Ultrasonic ball/wedge bonding of fine copper wire

TWI Bulletin, March 1986

S T Riches

S T Riches, MA, is a Research Engineer in the Sheet and Precision Processes Department.

Copper wire is one of the most promising alternatives to the Au and Al-1%Si wires currently used for electronic device interconnection. This article presents the results of welding trials to investigate the feasibility of ultrasonic ball/wedge bonding 25µm diameter Cu wires to Al thin film metallised Si wafers, and wedge bonding to Au and Cu thick film on alumina. Also included are the results of environmental tests involving exposure to dry heat.

Development of VLSI (very large scale integration) circuits for the next generation of microelectronic devices has led to a dramatic increase in the number of functions carried out on a Si chip. There is consequently an increase in the number of interconnections to the chip and a reduction in the area available for bonding pads. This has resulted in a drive to reduce the size of the bonding pads (down to 50µm square) and consequently smaller diameter wires (less than 25µm diameter) have to be used for device interconnection. Although Au and Al-1%Si wires of 25-33µm diameter have been used successfully for several years, there is concern that, when smaller diameters are necessary, the strength of the wires and the bonds may become a critical factor during processing and service. Therefore, alternative materials are being sought which have properties capable of withstanding the rigours of automatic wire bonding (at speeds of up to 16 bonds/sec) and retaining enough strength for service.

The most promising of the materials for this purpose is Cu as it has greater tensile strength and ductility than Au and Al-1%Si. It also offers a higher thermal and electrical conductivity than Au and Al-1% Si, which allows use of smaller diameter wires for similar current carrying capabilities and improved thermal dissipation from the device at a lower cost than Au wire ( Table 1 ).

Table 1 Comparative properties of materials used for interconnection of microelectronic devices

Source: Metallic materials specification handbook, Ross, 1980, and various wire manufacturers (1982)

The long term performance of an integrated circuit is reliant on the stability of the materials to the package environments. For Cu wire an equal resistance to corrosion for a Cu ball bond/Al thin film joint in both dry heat (for hermetically sealed packages) and damp heat (for plastic encapsulated packages) cannot be attained when compared with the mono-metallic A1 ball or wedge bond/Al thin film joint. However, the rate of Cu-Al intermetallic formation in dry heat conditions should be less than an Au ball bond/Al thin film joint which seriously degrades device characteristics above 150°C. Also, on exposure to humidity, a Cu ball bond/Al thin film joint should exhibit less severe galvanic corrosion of the Al thin film than an Au-Al joint because Cu is less noble than Au.

Despite the advantages detailed above, Cu has not been readily adopted in the semiconductor industry as it is more difficult to weld ultrasonically to thin and thick film circuitry than Au or Al-1% Si and the long term reliability of Al-Cu joints has recently been questioned in studies on Al wire bonds to Cu substrates. [1] There is also concern about the oxidation of Cu which occurs above 150°C.

The work reported here was conducted to establish the feasibility of, and to optimise procedures for, ultrasonic ball/wedge bonding of 25µm diameter Cu wire to Al thin film metallised Si wafers and wedge bonding to Au and Cu thick film on Al 2 O 3 . This involved extensive welding trials on manual wire bonding equipment and an investigation into degradation of Cu ball bonds/Al thin film joints and Cu wedge bonds/Au thick film joints on exposure to 200°C for up to 1000hr in dry heat.

Welding trials

Ball bonding

Ball forming trials were conducted on 25µm diameter Cu wire using ball forming equipment developed for Al-1%Si wire which utilises a spark discharge in an inert gas shield (Ar). Balls of 2.2 x the wire diameter (55µm) were produced with the following ball forming conditions: wire projection 150-200µm; spark gap 75-100µm; wire positive; voltage 560V; capacitance 0.2µF; circuit resistance 200 Ω; atmosphere Ar 1 litre/min. The size and shape of these balls were considered sufficiently consistent to enable ball bonding to be carried out.

