A NEW HIGH STRENGTHGOLD BOND WIRE

Giles Humpston e'r David M. Jacobson,

Hrst Research Centre,GEC-Marconi Ltd.,

Wembley, Middlesex HA9 7PPUnited Kingdom

As semiconductor technology develops, conventionalbond wires are reaching the limits of their capability.

Industry is demanding finer diameter wire coupled withhigher strength, and the retention of this strength at elevated

temperatures would be a decided advantage.

It has been demonstrated that a fine wire of the compositionAu-lwt.%Ti, which was originally formulated for high caratjewellery, can be endowed with these beneficial properties by

appropriate thermomechanical treatments. A three-foldincrease in the strength of 25 µm diameter wire over that

of conventional gold wire has been achieved.The mechanical properties are stable even when the

wire is subjected to heating at 400 °C for over a year.

Wire based on the Au-1 wt. %Ti alloy can be madecomparable to that of the standard gold products in terms

of its electrical properties and bonding characteristics.Moreover, it is more resilient to the demands of the

fabrication process.

132 Gold Bull., 1992, 25 (4)

Figure 1A silicon semiconductor chip soldered into a gold meta

recess in a ceramic package. This style of componentfinished by connectingfine gold wires between

contact pads on the chip and terminals in the packaand then soldering a lid on

INTRODUCTION

Electronic circuitry normally contains semiconductorchips, which are elaborate miniature circuits processedin thin slices of silicon or gallium arsenide, typically1 to 500 mm2 in size. These semiconductor chips arehoused in hermetic packages to protect them frommoisture and mechanical damage. A ceramic packagecontaining a chip, before attachment of the lid, isshown in Figure 1, Three main methods are used formaking electrical connection between contact padson the microchips and the external circuitry, namelywire bonding, tape automated bonding (TAB) andflip-chip bonding.

Wire bonding entails using fine wires, typically33 sm diameter or less, to connect the bond pads onsemiconductor devices to the headers on their pack-ages or to the tracks on circuit boards in hybrid assem-

blies. In this method, the bond pads on the devices andthe headers on the packages or circuit boards are met-allised with either gold or an aluminium alloy, de-pending on the intended application of the device.The ends of the fine wire interconnect are attached tothe appropriate device pad and header by micro-weld-ing. The sequence of steps involved in making an

individual wire bond are illustrated schematically inFigure 2. A semiconductor chip connected to its pack-age with gold bond wires is shown in Figure 3.

In TAB, the device is joined using a `gang' bondingoperation to a set of cantilever beams on a polymerictape (inner lead bonding) and subsequently these beamleads are attached to the circuit board or package (outerlead bonding). In flip-chip bonding, the silicon chip ismounted face down and its bond pads are directlysoldered to those of the substrate. Representative con-figurations for these two types of interconnection areshown in Figures 4, 5 and 6 (pages 135, 136), respec-tively.

The principal features of the three interconnectiontechnologies referred to above are listed in Table 1(page 137). Despite the apparent technical merits ofthe TAB and flip-chip processes, the wire bondingmethod has the commercial advantages of low set-up

costs and adaptability to changing com-ponent design.

The materials traditionally used forbond wire are gold, aluminium and oc-casionally copper, containing minor ad-ditions of other elements. This choice hasbeen determined by a diversity of re-quirements, not least ot which is theability to prepare essentially continuouslengths of wire in suitable diameters,which may be as little as 4 µm, by eitherdrawing or extrusion. Gold has specialmerits as bond wire including good resis-tance to corrosion, high electrical con-ductivity and the relative ease with whichit can be bonded into position in anambient environment by standard mi-

lised cro-welding techniques. For these rea-is sons, more than 90 % of the wire bonds

produced each year are made with goldge wire [1],

The versatility of wire bonding is likelyto ensure that its use will continue and

even grow for the foreseeable future [2]. Indeed, theconsumption of fine gold wire by the electronics indus-try has increased by roughly 20 % per annum since1970 and, despite the fineness of the product and theminuteness of the quantity of gold used in an inter-connect (typically 20 µg), sales of gold bond wireexceeded $ 150 million in 1991 [3].

