Table Of ContentVolume103,Number2,March–April1998
Journal of Research of the National Institute of Standards and Technology
[J.Res.Natl.Inst.Stand.Technol.103,177(1998)]
Degradation of GaAs/AlGaAs Quantized Hall
Resistors With Alloyed AuGe/Ni Contacts
Volume 103 Number 2 March–April 1998
Kevin C. Lee Carefultestingoveraperiodof6yearsofa protectingQHRdeviceswithalloyed
numberofGaAs/AlGaAsquantizedHallre- AuGe/Nicontactsfromdegradation:the
sistors(QHR)madewithalloyedAuGe/Ni heterostructurecanbeleftunpassivated,but
National Institute of Standards and contacts,bothwithandwithoutpassivating thealloyedcontactscanbecompletelycov-
Technology, siliconnitridecoatings,hasresultedinthe eredwithaverythick(>3(cid:109)m)coatingof
identificationofimportantmechanismsre- gold;ortheGaAscaplayercanbecare-
Gaithersburg, MD 20899-0001
sponsiblefordegradationintheperfor- fullyetchedawayafteralloyingthecontacts
manceofthedevicesasresistancestan- andpriortodepositingapassivatingsilicon
dards.Coveringthecontactswithafilm, nitridecoatingovertheentiresample.Of
suchasalow-temperaturesiliconnitride, thetwo,thelatterismorechallengingto
thatisimpervioustohumidityandother effect,butpreferablebecauseboththecon-
contaminantsintheatmosphereprevents tactsandtheheterostructureareprotected
thecontactsfromdegrading.Thedevices fromcorrosionandoxidation.
coatedwithsiliconnitrideusedinthis
study,however,showedtheeffectsofa Keywords: alloyedcontacts;contact
conductingpathinparallelwiththe degradation;GaAs;gold-germanium-nickel;
2-dimensionalelectrongas(2-DEG)at ohmiccontacts;passivation;quantizedHall
temperaturesabove1.1Kwhichinterferes resistor;quantumHalleffect;2-dimensional
withtheiruseasresistancestandards. electrongas.
Severalpossiblecausesofthisparallel
conductionareevaluated.Onthebasisof
thiswork,twomethodsareproposedfor Accepted: November18,1997
1. Introduction
Quantized Hall resistors (QHRs) made with alloyed alsooccuroveraperiodofmanyyearsashumidityand
AuGe/NiohmiccontactstoGaAs/AlGaAsheterostruc- atmospheric contaminants corrode or oxidize the con-
tures are quite widely used as resistance standards by tacts and the heterostructure.
many national standards laboratories [1, 2]. These Inpreviouswork[4]itwasshownthatbondingwires
devices are repeatedly cooled and warmed between to the contact pads directly over the heterostructure
room temperature and temperatures below 1.4 K over resultsintheformationofelectricallyactivedefectsin
periods of many years. Degradation or failure of the the fragile heterostructure beneath the contacts, which
devices during cooling or use is costly, for the labora- increases the contact resistances and degrades the
tory’s calibration schedule is delayed, both liquid performance of the device. This degradation can be
helium and staff time are lost while the device is eliminatedbydepositingbondingpadsthatextendover
replaced or repaired, and a lengthy testing procedure both the contacts on the heterostructure and the semi-
must be performed to certify a new or repaired device insulating substrate to permit wires to be bonded over
as a resistance standard [3]. It is therefore of great thesubstrate.Anydamagetothesubstratecausedbythe
importancethatthedevicesusedasresistancestandards high pressures created during the bonding process will
be as reliable and resistant to degradation as possible. then not affect the sensitive ohmic contact to the
Degradation of the devices can result from processing heterostructure.
steps used to mount the devices in packages, but can
177
Report Documentation Page Form Approved
OMB No. 0704-0188
Public reporting burden for the collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and
maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information,
including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington
VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to a penalty for failing to comply with a collection of information if it
does not display a currently valid OMB control number.
1. REPORT DATE 3. DATES COVERED
NOV 1997 2. REPORT TYPE 00-00-1997 to 00-00-1997
4. TITLE AND SUBTITLE 5a. CONTRACT NUMBER
Degradation of GaAs/AlGaAs Quantized Hall Resistors With Alloyed
5b. GRANT NUMBER
AuGe/Ni Contacts
5c. PROGRAM ELEMENT NUMBER
6. AUTHOR(S) 5d. PROJECT NUMBER
5e. TASK NUMBER
5f. WORK UNIT NUMBER
7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION
National Institute of Standards and Technology,Gaithersburg,MD,20899 REPORT NUMBER
9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSOR/MONITOR’S ACRONYM(S)
11. SPONSOR/MONITOR’S REPORT
NUMBER(S)
12. DISTRIBUTION/AVAILABILITY STATEMENT
Approved for public release; distribution unlimited
13. SUPPLEMENTARY NOTES
14. ABSTRACT
15. SUBJECT TERMS
16. SECURITY CLASSIFICATION OF: 17. LIMITATION OF 18. NUMBER 19a. NAME OF
ABSTRACT OF PAGES RESPONSIBLE PERSON
a. REPORT b. ABSTRACT c. THIS PAGE Same as 24
unclassified unclassified unclassified Report (SAR)
Standard Form 298 (Rev. 8-98)
Prescribed by ANSI Std Z39-18
Volume103,Number2,March–April1998
Journal of Research of the National Institute of Standards and Technology
Theworkreportedinthispaperconcernsthecauses should vanish as is required of standards-quality QHR
of long-term degradation in QHR devices that occurs devices.Thenitridecoatingshouldthenprotectboththe
over a period of many years. Quantized Hall resistance contacts and the heterostructure from corrosion and
deviceswithalloyedAuGe/NiohmiccontactsonGaAs/ degradation.
