Wassdahl, N., Rubensson, J. E., Bray, G.,Mortensson, N., Nordgren, J., Nyholm, R., and Cramm, S.,Extended Abstracts, High-Temperature Superconductors II,April 5-9, 1988, Reno, NY.

Tsang, K.-L., Zhang, C. H., Callcott, T. A., Canfield, L. R.,Ederer, D. L., Blendell, J. E., and Clark, C. W., Drexel High-T. Conference Proceedings (July 1987).

Ederer, D. L., Canfield, L. R., Callcott, T. A., Tsang, K.-L.,Zhang, C. H., Arakawa, E. T., SPIE, 911, X-Ray and VUVInteraction Data Bases, Calculations and Measurements, p. 75(1988).

Tsang, K.-L., Zhang, C. H., Callcott, T. A., Canfield, L. R.,Ederer, D. L., Blendell, J. E., Clark, C. W., Wassdahl, N.,Rubensson, J. E., Bray, G., Mortensson, N., Nordgren, J.,Nyholm, R., and Cramm, S., J. Phys., Colloque C9 supple-ment au n'12, 48 (December 1987).

Peterson, R. L., submitted to Phys. Rev. B.

177



Katayama-Yoshida, H., Hirooka, T., Oyamada, A., Okabe, Y.,Takahashi, T., Sasaki, T., Ochiai, A., Suzuki, T., Mascaren-has, A. J., Pankove, J. I., Ciszek, T. F., Deb, S. K., Goldfarb,R. B., and Li, Y., Physica C 156, 481 (1988).

Chen, D.-X., and Goldfarb, R. B., submitted to the Proceedingsof the March 1989 Meeting of the American Physical Society.

Peterson, R. L., and Ekin, J. W., Physica C (1989).Casella, R. C., Solid State Communications (in press) 1989.De Reggi, A. S., Chiang, C. K., Swartzendruber, L. J., and

Davis, G. T., High-T. Superconductors: Magnetic Interac-tions (World Scientific Publishing Co., Teaneck,N. J. 1989).

Swartzendruber, L. J., Bennett, L. H., and Gallo, C. F., High-T.Superconductors: Magnetic Interactions (World ScientificPublishing C., Teaneck, N. J. 1989).

Rubinstein, M., Swartzendruber, L. J., Bennett, L. H., Chaki, T.K., Harford, M. Z., Wolf, S. A., and Edelstein, A. E., High-ET.Superconductors: Magnetic Interactions (World ScientificPublishing Co., Teaneck, N. J. 1989).

Wu, M. K., Shull, R. D., Swartzendruber, L. J., Chiang, C. K.,Peters, P. N., and Huang, C. Y., High-T, Superconduc-tors: Magnetic Interactions (World Scientific Publishing Co.,Teaneck, N. J. 1989).

Kaiser, D. L., Gayle, F. W., Roth, R. S., and Swartzendruber,L. J., J. Materials Res., submitted 1989.

Swartzendruber, L. J., and Bennett, L. H., J. Phys., in press1988.

Swartzendruber, L. J., and Bennett, L. H., J. Superconductiv-ity, in press 1988.

Wong-Ng, W., Cook, L. P., Chiang, C. K., Swartzendruber, L.J., and Bennett, L. H., Mater. Res. Soc., in press 1989.

178



Calibration of Voltage Transformersand High- Voltage Capacitors at NIST


William E. Anderson The National Institute of Standards and 100 V to 170 kV at 60 Hz depending onTechnology (NIST) calibration service the nominal capacitance. Calibrations

National Institute of Standards for voltage transformers and high- over a reduced voltage range at otherand Technology, voltage capacitors is described. The ser- frequencies are also available. As in theGaithersburg, MD 20899 vice for voltage transformers provides case with voltage transformers, these

measurements of ratio correction factors voltage constraints are determined byand phase angles at primary voltages up the facilities at NIST.to 170 kV and secondary voltages aslow as 10 V at 60 Hz. Calibrations atfrequencies from 50-400 Hz are avail- Key words calibration; capacitors; dissi-able over a more limited voltage range. pation factor; electric power; electricalThe service for high-voltage range. standards; NIST services; voltage trans-

Thesevie or ig-vltgecapacitors formers.provides measurements of capacitanceand dissipation factor at applied voltagesranging from Accepted: February 15, 1989

1. Introduction

This paper describes the National Institute ofStandards and Technology (NIST) methodologyfor calibrating high-voltage capacitors and trans-formers. This should benefit NIST clients inseveral ways. First, by understanding how NISTmakes these measurements, the clients mightbe able to define weaknesses in their own mea-surement procedures and correct them. Second,the clients should be able to make better use ofthe data in the calibration report (e.g., to under-stand what is meant by the uncertainty statement).Third, the clients should be able to better specifythe required test conditions so that informationmore pertinent to their needs can be obtained at alower cost.

This paper describes two different calibrationservices: high-voltage capacitors and voltagetransformers. At NIST these two services are per-formed using the same equipment. In fact, in order

to calibrate a voltage transformer, one of the stepsis to measure the ratio of two capacitors. The twoservices are therefore discussed in parallel.

There are several different ways to measure theratio and phase angle of a voltage transformer.Harris [1] categorizes them as the direct versuscomparative methods and within these two classifi-cations either the deflection or null measurementtechnique. A direct measurement is defined here asa measurement in which the quantity of interestcan be determined without a comparison to someabsolute standard.

In the "direct deflection method" the primaryand secondary voltage vectors are each directlymeasured. This approach is, in general, of mostvalue for lower voltage transformers (i.e., primaryvoltages of order 100 V). Even then more accurate,less difficult measurements can be made using oneof the other techniques.

179



In the past NIST had used a "comparative nullmethod" to calibrate voltage transformers. The un-known transformer was compared to a NIST refer-ence transformer using a voltage comparatorconsisting of a variable resistive divider and a mu-tual inductor. Reference transformers were avail-able with ratios ranging from 1/1 up to 2000/1.Measurement uncertainties in the comparison ofthe unknown transformer with the reference trans-former were ±0.01% for ratio and ±0.3 minutesfor phase angle. The ratio and phase angle of thereference transformers were known to about thesame accuracy. There are several disadvantages tothis approach. Since the comparator has a limitedrange, several reference transformers must beavailable to cover the anticipated users' needs. Theratio and phase angles of each one of these trans-formers must be carefully determined over thesecondary voltage range of interest. These trans-formers then have to be rechecked at regular inter-vals to determine if the ratios and phase angleshave changed.

If a direct measurement method were availablethat was sufficiently accurate and straightforwardto make the calibration of these reference trans-formers a simple task, then that method could beused to measure the client's transformer directly.At NIST, the "direct null method" in use origi-nally involved balancing the secondary of the ref-erence transformer against the output of a resistivedivider used in conjunction with a variable mutualinductor to provide phase angle balance. Such ameasurement was difficult because the resistive di-vider ratio changed with heating. Since the late1960s a "direct null method" has been availablethat is straightforward and accurate and is nowused at NIST in place of comparative methods us-ing reference transformers.

Capacitors are invariably measured by balancingthe unknown capacitor against a known standardusing some type of bridge arrangement. There area variety of such bridges described in the literature[2]. The one most used in high-voltage applicationsin the last 60 years is the Schering bridge (fig. 1).The two high-voltage arms of this bridge consist ofthe standard and unknown capacitors. The twolow voltage arms are resistors (one has a parallelcapacitor for phase angle balance).

The main limitation of the Schering bridge isthat the low side of the unknown and standard ca-pacitors are not at ground potential at bridge bal-ance. Therefore, without carefully guarding thebridge compollents, stray currents can affect thebridge accuracy. The voltage applied to the shieldsto eliminate these stray currents must be adjusted

for both magnitude and phase. Unfortunately thisprocedure is not perfect and bridge accuracy isconsequently affected. Another limitation of theSchering bridge is the inherent inaccuracy of theresistance ratio of the two low-voltage arms.

Figure 1. Schering bridge.

The current comparator bridge developed byKusters and Petersons [3] allows the intercompari-son of two capacitors with their low-voltage termi-nals at ground potential, thereby eliminating themain objection in using the Schering bridge. Thisbridge, used in both voltage transformer and ca-pacitor calibrations, will be described in some de-tail in section 4. There is an important distinctionbetween the calibration of voltage transformersand capacitors at NIST. The voltage transformercalibration is of the direct null type, and the capac-itor calibration is of the comparative null type. Inother words, the accuracy of the capacitance mea-surements ultimately depends on the uncertainty inassigning a value to a standard capacitor. The stan-dard capacitor used in this service is directly trace-able to the calculable cross capacitor [4] which, inturn, is known in terms of the fundamental unit oflength.

The remainder of this paper is divided into thefollowing subject areas: voltage transformers andcapacitors covered by the service, measurementmethodology, measurement instrumentation, andanalysis of uncertainties. The contents of this paperplus the cited references should provide the readerwith a fairly complete description of the voltagetransformer and high-voltage capacitor calibrationservices at NIST.

180



2. Range of Services

The NIST measurement capabilities are summa-rized in table 1 and discussed in more detail below.

Table 1. Measurement capability

Voltage transformers-60 Hz

Primary voltage Secondary voltage Phase angle50-170,000 V rms > 50 V rms < 11 mrad

Capacitors-60 HzApplied voltage' Capacitance Dissipation factor

50-170,000 V rms 10 pF-0.001 F <0.011

'Total power must be less than 50 kVA.

2.1 Voltage Transformers

Presently, voltage transformers (assuming theyare of sufficient quality to be used as laboratorystandards) with primary voltages up to 170 kV at afrequency of 60 Hz can be calibrated at NIST. Thismaximum voltage is imposed by the supply trans-former and not by limitations in the measurementinstrumentation. Therefore, this constraint shouldnot be considered rigid and clients should contactthe NIST about present physical limitations.

The largest portion of the voltage transformerssubmitted to NIST are calibrated with total esti-mated uncertainties of ±300 parts per million(ppm) in ratio, and ±0.3 mrad in phase angle.These transformers are of sufficient quality to beconsidered transfer standards. Historically thesetransformers have shown excellent long-term sta-bility, rarely changing by more than 100 ppm inratio or 0.1 mrad in phase (at or below rated bur-den) for periods as long as 30 years or more. Ingeneral, the voltage and burden dependence ofthese transformers are the major contributors tothe measurement uncertainties. These uncertainties(±300 ppm for ratio, ±0.3 mrad for phase angle)meet the accuracy requirements of most NISTclients.

Voltage transformers of a higher accuracy classoften serve as transfer standards for manufacturersof voltage transformers and voltage transformertest sets (voltage comparators). The estimated un-certainties for these transformers are ± 100 ppm inratio, and ±0. I mrad in phase angle. They are gen-erally designed for use with very small burdens(<15 volt-amperes).

The above discussion for voltage transformersassumes a voltage at a frequency of 60 Hz. The Na-tional Institute of Standards and Technology hassome capability to calibrate voltage transformers

from about 50 Hz to 400 Hz (at the lower voltageand power ranges). Such calibrations are infre-quent and clients interested in these voltage rangesand measurement uncertainties should contactNIST directly.

2.2 Capacitors

The maximum voltage for capacitor calibrationsis presently 170kV at 60Hz. The restrictions areimposed by the supply transformer and not by limi-tations in the measurement instrumentation. There-fore, this constraint should not be considered timeinvariant and clients should contact NIST aboutpresent physical limitations.

The maximum power available is 50 kVA (i.e.,C<50,000/{2ir60V 2} where V is the appliedvoltage and C is the capacitance). In order to ener-gize the capacitors a resonant circuit is often re-quired to couple the necessary energy into theclient's capacitor. Since this requires the availabil-ity of an assortment of series and parallel inductorsand capacitors, there are undoubtedly some capaci-tors that, despite having a burden of less than50 kVA, cannot be calibrated. The client shouldcontact NIST before submitting a capacitor for cal-ibration. As with voltage transformers, NIST re-stricts its calibration services to those devices ofsufficient quality to be used as transfer standards.This in general depends upon the stability of thecapacitor (i.e., whether the measured capacitanceand dissipation factor are intrinsic properties of thedevice itself or instead are largely a function ofconditions at the time of the calibration). For ex-ample, small two-terminal capacitors less than10,000 pF) may be significantly influenced by straycapacitance in the measurement circuit. There arecases, however, where one component (capaci-tance or dissipation factor) is stable and the other isnot. For example, power factor capacitors oftenhave relatively stable dissipation factors but havecapacitances that vary significantly with appliedvoltage (even demonstrating hysteresis effects) andtemperature. In this case a calibration of dissipationfactor would be meaningful. It also is importantthat the capacitors have connectors' that aregenerally available, e.g., BNC, GR, UHF, BPO, orType N.

