376 376 其中 A7 B7 五、發明説明（/ ) 本發明是有關於一種即時檢測單一材料層厚度之方 法，且特別是有關於一種化學機械硏磨法（chemical-mechanical polishing, CMP)製程終點檢測的方法 ° 化學機械硏磨法，簡稱CMP，是現今唯一能提供VLSI， 甚至ULSI製程，u全面性平坦化(global planarization)"的一 種技術。而在半導體製程進入深次微米領域後’ CMP已成 爲一習用之製程技術，其中終點檢測（endpoint detection, EPD)是擴大CMP製程參數範圍（process window)與克服量 產穩定性最重要的關鍵問題。 目前習用之介電層終點檢測技術中最具有潛力的是光 學方式，其原理是量測反射激光強度對時間的周期性變 化，配合觀察窗邏輯(window logic)尋找曲線斜率轉折點， 來決定CMP製程的”解析終點'analytical endpoint)，此方 法的優點是即時性、非接觸式且雜訊較少。 以下敘述現有終點檢測之原理及所遭遇的問題。 以單純的介電層（包含二氧化矽層或氮化矽層)CMP爲 例，因爲沒有碰觸到不同介面而產生出的強烈訊號變化， 因此必須由反射激光強度本身變化的特性來決定製程終 點。請參照第1圖，反射激光強度/，可由光學干涉原理 得到： cos⑹............(1)376 376 of which A7 B7 V. Description of the invention (/) The present invention relates to a method for detecting the thickness of a single material layer in real time, and in particular to a chemical-mechanical polishing (CMP) process endpoint detection Method ° Chemical mechanical honing method, abbreviated as CMP, is the only technology that can provide VLSI and even ULSI processes, u global planarization ". After the semiconductor process enters the deep sub-micron field, CMP has become a customary process technology. Among them, endpoint detection (EPD) is the most important key issue to expand the CMP process window and overcome the stability of mass production. . Among the currently used dielectric layer endpoint detection technologies, the most promising is the optical method. The principle is to measure the periodic change in the intensity of the reflected laser light over time, and cooperate with window logic to find the turning point of the curve slope to determine the CMP process. "Analytical endpoint", the advantages of this method are immediacy, non-contact and less noise. The principle of existing endpoint detection and the problems encountered are described below. A simple dielectric layer (including silicon dioxide) Layer or silicon nitride layer) CMP as an example, because there is no strong signal change caused by touching different interfaces, the end point of the process must be determined by the characteristics of the reflected laser intensity itself. Please refer to Figure 1, reflected laser intensity /, Can be obtained from the principle of optical interference: cos⑹ ............ (1)
Amd. Λ c〇s(ar 木紙悵尺廋適用中囤围家標？( C'NS } Λ4規格（2〗0Χ297公釐） (誚先閲讀背面之注意事項再填寫本頁) -'φ 3766t A7 B7 五、發明説明（之） ~ 上述/爲反射激光強度，r爲時間，心爲一次反射激 光的強度’ A爲二次反射激光的強度，W爲該材料層的折 射率’ .J爲該材料層的厚度，^爲該入射激光的波長， ώΤβ.爲折射角。 而在扣切冗，仿=〇, 1，2,……，所對的厚度4可由方 程式(2)得到： = mZ0cos(aK/)Amd. Λ c〇s (ar wood paper 怅 廋 廋 is suitable for the standard of house hoarding? (C'NS} Λ4 specification (2〗 0 × 297 mm) (诮 Please read the precautions on the back before filling in this page) -'φ 3766t A7 B7 V. Description of the invention (of) ~ The above is the reflected laser intensity, r is the time, the heart is the intensity of the primary reflected laser 'A is the intensity of the secondary reflected laser, and W is the refractive index of the material layer'. J Is the thickness of the material layer, ^ is the wavelength of the incident laser light, and Τβ is the refraction angle. In the case of slicing, imitating = 0, 1, 2, ..., the thickness 4 can be obtained from equation (2): = mZ0cos (aK /)
方程式（1)顯示反射激光強度對時間（I_t曲線）呈現餘弦 函數的關係’實際之圖形如第2圖所示，且I-t曲線與介 電層厚度有關’故在理論上可由此I_t曲線即時決定出CMP 硏磨中介電層的厚度及製程終點。 一種終點檢測的方法爲，先試拋一片晶圓以取得反射 激光強度對硏磨時間的完整曲線圖形（這裏所選的硏磨時 間應涵蓋整個可能的製程範圍）。然後依製程需求決定出 終點厚度，並且在第2圖中的I-t曲線上找出對應終點厚 度的曲線位置即可。之後，硏磨後續晶圓時，只要對照試 拋晶圓與正在硏磨的晶圓之I-t曲線，就可以決定出硏磨 製程的終點。 而在實際施行上，因硏磨液流體分布及每片晶片差異 性等因素影響，每片晶片對應的曲線_幅變化都不盡相 同’且由於每片晶片起始的介電層厚度略有不同，曲線在 時間軸上會有不同的位移，所以很難由上述方法去找出製 程終點。此外對不同的製程需求，可能有不同的終點厚度， 4 本紙張尺度州中家标今（（'NS ) Λ4規格（2]0X297公釐） " ί I—------裝----^----訂------線 (对先閲讀背'&之注意事項再填寫本頁) 3 B7 . . -- ~ 五、發明説明（彡） 在曲線上對應的終點位置也會不一樣’因此必須有一套確 實可行的法則來輔助判斷。 由，於上述種種問題’使得我們無法單純由反射激光的 強度値或固定的斜率位置來決定出製程終點。因此’有另 一種解決方法被提出。