JP4642276B2 - Semiconductor device manufacturing method and recording medium - Google Patents

Description

によって表される。ただし、Ｃｍは、ジシラン分圧Ｐ_Si，ゲルマン分圧Ｐ_Ge及び基板温度Ｔと、組成比との関係を表す関数である。
【００６２】
また、式（２）における変数ｔ（ｄ）は、結晶成長の開始時刻ｔを０として、下記式（３）

により求めることができる。ただし、ｇｒは、ガス分圧及び基板温度と成長速度との関係を表す関数である。
【００６３】
したがって、関数Ｃｍ及びｇｒが既知で、結晶成長の開始から終了までの任意の時刻のジシラン及びゲルマンの分圧及び基板温度がわかれば、Ｇｅ組成比（組成率）の深さ方向のプロファイルを規定する値ｘ（ｄ）を求めることができる。つまり、プロセス条件からサンプル構造を推定することができる。なお、具体的に、これらの関係式（２），（３）は、成長条件を変えてサンプルを作成することによって求めることができる。
【００６４】
また、関係式（２），（３）は、成長機構等の情報の蓄積によって、より高精度のものに置き換えることができる。また、構造に関しても、対象となる要素の構造すべてではなく、例えばＧｅ組成率のプロファイルなどの特定の構造の評価に必要な程度の精度さえ得られればよいので、式（２），（３）よりも簡単化された近似式を用いることもできる。
【００６５】
なお、上記式（２），（３）を用いて連立方程式を解いて、解析的にＰ_Si（τ），Ｐ_Ge（τ），Ｔ（τ）からｘ（ｄ）を求めてもよいが、式（２），（３）自体が複雑なために、解が求められない場合もあり得る。
【００６６】
そこで、関係式（２），（３）から各変数の値を求めて行く数値計算的取り扱いを行なうことも可能であり、この数値計算的取り扱いを行なった方が簡便ではある。つまり、解析的取り扱いをすれば、任意の深さにおける膜の組成比（組成率）を計算することができるが、実用上は、組成の深さ方向におけるプロファイルから分光エリプソメトリ法を用いて得られるΔ，Ψのスペクトルの理論予想値を計算する過程において、信頼しうる予想値を得るために必要な程度の精度で、組成の深さ方向におけるプロファイルが求まればよい。
【００６７】
そこで、計算を簡略化するために、結晶成長プロセスの開始から完了までを、ステップ内では成長条件が変化しないとみなせる基本ステップの連続とみなし、各ステップにおけるガス圧力と基板温度と各ステップの時間長さとを与えれば、関数Ｃｍ及びｇｒの関係式から各ステップで成長された膜の組成と厚み（この場合は、ステップを行なう時間と成長速度との単純な積になる）とが求まり、組成のプロファイルを求めることができる。
【００６８】
実際、ＳｉＧｅ傾斜組成層を有するＨＢＴ構造のように、近似的に連続的に組成が変化するプロファイルを有する膜をエピタキシャル成長させる場合にも、現実にソースガスの分圧を近似的に連続的に変化させるのは困難であるために、短いステップごとにガス分圧を変化させる方法を採用することがほとんどである。その場合、各ステップ内では、プロセス条件は通常一定とみなすことができるので、各ステップごとの組成と成長速度・膜厚を計算する方法の方が実用的である。
【００６９】
上述のように、結晶層の成長開始から成長終了までのプロセス条件（ガス分圧，基板温度）がわかれば、サンプル構造を推定することができる。このように、プロセス条件からサンプル構造を推定するプログラムは、一般にはプロセスシミュレータと呼ばれている。したがって、本明細書においても、プロセスシミュレータをこのプログラムの意味で用いることにする。
【００７０】
−製造プロセスの基本手順−
図１は、本発明の実施形態におけるサンプル評価及び薄膜の製造プロセスにおける管理の手順を示すフローチャートである。
【００７１】
まず、ステップＳＴ１０１で、１ロット５０枚のウェハの製造プロセス（プロセスＰ）を行なうことを想定した場合に、最初の１枚のウェハ（評価用ウェハ）を、評価方法Ｍ（例えば分光エリプソメトリ法）により測定し、物理量の実測値（例えばΔ，Ψスペクトル）を得る。
【００７２】
一方、ステップＳＴ１０２で、プロセス条件の初期推定値を設定し、ステップＳＴ１０３で、プロセスシミュレータによるサンプル構造の推定を行なった後、ステップＳＴ１０４で、物理量測定値の理論予想値を計算する。
【００７３】
そして、ステップＳＴ１０５で、物理量の実測値と理論計算値とを互いに比較する。このときに、実測値と理論予想値との差の程度を評価するための評価値を計算する。
【００７４】
次に、ステップＳＴ１０６で、評価値が極小値か否かを判別し、評価値が極小でなければステップＳＴ１０７で新しい推定値の設定を行なってからステップＳＴ１０３の処理に戻って、ステップＳＴ１０４〜ＳＴ１０６の処理を繰り返す。
【００７５】
そして、ステップＳＴ１０６の判別において評価値が極小であると判別されると、ステップＳＴ１０８に進んでサンプル構造の推定値を決定した後、ステップＳＴ１０９で、サンプル構造が適正範囲か否かを判別する。この判別の結果、サンプル構造が適正範囲内にあれば、ステップＳＴ１１０に進んで、設定したプロセス条件のままで、次の処理を行なう。
【００７６】
一方、ステップＳＴ１０９における判別の結果、サンプル構造が適正範囲内にない場合には、ステップＳＴ１１１に移行して、ステップＳＴ１０８で推定したプロセス条件のうち，適正範囲からはずれているものを補正して、評価用ウェハ以外の他のウェハのプロセスを行なう。ただし、評価用ウェハについても再度プロセスを行なってもよい。
【００７７】
−フローチャートの各ステップの説明−
以下、以上の処理の具体的な内容を、傾斜組成ＳｉＧｅ−ＨＢＴ構造の評価を行う場合を例にとって説明する。
【００７８】
ステップＳＴ１０１の物理量の実測値は、光学的評価方法を用いるものとし、本実施形態においては、分光エリプソメトリ法を用いる。本発明は、例えば分光反射率測定法などの他の光学的評価方法を用いても実施が可能である。特に、物理量としては、光の波長を変数としたときに変数が複数の実数を採りうるものであることが好ましい。複数の実測値が得られないと、最小自乗法などを用いた実測値と推定値とのフィッティング処理を高い精度で行なうことができないからである。
【００７９】
本実施形態においては、エピタキシャル成長されたＳｉＧｅ膜を有するサンプルを、分光エリプソメータ法により測定し、複数の波長でのΨ，Δのスペクトルを得ておく。ＳｉＧｅ膜の光学的定数のＧｅ組成率依存性は、短波長領域で顕著であるが、波長の短い光はＳｉに吸収されやすいので、Ｇｅ組成率のプロファイル評価を行うには紫外から可視光領域にわたって測定を行っておくことが好ましい。
【００８０】
図２は、分光エリプソメトリ法による測定の結果、得られるΔ，Ψスペクトルの例を示す図である。上述のように、スペクトルのサンプルに直線偏光した光を照射して、得られた反射光において、電場ベクトルが入射面に平行となっているものをｐ偏光成分、垂直となっているものをｓ偏光成分とし、それぞれの複素反射率をＲｐ, Ｒｓとした場合、ρ≡Ｒｐ／Ｒｓはやはり複素数となる。従って、ρは二つの実数Ψ，Δを用いてρ≡tan ΨeiΔと表現することができる。図２は、このΨ，Δの二物理量を各波長の光について測定して得られたスペクトルである。
