ULSI嵌入式銅導線之無研磨粒子平坦化技術

Abstract

為提升元件之效能，在新的ULSI積体電路製程中，以銅導線取代原始之鋁導線已成為不可避免的趨勢，這主要是由於銅金屬具有較低之電阻值及較高抵抗電子遷移的能力。然而，在多層銅金屬導体連線的製作上，化學機械研磨技術雖是達成全域性平坦化的關鍵技術，但卻仍有許多問題等待解決，例如刮痕、磔陷、以及研磨之均勻度等等。在此論文中，將就銅金屬鑲嵌式製程的平坦化過程中所產生之刮痕缺陷的解決方式作一深入的探討。 在一般銅金屬化學機械研磨過程中，在第一個階段的研磨所使用的研磨粒子多為高溫燒結後所得之氧化鋁粉，由於氧化鋁粉的体積大且表面形態較不規則，在研磨過後表面會留下既多且深的刮痕，導致元件的損壞，為求得無刮痕缺陷之平坦表面，論文中使用的第一個方式是採用無研磨粒子之化學機械研磨平坦化技術，此技術之平坦化效能的主要影響因素在於研磨墊與研磨過程中銅金屬表面所形成之抗腐蝕保護層的性質。為求得較好之平坦化效果，故必須採用較硬且較不易產生形變的研磨墊才能有效的移除凸起處之抗腐蝕保護層，借此達成凹凸位置上移除速率明顯的不同，進而達到平坦化的要求。因此，實驗中所選用的研磨墊為較硬之Rodel IC – 1400,在所有實驗中，氧化劑皆固定為硝酸溶液，而抑制劑則有苯基疊氮(BTA)以及聚乙醇(PEG)兩種。由於在單純僅有笨基疊氮的硝酸溶液下，所得到之研磨均勻度均不理想，這主要是由於銅金屬表面與笨基疊氮反應形成斥水性質的抗腐蝕保護層，導致研磨液在基材上的潤濕效果不均，故為提升研磨速率的均勻度，加入聚乙醇混合以增加研磨液的潤濕能力，亦在論文中有進一步的探討。 此外，為達成無刮痕之平坦表面，論文中所採用的第二個方式是以表面改質過後的研磨粒子替代原始的氧化鋁粉末，此改質研磨粒子是借由凝膠方式形成圓形之氧化矽懸浮粒子，再接以化學方式在氧化矽粒子上沉積一層氧化鋁，希望借此能夠同時保有氧化鋁粉末的化學性質以及經由圓形的外貌避免因不規則形貌所造成的刮痕。在這階段的實驗中，所採用的氧化劑為雙氧水，借由電化學的分析可以得知，隨雙氧水濃度的增加，銅金屬的氧化態會有所不同，實驗目地主要是在探討刮痕的存在原因以及銅金屬其不同氧化態對於平坦化效能的影響。 至於論文的最後一個部分，則是以氧化矽膠体在碘酸鉀為氧化劑的條件下混合為研磨液，除了探討碘酸鉀分子的作用外，對於銅、鉭的選擇比與研磨後碟深的變化作更進一步的了解。經由實驗的結果可以得知，若銅導線與阻障層金屬的研磨選擇比可以控制在理想值1，則在第二段的化學機械研磨過後，碟深的值將不會有明顯的改變，也就是說，若銅、鉭、以及介電層材質的選擇比接近1，則對於研磨終點的控制，將可容許更大的誤差。

For improving the effectiveness of semiconductor device, instead of aluminum interconnection by copper material has become necessary in ULSI IC process. It is because copper has higher electron migration and lower resistance. However, although chemical mechanical polishing is the important technology to approach globe planar polished copper surface, there are still many issues existing for solving, like scratches, dishing, and uniformity. In this study, we discuss the way for approach the goal of scratch – free surface in copper damascene CMP. Generally, the abrasives applied to first step copper CMP are almost pure aluminum powders. Owing to its large size and irregular morphology, there will exist much and deep scratches on polished surface leading to damage of device. For achieving the goal of scratch free, the first way is applying abrasive free polishing instead of pure aluminum system. The planarization efficiency of abrasive free polishing is affected by the properties of polishing pad and passivation. The harder pad has less deformation effect during polish and faster remove of passivation, those will lead to clearly different removal rate between protrudent and recess areas. By the way, the required planar surface will be achieved. Therefore, in all experiments the polishing pad is used by Rodel IC – 1400. Besides, the oxidizer is fixed to be pure nitric acid aqueous, the inhibitors are selected by BTA and PEG. While abrasive free slurry formulated by nitric acid and BTA, the uniformity of polishing rates are not satisfied. By means of the observation of hydrophobic phenomenon on polished copper surface, it impacted that the surface tension existing between slurry and copper surface is too small leading to bad wettability of abrasive free slurry. For improving the wettability of abrasive free slurry, surfactant PEG is added into original polishing slurry. Besides, the second way to achieve scratch free copper surface is applied modified colloidal silica instead of pure aluminum powder. From sol – gel production, silica abrasives would have higher suspension and become circular morphology. Furthermore, the colloidal silica abrasives were coated a thin alumina layer that would retain the chemical characteristic of aluminum powders. By the modified surface, we hope to avoid the scratches produced from irregular morphology. In this partial section, the oxidizer is fixed to be hydrogen peroxide. From the analysis of electrochemical measurement, it was found that copper would be oxidized to different oxidizing state as the concentration of hydrogen peroxide increase. Unfortunately, the scratches still exist after polish. The goal of experiment is to find out the reason of scratch existance and the affection of copper oxidized behavior to planarization efficiency. Finally, we discuss the potassium iodate system with colloidal silica abrasives in order to understand the relation between selectivity and dishing. From the results of experiments, if the selectivity of tantalum to copper closes to 1, the dishing shift after second step polish would be nearly unapparent. It impacted that if the selectivity of copper, tantalum, and dielectric layer closes to 1, the control of end point could be neglected.