'Certain commercial products are identified to adequatelyspecify the experimental procedure. In no case does such identi-fication imply recommendation by NIST, nor does it imply theproducts are the best available.

181



The most accurate capacitor calibrations have anuncertainty of +25 ppm for capacitance and an un-certainty of ±5 X 10- for dissipation factor. Forcapacitors with large dissipation factors, the dissi-pation factor uncertainty is generally at least ± 1%of the measured value ±5 X 106. The uncertaintyin the capacitance value and the dissipation factorcan be largely a function of the stability of the ca-pacitor.

V

3. Measurement Methodology3.1 Basic Measurement Circuits

The current comparator bridge used to calibratevoltage transformers and high-voltage capacitorswill be discussed in considerable detail in section 4.A brief discussion of this bridge will be presentedhere in order to facilitate understanding of theNIST measurement methodology. A simplified cir-cuit for measuring the ratio of two capacitors isshown in figure 2. (The active circuitry to achievedissipation factor balance is not included.) At bal-ance

V2rrfC Ns_ V2rfC5N.ND ND (1)

where f is the frequency. This can be rewritten

N.C C (2)

The simplified circuit for measuring the ratio ofvoltage transformers is shown in figure 3. At bal-ance

V,27rrfCpN _ VY27]fCN (3)Nd Nd

or,

V, _ NXC (4)Vs N.Cp-

The ratio of the two capacitors in eq (2) can bemeasured using the circuit of figure 2.

The measurement of a voltage transformer or acapacitor both involve the measurement of the ra-tio of two standard capacitors. The measurementof capacitors will be discussed below followed by adiscussion on the measurement of voltage trans-formers.

Figure 2. Basic measurement circuit for the calibration of ahigh-voltage capacitor.

Vp

Figure 3. Basic measurement circuit for the calibration of avoltage transformer.

3.2 Capacitors

3.2.1 General Measurement Technique Capacitorsare measured by balancing the current through thecapacitor under test against the current through astandard air or compressed gas capacitor as shownin figure 2. Large capacitors (> 1 MF) necessitate afour-terminal measurement as shown in figure 4.This measurement will be discussed in section 4.The four-terminal measurement eliminates the ef-fect of leads in the measurement of capacitance anddissipation factor..3.2.2 Information Necessary to Initiate Calibra-tion The client usually only needs to specify thevoltage and the frequency. For small capacitors(10,000 pF or less), it is essential that the low-voltage electrode and the conductor leading to themeasurement instrumentation be shielded by agrounded conductor. Otherwise, the stray capaci-tance may cause significant measurement error.The National Institute of Standards and Technol-ogy requires some sort of standard connector(BNC, UHF, GR, BPO, or Type N) at the low-voltage terminal in order to connect to the mea-surement system. Larger capacitors do not need to

182



be shielded but must be measured as a four terminaladmittance because of the non-negligible leadimpedance. A description of how this measurementis done will be covered in section 4. Capacitorsmust be stable and reproducible in order to be con-sidered standards and hence warrant a NIST cali-bration. Power factor capacitors (large capacitorsused to tune distribution lines, etc.) are often spe-cial cases. Their dissipation factors (in-phase com-ponent of the current divided by quadraturecomponent) are often quite stable but their capaci-tance values are often not. Because of the impor-tance of these capacitors to the electrical industry,they are often acceptable for calibration eventhough they do not meet normal stability require-ments.

Ir0

Umij

Nx Ns

Figure 4. Basic measurement circuit for the four-terminalcalibration of large capacitors.

Although the instrumentation has been used tocalibrate a million-volt standard capacitor at ratedvoltage, the instrumentation does impose some lim-itations on the voltage applied to the capacitor.The only limitation on the maximum voltage is thatthe current through the standard capacitor shouldbe no larger than 10 mA. In order to have reason-able sensitivity, the current should be at least10 ,A. The current through the client's capacitorcan range from 10 pA to 1000 A.3.2.3 Voltage Dependence For the calibration ofboth capacitors and voltage transformers, thevoltage coefficient of the standard capacitor is im-portant. The unit of capacitance at NIST is main-tained at low voltage. This value must betransferred to the high-voltage standard capacitorsat their working voltages. At NIST, considerablework was done to modify a commercial high-voltage standard capacitor to minimize its voltagecoefficient and to determine the magnitude of thatvoltage coefficient [5]. The National Institute of

Standards and Technlogy was able to demonstratethat, if care is taken, a well-designed standard ca-pacitor should change capacitance by only a fewppm from 0 to 300 kV. A more recent paper alsodiscusses the problem of the voltage dependence ofstandard capacitors and describes an internationalcomparison of high-voltage capacitor measure-ments [6]. (This paper also discusses the effect ofshipping and handling on the measured capacitanceof a commercial standard capacitor.) The voltagedependence of a compressed gas capacitor princi-pally arises from the coulombic attraction of thetwo electrodes and is hence quadratic in nature.The capacitor should be expected to vary onlyslightly at lower voltages. Therefore, a capacitorrated at 200 kV should be quite effective in measur-ing the voltage dependence of another capacitorrated at 20 kV.3.2.4 Temperature Dependence Another concernis the temperature dependence of the high-voltagestandard capacitor. The typical dependence isabout +20 ppm/'C. This dependence arises solelyfrom the thermal expansion of the components ofthe capacitor. Since C is directly proportional tothe electrode area and inversely proportional to theelectrode separation, the thermal coefficient of thestandard capacitor is proportional to the linear co-efficient of expansion. Although the laboratories atNIST are fairly stable in temperature, the compari-son of the high-voltage standard capacitor to thelow-voltage standard (which has a thermal coeffi-cient of 2 ppm/'C) is done at the beginning andconclusion of the measurement process. The aver-age value is then used in order to minimize theproblem associated with this thermal drift.3.2.5 Gas-Density Dependence Compressed gasstandard capacitors can have an additional sourceof error associated with gas leakage. Values ofaClaP (to first order in pressure) measured at atemperature of 22.8 'C are shown in table 2 forthree different gases [6].

Table 2. Gas density dependence

Gas C/aP at T=22.8 'C(units of picofarads/pascal)

SF6 (2.012±0.022)X 10-6 +[(5.1±0.@6)X 10- 3]PC0 2 (0.903+0.015)X 10-6+[(l.4+0.4)X 10-13 ]PHe (0.0750.004) x 10-6+ [(0.2+0. 1)x 10-'31p

The gas pressure, P, is in units of pascals and thecapacitance in picofarads. For a 100-pF capacitorwith SF6 as the dielectric gas, a 1-psi (6900-Pa)

183

.,



leak would cause the capacitance to decrease byabout 140 ppm. It must be stressed that this changeis valid only if the pressure change is caused by theloss of gas and not by the lowering of the gas tem-perature. As can be seen in table 2, the gas densitycoefficient is largest for SF6 . Clients using com-pressed gas capacitors for standards might be ad-vised to monitor the gas pressure with a goodquality pressure gauge. Leaking SF6-filled capaci-tors should be checked often against a good low-voltage standard.

3.3 Voltage Transformers

3.3.1 Information Necessary to Initiate Calibra-tion In order to calibrate a voltage transformer,several different parameters must be specified:frequency; windings and/or range; secondaryvoltage; and burden or impedance across the sec-ondary winding. In some cases, for example whenthere is a tertiary winding, additional parametersmay be required.3.3.2 Labeling of Terminals There are some stan-dard conventions as to which of the primary andsecondary taps are to be at low or ground potentialand which are to be at rated voltage. Some trans-formers have one tap of the secondary and one tapof the primary winding marked by a "+". Thesetwo taps are connected together and to ground po-tential. Some transformers use the designators Hi,H2 for the primary taps, and Xl, X2 (and Yl andY2 for the transformers with two secondaries) forthe secondary taps. Sometimes the secondarywinding has a third tap, X3. By convention the pri-mary and secondary taps with the largest numberare connected together and to ground. If the clientwants some other arrangement, NIST should benotified prior to the calibration.3.3.3 Load Imposed by NIST Measurement Sys-tem The basic measurement circuit is shown infigure 5. The two capacitors shown are three-terminal standard capacitors. Their dissipation fac-tors are typically less than 5X lO-5. The capacitorconnected to the secondary usually has the nominalvalue of 1000 pF. Therefore, for 60-Hz measure-ments, the capacitor imposes a negligible load(2.7 Mfl or 0.005 volt-amperes at 120 V) on thevoltage transformer. Negligible in this case meansthat the effect of this burden on the measured ratioand phase angle can not be observed at the ppmlevel. The digital voltmeter (DVM) in figure 5 hasan estimated uncertainty of less than _0.5% of thereading and measures true-rms ac volts. The inter-nal impedance of the DVM is equal to or greaterthan one megohm.

Figure 5. Basic measurement circuit for the calibration of avoltage transformer with a digital voltmeter (DVM) andsecondary burden.

3.3.4 Possible Errors Caused by ImproperWiring The wiring of the circuit shown in figure5 is critical. For example, it is important that thetwo capacitors be connected directly to the pri-mary and secondary terminals of the transformer.Consider instead figure 6. The capacitor Cs is con-nected to the burden and the DVM instead of di-rectly to the secondary terminal of the transformer.If the secondary burden were an ANSI standardburden ZZ (36 £7 at 120 V, see table 3) and the re-sistance of the lead connecting the burden to thetransformer were 10 mfZ, the incorrect wiringshown in figure 6 would cause a error in the trans-former ratio measurement of about 0.03%. Forhigher impedance burdens this becomes less of aproblem but, in general, one must take precautionsto avoid including the voltage drop in the lead con-necting the transformer to the burden as part of thevoltage on the transformer secondary winding tobe measured.

Figure 6. Measurement circuit for the calibration of a voltagetransformer. Connection of low-voltage capacitor as shown isincorrect.

184



Table 3. ANSI standard burdens

ANSI burden Volt-amperes Power factor (lagging)

W 12.5 0.10X 25 0.70M 35 0.20Y 7 0.85Z 200 0.85

ZZ 400 0.85

Another major concern in the measurement ofthe ratio and phase angle of a voltage transformeris the proper definition of the ground point and theavoidance of ground loops. This can best be illus-trated by a few examples. In figure 7, some com-mon mistakes are shown. The transformer isenergized in such a manner that significant currentis forced to flow between the transformer groundand the circuit ground. The resulting voltage dropin the lead connecting the transformer and groundwill be part of the ratio and phase angle measured.The high-voltage capacitor is not connected di-rectly to the primary of the transformer under test.The measurement of the ratio and phase angle,therefore, includes the effect of the voltage drop inthe lead between the point where the capacitor isconnected to the power source and the trans-former. In addition, as there are three different"ground" points in the circuit and it is not, in gen-eral, possible to know the voltages and impedancesbetween these points, a measurement error is prob-able.

In figure 8 the problem has been eliminated bydefining the low-voltage terminal of the trans-former as ground. Although this point may signifi-cantly differ from the building or utility ground,from the measurement point of view this is the cor-rect ground. It is important that the shields of thethree-terminal capacitors, the bridge detectorground, and all other measurement grounds eachbe connected directly to this point.

In figure 5 the preferred method of wiring avoltage transformer calibration circuit is shown.The client's transformer is connected in such a waythat the energizing current does not flow betweenthe transformer and the measurement ground. Allmeasurement grounds are connected to the trans-former ground point. The two capacitors are con-nected directly to the primary and secondaryterminals of the transformer. Only one ground is

used in the circuit. While it is not always possibleto connect the transformer as in figure 5, this is thebest choice. Otherwise tests are required to ensurethat systematic errors are not compromising themeasurement results.

Figure 7. Measurement circuit for the calibration of a voltagetransformer. Grounds are poorly defined.

Figure 8. Measurement circuit for the calibration of a voltagetransformer. Measurement ground is defined. Transformer exci-tation current flows from the measurement ground to buildingground.

3.3.5 Burdens The burden attached to the sec-ondary of the client's transformer (as shown in fig.5) is specified by the client. In general this wouldnot be the burden corresponding to the maximumvolt-ampere rating of the transformer but insteadwould be equal to the burden attached to the trans-former in its intended use. For example, if thetransformer will only have a digital voltmeter at-tached to its secondary, a calibration with a sec-ondary impedance of one megohm would be moreuseful than one with an ANSI ZZ burden attached.Since the ANSI burdens are often requested, theyare summarized in table 3 [7]. By convention theseburdens are defined for a frequency of 60 Hz only.

185



3.3.6 Substitute Burdens If the client of the ser-vice does not send the secondary burden with thetransformer, the National Institute of Standardsand Technology will provide the burden. It is notpractical to have available and adequately charac-terized all of the anticipated burdens. Fortunatelythis is not necessary. If the ratio and phase angle ofa transformer is known for two different burdenvalues, the ratio and phase angle at any other bur-den can be calculated (with certain limitations) [8].A derivation of the formulas relating the ratios andphase angles at zero and some other known burdenvalue are given in the appendix and presented inabbreviated form below.

