JP2004101510A - Method of and apparatus for spectroscopic measurement using pulsed light - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To obtain a spectrum which does not include an interference pattern by analyzing transmitted time-series pulsed light or reflected time-series detection pulsed light, including pulses multiply reflected within a parallel flat sample, as incoherent light. <P>SOLUTION: The measuring apparatus is configured, such that the parallel flat sample 20 is irradiated with terahertz pulsed light, and detection pulsed light reflected from or transmitted through the sample 20 is detected. The apparatus separates and extracts each pulse from the detection pulsed light, including the pulses multiply reflected within the sample 20. The apparatus applies Fourier transformation to each separated waveform, sums their absolute values, and applies Fourier transformation on reference light to calculate a reference absolute value. The apparatus takes the ratio of the absolute value and the reference absolute value with respect to the detection pulse for incoherent analysis, to obtain a spectrum which does not contain interference patterns. <P>COPYRIGHT: (C)2004,JPO

Description

【数２】

で与えられる。
【００２５】
通常、試料を透過した光の振幅透過率ｔは、
【数３】

で表される。１行目の右辺第１項は直接透過した光を表し、第２項および第３項はそれぞれ、試料内部で１回多重反射した透過光と２回多重反射した透過光を表している（図６（ｂ）参照）。ここで、θは、
【数４】

で与えられる。ｃは真空中の光速、ｄは試料の厚さである。
【００２６】
また、試料で反射した光の振幅反射率ｒは、
【数５】

で表される。１行目の右辺第１項は試料の表面で反射した光を表し、第２項は裏面で反射した光を表している。第３項および第４項はそれぞれ、試料内部で１回多重反射した光と２回多重反射した光を表している（図６（ｂ）参照）。式（３）および式（５）は、各光の位相情報を保ったままそれぞれの振幅が足し合わされた場合の、振幅透過率および振幅反射率に対応している。すなわち、各光がコヒーレントとして取り扱われている。
【００２７】
これに対して、裏面反射および各多重反射の振幅の絶対値を取って位相情報を取り除き、その大きさのみを足し合わせると、各光がインコヒーレントな場合に対応する。したがって、式（３）および式（５）をそれぞれ、
【数６】

【数７】

とすると、各光がインコヒーレントな場合の振幅透過率および振幅反射率に対応する。
【００２８】
式（３）と式（５）、および式（６）と式（７）は無限回までの多重反射をすべて含んでいる。コヒーレントな場合の式（３）と、インコヒーレントな場合の数式（６）を用いて計算したシリコンウェハの振幅透過率を図７に示す。ここから明らかなように、コヒーレントな場合に見られる干渉パターンが、インコヒーレントな場合では、まったく見られない。このことから、測定された各光をインコヒーレントとして解析することができれば、多重反射を考慮に入れても、干渉パターンを含まないスペクトルを測定することができる。
【００２９】
通常、テラヘルツ時間領域分光法で得られた測定波形を解析する場合、裏面反射光および各多重反射光はコヒーレントとして解析される。しかし、テラヘルツ時間領域分光法ではフェムト秒パルス光源が使用される。そこで、平行平板試料からの時系列透過パルス列あるいは時系列反射パルス列に含まれる時系列波形のそれぞれのパルスを抽出することにより、検出パルス光に対してインコヒーレントな解析処理を実行することが可能になる。
【００３０】
図３で示される透過時系列波形ｆ（ｔ）を、図８（ａ）〜（ｄ）で示される個々の波の足し合わせ、あるいは図４で示される反射時系列波形ｆ（ｔ）を、図９（ａ）〜（ｄ）で示される個々の波の足し合わせであると考える。すなわち、ウェハを透過する透過パルスは図８（ａ）〜（ｄ）の個々パルスを含む時系列パルスである。一方、ウェハを反射する反射パルスは図９（ａ）〜（ｄ）の個々パルスを含む時系列パルスである。
【００３１】
すなわち、透過パルスは、ウエハを透過するパルスを表す図８（ａ）の波をａ（ｔ）、１回目の多重反射を経て透過するパルスを表す図８（ｂ）の波をｂ（ｔ）、２回目の多重反射パルスを表す図８（ｃ）の波をｃ（ｔ）、３回目の多重反射パルスを表す図８（ｄ）の波をｄ（ｔ）とする。一方、反射パルスはウェハ表面での反射パルスを表す図９（ａ）の波をａ（ｔ）、裏面での反射パルスを表す図９（ｂ）の波をｂ（ｔ）、１回目の多重反射パルスを表す図９（ｃ）の波をｃ（ｔ）、２回目の多重反射パルスを表す図９（ｄ）の波をｄ（ｔ）とする。透過時系列パルスまたは反射時系列パルスを示すｆ（ｔ）は、
【数８】

で与えられる。
【００３２】
またａ（ｔ）、ｂ（ｔ）、ｃ（ｔ）、ｄ（ｔ）の各パルスのフーリエ変換をそれぞれ、Ａ（ω）、Ｂ（ω）、Ｃ（ω）、Ｄ（ω）とする。このとき、ｆ（ｔ）のフーリエ変換Ｆ（ω）は、
【数９】

とすることができる。これとリファレンス光のフーリエ変換との比をとったものは、無限回の多重反射を含む振幅透過率（式（３））または振幅反射率（式（５））を、有限回の多重反射で打ち切ったものに対応する。なお、リファレンス光は、透過型では、試料を取り除いた状態でフェムト秒パルス光源１からテラヘルツパルス光を検出器１１へ照射したときに検出される光の波形である。一方、反射型では、計測対象試料に代えて反射率１００％の基準試料を配置してフェムト秒パルス光源１からテラヘルツパルス光を検出器１１へ照射したときに検出される光の波形である。
【００３３】
そこで、各波をフーリエ変換し、その絶対値をとって足し合わせる場合を考える。このときの各波の足し合わせＦｉｎｃｏｈ（ω）は、
【数１０】

