JP4476551B2 - Plasma processing apparatus and processing method - Google Patents

Description

【００４８】
データベース２４には、フィルタベクトルｆを蓄積でき、演算処理部２３で出力信号ｚを算出するときにフィルタベクトルｆを演算処理部２３に送ることができる。
【００４９】
次に、信号フィルタの作成方法を示す。演算処理部２３の入力信号は数個であることもあるが、前記のように発光スペクトルなどを入力に採る場合には数百から数千の入力信号となる。そのため信号フィルタの生成には多変量解析を用いるのが有効である。本実施例では多変量解析の一つである主成分解析を用いる。主成分解析は統計処理の一般的な手法であり、例えば、上記非特許文献１に具体的な計算手法が記されている。
【００５０】
前記入力信号がｎ個ある入力信号ベクトルｓを行にとり、ｍ個の取得された入力信号ベクトルを列の方向に重ねていくと下記式（２）のようなｍ行ｎ列の信号行列Ｓが得られる。
【００５１】
【数２】

【００５２】
次に、信号行列Ｓの相関行列あるいは共分散行列Ｘを作成し、ｎ行ｎ列の対称行列であるＸの固有値解析を行うと、ｎ個の正の実数の固有値｛λ１、λ２、…、λｎ｝と各固有値に対応するｎ個の固有ベクトル｛ｙ１、ｙ２、…、ｙｎ｝が得られる。この際、固有値は値が大きい順にλ１、λ２、λ３、…、λｎと並べることとする。
【００５３】
固有ベクトルｙは、それぞれｎ行の一次元ベクトルである。この固有ベクトルを前記のフィルタベクトルｆとして用い、式（１）の計算をすることによりｎ個の入力信号を一つの装置状態信号に変換する信号フィルタとなる。
【００５４】
さらに、演算処理装置２３において生成された出力信号ｚには、後述するように、さらなる演算処理が行なわれ、その演算結果（出力信号）は判断部２５に送られる。判断部２５では、その演算結果（出力信号）を基に装置制御のための判断を行ない、シーズニングを続行するか否かの装置制御信号を生成できる。判断部２５によって、生成された装置制御信号は、装置制御部２６に送られる。
【００５５】
装置制御部２６は、プラズマエッチング装置の制御を行うものであり、例えばシーズニングを実行せよという装置制御信号を受ければシーズニングの続行の動作を行い、シーズニングを終了せよという装置制御信号を受ければシーズニングの終了動作を行う。
【００５６】
次に図２〜図７および図１４を用いて、第１実施例における処理について具体的に説明する。
【００５７】
図２において、まず、シーズニングにおける１ロットのダミー処理の発光データの取得動作を行う（ステップＳ２１）。これは前述したように、受光部２０によって採光したプラズマ発光が分光部２２において、あるサンプリング間隔で多チャンネルの信号に変換され、その信号が演算処理部２３に送られるまでの過程を指す。なお多チャンネルの信号からなるプラズマ発光データは、現在進行中のひとまとまり（ここではロット）のダミー処理が終了するまでデータベース２４に蓄積される。
【００５８】
次に、プラズマ発光を示す数値データの中から有効部分の抽出を行なう（ステップＳ２２）。主成分解析は大きな変動を抽出する性質を持っている。そのため、プラズマの着火から消失までの一連の発光データを主成分解析すると、プラズマの着火時もしくは消失時におけるプラズマの大きな変動を拾ってしまう。また同様にプラズマの着火から消失までの一連の発光データを主成分解析すると、処理シーケンスが異なる処理条件を持つ複数のステップから成る場合に、ステップ間の移行の際に処理条件の変化に起因するプラズマの大きな変動を拾ってしまう。これらの大きな変動は、小さいが処理室内の状態の推移を示す重要な変動を埋もれさせてしまう事態を引き起こす。これを防止するためにはプラズマ着火時・消失時およびステップ間移行時を除いたプラズマ発光データのみを用いて主成分解析を行なうことが必要である。
【００５９】
図３および図４を用いて、有効部分の抽出動作Ｓ２２について具体的に説明を行なう。図３は、１枚のダミーウェハを用いて行なったダミー処理のプラズマ発光を、着火から消失まで１秒間隔でサンプリングし、そのプラズマ発光データを主成分解析した場合の第１主成分得点を示すグラフである。領域３０はプラズマ着火およびその直後の不安定に起因するプラズマの変動を示している。また、領域３１および領域３２は、ダミー処理の第１ステップおよび第２ステップにおけるプラズマの変動を示している。また、領域３３はプラズマの消失に起因するプラズマの変動を示している。この図３に示したグラフにおいては、第１ステップおよび第２ステップ中におけるプラズマの変動がほとんど把握できていないことがわかる。
【００６０】
次に、図４に、プラズマの着火から１７秒後から５２秒後までのプラズマ発光データを主成分解析した場合の第１主成分得点の挙動を表すグラフを示す。領域３２´はダミー処理の第２ステップにおける処理に相当する。この図４から、第２ステップにおけるプラズマ変動が把握できることがわかる。これにより、処理室内状態の微小な変動を把握することができる。なお、プラズマ発光データのうち、プラズマ着火から何秒から何秒までが有効部分であり主成分解析に用いるか、もしくはプラズマ着火からデータの何サンプリング目から何サンプリング目までが有効部分であり主成分解析に用いるかは、予めデータベース２４に登録されており、演算処理部２３において主成分解析を行う際に、プラズマ発光データの有効部分の指定が行われる。
【００６１】
なお、シーズニングにおけるダミー処理は全く同じ処理シーケンスで繰り返し行なうため、予めプラズマ発光データの有効部分を指定し、データベース２４に登録していれば、以後のダミー処理においても同じものを使うことができる。
【００６２】
また、ここでは、ダミー処理の第２ステップをプラズマ発光データの有効部分としたがそれに限るものではない。例えば、ダミー処理の第１ステップをプラズマ発光データの有効部分としてもよい。
【００６３】
次に、ステップＳ２２で得られた有効部分におけるプラズマ発光データを用いて主成分解析を行ない、出力信号ｚの算出を行なう（ステップＳ２３）。ここでは主成分解析によって得られた第１主成分得点が出力信号ｚとなる。
【００６４】
次に、算出動作Ｓ２３の主成分解析によって算出された第１主成分得点の、１回のダミー処理における平均値の算出を行なう（ステップＳ２４）。主成分得点は図４に示したように、あるサンプリング間隔（ここでは１秒）で得られたデータである。各々のダミー処理において、この第１主成分得点の平均値を算出する。ｋ枚目のダミー処理において、ｉ回目のサンプリングにおける第１主成分得点をａ_ｉとし、ダミー処理のプラズマ発光データの有効範囲がＮ回のサンプリングによる第１主成分得点ａ_ｉから成れば、ｋ枚目のダミー処理における第１主成分得点の平均値Ｐ_ｋは、下記式（３）で表される。
【００６５】
【数３】

【００６６】
図５は、ｋ枚目のダミー処理における第１主成分得点の平均値Ｐ_ｋの１ロット２５枚のダミー処理の４ロット分の推移を示したものである。この図５によってシーズニングによって処理室３内の状態が推移していることがわかるが、このままではシーズニングの終点は検出できない。
【００６７】
次に、各々のダミー処理の第１主成分得点の平均値Ｐ_ｋにおいて、注目しているロットの第１主成分得点と、その前のロットの第１主成分得点との差を算出する（ステップＳ２５）。ｉロット目のｊ枚目のダミー処理における第１主成分得点の平均値ＰをＰ_{ｉ , ｊ}とすると、ｉロット目のｊ枚目のダミー処理における第１主成分得点Ｐ_{ｉ , ｊ}とｉ−１ロット目のｊ枚目のダミー処理における第１主成分得点Ｐ_{ｉ -1, ｊ}との差Ｄ_{ｉ , ｊ}は,下記式（４）で表される。
【００６８】
【数４】

【００６９】
シーズニング処理によって処理室内の状態が量産安定時に近付けば、ロット間においてダミー処理が再現性良く行なわれる。そのため、処理が終わったロットにおける第１主成分得点Ｐ_i,jと、直前のロットにおける第１主成分得点Ｐ_{i-1, ｊ}との差Ｄ_i,jによって、ダミー処理のロット間再現性を評価することができ、その結果シーズニングが十分であるか否かの判定、つまりシーズニング終点判定を行なうことができる。
【００７０】
図６にＤ_i,jの挙動を表すグラフを示す。この図６から、ダミー処理の進行と共にＤ_i,jの値がロット間で全体的に下がっており、また、Ｄ_i,jのロット内のばらつきが小さくなっていることがわかる。
【００７１】
なお、ダミー処理の１ロット目には、Ｄ_i,jを算出するこのステップＳ２５は省かれる。
【００７２】
次に、Ｄ_i,jのロット内の平均値、Ｄ_i,jのロット内の最大値と最小値との差、Ｄ_i,jのロット内の標準偏差を算出する（ステップＳ２６）。図６に示したＤ_i,jがロット間で全体的に下がる挙動は、下記式（５）に示すｎ回のダミー処理から成るロット内のＤ_i,jの平均値ＡＶＥで評価することができる。なお、式（５）において、ＡＶＥの添え字のｉはｉロット目のロット内のＤ_i,jの平均値ＡＶＥであることを示している。
【００７３】
【数５】

【００７４】
さらに図６に示したＤ_i,jのロット内の変動は、下記式（６）で示されるｉロット目のロット内のＤ_i,jの最大値と最小値との差ＭＡＸＭＩＮ_ｉ、および下記式（７）で示されるｉロット目のロット内のＤ_i,jの標準偏差σ_ｉで評価することができる。
【００７５】
【数６】

