1336778 九、發明說明： 【發明所屬之技術領域】 * 本發明係關於圖像缺陷檢查裝置、圖像缺陷檢查系統及 圖像缺陷檢查方法；其係比較將檢查對象攝像後之檢查圖 像及與此檢查圖像應為本來相同之參考圖像，將彼此不同 之部分作為缺陷而檢測者；且係特別關於如下圖像缺陷檢 查裝置、圖像缺陷檢查系統及圖像缺陷檢查方法，其係為 了於半導體製造步驟檢測在半導體晶圓上所形成之半導體 # 電路圖案之缺陷，而將半導體電路晶圓表面攝像，將該攝 像圖像與參考圖像作比較，將彼此不同之部分作為缺陷而 檢測者。 【先前技術】 本發明之對象為圖像缺陷檢查裝置、圖像缺陷檢查系統 及圖像缺陷檢查方法；其係將檢查對象攝像後之檢查圖像 及與此檢查圖像應為本來相同之參考圖像作比較，將彼此 不同之部分作為缺陷而檢測者。在此，係以外觀檢查裝置 • (inspection machine)為例作說明，但本發明並不受限於 此，而該外觀檢查裝置係於半導體製造步驟檢測形成於半 導體晶圓上之半導體電路圖案之缺陷者。 一般之外觀檢查裝置，係從垂直方向將對象表面予以照 明，以捕捉其反射光之像的明視野檢查裝置。然而，亦使 用不直接S捉照明光之暗視野檢查裝置。如為暗視野檢查 裝置之情形，係配置從斜方向或垂直方向照明對象表面而 不檢測正反射之感知器，藉由將照明光之照射位置依序掃 113715.doc !336778 描’以獲得對象表面之暗視野像。因此，在暗視野裝置方 面，亦有不使用影像感知器之情形，此當然亦為發明之對 象。如上述般，如為將應為相同之2個圖像（信號）之對應部 分作比較，將差之大的部分判定為缺陷之圖像處理方法及 裝置的話，無論何種圖像處理方法及裝置，本發明均可適 用。 在半導體製造步驟方面，係形成多個晶片（晶粒）於半導 體晶圓上。在各晶粒上係反覆形成多層之圖案。所完成之 晶粒係藉由探針器及測試器作電性檢查，將不良晶粒從組 裝步驟排除。在半導體製造步驟方面，良率相當重要，上 述電性檢查之結果係被回饋至製造步驟，而使用於各步驟 之ΐ理上。然而，半導體製造步驟係由多個步驟所構成， 從開始製造起至實施電性檢查為止需要相當長的時間，因 此，當藉由電性檢查而察覺步驟有問題發生時，也已處於 複數個晶圓處理之途中，故無法將檢查結果充分發揮於良 率之提昇上。基於上述原因，而實施圖案缺陷檢查，其係 檢查途中之步驟所形成之圖案，而檢測缺陷者。如於全部 步驟中之複數個步驟實施圖案缺陷檢查，則可檢測在前面 之檢查後所產生之缺陷，將檢查結果迅速反映於步驟管理 上。 圖1係外觀檢查裝置之區塊圖，其係本發明專利申請人 於曰本特願20〇3-188209(下述專利文獻υ中所提出者。如 圖示般，在可往2次元或3次元方向自由移動之載物台丨之 上面，係設置著試料台（吸盤載物台）2。在此試料台之上， 1137l5.doc 1336778 係載置作為檢查對象之半導體晶圓3並予以固定。在載物 台之上部，係設置使用1次元或2次元之CCD照相機等所構 成之攝像裝置4;攝像裝置4係產生圖案之圖像信號，而該 圖案係形成於半導體晶圓3上者。 如圖2所示般，於半導體晶圓3上，複數之晶粒3a係往X 方向及Y方向分別反覆，呈矩陣狀排列。由於各晶粒形成 相同圖案，因此，一般係將鄰接之晶粒之對應的部分之圖 像進行比較。如兩方之晶粒並無缺陷，則灰階差小；如一 方有缺陷’則灰階差比臨限值為大，因此，藉由將此灰階 差與特定之檢測臨限值作比較，而將缺陷予以檢測（單檢 測）。但光如此並無法了解哪一方之晶粒有缺陷，因此， 進一步進行鄰接於不同側之晶粒的比較，如相同部分之灰 階差大於臨限值，則可知該晶粒有缺陷（雙檢測）。 攝像裝置4包含以TDI攝像元件等所實現之！次元之ccd 照相機；移動載物台1，使照相機對半導體晶圓3往χ方向 或Y方向以一定速度作相對性移動（掃描）。藉由丨次之在主 掃描方向的掃描，把如下帶狀區域（稱為帶區：swath)攝像 後之圖像信號80，係被從攝像裝置4輪出；此圖像信號被 變換為多值之數位化信號（灰階信號），並記憶於圖像記憶 部5;而該帶狀區域係與圖3(A)中以虛線作區隔顯示之攝 像裝置4之攝像寬度ws為相同寬度。 八後持績進行掃描，當從晶粒3a至隔鄰之晶粒3b之分 為止產生灰階信號（檢查圖像信號）後，差分檢測部6係從 圖像記憶部5讀出如下資料’並將之輪人至差分檢測部6; M3715.doc 1336778 該資料係：如圖3(B)所示緊鄰之2個晶粒3&及31)之各自相 同部分的小部分圖像81及82(邏輯圖框）之灰階信號（基準圖 像信號）。實際上，有實施微細之較準處理等，但在此不 作詳細說明。 差分檢測部6係被輸入鄰接之2個晶粒3a&3b之相同部分 的部分圖像81及82之灰階信號，將其一方作為檢查部分圖 像，將他方作為參考圖像，將彼此之對應像素彼此的灰階 號之差（灰1¾差）予以運算，並輸出至檢測臨限值計算部7 及缺陷檢測部8。檢測臨限值計算部7係依據灰階差之分 布自動决疋檢測臨限值，並輸出至缺陷檢測部8。缺陷 檢測部8係將灰階差已決定之臨限值進行比較，判斷是否 為缺fe。接著，缺陷檢測部8係針對被判定為缺陷之部 分’依照各個缺陷，輸出缺陷資訊，纟包含：該缺陷之位 置、灰階差、檢測時之檢測臨限值、及決定該檢測臨限值 之際所使用之缺陷檢測參數等。 其後’ A 了更詳細檢查被判定為缺陷之部分，缺陷資訊 係被供應至自動缺陷分等（ADC)裝置（未圖示卜在自動缺 陷分等裝置巾，係實拖如下缺时等處理：制定為缺陷 之部分是否為影響良率之真正缺陷；或為因攝像圖像之雜 訊等之影響而導致錯誤檢測之疑似缺陷；或是判定其為何 種缺陷（布線短路 '圖案缺損或微粒等）(> 、’” 在此缺陷分等處理方面，由於有必要將缺陷部分作詳細 檢查，因此需要較長處理時間。基於此因，在進行缺陷判 定之際，必須合乎真正缺陷無漏失且除真正缺陷以外= U3715.doc 1336778 判定為缺陷之要求。 基於此因，臨限值之設定係成為一大問題。如將臨限值 設定得較小’則被判定為缺陷之像素（pixel)增多，使得非 真正缺陷之部分亦遭判定為缺陷；其結果將導致缺陷分等 處理所需時間變長之問題。相對的，如將臨限值設定得過 大’則使真正缺陷亦被判定為非缺陷，而導致檢查不夠完 整之問題》 在本發明專利申請人於日本特願2003-188209(下述專利 文獻1)所提出之外觀檢查裝置方面，檢測臨限值計算部7 係依據檢查部分圖像及參考圖像之對應像素彼此的灰階差 之分布，而決定檢測臨限值。 當從差分檢測部6輸入被對比之2個部分圖像8 1及82所含 之各像素的灰階差信號，則檢測臨限值計算部7係作成如 圖4(A)所示之直方圖。接著，算出如圖4(B)所示之累積頻 度’在此灰階差之分布係遵照特定之分布的假設下，將變 換累積頻度予以算出，而其係已作變換使累積頻度對灰階 差呈線性關係者（參考圖4(C))。 其後’算出此變換累積頻度之近似直線，根據此算出之 近似直線，從特定之累積頻度之值，遵照特定之算出方 法’而決定臨限值。譬如’在圖4(C)之例中，假設所算出 之近似曲線之斜率為a、對縱轴之近似曲線的截距（亦即， 灰階差成為0之累積頻度）為b(可稱此斜率a及截距b為缺陷 檢出參數），則臨限值T係藉由下式（1)算出。 T=(Pl-b+VOP)/a+HO (1) 1137l5.doc I336778 在此，pi係對應於特定之累積機率（p)之累積頻度； VOP、HO係特定之感度設定參數。 [專利文獻1]日本特開2004-1 77397號公報 [專利文獻2]曰本特開平4-107946號公報 [專利文獻3]日本特許第2996263號公報 [專利文獻4]日本特開2002-22421號公報 [發明所欲解決之問題] 在將半導體晶圓表面攝像後之實際之檢查圖像方面，即 便是將以相同製程所形成之晶圓在相同狀態下攝像後之檢 查圖像，亦可能產生所謂「顏色不均」現象，其係：成為不 同明度（灰階值）之圖像；或是即使在1個晶圓内，原本應以 相同明度攝像之區域卻被以不同明度攝像。 如上述般’如使用顏色不均之檢查圖像實施缺陷檢查， 則連原本非缺陷之部分亦灰階差變大，被作為缺陷檢測而 導致疑似缺陷之增大。 然而’在半導體晶圓表面之攝像圖像產生顏色不均之情 形時，由於相較於進行缺陷檢查之圖像單位（邏輯圖框）之 大小，明度之對晶圓表面上之位置之差異的變動係較為緩 和’因此，即使觀察1個邏輯圖框内之灰階值之變動，並 無法檢測顏色不均之有無。 有鑑於上述待解決之問題’本發明之目的為，在圖像缺 陷檢查上’減少因檢查圖像產生之顏色不均所產生之疑似 缺陷；該圖像缺陷檢查係’將檢查對象攝像後之檢查圖 像、及與此檢查圖像應為本來相同之參考圖像作比較，將 113715.doc 1336778 彼此不同之部分作為缺陷而檢測者。 【發明内容】 $了達成上述目的’在本發明中，係依照檢查對象上之 寺疋大J之各區域，將依據將此等各區域攝像後之圖像所 a之像素之像素值而定之參考值，予以決定針對包含複 數個此特定大小之區域而構成之巨集區域，將就巨集區域 所含之上述特定大小之各區域所決定之參考值的分布資 訊，予以決定，於巨集區域依據已決定之分布資訊而變更 缺陷檢測條件，實施缺陷檢測。 又，在本發明中，就檢查對象上之特定大小之各區域， 在此等各區域進行測定與檢查對象有關之特定測定值，將 依據已測定之前述特定測定值而定的參考值，予以分別決 定；針對包含複數個此特定大小之區域而構成之巨集區 域將就巨集區域所含之上述特定大小之各區域所決定之 前述參考值的分布資訊，予以決定，依據在巨集區域所決 定之分布資訊而改變缺陷檢測條件，實施缺陷檢測。 [發明之效果] 針對根據就檢查對象上之特定大小之各區域而取得之像… 素值、檢查對象之測定值而決定之參考值，藉由算出比此 特定大小之區域更廣之巨集區域中之分布資訊，而可檢測 在將檢查對象攝像後之檢查圖像之廣大範圍所產生之顏色\ 不均。 接著’依據已算出之分布資訊而改變缺陷檢測條件，實y 施缺陷檢測；藉此可減少因檢查圖像所產生之顏色不均而// U3715.doc 11 1336778 產生之疑似缺陷。 【實施方式】 以下’參考附圖，針對本發明之實施例作說明。圖5係 本發明之半導體電路用之圖像缺陷檢查裝置之第i實施例 之區塊圖。圖5所示圖像缺陷檢查裝置1〇係具有與參考圖1 所說明之先前之外觀檢查裝置1〇等似之結構，因此，對相 同構成要素賦予相同元件符號，但省略其說明。 藉由移動承載檢查對象（半導體晶圓3)之載物台丨，使攝 像裝置4對半導體晶圓3往χ方向或γ方向以一定速度作相 對f生掃描，藉由1次之在主掃描方向的掃描，可獲得將如 下帶狀區域攝像後之檢查圖像，此檢查圖像信號係被變換 為多值之數位化信號（灰階信號），並記憶於圖像記憶部5 ; 而該攝像裝置4係包含以TDI攝像元件等所實現之i次元之 CCD照相機者；而該帶狀區域係與如圖3(A)中以虛線區隔 顯示之攝像裝置4的攝像寬度Ws為相同寬度。 回到圖5，如仍持續進行掃描，則從晶粒3a至隔鄰之晶 粒3b之分為止，產生灰階信號（檢查圖像信號p差分檢測 部6係從圖像記憶部5讀出如下資料，並將之輸入至差分檢 測部6，該資料係：如圖3(B)所示緊鄰之2個晶粒3&及31(之 各自相同部分的小部分圖像8丨及82(邏輯圖框）之灰階信號 (基準圖像信號）。 差分檢測部6係被輸入鄰接之2個晶粒3a及3b之相同部分 的。卩为圖像81及82之灰階信號’將其一方作為檢查部分圖 像，將他方作為參考圖像，將彼此分別之對應像素彼此的 I137l5.doc 12 1336778 灰階彳S號之差（灰階差）予以運异’並輸出至檢測臨限值計 算部7及缺陷檢測部8。 在此，部分圖像81係將圖6中以虛線所夾之區域間所示 之寬度Ws的帶狀檢查圖像80(帶區），如一點劃線所示般被 分割切出為複數個。基於此因，該等部分圖像81係成為與 本發明之申凊專利範圍有關之檢查部份圖像。又，被與部 分圖像8 1作比較之部分圖像82，係成為與檢查部份圖像作 比較之參考圖像。 再者，部分圖像81係將檢查對象（半導體晶圓3)就特定 大小之各區域攝像後之圖像，因&，其係成為就與本發明 之申請專利範圍有關之檢查對象上之特定大小之各區域， 將各區域攝像後之圖像。 又’部分圖像㈣將檢查圖像就特^大小分割而切出， 因此，其等係成為與發明之中請專利範圍有關之圖像區 塊0 又，在本實施例之圖像缺陷檢查裝置1〇中，係將緊鄰之 2個晶粒之邏輯圖框之-方作為參考圖像使用。然而以 過去所攝像之理想之晶圓3的攝 J僻彳冢圖像之樣本來取代前 者，作為參考圖像使用亦可；又，將複數個晶旧之攝像 圖像二樣本，譬如將相當於平均為各個像素而製作之樣本 (所明黃金影像）作為參考囷像 sri 豕便用亦可；或是從在半導體 曰日圓3之表面所形成之圖荦 ^ 系之。又β十資料（譬如，CAD資料） 荨，將藉由模擬以理想方式所制 飞斤製作之圖像作為參考圖像使 用亦可。 I13715.doc -13* 1336778 檢測臨限值計算部7係依照灰階差之分布，自動決定缺 陷檢測條件之檢測臨限值，將之輸出至缺陷檢測部8。 藉由檢測臨限值計算部7之檢測臨限值的決定，係如參 考圖4(A)〜圖4(C)所說明般，從差分檢測部6輸入被比對之 2個部分圖像8 1及82所含之各像素的灰階差信號，作成直 方圖（參考圖4(A))，接著’算出其累積頻度（參考圖4(B)); 在被輸入之灰階差之分布係遵照特定之分布的假設下，算 出變換累積頻度’而其係已作變換使累積頻度對灰階差呈 線性關係者（參考圖4(C))。其後，算出此變換累積頻度之 近似直線，依據此算出之近似直線（亦即，算出缺陷檢測 參數.近似曲線之斜率a及截距b)，根據此算出之近似直 線，從特定之累積頻度之值，遵照上述之特定之算出方 法，而決定臨限值。 缺陷檢測部8係將已決定灰階差之臨限值作比較，並判 定是否為缺陷。此外，缺陷檢測部8係針對已被判定為缺 陷之部分，依照各個缺陷，輸出缺陷資訊，其包含：該缺 陷之位置、灰階差、檢測時之檢測臨限值、及缺陷檢測參 數等。 圖像缺陷檢查裝置10包含參考值決定部21，其係就各個 部分圖像（邏輯圖框）81，依據此等部分圖像所含之像素的 像素值，將遵照特定之決定方法所定之參考值分別予以決 定者；如圖6所示般，部分圖像81亦為將檢查圖像⑽就特 定之大小而分割成之圖像區塊。 參考值決定部21例如可以圖像區塊（各部分圖像8ι)所含 1137l5.doc 14 1336778 =數：像素之平均值、分散值、最大值或最小值為參考 值進订決定。或是參考值決定部21可以如下者為參考值進 灯決定：譬如，各部分圖像81所含之全部像素之平均值、 分散值、最大值、最小值；或此等最大_ 值或差。 阻<> r间 又’參考值決定部21可以如下者為參考值進行決定：在 圖像區塊（各部分圖像81)所含之全部像素中，存在於部分 圖像81内之特定範園之像素之像素值的平均值、分散值、 最大值、最小值；或此等最大值與最小值之令間值或差。 或是將存在於部分圖像81内之特定位置之像素之像素值作 為參考值進行決定亦可。 再者’參考值決定部21可以如下者作為參考值進行決 定：各部分圖像81及其與在差分檢測部6中應被比較之參 考圖像之間的差圖像所含之全部像素的平均值、分散值、 最大值最j、值，或此等最大值與最小值之令間值或差。 或是將存在於此差圖像中之特定範圍之像素的像素值之平 句值刀政冑I大值、最小值或此等最大值與最小值之 中間值或差，作為參考值進行決定亦可。 回到圖#者，圖像缺陷檢查裝置1〇包含分布資訊決 定部22 ’其係針對包含複數個圖像區塊（在上述之例中係 部分圖像81)而構成之巨集區域，將依照巨集區域所含之 各圖像區塊而決定之參考值的分布資訊予以算出者。而該 圖像區塊係算出參考值之單位。 在此，上述巨集區域可設定為：譬如在圖7中，如二點 113715.doc ^36778 劃線所示之包含複數個部分圖像8丨而構成之區域。分布資 訊決定部22係將與某一巨集區域有關之分布資訊，作為依 就此巨集區域所含之複數之各圖像區塊（亦即部分圖像8工） 而决疋之參考值之集合而決定。又，在此等複數個參考值 之中’作為預定為本來相同值之參考值彼此之分散值而決 定亦可。 再者’圖像缺陷檢查裝置1〇更包含缺陷輸出可否判定部 23 ’其係依據在各巨集區域所決定之分布資訊’將各巨集 區域中之缺陷檢測條件分別再設定，在再設定後之缺陷檢 測條件下，將在各巨集區域藉由缺陷檢測部8所檢測之缺 陷進行判定輸出之可否者。 缺陷輸出可否判定部23係譬如輸入與在缺陷檢測部8已 檢測之與各缺陷有關之缺陷資訊，將各自之缺陷之缺陷資 sfl所含之檢測臨限值’依據針對該缺陷所屬之巨集區域所 決定之分布資訊作修正’藉此進行再設定；將該缺陷之缺 陷資訊所含之此缺陷之灰階差、與再設定後之檢測臨限值 作比較，如灰階差超過檢測臨限值之情形，則將該缺陷資 訊作為真正缺陷而允許輸出。相對的，如灰階差未超過檢 測臨限值之情形，則將該缺陷資訊作為疑似缺陷，而禁止 輸出。 又’缺陷輸出可否判定部23係譬如將各自之缺陷之缺陷 資訊所含之缺陷檢測參數（譬如，上述近似直線之斜率a及 截距b)，依據針對該缺陷所屬之巨集區域所決定之分布資 訊作修正，藉由修正後之近似曲線將檢測臨限值再設定’ 113715.doc •16· 1336778 將該缺陷之缺陷資訊所含之此缺陷之灰階差、與再設定後 之檢測Ba限值作比較，以判定該缺陷資訊之輸出的可否亦 可。 如上述般，將於缺陷檢測部8檢測之各缺陷，在缺陷輸 出可否判定部23再設定後之缺陷檢測條件下作再度檢測， 藉此依據對各巨集區域所決定之分布資訊實施缺陷檢測條 件之變更。 如與某一巨集區域有關之分布資訊被作為如下參考值之 集合而決定之情形時，缺陷輸出可否判定部23係將缺陷檢 測條件再設定，以使缺陷檢測感度隨著在分布資訊所含之 各參考值内預定為本來相同值之參考值彼此之參差不齊的 增大而降低；而該參考值係藉由分布資訊決定部22就此巨 集區域所含之各圖像區塊決定者。譬如，將檢測臨限值再 設定’以使之隨著被預定本來為相同值之參考值彼此之參 差不齊的增大而增大。 缺陷輸出可否判定部23譬如可將如下參考值作為上述預 定為本來相同值之參考值；而該參考值係關於複數個晶片 (晶粒）3a針對將晶片内之相同部分攝像後之各自之部分圖 像8 1分別決定者。接著，亦可將檢測缺陷條件再設定，以 使缺陷檢測感度隨著參考值彼此之參差不齊的增大而降 低。 與某一巨集區域有關之分布資訊，在如下參考值之中， 被作為預定為本來相同值之參考值彼此之分散值而決定之 情形時，缺陷輸出可否判定部23係將檢測缺陷條件再設 113715.doc 17 1336778 定，以使缺陷檢測感度隨著分布資訊的增大而降低；而該 參考值係藉由分布資訊決定部22就此巨集區域所含之各圖 像區塊而決定者。 再者，上述及下述之說明中，係將算出參考值之單位 (檢查圖像）進行分割後之圖像區塊之單位，作為差分檢測 部6之進行1次圖像比較之單位（部分圖像，即邏輯圖框）； 除此之外’圖像區塊之大小（單位）可作自由決定。 譬如，上述圖像區塊係可將形成於半導體晶圓表面之晶 片（晶粒）3a之一個分攝像後之圖像作為一單位（亦即，如同 將1個晶片3a分攝像後之圖像作為H固圖像區塊）而決定圖 像區塊之單位*在發生明度差呈緩和變化之彩色不均的情 形時’雖有必要於檢查圖像中之廣範圍取得分布資訊，但 藉由如上述般設定較大之圖像區塊，則可節約參考值之計 算量。 如上述般，如就將晶片之一個分攝像後之各圖像決定各 圖像區塊’則將各晶片攝像後之圖像係被預定為本來相 同，因此’可決定圖像區塊使各圖像區塊成為本來相同圖 像，或是可決定參考值之決定方法使針對各圖像區塊所決 定之參考值成為本來相同’因而可預定參考值為相同值。 再者，如於檢查對象之半導體晶圓表面形成之圖案係藉 由光學或電子射束之曝光步驟（微影步驟）所形成之情形 時’在將圖案曝光於半導體晶圓表面上之際，有可能在晶 圓表面上曝光圖案以未成像狀態（失焦狀態）被曝光而產生 不良品。在於此種失焦狀態下被曝光且形成圖案之半導體 113715.doc -18 - 1336778 晶圓之情形，在其表面之攝像圖像亦可觀測到顏色不均。 此顏色不均，係就1次之標線拍攝所曝光之各範圍（亦即， 以1片曝光遮罩同時曝光之各範圍），以明度差呈現變化之 樣態產生，因此，在以〖次之標線拍攝將複數個晶片（晶粒） 曝光之情形時，有必要在更廣之範圍取得分布資訊。 基於此因’參考值決定部21將檢查對象（在微影步驟形 成圖案之半導體晶圓表面）攝像後之前述檢查圖像之以1次 之標線拍攝所曝光之範圍的圖像，作為圖像區塊單位而決 定亦可。 如上述般，如就以1次之標線拍攝所曝光之範圍的各圖 像來決定各圖像區塊，則以1次之標線拍攝所曝光之範圍 的圖像係被預定為本來相同，因此，可決定圖像區塊使各 圖像區塊為本來相同圖像，或是可決定參考值之決定方法 使針對各圖像區塊所決定之參考值為本來相同，因而可預 定參考值為相同值。 圖8係本發明之半導體電路用之圖像缺陷檢查裝置之第2 實施例之區塊圖。在與本實施例有關之圖像缺陷檢查裝置 10中，參考值決定部21係就各圖像區塊，將缺陷檢測條件 之檢測臨限值作為參考值而決定；而該缺陷檢測條件係在 個別之圖像區塊内之缺陷檢測上所使用者。 如上述般，檢測臨限值計算部7係依據各檢查部分圖像 81所含之各像素之像素值與基準圖像之各像素之像素值之 灰階差的分布而決定檢測臨限值；因此，參考此檢測臨限 值之分布狀態，亦可檢測檢查圖像所產生之顏色不均。 1137l5.doc -19- 1336778 基於此因，在圖8所示之例中，參考值決定部21係從缺 陷檢測部8輸入在圖像區塊（各檢查部分圖像81)所檢測之缺 陷資訊’將個別之缺陷資訊所含之該缺陷之檢測時所使用 之檢測臨限值作為參考值而決定。 或是參考值決定部21可就各圖像區塊，將決定檢測臨限 值之際所使用之缺陷檢測參數（在參考圖4所說明之檢測臨 限值決定方法之例中，係近似曲線之斜率a及截距b)作為 參考值而決定；而該檢測臨限值即為於個別之圖像區塊内 之缺陷檢測上所使用之缺陷檢測條件。 此一情況’在圖8之例中，參考值決定部21係從缺陷檢 測部8輸入在圖像區塊（各檢查部分圖像8丨）所檢測之缺陷檢 測參數，將個別之缺陷資訊所含之該缺陷之檢測時所使用 之缺陷檢測參數作為參考值而決定。 圖9係本發明之半導體電路用之圖像缺陷檢查裝置之第3 實施例之區塊圖。在圖5及圖8所示之實施例中，檢測臨限 值計算部7係自動算出檢測臨限值；缺陷檢測部8係在已決 定之檢測臨限值之下，先進行缺陷檢測；其後，在缺陷輸 出可否判定部23將檢測臨限值再設定，實施可否作為檢查 結果而輸出之缺陷的再度判斷。 在與本實施例有關之圖像缺陷檢查裝置1〇中，在缺陷檢 測部8實施缺陷檢測之前，分布資訊決定部22係先決定上 述之分布資訊，根據此分布資訊，檢測臨限值計算部7在 決定缺陷檢測條件之檢測臨限值後，實施藉由缺陷檢測部 8之缺陷檢測。 1137l5.doc -20· 1336778 在此’檢測臨限值計算部7亦可僅根據上述分布資訊， 而決定檢測臨限值或上述缺陷檢測參數；或是參考圖4遵 照上述檢測臨限值之算出方法先決定檢測臨限值之初期 值，依據分布資訊將之修正亦可。 此時，如與某一巨集區域有關之分布資訊，被作為如下1336778 IX. Description of the invention: [Technical field to which the invention pertains] The present invention relates to an image defect inspection device, an image defect inspection system, and an image defect inspection method, which compares an inspection image after an image of an inspection object and The inspection image should be the same reference image, and the different portions are detected as defects; and the image defect inspection device, the image defect inspection system, and the image defect inspection method are specifically In the semiconductor manufacturing step, the defect of the semiconductor # circuit pattern formed on the semiconductor wafer is detected, and the surface of the semiconductor circuit wafer is imaged, the captured image is compared with the reference image, and the different portions are detected as defects By. [Prior Art] The object of the present invention is an image defect inspection device, an image defect inspection system, and an image defect inspection method; the inspection image after the inspection object is imaged and the same reference as the inspection image The images are compared, and the different parts are detected as defects. Here, an inspection apparatus is taken as an example, but the invention is not limited thereto, and the visual inspection apparatus detects a semiconductor circuit pattern formed on a semiconductor wafer in a semiconductor manufacturing step. Defective. A general visual inspection device is a bright-field inspection device that illuminates a surface of a subject in a vertical direction to capture an image of the reflected light. However, a dark field inspection device that does not directly capture the illumination light is also used. In the case of a dark field inspection device, a sensor that illuminates the surface of the object from an oblique direction or a vertical direction without detecting a specular reflection is configured by sequentially scanning the illumination position of the illumination light 113715.doc !336778 to obtain an object. The dark field of view of the surface. Therefore, in the case of the dark field device, there is also a case where the image sensor is not used, which is of course an object of the invention. As described above, if the image processing method and apparatus for determining the difference between the two images (signals) that are to be identical and the difference is determined as the defect, no matter what image processing method and The device and the present invention are applicable. In terms of semiconductor fabrication steps, a plurality of wafers (dies) are formed on the semiconductor wafer. A pattern of a plurality of layers is repeatedly formed on each of the crystal grains. The completed die is electrically inspected by a probe and a tester to remove defective grains from the assembly step. The yield is quite important in the semiconductor manufacturing steps, and the results of the above electrical inspection are fed back to the manufacturing steps and used in the processing of each step. However, the semiconductor manufacturing step is composed of a plurality of steps, and it takes a long time from the start of manufacture to the implementation of the electrical inspection. Therefore, when it is detected by the electrical inspection that there is a problem in the step, it is already in a plurality of In the middle of wafer processing, the inspection results cannot be fully utilized to improve the yield. For the above reasons, a pattern defect inspection is performed, which is a pattern formed by the steps in the inspection, and the defect is detected. If the pattern defect inspection is performed in a plurality of steps in all the steps, the defects generated after the previous inspection can be detected, and the inspection results are quickly reflected in the step management. Figure 1 is a block diagram of an appearance inspection device, which is claimed by the applicant of the present invention in Japanese Patent Application No. 20-38-108209 (the following patent document). As shown in the figure, it can be in 2 dimensions or A sample stage (suction tray stage) 2 is disposed on the upper surface of the stage in which the three-dimensional direction is freely movable. On the sample stage, 1137l5.doc 1336778 is placed on the semiconductor wafer 3 to be inspected and placed thereon. In the upper part of the stage, an imaging device 4 configured by using a 1- or 2-dimensional CCD camera or the like is provided; the imaging device 4 generates an image signal of a pattern, and the pattern is formed on the semiconductor wafer 3. As shown in Fig. 2, on the semiconductor wafer 3, a plurality of crystal grains 3a are arranged in a matrix in the X direction and the Y direction, respectively. Since the crystal