Gold Bull., 1992, 25 (4) 133

Wire spoot Capillary

u tool

Flame -off _ Gold

electrode f balt

Packsgeheoder

Bond pads

Silicon chip ] )

Ceramic packageLead

TootBalt bond movemeni

a o —^

liii

Wire loop

II

Wedgebond

Figure 2

Schematic illustra-tion of the sequenceofsteps involved in

making a singleinterconnect using

bond wire:

Step 1:

A spark or smallflame is used to

locally melt the endof the wire so as to

form a spherical baltthat is approximately

twice the diameterof the wire

Step 2:

The balt is thermo-sonically welded to

an aluminiummetallised pad on the

semiconductor

Step 3:

A loop of wire isformed as the

bonding capillarymoves across to thecontact pad of thedevice package or

circuit board

Step 4:

The wire is thermo-sonically welded tothe gold metallisedpad of the package

Step 5:

A sharp edge on thetool is used to cut thewire, leaving a lengthprotruding from the

capillary that is of thecorrect length to form

the next hall

The value of fine bond wire is dominatedby the manufacturing cost rather than the met-al content. This is highlighted in Figure 7(page 138), which shows the price mark-upof precious metal fine wire in relation to goldbullion. Gold wire of 25 tm diameter com-mands a price that is roughly eight times thebullion price, but this factor rises to almost ahundred for 10 4m diameter wire, reflectingthe technical difficulties in fabricating suchfine wire and the higher processing costs. Thisexplains the relatively small price differentialbetween gold, copper and aluminium wire de-spite the approximate 60,000:6:1 ratio of thevalue of the respective metals on a volume basis.

LIMITATIONS OF EXISTING

BOND WIRE MATERIALS

The functional requirements of bond wireused in semiconductor manufacture are asfollows:

a) High electrical conductivityb) Ease of weldingc) Adequate strength to permït use in mod-

ern high-speed bonding machinesd) Retention of strength during the heating

routines encountered through devicepackaging and use.

In order to meet these requirements, existingbond wires comprise alloys of gold, alu-minium and copper that exploit solid solu-tion strengthening and work hardeningmechanisms to attain adequate mechanicalproperties. However, this approach is in-creasingly less able to meet the Bemand forfiner diameter wires in order to achieveincreased miniaturisation coupled withboosted rates of production, which meansusing faster mïcrowelding machines. Rapidmanipulation of the wire imposes signifi-cant forces on it; modern automated bond-ing machines can form a wire loop with abond at each end in less than 0.1 second„It is perhaps significant that, in recent years,

134 Gold Bull., 1992, 25 (4)

Figure 3

A close-up view of the gold bond wires ofa silicon chiphoused in a ceramic package

there has been a tendency to revert to 33 µm diameterwire in place of 25 gm gauge in order to cope with thehigh stresses imposed on the wire by modern bondingmachines, notwithstanding the requrement for finerdiameter wires consistent with higher interconnectdensities.

Furthermore, ongoing progress insemiconductor technology means thatthe upper operating temperature of de-vices is being progressively raised whichreduces the need for forced cooling withits associated space and colt penalties.Devices capable of continuous serviceat 300 °C are currently available andmicron valves [4] are likely to be capableof operating in even harsher environ-ments. Pure gold will anneal throughrecrystallisation and regrowth of the de-formed grains at temperatures as low as100 °C. When the grains grow to dimen-sions that are comparable to the diameterof the wire, the strength and fatigue prop-erties of the wire greatly deteriorate.

The purpose of introducing solid so-lution strengthening elements is not onlyto improve the strength of the wire but,more particularly, to increase the recrys-

tallisat on temperature of the Gold workedmicrostructure. By a judicious selection ofthe doping elements, it is possible to in-crease the recrystallisation temperature ofgold bond wire up to a maximum ofabout 350 °C without sign ficantly de-grading its electrival characteristcs [5].At first sight this might seem adequatebut recrystallisation temperature is de-fined as `the temperature at which recrys-tallisation is complete within one hour',This is an insufficiently demanding cri-terion for fine wire interconnects becausethe wire is often required to maintain itsmechanical integrity over a long servicelife. The aluminium and copper alloysused for bond wire anneal at tempera-tures comparable to those for the goldalloys, so that all commercially availablebond wire will soften progressivelythrough the Toss of its work-hardened

microstructure and grain growth. A collation of dataculled from the literature, presented in Figure 8 (page139), shows that when exposed to temperatures above300 °C, conventional gold alloys used for bond wire losehalf of their strength in under one hour.

