Factors affecting the long-term stability of Cu/Al ball bonds subjected to standard and extended high temperature storage

https://doi.org/10.1016/j.microrel.2010.06.015Get rights and content

Abstract

This paper presents the findings of work performed on 20-μm diameter copper wire of five different wire types from three suppliers. Gold wire is the control. The test die was mounted on BT (B (Bismaleimide) and T (Triazine)) resin substrates. The bonding parameters were optimized for each wire used. Part of the optimization process involved monitoring the flatness of the bonded ball and the amount of aluminum remaining under the bond. The crystal structure of each type of interconnect was examined using composite imaging techniques. Visual data such as ball size, thickness, and shape were collected. First and second bonds were subjected to destructive testing, such as ball shear and wire pull, throughout the preparation process. The samples were then subjected to an industry-standard, high temperature stress test to determine the long-term stability of the interface of each wire type. Data for all read points are presented on all tests performed and provide useful information on the material and process set best suited for long term reliability.

Introduction

Wire bonding remains the preferred method of making an electrical connection between an integrated circuit or discrete silicon die and a lead frame or substrate. Industry forecasts show that this will continue with a compound annual growth rate (CAGR) of 7.2% through 2013 [1]. Fig. 1 shows the trend for various package types for the next 5 years.

Packaging costs are under constant pressure and the high price of gold makes it difficult to achieve the targeted cost reductions and maintain reasonable profit margins. A suitable alternative to gold is needed. Copper, as an alternative to gold, has been under investigation for more than 20 years. A few early references are included here [2], [3], [4], [5]. Fig. 2 shows some early work on a device using 76 μm copper wire. The date in Fig. 2b is March 23, 1996.

A comparison of the material properties of gold and copper, as shown in Table 1 [6], reveals that copper is better than gold with respect to electrical conductivity (DC). The inherent stiffness of the copper wire also makes long wires with small diameters more resistant to wire sweep during molding. The major drawbacks are in the hardness of the free air ball (FAB) and finished bond hardness. These can give rise to cratering and added stress resulting in lifted bonds. The harder FAB makes bonding on thin metallization extremely challenging, especially when there is low k material under the pad metallization [7]. However, current wire bond platforms can be programmed to minimize this impact. The risk is mitigated further when the process is co-worked with the Wafer fabrication operation and the pad designs are slightly modified.

The major issue with substituting copper wire for gold in today’s assembly process flow is one of reliability. For decades the industry has faced various reliability hurdles involving gold wire. Kirkendall voiding [8], corrosion of copper doped aluminum bond pads [9], and bromine induced failures [10], to name a few. Modifying the dopants in the wire mitigated the Kirkendall issues, minimizing the copper content in the aluminum films eliminated the pad corrosion issue on unbonded pads, and removal of bromine from plating chemistries fixed the Au/Al corrosion problem. The one issue still facing the successful implementation of copper wire as a mainstream material is still corrosion. Lifted ball bonds are still the major failure mode seen during reliability testing.

Section snippets

Objective

The objective of this work is to understand the impact of wire purity and coating (palladium) on the performance of copper wire under high temperature aging in air. There is little information in the literature on such a comparison. Understanding the behavior of different wires under this test condition will enable the user to make better decisions as to which material to select for the application.

Methodology

The wire bonding parameters for each wire type were optimized to obtain a flat interface between

Mechanical test results

Examples of the bonded ball and stitch bond shapes are shown Fig. 3 along with the measurements for the bonds. An output from the Zygo profiling system showing the bonded ball flatness and aluminum remnant is shown in Fig. 4.

The destructive test results are shown Fig. 5, Fig. 6. Fig. 5 is the graph for all the wire pull data for all six wires in the study. These curves represent the minimum, maximum, and average values of each set of data taken at the time indicated. The values are plotted on

Ion polished cross-section results

Representative samples were collected at time zero and at 1000 h for cross-sectional analysis. The samples were prepared using a special encapsulating material and were then sectioned using an ion polishing system by JEOL. This system provides a surface virtually free of defects that are normally present from conventional sectioning and polishing methods. The timing was such that the samples were moved from the polisher to the FESEM immediately upon breaking vacuum in the polisher. Thus, the

Discussion

The degradation of the wire pull values over time as shown in Fig. 5 reflects the affect of high temperature annealing of the wire. This is confirmed in the color-enhanced images taken after 1000 h where sizes and shapes of the crystal can be observed. The effect is more noticeable as the wire purity increases. This is obvious when one compares the color-enhanced composite images of Cu wire 3 at time zero and 1000 h, as shown in Fig. 11. All aspects of the original crystallography have changed,

Conclusions

This experiment was designed to examine the impact of wire purity on the isothermal aging of copper wire at 150 °C for 2000 h, using gold wire as a control. Precautions were taken to keep the samples and experimental apparatus free of moisture and halogens. Significant sampling and destructive physical analyses of the bonded balls was conducted at specified read points of the test period.

