11638twf.doc/006 玖、發明說明： 【發明所屬之技術領域】 本發明是有關於一種製程最佳化（Process optimum)的方 法，且特別是有關於一種使用動態製程裕度檢查（dynamic process window check)法將製程最佳化的方法。 【先前技術】 在半導體製程中，其整個製程係包括許多的子製程，在 對每一個子製程作分析的情況下，係可以發現在每一個子製 程中可以分析出許多的影響因子，而其中的每一道子製程中 的影響因子都可能會對於產品的品質與特性造成影響。而且， 除了單一製程的變動所造成的影響之外，製程機台的變動例 如是控制單元零點飄移、正常的調機、化學品濃度產生變化… 等等，同樣會對產品的品質與特性造成影響，因此，係需要 對上述各種會對製程良率的影響因子進行分析、調整與測試， 以使製程能夠最佳化，進而提高產品的良率。 第1圖至第4圖係說明習知一種使用靜態製程裕度檢查 (Static process window check)法將製程最佳化的方法，如前所 述，於整個的製程中係包含許多的子製程，請參照第1圖， 對其中的一個子製程而言，係可以將之視爲一無限延伸的平 面1〇〇上所出現特定形狀的洞，其中此洞即是由子製程中之 所有影響因子交集而形成。 接著，請參照第2圖，提供一個固定不動的平面200， 其中此平面上係描繪出複數個同心圓，而此些同心圓係可以 代表產品的良率與特性（例如是電性、幾何形狀等等），並且同 心圓的圓心表示良率達100%以及特性的目標値。 接著，請參照第3圖，將平面200置於所有製程平面的 最後，並從製程的一開始沿著第一個子製程往最後一個子製 11638twf.doc/006 程(於第3圖僅繪示出平面10(^至1004)到平面200看，係可 以在平面上看到一個特定的區域’此特定區域係可以表示爲 產品經過不同子製程後最終的良率與特性表現，此區域亦可 以表示整個製程的整體製程裕度。 接著，請參照第4圖，在進行整個製程之製程裕度檢查 時，一般而言僅考慮到單一個子製程，並不考慮所有子製程 同時變動的結果，如第4圖所示，在子製程平面10〇1至10〇n 之中，係僅對子製程平面1〇〇η·2本身單獨進行參數的調整與測 試，然後再進行整體製程裕度的檢查。 然而，上述使用靜態製程裕度檢查法以求取最佳化製程 裕度的方法，在其他子製程同時有較大變動時不一定適用， 從而使得所得到的檢查結果不一定正確，因此，在產品的製 程良率逐漸提升的同時，會造成整體的最佳製程裕度可能必 須重新確認。 【發明內容】 因此，本發明的目的就是在提供一種製程最佳化的方法 與動態製程裕度檢查法，能夠允許所有的子製程與所有的參 數同時進行調整，並能夠得到較準確的最佳化結果。 本發明提供一種製程最佳化的方法’此方法包括下列步 驟：（a)選取一段時間的產品，（b)將產品依良率的高低分爲複 數個良率分組，（c)選擇預定進行製程最佳化之參數’（d)個別 求取良率分組之參數的平均値，（e)求取預定良率到達1〇〇%之 參數的平均値，（f)以預定良率到達1〇0%之參數的平均値取代 原來之數値。 在上述製程最佳化的方法中，其中在⑴以預定良率到達 100%之參數的平均値取代原來之數値之後，更包括進行一判 斷步驟，以判斷產品的良率是否提高。並且在判斷步驟判斷 11638twf.doc/006 產品的良率未提高時，重新選取另一段時間的產品，並重複(b) 至（f)的步驟。 本發明提供一種動態製程裕度檢查法，適於對一製程進 行最佳化，其中此製程係由複數個子製程所組成，且每一子 製程具有一製程裕度，此動態製程裕度檢查法包括：於此製 程中選定參數，然後將選定的參數最佳化，其後檢查所有子 製程之製程裕度，以得到製程之整體製程裕度，其特徵在於 子製程之製程裕度係隨著參數最佳化步驟而產生相對應之變 動。 並且，於此動態製程裕度檢查法中，其中參數最佳化的 方法係能夠採用下述的方法以進行：（a)選取一段時間的產品， (b)將產品依良率分爲複數個良率分組，（c)個別求取良率分組 之參數的平均値，（d)求取預定良率到達100%之參數的平均 値，（e)以預定良率到達100%之參數的平均値取代原來之數 値。 由上述可知，由於本發明之製程最佳化的方式與動態製 程裕度檢查法，係藉由持續收集產品之經由實際製程的良率 以及相關的參數數値的情況下，選定一段時間的產品後，再 利用良率以及相關的參數數値以進行製程的最佳化。由於製 程最佳化所使用的數據（良率）是實際經由每一道製程所得的結 果，因此，在進行最佳化時，係能夠反映出整個製程中每一 個子製程之製程裕度的變動，從而使得本發明能夠允許所有 的子製程與所有的參數同時進行調整，並能夠得到較準確的 最佳化結果。 而且，於本發明之製程最佳化的方式與動態製程裕度檢 查法中，對單一個參數而言，只需考慮此參數的數値與產品 良率的關係，就能夠將此參數最佳化，因此，本發明係能夠 11638twf.doc/006 以相當簡單的方式對所有子製程與所有參數進行最佳化，同 時能夠兼顧最佳化結果的準確性。 爲讓本發明之上述和其他目的、特徵、和優點能更明顯 易懂，下文特舉一較佳實施例，並配合所附圖式，作詳細說 明如下： 【實施方式】 第5圖所繪示爲依照本發明一較佳實施例之一種使用動 態製程裕度檢查（dynamic process window check)法以將製程最 佳化的流程圖。 請參照第5圖的步驟S302,首先選取一段時間的產品及 良率，其中此處之產品的定義係表示經過整個製程的每一道 子製程所製造出來的成品。 接著，請參照第5圖的步驟S304，對所選擇時間內的產 品依照良率的高低加以分組，在此分組過程中，其中所分出 之良率組別越多的話，則後續最佳化的結果將會越準確，並 且，在每一個良率分組中，其最高良率與最低良率的差値(YD) 不可過大，而且此差値（YD)較佳爲小於所有產品之最高良率 與最低良率差値的4分之1，尙且，此差値（YD)對每一個良率 分組較佳爲相等。 接著，請參照第5圖的步驟S306，找出產品欲做最佳化 製程中可能的參數FE，然後，對每一個良率分組的產品所相 對應的所有FE値，求出每一個良率分組（亦即是不同良率値） 的平均値F A。11638twf.doc / 006 发明, Description of the invention: [Technical field to which the invention belongs] The present invention relates to a method for process optimization, and in particular, to a method using dynamic process margin checking (dynamic process window) check) Method to optimize the process. [Previous technology] In the semiconductor process, the entire process system includes many sub-processes. In the case of analyzing each sub-process, the system can find that many influence factors can be analyzed in each sub-process, and among them The impact factors in each sub-process may affect the quality and characteristics of the product. In