Inner leads bond to0contact pads on the

semiconductor

Polymeric tapecontaining sprocketdrive holes

Window throuah whichouter lead bond is made

Figure 4Schematic illustration of tape automated bonding (TAB):

A semiconductor device with beam leads attached,held on polymeric tape

Gold Bull., 1992, 25 (4) 135

Heated bonding tool 1

Toolmovement

— Beams --Solder

I - Silicon chip

Tape leadframe

Gold bumps Bond pad

Solder reflowed on inner lead

Bondingtooll2

Substrate

Solder reflowed on outer lead

Figure 5The bonding sequence

used in TAB

Step 1:Start of the bonding cycle

Step 2:Inner ends of the beam leadsare joined to the component,which then appears as shown

in Figure 4, and in which stateit can be electrically tested

Step 3:Outer lead bonding operation,

at the completion ofthe joining cycle

that possess electrical conduc-tivities close to those of gold,copper and aluminium whichare currently used in this appli-cation.The candidate materialsmust also be readily fabricatedinto fine wire, and be compat-ible with existing microweld-ing machines if they are to beadopted readily by the semicon-ductor industry. This article de-scribes the development of anew gold alloy bond wire that iscapable of meeting this set ofrequirements.

REVIEW OFMETALLURGICALSTRENGTHENINGMECHANISMS

Any enhancement in the reststance of bond wire toelevated temperatures can only be achieved by chang-ïng to new materials which can be strengthened bydifferent mechanisms. The choice is limited to metals

Figure 6

Schematic illustrationof flip-chip bonding:

Step 1:The contact pads on the

semiconductor device are onthe same face as

the electronic circuit

Step 2:The pads are plated

selectively with solder

Step 3:By inverting the chip these pads

are made to register withcontact pads on the circuit

board and the joints areformed by reflowing the solder

As a first step towards developing a new high strengthbond wire, it is necessary to select a material that is intrin-sically strong, or can be strengthened, and which willretain the enhanced strength when exposed to ele-

vated temperatures. Intrinsi-cally strong metals that can bedrawn to fine wire such as steeland tungsten are not amenableto micro-welding and have lowelectrical conductivity, so thatattention focused instead onmethods of strengthening gold.

Work hardening, achievedby mechanically deforming themetal so as to create disloca-tion tangles and other defectsin the material, and solid so-lution strengthening, involv-ing sub-percentage additionsof a second constituent, areused in the existing range ofbond wire metals. Strengthen-ing of the wire by work harden-

136 Gold Bull., 1992, 25 (4)

ing through cold drawingor extrusion operations is re-moved by relatively short,low temperature heat-treat-ments. On the other hand,solid solution strengthen-ing is only capable of mod-erate improvements in me-chanical propertjes.

More substantial im-provements in strength canbe achieved by introducingparticles of a second phaseinto the host metal. Thereare three different ways inwhich this can be done.One is to create a fine dis-persion of intermetallic pre-cipitates in the metal to pro-duce what is known as pre-cipitation strengthening. Inorder to obtain the neces-sary fine dispersion of thesecond phase, the requisite

Table 1Comparison of semiconductor chip interconnection technologies,

with representative data

TapeWire automated Flip chip

Parameter bonding bonding bonding

Gold or Copper beamsJoining materials aluminium attached with Pb-5Sn solder

fine wire Au-20Sn solder

Bond strength as made,g 10 50 30

After burn-in testing, g

element is able to diffuse to the surface. This can resultin a fine internal dispersion of stable oxide particles.Once formed they are stable because zirconia is insol-uble in platinum.