【００８１】
次に、ステップＳＴ１０２では、プロセス条件の初期値を推定する。ここで、プロセス条件の設定時には、基板温度と、ジシランおよびゲルマンの流量の３つの値を指定すれば、成長速度とＧｅ組成率が求まるので、傾斜組成を有するＳｉＧｅ膜を有する積層膜のような希望するＧｅ組成率の台形プロファイルが得られるように、基板温度およびガスの流し方が決定されている。したがって、まず、所望のＧｅ組成率のプロファイル構造を設定し、このプロファイル構造が得られるプロセス条件を定めておく必要がある。
【００８２】
図３（ａ）〜（ｄ）は、それそれ順に、傾斜組成ＳｉＧｅ−ＨＢＴの積層膜における設計されたＧｅ組成率のプロファイル，条件がはずれた場合のプロファイル，プロセスシミュレータを用いた条件の推定用プロファイル，補正後のサンプル構造のプロファイルを示す図である。
【００８３】
図３（ａ）に示すように、傾斜組成ＳｉＧｅ−ＨＢＴの場合、すでに説明したように、ベース層にキャリアを加速するための内部電界を発生させるために、台形のＧｅ組成率のプロファイルを設計する。本実施形態では、ＳｉＧｅバッファ層の厚みを４０ｎｍとし、ＳｉＧｅ傾斜組成層の厚みを４０ｎｍとし、Ｓｉキャップ層の厚みを５０ｎｍとして、合計１３０ｎｍの積層膜をＳｉ基板上にエピタキシャル成長させるものとする。そして、ＳｉＧｅバッファ層のＧｅ組成率を均一な１５％とする。ＳｉＧｅ傾斜組成層のＧｅ組成率を、ＳｉＧｅバッファ層に隣接する部分では１５％とし、Ｓｉキャップ層に隣接する部分では０％として、ＳｉＧｅ傾斜組成層中では、Ｇｅ組成率を近似的に直線的（正確には段階的）に変化させる。
【００８４】
図４は、図３（ａ）に示す台形プロファイル構造を得るためのプロセス条件を示す図である。通常、制御を容易にするために、成長中は基板温度Ｔを一定とし、ジシラン（Ｓｉ₂ Ｈ₆ ）の流量を一定とし、ゲルマン（ＧｅＨ₄ ）の流量のみ変調をかけることにより、流量比を制御している。つまり、ゲルマンの流量を段階的に減少させることにより、ＳｉＧｅ膜のＧｅ組成率を段階的に変化させるのである。この時、ゲルマン流量を減らしていけばＧｅ組成率を下げることができるが、同時に成長速度が低下してくるので、求めるプロファイルとなるようにゲルマンの流量変調をおこなう。
【００８５】
図５は、ステップＳＴ１０３におけるプロセスシミュレータによるサンプル構造の推定手順を示すフローチャートである。
【００８６】
まず、ステップＳＴ１５０で、エピタキシャル成長プロセスの開始から終了までを、各ステップ内ではプロセス条件が一定とみなせる基本ステップの流れとして表し、ステップＳＴ１５１で、各ステップにおけるガス圧力（もしくは流量），基板温度及びステップの時間長さを与える。具体的には、図４に示すようなプロセス条件の流れに沿って、プロセス条件を構成するパラメータ（この例では、ゲルマンの流量だけ）が変化する時刻ｔ０，ｔ１，ｔ２，ｔ３，…を境界として、複数の基本ステップを定める。
【００８７】
次に、ステップＳＴ１５２で、原料ガスの圧力比（又は流量比）から、そのステップで成長する結晶層の組成（Ｇｅ組成率）を計算する。組成が基板温度依存性を持つ場合には、それも考慮して計算する。つまり、上述の式（２）に基づいて各ステップにおけるＧｅ組成率ｄ（ｔ）を求めることに対応する。
【００８８】
そして、ステップＳＴ１５２で、原料ガスの圧力及び基板温度から、そのステップで成長する結晶層の厚み（成長速度と時間の積）を計算し、ステップＳＴ１５４で、すべてのステップにおける層の組成，厚みの計算結果から最終的にできあがる構造を計算する。
【００８９】
図６は、プロセスシミュレータによって推定されたＧｅ組成率のエピタキシャル層厚み方向におけるプロファイルを示す図である。同図に示すプロファイルは、輪郭をみるとやや上に凸の形状を有していることがわかる。
【００９０】
図７は、ＳｉＧｅ層におけるＧｅ組成率のゲルマン流量比依存性を示す図である。Ｇｅ流量比とは、ジシラン（Ｓｉ₂ Ｈ₆ ）とゲルマン（ＧｅＨ₄ ）との合計流量に対するゲルマンの流量比である。ただし、ＳｉＧｅ層の成長速度は、基板温度およびガス流量に依存する。ＵＨＶ−ＣＶＤ法の場合、基板温度が比較的低温であるために、ある一定量以上の原料ガスを流しても、基板表面での反応が追いつかず成長速度がガス流量に依存しなくなる反応律速領域に容易に突入する。
通常のプロセスはこの反応律速領域でのみ行っている。
【００９１】
反応律速領域では、成長速度の温度依存性はいわゆる熱励起過程型となり、成長速度をｇ、基板温度をＴ，活性化エネルギーをＥａとした場合、反応速度ｇと温度Ｔとの間には下記
ｇ∝ exp（−Ｅａ／Ｔ＊ｋ）
の関係がある。ただし、ｋはボルツマン定数である。
【００９２】
ここで、ＳｉＧｅ層のエピタキシャル成長においては、活性化エネルギーＥａがＧｅ組成率依存性をもつという特徴があり、Ｇｅ組成率が高くなるほど活性化エネルギーＥａの値が小さくなる。その結果、同一の温度ではＧｅ組成率が高くなるほどＳｉＧｅ層の成長速度が増大し、成長速度の基板温度依存性が小さくなる。
【００９３】
図８は、ＳｉＧｅ層の成長速度（ｎｍ／ｍｉｎ）のゲルマン流量比依存性を示す図である。図８においては、基板温度５８７℃の時の成長速度とゲルマンの流量比との関係が示されている。同図に示されるように、ゲルマン（ＧｅＨ₄ ）の流量比が高くなるほど成長速度が大きくなっている。なお、６００℃近傍においては、基板温度が１℃（１degree）変化すると、Ｓｉ層（Ｇｅ流量比が０の点）の場合には３％程度の成長速度の変化が起こるのに対し、Ｇｅ組成率１５％のＳｉＧｅ層では成長速度変化は約２％程度にとどまることがわかっている。
【００９４】
以上のように、プロセス条件として、基板温度、ジシラン流量、ゲルマン流量の３つを設定し、これらの３つの量の成膜プロセス中の任意の時間における値を指定すればできあがる構造を計算することができる。ただし、実際のプロセスにおいて、プロセス条件が設定条件からずれているとき、ガス流量が設定値からずれていることはまれで、ほとんどの場合、基板温度が設定値からずれている。また、基板温度は時間とともに変化することはほとんどないので、構造推定のパラメータとしては、基板温度を一定の値の変数とする１つの変数を設定するだけで十分なことが多い。 もちろん、状況によっては、ガス流量がずれる場合や、基板温度が時間とともに変化するような場合もあるので、かかる場合には、それらの値もフィッティングパラメータとすることができる。このように、プロセス条件から構造を推定する方法を用いれば、実際に作成され得る構造をすべてパラメータで表現することが可能になる。
【００９５】
なお、プロセス条件から構造を推定する方法としては、先に述べたような一定の成長温度と、一定のガス流量比と、成長速度と、Ｇｅ組成率との関係を明らかにしておき、任意の成長条件における構造を計算する方法をとることができるが、この方法とは別に、あらかじめプロセス条件を幾通りかに分けてサンプルを作成しておき、その時にできあがる構造とプロセス条件値の関係を求めておいて、サンプルを作成した以外の条件におけるサンプル構造を補間により求める方法をとることもできる。プロセスパラメータの数が少ない場合にはこの方法が簡便である。
【００９６】
