PLASMA ENHANCES PACKAGE YIELDS AND RELIABILITY
BY JAMES D. GETTY
To ensure high device reliability and minimize manufacturing costs, it is important to optimize the wire bonding process for good bond strengths and yields. Poor bond strengths and low yields often are due to upstream contamination sources or the materials selection in advanced packaging. Gas plasma technology can be used to clean pads prior to wire bonding for improved bond strengths and yields.
Gas plasma is a powerful, efficient resource that can dramatically improve the manufacturability, reliability, and yield of advanced packages. Plasma is used to improve the pull strength and uniformity of wire bonds, increase fillet height, fillet uniformity, and underfill adhesion for flip chip devices, and to alter surfaces for better adhesion in mold and encapsulation processes. Numerous factors dictate the effectiveness of a plasma process, including choice of chemistry, process parameters, power, time, part placement, and electrode configuration. Electrode configuration and process chemistry are key factors when considering a plasma process for specific packaging applications. Successful implementation of plasma in packaging these devices requires an understanding of both the device to be packaged, including its materials of construction, the preprocessing steps, and any sensitivity, as well as plasma technology itself.
Thin-film Metallization
In today's advanced substrate technologies, low-cost substrates are manufactured using very thin gold (Au) plating, typically on nickel (Ni) or palladium (Pd) metallization. The gold thickness is very thin, typically less than 50 nm. The presence of the epoxy can lead to poor wire bond pull strengths and bonding yields. The thickness of this Au becomes a challenge to the plasma system when considering its use for bond pad contamination removal, because of epoxy bleed-out from the die attach step. The challenge is to successfully remove the organic resin bleed with plasma, without damaging or removing the thin-film Au required for the wire-bonding step.
Two plasma modes can be used to treat the substrate prior to wire bonding: direct or downstream plasma. The direct plasma mode uses an energy source to ionize and dissociate a source gas, creating a gas plasma that consists of physically and chemically active components. Samples to be plasma-treated are placed directly in the gas discharge, on or near the electrode plates of the system, with full exposure to the working species of the plasma (i.e., ions, free radicals, and byproducts). The type of working species that the substrate is exposed to is a function of the source gas selected. For example, if argon (Ar) is used as a source gas, the plasma-generated Ar ions impact the substrate surface and remove the organic residue via a sputtering mechanism. For example, a quad flat no-lead (QFN) package with 25 nm of Au on Pd was evaluated for wire bond improvement with and without Ar-direct plasma. The die was attached with conductive epoxy, oven cured, direct plasma treated, and wire bonded with 25-µm wire. A statistically valid set of samples yielded a mean pull strength of 10.00 g, with a CpK of 2.07 with plasma, as opposed to a mean pull strength of 3.89 with a CpK of 0.03 without plasma. With tightly controlled process conditions, therefore, direct plasma can be used to dramatically improve wire bond pull strengths.
 Close-up: Gas plasma vacuum chamber. |
Oxygen can also be selected as the source gas. In this case, the direct plasma-generated active species include oxygen ions, oxygen radicals, and byproducts such as ozone. The oxygen radicals generated in the plasma oxidize the organic resin, producing gas phase carbon dioxide and water with a slight assist from the oxygen ions. An alternative to direction plasma is downstream ion-free plasma (IFP). IFP plasma is a pure chemical plasma, free of both ions responsible for the physical component and photons. The IFP process consists of the generation of active species upstream of the sample processing area, followed by diffusion of active species through a gas baffle assembly. The gas baffle removes the ions, electrons, and photons, allowing the substrate to be exposed only to the radicals and byproducts generated in the upstream plasma. The downstream plasma mode is used when the substrate or die is sensitive to the exposure of ions or photons generated in the direct plasma.
 Table 1. A tightly controlled plasma process ensures that all organic resin bleed is removed, without sputtering thin-film Au on the bond pad. |
When processing substrates with thin metallization, downstream and direct plasma can be considered. In a single plasma cycle, it is possible to remove all of the gold on the substrate bond pads, which dramatically and negatively impacts the wire bond pull strengths. For example, identical QFN packages with wire bond pads consisting of 25 nm of Au on Pd were die attached with conductive adhesive, oven cured, direct plasma treated using Ar-source gas under different power and time conditions, and wire bonded with 25-µm wire. Table 1 illustrates the importance of tightly controlling the plasma process to ensure that all the organic resin bleed is removed without sputtering the thin film gold on the bond pad. The pull strength data was collected with a constant plasma power, pressure, and Ar source gas, while varying the plasma process time. The “Under Treated” sample showed some improvement over the “No Plasma” sample. When compared to the “Optimized” conditions, the process did not remove all of the epoxy bleed. An additional set of experiments was conducted to illustrate the importance of tightly controlling the plasma process. The “Over Treated” sample yielded poor pull strengths because of the removal of the thin-film Au bond pad material.