Metals may also be strengthened by the deliberatemixing in of much larger (0.1 to 100 µm diameter),insoluble, particles and these are generally referred to as metalmatrix composite materials (MMCs) , The strengtheningmechanism in this case is different and, in any case, is notapplicable to fine bond wire, which can have a diameter ofas little as 4 µm for microwave applications.

The attainment of dispersion strengthened highgold alloys of bulk geometry by internal reaction atelevated temperature is not likely to be viable becauseoxygen and other constituents of the atmosphere arenot soluble to any appreciable extent in gold [11]. Thisappeared to leave precipitation strengthening as theonly foreseeable means for producing high strengthgold- based bond wire.

characteristics are obtained via a precipitation strengthen-ingmechanism [13, 14, 15, 16].

The hartlening characteristics of this precipitationstrengthened alloy in bulk form are well establishedand are indicated in Figure 9. The curves representedindicate that the alloy has an exceptionally high resis-tance to over-aging at temperatures below 400 °C. Thisproperty, in particular, suggested that the Au- lwt.%Tialloy would be a promising candidate for an improvedbond wire. Furthermore, the high gold content of thealloy made it likely that wire produced from it wouldmatch conventional gold alloy bond wire in its func-tional characteristics. Accordingly, ït was decided tocarry out a detailed evaluation of the Au- lwt.%Ti alloyas fine wire.

FINE WIRE OF THE Au-Iwt.%Ti ALLOY

PRECIPITATION STRENGTHENEDHIGH-GOLD JEWELLERY ALLOY

In 1988, the development of a new gold alloy, of com-position Au- lwt.%Ti, for high carat jewellery applica-tions was reported [12]. This alloywith a 990 millesimalfineness (i.e. 99.0 % gold) can be endowed with themechanical properties of a 9carat alloy while retaining thehue and surface brilliance of 24carat gold [ 13] . These improved

Figure 7

Added value ofgold alloy bondwire in terms ofa mark-up onthe goldprice, as a function ofwire diameter. The increasing

pricepremium of wire of reduceddiameter reflects the escalating

processing effort needed

A 1 mm diameter rod of the Au-lwt.%Ti alloy,supplied to the jewellery specification, was heat-treated at 800 °C for one hour in high vacuum, toplace the titanium in solid solution. It was thenwater quenched, and cold-drawn to 25 µm diame-ter in 20 stages. No intermediate annealing treat-ments were used as these would have initiated theprecipitation process.

1000

x

100

-a- 10

11 10 100 1000

Wire diameter, µm

138 Gold Bull., 1992, 25 (4)

20

15

200 °C

]o-350 °C

5 500 °C

800 °C

01E-3 0.1 10 1000 1E6

Heat-treatment time, s

Figure 8Annealing characteristics ofdoped gold bond wire alloys.The short heating times wereachieved using an electricalpulseheating method

By contrast, a conventional gold alloy for bondwire requires upwards of 120 drawing stages and severalintermediate anneals to prevent breakage, emphasis-ing that the Au-lwt.%Ti wire possesses a superiorresistance to necking and consequential fracture.

Samples of the fine wire were heat-treated in air ata range of temperatures between 25 and 400 °C toinduce precipitation strengthening. Strength data ob-tained from these samples revealed the following dis-tinctive features:a) At high and low precipitation treatment tempera-

tures, the wire exhibits classical precipitationstrengthening behaviour as shown by the data

presented in Figure 10, ob-tained at room temperatureand 400 °C. The heat-treat-ment required to produce themaximum strengthening effectis governed by the the tempera-ture and time of the process.