addition, in addition to the effects caused by changes in a single process, changes in the process equipment such as zero drift of the control unit, normal machine adjustment, changes in chemical concentration, etc., will also affect the quality and characteristics of the product. Therefore, it is necessary to analyze, adjust and test the above-mentioned various influencing factors on the yield of the process in order to optimize the process and thereby improve the yield of the product. Figures 1 to 4 illustrate a conventional method for optimizing a process using a static process window check method. As mentioned earlier, the entire process includes many sub-processes. Please refer to Figure 1. For one of the sub-processes, it can be regarded as a hole of a specific shape on an infinitely extending plane 100, where the hole is the intersection of all the influence factors in the sub-process And formed. Next, please refer to FIG. 2 to provide a fixed plane 200, where a plurality of concentric circles are depicted on the plane, and these concentric circles can represent the yield and characteristics of the product (for example, electrical, geometric shapes) Etc.), and the center of the concentric circle represents the target of 100% yield and characteristics. Next, referring to Figure 3, place plane 200 at the end of all process planes, and follow the first sub-process from the beginning of the process to the last sub-process 11638twf.doc / 006 process (only drawn in Figure 3 Show plane 10 (^ to 1004) to plane 200. You can see a specific area on the plane. 'This specific area can be expressed as the final yield and characteristics of the product after different sub-processes. This area is also It can represent the overall process margin of the entire process. Next, please refer to Figure 4, when checking the process margin of the entire process, generally only a single sub-process is considered, and the results of all sub-processes changing at the same time are not considered. As shown in Figure 4, among the sub-process planes 1001 to 100n, only the sub-process plane 100n · 2 itself is individually adjusted and tested for parameters, and then the overall process margin is performed. However, the above-mentioned method using the static process margin inspection method to obtain an optimized process margin may not be applicable when other sub-processes have large changes at the same time, so that the obtained inspection results may not be necessarily Indeed, as the process yield of the product is gradually improved, the overall optimal process margin may have to be reconfirmed. [Summary of the Invention] Therefore, the object of the present invention is to provide a method and method for process optimization. The dynamic process margin check method can allow all sub-processes to be adjusted simultaneously with all parameters, and can obtain more accurate optimization results. The present invention provides a method for process optimization. This method includes the following steps: ( a) Select products for a period of time, (b) Divide the product into multiple yield groups according to the yield rate, (c) Select the parameters that are scheduled to be optimized for the process' (d) Determine