Reduction of ultrasonic pad stress and aluminum splash in copper ball bonding

https://doi.org/10.1016/j.microrel.2010.06.002Get rights and content

Abstract

Given the cost and performance advantages associated with Cu wire, it is being increasingly seen as a candidate to replace Au wire for making interconnections in first level microelectronics packaging. A Cu ball bonding process is optimized with reduced pad stress and splash, using a 25.4 μm diameter Cu wire. For ball bonds made with conventionally optimized bond force and ultrasonic settings, the shear strength is ≈140 MPa. The amount of splash extruding out of bonded ball interface is between 10 and 12 μm. It can be reduced to 3–7 μm if accepting a shear strength reduction to 50–70 MPa. For excessive ultrasonic settings, elliptical shaped Cu bonded balls are observed, with the minor axis of the ellipse in the ultrasonic direction and the major axis perpendicular to the ultrasonic direction. To quantify the direct effect of bond force and ultrasound settings on pad stress, test pads with piezoresistive microsensors integrated next to the pad and the real-time ultrasonic force signals are used. By using a lower value of bond force combined with a reduced ultrasound level, the pad stress can be reduced by 30% while achieving an average shear strength of at least 120 MPa. These process settings also aid in reducing the amount of splash by 4.3 μm.

Introduction

Thermosonic Au wire bonding is the most common first level interconnection technology used in microelectronics [1], [2], [3]. It is used to weld microwires to metallized pads of integrated circuits. In the microelectronics industry, there is a continuous push towards higher performance and lower costs [1], [2], [3], [4]. This has led to an increased interest towards the development of lower-cost bonding wire materials. Compared to Au wire, Cu wire has lower cost, superior electrical and thermal conductivities, and higher mechanical strength [5], [6], [7], [8]. However, since Cu is harder than Au, higher normal and ultrasonic forces are often used, resulting in bonds stronger than Au [9]. This results in ≈30% higher stress [10], [11] delivered to the bond pad during Cu ball bonding than during Au ball bonding. The higher pad stress increases the likelihood of underpad damage, such as pad peeling [12], [13] or bulk silicon cratering [9], [14], [15], [16], particularly in the case of sensitive substrates, such as chips with low-k dielectrics.

Another effect of high bonding stress specifically observed during ball bonding of Cu on Al bond pad is the squeezing of Al pad metal from the peripheries of the ball bond [17], [18], [19]. The Al material squeezing (or splash) occurs in the ultrasonic direction. Splash is not desired as it results in localized pad thinning [19], which can reduce bond reliability.

There are several different approaches in reducing underpad stress and therefore limiting underpad damage during thermosonic bonding with Cu wire, e.g., by using softer Cu wire types [10], [11], producing softer free air balls (FABs) [16], [17], [18], [19], [20], [21], [22], [23], optimizing the bonding parameters [11], [24], [25], using higher frequency ultrasound (US) transducers [26], modifying the bond pad thickness and design [26], [27], using harder bond pads, such as Ni/Pd, Ni/Au, and Ni/Pd/Au [27], or by using pre-US (ultrasound applied during impact portion of bonding) [28].

Previous studies reported that a double stage bonding process [24], [25], in which an impact force (IF) that is higher than the bond force (BF) is used, aids in minimizing chip damage. Recently, a new fast method to reduce the extra stress observed during Cu ball bonding was reported [11]. It was found that an US level, about 15% lower than the conventionally optimized level can be used to obtain Cu ball bonds of comparable geometry and strength to that of Au ball bonds, while reducing the stress gap by ≈40%. The study considered the optimization of the US parameter only. The effect of BF on bonding quality and pad stress was not investigated. The present work addresses the question: what is the synergistic effect of BF and US on the bonding quality, splash, and the stress delivered to the bond pad? Parts of this work have been presented in [29]. Splash measurements and further discussions are added.

