201101400 六、發明說明： 【發明所屬之技術領域】 本發明大體上係關於半導體裝置製造過程中之檢測及 方法工具’用來確保品質及增加良率。 【先前技術】 在微影半導體裝置製造技術中，步進器能準確地將半 〇 導體基板或晶圓上之倍縮光罩的影像聚焦是刻不容緩的。 當倍縮光罩的影像不在焦點上時，亦即一種已知之失焦（ defocus )狀態，所產生之半導體裝置之結構可能有不正 確的尺寸及樣式。例如，其結構邊緣可能較爲擴大及不明 顯、有圓的或是過度切割的表面，而取代了通常爲直線之 更理想的幾何結構。此失焦狀態通常導致問題半導體裝置 有不良的功能及/或無法運作。因此失焦測量是重要的方 法’用來允許半導體裝置製造者確保步進器持續地將倍縮 〇 光罩影像聚焦於晶圓上，因此能使製程的良率更好及更有 利基。 曝光缺陷是就半導體裝置製造而言的另一個常見問題 。當使光阻層曝光的程度超出可接受的光量範圍時，可能 會使將成形在半導體基板上的型態不正確。因此，當缺陷 存在時之缺陷辨識有同等的重要性。 除了針對曝光或失焦缺陷來檢測基板或晶圓之外，重 • 要的是針對與製程或材料相關之缺陷的基板及晶圓的檢測 " ，通常將這種缺陷稱爲“巨觀”缺陷。巨觀缺陷通常的定義 -5- 201101400 是出現在基板上之具有尺寸約爲〇.5u至10u之碎屑、裂 . 痕、刮痕、疊層剝離 '及/或微粒。這種缺陷容易引起半 導體裝置的失效而且顯著地降低這種裝置之良率。注意： 巨觀缺陷的尺寸可以較上述之尺寸範圍高或低，上述之尺 寸範圍僅定義此種缺陷的標稱大小。 傳統上，已經使用指定的檢測系統來檢測巨觀缺陷， 這種檢測系統尙未能夠立即及可靠地辨識曝光缺陷或是失 焦缺陷的存在。辨識曝光缺陷或是失焦缺陷通常是使用光 學臨界尺度（optical critical dimension，OCD)技術於多 種精密量測工具如橢圓偏振儀（e 11 i p s 〇 m e t e r )、反射儀 (refl ectometer )、散射儀（SCatterometer)的其中任一 種。理想的情況是使用相同的光學系統來結合辨識曝光及 _ 失焦缺陷的存在與針對巨觀缺陷檢測基板兩種功能。 【發明內容】 用來辨識基板上缺陷的檢測系統之實施例包含光源， u 光源將光線引導至將檢測的基板上。第一偏振濾光片或偏 振板定位在光源和基板之間。第二偏振濾光片或分析器定 位在基板與光學感測器之間，光學感測器接收反射自該基 板的光線。偏振板及分析器彼此成一角度配置，使得光學 感測器所擷取之影像的影像強度與受檢測基板上之與偏振 有關之缺陷的存在至少有部分相關。與偏振有關之缺陷包 含失焦（defocus )及曝光缺陷。亦也可辨識主要尺度約 -爲入射光束之波長或更小之非失焦及曝光缺陷之缺陷。 -6- 201101400 光源可以是任一種有用的型態，包含（但非限定於此 )寬頻白熾光、或雷射。這些光源的其中任一種可爲閃光 而且可以被定位在包含入射法線角度之任何有用的入射角 度，用以將光線引導在基板表面。雷射可以是固定的、不 同的單色光或者可以經配置用以於多個不同標稱波長下輸 出光線。 當使用閃光照明時，閃光明滅的順序至少與基板相對 0 於檢測系統移動的速度有部分相關。此使得檢測系統可以 可靠地在適當的位置擷取基板影像。 光學感測器或是影像器可以爲單色光的充電式電容性 裝置（charged capacitance device, CCD )。在一些實例 中，光學感測器可以是Bayer-型態的彩色影像器或者是 3 -晶片設計。在另外的實例中，一個或更多個光源及/或 彩色濾光片可以被用來與單色光學感測器共同使用，以便 自基板獲取色彩資料。可使用面掃描及線掃描兩者。 〇 除了失焦及曝光缺陷之外，也可以辨識它種類型的缺 陷。這些他種類型之缺陷可以包含：凹洞、孔隙、碎屑、 裂痕、微粒、及刮痕。 依照本發明之檢測系統其運作係藉由配置光源將光線 引導至基板上。第一偏振濾光片定位在光源與基板之間， 而且置放光學感測器以接收反射自該基板的光線。第二偏 振濾光片置於基板與光學感測器之間，所以第一及第二偏 - 振濾光片互相呈經選擇之相對角度。檢測系統則被用來擷 取基板的影像，而且自這些影像產生比較性資料，用以辨 201101400 識基板上之曝光及/或失焦缺陷的存在（如果有的話）。 配置偏振濾光片以擷取所需之影像會涉及到將第一及第二 偏振濾光片一齊轉動至理想的檢測角度，而同時維持兩者 間之經選擇的相對角度。 取得比較性資料係首先針對各個所擷取之影像產生差 異性影像，而後再就整個差異性影像求取個別差異性影像 的像素強度差異的平均，以取得針對各個差異性影像的平 均影像強度。針對預定的臨限値來評估各個所擷取影像之 平均影像強度，基板上若有曝光及失焦缺陷則用以判斷基 板上曝光及失焦缺陷之至少其中之一的存在。 就基板中既知之曝光及失焦缺陷之至少其中之一的缺 陷程度來校正光學感測器之輸出，此校正係被用來判斷曝 光及失焦缺陷的適當程度。在一實施例中，校正涉及到擷 取校正基板之複數個影像，其中各個影像已蒙受既知程度 之失焦及曝光缺陷。如上述，差異性影像之產生係針對各 個經擷取之影像，在整個差異性影像求取差異性影像的像 素強度差異的平均，以獲得平均影像強度。針對各個具有 既知程度之失焦及曝光缺陷之經擷取之影像，記錄平均影 像強度値。使用者可以選擇任一所記錄之平均影像強度値 ，其將特定程度或大小的失焦及/或曝光缺陷値標示爲臨 限値；或者可以內插於這種記錄値之間；或者簡單地使用 記錄値作爲起始點，並將針對產品之特定的修正値加於起 始點。完全讓檢測系統的使用者定義失焦及/或曝光缺陷 之適當的臨限値。 -8 - 201101400 • 產生差異性影像會涉及到以一個像素接著一個像素爲 基礎來將複數個經擷取影像平均，用以求取平均後的影像 。再從各個所擷取之影像去除平均後的影像，並以一個像 素接著一個像素爲基礎產出差異性影像，這些差異性影像 可以被視爲像素強度値之陣列、或者爲像素強度値之差異 的陣列。 針對失焦及/或曝光缺陷之基板檢測可以與針對其他 0 缺陷例如凹洞、孔隙、碎屑、裂痕、微粒、及刮痕等檢測 同時進行。或者，可以相繼進行針對這些個別缺陷型態的 檢測、或者甚至以時間偏移的方式進行，亦即以明顯彼此 分隔之時間來進行。 本發明之另一實施例，影像分析技術，如空間圖案辨 識（spatial pattern recognition，SPR)可以被用來分析差 異性影像以辨識基板上之層邊界（boundaries of layers) 。注意，如前述般之層邊界可以是基板部分的部分材料層 〇 ，或者是相關於爲基板無關部分之殘餘物的材料層，亦即 該材料層可以爲一種型態的污染物或另一種型態的污染物 【實施方式】 後文中將詳細說明本發明，並將參照構成本發明之一 部分的圖式’圖式以圖示方式呈現可實施本發明之實施例 。在圖式中’相同的數字代表實質上爲由多種視角呈現的 相同成分。以下將詳細說明這些實施例，使得精熟此技術 -9 - 201101400 的人士得以能實施本發明。可以使用其他的實施例， 在不背離本發明之範疇下進行結構上、邏輯上、及電 的變化。因此，以下的詳細說明不在於限制，而且本 之範疇係以申請專利範圍及其同等物來加以界定。 本發明涉及一種方法及設備，藉由測量反射自基 面的偏振光線的改變，來判斷在半導體基板中曝光及 微影缺陷的存在。爲了簡化以下的討論，「失焦 defocus ) —辭在本文將代表曝光及失焦缺陷兩者， 一個給定的基板可能有一種或是另一種缺陷、或是兩 有。再者，「失焦」一辭將廣泛的包括在檢測下之基 任一缺陷或是基板之不理想的部份，其特性將同樣導 光及/或失焦缺陷，而且可以被本發明之檢測系統辨 者特徵化。雖然精熟此領域的人士知道失焦缺陷的其 點也會影響這種偏振改變的本質及程度，但一般而言 焦缺陷是與偏振有關的特性，亦即，失焦缺陷將引起 光線在偏振狀態的改變。 此處所使用的「基板」一詞將用來包含可以以本 之檢測系統所檢視之任一材料或結構。特別是，「基 將包含任何構造、型態或材料之半導體晶圓，包含（ 只限定於）整個晶圓、未經圖案化之晶圓、圖案化之 、部分圖案化之晶圓、全部或部分圖案化之破損晶圓 經圖案化之破損晶圓、以任何形式或是在任何支撐機 之已經被切割之晶圓’包含膜層框架（film frames JEDEC 托架（JEDEC trays ) 、Auer 船形容器（ 可以 性上 發明 板表 失焦 j ( 雖然 者皆 板之 致曝 識或 他觀 ，失 反射 發明 板」 但非 晶圓 '未 構上 ),、 Auer •10- 201101400 • boats)、凝膠（gel)或疊片包裝（waffle packs)中之晶 粒、通常稱爲mcm之多晶片模組等等。基板一詞及晶圓 一詞在本文中可以交替使用。 「巨觀缺陷」一詞在此處將包含所有不希望出現在基 板上之實際上與偏振無關之特徵。如上述般，巨觀缺陷在 傳統上被稱之爲凹洞、孔隙、碎屑、裂痕、微粒、及刮痕 等等。注意，在一些情形下，巨觀缺陷的尺寸可以接近用 0 於檢測之入射光線的波長。在這些情形下，巨觀缺陷可能 會影響在其上反射之光線的偏振狀態。 參照圖1，成像系統8之一實施例包含照明器1 0、偏 振片12、分析器14、及光學感測器1 6。照明器1 0引導 光線沿著光學路徑P至偏振片1 2上，偏振片1 2實際上僅 穿透具有預定偏振角度的光線。經由偏振片12穿透的光 線而後入射在基板S上。在一實施例中，基板S是矽晶圓 或是部分矽晶圓，其具有形成於其上之結構。在一些實施 Q 例中，這些結構形成在基板s上之一種或更多種半導體裝 置。基板s上亦可以形成其他的機構或是結構。 如圖1中所示，光學路徑P位於相對於基板S之非法 線入射角之處。在一些實施例中，可以裝設照明器1 0、 偏振片1 2、分析器1 4、及光學感測器1 6及其他相關的光 學元件如物鏡等等，以便調整基板S上的光線入射角。調 整系統8中光線之入射角度的裝置機構型態已爲精熟此領 - 域的人士所知’此種裝置機構可包含裝設板，系統8之光 ·-學元件裝設於裝設板上，裝設板可以藉由轉動機構轉動， -11 - 201101400 轉動機構可以是一個或更多個致動器。入射角度可以是固 定的（如圖所示）或是針對各種產品裝設而調整。再者， 在一些實施例中，在檢測期間可以視需要調整入射角度。 照明器1 〇可以爲任何一種有用的型態，包含寬頻白 光、具有固定波長輸出之雷射、用以輸出多個波長之雷射 、或是複數個用以沿著光學路徑引導光線之雷射。照明器 所需之強度可以視系統8所導向之應用而定。在一些應用 中，需要高強度的照明，相反的，在另一些應用中需要的 是相對而言較低的強度。照明器1 〇可以用來提供實際上 爲保持一定的輸出、或者提供閃光以便凍結系統8內基板 S的運動，而可以快速擷取基板S的影像。 入射在基板S上的光線從基板反射，而且此經反射的 光線入射在分析器14，分析器1 4爲與偏振片1 2類似的 偏振光學元件，偏振片1 2僅能使具有預定偏振角度的光 線通過。通過分析器14的光線入射在光學感測器16上， 光學感測器1 6擷取基板S的影像。雖然可以使用任一種 可以灰階或是以色彩產生像素強度値之兩維陣列之裝置’ 如線掃描、或是時間延遲積分成像（time delay integration, TDI )裝置、或CMOS光學感測器陣列，然而 在一實施例中，光學感測器1 6爲二維電子式光學感測器 ，如電荷耦合裝置（CCD )。在一實施例中，光學感測器 1 6爲單色光光學感測器，其中光學感測器之2D像素陣列 的各像素記錄0-256的灰階値，這些像素共同表現基板S 之影像。當使用單色光光學感測器時，一個或更多個彩色 -12- 201101400 . 濾光片1 8可以定位在介於照明器1 〇與光學感測器1 6之 間的光學路徑S中，以便只能使對應於彩色濾光片之波長 範圍內之光線通過。在另一實施例中，光學感測器可以爲 Bayer類型、或3 -晶片類型之彩色光學感測器，這些感 測器具有個別的光學感測器，各自用於不同的色彩，例如 一種感測器用於紅光、一種感測器用於藍光以及一種用於 綠光。 0 精熟此技術的人士將了解到：上述說明之系統8的基 本元件將與其他光學元件共同使用，或者不與其他光學元 件共同使用，且基本元件包含但非侷限於光學濾光片、透 鏡、鏡子、延遲器（retarder)及調變器（modulator)。 經調整後可用以實現本發明的一種檢測系統爲Rudolph Technologies Inc. of Flanders，New Jersey 所出品，商品 名爲WaferViewTM。再者，應理解的是，系統8可用以進 行多項功能，這些功能可以在同一時間或是間隔一些時間 Q 進行。例如，系統8可適於用以進行巨觀缺陷的檢測以及 用於失焦缺陷的檢測。再者，系統8可用以進行巨觀缺陷 的檢測’接著進行用於失焦缺陷的檢測，或者可以同時進 行兩種檢測。 彩色濾光片18可以用於系統8中，如圖1中之示意 圖。一個或更多個彩色濾光片18可以被置放在介於偏振 片12與基板S之間 '介於基板S與分析器14之間、介於 - 照明器1 〇與偏振片1 2之間、或者介於分析器1 4與光學 - 感測器1 6之間。在一實施例中，彩色濾光片1 8可以爲本 -13- 201101400 技術中之既知型式之濾光輪（filter wheel )，其中一彩色 濾光片群組的其中之一個彩色濾光片被固定在位於光學路 徑P中之轉動輪，因此可以選擇性地將彩色濾光片1 8橫 跨光學路徑P而定位。在另一實施例中，可移除之濾光片 架可位於光學路徑P中，允許不同的彩色濾光片位於光學 路徑P中。在另一實施例中，固定的彩色濾光片可以裝設 在光學路徑P中。應理解的是：可以使用任一濾光片媒介 或是機構作爲彩色濾光片’只要其適合用於選擇性地通過 預定波長或波長範圍。 在一些實施例中，理想上可以就預定之彩色頻道（ color channel)分離光學感測器16的輸出，其中的「彩 色頻道」之定義爲預定之波長或波長範圍。如上述所建議 的，彩色頻道的分離可以使用彩色濾光片、使用具有直接 可分辨個別彩色頻道的彩色光學感測器’如3 -晶片光學 感測器及Bayer光學感測器、或是使用在預先選定之波長 範圍內輸出光線的照明器1 〇。應知的是：一些基板S對 於某些波長或色彩範圍可以部分或全部的穿透。僅以一例 而言，某個基板可以穿透或者破壞性干涉入射在基板上之 具有約以4 7 5 nm爲中心之波長之所有入射藍光的其中大 多數光線，但是反射大部分具有約以700 nm爲中心之波 長的紅光。在此例中，有用的是可以利用光學感測器16 輸出訊號，此輸出訊號係源於入射在光學感測器1 6上的 紅光。與個別彩色頻道相關的資料使用將取決於系統8將 檢驗何種特徵。在一些實施例中，某些半導體基板，亦即 -14- 201101400 產品，其特徵將傾向於以既知方式反射光線，因此可以特 別針對某些產品裝設檢測系統8，用以最佳化產品之檢測 〇 圖2所示爲本發明之另一實施例，其中檢測系統3 〇 以一般的方式配置，具有照明器3 2，引導光線沿著光學 路徑P通過偏振片36、濾光片40(選擇性的）、及分光 器42至基板上。在路徑P上反射自基板S之光線由分光 0 器42所引導，經過濾光片40 (選擇性的）及分析器3 8 至光學感測器34。除了在分光器的存在及入射角度的差 異之外’系統8及3 0實質上爲相似。在此實施例中，光 學路徑P實質上垂直於基板S。 ' 已被觀察到的是，如同於失焦缺陷將改變形成於基板 S上之結構幾何般，失焦缺陷亦將改變基板S的反射率。 改變基板S反射率之其他因素爲其他膜層的性質及入射在 基板上之光線的波長、偏振、及入射角度。使用依照本發 Q 明之不同實施例的檢測系統，如系統8或3 0，可以區辨 失焦缺陷所導致之反射率的改變，而且可以快速及可靠的 區辨。 一般而言，來自照明器1 0及3 2的光線經偏振片1 2 及36偏振成一預定角度P並以特定角度β入射在基板上 σ 反射時’基板S將改變入射光線的偏振狀態，偏振狀 •態與數個特徵相關，特別是失焦。經過偏振的變化，可以 取得關於基板S中之失焦缺陷的資訊。經反射之光線通過 -15- 201101400 分析器14 ’並入射在光學感測器丨6上。分析器14之配 置（後文將有詳細說明）有助於確保取自於光學感測器 1 6之資料包含與反射光之振幅及偏振變化兩者有關的資 訊，而且特別是關於基板S上之失焦缺陷存在的資訊。 在一實施例中，入射在基板S上之光線由偏振片1 2 及3 6線性偏振，此通常代表對於檢測問題的最簡單的解 答。在另一些實施例中，入射光以橢圓方式受偏振或者是 以圓形方式受偏振，視需要而定。藉由將分析器1 4及3 8 設定在相對於偏振片1 2及3 6之角度A，可以調變到達光 學感測器16，34的光線。介於偏振片12、36與分析器 14、38之間的角度被給定爲P-A。 現在參照圖3，一般而言，當光線反射自基板S時， 入射光的一部分Ep，以不同於其他部分的入射光的方式 被反射。入射光的一部分Ep反射自基板S表面上之與偏 振無關的部分，在其偏振上沒有任何顯著的改變，如圖3 中所示之E!及E2。可以在半導體基板S上發現的這種特 徵之一些例子包含（但非限定在）基板中之看起來是缺陷 的亮及暗，如碎屑、裂痕、刮痕、凹洞、孔隙、及微粒。 經反射之入射光線的另一部分E3反射自形成於基板上之 與偏振有關的特徵，這些特徵將會改變入射光的極性。在 半導體基板S上發現的一些與偏振有關之特徵或結構的例 子包含（但非限定在）：線性結構、導體、連接線、通孔 及緯線（street )。還有另一部分的入射光反射自基板S 表面之結構或特徵，這些改變反射光線之偏振的結構或特 -16 - 201101400 • 徵係爲蒙受失焦缺陷。此光線E4偏振狀態與E3不同。反 射光E4的結構可以與上述反射光線&之標稱結構相似或 相同，除非它們蒙受失焦缺陷，失焦缺陷的程度將影響光 線E4的強度結構遭受到失焦缺陷的例子示於圖8的測試 結構74。 圖8爲部分基板之示意圖。在此例中，一個半導體晶 圓W’具有數個形成於其上之積體電路裝置（ICdevice) 0 70。積體電路裝置70以網格方式排列，網格之間具有間 隔，俗稱爲街道72。數個測試結構74形成於街道72中 。在各個1C裝置70上，也有形成於其上的結構76，結 構76構成1C裝置70的主動電路的一部分，而且遭受到 失焦缺陷。應注意的是，在某些案例裡，整個1C裝置70 、或至少是1C裝置70的大部份，可能有一個或更多個結 構76遭受到失焦缺陷。通常以使測試結構74及結構76 兩者相似的方式將其形成。