The voltage transformer will be represented asan ideal transformer with some unknown seriesoutput impedance Z0, as shown in figure 9. Themodel has been shown to be sufficiently accurateexperimentally. The relationship between the inputvoltage Ej, and the output voltage with zero bur-den Ea, is:

E; = RCFoe-r= I Ef eflro, (5)

where N is the nominal (or turns) ratio of the trans-former, RCF is the ratio-correction factor(NXRCF = actual ratio) at zero burden, rJO is theangle by which the secondary voltage vector leadsthe primary voltage vector andj = V/iI. A similarrelationship exists between the input voltage E1 ,and the output voltage E., with secondary burdenC (having impedance Z) shown in figure 10:

E- = N R CFce l rc,1 (6)

where RCFC is the ratio correction factor with sec-ondary burden C and rF, is the corresponding phaseangle. If the transformer is measured at zero bur-den (RCFo and ro) and at burden T (RCF, and r,),the ratio correction factor and phase angle at bur-den C are approximately given by:

RCFc=RCFo+±§ [(RCF,-RCFo)B,cos(0,-0,,)+(r,-r o)sin(0t-Oo)], (7)

where B,= llZc is the burden in fl-1 of theimpedance Z., and

r0+ Bc [(r,- ro)cos(O,-03-(RCF,-RCFO)B,

sin(O, - Oj] * (8)

Ei

zoEo

Figure 9. Equivalent circuit of a voltage transformer.

Eo EC

Figure 10. Equivalent circuit of a voltage transformer withsecondary burden Z.

The power factor of burden C is cos0,:, RCFC isthe ratio correction factor calculated for burden C,and r,, is the angle by which the secondary voltageleads the primary voltage for burden C.

Equations (7) and (8) can be used to calculate theRCF and phase angle for some secondary burden,C, if the ratio correction factors and phase anglesare known at some other burden T, and at zeroburden. In practice, at NIST, capacitive burdensare used for the "T" or known burdens in eqs (7)and (8). The main reason is their stability. The heatgenerated in a large resistive burden, for example,is likely to cause the burden's impedance value tovary. Capacitors, in addition, are compact so eventhe ZZ burden in table 3 is easy to handle. AtNIST, capacitive burden boxes have been con-structed in a binary layout (fig. 11) so that capaci-tors from 1 to 32 pF can be switched in and outallowing any capacitance value from zero to63 gtF. Since a ZZ burden is equivalent to a 74 pFcapacitor at 120 V, two such burden boxes are suf-ficient for nearly all the calibrations at NIST.

Several approximations were made to derive eqs(7) and (8). The approximations relate to the rela-tive ratio of the transformer's output impedance Zoto the impedance of the secondary burden Z, or Zc.

186



The smaller this ratio, the more accurate are eqs (7)and (8). This ratio also affects the differences,RCFA-RCFo and r,-ro. If the ratio correctionfactor difference is 0.001 or less, and if the phaseangle difference is 1 mrad or less than eqs (7) and(8) should be accurate to within + 10 ppm for theratio correction factor and to within ± 10 Fzrad forthe phase angle if it is assumed that the ratio of theburdens is known with no more than ± 1 percentuncertainty. Data over the years has indicated thateqs (7) and (8) are always at least that accurate. Inorder to identify any problems, an extra measure-ment is made at a different secondary burden totest the predictive capabilities of eqs (7) and (8) forthe transformer under test. If a problem is discov-ered, the error budget is adjusted accordingly.

Figure 11. Capacitive burden box.

The above discussion might enable clients of thevoltage transformer calibration service to betterdesign their calibration requests. Using eqs (7) and(8), the client might be able to reduce the numberof measurements required. A note of caution is inorder. It is likely that using a zero burden resultand a 10 volt-ampere burden result to predict thetransformer's behavior at a ZZ burden may lead tolarge inaccuracies. The reasons are twofold. First,the differences RCF,-RCFO and F, - FO are likelyto be small for a burden as small as 10 volt-amperesand extrapolations can cause large errors. The sec-ond reason can be seen from figure 10. The highercurrent of the ZZ burden will cause ZA to heat upand increase in value, leading to errors if eqs (7)and (8) are used. Somewhat better results are likelyif one uses a ZZ burden result to predict a trans-former's behavior at 10 volt-amperes. However, itis best to choose burden T to have a volt-ampererating the same order of magnitude as the burdenof interest C. Also, the values in eqs (7) and (8) areall to be measured at the same frequency and at thesame secondary voltage.3.3.7 Harmonic Effects The measurement of theratio and phase angle of a voltage transformer canbe affected by the presence of harmonics in thevoltage waveform. If a tuned null detector is notused, the balance of a bridge circuit can be difficult

in the presence of harmonics and often a precisebalance is not possible resulting in increased mea-surement uncertainties. Harmonics can also lead toerrors in measuring the magnitude of the secondaryvoltage. For example, if an average reading, rmsscaled voltmeter measured a 100-V rms fundamen-tal with an in-phase 3-V rms third harmonic, themeter would read 101 V. Setting the voltage toread 100 V on the meter would result in a 1-V dis-crepancy between the intended and actual voltage.Many transformers have large enough voltage co-efficients for this 1-V error in the voltage setting tohave a non-negligible effect on the measured ratiocorrection factor and phase angle. If instead, a truerms voltmeter were used to measure this signal, themeasured voltage would be 100.045 V and the re-sulting error would be negligible. At NIST threedifferent steps are taken to lessen the effects of har-monics. The first is to try to minimize the harmoniccontent of the power supply. The supply used formost of the calibrations has a total harmonic distor-tion of order 0.2% of the fundamental. Second, atuned detector is used to assure that the balanceconditions are for the fundamental component ofthe voltage waveform. And third, all voltage mea-surements are made with true-rms voltmeters.3.3.8 Voltage Dependence of Standard Capaci-tor An additional measurement concern is thevoltage coefficient of the high-voltage standard ca-pacitor shown in figure 5. Although no absolutemeasurements are required to calibrate a voltagetransformer, the ratio of the two standard capaci-tors must be known. The problem is that the low-voltage standard capacitor typically has amaximum voltage rating of 500 V, and both the pri-mary of the transformer and the high-voltage stan-dard capacitor might be energized to 100 kV. Sincethe capacitor ratio measurement must be done atless than 500 V, the voltage dependence of thehigh-voltage capacitor is important. This problemwas discussed in section 3.2.

4. Measurement Instrumentation

The calibration of voltage transformers andhigh-voltage capacitors at NIST requires the com-bined use of standard capacitors and the currentcomparator bridge. Standard capacitors have beenthoroughly discussed in the literature [5, 6, 9,]. Thecare that must be taken with their use in these typesof measurements has been discussed above. Thecurrent comparator bridge will be discussed in thissection.

187



The current comparator bridge can be thoughtof as a voltage comparator transformer arm bridgein which the detector and power source have beeninterchanged. Traditionally, the disadvantage ofthe current comparator bridge versus the voltagecomparator bridge is the signal-to-noise level. Forhigh-voltage measurement applications, this is nolonger a problem. Kusters and Petersons were thefirst to develop this bridge for the comparison oftwo capacitors at high voltage [3]. A basic currentcomparator bridge is shown in figure 12. The cur-rent in the unknown capacitor, C., is balancedagainst the current in the standard capacitor, C,, byvarying the turns ratios, N, and N,,.

V

, X Cs

NX NS

Figure 12. Basic current comparator bridge.

Balance is achieved when the signal at the detec-tor D is equal to zero. At balance I,,N,=IN, or:

V27TfCN,,= V27nfC 5sN, (9)

wheref is the frequency. This balance equation canalso be expressed as:

V

Rs

Figure 13. Current comparator bridge with high-voltage resis-tor for in-phase current balance.

The current comparator shown in figure 14 pro-vides a satisfactory means of achieving both thein-phase and quadrature current balances. Thequadrature current balance is identical to that infigures 12 and 13 above. The in-phase current bal-ance is accomplished at low voltage with the aid ofan operational amplifier. The current from thestandard capacitor, after passing through the N,winding, goes to the inverting input of the opera-tional amplifier. This point is at virtual ground sothe capacitive current balance, eq (10), is not af-fected. The feedback capacitor Cf causes the outputvoltage of the operational amplifier to be a smallfraction (ClCf where Cf is approximately 10 pF)of the applied voltage and 7r radians out of phasewith it. The inductive voltage divider allows aknown fraction, a, of this output signal to beapplied across a standard resistor R. As can be seenfrom figure 14, the signal is first inverted before theresistor in order to have the correct phase rela-tionship with the unknown in-phase current.

V

(10).

The bridge shown in figure 12 has no means ofbalancing the in-phase current resulting from anon-ideal unknown capacitor C,. The current com-parator in figure 13 does have the capability of bal-ancing both the in-phase and quadraturecomponents of the capacitive current. The diffi-culty with the approach used in figure 13 is that theapplied high voltage is across the variable resis-tance R,. It is nearly impossible to design a stablehigh-voltage variable resistor with negligible phaseangle. Another means is necessary to balance thein-phase current, preferably at low voltage usingwell-characterized components.

Nx

R

Figure 14. Current comparator with superior in-phase currentbalance.

188

� CX - F��

T L�ID JC.= N� C�'Nx



It is necessary that the non-inverted signal be ap-plied to an identical standard resistor as shown infigure 14 so that the current from the standardwinding N, reaching the operational amplifier hasno phase defect. The in-phase current into the stan-dard winding N, is then equal to:

An= (aVCVCf) (11)R

Since the quadrature current 'outu= V2,rfC,, the dis-sipation factor is:

__Ii _ aVcsDF = Io 2-fl VCfCS (12)

transformer shown in figure 16, referred to as arange extender, increases the measurement rangeby a factor of 1000 allowing the comparison of twocurrents differing in magnitude by as much as afactor of a million. As with the transformers inter-nal to the current comparator bridge, the accuracyrequirements on the range extender are quite strin-gent. Further details on the design of a ppm cur-rent comparator and the specifics of NIST'scurrent comparator bridge are available in the liter-ature [10, 11].

V

orDF = 2RCa (13)

The resistor R can be chosen so that a is directreading in percent or milliradians.

In some cases, particularly for larger capacitors,it is necessary to make a four-terminal measure-ment. This is required when the lead and windingimpedances become a significant fraction of theimpedance to be measured. Figure 15 shows a cur-rent comparator bridge with this capability. Be-cause of the non-negligible lead and windingimpedance, there is some voltage e at the low-voltage terminal of the capacitor. This voltage sig-nal is inverted as shown in figure 15 and connectedto the N, winding through a capacitor C,'. The cur-rent through the unknown capacitor is:

-r1 '

e

Nx

modified for four-Figure 15. Current comparator bridgeterminal capacitance measurements.

(14)

The current reaching the N, winding is:

Is= j27rVCs-j2irfeC,'. (15)

If C,, is adjusted prior to the measurement to beequal to C. then eq (15) reduces to:

(16)

Comparing this with eq (14), the effect of thecompensation circuit has been to place the samevoltage across both the standard and unknown ca-pacitors. This is exactly what is required for leadcompensation.

Figure 16 shows the last enhancement of thebridge to be discussed. The National Institute ofStandards and Technology's current comparatorbridge has an internal range of 1000:1 (i.e., the max-imum value of Nd/NX is 1000). The external current

Figure 16. Current comparator bridgeextender.

with external range

189

V

sCS

1. =j2irf(V-e)Q,.

I� = j2 7rf(V - e) C. .

Volume 94, Number 3. May-June 1989


The current comparator bridge is quite straight-forward to use and has proven to be rugged inpractice. In order to monitor the behavior ofNIST's current comparator bridge, a check stan-dard is maintained. In this case, the check standardconsists of two high quality standard capacitors.The ratio of the two capacitors is measured quar-terly. For the last 8 years, this ratio has been stableto within about 20 ppm as can be seen in table 4.

