Gold/Chromium Metallizations forElectronic Devices

MANUFACTURING TECHNOLOGY AND BEHAVIOUR

Paul H. HollowayUniversity of Florida, Gainesville, FL., U.S.A.

The thin films of gold which are applied so extensively in electronic

devices are normally deposited over thin films of other metals in two,

three or even four component metallization systems. The problems

arising in such systems are vividly illustrated in this article on the

gold/chromium metallization system developed in large measure at th.e

Sandia Laboratories in Albuquerque, New Mexico.

Three main criteria are applied in the selection ofmetallization systems used in electronic devices.Firstly, these systems must have good conductivityto minimize power losses. Gold, with a resistivity of2.4 mn xcm, fulfils this condition and can toleratehigh current densities. Secondly, metallizations mustbe inert or remain passive in operating environmentsif a high level of reliability is to be achieved. In thislatter respect, gold is unequalled as an outer,layer formetallizations because it does not form corrosion pro-ducts in severe environments. Thirdly, metallizationsystems should be compatible with manufacturingprocesses. Anderson (1) and Mattox (2) have reviewedsome of these processes and their operatingparameters. Thin film systems can be categorized ac-cording to whether one, two, three or four corn-ponents are used (Table I) and, in general, it isdesirable to use the least number of metals in order tosimplify their preparation. Thus, the one-componentsystem of aluminium has been used extensively andwith great success in applications where corrosionresistance is not critical.

Apart from meeting the first two criteria listedabove, gold is also compatible with beam-leaded andlead frame inter- and intraconnection technologies, ithas substantial resistance to electromigration and canreadily be deposited in thin layers by a variety of wetor vacuum processes. However, although Mattox (2)lists gold as a one-component metallization, it is notfrequently used as such because it does not adherewell to oxide (or oxidized) surfaces (1,3). One or moremetals are normally used at the gold-substrate inter-face to ensure strong bonding. Of the two-componentsystems listed in Table I, gold/chromiumm?is the mostresistant to atmospheric corrosion (4 to 8) and itsresistivity is generally lower than that of gold-palladium/chromium because of alloying effects bet-ween gold and palladium (7). The presence ofpalladium or platinum as x in three-componentsystems of the gold/x/titanium type improves the cor-rosion resistance over that of gold/chromium (4), butalso increases the metallization resistivity becauseof interdiffusion between gold and palladium or=.:•=•platinum (9,10). Unalloyed palladium is not as inert

Table 1Thin Film Metallization Systems Classified According to the Numb"er of Components Used.

In part after 12)

One-component Two-component Three-component Four-componentsystems systems* systems* systems*

Al Au/Ti Au/Pd/Ti Au/Rh/Pd/TiCr Au/Nichrome®. Au/Pt/Ti Au/Ni/Cu/Ti

Nichrome® (NiCr) Au/Cr Pd/Cu/Nichrome®W Au/Mo Pd/Cu/TiMo Au/Nb Au/W/TiAu AuPd/Cr Au/Mo/Ti

*The sequence of the components is from the free surface towerds the substrate, the sequence of deposition during manufacture is therefore the

reverse of that indicated here.

99

Fig. 1 This enlarged photographof a typical hybrid microcircuit(HMC) made at SandiaLaboratories shows the whitealumina substrate (25 x 25 x 1.5mm), the gold/chromiummetallization and the grey-brown tantalum nitride thin filmresistors. Some of the metallizedconductor segments are inter -connected by thermocompres-sion bonded thin gold wires.The active electronic functionsof the circuit are performed bythe square appliqué parts whichappear as black or grey on thisphotograph. Electrical connec-tion to the device after encap-sulation is by means of the gold -plated lead frame which is ther-mocompression bonded to themetallized pads at the edges ofthe assembly

as gold and is unsuitable as a substitute for it in high-reliability devices.

While the gold/chromium system is relatively im-mune to corrosion in normal atmospheres, this is notthe case when even trace amounts of halogens are pre-sent (4,5,11,12). Furthermore, chromium is verymobile in the grain boundaries of gold. Both of thesephenomena may cause problems during manufactureof devices. Although gold/chromium is used tometallize integrated circuits (1,6,13) and capacitorterminations (14), the author's experience has beenprimarily with its use on hybrid microcircuits(HMC). This HMC technology, developed at SandiaLaboratories, will be used to illustrate typicalmanufacturing techniques and problems which maybe encountered with the gold/chromium system.

