JP3872879B2

JP3872879B2 - End point detection method and apparatus for plasma processing

Info

Publication number: JP3872879B2
Application number: JP30444297A
Authority: JP
Inventors: 地塩輿水; 進斉藤
Original assignee: Tokyo Electron Ltd
Current assignee: Tokyo Electron Ltd
Priority date: 1996-11-11
Filing date: 1997-11-06
Publication date: 2007-01-24
Anticipated expiration: 2017-11-06
Also published as: JPH10189552A

Description

【０００１】
【発明の属する技術分野】
本発明は、プラズマ処理の終点検出方法およびその装置に関する。
【０００２】
【従来の技術】
プラズマ処理装置、特に、エッチング装置は、従来から半導体製造工程あるいは液晶表示装置用基板の製造工程に広く適用されている。このようなエッチング装置は、例えば、互いに平行に配設された上部電極と下部電極とを備えており、上部電極と下部電極との間の放電によりエッチング用ガスからプラズマを発生させ、その活性種で被処理体としての上に酸化膜等の膜が形成された半導体ウェハの膜をエッチングするものである。このようなエッチング処理に際しては、エッチングの進捗状況を監視し、その終点をできるだけ正確に検出することにより、所定のパターン通りのエッチング処理を行うことが望まれている。
【０００３】
従来から、エッチングの終点を検出する方法には、質量分析、分光分析等の機器分析手法が用いられており、それらの中でも比較的簡易で高感度な分光分析が広く用いられている。エッチングの終点を検出する場合に分光分析法を用いるときは、具体的には、エッチング用ガス、その分解生成物もしくは反応生成物等のラジカル、またはイオン等の活性種のうち特定の種類の活性種を選択し、選択された活性種の発光の強度の時間に対する変化を測定している。この場合、選択する活性種は、エッチング用ガスの種類や被エッチング材により異なる。例えば、ＣＦ₄ 等のフルオロカーボン系のエッチング用ガスを用いてシリコン酸化膜をエッチングする場合には、その反応生成物であり、終点時に急激に発光強度が減少するＣＯ^* を用いている。この場合、ＣＯ^* からの時間に対する発光強度（２１９nmまたは４８３．５nmの波長を使用）のみ（一波長のみを使用）を測定し、その発光強度やその強度の一次微分値、二次微分値等の時間に対する変化量を比較することによりエッチングの終点を判定する方法と、特開昭６３−８１９２９に開示されているように、ＣＯ^* からの時間に対する発光強度と、２波長の基準光（特開昭６３−８１９２９においては、ヘリウム等の原子の発光スペクトル強度が７０６．５ｎｍ並びに６６７．８ｎｍの波長）の時間に対する発光強度を測定し、それぞれの発光強度比又はその比の一次微分値、二次微分値等の変化量を比較することによりエッチングの終点を判定する方法と、同じく２波長を利用する方法として、特開平５−２９２７６（対応米国特許Ｎｏ．５，３２２，５９０）にて開示されているように、ＣＯ^* からの時間に対する発光強度と上述の基準光の代わりに終点時に発光強度が急上昇するエッチングガスの分解物であるＣＦ^* の時間に対する発光強度とをそれぞれ測定し、それぞれの発光強度の比又はその比の一次微分値、二次微分値等の時間に対する変化量を比較することによりエッチングの終点を判定する方法が知られている。
【０００４】
一波長を用いる従来の終点検出方法では、プラズマの揺らぎ等に起因する発光強度変動により、プラズマ処理の終点が不明確になり正確に終点を検出することができない。二波長を用いる従来の終点検出方法では、反応生成物の活性種であるＣＯ^* の発光強度と基準光又は終点時に発光強度が急上昇するエッチングガスの分解物であるＣＦ^* の発光強度の終点時間に対する変化量がプラズマの揺らぎやチャンバや電極やウエハの温度変化、チャンバ壁に付着したデポ物等が原因で夫々異なることを考慮せず、単純に両者の発光強度比をとり、その比を使って終点検出を行うため、やはり正確に終点検出を行うことが困難である。
【０００５】
米国特許Ｎｏ．５，５６５，１１４で、本発明者達は、２波長を使用して夫々の波長における発光強度比を使用してエッチングの終点を検出する際に、両者の発光強度について時間に対する変化量を一致させて（即ち、両者の時間に対する発光強度の変化曲線の傾きを一致させる）から両者の比をとる思想を開示している。確かに、単純に両者の比を監視するよりは、この公報（米国特許Ｎｏ．５，５６５，１１４）で開示されているように、事前に両者の変化曲線の傾きを一致させた後に、両者の比を監視したほうがより正確に終点を検出することが可能である。
【０００６】
【発明が解決しようとする課題】
しかしながら、この米国特許で開示されている両者の傾きを一致させる方法は、終点前の指定区間における発光強度の変化曲線の平均値を求め、その指定区間での夫々の変化曲線について発光強度と平均値との差の絶対値の総和を求め、この総和を利用して（即ち、面積を計算して）いるために、ノイズに弱いという欠点がある。
本発明は、プラズマ状態の変動を許容して正確に終点検出を行うことができるプラズマ処理の終点検出方法およびその装置を提供することを目的とする。
【０００７】
【課題を解決するための手段】
本発明の第１の態様に係わるプラズマ処理の終点検出方法は、被処理体にプラズマを用いた処理を施す際に、前記プラズマ処理中の指定期間中及びこれ以降の、第１および第２の活性種の夫々特定波長を有する発光強度を検出し、この発光検出情報を出力する工程と、
これら発光検出情報に基づいて、発光強度と時間との関係において第１の活性種の近似関係式Ａ、即ち、Ｙ₁ ＝ａ₁ ×Ｘ＋ｂ₁、および第２の活性種の近似関係式Ｂ、即ち、Ｙ₂ ＝ａ₂ ×Ｘ＋ｂ₂（ここて゛、Ｙ₁ ，Ｙ₂ は、それぞれ第１および第２の活性種の発光光量の近似値を表し、Ｘは経過時間を表し、ａ₁ ，ａ₂ は一次係数を表し、ｂ₁ ，ｂ₂ はＹ切片を表す）を求める工程と、
前記第１の活性種の近似関係式Ａおよび前記第２の活性種の近似関係式Ｂを用いて、第１の活性種の擬似近似関係式Ａ´、即ち、Ｙ₁ ＝（ａ₁ ／ａ₂ ）×（Ｙ₂ −ｂ₂ ）＋ｂ₁を求める工程（ここでのＹ ₂ は実測値である）と、
前記第１の活性種の近似関係式Ａおよび前記第１の活性種の擬似近似関係式Ａ´から、両者の比、即ち、（式ＡにおけるＹ₁）／（式Ａ’におけるＹ₁）および前記比の微分値、即ち、ｄ（（式ＡにおけるＹ₁）／（式Ａ’におけるＹ₁））／ｄｔを求める工程と、
前記指定期間内の前記比（式ＡにおけるＹ₁）／（式Ａ’におけるＹ₁）および前記比の微分値ｄ（（式ＡにおけるＹ₁）／（式Ａ’におけるＹ₁））／ｄｔのそれぞれの平均値を求める工程と、
前記比を横軸にとり、前記比の微分値を縦軸にとり、前記指定期間内の前記比の平均値と前記比の微分値の平均値との交点を原点としたグラフに、所定領域を設定する工程と、
前記指定期間以降の処理中における前記第１および第２の活性種の発光検出情報を用いて比、即ち、（式ＡにおけるＹ₁）／（式Ａ’におけるＹ₁）および比の微分値、即ち、ｄ（（式ＡにおけるＹ₁）／（式Ａ’におけるＹ₁））／ｄｔを求め、求められた比および比の微分値の前記グラフにおける位置が前記所定領域から外れたときをプラズマ処理の終点として判定する工程と、を具備することを特徴としている。
【０００８】
本発明の別の態様に係わるプラズマ処理の終点検出方法は、被処理体にプラズマを用いた処理を施す際に、前記プラズマ処理中の指定期間、並びにそれ以降に第１および第２の活性種からの発光の特定波長での発光強度を光検出手段により逐次検出して、発光検出情報を出力する工程と、
前記指定期間内の発光検出情報に基づいて、発光強度と時間との関係において第１の活性種の近似関係式Ａ、即ち、Ｙ₁ ＝ａ₁ ×Ｘ＋ｂ₁、および第２の活性種の近似関係式Ｂ、即ち、Ｙ₂ ＝ａ₂ ×Ｘ＋ｂ₂（ここて゛、Ｙ₁ ，Ｙ₂ は、それぞれ第１および第２の活性種の発光光量の近似値を表し、Ｘは経過時間を表し、ａ₁ ，ａ₂ は一次係数を表し、ｂ₁ ，ｂ₂ はＹ切片を表す）を求める工程と、
前記第１の活性種の近似関係式Ａおよび前記第２の活性種の近似関係式Ｂを用いて、第１の活性種の擬似近似関係式Ａ´、即ち、Ｙ₁ ＝（ａ₁ ／ａ₂ ）×（Ｙ₂ −ｂ₂ ）＋ｂ₁を求める工程（ここでのＹ ₂ は実測値である）と、
前記第１の活性種の近似関係式Ａおよび前記第１の活性種の擬似近似関係式Ａ´から、両者の比、即ち、（式ＡにおけるＹ₁）／（式Ａ’におけるＹ₁）および前記比の微分値、即ち、ｄ（（式ＡにおけるＹ₁）／（式Ａ’におけるＹ₁））／ｄｔを求める工程と、
前記指定期間内の前記比（式ＡにおけるＹ₁）／（式Ａ’におけるＹ₁）および前記比の微分値ｄ（（式ＡにおけるＹ₁）／（式Ａ’におけるＹ₁））／ｄｔのそれぞれの平均値を求める工程と、
前記比を横軸にとり、前記比の微分値を縦軸にとり、前記指定期間内の前記比の平均値と前記比の微分値の平均値との交点を原点としたグラフに、前記指定期間以降の処理中における前記第１および第２の活性種の発光検出情報を用いて求められた比、即ち、（式ＡにおけるＹ₁）／（式Ａ’におけるＹ₁）および比の微分値、即ち、ｄ（（式ＡにおけるＹ₁）／（式Ａ’におけるＹ₁））／ｄｔの前記グラフにおける位置が前記原点から外れて再び横軸に近接するときをプラズマ処理の終点として判定する工程と、を具備することを特徴とする。
【００１０】
本発明のプラズマ処理の終点検出装置は、被処理体にプラズマを用いた処理を施す際にプラズマにより発生した第１および第２の活性種の特定波長での発光強度の時間に対する変化を検出して、発光検出情報する光検出手段と、
指定期間内の発光検出情報に基づいて、発光強度と時間との関係において第１の活性種の近似関係式Ａ、即ち、Ｙ₁ ＝ａ₁ ×Ｘ＋ｂ₁、および第２の活性種の近似関係式Ｂ、即ち、Ｙ₂ ＝ａ₂ ×Ｘ＋ｂ₂（ここて゛、Ｙ₁ ，Ｙ₂ は、それぞれ第１および第２の活性種の発光光量の近似値を表し、Ｘは経過時間を表し、ａ₁ ，ａ₂ は一次係数を表し、ｂ₁ ，ｂ₂ はＹ切片を表す）を求め、前記第１の活性種の近似関係式Ａおよび前記第２の活性種の近似関係式Ｂを用いて、第１の活性種の擬似近似関係式Ａ´、即ち、Ｙ₁ ＝（ａ₁ ／ａ₂ ）×（Ｙ₂ −ｂ₂ ）＋ｂ₁を求め（ここでのＹ_２は実測値である）、前記第１の活性種の近似関係式Ａおよび前記第１の活性種の擬似近似関係式Ａ´から、両者の比、即ち、（式ＡにおけるＹ₁）／（式Ａ’におけるＹ₁）および前記比の微分値、即ち、ｄ（（式ＡにおけるＹ₁）／（式Ａ’におけるＹ₁））／ｄｔを求め、前記指定期間内の前記比（式ＡにおけるＹ₁）／（式Ａ’におけるＹ₁）および前記比の微分値ｄ（（式ＡにおけるＹ₁）／（式Ａ’におけるＹ₁））／ｄｔのそれぞれの平均値を求める演算手段と、
前記比を横軸にとり、前記比の微分値を縦軸にとり、前記比（式ＡにおけるＹ₁）／（式Ａ’におけるＹ₁）の平均値と前記比の微分値ｄ（（式ＡにおけるＹ₁）／（式Ａ’におけるＹ₁））／ｄｔの平均値との交点を原点としたグラフを作成するグラフ化手段と、
所定領域を設定し、前記指定期間以降の処理中における前記第１および第２の活性種の発光検出情報を用いて比（式ＡにおけるＹ₁）／（式Ａ’におけるＹ₁）および比の微分値ｄ（（式ＡにおけるＹ₁）／（式Ａ’におけるＹ₁））／ｄｔを求め、求められた比（式ＡにおけるＹ₁）／（式Ａ’におけるＹ₁）および比の微分値ｄ（（式ＡにおけるＹ₁）／（式Ａ’におけるＹ₁））／ｄｔの前記グラフにおける位置が前記所定領域から外れたときをプラズマ処理の終点として判定する判定手段と、を具備することを特徴とする。
【００１１】
前記第１の活性種と第２の活性種として、これの発光強度が、指定期間以後のプラズマ処理の終点において、弱くなる活性種と、強くなる活性種とが使用されることが望ましい。
【００１２】
【発明の実施の形態】
以下、本発明の実施の形態のプラズマ処理の終点検出方法並びに装置を添付図面を参照して具体的に説明する。
プラズマ処理中、プラズマ状態は、印加電力、ガス流量、圧力、プラズマ温度等のさまざまな条件により影響を受けるので、不安定な状態である（プラズマの揺らぎ）。したがって、プラズマ処理の終点検出に使用する活性種の発光の強度も不安定である。
【００１３】
発光強度の変動には、図１に示すような周期的または単発のノイズ変動と、図２に示すような時間と共に徐々に増加または減少するドリフト変動の２つの要素がある。図１並びに図２において、縦軸は発光強度を、また横軸は時間を示す。このノイズ変動は、主に高周波電力、ガス流量、圧力のわずかな変動に起因するものであり、ドリフト変動は、主にプラズマ温度の時間変動に起因することが知られている。
【００１４】
ノイズ変動は、特開昭６３−８１９２９並びに特開平５−２９２７６に開示されているように、２波長を用いることにより、両波長の強度比を求めることで共通する変動成分を相殺でき、基本的に除去することができる。しかしながら、活性種からの発光と基準光又は他の活性種との比を単純にとり、その比を比較すると、ドリフト（全体的な波形の傾き）が異なる場合に、その変動成分を充分に除去することはできない。
【００１５】
本発明者らは、上記の点を鋭意検討し、米国特許Ｎｏ．５，５６５，１１４にて開示されているように波長の光の発光強度の微分値を時間的に合わせた状態で変化量を比較することにより、ドリフト変動を除去して正確な終点検出を行うことができることを見出した。しかしながら米国特許Ｎｏ．５，５６５，１１４にて開示されている２波長の光の発光強度の微分値を時間的に合わせる方法（即ち、２波長の夫々の発光強度変化曲線の傾きを一致させる）は、発明の背景で述べたように、面積計算を利用しているために、ノイズに弱いという欠点がある。そこで、本発明者達は、この欠点を考慮して、更に２波長の夫々の発光強度変化曲線の傾きを一致させる方法を改良し、ひいては、終点検出の精度向上を図ることとした。
【００１６】
次に、本発明の終点検出方法を具体的に説明する。
まず、被処理体にプラズマを用いた処理を施す際に、プラズマ処理中において指定期間（平均化時間）に、夫々特定波長（例えば、ピーク波長）を有する第１および第２の活性種の発光強度を光検出手段により逐次検出する。次いで、平均化時間内の第１および第２の活性種の発光検出情報（即ち、時間に対する発光強度の変化）に基づいて、発光強度と時間との関係における、第１の活性種の近似一次関係式Ａおよび第２の活性種の近似一次関係式Ｂを夫々求める。すなわち、縦軸を発光強度、横軸を時間にとった図３に示すように、平均化時間内の、第１の活性種の発光強度の変化波形１および第２の活性種の発光強度の波形２を用いて最小二乗法から、それぞれ実験式を求める。この場合には、実験式として近似一次関係式ＡおよびＢ（式１および式２）を求める。
【００１７】
Ｙ₁ ＝ａ₁ ×Ｘ＋ｂ₁ （Ａ） …式１
Ｙ₂ ＝ａ₂ ×Ｘ＋ｂ₂ （Ｂ） …式２
（式中、Ｙ₁ ，Ｙ₂ は、それぞれ第１および第２の活性種の発光光量を表し、Ｘは経過時間を表し、ａ₁ ，ａ₂ は一次係数を表し、ｂ₁ ，ｂ₂ はＹ切片（Ｙ−ｉｎｔｅｒｃｅｐｔ）を表す）
次いで、第１の活性種の近似一次関係式Ａおよび第２の活性種の近似一次関係式Ｂから、第１の活性種の擬似近似一次関係式Ａ´を求める。すなわち、式１のＸに式２のＸを代入して擬似近似一次関係式Ａ´（式３）を求める。
【００１８】
Ｙ_１＝（ａ_１／ａ_２）×（Ｙ_２−ｂ_２）＋ｂ_１（Ａ´） …式３
尚、式３（Ａ´）でのＹ ₂ は、第２の活性種の実測の発光光量である。尚、式（１）および式（３）はいずれも第１の活性種の発光光量Ｙ_１を表すものであり、この比（Ａ／Ａ´）（図３中に参考のために符号３で示すカーブ）を求める。さらに、この比（Ａ／Ａ´）の微分値（ｄ（Ａ／Ａ´）／ｄｔ）を求める。なお、この比（Ａ／Ａ´）は、２つの活性種が同じような発光強度変化の傾向（特徴）を示す場合には、ほぼ１となり、傾向が異なるのに従って１より大きいか小さくなる。すなわち、２つの活性種が異なる傾向を示す場合には、図３に示すように、この比の値は、終点前（後述するスロープスタート：Ｓ．Ｓの前）で一定であり、終点で大きく変化（増加または減少）し、終点後（後述するスロープエンド：Ｓ．Ｅの後）は再び一定となる。ここで、便宜上“終点”という用語を使用したが、これは、瞬間を指すのではなく、一定の期間（Ｓ．ＳからＳ．Ｅまでの時間）を指していることは、図３並びに後述する説明から理解され得よう。
【００１９】
次いで、上記に示す結果に基づいて終点を判定する。
このように、比（Ａ／Ａ´）についての波形は、終点手前と終点以降でそれぞれ一定に安定する（時間経過で変化しない）。したがって、この２つの一定値（終点手前と終点以降）間に閾値を設定することにより、この閾値検出で容易に終点の検出を行うことができる。ただし、このような終点の検出においては、比（Ａ／Ａ´）についての波形における変化の始めや終りの非常に近い位置に閾値を設定すると、安定して終点を検出することはできないことがある。このために閾値は、Ｓ．Ｓ．とＳ．Ｅ．との間で、両者から少し離れた期間に設定することが望ましい。
【００２０】
本発明においては、以下のように変化の始まり（スロープスタート：Ｓ．Ｓ．）と変化の終り（スロープエンド：Ｓ．Ｅ．）を精度良く検出して終点の判定を正確に行うことができる。
【００２１】
上記のようにして求めた発光強度の比（Ａ／Ａ´）およびその微分値、すなわち傾き（ｄ（Ａ／Ａ´）／ｄｔ）の時間変化を図４に示す。図４から分かるように、比（Ａ／Ａ´）については、終点前で一定であり、終点で大きく減少し、終点後は再び一定となる。傾き（ｄ（Ａ／Ａ´）／ｄｔ）については、終点前で一定であり、終点で大きく減少するピークを示し、終点後は終点前と同じように一定となる。
【００２２】
発光強度の比（Ａ／Ａ´）を横軸にとり、比の微分値（ｄ（Ａ／Ａ´）／ｄｔ）を縦軸にとり、比（Ａ／Ａ´）の平均値と比の微分値（ｄ（Ａ／Ａ´）／ｄｔ）の平均値との交点を原点とした２次元座標のグラフに値をプロットすると図５に示すようになる。時間の経過を曲線で示す図５から分かるように、終点以前は、座標中央に分布しており、終点付近で変化が始まると、座標の第１象限または第３象限（図５の場合には第３象限）に大きく飛び出してくる。そして、終点以降は横軸上に近付いてこの近くに分布する。このように、２次元座標上に表現することにより、比（Ａ／Ａ´）および比の微分値（ｄ（Ａ／Ａ´）／ｄｔ）の関係やそれぞれについての波形の状態変化を可視化でき、例えば、プログラム処理して画像として見ることができる。
【００２３】
終点の判定においては、まず、平均化時間内の比（Ａ／Ａ´）および比の微分値（ｄ（Ａ／Ａ´）／ｄｔ）のそれぞれの平均値および分散傾向を求める。平均化時間は、プラズマ処理の終点よりも前の期間で設定する。なお、平均化時間の開始点は、プラズマ処理開始時ではなく、プラズマが安定した時点に設定することが好ましい。
【００２４】
次いで、比（Ａ／Ａ´）および比の微分値（ｄ（Ａ／Ａ´）／ｄｔ）の分散傾向の情報から算出した後述する所定値から、グラフにおける初期の変動範囲（所定領域）を求める。この変動範囲（所定領域）は、図６に示すように、指定期間（平均化時間）の分散傾向の情報から算出した比（Ａ／Ａ´）の所定値と比の微分値（ｄ（Ａ／Ａ´）／ｄｔ）の所定値との二乗和の平方根ｒ₁ により設定されることが好ましく、実際には、分散傾向の情報から算出した比（Ａ／Ａ´）および比の微分値（ｄ（Ａ／Ａ´）／ｄｔ）の所定値の二乗和の平方根｛√［（比の所定値）² ＋（傾き（比の微分値）の所定値）² ］｝（Ｃ）を求め、その値を半径とし、平均値を原点Ｏとする、点線Ｐで示す円形の範囲である。したがって、終点以前は、この円Ｐ内に多く分布し、終点付近の変化が始まると、この円から徐々に離れていく。ここで、分散傾向の情報から算出した比（Ａ／Ａ´）の所定値として比と比の平均値との間の差の最大値を用い、比の微分値（ｄ（Ａ／Ａ´）／ｄｔ）の所定値として比の微分値と比の微分値の平均値との差の最大値を用いることができる。
【００２５】
上記の事実に基づいて、上記変動範囲を超えたときをスロープスタートとして判定する。ただし、変動範囲の近傍ではスロープスタートを正確に判定することができないので、例えば、グラフ上の変動範囲外で終点を判定する位置（値）を設定し、その位置から原点までの距離Ｌ１と、変動範囲の円の半径Ｌ２とを比較することによりスロープスタートを判定する。すなわち、平均化時間以降の処理中における第１および第２の活性種の発光検出情報を用いて比（Ａ／Ａ´）および比の微分値（ｄ（Ａ／Ａ´）／ｄｔ）を求め、グラフ上の位置から原点までの距離Ｌ１を求め、この距離と平均化時間内の発光検出情報で求められた変動範囲の半径Ｌ２とを比較する。
【００２６】
具体的には、終点を判定する位置に対応する比（Ａ／Ａ´）および比の微分値（ｄ（Ａ／Ａ´）／ｄｔ）の値と、それぞれの平均値（原点）との差の二乗和の平方根｛√［（比−比の平均値）² ＋（傾き−傾きの平均値）² ］｝（Ｄ）を求め、上記二乗和の平方根（Ｃ）との比（Ｄ／Ｃ）があらかじめ設定した閾値より大きくなったところをスロープスタートとして判定する。
【００２７】
一方、スロープエンドは、比の微分値（ｄ（Ａ／Ａ´）／ｄｔ）、すなわち傾きが再び横軸に近付く（傾きの変動範囲に近付く）ことにより判定する。すなわち、終点を判定する傾きとその分散傾向から求められた所定値とを比較して、これから得られた（（傾き−傾きの平均値）／所定値）を求め、この値があらかじめ設定した閾値より小さくなったところをスロープエンドとして判定する。なお、分散傾向から求められる所定値は、上述したようにして算出する。
【００２８】
スロープエンドを終点検出に使用する場合、指定期間以降の処理中における第１および第２の活性種の発光検出情報を用いて求められた比（Ａ／Ａ´）および比の微分値（ｄ（Ａ／Ａ´）／ｄｔ）のグラフにおける位置が原点から外れて再び横軸に近接するときをプラズマ処理の終点として判定する。
【００２９】