Section snippets

Experimental

Thermosonic ball-wedge bonding is performed using a 25 μm diameter Cu wire (MK Electron Co. Ltd., Yongin, Korea) on an automatic ESEC 3100 ball-wedge bonder (Besi Esec, Cham, Switzerland), having an ultrasonic frequency of 128 kHz. The bonding is performed at a nominal heater plate temperature of 150 °C, resulting in a temperature of ≈138 °C at the bond pad. A commercial ceramics bottleneck capillary having a hole diameter of 35 μm and a chamfer diameter of 51 μm is used. During the formation of

Conclusions

The conventional optimization aims at maximum bond shear strength for the targeted bond geometry. However, it also causes the highest ultrasonic stress on the pad. Therefore, a better optimization criterion could be adequately high shear strength, opposed to maximum shear strength. To this end, contour curves of the maximum ultrasonic force measured by the microsensor are shown in the US/BF process space. The results demonstrate that the underpad damage risks typical to a Cu ball bonding

Acknowledgements

This work is supported in part by the Natural Science and Engineering Research Council of Canada (NSERC), the Initiative for Automotive Manufacturing Innovation (IAMI) of Ontario and Ontario Centers of Excellence (OCE), Microbonds Inc. (Markham, Canada), and MK Electron Co. Ltd. (Yongin, Korea). Generous financial support in the form of Government of Canada’s NSERC Alexander Graham Bell Canada Graduate Scholarship and University of Waterloo President’s Graduate Scholarship is gratefully

References (36)

  • S. Kaimori et al.

    The development of Cu bonding wire with oxidation-resistant metal coating

    IEEE Trans Adv Packag

    (2006)
  • Deley M, Levine L. The emergence of high volume copper ball bonding. In: Proc SEMICON west conf;...
  • C.W. Tan et al.

    Cratering on thermosonic copper wire ball bonding

    J Mater Eng Perform

    (2002)
  • A. Shah et al.

    Low-stress thermosonic copper ball bonding

    IEEE Trans Electron Packag Manuf

    (2009)
  • C.M. Tan et al.

    Failure analysis of bond pad metal peeling using FIB and AFM

    IEEE Trans Compon Packag Manuf Technol

    (1998)
  • C.M. Tan et al.

    Effect of BOE etching time on wire bonding quality

    IEEE Trans Compon Packag Technol

    (1999)
  • C.W. Tan et al.

    Bond pad cratering study by reliability tests

    J Mater Sci: Mater Electron

    (2002)
  • Caers J, Bischoff A, Falk J, Roggen J. Conditions for reliable ball/wedge copper wire bonding. In: Proc IEEE/CHMT Eur...
  • Cited by (28)

    • Cu-Al intermetallic compound investigation using ex-situ post annealing and in-situ annealing

      2015, Microelectronics Reliability
      Citation Excerpt :

      The Al metallization at ball periphery will be thinned down leaving a very thin layer for diffusion to take place. This observation is in agreement with other published observations [21–23]. The overall IMC is observed to growth continuously in proportion to the annealing time with thickness ranging from a few nm grows to a few μm.

    • μ-Raman spectroscopy for stress analysis in high power silicon devices

      2014, Microelectronics Reliability
      Citation Excerpt :

      It can also be observed that the region of the crystal in contact with the wire bonding exhibits slight compression stress. The existence of this stress could be related to the ultrasonic thermo-compression process [17]. Stress map was also realized on an IGBT cross section in order to study the local residual stress issue from synthesis around IGBT elementary cells.

    • Role of impact ultrasound on bond strength and Al pad splash in Cu wire bonding

      2013, Microelectronics Reliability
      Citation Excerpt :

      It is not a failure by itself but can reduce bond reliability because of pad material thinning. Pad metal splash is of particular concern in today’s ultra fine pitch applications where it may cause shorting between the adjacent bonded balls pads [1,18,19]. Ball bonding can be described as a mixture of two processes types: Ultrasound Enhanced Deformation (UED) process and impact deformation process [9].

    • Copper wire bonding process characterization and simulation

      2020, CIPS 2020 - 11th International Conference on Integrated Power Electronics Systems
    View all citing articles on Scopus
    View full text