如方程式(2)所示’曲線上的每個極 値點所對應的厚度是固定的’因此由這些極値點的數目可 形成一種特徵點或指標’幫助找出最接近製程終點的極値 點202位置，通常這一點就稱爲”解析終點"(analytical endpoint) 204,因爲這是在曲線上可以明確定義出來的位 置，而實際的製程終點不必然與這些”解析終點”重合’當 兩者不相同時，則再加以數秒鐘的固定時間硏磨來達到製 程終點206，如第2圖所示。 理論上，這些極値點可由曲線斜率等於零的位置而決 定出來，不過因製程上轉盤轉速的限制，實際取樣頻率不 會太高，通常爲1次/秒，故不易由這種方式來精確計算出 來，較實用的方法是以斜率的正負變換來定出這些轉折 點，而第2圖中同時顯示決定解析終點常用的觀察窗邏輯 方法，其由連續數個斜率的變化來定義出曲線的轉折點。 不過這種簡化方法的缺點是精確性與再現性不佳。因爲要 避免雜訊干擾及曲線斜率變化所導致的誤判情形發生，故 必須以連續3〜4個觀察窗(window) 208(第2圖）才能來定義 出一個轉折點，所以精確性不佳。而且這種方法終究無法 保證檢測的成功率。 因此對單一介電層或淺渠溝隔離（shallow trench 本紙張尺度边中國1¾¾:揼々（（，ns ) λ4規格（2丨297公釐） J.---------^----^---ΐτ------^ (邻先閱讀背面之注意事項再填穷本頁) 3 7661 wt'l. doc/006 A7 __B7 _ 五、發明説明（¥) 1S〇iati〇n，STI) CMP製程而言，現有的終點檢測技術只能由 入射激光的波長，定義出最接近實際製程厚度的極値點位 置。而;且因材料本身、環境與製程上的變異，導致目前在 量產使用上所遭遇的問題是：一、因訊號會隨著材料不同 而有所差異，故無法保證檢測結果的成功率。二、對於不 同型式產品必須使用不同的條件來決定”解析終點”位置。 三、對於多轉盤硏磨製程，轉盤間的訊號位準落差問題會 影響訊號的再現性。四、無法適用於硏磨時間特別短，尤 其是只包含不到一個週期訊號的製程。 而造成這些問題的原因是因爲在訊號分析技術上無法 精確的定義出反射激光強度曲線的極値點，以及無法快速 而明確的判斷出曲線的斜率是上昇或下降。 有鑑於此’本發明的一目的就是要精確的定義出反射 激光強度曲線的極値點，藉此檢測介電層的厚度^ 本發明的另一目的就是要快速而明確的判斷出反射激 光強度曲線的斜率是上昇或下降，提昇檢測介電層厚度時 精確度。 本發明的再一目的就是要克服上述之問題，使得訊號 終點檢測可以應用於量產製程上。 根據本發明之上述目的’提出一種即時檢測材料層厚 度之方法。上述材料層底部具有一基底，並且材料層厚度 會隨時間而漸減，同時以一入射激光射向材料層而取得一 反射激光。此方法包括將反射激光的強度對時間之積分曲 線’除以反射激光的強度對時間之微分曲線和時間的乘 6 本紙张尺度1¾ /1]屮K B2家標彳（（、N'S ) 現栝（210x 297公釐) ' ------ L--Γ--------裝------訂------線 _銷先閲讀背面之注意事項再填寫本頁) ___ B7 五、發明説明（g ) 積，得到一I—DT轉換曲線，其中I-DT轉換曲線具有餘割 函數性質，藉此利用餘割函數具有明顯的峰値之特性，可 用於即I時檢測出材料層的厚度。 根據本發明之上述目的，提出一種化學機械硏磨法的 製程終點檢測之方法。此方法於化學機械硏磨法硏磨一材 料層時’同時以一入射激光射向材料層而取得一反射激 光；以及，將反射激光的強度對時間之積分曲線，除以反 射激光的強度對時間之微分曲線和時間的乘積，得到一 I-DT轉換曲線，其中I-DT轉換曲線具有餘割函數性質，而 利用餘割函數具有明顯的峰値之特性，可用於即時檢測出 材料層的厚度，藉此並可決定製程終點何時到達。 爲讓本發明之上述和其他目的、特徵、和優點能更明 顯易懂，下文特舉一較佳實施例，並配合所附圖式，作詳 細說明如下： 圖式之簡單說明： 第1圖顯示介電層CMP光學終點檢測之原理示意圖； 第2圖顯示由反射激光的強度對時間曲線決定製程終 點之示意圖； 第3圖顯示本發明將反射激光的強度對時間之積分曲 線除以微分曲線和時間的乘積(故稱爲I-DT轉換)所得到之 反射激光強度曲線的示意圖；以及 第4圖顯示應用本發明I-DT轉換於分析短硏磨時間 製程之示意圖° 圖式之標記說明： 7 本(，NS)Λ4規格（2丨0〆297公釐） r---------^----^---'ST------,^ (讳先閏讀背面之注意事項再填寫本頁) υΨύ^{Φ)Τφ'\Φ7 l Λ · ίΙ{ αφ t αφ A7 B7 五、發明説明（ 202:極値點 2Q4: it曲線解析終點 叫6:製程終點 208:觀察窗 3〇2:極値點 3〇4 Μ-DT轉換曲線解析終點 306 :製程終點 3〇8, 312, 314, 316 :峰値點 31〇 · I-t曲線解析終點 :特定時間 td _·時間差 實施例 以下敘述本發明之原理及改進方法。 本發明之原理是將反射激光的強度對時間之積分 (integration)曲線除以微分（differential)曲線和時間（time)的 乘積(故稱爲I-DT轉換），反射激光強度經I-DT轉換後對 時間之關係，如方程式(3)所示： ΤΓΖ ~ ~ ^ Λπ—csc(^)~ ~(~ττ)2................(3) 1 / II---_-----^----^---,玎------d. {邻先閱讀背面之注意事項再填寫本頁) 其中 4mtd COSi^ 上述/該反射激光強度，ί爲時間，Α爲一次反射激 光的強度’ A爲二次反射激光的強度，；J爲該材料層的折 射率，d爲該材料層的摩度，‘爲該入射激光的波長， 本紙ί長尺度適川中國围家標々（ΓΝί5 ) Λ4規格（2丨0><297公着） 411391 3766twfl .d〇c/〇〇6 . _ A7 —B7 —----- .. _ 一-一 五、發明説明（/y ) 爲折射角。 方程式(3)中的4是常數項，等號右邊第二項的影響 αφ 會隨時間增加而減小’主要的變化是由第一項而來，故此 轉換曲線對時間具有餘割(cosecant)函數的性質，如第3圖 所示。而經過此轉換後的重要特性是： 1·極値點位置明顯，可解決習知方法不易判斷極値點 的困擾。因爲在極値點時微分項等於零時，函數値趨近無 窮大’故I-DT轉換曲線可形成明確的峰値點，可精確的 定出I-t曲線極値點的位置。例如可以輕易地在第3圖中 找到峰値點308 ’作爲”解析終點”304。由此Ι-DT轉換曲 線所得之解析終點304位置相當於i_t曲線的極値點302， 其準確性非常高。相較於習知方法，本發明無須如習知單 純I-t曲線還需採用觀察窗邏輯法以連續3〜4個觀察窗才 能由I-t曲線的解析終點310定義出一個轉折點，並且本 發明方法I-DT轉換曲線的峰値點3〇8與κ曲線極値點3〇2 相對應。 2.函數値本身即代表斜率之正負，可解決習知方法精 確度不佳的問題。因爲積分項是單調遞增函數，函數値的 正負是由微分項而決定，故函數値本身就代表了斜率的正 負，可快速而明確的判斷出曲線是上昇或下降。 3，可提供更多的訊號特徵點以分析短硏磨時間的製 程。對於硏磨時間特別短，只包含不到一個完整週期訊號 的製程，因爲在強度曲線上參υπ到多二細之間的斜率 9 川屮阄 ® 家桴彳（rNS )規格（2I0X297^J~5 ----— ----：-----^------ΐτ------Φ (誚先閲讀背面之注意事項再填寫本頁) A7 B7 411331 3766lwfl .doc/006 五、發明説明（s ) 是相同的，所以習知方法(I-t曲線)不能精確的定義出在這 段區間的點。而經過轉換過的曲線(〗-DT曲線）’在這段區 間有二:個峰點可供利用，如第4圖所顯示’其對於硏磨時 間特別短，只包含不到—個週期訊號的製程’可提供較多 的資訊。 4. 可提供較習知方法更寬廣分析區域。因爲在$ 到0=細">之間這段區域曲線是平坦的’可消除因不同晶 片或不同產品所造成的差異，故可解決現有方法因爲強度 振幅變化所造成的檢測成功率不確定性’亦使得不同產品 可能用同一種法則來定義”解析終點” ’同時也提供了分析 此區域厚度的可能性。 5. 抗雜訊能力比習知方法佳，因爲雜訊將會被積分項 吸收。而且可以解決不同轉盤間訊號位準落差問題。 由以上可知，本發明將反射激光的強度對時間之積分 曲線除以微分曲線和時間的乘積(故稱爲I-DT轉換），如此 得到之轉換曲線對時間具有餘割函數的性質’如第3圖所 示。由於轉換函數具有峰値明顯，函數値本身正負代表強 度曲線斜率及峰値間曲線平坦等特性，能提供一種明確而 穩定的法則來尋找”解析終點”(analytical endpoint)，因而可 徹底解決以往終點檢測所遭遇到的問題。並且’本發明不 只適用於介電層CMP，亦可適用於相同原理之光擧裝置ΰ 以下詳細敘述本發明較佳實施例之έ卩時檢測材料層:厚 度之方法與CMP的製程終點檢測之方法。 首先，參照第3圖，說明本發明較佳實施例之〜種即 I----------t----r----il------線 (誚先閱讀背面之注意事項再填寫本頁} 木紙張尺度递川中阄闽家榡彳（)八4規格{ 2丨〇〆297公釐） ίΐ# 部屮-^(iif/Ju 於*;5τί1·印：^ 411331 3766twfl.d〇c/006 A7 ________B7 五、發明説明（7) 時檢測材料層厚度之方法。