AlGaAs heterostructures both with and without passi- Section2ofthispapergivesabriefdescriptionofthe
vating silicon nitride coatings were studied. Enlarged samplesusedinthisstudy.Thedetailsoftheprocedure
bonding pads were deposited over the alloyed contacts usedtomountthesamplesforquantumHalleffectmea-
andwireswerebondedtothepadsoverthesubstrate,so surementsaregiveninSec.3.Asummaryoftheresults
the sensitive alloyed contacts were not exposed to any ofQHEtestsonboththesamplescoveredwithapassi-
mechanical stresses. When the samples were tested, it vatingsiliconnitridecoatingandthosewithoutisgiven
was found that those with silicon nitride coatings had inSec.4.InSec.5thecausesofdegradationofunpro-
verylowcontactresistancesandwereofveryhighqual- tected AuGe/Ni contacts are discussed, and in Sec. 6
ity, but the minima in the voltages measured between two methods are proposed for preventing this degrada-
probesonthesamesideoftheHalldevice(V)underthe tion from occurring.
x
conditions required to observe the quantum Hall effect
(QHE)didnotvanishattemperaturesabove1.1Kasis 2. Origin and Design of the QHE Devices
required for use as resistance standards [3]. While the
minimainV didvanishforsamplesthatwerenotcoated In 1990, the EUROMET consortium of European
x
withsiliconnitrideandwhichhadbeenstoredinplastic national standards laboratories, in conjunction with the
petridishesforover6yearsinanunregulatedlaboratory Bureau International des Poids et Measures (BIPM) in
environment, these samples were found to have higher France,theNationalInstituteofStandardsandTechnol-
contact resistances and somewhat more nonuniform ogy (NIST) in the USA, and the National Research
electron concentrations than the coated samples. The Council(NRC)inCanada,letacontractwiththeLimeil
higher contact resistances are attributed to the fact that GaAs Foundry of the Laboratoires d’Electronique
the AuGe/Ni contacts on the uncoated samples were Philips (LEP)1 in France to produce quantized Hall
exposed directly to corrosive compounds in the labora- resistancedevices(seeRef.[1]).LEPgrewaGaAs/Al-
tory atmosphere, where the temperature and humidity GaAs heterostructure, a schematic cross-section of
variedoverawiderange(18(cid:56)Cto30(cid:56)C,10%to70% whichisshowninFig.1a,usingthetechniqueofMetal-
relativehumidity).Corrosionofunprotectedmetalcon- Organic Vapor Phase Epitaxy (MOVPE). The top layer
tacts under these conditions has been reported in the of the heterostructure is a GaAs cap layer doped with
literature [5, 6, 7]. The higher nonuniformity of the silicon; below it are a donor layer and spacer layer of
electron concentration in the samples without a silicon Al Ga As. The donor layer is doped with silicon
0.28 0.72
nitridecoatingisattributedtononuniformoxidationof atoms,butthespacerisnot.The“bufferlayer”iscom-
the exposed top surface of the heterostructure. posed of two layers of GaAs separated by a layer of
These observations indicate that it is necessary to Al Ga As,allundoped.Thelocationofthe2-dimen-
0.1 0.9
protect the AuGe/Ni ohmic contacts on GaAs/AlGaAs sionalelectrongas(2-DEG)responsibleforthequantum
QHE devices from the atmosphere in order to ensure Halleffectisshownbytheblacklinelabeled“2-DEG.”
their long-term reliability. Two techniques for protect- The Al Ga As layer in the buffer layer helps isolate
0.1 0.9
ing the contacts are proposed as a result of this work. thechannelinwhichthe2-DEGresidesfromdefectsin
The simplest is to cover the contacts completely with a thesubstrate,andisalsointendedtominimizetheinjec-
coating of gold greater than 3 (cid:109)m in thickness. Such a tion of high velocity electrons (also called “hot elec-
coating has been shown to prevent corrosion in metal trons”) from the 2-DEG into the buffer layer and
contacts [5]. The other is to cover the samples with a substrate,aproblemthatismuchmoreseverewithhigh
silicon nitride layer deposited using a low temperature electron mobility transistors (made using this same
chemical vapor deposition (LTCVD) technique, as was design of heterostructure) than with QHE devices [8].