grains are formed in the same pattern, they are generally adjacent. The images of the corresponding portions of the crystal grains are compared. If the two crystal grains have no defects, the gray scale difference is small; if one of the defects is defective, the gray scale difference is larger than the threshold value, and therefore, This grayscale difference is compared to a specific detection threshold, and will The trap is detected (single detection), but the light does not know which one of the crystal grains is defective. Therefore, the comparison of the crystal grains adjacent to the different sides is further performed. If the gray level difference of the same portion is greater than the threshold value, it is known. The crystal grain is defective (double detection). The image pickup device 4 includes a ccd camera realized by a TDI image sensor or the like; the moving stage 1 is such that the camera has a certain speed toward the semiconductor wafer 3 in the χ or Y direction. Relatively moving (scanning). The image signal 80 after being imaged by the following strip-shaped area (referred to as swath) is rotated from the imaging device 4 by scanning in the main scanning direction. The image signal is converted into a multi-valued digitized signal (grayscale signal) and stored in the image memory unit 5; and the strip-shaped region is compared with the image pickup device shown by the dotted line in FIG. 3(A). The imaging width ws of 4 is the same width. After the eight-period performance is scanned, when the gray-scale signal (inspection image signal) is generated from the die 3a to the adjacent die 3b, the difference detecting section 6 is a slave image. The memory unit 5 reads the following information 'and The wheel-to-human differential detection unit 6; M3715.doc 1336778 This data is a small portion of images 81 and 82 of the same portion of the two adjacent crystal grains 3& and 31) as shown in Fig. 3(B) (logic Grayscale signal (reference image signal) of the frame). In fact, there is a fine-grained comparison process, but it will not be described in detail here. The difference detecting unit 6 receives the gray scale signals of the partial images 81 and 82 of the same portion of the adjacent two crystal grains 3a & 3b, and uses one of them as the inspection portion image, and uses the other as the reference image to The difference (gray difference) of the gray scale numbers of the corresponding pixels is calculated and output to the detection threshold calculation unit 7 and the defect detection unit 8. The detection threshold calculation unit 7 outputs the automatic detection detection threshold value based on the gray scale difference, and outputs it to the defect detecting unit 8. The defect detecting unit 8 compares the threshold values determined by the gray level difference, and determines whether or not the defect is missing. Next, the defect detecting unit 8 outputs defect information according to each defect for the portion determined to be defective, and includes: the position of the defect, the grayscale difference, the detection threshold at the time of detection, and the determination of the detection threshold. Defect detection parameters used at the time. After that, A's more detailed inspection of the part determined to be defective, the defect information is supplied to the automatic defect classification (ADC) device (not shown in the automatic defect classification device, etc. : Whether the part of the defect is a real defect affecting the yield; or a suspected defect caused by the noise of the captured image, etc.; or determining the defect (wiring short circuit 'pattern defect or Particles, etc. (>, '" In this defect classification process, it is necessary to perform a detailed inspection because it is necessary to perform a detailed inspection. Therefore, in order to determine the defect, it is necessary to meet the true defect. Missing and except for the real defect = U3715.doc 1336778 The requirement for the defect is determined. Based on this, the setting of the threshold is a big problem. If the threshold is set smaller, the pixel is determined to be defective ( Increased pixel, so that the part of the non-true defect is also judged as a defect; the result will lead to the problem that the time required for the defect grading becomes longer. In contrast, if the threshold is set If it is too large, the problem that the true defect is also judged to be non-defective, and the inspection is not complete is inspected by the applicant of the present invention in Japanese Patent Application No. 2003-188209 (Patent Document 1 below) The threshold value calculation unit 7 determines the detection threshold value based on the distribution of the gray scale differences between the corresponding pixels of the inspection portion image and the reference image. When the contrasted partial detection image is input from the difference detection portion 6 The grayscale difference signals of the respective pixels included in 1 and 82 are subjected to a histogram as shown in Fig. 4(A), and then the cumulative frequency as shown in Fig. 4(B) is calculated. 'The distribution of the gray-scale difference is calculated according to the assumption of the specific distribution, and the conversion cumulative frequency is calculated, and the system has been transformed so that the cumulative frequency has a linear relationship with the gray-scale difference (refer to FIG. 4(C)). Then, 'the approximate straight line of the cumulative frequency of the transformation is calculated, and based on the approximate straight line calculated therefrom, the threshold value is determined from the value of the specific cumulative frequency according to the specific calculation method'. For example, in the example of Fig. 4(C) , assuming the approximate curve calculated The intercept is a, the intercept of the approximate curve of the vertical axis (that is, the cumulative frequency at which the gray-scale difference becomes 0) is b (it can be said that the slope a and the intercept b are the defect detection parameters), and the threshold T It is calculated by the following formula (1): T = (Pl - b + VOP) / a + HO (1) 1137l5. doc I336778 Here, pi corresponds to the cumulative frequency of the specific cumulative probability (p); VOP, [Patent Document 1] Japanese Laid-Open Patent Publication No. Hei. No. Hei. No. Hei. No. Hei. No. Hei. No. Hei. 4] Japanese Laid-Open Patent Publication No. 2002-22421 [Problems to be Solved by the Invention] In the actual inspection image after imaging the surface of a semiconductor wafer, even a wafer formed by the same process is imaged in the same state. After the inspection image, there may be a phenomenon of "color unevenness", which is an image of different brightness (grayscale values); or an area that should be photographed with the same brightness even in one wafer. It was filmed with different brightness. As described above, if the defect inspection is performed using the inspection image of the color unevenness, the grayscale difference is also increased even in the original non-defective portion, and the defect detection is caused to cause an increase in the suspected defect. However, when the image of the semiconductor wafer surface is uneven in color, the difference in the position of the brightness on the surface of the wafer is different from the size of the image unit (logical frame) for performing the defect inspection. The change is more moderated. Therefore, even if the change of the grayscale value in one logical frame is observed, the presence or absence of color unevenness cannot be detected. In view of the above-mentioned problems to be solved, 'the object of the present invention is to reduce the suspected defect caused by the color unevenness generated by the inspection image on the image defect inspection; the image defect inspection system will detect the object after the inspection The inspection image, and the inspection image to be compared with the original reference image, are compared, and the different portions of 113715.doc 1336778 are detected as defects. SUMMARY OF THE INVENTION In order to achieve the above object, in the present invention, according to the regions of the temples on the inspection object, the pixel values of the pixels of the image a obtained by the respective regions are determined. The reference value is determined to be a macro region composed of a plurality of regions of the specific size, and the distribution information of the reference values determined by the respective regions of the specific size included in the macro region is determined. The area changes the defect detection condition based on the determined distribution information, and performs defect detection. Further, in the present invention, each region of a specific size on the inspection target is subjected to measurement of a specific measurement value relating to the inspection target in each of the regions, and a reference value based on the measured specific measurement value is used. Determining separately; the macro region formed by the plurality of regions of the specific size is determined based on the distribution information of the reference values determined by the respective regions of the specific size included in the macro region, and is determined according to the macro region. Defect detection conditions are changed by the determined distribution information, and defect detection is performed. [Effects of the Invention] The reference value determined based on the image value of the image obtained from each region of the specific size on the inspection target and the measurement value of the inspection target is calculated by calculating a macro larger than the region of the specific size. The distribution information in the area can detect the color \ unevenness generated by the wide range of the inspection image after the inspection object is imaged. Then, the defect detection condition is changed according to the calculated distribution information, and the defect detection is performed; thereby, the color defect caused by the inspection image can be reduced, and the suspected defect generated by U3715.doc 11 1336778 can be reduced. [Embodiment] Hereinafter, embodiments of the invention will be described with reference to the accompanying drawings. Fig. 5 is a block diagram showing an i-th embodiment of an image defect inspection apparatus for a semiconductor circuit of the present invention. The image defect inspection apparatus 1 shown in Fig. 5 has a configuration similar to that of the prior art visual inspection apparatus 1 described above with reference to Fig. 1. Therefore, the same components are denoted by the same reference numerals, and the description thereof will be omitted. By moving the carrier 承载 of the inspection object (semiconductor wafer 3), the imaging device 4 scans the semiconductor wafer 3 at a constant speed in the χ or γ direction, by one time in the main scanning. Scanning of the direction, an inspection image obtained by imaging the following strip-shaped area is obtained, and the inspection image signal is converted into a multi-valued digital signal (gray scale signal) and stored in the image memory unit 5; The imaging device 4 includes an i-dimensional CCD camera realized by a TDI imaging element or the like; and the band-shaped region is the same width as the imaging width Ws of the imaging device 4 displayed by a broken line in FIG. 3(A). . Referring back to FIG. 5, if the scanning is continued, a gray scale signal is generated from the die 3a to the adjacent die 3b (the inspection image signal p difference detecting unit 6 is read from the image memory unit 5). The following information is input to the difference detecting unit 6, which is: two crystal grains 3 & and 31 adjacent to each other as shown in Fig. 3 (B) (small partial images 8 and 82 of the same portion) Gray scale signal (reference image signal) of the logic frame. The difference detecting unit 6 is input to the same portion of the two adjacent crystal grains 3a and 3b. 卩 is the gray-scale signal of the images 81 and 82' One side is used as a part of the inspection image, and the other side is used as a reference image, and the difference (gray scale difference) between the corresponding pixels of the respective pixels of the I137l5.doc 12 1336778 is shifted and is output to the detection threshold. The calculation unit 7 and the defect detection unit 8. Here, the partial image 81 is a strip-shaped inspection image 80 (band) having a width Ws indicated by a broken line in Fig. 6, such as a one-dot chain line. The partial image 81 is divided into a plurality of patents based on the reason. In addition, the partial image 82 is compared with the partial image 81 to be a reference image for comparison with the inspection partial image. Furthermore, the partial image 81 is checked. The object (semiconductor wafer 3) is imaged by each region of a specific size, and is used as an area of a specific size on the inspection object related to the scope of the patent application of the present invention. The image after the image is also cut out by the partial image (4). Therefore, it is an image block 0 related to the patent scope of the invention. In the image defect inspection apparatus, the square of the logical frame of the two adjacent crystal grains is used as a reference image. However, the ideal image of the wafer 3 in the past is taken. The sample is used to replace the former, and can be used as a reference image; in addition, two samples of the image of the old crystal image, for example, a sample corresponding to the average pixel for each pixel (the illustrated gold image) is used as a reference image. Sri can be used as well; or from The figure formed by the surface of the conductor 曰 yen 3 is also used. The β 资料 data (for example, CAD data) 使用 will be used as a reference image by simulating the image produced by the ideal method. I13715.doc -13* 1336778 The detection threshold calculation unit 7 automatically determines the detection threshold of the defect detection condition based on the distribution of the gray scale difference, and outputs it to the defect detection unit 8. By detecting the threshold calculation unit The detection threshold value of 7 is determined by referring to FIG. 4(A) to FIG. 4(C), and each of the two partial images 8 1 and 82 that are compared is input from the