此方法中，材料層厚度會隨時 間而漸減。並且在材料層厚度會隨時間而漸減時，以一入 射激光^射向材料層而取得一反射激光。將反射激光的強度 對時間之積分（f/汾）曲線，除以反射激光的強度對時間之 微分（$)曲線和時間⑺的乘積，得到一Ι-DT轉換曲線：其 中’ I-DT轉換曲線具有複數個峰値點3〇8, 312, 314, 316 , 並且這些峰値點分別對應材料層之複數個預測厚度。例 如，當爲正整數)時，代表轉換曲線上的複 數個峰値點，並且每一個峰値點所對應之材料層的預測厚 度4可表示爲：。當換曲線的—特 疋峰値點308發生時，特定峰値點308所對應的該材料層 之一特定預測厚度爲此時材料層的即時厚度。而，當μΓ)Τ 轉換曲線的特定峰値點308出現一特定時間後，材料 層的即時厚度相當於，特定峰値點所對應的材料層之特定 預測厚度減去特定時間t。％.與材料層的一即時厚度遞減速 率之乘積。上述即時厚度遞減速率相當於I_DT轉換曲線 的特定峰値點308與其前一同號峰値點316分別對應之材 料層之預測厚度的差除以特定峰値點308與其前一同號峰 値點316發生的時間差td。 再來’參照第3圖’說明本發明較佳實施例之一種cmp 的製程終點檢測之方法。當一材料層被由一初始厚度硏磨 成一終點厚度時’此方法可以檢測硏磨製程的終點。當材 料層被化學機械硏磨法硏磨時’同時以一入射激光射向材 .·· 11 ---- _· - .. a, ,,, ,___ M|| a mi b 本紙张尺庋滴川中固段家榲.今（（'NS ) M規格（210X297公釐） I---^------裝----_----訂------線 (对先閱讀背面之注意事項再填寫本頁) 411331 3766twl'l .doc/〇〇^ ^ _________B7__ 五、發明説明（/〇 ) ~~〜 料層而取得一反射激光。將反射激光的強度對時間之檳分 (fWi)曲線’除以反射激光的強度對時間之微分（尝）曲線和 時間（d的乘積，得到—I_DT轉換曲線，其中I-DT轉換曲 線具有複數個峰値點308，312，314，316，並且這些峰値黧占 分別對應材料層之複數個預測厚度。例如，當&焉 正整數)時，代表i—DT轉換曲線上的複數個峰値點，並且 每一個峰値點所對應之材料層的預測厚度必可表示爲. , mA0 cos(a ) <=—。從Ι-DT轉換曲線的複數個峰値點中選定 一特定峰値點308，其中此特定峰値點308所對應的材料 層之一特定預測厚度不小於且最接近材料層之終點厚度。 上述特疋峰値點的選定，係選定適當的正整數仿，使得材 料層之特定預測厚度心不小於且最接近材料層之終點厚 度。而S I-DT轉換曲線的特定峰値點308出現後，再經 一特定時間.的硏磨就停止硏磨製程，而獲得化學機械 硏磨的製程終點。上述特定時間相當於Ι-DT轉換曲線 的特定峰値點308所對應之材料層的特定預測厚度減去材 料層的終點厚度再除以一即時化學機械硏磨速率。而，上 述即時化學機械硏磨速率相當於I-DT轉換曲線的特定峰 値點308與其前一同號峰値點316分別對應之材料層之預 測厚度的差除以特定峰値點308與其前一同號峰値點316 發生的時間差td。若I-DT轉換曲線的特定峰値點所對應之 材料層的特定預測厚度等於材料層的終點厚度，則上述特 定時間爲零。也就是說’當I-DT轉換曲線的特定峰値點 12 本紙張尺度瑱圯中阐K家標泠(rNS ) A4im ( 2I0X 297/^¾ ) ~~ ---- (¾先聞讀背面之注意事唄再項寫本頁』 訂 411391 3760twll.doc/006 A7 _B7 五、發明説明（"） —-- 所對應之材料層的特定預測厚度等於松^ ~ 八何+4層的終點厚度 時’上述特定峰値點發生即爲化學機械硏磨的製程終點。 雖然本發明已以較佳實施例揭露如上，然其並非用以 限定本發明，任何熟習此技藝者，在不脫離本發明之精神 和範圍內，當可作各種之更動與潤飾，因此本發明之保護 範圍當視後附之申請專利範圍所界定者爲準。 ilhlfl· D n - n ϋ n n---I 丁 ——1 ί, ' n U5, ,-¾ ^ (《先背面之注悫事項再填寫本萸) 4?^.部十"ί?··:^^,-Η π;1'^ Ac 竹.^印t 木紙张尺度適圯屮闽囤家標今<(^)以规梠（21(^ 297公釐）Equation (1) shows the relationship between the reflected laser intensity and time (I_t curve) shows a cosine function. The actual graph is shown in Figure 2, and the It curve is related to the thickness of the dielectric layer. The thickness of the dielectric layer in the CMP honing and the end of the process are obtained. One method of endpoint detection is to first test a wafer to obtain a complete curve of the reflected laser intensity versus honing time (the honing time selected here should cover the entire possible process range). Then determine the thickness of the end point according to the process requirements, and find the curve position corresponding to the end point thickness on the I-t curve in Figure 2. Later, when honing subsequent wafers, as long as the I-t curve of the test wafer and the wafer being honing are compared, the end point of the honing process can be determined. In actual implementation, due to factors such as the distribution of the honing fluid and the difference between each wafer, the corresponding curve and amplitude of each wafer vary, and because the thickness of the dielectric layer at the beginning of each wafer is slightly different, Different, the curve will have different displacements on the time axis, so it is difficult to find the end point of the process by the above method. In addition, for different process requirements, there may be different end thicknesses. The paper size of the state-of-the-art in this paper (('NS) Λ4 specification (2) 0X297 mm) " ί I —------ 装- --- ^ ---- Order ------ line (for the notes on reading the back '& then fill out this page) 3 B7..-~ V. Description of the invention (彡) Correspondence on the curve The end position of the laser beam will also be different. 