metallization will yield before much energy can be transmitted to the under- lying substrate);
2. the delay in the onset of cratering for thicker metallizations (a longer period of time would be required to penetrate the thicker metallization and alloy with the underlying silicon substrate);
3. the effect is significantly less for pads over SiO2(Au-A1 intermetallic does not form as good a bond to SiO2as it will to bare silicon); and
4. as the weld area increases with additional intermetallic formation, the esti- mated stress concentration factors become significantly reduced, thus explaining the reduction in cratering in strong bonds after prolonged thermal aging.
In summarizing this cratering effect, a thicker bond pad metallization, a larger weld area (high ultrasonic energy and stage temperatures), and flatter bonds (high force) will produce bonds that are more resistant to any form of shear-induced cratering for bond pads on SiO2. Weak bonds are to be avoided since they can crater at much lower shear values. Preliminary information suggests that for bonding above multilevel oxides such as PSG, BPSG, and other low-strength chemical vapor (CVD) deposited oxides, thicker pad metal- lization and the combination of low power, short dwell, hot stage temperature, and high force will be required to produce reasonably strong bonds and reduce damage to the underlying dielectric layers.
Historically, ball bonding standards required relatively large balls com- pared to the wire diameter (i.e. greater than or equal to 2.5 times the wire diameter) yielding effective bond areas five to six times greater than the wire cross-sectional area (see Fig. 4.2). Today ball bonds are small nail heads with sizes down to approximately 25mm in diameter (for 15 mm diameter wire).
Thus, the typical modern ball diameter standard requires the ‘‘ball’’ to be about 1.4–1.5 times the wire diameter (D). Such diameters would yield effec- tive bond cross-sectional areas (wire to pad) [12] of about twice that of the wire and thus should still be robust enough to avoid ball lifts under pull testing on well made bonds. Ball (nail head) diameters down to 1.2D have been seen in production. Figure 4.23 is a scanning electron microscope photomicrograph of fine pitch thermosonic ball bonds.
In the ultrasonic bonders, the historical deformation of the bond foot was 1.5 times the wire diameter. See Fig. 4.3. Today that deformation is down to about 1.1–1.2 times the wire diameter in fine pitch applications. Thus, for the same wire diameter wedge-wedge bonds can be placed closer together provided the required bond wire geometry (height, length, first bond- second bond location, etc.) can be accomplished with the ultrasonic bonders in line step. Figure 4.24 us a scanning electron microscope photomicrograph of fine pitch ultrasonic wedge bonds.
As mentioned above, today’s packaging environment is creating new geo- metrical challenges for wire bonding. Not only are we seeing multi-tiered pad arrangements (two, three and sometimes four tiers as discussed above, See Fig. 4.6) but also the requirement for low profile bonds necessitated by the wide spread deployment of stacked packages. Stacked packages typically have one of two staked die configurations: (1) pyramid or (2) overhanging die of the
Fig. 4.23 Scanning electron photomicrograph of ultrafine pitch (55mm) thermosonic ball bonding.
The bonds were made on a K&S Model 8020 automatic ball bonder using 23mm (0.9 mil) diameter gold alloy wire. Pad metallization was Al + 1% Si + 2% Cu on SiO2 with nominal 1mm thickness. (Photomicrograph courtesy of L. Levine, K&S)
same size. These two stacking configurations are shown schematically in Fig.
4.25. Stacking requires special wirebond profiles and low loop height. As die thickness decreases the spacing between the loops of the different tier wirebonds must decrease proportionally to avoid wire shorts between the different wiring layers. The top layer loop also must remain low to avoid wire exposure during molding. The maximum loop height should be no higher than the thickness of the die to maintain an optimal gap between the wire tiers. For example, if the die thickness is 100mm, the optimal loop height would be 100mm or less.
Some situations even require reverse bonding (i.e. the ball is on the package substrate and the tail bond is on the die). In a normal ball bonding process the ball is placed on the die contact pad and then after wire looping to the second bonding location the tail or stitch bond is formed on the substrate or package contact. In a reverse ball bonding process a stud bump is placed on the chip Fig. 4.24 Scanning electron photomicrograph of ultrafine pitch (40mm) wedge bonds. The bonds were made on a K&S Model 8060 automatic wedge bonder using 20mm (0.8 mil) diameter gold alloy wire. Pad metallization was Al + 1% Si + 2% Cu on SiO2with nominal 1mm thickness. (Photomicrograph courtesy of L. Levine, K&S)
contact pad. This bump provides elevation above the chip and acts as a force distributor for the stitch bonding process to come. Next the chip is wire bonded with the ball bond placed on the substrate and the tail bond placed on the bumped chip bonding pad. Loop height with reverse bonding can be less than 75mm. Over hanging thin die (thickness down to 50mm) require special bonding techniques die to die flexure (bending) upon application of bonding force. The use of delayed application of ultrasonic energy after capillary touchdown is a must.