The increased price of gold is making gold ball bonding, a mainstay process in package assembly, cost prohibitive. Copper provides a viable alternative, but also poses challenges due to physical properties. With proper equipment adaptation, copper is on its way to replacing gold as the material-of-choice for fine- and ultra-fine pitch packaging.
BY CHRISTOPHER BREACH, Ph.D, Oerlikon Esec
Thermosonic ball bonding is a major interconnect process in microelectronics packaging. However, the primary wire material used in fine-pitch (FP) and ultra-fine pitch (UFP) ball bonding is gold, and due to drastic price increases, gold ballbonding has become a costly process that has a considerable economic effect on package assembly. An alternative wire material to gold is copper, which is much cheaper ($20/kg vs. $28,000/Kg) and has several technical benefits including better electrical conductivity. It has been widely used in discrete and power devices with wire diameters typically larger than 30μm for many years.1,2 With only minor differences in the drawing process used to manufacture the wires, there are potentially huge cost savings.
Figure 1. Calculated graphs of (a) electrical conductivity (b) fusing current for Cu and Au wires.
However copper wire behaves quite differently than gold due to its different physical properties, some of which are beneficial and others detrimental to bonding performance.
Electrical Properties of Cu and Au
Figure 1a shows that the electrical conductivity of Cu is significantly better than Au. Another property of interest is the fusing current, defined as the current required to melt the wire at its center.3 The graph in Figure 1b shows how fusing current varies with wire diameter and wire length in air and is calculated using the procedure in reference 3.3 Copper wire has a significantly larger fusing current than gold over a range of wire lengths, although the difference becomes much less significant in long wires. However, when wires are surrounded by molding compound, fusing current generally increases significantly due to the high thermal conductivity of the compound (relative to air). In general, Cu wire is capable of carrying higher currents than the same diameter Au wire. Of course, the specific current carrying properties of Cu wire will depend on the nature of the wire i.e. the purity. Increasing amounts of impurities and alloy/micro-alloy content reduce electrical conductivity.4
Ball–wedge Bonding with Cu Wire
Copper oxidizes easily in air, so Cu wires must be stored in packaging that reduces exposure to oxygen from the environment. However, once removed from the package and placed on a wire bonder, spools of Cu wire become exposed to air and can therefore oxidize. The useful life of Cu wire, when placed on a bonding machine, is generally 3-4 days, although studies show that it can bond even after longer exposure times. However, oxidation during ball formation is a more serious issue that is prevented by forming the free air ball (FAB) and bonding in an inert atmosphere, which requires some minor modifications to the ball bonder in the form of a copper kit. The kit allows the use of inert gases to shield the Cu from exposure to air. Gases used include N2 but ‘forming’ gas, a mixture of 5%H2 in N2 is more commonly used because H2 dissociates into atomic hydrogen, which acts as an oxygen getter, effectively removing any trace amounts of oxygen. To surround the wire with forming gas during FAB formation, the electronic flame-off (EFO) electrode has a shroud that the wire enters (Figure 2a) where it is surrounded by forming gas prior to forming the FAB (Figure 2b).
Figure 2. Pictures of the shroud around the EFO electrode (a) prior to formation of the FAB (b) FAB formation.
There are still some problems with the Cu bonding process. The hardness of Cu wire generally requires high ultrasonic power to soften the ball more, but also softens the bond-pad metallization leading to ‘squeeze out’ and inconsistent metallization thickness under bonded balls. This leads to reliability issues. The overall hardness of copper balls also leads to high risks of cratering. These challenges can be partially overcome by design and manufacturing of Cu wire to make it softer (although it cannot be as soft as gold) and by manipulation of the process parameters.
