Reading time ( words)
There are only two common and fundamental ways of making interconnections to an integrated circuit package, direct attach, exemplified by flip chip technology and wire bonding. There are other technologies of long standing as well, such as tape automated bonding (TAB), where the IC is connected directly to a lead frame in a gang bonded interconnection fashion. There are some interesting and quite non-standard ways of accomplishing interconnection also.
One example of a non-standard chip interconnection is exemplified by those interconnection strategies wherein the interconnection is made by plating directly to the contacts on the IC. This type of interconnection has been visited upon and investigated a number of times over the last 15 to 20 years. The broad brush term being applied to this alternative method for chip interconnection is "embedded active." While the results of these latest efforts will not likely see immediate broad scale use for what benefits they may ultimately hold, traditional technologies such as wire bonding appear ready to hold the line for chip interconnection until the time, when and/or if these new technologies step up. <?xml:namespace prefix = o ns = "urn:schemas-microsoft-com:office:office" />
Wire bonding interconnection to the integrated circuit chip is one of the earliest chip interconnection technologies born in the late 1950s shortly after the invention of the IC itself. It is generally accepted that Bell Labs was the birthplace of wire bond technology, however it was Fred Kulicke and Al Soffa who turned the technology into a thriving business now known as Kulicke & Soffa, Inc. (Willow Grove, PA).
Wire bonding is not a monolithic technology. When various materials and processes that can be used are factored in, the potential number of permutations can be rather large. For example, while it is generally accepted that there are two basic forms of wire bonds, wedge bonds and ball bonds, there are two ways to make ball bonds, thermocompression (which has fallen in use do to the slowness of the process and the high temperature of ~ 300*C that is required) and thermosonic which is widely used.
Moreover there are a number of different wire materials possible and available including gold and aluminum, which are primary, but also copper and silver. Most wires are not pure metals but have small amounts of other elements to improve certain properties, such as strength. Aluminum, for example is often alloyed with small amounts of silicon. There are also a range of finishes to which the connections are made that can add to the matrix but aluminum and gold are the most common.
Among wire bonding technologies, copper ball bonding has enjoyed increased attention in recent years to due to both its economic advantage (with the price of gold now approaching its historic high) and its greater resistance to wire sweep, a condition wherein wires sag and short due to intrinsic stress or as a result of the molding process (leaning of the stress relief loop until it touches an adjacent bond wire).
When it comes to interconnecting ICs, the vast majority of interconnections to the chip are made with ball bonds. Wedge bonding, which is largely carried out with aluminum wire, is still popular for low cost applications. While ball bonds dominate chip connections, most second bonds to the package are made using stitch bonds that greatly resemble wedge bonds in many ways.
Because wire bonds are often made between dissimilar metals (e.g. gold wire and aluminum pad), the bonds made are largely comprised of intermetallic layers, which have properties that can be quite different from either of the metals.
The limitation of spacing between bonds is a critical factor when it comes to chip design. Ball bonding technology uses a ceramic wire capillary that allows for very close bonds. The current limits are now less than 50um and are projected to reach 25um limits not yet being approached using wedge bonding tools.
As mentioned earlier, the actual bonding process is carried out in one of two manners. One is ultrasonic or wedge-wedge bonding, where the bond is formed as a wedge bond by means of pressure and the vibration of ultrasonic energy. This process is carried out at room temperature. The other method is thermosonic or ball-wedge bonding which is performed on a heated stage (~150*C). The bonding is formed by coupling the ball to the pad with modest pressure and ultrasonically vibrating the capillary.
While wire bonding is amazingly reliable with few failures among the literally trillions of wires bonded every year, failures do occur. At one point in time past, there were some hard won lessons relative to the reliability that gave rise to increased understanding of the nature of wire bonding and material choices. "Purple Plague" is a colorful name for what was once a real reliability headache caused by the rapid diffusion of gold and aluminum into one another, resulting in the formation of a characteristically purple gold-aluminum alloy and the presence of Kirkendall voids, which caused bonds to lift.
While Purple Plague is rare today, some of the more common failures encountered now include bond lifts due to weak bonding or contamination, tail shorts and cratering wherein the pad is damaged in the bonding process. With low K materials there is an added concern as these materials tend to be rather brittle increasing the potential for bond pad damage and lifting. Wire shorting was once a concern, however, a new company, Microbonds, Inc., (Markham, Ontario, Canada) has developed a method for bonding insulated wires which not only solves the shorting problem but opens up a whole new world of potentially novel interconnection solution opportunities.
In summary, wire bonding remains the work horse of the IC assembly industry even as flip chip continues to gather broader appeal. Given the versatility and history of reliability of wire bonding and some of the recent innovations, it is likely that wire bonding will be around for a long time to come.
1) Harman, G., Wire Bonding in Microelectronics Materials, Processes, Reliability and Yield, 2nd Edition, McGraw Hill, 1997