the yield group parameters individually (E) find the average 値 of the parameters with a predetermined yield of 100%, and (f) replace the original 値 with the average 値 of the parameters with a predetermined yield of 100%. In the optimization method, after replacing the original number with an average value of a parameter that reaches 100% with a predetermined yield, a judgment step is further included to determine whether the yield of the product has increased. When the determination step determines that the yield of the 11638twf.doc / 006 product has not improved, reselect the product for another period of time and repeat the steps (b) to (f). The present invention provides a dynamic process margin inspection method, which is suitable for Optimize a process, where the process is composed of multiple sub-processes, and each sub-process has a process margin. The dynamic process margin check method includes: selecting parameters in this process, and then selecting the selected Parameter optimization, and then check the process margins of all sub-processes to obtain the overall process margin of the process, which is characterized in that the process margins of the sub-processes change correspondingly with the parameter optimization steps. In this dynamic process margin checking method, the method of parameter optimization can be performed by the following methods: (a) selecting a product for a period of time, (b) dividing the product into a plurality of good products according to a good yield. Rate grouping, (c) individually obtain the average value of the parameters of the yield grouping, (d) obtain the average value of the parameters that reach the predetermined yield of 100%, (e) average the parameters of the rate that reach 100% of the predetermined yield, Replace the original The number 値. From the above, it can be known that due to the method for optimizing the process of the present invention and the dynamic process margin checking method, the product is selected for a period of time by continuously collecting the product's yield through the actual process and the number of related parameters. After that, the yield and related parameters are used to optimize the process. Since the data (yield) used for process optimization is the result obtained through each process, the optimization of the process can reflect the changes in the process margin of each sub-process in the entire process. Therefore, the present invention can allow all sub-processes to be adjusted simultaneously with all parameters, and can obtain more accurate optimization results. Moreover, in the process optimization method and dynamic process margin checking method of the present invention, for a single parameter, only the relationship between the number of this parameter and the product yield rate can be considered to optimize this parameter. Therefore, the present invention is capable of optimizing all sub-processes and all parameters in a relatively simple manner at 11638twf.doc / 006, while taking into account the accuracy of the optimization results. In order to make the above and other objects, features, and advantages of the present invention more comprehensible, a preferred embodiment is given below in