difference detecting unit 6 . The grayscale difference signal of the pixel is made into a histogram (refer to FIG. 4(A)), and then 'the cumulative frequency is calculated (refer to FIG. 4(B)); the distribution of the grayscale difference input is in accordance with the assumption of the specific distribution Next, the conversion cumulative frequency ' is calculated and the system has been transformed so that the cumulative frequency has a linear relationship with the gray level difference (refer to FIG. 4(C)). Then, an approximate straight line of the cumulative frequency of the conversion is calculated, and an approximate straight line calculated based on this (that is, a defect detection parameter is calculated. The slope a and the intercept b of the approximate curve) are calculated, and the approximate straight line calculated therefrom is used for the specific cumulative frequency. The value is determined by the specific calculation method described above. The defect detecting unit 8 compares the threshold values of the determined gray scale differences and determines whether or not it is a defect. Further, the defect detecting unit 8 outputs defect information in accordance with each defect for the portion which has been determined to be defective, and includes the position of the defect, the gray level difference, the detection threshold value at the time of detection, and the defect detection parameter. The image defect inspection apparatus 10 includes a reference value determining unit 21 for each partial image (logical frame) 81, and based on the pixel values of the pixels included in the partial images, the reference is determined in accordance with the specific determination method. The values are determined separately; as shown in Fig. 6, the partial image 81 is also an image block into which the inspection image (10) is divided into a specific size. The reference value determining unit 21 may, for example, determine the average value, the dispersion value, the maximum value, or the minimum value of the image block (each partial image 8) including the reference value. Alternatively, the reference value determining unit 21 may determine the reference value as follows: for example, the average value, the dispersion value, the maximum value, and the minimum value of all the pixels included in each partial image 81; or such maximum _ value or difference . The reference value determination unit 21 may determine the reference value as follows: among all the pixels included in the image block (each partial image 81), it exists in the partial image 81. The average value, dispersion value, maximum value, and minimum value of the pixel values of the pixels of a particular range; or the inter-value or difference between the maximum and minimum values. Alternatively, the pixel value of the pixel existing at a specific position in the partial image 81 may be determined as a reference value. Further, the 'reference value determining unit 21' can determine as a reference value: all the pixels included in the difference image between the partial image 81 and the reference image to be compared in the difference detecting portion 6 Average, scatter value, maximum value j, value, or inter-order value or difference between these maximum and minimum values. Or determining whether the value of the pixel value of the pixel value of the pixel in the specific range in the difference image is a large value, a minimum value, or an intermediate value or a difference between the maximum value and the minimum value, and is determined as a reference value. Also. Returning to FIG. #, the image defect inspection apparatus 1 includes a distribution information determining unit 22' for a macro region including a plurality of image blocks (in the above-described example partial image 81), The distribution information of the reference value determined according to each image block included in the macro region is calculated. The image block is the unit of the reference value. Here, the above-mentioned macro region can be set as follows: for example, in FIG. 7, an area including a plurality of partial images 8A as indicated by a two-point 113715.doc ^36778 scribe line. The distribution information determining unit 22 sets the distribution information relating to a certain macro region as a reference value for each of the plurality of image blocks (that is, part of the image) included in the macro region. Determined by the collection. Further, among the plurality of reference values, 'the value of the reference values which are predetermined to be the same value may be determined as the dispersion value of each other. Further, the 'image defect inspection apparatus 1' further includes a defect output availability determination unit 23' that resets the defect detection conditions in the respective macro regions based on the distribution information determined in each macro region, and resets them. In the subsequent defect detection condition, it is determined whether or not the output is determined by the defect detected by the defect detecting unit 8 in each macro region. The defect output availability determining unit 23 inputs, for example, the defect information related to each defect detected by the defect detecting unit 8, and the detection threshold included in the defect sfl of each defect is based on the macro set for the defect. The distribution information determined by the area is corrected by 're-setting; the gray-scale difference of the defect included in the defect information of the defect is compared with the reset detection threshold, if the gray-scale difference exceeds the detection threshold In the case of a limit, the defect information is allowed to be output as a true defect. In contrast, if the gray level difference does not exceed the detection threshold, the defect information is regarded as a suspected defect and the output is prohibited. Further, the defect output determination unit 23 determines the defect detection parameters (for example, the slope a and the intercept b of the approximate straight line) included in the defect information of each defect, depending on the macro region to which the defect belongs. The distribution information is corrected, and the detection threshold is reset by the corrected approximation curve. 113117.doc •16· 1336778 The gray level difference of the defect contained in the defect information of the defect and the detection after resetting Ba The limits are compared to determine the availability of the defect information output. As described above, each defect detected by the defect detecting unit 8 is re-detected under the defect detecting condition after the defect output possibility determining unit 23 resets, thereby performing defect detection based on the distribution information determined for each macro region. Changes in conditions. When the distribution information relating to a certain macro region is determined as a set of reference values as follows, the defect output possibility determination unit 23 resets the defect detection condition so that the defect detection sensitivity is included in the distribution information. Each of the reference values is determined to be reduced by a difference in the reference values of the same value, and the reference value is determined by the distribution information determining unit 22 for each image block included in the macro region. . For example, the detection threshold is set again so as to increase as the reference values which are predetermined to be the same value increase with each other. The defect output possibility determination section 23 may, for example, use the following reference value as a reference value of the above-mentioned predetermined same value; and the reference value relates to a plurality of wafers (die) 3a for respective portions of the same portion in the wafer Image 8 1 is determined separately. Then, the detection defect condition can be reset again so that the defect detection sensitivity is lowered as the reference values are unevenly increased. When the distribution information relating to a certain macro region is determined as the dispersion value of the reference values of the same value for the same value among the following reference values, the defect output availability determination unit 23 will detect the defect condition again. 113715.doc 17 1336778 is set such that the defect detection sensitivity decreases as the distribution information increases; and the reference value is determined by the distribution information determining unit 22 for each image block included in the macro region. . In the above description and the following description, the unit of the image block in which the unit of the reference value (inspection image) is divided is used as the unit for performing the image comparison of the difference detecting unit 6 (partial) Image, ie logical frame); In addition to this, the size (unit) of the image block can be determined freely. For example, the image block can take an image of one of the wafers (grains) 3a formed on the surface of the semiconductor wafer as a unit (that is, an image obtained by dividing the image of one wafer 3a). As the H-solid image block, the unit of the image block is determined. * When the color difference of the lightness difference is moderately changed, it is necessary to obtain the distribution information in the wide range of the inspection image, but by If a larger image block is set as described above, the calculation amount of the reference value can be saved. As described above, if each image block is determined by one image of one of the wafers, the image images obtained by imaging each wafer are predetermined to be the same, so that the image blocks can be determined. The image block becomes the same image, or the reference value can be determined by making the reference value determined for each image block the same. Thus, the reference value can be predetermined to be the same value. Furthermore, when the pattern formed on the surface of the semiconductor wafer of the inspection object is formed by an exposure step (lithography step) of optical or electron beam, when the pattern is exposed on the surface of the semiconductor wafer, It is possible that the exposure pattern on the surface of the wafer is exposed in an unimaged state (out of focus state) to cause defective products. In the case of a wafer that is exposed and patterned in such a defocused state, in the case of a wafer, a color unevenness can also be observed on a captured image of the surface. This color is uneven, which is the range in which the exposure is taken for the first time (that is, the range of simultaneous exposure with one exposure mask), and the difference is produced by the difference in brightness. Therefore, When the second reticle is used to expose a plurality of wafers (grains), it is necessary to obtain distribution information in a wider range. Based on this, the reference value determining unit 21 captures an image of the exposure range of the inspection image after the inspection target image (the surface of the semiconductor wafer on which the lithography step is patterned) by one time, as a map. It is also possible to decide like a block unit. As described above, if each image block is determined by taking each of the images in the range of exposure by the reticle of one time, the image of the range exposed by the reticle of one time is predetermined to be the same. Therefore, the image block may be determined such that each image block is the same image, or the reference value may be determined so that the reference value determined for each image block is the same, and thus the reference may be predetermined. The value is the same value. Fig. 8 is a block diagram showing a second embodiment of the image defect inspection apparatus for a semiconductor circuit of the present invention. In the image defect inspection apparatus 10 according to the present embodiment, the reference value determining unit 21 determines the detection threshold value of the defect detection condition as a reference value for each image block; and the defect detection condition is The user in the individual image block is detected by the defect. As described above, the detection threshold calculation unit 7 determines the detection threshold based on the distribution of the grayscale difference between the pixel value of each pixel included in each inspection portion image 81 and the pixel value of each pixel of the reference image; Therefore, referring to the distribution state of the detection threshold, it is also possible to detect the color unevenness generated by the inspection image. 1137l5.doc -19- 1336778 Based on this, in the example shown in FIG. 8, the reference value determining unit 21 inputs the defect information detected in the image block (each inspection portion image 81) from the defect detecting unit 8. 'The detection threshold used in the detection of the defect contained in the individual defect information is determined as a reference value. Alternatively, the reference value determining unit 21 may determine the defect detection parameter used when determining the detection threshold for each image block (in the example of the detection threshold determination method described with reference to FIG. 4, an approximate curve) The slope a and the intercept b) are determined as reference values; and the detection threshold is the defect detection condition used for defect detection in individual image blocks. In the example of FIG. 8, the reference value determining unit 21 inputs the defect detecting parameter detected in the image block (each inspection portion image 8A) from the defect detecting unit 8, and the individual defect information is The defect detection parameter used in the detection of the defect is determined as a reference value. Fig. 9 is a block diagram showing a third embodiment of the image defect inspection apparatus for a semiconductor circuit of the present invention. In the embodiment shown in FIGS. 5 and 8, the detection threshold calculating unit 7 automatically calculates the detection threshold; the defect detecting unit 8 performs defect detection first under the determined detection threshold; Then, the defect output possibility determination unit 23 resets the detection threshold value, and performs re-determination of whether or not the defect output as the inspection result can be determined. In the image defect inspection apparatus 1 according to the present embodiment, before the defect detecting unit 8 performs the defect detection, the distribution information determining unit 22 first determines the above-described distribution information, and based on the distribution information, detects the threshold calculating unit. 7 After the detection threshold of the defect detection condition is determined, the defect detection by the defect detecting unit 8 is performed. 1137l5.doc -20· 1336778 Here, the detection threshold calculation unit 7 may determine the detection threshold or the defect detection parameter based only on the distribution information; or refer to FIG. 4 to calculate the detection threshold. The method first determines the initial value of the detection threshold, and corrects it according to the distribution information. At this time, the distribution information related to a certain macro area is as follows