次に、ステップＳＴ１０４では、図８に示すＧｅ組成率のプロファイル構造に基づいて、分光エリプソメトリ法により、このような構造を有する積層膜の光学的評価を行うと得られるであろうΔ，Ψのスペクトル（物理量の理論予想値）を計算する。つまり、Ｇｅ組成率のプロファイルに応じて光学定数の深さ方向におけるプロファイルが計算でき、その光学定数のプロファイルに応じて、この積層膜の分光エリプソメトリ法又は分光反射率測定法による測定値の理論予想値を計算することができる。
【００９７】
この理論予想値の具体的な計算方法については、Azzam et. al. による“Ellipsometry and Polarized Light (Elsevier Science Ltd Published 1987 ）”に詳述されている。その計算原理は、サンプルを、層内では光学定数が一定である多数の層からなる積層膜であるとして、各層における光の伝播及び各界面での反射率及び透過率を各層の光学定数から求めることにある。つまり、各層の組成と光学定数（屈折率ｎ，消光係数ｋ）との関係が既知であれば、プロセスシミュレータで推定した積層膜の構造を光学的に評価したときの物理量の測定値を予想することができる。
【００９８】
ＳｉＧｅ又はＳｉＧｅＣの組成と光学定数との関係は、例えば論文“R.T.Carline et.al. Appl.Phys.Lett. 64 No.9 p.1114-1116,1994 ”に記載されており、任意の組成から光学定数を計算するアルゴリズムについては、例えば論文“Snyder et.al. Appl.Phys. 68 No.11 p.5925-5926,1990 ”に記載されている。
【００９９】
したがって、プロセスシミュレータにプロセス条件値を入力すれば、サンプル構造を計算することができ、そのサンプル構造から得られる分光エリプソメトリ法による測定値の理論予想値を計算することができる。
【０１００】
図９は、本発明者達が図６に示すＧｅ組成率のプロファイルを有する積層膜からシミュレーションを行なったΔ，Ψのスペクトル（理論予想値）である。なお、図９に示すスペクトルのシミュレーションにおいては、用いたエリプソメトリ分光装置固有の装置定数を考慮に入れている。したがって、同じ構造の積層膜について、他のエリプソメトリ分光装置を用いて測定することを前提とする場合には、図９に示すと同じスペクトルが得られるとは限らない。
【０１０１】
次に、ステップＳＴ１０５では、図２に示す実際の測定値と図９に示す理論予想値とを比較する。そして、例えば式（１）に示す分散値σを評価値として求め、ステップＳＴ１０６で評価値が極小か否かを判別して、評価値が極小になるまではステップＳＴ１０７からステップＳＴ１０３に戻る処理を繰り返す。このとき、ステップＳＴ１０７の処理では、新しい推定値として、基板温度Ｔのみを図４に示す６００℃から例えば１℃だけ変化させたプロセス条件を設定し、このプロセス条件に基づいてステップＳＴ１０３におけるサンプル構造の推定を行なう。
【０１０２】
プロセス条件としては、基板温度，ジシラン流量，ゲルマン流量の３つを設定しているので、ステップＳＴ１０７においても、これらの３つの量の成膜プロセス中の任意の時間における値を新たに設定することもできる。ただし、上述のように、実際のプロセスにおいて、プロセス条件のうち設定からずれるのはほとんどの場合基板温度Ｔであるので、この例では、構造推定のパラメータとして、基板温度Ｔの１変数だけを用いる。
【０１０３】
そして、急速降下法を用いる場合には、ステップＳＴ１０７の処理で、基板温度Ｔを変化させたときの評価値の微分係数に基づいて、評価値が極小に達する確率がもっとも大きいであろうと思われる方向に、基板温度Ｔの値が補正され、評価値が極小に収束することになる。
【０１０４】
次に、ステップＳＴ１０８では、ステップＳＴ１０１で求められた物理量（図２に示すΔ，Ψのスペクトル）に基づいて、もっとも確からしいプロセス条件の推定値が求められる。そして、ステップＳＴ１０９の判別で、プロセス条件が適正範囲（例えば６００℃±０．５℃）からはずれている場合には、ステップＳＴ１１１で、その条件を補正する。つまり、ステップＳＴ１０９からＳＴ１１０又はＳＴ１１１に進む処理により、プロセス条件を修正するか否かを決定することになる。
【０１０５】
例えば、図２に示される物理量の実測値が、基板温度Ｔが６００℃よりも２℃低い条件でエピタキシャル成長された積層膜から得られたものであると推定された場合、基板温度Ｔを規定するパラメータを基板温度Ｔが２℃高くなるように変更して、以後のウェハについてプロセスを行なう。これにより、異常原因に対し適切な対処が行えるために、次回からは求めるプロファイルを持った構造を形成することができる。
【０１０６】
−実施形態の効果−
実際のトランジスタを作成するための結晶成長プロセスの場合、結晶成長までにすでに基板内には絶縁膜やポリシリコン層などが形成されている。このようにパターンがすでに形成された基板の場合、熱輻射率などの違いから、パターンが形成されていないＳｉ基板と同じ加熱の仕方を行っても同じ温度にはならない。また、同じ種類のパターンを有する基板でも、絶縁膜層の厚みの工程ばらつきにより成長温度は変化する。従って、実際に評価しなければならないサンプルは、温度がずれた状態で成膜された，Ｇｅ組成率のプロファイルに曲線部分をもつものとなっていることが多い。また、基板温度を面内で均一とすることは非常に困難であり、基板面内で温度がそれぞれ違った状態で成膜された部分が形成されることになる。その場合、従来の方法では、図１１（ｃ）に示すような，実際には積層膜が有しないであろうと考えられる構造を前提として、分光エリプソメトリ法などによる物理量の測定値の理論予想値を推定している。よって、図１１（ｄ）に示すように、設計からずれた物理的構造（例えば、Ｇｅ組成率のプロファイル構造）しか得られないことになる。
【０１０７】
それに対し、本実施形態によると、以下のように、実際に起こりうる構造を予想して、温度が標準条件からずれていることを把握し、それに基づく条件補正が可能になる。
【０１０８】
たとえば、図３（ａ）に示す台形プロファイル構造を設計値としてプロセス条件を設定したところ、図３（ｂ）に示すプロファイル構造が得られている場合に、ステップＳＴ１０７からステップＳＴ１０３に戻ってステップＳＴ１０３〜ＳＴ１０６の処理を繰り返す段階で、現実に積層膜が有しているであろうもっとも確からしいプロファイル構造が得られる。
【０１０９】
つまり、図３（ｃ）に示すように、基板温度Ｔが設定値よりも高いときに発生する，上に凸の形状か、基板温度Ｔが設定値よりも低いときに発生する，下に凸の形状かという情報を含め、積層膜中の各層（ＳｉＧｅバッファ層，ＳｉＧｅ傾斜組成層及びＳｉキャップ層）の組成及び膜厚に関する情報が得られる。そして、その情報に基づいてプロセス条件を補正するので、以後のプロセスにおいては、図３（ｄ）に示すようなほぼ設計通りの台形プロファイルを有する積層膜を形成することができる。
【０１１０】
言い換えると、本実施形態では、実際に積層膜が有していると推定されるプロファイル構造を前提として、分光エリプソメトリ法などによる物理量の測定値の理論予想値を推定する。その結果、設計通りの構造（ここでは、Ｇｅ組成率の台形プロファイル構造）を得ることができる。
【０１１１】