donewiththepassivatedsamplesusedinthisstudy.The The EUROMET committee provided LEP with a
nonzero minima in V measured on these passivated pattern for a Hall bridge and ohmic contacts, shown in
x
samples do not appear to be due to current flowing Fig. 1b, based on designs used at European standards
throughthenitride,ashaspreviouslybeensupposed[1],
butismostlikelyduetocurrentflowinginthedegener- 1Certaincommercialequipment,instruments,ormaterialsareidenti-
fiedinthispapertofosterunderstanding.Suchidentificationdoesnot
ately doped GaAs cap layer. If this cap layer is grown
implyrecommendationorendorsementbytheNationalInstituteof
without donor impurities or is etched off after alloying
Standards and Technology, nor does it imply that the materials or
the contacts and prior to depositing the nitride, this equipment identified are necessarily the best available for the
conduction should be eliminated, and the V minima purpose.
x
178
Volume103,Number2,March–April1998
Journal of Research of the National Institute of Standards and Technology
width, and the pad neck is about 50 (cid:109)m in width, as
indicated in Fig. 1c. The AuGe/Ni alloyed contacts are
175 (cid:109)m square, and extend about 7.5 (cid:109)m beyond the
edgeoftheheterostructuremesa.TheTi/Pt/Aubonding
pads are 152 (cid:109)m square, and were deposited over the
AuGe/Nialloyedcontacts,entirelywithinthemesa:the
edgeoftheTi/Pt/Aucontactisabout4(cid:109)minsideofthe
edge of the heterostructure mesa.
On half of the devices, a protective 165 nm thick
silicon nitride coating was applied using low-tempera-
ture chemical vapor deposition (LTCVD). This nitride
coatingcoverstheentiresample,exceptforholesexpos-
ingalloftheTi/Pt/Aubondingpadswiththeexception
of a 2.5 (cid:109)m wide rim around the edge of the bonding
pads, which rim lies under the nitride. The hole over
each potential probe, shown in Fig. 1c, is 147 (cid:109)m
square.
These devices were given to the individual national
standards laboratories to be mounted and tested. In the
autumn of 1990, NIST received 30 devices without the
nitridecoatingand30deviceswiththenitridecoating.
These devices were used in the work reported in this
paper.
3. Procedure Used to Mount the Devices
In order to use these devices as resistance standards,
they must be mounted in packages, typically 12-pin,
nonmagnetic“headers,”thatfitintosocketsintheprobe
ofacryogenicsystemthatcancoolthemtotemperatures
below 1.2 K in magnetic flux densities of between 5 T
and8TrequiredtoobservethequantumHalleffect.In
principle, the procedure for doing this is quite simple:
the sample is attached to the package using epoxy, and
wires are bonded between the pads on the sample and
theheaderpinsusingstandardwirebondingtechniques.
Fig.1. DesignofLEPsamples.(a)Schematiccrosssectionofthe Ifoneusesverysoft,12(cid:109)mdiameterwirewhenbond-
heterostructurefromwhichtheEUROMETsamplesweremade(from
ing,thissimpleprocedurecanprobablybeusedwithout
Ref. [1]) (b) Scale drawing of the mask used by LEP to make the
harming the contacts, for the forces generated when
EUROMETsamples.(c)Detailoftheregioncontainingtheohmic
contact on a potential pad on a sample coated with silicon nitride. such wire is bonded to pads on the GaAs/AlGaAs
heterostructure are quite small. Such small diameter
laboratories.ThelengthoftheHallbarisapproximately wireis,however,veryfragile,andcaneasilybebroken
2.7 mm and its width is about 0.4 mm. The Limeil bygustsofheliumgasorotherstressesgeneratedwhen
FoundryetchedtheHallbarpatternintotheheterostruc- the sample is cooled to cryogenic temperatures. Larger
ture,andalloyedgold-germanium-nickelcontactstothe diameterwire,suchasthe25(cid:109)mdiameterwireusedin
samples. Ti/Pt/Au bonding pads were deposited on top this study, is much sturdier and can better withstand
oftheAuGe/Nipads(seeFig.1c).Thegreysquaresin the stresses generated during cooling, but such
Fig.1bindicatethelocationsofthealloyedgold-germa- wire requires the use of higher bonding forces which
nium nickel contacts. The labels D and S in the figure generate greater stresses on the semiconductor under-
identify, respectively, the drain and source contacts neaththepadtowhichthewireisbonded.Asdiscussed
through which current flows, and the pads labeled in an earlier paper [4], the heterostructure is extremely
P1...P6identifythepotentialprobes.Theheterostruc- fragile, and quantized Hall resistance devices are
turemesabeneaththepotentialpadsisabout160(cid:109)min extremely sensitive to the slightest damage in the
179
Volume103,Number2,March–April1998
Journal of Research of the National Institute of Standards and Technology
contact region. Hence, the forces generated during passivated samples. The metal film did not adhere as
bonding of 25 (cid:109)m diameter wires create electrically welltotheGaAssubstrateontheunpassivatedsamples,
active defects in the heterostructure underneath the andtheagitationrequiredtofragmentthefree-standing
bonding pad even when the lightest bonding pressures gold film that had been deposited over the black wax
areused,resultinginameasurableincreaseintheresis- alsotendedtoremovetheevaporatedfilmfromoverthe
tance of the contacts. alloyed contacts, where it was supposed to remain.