'There must be a set of practical rules to assist in the judgment. Because of the above problems,' it is impossible for us to determine the process end point solely by the intensity of the reflected laser light or a fixed slope position. Therefore ' Another solution is proposed. As shown in equation (2), 'the thickness of each pole on the curve is fixed', so the number of these poles can form a characteristic point or index to help find The position of the extreme point 202 closest to the process end point is usually called "analytical endpoint" (analytical endpoint) 204, because this is a position that can be clearly defined on the curve, and the actual process end point is not necessarily related to These "analytic end points" coincide when the two At the same time, a few seconds of constant time honing is used to reach the end of the process 206, as shown in Figure 2. In theory, these extreme points can be determined by the position where the slope of the curve is equal to zero, but due to the speed of the turntable on the process Limitation, the actual sampling frequency will not be too high, usually 1 time per second, so it is not easy to accurately calculate this way. The more practical method is to determine these turning points by the positive and negative transformation of the slope. The observation window logic method commonly used to determine the end of the analysis is to define the turning point of the curve by several consecutive slope changes. However, the shortcomings of this simplified method are poor accuracy and reproducibility. Because noise interference and the curve must be avoided The misjudgment caused by the change in slope occurs, so you must define a turning point with 3 to 4 consecutive observation windows (208) (Figure 2), so the accuracy is poor. And this method cannot guarantee the detection after all Success rate. Therefore, for a single dielectric layer or shallow trench isolation (shallow trench, this paper scales China 1¾¾: 揼 々 ((, ns)) λ4 specification (2 丨297 mm) J .--------- ^ ---- ^ --- ΐτ ------ ^ (Read the notes on the back first and then fill in the last page) 3 7661 wt ' l. doc / 006 A7 __B7 _ 5. Description of the invention (¥) 1Sioation (STI) For CMP process, the existing endpoint detection technology can only define the electrode closest to the actual process thickness by the wavelength of the incident laser. The location of the puppet point. And because of the variation of the material itself, the environment and the process, the problems currently encountered in mass production and use are: First, because the signal will vary with different materials, the test results cannot be guaranteed Success rate. Second, for different types of products, different conditions must be used to determine the "analysis end point" position. 3. For multi-turntable honing process, the problem of signal level difference between turntables will affect the reproducibility of signals. Fourth, it can not be applied to the honing time is particularly short, especially the process that contains less than one cycle signal. The reason for these problems is that it is impossible to accurately define the extreme points of the reflected laser intensity curve in the signal analysis technology, and it is impossible to quickly and clearly determine whether the slope of the curve is rising or falling. In view of this, an object of the present invention is to accurately define the extreme points of the reflected laser intensity curve, thereby detecting the thickness of the dielectric layer. Another object of the present invention is to quickly and clearly determine the reflected laser intensity. The slope of the curve is rising or falling, which improves the accuracy when detecting the thickness of the dielectric layer. Another object of the present invention is to overcome the above problems, so that the detection of signal endpoints can be applied to mass production processes. According to the above object of the