【００７６】
【数７】

【００７７】
次にシーズニングが十分か否かの判定を行なう（ステップＳ２７）。
【００７８】
図７（ａ）にＤ_i,jのロット内の平均値ＡＶＥ_ｉを示すグラフを、図７（ｂ）にＤ_i,jのロット内の最大値と最小値との差ＭＡＸＭＩＮ_ｉを示すグラフを、図７（c）にＤ_i,jロット内の標準偏差σ_ｉを示すグラフをそれぞれ示す。これらからシーズニングの進行に伴ってそれぞれの値が小さくなっていることがわかる。これは、ウェットクリーニングＳ１によって変化した処理室３内の状態が、シーズニングによって量産安定時に近づいたため、ダミー処理においてロット間で再現性良くプラズマ処理が行なわれるようになり、それを反映した第１主成分得点のＤ_i,jの値がシーズニングの進行と共に小さくなったことを示している。
【００７９】
また、第１主成分得点を基に算出されたＡＶＥ_ｉ、ＭＡＸＭＩＮ_ｉ、およびσ_ｉが、予め設定された閾値以下になっていれば、処理室内の状態が量産安定時に近づいた、つまりシーズニングが終点に達したと判断すれば良い。
【００８０】
例えば、この実施例においては、ＡＶＥ_ｉにおいてはＡＶＥ_ｉについて予め設定された閾値４０以下に、ＭＡＸＭＩＮ_ｉにおいてはＭＡＸＭＩＮ_ｉについて予め設定された閾値４２以下に、σ_ｉについてはσ_ｉについて予め設定された閾値４４以下にそれぞれなっているために、終点に達したと判断した。
【００８１】
図７に示したＡＶＥ_ｉ、ＭＡＸＭＩＮ_ｉ、σ_ｉは、それぞれの値がｉ=４において閾値よりも小さくなっているため、４ロット目のダミー処理によってシーズニングの終点に達した、つまりこのシーズニングは４ロット目のダミー処理で十分と判断できる。
【００８２】
なお、ＡＶＥ_ｉ、ＭＡＸＭＩＮ_ｉ、およびσ_ｉの３つの値のうち、ＡＶＥ_ｉは負の値をとり得る。ＡＶＥ_ｉが負の値をとった場合にはシーズニングの進行と共に絶対値が小さくなる挙動を示すため、ＡＶＥ_ｉの絶対値が予め決められた閾値以下になっていれば、処理室内の状態が量産安定時に近づいた、つまりシーズニングが終点に達したと判断すれば良い。
【００８３】
本第１実施例においてはＡＶＥ_ｉ、ＭＡＸＭＩＮ_ｉ、およびσ_ｉの３つの値が全てｉ=４においてそれぞれの閾値よりも小さくなったため、４ロット目のダミー処理によってシーズニングの終点に達したと判断したが、これら３つの値の全てが閾値よりも小さくなったときにシーズニング終点と見なすと限るものではない。例えばＡＶＥ_ｉ、ＭＡＸＭＩＮ_ｉ、およびσ_ｉのうち１つでも閾値よりも小さくなればシーズニング終点と見なしても構わないし、例えばＡＶＥ_ｉ、ＭＡＸＭＩＮ_ｉ、およびσ_ｉのうち２つがそれぞれの閾値よりも小さくなればシーズニング終点と見なしても構わず、これらは本発明の実施者が選択できる。
【００８４】
次に、判定動作Ｓ２７においてシーズニングが十分と判定された場合はシーズニングの終了動作を行なう（ステップＳ１０）。このときは図１の判断部２５からシーズニングを終了せよという信号が装置制御部２６に送られ、シーズニングの終了動作Ｓ１０を行ない、製品基板着工Ｓ１１を行なえるようにする。また、このときにシーズニングが終了したことを作業者に知らせてもよい。
【００８５】
一方、判定動作Ｓ２７においてシーズニングが不十分と判定された場合は、図１の判断部２５からシーズニングを続行せよという信号が装置制御部２６に送られ、シーズニングの続行を行なう（ステップＳ２８）。
【００８６】
このときはまたステップＳ２１に戻り、１ロットのダミー処理の発光データの取得を行なう。
【００８７】
以上に示したシーケンスを行なうことにより、作業者が手作業で確認処理Ｓ９を行なうことなくシーズニングが十分か否かを判定することが可能となる。また、必要最小限のロット数のダミー処理によってシーズニングを終了させることが可能となるため、ダミー基板の使用量の低減、およびプラズマ処理装置の稼働率の向上が実現できる。
【００８８】
また、第１実施例に示したシーズニングの終点判定においては、ｉロット目のｊ枚目のダミー処理における第１主成分得点Ｐ_ｉとｉ-１ロット目のｊ枚目のダミー処理における第１主成分得点Ｐ_i-1,jとの差Ｄ_i,jを用いた。すなわち連続する一連のプラズマ処理における発光データから得られた出力信号ｚのロット間の差を用いているため、ウェットクリーニングＳ１の仕上がりのばらつきがシーズニングの終点検出の精度に与える影響は小さいものになる。
【００８９】
なお、本第１実施例ではステップＳ２３において、現在進行中のシーズニングでのダミー処理の発光を対象に主成分解析を行うことでフィルタベクトルｆを算出し、その結果、出力信号ｚを算出したが、それに限るものではない。ダミー処理の発光から得られた入力信号ベクトルｓおよび、予め設定されたフィルタベクトルｆを用いて式（1）から出力信号ｚを算出することもできる。
【００９０】
その際に用いるフィルタベクトルｆは、前回のウェットクリーニング直後のシーズニングでのダミー処理の発光を対象に主成分解析を行って得られたものを使用できる。また、前回のウェットクリーニングではなく、２回前あるいは３回前のウェットクリーニング直後のシーズニングでのダミー発光を対象に主成分解析を行うことで得られたフィルタベクトルｆを使用してもよい。あるいは、プラズマ処理装置メーカーが予めフィルタベクトルｆを設定しておくこともできる。
【００９１】
また、本第１実施例では、１ロット２５枚のダミー処理をひとまとまりとしたが、それに限るものではない。ウェハの１ロットが２５枚から成っていたとしても、シーズニング処理において例えば５枚のダミー処理ごとにドライクリーニングが行われている場合、すなわちダミー処理Ｓ８-２〜８-２´´あるいはダミー処理Ｓ８-３〜８-３´´が５枚のダミー処理から構成されている場合には、ダミー処理のひとまとまりは、５枚となる。その場合は、ひとまとまりのダミー処理の発光データの取得動作Ｓ２１からシーズニングが十分か否かの判定を行う判定動作Ｓ２７までは、５枚のダミー処理ごとに行われる。
【００９２】
さらに、各々のダミー処理Ｓ８−２〜Ｓ８−２´´の後に枚葉クリーニングが行われている場合でも、ダミー処理Ｓ８−３〜Ｓ８−３´´が５枚のダミー処理から構成されている場合には、ダミー処理のひとまとまりは、５枚となる。その場合は、ひとまとまりのダミー処理の発光データの取得動作Ｓ２１からシーズニングが十分か否かの判定を行う判定動作Ｓ２７までは、５枚のダミー処理ごとに行われる。
【００９３】
また、本第１実施例では算出動作Ｓ２３の出力信号ｚの算出において主成分解析を行ない、出力信号ｚとして第１主成分得点のみを用いたが、それに限るものではない。複数の主成分得点を出力信号ｚとして用いることもできる。
【００９４】
以下、本発明の第２の実施例について図８〜１１を用いて説明する。第１実施例では出力信号ｚとして第１主成分得点のみを用いたが、本実施例では第２主成分得点および第３主成分得点を使用するものである。主成分解析では高次の主成分ほどノイズ成分が多くなり、処理室内状態のモニタリングの有効性が低くなるため、本実施例では第３主成分までを使用する。第２主成分得点を用いて、第１実施例と同様の方法で得られたＡＶＥ_ｉを図８（ａ）に、第２主成分のＭＡＸＭＩＮ_ｉを図８（ｂ）に、第２主成分のσ_ｉを図８（ｃ）に、それぞれ示す。また、第３主成分得点を用いて、第１実施例と同様の方法で得られたＡＶＥ_ｉを図９（ａ）に、第３主成分のＭＡＸＭＩＮ_ｉを図９（ｂ）に、第３主成分のσ_ｉを図９（ｃ）に、それぞれ示す。これらの値も図７に示した第１主成分のＡＶＥ_ｉ、ＭＡＸＭＩＮ_ｉ、σ_ｉと同様にシーズニングが進行するほど低下していることがわかる。これらの値を予め設定された閾値と比較し、閾値以下になっていれば、処理室内の状態が量産安定時に近づいた、つまりシーズニングが終点に達したと判断すれば良い。
【００９５】
なお、第１主成分のＡＶＥ_ｉに対して設定された閾値４０と、第１主成分のＭＡＸＭＩＮ_ｉに対して設定された閾値４２と、第１主成分のσ_ｉに対して設定された閾値４４と、第２主成分のＡＶＥ_ｉに対して設定された閾値４６と、第２主成分のＭＡＸＭＩＮ_ｉに対して設定された閾値４８と、第２主成分のσ_ｉに対して設定された閾値５０と、第３主成分のＡＶＥ_ｉに対して設定された閾値５２と、第３主成分のＭＡＸＭＩＮ_ｉに対して設定された閾値５４と、第３主成分のσ_ｉに対して設定された閾値５６とは全て同じである必要は無く、それぞれに対して適した閾値を設定することが望ましい。
【００９６】
また、第１主成分のＡＶＥ_ｉ、ＭＡＸＭＩＮ_ｉ、σ_ｉ、および第２主成分のＡＶＥ_ｉ、ＭＡＸＭＩＮ_ｉ、σ_ｉ、および第３主成分のＡＶＥ_ｉ、ＭＡＸＭＩＮ_ｉ、σ_ｉの値が全て、それぞれ予め設定する閾値よりも小さくなった場合にシーズニングが十分であるとみなしても良いが、その限りではない。例えば、図１０に示すグラフのように２つの閾値５８および閾値５８´を設定し、ＡＶＥ_ｉ、ＭＡＸＭＩＮ_ｉ、σ_ｉの値が閾値５８以上であれば０点、閾値５８未満かつ閾値５８´以上であれば１点、閾値５８´未満であれば２点の得点を与え、図１１示した表のように、その合計得点が予め設定された閾値（例えば１２点）よりも大きければシーズニングが十分であると判定することもできる。
【００９７】
本第２実施例においては、ｉ=２のときに合計得点６０（ここでは１点）。ｉ＝３のときに合計得点６１（ここでは４点）、ｉ＝４のときに合計得点６２（ここでは１７点）となり、合計得点があらかじめ設定された閾値（ここでは１２点）よりも大きくなったのがｉ＝４であるため、４ロット目でシーズニングが終点に達したと判断できる。
【００９８】
以上示した、本第２実施例のように、複数の出力信号ｚから得られたＡＶＥ_ｉ、ＭＡＸＭＩＮ_ｉ、σ_ｉを用いることによりシーズニングが十分であるか否かの判定の精度を向上させることができる。
【００９９】
なお、第２実施例では閾値は数値を設定していたが、それに限るものではなく、割合を予め設定し、それを用いて閾値を決定しても良い。例えば、ｉ＝２におけるＡＶＥ_ｉ、ＭＡＸＭＩＮ_ｉ、σ_Iの５０％の値を閾値５８に（設定割合５０％）、２０％の値を閾値５８´（設定割合２０％）にし、それらの閾値を用いて以後のＡＶＥ_ｉ、ＭＡＸＭＩＮ_ｉ、σ_Iに得点を与えてもよい。
【０１００】
なお、本第２実施例では、第１主成分得点および第２主成分得点および第３主成分得点から得たＡＶＥ_ｉ、ＭＡＸＭＩＮ_ｉ、σ_ｉに対して２つの閾値５８および閾値５８´を設定したが、閾値の個数は２個に限るものではない。
【０１０１】
また、それぞれの閾値と比較して与える得点は、ＡＶＥ_ｉ、ＭＡＸＭＩＮ_ｉ、σ_ｉに対して全て同一（ここでは０点、１点、２点）のものとしたが、それに限るものではない。例えば第２主成分得点から得られるＡＶＥ_ｉ、ＭＡＸＭＩＮ_ｉ、σ_ｉの重要度が他の主成分得点から得られるＡＶＥ_ｉ、ＭＡＸＭＩＮ_ｉ、σ_ｉの重要度よりも高ければ、第２主成分得点から得られるＡＶＥ_ｉ、ＭＡＸＭＩＮ_ｉ、σ_ｉに与える得点を大きくしても構わない（例えば０点、２点、４点）。また同様にＡＶＥ_ｉ、ＭＡＸＭＩＮ_ｉ、σ_ｉのうち、ＡＶＥ_ｉの重要度がＭＡＸＭＩＮ_ｉおよびσ_ｉの重要度よりも高ければ、ＡＶＥ_ｉに対して与える得点を大きくしても構わない（例えば０点、２点、４点）。
【０１０２】
なお、本第２実施例では出力信号ｚとして第１主成分得点および第２主成分得点および第３主成分得点を用いたが、使用する主成分得点は３つに限るものではない。
【０１０３】
なお、本第２実施例では、ステップＳ２３において、現在進行中のシーズニングでのダミー処理の発光を対象に主成分解析を行うことでフィルタベクトルｆを算出し、その結果、出力信号ｚを算出したが、それに限るものではない。ダミー処理の発光から得られた入力信号ベクトルｓ、および予め設定された複数のフィルタベクトルｆを用いて式（１）を用いることによって、複数の出力信号ｚを算出することもできる。
【０１０４】
その際用いる複数のフィルタベクトルｆは、前回のウェットクリーニング直後のシーズニングでのダミー処理の発光を対象に主成分解析を行い得られたものを使用できる。また、前回のウェットクリーニングではなく、２回前あるいは３回前のウェットクリーニング直後のシーズニングでのダミー処理の発光を対象に主成分解析を行うことで得られた複数のフィルタベクトルｆを使用してもよい。あるいは、プラズマ処理装置メーカーが予めフィルタベクトルｆを複数設定しておくこともできる。
【０１０５】
次に、本発明の第３実施例について図１および図１２を用いて説明する。
【０１０６】
本第３実施例は、ウェットクリーニング後の真空引きの開始から製品基板処理着工Ｓ１０までを、作業者の手作業を行なうことなく自動的に行なうものであり、真空引きを行う過程と、真空度が十分か否かを判断する過程と、装置異常が無いか否かを判断する過程とを、前記した処理に付加することで実現される。
【０１０７】
まず、プラズマ処理室３のウェットクリーニングを行なう（ステップＳ１）。
【０１０８】
その後処理室蓋２を閉じた後に、排気口１０に接続された真空ポンプ（図示しない）を始動し、処理室３の真空引き動作Ｓ２を開始する。
【０１０９】
その後、予め設定された真空度に達したか否かの確認動作Ｓ３を自動的に行なう。これは処理室３に接続された圧力計によって処理室３内の圧力を自動的に測定することによって行なわれる。このとき予め設定された真空度に達していなければ、さらに真空引きを続行する。さらに、予め設定された時間、真空引き動作Ｓ２を行なった後においても、予め設定された真空度に達していなければ自動的に警報動作Ｓ４を行う。これはブザーを鳴らしたり、プラズマ処理装置に設置されたモニタ画面に警報を表示しても良いし、またプラズマ処理装置がコンピュータネットワークに接続されていれば離れた場所にいるエンジニアにＥ−Ｍａｉｌを送信しても良いし、ページャーなどで知らせても良い。また警報を発するだけではなく、処理室３に接続された真空ポンプ（図示しない）の作動を止めても良い。
【０１１０】
確認動作Ｓ３において予め設定された真空度に達していれば、次にプラズマ処理装置に異常が無いか否かの確認動作Ｓ５を自動的に行なう。
【０１１１】
この確認動作Ｓ５の一例として、処理室３のリークチェック、つまり単位時間あたりの空気の混入量の確認動作が挙げられる。この処理室３への単位時間当たりの空気の混入量Ｌは、処理ガス導入管バルブ７とマスフローコントローラ８および排気管バルブ１４を閉じ、排気管バルブ１４を閉じている時間Ｔおよび排気管バルブ１４を閉じている間の処理室３の圧力上昇ΔＰおよび処理室３の容積ｖを用いて下記式（８）を用いて算出される。
【０１１２】
【数８】

