Wire Bonding
TASA ID: 4595
The dominant process for interconnecting semiconductor chips to the outside world is an ultrasonic welding process called wire bonding. More than 90% of the chip interconnections produced annually (more than 15 trillion wires) are produced with this process. Welding is a process where an intermetallic alloy is formed from the materials to be joined. Generally, intermetallic alloys are stronger and also more brittle than their constituents. Welding is superior to other joining methods, such as soldering which require that a low melting temperature material melt and solidify within the joint. Low melting temperature materials, such as solders have significantly lower strength and are more subject to creep and fatigue failures than intermetallics. There are two major variations of the wire bonding process, ball bonding and wedge bonding. Ball bonding is the larger portion with about 90% of the entire wire bonding market. It is faster, the fastest ball bonders can bond more than 20 wires/second compared to less than 10 wires/sec for wedge bonding. Ball bonding also has more advanced capabilities than wedge bonding. However, ball bonding is limited to wires below approximately 50µm in diameter. All interconnections that require larger diameter wire are produced by wedge bonding Al or Cu, using either round wire or ribbon (a flattened form of round wire).
During the past five years, there has been a major transition in our industry from ball bonding with gold wire to the use of copper, palladium coated copper or silver wire. This year will be the first year where market share for gold wire falls below 50%. Cost, yield and reliability have all played a major part in this transition. In 2009 when gold rose in price above $1000/ Tr. Oz in 2009 and remained there, gold reduction became a mandate in semiconductor packaging. Gold wire represented a large portion of the gold used in semiconductor packaging. Copper had been discussed1 and demonstrated since the early 1980s but had not been widely adopted. Copper was more difficult to bond and had package reliability issues. As these issues (optimum bond pad metallization, encapsulation chemistry for long-term reliability, bonder recipe improvements) were resolved, the transition became a stampede and in five years became a new paradigm. Silver is also less expensive than gold. Silver is used for bonding LED devices because it has better reflectivity properties than either copper or gold. Early problems with silver wire in 850C/85%RH testing were resolved using Ag-Pd alloy wire. Silver market share is now approaching 10%.
Figure 1 is a photo of the bond head with capillary, wire and EFO (Electronic Flame OFF) wand. In ball bonding the tip of a fine diameter metallic wire (protruding from the capillary) is melted by a spark from the EFO. Surface tension in the metallic liquid pulls the liquid into a sphere, the sphere solidifies with more than 80% of the heat transferring back into the wire. This leaves a short region above the ball, called the HAZ (Heat Affected Zone) that has been rapidly heated to just below the melting temperature and then cooled rapidly to near room temperature. The HAZ is the weakest portion of the wire. The bondhead, with capillary and ball dangling below it, descends at a high speed towards the surface (normally the bond pad on a die). At a programmed height above the surface the bond pad velocity transitions to a slower, constant velocity and the bonder begins searching for the surface (surface height can vary due to the many tolerances from material and prior operations). Surface detection can occur by a number of methods including mechanically opening a contact spring as in older machines or high speed sensing of a current rise in a voice coil motor when the coil stalls on contact. After contact detection, the bond head continues downward to apply a programmable force on the ball. Ultrasonic energy from a piezo-electric transducer is added for a programmable time (8-12 mille-seconds is typical for a high-speed ball bonder). The die and substrate are normally heated to 125-2000C depending on the process and materials. These four factors: ultrasonic energy, bond force, heat and time constitute the principal variables for ultrasonic weld formation.
After completing the ball bond cycle, the bond head rises and a series of very precise coordinated motions occur, forming a loop between the ball bond and the second bond. Loop height and uniformity are very important packaging requirements. The demand for thin packages and stacked die packages that are also, as thin as possible, lead to the development of improved bond head control algorithms and many new loop shape options. Memory devices often have their bond pads located down the center of the die surface rather than around their periphery. This allows better signal and voltage distribution and results in faster devices that command premium values. The left panel of Figure 2 is a photo of low loop wires for memory. These loops rise to a low height and then travel parallel to the edge of the die where they descend to second bond. Stacked die packages (right panel of figure 2) often employ a hybrid bond called a stand-off- stitch (SOS). In a SOS bond, a ball is formed and bonded with the wire intentionally broken in the HAZ. Another ball is formed and bonded to the substrate side of the package. The stitch (second bond) side of the wire is then bonded on top of the original ball. Because it requires the formation of three bonds rather than two, the SOS bond is approximately 40% slower than a standard bond, but it provides the lowest loop height available. Every smart phone (more than 1 billion annually) has at least one stacked die package. Stacked dies, because each die can contain a separate technology (analog, digital, memory, rf), enable integration of the entire system within the package. Earlier attempts to integrate all of these technologies on the same chip proved costly and decreased reliability. Joining the technologies by stacking them within the package became the dominant method.
Second bond is formed by a different portion of the capillary tip than the ball bond. Figure 3 is an illustration of a capillary tip and the portions of the tip that produce the ball bond, the loop and the second bond. In forming the second bond, the capillary face and outer radius are pressed on the top of a round wire. The combination of ultrasonic energy, bond force, heat and time deform the round wire into the fishtail shape and form the initial intermetallic bond.