であり、これとリファレンス光のフーリエ変換の絶対値との比をとると、インコヒーレントな場合の振幅反射率（式（６））または振幅反射率（数（７））を、有限回の多重反射で打ち切ったものに対応する。すなわち、測定結果の各パルスを個別にフーリエ変換し、個々の振幅の絶対値を足し合わせることで、各パルスをインコヒーレントとして解析することができる。
【００３４】
以上のような解析手法は、光源がパルス光であるときに可能になる。例えば、従来のフーリエ変換赤外分光法では、通常、光源として連続光が用いられている。この場合、各多重反射パルスを個々に分離することはできない。すなわち、コヒーレントとして解析すべきか、インコヒーレントとして解析すべきかは光源のコヒーレンス長により一意的に決定される。したがって、本発明による分光計測方法および装置は、テラヘルツパルス光を用いたものに限らず、その他のパルス光を用いた分光計測方法および装置に適用することができる。
【００３５】
ここで、本実施の形態で採用されている、計測時系列波形Ｅｓａｍ（ｔ）に基づいて被測定物２０のスペクトルを解析する手順を図１０のフローチャートを参照して説明する。
発生器７と検出器１１との間の光路上に被測定物２０を配置しない状態で、テラヘルツパルス光の電場強度の基準時系列波形Ｅｒｅｆ（ｔ）を予め計測しておく（ステップＳ１）。計測結果はＡ／Ｄ変換器２２でデジタルデータに変換され、メモリ２３Ａに記憶される。制御・演算処理部２３はメモリ２３Ａに記憶した基準時系列波形Ｅｒｅｆ（ｔ）を読み出してフーリエ変換を実行する。さらに、フーリエ変換した結果の絶対値（基準絶対値と呼ぶ）を算出して記憶する（ステップＳ２）。
【００３６】
発生器７と検出器１１との間の光路上に被測定物２０を配置した状態で、テラヘルツパルス光の電場強度の計測時系列波形Ｅｓａｍ（ｔ）を計測する（ステップＳ３）。計測結果はＡ／Ｄ変換器２２でデジタルデータに変換され、メモリ２３Ａに記憶される。この計測時系列波形Ｅｓａｍ（ｔ）は概念図として図３に示すような時系列波形である。制御・演算処理部２３はメモリ２３Ａに記憶した計測時系列波形Ｅｓａｍ（ｔ）を読み出し、計測時系列波形Ｅｓａｍ（ｔ）を構成する図８（ａ）〜（ｄ）に示すパルス波形をそれぞれ分離抽出する（ステップＳ４）。分離抽出した各パルス波形に対してフーリエ変換を順次実行してそれらの絶対値を算出し、その総和（式（１０）に対応する総和値）を算出する（ステップＳ５）。この総和値と基準絶対値との比を取る（ステップＳ６）。これにより、各パルスをインコヒーレントとして解析することができる。
【００３７】
第１の実施の形態の分光計測装置では次の点も考慮している。
測定系には平板試料による多重反射だけでなく、光源もしくは検出器による多重反射が存在する場合がある。テラヘルツ時間領域分光法で使用するテラヘルツパルス光源（図１の光発生器７）は、アンテナ基板（ガリウム砒素基板）でテラヘルツパルス光を生成し、テラヘルツパルス光検出器（図１の光検出器１１）は、アンテナ基板（ガリウム砒素基板）でテラヘルツパルス光を検出する。したがって、光源７および検出器１１に用いているアンテナ基板（ガリウム砒素基板）の厚さと屈折率、および測定する平板試料２０の厚さと屈折率の関係から、試料２０内部での多重反射を経た透過光もしくは反射光が、アンテナ基板７，１１による多重反射光と同じタイミングで測定される場合がある。この場合、試料２０の多重反射とアンテナ基板７，１１の多重反射を分離することができず、検出パルス波形をインコーヒレントとして解析する上記手法を採用することができない。
【００３８】
そこで、上記解析手法を用いるためには、アンテナ基板７，１１と平板試料２０の厚さおよび屈折率の関係から、両者の多重反射が分離できるように、アンテナ基板７，１１および平板試料２０の厚さを最適化する必要がある。例えば、厚さ１０００μｍのシリコンウェハを測定する場合を考える。
【００３９】
テラヘルツパルス光源７および検出器１１のアンテナ基板として用いているガリウム砒素の屈折率は約３．６であり、シリコンの屈折率は約３．４である。したがって、アンテナ基板７，１１の厚さが９４５μｍ程度だと、シリコンウェハ２０の裏面反射光および多重反射光とアンテナ基板７，１１での多重反射光が、ほぼ同じタイミングで測定される。そこで、例えばアンテナ基板７，１１の厚さを測定試料２０の４分の１である２３５μｍ程度にしたとする。この場合、アンテナ基板７，１１による４回目の多重反射が、シリコンウェハ２０の裏面反射とほぼ同じタイミングで測定される。
【００４０】
しかし、アンテナ基板７，１１による４回目の多重反射は十分減衰し、シリコンウェハ２０での裏面反射に比べて弱いため、無視することができる。また、テラヘルツパルスの幅は約１ｐｓであり、この間に光は約３００μｍ進む。この距離は、アンテナ基板７，１１を往復する光路長にアンテナ基板７，１１の屈折率を掛けたもの（約１７００μｍ）よりも、十分短い。したがって、各パルス光は空間的に離れており、アンテナ基板７，１１の多重反射と平行平板試料２０の多重反射を分離して解析することができる。
【００４１】
−第２の実施の形態−
これまで説明してきたインコヒーレントな解析手法は、テラヘルツ光発生器のアンテナ基板７やテラヘルツ光検出器のアンテナ基板１１の厚さを変更することによりアンテナ基板からの多重反射を予め除去して、平行平板試料２０の多重反射のみを取り出して解析するものである。しかし、アンテナ基板７，１１の厚さを変更することは、平行平板試料２０の厚さが変わる度に行わなければならないので、手間も費用もかかる。なお、テラヘルツ光発生器やテラヘルツ光検出器では、アンテナ基板７，１１以外にも多重反射をもたらす要因はあるが、アンテナ基板７，１１が最大の要因なので、これらに代表させる。
【００４２】
以下、アンテナ基板７，１１の厚さを変更することなく、アンテナ基板７，１１による多重反射があってもこの影響を除去し、平行平板試料２０内部での多重反射を経た各パルスを分離する手法について説明する。
【００４３】
図１５は、テラヘルツ時間領域分光法により得られる透過時系列波形の概念図である。以下、透過時系列波形の縦軸が振幅、横軸が時間を表す。図１５（ａ）は、アンテナ基板７，１１による多重反射を全く含まない理想的な透過時系列波形Ａである。透過時系列波形Ａにおいて、一番左のパルス▲１▼は、シリコンウェハ２０の内部での反射なしに透過したもの、パルス▲２▼は、ウェハ２０の内部で１回多重反射して透過したもの、パルス▲３▼は、ウェハ２０の内部で２回多重反射して透過したもの、パルス▲４▼は、ウェハ２０の内部で３回多重反射して透過したものである。このような理想的な場合は、時間領域で各パルスを分離することは容易であり、直ちに前述したインコヒーレントな解析手法が適用できる。
【００４４】
図１５（ｂ）は、シリコンウェハ２０を配設しないで得られた透過時系列波形Ｂ、すなわち基準時系列波形Ｂである。この基準時系列波形Ｂには、直接にテラヘルツ光発生器からテラヘルツ光検出器へ入射したパルス▲５▼と、アンテナ基板７，１１による多重反射パルス▲６▼および▲７▼が現れる。
【００４５】
図１５（ｃ）は、シリコンウェハ２０を配設して得られた現実的な透過時系列波形Ｃ、すなわち計測時系列波形Ｃである。これは、図１５（ｂ）の基準時系列波形Ｂに図１５（ａ）の理想的な透過時系列波形Ａが重畳したものである。直接この計測時系列波形Ｃから透過時系列波形Ａの各パルスを分離することは極めて困難である。従って、透過時系列波形Ａに現れた多重反射の影響と基準時系列波形Ｂに現れた多重反射の影響を分離するためには、前述したように、アンテナ基板７，１１の厚さを変更する必要があった。
【００４６】
本実施の形態では、デコンボリューション（ｄｅｃｏｎｖｏｌｕｔｉｏｎ）を利用することにより、基準時系列波形Ｂと計測時系列波形Ｃの両方に含まれる信号成分を除去する手法を採っている。この信号成分は、具体的には、アンテナ基板７，１１の多重反射によるものである。
今、透過時系列波形Ａを入力信号ｈ（ｔ）、基準時系列波形Ｂを装置関数ｆ（ｔ）、計測時系列波形Ｃを出力信号ｇ（ｔ）と考えると、基準時系列波形Ｂと計測時系列波形Ｃは、実際に測定されているので、装置関数ｆ（ｔ）と出力信号ｇ（ｔ）は既知である。デコンボリューションは、装置関数ｆ（ｔ）と出力信号ｇ（ｔ）が既知であるときに、入力信号ｈ（ｔ）を求める計算法である。デコンボリューションの計算には、高効率の演算アルゴリズムである高速フーリエ変換（ＦＦＴ：Ｆａｓｔ　Ｆｏｕｒｉｅｒ　Ｔｒａｎｓｆｏｒｍ）を使うと便利である。
【００４７】
先ず、入力信号ｈ（ｔ）と装置関数ｆ（ｔ）のコンボリューションが出力信号ｇ（ｔ）であるから、デコンボリューションはこの逆演算である。デコンボリューションを計算するためには、一般に、フーリエ変換を利用するのが効率的である。入力信号ｈ（ｔ）、装置関数ｆ（ｔ）および出力信号ｇ（ｔ）は、総て時間ｔの関数であるから、フーリエ変換すると、それぞれＨ（ω）、Ｆ（ω）およびＧ（ω）となり、これらは総て周波数ωの関数であり、式１１が成立する。
【数１１】
Ｈ（ω）＝Ｇ（ω）／Ｆ（ω）　　　　・・・（１１）
Ｈ（ω）は、Ｇ（ω）にＦ（ω）の逆数を乗じたものであるから、インバース・フィルタに相当する１／Ｆ（ω）を用いてＨ（ω）を算出したことになる。式１１により求めたＨ（ω）をフーリエ逆変換すれば、求めたい入力信号ｈ（ｔ）を算出できる。
従って、シリコンウェハ２０からのみのパルス▲１▼〜▲４▼を完全に分離することができる。
【００４８】
ここで、式１１のＨ（ω）は、Ｇ（ω）をＦ（ω）で除したものであるから、Ｈ（ω）をフーリエ逆変換した入力信号ｈ（ｔ）には、計測時系列波形Ｃと基準時系列波形Ｂとの比を求めることに相当する演算が含まれている。これは、振幅透過率に相当する量になっている。
【００４９】
上述したインバース・フィルタは、信号強度に対してノイズが大きい周波数帯域では十分に機能しないという欠点がある。すなわち、ノイズのみが強調され、式１１の分子、分母で表される信号成分がともに小さくなるために、計算結果は不正確になる恐れがある。
この欠点を除くために、インバース・フィルタの代わりに、ウィナー・ヘルストローム・フィルタ（Ｗｉｅｎｅｒ−Ｈｅｌｓｔｒｏｍ　Ｆｉｌｔｅｒ）を用いるとよい。ウィナー・ヘルストローム・フィルタを用いた場合は、式１１に対応する式１２が成立する。
【数１２】
Ｈ（ω）＝Ｆ（ω）^＊・Ｇ（ω）／（｜Ｆ（ω）｜^２＋｜Ｎ（ω）｜^２）　・・・（１２）
Ｆ（ω）^＊は複素共役を表し、Ｎ（ω）はノイズの周波数スペクトルである。式１１と同様に、式１２のＨ（ω）をフーリエ逆変換すれば、求めたい入力信号ｈ（ｔ）を算出できる。このｈ（ｔ）は、Ｓ／Ｎ比が統計的に最大となるので、ノイズが含まれている周波数帯域でもノイズの影響を抑えた値を算出することができる。
【００５０】
本実施の形態では、テラヘルツ光発生器やテラヘルツ光検出器に起因する多重反射パルスを除去して、被測定物（シリコンウェハ）２０からのみの各パルスを分離抽出することができる。これにより、テラヘルツ光発生器やテラヘルツ光検出器のアンテナ基板７，１１の厚さを被測定物２０の厚さに応じて変更する手間を省くことができる。
【００５１】
本実施の形態では、上述した第１の実施の形態、すなわち被測定物２０を透過した透過光に関する計測時系列波形の多重反射パルスの分離について説明したが、後述する被測定物２０から反射する反射光に関する計測時系列波形の多重反射パルスの分離についても適用できる。
【００５２】
−第３の実施の形態−
図１１は、本発明の第３の実施の形態による分光計測装置を模式的に示す概略構成図である。図１１おいて、図１中の要素と同一又は対応する要素には同一符号を付し、その重複する説明は省略する。
【００５３】
本実施の形態による分光計測装置が前述した第１の実施の形態と異なる点は、以下の通りである。すなわち、第１の実施の形態では、被測定物２０の透過光が検出器１１で検出されるのに対し、本実施の形態では、被測定物２０の反射光が検出器１１で検出される点と、制御・演算処理部２３の動作である。
【００５４】
上述した第１の実施の形態では、被測定物２０の各測定部位について、透過光に関する計測時系列波形Ｅｓａｍ（ｔ）を取得していたのに対し、本実施の形態では、同様の手法により、被測定物２０の各測定部位について、反射光に関する計測時系列波形Ｅｓａｍ’（ｔ）を取得する。
【００５５】
被測定物２０で反射するテラヘルツパルス光には、図１２（ａ）〜図１２（ｄ）に示される様々なパターンの反射光が含まれる。図１２（ａ）は、被測定物２０の入射面（検出器１１側の面）でのみ反射する光（内部で１回も反射せずに、多重反射でない反射光）、図１２（ｂ）は、被測定物２０の検出器１１と反対側の面で１回のみ反射する光（内部で１回反射するが、多重反射でない反射光）、図１２（ｃ）は、被測定物２０の内部で３回反射した光（１回多重反射した光）、図１２（ｄ）は、被測定物２０の内部で（２ｋ−１）回反射した光（（ｋ−１）回多重反射した光）をそれぞれ示している。ここで便宜上、ｋを０以上の整数とし、図１２（ｂ）に示した光を０回多重反射した光、図１２（ａ）に示した光を−１回多重反射した光と呼ぶことにする。なお、図１２において、反射回数は、被測定物２０の内部での反射回数を示している。
【００５６】
テラヘルツ時間領域分光法により得られる反射時系列波形ｆ（ｔ）の概念図は図４に示したとおりである。図４で示される測定時系列波形ｆ（ｔ）を、図９（ａ）〜（ｄ）で示される個々の波の足し合わせであると考える。すなわち、平行平板試料２０の表面での反射パルスを表す図９（ａ）の波をａ（ｔ）、裏面での反射パルスを表す図９（ｂ）の波をｂ（ｔ）、１回目の多重反射パルスを表す図９（ｃ）の波をｃ（ｔ）、２回目の多重反射パルスを表す図９（ｄ）の波をｄ（ｔ）とする。これら図９（ａ）〜（ｄ）に示す各波形をフーリエ変換してそれらの絶対値の総和（総和値）をとる。そして、リファレンス光に対してフーリエ変換を行い、その絶対値（基準絶対値）をとる。検出パルス光に関する総和値と基準絶対値の比を算出して平行平板試料のスペクトルを解析する。
【００５７】
第３の実施の形態で採用されている、計測時系列波形Ｅｓａｍ’（ｔ）に基づいて被測定物２０のスペクトルを解析する手法について図１３のフローチャートを参照して詳細に説明する。発生器７と検出器１１との間の光路上に、被測定物２０に代えて、既知の複素屈折率及び既知の厚みを持つ試料を配置する。この試料の反射率はほぼ１００％であることが好ましい。この状態で、計測時系列波形Ｅｓａｍ’（ｔ）を取得する場合と同様に、反射光に関するテラヘルツパルス光の電場強度の基準時系列波形Ｅｒｅｆ’（ｔ）を計測する（ステップＳ１１）。計測結果はＡ／Ｄ変換器２２でデジタルデータに変換され、メモリ２３Ａに記憶される。制御・演算処理部２３はメモリ２３Ａに記憶した基準時系列波形Ｅｒｅｆ’（ｔ）を読み出してフーリエ変換を実行する。さらに、フーリエ変換した結果の絶対値（基準絶対値と呼ぶ）を算出して記憶する（ステップＳ１２）。
【００５８】
発生器７と検出器１１との間の光路上に被測定物２０を配置した状態で、テラヘルツパルス光の電場強度の計測時系列波形Ｅｓａｍ’（ｔ）を計測して記憶する（ステップＳ１３）。計測結果はＡ／Ｄ変換器２２でデジタルデータに変換され、メモリ２３Ａに記憶される。この計測時系列波形Ｅｓａｍ’（ｔ）は概念図として図４に示すような時系列波形である。制御・演算処理部２３はメモリ２３Ａに記憶した計測時系列波形Ｅｓａｍ’（ｔ）を読み出し、計測時系列波形Ｅｓａｍ’（ｔ）を構成する図９（ａ）〜（ｄ）に示すパルス波形をそれぞれ分離抽出する（ステップＳ１４）。分離抽出したパルス波形に対してフーリエ変換を順次実行してその絶対値を算出し、その総和（式（１０）に対応する総和値）を算出する（ステップＳ１５）。この総和値と基準絶対値との比を取る（ステップＳ１６）。これにより、各パルスをインコヒーレントとして解析することができる。
【００５９】
本発明の効果を確認するために、実際にシリコンウェハを反射測定し解析した。図１４（ａ）にシリコンウェハの反射時系列波形を示す。図１４（ａ）のグラフから、シリコンウェハの測定波形が、２回目の多重反射までのパルスを含んでいることが分かる。この測定波形を、通常の解析手法で解析した結果が図１４（ｂ）のグラフであり、本解析手法を用いて解析した結果が図１４（ｃ）のグラフである。これより、実際の測定においても本解析手法を用いることで、多重反射を含み、かつ干渉パターンのない解析結果を得られるようになることがわかる。
【００６０】
ここで、インコヒーレントな解析処理とは、解析処理後に干渉パターンのないスペクトルを得るような処理である。したがって、インコヒーレントな解析処理は上記実施の形態に限定されない。たとえば、平行平板試料からの時系列検出パルス光をパルス毎に分離し、分離した複数のパルスと参照パルスとに基づいた解析処理を行う処理であればどのような手法でも良い。一例としては、分離した各パルスに対して周波数分析を行い、参照パルスに対して周波数分析を行い、分離した各パルスに対する周波数分析結果と参照パルスに対する周波数分析結果に基づいた解析処理である。
【００６１】
また以上では、分離した各パルスに対してフーリエ変換を行い、各パルスに対するフーリエ変換結果の絶対値の総和（総和値）を算出し、参照パルスに対してフーリエ変換を行い、フーリエ変換結果の絶対値（参照値）を算出し、総和値と参照値の比を算出して検出パルス光のスペクトルを得るようにした。しかし次のような処理でも良い。
（１）分離した各パルスに対してフーリエ変換を行い、各パルスに対するフーリエ変換結果の絶対値を算出する。
（２）参照パルスに対してフーリエ変換を行い、フーリエ変換結果の絶対値（参照値）を算出する。
（３）分離パルスの絶対値と参照値の比を算出して検出パルス光のスペクトルを得る。この場合、算出される比は複数得られるので、それらの総和を用いても良いし、個々の比を用いても良い。
【００６２】
なお、本発明はテラヘルツパルス光以外の種々のパルスの分光計測に適用することができる。
【００６３】
【発明の効果】
本発明によれば、測定結果における各パルスに対してインコヒーレントな解析処理を行うことにより、多重反射を含んだ場合においても、干渉パターンを含まないスペクトルを得ることが可能になる。
また、本発明によれば、テラヘルツ時間領域分光法の光源として用いているパルス光の特徴を活かすことにより、測定結果の各パルスに対してインコヒーレントな解析処理を行うので、多重反射したパルスや裏面反射したパルスを含んでいる場合にも、干渉パターンを含まないスペクトルを得ることが可能になる。
さらに、本発明によれば、平行平板試料以外のものに起因する多重反射による測定結果への影響を除去するようにしたので、測定結果の各パルスに対してインコヒーレントな解析処理を行うことができる。
【図面の簡単な説明】
【図１】図１は、本発明の第１の実施の形態による透過型テラヘルツ分光計測装置を示す概略構成図
【図２】図２は、本発明の第１の実施の形態における平行平板試料の透過パルスを説明する図。
【図３】図３は透過時系列波形の概念図
【図４】図４は反射時系列波形の概念図
【図５】図５は、干渉パターンを含む透過率もしくは反射率の解析結果を示す図
【図６】図６（ａ）は屈折率の異なる界面での透過と反射を説明する図
図６（ｂ）は各多重反射回数における透過光と反射光を説明する図
【図７】図７は、コヒーレントとインコヒーレントな場合における、無限回の多重反射を含んだ透過光の複素振幅の計算結果を示す図
【図８】図８（ａ）〜（ｄ）は、透過時系列波形をインコヒーレントとして取り扱う場合の分離した波形の概念図
【図９】図９（ａ）〜（ｄ）は、反射時系列波形をインコヒーレントとして取り扱う場合の分離した波形の概念図
【図１０】図１０は、本発明の第１の実施の形態による透過型テラヘルツ分光計測装置の制御・演算処理部の動作を示す概略フローチャート
【図１１】図１１は、本発明の第３の実施の形態による反射型テラヘルツ分光計測装置を示す概略構成図
【図１２】図１２は、本発明の第３の実施の形態における平行平板試料の反射パルスを説明する図。
【図１３】図１３は、本発明の第３の実施の形態による反射型テラヘルツ分光計測装置の制御・演算処理部の動作を示す概略フローチャート
【図１４】図１４（ａ）は、実際に測定されたシリコンウェハの反射時系列波形図
図１４（ｂ）は、通常の解析手法により得られるシリコンウェハの解析結果を示す図
図１４（ｃ）は、本解析手法により得られるシリコンウェハの解析結果を示す図
【図１５】図１５（ａ）は、アンテナ基板７，１１による多重反射を全く含まない理想的な透過時系列波形図
図１５（ｂ）は、平行平板試料２０を配設しないで得られた透過時系列波形Ｂ、すなわち基準時系列波形図
図１５（ｃ）は、平行平板試料２０を配設して得られた現実的な透過時系列波形Ｃ、すなわち計測時系列波形図
【符号の説明】
１：フェムト秒パルス光源　　　　７：テラヘルツ光発生器
９：可動鏡　　　　　　　　　　１１：テラヘルツ光検出器
２０：平行平板試料　　　　　　　２３：制御・演算処理部[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a pulse spectrometry method and apparatus for obtaining a spectrum that does not include an interference pattern while taking into account multiple reflection inside a sample when analyzing a spectrum of a parallel plate sample.
[0002]
BACKGROUND OF THE INVENTION