Topreventthisdamage,largebondingpads,overlap- Thisproblemisreadilysolvedbyapplyingphotoresist
ping both the alloyed contact and a substantial part of tothesampleandusingthe“lift-off”procedure[9].As
thesubstrate,weredeposited,andwireswerebondedto shown in Fig. 2b, this results in a discontinuous metal
these pads over the substrate, rather than over the het- film. When the photoresist is dissolved in acetone, the
erostructure.Defectscreatedinthesubstrateduetothe metal between the bonding pads is removed easily
high pressures required to bond wires to the bonding without the need for vigorous ultrasonic agitation.
pads therefore do not affect the quality of the ohmic
contact, because the substrate is semi-insulating and
does not carry any current during normal operation of
the device. All of the devices used in this study have
beenmountedusingproceduresbasedonthis“enlarged
bondingpad”principle.Thenextsection(Sec.3.1)dis-
cusses some of the challenges experienced with pro-
cesses used in early trials, and Sec. 3.2 describes the
finalprocedureusedtomountthemajorityofthesam-
ples used in this study.
3.1. Principles of the Procedure
There are a number of ways in which this “enlarged
bondingpad”principlecanbeimplemented.Allproce-
dures must, however, ensure that:
• bonding pads are deposited over the alloyed con-
tactsandmakegoodelectricalcontactwiththem—there
mustbenoorganiccontaminationonthecontactsprior
to deposition of the bonding pads;
• the bonding pads are very adherent to the sub-
strate and are thick enough so stresses created during
bonding do not cause the bonding pads to tear away
from the substrate;
• and the method used to define the bonding pads
does not disturb the alloyed contacts.
In early experiments, Apiezon W (black wax made by
Apiezon products, London, UK) was applied to the re-
gionsofthesampletowhichthemetalfilmthatwould
form the bonding pad would not adhere, the metal was
deposited,andtheblackwaxwasdissolved.Becausethe
evaporatedgoldfilmcontinuouslycoatedthesampleas
shown in Figure 2a, dissolution of the black wax in Fig.2. Illustrationoftwodifferentmethodsofdefiningapatternof
enlargedbondingpadsontheLEPsamples:(a)Applicationofablack
solvent did not separate the adjacent bonding pads.
waxpatternresultsinacontinuousmetalfilmthatcoversboththe
Vigorouslyagitatingthesampleinanultrasoniccleaner
sampleandthewaxmask.Themetalbetweenthebondingpadsmust
successfullyfragmentedthefree-standingfilmthathad beremovedbyvigorousagitationinanultrasoniccleaner.(b)Appli-
covered the black wax, thus resulting in well-defined cationofphotoresistresultsinadiscontinuousmetalfilmthatcanbe
bonding pads over each ohmic contact. Because the simply lifted off by dissolving the resist in acetone, resulting in
well-definedbondingpadswithouttheneedforvigorousultrasonic
evaporated metal adhered very strongly to silicon
agitationofthesample.
nitride, this patterning technique worked well for the
180
Volume103,Number2,March–April1998
Journal of Research of the National Institute of Standards and Technology
The metal film has to be fairly thick (a minimum of 3.2. Annotated Processing Sequence
0.3(cid:109)mandpreferablythicker)inorderthatitbeableto
absorbenoughofthebondingstressestominimizedam- Theprocedureusedtomountthesamplesinthiswork
age to the substrate, and to prevent the stress at the is based upon the principles described in the previous
metal-substrateinterfacefromreachingavalueatwhich section, namely:
themetalwillshearandpeelawayfromthesubstrate.In 1. The samples must only be cleaned in inert sol-
addition, for this procedure to work properly, the resist vents with a minimum of agitation—the samples must
mustbeconsiderablythickerthanthemetalinorderthat neverbeexposedtoanyaqueousorcausticorcorrosive
the metal not form a continuous bridge over the resist. solutions, even those that are not water-based;
This requires the application of uniformly thick resist 2. Thick bonding pads, covering both the alloyed
films greater than 1 (cid:109)m in thickness to chips 0.2 mm contact and the heterostructure, must be deposited
thick with width and length of 1.1 mm and 2.9 mm, through a “see-through” mask;
respectively, without the formation of a bead of resist 3. Exposure of the sample to temperatures of
aroundtheedgeofthechip(commonlyreferredtoasan 200 (cid:56)C or higher during bonding and curing of the
“edgebead”).Theabsenceoftheedgebeadwaspartic- epoxy should be minimized.