例如，作爲1C裝置70的主動 Q 電路的一部分的結構76，如果有一系列的線性結構形成 於其中時，測試結構74將會以相同方式被形成，以致於 測試結構74的特性將可指示結構76的特性。 在一些實施例中，測試結構74包括一系列的格狀物 78，各個格狀物具有形成於其內的週期性結構。圖9爲具 有示範性質的測試結構74的照片，此測試結構74具有數 個不相連區域78。各個不相連區域78具有就名義上來說 - 是不同的臨界尺度，意即，具有特定的間距或是臨界尺度 的周期性結構。注意：測試結構74的淨效應是用來提供 -17- 201101400 的 結 影 驟 測 路 而 ( 度 指 反 均 相 以 獲 有 7 6 變 尺 取 之 値 用 某些訊息，而這些訊息通常是只由各個產品晶圓上 FEM晶圓所提供的。因爲在某些臨界尺度上的周期性 構將對於從該周期性結構反射而來的光的偏振狀態產生 響，測試結構74對於評估微影製程中執行圖案形成步 的步進器（stepper)的運作變化將有所助益。再者，當 試結構74之不相連區域78與位於1C裝置70的主動電 中的結構7 6足夠相似，且假設存在著額外的製程變異 可能影響入射光的偏振狀態改變時，例如：構成基板W 1C裝置70形成於其上）之疊層的材料性質或是相對厚 ，吾人可以獲得對於結構76之臨界尺度數値的一般性 示。 如同入射於基板表面的光的偏振狀態改變可能導致 射光強度的變化般，在一些例子裡，某個區域7 8的平 灰階値可以直接且毫無模擬兩可地與該區域的臨界尺度 關聯。根據該結構76的平均灰階値，這種相關性就可 被用來辨識結構76的臨界尺度。雖然，以這種方式所 致的區域76的臨界尺度値可能不是非常精確，利用具 代表多重臨界尺度的測試結構74來評估基板上的結構 的臨界尺度，對於取得1C裝置70中的臨界尺度的任何 化大小的訊息則是有用的。例如，除了「絕對的」臨界 度或是替代評估「絕對的」臨界尺度之外，利用測量選 之測試結構76 (或是不相連區域78 )的灰階値與接續 受檢測或被成像之結構76 (或是測試結構74 )的灰階 之間的差異大小，對於臨界尺度有指示性的灰階訊息可 -18- 201101400 來辨識臨界尺度的變化大小。 參照圖4可發現，對於沒有通過分析器14、3 8的光 線而言，在反射光E 3及E 4強度之間沒有對比。可以理解 的是，在這種情況下辨識失焦缺陷是困難的。然而，一旦 反射光E!、E2、E3、及E4通過經過適當配置的分析器I4 、38取得光訊號E’1、E’2、E’3、及E’4，光訊號E’3及 E ’ 4間的對比程度足夠取得關於失焦缺陷存在的有用資訊 0 。見圖5。 在一實施例中，在偏振片1 2與分析器1 4之間的角度 P-A係以實驗方式決定。現在參照圖6，利用定位在系統 8內之測試基板S以作爲檢測（步驟5 0 )，照明器10 ( 在一實施例中爲閃光照明器）被設定在預定之照明水準（ 步驟52)，照明器1〇接近其最高亮度輸出爲較佳，但可 爲任一適合的亮度。來自照明器或光源10之光Ep經由偏 振片12被引導至基板S上方。接著，使偏振片12設定在 〇 角度P (步驟54)。在一實施例中，偏振片12之角度被 導向於與基板上存在之任一線性結構實質上垂直。將被理 解的是：當基板S是半導體裝置形成於其上之半導體晶圓 時（在任何完成狀態下），這種結構通常但非總是具有顯 著的線性結構。反射光（E!、E2、E3、及E4 )通過分析器 14 ’用以取得在光學感測器16上之光學訊號、E’2、 E’3、及E’4。在一些情況下，基板S上沒有可供區辨線性 - 結構之方位，或者沒有形成於基板S上的線性結構，可針 ' 對偏振片1 2選擇任意角度P。 -19- 201101400 分析器1 4接下來轉動至角a (步驟56 )，以致於有 足夠的照明可以到達光學感測器1 6，以允許針對巨觀缺 陷或是與偏振無關之缺陷來檢測基板S，如美國專利字號 6324298、6487307、及6826298，這些專利與本專利申請 共同被持有’且爲供參考而將其納入。注意：當照明不止 針對缺陷檢測基板S，亦利用良好品質檢測之訊號對噪音 的比値來滿足系統8的終端使用者，於檢驗中不會發生假 的正値及錯過的缺陷等顯著的錯誤時，分析器1 4將被視 爲位在正確的角度A。系統8之訊號對噪音的比値係以已 知方式分析光學感測器1 6之輸出而決定。 一旦偏振片及分析器的定位角度P及A爲已知時， 偏振片及分析器經由一連串的檢測角度一起被轉動（步驟 58 )，偏振片與分析器（P-A )之間的相對角度保持爲實 質上固定，轉動至相對於基板S之理想的角度位置，此角 度位置將提供光線訊號Eh及E’4間如上述般之足夠的對 比。在轉動偏振片12及分析器14的過程中，記錄光學感 測器1 6上之入射光的強度。針對各個偏振片1 2及分析器 1 4所轉動之檢測角或檢測位置記錄在光學感測器1 6之光 強度。在一實施例中，以或多或少爲連續的方式轉動偏振 片1 2及分析器1 4，而且以少量增加偏振片1 2及分析器 14的轉動，記錄偏振片12及分析器14的位置及存在於 光學感測器1 6之光強度。分析偏振片1 2及分析器1 4轉 動期間所獲得的資料’就反射光線E ’ 3及E ’ 4之間的對比 用以辨識偏振片與分析器的最佳檢測角或檢測位置。 -20- 201101400 • 針對角度P及A之最佳配置的辨識過程可以是手動 的，其中系統8的使用者將偏振片12及分析器14轉動經 選取之角度範圍，而光學感測器1 6記錄影像資料，影像 資料由適當型態的電腦C進行處理，用以決定最適角度 P-A。或者’及較佳的是，將偏振片12及分析器14自動 化，使前述之電腦可以控制它們的轉動，同時在不同角度 P-A處記錄來自於光學感測器1 6之資料。偏振片與分析 ^ 器之自動化已爲精熟此技術的人士所熟知。如上述所建議 般，判斷偏振片與分析器間之最適角度P-A的過程可能需 要在以下將說明之步驟前和後進行多次重複（iteration ) 。例如，一旦完成全程的校正/設定步驟，額外執行多次 ' 的全程校正/設定步驟是有用的，用以判斷最後的系統設 定爲最佳化。 一旦如圖6所說明般已經適當地設定好系統8，可以 進行針對失焦缺陷的檢測、或是有需要的話，進行針對其 Q 他缺陷之檢測。然而，首先必須校正系統8。使用聚焦曝 光矩陣（focus exposure matrix, FEM)晶圓進行校正。 FEM是一種基板，在其上方有多數個已經成形的圖案或 結構，每一個都有不同的聚焦位置及曝光時間。FEM通 常用於半導體裝置之製造用以作爲校正光學微影（ photolithography)工具的一部分。FEM將形成於基板S 上之圖案及結構之結構性變化具體化，這些變化係源自於 . 焦點位置及曝光的改變。在基板S的檢測期間，得自於 FEM之失焦及曝光資料被用來作爲比較資料（comparator 201101400 )。注意：形成於FEM上之圖案可以不同於形成於被檢 測之基板S上之圖案，但以相同爲較佳。 用於校正目的之失焦資料的取得方法與用於檢測目的 之失焦資料取得方法相同，將說明用於校正目的之失焦資 料的取得方法以作爲檢測程序的一部分。將在適當之處標 示校正與檢測程序間的差異。 在檢測期間’取得將被檢測之型態的基板S，並置於 晶圓支持台或頂部平板上（圖未示出），晶圓支持台或頂 部平板上以習知方式使基板相對於檢測系統8之光學裝置 (optics )移動。在一些實施例中，基板s (產品或FEM )一個一個被檢測。在一實施例中，其中基板S是晶圓， 半導體裝置形成於晶圓上，以晶粒層次爲基礎進行基板S 之檢測’亦即，以感測器1 6使基板S上個別晶粒之影像 成像’而且以下述方式處理這些影像。在其他實施例中， 以視野爲基礎進行檢測。系統8之光學裝置經配置用以擷 取視野影像，這些影像的尺寸可不同於個別晶粒或者個別 的步進器攝像的尺寸。當系統8之視野小於個別晶粒時， 可將多個視野串接（stitch )起來形成個別晶粒的複合式 影像（composite images)。相同的串接技術應該用來形 成整個步進器攝影的複合影像。應理解的是，形成複合式 影像的串接影像技術已爲此技術領域所既知之技術。 在其他實施例中’當視野大於個別晶粒或者大於步進 器攝像時，可以裁去所形成之影像以顯示一個或更多個晶 粒或步進器攝像。通常有用的是’不要裁去過大影像’以 -22- 201101400 • 便包含得自於個別步進器攝像的多個晶粒，因爲由第一步 進器攝像所產生的晶粒可能是可被接受的，而那些由第二 步進器攝像所產生的晶粒可能是有缺陷的。裁去影像的技 術爲精熟此項技術的人士所既知。 在其他實施例中，首先擷取整個基板S之影像來完成 基板S之檢測。當基板S相對而言爲小時，可以使用系統 8以面掃描的運作原則來擷取。當基板S大於面掃描系統 0 8的視野時，可以取得基板S的多個影像並且將其如上述 般串接起來。串接可以與線掃描合倂進行，亦可與面掃描 檢測系統8合倂進行，精熟此項技術的人士將能立即了解 〇 ' 一旦如上述般取得適當的偏振片/分析器角度P-A， 基板S上個別晶粒之影像由光學感測器1 6 (步驟60 )擷 取。見圖7。以下說明之校正及檢測程序將以一個晶粒接 著一個晶粒爲基礎進行，雖然應理解的是，亦可以以其他 Q 基礎進行。使用者可以選取成像之基板s上之晶粒，使用 者判斷晶粒確實沒有如碎屑、裂痕、凹洞、顏色變異、微 粒等等缺陷用以形成模型。這種判斷完全取決於系統8的 使用者，而且可以有極大的變異，取決於基板S上之性質 及產品之使用方式。例如，使用者檢測的基板S具有形成 於其上方的半導體裝置，此半導體裝置想被作爲心律調整 器而使用，使用者對於用於產生模型之晶粒上的缺陷數目 • 會有非常嚴峻的標準。相反的，使用者檢測的基板S具有 " 相同的形成於其上方的半導體裝置，想要用於在廉價及可 -23- 201101400 棄式的消費者產品，使用者很可能願意接受有較多數目的 缺陷之晶粒作爲模型之用。簡言之，針對產生模型的目的 而言，一·個「好」的晶粒該如何疋義’取決於系統8之使 用者的判斷。針對產生模型的目的而言’雖然想像到的是 取得基板S上所有好晶粒的影像，然而通常只能得到就統 計上而言有意義數量的好晶粒；一般而言此數量小於整體 好晶粒的數量，而且在約爲10至15。如果取樣統計上有 意義之樣本，則最少應該選取具有不多於隨機出現之缺陷 的晶粒，因爲少數的、隨機出現的缺陷不太會對於檢測有 顯著的影響，應理解的是，多量的、非隨機缺陷將更有可 能扭曲檢測程序。 有用的模型亦可以使用自動方法來取得。例如’控制 系統8的電腦C可以以隨機方式選取統計上爲有意義數量 的晶粒數目而且可以擷取影像。這些影像被用來構成模型 ，用此模型來檢測構成模型的個別影像。當選取的晶粒在 使用者所選擇的標準下是有缺陷時，由另外以隨機方式選 取的晶粒來置換缺陷影像。將重複此程序多次，直到產生 一個適合的模型爲止。注意，以手動方式或是自動方式所 產生的模型會保持在靜態，亦即不會隨時間而改變，應隨 檢測的進行，經過一段時間後把新的、好的晶粒加入模型 以修改模型。 因爲「模型」一詞對於精熟此項技術的不同人士來說 可以有不同的意義，所以應澄清此處所使用的辭彙；「重 要晶粒」（golden die)或是「重要參照物」被用來描述 -24- 201101400 • 一種影像，將數個晶粒所對應的像素値加總，並將些數値 的平均可以求得影像構成像素之強度値。所以’ 「重要晶 粒」只是經平均後的晶粒影像。「模型」一詞較「重要晶 粒」或是「重要參照物」更廣泛，在一些情形下，將不包 含或使用重要晶粒或是重要參照物資訊。 重要晶粒使用於缺陷檢測的一個實施例（步驟62 ) 。同樣的’重要晶粒可至少構成使用於巨觀缺陷檢測之模 0 型的基礎的一部分。然而，通常針對影像中之各個像素定 義像素強度臨限値使用於巨觀缺陷檢測中之模型的移動超 越單純的重要晶粒。在巨觀缺陷的檢測中，若評估時，像 素強度値被發現落在由臨限値界定的範圍之外時，這些像 ' 素被視爲是有缺陷的。這些臨限値本身可以是單純的從重 要晶粒計算得到的標準差，但是更常的是包含將不同型態 的基板、變化及特徵列入考慮的加權値，而且使用者可以 定義將施加於基板上所形成之產品的標準。應理解的是， 〇 用於基板檢測之模型可以以無數種方式形成，而且在任何 方式中可以或不必根據重要晶粒，對於巨觀缺陷檢測模型 的唯一要求是：最後檢測得出的結果對於系統8的使用者 而言是令他滿意的。當進行巨觀缺陷檢測時，取得用於這 種檢測的適當模型（步驟6 4 )的時間可以與重要晶粒產 生的時間或多或少相同。如箭號65所標示，在一些情況 下模型的形成係利用重要晶粒之資訊。一旦模型產生，接 • 著比較模型與擷取影像，用以辨識缺陷（步驟72 )。 " 在針對失焦缺陷之檢測期間，前述步驟中所取得之重 -25- 201101400 要晶粒影像被使用來移除擷取影像的背景，而產生所謂的 差異性影像（differential image)(步驟66)。差異性影 像由重要晶粒影像之個別對應的像素値與在檢測下之晶粒 影像的差異所組成。構成差異性影像之像素強度値可以是 正値、負値、或是零，在整個差異性影像（步驟68 )上 加總及平均。產生的平均値則與得自於FEM檢測之相似 平均値比較，用以判斷該平均値是否超過由系統使用者所 預先設定的臨限値。在一些實施例中，可以直接比較由 FEM所得到之差異性影像平均値與由檢測所得到的差異 性影像平均値，用以決定晶粒中是否存有無法接受程度的 失焦缺陷。 如同精熟此項技術的人士所了解的，偏振片1 2及分 析器1 4可以彼此成一角度配置，用以避免通過所有光線 ，或者用以允許通過所有光線。在本發明之一實施例中， 偏振片1 2及分析器1 4彼此成一角度配置，用以避免所有 光線E!及E2之通過。在此實施例中’且當基板S不影響 反射光的偏振狀態時，光學感測器1 6沒有記錄到影像。 然而，因爲改變偏振的特徵通常存在於基板S上，而且通 常至少會存在某些程度的失焦缺陷’因此光線E3及e4將 會入射在光學感測器1 6上。 相對於分析器1 4之偏振片1 2的角度定位將通常取決 於被檢測基板S的性質，雖然可以利用到其他性質，包含 (但不限定）光源1 2的性質、光學系統的物理性質等等 。在一實施例中，偏振片1 2的偏振角度約與欲加以檢測 -26- 201101400 • 之基板的線性結構呈45度角。因此’可知在一些實施例 中，分析器1 4的偏振角度會改變’視欲加以檢測之基板 的性質而定。 在一些實施例中，將使用多重掃描檢測，用以判斷基 板S上之缺陷的存在。在一實施例中，利用裝置中之分析 器14及偏振片12產生第一通過，該裝置針對巨觀缺陷檢 測通過不足光線。第一通過僅在於判斷失焦或曝光缺陷是 0 否存在於基板之成像區域中，通常爲一個或更多個晶粒或 步進攝像。第二通過涉及發現巨觀缺陷如碎屑、裂痕、微 粒、孔隙、及刮痕，並使偏振片1 2及分析器1 4配置成允 許通過更多的光線量。 ' 在另一個實施例中，系統8可以用來偵測基板S上薄 膜的厚度變化、或者偵測基板上薄膜的存在。在一些情況 下，基板S的全部或部分在製程程序之後會殘留不需要的 薄膜。當以適當方式安排時，基板S的差異性影像將辨識 Q 出殘留薄膜的位置及殘留的程度。 結論 此處雖然已以圖示並說明本發明之特定的實施例，對 於精熟此項技術的人士而言應理解的是，用來達到相同目 的之任何經過計算的配置可以替代此處所示之特定的實施 例。精熟此項技術的人士會清楚地瞭解本發明之許多修改 - 。因此，本申請案在於涵蓋本發明之任何修改及變異。本 - 發明僅由申請專利範圍所界定。 -27- 201101400 【圖式簡單說明】 圖1爲本發明之成像系統之一實施例的示意圖’該系 統具有非90度角的入射標稱角度（nominal angle)。 圖2爲本發明之成像系統之一實施例的示意圖’該系 統具有實質上爲90度角的入射標稱角度。 圖3爲以向量型態呈現自基板所反射之所反射光線的 相對成分。 圖4係顯示所反射之光線通過適當配置之分析器之前 的相對成分。 圖5係顯示所反射之光線通過適當配置之分析器之後 的相對成分。 圖6爲流程圖，顯示用於檢測之檢測系統的設定方法 〇 圖7爲流程圖，顯示檢測基板的方法。 圖8爲示意圖，顯示半導體基板上之積體電路裝置之 陣列或晶片。 