The copper wire purity had a direct impact on the wire pull performance over the test period. The 3 N wire

References (17)

  • C.D. Breach et al.

    Microelectron Reliab

    (2004)
  • X.R. Zhang et al.

    Thin Solid films

    (2006)
  • H. Xu et al.

    A re-examination of the mechanism of thermosonic copper ball bonding on aluminium metallization pads

    Scripta Mater

    (2009)
  • The worldwide IC packaging market, 2009 Edition. Electronic Trend Publications. p. 2–3 [Table...
  • Levine L, Shaeffer M. Copper ball bonding, semiconductor international; August, 1986. p....
  • J. Onuki et al.

    Investigation of the reliability of copper ball bonds to aluminum electrodes

    IEEE Trans-CHMT-10

    (1987)
  • S.L. Khoury et al.

    Comparison of copper and gold wire bonding on integrated circuit devices

    IEEE Trans CHMT

    (1990)
  • Toyozawa K, Fujita K, Minamide S, Maeda T. Development of copper wire bonding application technology. In: Proceedings...
There are more references available in the full text version of this article.

Cited by (17)

  • Corrosion-induced degradation and its mechanism study of Cu–Al interface for Cu-wire bonding under HAST conditions

    2020, Journal of Alloys and Compounds
    Citation Excerpt :

    The electrolytes used for this study included 25 ppm, 100 ppm, and 150 ppm of chloride ion concentration with pH = 7 for different wire samples (Cu, PdCu, Ag alloy, Au) as well as PH = 5.5, PH = 6.33 and PH = 7.42 at 100 ppm of chloride ion concentration for Cu wire sample only. Interfacial IMC formation and its microstructure evolution in Cu–Al binary system under isothermal aging test of HTSL test were intensively investigated by many researchers [10,28–36]. To compare with bHAST, it takes EMC-B sample as a control experiment annealing for 1000 h at 150 °C and its TEM mapping result is shown in Fig. 1a. It’s in good agreement with previous studies which confirmed by EDX analysis and its phase diagram showing the dominant lamellar IMC structure with CuAl2 on the side of the Al metallization pad and Cu9Al4 on the side of the Cu bonded ball [27].

  • Cu-Al interfacial formation and kinetic growth behavior during HTS reliability test

    2019, Journal of Materials Processing Technology
    Citation Excerpt :

    It’s greatly contingent on oxide breakdown due to ultrasonic vibrations at certain temperatures and pressures during electrical spark induced free air ball formation. Vath Iii et al. (2011) have corroborated that Cu wire purity would also impact the change of Cu grain morphology after thermal annealing process and the degree of negative Al oxide disruption during bonding, suggesting that high purity of Cu wire (5 N) with larger crystals restructuring contributed to the sharp decline of the mechanical wire pull results. In the parts of the Cu-Al interface where there are large grains, the IMC formation proceeds more slowly due to less disruption of the native Al oxide.

  • Microstructural evaluation of interfacial intermetallic compounds in Cu wire bonding with Al and Au pads

    2014, Acta Materialia
    Citation Excerpt :

    Wire bonding is a technique to attach fine wire, usually 17.8–76.2 μm (0.7–3 mils) in diameter, from one connection pad to another, completing the electrical parts in an electronic circuit [1–3]. Wire bonding using Cu wire offers a number of advantages over the wire bonding using Au wire, since the Cu wire has better electrical and thermal properties [4–11]. However, a number of problems limit the use of Cu wire, particularly because Cu wire bonding to different pad materials tends to produce a variety of intermetallic compounds (IMCs), reducing both the efficiency and the lifetime of the device’s functionality.

  • The influence of Pd on the interfacial reactions between the Pd-plated Cu ball bond and Al pad

    2013, Surface and Coatings Technology
    Citation Excerpt :

    Nevertheless, the industry is searching for an alternate bonding material in view of the booming price of Au. Compared with gold, copper has better electrical conductivity and mechanical properties [1–3]. Additionally, the growth rate of CuAl IMCs is much slower than that of AuAl IMCs [4,5], which may lead to significant contribution to reliability and lifetime in electronic devices.

  • Accelerated lifetime estimation of thermosonic Cu ball bonds on Al metallization

    2013, Microelectronic Engineering
    Citation Excerpt :

    To evaluate the bonding quality of devices subjected to elevated temperatures during service life, above mentioned tests are conducted after high temperature storage (HTS) to accelerate the IMC formation and thus its microstructural evolution. Though these thermal treatments result in a degradation of the bonding interface due to IMC formation, such as voiding, cracking and corrosion, shear tests have shown contradictory results: recorded shear strength values are increased at elevated HTS duration [8]. It can be summarized that reliability assessment of ball bonded interconnects by simple shear or pull test might be encountered with some uncertainties.

View all citing articles on Scopus
View full text