HMC Manufacturing ProcessA typical HMC, as shown in Figure 1, consists of

an alumina substrate, 25 x 25 x 1.5 mm in size, uponwhich tantalum nitride is reactively sputtered to forenthin film resistors. The conductors and bonding padsconsist of a chromium adhesion layer some 30 nmthick, with a gold layer, some 3 to 6µm thick, sequen-tially deposited on top of the tantalum nitride. In-traconnections are made by thermocompression (TC)bonding (15) fine gold wires or gold beam-leadeddevices. Interconnections are made with TC bondedgold-plated lead frames.

The HMC technology developed at the SandiaLaboratories consists of the following processingsteps:(1)Cleaning of the alumina substrates(2)Sputtering of tantalum nitride thin film resistor

material (50 nm)(3)Evaporation of chromium and gold film con-

ductors (30 nm chromium, 3 000 to 6 000 nmgold) in succession

(4)Resist application, conductor definition andetching

(5)Resist application, resistor definition and etching(6)Resistor stabilization by heating in air for.hours

at 300°C(7)Substrate sizing with a laser scribe and breaking(8)Cleaning(9)Bonding of lead frames

(10)Assembly of appliqué parts and beam-leadeddevices, bonding of interconnecting wires

(11)Electrical testing(12)Attaching of the lid using an epoxy adhesive.

This technology is further illustrated by thematerial-flow diagram of Figure 2. The conductorpattern is defined using a potassium iodide solutionsaturated with iodine (KI + 1 2) to etch gold and aceric ammonium nitrate (CAN) solution to etchchromium (17). Additional details on these steps maybe found in reviews (16,17,18) or in actual processspecifications (19).

100

Substrate cleaned after step (1)

Gold deposited after step (3)

TaN deposited after step (2)

Conductor mask used in step (4)

Chromium deposited in step (3)

Conductor pattern photolithographedin step (4)

5..5,

Conductor pattern etched Resistor-conductor mark Resistor-conductor pattern Resistor-conductor patternafter step (4) used in step (5) photolithographed in step (5) etched after step (5)

---- --__

Substrate sized and lead frame Beam-leaded devices, appliqué parts and Completed and sealed deviceattached after step (9) interconnecting wires bonded after step (10) after step (12)

Fig. 2 Illustration of some of the processing steps in the manufacturing'technology of HMC's with gold/chromiummetallization which was developed at Sandia Laboratories in Albuquerque, NM., U.S.A. Additional details on..the processare given in the teat of this article

Deposition of the gold/chromium metallization iscarried out on carefully stored, oxidized tantalumnitride surfaces without breaking vacuum and atpressures (prior to evaporation) below 3 x 10 -4 Pa.Data by Speight et al. (4) suggest that the contactresistance of these metallizations may be sensitive tothe background pressure during deposition.

Substrate temperatures ranging from 200 to 350°Chave been used with good results for the deposition ofgold/chromium. Data by Clay et al. (20) show that thechromium resistance decreased with increases insubstrate temperature while the reverse was observedfor the gold. These effects result from recovery of thechromium film and from enhanced diffusion ofchromium into the gold film with increasingtemperature respectively. The grain size of the goldincreased with the substrate temperature, probablybecause the resulting higher atomic mobility favoursgrowth but not nucleation. The optimum substratetemperature seemed to be about 300°C (20).

Since chromium is only used as an adhesion layer,its thickness can be minimal. Weaver and Hill (21)

found that 40 nm of chromium was required for goodadhesion of gold to alumina, while Haq et al. (3)reported good adhesion for only 1 nm of chromiumand studies at Sandia Laboratories (11) indicated goodadhesion for 10 nm. In practice the chromium filmthickness is therefore specified to be 30 ± 10 nm toallow for process variation. Clay et al. (20) report thatthe optimum deposition rate is from 0.1 to 0.3 nm/s.If extremely high deposition rates or very thick films(more than 100 nm) are used, the films may containstresses of such magnitude as to cause delamination (2).