參考值之集合而被決定之情形時，檢測臨限值計算部7係 在分布資訊所含之各參考值之中’進行設定缺陷檢測條 件，以使缺陷檢測感度隨著預定為本來相同值之參考值彼 此之參差不齊的增大而降低；而該參考值係藉由分布資訊 決定部22依照此巨集區域所含之各圖像區塊而決定者。譬 如’將檢測臨限值進行再設定，以使之隨著此參差不齊的 增大而增大。又’如與某一巨集區域有關之分布資訊，在 如下各參考值之中，被作為預定為本來相同值之參考值彼 此之分散值而被決定之情形時，檢測臨限值計算部7係進 行設定缺陷檢測條件，以使缺陷檢測感度隨著分布資訊的 增大而降低，而该各參考值係藉由分布資訊決定部依照 此巨集區域所含之各圖像區塊而決定者。 在此，如決定參考值之一單位（圖像區塊），被作為如下 圖像區塊而決定之情料，可將如下參考值，作為上述預 定為本來相同值之參考值；而該圖像區塊係如檢查部分圖 像（邏輯圖框）般，係具有比將1個晶片攝像後之像素數為小 之像素數者；而該參考值係關於複數個晶片（晶粒）3a，針 對將a曰片内之相同部份攝像後之個別之圖像區塊分別決定 者。 、 U3715.doc -21 - 丄划6778 又’如圖像區塊係就將1個晶片攝像後之各圖像、或就 將1次之標線拍攝所曝光之範圍之圖像之一攝像後之各圓 像而破決定 < 情形日夺，由於可決定區塊使各圖$區塊成為 本來相同之圖像，或是可決定其決定方法使針對各圖像區 塊所決定之參考值成為本來相同值，因而針對各圖像區塊 分別決定之參考值可預定為相同值。 藉由分布資訊決定部22所應取得之分布資訊，係顯示檢 查對象在較廣區域中之明度差的分布者，因此，參考值決 疋部21用於決定參考值所需之圖像解析度，係比缺陷檢測 邛8用於進行缺陷檢測所需之圖像解析度為低即足夠◊基 於此因，參考值決定部21在決定參考值之際所使用之攝像 圖像，可為與缺陷檢測部8在進行缺陷檢測之際所使用之 檢查圖像為不同之圖像。 亦即，參考值決定部2丨可依據半導體晶圓3之攝像圖像 之像素之像素值而決定參考值，而該攝像圖像係與缺陷檢 測部8在進行缺陷檢測之際所使用之檢查圖像為不同者。 s如，在圖9之例中，在將缺陷檢測部8在進行缺陷檢測 之際所使用之檢查圖像記憶於第丨圖像記憶部5的同時，將 參考值決定部21在決定參考值之際所使用之攝像圖像記憶 於第2圖像記憶部25。再者，第2圖像記憶部25係從攝像裝 置4所輸出之圖像信號中，進行記憶以特定之間距將像素 間隔而设後之攝像圖像。或是不將第2圖像記憶部25與第i 圖像記憶部5作分別設置，參考值決定部21將記憶於第丄圖 像記憶部5之圖像信號以特定之間距將像素間隔而設輸入 1137l5.doc •22· 1336778 亦可。 又，譬如，在第2圖像記憶部25中，亦可使用攝像裝置 4，以比缺陷檢測部8作缺陷檢測時所使用之檢查圖像予以 攝影之更高速且低倍率進行攝像，並將該攝像圖像記憶》 圖10係本發明之半導體電路用之圖像缺陷檢查裝置之第 4實施例之區塊圖。與本發明有關之圖像缺陷檢查裝置1〇 在第1攝像裝置4之外，另包含第2圖像記憶部24，其係將 參考值決定部2 1在決定參考值之際所使用之攝像圖像予以 攝像者；而第1攝像裝置4將缺陷檢測部8在進行缺陷檢測 之際所使用之檢查圖像予以攝像者。 亦即’參考值決定部21係依據藉由第2攝像裝置24所攝 像之攝像圖像所含之像素之像素值，進行決定參考值；在 獲得檢查圖像之第1攝像裝置4之外，第2攝像裝置24係用 於將半導體晶圓3攝像而設置者。 藉由在將檢查圖像予以攝像之第丨攝像裝置4之外，另設 置第2攝像裝置24，可容易構成圖像缺陷檢查裝置1〇，使 參考值決定部21、分布資訊決定部22係與藉由缺陷檢測部 8之缺陷檢測平行，分別進行決定參考值及分布資訊。再 者，基於上述理由，第2攝像裝置24亦可為以比第i攝像裝 置4更低解析度或更低倍率進行攝像之攝像裝置。 圖11係本發明之半導體電路用之圖像缺陷檢查系統之第 1實施例之區塊圖。參考值決定部21及分布資訊決定部22 用於为別進行決定參考值及分布資訊所使用之攝像圖像， 無需限定於藉由設置於圖像缺陷檢查裝置1〇之攝像裝置所 113715.doc •23· 1336778 攝像者。 亦即，在以圖像缺陷檢查裝置10進行半導體晶圓3之外 觀檢查之前，參考值決定部21及分布資訊決定部22亦可使 用藉由攝像裝置將半導體晶圓3攝像後之攝像圖像，分別 決定上述參考值及上述分布資訊；而該攝像裝置係有別於 設置於圖像缺陷檢查裝置1〇之攝像裝置，另作準備者。 基於此因，圖11所示之圖像缺陷檢查系統包含圖像缺陷 檢查裝置10 ;及在此圖像缺陷檢查裝置10之外，另包含取 得半導體晶圓3之攝像圖像之攝像裝置50 ;而該攝像圖像 係用於決定上述參考值及上述分布資訊之用者，而該參考 值及該分布資訊係圖像缺陷檢查裝置丨〇之檢測臨限值計算 部7進行決定其缺陷檢測條件之際所使用者。 攝像裝置50包含將檢查對象（半導體晶圓3)攝像之攝像 裝置51，及用於將藉由攝像裝置51所攝像之攝像圖像等之 資料輸出至圖像缺陷檢查裝置10之資料輸出部52。 另一方面，在圖像缺陷檢查裝置10中係設有用於將攝像 裝置50所輸出之攝像圖像資料輸入之資料輸入部％。攝像 裝置50侧之資料輸出部52、與圖像缺陷檢查裝置1〇側之資 料輸入部26之間之資料往返，可以經由有線或無線之信號 傳達路徑之連線方式進行，或是藉由可撓式光碟、CD_ ROM、可移動媒體等之資訊記憶媒體之離線方式進行。 此外，設於圖像缺陷檢查裝置10之參考值決定部21，係 依照檢查對象（半導體晶圓3)上之各特定大小之區域，依據 攝像圖像所含之像素值’與上述相同般，分別進行決定參 113715.doc -24- 1336778 考值；而該攝像圖像係攝像裝置5〇將此各區域攝像而成 者。 又，分布資訊決定部22係針對在半導體晶圓3上被決定 而包含複數個上述特定大小之區域之巨集區域，決定參考 值之分布資訊，而該參考值係、依照此巨集區域所含之上述 各特定大小之區域而決定者。When the set of reference values is determined, the detection threshold calculation unit 7 performs a set defect detection condition among the reference values included in the distribution information so that the defect detection sensitivity is the same as the predetermined value. The reference values are decreased by increasing the jaggedness of each other; and the reference value is determined by the distribution information determining unit 22 in accordance with each image block included in the macro region.譬 For example, 'Reset the detection threshold to increase it as the jaggedness increases. In the case where the distribution information relating to a certain macro region is determined as the dispersion value of the reference values of the same value for the same value among the following reference values, the detection threshold calculation unit 7 Setting the defect detection condition so that the defect detection sensitivity decreases as the distribution information increases, and the reference values are determined by the distribution information determining unit according to each image block included in the macro region. . Here, if one unit (image block) of the reference value is determined as the image block as follows, the following reference value may be used as the reference value of the predetermined value which is predetermined to be the same; The image block system has a pixel number smaller than the number of pixels after imaging one wafer, and the reference value is related to a plurality of wafers (grains) 3a, as in the inspection of the partial image (logical frame). The individual image blocks after imaging the same portion in the a slice are determined separately. , U3715.doc -21 - 6 6778 and 'If the image block system is imaged after capturing one chip, or one of the images of the range exposed by the reticle The round image is determined to be <RTIgt; </ RTI> <RTIgt; </ RTI> <RTIgt; </ RTI> <RTIgt; </ RTI> <RTIgt; </ RTI> <RTIgt; </ RTI> <RTIgt; The original values are the same, and thus the reference values determined for each image block can be predetermined to be the same value. The distribution information to be acquired by the distribution information determining unit 22 displays the distribution of the brightness difference of the inspection target in a wide area. Therefore, the reference value decision unit 21 is used to determine the image resolution required for the reference value. It is sufficient that the image resolution required for the defect detection is low, and the image capturing image used by the reference value determining unit 21 when determining the reference value may be a defect. The inspection image used by the detecting unit 8 when performing defect detection is a different image. In other words, the reference value determining unit 2 determines the reference value based on the pixel value of the pixel of the captured image of the semiconductor wafer 3, and the captured image is checked by the defect detecting unit 8 when performing defect detection. The images are different. In the example of FIG. 9, the inspection image used when the defect detecting unit 8 performs the defect detection is stored in the second image storage unit 5, and the reference value determining unit 21 determines the reference value. The captured image used at the time is stored in the second image storage unit 25. Further, the second image storage unit 25 stores a captured image in which the pixels are spaced apart by a predetermined interval from the image signals output from the imaging device 4. Alternatively, the second image storage unit 25 and the i-th image storage unit 5 are not provided, and the reference value determination unit 21 divides the image signals stored in the second image storage unit 5 by a specific interval. Set the input 1137l5.doc •22· 1336778. Further, for example, in the second image storage unit 25, the imaging device 4 can be used to capture images at a higher speed and at a lower magnification than the inspection image used for the defect detection by the defect detecting unit 8 Fig. 10 is a block diagram showing a fourth embodiment of the image defect inspection apparatus for a semiconductor circuit of the present invention. The image defect inspection apparatus 1 according to the present invention includes a second image storage unit 24 in addition to the first imaging device 4, and is a camera used by the reference value determining unit 2 when determining a reference value. The image is imaged by the first imaging device 4, and the inspection image used by the defect detecting unit 8 when detecting the defect is imaged. In other words, the reference value determining unit 21 determines the reference value based on the pixel value of the pixel included in the captured image captured by the second imaging device 24, and the first imaging device 4 that obtains the inspection image, The second imaging device 24 is used to set the semiconductor wafer 3 to be imaged. By providing the second imaging device 24 in addition to the second imaging device 4 that images the inspection image, the image defect inspection device 1 can be easily configured, and the reference value determining unit 21 and the distribution information determining unit 22 can be configured. The reference value and the distribution information are determined separately in parallel with the defect detection by the defect detecting unit 8. Further, for the above reasons, the second imaging device 24 may be an imaging device that performs imaging at a lower resolution or lower magnification than the i-th imaging device 4. Figure 11 is a block diagram showing a first embodiment of an image defect inspection system for a semiconductor circuit of the present invention. The reference value determining unit 21 and the distribution information determining unit 22 are used for the image capturing image used for determining the reference value and the distribution information, and are not necessarily limited to the image capturing device provided by the image defect detecting device 1113. •23· 1336778 Photographer. In other words, before the visual inspection of the semiconductor wafer 3 by the image defect inspection apparatus 10, the reference value determining unit 21 and the distribution information determining unit 22 may use a captured image obtained by imaging the semiconductor wafer 3 by the imaging device. The reference value and the distribution information are respectively determined, and the imaging device is different from the imaging device provided in the image defect inspection device 1 and is prepared as a preparer. Based on this, the image defect inspection system shown in FIG. 11 includes an image defect inspection device 10; and an image pickup device 50 that acquires a captured image of the semiconductor wafer 3 in addition to the image defect inspection device 10; The captured image is used to determine the reference value and the distribution information, and the reference value and the distribution information are detected by the detection threshold calculation unit 7 of the image defect inspection device 决定 to determine the defect detection condition. The user at the time. The imaging device 50 includes an imaging device 51 that images an inspection target (semiconductor wafer 3), and a material output portion 52 that outputs information such as a captured image captured by the imaging device 51 to the image defect inspection device 10. . On the other hand, the image defect inspection device 10 is provided with a data input unit % for inputting captured image data output from the imaging device 50. The data output unit 52 on the imaging device 50 side and the data input unit 26 on the side of the image defect inspection device 1 can be reciprocated via a wired or wireless signal transmission path, or by Offline operation of information memory media such as flexo discs, CD_ROMs, and removable media. Further, the reference value determining unit 21 provided in the image defect inspection device 10 is the same as the above in accordance with the pixel value included in the captured image in accordance with the region of each specific size on the inspection target (semiconductor wafer 3). The decision value of 113715.doc -24 - 1336778 is separately determined; and the camera image is obtained by imaging the respective regions. Further, the distribution information determining unit 22 determines the distribution information of the reference value for the macro region including the plurality of regions of the specific size determined on the semiconductor wafer 3, and the reference value is based on the macro region. It is determined by the above-mentioned specific size regions.
此外，檢測臨限值計算部7係依據在巨集區域所決定之 刀布資訊，將缺陷檢測條件之檢測臨限值予以改變。此 時，檢測臨限值計算部7係改變檢測臨限值，以使缺陷檢 測感度隨著在分布資訊所含之各參考值内狀為本來相同 值之參考值彼此之參差不齊的增大而降低。譬#，改變檢 測6«限值，使之隨著此參差不齊的增大而增大。 在此，決定參考值之一單位（上述特定大小之區域广如 被作為比晶片1個更小之區域而決定之情形時，亦可將如 下參考值’作為上述預定為本來相同值之參考值；而該參 考值係關於複數個晶片（晶粒）3a，針㈣晶片内之相同部 分攝像後之圖像分別決定者H上料定大小之區域 係就1個晶片之各範圍、或就以卜欠標線拍攝所曝光之各範 圍而被決；t之情形時’由於可決定上述特定大小之區域， 使將上述特定大小之區域攝像後之圖像成為本來相同之圖 像，或是可決定其決定方法使針對各區域所決定之參考值 成為本來相同值，因而此種參考值可預定為相同值。 圖12係本發明之半導體電路用之圖像缺陷檢 2實施例之區塊圖。在本實施例中 查系統之第 參考值決定部5 3係設 113715.doc -25- 1336778 置於攝像裝置50侧；而參考錢定部53係就檢查對象（半 導體晶圓3)上之特定大小之各區域’依據攝像裝置η將此 各區域攝像後之攝像@像所+的像素值而分別決定參考值 者。 接著，資料輸出部52係以參考值資料取代攝像圖像資 料，將之輸出至圖像缺陷檢查裝置1〇，圖像缺陷檢查裝置 10側之資料輸入部26係將之輸入。 圖像缺陷檢查裝置1G側之分布資訊決定部22係針對在半 導體晶圓3上被決定而包含複數個上述特定大小之區域之 巨集區域，就此巨集區域所含之上述特定大小之各區域， 進行決定攝像裝置50側之參考值決定部53所決定之參考值 之分布資訊。接著，參考_與上述所說明之方法相同， 檢測臨限值計算部7係依據已決定之分布資訊，將缺陷檢 測條件之檢測臨限值予以改變。 圖13係本發明之半導體電路狀圖像缺陷檢查系統之第 3實施例之區塊圖。在本實施例中，更進一步在攝像裝置 50側設置分布資訊決定部54，其係決定參考值之分布資訊 者，該參考值係㈣&含複數個特定大小之區域而構 成之巨集區域，就包含於此巨集區域之上述特定大小之各 區域而決定者。 接著，資料輸出部52將此分布資訊輸出至圖像缺陷檢查 裝置10，圖像缺陷檢查裝置10侧之資料輸入部26係將之輸 入0 參考圖11與上述所說明之方法相同，圖像缺陷檢查裝置 113715.doc -26· 1336778 ι〇側之檢測臨限值計算部7係依據從攝像裝置50輸入之分 布資訊，將缺陷檢測條件之檢測臨限值予以改變。 圖14係本發明之半導體電路用之圖像缺陷檢查系統之第 4實施例之區塊圖。 在將半導體晶圓3攝像後之檢查圖像所產生之顏色不 均，係藉由形成於半導體晶圓3表面之絕緣層等之透明或 半透明之膜的膜厚變動所產生者。於晶圓3之各部位進行 測定形成於半導體晶圓3表面之膜的膜厚，藉由將此測定 值作為參考值作成其分布資訊，亦可檢測顏色不均。 亦即，在以圖像缺陷檢查裝置1〇進行半導體晶圓3之外 觀檢查之前，參考值決定部21及分布資訊決定部22亦可使 用藉由膜厚測定裝置所測定之半導體晶圓3各部位之膜厚 =料，分職定上述參考值及上述分布f訊；而該膜厚測 疋裝置係在圖像缺陷檢查裝置1〇之外另準備者。 基於此因’圖14所不圖像缺陷檢查系統包含圖像缺陷檢 查裝置10’及測定形成於半導體晶圓3表面之膜厚之膜厚 值的膜厚測定裝置6G;而該膜厚之膜厚值係用於決定上述 參考值及上述分布資訊之用纟，而該參考值及該分布資訊 係圖像缺陷檢查裝置1G之檢測臨限值計算部7進行決定其 缺陷檢測條件之際所使用者。 膜厚測A裝置60包含膜厚測定部61，其係測定形成於半 導體晶圓3表面之絕緣層等之透明或半透明之媒的膜厚 者；及資料輸出部62 ’其係用於將此臈厚測定資料輸出至 圖像缺陷檢查裝置10者。 113715.doc •27- 1336778 另一方面’在圖像缺陷檢查裝置10中設有資料輸入部 26，其係用於將從膜厚測定裝置6〇輸出之膜厚測定資料予 以輸入者。膜厚測定裝置60側之資料輸出部62、與圖像缺 陷檢查裝置1 0側之資料輸入部26之間的資料往返，可以上 述例示之連線方式或離線方式進行。 接著，設置於圖像缺陷檢查裝置1〇之參考值決定部21， 係就檢查對象（半導體晶圓3)上之各特定大小之區域，依據 膜厚測定裝置60在此各區域所測定之膜厚測定值，如上述 般分別決定參考值。 又，分布資訊決定部22係針對在半導體晶圓3上被決定 而包含複數個上述特定大小之區域之巨集區域，將就此巨 集區域所3之上述各特定大小之區域而決定之參考值之分 布資訊，予以決定。 接著，檢測臨限值計算部7係依據於巨集區域所決定之 分布資訊，將缺陷檢測條件之檢測臨限值予以改變。此 時，檢測臨限值計算部7係改變檢測臨限值，以使缺陷檢 測感度隨著在分布資訊所含之各參考值内狀為本來相同 值之參考值彼此之參差不齊的增大而降低。譬如，改變檢 測臨限值，使之隨著此參差不齊的增大而增大。 在此，如決定參考值之一單位（上述特定大小之區域）， 被作為比晶片1個更小之區域而決定之情形時，可將如下 參考值，作為上述被預定為本來相同值之參考I;而 考值係係關於複數個晶片（晶粒）3a，針對在晶片内之相同 部位所測定之測定值分別決定者。