present invention ', a method for detecting the thickness of a material layer in real time is proposed. The bottom of the material layer has a substrate, and the thickness of the material layer gradually decreases with time, and an incident laser light is incident on the material layer to obtain a reflected laser light. This method includes dividing the integral curve of the intensity of the reflected laser light with respect to time by the differential curve of the intensity of the reflected laser light with respect to time and time multiplied by 6. Paper size 1¾ / 1] 屮 K B2 house standard 彳 ((, N'S) now 栝(210x 297 mm) '------ L--Γ -------- install -------- order ------ line_pin first read the notes on the back before filling (This page) ___ B7 V. Product description (g) product to obtain an I-DT conversion curve, where the I-DT conversion curve has the property of cosecant function, so that the use of the cosecant function has obvious peak chirp characteristics, which can be used for That is, the thickness of the material layer is detected at I. According to the above object of the present invention, a method for detecting the end of a process by a chemical mechanical honing method is proposed. In this method, when a material layer is honed by a chemical mechanical honing method, an incident laser light is incident on the material layer to obtain a reflected laser light; and the integral curve of the intensity of the reflected laser light against time is divided by the intensity of the reflected laser light The product of the time differential curve and the product of time yields an I-DT conversion curve, where the I-DT conversion curve has the property of cosecant function, and the use of the cosecant function has obvious characteristics of peak chirp, which can be used to detect the material layer in real time Thickness, which can determine when the end of the process is reached. In order to make the above and other objects, features, and advantages of the present invention more comprehensible, a preferred embodiment is given below in conjunction with the accompanying drawings for detailed description as follows: Brief description of the drawings: FIG. 1 The principle diagram of the CMP optical endpoint detection of the dielectric layer is shown. Figure 2 shows the schematic diagram of the process end point determined by the reflected laser intensity versus time curve. Figure 3 shows the present invention dividing the integral curve of the reflected laser intensity versus time by the differential curve. A schematic diagram of the reflected laser intensity curve obtained by the product of time and time (hence the name I-DT conversion); and Figure 4 shows a schematic diagram of the process of analyzing the short honing time by applying the I-DT conversion of the present invention ° : 7 books (, NS) Λ4 specifications (2 丨 0〆297 mm) r --------- ^ ---- ^ --- 'ST ------, ^ (tab first (Read the notes on the back and fill in this page) υΨύ ^ {Φ) Τφ '\ Φ7 l Λ · ίΙ {αφ t αφ A7 B7 V. Description of the invention (202: Extreme point 2Q4: It curve analysis end point is 6: manufacturing process End point 208: Observation window 30: End point 304 M-DT conversion curve analysis end point 306: Process end point 308, 312, 314, 3 16: Peak point 31. It curve analysis end point: specific time td _ time difference Example The following describes the principle and improvement method of the present invention. The principle of the present invention is to divide the integration curve of the intensity of the reflected laser light against time. With the product of the differential curve and time (hence the name I-DT conversion), the relationship between reflected laser intensity and time after I-DT conversion is shown in equation (3): ΤΓZ ~ ~ ^ Λπ --Csc (^) ~ ~ (~ ττ) 2 ...... (3) 1 / II ---_----- ^ ---- ^- -, 玎 ------ d. {Please read the notes on the back before filling this page) where 4mtd COSi ^ above / the reflected laser intensity, ί is time, Α is the intensity of the reflected laser once, 'A is The intensity of the secondary reflected laser, J is the refractive index of the material layer, d is the friction of the material layer, 'is the wavelength of the incident laser light, the paper is a long-size paper suitable for the Chinese family standard (ΓΝί5) Λ4 specification (2 丨 0 > < 297) 411391 3766twfl .doc / 〇〇6. _ A7 —B7 —----- .. _ Explanation of the invention (/ y) is the refraction angle. The 4 in equation (3) is a constant term. The influence of the second term to the right of the equal sign αφ will decrease with time. The main change is from the first term, so the conversion curve has a cosecant to time. The nature of the function is shown in Figure 3. The important characteristics after this conversion are: 1. The location of the extreme point is obvious, which can solve the problem that the conventional method is difficult to judge the extreme point. Because at the extreme point the differential term is equal to zero, the function 値 approaches infinity ', so the I-DT conversion curve can form a clear peak 値 point, which can accurately determine the position of the extreme point of the I-t curve. For example, the peak point 308 'can be easily found in Fig. 3 as the "analysis end point" 304. The analytical end point 304 obtained from the 1-DT conversion curve is equivalent to the extreme point 302 of the i_t curve, and its accuracy is very high. Compared with the conventional method, the present invention does not need to use a simple It curve as in the conventional method, but also needs to use the observation window logic method for 3 to 4 consecutive observation windows in order to define a turning point from the analytical end point 310 of the It curve, and the method I- The peak point 308 of the DT conversion curve corresponds to the extreme point 302 of the κ curve. 2. Function 値 itself represents the positive and negative slope, which can solve the problem of poor accuracy of the conventional method. Because the integral term is a monotonically increasing function, the positive and negative of the function 値 is determined by the differential term, so the function 値 itself represents the positive and negative slopes, and it can quickly and clearly determine whether the curve is rising or falling. 3. It can provide more signal characteristic points to analyze the process of short honing time. For the honing time is very short, it only contains less than one complete cycle signal, because the slope on the intensity curve is between 9 and π. 9 Chuanxiong® Home Furnishing (rNS) specification (2I0X297 ^ J ~ 5 ----— ----: ----- ^ ------ ΐτ ------ Φ (诮 Please read the notes on the back before filling in this page) A7 B7 411331 3766lwfl .doc / 006 Fifth, the invention description (s) is the same, so the conventional method (It curve) cannot accurately define the point in this interval. The converted curve (〖-DT curve) 'is in this interval. There are two peak points available. As shown in Figure 4, 'It is particularly short for honing time and only contains less than one cycle of the process' can provide more information. 