すなわち、半導体デバイスの特性の向上及び特性のバラツキの低減を図ることができる。
【０１１２】
（その他の実施形態）
なお、評価値の計算のアルゴリズムとしては、基本的に従来の方法のアルゴリズムを用いることができ、た式（１）に示す分散値以外の評価値を用いることもできる。また、極小値の判定アルゴリズムとして、急速降下法に代えて、他の既存の極小値判定のためのアルゴリズムを用いることができる。
【０１１３】
また、ステップＳＴ１０９の判別処理は、デバイス構造（Ｇｅ組成率のプロファイル構造）の推定値が許容範囲内にあるかどうかの判断に代えてもかまわない。
【０１１４】
さらに、評価装置と成膜処理を行なう製造装置間に通信機能をもたせ、評価結果に基づいて、自動的に次のプロセスを補正するようにしてもよい。これは、評価装置を製造装置と共通の筐体内に収納するような構成をとった場合に、特に有効である。
【０１１５】
このようなプロセスシミュレーションによる構造推定と、物理量の理論予想値と実測値の比較をおこなうフィッティング手順は、物理量の実測値を測定する装置と別の装置でおこなうことも可能ではあるが、測定器内にこのフィッティング手順をおこなうアルゴリズムを格納したハードディスクを内蔵したコンピュータを配備しておくと、物理量の測定から最終サンプル構造の推定まで一貫しておこなえるので好ましい。
【０１１６】
また、図１に示すフローチャートに示す処理は、その手順を記録媒体に記録しておくことにより自動的に行なうことができる。例えば、ステップＳＴ１０２〜ステップＳＴ１０７の手順をコンピュータで読みとり可能な記録媒体にプログラムとして記録しておくことにより、制御対象の構造が形成されたプロセス条件を自動的に推定することができる。記録媒体としては、例えば、ＣＤ−ＲＯＭ，磁性体を利用した磁気テープ，ＦＤなどの他、ＥＥＰＲＯＭなどの不揮発性メモリ、ＤＶＤなどがあり、いずれを用いてもよい。
【０１１７】
なお、本発明の実施に際し、サンプル構造のすべての部分をプロセスシミュレータで行う必要は無い。例えば傾斜組成ＨＢＴ構造の場合、ＳｉＧｅバッファ層部分や、Ｓｉキャップ層のように均一組成の膜となるような部分は、従来のように膜厚や組成率を直接パラメータとして用い、ＳｉＧｅ傾斜組成層のみプロセスシミュレータを用いたほうが簡便となる場合がある。この場合でも、従来の方法に比べ、ＳｉＧｅ傾斜組成層のプロファイルを適切に表現できるようになるために、評価精度は従来の方法に比べ格段に向上する。
【０１１８】
上記実施形態では、ＵＨＶ−ＣＶＤ法によるＳｉＧｅ傾斜組成ＨＢＴ構造を作成し、分光エリプソメトリ法で評価する場合を示したが、本発明はこの組み合わせに限定されるものではない。たとえば分光反射率測定法と分光エリプソメトリ法とは共通点が多いので、図１のフローチャートに示す手順のうち、実測し理論予想値を計算する物理量を、Ψ，Δから反射率に置き換えるだけでほとんど同じように実施することができる。それ以外の評価方法でも、サンプル構造を推定し、物理量の理論予想値の計算さえできれば本発明の実施は可能である。
【０１１９】
また、ＵＨＶ−ＣＶＤ法でなくても、プロセス条件とそのプロセス条件を用いると形成されるであろう構造との関係が明らかになっており、プロセスシミュレータが開発できるようなプロセスであれば、本発明の実施が可能である。例えば、ＬＰ−ＣＶＤ法、ＭＢＥ法によるエピタキシャル成長などでは容易に実施可能である。また、不純物拡散や活性化、シリサイド形成のためのＲＴＡ処理ではプロセス条件が、温度と時間といった非常に少数のパラメータで制御されるのに対し、形成される構造は非常に複雑なプロファイルをもつので、本発明の実施に適している。
【０１２０】
また、酸化膜や窒化膜といった単純な構造をもつ膜でも、本発明の方法によって補正すべきプロセス条件値が容易に求まるので、本発明を適用することにより、プロセスの改善効果を得ることができる。
【０１２１】
ただし、本発明は、特に複数の元素によって構成され、しかもその組成が変化する構造を有するものに適用することに、特に意義がある。このような構造の場合には、組成が変化すると成長状態（例えば成長速度）もそれに応じて変化することが多いので、従来の方法では正確な構造を把握したり、正確な条件補正を行なうことが困難である。それに対し、本発明のごとく、プロセス条件を変化させたときに実際に起こりうる物理量（構造）を把握してから、これに基づいて物理量の測定値を予測することにより、複雑に変化する膜などの構造をほぼ設計通りに制御することが可能になる。かかる複数の元素によって構成される膜の例としては、ＳｉＧｅ結晶膜の他、ＳｉＧｅＣ結晶膜，ＡｌＧａＡｓ結晶膜，Ｓｉ_y Ｃ_1-y 膜，ＩｎＰ膜などの化合物半導体膜があり、いずれの半導体膜に対しても本発明を適用することができる。
【０１２２】
【発明の効果】
本発明によると、半導体デバイスの要素の構造を現実に起こりうる構造を媒介として光学的評価法による実測値を予想し、この予想値を実測値と比較して半導体デバイスの要素の構造を評価するので、実際に起こりうる構造を反映した正確な構造評価，これを利用した半導体デバイス及びこれを自動的に行なうための記録媒体の提供を図ることができる。
【図面の簡単な説明】
【図１】本発明の実施形態におけるサンプル評価及び薄膜の製造プロセスにおける管理の手順を示すフローチャートである。
【図２】分光エリプソメトリ法による測定の結果得られるΔ，Ψスペクトルの例を示す図である。
【図３】（ａ）〜（ｄ）は、それそれ順に、傾斜組成ＳｉＧｅ−ＨＢＴの積層膜における設計されたＧｅ組成率のプロファイル，条件がはずれた場合のプロファイル，プロセスシミュレータを用いた条件の推定用プロファイル，補正後のサンプル構造のプロファイルを示す図である。
【図４】図３（ａ）に示す台形プロファイル構造を得るためのプロセス条件を示す図である。
【図５】プロセスシミュレータによるサンプル構造の推定手順を示すフローチャートである。
【図６】プロセスシミュレータによって推定されたＧｅ組成率のエピタキシャル層厚み方向におけるプロファイルを示す図である。
【図７】ＳｉＧｅ層におけるＧｅ組成率のゲルマン流量比依存性を示す図である。
【図８】ＳｉＧｅ層の成長速度のゲルマン流量比依存性を示す図である。
【図９】図６に示すＧｅ組成率のプロファイルを有する積層膜からシミュレーションを行なったΔ，Ψのスペクトルである。
【図１０】従来のサンプル評価及び薄膜の製造プロセスにおける管理の手順を示すフローチャートである。
【図１１】（ａ）〜（ｄ）は、それそれ順に、傾斜組成ＳｉＧｅ−ＨＢＴの積層膜における設計されたＧｅ組成率のプロファイル，条件がはずれた場合のプロファイル，一律構造を仮定した従来の推定プロファイル，補正後のサンプル構造のプロファイルを示す図である。[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a structure evaluation method, a semiconductor device manufacturing method, and a recording medium used for managing a manufacturing process of an element of a semiconductor device.