Table 4. Check standard history

Date Capacitance ratio Date Capacitance ratio

6/80 1.000025 10/84 1.0000326/81 1.000028 4/85 1.0000419/82 1.000027 6/85 1.0000421/82 1.000027 10/85 1.0000414/82 1.000026 12/85 1.0000427/82 1.000026 1/86 1.0000419/82 1.000028 5/86 1.0000441/83 1.000030 7/86 1.0000443/83 1.000031 10/86 1.0000446/83 1.000033 2/87 1.0000448/83 1.000031 7/87 1.000046

12/83 1.000031 12/87 1.0000441/84 1.000032 4/88 1.0000405/84 1.000033 11/88 1.000046

divider and a resistor as shown. The advantage ofthis circuit is that the voltage across the resistor issmall (-0.3 V). However, because of the smallvoltage, any error voltage, c, at the low side of theresistor, R, becomes important. The in-phase cur-rent entering the Nx winding is:

aV-eR (17)

where a is the ratio of the inductive voltage di-vider (a.<I). The dissipation factor lin/I/u is thenequal to:

DF = [ aV-e J.(V-e)R2iffC,,J (18)

The effect of e can be significant at the ppm leveland needs to be eliminated. The circuit in figure 18is identical to that in figure 17 except that the inputof the inductive voltage divider isgrounded. The dissipation factor in this case isthen:

The drift can readily be attributed to the twocapacitors. The 9 ppm change between 10/84 and4/85 occurred apparently after one of the capaci-tors had been used for another purpose. An inde-pendent measurement of that capacitor verified thechange. While the use of this check standard can-not prove that the bridge is still working to theppm level, it can alert the user of changes largeenough to affect calibration results. Of course,since the two capacitive currents are largely bal-anced using stable passive components (i.e., trans-former windings), one expects that the bridgeshould be stable. It should be noted that if a trans-former winding were to become open or short cir-cuited the result would be dramatic and readilyobserved by the operator.

The situation with the dissipation factor (or in-phase current) balance is different as active compo-nents play an important role. Also, it is difficult todesign a stable dissipation factor standard to act asa check standard. This problem has been overcomeby using the circuit in figure 17. Standard capaci-tors are connected to the standard and unknownsides of the bridge. The known in-phase current isapplied with the use of the inductive voltage

DF 0 = [ (V-E)R27TfCY (19)

V

Cx

CURRENT II COMPARATOR I-

_ BRIDGE I_17-

Figure 17. Circuit for checking operation of dissipation factormeasurement of current comparator bridge.

Since &CV subtracting eq (19) from eq (18) oneobtains:

DFm = DF-DF. = a2trRC, - (20)

190



V

K CS,

X_ Cs__ ____1

I CURRENT I

i COMPARATOR IBRIDGE I

L --- 1----

Figure 18. Circuit for checking operation of dissipation factormeasurement of current comparator bridge. Input is groundedin order to measure c in eq (18).

At NIST typical values of a are 0.003, 0.0003,-0.0003, -0.003. With a 1 Mfl resistor and a1000-pF standard capacitor this enables a near fullscale test of the dissipation factor on its fourranges. Recent results are shown in table 5. Thedissipation factor values are all in units of percent.

Agreement between the calculated values in eq(20) and the corrected measurement DFm (the lasttwo columns) are well within +-0.2% of the mea-sured value. This check is performed at approxi-mately 6-month intervals.

It is further proposed that an additional checkstandard be obtained and measured quarterly.Specifically, a voltage transformer measured regu-larly at a ratio of 10:1 would give an additionalcheck on the phase angle circuitry and on the bridgewindings at something other than a 1:1 ratio.

5. Measurement Uncertainties5.1 Voltage Transformers

The records of the National Institute of Stan-dards and Technology show examples of voltagetransformers that have been calibrated at 5-yearintervals over a period of 30 to 40 years. Invariablythe original uncertainty statement covers any vari-ation in ratio correction factor and phase angle ob-served over this period of time. Voltagetransformers are often used by the client in con-junction with other equipment to measure somequantity. For example, used with a current trans-former and watthour meter, a voltage transformercan help provide a measure of the energy con-sumed by a large power transformer. Thus it is im-portant to the clients of this calibration service to

Table 5. Dissipation factor check standard

Measured Correction Corrected Theoretical

Date a (DF) (DFo) (DFm) (a/27rfRC)

7/82 0.00030.003

-0.003-0.0003

3/83 0.00030.003

-0.003-0.0003

10/83 0.00030.003

-0.003-0.0003

1/84 0.00030.003

-0.003-0.0003

5/84 0.00030.003

-0.003-0.0003

11/84 0.00030.003

-0.003-0.0003

4/85 0.00030.003

-0.003-0.0003

12/85 0.00030.003

-0.003-0.0003

11/86 0.00030.003

-0.003-0.0003

7/87 0.00030.003

-0.003-0.0003

12/87 0.00030.003

-0.003-0.0003

8/88 0.00030.003

-0.003-0.0003

11/88 0.00030.003

-0.003-0.0003

0.080030.7982

-0.7979-0.07959

0.081350.81455

-0.81475-0.08160

0.079520.7959

-0.7960-0.07965

0.081740.8148

-0.8140-0.08110

0.080900.8076

-0.8071-0.08050

0.080000.8000

-0.8000-0.08000

0.080600.8056

-0.8055-0.08050

0.080700.8076

-0.8076-0.08070

0.080290.8049

-0.8054-0.08067

0.080310.8051

-0.8055-0.08071

0.080600.8053

-0.8051-0.08040

0.080100.8037

-0.8043-0.08070

0.081420.8140

-0.8140-0.08135

0.00020.00020.00020.0002

-0.00014-0.00015-0.00015-0.00014

-0.0001-0.0001-0.0001-0.0001

0.000290.000290.000290.00029

0.00020.00020.00020.0002

0.00000.00000.00000.0000

0.00000.00000.00000.0000

-0.0001-0.0001-0.0001-0.0001

-0.00022-0.00022-0.00022-0.00022

-0.0002-0.0002-0.0002-0.0002

0.00010.00010.00010.0001

-0.0003-0.0003-0.0003-0.0003

0.00000.00000.00000.0000

0.079830.7980

-0.7981-0.07979

0.081490.8147

-0.8146-0.08146

0.079620.7960

-0.7959-0.07955

0.081450.8145

-0.8143-0.08139

0.080700.8074

-0.8073-0.08070

0.080000.8000

-0.8000-0.08000

0.080600.8056

-0.8055-0.08050

0.080800.8077

-0.8075-0.08060

0.080510.8051

-0.8052-0.08045

0.080510.8053

-0.8053-0.08051

0.080500.8052

-0.8052-0.08050

0.080400.8040

-0.8040-0.08040

0.081420.8140

-0.8140-0.08135

0.079770.7977

-0.7977-0.07977

0.081470.8147

-0.8147-0.08147

0.079590.7959

-0.7959-0.07959

0.081430.8143

-0.8143-0.08143

0.080710.8071

-0.8071-0.08071

0.079970.7997

-0.7997-0.07997

0.080590.8059

-0.8059-0.08059

0.080710.8071

-0.8071-0.08071

0.080460.8046

-0.8046-0.08046

0.080450.8045

-0.8045-0.08045

0.080390.8039

-0.8039-0.08039

0.080300.8030

-0.8030-0.08030

0.081280.8128

-0.8128-0.08128

191



obtain a meaningful uncertainty statement that re-flects the contribution the voltage transformerwould make to their total error budget.

As mentioned earlier in this paper, voltage trans-formers calibrated at NIST generally fall into twoaccuracy classes: +±0.03% uncertainty for ratiocorrection factor, ±0.3 mrad for phase angle; and±0.01% for ratio correction factor, ±0.1 mrad forphase angle. While it would be possible in somecases to report smaller uncertainties to the clientsby more thorough determinations of such parame-ters as voltage coefficients, proximity effects, andburden dependencies, the present service providesan economical way to present meaningful errorstatements to the clients and meets their needs.

The analysis of the uncertainties for the ratiocorrection factor measurements are summarized intable 6. The units are in ppm. The values in paren-theses apply to the higher accuracy voltage trans-formers described in section 2.1. The uncertaintiesfor the phase angle measurement of voltage trans-formers are the same as is shown in table 6 exceptthe units are microradians instead of ppm.

Table 6. Contributions to uncertainty

UncertaintiesRandom Systematic

Bridge measurement ±2 (±2) ±75 (±25)Secondary voltage setting ±50 (± 10)Burden setting +50 (± 10)Transformer self-heating ±75 (±20)Capacitance ratio measurement ±2 (±2) ± 5 (± 5)

To calculate the uncertainties reported to theclient, the systematic uncertainties tabulated aboveare algebraically summed and added to three timesthe root sum of squares of the random uncertain-ties. The results are shown in table 7.

Table 7. Total estimated uncertainties

Ratio correction factor ±0.03% (±0.01%)Phase Angle ±0.3 mrad (±0.1 mrad)

The values in table 6 are approximate. Sometransformers demonstrate stronger voltage depen-dences than others or stronger burden depen-dences. In some cases the values in table 7 must beadjusted for such transformers. The purpose of theabove tables is to give the users an idea of thesources of errors and how they are used to calcu-late an uncertainty statement.

Since most of the sources of uncertainty pre-sented in table 6 originate from the transformer un-der test, NIST could in principle measure a nearlyideal voltage transformer to much better accuracythan shown in table 7. Such a test would be expen-sive because of the time-consuming care thatwould be required.

5.2 Capacitors

The National Institute of Standards and Tech-nology has the capability to measure the ratio oftwo capacitors to an estimated systematic uncer-tainty of ± 1 ppm and ±+ X 10'6 + 1% of the mea-sured value for the relative dissipation factor. Thevalues of the standard capacitors used for thesecomparisons are known to +10 ppm for capaci-tance (+1 X 10-6 for dissipation factor). The ran-dom uncertainty associated with the capacitancemeasurement is ± 1 ppm and ± I X 10-6 for dissipa-tion factor. Conservatively then, NIST could cali-brate a client's capacitor to an overall uncertaintyof ±15 ppm in capacitance and ±5X 106 +1% ofthe value for dissipation factor. In general, thequoted uncertainty is always larger than this exceptfor low-voltage standard capacitors similar to thoseused at the National Institute of Standards andTechnology. (Low-voltage standard capacitors arein general calibrated elsewhere at NIST. The ser-vice described here provides higher voltage cali-bration of these same capacitors.)

The uncertainty statements for high-voltagestandard capacitors and power-factor capacitorsdepend on the stability of these devices during thecourse of the NIST measurements. The stability isinfluenced by both the voltage dependence of thedevice and self-heating (i.e., the capacitance anddissipation factors vary as the internal energy dissi-pated heats the device). Self-heating effects aremore important for power-factor capacitors. Somepower-factor capacitors demonstrate significanthysteresis effects. Assigning an uncertainty state-ment to these measurements depends on thespecific behavior of the capacitor. If self-heating isa problem the calibration report clearly mustspecify the amount of time the capacitor was ener-gized before the measurement was made. If hys-teresis effects are detected they are so noted.Because of the nature of most of these devices, thecalibration reports for capacitors usually include astatement of the form: "the estimated uncertaintiesquoted apply to the above tabulated values andshould not be construed as being indicative of thelong-term stability of the device under test." This

192



statement is also important for the compressed gasinsulated capacitors whose values might changesignificantly by handling during shipping.

The actual uncertainty quoted to the client isderived by algebraically summing the systematicuncertainties and adding three times the root meansum of squares of the random uncertainties. For thecapacitance measurement of compressed gas insu-lated capacitors, the measurement uncertainty willinclude a 20 ppm contribution because of the possi-ble 1 K variation in temperature of the NISTvoltage transformer laboratory. For power-factorcapacitors the self-heating variations will dominateambient temperature effects.

Setting Z0 equal to Ro+jXo and Z, equal to RC ±IXc,eq (25) becomes:

(26)E = E 1+ R+iXE. Eo I R. +jX. I

Taking the absolute value of both sides of eq (26),one finds that:

(27)

where it has been assumed that both R0 and X0 aremuch less than Zr so that terms of order[(R0R 0+Xox)/(Rg+X~C)]2 and higher have been ne-glected. Using eqs (21) and (26), one obtains

6. Appendix

The voltage transformer will be represented asan ideal transformer with some unknown seriesoutput impedance Zo, as shown in figure 9. Themodel has been shown to be sufficiently accurateexperimentally. The relationship between the inputvoltage Ej and the output voltage with zero burdenEo is:

(21)-= NRCF~rirO,Eo

E' = IE, e"irO [I +Ro+0 Xo Ior

Ei = I Ei I e r, [i +(Ro+JXo)(R x) ]

This can also be expressed as

I Ej Ie -arc = I sE| e -"O [.., ] .

where N is the nominal (or turns) ratio of the trans-former, RCFO is the ratio-correction factor(N x R CFo= actual ratio) at zero burden, rO is theangle by which the secondary voltage vector leadsthe primary voltage vector, and j =\/Cj. A simi-lar relationship exists between the input voltage Ejand the output voltage EC, with secondary burdenC (having impedance ZC) shown in figure 10:

E- =NRCFCe-rTE.