There can also be issues with wire tearing after wedge bonding, which is relatively easy at large wire diameters (>30μm) but can be inconsistent at fine diameters (=25μm), often leaving an inconsistent tail. This behavior may be related to the larger grain size of copper wires, which is typically on the order of a few microns compared with gold wires that have thin elongated grains that can be a micron or less in thickness.2 Finer copper wires may have fewer grains, and the wire’s tearing strength can be more dependent on grain orientation compared with larger diameter wire, leading to variations in wire deformation, tearing strength, and tail length. Wedge bonding dimensional consistency can also be affected, because wedge-bond formation depends on the deformation of a fewer number of larger grains of varying orientation that affects the tearing strength and sometimes can lead to inconsistent wedge bonds.5 The origin of small balls can also have the same root cause: variations in tail length can result from different tearing behavior of the wire resulting in small balls during the next cycle of FAB formation.5
Figure 3. Graph of intermetallic thickness versus square root of time during aging at 175°C calculated from data.7
There is a trend towards elimination plasma cleaning in the copper bonding process to save on assembly costs, particularly with cheap, low-pin-count packages assembled in low-cost regions. However, such packages also tend to use cheap, low-grade epoxies and plastic materials that have lower purity and higher outgassing. Without plasma cleaning, Cu wire bonding becomes more difficult because of surface contamination that must be removed during the bonding process.
Intermetallic Growth and Reliability
Whether bonding Cu or Au wire on Al alloy metallization, the bond formed is due to a new phase, Cu–Al or Au–Al respectively, generally known as intermetallic coverage.1 Copper forms an extremely thin layer of Cu–Al intermetallic phase, much thinner than the Au–Al phase formed during gold ball bonding.2 During later processing steps such as encapsulation and surface mounting, packages are exposed to elevated temperatures in the range 175°C (e.g. curing of encapsulants) to 260°C (peak temperature during reflow sustained for several minutes). These additional thermal processes allow the initial intermetallic phases to grow. Cu–Al intermetallic phases grow extremely slowly in comparison to Au–Al phases, offering the potential of higher reliability.1,2 Intermetallic growth between Cu and Al bondpads is slow compared with Au and Al (Figure 3). In addition, Figure 4 shows that the resistivity of Cu–Al intermetallics is much less than Au–Al intermetallics.
Figure 4. Bar graphs showing electrical resistivity of (a) Au-Al and (b) Cu-Al intermetallics. Note the much lower electrical resistivity of Cu-Al compounds.8
Copper ball bonding is becoming more widespread primarily because of cost savings and potentially higher reliability. There is a trend towards conversion of existing low-pin-count packages from gold to copper wire, but there are still process issues that need to be overcome before the process becomes mainstream for Advanced Packaging. However, it is expected that lowering of costs will eventually force the adoption of Cu ball bonding.
- F. Wulff, C. D. Breach, Saraswati and D. Stephan. “Characterisation and comparison of intermetallic growth in copper and gold ballbonds an aluminium metallization”, EPTC 2004, Dec 8-10, Singapore.
- F. Wulff, C. D. Breach, Saraswati, K. Dittmer, M. Garnier. “Further characterisation and comparison of intermetallic growth in copper and gold ballbonds an aluminium metallization”, SEMICON Tech. Symp. S2 Proc., May 6th, 2004 Singapore.
- B. Krabbenborg. “High current bond design rules based on bond pad degradation and fusing of the wire.” Microelectronics Reliability 39 (1999) 77.
- P.L. Rossiter. “The electrical resisitivity of metals and alloys”, Cambridge Solid State Science Series, Cambridge University Press (1987).
- J. Beleran, A. Turiano, D. R.M. Calpito, D. Stephan, F. Wulff, C. D. Breach. “Tail pull strength of Cu wires on gold and silver-plated bonding leads”, SEMICON Tech. Symp. S2 Proc., Singapore (2005).
- J.F.M.J. Caers, A. Bischoff, K. Falk and J. Roggen. “Conditions for reliable ball-wedge copper bonding.” Japan Int’l Electronics Mfg. Tech. Symp. (1993).
- H.J. Kim, J. Y. Lee, K. W. Paik, , K. W. Koh, J. Won, S. Choe, et.al.; “Effects of Cu/Al intermetallic compound (IMC) on copper wire and aluminum pad bondability”, IEEE Trans. 26(2) (2003) 267.
- E. Zschech, Ch. 6 in Baugruppen-Technologie Der Elektronik. Ed W. Scheel. Verlag Technik Berlin 1997.
CHRISTOPHER BREACH, Ph.D., V.P., wire bond strategic business unit, may be contacted at Oerlikon Assembly Equipment Pte. Ltd., Singapore Science Park II, #03-10 Capricorn Building, Singapore 117528; 65/6303 7000; email@example.com