Yoshinori Endo

Process optimization and reliability study for Cu wire bonding advanced nodes

Cu wire bonding has taken over Au wire bonding due to its cost savings and other performance advantages such as higher mechanical strength for complex looping and better electrical performance. Currently, Cu wire bonding of 28nm node devices has been realized in high volume production [1]. To meet the challenges of device reliability for these advanced node devices, we need to understand how to achieve the optimal wire bonding results for the best condition for reliability tests. Previous reliability study by the authors identified one of the key factors in achieving good wire bonding reliability outcome is the wire type [2]. Pd coated Cu wire (PdCu) wire showed much better performance than Bare Cu wire. In this paper, traditional bare Cu wire, AuPd coated Cu wire and Pd doped Cu wire are compared. Biased HAST testing is one of the most challenging reliability tests for Cu wire bonded packages. Biased HAST test was performed for up to 336 hours. A medium grade molding compound was used. Coated AuPdCu wire shows the best performing wire in our study. The Pd doped Cu wire showed much better reliability than traditional bare Cu wire, indicating that the Pd doped Cu wire is a lower cost alternative to Pd coated Cu wire. In advanced node devices such as 28nm and 20nm, the pitch is further reduced to as low as 45um with the bond pad opening of around 40um. Two different wire diameters (15um and 18um) and 6 different ball diameters are tested to understand the correlation between bonded ball diameter, ball to pad contact area and reliability. ProCu5 process was used and all cells achieved over 90% IMC. Bonded ball diameters of 34um and larger passed the reliability tests for both 96 hours and 168 hours. The two smallest bonded ball diameters (30 and 32um) saw low level of failures at 96 hours. This confirms that contact area is an important factor in reliability testing. A 34um bonded ball is suitable for 45um pitch application, which means that reliability can be achieved with a medium grade molding compound for a 45um pitch application. Finer than 45um pitch devices may need higher grade molding compound to ensure reliability. Another objective of this paper is to understand the critical wire bonding responses for good reliability outcome. The wire bonding responses studied here include contact area between ball and pad, intermetallic coverage percentage (IMC%), and Al remain thickness. We found that higher values for contact area and IMC% result in increased reliability. An IMC% value of 85% is generally recommended to ensure an acceptable reliability outcome. The Al remain % is dependent on the application and the process being used. With an optimal process, Al remain around 50% of the initial Al thickness shows good reliability outcome. A less optimal process tends to require less Al remain % to pass reliability.

Molded reliability study for different Cu wire bonding configurations

With the transition from Au wire bonding to Cu wire bonding, both bare Cu wire and Pd coated Cu (PCC) wire are now commonly used. Humidity reliability is particularly challenging for Cu wire bonding. In this paper, 18 um (0.7 mil) bare Cu and PCC wire are used in the wire bonding and HAST reliability test. Various factors can influence the HAST reliability outcome. The key factors examined in this paper include: wire bonding parameters such as ultrasonic energy, force and time; wire type including bare Cu wire and PCC wire; cover gas type including forming gas and nitrogen for PCC wire; molding compound; and plasma cleaning. From our HAST test results, the most influential factor is the wire type. PCC wire exhibited much better HAST results than Cu wire. Using the recently developed Cu wire bonding processes, bonding parameters played a secondary role in the reliability outcome. For PCC wire, the lowest and highest ultrasonic energy settings both passed 336 hours with IMC at 71% and 94%. The IMC%, a key factor used in evaluating wire bonding quality, does not seem to correlate to the reliability outcome for PCC wire. A higher IMC% does help bare Cu wire HAST results. Our tests indicated that there was no difference between forming gas and Nitrogen in the HAST testing. A plasma cleaning time of 20 sec provided better HAST results than no plasma cleaning or a long plasma cleaning time. A time interval study was carried out to study the failure mechanism. Samples at different HAST time intervals were cross sectioned and analyzed using transmission electron microscopy (TEM). The different stages include before HAST but after molding and post molding cure (PMC), HAST test at 48hr, 96 hr, 168 hr and 336 hr. A typical failure starts with Cu nano-particles and Al oxides formation at the Cu rich IMC layer at 48hr. Over time, the Cu-Al IMC corrodes and more Cu and Al oxides form and leave small voids, and the voids can develop into continuous gaps (cracks). The gap normally starts at the edge of the bond and propagates inward. When the gap propagates all the way to the center of the bond, an open failure will occur. We found two primary factors that influence the HAST outcome. The first factor is the presence of the Pd elements. For the PCC wire samples, even though Pd is absent at the ball and pad interface immediately after wire bonding, it tends to diffuse into the interface during HAST testing. For a good interface with no cracks, a higher concentration of Pd is present and Pd-Cu-Al IMC is formed. At the failed bond, Pd elements are absent. This indicates that the Pd element has the effect of slowing or preventing the Cu-Al IMC corrosion or Pd-Cu-Al IMC is more corrosion resistant. That explains why PCC wire has a better HAST test outcome. The second factor is the concentration of chlorine. At failed bonds, a much higher concentration of chlorine was found, which implies the role of chlorine in the IMC corrosion. Controlling the chlorine level, and mobility, are key considerations in molding compound formulation.