この場合、スロープスタートを終点検出に使用する場合と同様に、変動範囲を設定して、指定期間以降の処理中における第１および第２の活性種の発光検出情報を用いて比および比の微分値を求め、求められた比および比の微分値のグラフにおける位置が変動範囲（所定領域）内に入ったときをプラズマ処理の他の終点として判定しても良い。なお、変動範囲の設定は、スロープスタートを終点検出に使用する場合と同様である。
【００３０】
スロープエンドを終点検出に使用する場合においても、スロープスタートを終点検出に使用する場合と同様に、順次新たな原点や変動範囲を設定して、複数回の終点検出を行っても良い。
【００３１】
このように、終点の判定において、閾値との直接比較を行わず、プラズマ処理の平均化時間内の情報から得られた所定値を終点の判定に利用することにより、２波長を用いた従来の終点検出方法で除去できないドリフト変動を充分に除去することができる。したがって、Ｓ／Ｎ（シグナル・ノイズ）が悪い状態であっても正確に終点を検出することができる。なお、実際のプラズマ処理の終点は、スロープスタートの位置としても良く、スロープエンドの位置としても良い。この選定はプラズマ処理の目的や条件等により適宜行う。
【００３２】
本発明の終点検出方法においては、図７の（Ａ）に示すように、発光強度の比の時間変化の波形が終点近傍で上昇する形状である場合にも、図７の（Ｂ）に示すように、上記と同様にしてスロープスタートとスロープエンドを判定することができる。この場合には、結果の曲線が座標の第１象限に大きく飛び出してくることが認識できよう。
【００３３】
異なる種類の２つの膜をエッチングしてホールを形成する場合、例えば、図８に示すように、シリコン基板４上にＳｉＯ₂ 膜５とＳｉ₃ Ｎ₄ 膜６とを順次形成したものをエッチングする場合には、図９の（Ａ）に示すように、発光強度の比の時間変化の波形が１回目、即ち、１段目の終点近傍で上昇し、２回目、即ち、２段目の終点近傍で下降する形状となる。この場合には、図９の（Ｂ）に示すように、上記のようにして１回目の終点を検出した後、求められた比（Ａ／Ａ´）および比の微分値（ｄ（Ａ／Ａ´）／ｄｔ）のグラフにおける位置がグラフの横軸と再び交わる点を新たな原点Ｏ₂ とし、分散傾向の情報から算出された比（Ａ／Ａ´）の所定値および比の微分値（ｄ（Ａ／Ａ´）／ｄｔ）の所定値から他の所定領域（変動範囲）Ｒ₂ を設定し、第１および第２の活性種の発光検出情報を用いて比（Ａ／Ａ´）および比の微分値（ｄ（Ａ／Ａ´）／ｄｔ）を求め、求められた比（Ａ／Ａ´）および比の微分値（ｄ（Ａ／Ａ´）／ｄｔ）のグラフにおける位置が他の所定領域Ｒ₂ から外れたときをプラズマ処理の２回目の終点（スロープスタート）として判定する。なお、図９の（Ｂ）中、Ｏ₁ は１回目の終点検出の際の原点であり、Ｒ₁ は１回目の終点検出の際の変動範囲である。また、２回目のスロープスタートおよびスロープエンドの判定方法は上記と同様に行う。
【００３４】
この結果、Ｓｉ₃ Ｎ₄ 膜６にホールを形成するためのエッチングの最終点と、これに続くＳｉＯ₂ 膜５にホールを形成するためのエッチングの最終点とを夫々認識することができる。
このような原点移動を行ってスロープスタートおよびスロープエンドの判定を繰り返す技術は、図１０に示すような、段差がある基板７上に形成された部分的に厚さが異なるＳｉＯ₂ 膜５にアスペクト比が異なる穴８ａ〜８ｃを形成する場合にも適用できる。この場合においても、図１１の（Ａ）および図１１の（Ｂ）に示すように、上記と同様に原点移動を行うことにより、複数回のスロープスタートおよびスロープエンドの判定を行うことができ、正確に各ステップのエッチング処理の終点検出を行うことができる。なお、図１１の（Ｂ）中、Ｏ₁ は１回目、即ち、１段目の終点検出の際の原点、Ｏ₂ は２回目、即ち、段目の終点検出の際の原点、Ｏ₃ は３回目、即ち、３段目の終点検出の際の原点を夫々示し、Ｒ₁ は１回目の終点検出の際の変動範囲、Ｒ₂ は２回目の終点検出の際の変動範囲、Ｒ₃ は３回目の終点検出の際の変動範囲を、夫々示す。
【００３５】
次に、本発明のプラズマ処理の終点検出方法の効果を明確にするために行った実施例について説明する。
図１２は本発明に係る終点検出装置を備えたプラズマ処理装置、例えば、プラズマエッチング装置、の構成を概略的に示す図である。プラズマ処理装置１０は、例えばアルミニウム等の導電性材料からなる処理室１１と、この処理室１１内に配設されかつ被処理体である半導体ウェハＷを載置する載置台を兼ねた下部電極１２と、この下部電極１２の上方に下部電極１２と離隔して配設された上部電極１３とを備え、これら電極間にプラズマ発生領域が規定されている。
【００３６】
処理室１１の上部には、処理ガス、例えば、ＣＦ₄ 等のフルオロカーボン系のエッチング用ガスを導入するためのガス導入管１４が接続されている。処理室１１の側壁には、生成ガスを排出するための排気管１５が接続されている。下部電極１２は接地されており常時グランド電位に保たれている。また、上部電極１３は高周波電源１６に接続されており、高周波電源１６から高周波電圧を印加して下部電極１２と上部電極１３との間で放電させることにより、エッチング用ガスを活性化してラジカル種、イオン種等の活性種からなるプラズマＰをプラズマ発生領域に発生させる。
【００３７】
処理室１１の側壁には、石英ガラス等の透明体からなる監視用窓１７が取り付けられており、この窓１７からプラズマＰの発光スペクトルを透過させ、この透過光を分析することによりエッチングの進捗状況を監視する。窓１７の外部には、透過光を集光するためのレンズ２１が配設され、さらにレンズ２１の後段には、レンズ２１によって集光された光を検出して光電変換する光検出器２２が配設されている。この光検出器２２は、例えば、１対の干渉フィルタまたは分光器と、１対のフォトマルまたはフォトダイオードとから構成されており、２つの特定波長の光を干渉フィルタまたは分光器で夫々分光した後、分光した特定波長の光を夫々光電変換して、時間に対しての発光強度の変化を表す電気信号として送信する。この光検出器２２から送信される発光強度の時間に対する変化に対応した２つの電気信号に基づいて後述の演算装置３０でエッチングの終点を検出し、終点を検出した時点で制御信号を制御装置４０に送信し、この制御装置４０を介してプラズマ処理装置１０を制御して、即ち、高周波電源１６の発信を停止させてエッチングを終了させる。
【００３８】
ここで、レンズ２１の位置は、レンズ移動手段（図示せず）により適宜垂直方向に移動することができる。例えば、半導体基板上に形成された膜をホール形成のためにプラズマエッチングする場合において、特定波長を有する発光スペクトルを検出するとき、膜の上面から反射した光と、膜の下面（半導体基板と膜との界面）から反射した光とが干渉して光検出器２２に入ると、正確に発光スペクトルの発光強度を検出することができなくなる可能性がある。このような干渉光が入射しないように、この実施の形態では、レンズ移動手段によりレンズの焦点位置をずらせることができる。
【００３９】
次に、本発明に係る終点検出を行う演算装置３０について説明する。演算装置３０は、図１３に示すように、光検出器２２からの入力信号、すなわち第１の特定波長の光の発光強度の変化を表す信号と、第２の特定波長の光の発光強度の変化を表す信号の夫々の情報を演算して、両発光強度の変化の比およびこの比の微分値（傾き）を抽出する、即ち、計算する要素抽出器３１と、発光強度の比（Ａ／Ａ´）を横軸にとり、比の微分値（ｄ（Ａ／Ａ´）／ｄｔ）、即ち、傾きを縦軸にとり、比（Ａ／Ａ´）の平均値と比の微分値（ｄ（Ａ／Ａ´）／ｄｔ）の平均値との交点を原点としたグラフに、上記比と比の微分値の時間に対する変化をプロットして、図５に示すようなグラフを作成するグラフ化器３２と、作成されたグラフからスロープスタートを判定するスロープスタート判定器３３と、作成されたグラフからスロープエンドを判定するスロープエンド判定器３４とにより構成されている。この好ましい実施の形態においては、前述した積層膜や段差を有する基板上の膜の処理の際に作成されたグラフにおける原点を移動する原点移動器３５が、判定器３３，３４の出力側に設けられている。これら原点移動器は、上述したような原点を移動させる場合にのみ駆動されるものであり、要素抽出器３１に、同様の操作を繰り返すように指令する。
【００４０】
さらに詳しくは、要素抽出器３１においては、次の演算処理が行われる。
（１）指定期間（平均化時間）内の第１および第２の活性種の発光検出情報に基づいて、発光強度と時間との関係において第１の活性種の近似一次関係式Ａおよび第２の活性種の近似一次関係式Ｂを求める。
（２）第１の活性種の近似一次関係式Ａおよび第２の活性種の近似一次関係式Ｂを用いて、即ち、第２の活性種の近似一次関係式Ｂの時間経過成分を第１の活性種の近似一次関係式Ａの時間経過成分に代入して、第１の活性種の擬似近似一次関係式Ａ´を求める。
（３）第１の活性種の近似一次関係式Ａおよび第１の活性種の擬似近似一次関係式Ａ´から、両者の比（Ａ／Ａ´）および比の微分値（ｄ（Ａ／Ａ´）／ｄｔ）を求める。
（４）平均化時間内の比（Ａ／Ａ´）および比の微分値（ｄ（Ａ／Ａ´）／ｄｔ）のそれぞれの平均値および分散値を求める。
【００４１】
グラフ化器３２、スロープスタート判定器３３、およびスロープエンド判定器３４では、それぞれ上述したような処理を行って、それぞれグラフ作成、スロープスタート判定、およびスロープエンド判定を行う。
【００４２】
スロープスタート判定およびスロープエンド判定の判定結果は、制御装置４０へ送信され、その判定結果の信号に基づいて制御装置４０を介して高周波電源１６等を制御することによりエッチング処理を制御する。なお、スロープエンド判定の判定結果は、必要に応じて原点移動器３５に送られ、その原点移動がなされた結果の信号もまた制御装置４０に送られる。
【００４３】
上記構成を有するプラズマ処理装置（プラズマエッチング装置）を用いてシリコン基板上に形成されたＳｉＯ₂ 膜をＣＦ₄ ガスを用いて実際にエッチングする場合について説明する。ここでは、第１の活性種としてＳｉＯ₂ のエッチングにより生成するＣＯ分子を用い、第２の活性種としてイオン化されたエッチャントであるＣＦ₂ ^* 分子を用いる。また、ＣＯ分子はスペクトロスコープで分析し、ＣＦ₂ ^* 分子はオプトフィルタでフィルタリングした。なお、ＣＯ分子の使用発光波長は約２１９ｎｍであり、ＣＦ₂ ^* 分子の使用発光波長は約２６０ｎｍである。
【００４４】
まず、光検出器２２により検出され、フィルタリングされた光信号を、光電変換器（図示せず）により電気信号に変換し、その電気信号はプリアンプ（図示せず）により増幅する。
【００４５】
次いで、ＣＯ分子の電気信号のプリアンプの増幅率を調節してＣＦ₂ ^* 分子と同じレベルにＣＯ分子の電気信号を増幅する。この段階で両電気信号はアナログ信号である。
【００４６】
次いで、サンプリングサイクルの２倍以上の周波数、この場合２０Ｈｚ以上を有するノイズ成分をフィルタでカットする。その後、両電気信号を０．１秒のサイクルでサンプリングし、Ａ／Ｄ変換器によりデジタル化する。
【００４７】
次いで、デジタル化された両電気信号を動的な平均化方法により平滑化する。これは、いわゆるローパスフィルタの効果を有し、高周波数ノイズ（すなわちランダムノイズ）を含まない比較的平滑な信号が得られる。
【００４８】
次いで、要素抽出器３１において、ＣＯ分子の近似一次関係式ＡおよびＣＦ₂ ^* 分子の近似一次関係式Ｂを求める。これらの発光強度の時間変化及び一次関係式を示す直線を図１４に示す。また、これらを用いてＣＦ₂ ^* 分子の近似一次関係式ＢからＣＯ分子の擬似近似一次関係式Ａ´を求める。このこれら関係式Ａ，Ｂ，Ａ´に対応する発光強度の時間変化を図１５に示す。
【００４９】
近似一次関係式Ａおよび擬似近似一次関係式Ａ´は、いずれもＣＯ分子の発光光量（強度）を表すものであり、その比（Ａ／Ａ´）および比の微分値（ｄ（Ａ／Ａ´）／ｄｔ）を求める。比（Ａ／Ａ´）の時間変化を図１６に示す。図１６には、比較のために、ＣＯ分子の近似一次関係式ＡおよびＣＦ₂ ^* 分子の近似一次関係式Ｂの単なる比（Ａ／Ｂ）の時間変化を併記した。図１６から明らかなように、本発明の方法によれば、終点前後において波形が平坦であるので、終点検出を正確に行うことができる。一方、単なる比（Ａ／Ｂ）の場合には、波形が安定せず、ドリフト変動により終点を正確に検出することができない。
【００５０】
次いで、比（Ａ／Ａ´）を横軸とし、比の微分値（ｄ（Ａ／Ａ´）／ｄｔ）を縦軸とし、指定期間（平均化時間）内の情報から算出した比および比の微分値の平均値を原点としたグラフを作成する。グラフ上に比および比の微分値をプロットすると、図１７に示すようになる。
【００５１】
さらに、グラフ上において、平均化時間内に求められた分散傾向の情報から算出された比（Ａ／Ａ´）および比の微分値（ｄ（Ａ／Ａ´）／ｄｔ）の所定値から、グラフにおける初期の変動範囲を求める。
【００５２】
次いで、グラフ上の変動範囲外で終点を判定する位置（値）を設定し、その位置から原点までの距離と、変動範囲の円の半径とを比較することにより、スロープスタート判定器３３においてスロープスタートを判定する。すなわち、平均化時間以降の処理中におけるＣＯ分子およびＣＦ₂ ^* の発光検出情報を用いて比（Ａ／Ａ´）および比の微分値（ｄ（Ａ／Ａ´）／ｄｔ）を求め、グラフ上の位置から原点までの距離を求め、この距離と平均化時間内の発光検出情報で求められた変動範囲の半径とを比較する。
【００５３】
具体的には、終点を判定する位置に対応する比および比の微分値の値と、それぞれの平均値（原点）との差の二乗和の平方根（Ｄ）を求め、上記円の半径（Ｃ）との比（Ｄ／Ｃ）があらかじめ設定した閾値より大きくなったところをスロープスタートとして判定する。
【００５４】
一方、スロープエンドは、比の微分値、すなわち傾きが再び横軸に近付くことにより判定する。すなわち、終点を判定する傾きとその分散傾向に基づく所定値とを比較して、これから得られた（（傾き−傾きの平均値）／所定値）を求め、この値があらかじめ設定した閾値より小さくなったところをスロープエンド判定器３４によりスロープエンドとして判定する。
【００５５】
上述した本発明の終点検出方法によれば、発光光量（開口率）を従来に比べて３倍（１％から３％）に向上させることができる。
上記実施例では、第１の活性種としてＣＯ分子を用い、第２の活性種としてＣＦ₂ ^* 分子を用いた場合について説明しているが、本発明は、第１の活性種および第２の活性種として他の活性種を用いた場合にも適用することができる。
【００５６】
上記実施形態においては、プラズマ処理がエッチングである場合について説明しているが、本発明はプラズマ処理がＣＶＤ（Chemical Vapor Deposition ）やＰＶＤ（Physical Vapor Deposition ）等のプラズマを使用した処理の場合についても同様に適用することができる。
【００５７】
上記実施の形態では、第１の活性種および第２の活性種の特定波長での発光スペクトルの発光強度の時間に対する変化に近似する実験式を近似一次関係式をしようしたが、実験式は、これに限定されることはないことは、当業者にとって自明であろう。例えば、時間に対する変化が、楕円もしくは双曲線の一部に沿うように変化する場合には、これら二次関係式を使用することができる。即ち、本発明においては、発光強度の変化に最も近似した実験式を使用することが、より精度の良い測定をするために好ましい。
【００５８】
前記第１の活性種と第２の活性種として、これの発光強度が、指定期間以後のプラズマ処理の終点において、図４に示すように弱くなる活性種と、図７の（Ａ）に示すように強くなる活性種とが使用されることが、感度を高めるために好ましい。
【００５９】
以上説明したように本発明によれば、閾値との直接比較を行わず、プラズマ処理の指定期間内の分散傾向の情報から得られた所定値を終点の判定に利用しているので、２波長を用いた従来の終点検出方法で除去できないドリフト変動を充分に除去することができる。したがって、Ｓ／Ｎが悪い状態であっても正確に終点を検出することができる。このように、正確にプラズマの終端を検出できるので、発光光量（開口率）を従来に比べて３倍に向上させることができる。
【図面の簡単な説明】
【図１】プラズマ発光強度のノイズ変動を説明するための図。
【図２】プラズマ発光強度のドリフト変動を説明するための図。
【図３】第１および第２の活性種の近似一次関係式並びに第１の活性種の擬似近似一次関係式を説明するための図。
【図４】発光強度（比）と傾き（微分値）との時間変化を示す特性図。
【図５】発光強度（比）を横軸とし、傾き（微分値）を縦軸としたときの、図４に示す値をプロットしたグラフ。
【図６】スロープスタートを説明するための図。
【図７】（Ａ）は発光強度（比）および傾き（微分値）の時間変化の他の例を示す特性図、そして、（Ｂ）は発光強度（比）を横軸とし、傾き（微分値）を縦軸とした、（Ａ）に示す値をプロットしたグラフ。
【図８】被エッチング体の一例を示す図。
【図９】（Ａ）は、発光強度（比）および傾き（微分値）の時間変化の他の例を示す、図７の（Ａ）と同様の特性図、そして、（Ｂ）は発光強度（比）を横軸とし、傾き（微分値）を縦軸とした、（Ａ）に示す値をプロットしたグラフ。
【図１０】被エッチング体の他の例を示す断面図。
【図１１】（Ａ）は、は発光強度（比）および傾き（微分値）の時間変化の他の例を示す特性図、そして、（Ｂ）は発光強度（比）を横軸とし、傾き（微分値）を縦軸とした、（Ａ）に示す値をプロットしたグラフ。
【図１２】本発明の終点検出装置を有するプラズマエッチング装置を概略的に示す図。
【図１３】図１２に示すプラズマエッチング装置における終点検出装置を説明するための図。
【図１４】ＣＯ分子とＣＦ₂ ^* の発光強度の時間変化を示す特性図。
【図１５】演算後のＣＯ分子とＣＦ₂ ^* の発光強度の時間変化を示す特性図。
【図１６】ＣＯ分子とＣＦ₂ ^* の発光強度の変化率（比）の時間変化を示す特性図。
【図１７】ＣＯ分子とＣＦ₂ ^* の発光強度の変化率（比）を横軸とし、その傾き（微分値）を縦軸としたときのグラフ。
【符号の説明】
１０…プラズマ処理装置、２２…光検出器、３０…演算装置、３１…要素抽出器、３２…グラフ化器、３３…スロープスタート判定器、３４…スロープエンド判定器、３５…原点移動器、４０…制御装置。[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a plasma processing end point detection method and apparatus.
[0002]
[Prior art]
2. Description of the Related Art Conventionally, plasma processing apparatuses, particularly etching apparatuses, have been widely applied to semiconductor manufacturing processes or liquid crystal display substrate manufacturing processes. Such an etching apparatus includes, for example, an upper electrode and a lower electrode arranged in parallel to each other, generates plasma from an etching gas by discharge between the upper electrode and the lower electrode, and activates the active species. Then, a film of a semiconductor wafer having a film such as an oxide film formed thereon as an object to be processed is etched. In such an etching process, it is desired to perform the etching process according to a predetermined pattern by monitoring the progress of etching and detecting the end point as accurately as possible.
[0003]