【０１１３】
この（式８）によって算出される、単位時間あたりの空気混入量Ｌが、予め設定された閾値よりも小さければ、リークチェックは合格となる。
【０１１４】
また、この確認動作Ｓ５の他の一例として、処理室３に導入する処理ガスの流量を調節するマスフローコントローラ８の確認動作が挙げられる。これはマスフローコントローラ８に所定の流量だけ処理室３に導入するように設定し、また同時に、処理室３に接続された圧力計１５によって、処理ガス６の導入による圧力上昇を求めることによって、マスフローコントローラ８が設定された流量の処理ガスを正しく処理室３に導入しているか否かの確認を行なうものである。このマスフローコントローラ８への流量設定値と、処理室３の圧力上昇から算出される処理ガス６の流量との差が、予め設定された閾値よりも小さければ、マスフローコントローラ８のチェックは合格となる。
【０１１５】
確認動作Ｓ５によってプラズマ処理装置に異常があると判断された場合は、自動的に警報を発する（ステップＳ４´）。これはブザーを鳴らしたり、プラズマ処理装置に設置されたモニタ画面に警報を表示しても良いし、またプラズマ処理装置がコンピュータネットワークに接続されていれば離れた場所にいるエンジニアにＥ−Ｍａｉｌを送信しても良いし、ページャーなどで知らせても良い。また、警報を発するだけではなく、処理室３に接続された真空ポンプ（図示しない）の作動を止めても良い。
【０１１６】
確認動作Ｓ５によってプラズマ処理装置に異常が無いと判断された場合は、プラズマ処理装置に予めセットされていたダミー基板が処理室３内に自動的に搬入され、シーズニングを自動的に開始する。まずドライクリーニングＳ６−２を行ない、その後ダミー処理Ｓ８−２〜Ｓ８−２´´を行なう。このダミー処理Ｓ８−２〜Ｓ８−２´´におけるプラズマ発光を受光部２０によって採光し、第１〜第２実施例で示したような処理を行なうことで出力信号ｚが算出される。
【０１１７】
次に、第１〜第２実施例で示したような処理を行なうことでシーズニングが十分か否かの判定処理（ステップＳ２７）を自動的に行なう。
【０１１８】
なお、ダミー処理がひとまとまり１回目ならば、この判定処理Ｓ２７は省かれ、自動的にドライクリーニングＳ６−２に移る。
【０１１９】
次に、判定処理Ｓ２７によってシーズニングが不十分と判断された場合には、自動的にドライクリーニングＳ６−２に移る。また、判定処理Ｓ２７においてシーズニングが十分と判定された場合はシーズニングの終了動作Ｓ１０を行なう。このときは図１の判断部２５でシーズニングを終了せよという信号が装置制御部２６に送られ、シーズニングの終了動作Ｓ１０を行ない、製品基板着工Ｓ１１を行なえるようにする。また、このときにシーズニングが終了したことを作業者に知らせてもよい。
【０１２０】
以上に示したシーケンスを行なうことにより、ウェットクリーニング後に真空引きを開始した後に、自動的にシーズニングの前準備およびシーズニングを行い、かつシーズニングが十分か否かを自動的に判断し、製品基盤の着工準備を行うことができる。そのため、従来は必要だった作業者による確認処理Ｓ９を行うことなく、シーズニングが十分か否かを自動的に判定することが可能となる。また、必要最小限の枚数のダミー処理によってシーズニングを終了させることが可能となるため、ダミー基板の使用量の低減およびシーズニングに要する時間の短縮によるプラズマ処理装置のスループットの向上が実現できる。
【０１２１】
なお、第１〜第３実施例においてはシーズニングの終点判定に使用するプラズマ発光データは、ダミー処理によるものだったが、それに限るものではない。例えば、各々のダミー処理の後に枚葉クリーニングを行なっていれば、その枚葉クリーニングにおけるプラズマ発光を用いて、第１〜第３実施例と同様の方法により、シーズニングが十分か否かの判定を行なうことができる。
【０１２２】
なお、第１〜第３実施例においては、シーズニングにおけるプラズマ処理のプラズマ発光を対象に行う多変量解析の方法として、主成分解析を用いたが、それに限るものではない。
【０１２３】
なお、第１〜第３実施例においては、プラズマエッチング装置を例に取り説明したが、本発明の適用はプラズマエッチング装置に限るものではなく、プラズマＣＶＤ装置など他のプラズマ処理装置にも適用できる。
【０１２４】
なお、第１〜第３実施例においてはプラズマ処理を行う基板として半導体基板を例にとり説明したが、本発明の適用は半導体基板に装置に限るものではなく、ＬＣＤ基板など他の基板のプラズマ処理にも適用できる。
【０１２５】
【発明の効果】
以上説明したように、本発明によればプラズマ発光を分光し、そのデータを対象にして多変量解析を行い、算出された出力信号について、前のひとまとまりのプラズマ処理での出力信号と差を取り、ひとまとまりの中での平均値、ひとまとまりの中での最大値と最小値との差、ひとまとまりの中での標準偏差を算出し、それらを予め設定された閾値と比較することによってシーズニングが十分であるか否かの判定を行なうことができる。これによりシーズニングでおこなうダミー処理枚数を低減できるプラズマ処理装置およびプラズマ処理方法を提供できる。
【図面の簡単な説明】
【図１】本発明の第１実施例を示すプラズマエッチング装置の側断面図。
【図２】本発明の第１実施例におけるプラズマ発光データの処理方法を示すフローチャート。
【図３】本発明の第１実施例における出力信号ｚ（第１主成分得点）の挙動を示すグラフ。
【図４】本発明の第１実施例における発光データの有効領域から得られた出力信号ｚ（第１主成分得点）の挙動を示すグラフ。
【図５】本発明の第１実施例における発光データの有効領域から得られた出力信号ｚ（第１主成分得点）の各々のダミー処理の平均値の挙動を示すグラフ。
【図６】本発明の第１実施例における出力信号ｚ（第１主成分得点）の各々のダミー処理での、平均値Ｐ_ｋのロット間の差から得られたＤ_i,jの挙動を示すグラフ。
【図７】本発明の第１実施例における出力信号ｚ（ここでは第１主成分得点）を基に算出されたＡＶＥ_ｉと、ＭＡＸＭＩＮ_ｉと、σ_ｉとの挙動を示すグラフ。
【図８】本発明の第２実施例における出力信号ｚ（ここでは第２主成分直点）を基に算出されたＡＶＥ_ｉと、ＭＡＸＭＩＮ_ｉと、σ_ｉとの挙動を示すグラフ。
【図９】本発明の第２実施例における出力信号ｚ（ここでは第３主成分得点）を基に算出されたＡＶＥ_ｉと、ＭＡＸＭＩＮ_ｉと、σ_ｉとの挙動を示すグラフ。
【図１０】本発明の第２実施例における出力信号ｚを基に算出されたＡＶＥ_ｉと、ＭＡＸＭＩＮ_ｉと、σ_ｉに対して得点を与える方法を示す図。
【図１１】本発明の第２実施例における出力信号ｚ（ここでは第１主成分得点および第２主成分得点および第３主成分得点）を基に算出されたＡＶＥ_ｉと、ＭＡＸＭＩＮ_ｉと、σ_ｉに対して与えた得点を用いてシーズニングが十分であるか否かを判定する方法を示す図表。
【図１２】本発明の第３実施例における、ウェットクリーニング後に行なう処理室の真空引きおよび真空度確認動作および装置異常確認動作およびシーズニングが十分であるか否かを判定を全て自動的に行なう方法を示したフローチャート。
【図１３】一般的に行なわれている製品基板処理の従来例を示すフローチャート。
【図１４】ウェットクリーニング後に行われているシーズニングの従来例を示すフローチャート。
【符号の説明】
１…処理室壁、 ２…処理室蓋、 ３…処理室、 ４…基板保持台、 ５…半導体基板、 ６…処理ガス、 ７…処理ガス導入管バルブ、 ８…マスフローコントローラ、 ９…処理ガス導入管、 １０…マイクロ波、 １１…プラズマ、１２…排気口、 １３…可変バルブ、 １４…排気管バルブ、 １５…圧力計、 ２…受光部、 ２１…光ファイバ、 ２２…分光部、 ２３…演算処理部、２４…データベース、 ２５…判断部、 ２６…装置制御部
Ｓ１ ウェットクリーニング、 Ｓ２…真空引き動作、Ｓ３…確認動作、 Ｓ４…警報動作、 Ｓ５…確認動作、 Ｓ６…ドライクリーニング、 Ｓ７…エージング、 Ｓ９…確認処理、 Ｓ１０…シーズニング終了動作、 Ｓ１１…製品基板着工、 Ｓ２１…発光データの取得、Ｓ２２…有効部分の抽出動作、 Ｓ２３…算出動作、 Ｓ２７…判定動作、 Ｓ２８…シーズニング続行。[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a plasma processing apparatus for processing a semiconductor substrate such as a semiconductor wafer, an LCD substrate, and the like, and a plasma processing method in the plasma processing apparatus.
[0002]
[Prior art]
Conventionally, plasma processing apparatuses such as a plasma etching apparatus using a reactive plasma and a plasma CVD (Chemical Vapor Deposition) apparatus have been used in the manufacturing process of a semiconductor substrate such as a semiconductor wafer and an LCD (Liquid Crystal Display) substrate.
[0003]
These plasma processing apparatuses have a processing chamber that accommodates an object to be processed such as a semiconductor substrate, and non-volatile reaction products generated during the plasma processing of the object to be processed are deposited on the wall of the processing chamber. In the subsequent processing, particles generated by the separation of the deposit from the processing chamber wall may fall and adhere to the surface of the object to be processed. The adhesion of particles causes a short circuit, disconnection, or etching residue of the integrated circuit wiring provided on the surface of the object to be processed, resulting in a semiconductor device failure, and thus a decrease in yield in semiconductor manufacturing.
[0004]
In order to prevent this, in manufacturing semiconductors and LCDs, a process called dry cleaning such as gas cleaning using reactive gas or plasma cleaning using reactive plasma is periodically performed without returning the processing chamber to atmospheric pressure. In addition, the processing chamber is cleaned by removing the reaction products deposited on the processing chamber wall.
[0005]
Here, an example of plasma etching processing of a semiconductor substrate in mass production will be described with reference to FIG.
[0006]
First, the processing chamber of the semiconductor processing apparatus is dry cleaned (step S6). Thereby, the reaction product deposited on the inner wall of the processing chamber is removed. The dry cleaning S6 is often performed by installing a substrate called a dummy substrate, which is not a product substrate, in the processing chamber, but may be performed without installing the dummy substrate in the processing chamber.
[0007]
Next, a process called aging is performed (step S7). In general, since dry cleaning S6 uses a gas different from the processing gas used for the plasma etching process, the reaction product from the dry cleaning S6 remains on the inner wall of the processing chamber, and the state of the processing chamber changes. Therefore, processing is performed using processing conditions that are the same as or close to those of normal plasma etching processing in order to condition the processing chamber changed by the dry cleaning S6. The aging S7 is often performed with the dummy substrate installed in the processing chamber, but may be performed without installing the dummy substrate in the processing chamber.
[0008]
After the aging S7, if the dummy substrate is installed in the processing chamber, the dummy substrate is unloaded from the processing chamber, the product substrate is loaded into the processing chamber, and product substrate processing is performed by plasma etching (step S8). After the product substrate processing S8 is completed, the product substrate is carried out from the processing chamber, another product substrate is carried into the processing chamber, and the product substrate processing is performed (step S8 ′). This substrate processing is performed for a set number of sheets (steps S8 to S8 ″).
[0009]
The plasma etching process of the semiconductor substrate is performed by repeating these dry cleaning S6, aging S7, and product substrate processing S8 to S8 ″.
[0010]
Here, a supplementary explanation will be given regarding the number of continuous processing of substrate processing. In general, a plurality of semiconductor substrates are handled in a cassette. A group of a plurality of cassettes is sometimes called a lot. The number of semiconductor substrates in one lot is 25 for a 200 mm diameter semiconductor substrate and 13 for a 300 mm diameter semiconductor substrate. For this reason, the number of substrate processing batches described above is often the same as the number of semiconductor substrates in one lot. However, the number of semiconductor substrates in one lot may be different from the number of batches of product substrate processing S8 to S8 ″ described above. Further, when a plurality of products are processed by the same plasma processing apparatus, the number of the product substrate processes S8 to S8 ″ described above may differ depending on the product.
[0011]
Here, dry cleaning is not performed after each processing of the product substrate processing S8 to S8 ″, but dry cleaning, which is shorter in time than the dry cleaning S6 after each processing of the product substrate processing S8 to S8 ″, so-called “cleaning”. In some cases, single wafer cleaning is performed.
[0012]
As described above, the reaction product adhering to the processing chamber wall by the product substrate processing S8 to S8 '' is removed by performing the dry cleaning S6, but the deposits that could not be removed by the dry cleaning S6, After opening the processing chamber to the atmosphere, the worker manually cleans the processing chamber wall with water, alcohol, etc., so-called wet cleaning or manual cleaning is removed to clean the processing chamber. .
[0013]
In addition, since reactive plasma is used in the plasma processing apparatus as described above, components in the processing chamber may be chemically and thermally damaged, and may be worn or broken. For this reason, it is necessary to periodically replace components in the processing chamber, and components that have reached a predetermined life are replaced during wet cleaning.
[0014]
Since the state of the processing chamber wall immediately after wet cleaning is different from that when mass production is stable, in the plasma etching apparatus, the etching rate, the etching rate distribution in the surface of the semiconductor substrate, the etching selectivity between the object to be etched and the mask or the base, that is, There arises a problem that the processing performance such as the etching rate ratio and the processing shape by etching changes. Similarly, in the plasma CVD apparatus, there arises a problem that the processing performance such as the film forming speed, the film forming speed distribution in the semiconductor substrate surface, and the film quality changes.
[0015]
In mass production of semiconductors and LCDs, a process called seasoning is generally performed in order to prevent the occurrence of these problems, and the processing chamber state changed by wet cleaning is brought closer to the time of stable mass production. This seasoning is often performed by simulating the processing of the semiconductor substrate shown in FIG.
[0016]
Here, an example of wet cleaning and seasoning processing in a plasma etching apparatus for processing a semiconductor substrate will be described with reference to FIG. Seasoning includes dry cleaning, processing of a dummy substrate (a semiconductor substrate other than a product substrate), and confirmation processing of whether seasoning is sufficient.
[0017]
First, the plasma processing chamber of the plasma processing apparatus is wet cleaned (step S1).
[0018]
Thereafter, the processing chamber is evacuated, and after a predetermined degree of vacuum is reached, dry cleaning is performed (step S6-2). The organic solvent remaining on the processing chamber wall by the wet cleaning is removed by this dry cleaning S6-2.
[0019]
Next, an etching process (hereinafter referred to as a dummy process) simulating an actual process is performed using a dummy substrate (step S8-2). After performing this dummy process S8-2, the dummy substrate is unloaded from the process chamber, another dummy substrate is loaded into the process chamber, and the same dummy process as step S8-2 is performed (step S8-2 '). This dummy process is performed as a group (for example, 25 sheets) (step S8-2 ″).
[0020]
After performing the dummy processes S8-2 to S8-2 '', dry cleaning is performed (step S6-2). The reaction product adhering to the processing chamber wall by the dummy processing becomes a volatile compound by the dry cleaning S2, and is removed from the processing chamber. These processes, that is, the dry cleaning S6-2 and the dummy processes S8-2 to S8-2 ″ are repeated a predetermined number of times.
[0021]
Thereafter, a confirmation process is performed to determine whether seasoning is sufficient (step S9). For example, a predetermined film (for example, SiO₂Etching is performed using a wafer whose surface is coated with (silicon oxide)), and the average value of the etching rate within the wafer surface and the uniformity within the wafer surface are obtained. If the average value of the etching rate within the wafer surface and the uniformity within the wafer surface satisfy predetermined criteria, it is considered that seasoning is sufficient, and if the standard is not satisfied, seasoning is considered insufficient.
[0022]
If it is determined that the seasoning is sufficient in the confirmation process S9, the seasoning end operation (step S10) is performed, the product substrate process is started (step S11), and the plasma process of the product substrate is started.