再者，如上述特定大小 U3715.doc -28- 1336778 之區域係就i個晶片之各範圍、或就丨次之標線拍攝所曝光 之各範圍而決定之情形時，由於可決定各區域中之測定部 位（譬如，在各晶片或各標線拍攝内之相同位置彼此進行 測定等），使在上述特定大小之區域内測定之測定值成為 本來相同之測定值，或是可決定其決定方法使針對各區域 所決定之參考值成為本來相同值，因而此種參考值可預定 為相同值。 圖15係本發明之半導體電路用之圖像缺陷檢查系統之第 5實施例之區塊圖。在本實施例中，係將參考值決定部63 設置於膜厚測定裝置60侧；參考值決定部63係就檢查對象 (半導體晶圓3)上之特定大小之各區域，依據膜厚測定值， 將參考值分別決定者；而該膜厚測定值係膜厚測定部6 i在 此各區域所測定者。 接著’資料輸出部62係以參考值資料取代膜厚測定值資 料，將之輸出至圖像缺陷檢查裝置10;圖像缺陷檢查裝置 10側之資料輸入部26係將之輸入。 圓像缺陷檢查裝置1 0側之分布資訊決定部22，係針對在 半導體晶圓3上被決定而包含複數個上述特定大小之區域 之巨集區域，就此巨集區域所含之上述特定大小之各區 域’將攝像裝置60側之參考值決定部63所決定之參考值之 分布資訊，予以決定。 圖像缺陷檢查裝置10側之檢測臨限值計算部7係依據已 決定之分布資訊，如參考圖14所說明之上述方法般，將缺 陷檢測條件之檢測臨限值予以改變。 1137I5.doc -29- 1336778 圖16係本發明之半導體電路用之圖像缺陷檢查系統之第 6實施例之區塊圖。在本實施例中，更進一步將分布資訊 決定部64設置於膜厚測定裝置6〇側；分布資訊決定部“係 針對包含複數個上述特定大小之區域而構成之巨集區域， 將就此巨集區域所含之上述各特定大小之區域而決定之參 考值之分布資訊，予以決定者。 接著，資料輸出部62係將此分布資訊輸出至圖像缺陷檢 查裝置ίο ’·圖像缺陷檢查裝置10側之資料輸入部26係將之 輸入。 圖像缺陷檢查裝置1〇側之檢測臨限值計算部7係依據從 膜厚測定裝置60輸入之分布資訊，如參考圖14所說明之上 述方法般，將缺陷檢測條件之檢測臨限值予以改變。 圖17係本發明之半導體電路用之圖像缺陷檢查系統之第 7實施例之區塊圖。如在將半導體晶圓3表面攝像後之攝像 圖像產生明度差（顏色不均）之情形時，則形成於半導體晶 圓3上之圖案之最小尺寸之臨界尺寸（criUcai 發生參差不齊，因此，以掃描型電子顯微鏡等測定形成於 半導體晶圓3表面之各部位之臨界尺寸的大小，藉由將此 測定值作為參考值作成其分部資訊，亦可檢測顏色不均。 亦即’在以圖像缺陷檢查裝置丨0進行半導體晶圓3之外 觀檢查之前，參考值決定部21及分布資訊決定部22亦可使 用藉由掃描型電子顯微鏡裝置所測定之半導體晶圓3各部 位之臨界尺寸測定值’分別決定上述參考值及上述分布資 訊，而該掃描型電子顯微鏡裝置係在圖像缺陷檢查裝置1〇 II3715.doc -30- 1336778 之外另準備者。 基於此因，圖17所示之圓像缺陷檢查系統包含圖像缺陷 檢查裝置10，及掃描型電子顯微鏡裝置7〇，其係測定形成 於半導體晶圓3表面之圖案之臨界尺寸者，而而該圖案之 臨界尺寸係用於決定上述參考值及上述分布資訊之用者， 而該參考值及該分布資訊係圖像缺陷檢查裝置丨〇之檢測臨 限值5十算部7進行決定其缺陷檢測條件之際所使用者。 掃描型電子顯微鏡裝置70包含：電子槍71，其係產生照 射於檢查對象（半導體晶圓3)之電子射束EB者；偏向器 72，其係用於將電子射束EB在半導體晶圓3上掃描者；電 子檢測器73，其係檢測在半導體晶圓3上反射之電子射束 EB者；信號處理電路74，其係將檢測電子射束EB之電子 檢測器73的電流強度信號變換為數位型式之強度信號者； 圖像產生部75，其係根據此強度信號及電子射束EB之掃描 位置，以產生半導體晶圓3表面之高倍率圖像者；CD測定 部76，其係將顯現於圖像產生部75所產生之圖像的圖案之 臨界尺寸，予以測定者；及資料輸出部77，其係用於將此 臨界尺寸測定值資料輸出至圖像缺陷檢查裝置1〇者。 另一方面，在圖像缺陷檢查裝置1〇中設有資料輸入部 26，其係用於將從膜厚測定裝置6〇輸出之臨界尺寸測定值 資料予以輸入者。掃描型電子顯微鏡裝置7〇侧之資料輸出 部77、與圖像缺陷檢查裝置1〇側之資料輸入部％之間的資 料往返，可以上述例示之連線方式或離線方式進行。 接著，設置於圖像缺陷檢查裝置1〇之參考值決定部以， 1137I5.doc -31 · 係就檢查對象（半導體晶圓3)上之各特定大小之區域，依據 掃描型電子顯微鏡70在此各區域所測定之臨界尺寸測定 值，如上述般分別決定參考值。 又，分布資訊決定部22係針對在半導體晶圓3上被決定 而包含複數個上述特定大小之區域之巨集區域，將就此巨 集區域所含之上述各特定大小之區域而決定之參考值之分 布資訊，予以決定。 接著’檢測臨限值計算部7係依據於巨集區域所決定之 分布資訊，將缺陷檢測條件之檢測臨限值予以改變。此 時’檢測臨限值計算部7係改變檢測臨限值，以使缺陷檢 測感度隨著在分布資訊所含之各參考值内預定為本來相同 值之參考值彼此之參差不齊的增大而降低。譬如，改變檢 測臨限值，使之隨著此參差不齊的增大而增大。 在此’如決定參考值之一單位（上述特定大小之區域）， 被作為比晶片1個更小之區域而決定之情形時，可將如下 參考值’作為上述被預定為本來相同值之參考值；而該參 考值係係關於複數個晶片（晶粒）3a，針對在晶片内之相同 部位所測定之測定值分別決定者。再者，如上述特定大小 之區域係就1個晶片之各範圍、或就1次之標線拍攝所曝光 之各範圍而決定之情形時，由於可決定各區域中之測定部 位（譬如’在各晶片或各標線拍攝内之相同位置彼此進行 測疋等）’使在上述特定大小之區域内測定之測定值成為 本來相同之測定值，或是可決定其決定方法使針對各區域 所決定之參考值成為本來相同值，因而此種參考值可預定 113715.doc 1336778 為相同值。 ‘ 圖18係本發明之半導體電路用之圖像缺陷檢查系統之第 • 8實施例之區㈣。在本實施例中，係將參考值決定部78 設置於掃描型電子顯微鏡裝置7〇;參考值決定部78係就檢 查對象(半導體晶圓3)上之各特定大小之區域，依據掃描型 電子顯微鏡裝置7 0在此各區域所測定之臨界尺寸測定值， 分別決定參考值者。 接著’資料輸出部77仙參考值㈣取代臨界尺寸測定 I資料，將之輸出至圖像缺陷檢查裝置ig;圖像缺陷檢查 裝置10側之資料輸入部26係將之輸入。 圖像缺陷檢查裝置10側之分布資訊決定部22，係針對在 半導體晶圓3上被決定而包含複數個上述特定大小之區域 的巨集區域，就此巨集區域所含之上述特定大小之各區 域’將掃描型電子顯微鏡裝置7〇側之參考值決定部78所決 定之參考值之分布資訊，予以決定。 ’圖19係本發明之半導體電路用之圖像缺陷檢查系統之第 9實施例之區塊圖。在本實施例中，更進一步將分布資訊 決定部79設置於掃描型電子顯微鏡裝置7〇側；分布資訊決 定部79係決定參考值之分布資訊者，而該參考值係針對包 含複數個上述特定大小之區域而構成之巨集區域，就此巨 集區域所含之上述特定大小之各區域而決定者。 接著，資料輸出部77係將此分布資訊輸出至圖像缺陷檢 查裝置10 ;圖像缺陷檢查裝置1〇側之資料輸入部26係將之 輸入0 113715.doc -33- 1336778 圖像缺陷檢查裝置ίο側之檢測臨限值計算部7係依據從 ‘ 膜厚測定裝置60輸入之分布資訊’將缺陷檢測條件之檢測 臨限值予以改變。 圖20係本發明之半導體電路用之圖像缺陷檢查裝置之第 5實施例之區塊圖。 在本實施例之圖像缺陷檢查裝置1〇中，首先，針對藉由 缺陷檢測部8所檢測之各個缺陷，缺陷資訊作成部9係以預 先決定之型式將缺陷資訊作成。 Φ 在此，係將缺陷資訊之型式先作決定，以包含如下與被 檢測之缺陷有關之資訊：檢測位置（缺陷位置）、於缺陷位 置之檢查部分圖像與基準圖像之灰階差、在檢查部分圖像 内含有缺陷位置之像素之部分圖像、在檢測缺陷之際所使 用之檢測臨限值或缺陷檢測參數。此等部分圖像、檢測臨 限值或缺陷檢測參數係可依據對應於檢查圖像8 〇上之各部 位（在此例中，係檢測缺陷之部位）而定之像素或圖像之像 素值予以決定；又，由於是依據對應於檢查圖像80上之各 • 部位而決定之像素或像素之像素值而變動之資訊，因此， 在以下之本發明專利說明書中，係以「像素值相關資訊」表 示。 接著，根據此像素值相關資訊，針對各缺陷資訊，決定 特定之參考值；進而決定各巨集區域中之參考值之分布資 訊，依據分布資訊，將各巨集區域中之缺陷檢測條件分別 再。X疋，並判疋在各巨集區域所檢測之各缺陷資訊之輸出 可否。 由於參考值分別反映檢查圖像上之複數部位的像素值， 113715.doc •34- 1336778 因此’如涵蓋檢查圖像之大約全區域作成缺陷資訊的話， 藉由決定具有某一程度寬度之巨集區域之參考值的分布資 訊，觀察某一程度寬度之範圍内之的像素值（灰階值）之變 動，則可檢測顏色不均的有無。 接著’將缺陷檢測條件再設定（如為分布參差不齊較大 之顏色不均較大的巨集區域，則設定缺陷檢測感度較低之 缺陷檢測條件；如為顏色不均較小的巨集區域，則設定缺 仏測感度較尚之缺陷檢測條件等），以隨著此分布資訊 而變化；藉由於再設定後之缺陷檢測條件下，判定缺陷資 訊之輸出的可否’則可防止因顏色不均所產生之疑似缺陷 之輸出。 以下’詳述本實施例之圖像缺陷檢查裝置1〇之各部的動 作。再者’本實施例之圖像缺陷檢查裝置1〇係具有等似參 考圖5之上述圖像缺陷檢查裝置之結構，因此，針對相同 之結構要素賦予相同元件符號，但與被賦予相同元件符號 之結構要素有關之動作，則省略其說明。 缺陷檢測部8係藉由比較差分檢測部6所輸出之灰階差及 檢測臨限值計算部7所輸出之檢測臨限值，而檢測缺陷； 將與檢測之缺陷有關之資訊輸出至缺陷資訊作成部9，而 °玄與檢測之缺陷有關之資訊係檢測位置（缺陷位置）、於缺 陷位置之檢查部分圖像與基準圖像之灰階差等。缺陷資訊 作成°卩9為了將所輸入之與缺陷有關之資訊輸出至其他裝 置’因此將包含此資訊之缺陷資訊，遵照預先決定之格 式’就各缺陷予以作成；而其他裝置係，利用與被檢測之 113715.doc -35- 1336778 缺陷有關之資訊的自動缺陷資訊分等裝置、顯示裝置、及 飼服器等裝置。 此缺陷資訊之格式可以如下方式予以定義：將為了使用 於自動缺陷資訊分等裝置之缺陷分等的各種資訊包含於缺 陷資訊中。譬如，在缺陷資訊中亦包含下列上述像素值相 關資訊：在檢查部分圖像81内含有缺陷位置之像素的部分 圖像、進行檢測缺陷之際所使用之檢測臨限值及缺陷檢測 參數。 可含於缺陷資訊中之像素值相關資訊，譬如可為：被檢 測缺陷資訊所示之缺陷的檢查部分圖像8丨；或在對應於檢 查部分圖像8 1之參考圖像82内，含有此缺陷位置之像素的 部分圖像。又，像素值相關資訊可為：檢測缺陷資訊所示 之缺陷的檢查部分圖像81 ;或對應於檢查部分圖像81之參 考圖像82。 再者’像素值相關資訊可為如下兩者之間的差圖像：被 檢測缺陷資訊所示之缺陷的檢查部分圖像8丨、對應於此檢 查部分圖像8 1之參考圖像82。像素值相關資訊可為如下兩 者之間的差圖像：在被檢測缺陷資訊所示之缺陷的檢查部 分圖像81内之含有缺陷位置之像素的部分圖像、在對應於 此檢查部分圖像8 1之參考圖像82内之含有缺陷位置之像素 的部分圖像。此外，如包含在檢測缺陷資訊所示之缺陷之 際所使用之檢測臨限值、缺陷檢測參數亦可。缺陷資訊作 成部9係將包含上述像素值相關資訊之缺陷資訊輸出至參 考值決定部2 1。 H3715.doc -36- 1336778 參考值決定部21係依據缺陷資訊所含之像素值相關資 訊，就各缺陷資訊（亦即，就各像素值相關資訊），進行決 定特定之參考值》 ~ 譬如，參考值決定部21在被提供檢測缺陷資訊所示之缺 陷的檢查部分圖像81、在對應於檢查部分圖像81之參考圖 像82内之含有缺陷位置之像素的部分圖像，作為像素值相 關資訊之情形時，可將此部分圖像之任一方或兩方所含之 像素值之平均值、分散值、最大值、最小值、或此等最大 值與最小值之中間值或差作為參考值而決定。或是參考值 決定部21可將下列兩者間之差圖像所含的像素值之平均 值、分散值、最大值、最小值、或此等最大值與最小值之 中間值或差，作為參考值而決定；而該兩者係：檢查部分 圖像81内之含有缺陷位置之像素的部分圖像、在對應於此 檢查部分圖像81之參考圖像82内之含有缺陷位置之像素的 部分圖像。 如被提供下列兩者之間的差圖像作為像素值相關資訊之 情形時，參考值決定部21可將此差圖像所含之像素值之平 均值、分散值、最大值、最小值、或此等最大值與最小值 之中間值或差’作為參考值而決定；而該兩者係：檢查部 分圖像81内之含有缺陷位置之像素的部分圖像、對應於此 檢查部分圖像81之參考圖像82内之含有缺陷位置之像素的 部分圖像。 如被提供檢測缺陷資訊所示之缺陷的檢查部分圖像8 j、 對應於此檢查部分圖像8 1之參考圖像82作為像素值相關資 113715.doc -37- 1336778 訊之情形時，可將如下所述者作為參考值而決定：此等圖 像之任一方或兩方之圖像内之特定位置之像素值；此等圖 像之任一方或兩方之圖像所含之像素值之平均值、分散 值、最大值、最小值、或此等最大值與最小值之中間值或 差。或是可將如下所述者作為參考值而決定：檢查部分圖 像81及對應於此檢查部分圖像81之參考圖像82之間之差圖 像内之特定位置之像素值；差圖像所含之像素值之平均 值、分散值、最大值、最小值、或此等最大值與最小值之 中間值或差。 如被提供檢測缺陷資訊所示之缺陷的檢查部分圖像8丨及 對應於此之參考圖像82之間的差圖像作為像素值相關資訊 之情形時，可將如下所述者作為參考值而決定：差圖像内 之特定位置之像素值；差圖像所含之像素值之平均值、分 放值、最大值、最小值、或此等最大值與最小值之中間值 或差。 如被提供檢測缺陷資訊所示之缺陷之際時所使用之檢測 臨限值、缺陷檢測參數作為像素值相關資訊之情形時，可 將此檢測臨限值、缺陷檢測參數作為參考值而決定。 在決定參考值之際，參考值決定部21可就各檢查部分圖 像8丨、或就各部分圖像，而決定丨個參考值；而該各部分 圖像係將檢查圖像就各特定之像素數分割而成者。在本發 明專利說明書中，有時將上述作為決定參考值之單位的部 分圖像稱為「參考值決定單位圖像」。譬如，在某一參考值 決定單位®像㈣測複數個缺陷之情料，參考值決定部 H37I5.doc •38- 1336778 21可將關於此等缺陷中任一個而決定之參考值作為代表 值，而決定各參考值決定單位圖像之參考值。或是可將關 於參考值決定單位圖像内之各個缺陷而決定之參考值彼此 之特定運算值’予以求出，作為各參考值決定單位圖像之 參考值。 回到圖20，分布資訊決定部22係針對特定大小之圖像區 域（巨集區域），將關於各缺陷而決定之參考值之分布資 訊，予以決定。而該特定大小之圖像區域係在將半導體晶 圓3攝像後之檢查圖像内被決定者；而該各缺陷係在此巨 集區域所含之檢查部分圖像81所檢測者。針對巨集區域之 說明係如上述例示所示。 为布資訊決定部22可就某巨集區域所含之複數個檢查部 分圖像8 1，將關於在此檢查部分圖像8丨上檢測之各缺陷之 任一個而決定之參考值之分散值，作為關於巨集區域之參 考值之分布資訊而決定。或是可將關於在此巨集區域所含 之各檢查部分圖像81上檢測之缺陷而決定之參考值之集 σ ’作為分布資訊而決定。 又，如上述般，在就各參考值決定單位圖像決定參考值 之If形時，係將巨集區域作為比1個參考值決定單位圖像 更大之區域而決定；將就此巨集區域中之各參考值決定單 位圖像而決定之參考值之分布資訊，予以決定。 此外’缺陷輸出可否判斷部23係依據針對巨集區域而決 定之分布資訊’進行再設定作為缺陷檢測條件之檢測臨限 值’於再設定後之缺陷檢測條件下，在此巨集區域所含之 113715.doc •39· 1336778 各檢查部分圖像8 1所檢測之缺陷資訊的輸出可否，予以判 定。 缺陷輸出可否判斷部23係譬如輸入與在缺陷檢測部8所 檢測之各缺陷有關之缺陷資訊，將個別之缺陷的缺陷資訊 所含之檢測臨限值’藉由依據針對此缺陷所屬之巨集區域 而決定之分布資訊作修正，而再設定；此外，將缺陷資訊 所含之此缺陷之灰階差與再設定後之檢測臨限值作比較， 如灰階差超過檢測臨限值之情形，則將該缺陷資訊作為真 正缺陷而允許輸出；相反的，如灰階差未超過檢測臨限值 之情形’則將該缺陷資訊作為疑似缺陷而禁止輸出。 又’缺陷輸出可否判斷部23係譬如可將個別之缺陷的缺 陷資訊所含之缺陷檢測參數（譬如，上述斜率a及截距b)， 依據針對此缺陷所屬之巨集區域而決定之分布資訊作修 正’藉由修正後之近似曲線將檢測臨限值再設定，將該缺 陷資訊所含之此缺陷之灰階差與再設定後之檢測臨限值作 比較’將該缺陷資訊之輸出可否，予以判定。 如上述般’缺陷輸出可否判斷部23係於再設定後之缺陷 檢測條件下，將在缺陷檢測部8所檢測之各缺陷予以再度 檢測，藉此缺陷檢測條件係依據針對巨集區域而決定之分 布資訊而變更。 與某一巨集區域有關之分布資訊，如被作為就此巨集區 域所含之各檢查部分圖像81而決定之參考值的集合而決定 之情形時’缺陷輸出可否判斷部23係將缺陷檢測條件再設 定’以使缺陷檢測感度隨著在分布資訊所含之各參考值内 "3715.doc •40- 1336778 預定為本來相同值之參考值彼此之參差不齊的增大而降 低。譬如，依據預定為本來相同值之參考值彼此之參差不 齊的增大而將檢測臨限值再設定。 缺陷輸出可否判斷部23係譬如可將如下參考值，作為上 述預定為本來相同值之參考值；而該參考值係關於譬如複 數個晶片（晶粒）3a，針對將晶片内之相同部份攝像後之個 別之部分圖像81(上述之參考值決定單位圖像亦可）分別決 疋者。接著，可將缺陷檢測條件再設定，以使缺陷檢測感 度隨著此等參考值彼此之參差不齊的增大而降低。 再者，與某一巨集區域有關之參考值之分布資訊，如被 作為就此巨集區域所含之各檢查部分圖像81而決定之參考 值的分散值而決定之情形時，缺陷輸出可否判斷部23係將 缺陷檢測條件再設定，以使缺陷檢測感度隨著如下分布資 訊的增大而降低；而該分布資訊係作為預定為本來相同值 之參考值彼此之分散值而決定者。 如上述般，在本圖像缺陷檢測裝置丨〇中，由於就比檢查 部分圖像81(邏輯圖框）更大（廣）之各巨集區域，可獲得檢 查部分圖像81、參考圖像所含之像素值之平均值等之分布 資訊’因此，觀察廣巨集區域中之灰階值的變動，則可檢 測顏色不均的有無；而檢查部分圖像81係用於進行缺陷檢 測之圖像比較之單位。 接著，依據此種分布資訊將缺陷檢測條件再設定，藉由 在再S又疋後之缺陷檢測條件下，判定缺陷資訊之輸出可 否’則可防止疑似缺陷的輸出，而疑似缺陷係檢查圖像所 113715.doc • 41 1336778 產生之顏色不均所產生者。 再者，如參考圖4(A)〜圖4(C)所說明般，檢測臨限值計 算部7係根據檢查部分圖像81及參考圖像之像素值，進行 汁算檢測臨限值、缺陷檢測參數。基於此因，此種檢測臨 限值、缺陷檢測參數之分布資訊亦可再被利用於顏色不均 之檢測上。此時，缺陷輸出可否判斷部23可將如下檢測臨 限值或缺陷檢測參數作為上述預定為本來相同值之參考 值；而該檢測臨限值或缺陷檢測參數係關於複數個晶片 (晶粒）3a，針對將晶片内之相同部份攝像後之個別之部分 圖像8 1所算出者。 再者，缺陷資訊作成部9可就檢查圖像8〇上預先決定之 複數之各部位，無論各部位缺陷之有無，均將上述之像素 值相關資訊予以決定，作成包含此像素值相關資訊之虛擬 缺陷資訊。藉此則無論缺陷之有無，參考值決定部21均可 將檢查圖像80上預先決定之複數之各部位的參考值，予以 決定。譬如，將均一分散於檢查圖像整體之定點，作為決 疋像素值相關資訊之部位予以決定，則可取得在均一分散 於檢查圖像80整體之定點上之參考值的分布資訊。 此時，作為像素值相關資訊，缺陷資訊作成部9係譬如 可在缺陷資訊中包含：位於檢查圖像8〇上之預先決定之部 位的檢查部分圖像8 1、及與之對應之參考圖像82 ;或是包 含位於檢查圖像80上之預先決定之部位的檢查部分圓像^ 及與之對應之參考圖像82間之差圖像亦可。或是作為像素 值相關資訊，缺陷資訊作成部9可在缺陷資訊中包含.針 113715.doc • 42- 1336778 對位於檢查圖像80上之預先決定之部位的檢查部分圖像^ 所決定之檢測臨限值、缺陷檢測參數。 接著’如被提供位於檢查圖像80上之預先決定之部位的 檢查部分圖像81、及與之對應之參考圖像82作為像素值相 關資訊之情形時，參考值決定部21可將如下所述者作為夂 考值而決定：此等圖像之任一方或兩方之圖像内之特定位 置之像素值；此等圖像之任一方或兩方之圖像所含之像素 值之平均值、分散值、最大值、最小值、或此等最大值與 最小值之中間值或差。或是可將如下所述者作為參考值而 決定：檢查部分圖像81及對應於此檢查部分圖像81之參考 圖像82間之差圖像内之特定位置之像素值；差圖像所含之 像素值之平均值、分散值、最大值、最小值、或此等最大 值與最小值之中間值或差。 如被提供位於檢查圖像80上之預先決定之部位的檢查部 分圖像81及與之對應之參考圖像_之差圖像作為像素值 相關資訊之情形肖，可冑差圖像内之特定位置之像素值、 差圖像所含之像素值之平均值、分散值、最大值、最小 值、或此等最大值與最小值之中間值或差，作為參考值而 決定。 ^被提供針對位於檢查圖像8〇上之預先決定之部位的檢 P刀圖像8 1而決定之檢測臨限值、缺陷檢測參數作為像 素值相關資訊之情形時，可將此等檢測臨限值、缺陷檢測 參數作為參考值而決定。 此芋s如，可將虛擬之缺陷資訊（亦即，作成1個像素 113715.doc -43- 1336778 值相關資訊、參考值之位置 形成β 辨為將+導體晶圓3上作複數 阳方C日日粒）3a内之特定仂 mM Μ ^ 夂位置攝像後之位置。在發生 月度差呈緩和變化之彩色 Ρ, ^ „ 巴+均的情形時，雖有必要於檢查 圖像中之廣範圍取得分布資訊 夂γ阁 仁精由如上述般就較大之 谷圍而決定參考值，則 可值則可卽約參考值之計算量。 如上述般’如將作成參考值 值之像素值等的所得位置決定 ::晶片之相同位置’則可將各晶片攝像後之圖像預定為 相冋’因此，成為作成參考值之基礎的像素值相關資 訊’亦可被預定為本來相同值，故參考值可預定為本來相 同值。 再者’如於檢查對象之半導體晶圓表面形成之圖案係藉 由光學或電子射束之曝光步驟（微影步驟）所形成之情形 時，缺陷資訊作成部9可將虛擬之缺陷資訊亦即，作成“固 像素值相關資訊、參考值之位置作為如下位置：將檢查對 象將在微影步驟形成有囷案之半導體晶圓表面攝像後之、 將前述檢查圖像之在丨次標線拍攝所曝光之範圍内之特定 位置攝像後的位置。 如上述般’如將作成參考值之像素值等的所得位置決定 為在各標線拍攝所曝光之範圍的相同位置，則將各標線拍 攝所曝光之範圍攝像後之圖像係可預定為本來相同，因 此’成為作成參考值之基礎的像素值相關資訊，亦可預定 為具有本來相同值，故參考值可預定為本來相同值。 [產業上之可利用性] 本發明可利用於圖像缺陷檢查裝置、圖像缺陷檢查系統 113715.doc • 44· 1336778 及圖像缺陷檢查方法，其係將檢查對象攝像後之檢查圖像 及/、此檢查圖像應為本來相同之參考圖像作比較，將彼此 不同之部分作為缺陷而檢測者。此外，可特別利用於如下 圖像缺陷檢查裝置、圖像缺陷檢查系統及圖像缺陷檢查方 法，其係為了於半導體製造步驟檢測在半導體晶圓上所形 成之半導體電路圖案之缺陷，而將半導體電路晶圓表面攝 像，將該攝像圖像與參考圖像作比較，將彼此不同之部分 作為缺陷而檢測者。 【圖式簡單說明】 圖1係先前之半導體電路用之外觀檢查裝置之區塊圖。 圖2係顯示半導體晶片上之晶粒之排列之圖。 圖3(A)係圖1之攝像裝置對晶圓作相對掃描時所獲得之 攝像圖像之說明圖；（B)係在圖像缺陷檢查上被比較之檢 查部分圖像及參考圖像之說明圖。 圖4(A)-(C)係先剛之圖像缺陷檢查方法之說明圖。 圖5係本發明之半導體電路用之圖像缺陷檢查裝置之第^ 實施例之區塊圖。 圖6係邏輯圖框之說明圖。 圖7係巨集區域之說明圖。 圖8係本發明之半導體電路用之圖像缺陷檢查裝置之第2 實施例之區塊圖。 圖9係本發明之半導體電路用之圖像缺陷檢查裝置之第3 實施例之區塊圖。 圖10係本發明之半導體電路用之圖像缺陷檢查裝置之第 113715.doc -45- 1336778 4實施例之區塊圖。 圖11係本發明之半導體電路用之圖像缺陷檢查系統之第 1實施例之區塊圖。 圖12係本發明之半導體電路用之圊像缺陷檢查系統之第 2實施例之區塊圖。 圖1 3係本發明之半導體電路用之圖像缺陷檢查系統之第 3實施例之區塊圖。 圖14係本發明之半導體電路用之圖像缺陷檢查系統之第 4實施例之區塊圖。 圖15係本發明之半導體電路用之圖像缺陷檢查系統之第 5實施例之區塊圖》 圖16係本發明之半導體電路用之圖像缺陷檢查系統之第 6實施例之區塊圖。 圖17係本發明之半導體電路用之圖像缺陷檢查系統之第 7實施例之區塊圖。 圖18係本發明之半導體電路用之圖像缺陷檢查系統之第 8實施例之區塊圖。 圖1 9係本發明之半導體電路用之圖像缺陷檢查系統之第 9實施例之區塊圖》 圖20係本發明之半導體電路用之圖像缺陷檢查裝置之第 5實施例之區塊圖。 【主要元件符號說明】 1 載物台 2 試料台 113715.doc -46· 1336778 3 半導體晶圓 4 攝像裝置 5 圖像記憶部 6 差分檢測部 7 檢測臨限值計算部 8 缺陷檢測部 10 圖像缺陷檢查裝置 21 參考值決定部 22 分布資訊決定部 113715.doc -47-Further, the detection threshold calculation unit 7 changes the detection threshold of the defect detection condition based on the knife information determined in the macro region. At this time, the detection threshold calculation unit 7 changes the detection threshold so that the defect detection sensitivity increases in accordance with the reference values of the same value in the respective reference values included in the distribution information. And lower.譬#, change the detection 6« limit so that it increases with this jagged increase. Here, when one unit of the reference value is determined (the area of the specific size is determined to be smaller than the area smaller than the wafer, the following reference value ' may be used as the reference value of the predetermined value of the predetermined value. And the reference value is related to a plurality of wafers (grains) 3a, and the image of the same portion in the wafer (four) wafer is determined by the area where the size of the wafer is determined to be the range of one wafer, or It is determined that the range of exposure is underlined; in the case of t, 'because the area of the specific size can be determined, the image captured by the specific size area is the same image, or can be determined. The determination method is such that the reference values determined for the respective regions become the same values, and thus such reference values can be predetermined to be the same value. Fig. 12 is a block diagram showing an embodiment of the image defect detection 2 for the semiconductor circuit of the present invention. In this embodiment, the first reference value determining unit 53 of the system is set to 113715. Doc -25 - 1336778 is placed on the side of the imaging device 50; and the reference money portion 53 is a camera image of the specific size of the inspection target (semiconductor wafer 3). The pixel value of + is determined by the reference value. Next, the data output unit 52 replaces the captured image data with the reference value data, and outputs it to the image defect inspection device 1A, and the data input unit 26 on the image defect inspection device 10 side inputs it. The distribution information determining unit 22 on the image defect inspection device 1G side is a macro region that is determined on the semiconductor wafer 3 and includes a plurality of regions of the specific size, and the regions of the specific size included in the macro region are The distribution information of the reference value determined by the reference value determining unit 53 on the imaging device 50 side is determined. Next, with reference to the method described above, the detection threshold calculating unit 7 changes the detection threshold of the defect detecting condition based on the determined distribution information. Figure 13 is a block diagram showing a third embodiment of the semiconductor circuit image defect inspection system of the present invention. In the present embodiment, the distribution information determining unit 54 is further provided on the side of the imaging device 50, which determines the distribution information of the reference value, which is a macro region composed of a plurality of regions of a specific size, It is determined by including each of the above-mentioned specific sizes of the macro region. Next, the data output unit 52 outputs the distribution information to the image defect inspection device 10, and the data input unit 26 on the image defect inspection device 10 side inputs the same as that described above with reference to FIG. 11 and the image defect. Inspection device 113715. Doc -26· 1336778 The detection threshold calculation unit 7 on the ι〇 side changes the detection threshold value of the defect detection condition based on the distribution information input from the imaging device 50. Figure 14 is a block diagram showing a fourth embodiment of the image defect inspection system for a semiconductor circuit of the present invention. The color unevenness of the inspection image after the semiconductor wafer 3 is imaged is caused by variations in the film thickness of the transparent or translucent film formed on the surface of the semiconductor wafer 3 or the like. The film thickness of the film formed on the surface of the semiconductor wafer 3 is measured at each portion of the wafer 3. By setting the measured value as a reference value, color unevenness can be detected. In other words, the reference value determining unit 21 and the distribution information determining unit 22 may use the semiconductor wafers 3 measured by the film thickness measuring device before the visual inspection of the semiconductor wafer 3 by the image defect inspection device 1 The film thickness of the part = material, the above reference value and the above-mentioned distribution f message are divided, and the film thickness measuring device is prepared in addition to the image defect inspection device 1 . The image defect inspection apparatus 10' includes an image defect inspection apparatus 10' and a film thickness measuring apparatus 6G that measures the film thickness of the film thickness formed on the surface of the semiconductor wafer 3; The thickness value is used for determining the reference value and the distribution information, and the reference value and the distribution information are used when the detection threshold calculation unit 7 of the image defect inspection device 1G determines the defect detection condition. By. The film thickness measuring apparatus A 60 includes a film thickness measuring unit 61 that measures the film thickness of a transparent or translucent medium formed on an insulating layer or the like on the surface of the semiconductor wafer 3, and a data output unit 62' for This thick measurement data is output to the image defect inspection device 10. 