4. Can provide more knowledge The method has a wider analysis area. Because the area curve between $ and 0 = fine " > is flat, it can eliminate the differences caused by different chips or different products, so it can solve the existing method because of the intensity amplitude change. Caused by the uncertainty of the detection success rate 'also makes it possible for different products to use the same rule to define "analytic endpoints" 'It also provides the possibility to analyze the thickness of this area. 5. The anti-noise ability is better than the conventional method, because the noise will be absorbed by the integral term. And it can solve the problem of signal level difference between different turntables. From the above we can know In the present invention, the integral curve of the intensity of reflected laser light versus time is divided by the product of the differential curve and time (hence the name I-DT conversion). The conversion curve obtained in this way has the property of a cosecant function with respect to time. Since the conversion function has obvious peaks, the function 値 itself represents the slope of the intensity curve and the flatness of the curve between the peaks. It can provide a clear and stable rule to find the "analytical endpoint", so it can be completely solved. Problems encountered in the past endpoint detection. And 'The present invention is not only applicable to the dielectric layer CMP, but also to the light lifting device of the same principle. The following describes the material layer detection thickness in the preferred embodiment of the present invention in detail: Method and CMP process end point detection method. First, referring to FIG. 3, a description of the preferred embodiment of the present invention, which is I ------------ t ---- r ---- il ------ line (诮 Please read the precautions on the back before filling in this page} Wood paper scale delivery Sichuan Minjia 阄 () 8 4 specifications {2 丨 〇 〆297 mm) ίΐ # 部 屮-^ (iif / Ju Yu *; 5τί1 · India: ^ 411331 3766twfl.d〇c / 006 A7 ________B7 V. Method for detecting the thickness of the material layer in the description of the invention (7). This method The thickness of the material layer will gradually decrease with time, and when the thickness of the material layer will gradually decrease with time, an incident laser light will be incident on the material layer to obtain a reflected laser light. Divide the integral (f / fen) curve of the intensity of the reflected laser with time by the product of the differential ($) curve of the intensity of the reflected laser with time and the product of time ， to obtain a 1-DT conversion curve: where 'I-DT The curve has a plurality of peak points 308, 312, 314, 316, and these peak points correspond to a plurality of predicted thicknesses of the material layer, respectively. For example, when it is a positive integer), it represents a plurality of peak points on the conversion curve, and the predicted thickness 4 of the material layer corresponding to each peak point can be expressed as:. When the curve-specific peak peak point 308 occurs, the specific predicted thickness of one of the material layers corresponding to the specific peak peak point 308 is the instant thickness of the material layer at this time. However, when the specific peak point 308 of the μ) T conversion curve appears for a specific time, the instantaneous thickness of the material layer is equivalent to the specific predicted thickness of the material layer corresponding to the specific peak point minus the specific time t. %. Product of the instantaneous thickness deceleration rate of the material layer. The above-mentioned instant thickness deceleration rate is equivalent to the difference between the predicted thickness of the material layer corresponding to the specific peak point 308 of the I_DT conversion curve and its front peak point 316 divided by the specific peak peak point 308 and the front peak point 316. Time difference td. Next, referring to FIG. 3, a method for detecting a process endpoint of a cmp according to a preferred embodiment of the present invention will be described. When a material layer is honed from an initial thickness to an end thickness', this method can detect the end of the honing process. When the material layer is honed by the chemical mechanical honing method, 'at the same time, an incident laser is irradiated on the material ... 