| dc.description.abstract | The goal of this thesis is on improving the understanding of mechanical and tribological mechanisms in microelectronic wire bonding. In particular, it focusses on the development and application of quantitative models of ultrasonic (US) friction and interfacial wear in wire bonding. Another objective of the thesis is to develop a low-stress Cu ball bonding process that minimizes damage to the microchip. These are accomplished through experimental measurements of in situ US tangential force by piezoresistive microsensors integrated next to the bonding zone using standard complementary metal oxide semiconductor (CMOS) technology. The processes investigated are thermosonic (TS) Au ball bonding on Al pads (Au-Al process), TS Cu ball bonding on Al pads (Cu-Al process), and US Al wedge-wedge bonding on Al pads (Al-Al process).
TS ball bonding processes are optimized with one Au and two Cu wire types, obtaining average shear strength (SS) of more than 120 MPa. Ball bonds made with Cu wire show at least 15% higher SS than those made with Au wire. However, 30% higher US force induced to the bonding pad is measured for the Cu process using the microsensor, which increases the risk of underpad damage. The US force can be reduced by: (i) using a Cu wire type that produces softer deformed ball results in a measured US force reduction of 5%; and (ii) reducing the US level to 0.9 times the conventionally optimized level, the US force can be reduced by 9%. It is shown that using a softer Cu deformed ball and a reduced US level reduces the extra stress observed with Cu wire compared to Au wire by 42%.
To study the combined effect of bond force (BF) and US in Cu ball bonding, the US parameter is optimized for eight levels of BF. For ball bonds made with conventionally optimized BF and US settings, the SS is ≈ 140 MPa. The amount of Al pad splash extruding out of bonded ball interface (for conventionally optimized BF and US settings) is between 10–12 µm. It can be reduced to 3–7 µm if accepting a SS reduction to 50–70 MPa. For excessive US settings, elliptical shaped Cu bonded balls are observed, with the major axis perpendicular to the US direction. By using a lower value of BF combined with a reduced US level, the US force can be reduced by 30% while achieving an average SS of at least 120 MPa. These process settings also aid in reducing the amount of splash by 4.3 µm.
The US force measurement is like a signature of the bond as it allows for detailed insight into the tribological mechanisms during the bonding process. The relative amount of the third harmonic of US force in the Cu-Al process is found to be five times smaller than in the Au-Al process. In contrast, in the Al-Al process, a large second harmonic content is observed, describing a non-symmetric deviation of the force signal waveform from the sinusoidal shape. This deviation might be due to the reduced geometrical symmetry of the wedge tool. The analysis of harmonics of the US force indicates that although slightly different from each other, stick-slip friction is an important mechanism in all these wire bonding variants.
A friction power theory is used to derive the US friction power during Au-Al, Cu-Al, and Al-Al processes. Auxiliary measurements include the current delivered to the US transducer, the vibration amplitude of the bonding tool tip in free-air, and the US tangential force acting on the bonding pad. For bonds made with typical process parameters, several characteristic values used in the friction power model such as the ultrasonic compliance of the bonding system and the profile of the relative interfacial sliding amplitude are determined. The maximum interfacial friction power during Al-Al process is at least 11.5 mW (3.9 W/mm²), which is only about 4.8% of the total electrical power delivered to the US transducer. The total sliding friction energy delivered to the Al-Al wedge bond is 60.4 mJ (20.4 J/mm²).
For the Au-Al and Cu-Al processes, the US friction power is derived with an improved, more accurate method to derive the US compliance. The method uses a multi-step bonding process. In the first two steps, the US current is set to levels that are low enough to prevent sliding. Sliding and bonding take place during the third step, when the current is ramped up to the optimum value. The US compliance values are derived from the first two steps. The average maximum interfacial friction power is 10.3 mW (10.8 W/mm²) and 16.9 mW (18.7 W/mm²) for the Au-Al and Cu-Al processes, respectively. The total sliding friction energy delivered to the bond is 48.5 mJ (50.3 J/mm²) and 49.4 mJ (54.8 J/mm²) for the Au-Al and Cu-Al processes, respectively.
Finally, the sliding wear theory is used to derive the amount of interfacial wear during Au-Al and Cu-Al processes. The method uses the US force and the derived interfacial sliding amplitude as the main inputs. The estimated total average depth of interfacial wear in Au-Al and Cu-Al processes is 416 nm and 895 nm, respectively. However, the error of estimation of wear in both the Au-Al and the Cu-Al processes is ≈ 50%, making this method less accurate than the friction power and energy results. Given the error in the determination of compliance in the Al-Al process, the error in the estimation of wear in the Al-Al process might have been even larger; hence the wear results pertaining to the Al-Al process are not discussed in this study. | en |