[0023]
On the other hand, if it is determined that the seasoning is insufficient in the confirmation process S9, the same dry cleaning as in step S6-2 is further performed (step S6-3), and then the same as in steps S8-2 to S8-2 ″. Dummy processing is performed for a set number of sheets (steps S8-3 to S8-3 ″), and then confirmation processing is performed to determine whether seasoning is sufficient (step S9).
[0024]
In the seasoning, these processes S6-3 to S8-3 '' are performed until the seasoning is considered sufficient in the confirmation process S9.
[0025]
Here, aging is not performed after dry cleaning S6-2 and step S6-3, but aging as in step S7 of FIG. 13 may be performed after dry cleaning S6-2 and S6-3. is there.
[0026]
Here, dry cleaning is not performed after each of the dummy processes S8-2 to S8-2 ′ and S8-3 to S8-3 ″, but each dummy process S8-2 to S8-2 ′ is performed. , S8-3 to S8-3 ″ may be followed by dry cleaning that is shorter than dry cleaning S6-2 and S6-3, so-called single wafer cleaning.
[0027]
In the prior art described above, the number of dummy processes to be performed in seasoning has been determined based on experience. For example, if the dummy process in seasoning is determined to be 5 lots in advance, seasoning consisting of 5 lots of dummy processes is performed after the wet cleaning, and then the operator performs a confirmation process S9 to determine the film formation speed and the like. Evaluation of processing performance such as etching rate or in-plane uniformity was performed to determine whether seasoning was sufficient, that is, whether the condition of the processing chamber after wet cleaning was approaching when mass production was stable. .
[0028]
When seasoning is insufficient, the desired processing performance cannot be obtained in the processing of the product substrate, resulting in a decrease in yield in mass production. For this reason, dummy processing in seasoning is often estimated to be larger than the minimum required number. This extra dummy process causes a decrease in the throughput of the semiconductor manufacturing apparatus, which is undesirable because it deteriorates the production efficiency. In recent years, the diameter of semiconductor substrates has been increased, and the cost per dummy substrate has increased, and it has been desired to reduce the amount of dummy substrates used for dummy processing.
[0029]
On the other hand, variations in the finish of the wet cleaning S1 affect the processing chamber state in some cases, even when predetermined seasoning is performed, not approaching when mass production is stable. In this case, as shown in FIG. 14, it is necessary for the operator to repeatedly perform the dry cleaning S6-3, the dummy processes S8-3 to S8-3 ″, and the confirmation process S9. Cause.
[0030]
During seasoning, if it is possible to determine whether seasoning is sufficient and finish with the minimum required seasoning, the amount of dummy substrate used can be minimized, so seasoning is sufficient when mass production is stable. There is a need for a method for determining whether or not the vehicle is approaching, that is, a method for determining the end point of seasoning.
[0031]
The technology for determining the end point in the plasma processing of the product substrate is that the same substrate as the product substrate is processed in advance, the principal component analysis of the plasma emission spectrum is performed, and in the subsequent actual product substrate processing, the measured emission spectrum is measured. It has been proposed to perform principal component analysis and detect the end point of processing from changes in the principal component score (see, for example, Patent Document 1). This technique is an effective means for detecting the end points of the processing of each product substrate, but cannot detect the end points of the seasoning process, which is a continuous process using a plurality of substrates.
[0032]
On the other hand, as a technique for determining the end point of seasoning, the principal component analysis is performed on the electrical data of the high-frequency power supply of the stabilized processing apparatus, the calculated principal component score and residual score for the reference, and an arbitrary It has been proposed that principal component analysis is performed on a plurality of electrical data in the processing chamber state, and the state in the processing chamber is detected by comparing the calculated principal component score and residual score ( For example, see Patent Document 2). This technique is an effective means for detecting the state in the processing chamber that changes due to seasoning after wet cleaning. However, since wet cleaning is performed manually, there is a large variation in the finish, and as a result, a large variation occurs in the processing chamber state due to seasoning. In other words, the transition of the processing chamber state in different seasonings is poorly reproducible, and the end point of any seasoning can be determined by comparing the principal component score and residual score for the reference with the principal component score and residual score in any seasoning. In some cases, sufficient accuracy cannot be obtained.
[0033]
[Patent Document 1]
JP 2000-331985 A
[Patent Document 2]
Japanese Patent Laid-Open No. 2002-18274
[Non-Patent Document 1]
Okuno Tadaichi et al., Multivariate Analysis, Nikka Giren Publisher (1971)
[0034]
[Problems to be solved by the invention]
An object of the present invention is to have a function of accurately determining the end point of seasoning, that is, whether or not seasoning is sufficient, using data derived based on plasma emission data in seasoning performed immediately after wet cleaning. A plasma processing apparatus and a processing method are provided.
[0035]
[Means for Solving the Problems]
In the plasma processing apparatus having a processing chamber for processing a substrate, the object is to perform arithmetic processing using a light receiving unit that monitors plasma emission, a spectroscopic unit that splits the plasma emission, and a signal obtained by the spectroscopic unit. An arithmetic processing unit that generates an output signal, a determination unit that performs a determination based on the output signal and generates a device control signal, and a device control unit that controls the plasma processing apparatus based on a device control signal from the determination unit The plasma emission in a batch of plasma processing during seasoning is monitored, multivariate analysis is performed on the obtained plasma emission data, and the output signal obtained in the multivariate analysis is output to the previous batch. Taking the difference from the output signal at, the average value of the difference in the group, and the difference between the maximum and minimum values in the group, A standard deviation of within the human Matomarimasu Rabbi, by comparing the the values preset threshold is achieved.
[0036]
In addition, in the plasma processing apparatus and the plasma processing method, the object is to take a difference between the output signal and the output signal in the previous batch for a plurality of the output signals, This is accomplished by determining the mean value at the point, the difference between the maximum and minimum values in the group, and the standard deviation in the group, and comparing these values to a preset threshold .
[0037]
Further, the object is to take a difference between the plurality of output signals and the output signal in the previous batch for the plurality of output signals in the plasma processing apparatus and the plasma processing method, and to calculate an average value of the differences in the batch, And find the difference between the maximum and minimum values in a group, and the standard deviation in the group, and compare these values with multiple preset thresholds for each value. This is accomplished by calculating a score and comparing the total value to a preset threshold.
[0038]
Further, the object is to calculate a single or a plurality of output signals using a pre-registered filter vector or a plurality of filter vectors in the plasma processing apparatus and the plasma processing method, and to obtain a difference from the output signal in the previous batch. Taking the mean value of the difference in a group, the difference between the maximum and minimum values in the group, and the standard deviation in the group, and setting these values in advance This is accomplished by comparing with one or more thresholds.
[0039]
Further, the above object is achieved by automatically performing the processing chamber evacuation and the degree of vacuum confirmation operation, the apparatus abnormality confirmation operation and the above processing method in the above plasma processing apparatus and plasma processing method.
[0040]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, the first embodiment of the present invention will be described in detail with reference to FIGS. The first embodiment acquires plasma emission data of dummy processing in seasoning, and uses principal component analysis as a method of multivariate analysis performed on the plasma emission data. A group of plasma processing is composed of one lot of dummy processing consisting of 25 sheets.
[0041]
FIG. 1 is a diagram showing a configuration of a microwave ECR (Electron Cyclotron Resonance) plasma etching apparatus to which a first embodiment of the present invention is applied.
[0042]
The microwave ECR plasma etching apparatus includes a processing chamber 3 including a processing chamber wall 1 and a processing chamber lid 2, a substrate holder 4, a processing gas introduction pipe valve 7, a mass flow controller 8, a processing gas introduction pipe 9, Exhaust port 12, variable valve 13, exhaust pipe valve 14, pressure gauge 15, light receiving unit 20, optical fiber 21, spectroscopic unit 22, arithmetic processing unit 23, database 24, and determination unit 25 The apparatus control unit 26 is configured.
[0043]
A processing chamber lid 2 is installed on the processing chamber wall 1, and a substrate holding table 4 is provided in a processing chamber 3 constituted by this, and a semiconductor substrate (object to be processed) 5 is provided on the substrate holding table 4. Is placed.
[0044]
The processing gas 6 used for the plasma etching is introduced into the processing chamber 3 through the processing gas introduction pipe valve 7 and the mass flow controller 8 for adjusting the flow rate of the processing gas 6 and the processing gas introduction pipe 9. A microwave 10 transmitted by a microwave transmitter (not shown) is introduced into the processing chamber 3, and a plasma 11 is generated by the interaction between the magnetic field generated by a coil (not shown) and the microwave 10. Volatile substances generated by the reaction in the processing gas 6 and the etching process are discharged from the exhaust port 12. A vacuum pump (not shown) is connected to the tip of the exhaust port 12 to reduce the pressure in the processing chamber 3. Further, a variable valve 13 and an exhaust pipe valve 14 are connected between the exhaust port 12 and a vacuum pump (not shown) so that the variable valve 13 can adjust the pressure in the processing chamber 3 during the plasma etching process. It has become. In addition, whether or not the processing chamber 3 is evacuated can be switched by opening and closing the exhaust pipe valve 14. The pressure in the processing chamber 3 is measured by a pressure gauge 15 connected to the processing chamber wall 1.
[0045]
A light receiving unit 20 is installed on the processing chamber wall 1 so as to receive light emitted from the plasma 11, and the plasma emission collected by the light receiving unit 20 is guided to the spectroscopic unit 22 through the optical fiber 21. The plasma light is spectrally separated by the spectroscopic unit 22, and a multi-channel signal (for example, a wavelength of 200 nm to 800 nm) is periodically provided at a certain sampling interval (for example, 1 second) by a CCD (Charge-Coupled Devices) incorporated in the spectroscopic unit 22. 1024 channel signals).
[0046]
Multi-channel signals are input to the arithmetic processing unit 23. The arithmetic processing unit 23 has a function of converting the multi-channel signal input from the spectroscopic unit 22 into several signals. For example, assuming that the signal from the spectroscopic unit 22 is n channels (for example, 1024 channels), these are combined into one input signal vector s and described as s = {s1, s2,..., Sn}. Assume that a filter vector f representing a signal filter is defined as f = {f1, f2,..., Fn} at this time. At this time, the arithmetic processing unit 23 calculates one output signal z by calculating the inner product of the input signal vector s and the filter vector f using the following formula (1).
[0047]
[Expression 1]