Terahertz time-domain spectroscopy using terahertz pulse light as a light source is known. When a transmission time series waveform or a reflection time series waveform of a parallel plate sample is measured by this terahertz time domain spectroscopy, a pulse reflected multiple times inside the sample is seen in the transmission measurement, and a pulse reflected inside the sample is reflected in the reflection measurement. The pulse reflected multiple times is seen at. In this measurement waveform, the phase difference between each pulse is constant, that is, each pulse is in a coherent state. Therefore, when the analysis is performed including the pulse reflected by the multiple reflection and the pulse reflected by the back surface, an interference pattern appears in the analysis result. This interference pattern largely contributes from the thickness of the sample, and does not reflect only the physical quantity inside the sample. For example, when exactly the same sample is measured except for the thickness, different interference patterns appear in both analysis results.
[0003]
The present invention provides a pulse that obtains a spectrum that does not include an interference pattern by performing incoherent analysis processing on transmission time-series pulse light or reflection time-series detection pulse light that includes a pulse that is multiple reflected inside a parallel plate sample. An object of the present invention is to provide a spectroscopic measurement method and apparatus using light. Further, the present invention provides a method for analyzing spectroscopic measurement data in which an influence on a measurement result due to multiple reflection caused by a material other than a parallel plate sample is removed.
[0004]
[Means for Solving the Problems]
(1) A spectroscopic measurement method using pulsed light according to the present invention includes irradiating a parallel plate sample with pulse light, detecting a detection pulse light reflected or transmitted from the parallel plate sample irradiated with the pulse light, and detecting the parallel plate sample. By performing incoherent analysis processing on coherent detection pulse light including internally multiple reflected pulses, a spectrum that does not include an interference pattern is obtained.
(2) When the present invention is applied to a spectroscopic measurement method using terahertz pulsed light, incoherent analysis processing includes processing for separating a time-series detection pulsed light from a parallel plate sample for each pulse, and a plurality of separated pulsed lights. An analysis process based on the pulse and the reference pulse may be included. Alternatively, the incoherent analysis process includes a process of separating time-series detection pulse light from a parallel plate sample for each pulse, a process of performing frequency analysis on each separated pulse, and a frequency analysis of a reference pulse. Processing to be performed and analysis processing based on the frequency analysis result for each separated pulse and the frequency analysis result for the reference pulse can be included.
(3) The incoherent analysis processing preferably separates the time-series detection pulse light from the parallel plate sample for each pulse, performs Fourier transform on each pulse, and calculates the absolute value of the Fourier transform result for each pulse. The sum (sum value) is calculated, the Fourier transform is performed on the reference pulse, the absolute value (reference value) of the Fourier transform result is calculated, the ratio of the sum value to the reference value is calculated, and the spectrum of the detected pulse light is calculated. It is a process to obtain.
(4) A spectrometer using pulsed light according to the present invention comprises: an irradiating means for irradiating a parallel plate sample with pulsed light; and a detecting means for detecting a detected pulsed light reflected or transmitted from the parallel plate sample irradiated with the pulsed light. And an analyzing means for performing an incoherent analysis process on a coherent detection pulse light including a pulse multiply reflected inside the parallel plate sample, thereby obtaining a spectrum without an interference pattern. .
(5) A spectrometer that applies the present invention to a spectrometer using terahertz pulse light includes a terahertz pulse light generator that irradiates terahertz pulse light to a parallel plate sample, and a parallel plate sample that is irradiated with terahertz pulse light. The terahertz pulse light detector that detects the detection pulse light that is reflected or transmitted from the sample, and the incoherent analysis processing for the coherent detection pulse light that includes the pulse that is multiple-reflected inside the parallel plate sample, the interference pattern is formed. Analysis means for obtaining a spectrum that does not include the spectrum.
(6) analyzing means for separating the time-series detection pulsed light from the parallel plate sample for each pulse, and processing means for obtaining a spectrum without an interference pattern based on the plurality of separated pulses and the reference pulse; Can be included. Alternatively, the analyzing means includes a separating means for separating the time-series detection pulsed light from the parallel plate sample for each pulse, a first frequency analyzing means for performing a frequency analysis on each of the separated pulses, and a frequency analyzing means for the reference pulse. A second frequency analysis unit for performing analysis and a processing unit for obtaining a spectrum that does not include an interference pattern based on a frequency analysis result for each separated pulse and a frequency analysis result for a reference pulse can be included.
(7) Preferably, the analyzing means is a separating means for separating the time-series detected pulsed light from the parallel plate sample for each pulse, performs a Fourier transform on each pulse, and calculates an absolute value of a Fourier transform result for each pulse. First calculating means for calculating the sum (total value), second calculating means for performing Fourier transform on the reference pulse and calculating the absolute value (reference value) of the Fourier transform result, and a ratio of the sum value to the reference value And a third calculating means for obtaining the spectrum of the detected pulse light.
(8) Preferably, the thickness of the substrate of the terahertz pulse light generator, the thickness of the substrate of the terahertz pulse light detector, and the thickness of the parallel plate sample are respectively multiplied by each substrate and multiplexed by the parallel plate sample. Determine that reflections are separated.
(9) In the method for analyzing spectroscopic measurement data using pulsed light according to the present invention, a detection pulse light reflected or transmitted from a parallel plate sample irradiated with pulsed light, including a pulse multiple reflected inside the sample, is detected. The measurement time-series waveform is obtained by the detected pulse light, and the pulse light is irradiated in a state in which a perfect reflection mirror is provided in place of the parallel plate sample. Irradiate pulsed light without placing a flat sample to obtain a reference time-series waveform from the directly detected pulse light, obtain the deconvolution of the measured time-series waveform and the reference time-series waveform, and multiplex each on the time axis. It is characterized in that a time-series waveform in which the reflected pulse light is separated is obtained.
(10) In the deconvolution, it is preferable to perform a predetermined operation after performing a Fourier transform of the measured time-series waveform and the reference time-series waveform, and to perform an inverse Fourier transform of the operation result.
(11) The calculation is an operation using an inverse filter, considering the time series waveform only from the parallel plate sample as an input signal, the measurement time series waveform as an output signal, and the reference time series waveform as an apparatus function. You may. In addition, a Wiener-Hellstrom filter may be used instead of the inverse filter. These methods of analyzing spectroscopic measurement data can also be applied to a case where terahertz pulse light is used as pulse light.
(12) In the spectroscopic measurement method using pulsed light according to the present invention, the time-series waveform (input signal) in which each multiple reflected pulsed light is separated on the time axis obtained in (9) to (11). The incoherent analysis processes (1) to (8) described above are performed to obtain a spectrum that does not include an interference pattern.
[0005]
BEST MODE FOR CARRYING OUT THE INVENTION
-1st Embodiment-
A first embodiment of a spectrometer using terahertz pulse light according to the present invention will be described with reference to FIGS. The spectrometer according to the first embodiment performs an incoherent analysis process on a coherent transmission time-series pulse light including a pulse multiply reflected inside a parallel plate sample irradiated with a terahertz pulse light. , To obtain a spectrum that does not include an interference pattern.
[0006]
FIG. 1 is a schematic configuration diagram of a transmission-type spectrometer using terahertz pulsed light according to the first embodiment of the present invention. In the spectrometer according to the present embodiment, as shown in FIG. 1, a femtosecond pulse light L1 emitted from a femtosecond pulse light source 1 composed of a laser light source or the like is converted into two pulse lights L2 and L2 by a beam splitter 2. It is divided into L3.
[0007]