圖9爲測試結構之黑白或灰階影像，此測試結構由數 個具有不同臨界尺度之臨界尺度格狀物（CD box )所組成 【主要元件符號說明】 S :基板 P :光學路徑 -28- 201101400 電腦 Ο 成像系統 :照明器 :偏振片 :分析器 :光學感測器 :彩色濾光片 :檢測系統 :照明器 :光學感測器 :偏振片 :分析器 :濾光片 :分光器 :積體電路裝置 :測試結構 :街道 :格狀物 76 :結構201101400 VI. Description of the Invention: [Technical Field of the Invention] The present invention relates generally to the detection and method tools used in the manufacture of semiconductor devices to ensure quality and increase yield. [Prior Art] In the lithography semiconductor device manufacturing technology, it is imperative that the stepper can accurately focus the image of the reticle on the semiconductor substrate or the wafer. When the image of the reticle is not in focus, i.e., a known defocus state, the resulting semiconductor device structure may have an incorrect size and pattern. For example, the edges of the structure may be enlarged and unobtrusive, rounded or overcut, replacing the more desirable geometry, usually straight. This out-of-focus condition typically causes the problematic semiconductor device to have poor functionality and/or inoperability. Therefore, the out-of-focus measurement is an important method to allow the semiconductor device manufacturer to ensure that the stepper continuously focuses the reticle reticle image on the wafer, thereby making the process yield better and more profitable. Exposure defects are another common problem in the fabrication of semiconductor devices. When the degree of exposure of the photoresist layer is out of the acceptable range of light amount, the pattern to be formed on the semiconductor substrate may be incorrect. Therefore, the identification of defects when defects exist is of equal importance. In addition to detecting substrates or wafers for exposure or out-of-focus defects, the detection of substrates and wafers for process or material-related defects is often referred to as "maize". defect. The general definition of giant defects -5- 201101400 is on the substrate with a size of about 〇. 5u to 10u of debris, cracks. Traces, scratches, laminate peeling 'and/or particles. Such defects tend to cause failure of the semiconductor device and significantly reduce the yield of such devices. Note: The size of the macroscopic defect can be higher or lower than the above-mentioned size range. The above size range only defines the nominal size of such a defect. Traditionally, a specified detection system has been used to detect macroscopic defects, which have not been able to immediately and reliably identify the presence of exposure defects or out-of-focus defects. Identifying exposure defects or out-of-focus defects is usually done using optical critical dimension (OCD) techniques for a variety of precision measurement tools such as elliptical polarimeter (e 11 ips 〇meter), reflectometer (refl ectometer), scatterometer ( Any of the SCatterometers. Ideally, the same optical system is used in combination with the identification of exposure and the presence of _ out of focus defects and the detection of substrates for giant defects. SUMMARY OF THE INVENTION An embodiment of a detection system for identifying defects on a substrate includes a light source that directs light onto the substrate to be detected. A first polarizing filter or polarizing plate is positioned between the light source and the substrate. A second polarizing filter or analyzer is positioned between the substrate and the optical sensor, and the optical sensor receives light reflected from the substrate. The polarizing plate and the analyzer are disposed at an angle to each other such that the image intensity of the image captured by the optical sensor is at least partially related to the presence of polarization dependent defects on the substrate being detected. Polarization-related defects include defocus and exposure defects. It is also possible to identify defects whose primary dimension is about the wavelength of the incident beam or less, which is not out of focus and exposure defects. -6- 201101400 The light source can be of any useful type including, but not limited to, broadband incandescent light, or laser. Any of these sources may be flashing and may be positioned at any useful angle of incidence including the angle of incidence normal to direct light onto the surface of the substrate. The laser can be a fixed, different monochromatic light or can be configured to output light at a plurality of different nominal wavelengths. When flash illumination is used, the order in which the flash is extinguished is at least partially related to the speed at which the substrate is relative to the movement of the detection system. This allows the detection system to reliably capture the substrate image at the appropriate location. The optical sensor or the imager can be a monochromatic light rechargeable capacitive device (CCD). In some examples, the optical sensor can be a Bayer-type color imager or a 3-wafer design. In another example, one or more light sources and/or color filters can be used in conjunction with a monochrome optical sensor to obtain color data from the substrate. Both face scan and line scan can be used. 〇 In addition to the out-of-focus and exposure defects, it is also possible to identify its type of defect. These types of defects can include: cavities, voids, debris, cracks, particles, and scratches. The detection system in accordance with the present invention operates by directing light onto a substrate by configuring a light source. A first polarizing filter is positioned between the light source and the substrate, and an optical sensor is placed to receive light reflected from the substrate. The second polarizing filter is disposed between the substrate and the optical sensor, so the first and second polarizing filters are at a selected relative angle to each other. The inspection system is used to capture images of the substrate and to generate comparative data from these images to identify the presence of exposure and/or out-of-focus defects (if any) on the 201101400 substrate. Configuring the polarizing filter to capture the desired image involves rotating the first and second polarizing filters together to a desired angle of detection while maintaining a selected relative angle therebetween. The comparative data is obtained by first generating a difference image for each captured image, and then obtaining an average of the pixel intensity differences of the individual difference images for the entire difference image to obtain an average image intensity for each difference image. The average image intensity of each captured image is evaluated for a predetermined threshold, and if there is an exposure or out of focus defect on the substrate, it is used to judge the presence of at least one of the exposure and the out-of-focus defect on the substrate. The output of the optical sensor is corrected for the degree of defect of at least one of the known exposure and out-of-focus defects in the substrate, which is used to determine the appropriate level of exposure and out-of-focus defects. In one embodiment, the correction involves capturing a plurality of images of the calibrated substrate, wherein each image has been subjected to a known degree of out-of-focus and exposure defects. As described above, the difference image generation is obtained by averaging the difference in pixel intensity of the difference image for the entire difference image for each captured image to obtain an average image intensity. The average image intensity 値 is recorded for each captured image with a known degree of out-of-focus and exposure defects. The user can select any of the recorded average image intensities 値 which indicate a certain degree or magnitude of out-of-focus and/or exposure defects 临 as a threshold 値; or can be interpolated between such recording ticks; or simply Use record 値 as a starting point and add specific corrections for the product to the starting point. The user of the detection system is completely defined as the appropriate threshold for defocusing and/or exposure defects. -8 - 201101400 • Generating a difference image involves averaging a plurality of captured images based on one pixel by one pixel to obtain an averaged image. The averaged image is removed from each captured image, and the difference image is generated on a pixel by pixel basis. The difference image can be regarded as an array of pixel intensity, or a difference in pixel intensity. Array. Substrate detection for defocus and/or exposure defects can be performed simultaneously with detection of other 0 defects such as pits, voids, debris, cracks, particles, and scratches. Alternatively, the detection of these individual