Both exponentials have arguments much less thanone so that discarding quadratic and higher orderterms and equating the imaginary components ofthe left and right sides of eq (10) one obtains

rI E I F I E [r XCR--XoR 1|E I C~IE-IL °+2R+XC2 (31)

or from eq (5) assuming Zo<Z 0

(22)

where RCFC is the ratio correction factor with sec-ondary burden C and r, is the corresponding phaseangle.

Equating the current through ZO and ZC in figure10, one obtains

Eo-E. -EcZo Z.

or

EoZc=ZO+Zc.

This can be rewritten in the following form:

E. E, (I + )

(23)

XoRc2XCRo (32)

The resistive and reactive components of the bur-den C can be expressed as

RC = N/-cosXcosOcand

(33)

Xc = V/icinkjsinO~, (34)

where cosO. is the power factor of the burden C.(24) From eqs (21) and (22)

(25)and

|Efj = NRCFO

|E-| = NR CFC

(35)

(36)

193

(28)

(29)

(30)

111=111 [,+RI.+EC Eo R 2 + X.,



Using eqs (27) and (33)-(36) one obtains

RCF 0 =RCFo [1 + I .- I (Rocos&c+XosinOc]. (37)

For the purposes of this discussion, it will be as-sumed that burden C (having impedance Z4) aboveis the burden for which the ratio correction factorand phase angle are to be calculated. The ratio-cor-rection factor and phase angle must be known forsome other burden T, which shall be designated ashaving impedance Z. Using eq (25) and substitut-ing burden T for burden C:

Z° = [E / -1 Zt -

or using eq (22)

Z =Z,[R CFe I{r,-r-'-RCF]/RCF 0.

(38)

where B0 =l/Z0 is the burden in fl` of theimpedance Z.. Since the second term in eq (47) rep-resents a small correction to the first and sinceRCFo is approximately equal to one, RCFO has beendropped from the second term of eq (48). Using eqs(32)-(34)

rczro- I (XOcosOc-ROsinO6).

Using eqs (43)-(46) and (49)

rc0ro+ (I Bj ) [(r1-ro cos(o,-6c)

-(RCF 1-RCF0 )sin(6,-O6)]

or

rczrZo+ tB)[(x r-ro)cos(O -03)

-3(RCF1-RCFo)sin(6,-O 6] (51)

Neglecting second order and higher terms

ZozZtER CF~-RCFo+i(Po-PW/iRCFo.

Using the facts that

Z4 = f Zt(cos6t+jsinOt)

and and 4~Z = Ro+jXo x

one finds

Roz (RCJ) [(RCF, -RCFo)cosO6

+ (r, - lro)sinOjand

XO(zaK, ) [(rO- Pt)CosOt

+ (RCF,-RCFo)sin0d.

(40)

since RCFo is approximately equal to one.Equations (48) and (51) can be used to calculate

the RCF and phase angle for some secondary bur-den C, if the ratio correction factors and phase an-gles are known at some other burden T, and at zeroburden.

(41)

(42)

Using eqs (37), (43), and (44) and the relations:

cosO, cosOt+ sin&: sinO6= cos(6, -03 (

cosO0 sinO-sin& cos0t=sin(0,-O3 (

7. Acknowledgments

The author would like to thank Oskars Peter-sons, Chief of the Electrosystems Division, whohas been the source of nearly all the author'sknowledge on the calibration of voltage transform-

43) ers and capacitors. The author would also like toacknowledge the work of Barbara Frey andRoberta Cummings who helped prepare thismanuscript. Last, but certainly not least, the authorwould like to acknowledge both the "old-timers"here at the National Institute of Standards andTechnology, who began the tradition of excellencein measurements, and the present calibration staff

45) who are attempting to carry on this tradition under46) vastly different constraints.

one finds

RCF0 RCFo+ (B; F}[(RCFt-RCFo)

cos(0,- O,)+( -I't-ro)sin(6,- 0)] (47)or

About the author: William E Anderson is a physicistin the Electrosystems Division of the NIST Center forElectronics and Electrical Engineering.

RCFczRCFo+(B-) [(RCF,-RCF.)

c o s ( 8 - O c + (1r t - r O) Si n( O t- 8 c)] (48)

194

(49)

(50)



8. References

[1] Harris, F. K., Electrical Measurements, John Wiley &Sons, New York (1966) pp. 576-577.


[3] Kusters, N. L., and Petersons, O., Trans. Commun. Elec-tron. (U.S.) CE-82 606 (1963).

[4] McGregor, M. C., Hersh, J. F., Cutkosky, R. D., Harris, F.K., and Kotter, F. R., Trans. on Instrum. (U.S.) 1-7 No. 3and 4 (1958).

[5] Hillhouse, D. L., and Peterson, A. E., IEEE Trans. In-strum. Meas. (U.S.) IM-22 406 (1973).

[6] Anderson, W. E., Davis, R. S., Petersons, O., and Moore,W. J. M., IEEE Trans. Power Appar. Syst. (U.S.) PAS-971217 (1973).

[7] IEEE Standard Requirements for Instrument Transform-ers, American National Standards Institute, ANSI/IEEEC57.13-1978 32 (1978).

[8] IEEE Standard Requirements for Instrument Transform-ers, American National Standards Institute, ANSI/IEEEC57.13-1978 45 (1978).


[10] Petersons, O., and Anderson, W. E., IEEE Trans. Instrum.Meas. (U.S.) IM-24 4 (1975).

[11] Petersons, O., A Wide-Range High-Voltage CapacitanceBridge with One PPM Accuracy, D.Sc. dissertation,School of Engineering and Applied Sciences, GeorgeWashington University, Washington, DC (1974).

195



Consensus Values, Regressions,and Weighting Factors


Robert C. Paule and An extension to the theory of consensus Consensus values are described for bothJohn Mandel values is-presented. Consensus values the case of the weighted average and

are calculated from averages obtained the weighted regression.National Institute of Standards from different sources of measurement.and Technology, Each source may have its own variabil- KGaithersburg, MD 20899 ity. For each average a weighting factor K ey words: compon ents of varianceGaithersburg, MD 20899 is calculated, consisting of contributions (within- and between-groups); consensus

from both the within- and the between- values; convergence proof; Taylor se-source variability. An iteration proce- ries; weighted average; weighted leastdure is used and calculations] details are squares regression.presented. An outline of a proof for theconvergence of the procedure is given. Accepted: February 10, 1989

2. Review

The problem of computing consensus valueswhen the errors of measurement involve both in-ternal (within group) and external (between group)components has been discussed in a number of pa-pers [1-4]. The present authors have studied thecase of a simple weighted average, as well as thatin which the measured quantity y is a linear func-tion of a known variable x.

In the present paper we extend our results tocases in which the error standard deviations arefunctions, of known form, of the x-variables. Wealso provide an outline of a proof for conver-gence of the iterative process described in refer-ence [1].

While our procedure is entirely reasonable, andresults in acceptable values, we have no mathemat-ical proof that the weights, which we calculatefrom the data, are optimal in any well-defined theo-retical sense. The problem has been recognized inthe literature [5], but we know of no attempt toprovide the proof of optimality.

If w( denotes the weight (reciprocal variance) ofa quantity, Y. then the general equation for aweighted average is:

m

Xwi Y

If X equals the average of ni results from group i(i =I to mi), then

where

2O-i 2Var (19)= Ob

nt

aO-w = the component of standard deviationwithin group i (the o-; value can beestimated from the ni results withineach group)

a-b = the component of standard deviationbetween groups.

197

1. Introduction



Then the weight co) of 1i is equal to:

1 _ 1W I = I

Var ( Y1) (i 2ni

The weight equation, xi = l/Var( 1), yields:

WiVar( 9)=1or

Var(W 1i).=I -

or

(2)

Generally, all a-values, and consequently the xivalues are unknown. The a-, can be estimated (assi) from the replicate measurements. We derive anestimate for o-b and consequently for the wi by us-ing the quantity

m

Xci(t 7 ~)2

Var (<i 5) = rm-1

which we equate to unity. Thus we have

m

i=lI

m-1 1. (3)

Equation (3) is used in "reverse fashion" to esti-mate the w1 and 7 from the sample data. This ispossible if in eq (2), the a-,, are estimated from thewithin-group variability, so that the only unknownis a-b. Note in eq (3) that ab is embedded withineach weight and therefore within 7. The esti-mated o-. and o-b can also be used to estimate thestandard deviation of the weighted average, whichis equal to 1/\/-7i. Henceforth, we use the symboloil for the sample estimate of ci1.

The same general reasoning holds for theweighted regression case. The variance of a simpleweighted average is replaced by the residual meansquare from a weighted least squares regression.For a regression with m groups and p coefficientsthe analogue of eq. (3) is

m

_ P = 1, (4)

where cxi is given by eq (2) and Rf is the fittedvalue.

We now describe the case of a weighted regres-sion with p =2. The fitted value 1i, for the ithgroup can be written as follows:

Zi=&+ Ax, (5)

k5= Y +j (Xi-X), (5')

where X is a weighted average analogous to theweighted average described by eq (1), and aand 4 are weighted least squares estimates of thecoefficients, a and j8. Again, the only unknown iso-b, which can now be estimated from sample databy use of eq (4).

A direct solution for o-b in either eq (3) or (4)would be extremely complicated since ci, Y,and X, all contain a-b. The number of terms m, inboth equations will vary depending on the numberof groups in a particular sample data set. Further-more, for the regression case, the 4 and X alsodepend on arb. Therefore an iterative solution wasproposed in reference [1]. This iterative procedureis central to the practical solution of either eq (3) or(4). In order that this paper be self-contained, webriefly review the iterative procedure for the re-gression case using eq (4) with p = 2.

3. Iteration Procedure

We define the function:

mF(sb)= nii (15 - fj)2 (m -2).

i=1(6)

In view of eqs (2) and (4), the estimate s2 of a2-

must be such that F(sb) = 0. For ease of notation letsb =1u. Start with an initial value, vo0=, and calcu-late an initial set of weights and then evaluate eq(6). In general, F(sb) will be different from zero. Itis desired to find an adjustment, do, such thatF(uo+dv)=0. Using a truncated Taylor series ex-pansion, one obtains:

F(vo + du) =FO+ (' dv = O

and dv =-Fo/(a-

Evaluating the partial derivative in this equation,one obtains:

dto=Fo/[ ci' ( _ (7)

The adjusted (new) value for v is:

New vo=Old vo+dv. (8)

This new value is now used and the procedure isiterated until dv is satisfactorily close to zero.

198



The iterative procedure is easily adapted to thecomputer. The programming steps are as follows:

1. Evaluate the s5 from the individual groups ofdata.

2. Start the iteration process with a value of v0 justslightly over zero.

3. Evaluate eq (2) to get estimates of o,1.4. Fit eq (5) by a weighted least squares regression

of Yi on Xi, and get estimates of the Y1.5. Use eq (6) to evaluate Fo. If FD<O, then stop the

iteration and set v=0. If not, continue with 6.6. Use eq (7) to evaluate dv.7. If dv is positive and small enough to justify stop-

ping, then stop. If it is positive, but is not smallenough, repeat steps 3-7 [using the new v0 fromeq (8)].

The consensus values are the final coefficients ofthe regression equation. One is also interested inthe final v =sb' value since this is needed to charac-terize the imprecision of the fit.

For the case of a weighted average [see eq (1)]the above iteration steps are the same, except thatin place of step 4, 7 is calculated by eq (1), andsteps 5 and 6 use 7 in place of 1i, and unity isused for the p value. The authors have frequentlyused this procedure for the evaluation of StandardReference Materials [6].

desired between-group component of variance isthus:

s32 =v'(c+dXi)2 . (10)

The weights estimated by eq (2) would then be:

oi, =1

(2')( S+v'(c+dXi)2)ft

This newly defined weight can be used in the itera-tion process. The iteration process proceeds as be-fore, but now the adjustable iteration parameter v'is the multiplier needed to make eq (4) true, that is,to make it consistent with the sample data sets. Thedenominator of eq (7) which is used in iterationstep 6 for calculating do, needs to be slightly modi-fied since the derivative of F with respect to v nowcontains the function described by eq (10).

do' = (7')m3 0? (C i +dX) 2( Y-i )2i=l

All other steps in the iteration process are thesame. The final between-group components ofvariance will be described by eq (10).

4. Theoretical Extensions

Once one recognizes the between- as well as thewithin-group component of variance in the evalua-tion of consensus values, one can begin to considerfunctional forms for these components. The within-group component can be of any form, and can beeasily handled since the appropriate sample valuesof the component are simply substituted into theweights described by eq (2). Thereafter, this com-ponent does not affect the iteration procedure. Seefor example reference [7], where the within com-ponent of variance refers to a Poisson process. Thebetween-group component, however, affects the it-eration procedure and must be handled more care-fully. As an example, consider the case where thebetween-group component of standard deviation isbelieved to be a linear function of the level of Xi:

a-bzY+SX. - (9)

Let us assume that we have preliminary esti-mates, c and d for the y and 8 coefficients. Supposefurther that we wish to adjust the estimated valueof the variance by a fixed scale factor, say v'. The

5. Example

The iteration process will be used to fit the dataof table 1 to a straight line. These are real datataken from a large interlaboratory study for the de-termination of oxygen in silicon wafers.