One of the divided pulse lights L2 becomes a pump pulse (pulse excitation light) for exciting the terahertz light generator (light source) 7 and causing the generator 7 to generate terahertz pulse light. After being pumped by the chopper 3, the pump pulse L2 is guided to the terahertz light generator 7 through the plane reflecting mirrors 4, 5, and 6. The other divided pulse light L3 becomes a probe pulse (sampling pulse light) for determining the timing for detecting the terahertz pulse light. The probe pulse L3 is guided to a terahertz photodetector 11 via a plane reflecting mirror 8, a movable mirror 9 formed by combining two or three plane reflecting mirrors, and a plane reflecting mirror 10.
[0008]
The movable mirror 9 arranged on the optical path of the probe pulse L3 can be moved by the moving mechanism 12 in the arrow X direction. The moving mechanism 12 is controlled by the control / arithmetic processing unit 23. The optical path length of the probe pulse L3 changes according to the amount of movement of the movable mirror 9, and the time for the probe pulse L3 to reach the detector 11 is delayed. That is, in the present embodiment, the movable mirror 9 and the moving mechanism 12 constitute a time delay device that changes the time when the probe pulse L3 reaches the detector 11.
[0009]
On the other hand, when the generator 7 is excited by the pump pulse L2, the generator 7 emits terahertz pulse light L4. As the terahertz pulse light L4, approximately 0.1 × 10 ¹² From 100 × 10 ¹² Light in the frequency range up to Hertz is desirable. The terahertz pulse light L4 is condensed at a light condensing position via curved mirrors 13, 14 such as parabolic mirrors.
[0010]
In the present embodiment, a measurement site of the device under test 20 is arranged at the light-collecting position. Here, it is assumed that the DUT 20 is a plate-like sample having a parallel plate with a known thickness d and is made of a uniform substance. However, the DUT 20 is not limited to this. For example, when the terahertz pulsed light is locally irradiated, the thickness d may be different for each measurement site if known.
[0011]
The device under test 20 is arranged such that the optical axis of the terahertz pulse light L4 with respect to the device under test 20 substantially matches the normal to the surface of the device under test 20. To improve the measurement accuracy, it is better that the incident angle distribution is narrow. Therefore, it is better that the angle θ between the outermost ray of the terahertz pulsed light L4 incident on the DUT 20 and the optical axis is smaller. On the other hand, in order to increase the spatial resolution as an image, the larger the angle θ, the better. Therefore, it is necessary to determine the angle θ as necessary.
[0012]
When an image is not necessary, although not shown in the drawings, it is better to employ an irradiation optical system in which the terahertz pulse light L4 is incident on the DUT 20 as parallel light to improve the measurement accuracy. More preferred above. In the following description, it is assumed that the terahertz pulse light is perpendicularly incident on the DUT 20, but the case of oblique incidence can be similarly considered.
[0013]
In the present embodiment, the DUT 20 can be moved two-dimensionally in a direction within the plane of the DUT 20 by a moving mechanism 26 such as a stage. Thereby, the measurement site of the DUT 20 can be two-dimensionally scanned along the surface of the DUT 20.
[0014]
The terahertz pulse light L5 transmitted through the device under test 20 is detected by the detector 11 via curved mirrors 15, 16 such as parabolic mirrors. The terahertz pulse light L5 detected by the detector 11 is converted into an electric signal and sent to the amplifier 21.
[0015]
The repetition period of the femtosecond pulse light L1 emitted from the femtosecond pulse light source 1 is on the order of several KHz to MHz. Therefore, the terahertz pulse light L4 radiated from the generator 7 is also radiated in the order of several KHz to MHz. The existing detector 11 cannot measure the waveform of the terahertz pulse light instantaneously with its shape.
[0016]
Therefore, in the present embodiment, a time delay is provided between the pump pulse L2 and the probe pulse L3 by utilizing the fact that the terahertz pulse light L4 having the same waveform is repeatedly emitted from several KHz to the order of MHz. A method of measuring the waveform of the pulse light L5, that is, a so-called pump-probe method is employed. That is, by delaying the timing of activating the terahertz light detector 11 by τ seconds with respect to the pump pulse L2 of activating the terahertz light generator 7, the electric field intensity of the terahertz pulse light L5 at the time delayed by τ seconds is reduced. It can be measured by the detector 11. In other words, the probe pulse L3 gates the terahertz photodetector 11.
[0017]
By gradually moving the movable mirror 9, the delay time τ can be gradually changed. By shifting the timing of gate application by the above-described time delay device, the electric field intensity at each delay time τ of the repeatedly arriving terahertz pulse light L5 can be sequentially obtained from the detector 11 as an electric signal. Thereby, the time-series waveform E (t) of the electric field intensity of the terahertz pulse light L5 can be measured.
[0018]
Note that the terahertz photodetector 11 generates photoexcited carriers only when receiving the probe pulse L3. At this time, if an electric field of the terahertz pulse light is applied, a photoconductive current flows in proportion to the electric field. The current J (τ) measured at this time can be expressed using the electric field E (t) of the terahertz pulse light and the photoconductivity g (t−τ) of the photoexcited carrier. That is, J (τ) = ∫E (t) g (t−τ) dt. Since the photoconductivity g (t−τ) has a delta function characteristic, the measured current value J (τ) can be considered to be proportional to the electric field intensity E (t) of the arriving terahertz pulse light L5. . The electric signal sent from the detector 11 to the amplifier 21 is amplified by the amplifier 21 and then A / D converted by the A / D converter 22.
[0019]
In the present embodiment, when measuring the time-series waveform E (t) indicating the electric field intensity of the terahertz pulse light L5, the control / arithmetic processing unit 23 transmits a control signal to the moving mechanism 12 and gradually reduces the delay time τ. The data sequentially input from the A / D converter 22 is stored in the memory 23A while changing to. Thus, the entire data indicating the time-series waveform E (t) of the electric field intensity of the terahertz pulse light L5 transmitted through a certain measurement site of the device under test 20 is stored in the memory 23A. Thereafter, the control / arithmetic processing unit 23 transmits a control signal to the moving mechanism 26 and sequentially scans the measurement site of the DUT 20 two-dimensionally, thereby obtaining a time-series waveform E (t) for each measurement site. Is stored in the memory 23A.
[0020]
In the present embodiment, measurement is performed in a state where the DUT 20 is arranged on the optical path between the generator 7 and the detector 11 (in the present embodiment, the focus position of the terahertz pulse light L4 shown in FIG. 1). The time-series waveform E (t) of the electric field intensity of the terahertz pulsed light is referred to as a measured time-series waveform Esam (t).
[0021]
The terahertz pulse light transmitted through the device under test 20 includes various patterns of transmitted light shown in FIGS. 2A to 2D. 2A is light that is transmitted without being reflected inside the DUT 20 (transmitted light that is not multiple reflection), and FIG. 2B is reflected twice that is transmitted inside the DUT 20. FIG. 2C shows light (light that has been multiply reflected once) and FIG. 2C shows light that has been reflected and transmitted four times inside the device under test 20 (light that has been multiply reflected twice). FIG. Light that is reflected and transmitted 2k times inside the object 20 (light that is k-times multiple reflected) is shown. Here, k is an integer of 0 or more.
[0022]
FIG. 3 is a conceptual diagram of a transmission time-series waveform obtained by terahertz time-domain spectroscopy, and FIG. 4 is a conceptual diagram of a reflection time-series waveform. When the Fourier transform F (ω) of the waveform f (t) measured in this manner is calculated, and the ratio between the Fourier transform and the Fourier transform of the reference light is calculated, the transmittance or the reflectance is analyzed. Due to the effect of multiple reflection, an interference pattern as shown in FIG. 5 appears. This interference pattern largely contributes from the thickness of the sample, and does not reflect only the physical quantity inside the sample.
[0023]
Therefore, as shown in FIG. 6A, consider a case where light is vertically incident on a sample (refractive index n1) placed in a medium having a refractive index n0. Here, in the drawing, t0, t1 and r0, r1 are the amplitude transmittance and the amplitude reflectance at the interface between the medium having the refractive index n0 and the sample. t0 represents the amplitude transmittance of light incident on the sample from the medium having the refractive index n0, and t1 represents the amplitude transmittance of light emitted from the sample to the medium having the refractive index n0. R0 represents the amplitude reflectance of light reflected by the sample, and r1 represents the amplitude reflectance of light reflected by a medium having a refractive index n0.
[0024]
In this case, the amplitude transmittance and reflectance at each interface are
(Equation 1)