conjunction with the accompanying drawings to describe in detail as follows: [Embodiment Mode] FIG. 5 Shown is a flow chart for optimizing a process using a dynamic process window check method according to a preferred embodiment of the present invention. Please refer to step S302 in FIG. 5, first select the product and yield for a period of time. The definition of the product here refers to the finished product manufactured through each sub-process of the entire process. Next, referring to step S304 in FIG. 5, the products in the selected time are grouped according to the high and low yield rates. During this grouping process, the more the yield rate groups are divided, the subsequent optimization will be. The more accurate the results will be, and in each yield group, the difference between the highest yield and the lowest yield (YD) should not be too large, and the difference (YD) is preferably smaller than the highest yield of all products. The rate is one-fourth of the lowest yield rate, and this rate (YD) is preferably equal for each yield group. Next, please refer to step S306 in FIG. 5 to find possible parameters FE in the product to be optimized, and then, for all yields corresponding to each yield grouped product, find each yield The mean (FA) of the grouping (ie, different yields).
接著，請參照第5圖的步驟S308，由上述各個分組（良率） 的FA値，計算出當預定良率到達1〇〇%時的FA値，其中於 此步驟中求出預定良率到達100%時的FA値的方法，例如是 藉由上述取得之良率與對應之FA値的數據，求取良率對FA 11638twf.doc/006 値的關係式，然後再將100%的良率帶入關係式中以求出預定 良率到達100%時的FA値。 接著，請參照第5圖的步驟S310，將製程參數FE的原 來數値以預定良率到達100%時的FA値取代。 接著，請參照第5圖的步驟S312，判斷在以預定良率到 達100%時的FA値取代先前的製程參數後，所得到的結果是 否能夠符合預期，亦即是判斷在以預定良率到達1〇〇%時的FA 値取代原來的製程參數後良率是否會提高。 接著，請參照第5圖的步驟S316，當結果符合預期時， 即表示此產品的製程已經被最佳化。 而且，於上述第5圖的步驟S312中，當判斷結果不能符 合預期時，則進行步驟S314，重新選取相對應的一段包含新 估計値的時間與良率，再接續步驟S304至S310以進行此產 品之製程參數的最佳化。 爲了更詳細說明本發明之使用動態製程裕度檢查法以將 製程最佳化的方法，以下特舉出一實例以做進一步的說明。 於本實例中，對一個已經由每一道子製程所製造出來的 產品而言，其中預定使製程最佳化的可能參數係表示爲參數A 與參數B，並且參數A之數値的變動係爲每隔1秒鐘以100 的變動値，進行 900(A1)、1000(A2)、1100(A3)、1000(A2)、 900(A1)……的循環變動，而參數B之數値的變動係爲每隔3 秒鐘以 50 的變動値，進行 500(B1)、550(B2)、600(B3)、550(B2)、 5〇〇(Β1)……的循環變動。 請參照表1，表1係表示參數A與參數B在不同的變動 數値的組合下，所得到之此產品的良率組合表。 11638twf.doc/006 參數A 參數B A1 A2 A3 B1 85% 80% 70% B2 90% 85% 75% B3 80% 70% 55% 1222151 在經由確認參數A與參數B之每一秒的位置，係可以得 到如表2所示的參數A與參數B隨著時間的數値變動以及能 得到的相對應良率。 表2 時間（秒） 參數A 參數B 良率（％) 1 A1 900 B1 500 85 2 A2 1000 B1 500 80 3 A3 1100 B1 500 70 4 A2 1000 B2 550 85 5 A1 900 B2 550 90 6 A2 1000 B2 550 85 7 A3 1100 B3 600 55 8 A2 1000 B3 600 70 9 A1 900 B3 600 80 10 A2 1000 B2 550 85 11 A3 1100 B2 550 75 12 A2 . 1000 B2 550 85 13 A1 900 B1 500 85 14 A2 1000 B1 500 80 15 A3 1100 B1 500 70 16 A2 1000 B2 550 85 二 二 ·· ·· • ·· 1222151 11638twf.doc/006 於本實例中，由表1與表2所提供之參數A跑為# 默Λ與参數B隨 著時間的變動値與良率的關係觀之，就整體而言，#& η $ 參數Α與參數Β的良率，個別分爲55%、70%、 乙 /〇、80%、 85%、9〇%等六個良率分組’接著’由表2所提供的數 取一段預定的時間後，針對參數A與參數B，個別將每_^ 良率分組的變動値(FE)求出平均値(FA)。亦即是對參^ 4敷A與參 數B而言，係能夠個別求出相對應良率55%、70%、^^。/ /5〇/〇、80%、 85%、90%的平均値。 接著，請參照第6圖，以良率爲橫軸，參數a的平纟勺官 爲縱軸，將上述求得之參數A的平均値繪示於第 间2均値 1®ί 中，、、丨 求出參數A之良率對平均値的關係式如第6圖所$，t 0 係式的求得方法例如是可以使用趨勢線預測法，於本胃 所求得的關係式係表示如下式（1): ’ y = -5·5351χ+1433·1 , 接著，請參照第7圖，以良率爲橫軸，參數β的平纟巧値 爲縱軸，將上述求得之參數Β的平均値繪示於第7圖中 求出參數B之良率對平均値的關係式如第7圖所承，#+ _ 係式的求得方法例如是可以使用趨勢線預測法，於本胃 所求得的關係式係表示如下式(2): 、 y = -1.4703x+668.16 ,··(2) 在個別求出參數Α之良率對平均値的關係式與參數β 2 良率對平均値的關係式後，接著個別藉由上述關係式；求&# 良率到達100時參數Α與參數Β的平均値，其中良率到達ι〇〇 時的平均値，例如是將良率以100%代入上述關係式中以求 得，於本實例中，參數A在良率到達100%時的平均値爲880， 而參數B在良率到達100時％的平均値爲521。 