ularlyimportant,forthecontactpadswerelessthan100
The processing steps in the procedure used in this
(cid:109)m from the edge of the chip.
work,hereafterreferredtoasthe“optimizedprocedure”
While this task is rather challenging, a technique for
are now described.
doing it reliably was developed. A depression with
1. Teflon FEP (Fluorinated Ethylene Propylene)
nearly the exact dimensions of the chip was made in a
beakers were used to clean the samples. The beakers
glassplateandthesamplewasaffixedtothisplatewith
werefirstcleanedbyboilinginthemasolutionof1part
a minute quantity of photoresist. The plate was then
98 % HSO to 1 part 30 % HO to between 3 and 5
heated very slowly to drive the solvent out of the pho- 2 4 2 2
parts by volume of deionized water (with resistivity of
toresist without causing the formation of bubbles that
18M(cid:86)cm)foratleast10min,followedbyathorough
would force the sample up out of the depression in the
rinse with filtered, deionized water. Note that this
plate.Thesurfaceofthesamplewasthenpreciselyeven
process ensures that the beakers are clean before the
withthetopsurfaceoftheglassplate,sotheresistfilm
samples are placed in them. This step is necessary
on the sample was quite uniform. This technique was
because not only are the AuGe/Ni contacts very sensi-
successfullyusedtomountafewofthesamplescoated
tive to corrosion in strong cleaning solutions such as
with a passivating nitride layer without significant
wafer detergents, but GaAs itself is oxidized in almost
degradation of the contacts. When attempts were made
all aqueous cleaning solutions.2 Thus, the only agents
to mount unpassivated samples using this technique,
thatcanbeusedtocleanthesamplesaresolvents,which
however, the resistances of the contacts were quite
arenotparticularlystrongcleaningagents,socaremust
noticeably increased, even at room temperature. The
betakentoensurethatthesamplesnevergetverydirty.
alloyed AuGe/Ni contacts on the unpassivated samples
2. ThechipwiththeQHRdeviceandaglasscarrier
weredirectlyexposedtoalloftheprocessingsolutions,
plate3 with evaporated gold bonding pads (made sepa-
anditwasfoundthattheyaredegradedquitenoticeably
rately) are placed in separate beakers and cleaned by
by exposures for periods of time as short as a few
minutes to several different varieties of photoresist
developer made by different manufacturers, as well as
2TheGaAsisreadilyandseverelyoxidizedandcorrodedinwater,
by photoresist remover.
particularlyatelevatedtemperatures—seeRef.[29].
In view of the extreme sensitivity of the exposed 3Thesampleswereattachedtoaglasscarrierplateratherthandirectly
alloyed contacts on the unpassivated samples to corro- to the header, to facilitate removal of the sample from the header
sion,thebondingpadscanonlybeappliedbyevaporat- withoutdisruptingthewiresbondedtothepadsonthesample.Wires
ingthemetalthrougha“seethrough”mask,athinmetal werebondedbetweenthesampleandpadsonthecarrierplate,soonly
thebondsbetweenthepadsonthecarrierplateandtheheaderpins
foil with holes etched through it in the appropriate
would have to be broken when the sample was removed from the
places, that is placed in proximity to the sample while
header.Bothofthesebondsaremoreeasilyreplacedthanthebonds
themetalisevaporated.Thesamplecanonlybecleaned tothepadsonthefragileGaAschip.
in inert solvents, such as xylenes, tricholorethylene, or
acetone,forbriefperiods,withrelativelylittleultrasonic
agitation.
181
Volume103,Number2,March–April1998
Journal of Research of the National Institute of Standards and Technology
firstheatinginxylenes4for10min,andthenagitatingin The sample is protected behind an aluminum plate
an ultrasonic cleaner for 5 min. The xylenes are then while the filaments are being heated to evaporation
decanted, and the procedure repeated with trichloro- temperature.Chromiumisdepositedonthesamplefirst,
ethylene,acetone,andfinallymethanol.Thesamplewas typically for 2 min to 4 min with a power of 154 VA,
thenboiledbrieflyinacetone,theacetonedecanted,and resulting in Cr films on the sample that are between
the samples blown dry with filtered dry nitrogen. 17nmand35nmthick.Thesampleisthenrotateduntil
3. AweightofAuwiresufficienttoproduceacoat- it is over the gold filament, and deposition of the gold
ingonthesampleofatleast340nm(preferablyatleast iscommencedimmediately.Thegoldchargeisevapo-
0.5(cid:109)mthick)iscut,woundintoasmallballabout5mm ratedtocompletion,typicallyfor15minto30minwith
in diameter, and cleaned by etching in a solution of 1 a filament power between 350 VA and 400 VA. The
part 98 % HSO to 1 part 30 % H O to 5 parts by final thickness of the Au coating is determined by the
2 4 2 2
volume of water. This solution is first heated on a hot- weight of the gold charge as described in paragraph 3
platetobetween80(cid:56)Cand100(cid:56)Cforabout5min,and above, and was typically between 320 nm and 340 nm
thenplacedinanultrasoniccleaner,whereitisagitated in these experiments.