(Equation 2)

Given by
[0025]
Usually, the amplitude transmittance t of the light transmitted through the sample is
[Equation 3]

Is represented by The first term on the right side of the first row represents light that has directly transmitted, and the second and third terms respectively represent transmitted light that has undergone multiple reflection once and transmission light that has undergone multiple reflection twice inside the sample (FIG. 6 (b)). Where θ is
(Equation 4)

Given by c is the speed of light in a vacuum, and d is the thickness of the sample.
[0026]
The amplitude reflectance r of the light reflected by the sample is
(Equation 5)

Is represented by The first term on the right side of the first row represents light reflected on the front surface of the sample, and the second term represents light reflected on the back surface. The third and fourth terms respectively represent light that has been multiply reflected once and light that has been multiply reflected twice inside the sample (see FIG. 6B). Equations (3) and (5) correspond to the amplitude transmittance and the amplitude reflectance when the respective amplitudes are added together while maintaining the phase information of each light. That is, each light is treated as coherent.
[0027]
On the other hand, taking the absolute values of the amplitudes of the back reflection and the multiple reflections to remove the phase information and adding only the magnitudes corresponds to the case where each light is incoherent. Therefore, equations (3) and (5) are
(Equation 6)

(Equation 7)

Then, it corresponds to the amplitude transmittance and the amplitude reflectance when each light is incoherent.
[0028]
Equations (3) and (5) and Equations (6) and (7) include all multiple reflections up to infinity. FIG. 7 shows the amplitude transmittance of the silicon wafer calculated using Equation (3) for the coherent case and Equation (6) for the incoherent case. As is clear from this, the interference pattern seen in the coherent case is not seen at all in the incoherent case. From this, if each measured light can be analyzed as incoherent, a spectrum that does not include an interference pattern can be measured even if multiple reflections are taken into account.
[0029]
Normally, when analyzing a measurement waveform obtained by terahertz time domain spectroscopy, the backside reflected light and each multiple reflected light are analyzed as coherent. However, terahertz time-domain spectroscopy uses femtosecond pulsed light sources. Therefore, it is possible to execute incoherent analysis processing on the detected pulse light by extracting each pulse of the time series waveform included in the time series transmitted pulse train or time series reflected pulse train from the parallel plate sample. Become.
[0030]
The transmission time-series waveform f (t) shown in FIG. 3 is added to the individual waves shown in FIGS. 8A to 8D, or the reflection time-series waveform f (t) shown in FIG. This is considered to be the sum of the individual waves shown in FIGS. 9 (a) to 9 (d). That is, the transmitted pulse transmitted through the wafer is a time-series pulse including the individual pulses shown in FIGS. On the other hand, the reflected pulses reflecting the wafer are time-series pulses including the individual pulses shown in FIGS.
[0031]
That is, as the transmission pulse, the wave of FIG. 8A representing the pulse transmitted through the wafer is represented by a (t), and the wave of FIG. 8B representing the pulse transmitted through the first multiple reflection is represented by b (t). The wave of FIG. 8C representing the second multiple reflection pulse is represented by c (t), and the wave of FIG. 8D representing the third multiple reflection pulse is represented by d (t). On the other hand, as for the reflection pulse, the wave of FIG. 9A representing the reflection pulse on the wafer surface is represented by a (t), and the wave of FIG. 9B representing the reflection pulse on the back surface is represented by b (t). The wave in FIG. 9C representing the reflected pulse is denoted by c (t), and the wave in FIG. 9D representing the second multiple reflected pulse is denoted by d (t). F (t) indicating a transmission time series pulse or a reflection time series pulse is
(Equation 8)

Given by
[0032]
Further, the Fourier transform of each pulse of a (t), b (t), c (t), and d (t) is A (ω), B (ω), C (ω), and D (ω), respectively. . At this time, the Fourier transform F (ω) of f (t) is
(Equation 9)

