Not only more conductive and considerably cheaper than gold, copper uses the same ball bonding process.

The most common IC conductor metals today are gold, aluminum, silver alloy and copper. Gold is the most widely used metal for IC wire bonding because of its resistance to surface corrosion and high productivity through the gold ball bonding process. However, the price of gold has risen more than 50% since June 2005, spurring interest in a replacement that could reduce cost without diminishing conductivity, chip functionality and reliability.

Heavy copper wire, >0.002" diameter, has been used on power devices since 2001.¹ Aside from the lower cost, copper wire lowers electrical resistance, and increases the maximum allowable current and thermal stability at high temperatures.²

For 0.001" copper wire development, the main driving factor is the replacement of 0.001" gold wire for cost-reduction purposes. Table 1 compares copper wire properties, setup requirements and reliability performance to those of gold wire.


One of the most important properties is electrical conductivity. Based on the electrical resistivity measured, 0.001" copper wire is ~26% more conductive than gold wire. Because of this, copper wire also offers the benefit of reducing bond pad size/die size by replacing current gold wire diameter with a smaller copper wire diameter at an equivalent performance.

A well-known property of copper is its hardness. The copper free air ball (FAB) is already ~30% harder than a gold FAB. Finite element modeling studies have shown both copper and gold wire harden even more during ball bond deformation.4 The change in hardness, ~20Hv, of gold during ball bonding is not significant enough to induce bond pad damage. However, the change in hardness, ~48Hv, of copper is enough to induce bond pad metal lift, or worse, silicon damage. Recent developments in copper bonding, like the availability of soft copper wires and the understanding of the effects of FAB parameters on ball hardness, have made it possible to overcome this.⁵

Tensile strength and elongation are important properties to check, as the bonding wire will undergo a number of tensile stresses during bonding, molding, board mounting and actual use. The tensile strength of 0.001" copper wire and gold wire is comparable. Copper wire has a higher percent of elongation than does gold wire, which means it can withstand plastic deformation longer without rupture.

Copper wire bonding can use the same infrastructure as gold wire bonding, provided the equipment can be retrofitted with forming gas supply, and the electronic flame-off (EFO) current supply is adequate (Table 2). Standard bonding tools or capillary can be used for copper wire.


The major addition is the forming gas supply at the electronic EFO firing area to prevent copper wire oxidation to enable FAB formation (Figures 1 and 2). The EFO wand and the capillary tip for copper bonding are within an enclosure to contain the forming gas supply within the area where the EFO flame-off is emitted and where the FAB is formed.


Two methods are used in the ball bonding process:

1) Thermosonic bonding is, at present, the most popular approach used in gold ball bonding. This involves the rubbing of the bond wire onto the heated bond pad with a high energy that removes any surface impurities and brings together atoms of both metals close enough to form a good bond. A tool that vibrates at ultrasonic frequency provides this energy.

2) Thermocompression bonding is used on sensitive devices, taped-lead frames and copper bonding. The FAB formed is simply pressed onto the bond pad to make a joint. Heat and pressure are used to make a good weld. Since copper wire is harder than gold wire, this bonding method is effective in preventing aluminum pad splashing, silicon damage or cratering.

Current wire bond equipment is designed for thermosonic bonding. However, it is just as easy to switch it to thermocompression bonding by simply setting the ultrasonic generator to zero, which turns it off.

Another difference in the wire bond setup of gold and copper bonding is the EFO current setting. The EFO current for copper wire is four times higher than the EFO current for gold wire. Evaluations on the effects of EFO current on copper FAB hardness show that higher EFO current results in a softer copper FAB, which in turn facilitates an easier ball bond formation.⁵ A higher EFO current equates to a higher temperature, which contributes to a softer FAB.

The much-promoted advantage of copper wire bond is its better reliability performance in terms of a high-temperature storage test. This is attributed to a significantly slower reaction rate of CuAl IMC formation as compared to Au/Al IMC formations (Figure 3, Table 3).⁶


Gas mixture/flow rate. Determining the suitable forming gas mixture and the corresponding flow rate is a prerequisite for the copper bonding process. The forming gas mixture in copper bonding is a combination of nitrogen and hydrogen gases. This provides the inert environment needed for FAB formation.

Establishing the gas flow rate is also critical. Too little may not be enough to provide a sufficient inert surrounding during FAB formation, which may lead to deformed FAB or oxidized FAB. Too much will result in a high gas consumption rate and waste.

