Moisture Induced Corrosion in Gold and Copper Ball Bonds

  • 格式:pdf
  • 大小:1.19 MB
  • 文档页数:11

Moisture Induced Corrosion in Gold and Copper Ball Bonds 1C. D. Breach, 2Wai Mun Tee, 2 Teck Kheng Lee and 3R. Holliday

1ProMat Consultants, 160 Lentor Loop, #08-05 Tower 6, Singapore 789094

cbreach@promat-consulting.com 2ITE College Central, 20 Yishun Ave 9, Singapore 768892

3World Gold Council, 10 Old Bailey, London EC4M 7MG, United Kingdom

richard.holliday@gold.org

Abstract High gold prices have led to renewed interest in replacing gold with copper in existing packages and new packages in order to save costs. Although reliability is often cited as a reason for using copper, the main driving force for its use is undoubtedly cost. Perceptions that copper wire is more reliable are based on the notion that the intermetallics grow more slowly and that thinner intermetallics are more reliable and yet old data tend to support the idea that copper is as reliable as gold. More recently however, copper ball bonds on aluminium metallization have been found to fail more than gold during temperature cycling (TMCL) and pressure cooker testing (PCT) [1-3]. The key feature of these tests is the presence of moisture than appears to accelerate corrosion. A proposed solution to this problem is Pd-coated Cu wire, which looks to be a promising but relatively untested solution at present [1-4]. While Pd-coated copper wire may improve reliability, cost-savings will be less compared with bare copper wire. It is important to note however that a recent SEMI survey shows that industry is generally not as confident in copper as wire equipment and materials suppliers [5]. Localized corrosion of aluminium bond pads is well known in microelectronics packaging and in Al-Cu and Al-Cu-Si bond pads, CuAl2 acts as a cathode and aluminium corrodes in the presence of water (electrolyte) [6]. The situation is complicated by the presence of chlorine and other ionic contaminants in addition to moisture. In copper ball bonds, intermetallic coverage in as-bonded balls is very difficult to see but it appears that CuAl2 and Cu9Al4 are the compounds that form initially [7]. Intermetallic growth of Cu-Al compounds is slow compared to Au-Al compounds, which means that in finished packages, a large amount of Al remains whereas in gold ball bonds, with thin 1µm bond pads, aluminium can easily be consumed after encapsulation and moulding and surface mounting. It is plausible that during extended periods of exposure to moisture, stress and ionic contaminants, Cu ball bonds may be more susceptible to localized corrosion than Au ball bonds because slow intermetallic growth permits aluminium corrosion. This paper discusses ball bond corrosion and suggests that for high reliability applications or applications in moist environments, it may be necessary to accelerate the growth of Cu-Al intermetallics to mitigate potential corrosion and failure.

1.0 Introduction The great amount of ongoing debate on the potential of copper to completely replace gold wire in electronics packaging is by itself testament to the fact that copper wire is not a drop in replacement for gold. Individual opinions tend to vary from (a) unwavering and unshakeable blind faith that

copper will entirely replace gold to (b) more measured and scientific opinions that recognise the benefits and disadvantages of each material and the difficulties that are often encountered when implementing copper on mass production scale. Fortunately, according to a recent SEMI survey [5] the electronics packaging industry generally seems to be taking a cautiously optimistic approach to copper (approach (b)). Assuming that the prevailing attitude towards changing to copper wire in most companies is (b), there are two basic routes that can be taken to further understand the differences between copper and gold wire performance. The first is to use standard tests to compare the performance such as HTST, TMCL, PCT and so forth and demand only that these tests are passed. The second approach is to perform standard and non-standard tests and push a wire material to its limit and understand what determines the performance limits of a particular material. Alternatively, both approaches can be taken. With either approach, the amount of experimental work necessary to cover even the most basic of test conditions can be rather large if parameters such as wire type, wire supplier, moulding compound type (supplier), HTS temperature, PCT, etc. are considered. The large number of variables coupled with a trend over the past thirty years towards smaller and smaller lot sizes for evaluation of reliability makes it difficult to properly assess reliability except where gross failures are expected. Therefore it is no small task to properly compare reliability performance. Rather than try to address all issues in the copper versus gold wire debate, this study presents preliminary results that focus on corrosion effects on copper and gold reliability. Copper wire is considered metallurgically to be rather unreactive but it is able to react with moisture and oxygen and therefore relative to gold wire it is prone to corrosion. Even so, corrosion of copper is sensitive to the moisture level, pH and contact with other materials, which can determine which material corrodes. In ball bonds the situation is more complex because concerns are not just with corrosion and oxidation of the wire but with corrosion of materials in contact at elevated temperature in the presence of moisture, which can result in galvanic corrosion. Galvanic corrosion can occur between CuAl2 precipitates in bond pads and the Al matrix and it is known that CuAl2 is known acts as a cathode and having a higher (more positive) corrosion potential than aluminium a current flows that causes Al to react and form corrosion products [6]. A similar effect is known to occur in copper – aluminium alloys that contain CuAl2 precipitates by design for the purpose of strengthening where Al corrosion can sometimes be observed and can be a serious reliability issue [8]. The results presented in this paper are part of an overall effort to understand the reliability of copper and gold wires