It can be. The ratio between this and the Fourier transform of the reference light is obtained by converting the amplitude transmittance (Equation (3)) or the amplitude reflectance (Equation (5)) including infinite multiple reflections into a finite number of multiple reflections. Corresponds to censored. In the case of the transmission type, the reference light has a waveform of light detected when the terahertz pulse light is irradiated from the femtosecond pulse light source 1 to the detector 11 with the sample removed. On the other hand, in the reflection type, it is a light waveform detected when a terahertz pulse light is irradiated from the femtosecond pulse light source 1 to the detector 11 by arranging a reference sample having a reflectance of 100% instead of the measurement target sample.
[0033]
Therefore, a case is considered in which each wave is subjected to Fourier transform, and its absolute value is taken and added. The sum Fincoh (ω) of each wave at this time is
(Equation 10)

By taking the ratio of this to the absolute value of the Fourier transform of the reference light, the amplitude reflectivity (Equation (6)) or the amplitude reflectivity (Equation (7)) in the incoherent case is multiplexed finite times. Corresponds to one censored by reflection. That is, each pulse of the measurement result is individually subjected to Fourier transform, and the absolute value of each amplitude is added, whereby each pulse can be analyzed as incoherent.
[0034]
The above analysis method becomes possible when the light source is pulsed light. For example, in the conventional Fourier transform infrared spectroscopy, continuous light is usually used as a light source. In this case, each multiple reflection pulse cannot be separated individually. That is, whether to analyze as coherent or incoherent is uniquely determined by the coherence length of the light source. Therefore, the spectroscopic measurement method and apparatus according to the present invention are not limited to those using terahertz pulsed light, but can be applied to other spectroscopic measurement methods and apparatuses using pulsed light.
[0035]
Here, a procedure of analyzing the spectrum of the DUT 20 based on the measurement time-series waveform Esam (t) employed in the present embodiment will be described with reference to the flowchart of FIG.
A reference time-series waveform Eref (t) of the electric field intensity of the terahertz pulse light is measured in advance in a state where the device under test 20 is not arranged on the optical path between the generator 7 and the detector 11 (Step S1). The measurement result is converted into digital data by the A / D converter 22 and stored in the memory 23A. The control / arithmetic processing unit 23 reads out the reference time-series waveform Eref (t) stored in the memory 23A and executes a Fourier transform. Further, an absolute value (referred to as a reference absolute value) of the result of the Fourier transform is calculated and stored (step S2).
[0036]
The measurement time-series waveform Esam (t) of the electric field intensity of the terahertz pulse light is measured with the device under test 20 arranged on the optical path between the generator 7 and the detector 11 (step S3). The measurement result is converted into digital data by the A / D converter 22 and stored in the memory 23A. This measurement time-series waveform Esam (t) is a time-series waveform as shown in FIG. 3 as a conceptual diagram. The control / arithmetic processing unit 23 reads the measured time-series waveform Esam (t) stored in the memory 23A, and separates the pulse waveforms shown in FIGS. 8A to 8D constituting the measured time-series waveform Esam (t). Extract (step S4). Fourier transform is sequentially performed on each of the separated and extracted pulse waveforms to calculate their absolute values, and their sum (total value corresponding to equation (10)) is calculated (step S5). The ratio between the total value and the reference absolute value is calculated (step S6). Thereby, each pulse can be analyzed as incoherent.
[0037]
The spectrometer according to the first embodiment also considers the following points.
In some measurement systems, not only multiple reflections due to a flat sample but also multiple reflections due to a light source or a detector exist. A terahertz pulse light source (the light generator 7 in FIG. 1) used in the terahertz time domain spectroscopy generates a terahertz pulse light with an antenna substrate (gallium arsenide substrate) and a terahertz pulse light detector (the light detector 11 in FIG. 1). ) Detects terahertz pulse light with an antenna substrate (gallium arsenide substrate). Therefore, from the relationship between the thickness and the refractive index of the antenna substrate (gallium arsenide substrate) used for the light source 7 and the detector 11 and the relationship between the thickness and the refractive index of the flat plate sample 20 to be measured, the transmission through the multiple reflection inside the sample 20 is performed. The light or the reflected light may be measured at the same timing as the multiple reflected light from the antenna substrates 7 and 11. In this case, the multiple reflection of the sample 20 and the multiple reflection of the antenna substrates 7 and 11 cannot be separated, and the above-described method of analyzing the detected pulse waveform as incoherent cannot be adopted.
[0038]
Therefore, in order to use the above analysis method, the antenna substrates 7 and 11 and the flat plate sample 20 are separated from each other based on the relationship between the thickness and the refractive index of the antenna substrates 7 and 11 and the flat plate sample 20 so that multiple reflections of the two can be separated. Thickness needs to be optimized. For example, consider the case of measuring a 1000 μm thick silicon wafer.
[0039]
Gallium arsenide used as the antenna substrate of the terahertz pulse light source 7 and the detector 11 has a refractive index of about 3.6, and silicon has a refractive index of about 3.4. Therefore, when the thickness of the antenna substrates 7 and 11 is about 945 μm, the reflected light and the multiple reflected light on the back surface of the silicon wafer 20 and the multiple reflected light on the antenna substrates 7 and 11 are measured at almost the same timing. Therefore, for example, it is assumed that the thickness of the antenna substrates 7 and 11 is set to about 235 μm, which is a quarter of the measurement sample 20. In this case, the fourth multiple reflection by the antenna substrates 7 and 11 is measured at substantially the same timing as the back reflection of the silicon wafer 20.
[0040]
However, the fourth multiple reflection by the antenna substrates 7 and 11 is sufficiently attenuated and weaker than the rear surface reflection on the silicon wafer 20, and can be ignored. The width of the terahertz pulse is about 1 ps, during which the light travels about 300 μm. This distance is sufficiently shorter than the value obtained by multiplying the optical path length reciprocating between the antenna substrates 7 and 11 by the refractive index of the antenna substrates 7 and 11 (about 1700 μm). Therefore, each pulse light is spatially separated, and the multiple reflection of the antenna substrates 7 and 11 and the multiple reflection of the parallel plate sample 20 can be separated and analyzed.
[0041]
-2nd Embodiment-
The incoherent analysis method described so far removes multiple reflections from the antenna substrate in advance by changing the thickness of the antenna substrate 7 of the terahertz light generator and the antenna substrate 11 of the terahertz photodetector, thereby reducing the parallelism. Only multiple reflections of the flat plate sample 20 are taken out and analyzed. However, changing the thickness of the antenna substrates 7 and 11 has to be performed every time the thickness of the parallel plate sample 20 changes, which is troublesome and costly. In the terahertz light generator and the terahertz light detector, there are factors other than the antenna substrates 7 and 11 that cause multiple reflection. However, the antenna substrates 7 and 11 are the largest factors, and therefore are represented by these factors.
[0042]
Hereinafter, without changing the thicknesses of the antenna substrates 7 and 11, even if there are multiple reflections by the antenna substrates 7 and 11, the influence is removed, and each pulse that has undergone multiple reflection inside the parallel plate sample 20 is separated. The method will be described.
[0043]
FIG. 15 is a conceptual diagram of a transmission time-series waveform obtained by terahertz time-domain spectroscopy. Hereinafter, the vertical axis of the transmission time-series waveform indicates amplitude, and the horizontal axis indicates time. FIG. 15A shows an ideal transmission time-series waveform A that does not include any multiple reflection by the antenna substrates 7 and 11. In the transmission time-series waveform A, the leftmost pulse (1) is transmitted without reflection inside the silicon wafer 20, and the pulse (2) is transmitted once after multiple reflection inside the wafer 20. The pulse {circle around (3)} is transmitted twice after multiple reflection inside the wafer 20, and the pulse {circle around (4)} is triple reflected inside the wafer 20 and transmitted. In such an ideal case, it is easy to separate each pulse in the time domain, and the above-described incoherent analysis method can be applied immediately.
[0044]
FIG. 15B shows a transmission time series waveform B obtained without disposing the silicon wafer 20, that is, a reference time series waveform B. In this reference time-series waveform B, a pulse (5) directly incident on the terahertz light detector from the terahertz light generator, and multiple reflection pulses (6) and (7) by the antenna substrates 7 and 11 appear.
[0045]
FIG. 15C shows a realistic transmission time-series waveform C obtained by disposing the silicon wafer 20, that is, a measurement time-series waveform C. This is obtained by superimposing the ideal transmission time-series waveform A of FIG. 15A on the reference time-series waveform B of FIG. It is extremely difficult to directly separate each pulse of the transmission time series waveform A from the measurement time series waveform C. Therefore, in order to separate the influence of multiple reflections appearing in the transmission time series waveform A from the influence of multiple reflections appearing in the reference time series waveform B, the thickness of the antenna substrates 7 and 11 is changed as described above. Needed.
[0046]
In the present embodiment, a technique is employed in which signal components included in both the reference time-series waveform B and the measurement time-series waveform C are removed by using deconvolution. This signal component is specifically caused by multiple reflections of the antenna substrates 7 and 11.
Now, assuming that the transmission time series waveform A is an input signal h (t), the reference time series waveform B is a device function f (t), and the measurement time series waveform C is an output signal g (t), Since the measurement time-series waveform C is actually measured, the device function f (t) and the output signal g (t) are known. Deconvolution is a calculation method for obtaining the input signal h (t) when the device function f (t) and the output signal g (t) are known. For the calculation of deconvolution, it is convenient to use fast Fourier transform (FFT), which is a highly efficient operation algorithm.
[0047]
First, since the convolution of the input signal h (t) and the device function f (t) is the output signal g (t), deconvolution is the inverse operation of this. In order to calculate the deconvolution, it is generally efficient to use the Fourier transform. Since the input signal h (t), the device function f (t) and the output signal g (t) are all functions of the time t, they are H (ω), F (ω) and G (ω ), Which are all functions of the frequency ω, and Equation 11 holds.
[Equation 11]
H (ω) = G (ω) / F (ω) (11)
Since H (ω) is obtained by multiplying G (ω) by the reciprocal of F (ω), H (ω) is calculated using 1 / F (ω) corresponding to an inverse filter. . The input signal h (t) to be obtained can be calculated by performing an inverse Fourier transform on H (ω) obtained by Expression 11.
Therefore, the pulses (1) to (4) only from the silicon wafer 20 can be completely separated.
[0048]
Here, since H (ω) in Equation 11 is obtained by dividing G (ω) by F (ω), the input signal h (t) obtained by inversely transforming H (ω) includes a measurement time series An operation corresponding to obtaining a ratio between the waveform C and the reference time-series waveform B is included. This is an amount corresponding to the amplitude transmittance.
[0049]
The above-mentioned inverse filter has a drawback that it does not function sufficiently in a frequency band where noise is large with respect to signal strength. That is, only the noise is emphasized, and the signal components represented by the numerator and the denominator of Expression 11 are both reduced, so that the calculation result may be inaccurate.
In order to eliminate this disadvantage, a Wiener-Helstrom filter may be used instead of the inverse filter. When the Wiener-Herstrom filter is used, Expression 12 corresponding to Expression 11 is established.