在上述將製程參數最佳化的方法中，由於所使用的數據 10 1222151 11638twf.doc/006 (良率）是實際經由每一道子製程所得的結果，因此，在製程參 數的最佳化時，實際上已經考慮到每一個子製程所會產生的 製程裕度變動。 接著，請參照第8圖，在進行整個製程之整體製程裕度 的檢查時，如前所述，其中每一個子製程子製程l〇〇n 之製程裕度的變動係能夠同時被考慮，亦即是表示本發明進 行製程裕度檢查的方法，是屬於一種能夠同時考慮到各個子 製程之製程裕度變動的動態製程裕度檢查法。 於上述實例中，其中所揭示之良率對參數數値的關係式 係爲一次方程式，然而本發明並不限定於此，於本發明之製 程最佳化的方法中，其良率對參數數値的關係式，視實際狀 況亦可以是一次以上的方程式。 上述所揭示之製程最佳化的方法，於實際的一種應用範 例中，係能夠應用於製程的微調(Fme Time)。首先，在進行產 品之整個製程時，持續的收集/記錄經過完整製程之產品的良 率，同樣的，針對各個子製程中可能需要調整的參數，亦持 續的收集其相對應的數據。而且，上述產品的良率以及相關 參數之相對應數値，例如是可以進一步藉由構成資料庫的方 式以存在。 接著，如第5圖之步驟S302至步驟S310所述的進行選 定參數的計算，以計算出良率到達1〇〇%時此些參數的數値（平 均値），接著，將所求得之良率到達100%的數値反饋（Feedback) 至製程中，如此則能夠將整個製程調整/最佳化爲朝向使良率 到達100%，亦即是提高良率的方向進行。 其中，上述諸如產品之資料範圍的選定、擷取（步驟 S302)，良率的分組（步驟S304)、每一良率分組之平均値的求 取（S306)、良率到達100%的數値的求取（S308)以及數値的反 11 1222151 11638twf.doc/006 饋（S310)等步驟除了可以逐步計算之外，亦可以在製程的控制 系統中建立適當的計算程式或軟體，以將上述之各項計算步 驟整合於其中，如此則能夠對此產品的整個製程自動的進行 微調以及最佳化。 由上述可知，本發明至少具有下述優點： 1. 由於本發明之製程最佳化的方式與動態製程裕度檢查 法，係藉由持續收集產品之經由實際製程的良率以及相關的 參數數値的情況下，選定一段時間的產品後，再利用良率以 及相關的參數數値以進行製程的最佳化。由於製程最佳化所 使用的數據（良率）是實際經由每一道製程所得的結果，因此在 進行最佳化時，係能夠反映出整個製程中每一個子製程之製 程裕度的變動，從而使得本發明能夠允許所有的子製程與所 有的參數在同時調整，並能夠得到較準確的最佳化結果。 2. 而且，於本發明之製程最佳化的方式與動態製程裕度 檢查法中，對單一個參數而言，只需考慮此參數的數値與產 品良率的關係，就能夠將此參數最佳化，因此，本發明係能 夠以相當簡單的方式對多個子製程或是多個參數進行最佳 化，同時能夠兼顧最佳化結果的準確性。 雖然本發明已以一較佳實施例揭露如上，然其並非用以 限定本發明，任何熟習此技藝者，在不脫離本發明之精神和 範圍內，當可作些許之更動與潤飾，因此本發明之保護範圍 當視後附之申請專利範圍所界定者爲準。 【圖式簡單說明】 第1圖所繪示爲一個子製程之製程裕度平面的示意圖。 第2圖所繪示爲代表產品的良率與特性的平面的示意 圖。 12 1222151 11638twf.doc/006 第3圖所繪示爲複數個子製程之製程裕度平面重疊所得 之整體製程裕度的示意圖。 第4圖所繪示爲習知以靜態製程裕度檢查法以檢查整體 製程裕度的示意圖。 第5圖所繪示爲依照本發明一較佳實施例之一種利用動 態製程裕度檢查法使製程最佳化的流程圖。 第6圖所繪示爲產品之良率對參數A之數値的關係式的 示意圖。 第7圖所繪示爲產品之良率對參數A之數値的關係式的 示意圖。 _ 第8圖所繪示爲依照本發明一較佳實施例之一種利用動 態製程裕度檢查法以檢查整體製程裕度的示意圖。 【圖式標記說明】 100!、1002、1003、1004、100n_2、lOOw、l〇〇n :表不子 製程的平面 - 200 :表示良率或特性的平面 S302、S304、S306、S308、S310、S312、S314、S316 : 步驟 13Next, referring to step S308 in FIG. 5, the FA 値 when the predetermined yield reaches 100% is calculated from the FA 値 of each group (yield), and in this step, the predetermined yield is reached. The method of FA 値 at 100% is, for example, to obtain the relationship between the yield and FA 11638twf.doc / 006 藉 using the obtained yield and the corresponding FA 上述 data, and then the 100% yield Bring it into the relationship to find FA 値 when the predetermined yield reaches 100%. Next, referring to step S310 in FIG. 5, the original number 値 of the process parameter FE is replaced by FA 値 when the predetermined yield reaches 100%. Next, please refer to step S312 in FIG. 5 to determine whether FA 値 when the predetermined process yield reaches 100% replaces the previous process parameters, and whether the obtained result can meet the expectations, that is, it is judged that the process arrives at the predetermined yield. Will FA 时 at 100% improve the yield after replacing the original process parameters? Next, please refer to step S316 in FIG. 5. When the result meets expectations, it means that the manufacturing process of this product has been optimized. Moreover, in step S312 in the above FIG. 5, when the judgment result fails to meet the expectations, step S314 is performed, and a corresponding period including the newly estimated time and yield is