The shear strength results obtained for 25µm diameter Cu wire ball bonded to Al thin film on Si wafers are summarised in Table 2. The effect of vibration amplitude on ball bond shear strength for trials with substrate temperatures of 20°C (wire tensile strength 173mN) and 150°C (wire tensile strength 241mN) is shown in Fig.1. The bond strengths for both wires were similar over the range of amplitudes studied and both have the same tolerance range (0.4-1.0µm) with the shear strength at the chosen optimum condition being about 600mN. This compares favourably with shear strengths obtained for similar size ball bonds with Au and Al-1%Si which averaged about 400mN. The bonding force at the optimum condition was 600mN compared to 300-400mN for Au and Al-1%Si wires [2] which is within the range of forces covered by standard bonding machines.

Table 2 Summary of ball bonding trials for 25µm diameter Cu wires to Al thin film metallised Si

Fig.1. Variation of shear strength with vibration amplitude for 25µm diameter Cu wire ball bonded to Al thin film metallised Si, bond force 600mN, bond time 30msec:

a) Substrate temperature 20°C, wire tensile strength 173mN;

b) Substrate temperature 150°C, wire tensile strength 241 mN

On 90° pull testing, the wire failed above the ball in all cases on bonds at the optimised conditions, with strengths of about 90% of the parent wire strength ( Table 2) as compared with approximately 60% for Au and Al-1%Si wire. ^[2] This suggests that Cu wire could be less affected by the wire softening that occurs on Au wire during automatic wire bonding especially above the ball bond.

Ball bond sizes at the highest vibration amplitude did not exceed 4 x wire diameter, which corresponds to the usual bond pad size if 25µm diameter wire is used, and the bond size at optimum was about 3.5 x wire diameter ( Fig.2). However, extrusion of the unalloyed Al thin film metallisation from underneath the ball bond was apparent in all trials. Visual examination of the fracture surfaces suggested that the Cu ball bond did not totally penetrate the Al thin film as far as the underlying Si, although further work is necessary to confirm this. The extrusion appeared to be related to the difference in hardness between the Cu ball bond and Al thin film as extrusion was eliminated when Cu was ball bonded to an alloyed Al thin film (Al-1%Si-4%Cu): Si and Cu are the two most frequent alloying additions to Al thin films for integrated circuits and both have a significant hardening effect on the film, the exact amount of hardening depending on the alloy composition and processing route. Limited work was conducted on this aspect and further trials are required to establish optimum conditions for Cu wire ball bonded to alloyed Al thin film metallisations.

Fig.2. Typical ball bond in 25µm diameter Cu wire to Al thin film metallised Si. Vibration amplitude 0.7µm, bond force 600mN, bond time 30msec, substrate temperature 20°C

Wedge bonding

In addition to bonding to thin film metallised pads on the Si chip, the other end of the wire has to be connected to the external circuitry. Capillary wedge bonding trials were therefore conducted on Au and Cu thick films on Al ₂ O ₃ : Au thick film is extensively used in hybrid electronic manufacture and Cu thick film is being developed as an alternative to Au and Ag-Pd films because of its improved conductivity and lower costs.

The capillary wedge bond 30° pull strengths obtained for Cu wire, with a tensile strength of 175mN, bonded to Au thick film on Al ₂ O ₃ at a substrate temperature of 20 and 150°C, in relation the wire) are shown in Fig.3 and Table 3. A heated substrate permitted use of lower bonding forces (600mN for 150°C, 900mN for 20°C) and free standing vibration amplitude (1.1µm for 150°C, 1.8µm for 20°C) for the optimum conditions. Both these conditions achieved approximately 140mN (80% wire strength) in parallel and transverse vibrations, with all the wires failing at the heel of the bonds made at 20°C, although there were occasional film failures for bonds made at 150°C. These did not show any significant strength reduction compared with the heel failures and the results again compared favourably with the 30° pull strengths for 25µm diameter Au and Al-1%Si wires which typically lie between 50-90mN (40-70% wire strength). ^[2]