forbetween10minand15min.thewireisthenboiled Gold will not adhere to most substrates, including
inmethanolandblowndry.Inthiswork,anNRC3114 GaAs, without the presence of some other element,
Vacuum Coater with a source-to-sample distance of usually a transition metal like chromium, titanium, or
about17cmwasused.Inthisevaporatora1gchargeof tungsten. Chromium was used in these experiments
gold produces a gold coating about 0.33 (cid:109)m thick. becauseofthereadycommercialavailabilityofconve-
4. Aglassmicroscopeslideisheatedto90(cid:56)Cona nientlyusedchromium-platedtungstenrodevaporation
small hotplate and the sample is mounted on it with sources.
black wax. The slide is then placed in a specially Experiencehasshownthatevenwithalowpressure
designed device, in which the “see-through” mask, in the evaporation chamber and a liquid nitrogen trap
madefroma25(cid:109)mthickbrassfoil,isclampedbetween between the diffusion pump and the chamber, the
two support plates, and the sample is held in close strength with which the Au film adheres to the
proximity to, but not in direct contact with, the mask. chromium layer decreases markedly as the time
5. Thesampleisplacedintheevaporator.Thegold betweentheendofthechromiumdepositionandstartof
wireisheatedbrieflywithatorchuntilitglowsreddish- thegolddepositionincreases.Byheatingbothfilaments
orange,andthenplacedona50mmlong,250(cid:109)mthick totheevaporationtemperaturenearlysimultaneouslyso
molybdenum strip with a thin alumina coating. A that the deposition of the gold film could be started
50 mm long chromium-plated tungsten rod is also immediately after deposition of the chromium,
heatedtoredheatwithatorch,andisthenplacedinthe the adhesion of the bonding pads to the sample was
secondfilamentbayintheevaporator.Heatingtheevap- maximized.
oration charges immediately prior to installing them in 7. Afterthebondingpadshavebeendeposited,the
theevaporatorremovesanyresidualorganiccontamina- sampleisremovedfromtheevaporatorandcleanedby
tion,drivesoffadsorbedwatervapororsolventresidues, heatinginxylenesforabout5min,andagitatingitinan
and minimizes the amount of gas evolved from them ultrasonic cleaner for another 5 min. The xylenes are
when they are heated in vacuum prior to coating the then decanted, the sample boiled in trichloroethylene
samples. forafewminutes,andthenblowndrywithfiltereddry
6. Theevaporatorisevacuatedforabout1huntila nitrogen.
base pressure of between 20 (cid:109)Pa and 65 (cid:109)Pa 8. Asmallquantityofepoxyisthenappliedtothe
(0.15(cid:109)Torrto0.49(cid:109)Torr)isreached.Theliquidnitro- glass carrier plate and the sample placed in the epoxy.
gentrapbetweenthediffusionpumpandthechamberis Theglassplateisthenplacedonahotplateat165(cid:56)Cfor
then filled. Cooling water is supplied to the current 5 min to cure the epoxy. The epoxy must meet rather
feedthroughs to prevent them and the baseplate from demanding requirements: it must be strongly adherent
heating up during the evaporation. Separate power at temperatures of about 200(cid:56)C to ensure that it holds
supplies are used to supply current to the Cr and Au the sample firmly during wire bonding in order to
filaments,andbothfilamentsareheatedsimultaneously. minimizethedamagetothesubstratethatwouldresult
frommovementsofthesampleduringbonding,andyet
mustalsonotcrackorfailatcryogenictemperaturesof
4Allsolventsusedinthisworkwereeither“electronicgrade”(low
1.2 K or less at which the quantized Hall resistors are
concentrationofheavymetalsandfilteredtoremoveparticulatecon-
operated.
tamination)or“reagentgrade.”
182
Volume103,Number2,March–April1998
Journal of Research of the National Institute of Standards and Technology
Trialwasmadeofanumberofdifferentcommercial 2. the carrier plate with attached sample was then
epoxies. Conductive epoxies from two different manu- cleaned in solvents (step 2 above);
facturers,oneasinglecomponentepoxyandtheothera 3. thesamplewasbakedforabout1hourat200(cid:56)Cto
two-component epoxy, were tried in several experi- desorbwatervapor,andplacedinthemaskholder,
ments. The single component, silver-filled conductive 4. thechromiumandgoldfilmswereevaporated;and,
epoxywasfoundtoadherewellatalltemperaturesand 5. theglasscarrierplatewasimmediatelymountedin
wasextremelyconvenienttouse,butitshighconductiv- theheaderandthewiresbondedwithoutpost-evap-
itymeantthatextremecarehadtobeexercisedtoensure oration cleaning.
thatnoneoftheepoxytouchedanyofthebondingpads.