Gold thicknesses of 3 µm are normally used fortypical HMC's although 6 pm films are used for radiofrequency applications. The gold film must be sufl -i-ciently thick to minimize chromium diffusion to thefree surface (see below) and to minimize the numberof pinholes through which corrosion could occur(12,22). Optimum deposition rates are about 2 to 3nm/s (20) although rates up to 10 nm/s have been suc-cessfully used. As for chromium, internal stresses canbe excessive in thick gold or in films deposited at highrates (2): The deposited films have a rough nodular

101

Fig. 3 Scanning electron micrograph of a thin film of golddeposited onto an alumina substrate. The nodulartopography of the gold is related to that of the substratebut does not reproduce it in every detail. Also, individualgold nodules are not necessarily single grains

appearance (Figure 3) which is similar to that of thealumina substrate. It is often assumed that thenodules replicate the substrate and that they corres-pond to individual gold grains — neither is true.While the substrate topography certainly influencesthe gold nodule size, other factors such as substratetemperature, deposition rate and chromium thicknessalso affect the topography of the gold (23).

By thinning gold films and examining them withtransmission electron microscopy, Holloway andNelson have shown that Bach gold nodule can consistof several grains (23). Thus, the microstructure of thegold film is related to processing variables bymechanisms which are stijl poorly understood.

Chromium Diffusion Effects

While the manufacturing procedure detailed aboveproved generally satisfactory, the metallizationresistivity increased (see Table II) during thestabilization step (6) and problems were encounteredat steps (9) and (10) during TC bonding of lead framesand fine wires (17,20,24,25,26). The origin of thebonding problems was also traced to step (6), inwhich substrates with resistors and conductors defin-ed are heated in air at 300°C for 2 hours to oxidizethe tantalum nitride films. This treatment stabilizestheir resistance against change during their 20-yearexpected lifetime. Data in Table II show that beforestabilization, the average 90° peel strength (25,27) ofbonded lead frames was 8.4 N and all failures werelocated at the heel (that is, at the edge of the bondingCool deformation area). After resistor stabilization, theaverage peel strength was 5.8 N and most failureswere delaminations (that is, separation at the originalbonding interface). Auger electron spectra shown inFigure 4 clearly demonstrated that the resistorstabilization treatment led to the formation of Cr20 3on the gold surface. The oxide layer was determinedto be approximately 0.8 nm thick by a signal intensityanalysis (28). To verify that such a thin film coulddegrade TC bondability, thin layers of Cr 20 3 werevapour-deposited onto gold and their effect upon leadframe peel strength was measured. The results (TableII) show less effect for 0.6 but more for 1.5 nm ofvapour-deposited Cr 20 3 as compared to the resistorstabilization treatment. It can therefore be concludedthat films of Cr 20 3, 1 nm or less in thickness, canseriously degrade the TC bondability of gold.

In step (6) of the manufacturing process, Cr 203must form on the free surface of the metallization bydiffusion of chromium through gold. Such diffusionhas been studied by resistivity (29,30), work function

TabEffect of Hybrid Microcircuit Res !tor, StabilizContamination by Cry03hon a• Bon-dability of G

of Gold C

letlation Treatment 12 Hours at 300°C in Air) and ofold/Chrom Miuín etallizations and thé Resistiviiyónductors

Averàge TC tCondition lead frame fail

peel strength m

N.

as deposited 8.4 100°/resistor stabilized 5.8 90% dek

Cr203 deposited

80% delamiion gold

20% he

Cr203; deposited

mu ro aeianon gold

Cr203 Goldthickness metallization

as determined resistivityby AES

nm µS2 x cm

cel 0 3

sliatión 0.8' 3.7

iation 0.6el

nation ` 1.5

ode

oh

102

.AIR

;:..Au

3Y, eV

Fig. 4 Auger electron spectra from the surfaceof a gold/chromium metallization (a) asdeposited and (b) after stabilization of the AS DEPOSITEDtantalum nitride resistors by heating at 300°C / xtofor 2 hours in air. During the heat treatment, Xt /quite significant amounts of chromium have j, b

(31), Rutherford backscattering (32), electronmicroscopy (33), electron probe microanalysis (34)and in particular detail using Auger electron spec-troscopy (AES) and ion scattering spectroscopy(35,36). The resulting data clearly establish that grainboundary diffusion is the dominant transportmechanism by which chromium migrates to the sur-face of the gold where it forms Cr 20 3 (23,35,36).Grain boundary diffusion also explains whymetallization resistivity increases so dramaticallyduring resistor stabilization. Chromium rapidlymoves throughout the thickness of the gold films viagrain boundaries from where some of it also diffusesinto the bulk of the gold grains. This bulk contamina-tion of gold by chromium results in resistivityincreases of 25 per cent in only 2 hours at 300°C.