[0002]
[Prior art]
In recent years, a process for forming a thin film such as an oxide film, a nitride film, or a polysilicon film on a substrate is frequently used in the manufacture of semiconductor devices. In order to create a device using these thin films as elements and obtain desired characteristics, the film thickness and physical properties of the thin films need to be within a predetermined range. In general, the physical properties and thickness of a thin film vary depending on the conditions of the process for forming the thin film (hereinafter referred to as thin film process) and the time for which the thin film is formed. Evaluate whether or not they have physical properties. In the device mass production process, if it is found from the evaluation results that the formed thin film does not have the desired film thickness and physical properties, it is necessary to change the process conditions.
[0003]
Here, in the case of a conventional thin film process, a thin film formed in a single process is basically a substantially homogeneous film in which the composition and other physical properties do not change greatly in the depth direction. On the other hand, there are many examples in which a polysilicon film serving as a gate electrode is stacked on a silicon oxide film serving as a gate insulating film like the gate portion of a Si-MOS transistor. However, even in these cases, the composition is within each layer. It is almost uniform and the interface between each layer is clear in most cases.
[0004]
In addition, there is an optical evaluation method as a technique for evaluating the film thickness and composition of each layer of a single layer film and a multilayer film on a substrate. Spectral ellipsometry and spectral reflectance measurement methods are widely used as examples of optical evaluation methods. It has been.
[0005]
Spectral reflectance measurement method is an evaluation method that irradiates a sample with light and obtains the ratio (reflectance) between the intensity of the light irradiated to the sample and the intensity of the light reflected from the sample in each spectral wavelength range. is there.
[0006]
The spectroscopic ellipsometry method is an evaluation technique that irradiates a sample with linearly polarized light and obtains information about the sample from changes in the polarization state of reflected light. Of the linearly polarized light, when the electric field vector is parallel to the incident surface, the p-polarized light component is perpendicular, and the s-polarized light is the perpendicular one, and ρ and Rs are the complex reflectivities. ≡Rp / Rs is still a complex number. Therefore, ρ can be expressed as ρ≡tan ΨeiΔ using two real numbers ψ and Δ. The spectroscopic ellipsometry method obtains a spectrum by measuring these two physical quantities of Ψ and Δ for each wavelength of light.
[0007]
A common feature of these optical evaluation methods is that the phase and reflectance of the light change depending on the combination of the optical constants (refractive index n, extinction coefficient k) of the substance through which light passes. Information is included. In addition, in the optical information extracted from the object to be measured, the interference effect of light appears prominently, so the measurement result often changes greatly depending on the film thickness of the thin film, and by any of the above evaluation methods, Information such as the thickness of the thin film can also be obtained.
[0008]
However, in the reflectance measurement method or the spectroscopic ellipsometry method, the measured physical quantities (reflectance in the case of the reflectance measurement method, Ψ, Δ in the case of the spectroscopic ellipsometry method) are all on the path through which light passes. The effects of these substances are included, and these effects cannot be directly extracted as individually separated information.
[0009]
Therefore, when measuring a sample using the spectral reflectance measurement method or the spectroscopic ellipsometry method and evaluating the film thickness and characteristics of the thin film, the following procedure is to compare the actual measurement value with the expected measurement value. It is necessary to step on.
[0010]
FIG. 10 is a flowchart showing a management procedure in a conventional sample evaluation and thin film manufacturing process.
[0011]
First, in step ST201, the sample A created by a certain process P is measured by the evaluation method M, and an actual measured value (for example, Δ, Ψ) is obtained.
[0012]
On the other hand, in step ST202, a geometric model of the sample structure is set. In step ST203, an initial estimated value that defines the sample structure is set, and in step ST204, a theoretical predicted value of the physical quantity measurement value is calculated. That is, when optical evaluation is used, the structure (n, k profile) of the measurement sample is assumed, and the theoretical prediction of the physical quantity measurement value that will be obtained when the n, k profile is evaluated by the evaluation method M. Calculate the value.
[0013]
In step ST205, the actual measured value of the physical quantity and the theoretical predicted value are compared with each other. At this time, an evaluation value for evaluating the degree of difference between the actual measurement value and the theoretical prediction value is defined.
[0014]
Next, in step ST206, it is determined whether or not the evaluation value is a minimum value. If the evaluation value is not the minimum value, a new estimated value is set in step ST207, and then the process returns to step ST204 to return to step ST204. Repeat the process of ~ ST206.
[0015]
If it is determined in step ST206 that the evaluation value is minimal, the process proceeds to step ST208 to determine an estimated value of the sample structure, and then in step ST209, it is determined whether or not the sample structure is in an appropriate range. If the sample structure is within the appropriate range as a result of this determination, the process proceeds to step ST210, and the next process is performed with the set process conditions.
[0016]
On the other hand, if the result of determination in step ST209 is that the sample structure is not within the appropriate range, the process proceeds to step ST211 to determine whether or not the geometric model of the sample structure is appropriate. If the geometric model of the sample concept is appropriate, the process proceeds to step ST212 to estimate the cause of the structure abnormality and take measures such as changing the temperature, time, gas flow rate, and the like.
