標題: 超大型積體電路嵌入式導線製程的研究
Investigations on Damascene Process for ULSI Interconnects
作者: 邱紹裕
Shao-Yu Chiu
張 翼
馮明憲
Dr. Yi Chang
Dr. Ming-Shainn Feng
材料科學與工程學系
關鍵字: 嵌入式製程;銅;電鍍;電漿離子佈植;全面平坦化製程;化學機械研磨後清洗製程;Damascene process;Cu;Electroplating;PIII;Global planarization process;Post CMP cleaning
公開日期: 2005
摘要: 在半導體製造工業中,當元件大小縮減到深次微米尺寸時,高密度的多層銅金屬導體連線技術開發是因銅金屬相較於傳統得鋁合金製成有較低的電阻值以及較高得抵抗電致遷移能力,進而增進超大型積體電路的元件運算速度。為了因應銅導線製程的整合考慮,鑲嵌式金屬導線製程是目前公認為最實用的後段銅導線製程設計。因此,本論文將針對鑲嵌式金屬導線製程中,銅電鍍製程、多步驟金屬化學機械研磨平坦化製程及後清洗製程的效能改善,逐一深入探討與研究。 為了提升高深寬比溝槽的填充能力,本論文利用電子顯微鏡的剖面圖以監測銅鍍層由”溝槽底端往上:Δy ”及”溝槽側壁往內:Δx”的填充輪廓所定義的鍍液填充率(Δy/Δx),得到最佳化的鍍液添加物含量; 於銅酸性電鍍液中添加聚乙烯二醇(polyethylene glycol)、氯離子及2-mercaptopyridine,可在0.15μm,深寬比6的銅導線中得到超填充(superfilling)間隙的效果。超填充動態機制可利用吸附擴散的模型進一步瞭解添加劑於鑲嵌結構中的選擇性抑制梯度效應。另一方面,我們亦提出一種有效的晶種層沉積技術-電漿離子佈植Pd系統 (Pd PIII),預鍍一層Pd導電晶種層,再利用銅電鍍沈積銅金屬,可得無缺陷的間隙填充能力且具有良好的階梯覆蓋性。此技術需要一臨界的佈植量以驅動銅電鍍沈積;佈植Pd時,使用較高的基板電壓可提升銅電鍍沈積階梯覆蓋性,以達到由溝槽底端往上沉積(超填充)的現象,且可使銅膜有(111)的優選方向。 為建立鑲嵌製程的全面平坦化技術,我們利用新的電化學即時量測系統設計來闡明鋁化學機械研磨的機制。研究中藉由改變研磨漿料中雙氧水濃度及酸鹼值,瞭解鋁及鈦金屬薄膜在磷酸-雙氧水為主的研磨漿料系統中金屬氧化物生成與磨耗的機制。從電化學量測的結果得知,鋁金屬薄膜的研磨速率主要是控制於氧化物鈍化層的機械磨除及化學溶解,而鈦金屬薄膜的研磨速率則是受氧化物鈍化層的形成速率控制。藉此,鋁化學機械研磨的研磨選擇率,就可由研磨漿料中雙氧水濃度及酸鹼值的微調來有效控制。 為了更瞭解銅化學機械研磨的化學磨耗機制,我們利用不同鍛燒製備的次微米氧化鋁粉末,由其不同的粉末晶相含量及顆粒大小,來探討銅薄膜在不同研磨漿料環境中,化學腐蝕與粉體機械磨耗的交互作用。在高銅溶解率的研磨漿料中,銅研磨速率會隨粉體大小呈線性變化,我們可視為壓痕控制模型(indentation-limited model)。相反的,如果銅薄膜在低銅溶解率的研磨漿料中研磨,比如只單用純水或雙氧水,氧化鋁粉體在溶解控制模型(dissolution-limited model)中的研磨特性,不但可由低銅研磨速率得到印證,且在低溫相氧化鋁存在一臨界研磨速率而在純α相氧化鋁會急遽轉為磨除率飽合的現象。而在含腐蝕抑制劑的研磨漿料中,銅研磨速率在複合相氧化鋁粉體中呈反轉的現象,可由接觸面積控制(contact surface-limited)機制進一步來解釋。 銅鑲嵌製程的研究中,多步驟的銅化學機械研磨利用不同的銅研磨速率及研磨墊的選擇,可以有效減少研磨後銅金屬層的高低差。而在二氧化矽膠體的研磨漿料系統中,我們發現銅金屬薄膜、鉭金屬薄膜和二氧化矽的研磨行為和所添加的鹼有關。添加的鹼其陽離子大小會改變研磨漿料的列塔電位(zeta potential),而改變被研磨材料的研磨速率。因此在第二步驟的銅鑲嵌化學機械研磨製程中,可添加氫氧化鉀鹼液其所提供的鉀離子較小,可使鉭金屬薄膜及二氧化矽對銅金屬薄膜的研磨選擇比提高,有利於減少銅碟陷(dishing)的現象。 最後,我們提出了優化之銅化學機械研磨後中性清洗液的配方,其組成由銨基之鹼性水溶液添加離子羧酸及腐蝕抑制劑,此配方不但可有效去除氧化鋁粉體及二氧化矽膠體粒子吸附於研磨後的晶圓表面,同時亦可使銅金屬導線避免腐蝕。另外,我們研究發現,山梨糖醇(D-sorbitol)和聚乙烯醇(PVA)水溶液的清洗配方配合PVA刷子的刷洗製程,在二氧化矽膠體的研磨漿料的後清洗中,亦可有效的去除粉體的粒子的吸附。
In semiconductor manufacturing, as device dimensions continue shrinking into deep sub-micro regime, high packing density Cu multilevel interconnection technology has been developed due to its lower resistance and parasitic capacitance for increasing ULSI operating speed, compared to the conventional aluminum-alloy metallization. As considering the integration of Cu metallization, the damascene process has been expected to be very promising for fine Cu feature in IC backend metallization process. Therefore, in this thesis, the major investigations in the damascene process are carried out by improving the performance of the Cu electroplating deposition, the multi-steps metal CMP planarization and the following post CMP cleaning. To enhance the capability of gap-filling into high-aspect-ratio vias and trenches, it should to be optimized with additives by monitoring the filling ration Δy/Δx between ‘‘bottom-up’’ with ‘‘sidewall shift’’ from the cross section of a partially filled copper profile and achieved the superfilling performance for 0.15μm vias with aspect ratio 6 by an acid-copper electrolyte with polyethylene glycol, Cl﹣, and 2-mercaptopyridine (2-MP). The superfilling dynamics was explored with the adsorption-diffusion model, which explains the behavior of additives providing selective inhibition gradient within the damascene feature. On the other hand, we also propose an effective seeding technology, plasma immersion ion implantation of palladium (PIII Pd), to achieve defect-free gap-filling for copper electroplating (Cu-ECP). It was