The mechanical and other materials properties of the ball and the wire are significantly different. Second bond is more diffusion controlled than the ball bond.
Wire Bond Failure Mechanisms
Semiconductor packages must normally pass a battery of short and long-term reliability testing during package qualification prior to market introduction. Once manufacturing and sales begin, mechanical testing is commonly done on each material lot. Mechanical testing normally consists of both wire bond pull testing and shear testing. Because the weld areas for both the ball bond and the second bond are several times larger in cross section than the wire cross section, the pull test is not capable of testing the strength of either bond (the wire breaks first). However, it is capable of detecting very poor bonds, wire damage, damage to the HAZ or a second bond that has been over-deformed and that has a thin cross section at the heel of the bond. The pull test measurement can be understood from a simple resolution of forces. However, once a history of data exists and SPC has been established the use of control charts can be a very powerful quality tool. The shear test is capable of measuring ball bond strength and should be a standard test for each lot. Average shear strength of 5.5g/mil2 (85MPa) meets the JESD-22-B116A standard for shear testing required by the automotive industry.
The life, and subsequent failure of gold ball bonds on aluminum bond pads by Kirkendall voiding has been well-documented. At a temperature above 1500C, for some packages this can occur quickly and catastrophically. Bonds literally fall off with almost no stress. New 99.9% gold alloy wires (standard gold bonding wire is 99.99% gold) with additional impurities added to stabilize intermetallic formation can improve reliability. Gold ball bonds on gold bond pads in high temperature environments do not exhibit the problem.
Analysis of intermetallic coverage and morphology should be a standard part of qualification testing and be repeated periodically through the life of a product. Aluminum bond pads can be easily etched with sodium hydroxide or potassium hydroxide to release the bonded balls. Etching will not remove the intermetallic on the bottom of the balls. The balls can be flipped with a dental pick or the die paddle tie bars can be removed to reveal the bottom side of the balls. Intermetallic coverage should exceed 80% as bonded. Figure 4 demonstrates the evolution of bond coverage as a function of bonding time. After 16ms bond time, the intermetallic coverage is over 80%. To expose the bonds in encapsulated packages, it is often necessary to remove the encapsulating material with hot, fuming nitric acid. This will reveal gold ball bonds but will immediately attack copper bonds. Several techniques, such as laser ablation and very controlled etching in an inert atmosphere have been used for copper ball bonds.
Copper-aluminum intermetallic requires both a higher formation temperature and longer time (slower growth rate) than gold aluminum. Therefore, copper ball bonds can be more reliable than gold bonds at high temperature. Encapsulation to protect copper bonds is critical. The presence of Cl- ions is auto catalytic to copper. Chlorine corrodes copper and then is released to continue corrosion. Molding compounds that contain less than 30ppm chlorine and have a controlled pH of 4-6 are now available for copper and are necessary for high reliability products2.
Figure 5 shows SEM photos of two failure modes that can occur as a result of wire bonding. Normally, ultrasonic energy is the most aggressive variable affecting bond pad failures but poor design of the bond pads is also a root cause. Designed in reliability resulting from careful DOEs and the development of internal design guidelines focused on the use of robust bond pad structures cannot be ignored. Modern bond pads often contain multi-level stacks of metal and dielectric layers. In some cases, low-k dielectrics with poor mechanical stability are required for functionality. Instances of failures in layers below the surface, allowing electro-migration and eventually resulting in interlayer shorts are well-documented. Often these failures can occur while the top metal layers and wire bond are unaffected3. They are difficult to detect and analyze. A team approach, involving FAB, Assembly and Reliability engineers must focus on development of pad structures that not only can achieve electrical design requirements but be robust enough to withstand manufacturing and reliability.
Conclusion
Wire bonding continues to be the lowest cost, highest reliability, most flexible semiconductor interconnection method. It continues to reinvent itself, as new demands are understood machine, wire, tool and end users come together to find solutions that enable successful implementation of the new requirements. Each new generation of devices has required increased capabilities for both manufacturing and metrology. Wire bonding has met these challenges and added the capabilities necessary for its continued growth as the leading semiconductor interconnection method.
[1] M. Sheaffer, L. Levine and B. Schlain, “Optimizing the wire bonding process for copper ball bonding using classic experimental designs,” Proc. IEMT, pp 103-108, Sept. 1986
[2] F. Carson, “Copper wire interconnect has arrived,” Chip Scale Review, Jan/Feb 2011, pg 35-37
[3] S. Hunter, R. Gohnert, R.L. Warnick, A.J. Dutson, “A bond pad’s view of wire bonding,” IMAPS Wirebonding Workshop, San Jose, CA, Jan 2013
This article discusses issues of general interest and does not give any specific legal or business advice pertaining to any specific circumstances. Before acting upon any of its information, you should obtain appropriate advice from a lawyer or other qualified professional.
This article may not be duplicated, altered, distributed, saved, incorporated into another document or website, or otherwise modified without the permission of TASA.