The two-component conductive epoxy from the second This procedure was tested on one sample (designated
manufacturer suffered from the same problem and “E8”),buttheresistancesofthecontactsonthissample
furthermore tended to be crumbly at the bonding tem- werehigherthanthoseofothersamplesmountedusing
perature of 200 (cid:56)C, and so was judged not suitable for the normal procedure described above. While the con-
this application. tactsonthisonesamplemayhavebeenofpoorerquality
Inmostoftheseexperiments,atwo-component,non- before processing than those mounted with the opti-
conducting epoxy, EPOTEK H70E (made by Epoxy mized procedure, it is possible that some steps or steps
Technology, Billerica, MA, USA) was used. It was in the streamlined procedure degraded the contacts. If
foundthatifcarewastakentopreparetheepoxyusing this were the case, the most likely causes for the
equal weights of the two components [10], the epoxy degraded contacts would be residues from the epoxy
remainedstronglyadherenttothesampleovertheentire (applied in the first step) not removed by cleaning in
temperaturerangefrom200(cid:56)Ctolessthan1.2K.Even solvents(secondstep)whichmayhavecontaminatedthe
though the epoxy is insulating, care has to be taken to AuGe/Ni contacts resulting in poor electrical contact
ensure that not too much epoxy is used so that neither between the evaporating bonding pads and the alloyed
the epoxy nor its residues contaminate the tops of the contacts.Thehigh-temperaturebakingstep(step3)may
bondingpadsduringcuring,asthismakesitdifficultto have led to some change in the composition of the
bond wires to the pads [11]. contacts that may have increased their resistances. For
9. The carrier plate was then attached to a clean thesereasons,the“streamlinedprocedure”wasnotused
TO-8 header, also with epoxy, and gold wires with a to mount more than the single test sample.
diameter of 25 (cid:109)m, a tensile strength of 5.9 cN
(centinewton), and “4% elongation” were bonded
4. Results
betweenthebondingpadsonthesampleandthepadson
the carrier plate, and then between the pads on the
The optimized procedure described in the previous
carrier plate and the head pins. The sample was main-
section was used to mount three LEP samples without
tainedat200(cid:56)Cduringbonding.Thebondingtoolwas
passivatingsiliconnitridecoatingswithserialnumbers
pressed against the sample and ultrasonic power
E5,E6,andE7.OneunpassivatedLEPsample(E8)was
(approximately250mW)wasappliedforabout100ms.
mounted using the “streamlined procedure” described
Thetoolwaspressedontothepadsonthesamplewith
above.Twopassivatedsamples(serialnumbersE5Cand
a force of between 25 cN and 30 cN; when bonding to
E7C)hadbeenmountedin1993usingatechniqueiden-
the header pins and the pads on the glass carrier plate,
ticaltotheoptimizedprocedure,exceptthatblackwax
bonding forces between 39 cN and 49 cN were used.
was used to define the bonding pads, as described in
Typically, one wire was bonded to each potential pad,
Sec. 2, rather than a “see-through” mask. Tests on
and two wires were bonded to each source and drain
sample E7C were reported in Ref. [4] in the section
contact pad, each to a different header pin (there were
entitled“EnlargingBondingPads.”Thesesampleswere
therefore two header pins connected to the source and
compared with one coated sample (E2C) and one
twotothedrain,sothatthesamplecouldcontinuetobe
uncoatedsample(E1)thatweremountedinheadersand
usedevenifoneofthepairofwiresgotbroken).Allthe
testedin1990shortlyafterthesampleswerereceivedat
wireswerebondedwithinaperiodof30minto45min.
NIST.Wireswereattachedtothecontactpadsonthese
Thisoptimizedprocedurewasusedtomountmostof
samplesbymeltingsmallbeadsofindiumontothegold
the samples used in this study. Figure 3 shows a pho-
pads directly over the heterostructure and pressing the
tographofatypicalsampleafterthecompletionofthis
gold wires into the beads. The results of the tests on
procedure. A “streamlined procedure” was developed,
these samples were reported in the “Soldering” section
in which:
of Ref. [4].
1. the sample was affixed to the glass carrier plate
with epoxy;
183
Volume103,Number2,March–April1998
Journal of Research of the National Institute of Standards and Technology
Fig. 3(a). Photograph of sample E8 mounted on a glass carrier plate in a
TO-8headerusingtheproceduredescribedinSec.3.
Fig.3(b). EnlargedviewofsampleE8showingtheenlargedbondingpadsandthewiresbondedtothem.