Another effect of grain boundary diffusion was firstnoticed by Kendrick (33) and then studied in detailby Hampy et al. (37). As a result of diffusion,chromium is preferentially removed from the adhe-sion layer where gold grain boundaries emerge. Anycounter-flux of gold is insufficient to compensate forthe chromium flux, thus a labyrinth of voids iscreated as shown in Figure 5. Since this labyrinth of

Fig. 5 Scanning electron micrograph of a thin film ofchromium on glass after a simulated resistor stabilizationheat treatment (2 hours at 300°C in air) and subsequentremoval of the gold layer that had initially been depositedabove it. During heat treatment, grain boundary diffusionof chromium through gold took place which resulted inpreferential removal of the former metal where gold grainboundaries emerged against the chromium adhesion layerat the metallization-glass interface. Thus, the pattern ofchromium islands separated by a labyrinth of voids, whichis arrowed here, resulted. After (37)

voids is connected to the environment at the edge ofconductors, corrosion of the remaining chromium`islands' may be greatly enhanced as discussed below.

To produce reliable HMC's, Cr 20 3 formation mustbe prevented or, if the oxide forms, it must bechemically removed from the gold surface prior toTC bonding. Holloway and Nelson (38) have shownthat subsequent chromium diffusion is considerablyslowed if gold deposition is interrupted and thedeposition chamber is filled with oxygen to pressuresof around 10 -3 Pa during step (3) of the manufactur-ing process. With the introduction of such a treat-ment, the lead frame peel strengths and failure modeswere the same before and after resistor stabilization.Furthermore, the gold resistivity was approximately2.6 rather than around 3.6 i 1 x cm after the standarddeposition procedure.

103

Edge Pinhole AttackAttack

Fig. 6 Evidente of undereutting and pinhole attack of agold/chromium metallization on glass by ceric ammoniumnitrate (CAN). Prior to the attack, this sample was suhmit-ted to a simulated resistor stabilization treatment of 2hours at 300°C in air. Data by Palmer and Wright (12) sug-gest that the extent of attack is lesser after stabilizationthan immediately after deposition but there are indicationsthat this differente may be temporary only X 120

06 WEEKS STORAGE BETWEEN STABILIZATION AND LEACHING•NO STORAGE BETWEEN STABILIZATION AND LEACHING

••Z

z •w •• ••

a Tw 5 • blá IjLL Jj, I

J 1 {1 Y

co 10 X 20 30 ".CHROMIUM LEACHED BV PREBOND ETCH, PER CENT

Fig. 7 Lead frame pull strength vs. percentage oftotal chromium removed by exposure of gold/30 nmchromium-metallized HMC's to ceric ammoniumnitrate (CAN) for 10 min. at room temperature forpre-bond etching. All samples were submtted to atantalum:. nitride resistor stabilization treatmentprior to being tested. Far leas chromium leaching,and coneommitant higher pull strengths, areobserved in samples etched immediately after thestabilization treatment than in samples stored for 6weeks in the manufacturing environment betweenthe two operations. After (12)

Chemical cleaning procedures also proved effectivein restoring the bondability of gold surfaces con-taminated with Cr20 3 (24,25). Both KI + I 2 andCAN may be used to improve bonding, but theirmechanisms of action are different. On the one hand,KI + I 2 preferentially attacks the gold at nodule boun-daries and restores bondability because the roughenedsurface allows more deformation during bonding.However, if the contamination layer is continuousand thick, KI + 1 2 may not be able to remove suffi-cient gold and restore bondability (25). On the otherhand, CAN actually removes the Cr 20 3 and evenremoves low concentrations of hydrocarbons that mayalso be present at the gold surface. Therefore withCAN, bondability is restored by re-establishing gold-gold contact.

