慣性的迷失(Lost in inertial) - III

 慣性的迷失(Lost in inertial) - III

現在矽基半導體元件的TEM分析主要分成二個主要階段：FIB試片製備和TEM分析。由於絕大多數的矽基半導體元件使用[001]晶圓製作，電晶體元件沿{110}方向排列。因此，在第一階段，FIB工程師習慣上沿某一[110]方向切出TEM試片。在第二階段，TEM工程師也習慣上，一定是先將TEM試片傾轉，使試片的[110]方向和電子束一致，然後再進行後續的分析。

Currently, TEM analyses of silicon based semiconductor devices are divided into two main steps, FIB sample preparation and TEM analysis. Since most of Si semiconductor devices are made by [001] Si wafers and transistors patterns are aligned along {110} directions. So, in the first step, FIB engineers are used to cutting chips along one of {110} directions. In the second step, TEM engineers are also used to tilt the specimen to [110] exact zone first, then proceed following analysis.

圖1(a)顯示一張FIB的二次電子影像，圖中紅色箭頭所指之處，第一層金屬(M1)和鎢栓(W)非常接近。由於解析度的問題，此影像無法確認第一層金屬和鎢栓是否已經接觸?因此必須進行TEM分析。這個特殊案子的TEM試片製作，理想上應該是照著如圖1(b)中虛線矩形的輪廓切挖成TEM試片，才能看清楚第一層金屬和鎢栓是否已經接觸。可是由於慣性思考使然，FIB工程師仍然照慣例，沿矽基板的[110]方向切挖，按照如圖1(c)中虛線矩形的輪廓做出TEM試片。

Figure 1(a) is an FIB SE image, the position pointed by the red arrow is a place where metal 1(M1) and a tungsten plug (W) close to each. However, it is hard to say whether they are touched or not due to the resolution limit of FIB. So, a cross-section TEM specimen has to be made for following TEM analysis. In this non-routine case, the TEM specimen should be cut as indicated by the dashed-line rectangle in Figure 1(b). But, following the inertial thinking, this TEM specimen was still cut along the [110] direction, as shown by the dashed-line rectangle in Figure 1(c) routine cases.

圖1.  FIB 二次電子影像。(a)紅色箭頭指處，第一層金屬(M1)和鎢栓(W)很接近；(b)虛線矩形代表理想TEM薄片輪廓示意圖；(c)虛線矩形代表實際TEM薄片輪廓示意圖。

如果TEM工程師也是照慣例，先傾轉TEM試片，使試片的[110]正極軸方向和電子束一致，然後拍攝影像，得到如圖2(a)的影像，得到的結論將是第一層金屬的腰部和鎢栓是接觸的。可是將TEM試片做適當的傾轉後，使觀察方向接近沿圖1(a)中的紅色點線AB方向時，可以發現第一層金屬的腰部和鎢栓雖然很接近，但是仍有幾奈米的間隙，如圖2(b)所示。從圖1(a)和圖2(b)得到的訊息，對於設計工程師或製程工程師而言，修正的工作有相當程度的差異。

If the TEM engineer did this case in conventional way, tilted the specimen to [110] exact zone condition, then took pictures, one of results is shown in Figure 2(a). The conclusion drawn from this image would be that the waist of metal 1 touched with the tungsten plug. But, if the specimen is tilted to a certain degree, and the viewing direction is nearly along the red dotted line AB in Figure 1(a). We will find that the waist of metal 1 does not touch the tungsten plug. There is a gap of several nano meters between them, as shown in Figure 2(b) by tilting the specimen correctly. What needed to be tuned in design or process would be significantly different from the information of Figure 2(a) or Figure 2(b).

圖2. 橫截面型TEM明場像。(a)沿[110]正極軸方向拍攝的TEM明場像，顯示第一層金屬(M1)和鎢栓(W)接觸；(b)沿[001]軸傾轉20幾度後拍攝的TEM明場像，顯示第一層金屬和鎢栓未接觸。

在歐美日已開發國家的科技業界，FIB和TEM操作人員的教育程度至少是碩士以上。在台灣，製造業的精神延伸到材料分析領域，專科生都可以被訓練操作FIB和TEM，人力成本大幅降低，形成一材料分析量產的產業。對於例行性的、大量的半導體元件分析，也不失為一降低成本的良策。但對於非例行性的案子，制式化的慣性操作，只是得到大量排列整齊的影像和若干組彩色的成份映像而已，TEM強大的分析功能並未被充分發揮，真正的材料訊息仍遺留在試片內。