11 ---- _ ·-.. a, ,,,, ___ M || a mi b庋 Drip Chuan Zhonggu section home. Today (('NS) M size (210X297 mm) I --- ^ ------ installation ----_---- order ------ line (Please read the precautions on the back before filling in this page) 411331 3766twl'l.doc / 〇〇 ^^ _________B7__ 5. Description of the invention (/ 〇) ~~~ Material layer to obtain a reflected laser light. Compare the intensity of the reflected laser light to Divided by time (fWi) curve 'divided by the product of the differential (taste) curve of the intensity of the reflected laser and time and time (d), we get the —I_DT conversion curve, where the I-DT conversion curve has a plurality of peaks 308, 312, 314, 316, and these peaks account for a plurality of predicted thicknesses corresponding to the material layer respectively. For example, when & (positive integer), it represents a plurality of peaks on the i-DT conversion curve, and each The predicted thickness of the material layer corresponding to the peak point must be expressed as., MA0 cos (a) < = —. A specific peak point 308 is selected from a plurality of peak points on the 1-DT conversion curve, where this The specific predicted thickness of one of the material layers corresponding to the fixed peak point 308 is not less than and closest to the end thickness of the material layer. The selection of the special peak point is based on the selection of an appropriate positive integer imitation, so that the specific predicted thickness center of the material layer Not less than and closest to the end thickness of the material layer. After the specific peak honing point 308 of the SI-DT conversion curve appears, the honing process is stopped after a certain time. The process of chemical mechanical honing is obtained. End point. The above specific time is equivalent to the specific predicted thickness of the material layer corresponding to the specific peak point 308 of the 1-DT conversion curve minus the end thickness of the material layer and then divided by an instant chemical mechanical honing rate. The mechanical honing rate is equivalent to the difference between the predicted thickness of the material layer corresponding to the specific peak point 308 of the I-DT conversion curve and its peak number 316, divided by the specific peak point 308 and the peak number 316. The time difference td occurred. If the specific predicted thickness of the material layer corresponding to the specific peak value of the I-DT conversion curve is equal to the end thickness of the material layer, then the above specific time is zero. It means' when the specific peak point of the I-DT conversion curve is 12 paper scales, and the K family standard (rNS) A4im (2I0X 297 / ^ ¾) ~~ ---- (¾ first read the back of the Attention, please write this page again ”Order 411391 3760twll.doc / 006 A7 _B7 V. Description of the invention (") --- The specific predicted thickness of the corresponding material layer is equal to the end thickness of the loose ^ ~ Bahe +4 layer The occurrence of the above-mentioned specific peak honing point is the end of the chemical mechanical honing process. Although the present invention has been disclosed as above with a preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art can make various modifications and retouches without departing from the spirit and scope of the present invention. Therefore, the present invention The scope of protection shall be determined by the scope of the attached patent application. ilhlfl · D n-n ϋ n n --- I 丁 ——1 ί, 'n U5,, -¾ ^ (《Notes on the back of the document before filling in this note） 4? ^. 部 十 " ί? ··: ^^,-Η π; 1 '^ Ac bamboo. ^ 印 t Wood paper size is suitable for Fujian storehouses < (^) according to the rules (21 (^ 297 mm)