Conventionally, instrumental analysis techniques such as mass spectrometry and spectroscopic analysis have been used as methods for detecting the end point of etching, and comparatively simple and highly sensitive spectroscopic analysis has been widely used. When spectroscopic analysis is used to detect the end point of etching, specifically, a specific type of active species such as an etching gas, radicals such as decomposition products or reaction products thereof, or active species such as ions. A species is selected, and the change in the intensity of light emission of the selected active species with respect to time is measured. In this case, the active species to be selected differs depending on the type of etching gas and the material to be etched. For example, CF_Four In the case of etching a silicon oxide film using a fluorocarbon-based etching gas such as CO, it is a reaction product of which CO emission intensity rapidly decreases at the end point.^* Is used. In this case, CO^* Measures only the emission intensity (using a wavelength of 219 nm or 483.5 nm) (using only one wavelength) with respect to time from the time, and the amount of change over time, such as the emission intensity and the first and second derivative values of the intensity And a method of determining the end point of etching by comparing the above and CO 2 as disclosed in JP-A-63-81929.^* The emission intensity with respect to the time from the light and the emission intensity with respect to the time of the reference light of two wavelengths (in Japanese Patent Laid-Open No. 63-81929, the emission spectrum intensity of atoms such as helium is 706.5 nm and 667.8 nm). JP-A-5-29276 discloses a method for determining the end point of etching by comparing the respective emission intensity ratios or the amount of change such as the primary differential value and the secondary differential value of the ratio, and a method using two wavelengths. (Corresponding US Pat. No. 5,322,590), as disclosed in CO^* Is a decomposition product of etching gas whose emission intensity rapidly increases at the end point instead of the above-mentioned reference light and the emission intensity with respect to time from^* A method for determining the end point of etching by measuring the emission intensity with respect to each time and comparing the ratio of each emission intensity or the amount of change of the ratio with respect to time, such as the first derivative value and the second derivative value, is known. ing.
[0004]
In the conventional end point detection method using one wavelength, the end point of the plasma processing becomes unclear due to emission intensity fluctuation caused by plasma fluctuation or the like, and the end point cannot be accurately detected. In the conventional end point detection method using two wavelengths, CO, which is an active species of the reaction product, is used.^* Is a decomposition product of etching gas whose emission intensity increases sharply at the reference light or end point.^* Without taking into account that the amount of change in the emission intensity for the end-point time differs due to plasma fluctuations, temperature changes in the chamber, electrode, or wafer, and deposits attached to the chamber wall, the emission intensity ratio between the two is simply calculated. However, since the end point is detected using the ratio, it is still difficult to accurately detect the end point.
[0005]
U.S. Pat. At 5,565,114, the present inventors matched the amount of change of both emission intensities with respect to time when detecting the etching end point using the emission intensity ratio at each wavelength using two wavelengths. The idea of taking the ratio between the two (that is, matching the slopes of the change curves of the light emission intensity with respect to both times) is disclosed. Certainly, rather than simply monitoring the ratio between the two, as disclosed in this publication (US Pat. No. 5,565,114), after matching the slopes of the two change curves in advance, both It is possible to detect the end point more accurately by monitoring the ratio.
[0006]
[Problems to be solved by the invention]
However, the method of matching the slopes of both disclosed in this US patent calculates the average value of the change curve of the emission intensity in the designated section before the end point, and calculates the average value of the emission intensity and the average for each change curve in the designated section. Since the sum of absolute values of differences from the values is obtained and this sum is used (that is, the area is calculated), there is a drawback that it is vulnerable to noise.
An object of the present invention is to provide a plasma processing end point detection method and apparatus capable of accurately detecting an end point while allowing fluctuations in the plasma state.
[0007]
[Means for Solving the Problems]
  In the plasma processing end point detection method according to the first aspect of the present invention, when the object to be processed is processed using plasma, the first and second during the specified period during the plasma processing and thereafter. Detecting the emission intensity of each of the active species having a specific wavelength, and outputting this emission detection information;
  Based on the light emission detection information, the approximate relational expression A of the first active species in the relationship between the light emission intensity and time, that is, Y₁ = A₁ × X + b₁, And the approximate relation B of the second active species, ie, Y₂ = A₂ × X + b₂(Here, Y₁ , Y₂ Are the light emission amounts of the first and second active species, respectively.approximationX represents elapsed time, a₁ , A₂ Represents the first order coefficient, b₁ , B₂ Represents a Y-intercept),
  Using the approximate relational expression A of the first active species and the approximate relational expression B of the second active species, a pseudo approximate relational expression A ′ of the first active species, that is, Y₁ = (A₁ / A₂ ) X (Y₂ -B₂ + B₁Process(Y here ₂ Is the measured value)When,
  From the approximate relational expression A of the first active species and the pseudo approximate relational expression A ′ of the first active species, the ratio between them, that is, (Y in Expression A)₁) / (Y in formula A '₁) And the differential value of the ratio, ie, d ((Y in Equation A₁) / (Y in formula A '₁)) / Dt obtaining step;
  The ratio within the specified period (Y in Equation A)₁) / (Y in formula A '₁) And the differential value d ((Y in Formula A)₁) / (Y in formula A '₁))) Obtaining the average value of / dt;
  A predetermined area is set in a graph with the ratio on the horizontal axis, the differential value of the ratio on the vertical axis, and the origin of the intersection of the average value of the ratio and the average value of the differential value of the ratio within the specified period. And a process of
  The ratio using the luminescence detection information of the first and second active species during the processing after the specified period, ie, (Y in Formula A₁) / (Y in formula A '₁) And the differential value of the ratio, ie d ((Y in equation A₁) / (Y in formula A '₁)) / Dt, and determining when the obtained ratio and the position of the differential value of the ratio in the graph deviate from the predetermined region as the end point of the plasma processing.
[0008]
  According to another aspect of the present invention, there is provided a method for detecting an end point of a plasma treatment, wherein a first active species and a second active species are applied during a designated period during the plasma treatment and thereafter after the treatment using plasma. A step of sequentially detecting light emission intensity at a specific wavelength of light emission from the light detection means, and outputting light emission detection information;
  Based on the light emission detection information within the specified period, the approximate relational expression A of the first active species in terms of the light emission intensity and time, that is, Y₁ = A₁ × X + b₁, And the approximate relation B of the second active species, ie, Y₂ = A₂ × X + b₂(Here, Y₁ , Y₂ Are the light emission amounts of the first and second active species, respectively.approximationX represents elapsed time, a₁ , A₂ Represents the first order coefficient, b₁ , B₂ Represents a Y-intercept),
  Using the approximate relational expression A of the first active species and the approximate relational expression B of the second active species, a pseudo approximate relational expression A ′ of the first active species, that is, Y₁ = (A₁ / A₂ ) X (Y₂ -B₂ + B₁Process(Y here ₂ Is the measured value)When,
  From the approximate relational expression A of the first active species and the pseudo approximate relational expression A ′ of the first active species, the ratio between them, that is, (Y in Expression A)₁) / (Y in formula A '₁) And the differential value of the ratio, ie, d ((Y in Equation A₁) / (Y in formula A '₁)) / Dt obtaining step;
  The ratio within the specified period (Y in Equation A)₁) / (Y in formula A '₁) And the differential value d ((Y in Formula A)₁) / (Y in formula A '₁))) Obtaining the average value of / dt;
  The ratio is plotted on the horizontal axis, the differential value of the ratio is plotted on the vertical axis, and the intersection point between the average value of the ratio and the average value of the differential value of the ratio in the specified period is the graph after the specified period. The ratio obtained by using the luminescence detection information of the first and second active species during the treatment of (1)₁) / (Y in formula A '₁) And the differential value of the ratio, ie d ((Y in equation A₁) / (Y in formula A '₁))) When the position of / dt in the graph deviates from the origin and approaches the horizontal axis again, the step of determining as an end point of the plasma processing is provided.
[0010]
  The plasma processing end point detection apparatus of the present invention detects a change with time of the emission intensity at a specific wavelength of the first and second active species generated by the plasma when the object to be processed is processed using plasma. A light detection means for detecting light emission detection information;