When CAN is used for pre-bond chemical cleaning,there is an obvious danger of undercutting or attack-ing the chromium adhesion layer through pinholes(Figure 6). Data by Holloway and Long (24,25) andPalmer and Wright (12) show that extensive reactiondoes not normally occur during pre-bond cleaning.However in some instances, CAN exposure mayremove over 10 per cent of the chromium present inthe adhesion layer (as measured by atomic dsorptionanalysis of the etchant) and result in poor lead framepeel strengths (Figure 7). Thus, under certainundefined conditions, gold/chromium films aresusceptible to excessive attack when CAN is used toremove Cr20 3 before TC bonding. Data by Mattoxand Sowell (39) suggest that this attack may be relatedto the porosity of the gold film which is controlled inpart by the alumina porosity.

Diffusion in the gold-chromium system has been aproblem in integrated circuits also. Magni et al. (13)have shown that gold diffuses rapidly throughchromium and alloys with underlying silicon. Thiscauses anomalous behaviour and excessive noise inelectronic devices.

CorrosionA film adhesion problem has sometimes been

observed after storage of gold/chromium metallizeddevices in the manufacturing environment. This pro-blem was characterized by very low strength failure ofTC bonded leads and by `peel-back' of gold films.AES data showed that the chromium layer was nolonger present and traces of cerium and halogens werefound on the delaminated gold and on the tantalumnitride surface. As pointed out earlier, halogensdramatically accelerate the corrosion of gold/chromium metallizations. Hampy et al. (11) haveshown that iodine vapours cause severe degradationeven in times as short as 1 hour at room temperature.Iodine vapours and moisture are always present in thephotolithographic environment that is an integral

104

part of electronic device processing. Data in Figure 7show that the strength of TC bonds decreased whenthese were performed on devices that had been storedbetween resistor stabilization and pre-bond etching.The facts that only samples stored for appreciabletimes exhibited failure and that failure rates peaked inspring and summer when humidity was high point toa significant part played by environmental conditions.Not unexpectedly, the chromium-depleted labyrinthobserved by Hampy et al. (37) increases the danger ofcorrosion. The small interconnecting network ofvoids allows liquid formation at lower relativehumidities and reduces the areas over which corro-sion must occur in order to cause loss of adhesion.Thus, chromium diffusion affects both bondabilityand resistance to corrosion.

These data show that controlled storage conditionsare necessary for gold/chromium metallization. In-déëd, no corrosion has been observed for HMC'sstored in halogen-free environments or in hermetical-ly sealed packages.

Other Intra- and InterconnectionsTC bonding of gold to gold is affected by con-

tamination irrespective of the metallization system.With regard to such problems, gold/chromiummetallizations do not exhibit unusual behaviour whencompared with other systems. For instánce, besidesthat by Cr 20 3, bond strength degradation byphotoresist residues only 1.0 nm thick has also beendemonstrated (40). Holloway and Bushmire haveshown that ozone and ultra-violet radiation plusozone remove photoresist residues from contaminatedsurfaces and restore their bondability. These authorshave also studied the susceptibility of different bon-ding techniques to contamination and found wirebonding to be most sensitive, while ultrasonic techni-ques were least sensitive (41). Jellison has studied theeffect of heating on the bondability of photoresist-contaminated substrates. His results show that thebond strength increases with heat treatment time afterbonding, apparently as the consequente of plasticdeformation recovery (42).

Yost et al. (43,44) have studied the lead-indiumsystem for Bolder attachment as an alternative to thetraditional lead-tin systems. Lead-indium alloysdissolve gold to a lesser extent and show good thermalfatigue lifetimes. Very few fatigue cracks and goodbond strength were observed after 500 cycles between- 55 and + 125°C (45).

Palmer et al. (46) have studied techniques to avoidthe formation of brittle intermetallic phases when agold metallization is bonded to aluminium-metallizedparts (the so-called 'purple plague' problem (47)).These authors have found that a small disk of Kovar(25 nickel/17 cobalt/0.2 manganese/bal. iron, per

cent) plated with gold on one side and aluminium onthe other can alleviate this problem. The Kovarserves as an effective diffusion barrier preventing theformation of brittle phases and/or Kirkendall voids.