Good bonding begins with the FAB formation. Inappropriate gas mixtures and flow rates could result in quality concerns such as lifted ball and lifted metal or productivity concerns such as EFO-open. An evaluation was conducted to select the appropriate gas mixture and the corresponding flow rate (Table 4).


The output responses checked were FAB formation (shape and condition), wire pull and ball shear.

A round and shiny (clean) FAB was achieved on all gas mixtures at a high flow rate setting, 944 sccm. FAB was also shiny (clean) on all gas mixtures at a low flow rate setting, 472 sccm, but the FAB did not seem to be fully formed or fully spherical in shape, which may cause lifted balls (Figure 4).


The 80N2/20H2 mixture resulted in ball shear readings with a wide standard deviation. Based on this and the gas mixture being flammable, the 80N2/20H2 was not selected.

Both the 90N2/10H2 and 95N2/5H2 gas mixtures had comparable results in terms of average ball shear and average wire pull (Figure 5). In this case, the 95N2/5H2 mixture was chosen as the optimum gas mixture for the copper bonding process, since it is safer (has lesser hydrogen content) and is a standard product/mixture of the gas supplier.


Consistent gas supply. After defining a sufficient gas flow rate setting, the next challenge is to ensure the gas supply is consistent and not fluctuating outside the recommended flow rate range. As mentioned earlier, insufficient gas flow may lead to potential bond failures such as lifted balls, bond pad metal lift and silicon damage/cratering. In a mass production setting, the approach must be a closed-loop process.

Thus, a digital gas flow controller was installed in the wire bond equipment to:

1. Automatically stop the bonding sequence and activate an alarm when the forming gas supply is below minimum setting to prevent bondability issues.

2. Automatically shut off forming gas supply during machine idle time.

Wire bonding methods. After creating a FAB that is ready for bonding, the next challenge is selection of the appropriate wire bond method: thermosonic bonding or thermocompression bonding.

Thermosonic bonding is currently used for gold ball bonding. However, this method is less effective on copper ball bonding versus gold ball bonding. This type of bonding requires mechanical rubbing of the already hard copper FAB onto the aluminum bond pad that has minimal resistance to such an abrasion. A considerable amount of aluminum splashing is evident at the ball bond periphery. This scrubbing motion can induce pad metal lift and/or silicon damage.

Thermocompression bonding, on the other hand, uses pressure/force to flatten the copper FAB onto the aluminum pad (Figure 6). This minimizes aluminum splashing and vibration on the pad, to prevent pad metal lift or silicon damage. However, this requires longer bonding – as much as 10% – compared to thermosonic gold bonding.


Both thermosonic and thermocompression bonding use heat to facilitate ball-to-pad adhesion. In this evaluation, a 200°Celsius bond temperature was used.

Eliminating pad metal lift during wire pull test. Pad metal lift break mode was encountered during the wire pull test. Although the affected unit had no silicon damage or crater after etch test and the frequency of occurrences was only 0.1%, pad metal lift break mode is a latent failure and an indicator of a non-robust process, so it must be eliminated.

An evaluation was conducted to determine the relationship between aluminum pad thickness and pad metal lift break mode. Four customer devices (three IC devices and one power device) having different aluminum pad thicknesses were evaluated. First and second bond parameter optimization was conducted on a test chip, with 1.3 µm aluminum pad thickness. The derived optimum parameter was verified and fine-tuned on each of the customer devices with aluminum pad thicknesses of 0.8 µm (device A), 1.0 µm (device B), 1.3 µm (device C) and 4.0 µm (device D). The main output response checked was wire pull break mode (pad metal lift) and crater test.

It was found that the 0.8 µm pad device had a 0.1% occurrence of pad metal lift. No pad metal lift was found on devices B, C and D with aluminum pad thickness equal to or greater than 1.0 µm. All devices passed the etch test with no cratering found, including the 0.8 µm pad device. This shows the minimum allowable aluminum pad thickness must be 1.0 µm to withstand the impact of the copper bond to prevent pad metal lift break mode during wire pull.

What about devices with less than 1.0 µm aluminum pad thickness? Could these devices still be bonded with copper wire? Soft copper wire was evaluated to check if a reduction in copper wire FAB hardness can eliminate pad metal lift at wire pull for devices with thin aluminum pads, less than 1.0 µm (Figure 7).


Two types of soft copper wire were used in the evaluation. Wire type A is 99.99% (4N) pure copper (Table 4). To achieve a softer copper wire, the supplier used a certain combination of impurities or dopants and annealing process to manipulate wire characteristics. Wire type B is 99.999% (5N) pure copper. The supplier used raw copper material with a higher purity (5N) than a standard raw copper material (4N).