Fig.3. Variation of 30° pull strength with vibration amplitude when capillary wedge bonding 25µm Cu wire to Au thick film on Al ₂ O ₃ :

a) Substrate temperature 20°C, vibrations parallel to the wire, bond force 900mN, bond time 86msec;

b) Substrate temperature 20°C, vibrations transverse to the wire, bond force 900mN, bond time 86msec

Table 3 Summary of capillary wedge bonding trials for 25µm diameter Cu wire* to Au thick film on Al 2 O 3 and Cu thick film on Al 2 O 3

Table 3 Summary of capillary wedge bonding trials for 25µm diameter Cu wire* to Au thick film on Al 2 O 3 and Cu thick film on Al 2 O 3 - continued

In contrast to the results obtained when ball bonding, the tensile strength of the wire appears to affect weldability. Trials using wire of tensile strength 240mN did not produce satisfactory wedge bonds as the required deformation to form a Cu wedge caused failure at the Au thick film/ceramic interface over a range of vibration amplitudes, forces, times and substrate temperatures. The difference in behaviour between the ball bonding and wedge bonding presumably arises because the ball, after melting during ball formation, is in an annealed condition, whereas the wire at the wedge bond is in a work hardened state. When the wire is too work hardened, as in the case of Cu wire with a tensile strength of 240mN, capillary wedge bonds cannot be consistently achieved. Substrate heating softens the wire and thus allows the lower power welding conditions as shown in the comparative trials at 20 and 150°C.

Wedge bonding trials to Cu thick film on Al 2 0 3 commenced at conditions based on those used for bonding to Au thick film at room temperature (vibration amplitude 1.1 - 1.5µm, force 900mN, time 86msec). However, bonding was found to be inconsistent at room temperature and it was necessary to use substrate heating to achieve satisfactory bonding. The capillary wedge bond strengths obtained at a substrate temperature of l00°C in relation to the vibration amplitude (parallel and transverse to the wire) are detailed in Fig.4 and Table 3. The optimum conditions gave strengths of approximately 110mN (65% wire strength) for both parallel and transverse vibrations to the wire. Most of the joints failed at the heel but there were some instances of joint failure although these did not have a significantly reduced strength compared with the heel failures. When a substrate temperature of 150°C was used, discoloration of the Cu thick film through oxidation was observed.

Fig.4. Variation of 30° pull strength with vibration amplitude when capillary wedge bonding 25µm Cu wire to Cu thick film on Al ₂ O ₃ substrate temperature 100°C, bond force 900mN, bond time 86msec:

a) Vibrations parallel to wire;

b) Vibrations transverse to wire

Environmental tests

Environmental tests were conducted to establish the reliability of circuits by Cu ball/wedge bonding. The tests consisted of exposing Cu wire, ball bonded to blanket Al thin film on Si and wedge bonded to blanket Au thick film, to a temperature of 200°C in air. Strength tests were performed after specimens had been subjected to 200°C for 16, l00 and l000hr, the heat tests being consecutive. Batches of six specimens were used, each with ten Cu ball/wedge bonds, and after heating, five specimens were shear tested and one was retained for metallurgical examination.

Table 4 Strength tests on ball bonds in 25µm diameter Cu wire to Al thin film on Si

* ns not significant to 5% level; 1% significant; 0.1% highly significant

The results of the strength tests on as-bonded and environmentally tested ball and wedge bonds are given in Tables 4 and 5 respectively. The strength tests showed that after l000hr at 200°C there is a reduction in the ball bonded shear strength to 55% of as-bonded value and in the wedge bonded 30°C pull strength to 56% of as-bonded value. The significance levels for the changes between the tests at 0, 16, 100, 1000hr are also given, from which it can be seen that all the changes in strength were highly significant. Scanning electron microscope examination of Cu ball bonds to Al thin film and Cu wedge bonds to Au thick film showed no signs of intermetallic formation. This compares favourably with Au-Al joints which readily form intermetallics at temperatures of 150° C and above. However, oxidation of Cu had occurred after 1000hr at 200°C and there was a considerable oxide layer surrounding the wire, as shown in Fig.5. This oxide is non-adherent and contributes little to the strength of the wire, so that the observed reduction in strength was probably mainly attributable to this. These results have been supported by tensile tests conducted on wires exposed to l000hr at 200°C which show a reduction in tensile strength to 75% of the as-received wire strength ( Table 6).