Table 1. Data used in example of iteration process

X Y, Y2 Y3

0.806 2.83 2.851.429 4.62 5.35 5.011.882 6.89 6.662.140 7.56 7.672.256 7.94 7.902.279 8.42 8.122.814 10.04 9.70 10.172.957 10.34 10.052.961 11.09 11.073.108 11.63 11.693.124 10.87 11.013.403 12.40 12.223.466 11.94 12.17 12.923.530 12.63 12.413.543 12.98 13.273.724 12.95 12.563.836 13.07 13.69 13.563.902 14.54 14.194.280 15.59 16.244.770 16.62 16.59

199



A preliminary examination of the data indicatesthat the within error has a constant standard devia-tion and that the between error has a standard devi-ation proportional to X. Thus, the error structurefor the example is given by the equation:

1

= i+ )

The figures support the assumpions made con-cerning the nature of the within and betweenerrors.

6. Appendix6.1 Sketch of Proof of Convergence

The general functional form of the F of eq (6) isas shown in figure 2a or 2b. It is because of thenature of these forms that convergence always oc-curs.

where v now stands for the product v'd2 of eq (2').From the replicates, the pooled within standard

deviation is readily calculated to be 0.265. The iter-ation process then yields the following results

(a)

fi = -0.0833+3.6085 XiF (SD)

ISwithin = 0.265

Sbetween = 0.0827Xi.

Figures la and lb show, respectively, the standarddeviations within, and the residuals ( i - i), asfunctions of Xi.

(a) 0.6

0.3

STO 0.

0EV

0.i

O.o

O.C

F (52)

* * *

***

*

r

± 2 3 4 5

LEVEL

0.Br

0.4_*

RESIDUAL 0.0

-O.4_

1 2 3 4 5

LEVEL

Figure 1. (a) Standard deviations within as a function of X(b) residuals of a function of X.

Figure 2. (a) Fas a function of sb where Fhas a positive root fors52 (b) F as a function of s5, where F does not have a positive rootfor sb.

If the functional form of F is as shown in figure2a, the previously described iterative procedure isused to determine the sb2 satisfying the equationF(sb2)=O. If an initial estimate of sb' is chosen that isvery slightly above zero, then convergence of theiteration process always occurs. This is a result ofthe fact that the first derivative of the function Fwith respect to sb2 is negative, and the secondderivative is positive. This means that each itera-tion will undershoot, since the iteration process ex-trapolates the slope of the F curve at the current sb2estimate to the F=0 value. Since each new itera-tion estimate of sb2 is the abscissa value of the inter-

200

O)

0

0

52

I - - - -,

0

S2

X *- - 1L_ - - _W - - - - -

1, * * 1



section of the tangent line with the F = 0 horizontalline, the iteration process will never overshoot andconvergence is obtained.

If the form is that of figure 2b, then there will beno positive solutions for the sb2 that is associatedwith the function F. This represents a situationwhere the variability between the sample groups isless than that expected from the variability withinthe sample groups. For this situation Fo is negativeand sb2 is set to zero (see iteration step 5).

The proof regarding the signs of the first andsecond derivatives of F with respect to sb follows.The simple regression case will be considered, withthe sb2 = v being constant. (The extension to thevariable sb' case is straightforward.)

6.2 Proof That the First Derivative of Fis Negative

An examination of eqs (1), (2), (5'), and (6) showshi and i to be functions of sb2. Equation 5' alsoindicates that , X , and /3 are functions of sb2.We start with the first derivative of the F of eq (6)

6.3 Proof That the Second Derivative of Fis Positive

The evaluation of the second derivative is in-volved and only an outline of the steps is presented.

d2Fd [d m

dd- o'- d i -o ) dod2 - dv _ >g2J

dv

(A2T)

where

d(_-Y)_ dY+ dXdvo ddo do

(A3)-(Xi-X ) dd.

Evaluation of the first two derivatives on the r.h.s.of eq (A3) yields:

dX _ A, Xi (Xj- ) and2dX

The derivative of ci will frequently be encounteredin the following material. At this point, it will beconvenient to note its value:

(A4)dv . W

where W=Y coi.

Evaluation of the last derivative of eq (A3) yields:

dwi 2

d =-i

Continuing, and making use of eq (5'):

E 4 i (X]-X ) _

I 1i(Xi_ X)2

dF m 2d -v I C(i (-1 X!))i= I

(Al)

The last two terms of eq (Al) each contain summa-tions that are equal to zero, so these terms dropout. Next, an examination of the remaining termshows that each product is a positive square, andthat the summation is preceded by a minus sign.Thus, the first derivative is negative.

d4 =

dov

. dco , (AS')

where

201

dFdo dv

+ 2 Z wj� (ki - kj)',1.

( M a)i ( Ri_ f,d I )2 - (m -2)i=1

2(!�d-� wX, ( ki - R,) .1.

. , = .

= (XI- XXY'- Y) Ido),

dZ wi (Xi - X )'i

O)i (XI - 1 ) _f,4

d I -1M �: O)i (Xi_ k )2

t

Z' do),t=1

dj coi (X, - k ) :P,i



The first term on the r.h.s. is the "total" weightedsum of squares of Z. The second term is theweighted sum of squares for the regression of Z onX. Therefore the difference between the two termsis a "residual" sum of squares:

&F = 2 jdY Z -(zj Z- )2, (A2)

Reassembling eq (A5'):

(A5)d# _du

&)Zi (Xi-X )2

At this point the running index t can be conve-niently changed back to the index i. Finally, thesubstitution of eq (A5) into (A3), and then eq (A3)into (A2') yields:

Tv2 2 [ ii (B_ fi)2-[ (2i2 y-Ri -

W

(xi - X -RI )2i (A2''

This second derivative eq (A2'') is a residualweighted sum of squares from regression. To seethis, let

Zi=&i(:Pi- i) (A6)

and substitute into eq (A2' '):

d 2 [ S@kxi -X)Zi]2]

2-~ =2 °)w( )2

= 2 [X;i (Z,- 2 )2-__I

where 2i is the fitted value of Z, in a weightedregression of Z on X. Thus, the second derivativeis positive for cx >0. The iteration process there-fore will never overshoot, and convergence is al-ways assured.

6.4 Extensions

The extension of the convergence proof to thevariable sb2 case is very similar to that given above.Two basic changes are needed. These changes,which introduce a function of Xi, are in the deriva-tive of o)i and in the definition of Z,.

- g(Xi) wi

and

Zi = O)i g(Xi) ( Z - ki) -

Equation (10) represents an example of a variablesb'. For that case,

g(Xi) = (c+dXi)2.

The reader may note that the new Zi contains Xi,and that Zi is regressed on Xi. The argument doesnot require that this regression "make sense", onlythat the sum of squares can be partitioned by a re-gression process. Again, convergence is obtained.

The weighted average is a special and simple ap-plication of the weighted regression case.

7. Acknowledgments

We wish to thank Miss Alexandra Patmanidou, agraduate student at Johns Hopkins University. Shenoted that whenever a negative correction in theiteration process is obtained, the process should beterminated and the between-group component ofvariance set to zero.

202

and

- (XI - ± ) GPI - -fl)Y, Wi (Xi - 1)2i

Doi (X!-.k) -i

dI (,), (X,_ V)2 _i

2 (X,_ f )(.?Y. cot _ t)t

I C,)Xxi - X )Z, 2r. I

Y, Coi(xi - X )2i



About the authors: Robert C Paule is a physical sci-entist and John Mandel is a consulting statistician atthe National Measurement Laboratory of NIST:

8. References

[1] Paule, R. C., and Mandel, J., Consensus Values and Weight-ing Factors, J. Res. Natl. Bur. Stand. (U.S.) 87, 377 (1982).

[2] Mandel, J., and Paule, R. C., Interlaboratory Evaluation of aMaterial with Unequal Numbers of Replicates, Anal. Chem.42, 1194 (1970), and Correction, Anal. Chem. 43, 1287(1971).

[3] Cohen, E. R., Determining the Best Numerical Values ofthe Fundamental Physical Constants, Proceedings of the In-ternational School of Physics (Enrico Fermi), CourseLXVIII (1980) pp. 581-619.

[4] Birge, R. T., Probable Values of the General Physical Con-stants, Rev. Mod. Phys. 1, 1 (1929).

[5] Cochran, W. G., The Combination of Estimates from Dif-ferent Experiments, Biometrics 10, 101 (1954).

[6] NIST Standard Reference Materials 1563, 1596, and 1647a.[7] Currie, L. A., The Limit of Precision in Nuclear and Ana-

lytical Chemistry, Nucl. Instr. Meth. 100, 387 (1972).

203



News Briefs

Developments

PARTICIPANTS WANTED FOR OSI, ISDNSECURITY PROGRAMOutside participants are invited by NIST to join ina cooperative research program relating to securityand management of computer networks that usethe Open Systems Interconnection (OSI) architec-ture or Integrated Services Digital Network(ISDN) communications services. NIST is lookingfor participants to provide funding, equipment,and/or staff. A major goal of the program is toexpedite the development and commercialavailability of OSI and ISDN security products. Aspart of the program, NIST will provide a facility todefine, develop, and test systems for a range oftelecommunications, network management, and se-curity services in a distributed information process-ing environment. For further details, write toNIST, B151 Technology Bldg., Gaithersburg, MD20899, Attn: Integrated OSI, ISDN, and SecurityProgram.

VENDORS PROVIDE WORKSTATIONS FOROSI SECURITY WORKThe Open Systems Interconnection (OSI) stan-dards being adopted by both government and in-dustry make it possible to interconnect computersystems manufactured by different vendors for datacommunications through networks. NIST, throughits National Computer Systems Laboratory, hasplayed a major role in developing these standards.Now NIST is working to develop ways to makesystems that are secure as well as open. To helpNIST develop security protocols for OSI, threemajor U.S. computer vendors have loaned equip-ment to the NIST OSI Security Laboratory. NIST

will use this equipment to perform research and todevelop specifications that can be used as the basisof Federal Information Processing Standards.(FIPS are developed by NIST for use by theFederal Government.)

STANDARD FOR INTERCHANGINGDOCUMENTS PROPOSEDTrying to interchange documents among differentdocument or text processing systems such as desk-top publishing systems can be a frustrating, some-times impossible, experience. A new FederalInformation Processing Standard (FIPS) formallyknown as Document Application Profile (DAP)for the Office Document Architecture (ODA) andInterchange Format Standard is being proposed byNIST and should make the process easier. (FIPSare developed by NIST for use by the federal gov-ernment.) The profile was developed by partici-pants, primarily vendors and users of computernetworks, of the long-running NIST Workshop forImplementors of Open Systems Interconnection,and is based on an international voluntary industrystandard. A copy of the proposed standard may beobtained from the Standards Coordinator (ADP),NIST, B64 Technology Bldg., Gaithersburg, MD20899.

MERCURY ION LASER-COOLED TO LIMITScientists at NIST's Time and Frequency Division,Boulder, CO, have succeeded for the first time inlaser-cooling a bound atomic ion to its fundamentallimit. "We pushed the atom into the ground state ofits confining well. That's the end of cooling for abound particle," says project leader David J.Wineland. Their finding is important for spec-troscopy, a study of the nature of matter throughvarious radiations it emits. One result may be thedevelopment of a highly sensitive spectrum ana-lyzer. A report on their work appears in the Jan.23, 1989, issue of Physical Review Letters. They

205



shined laser light on a mercury ion sideband fre-quency generated by the Doppler effect associatedwith thermal motion. The result was to reduce theion's kinetic energy, limit its movement, andsharpen its spectral features. The ion was confinedin a radio frequency trap.

NIST TO STUDY NEW POLYMER RESINSFOR INDUSTRYE. I. du Pont de Nemours & Co., Inc., has estab-lished a cooperative program at NIST to study thechemical and physical behavior of newly devel-oped methacrylate macromonomers, and to deter-mine how well the new materials blend with otherresins. The methacrylate macromonomers can becopolymerized by conventional methods or by anelectron beam process. Du Pont will supply NISTscientists with specially prepared samples of thematerials for examination by several analyticalmethods. The primary research tools will be small-angle neutron scattering (SANS), a techniquewhere low-energy neutrons from the NISTresearch reactor are used to characterize the struc-tures of materials on a nanometer scale, andsmall-angle x-ray scattering (SAXS) whichprovides information on the phase separationof molecules and polymer chain networks. TheNIST Research Associate Program provides anopportunity for scientists from industry, technicalsocieties, universities, and other organizations toconduct cooperative research on programs ofmutual interest.

