 |
| Steve DeCollibus, Manangin Editor, Semiconductor Packaging News
|
Semi recently announced the results of a study conducted by the World Gold Council. The study was conducted in order to to determine the extent of copper bonding wire programs in the industry and to identify the key issues and considerations related to decisions in selecting bonding wire material.
Companies surveyed included both integrated device manufacturers (IDMs) and fabless semiconductor companies. To get a better understanding of what the motivating factors for the study were I contacted Dr Richard Holliday, Director, Industrial at World Gold Council to get his input on the following questions.
Circuitnet: Dr. Holliday what was the impetus for your organization in carrying out this study?
 |
| Dr Richard Holliday, Director, Industrial at World Gold Council
|
Dr. Holliday: With a reported shift towards using copper in place of gold bonding wire in the semiconductor industry, we felt it important to clarify the extent to which this shift is actually taking place and find out what the industry actually thinks of copper as a replacement for gold.
In our opinion a balanced view of the merits of each metal are not being discussed and debated. According to some industry players, copper bonding wire can meet all the needs of the industry, yet the results of the survey suggest many people are not convinced that copper is the complete solution.
There are serious concerns with the reliability and yield of copper wire. We believe this survey should start the debate that has been notably absent up until now.
Circuitnet: It is understandable that the semiconductor industry is looking for more cost effective ways to manage costs and that the economic advantages of copper make it an attractive alternative to gold bonding. Is your organization conducting any research into compound alloys that could create a more reliable alternative?
Dr. Holliday: Firstly, in evaluating the costs associated with bonding wire selection it is recommended that the
total cost (rather than just the basic wire cost) of a copper solution versus a gold solution is considered. There is much more to the story than just the cost per meter of gold versus copper bonding wire.
Copper wire bonding can be slower due to longer bond formation resulting in lower units per hour production rates and the costs for infrastructure and consumable gases used to protect the free-air ball in copper bonding need to be accounted for during the costing process.
The costs associated with higher bonding temperatures and bonding forces also need to be considered. Reliability and package testing are more difficult with copper; using normal decapping methods based on nitric or sulphuric acid to remove the thermoset plastic leaving the gold wire intact can’t be used with copper.
Alternative laser decapping methods can be costly and time consuming. Copper’s inferior corrosion resistance means copper wire has a limited ‘floor life’ compared to gold. This may result in increased scrapping of material and associated costs. There is also the significant added cost of potential in-service chip failure.
As George G. Harman, widely considered to be the world's foremost authority on wire bonding, said in the recent 3rd Edition of his book Wire Bonding in Microelectronics,
"There is no question copper is a cheaper material but also one that brings new challenges to the bonding engineer. It may be a cost effective process in the eyes of a product manager but when the extra care and attention is required , it is weighed against the existing gold process and then we should ask ourselves, are the net savings worthwhile?"
Secondly, with the rising gold price it is important that wire suppliers assist their customers to reduce the impact of the metal price on the overall cost of gold bonding wire technology. Solutions include reducing the overall length of gold wire through optimization of loop height, reduction in wire diameter and design changes to minimize wire usage.
Reducing the amount of gold consumed, whilst maintaining the inherent reliability of a gold bonding wire solution, is a viable strategy for cost reduction. Solutions include:
- Reducing the overall length of gold wire through optimization of loop height
- Optimizing bonded ball volume
- Reduction in gold wire diameter
- Design changes to minimize wire usage (for example, using thinner wires for signal wires and thicker wires for power wires.) Through innovative thinking you can reduce costs and still maintain the inherent reliability and ease of use of gold. Dai Nippon, for example, have just described a metal sheet technology for QFPs that can reduce the volume of gold wire required.
It is important that wire suppliers assist their customers to reduce the impact of the metal price on the overall cost of gold bonding wire technology.
At present we are not looking at compound alloys, but are focusing our research efforts on understanding the failure mechanisms of gold and copper under pressure cooker testing (PCT) and temperature cycling (TCL).
If you look at the published literature there is actually very little documented on this, although one recent study with copper wires ≤25µm and bond pads ≤1µm did reveal problems with reliability in temperature cycling (TCL)1 We plan to publish our work in peer reviewed publications in due course.
1. T. Uno, S. Terashima, T. Yamada. IEEE Proc. ECTC (2009) 1486.
Circuitnet: What percent of the total gold market is focused on semiconductor manufacturing?
Dr. Holliday: In 2008 nearly 300 tonnes of gold was estimated to be used in the electronics industry including packaging, platings, thick film pastes etc. This represents 8% of total demand. Of this figure approximately 140 tonnes a year of gold bonding wire is used by the semiconductor industry.
Circuitnet: Can you compare the benefits of gold over copper in wire bonding?
Dr. Holliday: The clear benefit for the semiconductor industry is that gold has an extensive track record of use for many decades and each year millions of electronic devices trust in gold bonding wire to deliver reliable performance. Therefore there is substantial industry know-how of the material
Copper, on the other hand, has to-date limited proven performance in-service and there is substantially less industry knowledge of copper wire bonding. Copper reliability is still a significant issue for many applications; 2nd bond strength is often low in volume production and whilst 1st bond is often acceptable as bonded, it might be prone to failure in heat and humidity tests in some circumstances.
Stable wire bonding production is achievable with gold wire. Gold is softer and has a lower flow stress than copper so is less prone to bond pad cratering than copper and a robust 2nd bond is achieved. As copper wire is harder than gold wire, bond pad modification is required for some applications.
Gold wire bonding has larger process windows than copper – a high number of units per hour can be achieved with gold which allows higher productivity than copper and cost effective assembly operations. Copper wire bonding is slower due to longer bond formation causing lower production rates (units per hour).
Because of these concerns there are industry segments that will not even consider switching to copper. Automotive ICs is one obvious area where companies are very cautious in considering copper as automotive applications have very stringent reliability requirements.
DRAM memory is another area where companies will be reluctant to make the transition. DRAM is driven by high-volume production and high process yield, and gold wire processing is a known quantity. The same for stacked-die packages.
Circuitnet: Is there anything you would like to add?
Dr. Holliday: World Gold Council, is helping to train engineers and designers in the technology of wire bonding. At SEMICON China 2010 we are sponsoring a seminar entitled 'Materials Science of Wire Bonding & Wire Bond Reliability' in conjunction with SEMI. This will include best practice in wire bonding technology and materials selection issues.