found that a threshold dosage of PIII Pd seed is required to drive Cu-ECP. To enhance the gap-filling capability, a higher substrate bias of PIII Pd is suggested to achieve the bottom-up phenomena of Cu-ECP and obtain the Cu (111) formation of electroplated copper films. To develop the global planarization technology for damascene process, a novel in-situ electrochemical measuring system was established for exploring Al CMP dynamical corrosion and to elucidate the polishing mechanisms. Firstly, Al CMP is carried out to investigate the influences of H2O2 concentration, slurry pH, and surface oxide formation-abrasion mechanism on Al and Ti in the H3PO4-H2O2-based chemistry. From these electrochemical results, the removal rate of polishing aluminum is limited to its passivating oxide removal by mechanical abrasion or chemical dissolution and titanium is limited to the formation rate of surface oxide. Correspondingly, the removal selectivity issue of Al CMP can be effectively controlled by means of slurry formulation. In order to better understanding the tribo-chemical schemes of Cu CMP, the characteristics of abrasive studies focus on the results from the interaction of abrasive wearing and chemical corrosion of copper thin film in formulated slurries with sub-micron sized alumina abrasives of varying phases controlled mainly by calcinations condition. Polishing with high Cu dissolution rate formulation, Cu removal rates varied linearly with the particle size, suggesting an indentation-limited model. On the contrary, polishing in certain dissolution-limited circumstances such as DI water or H2O2 alone, the characteristics of alumina abrasives, do not manifest onto the low Cu removal rate, but exhibit a threshold removal behavior at low-temperature transition phase Al2O3 followed by a dramatically saturation removal rate at nearly mono α-Al2O3. By use of the corrosion-inhibited slurry, note that a contact surface-limited mechanism was first introduced for the inverse Cu polishing behavior with mixed-phase Al2O3. Higher Cu removal rates and lower roughness were due to the both effects of the higher reactive performance by the smaller abrasives and the higher mechanical abrasion power by the larger abrasives. From our preliminary investigations, multi-step Cu damascene CMP with different copper removal rates and polishing pads is used to eliminate topography efficiently. In colloidal-silica-based slurry, the polishing behaviors of copper, tantalum and silicon dioxide are found to relate to the kind of alkaline additives. The size of cations from alkaline additives influences the zeta potential of slurries, so as to vary the material removal rate. During the 2nd step of Cu damascene CMP, the addition of small-sized K+ from KOH provides high removal selectivity of tantalum/copper and oxide/copper, so as to benefit the reduction of copper dishing. Finally, a superior post Cu CMP cleaning chemistry in neutral pH is formulated by adding ionic carboxylic acids into ammonium-based solution with inhibitors, and developed to effectively remove not only alumina but colloidal silica particles left on the polished wafer, meanwhile not to cause copper wires corrosion. In another research, we also find that the use of D-sorbitol and PVA aqueous cleaning solution with sponge PVA brush scrubbing have good performance on colloidal silica particle removal ability on post Cu CMP cleaning.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT008618807
http://hdl.handle.net/11536/81902
Appears in Collections:Thesis