defect patterns may be performed sequentially, or even in a time-shifted manner, i.e., at a time that is clearly separated from each other. In another embodiment of the invention, image analysis techniques, such as spatial pattern recognition (SPR), can be used to analyze the difference images to identify the boundaries of the layers on the substrate. Note that the layer boundary as described above may be a partial material layer of the substrate portion or a material layer related to the residue of the substrate-independent portion, that is, the material layer may be one type of contaminant or another type. EMBODIMENT OF THE INVENTION [Embodiment] The present invention will be described in detail below, and embodiments of the present invention may be illustrated by way of illustration of the drawings. In the drawings, the same numbers represent substantially the same components that are presented by various angles of view. These embodiments will be described in detail below, so that those skilled in the art -9 - 201101400 can implement the present invention. Other embodiments may be utilized to carry out structural, logical, and electrical changes without departing from the scope of the invention. Therefore, the following detailed description is not to be considered in a limiting The present invention relates to a method and apparatus for determining the presence of exposure and lithographic defects in a semiconductor substrate by measuring changes in polarized light reflected from the substrate. In order to simplify the following discussion, "defocus defocus" - the words in this article will represent both exposure and out-of-focus defects, a given substrate may have one or another defect, or both. In addition, "out of focus" The phrase will cover a wide range of defects under the detection or an undesired portion of the substrate, the characteristics of which will be similar to light and/or out of focus defects, and can be characterized by the detection system of the present invention. . Although those skilled in the art know that the point of defocusing defects affects the nature and extent of this polarization change, in general, the focal defect is a polarization-dependent property, that is, the out-of-focus defect causes the light to be polarized. Change of state. The term "substrate" as used herein shall be used to encompass any material or structure that may be viewed by the present inspection system. In particular, "a semiconductor wafer containing any structure, type or material, including (only limited to) the entire wafer, unpatterned wafer, patterned, partially patterned wafer, all or Partially patterned damaged wafers are patterned wafers, wafers that have been cut in any form or on any support machine 'film frames JEDEC trays, Auer boat containers (You can invent the board table out of focus j (although the board is exposed or his view, the astigmatism invention board) but the non-wafer 'unstructured', Auer • 10-201101400 • boats), gel (gel) or wafers in waffle packs, commonly referred to as mcm multi-chip modules, etc. The term substrate and wafer are used interchangeably herein. The term "macroscopic defect" It will contain all the features that are not actually expected to appear on the substrate, which are not related to polarization. As mentioned above, giant defects are traditionally referred to as pits, voids, debris, cracks, particles, and scratches. Wait a minute. Attention, in In some cases, the size of the macroscopic defect may be close to the wavelength of the incident light detected by 0. In these cases, the macroscopic defect may affect the polarization state of the light reflected thereon. Referring to Figure 1, the imaging system 8 An embodiment includes a illuminator 10, a polarizer 12, an analyzer 14, and an optical sensor 16. The illuminator 10 directs light along the optical path P to the polarizer 12, and the polarizer 12 is actually only The light having a predetermined polarization angle is penetrated. The light transmitted through the polarizing plate 12 is then incident on the substrate S. In an embodiment, the substrate S is a germanium wafer or a partial germanium wafer having a photoresist formed thereon. Structures. In some implementations, these structures are formed on one or more semiconductor devices on the substrate s. Other mechanisms or structures may be formed on the substrate s. As shown in Figure 1, the optical path P is located opposite At the incident angle of the illegal line of the substrate S. In some embodiments, the illuminator 10, the polarizer 12, the analyzer 14, and the optical sensor 16 and other related optical components such as an objective lens may be provided. Wait, in order to adjust the base The angle of incidence of light on S. The type of device mechanism that adjusts the angle of incidence of light in system 8 is known to those skilled in the art - such device mechanisms can include mounting plates, the light of system 8 The component is mounted on the mounting plate, and the mounting plate can be rotated by the rotating mechanism. The -11 - 201101400 rotating mechanism can be one or more actuators. The incident angle can be fixed (as shown) or Adjusted for various product installations. Further, in some embodiments, the angle of incidence can be adjusted as needed during the inspection. Illuminator 1 〇 can be of any useful type, including broadband white light, with a fixed wavelength output. A laser that emits multiple wavelengths, or a plurality of lasers that direct light along an optical path. The intensity required for the illuminator can depend on the application to which the system 8 is directed. In some applications, high intensity illumination is required, and conversely, in other applications a relatively low intensity is required. The illuminator 1 〇 can be used to provide a motion for actually maintaining a certain output, or providing a flash to freeze the substrate S in the system 8, and can quickly capture an image of the substrate S. The light incident on the substrate S is reflected from the substrate, and the reflected light is incident on the analyzer 14, the analyzer 14 is a polarizing optical element similar to the polarizing plate 12, and the polarizing plate 12 can only have a predetermined polarization angle. The light passes through. Light passing through the analyzer 14 is incident on the optical sensor 16, and the optical sensor 16 captures an image of the substrate S. Although any device capable of producing a two-dimensional array of pixel intensity 灰 in grayscale or color can be used, such as line scan, or time delay integration (TDI) device, or CMOS optical sensor array, In one embodiment, however, optical sensor 16 is a two-dimensional electronic optical sensor, such as a charge coupled device (CCD). In one embodiment, the optical sensor 16 is a monochromatic optical optical sensor, wherein each pixel of the 2D pixel array of the optical sensor records a gray scale 0 of 0-256, and the pixels collectively represent the image of the substrate S . When using a monochromatic optical optical sensor, one or more colors -12-201101400. The filter 18 can be positioned in the optical path S between the illuminator 1 〇 and the optical sensor 16 so that only light rays in the wavelength range corresponding to the color filter can pass. In another embodiment, the optical sensor can be a Bayer type, or a 3-wafer type of color optical sensor having individual optical sensors, each for a different color, such as a sense The detector is used for red light, one sensor for blue light and one for green light. 