  Based on the light emission detection information within the designated period, the approximate relational expression A of the first active species in the relationship between the light emission intensity and time, that is, Y₁ = A₁ × X + b₁, And the approximate relation B of the second active species, ie, Y₂ = A₂ × X + b₂(Here, Y₁ , Y₂ Are the light emission amounts of the first and second active species, respectively.Approximate value ofX represents elapsed time, a₁ , A₂ Represents the first order coefficient, b₁ , B₂ Represents a Y-intercept), and using the approximate relational expression A of the first active species and the approximate relational expression B of the second active species, a pseudo approximate relational expression A ′ of the first active species, that is, , Y₁ = (A₁ / A₂ ) X (Y₂ -B₂ + B₁(Y here₂Is an actual measurement value), from the approximate relational expression A of the first active species and the pseudo approximate relational expression A ′ of the first active species, the ratio between them, that is, (Y in Formula A)₁) / (Y in formula A '₁) And the differential value of the ratio, ie, d ((Y in Equation A₁) / (Y in formula A '₁)) / Dt, and the ratio within the specified period (Y in Equation A)₁) / (Y in formula A '₁) And the differential value d ((Y in Formula A)₁) / (Y in formula A '₁)) / Dt for calculating each average value of / dt;
  The ratio is plotted on the horizontal axis, and the differential value of the ratio is plotted on the vertical axis.₁) / (Y in formula A '₁) And a differential value d of the ratio ((Y in Formula A)₁) / (Y in formula A '₁)) Graphing means for creating a graph with the origin at the intersection with the average value of / dt;
  A predetermined region is set, and the ratio (Y in the expression A is used) using the light emission detection information of the first and second active species during the processing after the specified period.₁) / (Y in formula A '₁) And the differential value d ((Y in Formula A)₁) / (Y in formula A '₁)) / Dt and the determined ratio (Y in Formula A)₁) / (Y in formula A '₁) And the differential value d ((Y in Formula A)₁) / (Y in formula A '₁)) / Dt determining means for determining when the position in the graph deviates from the predetermined region as an end point of the plasma processing.
[0011]
As the first active species and the second active species, it is desirable to use an active species whose emission intensity becomes weak and an active species whose emission intensity becomes weak at the end of the plasma treatment after a specified period.
[0012]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, an end point detection method and apparatus for plasma processing according to embodiments of the present invention will be described in detail with reference to the accompanying drawings.
During plasma processing, the plasma state is affected by various conditions such as applied power, gas flow rate, pressure, and plasma temperature, and thus is unstable (plasma fluctuation). Therefore, the emission intensity of the active species used for detecting the end point of the plasma treatment is also unstable.
[0013]
The fluctuation of the emission intensity has two elements: a periodic or single noise fluctuation as shown in FIG. 1 and a drift fluctuation that gradually increases or decreases with time as shown in FIG. 1 and 2, the vertical axis represents the emission intensity, and the horizontal axis represents time. This noise fluctuation is mainly caused by slight fluctuations in the high-frequency power, gas flow rate, and pressure, and the drift fluctuation is known to be mainly caused by time fluctuations in the plasma temperature.
[0014]
As disclosed in Japanese Patent Laid-Open No. 63-81929 and Japanese Patent Laid-Open No. 5-29276, the noise fluctuation can be canceled out by using two wavelengths to cancel the common fluctuation component by obtaining the intensity ratio of both wavelengths. Can be removed. However, when the ratio between the emission from the active species and the reference light or other active species is simply taken and the ratios are compared, if the drift (overall waveform slope) is different, the fluctuation component is sufficiently removed. It is not possible.
[0015]
The present inventors have intensively studied the above points, and have found that US Pat. As disclosed in US Pat. No. 5,565,114, the end point is accurately detected by removing the drift variation by comparing the amount of change in a state in which the differential value of the emission intensity of the light of the wavelength is matched with time. I found that I can do it. However, US Pat. No. 5,565,114 discloses a method for temporally matching the differential values of the light emission intensities of two wavelengths (that is, matching the slopes of the respective light emission intensity change curves of the two wavelengths). As described above, since area calculation is used, there is a drawback that it is vulnerable to noise. In view of this drawback, the present inventors further improved the method for matching the slopes of the respective emission intensity change curves of the two wavelengths to improve the accuracy of end point detection.
[0016]
Next, the end point detection method of the present invention will be specifically described.
First, when performing processing using plasma on an object to be processed, light emission of first and second active species each having a specific wavelength (for example, peak wavelength) during a specified period (average time) during plasma processing. The intensity is sequentially detected by the light detection means. Next, based on the luminescence detection information of the first and second active species within the averaging time (that is, the change in luminescence intensity with respect to time), the approximate primary of the first active species in the relationship between the luminescence intensity and time. Relational expression A and approximate primary relational expression B of the second active species are obtained respectively. That is, as shown in FIG. 3 where the vertical axis represents the emission intensity and the horizontal axis represents the time, the change waveform 1 of the emission intensity of the first active species and the emission intensity of the second active species within the averaging time are shown. Empirical formulas are obtained from the least square method using waveform 2. In this case, approximate primary relational expressions A and B (Expression 1 and Expression 2) are obtained as empirical expressions.
[0017]
Y₁ = A₁ × X + b₁   (A) ... Formula 1
Y₂ = A₂ × X + b₂   (B) ... Formula 2
(Where Y₁ , Y₂ Represents the amount of emitted light of the first and second active species, respectively, X represents the elapsed time, a₁ , A₂ Represents the first order coefficient, b₁ , B₂ Represents Y-intercept)
Next, a pseudo approximate primary relational expression A ′ of the first active species is obtained from the approximate primary relational expression A of the first active species and the approximate primary relational expression B of the second active species. That is, the pseudo approximate primary relational expression A ′ (Expression 3) is obtained by substituting X of Expression 2 into X of Expression 1.
[0018]
  Y₁= (A₁/ A₂) X (Y₂-B₂+ B₁(A ') ... Formula 3
  Y in formula 3 (A ') ₂ Is the measured light emission quantity of the second active species.It should be noted that the equations (1) and (3) are both emitted light amounts Y of the first active species.₁This ratio (A / A ′) (a curve indicated by reference numeral 3 for reference in FIG. 3) is obtained. Further, a differential value (d (A / A ′) / dt) of this ratio (A / A ′) is obtained. Note that this ratio (A / A ′) is approximately 1 when the two active species show the same tendency (characteristic) of change in emission intensity, and becomes larger or smaller as the trend is different. That is, when the two active species show different tendencies, as shown in FIG. 3, the value of this ratio is constant before the end point (slope start: before SS) described later, and is large at the end point. It changes (increases or decreases) and becomes constant again after the end point (after slope end: S.E described later). Here, for convenience, the term “end point” is used, but this does not indicate an instant but indicates a certain period (time from SS to SE) as shown in FIG. Can be understood from the explanation.
[0019]
Next, the end point is determined based on the result shown above.
As described above, the waveform of the ratio (A / A ′) is stabilized stably before the end point and after the end point (does not change over time). Therefore, by setting a threshold value between these two constant values (before and after the end point), the end point can be easily detected by this threshold value detection. However, in such detection of the end point, if the threshold is set at a position very close to the start or end of the change in the waveform of the ratio (A / A ′), the end point cannot be detected stably. is there. For this reason, the threshold is S.I. S. And S. E. It is desirable to set a period slightly away from both.
[0020]
In the present invention, the start of change (slope start: S.S.) and the end of change (slope end: S.E.) can be accurately detected and the end point can be accurately determined as follows. .
[0021]
FIG. 4 shows the time variation of the emission intensity ratio (A / A ′) and its differential value, that is, the slope (d (A / A ′) / dt) obtained as described above. As can be seen from FIG. 4, the ratio (A / A ′) is constant before the end point, greatly decreases at the end point, and becomes constant again after the end point. The slope (d (A / A ′) / dt) is constant before the end point and shows a peak that greatly decreases at the end point, and is constant after the end point in the same manner as before the end point.
[0022]