(Equation 12)
H (ω) = F (ω) ^* ・ G (ω) / (| F (ω) | ² + | N (ω) | ² ) (12)
F (ω) ^* Represents a complex conjugate, and N (ω) is a frequency spectrum of noise. Similarly to Expression 11, if H (ω) in Expression 12 is subjected to Fourier inverse transform, an input signal h (t) to be obtained can be calculated. Since h / t has a maximum S / N ratio statistically, a value that suppresses the influence of noise can be calculated even in a frequency band including noise.
[0050]
In the present embodiment, it is possible to separate and extract each pulse only from the device under test (silicon wafer) 20 by removing multiple reflection pulses caused by the terahertz light generator or the terahertz light detector. This can save the trouble of changing the thickness of the antenna substrates 7 and 11 of the terahertz light generator and the terahertz light detector according to the thickness of the device under test 20.
[0051]
In the present embodiment, the above-described first embodiment, that is, separation of multiple reflection pulses of a measurement time-series waveform related to transmitted light transmitted through the device under test 20 has been described. The present invention can also be applied to separation of multiple reflected pulses of a measurement time-series waveform related to reflected light.
[0052]
-Third embodiment-
FIG. 11 is a schematic configuration diagram schematically showing a spectrometer according to the third embodiment of the present invention. 11, elements that are the same as elements in FIG. 1 or that correspond to elements in FIG. 1 are given the same reference numerals, and overlapping descriptions are omitted.
[0053]
The difference between the spectrometer according to the present embodiment and the first embodiment described above is as follows. That is, in the first embodiment, the transmitted light of the DUT 20 is detected by the detector 11, whereas in the present embodiment, the reflected light of the DUT 20 is detected by the detector 11. And the operation of the control / arithmetic processing unit 23.
[0054]
In the above-described first embodiment, the measurement time-series waveform Esam (t) regarding the transmitted light is acquired for each measurement site of the DUT 20. In the present embodiment, a similar method is used. The measurement time series waveform Esam '(t) related to the reflected light is acquired for each measurement site of the device under test 20.
[0055]
The terahertz pulse light reflected by the device under test 20 includes various patterns of reflected light shown in FIGS. 12 (a) to 12 (d). FIG. 12A shows light that is reflected only on the incident surface (the surface on the detector 11 side) of the DUT 20 (reflected light that is not reflected once but internally and is not multiple reflection), and FIG. Is light reflected only once on the surface of the device under test 20 opposite to the detector 11 (reflected light that is reflected once inside but is not multiple reflection). FIG. The light reflected three times inside (light reflected multiple times once), and FIG. 12D is the light reflected (2k-1) times inside the DUT 20 (light reflected multiple times (k-1) times). ) Respectively. Here, for the sake of convenience, k is an integer of 0 or more, and the light shown in FIG. 12B is referred to as light that is multiply reflected 0 times, and the light shown in FIG. I do. In FIG. 12, the number of reflections indicates the number of reflections inside the DUT 20.
[0056]
The conceptual diagram of the reflected time-series waveform f (t) obtained by the terahertz time-domain spectroscopy is as shown in FIG. The measurement time-series waveform f (t) shown in FIG. 4 is considered to be the sum of the individual waves shown in FIGS. 9 (a) to 9 (d). That is, the wave of FIG. 9A representing the reflected pulse on the front surface of the parallel plate sample 20 is a (t), and the wave of FIG. 9B representing the reflected pulse on the back surface is b (t) for the first time. The wave in FIG. 9C representing the multiple reflection pulse is denoted by c (t), and the wave in FIG. 9D representing the second multiple reflection pulse is denoted by d (t). Each of the waveforms shown in FIGS. 9A to 9D is subjected to Fourier transform, and the sum of their absolute values (sum value) is obtained. Then, Fourier transform is performed on the reference light, and its absolute value (reference absolute value) is obtained. The spectrum of the parallel plate sample is analyzed by calculating the ratio between the total value and the reference absolute value of the detected pulse light.
[0057]
A method of analyzing the spectrum of the device under test 20 based on the measurement time-series waveform Esam '(t) employed in the third embodiment will be described in detail with reference to the flowchart in FIG. A sample having a known complex refractive index and a known thickness is arranged on the optical path between the generator 7 and the detector 11 instead of the object 20. Preferably, the reflectivity of this sample is approximately 100%. In this state, the reference time-series waveform Eref '(t) of the electric field intensity of the terahertz pulse light related to the reflected light is measured in the same manner as in the case of acquiring the measurement time-series waveform Esam' (t) (step S11). The measurement result is converted into digital data by the A / D converter 22 and stored in the memory 23A. The control / arithmetic processing unit 23 reads out the reference time-series waveform Eref ′ (t) stored in the memory 23A and executes a Fourier transform. Further, an absolute value (referred to as a reference absolute value) of the result of the Fourier transform is calculated and stored (step S12).
[0058]
With the device under test 20 arranged on the optical path between the generator 7 and the detector 11, the measurement time-series waveform Esam '(t) of the electric field intensity of the terahertz pulse light is measured and stored (step S13). . The measurement result is converted into digital data by the A / D converter 22 and stored in the memory 23A. This measurement time series waveform Esam '(t) is a time series waveform as shown in FIG. 4 as a conceptual diagram. The control / arithmetic processing unit 23 reads the measured time-series waveform Esam '(t) stored in the memory 23A, and converts the pulse waveforms shown in FIGS. 9A to 9D constituting the measured time-series waveform Esam' (t). Each is separated and extracted (step S14). Fourier transform is sequentially performed on the separated and extracted pulse waveforms to calculate their absolute values, and their totals (total values corresponding to equation (10)) are calculated (step S15). The ratio between the total value and the reference absolute value is calculated (step S16). Thereby, each pulse can be analyzed as incoherent.
[0059]
In order to confirm the effects of the present invention, a silicon wafer was actually measured for reflection and analyzed. FIG. 14A shows a reflection time-series waveform of a silicon wafer. From the graph of FIG. 14A, it can be seen that the measured waveform of the silicon wafer includes pulses up to the second multiple reflection. FIG. 14B is a graph showing the result of analyzing the measured waveform by a normal analysis method, and FIG. 14C is a graph showing the result of analysis using the present analysis method. From this, it can be seen that the use of the present analysis technique in actual measurement can also provide an analysis result that includes multiple reflections and has no interference pattern.
[0060]
Here, the incoherent analysis process is a process for obtaining a spectrum without an interference pattern after the analysis process. Therefore, the incoherent analysis processing is not limited to the above embodiment. For example, any method may be used as long as it separates the time-series detection pulse light from the parallel plate sample for each pulse and performs an analysis process based on the plurality of separated pulses and the reference pulse. As an example, a frequency analysis is performed on each separated pulse, a frequency analysis is performed on a reference pulse, and an analysis process is performed based on a frequency analysis result on each separated pulse and a frequency analysis result on a reference pulse.
[0061]
In the above description, Fourier transform is performed on each separated pulse, the sum of the absolute values of the Fourier transform results for each pulse (sum value) is calculated, Fourier transform is performed on the reference pulse, and the absolute value of the Fourier transform result is calculated. The value (reference value) was calculated, and the ratio of the total value to the reference value was calculated to obtain the spectrum of the detected pulse light. However, the following processing may be performed.
(1) Fourier transform is performed on each separated pulse, and the absolute value of the Fourier transform result for each pulse is calculated.
(2) Fourier transform is performed on the reference pulse, and the absolute value (reference value) of the Fourier transform result is calculated.
(3) The spectrum of the detected pulse light is obtained by calculating the ratio between the absolute value of the separation pulse and the reference value. In this case, since a plurality of calculated ratios are obtained, a total sum of them may be used, or individual ratios may be used.
[0062]
The present invention can be applied to spectroscopic measurement of various pulses other than the terahertz pulse light.
[0063]
【The invention's effect】
According to the present invention, by performing incoherent analysis processing on each pulse in the measurement result, it is possible to obtain a spectrum that does not include an interference pattern even when multiple reflections are included.
In addition, according to the present invention, by taking advantage of the characteristics of pulse light used as a light source in terahertz time-domain spectroscopy, incoherent analysis processing is performed on each pulse of a measurement result. Even when a pulse reflected from the back surface is included, a spectrum that does not include an interference pattern can be obtained.
Furthermore, according to the present invention, since the influence on the measurement result due to multiple reflections caused by things other than the parallel plate sample is removed, incoherent analysis processing can be performed on each pulse of the measurement result. it can.
[Brief description of the drawings]
FIG. 1 is a schematic configuration diagram showing a transmission type terahertz spectrometer according to a first embodiment of the present invention.
FIG. 2 is a diagram illustrating transmitted pulses of a parallel plate sample according to the first embodiment of the present invention.
FIG. 3 is a conceptual diagram of a transmission time-series waveform.
FIG. 4 is a conceptual diagram of a reflected time-series waveform;
FIG. 5 is a diagram showing an analysis result of transmittance or reflectance including an interference pattern;
FIG. 6A is a diagram for explaining transmission and reflection at interfaces having different refractive indexes.
FIG. 6B is a diagram illustrating transmitted light and reflected light at each multiple reflection number.
FIG. 7 is a diagram showing a calculation result of a complex amplitude of transmitted light including infinite multiple reflections in the case of coherent and incoherent cases.
FIGS. 8A to 8D are conceptual diagrams of separated waveforms when a transmitted time-series waveform is treated as incoherent.
FIGS. 9A to 9D are conceptual diagrams of separated waveforms when a reflected time-series waveform is treated as incoherent.
FIG. 10 is a schematic flowchart showing the operation of a control / arithmetic processing unit of the transmission type terahertz spectrometer according to the first embodiment of the present invention;
FIG. 11 is a schematic configuration diagram showing a reflection type terahertz spectrometer according to a third embodiment of the present invention.
FIG. 12 is a view for explaining reflected pulses of a parallel plate sample according to the third embodiment of the present invention.
FIG. 13 is a schematic flowchart showing an operation of a control / arithmetic processing unit of the reflection type terahertz spectrometer according to the third embodiment of the present invention.
FIG. 14 (a) is a reflection time-series waveform diagram of a silicon wafer actually measured.
FIG. 14B shows an analysis result of a silicon wafer obtained by a normal analysis method.
FIG. 14C shows an analysis result of a silicon wafer obtained by the present analysis method.
FIG. 15A is an ideal transmission time-series waveform diagram that does not include multiple reflections at all by the antenna substrates 7 and 11;
FIG. 15B is a transmission time series waveform B obtained without disposing the parallel plate sample 20, that is, a reference time series waveform chart.
FIG. 15C shows a realistic transmission time-series waveform C obtained by disposing the parallel plate sample 20, that is, a measurement time-series waveform diagram.
[Explanation of symbols]
1: Femtosecond pulse light source 7: Terahertz light generator
9: movable mirror 11: terahertz light detector
20: Parallel plate sample 23: Control / arithmetic processing unit

