微影製程疊對量測改善

目前半導體製程，在關鍵尺寸日益精密，越做越小的情況下，各層之間疊對的準確度也越來越重要，因為關鍵尺寸小，相對的layer 與layer 之間可以容忍的偏移量，也就變小，比如以往在0.6μm 的製程， layer 與layer 之間可以容忍的偏移量為0.15μm，現在已進步至0.18μm 的製程， layer 與layer 之間的允許偏移量，就會下修到0.06μm，而layer 與layer之間的對準狀況將其量化後的量測值，就稱為「overlay」。 Overlay 既然代表著前層與當層的對準狀況，因此量測的結果就會受到前層與當層圖形的影響，前製程所沉積的薄膜會因為生成的特性，影響到量測mark 的圖形，而當層因為是光阻定義的圖形，會較清晰，不影響量測，本研究主要在解決受到前製程所沉積的薄膜所影響的overlay 量測異常。因為受到前製程所沉積的二氧化矽薄膜影響，所形成的異常影像，稱之為double edge profile，此為造成的異常量測的主要原因，但這並非真正的異常，而是屬於量測所造成的false alarm，在現在重視產能產出的生產線，此種異常會延遲生產線上的順暢，同時也增加線上工程師異常處理的負擔，因此，此篇研究的主要內容就是要降低overlay 量測時因為double edge profile 造成false alarm 的發生。 Overlay 量測的準確性與overlay mark 影像清晰度有絕對的關係，所以本研究的實驗方法有三種，都是以改善影像的清晰度為目的： 1. 量測機台，量測程式參數的調整 (利用濾鏡產生不同光源) 機台所提供的不同濾鏡，可以提供不同的光源，不同光源對於影像清晰度有相關，本研究確認機台內波長以藍光最短，所以可以得到最佳影像。 2. 光罩的轉換 (reverse tone 與new mark type) 光罩的轉換，方法有二，第一reverse tone，先以類似於二氧化矽折射率n 值的光阻將外框覆蓋作平坦化，降低因為外框二氧化矽沉積所造成的折射反射效應，第二new mark type，利用深寬比的原理，減少沉積二氧化矽所造成的double edge profile 效應。 3. 製程改善 (clear out 外框二氧化矽) 量測異常是因為外框內沉積的二氧化矽所形成的double edge profile 所造成，利用矽與二氧化矽在四氟甲烷內氫的比例不同的蝕刻率原理，將外框內的二氧化矽清除。本研究所用的三方法，都有效改善overlay 量測；以量測機台參數做調整，可以得到立即的改善，此為最快也是以最小的改變所能達到最大的效果，適用於現有異常改善；以光罩的設計方面，可以根本根絕此量測問題，用於新產品改善最佳；以製程的改善，也能有效根絕異常情況，但以成本與風險考量來說，與上述兩方法相比，較不適用。

關鍵字

微影製程；疊對；光罩

並列摘要

In semiconductor industry, the critical dimension is increasingly smaller. In this case, between the layers overlay accuracy has become increasingly important. Because the critical dimensions is small.The offset spec between layers are becoming smaller. For example, in the past 0.6μm manufacturing process, the offset spec between layers are 0.15μm, However advances to 0.18μm manufacturing process, the offset spec between layers are shrinking to 0.08μm, the alignment accuracy between Layers measured values, is called the "overlay". Since the overlay mean alignment condition between layers. Therefore, the measurement result will be affected by the former layer pattern. Because of the characteristic for the film deposition will affect the measurement mark. When the layer defined as a resist pattern is clearer, so it does not affect the measure results, therefore in this research, we will find the resolution to solve the overlay abnormal measure results that affected by the film deposition for front layer. Because of the film deposition for front layer cause the abnormal measure image, this phenomenon we call double edge profile. It causes the abnormal measure result, however it is not true alarm, it is false alarm that cause by abnormal image. This abnormal phenomenon will add loading for the engineers and delay throughput. Therefore, in this research we will fine the way to solve the false alarm cause by double edge profile. The overlay measure accuracy has relation with definition for the overlay mark image. So in this research we have three methods to improve the definition for the mark. 1. Tool parameter adjust (different filter had different light) Different light has different wavelength, and different light will caused different image for the overlay mark. We find blue light has shortest wavelength and overlay mark in this condition has clear image. 2. Mask change (reverse tone and new mark type) For mask change, we have two way, first is reverse tone, we use the photoresist that has similar index of refraction with silica to cover the out mark for reducing double edge profile phenomenon. Second are new mark type, we use the principle for ascept ratio to reduce the double edge profile phenomenon by the film deposition. 3. Process improvement (clear out silica that fill in out mark) Because the silica causes abnormal measure result, we use the different etch ratio between silicon and silica to clear out the silica. The three ways in this research, all can solve the abnormal measure result that cause by double edge profile phenomenon. Use tool parameter we can get improvement immediately. It is suitable for improving abnormal condition now on. Use change mask type can solve this abnormal phenomenon completely. It is suitable for new product mask tape out. Process change, it can solve this abnormal phenomenon completely too. However it has high cost and high risk, compare with the other two methods, it is not suitable to improve.