113,715. Doc 27- 1336778 On the other hand, the image defect inspection apparatus 10 is provided with a data input unit 26 for inputting the film thickness measurement data output from the film thickness measuring device 6A. The data round-trip between the data output unit 62 on the side of the film thickness measuring device 60 and the data input unit 26 on the side of the image defect inspection device 10 can be performed by the above-described connection method or offline method. Then, the reference value determining unit 21 provided in the image defect inspection apparatus 1 is a film of each specific size on the inspection target (semiconductor wafer 3), and the film is measured by the film thickness measuring device 60 in each of the regions. The thickness measurement value determines the reference value as described above. Further, the distribution information determining unit 22 determines a reference value for each of the above-described specific size regions of the macro region 3 for a macro region including a plurality of regions of the specific size determined on the semiconductor wafer 3. The distribution information is determined. Next, the detection threshold calculating unit 7 changes the detection threshold of the defect detecting condition based on the distribution information determined by the macro region. At this time, the detection threshold calculation unit 7 changes the detection threshold so that the defect detection sensitivity increases in accordance with the reference values of the same value in the respective reference values included in the distribution information. And lower. For example, change the detection threshold to increase as the jaggedness increases. Here, when it is determined that one unit of the reference value (the area of the specific size described above) is determined to be smaller than the area of the wafer, the following reference value can be used as the reference for the above-mentioned same value. The evaluation value is determined for each of the plurality of wafers (grains) 3a for the measured values measured at the same portion in the wafer. Furthermore, as mentioned above, the specific size U3715. The area of doc -28- 1336778 is determined by the range of i wafers or the range of exposures taken by the reticle line, since the measurement sites in each region can be determined (for example, on each wafer) Or the same position in each of the reticle images is measured, etc., so that the measured value measured in the specific size region is the same measured value, or the determination method can be determined so that the reference value determined for each region is determined. It becomes the same value, and thus such a reference value can be predetermined to be the same value. Figure 15 is a block diagram showing a fifth embodiment of the image defect inspection system for a semiconductor circuit of the present invention. In the present embodiment, the reference value determining unit 63 is provided on the film thickness measuring device 60 side, and the reference value determining unit 63 is a region of a specific size on the inspection target (semiconductor wafer 3). The reference value is determined separately; and the film thickness measurement value is measured by the film thickness measurement unit 6 i in each of the regions. Next, the data output unit 62 replaces the film thickness measurement value data with the reference value data, and outputs it to the image defect inspection device 10; the data input unit 26 on the image defect inspection device 10 side inputs the data. The distribution information determining unit 22 on the side of the circular image defect inspection device 10 is a macro region including a plurality of regions of the specific size determined on the semiconductor wafer 3, and the specific size included in the macro region is Each area 'determines the distribution information of the reference value determined by the reference value determining unit 63 on the imaging device 60 side. The detection threshold calculation unit 7 on the image defect inspection apparatus 10 side changes the detection threshold value of the defect detection condition in accordance with the above-described method described with reference to Fig. 14 based on the determined distribution information. 1137I5. Doc -29- 1336778 Fig. 16 is a block diagram showing a sixth embodiment of the image defect inspection system for a semiconductor circuit of the present invention. In the present embodiment, the distribution information determining unit 64 is further disposed on the side of the film thickness measuring device 6; the distribution information determining unit "is a macro region including a plurality of regions of the specific size, and the macro is set. The information on the distribution of the reference values determined by the specific size regions included in the region is determined. Next, the data output unit 62 outputs the distribution information to the image defect inspection device ίο'·the image defect inspection device 10 The data input unit 26 is input to the side. The detection threshold calculating unit 7 on the side of the image defect inspection device 1 is based on the distribution information input from the film thickness measuring device 60, as described above with reference to FIG. Fig. 17 is a block diagram showing a seventh embodiment of the image defect inspection system for a semiconductor circuit of the present invention, as shown in the image after the surface of the semiconductor wafer 3 is imaged. When the image produces a difference in brightness (uneven color), the critical dimension of the minimum size of the pattern formed on the semiconductor wafer 3 (criUcai is uneven). Therefore, the size of the critical dimension of each portion formed on the surface of the semiconductor wafer 3 is measured by a scanning electron microscope or the like, and by using the measured value as a reference value to generate the segment information, it is possible to detect color unevenness. The reference value determining unit 21 and the distribution information determining unit 22 may use the respective portions of the semiconductor wafer 3 measured by the scanning electron microscope device before the visual inspection of the semiconductor wafer 3 by the image defect inspection device 丨0. The critical dimension measurement value 'determines the above reference value and the above distribution information, and the scanning electron microscope apparatus is in the image defect inspection apparatus 1 〇 II3715. Doc -30- 1336778 Other preparers. Based on this, the circular image defect inspection system shown in FIG. 17 includes the image defect inspection device 10 and the scanning electron microscope device 7A, which measure the critical size of the pattern formed on the surface of the semiconductor wafer 3, and The critical dimension of the pattern is used to determine the reference value and the distribution information, and the reference value and the distribution information are detected by the image defect inspection device 5 The user at the time of detecting the condition. The scanning electron microscope apparatus 70 includes an electron gun 71 that generates an electron beam EB that is irradiated onto an inspection target (semiconductor wafer 3), and a deflector 72 that uses the electron beam EB on the semiconductor wafer 3. a scanner; an electronic detector 73 for detecting an electron beam EB reflected on the semiconductor wafer 3; and a signal processing circuit 74 for converting a current intensity signal of the electron detector 73 detecting the electron beam EB into a digital position The intensity signal of the type; the image generating unit 75 is based on the intensity signal and the scanning position of the electron beam EB to generate a high-magnification image on the surface of the semiconductor wafer 3; the CD measuring unit 76 will appear The critical dimension of the pattern of the image generated by the image generating unit 75 is measured, and the data output unit 77 is used to output the critical dimension measured value data to the image defect inspection apparatus 1. On the other hand, the image defect inspection apparatus 1A is provided with a data input unit 26 for inputting the critical dimension measurement value data output from the film thickness measuring device 6A. The data round-trip between the data output unit 77 on the side of the scanning electron microscope apparatus 7 and the data input unit % on the side of the image defect inspection apparatus 1 can be performed by the above-described connection method or offline method. Next, the reference value determining unit is disposed in the image defect inspection device 1 to 1137I5. Doc -31 * The reference value is determined as described above based on the critical dimension measurement values measured by the scanning electron microscope 70 for each specific size region on the inspection target (semiconductor wafer 3). Further, the distribution information determining unit 22 is a reference value determined for each of the specific size regions included in the macro region for the macro region including the plurality of regions of the specific size determined on the semiconductor wafer 3. The distribution information is determined. Next, the detection threshold calculating unit 7 changes the detection threshold of the defect detecting condition based on the distribution information determined by the macro region. At this time, the detection threshold calculation unit 7 changes the detection threshold so that the defect detection sensitivity increases with the reference values of the same value which are predetermined to be within the respective reference values included in the distribution information. And lower. For example, change the detection threshold to increase as the jaggedness increases. Here, if one unit of the reference value (the area of the specific size mentioned above) is determined as a smaller area than the wafer, the following reference value ' can be used as the reference for the above-mentioned same value. The reference value is determined for each of the plurality of wafers (grains) 3a for the measured values measured at the same portion in the wafer. Further, when the region of the specific size is determined for each range of one wafer or for each range of exposure of the first reticle, the measurement site in each region can be determined (for example, The same position in each of the wafers or each of the reticle lines is measured, etc.) 