These two parameters are functionally related, as shownin Figure 11. It can be seen from this graph that theheat-treatment temperature required to obtain peakstrength is proportional to the logarithm of time, for both25 µm and 1 mm diameter wire. The similarity in therelationships for the two wire diameters indicates thatthe same strengthening mechanism is operative for both.Because it has been established that precipitation strength-ening accounts for the mechanica) properties of 1 mmdiameter wire, this same mechanism must also be respon-sible for the strengthening of 25 4m diameter wire.The offset with respect to time of the precipitationstrengthening data for the Au-lwt.%Ti alloy as fine

200

400 °C150

00oC> 600 °C

2 100ac

a

50

00.1 1 10 100 1000

Heat-treatment time, h

Figure 9Precipitation hardeningcharacteristics of theAu-1 wt. %Ti alloy, inbulk form, on heat-treatmentat 600, 500 and 400 °C.Prior to the aging treatment, thematerial was solution-treated byheating at 800 °C, in highvacuum, for one hour and waterquenched to room temperature

Gold Bull., 1992, 25 (4) 139

Figure 10

Precipitation strengtheningcharacteristics of the

Au-1 wt. %Ti alloy, in the formofa 25 µm diameter wire,

at 400 ° C, 200'C androom temperature.

Classicalprecipitation hardeningbehaviour is exhibited on

aging at 400°C and roomtemperature, but the sample of

fine wire aged at 200 ° Cshows a lower peak strength and

minimal tendency to over-age

1000 —

200 °C ^^i

a 800 `

400 °C 25 °Csrná 600

4

•H \

,°—' 400\^

20010 100 1000 1E4 ]E5

Heat-treatment time, h

b)

c) At temperatures below 400 °C but above about100 °C, there is a similar enhancement of strengthbut no over-aging is observed. This behaviour isconsistent with dispersion strengthening, as de-scribed above. The dispersed phase in this case islikely to be an oxide of titanium. The formation ofsuch compounds by reaction with the atmosphereis possible because the maximum distance the oxy-gen has to diffuse to reach the centre of the finewire is only 13 µm. This could account for disper-sion strengthening occurring in fine wire of thealloy while it is largely absent in thicker forms ofthe material. The transition to regular precipita-

600

V

o^ 400

aE

200

00.1 1 10 100 1000 1E4

Heat-treatment time, h

Figure 11

The relationship betweenaging heat-treatment time

and temperature to achievepeak tensile strength in the

Au-]wt. % Ti alloy when in theform of25 µm and 1 mm

diameter wires

wire, compared to that for the thicker wire, can beascribed to the annealing out of dislocations andpoint defects which are extensive in the heavilyworked fine wire. These annealing processes havelower activation energies than that required to initiateprecipitation strengthening. Hence, during theheat-treatment, these processes occur first andthere is a delay before precipitation strengtheningcommences in the 25 µm wire.Heat treatment at 400 °C and above results inthe strength of the wire increasing to a peak of1000 MPa, followed by a decline through over-aging. This compares with a typical peak value of275 MPa for conventionalgold bond wire of 25 tmdiameter.

o

0

^

0^

Dmm

140 Gold Buil., 1992, 25 (4)

tion strengthening above 400 °C must thereforemark the point at which the race of precipitation ofthe intermetallic phase overtakes that at which theminority phase within the alloy is able to react withthe atmosphere.

d) When the wire is left in an ambient environment,it again precipitation strengthens and over-ages inthe classical manner. One possible explanation forthis phenomenon, which is not observed in bulkalloys at ambient temperaturen, is that precipitationis occurring under the driving force of the storedenergy imparted to the material on cold workingto fine wire. This explanation is consistent with theobservation that pure gold (> 99.999 % purity),when subjected to heavy mechanica) cold work,recovers and recrystallises during storage at roomtemperature [17]. At ambient temperature, there isinsufficient activation for the development of adispersed refractory phase by reaction with theatmosphere.

formed. A reel of wire of the Au-lwt.%Ti alloy wasprepared as described in the previous section. Afterbeing drawn down to 25 µm, the wire was cut in twoand one of the two equal lengths was again solution-treated, quench-cooled and then aged in an ambientatmosphere, while the other half was similarly proc-essed hut in high vacuum throughout. The results aresummarised in Table 2 (page 142) and the aging curvengiven in Figure 12.

The second solution-treatment that was pro-vided removed the work hardening produced by thecold drawing to fine wire, thereby reducing thetensile strength of the wire. The strength of the wirethat had been heat-treated in high vacuum decreasedfrom 275 MPa to 90 MPa, which is close to thetensile strength of annealed gold. On the other hand,the wire exposed to the ambient environment didnot soffen to the same extent, with its strength de-creasing to 175 MPa. By implication, there must besome type of mechanism acting that either prevents thealloy from annealing completely or provider somestrengthening.

