Thermosonic Au ball bonding process investigated using microsensor and laser vibrometer

Wire bonding is the most widely used interconnection technology in microelectronics. Aggressive technology scaling has led to device miniaturization resulting in a need for ultra-fine pitch bonding. A deep understanding of the bond formation process will help improve wire bond process reliability and product yield. This paper reports on the study of ball bond (1st bond on the chip) formation process using state-of-the art piezoresistive microsensors integrated at the bonding pad and laser vibrometer. Au wire ball bonding is performed on a clean Al bond pad and a contaminated Al bond pad. A good quality bond is formed on the clean bonding surface while ball non-stick on pad (NSOP) is observed on the contaminated bonding surface. The microsensor data clearly shows the difference between a good quality bond and NSOP. For a good ball bond, the third harmonic of the ultrasonic force is high initially while the ultrasonic vibrations cause scrubbing of the native Al oxides. The signal then decreases as the bonding between Au and Al starts. In the case of NSOP, the third harmonic signal stays high throughout the duration of ultrasonic period, showing no signs of bond formation. It is found that the tool-tip amplitude measured using the laser vibrometer can be explained well with the ultrasonic force signal. When the magnitude of the ultrasonic force signal is low (i.e., less stress, therefore less bond formation), the tool-tip amplitude is high. As the bond starts to form, ultrasonic force rises, while the tool tip amplitude decreases. The third harmonic of the tool-tip amplitude has a similar trend as that of the ultrasonic force signal. A process DOE was performed with varying impact force, bonding force and ultrasonic current level. Microsensor data is collected for each setting including both the Y direction ultrasonic force and Z directional normal force. Process responses such as bonded ball diameter and shear strength are also collected. By correlating bonding response and sensor data, it shows that microsensor measurement can be very useful in understanding process response and optimizing process.