The ratio of emission intensity (A / A ′) is taken on the horizontal axis, the differential value of ratio (d (A / A ′) / dt) is taken on the vertical axis, and the average value of ratio (A / A ′) and the differential value of ratio When values are plotted on a graph of two-dimensional coordinates with the origin at the intersection with the average value of (d (A / A ′) / dt), the result is as shown in FIG. As can be seen from FIG. 5 that shows the passage of time as a curve, before the end point, it is distributed in the center of the coordinate, and when the change starts near the end point, the first quadrant or the third quadrant of the coordinate ( The third quadrant) jumps out greatly. And after the end point, it approaches the horizontal axis and is distributed near this. In this way, by expressing on two-dimensional coordinates, the relationship between the ratio (A / A ′) and the differential value of the ratio (d (A / A ′) / dt) and the waveform state change for each can be visualized. For example, the program can be processed and viewed as an image.
[0023]
In determining the end point, first, an average value and a dispersion tendency of the ratio (A / A ′) and the differential value (d (A / A ′) / dt) within the averaging time are obtained. The averaging time is set in a period before the end point of the plasma processing. The starting point of the averaging time is preferably set not at the start of the plasma processing but at the time when the plasma is stabilized.
[0024]
Next, an initial fluctuation range (predetermined region) in the graph is calculated from a predetermined value described later calculated from the dispersion tendency information of the ratio (A / A ′) and the differential value of the ratio (d (A / A ′) / dt). Ask. As shown in FIG. 6, the fluctuation range (predetermined region) includes a predetermined value of the ratio (A / A ′) calculated from the information on the dispersion tendency of the specified period (averaging time) and a differential value (d (A Square root of the sum of squares with a predetermined value of / A ′) / dt)₁ In practice, the square root of the sum of squares of a predetermined value of the ratio (A / A ′) and the differential value of the ratio (d (A / A ′) / dt) calculated from the information of the dispersion tendency {√ [(predetermined ratio)² + (Predetermined value of slope (differential value of ratio))² ]} (C) is obtained, the radius is the value, and the average value is the origin O, which is a circular range indicated by the dotted line P. Therefore, before the end point, many are distributed in this circle P, and when the change near the end point starts, it gradually moves away from this circle. Here, the maximum value of the difference between the ratio and the average value of the ratio is used as the predetermined value of the ratio (A / A ′) calculated from the dispersion tendency information, and the differential value of the ratio (d (A / A ′)) The maximum value of the difference between the differential value of the ratio and the average value of the differential value of the ratio can be used as the predetermined value of / dt).
[0025]
Based on the above fact, a time when the fluctuation range is exceeded is determined as a slope start. However, since the slope start cannot be accurately determined in the vicinity of the fluctuation range, for example, a position (value) for determining the end point outside the fluctuation range on the graph is set, and a distance L1 from the position to the origin is set. The slope start is determined by comparing with the radius L2 of the circle in the fluctuation range. That is, the ratio (A / A ′) and the differential value of the ratio (d (A / A ′) / dt) are obtained using the luminescence detection information of the first and second active species during the processing after the averaging time. Then, the distance L1 from the position on the graph to the origin is obtained, and this distance is compared with the radius L2 of the fluctuation range obtained from the light emission detection information within the averaging time.
[0026]
Specifically, the difference between the ratio (A / A ′) and the differential value (d (A / A ′) / dt) corresponding to the position where the end point is determined, and the average value (origin). Square root of sum of squares {√ [(average of ratio-ratio)² + (Slope-Average value of slope)² ]} (D) is obtained, and the slope start is determined when the ratio (D / C) to the square root (C) of the square sum is greater than a preset threshold value.
[0027]
On the other hand, the slope end is determined by the differential value of the ratio (d (A / A ′) / dt), that is, when the inclination approaches the horizontal axis again (approaches the fluctuation range of the inclination). That is, the inclination for determining the end point is compared with a predetermined value obtained from the dispersion tendency to obtain ((inclination−average value of inclination) / predetermined value), and this value is a preset threshold value. The smaller point is determined as the slope end. The predetermined value obtained from the dispersion tendency is calculated as described above.
[0028]
When the slope end is used for end point detection, the ratio (A / A ′) obtained using the luminescence detection information of the first and second active species during the processing after the specified period and the differential value of the ratio (d ( When the position in the graph of A / A ′) / dt) deviates from the origin and again approaches the horizontal axis, it is determined as the end point of the plasma processing.
[0029]
In this case, similarly to the case where the slope start is used for end point detection, the fluctuation range is set, and the ratio and the ratio differentiation are performed using the light emission detection information of the first and second active species during the processing after the designated period. A value may be obtained, and when the obtained ratio and the position in the graph of the differential value of the ratio are within the fluctuation range (predetermined region), it may be determined as another end point of the plasma processing. Note that the setting of the fluctuation range is the same as when the slope start is used for end point detection.
[0030]
Even when the slope end is used for end point detection, as in the case where the slope start is used for end point detection, a new origin and a variation range may be set sequentially to perform end point detection a plurality of times.
[0031]
In this way, in the determination of the end point, the conventional value using two wavelengths is used by using the predetermined value obtained from the information within the averaging time of the plasma processing for the end point determination without directly comparing with the threshold value. Drift fluctuations that cannot be removed by the end point detection method can be sufficiently removed. Therefore, the end point can be accurately detected even when the S / N (signal noise) is poor. Note that the actual plasma processing end point may be the slope start position or the slope end position. This selection is appropriately made according to the purpose and conditions of the plasma treatment.
[0032]
In the end point detection method of the present invention, as shown in FIG. 7A, even when the waveform of the temporal change in the ratio of the emission intensity is a shape that rises in the vicinity of the end point, it is shown in FIG. As described above, the slope start and the slope end can be determined in the same manner as described above. In this case, it can be recognized that the resulting curve jumps greatly into the first quadrant of coordinates.
[0033]
When holes are formed by etching two different types of films, for example, as shown in FIG.₂ Film 5 and Si_Three N_Four When etching the film sequentially formed with the film 6, as shown in FIG. 9A, the waveform of the temporal change in the ratio of the emission intensity increases at the first time, that is, near the end point of the first stage. It becomes a shape that descends for the second time, that is, near the end point of the second stage. In this case, as shown in FIG. 9B, after the first end point is detected as described above, the obtained ratio (A / A ′) and the differential value of the ratio (d (A / The point where the position in the graph of A ′) / dt) intersects the horizontal axis of the graph again is the new origin O₂ From the predetermined value of the ratio (A / A ′) calculated from the dispersion tendency information and the predetermined value of the differential value of the ratio (d (A / A ′) / dt), another predetermined region (variation range) R₂ And the ratio (A / A ′) and the differential value of the ratio (d (A / A ′) / dt) are obtained using the luminescence detection information of the first and second active species, and the obtained ratio ( A / A ′) and the differential value of the ratio (d (A / A ′) / dt) in the graph are in other predetermined regions R₂ Is determined as the second end point (slope start) of the plasma processing. In FIG. 9B, O₁ Is the origin for the first end point detection, R₁ Is a fluctuation range in the first end point detection. The second slope start and slope end determination method is performed in the same manner as described above.
[0034]
As a result, Si_Three N_Four The final point of etching to form holes in the film 6, followed by SiO 2₂ The final point of etching for forming holes in the film 5 can be recognized.
