銅金屬嵌入式導線之化學機械研磨技術研究

Abstract

銅金屬嵌入式導線製程之化學機械研磨技術研究 研究生：陳辰靜 指導教授 : 馮明憲 博士 蔡明蒔 博士 國立交通大學材料科學與工程研究所 摘 要 在ULSI積體電路製作上，銅因其有較低之電阻值以及較高抵抗電致遷移的能力，已成為未來金屬連線的另一選擇。而在多層金屬化導體連線的製作上，全域性平坦化就成為其製程中的關鍵技術之一，化學機械研磨是目前唯一可達全域性平坦化的技術。在此論文中，將就研磨液中的研磨粉體對於銅薄膜化學機械研磨，以及銅的鑲嵌式製程進行探討。 於銅薄膜之化學機械研磨之研究中，利用不同的氧化鋁粉末來探討銅薄膜在不同研磨液環境之下，其與研磨粉體之間的研磨作用。首先分別在純水與濃度為百分之十體積比的雙氧水溶液之下進行銅的化學機械研磨，發現在每個粉體的研磨結果均屬不佳，不僅銅的研磨速率低，移除率的不均勻度也都在百分之十五以上。當在雙氧水溶液中加入檸檬酸之後，銅的研磨情況則大大改變，主要是因為檸檬酸加速了氧化銅的溶解，再藉以粉體所提供的機械磨耗，增加了銅的移除率。在此，也發現以純α氧化鋁粉體為研磨粉體，有最高的磨除率而且銅表面在研磨後的表面粗糙度是所有粉體間最小的；其主要原因為化學溶解與機械磨耗同時作用於銅表面，恰達到平衡並得到良好的研磨結果。在以硝酸為氧化劑的溶液中 ，硝酸提供銅的腐蝕環境，而粉體在此提供的機械磨耗程度也隨著粉體的特性會有不同的研磨結果。另外，在硝酸溶液中加入苯基疊氮，Benzotriazole(BTA)，與檸檬酸之後，整個研磨結果就變的不一樣。這兩種添加劑在硝酸溶液中所扮演的角色為抑制硝酸對銅的腐蝕。研磨結果與粉體之間性質不同的差異性就變得不明顯；銅的移除主要控制在銅表面保護膜的生成與移除，粉體提供的機械磨耗就比較有限。所以可以得到一個結論：粉體在一良好的腐蝕環境之下，包括雙氧水與檸檬酸混合的溶液與硝酸溶液，會因α相、密度、粒徑的增加與比表面積的減少，有助於銅膜移除速率的提升。不過，若是在一中性溶液或抑制腐蝕的環境之下，粉體就不再類似前述的研磨行為，而是得到差異性不大的研磨結果。 在銅鑲嵌式製程研究中，我們做了一詳細的闡述。在實驗一開始，溝渠（trench）外的階層（step）必須先磨除掉，所以選擇對銅有一高移除率的研磨條件，而且有必要選擇一具有較大硬度、較小可壓縮性的研磨墊，如Rodel IC1400? 研磨墊，來避免較低的區域(low features)也被磨到，導致無法達到平坦化的目的。而當溝渠外的銅膜快被移除完時，我們選擇一低的銅移除速率，銅對鉭有高移除選擇比的條件，來完成溝渠外銅完全移除的的步驟。進入到第二步驟，即面臨到鉭、銅、氧化矽介電層移除的選擇比率問題。鑲嵌式製程結果的好與壞就決定在第二步驟中，如何選擇銅對鉭與氧化矽的移除選擇比。我們以氧化矽膠體（colloidal silica）與雙氧水混合的研磨液應用在此步驟，也以鉭與氧化矽的移除率比銅快為研磨條件，有助於降低原本在第一步驟就已造成的碟陷（dishing）現象。而且在過研磨（over-polish）的階段，得到碟陷現象更小的研磨結果。

Study on Chemical-Mechanical Polishing in the Copper Damascene Process Student: Cheng-Ging Chen Advisor: Dr. Ming-Shiann Feng Dr. Ming-Shih Tsai Institute of Materials Science and Engineering National Chiao-Tung University Abstract Under intensive investigation for Ultra-Large-Scale-Integration (ULSI), copper has emerged as an attractive, alternative choice for future interconnect applications owing to its low electrical resistivity and high electromigration resistance. On the manufacturing of multi-level metallization, globe planarization is one of the crucial techniques. Chemical mechanical polishing (CMP) is the only way known to achieve to the globe planarization. In this thesis, the characterization of abrasives for copper chemical mechanical polishing and the damascene process for Cu metallization are presented. In the study of chemical-mechanical polishing, the removal performances of Cu polished with various alumina abrasives and slurry formulations were investigated. While polishing copper film with various alumina abrasives separately in the DI water and 10 vol.% hydrogen peroxide solution, it is found that the polishing performances are poor, not only low removal rate, but also more than 15% removal non-uniformity. However, mixing the citric acid within the hydrogen peroxide solution, the polishing rate increases because of the dissolution of copper oxide by the citric acid and mechanical abrasion supplied by abrasives. Polishing copper with pure alpha alumina, in the slurry formulation of 10 vol.% H2O2/Citric acid obtained the highest removal rate and the smallest surface roughness, it is indicated that there is a balance between chemical dissolution and mechanical wear during polishing process such that an optimized performance is acquired. In the nitric acid-based slurries, with both H+ and NO3- present, it provides a corrosion environment for copper. The polishing results varied with the characteristics of alumina abrasives. Alternatively, with Benzotriazole (BTA) and citric acid presented in HNO3 based slurry, the behavior of abrasives during polishing would become diverse. The roles of these two additives play to form a native or nonnative passivation film on copper surface to inhibit the corrosion from HNO3, the differences of Cu removal with various abrasives become unapparent. Copper removal is conducted by formation and polishing removal the passivation layers; mechanical abrasion by abrasives was limited. Conclusions in this part are as the following: In the corrosion, e.g., the hydrogen peroxide solution mixed with citric acid or nitric acid-based slurries, no steady CuO forms, copper removal is enhanced with the increased alpha-phase content, density, and primary particle size, and the reduced of the BET surface area of the abrasives. Nevertheless, in the neutral solution or corrosion-inhibited environment, the abrasive behavior is no longer the same as described in the corrosion case. On the topic of Cu-damascene process, the detail procedures during CMP process were investigated. Topography planarization, end-point detection in the first step, and the removal selectivity between copper, tantalum, and silicon dioxide in the second step are discussed. During planarization of the topography of patterned wafer, phase 1 of the first step, high copper removal rate and hard polishing pad instead of soft pad were selected to eliminate topography efficiently. The deformation of pad under load is the most concern about whether the planarization could be achieved. Faster step height reduction occurs on the narrow space region in phase 1 of the first step. In order to release the process window for end-point detection, low copper removal rate and high removal selectivity of Cu to Ta would be required and carried out by polishing with optimized slurry formulation in the phase 2 of the first step. For the second step, the key issue is to find a slurry formulation with appropriate suitable removal selectivity of Cu/Ta/Oxide. The removal selectivities of Cu/Ta/Ox are 1: 3.7: 1.63 for slurry pH 7, and 1: 5.76: 2.76 for slurry pH 8. Copper dishing and oxide erosion were evaluated. Copper dishing resulted in the first step was reducing after the second step of damascene process and the amount of dishing for each measured linewidth was less than 1000 A. Oxide erosion was about less than 1000A for low pattern density region, but more than 1200 A for high pattern density region.