[0048]
The filter vector f can be stored in the database 24, and the filter vector f can be sent to the arithmetic processing unit 23 when the output signal z is calculated by the arithmetic processing unit 23.
[0049]
Next, a method for creating a signal filter will be described. The input signal of the arithmetic processing unit 23 may be several. However, when the emission spectrum or the like is used as input as described above, the input signal is several hundred to several thousand. Therefore, it is effective to use multivariate analysis to generate the signal filter. In this embodiment, principal component analysis which is one of multivariate analysis is used. Principal component analysis is a general method of statistical processing. For example, a specific calculation method is described in Non-Patent Document 1 described above.
[0050]
When an input signal vector s having n input signals is taken in a row and m acquired input signal vectors are overlapped in the column direction, a signal matrix S of m rows and n columns as shown in the following equation (2) is obtained. can get.
[0051]
[Expression 2]

[0052]
Next, when a correlation matrix or covariance matrix X of the signal matrix S is generated and eigenvalue analysis of X which is a symmetric matrix of n rows and n columns is performed, n positive real eigenvalues {λ1, λ2,. λn} and n eigenvectors {y1, y2,..., yn} corresponding to each eigenvalue are obtained. At this time, the eigenvalues are arranged as λ1, λ2, λ3,.
[0053]
Each eigenvector y is a one-dimensional vector of n rows. By using this eigenvector as the filter vector f and calculating the equation (1), it becomes a signal filter that converts n input signals into one device state signal.
[0054]
Further, the output signal z generated in the arithmetic processing unit 23 is subjected to further arithmetic processing as will be described later, and the calculation result (output signal) is sent to the determination unit 25. The determination unit 25 can make a determination for device control based on the calculation result (output signal) and generate a device control signal as to whether or not to continue seasoning. The device control signal generated by the determination unit 25 is sent to the device control unit 26.
[0055]
The apparatus control unit 26 controls the plasma etching apparatus. For example, when an apparatus control signal for executing seasoning is received, the operation for continuing seasoning is performed, and when an apparatus control signal for ending seasoning is received, seasoning is performed. Perform end operation.
[0056]
Next, the processing in the first embodiment will be described in detail with reference to FIGS.
[0057]
In FIG. 2, first, the light emission data acquisition operation for one lot of dummy processing in seasoning is performed (step S21). As described above, this refers to a process in which the plasma emission collected by the light receiving unit 20 is converted into a multi-channel signal at a certain sampling interval in the spectroscopic unit 22 and the signal is sent to the arithmetic processing unit 23. Note that the plasma emission data composed of multi-channel signals is accumulated in the database 24 until the dummy processing of the current batch (in this case, a lot) is completed.
[0058]
Next, an effective portion is extracted from numerical data indicating plasma emission (step S22). Principal component analysis has the property of extracting large fluctuations. Therefore, if a principal component analysis is performed on a series of light emission data from ignition to extinction of plasma, a large fluctuation of plasma at the time of ignition or extinction of plasma is picked up. Similarly, when a principal component analysis is performed on a series of light emission data from ignition to disappearance of plasma, when the processing sequence is composed of a plurality of steps having different processing conditions, it is caused by a change in the processing conditions during the transition between steps. Picking up large fluctuations in the plasma. These large fluctuations cause a situation in which important fluctuations that are small but indicate changes in the state of the processing chamber are buried. In order to prevent this, it is necessary to perform principal component analysis using only the plasma emission data excluding plasma ignition / disappearance and transition between steps.
[0059]
The effective portion extraction operation S22 will be specifically described with reference to FIGS. FIG. 3 is a graph showing the first principal component score when the plasma emission of the dummy process performed using one dummy wafer is sampled at intervals of 1 second from ignition to disappearance, and the plasma emission data is subjected to principal component analysis. It is. Region 30 shows plasma fluctuations due to plasma ignition and instability immediately thereafter. Regions 31 and 32 show plasma fluctuations in the first step and the second step of the dummy process. A region 33 shows the fluctuation of the plasma due to the disappearance of the plasma. In the graph shown in FIG. 3, it can be seen that the fluctuation of the plasma during the first step and the second step is hardly grasped.
[0060]
Next, FIG. 4 is a graph showing the behavior of the first principal component score when the principal component analysis is performed on the plasma emission data from 17 seconds to 52 seconds after the ignition of the plasma. The area 32 ′ corresponds to the process in the second step of the dummy process. It can be seen from FIG. 4 that the plasma fluctuation in the second step can be grasped. Thereby, the minute fluctuation | variation of a process chamber state can be grasped | ascertained. In the plasma emission data, how many seconds to how many seconds from the plasma ignition is the effective part and used for principal component analysis, or from what sampling to what sampling point of the data from plasma ignition to the effective part and the principal component Whether to use for analysis is registered in the database 24 in advance, and when the principal component analysis is performed in the arithmetic processing unit 23, the effective portion of the plasma emission data is designated.
[0061]
Since the dummy process in seasoning is repeatedly performed in exactly the same processing sequence, if the effective part of the plasma emission data is designated in advance and registered in the database 24, the same process can be used in the subsequent dummy processes.
[0062]
In addition, here, the second step of the dummy process is set as an effective portion of the plasma emission data, but is not limited thereto. For example, the first step of the dummy process may be an effective part of the plasma emission data.
[0063]
Next, principal component analysis is performed using the plasma emission data in the effective portion obtained in step S22, and the output signal z is calculated (step S23). Here, the first principal component score obtained by principal component analysis is the output signal z.
[0064]
Next, an average value in one dummy process of the first principal component score calculated by the principal component analysis of the calculation operation S23 is calculated (step S24). The principal component score is data obtained at a certain sampling interval (here, 1 second) as shown in FIG. In each dummy process, an average value of the first principal component scores is calculated. In the k-th dummy process, the first principal component score in the i-th sampling is a_iAnd the effective range of the plasma emission data of the dummy processing is the first principal component score a by N samplings_iThe average value P of the first principal component scores in the k-th dummy process_kIs represented by the following formula (3).
[0065]
[Equation 3]