The technique of repeating the origin start and the slope start and slope end determination is repeated as shown in FIG.₂ The present invention can also be applied when holes 8a to 8c having different aspect ratios are formed in the film 5. Also in this case, as shown in FIG. 11A and FIG. 11B, by performing the origin movement in the same manner as described above, a plurality of slope start and slope end determinations can be made, The end point of the etching process at each step can be accurately detected. In FIG. 11B, O₁ Is the origin when detecting the end point of the first stage, that is, the first stage, O₂ Is the origin at the second time, that is, the end point detection of the stage, O_Three Indicates the origin at the third time, that is, when the third stage end point is detected.₁ Is the fluctuation range when the first endpoint is detected, R₂ Is the fluctuation range at the time of the second end point detection, R_Three Indicates the variation range at the time of the third end point detection.
[0035]
Next, examples carried out to clarify the effect of the plasma processing end point detection method of the present invention will be described.
FIG. 12 is a diagram schematically showing a configuration of a plasma processing apparatus, for example, a plasma etching apparatus, provided with an end point detection apparatus according to the present invention. The plasma processing apparatus 10 includes, for example, a processing chamber 11 made of a conductive material such as aluminum, and a lower electrode 12 that also serves as a mounting table that is placed in the processing chamber 11 and mounts a semiconductor wafer W that is a target object. And an upper electrode 13 disposed above the lower electrode 12 so as to be spaced apart from the lower electrode 12, and a plasma generation region is defined between these electrodes.
[0036]
In the upper part of the processing chamber 11, a processing gas such as CF_Four A gas introduction pipe 14 for introducing a fluorocarbon-based etching gas such as the like is connected. An exhaust pipe 15 for discharging generated gas is connected to the side wall of the processing chamber 11. The lower electrode 12 is grounded and is always kept at the ground potential. Further, the upper electrode 13 is connected to a high frequency power source 16, and a high frequency voltage is applied from the high frequency power source 16 to discharge between the lower electrode 12 and the upper electrode 13, thereby activating the etching gas and radical species. Then, plasma P composed of active species such as ion species is generated in the plasma generation region.
[0037]
A monitoring window 17 made of a transparent material such as quartz glass is attached to the side wall of the processing chamber 11. The emission spectrum of the plasma P is transmitted through the window 17, and the progress of etching is analyzed by analyzing the transmitted light. Monitor the situation. A lens 21 for condensing the transmitted light is disposed outside the window 17, and a photodetector 22 that detects and photoelectrically converts the light collected by the lens 21 at the subsequent stage of the lens 21. It is arranged. The photodetector 22 includes, for example, a pair of interference filters or spectrometers and a pair of photomultipliers or photodiodes, and two light beams having specific wavelengths are respectively separated by the interference filters or spectrometers. Thereafter, the light having a specific wavelength that has been dispersed is photoelectrically converted and transmitted as an electric signal representing a change in emission intensity with respect to time. An end point of etching is detected by an arithmetic unit 30 described later based on two electrical signals corresponding to changes in emission intensity with respect to time transmitted from the photodetector 22, and a control signal is sent to the control unit 40 when the end point is detected. The plasma processing apparatus 10 is controlled via the control device 40, that is, the etching of the high-frequency power source 16 is stopped and the etching is terminated.
[0038]
Here, the position of the lens 21 can be appropriately moved in the vertical direction by a lens moving means (not shown). For example, when a film formed on a semiconductor substrate is subjected to plasma etching for hole formation, when detecting an emission spectrum having a specific wavelength, the light reflected from the upper surface of the film and the lower surface of the film (semiconductor substrate and film) If the light reflected from the interface) enters the photodetector 22, there is a possibility that the emission intensity of the emission spectrum cannot be detected accurately. In this embodiment, the focal position of the lens can be shifted by the lens moving means so that such interference light does not enter.
[0039]
Next, the arithmetic unit 30 that performs end point detection according to the present invention will be described. As shown in FIG. 13, the arithmetic unit 30 receives the input signal from the light detector 22, that is, the signal indicating the change in the emission intensity of the first specific wavelength light and the emission intensity of the second specific wavelength light. Each information of the signal representing the change is calculated to extract the ratio of the change in both emission intensities and the differential value (slope) of this ratio, that is, the element extractor 31 to calculate and the ratio of the emission intensities (A / A ′) is taken on the horizontal axis and the differential value of the ratio (d (A / A ′) / dt), that is, the slope is taken on the vertical axis, the average value of the ratio (A / A ′) and the differential value of the ratio (d ( A graphing device that creates a graph as shown in FIG. 5 by plotting the ratio and the change of the differential value of the ratio with respect to time on a graph whose origin is the intersection with the average value of A / A ′) / dt) 32, a slope start determination unit 33 for determining a slope start from the created graph, and a created graph Is constituted by a slope end determination unit 34 determines Luo slope end. In this preferred embodiment, an origin mover 35 that moves the origin in the graph created during the processing of the laminated film or the film on the substrate having steps is provided on the output side of the determiners 33 and 34. It has been. These origin movers are driven only when the origin is moved as described above, and instructs the element extractor 31 to repeat the same operation.
[0040]
More specifically, the element extractor 31 performs the following arithmetic processing.
(1) Based on the light emission detection information of the first and second active species within the specified period (average time), the approximate primary relational expression A and second of the first active species in the relationship between the emission intensity and time The approximate primary relational expression B of the active species is obtained.
(2) Using the approximate primary relational expression A of the first active species and the approximate primary relational expression B of the second active species, that is, the time course component of the approximate primary relational expression B of the second active species is the first Substituting into the time lapse component of the approximate primary relational expression A of the active species, a pseudo approximate primary relational expression A ′ of the first active species is obtained.
(3) From the approximate primary relational expression A of the first active species and the pseudo approximate primary relational expression A ′ of the first active species, the ratio (A / A ′) between them and the differential value of the ratio (d (A / A ') / Dt).
(4) The average value and the variance value of the ratio (A / A ′) and the differential value (d (A / A ′) / dt) within the averaging time are obtained.
[0041]
The graph generator 32, the slope start determination unit 33, and the slope end determination unit 34 perform the processes as described above, respectively, to perform graph creation, slope start determination, and slope end determination, respectively.
[0042]
The determination results of the slope start determination and the slope end determination are transmitted to the control device 40, and the etching process is controlled by controlling the high frequency power supply 16 and the like via the control device 40 based on the signal of the determination result. Note that the determination result of the slope end determination is sent to the origin moving device 35 as necessary, and a signal resulting from the origin movement is also sent to the control device 40.
[0043]
SiO formed on a silicon substrate using a plasma processing apparatus (plasma etching apparatus) having the above-described configuration₂ CF film_Four A case where etching is actually performed using a gas will be described. Here, as the first active species, SiO₂ CF, an etchant ionized as a second active species using CO molecules generated by etching₂ ^* Use molecules. CO molecules are analyzed with a spectroscope and CF₂ ^* The molecules were filtered with an optofilter. Note that the emission wavelength of CO molecules is about 219 nm, and CF₂ ^* The used emission wavelength of the molecule is about 260 nm.
[0044]
First, an optical signal detected and filtered by the photodetector 22 is converted into an electric signal by a photoelectric converter (not shown), and the electric signal is amplified by a preamplifier (not shown).
[0045]
Next, the gain of the preamplifier for the electrical signal of the CO molecule is adjusted to₂ ^* Amplifies the electrical signal of the CO molecule to the same level as the molecule. At this stage, both electrical signals are analog signals.
[0046]