selected again, and then steps S304 to S310 are performed to perform this. Optimization of product process parameters. In order to describe the method of using the dynamic process margin checking method to optimize the process in more detail, an example is given below for further explanation. In this example, for a product that has been manufactured by each sub-process, the possible parameters that are intended to optimize the process are expressed as parameters A and B, and the change in the number 値 of parameter A is With a change of 100 every 1 second, a cyclic change of 900 (A1), 1000 (A2), 1100 (A3), 1000 (A2), 900 (A1), etc. is performed, and the number of parameters B is changed. It is a cycle of 500 (B1), 550 (B2), 600 (B3), 550 (B2), 500 (B1), etc. with a change of 50 every 3 seconds. Please refer to Table 1. Table 1 shows the yield combination table of the product obtained under the combination of parameter A and parameter B with different variation numbers. 11638twf.doc / 006 Parameter A Parameter B A1 A2 A3 B1 85% 80% 70% B2 90% 85% 75% B3 80% 70% 55% 1222151 After confirming the position of parameter A and parameter B every second, it is The parameter A and parameter B shown in Table 2 can be obtained over time and the corresponding yield can be obtained. Table 2 Time (seconds) Parameter A Parameter B Yield (%) 1 A1 900 B1 500 85 2 A2 1000 B1 500 80 3 A3 1100 B1 500 70 4 A2 1000 B2 550 85 5 A1 900 B2 550 90 6 A2 1000 B2 550 85 7 A3 1100 B3 600 55 8 A2 1000 B3 600 70 9 A1 900 B3 600 80 10 A2 1000 B2 550 85 11 A3 1100 B2 550 75 12 A2. 1000 B2 550 85 13 A1 900 B1 500 85 14 A2 1000 B1 500 80 15 A3 1100 B1 500 70 16 A2 1000 B2 550 85 22 ······· 1222151 11638twf.doc / 006 In this example, the parameter A provided by Table 1 and Table 2 runs as # ΛΛ and the parameter Viewing the relationship between B and the yield over time, as a whole, the yield of # & η $ parameter A and parameter B are individually divided into 55%, 70%, B / 〇, 80%, 85%, 90%, and other six yield groups 'Next' After taking a predetermined period of time from the numbers provided in Table 2, for parameter A and parameter B, the change in each yield group of each ^^ (FE) Find the average 値 (FA). That is, for parameter A and parameter B, the corresponding yields of 55%, 70%, and ^^ can be obtained individually. / / 5〇 / 〇, 80%, 85%, 90% average 値. Next, referring to FIG. 6, with the yield as the horizontal axis and the horizontal axis of the parameter a as the vertical axis, the average value of the parameter A obtained above is plotted in the second interval 2 値 1®, The relationship between the yield of the parameter A and the average value is as shown in Figure 6. The method for obtaining the t 0 system is, for example, the trend line prediction method can be used to express the relationship system obtained in the stomach. The following formula (1): 'y = -5 · 5351χ + 1433 · 1. Then, referring to Fig. 7, the yield is on the horizontal axis, and the flatness of the parameter β is the vertical axis. The average 値 of Β is shown in Figure 7. The relationship between the yield of parameter B and the average 値 is obtained as shown in Figure 7. The method of obtaining the # + _ system formula is, for example, using the trend