0 Those skilled in the art will appreciate that the basic components of System 8 described above will be used with or without other optical components, and that the basic components include, but are not limited to, optical filters, lenses. , mirrors, retarders, and modulators. A detection system that can be adapted to implement the present invention is Rudolph Technologies Inc. Produced by Flanders, New Jersey, the product is called WaferViewTM. Again, it should be understood that system 8 can be used to perform a number of functions that can be performed at the same time or at intervals of time Q. For example, system 8 can be adapted to perform detection of macroscopic defects and for detection of out-of-focus defects. Furthermore, system 8 can be used to detect macroscopic defects' followed by detection for out-of-focus defects, or both can be performed simultaneously. Color filter 18 can be used in system 8, as shown in the schematic of Figure 1. One or more color filters 18 may be placed between the polarizer 12 and the substrate S' between the substrate S and the analyzer 14, between the illuminator 1 and the polarizer 12 Between, or between the analyzer 14 and the optical-sensor 16. In one embodiment, the color filter 18 may be a known filter wheel of the type of the technology of the present invention, wherein one of the color filters of the color filter group is fixed. In the rotating wheel located in the optical path P, it is thus possible to selectively position the color filter 18 across the optical path P. In another embodiment, the removable filter holder can be located in the optical path P, allowing different color filters to be located in the optical path P. In another embodiment, a fixed color filter can be mounted in the optical path P. It should be understood that any filter medium or mechanism can be used as the color filter as long as it is suitable for selectively passing a predetermined wavelength or range of wavelengths. In some embodiments, the output of the optical sensor 16 can desirably be separated for a predetermined color channel, wherein the "color channel" is defined as a predetermined wavelength or range of wavelengths. As suggested above, color channel separation can use color filters, use color optical sensors with directly distinguishable individual color channels such as 3-chip optical sensors and Bayer optical sensors, or use Illuminator 1 输出 that outputs light over a pre-selected wavelength range. It should be understood that some substrates S may penetrate partially or completely for certain wavelengths or ranges of colors. For example only, a substrate may penetrate or destructively interfere with most of the light incident on the substrate having all of the incident blue light having a wavelength centered at about 475 nm, but the reflection has a majority of about 700. The red light of the wavelength centered at nm. In this case, it is useful to use the optical sensor 16 to output a signal derived from the red light incident on the optical sensor 16. The use of data associated with individual color channels will depend on which features the system 8 will verify. In some embodiments, certain semiconductor substrates, i.e.,-14-201101400, will be characterized by a tendency to reflect light in a known manner, so that a detection system 8 can be specifically provided for certain products to optimize the product. Detecting Figure 2 shows another embodiment of the invention in which the detection system 3 is configured in a conventional manner with an illuminator 32 for directing light along the optical path P through the polarizer 36, filter 40 (selection And the optical splitter 42 onto the substrate. The light reflected from the substrate S on the path P is guided by the spectroscope 42, through the filter 40 (optional) and the analyzer 38 to the optical sensor 34. The systems 8 and 30 are substantially similar except for the presence of the beam splitter and the difference in angle of incidence. In this embodiment, the optical path P is substantially perpendicular to the substrate S. It has been observed that, as the defocus defect will change the geometry of the structure formed on the substrate S, the out-of-focus defect will also change the reflectivity of the substrate S. Other factors that change the reflectivity of the substrate S are the properties of other layers and the wavelength, polarization, and angle of incidence of the light incident on the substrate. Using a detection system in accordance with various embodiments of the present invention, such as system 8 or 30, can discern changes in reflectivity caused by out-of-focus defects and can be quickly and reliably identified. In general, the light from the illuminators 10 and 32 is polarized by the polarizers 1 2 and 36 to a predetermined angle P and incident on the substrate at a specific angle β. σ reflection when the substrate S will change the polarization state of the incident light, polarization The state is related to several features, especially out of focus. Information about the out-of-focus defects in the substrate S can be obtained by the change in polarization. The reflected light passes through the -15-201101400 analyzer 14' and is incident on the optical sensor 丨6. The configuration of the analyzer 14 (described in more detail below) helps to ensure that the information taken from the optical sensor 16 contains information relating to both the amplitude and polarization variations of the reflected light, and particularly with respect to the substrate S. Information on the existence of defocus defects. In one embodiment, the light incident on the substrate S is linearly polarized by the polarizers 1 2 and 36, which typically represents the simplest solution to the detection problem. In other embodiments, the incident light is polarized in an elliptical manner or polarized in a circular manner, as desired. The light reaching the optical sensors 16, 34 can be modulated by setting the analyzers 1 4 and 3 8 at an angle A relative to the polarizers 12 and 36. The angle between the polarizing plates 12, 36 and the analyzers 14, 38 is given as P-A. Referring now to Figure 3, in general, when light is reflected from the substrate S, a portion of the incident light Ep is reflected in a manner different from the incident light of the other portions. A portion of the incident light Ep is reflected from the portion of the surface of the substrate S which is not related to the polarization, and does not have any significant change in its polarization, as shown in Fig. 3, E! and E2. Some examples of such features that may be found on the semiconductor substrate S include, but are not limited to, light and dark, such as debris, cracks, scratches, pits, voids, and particulates, which appear to be defects in the substrate. Another portion of the reflected incident light E3 is reflected from polarization-related features formed on the substrate that will change the polarity of the incident light. Examples of some polarization-related features or structures found on the semiconductor substrate S include, but are not limited to, linear structures, conductors, connecting lines, vias, and streets. There is another portion of the structure or features of incident light that is reflected from the surface of the substrate S, which alters the structure of the polarization of the reflected light or the characteristic of the defocusing defect. This light E4 polarization state is different from E3. The structure of the reflected light E4 may be similar or identical to the nominal structure of the reflected light & unless they suffer from a defocusing defect, and the degree of the out-of-focus defect affects the strength structure of the light E4 to the defocusing defect. Test structure 74. Figure 8 is a schematic view of a portion of the substrate. In this example, a semiconductor