Conclusion

Reliable thin film conductors can be produced onmicroelectronic devices using the gold/chromiumsystem. The processing technology for this system iswell defined and easily accomplished. However, grainboundary diffusion of chromium and halogen-enhanced corrosion may constitute two major limita-tions for this metallization. Problems arising fromboth may be diminished or even avoided by properprocessing and by storage and operation in ap-propriate environments. Thus, the system is com-patible with the manufacture of high-reliabilitymicroelectronics.

AcknowledgementsMuch of the work reported here was performed at Sandia

Laboratories and was supported by the U.S. Department ofEnergy.

References1 J. C. Anderson, Thin Solid Films, 1972, 12, 1-152 D. M. Mattox, Thin Solid Films, 1973, 18, 173-1863 K. E. Haq; K. H. Behrndt and I. Kobin, 1. Vac. Sci. Technol.,

1969, 6, 148-1524 J. D. Speight, D. Gerstenberg and H. Basseches in 'Proc. 9th

Annu. Int. Reliab. Phys. Symp.', Las Vegas, 1971, pp.195-203

5 A. T. English and P. A. Turner, J. Electron. Mater., 1972, 1,1-9

6 L. E. Terry and R. W. Wilson, Proc. IEEE, 1969, 57,1580-1586

7 R. W. Berry, P. M. Hall and M. T. Harris, `Thin FilmTechnology', Van Nostrand-Reinhold, New York, 1968, p.337

8 J. J. Barb, IEEE Trans. Electron Devices, 1969, ED16, 351-3569 J. M. Poate, P. A. Turner and W. J. De Bonte, 7. Appl. Phys.,

1975, 46, 4275-428310 P. M. Hall, J. M. Morabito and J. M. Poate, Thin Solid Films,

1976, 33, 107-13411 R. E. Hampy, J. Villaneuva and F. P. Ganyard, Sandia

Laboratories•Rep. SAND 78-0646, July 1978*12 D. W. Palmer and R. E. Wright, Sandia Laboratories Rep,

SAND 76-0414, Sept. 1976*13 G. Majni, G. Ottavani and M. irudenziati, Thin Solid Films,

1976, 38, 15-1914 M. Tierman, U.S. Patent 3,638,035 (1972); see Chem. Abstr.,

1976, 11893815 R. F. Tylecote, `The Solid Phase Welding of Metals', St. Mar-

tin's Press, New York, 1968; Gold Bull., 1978, 11, (3), 74-8016 C. M. Tapp and T. A. Wiley, Sandia Laboratories Rep.

SLA-73-1063, June 1973*17 R. L. Long, Jr., Sandia Laboratories Rep. SC-DR-72-0887,

Dec. 1972*18 C. M. Tapp and R. K. Traeger, Sandia Laboratories Rep.

SLA-74-0009, June 1974*19 C. M. Tapp and D. J. Sharp, Sandia Laboratories Rep.

SLA-74-0300, June 1974*20 F. A. Clay, N. T. Panousis and R. W. Pierce in 'Proc. 24th

Electron. Components Conf.', Washington, D.C., 1974, pp.98-104; IEEE Trans. Parts Hybrids Packag., 1974, PHP10,258-262

21 C. Weaver and R. M. Hill, Philos. Mag., 1959, 4, 1107-111122 D. F. Weirauch, Ceram. Bull., 1974, 53, 26023 P. H. Holloway and G. C. Nelson, Sandia Laboratories Rep.

SAND 75-0216, May 1975*24 P. H. Holloway and R. L. Long, Jr., Sandia Laboratories Rep.

SLA-73-1049, March 1974*

105

25 P. H. Holloway and R. L. Long, Jr., IEEE Trans. ParisHybrids Packag., 1975, PHP11, 83-88

26 N. T. Panousis and H. B. Bonham in 'Proc. llth Annu.Reliab. Phys. Conf.', Las Vegas, 1973, pp.; 21-25

27 D. R. Johnson, Sandia Laboratories Rep. SC-DR-71-0539Dec. 1971*

28 P. H. Holloway, ,7. Vac. Sci. Technol., 1975, 12, 1418-142229 J. R. Rairden, C. A. Neugebauer and 9. A. Sigsbee, Metall.