'The measured value measured in the specific size area is the same measured value, or the determination method can be determined for each area. The reference value becomes the same value, so the reference value can be reserved 113715. Doc 1336778 is the same value. Fig. 18 is a section (4) of the eighth embodiment of the image defect inspection system for a semiconductor circuit of the present invention. In the present embodiment, the reference value determining unit 78 is provided in the scanning electron microscope device 7A; the reference value determining unit 78 is an area of each specific size on the inspection target (semiconductor wafer 3), in accordance with the scanning type electrons. The measurement value of the critical dimension measured by the microscope device 70 in each of the regions is determined by the reference value. Then, the data output unit 77 receives the reference value (4) instead of the critical dimension measurement I data, and outputs it to the image defect inspection device ig; the data input unit 26 on the image defect inspection device 10 side inputs the data. The distribution information determining unit 22 on the side of the image defect inspection device 10 is a macro region that is determined on the semiconductor wafer 3 and includes a plurality of regions of the specific size, and each of the specific sizes included in the macro region is The area 'determination information of the reference value determined by the reference value determining unit 78 on the side of the scanning electron microscope apparatus 7 is determined. Fig. 19 is a block diagram showing a ninth embodiment of the image defect inspection system for a semiconductor circuit of the present invention. In the present embodiment, the distribution information determining unit 79 is further disposed on the side of the scanning electron microscope device 7; the distribution information determining unit 79 determines the distribution information of the reference value, and the reference value is for including a plurality of the above specific The macro area formed by the size area is determined by each of the above-mentioned specific sizes included in the macro area. Next, the data output unit 77 outputs the distribution information to the image defect inspection device 10; the data input portion 26 on the side of the image defect inspection device 1 inputs it to 0 113715. Doc -33 - 1336778 The detection threshold calculation unit 7 on the image defect inspection device θ changes the detection threshold value of the defect detection condition based on the distribution information input from the "thickness measurement device 60". Figure 20 is a block diagram showing a fifth embodiment of the image defect inspection apparatus for a semiconductor circuit of the present invention. In the image defect inspection apparatus 1 of the present embodiment, first, the defect information creation unit 9 creates the defect information in a predetermined pattern for each defect detected by the defect detecting unit 8. Φ Here, the type of the defect information is first determined to include the following information related to the detected defect: the detection position (defect position), the gray level difference between the inspection portion image of the defect position and the reference image, A part of the image of the pixel at the defect position is included in the inspection partial image, and the detection threshold or defect detection parameter used when detecting the defect. These partial images, detection thresholds or defect detection parameters may be determined according to the pixel values of the pixels or images corresponding to the respective portions of the inspection image 8 (in this case, the portion detecting the defect). Further, since the information is changed according to the pixel value of the pixel or pixel determined corresponding to each part on the inspection image 80, the following patent specification of the present invention uses "pixel value related information". Said. Then, according to the pixel value related information, a specific reference value is determined for each defect information; and then the distribution information of the reference values in each macro region is determined, and the defect detection conditions in each macro region are respectively determined according to the distribution information. . X疋, and judge the output of each defect information detected in each macro area. Since the reference value respectively reflects the pixel value of the complex part on the inspection image, 113715. Doc •34- 1336778 Therefore, if the defect information of the entire region of the inspection image is made, the distribution information of the reference value of the macro region having a certain width is determined, and the extent of the width is observed. The variation of the pixel value (grayscale value) can detect the presence or absence of color unevenness. Then 'reset the defect detection condition (if it is a macro area with a large uneven distribution of color unevenness, set the defect detection condition with low defect detection sensitivity; if it is a macro with small color unevenness) In the area, the defect detection condition, such as the defect detection sensitivity, is set to be changed according to the distribution information; by determining the availability of the defect information under the defect detection condition after the resetting, it is possible to prevent the color from being The output of suspected defects generated. The operation of each part of the image defect inspection apparatus 1 of the present embodiment will be described in detail below. Further, the image defect inspection apparatus 1 of the present embodiment has a structure similar to that of the image defect inspection apparatus described above with reference to FIG. 5. Therefore, the same component symbols are given to the same structural elements, but the same component symbols are given. The description of the components related to the components will be omitted. The defect detecting unit 8 detects the defect by comparing the gray level difference output from the difference detecting unit 6 and the detection threshold value outputted by the detection threshold calculating unit 7, and outputs the information related to the detected defect to the defect information. The processing unit 9 is related to the detection position (defect position), the gray portion difference between the inspection portion image of the defect position and the reference image, and the like. Defect information is created in order to output the information related to the defect to other devices. Therefore, the defect information containing this information will be prepared in accordance with the pre-determined format for each defect; and other devices are utilized and used. Detected 113715. Doc -35- 1336778 Automatic defect information grading device, display device, and feeding device for information related to defects. The format of this defect information can be defined in such a manner that various information for the defect classification of the device used for the automatic defect information classification device is included in the defect information. For example, the defect information includes the following pixel value related information: a partial image of the pixel including the defect position in the inspection portion image 81, a detection threshold value for detecting a defect, and a defect detection parameter. The pixel value related information that may be included in the defect information, for example, may be: an inspection portion image of the defect indicated by the detected defect information; or a reference image 82 corresponding to the inspection portion image 81; A partial image of the pixel at this defect location. Further, the pixel value related information may be: an inspection portion image 81 that detects a defect indicated by the defect information; or a reference image 82 corresponding to the inspection portion image 81. Further, the 'pixel value related information' may be a difference image between the detected portion of the defect indicated by the detected defect information, and the reference image 82 corresponding to the inspection portion image 81. The pixel value related information may be a difference image between the partial image of the pixel having the defect position in the inspection portion image 81 of the defect indicated by the detected defect information, corresponding to the inspection portion map A partial image of a pixel containing a defect location in a reference image 82 like 8.1. In addition, the detection threshold and defect detection parameters used in the detection of defects indicated by the defect information may be included. The defect information creating unit 9 outputs the defect information including the pixel value related information to the reference value determining unit 21. H3715. Doc -36- 1336778 The reference value determining unit 21 determines the specific reference value for each defect information (that is, for each pixel value related information) based on the pixel value related information included in the defect information. For example, the reference value The determination unit 21 is provided with a partial image of the inspection portion image 81 indicating the defect indicated by the defect information and the pixel having the defect position in the reference image 82 corresponding to the inspection portion image 81 as the pixel value related information. In the case of the case, the average value, the dispersion value, the maximum value, the minimum value, or the intermediate value or difference between the maximum value and the minimum value of the pixel values included in one or both of the partial images may be used as reference values. And decided. Or the reference value determining unit 21 may use the average value, the dispersion value, the maximum value, the minimum value, or the intermediate value or difference between the maximum value and the minimum value of the pixel values included in the difference image between the following two The reference value is determined; and the two are: a partial image of the pixel including the defect position in the partial image 81, and a pixel having the defect position in the reference image 82 corresponding to the inspection partial image 81. Part of the image. When the difference image between the following is provided as the pixel value related information, the reference value determining unit 21 may average, disperse, maximize, and minimum the pixel values included in the difference image. Or the intermediate value or difference ' between the maximum and minimum values is determined as a reference value; and the two are: a partial image of the pixel containing the defect position in the partial image 81, corresponding to the inspection portion image A partial image of the pixel in the reference image 82 of 81 containing the location of the defect. The inspection portion image 8j, which is provided with the defect indicated by the detection defect information, and the reference image 82 corresponding to the inspection portion image 8 1 as the pixel value correlation 113715. In the case of doc -37- 1336778, the following can be used as a reference value: the pixel value of a specific position in the image of either or both of these images; either of these images Or the average value, dispersion value, maximum value, minimum value, or the median or difference between the maximum and minimum values of the pixel values contained in the two images. Alternatively, the following may be used as a reference value: the pixel value of the specific position in the difference image between the partial image 81 and the reference image 82 corresponding to the inspection partial image 81; the difference image The average, dispersion, maximum, minimum, or intermediate or difference between the maximum and minimum values of the pixel values contained. If the difference image between the inspection portion image 8丨 and the reference image 82 corresponding to the defect indicated by the defect information is provided as the pixel value related information, the following may be used as the reference value. It is determined that the pixel value of the specific position in the difference image; the average value of the pixel values contained in the difference image, the value of the distribution, the maximum value, the minimum value, or the intermediate value or difference between the maximum value and the minimum value. When the detection threshold and the defect detection parameter used in the detection of the defect indicated by the defect information are provided as the pixel value related information, the detection threshold and the defect detection parameter may be determined as reference values. When determining the reference value, the reference value determining unit 21 may determine one reference value for each of the inspection portion images 8 or for each partial image; and the partial image systems will check the images for each specific The number of pixels is divided. In the patent specification of the present invention, the partial image as the unit for determining the reference value may be referred to as a "reference value determination unit image". For example, in a certain reference value, the unit is determined as shown in (4), and the number of defects is measured, and the reference value determining unit H37I5. Doc •38- 1336778 21 can use the reference value determined by any of these defects as a representative value, and determine each reference value to determine the reference value of the unit image. Alternatively, the reference value determined by the reference value determining each defect in the unit image may be obtained as a reference value for determining the unit image as each reference value. Referring back to Fig. 20, the distribution information determining unit 22 determines the distribution information of the reference value determined for each defect for the image area (macro area) of a specific size. The image area of the specific size is determined in the inspection image after the semiconductor crystal 3 is imaged; and the defects are detected by the inspection portion image 81 included in the macro region. The description for the macro region is as shown in the above illustration. The cloth information determining unit 22 may determine the dispersion value of the reference value determined for each of the defects detected on the inspection portion image 8丨 for the plurality of inspection portion images 81 included in the certain macro region. It is determined as information on the distribution of reference values of the macro region. Alternatively, the set σ ' of reference values determined with respect to the defect detected on each of the inspection portion images 81 included in the macro region may be determined as distribution information. Further, as described above, when the If shape of the reference image is determined for each reference value, the macro region is determined as a larger area than the one reference value determination unit image; The distribution information of the reference value determined by each reference value in determining the unit image is determined. Further, the 'defective output availability determination unit 23 performs re-setting as the detection threshold value of the defect detection condition based on the distribution information determined for the macro region> under the defect detection condition after resetting, and is included in the macro region. 