On aging at 350 °C, in their respective environ-ments, the two wire samples behaved in a totally dif-

TESTS FORDISPERSIONSTRENGTHENINGIN FINE WIRE

The conformance of the heattreatment data on fine wire toclassical precipitation strength-ening behaviour in the highand ambient temperature re-gimes is sufficiently convincingto provide confidence aboutthe interpretation. On theother hand, the assumptionthat dispersion strengtheningoccurred in the intermediatetemperature range was not aswell supported and so there wasneed to obtain some additionalevidence. For this purpose, thefollowing experiment was per-

300

0

200/ ambient

t /highc 100 Vacuum

2

01 10 100 1000 1E4

Heat-treatment time, h

Figure 12Precipitation strengthening characteristics of25 tm diameter wireof the Au-1 wt. % Ti alloy, at 3500 C, in high vacuum and in air

Prior to the aging treatment, the two samples of wire were solution-treated,at 8000 C, for one hour, in high vacuum and in air respectively,

followed by a water quench

Gold Bull..1992, 25 (4) 141

25

20

15a0

10

U-

5

00 1 10 100 1000 1E4

Heat-treatment time, h

new bond wire

tt

.

conventional gold bond wire

Figure 13Failure load of25 gm diameterwire of half-hard pure gold andof the new Au-1 wt. %Ti, gold-

clad, bond wire as afunction ofheat-treatment time at 400 ° C.

Conventionalgold bond wirerapidly soffen in response to theheat-treatments associated withsemiconductor device packaging

and use, while the new bondwire is clearly more resisistant tosuch treatment. The initial dropin strength of the new bond wire

on heat-treatment is due toannealing of its gold cladding

ferent manner. The tensile strength of the wire that hadbeen aged in high vacuum increased three-fold to apeak of 240 MPa, after about 15 h heat treatment, andthen declined. This response was entirely characteristicof a precipïtation-strengthened material and thethree-fold strength enhancement was similar to thatmeasured for the alloy in bulk form (see Figure 9,page 139). Furthermore, the 15 h of heat-treatmentrequired to reach peak strength is in close agreementwith the duration predicted from Figure 11 (page140) for 1 mm diameter wire. As indicated in Fig-ure 12, the wire that was exposed to air throughout the

Table 2

solution and aging treatments was essentially unaf-fected by the heat-treatment even after 5,000 h at350 °C. The stability of the mechanical propertiesof the wire at this temperature suggests that the proc-essing that it had received had endowed it with disper-sion strengthening.

The ability to dispersion harden a Au-lwt.%Tialloy would appear to fly in the face of the fact thatoxygen has an extremely low solubility in gold. How-ever it is known that small additions of iron and tinsignificantly enhance the diffusion of oxygen intogold alloys, which occurs along grain boundaries

118]. The high defect density presentin the gold-titanium alloy matrix,when cold-worked to fine bondwire, might therefore provide aroute for oxygen diffusion. The dif-

ygfused-in ox en can then react ith

Tensile strength of 25 µm diameter wire of the Au-1 wt. % Ti alloyfollowing heat-treatment in air and high vacuum

Tensile strength, MPa, in:

AmbientHigh vacuum atmosphere

Sequentialprocessing steps

a) as draven 275 275b) 800 °C, for 1 h,

90 175quench cooledc) 350 °C, for 15 h 240 175d) 350 °C, for 5000 h 80 175

the titanium inside the alloy to forma distribution of fine oxide particles.In the gold-rich alloys containingsmall amounts of iron and tin, thedistance that the oxygen diffused inwas restricted to about 10 µm, whichis consïstent with dispersion strength-ening of the Au-lwt.%Ti alloy onlybeing evident in fine wire of this ma-terial. Furthermore, dispersionstrengthened gold-titanium alloyscontaining love concentrations of ti-

142 Gold Bull., 1992, 25 (4)

Figure 14

A thermomechanically strengthened wire of the Au-1 wt. % Ti alloy,25 gm in diameter, supporting a weight of 50 g.