[0017]
As a result of the determination in step ST211, when it is determined that the geometric model of the sample structure is not appropriate, the process proceeds to step ST213, a new geometric model is set, the process returns to step ST203, and the steps after step ST203 are performed. Repeat the process.
[0018]
Here, as the evaluation value used in step ST205, there is usually a function that is a positive real number and becomes smaller as the difference between the actual measurement value and the theoretical prediction value is smaller, and becomes 0 when the two values completely match. Used. Generally, as an evaluation value, a dispersion value σ obtained by adding the square of the difference between an actual measurement value and a theoretical prediction value expressed by the following formula (1) over all wavelengths.
σ = Σ {aj (Sj −Smodj) ² } (1)
Is often used. However, Sj is an actual measurement value of the physical quantity, and Smodj is a theoretical predicted value of the physical quantity. Moreover, aj is a weighting coefficient, and when the weighting coefficients aj are all 1, all the information on each wavelength is used for the same evaluation, but the weighting coefficient is increased so that the contribution of the wavelength at which the characteristic of the sample structure is likely to appear becomes large. In some cases, the value of aj is not 1.
[0019]
In the least square method, the assumed sample structure that minimizes the evaluation value is used as the measurement value. That is, the sample structure (n, k depth profile) that gives the same Ψ and Δ as the actual measurement is searched for, and the structure of the sample that gives the closest Ψ and Δ is used as the measurement value. However, in the case of spectroscopic ellipsometry, subtle changes in the sample, such as changes in the optical constants at the atomic layer level of the surface, change the measurement result of ΨΔ, so the theoretical expected values for any sample structure are calculated and It cannot be compared.
[0020]
Accordingly, in the actual evaluation, the sample structure is expressed by a small number of parameters, and a combination of parameter values that minimizes the evaluation value is obtained within a range of assumed values. The evaluation value is a function of these parameters. Generally, this function is complicated, and it is actually very difficult to obtain the minimum value. Therefore, the minimum value is used instead of the minimum value. If it is a minimum value, it can be obtained by an algorithm such as the fastest descent method. In these algorithms, an appropriate initial value is given to the parameter, a small change is given to the parameter value in the direction of decreasing the evaluation value, and a point where the evaluation value increases with any minute change, that is, a minimum point is obtained. . However, in the case of the method using the minimum point, the obtained minimum point is not necessarily the point that gives the minimum value.
[0021]
If a laminated structure of single composition films having a clear interface formed by a conventional thin film process is used, the thin film can be evaluated with relatively good reproducibility by the above method. This is because the structure that can be processed is simple, the structure of the thin film can be expressed with a relatively small number of parameters, and the evaluation value other than the combination of parameter values that gives the model structure of the thin film closest to the actual thin film structure. The reason is that the minimum point is hardly generated.
[0022]
Next, consideration will be given to the use of the above-described conventional optical evaluation method for evaluating the film thickness and characteristics of a crystal film containing a plurality of elements.
[0023]
In recent years, a technique different from a conventional thin film process, such as an epitaxial growth technique of a crystal layer, has begun to be used mainly for the production of HBT (hetero bipolar transistor) and the like. The epitaxial growth technique is a technique for growing a new crystal having a structure following the structure of an atom constituting the base crystal on the base crystal such as a substrate. In the case of this technique, the film thickness can be controlled with very high accuracy (generally about 1 nm, up to one atomic layer under special conditions). Further, when the crystal to be grown is made of a material such as SiGe that forms a mixed crystal in a wide composition ratio range, the composition ratio can be controlled. Therefore, by using these characteristics, it is possible to create a state in which the composition changes approximately continuously in the depth direction with an arbitrary profile. An example of a device that utilizes this property is graded composition SiGe-HBT. In the case of the gradient composition SiGe-HBT, the Ge composition is set to 0 in the emitter region, and the Ge composition is gradually increased in the base region. At this time, when the Ge composition ratio is increased, the band gap is narrowed, so that an electric field is generated in the direction of accelerating internal carriers. As a result, the base travel time of the carrier is shortened and the transistor can be operated at high speed.
[0024]
In such a SiGe graded composition HBT, the composition may be changed in the base region to have a triangular profile, but generally a trapezoidal profile with a buffer layer with a uniform Ge composition ratio is adopted. There are many.
[0025]
FIG. 11A is a diagram showing a profile of the Ge composition ratio in the depth direction of a stacked structure formed by depositing a SiGe graded composition layer and a Si cap layer on a SiGe buffer layer having a uniform composition. .
[0026]
As described above, since an epitaxial growth technique was used to create a structure in which the composition changes approximately continuously, an approximately continuously changing profile was evaluated, and the profile was When deviating from a predetermined range, a method of correcting is required.
[0027]
Therefore, an attempt has been made to evaluate the sample whose composition changes approximately continuously in the depth direction by the spectroscopic ellipsometry method. In the evaluation process by the spectroscopic ellipsometry method, Ψ and Δ can be measured even for a sample having an arbitrary composition profile. In addition, if the profile of a composition that changes approximately continuously is approximated as a thin film stack, calculation of theoretical predicted values of Ψ and Δ is possible.
[0028]
[Problems to be solved by the invention]
However, in the epitaxial growth of mixed crystals containing a plurality of elements such as SiGe films, there are the following problems.
[0029]
Unlike thin films with a uniform composition, the SiGe epitaxial growth film that can change the Ge composition ratio in the depth direction at the atomic layer level has an extremely large sample structure, but the set process conditions are accurate. If this is realized, the resulting thin film structure should almost match the intended thin film structure.
[0030]
However, for example, when a crystal is grown with the intention of producing a thin film having a trapezoidal gradient composition profile as shown in FIG. 11A, and the crystal is grown at a temperature different from the set temperature, the completed thin film It was found that the profile of the composition did not become an exact trapezoidal profile.
[0031]
The reason is considered to be that the crystal growth rate depends on the substrate temperature, and the dependency depends on the Ge composition ratio. When the Ge composition ratio of the growing SiGe crystal is increased in the growth stage of the SiGe film, the activation energy in the crystal growth is reduced, so that the change in the growth rate with respect to the change in the substrate temperature is smaller than that in Si. As a result, when crystal growth is performed under the condition that the Ge composition ratio becomes a trapezoidal profile at the reference temperature, the Ge composition ratio of the gradient composition SiGe crystal grown at a temperature higher than the reference temperature has a downwardly convex profile, The Ge composition ratio of the gradient composition SiGe crystal grown at a temperature lower than the reference temperature has an upwardly convex profile.
[0032]
FIG. 11B is a diagram showing a profile of the Ge composition ratio in the depth direction of the stacked structure having the gradient composition SiGe film grown at a temperature higher than the reference temperature.