line prediction method. The relationship obtained by the stomach is expressed by the following formula (2):, y = -1.4703x + 668.16, (2) The relationship between the yield of the parameter A and the average 値 and the parameter β 2 After the relationship between the rate and the average 接着, then individually use the above-mentioned relationship; find &# the average 値 of the parameter A and the parameter B when the yield reaches 100, where the average 値 when the yield reaches ι〇〇, for example It is obtained by substituting 100% of the yield rate into the above relational formula. In this example, the average value of the parameter A when the yield rate reaches 100% is 880, and the average value of the parameter B when the yield rate reaches 100% is 521. In the above method for optimizing process parameters, since the data used 10 1222151 11638twf.doc / 006 (yield) is the result obtained through each sub-process actually, therefore, when the process parameters are optimized, In fact, the process margin variation caused by each sub-process has been considered. Next, please refer to Figure 8. When checking the overall process margin of the entire process, as described above, the variation of the process margin of each sub-process sub-process 100n can be considered at the same time. That is to say, the method for inspecting the process margin of the present invention belongs to a dynamic process margin inspection method that can simultaneously consider the variation of the process margin of each sub-process. In the above example, the relationship between the yield and the parameter number 揭示 disclosed therein is a linear equation. However, the present invention is not limited to this. In the method for optimizing the process of the present invention, the yield versus the number of parameters The relationship of 値 can also be an equation more than once depending on the actual situation. The method of process optimization disclosed above can be applied to the fine adjustment of the process (Fme Time) in an actual application example. First of all, during the entire process of the product, continuously collect / record the yield of the product after the complete process. Similarly, for the parameters that may need to be adjusted in each sub-process, the corresponding data is also continuously collected. Moreover, the corresponding numbers of the yields and related parameters of the above products may exist, for example, by forming a database. Next, the calculation of the selected parameters is performed as described in steps S302 to S310 in FIG. 5 to calculate the number (average) of these parameters when the yield reaches 100%, and then, the obtained The feedback that the yield reaches 100% is fed into the process. In this way, the entire process can be adjusted / optimized to make the yield reach 100%, which is to increase the yield. Among them, the selection and retrieval of the above-mentioned data range of products (step S302), the grouping of yields (step S304), the calculation of the average value of each yielding group (S306), and the number of yields reaching 100% In addition to the steps of calculating (S308) and the inverse of the data 11 1222151 11638twf.doc / 006 (S310), in addition