利用OES光譜儀診斷作為電漿氮化閘極製程之氮濃度之應用

Abstract

半導體業在成本與效能的驅使下，電晶體密度不斷提高電晶體尺寸也不斷微小化，尋找閘極介電層高介電常數（high K）材料以有效降低漏電流問題，並利用金屬閘極（metal gate）縮小等效氧化層厚度，以提高電晶體速度，一直是半導體業界不斷努力的課題。但是這些先進的製程技術尚未成熟且需較高製造成本，所以目前仍以氮氧化矽（SiOxNy）的介電層薄膜為量產上的主要技術。 氮氧化矽中的氮濃度比例為電漿氮化閘極製程最重要的參數，但是由於這是個單一晶圓的集束型設備（single wafer clustered process）製程，在量產上遭遇到幾個困難，其一是當最後氮濃度結果有誤時，不易判定由哪一個製程反應室造成；即當電漿氮化製程在反應過程超過容許誤差時，並無法即時偵測。其二是傳統量化電漿氮化製程之方法為再氧化（Re-Oxidation），但因為需在晶圓上反覆鍍膜，故有耗時及較高成本之缺點。 為改善前述之缺點，本論文以5個不同波長的波段穿越濾光鏡（Band pass filter），分別為337.03 nm、357.38 nm、391.28 nm、590.43 nm及601.31 nm來偵測電漿氮化製程所產生之光放射光譜（Optical Emission Spectroscopy, OES），再以輝光強度與X-光光電子能譜儀測得的氮濃度做相關性（correlation）分析。實驗結果顯示337.03 nm波長的輝光強度與氮濃度有最高的相關性，推測此製程的氮濃度，主要是由此波長的氮離子所貢獻。因此，可以利用此波長的輝光強度來間接作為此製程中氮濃度的即時量化指標；進而延伸找到光譜訊號強度與部分的關鍵電性參數，如等效氧化層厚度，汲極飽和驅動電流和臨界電壓有線性關係存在。換言之，利用OES訊號可以幫助對元件電性參數找出最佳化的區間。而最重要的是此種預測方法具有及時與簡單之優點。

Due to the driving force of the manufacturing cost and device performance in the semiconductor industry, the transistor density has been continuously doubled in every year. Consequentially the transistor is also constantly down scaling. Therefore, looking for the material of gate dielectric with high dielectric constant (high k) to reduce the leak current and metal gate to decrease the effective oxide thickness (EOT) for high performance transistor are always the priority for semiconductor industry. Oxynitride is popularly used as gate dielectric in complementary metal oxide semiconductor (CMOS) process for device scaling before high K and metal gate being commercialized in semiconductor industry. One of the important technics to grow oxynitride is decoupled plasma Nitridation (DPN). For DPN process, the most important process parameter is the concentration of nitrogen in SiO2. However two major issues are introduced by cluster. One is that once the nitrogen concentration is out of expectation there is no way to decouple which chamber is the root cause. In other words, there is no in-situ monitoring techanic in place for DPN process. The other is that currently technic used to quantify nitrogen concentration is to use re-oxidation. However re-oxidation is more complicate and not cost effective. In this thesis, 5 band pass filters with different wave length are used to detect the spectrum intensity of species in DPN chamber. 5 of them are 337.03 nm、357.38 nm、391.28 nm、590.43 nm and 601.31 nm respectively. Thereafter the intensity of optical emission spectroscopy (OES) is used to correlate the Nitrogen concentration that measured by XPS. The results successfully demonstrate that the specific wavelength of 337.03 nm in the spectrum has the highest sensitivity and the best correlation to nitrogen concentration. Therefore it is also believed that the nitrogen concentration is dominated by this specific species. Furthermore, the result is extended and applied for optimization of the device performance from OES intensity. Results turning out show that some electrical properties of device correlate to OES intensity including equivalent oxide thickness, saturated drain current and threshold voltage. In other words, OES could be used to optimize device performance qualitatively. Finally this is an in-situ and simple methodology.