wafer W' has a plurality of integrated circuit devices (IC devices) 0 70 formed thereon. The integrated circuit device 70 is arranged in a grid manner with a space between the grids, commonly known as the street 72. A plurality of test structures 74 are formed in the street 72. Also on each of the 1C devices 70 is a structure 76 formed thereon that forms part of the active circuitry of the 1C device 70 and suffers from out-of-focus defects. It should be noted that in some cases, the majority of the 1C device 70, or at least the 1C device 70, may have one or more structures 76 suffering from out-of-focus defects. It is typically formed in a manner that makes both test structure 74 and structure 76 similar. For example, structure 76, which is part of the active Q circuit of 1C device 70, will have the test structure 74 formed in the same manner if a series of linear structures are formed therein, such that the characteristics of test structure 74 will indicate structure 76. Characteristics. In some embodiments, test structure 74 includes a series of lattices 78, each of which has a periodic structure formed therein. Figure 9 is a photograph of a test structure 74 having exemplary properties having a plurality of discrete regions 78. Each of the disconnected regions 78 has a nominally different - critical dimension, i.e., a periodic structure having a particular spacing or critical dimension. Note: The net effect of the test structure 74 is to provide a photographic snapshot of -17-201101400 (the degree refers to the anti-homogeneous to obtain some information using the 7 6 variable scale, and these messages are usually only Provided by the FEM wafer on each product wafer. Because the periodic configuration at certain critical dimensions will produce a response to the polarization state of the light reflected from the periodic structure, the test structure 74 is used to evaluate the lithography process. It will be helpful to perform operational changes in the stepper of the patterning step. Again, when the disconnected region 78 of the trial structure 74 is sufficiently similar to the structure 76 located in the active power of the 1C device 70, and Assuming that there is an additional process variation that may affect the change in the polarization state of the incident light, for example, the material properties of the stack on which the substrate W 1C device 70 is formed are relatively thick, we can obtain a critical dimension for the structure 76. The general indication of the number. As the polarization state of light incident on the surface of the substrate may result in a change in the intensity of the light, in some instances, the flat gray level 某个 of a certain region 78 may be directly and without simulation associated with the critical dimension of the region. . Based on the average gray scale 値 of the structure 76, this correlation can be used to identify the critical dimension of the structure 76. Although the critical dimension 区域 of the region 76 in this manner may not be very accurate, the critical dimension of the structure on the substrate is evaluated using a test structure 74 having a multi-critical dimension, for obtaining a critical dimension in the 1C device 70. Any size message is useful. For example, in addition to the "absolute" criticality or the alternative assessment of the "absolute" critical dimension, the grayscale 値 of the test structure 76 (or the unconnected region 78) selected by the measurement and the structure that is continuously detected or imaged are used. The difference in grayscale between 76 (or test structure 74), the grayscale information indicative of the critical dimension can be used to identify the magnitude of the change in the critical dimension. Referring to Figure 4, it can be seen that there is no contrast between the intensity of the reflected light E 3 and E 4 for the light that does not pass through the analyzers 14, 38. It will be appreciated that it is difficult to identify out-of-focus defects in this case. However, once the reflected light E!, E2, E3, and E4 are obtained by the appropriately configured analyzers I4, 38, the optical signals E'1, E'2, E'3, and E'4, the optical signal E'3 and The degree of contrast between E ' 4 is sufficient to obtain useful information about the existence of defocus defects. See Figure 5. In one embodiment, the angle P-A between the polarizer 12 and the analyzer 14 is experimentally determined. Referring now to Figure 6, using test substrate S positioned within system 8 for detection (step 50), illuminator 10 (in one embodiment, a flash illuminator) is set at a predetermined illumination level (step 52), It is preferred that the illuminator 1 is near its highest brightness output, but can be any suitable brightness. The light Ep from the illuminator or light source 10 is guided above the substrate S via the polarizing plate 12. Next, the polarizing plate 12 is set at the 角度 angle P (step 54). In one embodiment, the angle of the polarizer 12 is directed substantially perpendicular to any linear structure present on the substrate. It will be understood that when the substrate S is a semiconductor wafer on which the semiconductor device is formed (in any completed state), such a structure usually, but not always, has a remarkable linear structure. The reflected light (E!, E2, E3, and E4) is used by the analyzer 14' to obtain optical signals, E'2, E'3, and E'4 on the optical sensor 16. In some cases, there is no linearity-structured orientation on the substrate S, or a linear structure formed on the substrate S, which can be selected at any angle P for the polarizer 12. -19- 201101400 The analyzer 14 then turns to the angle a (step 56) so that there is sufficient illumination to reach the optical sensor 16 to allow detection of the substrate for macroscopic or polarization-independent defects. S, such as U.S. Patent Nos. 6,324,298, 6,487, 307, and 6, 826, 298, each of which is incorporated herein by reference in its entirety in its entirety in its entirety herein in Note: When the illumination is not only for the defect detection substrate S, but also the signal-to-noise ratio of the good quality detection is used to satisfy the end user of the system 8, no significant errors such as false positives and missed defects will occur during the inspection. When the analyzer 1 4 will be considered to be at the correct angle A. The signal-to-noise ratio of system 8 is determined by analyzing the output of optical sensor 16 in a known manner. Once the positioning angles P and A of the polarizer and the analyzer are known, the polarizer and the analyzer are rotated together via a series of detection angles (step 58), and the relative angle between the polarizer and the analyzer (PA) is maintained as Substantially fixed, rotated to a desired angular position relative to the substrate S, this angular position will provide sufficient contrast between the light signals Eh and E'4 as described above. During the rotation of the polarizer 12 and the analyzer 14, the intensity of the incident light on the optical sensor 16 is recorded. The light intensity of the optical sensor 16 is recorded for the detection angle or detection position of the rotation of each of the polarizing plate 1 2 and the analyzer 14. In one embodiment, the polarizer 12 and the analyzer 14 are rotated in a more or less continuous manner, and the rotation of the polarizer 12 and the analyzer 14 is increased by a small amount, and the polarizer 12 and the analyzer 14 are recorded. The position and the intensity of light present in the optical sensor 16. The data obtained during the polarization of the polarizer 12 and the analyzer 14 are analyzed as a comparison between the reflected rays E ′ 3 and E ′ 4 to identify the optimum detection angle or detection position of the polarizer and the analyzer. -20- 201101400 • The identification process for the optimal configuration of angles P and A can be manual, where the user of system 8 rotates polarizer 12 and analyzer 14 through a selected angular range, while optical sensor 16 The image data is recorded, and the image data is processed by a computer C of an appropriate type to determine the optimum angle PA. Alternatively, and preferably, the polarizer 12 and the analyzer 14 are automated so that the aforementioned computer can control their rotation while recording information from the optical sensor 16 at different angles P-A. The automation of polarizers and analyzers is well known to those skilled in the art. As suggested above, the process of determining the optimum angle P-A between the polarizer and the analyzer may require multiple iterations before and after the steps described below. For example, once the calibration/setting step of the entire process is completed, it is useful to perform a multiple-time full-scale correction/setting step to determine that the final system settings are optimized. Once the system 8 has been properly set up as illustrated in Figure 6, detection of out-of-focus defects can be performed, or if necessary, detection of its Q-defects can be made. However, system 8 must first be corrected. Calibration is performed using a focus exposure matrix (FEM) wafer. The FEM is a substrate with a plurality of already formed patterns or structures above it, each having a different focus position and exposure time. FEM is commonly used in the fabrication of semiconductor devices as part of a corrective photolithography tool. The FEM materializes the structural changes in the pattern and structure formed on the substrate S, and these changes are derived from . Focus position and exposure changes. During the detection of the substrate S, the out-of-focus and exposure data from the FEM were used as comparison data (comparator 201101400). Note that the pattern formed on the FEM may be different from the pattern formed on the substrate S to be inspected, but the same is preferable. The method of obtaining the out-of-focus data for the purpose of correction is the same as the method of obtaining the out-of-focus data for the purpose of detection, and the method of obtaining the out-of-focus material for the purpose of correction will be described as a part of the detection procedure. The difference between the calibration and the test procedure will be indicated where appropriate. During the detection period, the substrate S to be detected is taken and placed on a wafer support table or a top plate (not shown), and the substrate is opposed to the detection system in a conventional manner on the wafer support table or the top plate. The optical device (optics) of 8 moves. In some embodiments, the substrates s (product or FEM) are detected one by one. In one embodiment, wherein the substrate S is a wafer, and the semiconductor device is formed on the wafer, the detection of the substrate S is performed on the basis of the grain level', that is, the sensor 16 is used to make the individual crystal grains on the substrate S. Image imaging 'and these images are processed in the following manner. In other embodiments, the detection is based on a field of view. The optics of system 8 are configured to capture field of view images that may vary in size from individual dies or individual stepper imaging. When the field of view of system 8 is smaller than the individual dies, multiple fields of view can be stitched together to form composite images of individual dies. The same tandem technique should be used to create a composite image of the entire stepper photography. It should be understood that tandem image technology for forming composite images has been known to the art. In other embodiments, when the field of view is larger than the individual dies or larger than the stepper image, the resulting image can be cropped to display one or more crystal grains or stepper imaging. It is usually useful to 'do not cut too large images' to -22- 201101400. • It contains multiple dies from individual stepper cameras, because the grain produced by the first stepper camera may be Accepted, and those produced by the second stepper camera may be defective. The technique of cutting off images is known to those skilled in the art. In other embodiments, the image of the entire substrate S is first captured to complete the detection of the substrate S. When the substrate S is relatively small, the system 8 can be used to capture the principle of surface scanning. When the substrate S is larger than the field of view of the surface scanning system 08, a plurality of images of the substrate S can be taken and connected in series as described above. The tandem connection can be performed in conjunction with the line scan, or can be performed in conjunction with the face scan detection system 8. Those skilled in the art will be able to immediately understand that 'the appropriate polarizer/analyzer angle PA is obtained as described above. The image of the individual dies on substrate S is captured by optical sensor 16 (step 60). See Figure 7. The calibration and testing procedures described below will be based on a die bonded to a die, although it should be understood that other Q-based operations can be performed. The user can select the grain on the imaged substrate s, and the user judges that the grain does not have defects such as debris, cracks, pits, color variations, and particles to form a model. This determination is entirely dependent on the user of system 8, and can vary greatly depending on the nature of the substrate S and the manner in which the product is used. For example, the substrate S detected by the user has a semiconductor device formed thereon, which is intended to be used as a heart rate adjuster, and the user has a very severe standard for the number of defects on the die used to generate the model. . On the contrary, the substrate S detected by the user has the same semiconductor device formed thereon, and is intended to be used in a consumer product that is inexpensive and can be discarded, and the user is likely to accept more The grain of the target defect is used as a model. In short, the purpose of a "good" die for the purpose of generating a model depends on the judgment of the user of system 8. For the purpose of generating the model, 'though imagining is to obtain images of all the good grains on the substrate S, but usually only a statistically significant number of good grains are obtained; in general, this amount is smaller than the overall good crystal. The number of granules, and is about 10 to 15. If sampling statistically meaningful samples, at least the grains with no more than random defects should be selected, as a small number of randomly occurring defects will not have a significant impact on the detection. It should be understood that a large number of Non-random defects will be more likely to distort the test procedure. Useful models can also be obtained using automated methods. For example, computer C of control system 8 can select a statistically significant number of dies in a random manner and can capture images. These images are used to form a model that is used to detect individual images that make up the model. When the selected die is defective under the criteria selected by the user, the defective image is replaced by another randomly selected die. This procedure will be repeated multiple times until a suitable model is produced. Note that the model generated by manual or automatic mode will remain static, that is, it will not