Trans., 1971, 2A, 719-2230 G. L. Schnable and R. S. Keen, Tech. Rep. No. RADC-

TR-66-165, 1966*31 R. E. Thomas and G. A. Haas, Y. Appl. Phys., 1972, 43,

4900-490732 J. K. Hirvonen, W. H. Weisenberger, J. E. Westmoreland and

R. A. Meussner, Appl. Phys. Lett., 1972, 21, 37-3933 P. S. Kendrick, Nature, 1968, 217, 1249-125034 A. L. Pranatis, 'Diffusion Studies for Microcircuit Metalliza-

tions', Report of NRL Progress, Naval Research Laboratory,Washington, D.C., May 1971, p. 21

35 G. C. Nelson and P. H. Holloway in 'Surface Analysis Techni-ques for Metallurgical Applications', ASTM STP 596,American Society for Testing and Materials, Philadelphia,PA.,1976, pp. 68-78

36 P. H. Holloway, D. E. Amos and G. C. Nelson,,. Appl. Phys.,1976, 47, 3769-3775

37 R. E. Hampy, F. G. Yost and F. P. Ganyard, J. Vac. Sci.Technol. 1979, 16, 25-30

38 P. H. Holloway and G. C. Nelson, Thin Solid Films, 1976, 35,L13-L16

39 D. M. Mattox and R. R. Sowell in 'Proc. 7th Int. Vac. Congr.& 3rd Int. Conf. SoIid Surf.', Vierma, 1977, pp. 2659-2662

40 P. H. Holloway and D. W. Bushmire in 'Proc. 12th Int. Symp.Reliab. Phys.', Las Vegas, 1974, pp. 180-186

41 D. W. Bushmire and P. H. Holloway, in 'Proc. Int. Microelec-tron. Symp.', Vancouver, B.C., Int. Soc. for HybridMicroelectron., P.O. Box 3255, Montgomery, AL. 36109,U.S.A., 1975, pp. 402-408

42 J. L. Jellison in 'Proc. 25th Electron. Components Conf.',Washington, D.C., 1975, p. 271-277

43 F. G. Yost, Gold Bull., 1977, 10, (4), 94-10044 F. G. Yost, _7. Electron. Mater., 1974, 3, 353-36945 H. C. Olson and G. L. Knauss, Sandia Laboratories Rep.

SLA-73-1084, Oct. 1974*46 D. W. Palmer and F. P. Ganyard, IEEE Trans. Components

Hybrids Manuf. Technol., 1978,. CHMT1, 219-22247 E. Philofsky, Solid State Electron., 1970, 13, 1391-1399

*Available from National Technical Information Service, U.S.Department of Commerce, 5825 Port Royal Road, P.O. Box 1153,Springfield, VA. 22151, U.S.A.

Automatic Selective Brush Plating of GoldAPPLICATION TO TRANSISTOR AND INTEGRATED CIRCUIT HEADERS

The selective gold plating of intricately shaped articlespresents difficulties and for many years transistor headerswere barrel-plated despite the fact that gold was needed onlyon those areas where wire or wafer bonding was required.The restriction of gold plating essentially to these areas ismade possible, however, using equipment which has beendeveloped by the Auric Corporation of Newark, NJ.,(U.S.Patents 3,951,772 and 4,048,043).

In this, the headers are carried on a stainless steel bandwith the surface to be plated at a fixed height above a mov-ing feit belt which is kept impregnated with electrolyté fromthe gold plating bath through which it passes. A topperbrush serves as the cathode and platinum mesh as theanode. A proprietary electrolyte (Auric 609), which is a

neutral gold bath containing citrates and phosphates, or itsequivalent, is recommended for use with the equipment,which can be adapted to plate most types of headers. Theseare fed into the machine after cleaning by normal pro-cedures and drying, and after plating are unloadedautomatically into a container of distilled water. They arethen rinsed and dried.

Depending on the design of the units and the goldthickness desired, the equipment will plate 8 000 to 14 000headers per hour at an energy consumption'which is a frac-tion of that in barrel-plating. The resulting gold coatingshave been found to give higher bonding yields than thoseproduced by barrel-plating, and the equipment is approvedfor use by six major semiconductor manufacturers.

106