No pad metal lift occurred using soft copper wire A (Figure 8). This particular wire also showed a strong wire neck and good first bond and second bond adhesion, since 85% of the break mode was along the wire span, 11% heel break and 4% neck break. For standardization purposes and ease in material control and handling, soft copper wire A is the recommended wire type for copper bonding technology for all aluminum pad thicknesses.


Copper decapsulation. Decapsulation is the removal of the epoxy molding compound covering the IC package for close examination of the die condition, bond condition, epoxy, die pad and leads.

There are two types of decapsulation methods: wet and dry procedures. The wet procedure uses liquid chemicals to etch the encapsulant, while the dry procedure uses either a laser or a plasma etching method.⁷ The former is more common because it is faster and cheaper.

Decapsulating with copper wire bonds is more challenging than decapsulating with aluminum wire bonds because copper reacts with acid easily and corrodes together with the molding compound, very much unlike gold, which is resistant to corrosion. Defining the copper decapsulation technique is important not only for failure analysis but also as a critical tool in process development and DfM.

An acid decapsulation study for copper wire was used as reference for the decapsulation evaluation matrix for copper on a QFN package (Table 5).⁷ Decapsulation techniques may vary depending on mold compound type and thickness.


Concentrated sulfuric acid and hot fuming nitric acid were the decapsulation chemicals used. Input factors considered were the acid ratio, hot plate temperature, acid dripping and soak time. Visual condition of the wires, unit and percentage change in wire diameter were checked as output responses (Figure 9). Table 6 shows the optimum parameter derived for copper bond decapsulation.


Cost reduction assessment

The cost reduction derived from replacing 0.001" gold wire with 0.00l" copper wire was below the expected value. Certain logistics come into play. Forming gas supply, copper wire floor life control and wire bonding speed were just a few of the significant factors that added to the copper bonding process and assembly unit cost.

Copper wire is recommended for high-density (more than 200 wires) devices and for power devices more than 0.0015" wire diameter. These conditions are the sweet spot for copper wire bonding, where a minimum of 30% cost savings can be achieved.

For devices with fewer than 200 wires, the expected cost savings is at a 10% minimum.

Conclusions

Copper wire is a suitable replacement for gold wire based on comparisons made between properties and manufacturability requirements. Although copper wire has proven high-temperature storage performance, a comprehensive study must be pursued to establish robustness and define product life of 0.001" copper wire size and below under high-moisture environments/applications.

Solutions are available for the major challenges of copper wire in HVM:

1. Recommended forming gas mixture and flow rate: 95N2/5H2 at 944 sccm.
2. Closed-loop approach to gas supply control: installation of digital gas flow controller.
3. Recommended bonding method: thermocompression bonding.
4. Recommended Al pad thickness: >1.0 µm.
5. Pad metal lift solution for <1.0 µm aluminum pad: soft copper wire.
6. Recommended decapsulation technique: Please refer to Table 6.
7. Recommended application for maximum cost savings: high-density devices with >200 wires and on power devices using heavy wire diameters >0.0015".

Areas for further study are copper wire floor life/control and best storage condition.

Acknowledgments
The authors would like to thank Mike Goh from KNS for his support during the evaluations and all the copper wire suppliers who participated. The authors would also like to thank AIT’s QA-FA team, Denny Muharyadi and Tanti Rahayu, for defining the optimum decapsulation technique for copper wire.

Ed.: This article was originally published at SMTA International in October 2007 and is reprinted here with permission.

References

1. M. Deley and L. Levine, “The Emergence of High Volume Copper Ball Bonding,” KnS Technical Library, 2005.
2. M. Hundt, “Current Trends in Semiconductor Packaging,” SOURCE? June 2003.
3. Kulike and Soffa, Copper and Gold Wire Technical Data Sheets, 2007.
4. F. Wulff, et al, “Further Characterization of Intermetallic Growth in Copper and Gold Ball Bonds on Aluminum Metallization,” Semicon Singapore 2005.
5. M.H. Hong, et al, “Investigation Factors Affecting Bonded Ball Hardness on Copper Bonding,” SOURCE?
6. H.J. Kim, et al, “Effects of Cu/Al Intermetallic Compound (IMC) on Copper Wire and Aluminum Pad Bondability,” IEEE Transactions on Components and Packaging Technologies, vol. 26, no. 2, June 2003.
7. S. Murali and N. Srikanth, “Acid Decapsulation of Epoxy Molded IC Packages with Copper Wire Bonds,” IEEE transactions on Electronics and Packaging Manufacturing, vol. 29, no. 3, July 2006.