change with time. It should be carried out with the detection. After a period of time, new and good grains are added to the model to modify the model. . Because the term "model" can have different meanings for different people who are skilled in this technology, the vocabulary used here should be clarified; "golden die" or "important reference" is Used to describe -24- 201101400 • An image that adds the total number of pixels corresponding to several dies, and the average of these numbers can be used to determine the intensity of the image. Therefore, 'important crystals' are only averaged grain images. The term “model” is broader than “important crystals” or “important references.” In some cases, important grains or important reference information will not be included or used. Important grains are used in one embodiment of defect detection (step 62). The same 'important grain' can form at least part of the basis of the model 0 used for macroscopic defect detection. However, pixel intensity thresholds are typically defined for each pixel in the image, and the model used in macroscopic defect detection moves beyond the simple important grains. In the detection of macroscopic defects, if the pixel intensity 値 is found to fall outside the range defined by the threshold, the image is considered to be defective. These thresholds can themselves be purely standard deviations calculated from important grains, but more often include weighted defects that take into account different types of substrates, variations, and features, and the user can define the The standard of the product formed on the substrate. It should be understood that the model used for substrate inspection can be formed in a myriad of ways, and in any way may or may not be based on important grains, the only requirement for the macroscopic defect detection model is that the final test results are The user of system 8 is satisfied with him. When performing macroscopic defect detection, the time taken to obtain an appropriate model for such detection (step 64) may be more or less the same as the time at which important grains are produced. As indicated by arrow 65, in some cases the formation of the model utilizes information about important grains. Once the model is generated, the comparison model and the captured image are used to identify the defect (step 72). " During the detection of out-of-focus defects, the weights obtained in the previous steps - 25, 201101400 are used to remove the background of the captured image, resulting in a so-called differential image (steps) 66). The difference image consists of the individual pixel 重要 of the important grain image and the difference of the grain image under detection. The pixel intensity 构成 constituting the difference image may be positive, negative, or zero, summed and averaged across the difference image (step 68). The resulting average 値 is compared to a similar average 得 from the FEM test to determine if the average 超过 exceeds the threshold set by the system user. In some embodiments, the difference image average 値 obtained by the FEM and the difference image average 値 obtained by the detection can be directly compared to determine whether there is an unacceptable degree of defocus defect in the grain. As will be appreciated by those skilled in the art, the polarizer 12 and the analyzer 14 can be disposed at an angle to each other to avoid passing all of the light or to allow passage of all of the light. In one embodiment of the invention, the polarizer 12 and the analyzer 14 are disposed at an angle to each other to avoid passage of all of the rays E! and E2. In this embodiment' and when the substrate S does not affect the polarization state of the reflected light, the optical sensor 16 does not record an image. However, since the characteristics of changing polarization are usually present on the substrate S, and usually at least some degree of defocusing defects are present, the rays E3 and e4 will be incident on the optical sensor 16. The angular positioning of the polarizer 12 relative to the analyzer 14 will generally depend on the nature of the substrate S being tested, although other properties may be utilized including, but not limited to, the nature of the source 12, the physical properties of the optical system, and the like. Wait. In one embodiment, the polarization angle of the polarizer 12 is about 45 degrees from the linear structure of the substrate to be detected -26-201101400. Thus, it will be appreciated that in some embodiments, the polarization angle of the analyzer 14 will vary depending on the nature of the substrate to be detected. In some embodiments, multiple scan detection will be used to determine the presence of defects on the substrate S. In one embodiment, the first pass is generated by the analyzer 14 and the polarizer 12 in the device, which passes the insufficient light for the macroscopic defect detection. The first pass is only to determine whether the out-of-focus or exposure defect is 0 or not present in the imaging area of the substrate, typically one or more dies or step-by-step imaging. The second pass involves the discovery of macroscopic defects such as debris, cracks, microparticles, voids, and scratches, and the polarizer 12 and analyzer 14 are configured to allow more light to pass. In another embodiment, system 8 can be used to detect variations in the thickness of the film on substrate S or to detect the presence of a film on the substrate. In some cases, all or part of the substrate S may leave an undesired film after the process. When arranged in an appropriate manner, the differential image of the substrate S will identify the position of the residual film and the extent of residue. Conclusion While specific embodiments of the invention have been illustrated and described herein, it should be understood by those skilled in the art that any calculated configuration for the same purpose may be substituted. A specific embodiment. Those skilled in the art will be aware of many modifications of the invention. Therefore, this application is intended to cover any modifications and variations of the invention. This - invention is only defined by the scope of the patent application. -27- 201101400 BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a schematic view of an embodiment of an imaging system of the present invention. The system has an incident nominal angle of a non-90 degree angle. Figure 2 is a schematic illustration of one embodiment of an imaging system of the present invention. The system has an incident nominal angle of substantially 90 degrees. Figure 3 is a diagram showing the relative components of the reflected light reflected from the substrate in a vector form. Figure 4 shows the relative composition of the reflected light before passing through a suitably configured analyzer. Figure 5 shows the relative composition of the reflected light after passing through a suitably configured analyzer. Fig. 6 is a flow chart showing a setting method of a detecting system for detecting 〇 Fig. 7 is a flowchart showing a method of detecting a substrate. Figure 8 is a schematic view showing an array or wafer of integrated circuit devices on a semiconductor substrate. Figure 9 is a black-and-white or gray-scale image of the test structure. The test structure consists of several critical box sizes (CD boxes) with different critical dimensions. [Main component symbol description] S: Substrate P: Optical path-28- 201101400 Computer Ο Imaging System: Illuminator: Polarizer: Analyzer: Optical Sensor: Color Filter: Detection System: Illuminator: Optical Sensor: Polarizer: Analyzer: Filter: Beam Splitter: Product Body circuit device: test structure: street: lattice 76: structure