Resumen de Life prediction of copper wire bonds in commercial devices using principal component analysis (PCA)
Subramani Manoharan, Chandradip Patel, Steven Dunford, John Beshears, Patrick McCluskey
Copper has been widely adopted as a viable alternative to gold for wire bonding, due to its lower cost, better electrical and thermal conductivity, lower tendency for wire sweep due to a higher modulus and slower intermetallic compound (IMC) formation with aluminum bond-pad. However, these IMCs can cause an increase in resistance and separation between bonding interface which causes an open circuit. Wire bond and component reliability must consider all factors in combination with the change of wire bond materials and processes. How can users of off-the-shelf components determine the effects of wire bond material changes on the reliability of the components in their systems? How will these changes affect high reliability applications such as automotive, military and harsh environments? Although physics-of-failure based models can represent the failure mechanisms individually, the complex nature of the mixed failure mechanism has inhibited development of an analytical equation to represent it. In addition, geometrical and material characteristics of the entire package along with bonding process parameters play a crucial role in magnitude of stress experienced by these bonds, portraying difficulty in developing one such equation for use to all copper wire bonded devices. This study has aimed to develop a data-based predictive method based on the principal component analysis technique to provide an estimate of the wire bond time to failure for plastic encapsulate components. Extensive thermal cycling experiments have been performed on multiple commercial devices to obtain training data for the model. This method of wire bond fatigue and failure prediction is an adaptive technique which can achieve greater accuracy as more data is added. This method aims to provide a life-time estimate of copper wire bonded component with inputs such as initial geometrical, material properties, and wire bond related attributes for a specific package.