[0066]
FIG. 5 shows the average value P of the first principal component scores in the k-th dummy process._kThe transition of 4 lots of the dummy processing of 25 1 lots is shown. Although it can be seen from FIG. 5 that the state in the processing chamber 3 has changed due to seasoning, the end point of seasoning cannot be detected as it is.
[0067]
Next, the average value P of the first principal component scores of each dummy process_kThe difference between the first principal component score of the lot of interest and the first principal component score of the previous lot is calculated (step S25). The average value P of the first principal component score in the j-th dummy process for the i-th lot is P_{i , j}Then, the first principal component score P in the j-th dummy process of the i-th lot_{i , j}And the first principal component score P in the j-th dummy processing of the (i-1) th lot_{i -1, j}Difference D_{i , j}Is represented by the following formula (4).
[0068]
[Expression 4]

[0069]
When the seasoning process brings the state of the processing chamber close to when mass production is stable, dummy processing is performed with good reproducibility between lots. Therefore, the first principal component score P in the lot that has been processed_{i, j}And the first principal component score P in the immediately preceding lot_{i-1, j}Difference D_{i, j}Thus, the reproducibility between lots of dummy processing can be evaluated, and as a result, it is possible to determine whether seasoning is sufficient, that is, determine the seasoning end point.
[0070]
D in FIG._{i, j}The graph showing the behavior of is shown. As shown in FIG._{i, j}The value of is generally lower between lots, and D_{i, j}It can be seen that the variation within the lot is smaller.
[0071]
The first lot of dummy processing is D_{i, j}This step S25 of calculating is omitted.
[0072]
Next, D_{i, j}Average value in lots of D,_{i, j}Difference between the maximum and minimum values in a lot of_{i, j}The standard deviation in the lot is calculated (step S26). D shown in FIG._{i, j}Is generally reduced between lots, D in the lot consisting of n dummy processes shown in the following formula (5)_{i, j}The average value AVE can be evaluated. In equation (5), the subscript i of AVE is D in the i-th lot._{i, j}The average value AVE is shown.
[0073]
[Equation 5]

[0074]
Furthermore, D shown in FIG._{i, j}The fluctuation in the lot of D is the D in the i-th lot represented by the following formula (6)._{i, j}Difference between the maximum and minimum values of MAXMIN_i, And D in the i-th lot represented by the following formula (7)_{i, j}Standard deviation of_iCan be evaluated.
[0075]
[Formula 6]

[0076]
[Expression 7]