Table 5 Strength tests on wedge bonds in 25µm diameter Cu wire to Au thick film on Al 2 O 3

* 5% probably significant; 1% significant; 0.1% highly significant

Fig.5. Typical ball bond after shearing in 25µm Cu wire after 1000hr exposure at 200°C, showing oxide layer. Vibration amplitude 0.7µm, bond force 600mN, bond time 30msec, substrate temperature 20°C

Table 6 Tensile tests on 25µm diameter Cu wire after 0, 16, 100, 1000hr at 200°C (sample of 10 wires)

* ns not significant to 5% level; 0.1% highly significant

The formation of the oxide does not account totally for the reduction in bond strengths and differences in behaviour were also observed at the intermediary stages at 16 and 100hr at 200°C. Tables 4 and 5 show that at 16 and 100hr the ball bonds exhibited reduction to 90% and 76% of the as-bonded shear strength respectively, while the wedge bonds showed a reduction to 93% and 90% of the as-bonded 30° pull strength respectively. Wire tensile strengths on similar exposure to 16 and 100hr at 200°C did not exhibit any reduction in strength compared with the as-received wire.

A possible explanation of these differences can be proposed by considering the mechanical testing of the wires and bonds. In 30° pull strength testing of capillary wedge bonds, the wire itself is being tensile tested for heel failures. This strength is unlikely to be significantly affected by the thin adherent oxide layer which was present after 16 and 100hr and the observed reduction might be attributed to softening of the work hardened wire, which has been encountered in previous environmental trials on Au [3] and Al-1%Si [4] wires. After l000hr at 200°C, softening of the wire has continued, which, in combination with the oxide layer, has resulted in a strength reduction to 55% of the as-bonded pull strength. In addition to the above factors, shear strength could be affected by the surface condition ( e.g. an oxide layer) of the ball bond which may account for the faster reduction in strength after 16 and 100hr compared with wedge bonds.

The above results suggest that oxidation of the Cu is a major factor causing reduction in bond strengths of Cu ball bonds on Al thin film on Si, and of Cu wedge bonds on Au thick film on Al 2 O 3 exposed to 200°C for up to l000hr. To confirm this, further trials are necessary in non-oxidising gas ( e.g. N 2 ) which is commonly used in hermetic packages. However, effects of a damp heat environment, for example 85°C/85°%RH) on Cu ball/wedge bonds have not been investigated. This would have to be studied before Cu wire bonding could be accepted in non-hermetic packages ( e.g. plastic) where there is a likelihood of ingress of moisture.

Summary

Consistent ball and wedge bonds have been made with 25µm diameter Cu wire to Al thin film on Si, and Au thick film on Al 2 O 3 respectively, at room temperature. The ball bonds, at optimum conditions, had shear strengths of 600mN and 90° pull strengths of 90% parent material strength. The wedge bonds, at optimum conditions, had 30° pull strengths of 140mN (80% wire strength). Consistent wedge bonds were also made to Cu thick film on Al 2 O 3 at a substrate temperature of 100°C, and had strengths of 110mN (65% wire strength).

Environmental tests in dry heat at 200°C for up to 1000hr were conducted on Cu ball bonds to Al thin film on Si and Cu wedge bonds to Au thick film on Al 2 O 3 . The ball bonds showed a reduction in shear strength to 55% of the as-bonded strength and wedge bonds showed a reduction in 30° pull strength to 56% of the as-bonded strength. No evidence of intermetallic was found and it was thought that the reduction in strength was caused mainly by oxidation of Cu.

Acknowledgements

This work was carried out with the support of the Procurement Executive, Ministry of Defence, sponsored by DCVD. The assistance of the Design Authority, Mr G Binns, RSRE Malvern, is gratefully acknowledged. The author also acknowledges the work and advice of his colleagues R G Clements, M H Scott, N R Stockham and H M McMenamin.

References