NEW BUILDING CRITERIA FOR PRISONSOvercrowded and aging, deteriorating prisonshave produced a rapid increase in the United Statesin construction of new facilities. Over $1.5 billionwas spent to build new jails and prisons in both1985 and 1986. But little information exists specifi-cally for the special materials, equipment, and sys-tems used in these facilities. As a result, manycorrectional agencies have experienced equipmentand system performance problems leading to ex-pensive retrofits, repairs, or other fixes. In a projectfor the U.S. Department of Justice, researchers inthe NIST Center for Building Technology havedeveloped preliminary performance criteria to helpin the selection, application, and maintenance ofbuilding materials, equipment, and systems. A re-port covers criteria such as choosing a site, select-ing appropriate fencing and intrusion detectionsystems, and developing structural systems. Prelim-inary Performance Criteria for Building Materials,

Equipment and Systems Used in Detention andCorrectional Facilities (NISTIR 89-4027) is avail-able from the National Technical Information Ser-vice, Springfield, VA 22161 for $21.95 prepaid.Order by PB #89-1148514.

NIST STUDYING UNINTENTIONAL EEDFIRINGElectroexplosive devices (EEDs) are electricallyfired explosive initiators used in a wide variety ofapplications from triggering air bags in cars to sep-arating stages in rockets. EEDs are susceptible tounintentional triggering by electromagnetic (EM)fields such as those from local radio transmitters.To help cope with this problem, NIST engineersand statisticians have studied the statistical proba-bility of an EED firing when excited by an electro-magnetic pulse of given width and amplitude. TheNIST researchers have produced probability plots,called firing likelihood plots, which should assistelectrical engineers when designing applicationsfor EEDs. Methods of measuring the time and en-ergy required to fire an EED with a single currentpulse also are given. For a copy of this study, con-tact Fred McGehan, NIST, Division 360.2, Boul-der, CO 80303.

CD-ROM SPEECH DATABASE AVAILABLEAs part of its speech recognition research to helpcomputers become better listeners, researchers inthe NIST National Computer Systems Laboratoryhave produced the first speech database in thiscountry in CD-ROM (compact disc-read onlymemory) format. The database consists of digitizedspeech data for 420 talkers speaking 4,200 sen-tences. NIST has been working with private indus-try and the Defense Advanced Research ProjectsAgency to develop ways such as this database tomeasure the performance of speech recognitionsystems. A limited number of discs are availablefrom David Pallett, NIST, A216 TechnologyBldg., Gaithersburg, MD 20899; telephone: 301/975-2935.

NIST REPORT SUMMARIZES INVENTIONSPROGRAMA portable pothole patcher, a new composite mate-rial made of high-strength fibers, a new process forcontinuous casting of steel cylinders, and alightweight aluminum cylinder which makes itpractical to use natural gas as a vehicle fuel areamong the 400-plus inventions which have re-ceived support from the federal Energy-Related

206



Inventions Program. The program, which began in1975, is conducted jointly by NIST and the U.S.Department of Energy and aims at helping inven-tors get their ideas from the workshop to the mar-ketplace. NIST provides, at no cost to theinventor, evaluations of energy-related inventionsand recommends those it considers promising toDoE. In turn, DoE can provide financial supportor help in marketing an inventor's idea. A new re-port is available which describes the program aswell as the inventions which have been recom-mended for DoE support. Energy Related Inven-tions Program: A Joint Program of the Departmentof Energy and the National Institute of Standardsand Technology Status Report (NISTIR 88-4005)can be ordered from the National Technical Infor-mation Service, Springfield, VA 22161, for $36.95prepaid. Order by PB #89-141154.

DIAMOND FILMS PRODUCE NEW GEMSAncient alchemists did not succeed in changingbase metals into gold, but scientists today are ableto produce synthetic diamonds from common or-ganic materials. With modern technology, hydro-carbon vapors mixed with hydrogen can be madeto deposit a film of diamond on hot objects. Materi-als scientists at NIST are developing the measure-ment information that industry needs to producediamond films with many of the properties of natu-ral diamond. The physical and chemical propertiesof diamond make it a highly desirable material foraerospace products, electronics, and industrialequipment. At NIST, the scientists are evaluatingthe production of diamond films by a hot- filament,chemical vapor deposition (CVD) method. Otherstudies include measuring the thermal conductivityof diamond and developing a better understandingof how defects such as nitrogen impurities andcrystal lattice vacancies or voids can affect the per-formance of diamond films. For further informa-tion on the diamond film research program,contact Dr. Albert Feldman, NIST, A329 Materi-als Bldg., Gaithersburg, MD 20899; telephone:301/975-5740.

NEW WAY TO EVALUATE PROTECTIVECOATINGS ON METALSResearchers at the NIST Center for BuildingTechnology have developed a fast, reliable tech-nique for evaluating the performance of organiccoatings used for controlling metallic corrosion. Inaddition to being a threat to the safety and reliabil-ity of structures and products, metallic corrosion

annually costs the United States an estimated $160billion. About one-fourth of this cost is for paints,platings, or other surface coatings used to combatcorrosion. While other evaluation methods areavailable, they often are time consuming, some-times taking months, or may require expensiveequipment. The new NIST electrochemical tech-nique is quick (15 minutes to several hours depend-ing on the coating), reproducible, and causes verylittle perturbation to the coating. In addition, thetesting procedure is simple and uses commonlyavailable instrumentation. The method also is be-lieved to have other applications such as screeningnew coatings and corrosion inhibitors and evaluat-ing the effect of new surface preparation tech-niques on the performance of the coating and metalsystem. A report, An Electrochemical Techniquefor Rapidly Evaluating Protective Coatings onMetals (Technical Note 1253), is available from theSuperintendent of Documents, U.S. GovernmentPrinting Office, Washington, DC 20402, for $1.50prepaid. Order by stock no. 003-003-02910-6.

ABSTRACTS OF RECENT PUBLICATIONSAVAILABLEResearchers in high-temperature superconductivitywill be interested in a recent NIST publication thatlists abstracts of 61 NIST papers in this field be-tween March 1987 and May 1988. Topics includecritical current, crystal structure, electrical con-tacts, Josephson effect, and magnetic measure-ments. High-Temperature Superconductivity:Abstracts of NIST Publications, 1987-1988 (SP759) is available from the Superintendent of Docu-ments, U.S. Government Printing Office, Washing-ton, DC 20402. Order by stock number003-003-02902-5 for $2 prepaid.

COLLECTED PAPERS ON ION RESEARCHSome of the world's most advanced research onlaser cooling and storage of atomic ions is per-formed at NIST's Boulder, CO laboratories. A re-cent publication, Trapped Ions and Laser CoolingII (TN 1324), reproduces a number of papers ofthis Time and Frequency Division research groupand is a companion to an earlier collection of pa-pers. Subjects covered include spectroscopy andfrequency standards, quantum jumps, and nonneu-tral plasma studies. The publication is availablefrom the Superintendent of Documents, U.S.Government Printing Office, Washington, DC20402. Order by stock number 003-003-02918-1 for$10 prepaid.

207



CHEMICAL STRUCTURE OF DNA DAMAGEUNCOVEREDFor the first time, researchers have determined thechemical structure of a major type of DNA dam-age caused by oxygen-derived free radicals. Suchfree radicals, highly reactive groups of atoms withan unpaired electron and very short lifetimes, havebeen linked to cancer, among other ailments. Themethods used by the NIST researchers should helpscientists study this type of DNA damage in livingcells and gain deeper understanding of its biologi-cal effects. The NIST researchers uncovered thestructure of hydroxyl radical-induced DNA-protein cross-links, which is damage caused whenDNA forms a chemical bond with proteins insidethe cell nucleus. This damage to DNA eventuallycauses chemical changes in the cell that result, forexample, in altered proteins.

NIST, NSF PLAN JOINT NEUTRONRESEARCH FACILITYNIST and the National Science Foundation (NSF)have announced plans to develop a Center forHigh Resolution Neutron Scattering (CHRNS) forresearch in chemistry, physics, biology, and materi-als science. The center will include two state-of-the-art instruments to be built at the recentlydedicated NIST Cold Neutron Research Facility,with funding from NSF. The new instruments willprobe the microstructure and atomic and molecu-lar dynamics of a wide range of materials, and becompetitive with the best such facilities in theworld. The center will be managed as a nationalfacility, open to qualified users from universities,industries, government agencies, and nonprofit or-ganizations both U.S. and foreign. Proposals for re-search time will be evaluated on the basis ofscientific merit by a program advisory committee.The first instrument is expected to be complete andoperating within 2 years.

MEASURING HIGH-TEMPERATURESUPERCONDUCTORSNIST researchers have developed a novel appara-tus for variable-temperature measurements of high-temperature superconductors. Termed a cryogenicbathysphere, it can rapidly (in 10 minutes) test su-perconductors over a range of temperatures from300-4 K, or it can be used to stabilize the tempera-ture at a given value. The tiny (3-centimeter-width)device has no moving parts and can be used incompact spaces such as shipping Dewars andsmall-bore high-field magnets. The apparatus has

been tested successfully in liquid helium and liquidnitrogen by measuring the resistance-versus-tem-perature curves of several superconductors. A pa-per describing a prototype device is available fromFred McGehan, Division 360.2, NIST, Boulder,CO 80303.

STEEL IN FRACTURE TEST SETS U.S.RECORDJust how tough is a 6-inch thick piece of steel?NIST researchers performed a series of tests onthick steel plates to learn more about how cracksmight travel and stop in large pieces of metal. TheNIST research team learned that it took 5.94 mil-lion pounds of force in tension pulling force tofracture a 6-inch thick, 40-inch wide new steelplate: a U.S. record for fracture tests. All informa-tion so far from the tests indicates that the fracturetoughness of these steels at the point of crack arrestsignificantly exceeds minimum values used in appli-cable design codes and standards. This new infor-mation may be used to revise industry codes andstandards and will be useful in the design of alltypes of steel structures including bridges, ships,and buildings, as well as for equipment in industrialplants and utilities.

NIST INVITES VENDORS FOR GOSIPEVALUATION PROJECTThe National Computer Systems Laboratory atNIST is developing guidelines to help users evalu-ate different implementations of GOSIP (Govern-ment Open Systems Interconnection Profile)applications. To help expedite the project, vendorsare invited to lend to NIST software and hardwarewhich implements Message Handling Systems(MHS) and File Transfer, Access, and Manage-ment (FTAM) applications. Currently, GOSIPsupports both applications. GOSIP was approvedlast fall as a Federal Informaton Processing Stan-dard. (FIPS are developed by NIST for use by theFederal Government.) The standard defines a com-mon set of data communication protocols whichenables computer systems developed by differentvendors to communicate and enables the users ofdifferent applications on these systems to exchangeinformation. For further information on the evalua-tion project, contact Steve Trus, NIST, B225Technology Bldg., Gaithersburg, MD 20899; tele-phone: 301/975-3617.

208



NATIONAL EARTHQUAKE AWARENESSWEEKThe catastrophic earthquake which struck Arme-nia last December is a reminder of the threat tolives, property, economic activity, and national se-curity posed by earthquakes. To promote aware-ness of earthquakes, Congress designated the firstweek in April 1989 as National Earthquake Aware-ness Week. But, while earthquakes are an in-evitable hazard, they are not an inevitable disaster.Structures can be made earthquake resistant. Aspart of the National Earthquake Hazards Reduc-tion Program, NIST is working to improve theperformance of buildings and other structures sub-jected to earthquakes. NIST provides research andtechnical support for the development of improvedseismic design and construction practices. TheNIST program includes both laboratory researchand experiments and post-disaster investigations. Afact sheet on the program is available from theNIST Public Information Division, Jan Kosko,A909 Administration Bldg., Gaithersburg, MD20899; telephone 301/975-2762.

FIBER-MATRIX INTERFACE PROPERTIESVIA AN INSTRUMENTED INDENTERTECHNIQUEA commercial microhardness tester has been in-strumented to provide measurement of the fiber-matrix interface properties of a ceramic matrixcomposite. The technique uses a strain gage loadcell and a pair of capacitance probes to directlydetermine the force on and displacement of a fiberin the matrix. The system can be used to determineboth fiber-matrix debond strengths and interfacialfrictional stresses. These properties are importantfor determining the occurrence of matrix micro-cracking and, hence strain limits in the compositeas it is loaded in service.