113715. Doc •39· 1336778 The output of the defect information detected by each inspection part image 8 1 can be determined. The defect output possibility determination unit 23 inputs, for example, defect information related to each defect detected by the defect detecting unit 8, and detects the detection threshold included in the defect information of the individual defect by relying on the macro set for the defect The distribution information determined by the area is corrected and re-set; in addition, the gray level difference of the defect included in the defect information is compared with the reset detection threshold, for example, if the gray level difference exceeds the detection threshold Then, the defect information is allowed to be output as a true defect; on the contrary, if the gray level difference does not exceed the detection threshold, the defect information is prohibited from being output as a suspected defect. Further, the defect output determination unit 23 is configured to determine the defect detection parameters (for example, the slope a and the intercept b) included in the defect information of the individual defect based on the macro region to which the defect belongs. Correction 'Reset the detection threshold by the corrected approximation curve, compare the gray level difference of the defect contained in the defect information with the reset detection threshold. 'Can the output of the defect information be OK? , to be judged. As described above, the defect output possibility determination unit 23 re-detects each defect detected by the defect detecting unit 8 under the defect detection condition after resetting, whereby the defect detection condition is determined according to the macro region. Change information by distributing information. When the distribution information relating to a certain macro region is determined as a set of reference values determined for each of the inspection portion images 81 included in the macro region, the defect output determination unit 23 detects the defect. The condition is set again 'so that the defect detection sensitivity is within the reference values contained in the distribution information "3715. Doc •40- 1336778 It is intended that the reference values of the same value are reduced by the unevenness of each other. For example, the detection threshold is reset based on an increase in the reference value of the same value that is predetermined to be the same. The defect output availability determining unit 23 may, for example, use the following reference value as a reference value for the predetermined same value; and the reference value relates to, for example, a plurality of wafers (die) 3a for imaging the same portion of the wafer. The subsequent partial image 81 (the above reference value determines the unit image may also be determined). Next, the defect detecting condition can be reset so that the defect detecting sensitivity decreases as the reference values become uneven with each other. Further, when the distribution information of the reference value associated with a certain macro region is determined as the dispersion value of the reference value determined for each of the inspection portion images 81 included in the macro region, the defect output can be The determination unit 23 resets the defect detection condition so that the defect detection sensitivity decreases as the distribution information is increased as follows; and the distribution information is determined as a dispersion value of reference values that are predetermined to be the same value. As described above, in the image defect detecting apparatus ,, since the macro regions larger (wider) than the partial image 81 (logical frame) are inspected, the inspection portion image 81 and the reference image can be obtained. The distribution information of the average value of the pixel values included, etc. Therefore, it is possible to detect the presence or absence of color unevenness by observing the variation of the grayscale value in the wide macro region, and the inspection portion image 81 is used for defect detection. The unit of image comparison. Then, according to the distribution information, the defect detection condition is reset, and whether the output of the defect information can be determined by the defect detection condition after the S is further prevented, the suspected defect output can be prevented, and the suspected defect inspection image is prevented. 113715. Doc • 41 1336778 The resulting color unevenness is produced. Further, as described with reference to FIGS. 4(A) to 4(C), the detection threshold calculation unit 7 performs the juice calculation detection threshold based on the pixel values of the inspection portion image 81 and the reference image. Defect detection parameters. Based on this reason, the distribution information of such detection thresholds and defect detection parameters can be further utilized for the detection of color unevenness. At this time, the defect output availability determination unit 23 may use the detection detection threshold or the defect detection parameter as the reference value of the predetermined original value; and the detection threshold or the defect detection parameter is related to the plurality of wafers (die) 3a, which is calculated for the partial image 81 of the image obtained by imaging the same portion in the wafer. Further, the defect information creating unit 9 may determine each of the plurality of predetermined portions on the image 8〇, and determine the pixel value related information in accordance with the presence or absence of each part of the defect, thereby forming information related to the pixel value. Virtual defect information. Thereby, the reference value determining unit 21 can determine the reference value of each part of the plurality of predetermined numbers on the inspection image 80 regardless of the presence or absence of the defect. For example, by uniformly dispersing the fixed point of the entire inspection image as a part of the information relating to the pixel value, the distribution information of the reference value uniformly distributed over the entire fixed point of the inspection image 80 can be obtained. At this time, as the pixel value related information, the defect information creating unit 9 can include, for example, the inspection portion image 8 1 of the predetermined portion located on the inspection image 8〇, and the reference map corresponding thereto. An image such as 82; or a difference between the inspection portion circle image ^ and the reference image 82 corresponding to the predetermined portion of the inspection image 80 may be used. Or as the pixel value related information, the defect information creation unit 9 can be included in the defect information. Needle 113715. Doc • 42- 1336778 The detection threshold and defect detection parameters determined by the inspection part image ^ of the predetermined part located on the inspection image 80. Then, when the inspection portion image 81 of the predetermined portion on the inspection image 80 and the reference image 82 corresponding thereto are provided as the pixel value related information, the reference value determination portion 21 can The reference is determined as the value of the pixel: the pixel value of a specific position in the image of either or both of the images; the average of the pixel values contained in the image of either or both of the images Value, dispersion value, maximum value, minimum value, or the median or difference between these maximum and minimum values. Alternatively, the following may be used as a reference value: the pixel value of the specific position in the difference image between the partial image 81 and the reference image 82 corresponding to the inspection partial image 81; the difference image The average value, dispersion value, maximum value, minimum value, or the median or difference between these maximum and minimum values. If the image of the difference between the inspection portion image 81 of the predetermined portion on the inspection image 80 and the reference image _ corresponding thereto is provided as the pixel value related information, the specificity within the image may be The pixel value of the position, the average value of the pixel values contained in the difference image, the dispersion value, the maximum value, the minimum value, or the intermediate value or difference between the maximum value and the minimum value are determined as reference values. ^ When the detection threshold and the defect detection parameter are determined as the pixel value related information for the P blade image 81 located in the predetermined portion of the inspection image 8〇, the detection may be performed. The limit value and the defect detection parameter are determined as reference values. For example, the virtual defect information can be obtained (that is, made into 1 pixel 113715. Doc -43- 1336778 Value-related information, the position of the reference value The formation of β is determined by the position on the +conductor wafer 3 as a complex number of positive C 日 夂 ^ 内 in the position 3a. In the case where the monthly difference is a gradual change, in the case of +巴+均, it is necessary to obtain a distribution information in a wide range of images inspected. 夂 阁 阁 精 精 is as large as above. When the reference value is determined, the value can be calculated as the calculation value of the reference value. As described above, the position of the pixel value of the reference value is determined as follows: the same position of the wafer can be imaged by each wafer. The image is predetermined to be opposite. Therefore, the pixel value related information 'which becomes the basis of the reference value may be predetermined to be the same value, so the reference value may be predetermined to be the same value. Further, 'the semiconductor crystal as the inspection object When the pattern formed by the circular surface is formed by an exposure step (lithography step) of an optical or electron beam, the defect information creating portion 9 can create virtual defect information, that is, "solid pixel value related information, reference". The position of the value is as follows: the object to be inspected will be exposed after the lithography step is formed on the surface of the semiconductor wafer on which the stencil is formed, and the inspection image is exposed to the reticle line. Imaging position within a particular location. As described above, if the resulting position such as the pixel value of the reference value is determined to be the same position in the range in which each of the reticle is exposed, the image obtained by capturing the range of exposure of each reticle can be predetermined as Originally the same, therefore, the pixel value related information which becomes the basis of the reference value can also be predetermined to have the same value, so the reference value can be predetermined to be the same value. [Industrial Applicability] The present invention can be utilized in an image defect inspection device and an image defect inspection system 113715. Doc • 44· 1336778 and image defect inspection method, which is to check the inspection image after the object is inspected and/or, the inspection image should be compared with the same reference image, and the different parts are detected as defects. By. In addition, it can be particularly utilized in an image defect inspection device, an image defect inspection system, and an image defect inspection method for detecting a defect of a semiconductor circuit pattern formed on a semiconductor wafer in a semiconductor manufacturing step, and a semiconductor The surface of the circuit wafer is imaged, and the captured image is compared with a reference image, and portions different from each other are detected as defects. BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a block diagram of a prior art visual inspection device for a semiconductor circuit. Figure 2 is a diagram showing the arrangement of crystal grains on a semiconductor wafer. FIG. 3(A) is an explanatory diagram of a captured image obtained by the imaging device of FIG. 1 when the wafer is relatively scanned; and (B) is an image of the inspection portion and the reference image that are compared on the image defect inspection. Illustrating. 4(A)-(C) are explanatory views of the image defect inspection method of the first. Fig. 5 is a block diagram showing a second embodiment of an image defect inspection apparatus for a semiconductor circuit of the present invention. Figure 6 is an explanatory diagram of a logic frame. Fig. 7 is an explanatory diagram of a macro region. Fig. 8 is a block diagram showing a second embodiment of the image defect inspection apparatus for a semiconductor circuit of the present invention. Fig. 9 is a block diagram showing a third embodiment of the image defect inspection apparatus for a semiconductor circuit of the present invention. Figure 10 is a diagram 113715 of an image defect inspection apparatus for a semiconductor circuit of the present invention. Doc -45 - 1336778 4 block diagram of the embodiment. Figure 11 is a block diagram showing a first embodiment of an image defect inspection system for a semiconductor circuit of the present invention. Figure 12 is a block diagram showing a second embodiment of the imaging defect inspection system for a semiconductor circuit of the present invention. Fig. 1 is a block diagram showing a third embodiment of the image defect inspection system for a semiconductor circuit of the present invention. Figure 14 is a block diagram showing a fourth embodiment of the image defect inspection system for a semiconductor circuit of the present invention. Fig. 15 is a block diagram showing a fifth embodiment of the image defect inspection system for a semiconductor circuit of the present invention. Fig. 16 is a block diagram showing a sixth embodiment of the image defect inspection system for a semiconductor circuit of the present invention. Figure 17 is a block diagram showing a seventh embodiment of the image defect inspection system for a semiconductor circuit of the present invention. Figure 18 is a block diagram showing an eighth embodiment of the image defect inspection system for a semiconductor circuit of the present invention. Figure 9 is a block diagram of a ninth embodiment of an image defect inspection system for a semiconductor circuit of the present invention. Fig. 20 is a block diagram showing a fifth embodiment of an image defect inspection apparatus for a semiconductor circuit of the present invention. . [Main component symbol description] 1 Stage 2 Sample stage 113715. Doc -46· 1336778 3 Semiconductor wafer 4 Imaging device 5 Image memory unit 6 Difference detection unit 7 Detection threshold calculation unit 8 Defect detection unit 10 Image defect inspection device 21 Reference value determination unit 22 Distribution information determination unit 113715. Doc -47-