to the stepwise calculation, an appropriate calculation program or software can be established in the process control system to convert the above All calculation steps are integrated into it, so that the entire process of this product can be automatically fine-tuned and optimized. As can be seen from the above, the present invention has at least the following advantages: 1. Due to the process optimization method and dynamic process margin checking method of the present invention, the yield of the product through the actual process and the number of related parameters are continuously collected In the case of 値, after selecting a product for a period of time, the yield and related parameter numbers are used to optimize the process. Because the data (yield rate) used for process optimization is the result obtained through each process, the optimization of the process can reflect the change in the process margin of each sub-process in the entire process, so that Therefore, the present invention can allow all sub-processes and all parameters to be adjusted at the same time, and can obtain more accurate optimization results. 2. Moreover, in the process optimization method and dynamic process margin checking method of the present invention, for a single parameter, only the relationship between the number of the parameter and the product yield rate can be considered, and this parameter can be set. Optimization. Therefore, the present invention can optimize multiple sub-processes or multiple parameters in a relatively simple manner, while taking into account the accuracy of the optimization results. Although the present invention has been disclosed as above with a preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art can make some changes and retouch without departing from the spirit and scope of the present invention. The scope of protection of the invention shall be determined by the scope of the attached patent application. [Schematic description] Figure 1 shows a schematic diagram of the process margin plane of a sub-process. Figure 2 is a schematic diagram showing the yield and characteristics of the product. 12 1222151 11638twf.doc / 006 Figure 3 shows a schematic diagram of the overall process margin obtained by overlapping the process margin planes of a plurality of sub-processes. Figure 4 is a schematic diagram of the conventional process margin checking method to check the overall process margin. FIG. 5 is a flowchart of optimizing a process by using a dynamic process margin check method according to a preferred embodiment of the present invention. Figure 6 is a schematic diagram showing the relationship between the yield of the product and the number of parameters A. Fig. 7 is a schematic diagram showing the relationship between the yield of the product and the number of parameters A. _ Figure 8 is a schematic diagram of checking the overall process margin using a dynamic process margin inspection method according to a preferred embodiment of the present invention. [Explanation of Graphical Symbols] 100 !, 1002, 1003, 1004, 100n_2, 100w, 100n: planes representing sub-processes-200: planes representing yield or characteristics S302, S304, S306, S308, S310, S312, S314, S316: Step 13