[0077]
Next, it is determined whether seasoning is sufficient (step S27).
[0078]
D in FIG._{i, j}Average value AVE within a lot of_iA graph showing D is shown in FIG._{i, j}Difference between the maximum and minimum values in a lot_iA graph showing D is shown in FIG._{i, j}Standard deviation within lot_iThe graph which shows is shown, respectively. From these, it can be seen that the respective values become smaller as the seasoning progresses. This is because, since the state in the processing chamber 3 changed by the wet cleaning S1 has approached the time of stable mass production by seasoning, the plasma processing can be performed with good reproducibility between lots in the dummy processing, and this is reflected in the first main state. Component score D_{i, j}The value of became smaller as the seasoning progressed.
[0079]
Also, AVE calculated based on the first principal component score_i, MAXMIN_i, And σ_iHowever, if it is less than or equal to a preset threshold value, it may be determined that the state in the processing chamber has approached when mass production is stable, that is, seasoning has reached the end point.
[0080]
For example, in this embodiment, AVE_iAVE_iMAXMIN below the preset threshold of 40_iIn MAXMIN_iIs less than or equal to a preset threshold 42 for σ_iFor σ_iIt was determined that the end point had been reached because the threshold value was less than or equal to the preset threshold value 44.
[0081]
AVE shown in FIG._i, MAXMIN_i, Σ_iSince each value is smaller than the threshold value at i = 4, the end point of seasoning is reached by the dummy processing of the fourth lot, that is, it can be determined that the dummy processing of the fourth lot is sufficient for this seasoning.
[0082]
AVE_i, MAXMIN_i, And σ_iOf these three values, AVE_iCan be negative. AVE_iAVE takes a negative value, it shows a behavior in which the absolute value decreases with the progress of seasoning._iIf the absolute value of is less than or equal to a predetermined threshold value, it may be determined that the state in the processing chamber has approached the time when mass production is stable, that is, seasoning has reached the end point.
[0083]
In the first embodiment, AVE_i, MAXMIN_i, And σ_iSince all three values of i were smaller than their respective threshold values at i = 4, it was determined that the seasoning end point was reached by the dummy processing of the fourth lot, but all of these three values were smaller than the threshold values. Sometimes it is not limited to the end of seasoning. For example, AVE_i, MAXMIN_i, And σ_iIf one of them becomes smaller than the threshold value, it may be regarded as a seasoning end point, for example, AVE_i, MAXMIN_i, And σ_iIf two of them become smaller than the respective threshold values, they may be regarded as seasoning end points, and these can be selected by the practitioner of the present invention.
[0084]
Next, when it is determined that the seasoning is sufficient in the determination operation S27, the seasoning end operation is performed (step S10). At this time, a signal to end seasoning is sent from the determination unit 25 of FIG. 1 to the apparatus control unit 26, and the seasoning end operation S10 is performed so that the product substrate start S11 can be performed. At this time, the operator may be informed that seasoning has been completed.
[0085]
On the other hand, if it is determined in the determination operation S27 that seasoning is insufficient, the determination unit 25 in FIG. 1 sends a signal to continue the seasoning to the apparatus control unit 26, and the seasoning is continued (step S28).
[0086]
At this time, the process returns to step S21 to obtain light emission data for one lot of dummy processing.
[0087]
By performing the sequence described above, it is possible for the operator to determine whether seasoning is sufficient without manually performing the confirmation process S9. In addition, since the seasoning can be completed by the dummy processing with the minimum number of lots, the amount of dummy substrate used can be reduced and the operating rate of the plasma processing apparatus can be improved.
[0088]
In the seasoning end point determination shown in the first embodiment, the first principal component score P in the j-th dummy process of the i-th lot is used._iAnd the first principal component score P in the j-th dummy processing of the i-1 lot_{i-1, j}Difference D_{i, j}Was used. That is, since the difference between the lots of the output signal z obtained from the light emission data in a series of continuous plasma treatments is used, the influence of the variation in the finish of the wet cleaning S1 on the accuracy of the seasoning end point detection is small. .
[0089]
In the first embodiment, in step S23, the filter vector f is calculated by performing principal component analysis for the light emission of the dummy process in the ongoing seasoning, and as a result, the output signal z is calculated. It is not limited to that. The output signal z can also be calculated from the equation (1) using the input signal vector s obtained from the light emission of the dummy process and the preset filter vector f.
[0090]
The filter vector f used at that time can be obtained by performing principal component analysis for light emission of dummy processing in seasoning immediately after the previous wet cleaning. In addition, the filter vector f obtained by performing principal component analysis on the dummy light emission in the seasoning immediately after the wet cleaning two times before or three times before instead of the previous wet cleaning may be used. Alternatively, the filter processing apparatus f can set the filter vector f in advance.
[0091]
In the first embodiment, the dummy processing of 25 lots per lot is grouped, but the present invention is not limited to this. Even if one lot of wafers consists of 25 sheets, when the dry cleaning is performed every five dummy processes, for example, in the seasoning process, that is, the dummy process S8-2-8-2 ″ or the dummy process S 8. When -3 to 8-3 ″ are composed of five dummy processes, a group of dummy processes is five sheets. In this case, the process from the light emission data acquisition operation S21 for a group of dummy processes to the determination operation S27 for determining whether seasoning is sufficient is performed for every five dummy processes.
[0092]
Further, even when the single wafer cleaning is performed after each of the dummy processes S8-2 to S8-2 ″, the dummy processes S8-3 to S8-3 ″ are composed of five dummy processes. In this case, a group of dummy processes is five sheets. In this case, the process from the light emission data acquisition operation S21 for a group of dummy processes to the determination operation S27 for determining whether seasoning is sufficient is performed for every five dummy processes.
[0093]
In the first embodiment, the principal component analysis is performed in the calculation of the output signal z in the calculation operation S23, and only the first principal component score is used as the output signal z. However, the present invention is not limited to this. A plurality of principal component scores can also be used as the output signal z.
[0094]
Hereinafter, a second embodiment of the present invention will be described with reference to FIGS. In the first embodiment, only the first principal component score is used as the output signal z, but in this embodiment, the second principal component score and the third principal component score are used. In the principal component analysis, the higher-order principal component has more noise components, and the effectiveness of monitoring the processing chamber state becomes lower. Therefore, in the present embodiment, only the third principal component is used. AVE obtained by the same method as in the first embodiment using the second principal component score_iFIG. 8A shows the second main component MAXMIN._iFIG. 8B shows the σ of the second principal component._iAre shown in FIG. Also, AVE obtained by the same method as in the first embodiment using the third principal component score._iFIG. 9A shows the third main component MAXMIN._iFIG. 9B shows the third principal component σ._iAre shown in FIG. These values are also AVE of the first principal component shown in FIG._i, MAXMIN_i, Σ_iIt can be seen that the seasoning decreases as the seasoning progresses. These values are compared with a preset threshold value. If the value is equal to or less than the threshold value, it can be determined that the state in the processing chamber has approached the time when mass production is stable, that is, seasoning has reached the end point.
[0095]
AVE of the first main component_iAnd the threshold 40 set for the first principal component MAXMIN_iAnd the threshold 42 set for σ and the first principal component σ_iAnd the threshold 44 set for the AVE and the second principal component AVE_iAnd the threshold 46 set for the second principal component MAXMIN_iAnd the threshold 48 set for σ and the second principal component σ_iAnd the third principal component AVE_iAnd the third principal component MAXMIN._iAnd the third principal component σ_iIt is not necessary that the threshold values 56 set for all are the same, and it is desirable to set a suitable threshold value for each.
[0096]
Also, the first main component AVE_i, MAXMIN_i, Σ_i, And AVE of the second principal component_i, MAXMIN_i, Σ_i, And AVE of the third principal component_i, MAXMIN_i, Σ_iWhen all of the values are smaller than the preset threshold values, it may be considered that seasoning is sufficient, but this is not a limitation. For example, two threshold values 58 and 58 ′ are set as shown in the graph of FIG._i, MAXMIN_i, Σ_iIf the value is greater than or equal to the threshold value 58, a score of 0 points is given, if the value is less than the threshold value 58 and greater than or equal to the threshold value 58 ', one point is given, and if the value is less than the threshold value 58', two points are given. If the total score is larger than a preset threshold value (for example, 12 points), it can be determined that seasoning is sufficient.
[0097]
In the second embodiment, when i = 2, the total score is 60 (here, 1 point). When i = 3, the total score is 61 (here, 4 points), and when i = 4, the total score is 62 (here, 17 points), and the total score is larger than a preset threshold (here, 12 points). Since i = 4, it can be determined that seasoning has reached the end point in the fourth lot.
[0098]
AVE obtained from a plurality of output signals z as in the second embodiment described above._i, MAXMIN_i, Σ_iBy using this, it is possible to improve the accuracy of determination as to whether seasoning is sufficient.
[0099]
In the second embodiment, the threshold value is set to a numerical value. However, the threshold value is not limited to this, and the threshold value may be determined by setting a ratio in advance. For example, AVE at i = 2_i, MAXMIN_i, Σ_IThe value of 50% is set to the threshold value 58 (setting ratio 50%), the value of 20% is set to the threshold value 58 '(setting ratio 20%), and using these threshold values, the subsequent AVE_i, MAXMIN_i, Σ_IMay be scored.
[0100]
In the second embodiment, the AVE obtained from the first principal component score, the second principal component score, and the third principal component score._i, MAXMIN_i, Σ_iAlthough two threshold values 58 and 58 'are set for, the number of threshold values is not limited to two.
[0101]
The score given in comparison with each threshold is AVE_i, MAXMIN_i, Σ_iAre the same (here, 0 point, 1 point, 2 points), but are not limited thereto. For example, AVE obtained from the second principal component score_i, MAXMIN_i, Σ_iAVE whose importance is obtained from other principal component scores_i, MAXMIN_i, Σ_iAVE obtained from the second principal component score_i, MAXMIN_i, Σ_iYou may increase the score given to (for example, 0 points, 2 points, 4 points). Similarly, AVE_i, MAXMIN_i, Σ_iAVE_iImportance is MAXMIN_iAnd σ_iIf it is higher than the importance of AVE_iYou may increase the score given to (for example, 0 points, 2 points, 4 points).
[0102]
In the second embodiment, the first principal component score, the second principal component score, and the third principal component score are used as the output signal z, but the number of principal component scores to be used is not limited to three.
[0103]
In the second embodiment, in step S23, the filter vector f is calculated by performing principal component analysis on the light emission of the dummy process in the ongoing seasoning, and as a result, the output signal z is calculated. However, it is not limited to that. A plurality of output signals z can also be calculated by using equation (1) using the input signal vector s obtained from the light emission of the dummy processing and a plurality of preset filter vectors f.
[0104]
As the plurality of filter vectors f used at that time, those obtained by performing principal component analysis on the light emission of the dummy process in the seasoning immediately after the previous wet cleaning can be used. Also, by using a plurality of filter vectors f obtained by performing principal component analysis on the light emission of dummy processing in seasoning immediately after wet cleaning two times before or three times before instead of the previous wet cleaning. Also good. Alternatively, the plasma processing apparatus manufacturer can set a plurality of filter vectors f in advance.
[0105]
Next, a third embodiment of the present invention will be described with reference to FIGS.
[0106]
In the third embodiment, the process from the start of evacuation after wet cleaning to the product substrate processing start S10 is automatically performed without manual operation by the operator. This is realized by adding a process for determining whether or not the apparatus is sufficient and a process for determining whether or not there is an apparatus abnormality to the above-described processing.
[0107]
First, wet cleaning of the plasma processing chamber 3 is performed (step S1).
[0108]
Thereafter, after closing the processing chamber lid 2, a vacuum pump (not shown) connected to the exhaust port 10 is started to start the vacuuming operation S <b> 2 of the processing chamber 3.
[0109]
Thereafter, a confirmation operation S3 is automatically performed as to whether or not a predetermined degree of vacuum has been reached. This is done by automatically measuring the pressure in the processing chamber 3 with a pressure gauge connected to the processing chamber 3. At this time, if the preset vacuum degree is not reached, the evacuation is further continued. Further, even after the evacuation operation S2 is performed for a preset time, the alarm operation S4 is automatically performed if the preset vacuum degree is not reached. This may sound a buzzer or display an alarm on a monitor screen installed in the plasma processing apparatus. If the plasma processing apparatus is connected to a computer network, an E-Mail may be sent to a remote engineer. It may be transmitted or notified by a pager or the like. In addition to issuing an alarm, the operation of a vacuum pump (not shown) connected to the processing chamber 3 may be stopped.
[0110]
If the preset vacuum level is reached in the confirmation operation S3, the confirmation operation S5 is then automatically performed to determine whether or not there is any abnormality in the plasma processing apparatus.
[0111]
As an example of the confirmation operation S5, there is a leak check of the processing chamber 3, that is, a confirmation operation of the amount of mixed air per unit time. The amount L of air per unit time into the processing chamber 3 is determined by the time T during which the processing gas introduction pipe valve 7, the mass flow controller 8 and the exhaust pipe valve 14 are closed and the exhaust pipe valve 14 is closed, and the exhaust pipe valve 14. Is calculated using the following equation (8) using the pressure increase ΔP of the processing chamber 3 and the volume v of the processing chamber 3.
[0112]
[Equation 8]