Claims (18)

In a spectroscopic measurement method using pulsed light,
Irradiate the parallel plate sample with pulsed light,
Detecting detection pulse light reflected or transmitted from the parallel plate sample irradiated with the pulse light,
A spectroscopic measurement method using pulsed light, wherein incoherent analysis processing is performed on the coherent detection pulsed light including a pulse that is multiply reflected inside the parallel plate sample to obtain a spectrum that does not include an interference pattern. . In a spectroscopic measurement method using terahertz pulse light,
Irradiate the terahertz pulse light to the parallel plate sample,
Detecting the detection pulse light reflected or transmitted from the parallel plate sample irradiated with the terahertz pulse light,
Spectroscopic measurement using terahertz pulse light, wherein an incoherent analysis process is performed on the coherent detection pulse light including a pulse reflected multiple times inside the parallel plate sample to obtain a spectrum that does not include an interference pattern. Method. In the spectral measurement method according to claim 2,
The incoherent analysis processing includes:
The time-series detection pulse light from the parallel plate sample is separated for each pulse,
A spectral measurement method using terahertz pulse light, which is an analysis process based on a plurality of separated pulses and a reference pulse. In the spectral measurement method according to claim 2,
The incoherent analysis processing includes:
The time-series detection pulse light from the parallel plate sample is separated for each pulse,
Perform frequency analysis on each separated pulse,
Perform frequency analysis on the reference pulse,
A spectral measurement method using terahertz pulse light, which is an analysis process based on a frequency analysis result for each separated pulse and a frequency analysis result for a reference pulse. In the spectral measurement method according to claim 2,
The time-series detection pulse light from the parallel plate sample is separated for each pulse,
Fourier transform is performed on each pulse, and the sum of the absolute values (sum value) of the Fourier transform result for each pulse is calculated.
Perform a Fourier transform on the reference pulse, calculate the absolute value (reference value) of the Fourier transform result,
A spectral measurement method using terahertz pulsed light, wherein a spectrum of the detected pulsed light is obtained by calculating a ratio of the total value to a reference value. In any one of claims 2 to 5,
Each of the thickness of the substrate of the terahertz pulse light generator, the thickness of the substrate of the terahertz pulse light detector, and the thickness of the parallel plate sample is determined by multiple reflection by each substrate and multiple reflection by the parallel plate sample. A spectroscopic measurement method using terahertz pulsed light, characterized in that it is determined to be separated. In a spectrometer using pulsed light,
Irradiation means for irradiating the parallel plate sample with pulsed light,
A detector for detecting a detection pulse light reflected or transmitted from the parallel plate sample irradiated with the pulse light,
Analyzing means for obtaining a spectrum that does not include an interference pattern by performing an incoherent analysis process on the coherent detection pulse light including a pulse that is multiply reflected inside the parallel plate sample. A spectrometer using pulsed light. In a spectrometer using terahertz pulsed light,
A terahertz pulse light generator that irradiates a terahertz pulse light to a parallel plate sample, and a terahertz pulse light detector that detects a detection pulse light reflected or transmitted from the parallel plate sample irradiated with the terahertz pulse light,
Analyzing means for obtaining a spectrum that does not include an interference pattern by performing an incoherent analysis process on the coherent detection pulse light including a pulse that is multiply reflected inside the parallel plate sample. A spectrometer using terahertz pulse light. The spectrometer according to claim 8,
The analysis means,
Separation means for separating the time-series detection pulsed light from the parallel plate sample for each pulse,
A spectrometer using terahertz pulsed light, comprising: processing means for obtaining a spectrum that does not include an interference pattern based on a plurality of separated pulses and a reference pulse. The spectrometer according to claim 8,
The analysis means,
Separation means for separating the time-series detection pulsed light from the parallel plate sample for each pulse,
First frequency analysis means for performing frequency analysis on each of the separated pulses;
Second frequency analysis means for performing frequency analysis on the reference pulse;
A spectrometer using terahertz pulse light, comprising: processing means for obtaining a spectrum that does not include an interference pattern based on a frequency analysis result for each separated pulse and a frequency analysis result for a reference pulse. The spectrometer according to claim 8,
The analysis means,
Separation means for separating the time-series detection pulsed light from the parallel plate sample for each pulse,
First calculating means for performing Fourier transform on each pulse, and calculating the sum of the absolute values of the Fourier transform results (sum value) for each pulse;
Second calculating means for performing Fourier transform on the reference pulse and calculating an absolute value (reference value) of the Fourier transform result;
A third calculating means for calculating a ratio of the sum value and the reference value to obtain a spectrum that does not include an interference pattern, wherein the spectrometer uses terahertz pulsed light. The spectrometer according to any one of claims 8 to 11,
Each of the thickness of the substrate of the terahertz pulse light generator, the thickness of the substrate of the terahertz pulse light detector, and the thickness of the parallel plate sample is determined by multiple reflection by each substrate and multiple reflection by the parallel plate sample. A spectrometer using terahertz pulsed light, characterized in that it is determined to be separated. Detecting a detection pulse light reflected or transmitted including a pulse multiply reflected inside the parallel plate sample from the parallel plate sample irradiated with the pulse light, obtaining a measurement time-series waveform by the detection pulse light,
Pulse light is emitted by irradiating pulse light in a state where a perfect reflection mirror is provided in place of the parallel plate sample, and pulse light is detected by being reflected from the perfect reflection mirror or in a state where the parallel plate sample is not provided. Irradiates the reference time series waveform with the pulse light detected directly,
Analysis of spectrometry data using pulsed light, wherein deconvolution of the measured time-series waveform and the reference time-series waveform is obtained, and a time-series waveform obtained by separating each of multiple reflected pulse lights on a time axis is obtained. Method. The method for analyzing spectral measurement data according to claim 13,
The deconvolution is:
Perform a predetermined operation after performing the Fourier transform of the measured time series waveform and the reference time series waveform,
A method of analyzing spectral measurement data using pulsed light, comprising a step of performing an inverse Fourier transform of the operation result. In the method for analyzing spectral measurement data according to claim 14,
The predetermined operation is an operation using an inverse filter when the time series waveform only from the parallel plate sample is an input signal, the measurement time series waveform is an output signal, and the reference time series waveform is regarded as a device function. A method for analyzing spectroscopic measurement data using pulsed light, characterized in that: In the method for analyzing spectral measurement data according to claim 14,
The predetermined operation is a Wiener-Hellstrom filter when the time series waveform only from the parallel plate sample is an input signal, the measurement time series waveform is an output signal, and the reference time series waveform is regarded as an apparatus function. A method of analyzing spectroscopic measurement data using pulsed light, wherein the method is an operation using a pulse light. In the method for analyzing spectral measurement data according to any one of claims 13 to 16,
A method for analyzing spectroscopic measurement data using pulsed light, wherein terahertz pulsed light is used as the pulsed light. The incoherent signal according to any one of claims 1 to 12, wherein the time-series waveform obtained by separating each of the multiple reflection pulse lights on the time axis obtained by the method for analyzing spectral measurement data according to any one of claims 13 to 17 is applied. A spectral measurement method using pulsed light, wherein an analysis process is performed to obtain a spectrum that does not include an interference pattern.

Images