Loads up to 1 kg can be used and displacementsup to 25 micrometers can be obtained with this sys-tem. Sample sizes are typically 5X6X0.5-3 mmand require at least one well-polished surface. Thesystem can be used to perform either indentationpush-in or indentation push-out tests on a variety ofcomposites. The push-in test requires only that thefiber be pushed into the matrix while the push-outtest pushes the fiber out through the other side ofthe composite. Examples of typical materials char-acterized to date include CVI SiC/SiC fiber,borosilicate glass/SiC monofilament, and lithiumaluminosilicate glass-ceramic/SiC fiber.

NIST COLLABORATION WITH OAK RIDGENATIONAL LABORATORY ON NEUTRONSTANDARDSThe combined resources and measurement capabil-ities of NIST and the Oak Ridge National Labora-tory are being utilized to obtain a high-precisionmeasurement of the neutron interaction with theboron isotope with an atomic mass of 10. This in-teraction is one of the most widely used standardsin the determination of neutron flux. Because neu-trons are uncharged, they cannot be measured di-rectly. The neutron interaction with the boronisotope is easy to utilize because a gamma ray isreleased as a result of the interaction. The easilydetected gamma ray can then be used to indicatethe intensity of the neutron flux. Large uncertain-ties are now associated with the measurement forneutrons having energies above 500,000 electron-volts. NIST will take its carefully calibrated neu-tron flux detector to Oak Ridge where it will beused to calibrate the neutron flux generated by theOak Ridge Electron Linear Accelerator. The cali-brated neutron flux can then be used to make aprecise measurement of the boron interaction.

MAJOR NIST COLLABORATION TO STUDYNOVEL MAGNETIC SYSTEMSNIST Scientists have been participating in a multi-institutional effort at the National SynchrotronLight Source (NSLS) to study novel magnetic sys-tems created in situ by molecular beam epitaxy(MBE). To date, the work has concentrated on es-tablishing a spin-polarized, angle- and energy-re-solved, photoemission apparatus on the U-5beamiline of the uv storage ring.

This project is unique in a number of ways: (1)The research team consists of 10 principal investi-gators from eight institutions nationwide. Theseare national labs (NIST, Argonne, NRL, andNSLS), universities (Rice, U. Texas at Austin, andNorthwestern), and an industrial lab (AT&T Bell).(2) The beamline is the only spin polarized photoe-mission facility in the United States and the onlyone in the world with a movable spin analyzer topermit angular studies. (3) It is one of very fewbeamlines in the United States to have an MBEcapability. (4) The experiment is being carried outon the highest flux uv beamline at NSLS. It isbased on an undulator currently installed in astraight section of the ring. During the next 2 yearsthis same group, acting as an Insertion DeviceTeam, will install a new state-of-the-art undulatornow under construction.

209



The beamline has recently been used to take spinpolarized photoemission data. The object of futureexperiments will be to observe the magnetic prop-erties, e.g., anisotropy, Curie temperature, rema-nence magnetization, etc., as a function of layerthickness and growth methodology, and to corre-late them with the spin-dependent electronic struc-ture we measure. We expect this facility to greatlyextend our ability to study new and interestingmagnetic systems.

ATOMIC POSITIONS FROM X-RAYSTANDING WAVESRecent experiments carried out at the Cornell HighEnergy Synchrotron Source (CHESS) by NISTand CHESS scientists have demonstrated a newmethod of determining adsorbate positions on crys-tals with improved accuracy. x-ray standing wavesthat occur in the process of diffraction from thecrystal substrate excite the adsorbed atoms, whichfluoresce. Slight changes in crystal alignmenttranslate the standing waves and give the atomicregistration.

The novel element in these experiments is theapplication of the technique to the geometry ofglancing incidence. This gives accurate position in-formation parallel to the surface and allows for thepossibility that the measurement can be made notonly at a surface but at an interface between twodifferent crystalline materials.

PATENT APPLICATION ON NEW X-RAYDIFFRACtION DEVICEScientists from the Surface Science Division andthe Semiconductor Electronics Division have re-cently applied for a patent on a new method ofdetecting the Bragg diffraction condition of x raysincident on a crystal. The determination of diffrac-tion, an extremely common phenomenon in physicsexperiments, has always been made by monitoringthe existence and intensity of the diffracted beam.The novel approach takes advantage of largechanges in intensity and penetration depth of thex-ray fields inside the diffracting crystal itself. Asemiconductor detector is actually implanted insidethe diffracting crystal and responds to smallamounts of energy that are always absorbed in thediffraction process. The invention has potential ap-plications in such areas as medical angiography.

MAGNETIC THIN FILMS WITH LARGEPERPENDICULAR MOMENTSA class of ultrathin magnetic films has been discov-ered that is ferromagnetic at room temperature andhas both a large magnetic moment (high-spin state)and a large perpendicular magnetic anisotropyholding the moment normal to the thin film plane.Such perpendicular anisotropies are a muchsought-after property because of their potential forultrahigh-density information storage in advancedmagnetic-recording media.

A key ingredient in achieving these unusually fa-vorable properties has been the ability to optimizethe epitaxial growth conditions using the NIST-de-veloped technique of XPS forward-scattering crys-tallography. Optimum conditions consist ofdeposition of Fe on Cu at cryogenic temperatures,annealing at 350 K, and deposition of Cu on the Feat room temperature. This process can be repeatedcyclically to produce Cu-Fe superlattices, whichhave been found to retain the favorable magneticproperties.

This work is a result of collaborative researchbetween scientists at NIST, the Simon Fraser Uni-versity in Canada, and Cambridge University inEngland. The work highlights the important poten-tial of such artificially structured materials both forimproved understanding of the basic physics ofmagnetism and as useful novel materials for impor-tant technologies.

Calibration Services

INDUSTRY HELP REQUESTED ON COAXIALCONNECTORSNIST is interested in learning from the microwaveindustry of new metrology-grade precision coaxialconnectors developed by particular companies.Companies desiring NIST to provide calibrationservices for components with new connectors areencouraged to furnish NIST with appropriatecheck standards, air line impedance standards, andtest port adapters to use with NIST six-port mea-surement systems. NIST can then provide mea-surements traceable to these standards with anuncertainty determined by NIST. The instituteprovides calibration services for passive devicesover the frequency range from 50 MHz to 26.5GHz with plans to expand to 50 GHz. The stan-dards used by NIST to support coaxial impedance

210



measurements in this frequency range are lengthsof precision air-dielectric transmission lines. In de-veloping these standards, NIST tries to avoid usingcomponents or designs available from only onemanufacturer; NIST does not wish to favor or be-come dependent on any one manufacturer. Formore information, contact Ramon L. Jesch, NIST,Division 723.01, Boulder, CO 80303; telephone:303/497-3496.

NEW CALIBRATION SERVICES USERSGUIDE AVAILABLEThe new NIST Calibration Services Users Guide1989-1990 Special Publication 250 (SP 250) lists thecalibration services, special test services, and mea-surement assurance programs (MAPs) availablefrom NIST. The physical measurement servicesare designed to help the makers and users of preci-sion measurements achieve the highest possible lev-els of quality and productivity. The hundreds ofservices described in the guide are the most accu-rate calibrations of their type available in theUnited States. They directly link a customer's pre-cision equipment or transfer standards to nationalmeasurement standards. The calibrations and spe-cial tests include NIST services that check, adjust,or characterize instruments, devices, and sets ofstandards. The MAPs are quality control programsfor calibrating a customer's entire measurementsystem. The guide also lists NIST technical expertswho may be contacted for information on servicesand measurement problems. For information on theNIST measurement services program, or to obtaina copy of SP 250, contact the Office of PhysicalMeasurement Services, NIST, B362 Physics Bldg.,Gaithersburg, MD 20899; telephone: 301/975-2005.

NCSL AD HOC COMMITTEE 91.3 ON THECHANGE OF THE TEMPERATURE SCALEThe National Conference of Standards Laborato-ries (NCSL) ad hoc Committee on the Change ofthe Temperature Scale met in open session on Jan-uary 26, 1989, in Anaheim, CA, during the Mea-surement Science Conference. The committee wasformed last year to facilitate the change from theInternational Practical Temperature Scale of 1968(IPTS-68) to the new scale that will be imple-mented on January 1, 1990, and that will be knownas the International Temperature Scale of 1990(ITS-90). The meeting was well attended, both bycommittee members and by guests. Standards labo-ratories and instrument manufacturers were well

represented. The latest draft of the ITS-90 and theimplications of this new scale were discussed inconsiderable detail. Although the scale is not yet inits final form, it will be complete by September, atwhich time the CCT will meet and recommend itto the International Committee of Weights andMeasures (CIPM). The CIPM will then adopt it inOctober. Those involved in temperature measure-ments are being kept informed of changes in thescale, and will be told what steps they should taketo either implement the scale or ensure that theirmeasurements are on the ITS-90, beginning in1990.

Standard Reference Materials

IMPROVING LEAD-IN-FUEL ANALYSES ISAIM OF MATERIALSElevated lead levels in the environment could becaused by a number of sources, but one of the chiefculprits may still be the alkyl lead used as an addi-tive in gasoline, according to some reports. Be-cause the Environmental Protection Agencyregulates lead content in fuel emissions as part ofair quality standards, the petroleum industry needsaccurate measurement techniques to ensure com-pliance with federal emission limits. To aid indus-try in calibrating the instruments that analyze fuelsamples and to help confirm the accuracy of mea-surement techniques, NIST has developed four dif-ferent standard reference materials (SRMs).Individually, the SRMs consist of gasoline-like fuelin vials, with each SRM representing a differentcertified lead level (.0297, .0506, .0733, and 2.045grams per gallon of lead, respectively). By analyz-ing the SRM fuel the same way as a fuel sample, achemist can gauge how well analytical instrumentsand techniques are working at four differentlead levels. The new SRMs, numbered 2712through 2715, are available for $110 each from theOffice of Standard Reference Materials, NIST,B311 Chemistry Building, Gaithersburg, MD20899; telephone: 301/975-6776.

NEW MATERIALS CAN HELP GAUGE COALSULFUR CONTENTBecause sulfur emissions from coal-fired industrialplants are regulated by environmental agencies, itis important for coal and utility companies to know

211



how much sulfur is contained in a given coal batch.Likewise, coal companies and their customers needaccurate determinations of sulfur and ash contentalong with calorific value to set fair tonnage pricesof coal shipments. A new NIST Standard Refer-ence Material (SRM) can help boost the accuracyof all these measurements by allowing evaluationof laboratory methods and calibration of instru-ments used in coal analysis. The SRM, which con-sists of a 50-gram bottle of bituminous coal, iscertified for its sulfur and ash content as well as itscalorific value. Also included are non-certifiedvalues of 23 other elements. The new material(SRM 2692) costs $102 and is available from theOffice of Standard Reference Materials, NIST,B311 Chemistry Building, Gaithersburg, MD20899; telephone: 301/975-6776.

NEW AUSTRALIAN BAUXITE ORESTANDARD ISSUEDA new bauxite ore Standard Reference Material(SRM) is available from NIST for aluminum pro-ducers to use in analyzing raw materials. SRM 600,Bauxite from the Darling Range, Australia isthe fifth in a series of bauxite standards to be issuedby NIST under a cooperative program with indus-try through ASTM. The other SRMs are Arkan-sas, Surinam, Dominican, and Jamaican. SRM 600is in the form of a fine powder for use in validatingexperimental data and analytical methods. It pro-vides certified concentrations and estimated uncer-tainties for 15 inorganic constituents. SRM 600may be purchased for $94 per 60-gram unit fromthe Office of Standard Reference Materials, NIST,B311 Chemistry Building, Gaithersburg, MD20899; telephone: 301/975-6776.

Standard Reference Data

DIPPR DATABASE EXPANDED TO 1,023PURE COMPOUNDSMore than 250 pure chemical compounds havebeen added to a computerized database on the ther-modynamic and physical properties of chemicals.The database, DIPPR (Design Institute for Physi-cal Property Data), Data Compilation of PureCompound Properties, 1989, now contains infor-mation on 39 properties for 1,023 pure chemicalcompounds of high industrial priority. The data-base provides chemical engineers, manufacturers,and scientists in industry, government, and univer-sities with quick access to important informationon the behavior of substances and their reactions atvarious temperatures. The chemicals in the data-base were selected by the industry members of theAmerican Institute of Chemical Engineers'(AlChE) DIPPR group, and are considered to bethe most important ones to industry. For informa-tion on fees and license agreements for NIST Stan-dard Reference Database 11, DIPPR, DataCompilation of Pure Compounds, 1989, contact theOffice of Standard Reference Data, NIST, A323Physics Building, Gaithersburg, MD 20899; tele-phone: 301/975-2208.

212

Date post:	23-Aug-2020
Category:	Documents
Upload:	others
View:	0 times
Download:	0 times