覆晶封裝銲錫技術中熱時效對電遷移效應的影響

Abstract

隨著電子元件越做越小，每個銲錫球所需承受的電流密度就越大，因此研究如何讓銲錫球可以抵抗電遷移效應是非常重要的課題。由於在迴銲的時候，Under Bump Metallurgy (UBM)層通常是Ni或是Cu，並且會和Sn產生介金屬化合物(IMC)，這一層介金屬化合物，不但是銲接上去作為接合用的介面，更由於它是高電阻的物質，我們推測可以藉由此性質降低電流集中效應，增加銲錫球的使用壽命。因此我們使用熱時效的方式，增加IMC的厚度，使用的溫度約為銲錫溶點絕對溫度的百分之九十。我們使用薄膜UBM與厚膜UBM兩種試片。薄膜試片的結構是矽晶片上UBM層為Ti 0.1 micro-m /Cr-Cu 0.3 micro-m/Cu 0.7 micro-m 薄膜，在基板上的金屬層為Au 0.025 micro-m / Ni(P) 5 micro-m/Cu 20 micro-m，銲錫球為Sn63Pb37共晶銲錫結構由於溶點為183℃，所以我們使用的熱時效溫度是150℃﹔而厚膜試片結構是UBM層為Ti 0.5 micro-m/Cu 0.5 micro-m /Cu 5 micro-m /Ni 3 micro-m，而使用在板上的金屬層為Au 1 micro-m /electroless Ni 5 micro-m，銲錫球為Sn80Pb20的成分，溶點為208℃。所以使用的熱時效溫度為170℃。我們發現在薄膜試片中熱時效會降低銲錫球在電遷移測試當中的life time，而在厚膜試片當中，適度的熱時效則會增加銲錫球的life time，沒有熱時效的平均壞掉時間是430hr，而在熱時效25hr的時候壞掉時間可以達到858hr。我們推測薄膜試片由於熱時效在界面反應後容易造成孔洞，會使銲錫球降低電遷移測試的life time；而厚膜試片則會產生較厚且穩定介面的IMC在UBM與銲錫球中間，減緩因為電流集中效應產生的空孔，增加銲錫球抗電遷移能力。我們使用紅外線觀察試片在電遷移測試後的損壞的情況觀察是否是真的壞在銲錫球上面而不是鋁導線。我們發現不同的破壞機制在有熱時效與沒有熱時效在試片整個損壞前，由於厚膜試片有著Kelvin probes結構，所以我們可以使用四點量測去測量試片中因為電遷移孔洞造成銲錫球電阻增加情形，觀察其中的破壞機制，我們發現熱時效過後的試片，在電遷移測試中容易產生IMC橋連接晶片端與基版端使電子通過，可能也是增加試片life time的重要因素之一。

Due to the trend of miniaturization and the high performance, the dimension of solder bumps keeps decreasing and the current that each bump needs to carry keeps increasing, causing the current density in the solder bump to increase dramatically. As a result of the high current density, minimization of electromigration effect becomes the key to achieve high reliability. In flip chip technology, UBM (Under Bump Metallurgy) is often used to connect chip, solder and board. UBM is mostly composed of Ni or Cu which reacts with Sn to form IMC (Intermetallic compound). IMC is usually composed of Ni3Sn4 or Cu6Sn5 and is used to joint chip and solder. We conjecture IMC can reduce current crowding effect during electromigration test because of its higher resistivity than solder. For the reasons above, our goal is to make the IMC layer as smoothly as possible in order to achieve a longer life time for solder. In our experiment, two types of samples are used; both are aged at 90% of its melting point. One of the samples is thin-film UBM bump. Its structure is Ti 0.1 micro-m /Cr- Cu 0.3 micro-m/Cu 0.7 micro-m/Sn63Pb37/Au0.025micro-m / Ni(P) 5 micro-m/ Cu20 micro-m. Because solder ball is eutectic SnPb, its melting point is 183℃. We used 150℃ to age it. The other was thick -film bump, its structure was Ti 0.5 micro-m /Cu 0.5 micro-m/Cu5□micro-m /Ni3 micro-m/Sn80Pb20 /Au 1 micro-m/ electroless Ni 5 micro-m. Sn80Ni20 its melting point is 208℃. We used 170℃ to age it. At the end of the experiment, we found aging effect reduces solder ball life time under 0.28A and 150℃ during electromigration test in thin film structure. On the other hand, we observed an opposite result for thick-film bump structure. We speculated that voids form between UBM and solder in thin-film under thermal aging. However, complete and successive IMC formed in thick-film UBM. We used infrared ray to exam the sample failure in the Al trace or solder bump after electromigration test. Because thick-film UBM has Kelvin probes structure, we also used 4-point probe to measure bump resistance in thick-film bump. Before bump open, we found different failure mechanisms between aging-free and aging solder joint. Sample which age may form IMC bridge between chip and board during electromigration test. It may be an important reason to prolong the sample life time.