184
Volume103,Number2,March–April1998
Journal of Research of the National Institute of Standards and Technology
Allsamplesweretestedundertheconditionsrequired 1.2 (cid:51) 10–6 times the Hall resistance). The combined
to observe the quantum Hall effect. Because the super- standard uncertainty, including systematic effects, was
conductingsolenoidusedinthisexperimentwaslimited of the order of 1 (cid:109)V (about 6 (cid:51)10–6 times the Hall
to a maximum magnetic flux density of 8 T, only the resistance).
i=4plateauswereexaminedinthiswork.Underthese 2. Contact Resistances. With the magnetic flux
conditions, the sample resistance is V /I = R /4 = densitysettoavalueatthemiddleofthei=4plateau
H K
h/4e2 = 6453.20175 (cid:86), where V is the Hall voltage, I (between4.9Tand5.3T),thecontactresistanceofeach
H
isthecurrentthroughthedevice,R isthevonKlitzing contact was measured using a 3-terminal technique
K
constant, h is the Planck constant, and e is the elemen- similar to that described in Ref. [4]. A programmable
tary charge. current source passed current between the contact of
interest (denoted “A”) and a second contact (denoted
4.1 Measurements “B,”usuallythesourceordrain),andthepotentialwas
measured between the contact of interest and a third
Three different measurements were done to charac- contact (denoted “C”) that did not carry current and
terize the samples: was nominally at the same potential (i.e., on the same
1. Plateau Quality. The minima in V (measured side of the Hall bar) as the contact of interest. The
x
between pairs of contacts on the same side of the Hall contact resistance was determined by measuring the
bar) and the values of V (measured between pairs of voltageV withthecurrentsetatonevalue(I ),then
H AC AB
contacts on opposite sides of the Hall bar) were mea- increasingthecurrentbyanincrement(cid:68)I ,measuring
AB
suredwith25(cid:109)Aflowingthroughthesourceanddrain the voltage V (I +(cid:68)I ), and dividing the difference
AC AB AB
contacts(seeFig.1b).Themeasurementsystemusedin involtagesbythecurrentincrement.Inotherwords,the
this work is similar to that described in Ref. [12]: an “dynamic” contact resistance was calculated using the
electronic current source drove a current through the formula:
Hall device connected in series with a room-tempera-
ture10k(cid:86)referenceresistor(SerialNumberGR99).A
high-quality standard cell scanner was used to both 1 V (I +(cid:68)I )–V (I )
R (I) = » AC AB AB AC AB .
reversethecurrentdirectionandconnectahigh-stability AB,AC [dI /dV ] (cid:68)I
AB AC AB
8.5 digit digital voltmeter (DVM) alternately between
thereferenceresistorandtheHallresistor.Themeasure- (1)
mentsystemwasentirelyundercomputercontrol.Probe
voltage measurements at a given magnetic field were
madewithcurrentflowingalternatelyinoppositedirec- This definition of contact resistance was chosen for
tions(i.e.,whenthedrainwaspositiverelativetosource, severalreasons.Firstly,1/R istheslopeoftheI(V)
AB,AC
thecurrentwas+25(cid:109)Aandwhenthedrainwasnega- curve, which gives information about the nature of the
tiverelativetothesource,thecurrentwas–25(cid:109)A).The potentialbarrierbetweenthemetalandthesemiconduc-
voltages measured with current flowing in each direc- tor, and is commonly used to assess deviations of the
tion were averaged to eliminate the contribution of contact’s behavior from the ideal [31].
thermal voltages to the measured voltages [12]. While Secondly, the measured voltages are usually
this measurement system is capable of achieving quite verysmall,particularlywhensmallcurrentsarepassed
low uncertainties, as described in Ref. [12], several through the device, and can be less than thermal
modifications were made to the system to permit more voltages between the voltage probes. If the
rapid testing of samples. These modifications included common definition of “static” contact resistance, i.e.,
theuseoftheelectroniccurrentsourcewhichhadlower V (I )/I ,isused,nonnegligiblethermalvoltagescan
AC AB AB
leakageresistancesbetweenitsterminalsandearthand give rise to significant errors in the contact resistance.
higher noise than the mercury battery-powered current Because two voltages are subtracted to determine the
source used in that work, and the use of shorter mea- “dynamic” contact resistance defined in Eq. (1) above,
surementtimes.Theresolutionofthemeasurementsys- the thermal voltages cancel out, as long as they do not
tem was about 0.05 (cid:109)V for V measurements, and the driftsignificantlybetweenvoltagemeasurements.Since
x
standard uncertainty due to random effects was about the voltage measurements are made sequentially over a
0.05 (cid:109)V (which corresponds to 2 m(cid:86) at the 25 (cid:109)A short period of time (less than 1 min) the thermal
measurementcurrent).ForV measurements,thesystem voltageswillverylikelynotdriftsignificantly,andtheir
H
had a resolution of about 0.1 (cid:109)V (about 0.6 (cid:51)10–6 effect will cancel. It is for these two reasons that the
times the Hall resistance R /4), and a standard uncer- “dynamic”contactresistancewasusedasameasureof
K
tainty due to random effects of about 0.2 (cid:109)V (about contact resistance in this work.
185