Next, a noise component having a frequency more than twice the sampling cycle, in this case, 20 Hz or more, is cut with a filter. Thereafter, both electric signals are sampled at a cycle of 0.1 second and digitized by an A / D converter.
[0047]
Then, both digitized electrical signals are smoothed by a dynamic averaging method. This has the effect of a so-called low-pass filter, and a relatively smooth signal free from high-frequency noise (that is, random noise) can be obtained.
[0048]
Next, in the element extractor 31, the approximate linear relations A and CF of the CO molecule₂ ^* An approximate linear relational expression B of the molecule is obtained. FIG. 14 shows a straight line indicating the temporal change of these emission intensities and the primary relational expression. Also, using these, CF₂ ^* A pseudo approximate primary relation A ′ of the CO molecule is obtained from the approximate primary relation B of the molecule B. FIG. 15 shows temporal changes in emission intensity corresponding to these relational expressions A, B, and A ′.
[0049]
Both the approximate primary relational expression A and the pseudo-approximation primary relational expression A ′ represent the light emission amount (intensity) of CO molecules, and the ratio (A / A ′) and the differential value of the ratio (d (A / A ') / Dt). The time change of the ratio (A / A ′) is shown in FIG. For comparison, FIG. 16 shows the approximate primary relations A and CF of the CO molecule.₂ ^* The time change of the simple ratio (A / B) of the approximate primary relational expression B of the molecule is also shown. As is apparent from FIG. 16, according to the method of the present invention, the waveform is flat before and after the end point, so that the end point can be detected accurately. On the other hand, in the case of a simple ratio (A / B), the waveform is not stable, and the end point cannot be accurately detected due to drift fluctuation.
[0050]
Next, the ratio and ratio calculated from information within the specified period (averaging time) with the ratio (A / A ′) as the horizontal axis and the differential value of the ratio (d (A / A ′) / dt) as the vertical axis Create a graph with the average of the differential values of as the origin. When the ratio and the differential value of the ratio are plotted on the graph, it is as shown in FIG.
[0051]
Further, on the graph, from a predetermined value of the ratio (A / A ′) and the differential value (d (A / A ′) / dt) calculated from the information of the dispersion tendency obtained within the averaging time, Find the initial range of variation in the graph.
[0052]
Next, a position (value) for determining the end point outside the fluctuation range on the graph is set, and the slope start determination unit 33 compares the distance from the position to the origin with the radius of the circle in the fluctuation range, thereby determining the slope. Judge the start. That is, CO molecules and CF during processing after the averaging time₂ ^* The ratio (A / A ′) and the differential value of the ratio (d (A / A ′) / dt) are obtained using the light emission detection information, and the distance from the position on the graph to the origin is obtained, and this distance and averaged The radius of the fluctuation range obtained from the emission detection information within the time is compared.
[0053]
Specifically, the square root (D) of the sum of squares of the difference between the ratio corresponding to the position for determining the end point and the differential value of the ratio and the average value (origin) is obtained, and the radius of the circle (C ) And a ratio (D / C) greater than a preset threshold value is determined as a slope start.
[0054]
On the other hand, the slope end is determined when the differential value of the ratio, that is, the inclination approaches the horizontal axis again. That is, the inclination for determining the end point is compared with a predetermined value based on the dispersion tendency to obtain ((inclination−average value of inclination) / predetermined value), and this value is smaller than a preset threshold value. Then, the slope end determination unit 34 determines the slope end.
[0055]
According to the end point detection method of the present invention described above, the amount of emitted light (aperture ratio) can be improved by a factor of 3 (from 1% to 3%) as compared with the prior art.
In the above embodiment, CO molecules are used as the first active species, and CF is used as the second active species.₂ ^* Although the case where a molecule is used is described, the present invention can be applied to the case where another active species is used as the first active species and the second active species.
[0056]
In the above embodiment, the case where the plasma treatment is etching is described. However, the present invention also relates to the case where the plasma treatment is a treatment using plasma such as CVD (Chemical Vapor Deposition) or PVD (Physical Vapor Deposition). The same can be applied.
[0057]
In the above embodiment, the empirical formula that approximates the change with time of the emission intensity of the emission spectrum at the specific wavelength of the first active species and the second active species was used as an approximate primary relational expression. It will be obvious to those skilled in the art that the present invention is not limited to this. For example, when the change with respect to time changes along a part of an ellipse or a hyperbola, these quadratic relational expressions can be used. That is, in the present invention, it is preferable to use an empirical formula that most closely approximates the change in emission intensity in order to perform measurement with higher accuracy.
[0058]
As the first active species and the second active species, the active species whose emission intensity becomes weak as shown in FIG. 4 at the end point of the plasma treatment after the designated period, and shown in FIG. In order to increase sensitivity, it is preferable to use an active species that is so strong.
[0059]
As described above, according to the present invention, since the predetermined value obtained from the information of the dispersion tendency within the designated period of the plasma processing is used for determining the end point without performing a direct comparison with the threshold value. It is possible to sufficiently remove drift fluctuations that cannot be removed by the conventional end point detection method using the. Therefore, the end point can be accurately detected even when the S / N is poor. As described above, since the end of the plasma can be detected accurately, the amount of emitted light (aperture ratio) can be improved three times as compared with the conventional case.
[Brief description of the drawings]
FIG. 1 is a diagram for explaining noise fluctuation of plasma emission intensity.
FIG. 2 is a diagram for explaining drift fluctuation in plasma emission intensity.
FIG. 3 is a diagram for explaining an approximate primary relational expression of the first and second active species and a pseudo approximate primary relational expression of the first active species.
FIG. 4 is a characteristic diagram showing a temporal change in emission intensity (ratio) and slope (differential value).
FIG. 5 is a graph in which the values shown in FIG. 4 are plotted when the emission intensity (ratio) is on the horizontal axis and the slope (differential value) is on the vertical axis.
FIG. 6 is a diagram for explaining a slope start.
FIG. 7A is a characteristic diagram showing another example of temporal change in light emission intensity (ratio) and slope (differential value), and FIG. 7B is a graph with the light emission intensity (ratio) as a horizontal axis and the slope (differential) The value is plotted on the vertical axis, and the values shown in (A) are plotted.
FIG. 8 shows an example of an object to be etched.
9A is a characteristic diagram similar to FIG. 7A, showing another example of temporal changes in emission intensity (ratio) and slope (differential value), and FIG. 9B is emission intensity. A graph in which the values shown in (A) are plotted, with (ratio) on the horizontal axis and slope (differential value) on the vertical axis.
FIG. 10 is a cross-sectional view showing another example of an object to be etched.
11A is a characteristic diagram showing another example of the temporal change of the emission intensity (ratio) and the slope (differential value), and FIG. 11B is a slope with the emission intensity (ratio) as the horizontal axis. The graph which plotted the value shown to (A) on the vertical axis | shaft (differential value).
FIG. 12 is a diagram schematically showing a plasma etching apparatus having an end point detection apparatus according to the present invention.
13 is a diagram for explaining an end point detection apparatus in the plasma etching apparatus shown in FIG. 12;
FIG. 14: CO molecule and CF₂ ^* The characteristic view which shows the time change of the emitted light intensity.
FIG. 15: CO molecule and CF after operation₂ ^* The characteristic view which shows the time change of the emitted light intensity.
FIG. 16: CO molecule and CF₂ ^* The characteristic view which shows the time change of change rate (ratio) of the emitted light intensity.
FIG. 17: CO molecule and CF₂ ^* A graph in which the change rate (ratio) of the emission intensity is plotted on the horizontal axis and the slope (differential value) is plotted on the vertical axis.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 10 ... Plasma processing apparatus, 22 ... Photo detector, 30 ... Arithmetic unit, 31 ... Element extractor, 32 ... Graphizer, 33 ... Slope start determination device, 34 ... Slope end determination device, 35 ... Origin mover, 40 …Control device.