 Figure 1. QFN pull strengths under no plasma, direct plasma, and IFP oxygen conditions. |
Ion-free plasma can be used in cases where substrate metallization or the semiconductor device technology is sensitive to the direct plasma exposure. A thin-film Au QFN package was die attached with conductive adhesive, thermally cured, plasma treated under direct and ion-free plasma conditions, and wire bonded with 25-µm wire. With IFP plasma, the QFN is only exposed to the chemically active oxygen radicals, thus limiting the effect of sputtering the Au. Figure 1 displays pull strengths for these QFN packages under no plasma, direct oxygen plasma, and IFP oxygen plasma conditions. In both plasma cases, the wire bond pull strength and CpK improve dramatically when compared to the no-plasma condition. The direct plasma condition is slightly better, however, indicating that removal of the organic resin bleed is not the only mechanism for improved bond pull strength. Studies are underway to further understand this observation.
Additive Substrate Technology
There are three major metallization processes used in the manufacture of substrates: subtractive, additive, and semi-additive. Traditional methods use subtractive metallization, which involves the application of a blank metal followed by photolithography and metal etch to form the substrate traces. In additive plating, the metal traces are directly built on the substrate. Additive plating often is used because it offers advantages for small geometries required in high-density substrates. Two typical sources of contamination with additive plating can impact wire bonding pull strength and yield: organic contamination from the substrate manufacturing process and Ni diffusion from the plating. An appropriately configured plasma system can effectively treat these sources of contamination and improve wire bond yields.
An additive-plated substrate was used to study the effectiveness of plasma for improving wire bond pull strength under conditions of no plasma, oxygen-based, and Ar-based plasma. Results displayed in Table 2 indicate that both plasma processes significantly improve the pull strengths, while maintaining high CpK values. The pull strength data, however, does not indicate whether the pull strength improvement was due to the removal of organic contamination or the reduction of Ni on the bond pad surface.
 Table 2. Both oxygen-based and argon-based plasma processes improve the pull strengths, while maintaining high CpK values. |
To further understand the plasma-enhanced pull strength improvement, x-ray photoelectron spectroscopy (XPS) was used to evaluate the performance of the two plasma processes for removal of the organic and Ni contamination. Relative concentrations of carbon (C) , Ni, and Au were measured on the substrate bond pads (Table 3). The no-plasma condition data shows that the gold bond pad was contaminated with organic contamination, indicated by the high C content, as well as Ni. The oxygen-based process was efficient for the removal of the organic contamination via a chemical mechanism, but did not effectively treat the Ni. The Ar sputtering process was more efficient in removing the Ni contamination.
 Table 3. Relative amounts of carbon, nickel, and gold were measured on the substrate bond pads. |
Examining the oxygen-based data more closely, it can be concluded that the majority of the organic contamination that limits wire bond pull strength was effectively removed, and the remaining organic was adventitious C, as seen both by the increase in pull strengths displayed in Table 2 and the relative increase concentration of Au displayed in Table 3. The relative increase in the Ni content for the oxygen-based plasma was likely caused by the effective reduction of the top layer of organic contamination, and the exposure of the bond pad Ni contamination lying below the organic. The relative small amount of Ni, however, does not appear to be the significant factor in the pull strength improvement when comparing the no plasma condition in Table 2 to the two plasma conditions.
 Close-up: Wire bonding example. |
The Ar-based plasma uses a sputtering mechanism to remove both the organic and the Ni. The data in Table 3 illustrates both the reduction in the C levels, as well as Ni. It is likely that the slight improvement in the pull strengths for the Ar process, as displayed in Table 2, was due to the reduction in the Ni content on the bond pad. When considering the type of plasma chemistry required for the additive plated substrates, throughput requirements must be balanced with chemistry considerations. Typically, the chemically based processes (e.g., oxygen plasma) will provide shorter cycle times than those driven only by sputtering processes. In either case, the plasma process enables these substrates to be wire bonded.
Conclusion
Advanced packaging technologies continue to drive material innovations to satisfy the requirements of more functionality in smaller packages. The use of these new materials and processes, such as thin metallization, low-k dielectric materials, alternative substrate manufacturing techniques, and new coating technologies challenge traditional manufacturing methods. As a result, plasma processing is often required in wire bonding applications to enable acceptable pull strengths and improved bonding yields. Considerations for the contamination sources, and sensitivity of materials must be considered when configuring the plasma system. When optimally configured, plasma is an enabling technology for the enhancement of advanced package yields and reliability.
JAMES D. GETTY, Ph.D., director of applications and business development, may be contacted at March Plasma Systems, 2470-A Bates Ave., Concord, CA 94520; e-mail: [email protected].