[0033]
FIG. 11C is a diagram showing a Ge composition ratio profile when trapezoidal approximation is performed despite the fact that the film has the Ge composition ratio profile shown in FIG. 11B. That is, when the structure of the laminated film having the Ge composition ratio profile shown in FIG. 11B is fitted with a trapezoidal model using the spectroscopic ellipsometry method and adopting the film thickness and Ge composition ratio of each layer as parameters. Because the curve part of the actual profile cannot be expressed, the total film thickness is almost the same as the actual profile, but the structure with the profile of the inclined part different from the actual one is obtained as an estimated value become. As a result, it is determined that the thickness of the Si cap and the SiGe buffer layer is thinner than the actual thickness, the thickness of the SiGe gradient composition layer is determined to be much thicker than the actual thickness, and the growth time is evaluated as being longer than the actual one. .
[0034]
Here, since there is no geometric model that simply describes the profile of the Ge composition ratio as shown in FIG. 11B, the conventional process condition correction method simply corrects the growth time of each layer of the laminated film. The method will be used.
[0035]
FIG. 11D shows a profile of the Ge composition ratio in the depth direction of the laminated film formed by correcting the thickness of each layer of the laminated film by shortening the growth time based on the result of trapezoidal approximation. FIG. As shown in the figure, when the process conditions are corrected using the profile of the laminated film obtained by the trapezoidal approximation shown in FIG. 11C as an estimated value, the SiGe graded composition layer is evaluated as being thicker than the actual one. Since the correction is made, a laminated film having a SiGe graded composition layer thinner than the design value is formed.
[0036]
In other words, the conventionally used correction methods cannot appropriately represent the Ge profile by the geometric model, and therefore cannot be evaluated correctly, resulting in incorrect correction.
[0037]
As described above, as shown in FIG. 11A, there are four Ge composition ratio profile structures: the thickness of each of the Si cap layer, the gradient composition layer, and the SiGe buffer layer, and the Ge composition ratio in each layer. Only a film grown at a reference temperature can be expressed by a trapezoid determined by Therefore, even if the growth conditions are corrected based on the measured values of the thickness of the grown SiGe film, etc., as long as it is assumed that the Ge composition ratio profile is always trapezoidal, correct correction is performed. It is difficult.
[0038]
Therefore, if the number of parameters that define the Ge composition ratio profile of the film is increased and the Ge composition ratio profile is expressed by a polygon having more vertices than a trapezoid, the Ge composition ratio profile at an arbitrary temperature can be expressed. it can. However, if the number of parameters that define the profile of the Ge composition ratio is increased, the number of combinations of parameters that give the minimum value of the above-mentioned dispersion value x becomes extremely large, so that a correct estimated value can be obtained with a realistic calculation amount. It becomes difficult. Theoretically, if the number of parameters that define the structure of the geometric model, such as the Ge composition ratio profile, is increased, it is possible to express a sample structure that is closer to reality. However, when the number of parameters is increased, the combination of parameter values that gives the minimum evaluation value is not limited to one. As a result, the sample structure is almost the same, but there are subtle variations in Ψ and Δ measurements due to noise during measurement by the spectroscopic ellipsometry method, and subtle samples such as structures not included in the structural model, such as composition fluctuations at the interface. Due to the difference in structure, there is a problem that the measurement result changes greatly.
[0039]
An object of the present invention is to provide a structure evaluation method for obtaining a structure substantially as designed by grasping a process condition from a measured value of a sample through a sample structure and correcting the process condition using the result. A semiconductor device manufacturing method and a recording medium are provided.
[0040]
[Means for Solving the Problems]
The structure evaluation method of the present invention assumes a step (a) of obtaining a plurality of actual measurement values of physical quantities of elements of a semiconductor device by an optical evaluation method, and a process condition for forming the elements. Obtained by calculating the structure of the element formed through a process using process conditions by calculation (b), and evaluating the structure of the element obtained in step (b) by the optical evaluation method. Calculating a predicted value of a plurality of measured values of the physical quantity to be obtained, and estimating the structure of the element based on the plurality of actually measured values of the physical quantity of the element and the predicted value of the plurality of measured values Step (d).
[0041]
By this method, in step (d), based on the expected value of the measured physical quantity of the element based on the actual element structure obtained in step (b) and the actual measured value of the physical quantity, A probable structure is presumed. That is, unlike the conventional structure evaluation based on a uniform structure, it is possible to perform an accurate structure evaluation that reflects the structure of a physical quantity that changes in accordance with changes in process conditions.
[0042]
In the step (d), a numerical value for evaluating the difference between the actual measured values of the plurality of physical quantities and the predicted value of the plurality of measured values is calculated, and the step (b) is performed until the value becomes equal to or less than the threshold value. , (C), the structure evaluation is facilitated by using an algorithm such as a rapid descent method using the least square method.
[0043]
In the step (b), the structure evaluation can be performed easily and quickly by performing the calculation using a process simulator.
[0044]
An element is formed in advance by a process using a plurality of process conditions, the structure of the element is obtained by the optical evaluation method, and the structure of the element formed by the plurality of process conditions and the process condition is determined. In the step (b), the correlation is made into a database, and the structure of the element is obtained by calculation based on the correlation, whereby the structure evaluation can be performed more easily and quickly.
[0045]
When the process is an epitaxial growth process of a crystal film, particularly when the crystal film is a crystal film containing a plurality of elements, it is possible to exert a remarkable effect by applying the structure evaluation method of the present invention. .
[0046]
In the case where the crystal film is a crystal film containing Si and Ge and having a structure in which the band gap is inclined and changed, a structure evaluation that can be used for controlling the profile of the Ge composition ratio can be performed. In other words, since the crystal growth rate changes according to the Ge composition ratio, even if the Ge composition ratio profile does not have the inclined structure as designed, the Ge composition ratio profile that can actually occur is an estimate of the measured physical quantity. Since it is calculated as a value, an accurate profile of the Ge composition ratio in the crystal film from which the actual measurement value was obtained can be obtained by using the predicted value of the measurement value of the physical quantity.
[0047]
The optical evaluation method is preferably one of a spectral ellipsometry method and a spectral reflectance measurement method.
[0048]
The method for manufacturing a semiconductor device according to the present invention includes a step (a) of obtaining a plurality of actual measurement values of physical quantities of the above elements by optical evaluation method for one evaluation wafer among a plurality of wafers including the elements of the semiconductor device. (B) assuming process conditions for forming the element of the wafer for evaluation and calculating the structure of the element formed through a process using the assumed process condition; (C) calculating an expected value of a plurality of measured values of the physical quantity obtained when the structure of the element obtained in (b) is evaluated by the optical evaluation method, and the element of the evaluation wafer A step (d) of estimating a structure of the element based on the plurality of actually measured values of the physical quantity and an expected value of the plurality of measured values; and estimating the element of the evaluation wafer Determining whether or not to modify the process conditions of the process for at least a wafer other than the evaluation wafer among the plurality of wafers based on a difference between the structure and the design structure of the plurality of wafers ( e).
[0049]
This method makes it possible to change and set process conditions for other wafers after accurately grasping the structure of the elements of the evaluation wafer using the above-described structure evaluation method. Improvement and reduction in characteristic variation can be achieved.
[0050]
When the above process is an epitaxial growth process of a crystal film, particularly when it is a crystal film containing a plurality of elements, a remarkable effect can be exhibited by applying the method for manufacturing a semiconductor device of the present invention.