Measurements of the tensile strength of this wire indicate thatit could in principle support a load of 150 g

tanium (0.04 to 0.08 %) have been prepared by spray-ing the molten metal into air [19]. It is a prerequsite ofthis process that the spray is in the form of fine droplets,less than 5 µm diameter, in order to oxidise all of thetitanium in the short time that the alloy is hot.

Having established that strengthening of a Au-lwt.%Ti fine wire can be achieved by either the pre-

cipitation of a finely divided intermetallic phase or theformation of a dispersed refractory phase, or by a com-bination of both types of inclusion, the appropriateprocessing conditions for achieving particularstrengthening characteristics were established by ex-periment. Under a particular regime of thermome-chanical treatment it was found possible to selectively

develop the dispersed refractory phase inthe fine wire and thereby erevent loss ofstrength by over-aging.

A fabrication route was devised thatyielded fine wire with a tensilestrength close to 400 MPa, which re-mained stable at this value even afterbeing held for 10,000 h (one year) at400 °C and more than 1,000 h at 600°C[20] . This contrasts with the rapid de-cline in mechanical properties of a con-ventional gold bond wire, subject to thesame heating regime, as shown in Fig-ure 13.

The high strength that can be devel-oped in the Au-lwt.%Ti alloy by a suit-ably optimised thermomechanica) treat-ment is demonstrated in Figure 14,which shows a 25 µm diameter wire sup-porting a 50 g weight, following its heat-treatment in air at 400 °C for a year. Thiswire was fabricated with a remarkabletensile strength of 1540 MPa and anelongation to failure of 6 %. The attainedstrength is equivalent to that of pianowire, when scaled to 25 gin diameter. Bycontrast, pure gold wire would be inca-pable of supporting even the paper clipshown after being subjected to an identi-cal heat treatment!

Wire of this strength is not desirablefor bond wire because it possesses exces-sive stiffness and hardness for intercon-nection of semiconductor devices butcould find application in woven decora-tive products, as the 25 4m wire is actu-ally finer than a cotton thread. Figure 15shows a woven multistrand tube madewith 100 tm diameter wire of the Au-lwt.%Ti alloy.

Gold Bull., 1992, 25 (4) 143

SEMICONDUCTORINTERCONNECTION USINGTHE NEW BOND WIRE

Preliminary trials have been made using the new highstrength bond wire for making interconnects to semi-conductor devices. Bonds to aluminium and gold padshave been successfully made using the wire on conven-tional wire bonding machines, although the machinesettings needed to be slightly altered in order to obtainsatisfactory microwelds. However, by applying a clad-ding of pure gold to the wire, the Au-lwt.%Ti alloy ismade virtually indistinguishable from conventionaldilute gold alloy bond wire in both its electrical prop-erties and bonding characteristics. The latter embracedthermocompression wedge-wedge, thermosonicwedge-wedge and thermosonic ball-wedge weids. De-spite the enhanced tensile strength of this wire, thehardness of the Glad wire is comparable to that of theconventional product, making it suitable for intercon-nects to the relatively delicate integrated circuits of

gallium-arsenide. These chips are capable of operatingat higher temperatures than those of silicon and thebenefits of providing them with interconnects thatare mechanically stable at elevated temperatures areobvious.

CONCLUSIONS

The Au-lwt.%Ti alloy can be substantially strength-ened by a fine dispersion of two types of second phaseparticles. The alloy in bulk form can be reïnforced byparticles of an intermetallic phase that precipitates inresponse to a suitable heat treatment. When preparedin the form of a fine wire, less than 100 gm diameter,ït can also be strengthened through the formation of adispersed refractory phase in response to heating in anambient atmosphere. Thereby, fine wire can be madestable to extended heat treatment at elevated tempera-ture. Other characteristics of this material, namely its

Figure 15An 8-strand braid made with 100 µm diameter wire of the Au-1 uit. %Ti alloy:

The sample demonstrates the ability to produce coreless open basket braid with full control of the pitch anddiameter of the braid, using fine wire of this alloy. Conventional braided products

usually comprise 16 or more strands and can therefore be made correspondingly more elaborate.The braiding trial was carried out by G. Cooke of J.R. Cooke and Sons, Wotton-under-Edge, Gloucestershire

144 Gold Bull., 1992, 25 (4)

conductivity and weldability, are closely similar tothose of conventional dilute gold alloys used for bondwire. The beneficial properties of fine wire of theAu-lwt.%Ti alloy are currently being exploited ininterconnects for new generation semiconductor de-vices, and in particular for micron valves.