Claims

When performing processing using plasma on the object to be processed, the emission intensity having specific wavelengths of the first and second active species is detected during and after the specified period during the plasma processing, and this emission detection is performed. Outputting the information;
Based on the light emission detection information, in the relationship between the light emission intensity and time, the approximate relational expression A of the first active species, that is, Y ₁ = a ₁ × X + b ₁ , and the approximate relational expression B of the second active species, That is, Y ₂ = a ₂ × X + b ₂ (Here, Y ₁ and Y ₂ represent approximate values of the light emission amounts of the first and second active species, respectively, X represents an elapsed time, and a ₁ , a ₂ represents a primary coefficient, and b ₁ and b ₂ represent Y intercepts),
Using the approximate relational expression A of the first active species and the approximate relational expression B of the second active species, a pseudo approximate relational expression A ′ of the first active species, that is, Y ₁ = (a ₁ / a ₂ ) x (Y ₂ −b ₂ ) + b ₁ (where Y ₂ is an actual measurement value)
From the approximate relational expression A of the first active species and the pseudo approximate relational expression A ′ of the first active species, the ratio between them, that is, (Y ₁ in Formula A) / (Y ₁ in Formula A ′) and Obtaining a differential value of the ratio, that is, d ((Y ₁ in Formula A) / (Y ₁ in Formula A ′)) / dt;
The ratio within the specified period (Y ₁ in Formula A) / (Y ₁ in Formula A ′) and the differential value d ((Y ₁ in Formula A) / (Y ₁ in Formula A ′)) / dt Obtaining an average value of each of
A predetermined area is set in a graph with the ratio on the horizontal axis, the differential value of the ratio on the vertical axis, and the origin of the intersection of the average value of the ratio and the average value of the differential value of the ratio within the specified period. And a process of
Using the luminescence detection information of the first and second active species during the processing after the specified period, a ratio, that is, (Y ₁ in Formula A) / (Y ₁ in Formula A ′) and a differential value of the ratio, That is, d ((Y ₁ in the expression A) / (Y ₁ in the expression A ′)) / dt is obtained, and when the position of the obtained ratio and the differential value of the ratio in the graph deviates from the predetermined region, plasma is obtained. Determining the end point of the processing, and a method for detecting the end point of the plasma processing.

The ratio within the specified period (Y ₁ in Formula A) / (Y ₁ in Formula A ′) and the differential value d ((Y ₁ in Formula A) / (Y ₁ in Formula A ′)) / dt The method for detecting the end point of plasma processing according to claim 1, further comprising the step of obtaining each of the dispersion tendencies.

The predetermined area includes a predetermined value of the ratio (Y ₁ in Formula A) / (Y ₁ in Formula A ′) calculated from information on the dispersion tendency in the specified period and a differential value d of the ratio (in Formula A Y _₁₎ / (Y ₁ in formula a ')) / dt of the end point detection method of the plasma treatment according to claim 2, characterized in that it is set by the square root r ₁ of the sum of the squares of the predetermined value.

Using the maximum value of the difference between the ratio and the average value of the ratio as the predetermined value of the ratio (Y ₁ in Formula A) / (Y ₁ in Formula A ′) calculated from the information on the dispersion tendency, The maximum value of the difference between the differential value of the ratio and the average value of the differential value of the ratio as a predetermined value of the differential value d of the ratio d ((Y ₁ in Formula A) / (Y ₁ in Formula A ′)) / dt. The end point detection method for plasma processing according to claim 2, wherein the end point detection method is used.

In the step of determining as the end point of the plasma processing, a predetermined value of the ratio (Y ₁ in the expression A) / (Y ₁ in the expression A ′) calculated from the information of the dispersion tendency in the specified period and the differentiation of the ratio The square root r _{1 of the} sum of squares of a predetermined value d ((Y ₁ in Formula A) / (Y ₁ in Formula A ′)) / dt and the ratio (Y ₁ in Formula A) / (Formula A Y ₁ ) and differential ratio d ((Y ₁ in Formula A) / (Y ₁ in Formula A ′)) / dt in the graph are compared with the distance r ₂ between the origin and r _3. The plasma processing end point detection method according to claim 2, further comprising a step of determining when ₂ is larger than r ₁ as an end point of the plasma processing.

In the step of determining as the end point of the plasma processing, a predetermined value of the ratio (Y ₁ in the expression A) / (Y ₁ in the expression A ′) calculated from the information of the dispersion tendency in the specified period and the differentiation of the ratio The square root r ₁ of the sum of squares of a predetermined value of the value d ((Y ₁ in Formula A) / (Y ₁ in Formula A ′)) / dt is calculated, and the values of the horizontal axis component and the vertical axis component of the coordinates in the graph There endpoint detection method for a plasma treatment according to claim 2, characterized in that it comprises a step of determining the endpoint of plasma processing when either becomes larger than r _1.

The graph of the obtained ratio (Y ₁ in Formula A) / (Y ₁ in Formula A ′) and differential value d ((Y ₁ in Formula A) / (Y ₁ in Formula A ′)) / dt The origin at the point where the position in the graph intersects the horizontal axis of the graph again, and the predetermined value of the ratio (Y ₁ in Formula A) / (Y ₁ in Formula A ′) and the differential value d of the ratio ((Y ₁ in Formula A) ) / (Y ₁ in Formula A ′), another predetermined region is set from a predetermined value of dt), and the ratio is calculated using the light emission detection information of the first and second active species during the processing after the specified period. (Y ₁ in Formula A) / (Y ₁ in Formula A ′) and the differential value d ((Y ₁ in Formula A) / (Y ₁ in Formula A ′)) / dt were determined, and the ratio ( formula Y ₁ in a) / (formula a 'Y ₁ in) and a ratio of differential values d ((Y ₁ in formula a) / (formula a' Definitive Y ₁₎₎ / dt claim 1 or 2 of the plasma processing position in the graph, characterized by further comprising the step of determining the other endpoint of the plasma treatment when deviating from the other predetermined area of the End point detection method.

When performing processing using plasma on the object to be processed, the light detection means sequentially detects the emission intensity at a specific wavelength of light emission from the first and second active species during the specified period during the plasma processing and thereafter. Detecting and outputting light emission detection information;
Based on the light emission detection information within the specified period, the approximate relational expression A of the first active species in the relationship between the light emission intensity and time, that is, Y ₁ = a ₁ × X + b ₁ , and the approximation of the second active species Relational expression B, that is, Y ₂ = a ₂ × X + b ₂ (where Y ₁ and Y ₂ represent the approximate values of the light emission amounts of the first and second active species, respectively, X represents the elapsed time, a ₁ and a ₂ represent primary coefficients, and b ₁ and b ₂ represent Y intercepts),
Using the approximate relational expression A of the first active species and the approximate relational expression B of the second active species, a pseudo approximate relational expression A ′ of the first active species, that is, Y ₁ = (a ₁ / a ₂ ) x (Y ₂ −b ₂ ) + b ₁ (where Y ₂ is an actual measurement value)
From the approximate relational expression A of the first active species and the pseudo approximate relational expression A ′ of the first active species, the ratio between them, that is, (Y ₁ in Formula A) / (Y ₁ in Formula A ′) and Obtaining a differential value of the ratio, that is, d ((Y ₁ in Formula A) / (Y ₁ in Formula A ′)) / dt;
The ratio within the specified period (Y ₁ in Formula A) / (Y ₁ in Formula A ′) and the differential value d ((Y ₁ in Formula A) / (Y ₁ in Formula A ′)) / dt Obtaining an average value of each of
The ratio is plotted on the horizontal axis, the differential value of the ratio is plotted on the vertical axis, and the intersection point between the average value of the ratio and the average value of the differential value of the ratio in the specified period is the graph after the specified period. The ratio obtained by using the luminescence detection information of the first and second active species during the processing of, i.e., (Y ₁ in Formula A) / (Y ₁ in Formula A ′) and the differential value of the ratio, , D ((Y ₁ in Formula A) / (Y ₁ in Formula A ′)) / dt is determined as the plasma processing end point when the position in the graph deviates from the origin and again approaches the horizontal axis. A method for detecting the end point of plasma processing.

The ratio within the specified period (Y ₁ in Formula A) / (Y ₁ in Formula A ′) and the differential value d ((Y ₁ in Formula A) / (Y ₁ in Formula A ′)) / dt The method for detecting an end point of plasma processing according to claim 8, further comprising a step of obtaining each of the dispersion tendencies.

As the first active species and the second active species, active species whose emission intensity becomes weak and active species whose emission intensity becomes weak at the end point of the plasma treatment after the specified period are used. The method for detecting an end point of plasma processing according to any one of claims 1 to 9.

A light detection means for detecting light emission detection information by detecting a change with time of light emission intensity at a specific wavelength of the first and second active species generated by the plasma when the object to be processed is processed using plasma;
Based on the light emission detection information within the specified period, the approximate relational expression A of the first active species in the relationship between the light emission intensity and the time, that is, Y ₁ = a ₁ × X + b ₁ , and the approximate relationship of the second active species Formula B, that is, Y ₂ = a ₂ × X + b ₂ (where Y ₁ and Y ₂ represent approximate values of the light emission amounts of the first and second active species, respectively, X represents elapsed time, a ₁ and a ₂ represent first order coefficients, and b ₁ and b ₂ represent Y intercepts), and the approximate relational expression A of the first active species and the approximate relational expression B of the second active species are used. Then, the pseudo approximate relational expression A ′ of the first active species, that is, Y ₁ = (a ₁ / a ₂ ) × (Y ₂ −b ₂ ) + b ₁ is obtained (Y ₂ here is an actual measurement value). From the approximate relational expression A of the first active species and the pseudo approximate relational expression A ′ of the first active species, the ratio between them, that is, (Y ₁ in Formula A) / (in Formula A ′ Y ₁ ) and a differential value of the ratio, that is, d ((Y ₁ in Formula A) / (Y ₁ in Formula A ′)) / dt, and the ratio within the specified period (Y ₁ in Formula A) / (Y ₁ in the formula A ′) and a differential value d ((Y ₁ in the formula A) / (Y ₁ in the formula A ′)) / dt calculating means for calculating the respective average values;
The ratio is plotted on the horizontal axis, and the differential value of the ratio is plotted on the vertical axis. The average value of the ratio (Y ₁ in Formula A) / (Y ₁ in Formula A ′) and the differential value d of the ratio ((in Formula A) Y ₁ ) / (graphing means for creating a graph with the origin at the intersection with the average value of Y ₁ )) / dt in formula A ′;
A predetermined region is set, and the ratio (Y ₁ in Formula A) / (Y ₁ in Formula A ′) and the ratio of the first and second active species during the processing after the specified period are used. The differential value d ((Y ₁ in Formula A) / (Y ₁ in Formula A ′)) / dt is obtained, and the ratio (Y ₁ in Formula A) / (Y ₁ in Formula A ′) and the derivative of the ratio are obtained. Determining means for determining when the position of the value d ((Y ₁ in Formula A) / (Y ₁ in Formula A ′)) / dt deviates from the predetermined region as the end point of the plasma processing. An apparatus for detecting the end point of plasma processing.

The calculating means further, the ratio (Y in the formula A ₁₎ in the specified time period / (wherein A 'differential value of and the ratio Y ₁ in) d ((Y ₁ in formula A) / (formula A' Y ₁ )) / dt in each of the distribution tendencies, and the determination means calculates the ratio (Y ₁ in Formula A) / (Y in Formula A ′) calculated from the information on the dispersion tendencies. ₁ ) and a differential value of the ratio d ((Y ₁ in the expression A) / (Y ₁ in the expression A ′)) / dt has a function of setting the predetermined area. The plasma processing end point detection apparatus according to claim 11.

2. The plasma processing end point detection method according to claim 1, wherein the first active species is a CO molecule.