Engineers in charge of FIB and TEM are usually masters or doctors in TEM laboratories in developed countries, such as European, USA, and Japan. In Taiwan, the spirit of cost down in manufacture has been extended to the field of materials analysis. Many college students were trained to operate FIB and TEM. TEM analysis becomes an emerging industry in Taiwan due to the reduction in the labor costs. It is not bad for routine and mass TEM analysis of specified patterns in semiconductor devices. But, for non-routine cases, only mass aligned images and color composition maps were obtained from inertial standard operation. Useful information of materials was still not explored and powerful capability of TEM was not fully used.

慣性的迷失(lost in inertial)

     在日常例行性的交通中，從A處到B處，已經習慣某條行車路線後，如果臨時想去C處買個小東西，在前往途中，一不留神就會在必須轉彎的街口繼續沿舊路線前進。個人將此種行為稱之為慣性的迷失。對於許多人來說，慣性的迷失在日常生活中經常發生，大多無傷大雅。但是如果發生在工程案件和商業案件中，則會造成相當程度的損失，小則浪費幾天的人力物力，大則危及企業根本。

    After having driven from place A to place B daily for years, most people are gradually used to going to and fro along one or two routes. Occasionally, we may forget to make a turn at the right intersection when we want to buy something at the middle way, place C. I call this mistake to be “lost in inertial” that happens in daily life from time to time for many people. Most of them are harmless. However, when it happens in engineering cases and/or business cases, it will result in damages to some extent. The loss may be only a few of manpower and material resources, may be a huge damage to the enterprise.

廿幾年前在工研院材料所微結構分析實驗室工作。某天接到一個半導體元件良率的案子，客戶送來二組表面長有磊晶層的矽晶圓，經過相同的製程，用A組晶圓製作的元件的良率都低於50%，而用B組晶圓製作的元件的良率則高於95%。客戶要鑑定A組晶圓內的磊晶層是否有造成低良率的缺陷。利用TEM進行樣品的橫截面分析。第一天，完成A組晶圓的TEM試片製作，在TEM中，觀察到矽磊晶層和矽基板的界面有幾個數奈米大小的空孔。第二天，完成B組晶圓的TEM試片製作，但是在TEM中找不到矽磊晶層和矽基板的界面。估計沒有磨對位置，第三天，再製作第二個B組晶圓的TEM試片，仍然找不到矽磊晶層和矽基板的界面。

I received a case of yield issue when I worked at microanalysis laboratory in MRL, ITRI more than twenty years ago. Semiconductor devices made by two sets of epi Si wafers had big different yield under same processes. Set A had yield less than 50%, while set B had yield more than 95%. My mission was to characterize what kind of defect(s) in the epi layer. I used TEM to analyze the cross-section structure of these samples. A TEM specimen of A was made at the first working day, and several nano voids were observed at the interface of epi layer and Si substrate. A TEM specimen of B was made at the second day, but nothing except Si substrate was observed. With the assumption of grinding and polishing wrong place, a second TEM specimen of B was made at the third day. Unfortunately, there was still only one single crystal observed in TEM. The interface of epi Si layer and the Si substrate as observed in sample A was not found.

第四天，預定交付資料的前一天，完成第三個B組晶圓的TEM試片，在TEM中搜尋一遍後，仍然找不到矽磊晶層和矽基板的界面。此時已經晚上11點。疲乏的我到室外走走散心，順便思考今晚是否要研磨第四個B試片，以及如何向客戶提出延期的說詞。沈思中，忽然靈光一閃，摻雜濃度等級的矽同質磊晶層和矽基板之間，正常狀況下，本來就看不到界面。一直想要看到磊晶層和矽基板的界面是因為A試片和以往分析異質磊晶層樣品的經驗造成的慣性思維。

    The third TEM specimen of sample B was prepared and checked in the TEM at the fourth day, one day before the deadline. It was around 11:00pm, and there was still no solid data of sample B to deliver. Walked out TEM laboratory to take a rest and thought what was wrong with the TEM sample preparation of sample B as well as how to ask the customer to delay one or two days. During meditation, something came to my mind – nothing is correct. The interface of homogeneous epitaxial layer and the substrate is invisible when the epitaxial layer is well grown. I had been trapped in the inertial of previous knowledge from analyzing heterogenous epitaxial layers and sample A for all these days.