[0113]
If the air mixing amount L per unit time calculated by (Equation 8) is smaller than a preset threshold value, the leak check is passed.
[0114]
Another example of the confirmation operation S5 is a confirmation operation of the mass flow controller 8 that adjusts the flow rate of the processing gas introduced into the processing chamber 3. This is set so that the mass flow controller 8 introduces a predetermined flow rate into the processing chamber 3, and at the same time, the pressure increase due to the introduction of the processing gas 6 is obtained by the pressure gauge 15 connected to the processing chamber 3. The controller 8 checks whether or not the processing gas having the set flow rate is correctly introduced into the processing chamber 3. If the difference between the flow rate setting value to the mass flow controller 8 and the flow rate of the processing gas 6 calculated from the pressure increase in the processing chamber 3 is smaller than a preset threshold value, the mass flow controller 8 passes the check. .
[0115]
When it is determined by the confirmation operation S5 that the plasma processing apparatus is abnormal, an alarm is automatically issued (step S4 '). This may sound a buzzer or display an alarm on a monitor screen installed in the plasma processing apparatus. If the plasma processing apparatus is connected to a computer network, an E-Mail may be sent to a remote engineer. It may be transmitted or notified by a pager or the like. Further, not only an alarm may be issued, but the operation of a vacuum pump (not shown) connected to the processing chamber 3 may be stopped.
[0116]
When it is determined that there is no abnormality in the plasma processing apparatus by the confirmation operation S5, the dummy substrate set in advance in the plasma processing apparatus is automatically carried into the processing chamber 3, and seasoning is automatically started. First, dry cleaning S6-2 is performed, and then dummy processing S8-2 to S8-2 '' is performed. Plasma light emission in the dummy processes S8-2 to S8-2 '' is collected by the light receiving unit 20, and the process as shown in the first and second embodiments is performed to calculate the output signal z.
[0117]
Next, a process for determining whether seasoning is sufficient (step S27) is automatically performed by performing the processes as shown in the first to second embodiments.
[0118]
If the dummy process is performed for the first time, the determination process S27 is omitted, and the process automatically proceeds to dry cleaning S6-2.
[0119]
Next, when it is determined by the determination process S27 that seasoning is insufficient, the process automatically proceeds to dry cleaning S6-2. If the seasoning is determined to be sufficient in the determination process S27, the seasoning end operation S10 is performed. At this time, a signal to end seasoning is sent to the apparatus control unit 26 by the judgment unit 25 in FIG. 1, and the seasoning end operation S10 is performed so that the product substrate start S11 can be performed. At this time, the operator may be informed that seasoning has been completed.
[0120]
By performing the sequence shown above, after starting vacuuming after wet cleaning, it automatically prepares for seasoning and seasons, and automatically determines whether seasoning is sufficient, and starts the construction of the product base. Preparation can be done. Therefore, it is possible to automatically determine whether seasoning is sufficient without performing the confirmation process S9 by the operator, which was necessary in the past. In addition, since the seasoning can be completed by the minimum number of dummy processes, the throughput of the plasma processing apparatus can be improved by reducing the amount of dummy substrate used and shortening the time required for seasoning.
[0121]
In the first to third embodiments, the plasma emission data used for seasoning end point determination is based on dummy processing, but is not limited thereto. For example, if single wafer cleaning is performed after each dummy process, it is determined whether seasoning is sufficient by the same method as in the first to third embodiments using plasma emission in the single wafer cleaning. Can be done.
[0122]
In the first to third embodiments, the principal component analysis is used as a multivariate analysis method for plasma light emission of plasma processing in seasoning, but is not limited thereto.
[0123]
In the first to third embodiments, the plasma etching apparatus has been described as an example. However, the application of the present invention is not limited to the plasma etching apparatus, and can be applied to other plasma processing apparatuses such as a plasma CVD apparatus. .
[0124]
In the first to third embodiments, the semiconductor substrate is described as an example of the substrate to be subjected to the plasma processing. However, the application of the present invention is not limited to the semiconductor substrate, and the plasma processing of other substrates such as an LCD substrate is possible. It can also be applied to.
[0125]
【The invention's effect】
As described above, according to the present invention, plasma emission is dispersed, multivariate analysis is performed on the data, and the calculated output signal is compared with the output signal in the previous batch of plasma processing. By calculating the average value in the group, the difference between the maximum and minimum values in the group, the standard deviation in the group, and comparing them to a preset threshold It can be determined whether seasoning is sufficient. Accordingly, it is possible to provide a plasma processing apparatus and a plasma processing method that can reduce the number of dummy processing performed in seasoning.
[Brief description of the drawings]
FIG. 1 is a side sectional view of a plasma etching apparatus showing a first embodiment of the present invention.
FIG. 2 is a flowchart showing a plasma emission data processing method according to the first embodiment of the present invention.
FIG. 3 is a graph showing the behavior of an output signal z (first principal component score) in the first embodiment of the present invention.
FIG. 4 is a graph showing the behavior of an output signal z (first principal component score) obtained from an effective area of light emission data in the first embodiment of the present invention.
FIG. 5 is a graph showing the behavior of the average value of each dummy process of the output signal z (first principal component score) obtained from the effective region of the light emission data in the first embodiment of the present invention.
FIG. 6 shows an average value P in each dummy process of the output signal z (first principal component score) in the first embodiment of the present invention._kD obtained from the difference between lots_{i, j}The graph which shows the behavior of the.
FIG. 7 shows AVE calculated based on the output signal z (here, the first principal component score) in the first embodiment of the present invention._iAnd MAXMIN_iAnd σ_iThe graph which shows the behavior.
FIG. 8 shows AVE calculated based on the output signal z (here, the second principal component direct point) in the second embodiment of the present invention._iAnd MAXMIN_iAnd σ_iThe graph which shows the behavior.
FIG. 9 shows AVE calculated based on the output signal z (here, the third principal component score) in the second embodiment of the present invention._iAnd MAXMIN_iAnd σ_iThe graph which shows the behavior.
FIG. 10 shows AVE calculated based on the output signal z in the second embodiment of the present invention._iAnd MAXMIN_iAnd σ_iThe figure which shows the method of giving a score with respect to.
FIG. 11 shows AVE calculated based on the output signal z (here, the first principal component score, the second principal component score, and the third principal component score) in the second embodiment of the present invention._iAnd MAXMIN_iAnd σ_iThe chart which shows the method of determining whether seasoning is enough using the score given with respect to.
FIG. 12 is a method for automatically performing all of the determination of whether or not the evacuation of the processing chamber and the degree of vacuum confirmation, the apparatus abnormality confirmation and the seasoning performed after the wet cleaning are sufficient in the third embodiment of the present invention; The flowchart which showed.
FIG. 13 is a flowchart showing a conventional example of product substrate processing that is generally performed.
FIG. 14 is a flowchart showing a conventional example of seasoning performed after wet cleaning.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 ... Processing chamber wall, 2 ... Processing chamber lid, 3 ... Processing chamber, 4 ... Substrate holding stand, 5 ... Semiconductor substrate, 6 ... Processing gas, 7 ... Processing gas introduction pipe valve, 8 ... Mass flow controller, 9 ... Processing gas Introducing pipe, 10 ... microwave, 11 ... plasma, 12 ... exhaust port, 13 ... variable valve, 14 ... exhaust pipe valve, 15 ... pressure gauge, 2 ... light receiving part, 21 ... optical fiber, 22 ... spectral part, 23 ... Arithmetic processing unit, 24 ... database, 25 ... determination unit, 26 ... device control unit
S1 Wet cleaning, S2 ... Vacuum pulling operation, S3 ... Confirmation operation, S4 ... Alarm operation, S5 ... Confirmation operation, S6 ... Dry cleaning, S7 ... Aging, S9 ... Confirmation processing, S10 ... Seasoning end operation, S11 ... Product substrate start S21: Acquisition of light emission data, S22: Extraction operation of effective portion, S23: Calculation operation, S27: Determination operation, S28: Continue seasoning.

Claims (6)

A processing chamber for performing plasma processing on the substrate;
A light receiving unit for monitoring plasma emission in the processing chamber;
A spectroscopic unit that splits the plasma emission and converts it into a multi-channel signal;
An arithmetic processing unit that converts the multi-channel signal into a single or a plurality of output signals, and performs arithmetic processing on the output signals;
A database for storing filter vectors;
A determination unit for determining a state in the processing chamber using the calculation result;
In a plasma processing apparatus having an apparatus control unit that controls the operation of the plasma processing apparatus according to a signal from the determination unit,
When performing seasoning using a dummy substrate that is performed after wet cleaning,
The arithmetic processing unit outputs an output signal Pi, j generated by multivariate analysis of plasma emission data in the j-th dummy process of the i-th lot, and an output signal Pi in the j-th dummy process of the (i-1) -th lot. −1, j, and the difference between them is Di, j,
Calculating the average value AVEi of Di, j and the difference MAXMINi and standard deviation σi of the maximum and minimum values of a plurality of substrates to be continuously processed;
The determination unit, a plasma processing apparatus characterized by detecting the end point of the seasoning process by comparing the calculated value of the AVEi and MAXMINi and σi and a predetermined threshold value.   The determination unit calculates a score for each value by comparing the calculated values of AVEi, MAXMINi, and σi with a plurality of preset threshold values, and sets the total value as a preset threshold value. 2. The plasma processing apparatus according to claim 1, wherein an end point of seasoning is detected by comparing with.   The output signal Pi, j and the output signal Pi-1, j perform a plurality of samplings of plasma emission data during the dummy processing at predetermined intervals, and use the respective principal component analysis to obtain a plurality of first principal component scores. The plasma processing apparatus according to claim 1, wherein the plasma processing apparatus is an average value Pk of the calculated first principal component scores. A processing chamber for performing plasma processing on the substrate;
A light receiving unit for monitoring plasma emission in the processing chamber;
A spectroscopic unit that splits the plasma emission and converts it into a multi-channel signal;
An arithmetic processing unit that converts the multi-channel signal into a single or a plurality of output signals, and performs arithmetic processing on the output signals;
A database storing filter vectors;
A determination unit for determining a state in the processing chamber using the calculation result;
In a plasma processing method using a plasma processing apparatus having an apparatus control unit that controls an operation of the plasma processing apparatus according to a signal from the determination unit,
When performing seasoning using a dummy substrate that is performed after wet cleaning,
The arithmetic processing unit outputs an output signal Pi, j generated by multivariate analysis of plasma emission data in the j-th dummy process of the i-th lot, and an output signal Pi in the j-th dummy process of the (i-1) -th lot. −1, j, and the difference between them is Di, j,
A process of calculating an average value AVEi of Di, j and a difference MAXMINi and standard deviation σi of the maximum and minimum values of the plurality of substrates to be continuously processed;
The step of comparing the value with a preset threshold of the AVEi and MAXMINi and σi and the calculated plasma processing method using the plasma processing apparatus characterized by detecting the end point of the seasoning process. A processing chamber for performing plasma processing on the substrate;
A light receiving unit for monitoring plasma emission in the processing chamber;
A spectroscopic unit that splits the plasma emission and converts it into a multi-channel signal;
An arithmetic processing unit that converts the multi-channel signal into a single or a plurality of output signals, and performs arithmetic processing on the output signals;
A database storing filter vectors;
A determination unit for determining a state in the processing chamber using the calculation result;
In a plasma processing method using a plasma processing apparatus having an apparatus control unit that controls an operation of the plasma processing apparatus according to a signal from the determination unit,
The process of vacuuming after wet cleaning,
A process of automatically determining whether the degree of vacuum is sufficient,
A process of automatically judging whether there is a device abnormality,
When performing seasoning using a dummy substrate that is performed after wet cleaning,
The arithmetic processing unit outputs an output signal Pi, j generated by multivariate analysis of plasma emission data in the j-th dummy process of the i-th lot, and an output signal Pi in the j-th dummy process of the (i-1) -th lot. −1, j, and the difference between them is Di, j,
A process of calculating an average value AVEi of Di, j and a difference MAXMINi and standard deviation σi of the maximum and minimum values of the plurality of substrates to be continuously processed;
The step of comparing the value with a preset threshold of the AVEi and MAXMINi and σi and the calculated plasma processing method using the plasma processing apparatus characterized by detecting the end point of the seasoning process.   The output signal Pi, j and the output signal Pi-1, j perform a plurality of samplings of plasma emission data during the dummy processing at predetermined intervals, and use the respective principal component analysis to obtain a plurality of first principal component scores. 6. The plasma processing method using the plasma processing apparatus according to claim 4, wherein the plasma processing apparatus calculates and calculates an average value Pk of a plurality of calculated first principal component scores.

Images