ACKNOWLEDGEMENTS

The World Gold Council and GEC-Marconi Limitedare acknowedged for supporting the work. The contri-bution of the World Gold Council in supplying sam-ples of the Au- lwt.%Ti alloy is also gratefully acknow-ledged. Mr. G. Cooke of J. R. Cooke and Sons pro-duced the braided tube shown in Figure 14 (page 143).

REFERENCES

1. S. Prasad & A. Saboui, 'An improved wire bond pulltest', Solid State Technology, 1991, 34(6), 39-41

2. R. Bidin, `High pin count wirebonding: The challengefor packagïng', Solid State Technology, 1992, 35(5),75-77

3. `Gold 1991', Gold Fields Mineral Services Ltd., Lon-don, 1991N.A. Cade & R.A. Lee, `Vacuum microelectronics',GECjournalofResearch, 1990, 7(3), 129-138S. Tomiyama & Y. Fukui, `Gold bonding wire forsemiconductor applications', Gold Bull., 1982, 15(2),43-50

6. H. Izumi, M.. Bapna E.H. Greener & M. Meshii,`Precipitation hardening in gold-based iron alloys',Proc. Conf., Sixth International Congress for ElectronMicroscopy, 1966, Kyoto, 395-396

7. E. Raub & M. Engel, `The alloys of zirconium withcopper, silver and gold', Z. Metallkunde, 1948, 39(6),172-177

8. M. Graham, `Precipitation hardening in gold-titaniumalloys', Proc. Conf. Thirty-first Annual Meeting of theElectron Microscopy Society of America, 1973, NewOrleans, 148-149

9. G.L. Selman, J.G. Day & A.A. Bourne, `Dispersionstrengthened platinum', Platinum Metals Review, 1974,18(2), 46-56

10. A.E. Heywood & R.A. Benedek, `Dispersion strength-ened gold-platinum', Platinum Metals Review, 1982,26(3), 98-104

11. M. Poniatowski & M. Clasing, `Dispersion hardenedgold: A new material of improved strength at hightemperaturen', Gold Bull., 1972, 5(2), 34-36

12. A.M. Tasker, `The promise of 990 gold', Aurum, 1988,34(Summer), 62-67

13. G. Gafner, `The development of 990 gold-titanium: Itsproduction, use and properties', Gold Bull., 1989,22(4), 112-122

14. M.E. Graham, 'Precipitaion strengthening of an Au-4at.%Ti alloy', Ph.D. Thesis, North Western Univer-sity, Evanston, Illinois, USA, 1974

15. D.P. van Heerden et al., `Precipitation in rapidly solidi-fied Au-Ti alloys', Materzals Letters, 1991, 10(9), 425-430

16. `990 gold', Gold Technology, 1992, 6(5), 1-12 (fourarticles, entire issue)

17. H. Ramsey, 'Metallurgical behaviour of gold wire inthermal compression bonding', Solid State Technology,1973, 16(10), 43-47

18. H. Ohno & Y. Kanzawa, `Internal oxidation in goldalloys containing small amounts of Fe and Sn', j Mate-rials Science, 1983, 18(3),919-929

19. A.S. Darling, `Improvements in and relating to thedispersion strengthening of metals', United KingdomPatent No. 1 280 815, filed 14 July 1969

20. G. Humpston & D.M. Jacobson, 'Methods of makingelectrical conductors', United Kingdom Patent Appli-cation No. 2235211, United States Patent No5073210, European Patent Application No90305266.0

Gold Bull., 1992, 25 (4) 145