[0051]
When the crystal film includes Si and Ge and includes a structure in which the band gap changes in an inclined manner, the profile of the Ge composition ratio can be accurately controlled. In other words, since the crystal growth rate changes according to the Ge composition ratio, even if the Ge composition ratio profile does not have the inclined structure as designed, the Ge composition ratio profile that can actually occur is an estimate of the measured physical quantity. Since it is calculated as a value, an accurate profile of the Ge composition ratio in the crystal film from which the actual measurement value was obtained can be obtained by using the predicted value of the measurement value of the physical quantity.
[0052]
The recording medium of the present invention is a recording medium that can be incorporated into a computer used for evaluating the characteristics of the elements of a semiconductor device by an optical evaluation method, and a plurality of actually measured values of physical quantities of the elements of the semiconductor device are obtained. A procedure (a) for capturing, a process condition for forming the element, a procedure (b) for calculating the structure of the element formed through a process using the assumed process condition by calculation, A step (c) of calculating an expected value of a plurality of measured values of the physical quantity obtained when the structure of the element obtained in step (b) is evaluated by the optical evaluation method; and the plurality of physical quantities of the element A computer having recorded thereon a program for causing a computer to execute the step (d) for estimating the structure of the element based on the actually measured values of the measured values and the predicted values of the plurality of measured values Which is a recording medium that can be taken only.
[0053]
As a result, the above-described structural evaluation can be automatically performed using a computer.
[0054]
In the procedure (d), a numerical value for evaluating the difference between the actual measured values of the plurality of physical quantities and the predicted value of the plurality of measured values is calculated, and the procedure (b) is performed until the value becomes a threshold value or less. , (C), the structure of the element is preferably estimated.
[0055]
DETAILED DESCRIPTION OF THE INVENTION
-Basic concept of the structure evaluation method of the present invention-
In this embodiment, since a process simulator that calculates a profile of the Ge composition ratio actually created from the process conditions is used, the relationship between the process conditions, the growth rate, and the Ge composition ratio in the SiGe layer is clarified. There is a need. Here, the case of an epitaxial growth process of a crystal by the UHV-CVD method (ultra-high vacuum vapor phase chemical deposition method) will be described as an example.
[0056]
The epitaxial growth technique is a technique for growing a new crystal on the surface of an existing crystal body. In particular, the UHV-CVD method used in this embodiment is a kind of epitaxial growth process, and the degree of vacuum in the crystal growth apparatus is 10. ^-6 Pa-10 ^-7 In this technique, after improving to about Pa, a source gas is introduced into a crystal growth apparatus, and a crystal is grown by a chemical reaction between the heated substrate surface and the source gas.
[0057]
During the crystal growth, silane containing silicon (SiH _Four ) And disilane (Si ₂ H ₆ If a gas such as) is used as a source gas, silicon can be grown. In addition, together with the gas containing silicon, germanium containing germanium (GeH _Four When a gas such as) is introduced into the crystal growth apparatus, a SiGe crystal that is a mixed crystal of silicon and germanium can be grown. Further, in addition to a gas containing silicon and a gas containing germanium, SiH _Three CH _Three When a gas containing carbon such as is introduced into a crystal growth apparatus, a SiGeC crystal can be grown.
[0058]
In the case of SiGe crystal, in principle, the ratio (composition ratio) of each element contained in the mixed crystal can take an arbitrary value. The composition ratio of each element of the SiGe crystal is determined by the process conditions when the crystal is grown. In the case of the UHV-CVD method, the composition rate hardly depends on the substrate temperature and the pressure (total pressure) in the crystal growth apparatus, and the partial pressure of the silicon source gas in the crystal growth apparatus and the germanium source gas It is determined only by the partial pressure ratio.
[0059]
Usually, the crystal part once grown does not change until a process other than crystal growth such as high-temperature heat treatment is performed thereafter. For example, if a SiGe layer is first grown by simultaneously introducing disilane and germane into a crystal growth apparatus and then a Si layer is grown by introducing only disilane, the SiGe layer and the Si layer are formed on the substrate surface. A film having a double structure can be formed. If the flow ratio of disilane and germane is changed approximately continuously in time, a SiGe layer in which the Ge composition ratio changes approximately continuously in the depth direction can be formed. That is, the Ge composition ratio profile of the film grown by the epitaxial growth process is determined by the history of the process conditions when each layer is grown as if it were an annual ring. Therefore, in order to obtain a desired composition ratio profile, it is necessary to change the growth conditions from the start to the end of crystal growth approximately continuously.
[0060]
The composition ratio at a certain depth position of the epitaxially grown film is determined by the source gas partial pressure ratio when the depth portion of the film is growing. Whether it corresponds to the portion at the depth position cannot be grasped unless it is known how thick the film is grown before the portion at the depth position is grown.
[0061]
That is, in the SiGe film, the original substrate surface is taken as the origin, and the Ge composition ratio x (d) of the part at a distance d from the origin of the film is the time when the crystal of the part is grown is t (d). Then, the following formula (2)

Represented by Where Cm is the disilane partial pressure P _Si , German partial pressure P _Ge And a function representing the relationship between the substrate temperature T and the composition ratio.
[0062]
The variable t (d) in the equation (2) is expressed by the following equation (3) with the crystal growth start time t set to 0.

It can ask for. Here, gr is a function representing the relationship between the gas partial pressure, the substrate temperature, and the growth rate.
[0063]
Accordingly, if the functions Cm and gr are known, and the partial pressure of disilane and germane and the substrate temperature at any time from the start to the end of crystal growth are known, the profile in the depth direction of the Ge composition ratio (composition ratio) is defined. A value x (d) to be obtained can be obtained. That is, the sample structure can be estimated from the process conditions. Specifically, these relational expressions (2) and (3) can be obtained by preparing samples by changing the growth conditions.
[0064]
Further, the relational expressions (2) and (3) can be replaced with higher accuracy by accumulating information such as the growth mechanism. Further, regarding the structure, it is only necessary to obtain not only the entire structure of the target element but also the accuracy necessary for the evaluation of a specific structure such as a Ge composition ratio profile, so that the equations (2) and (3) It is also possible to use a simplified approximation formula.
[0065]
It should be noted that the simultaneous equations are solved using the above equations (2) and (3), and analytically P _Si (Τ), P _Ge Although x (d) may be obtained from (τ) and T (τ), the equations (2) and (3) themselves are complicated, so that there may be cases where a solution cannot be obtained.
[0066]
Therefore, it is also possible to perform numerical calculation handling in which the values of each variable are obtained from the relational expressions (2) and (3), and it is simpler to perform this numerical calculation handling. In other words, if it is analytically handled, the composition ratio (composition ratio) of the film at an arbitrary depth can be calculated. However, in practice, it can be obtained from the profile in the depth direction of the composition using the spectroscopic ellipsometry method. In the process of calculating the theoretical predicted values of the Δ and Ψ spectra obtained, the profile in the depth direction of the composition may be obtained with the accuracy required to obtain a reliable predicted value.
[0067]
Therefore, in order to simplify the calculation, from the start to the completion of the crystal growth process is regarded as a continuation of basic steps in which the growth conditions can be regarded as not changing, and the gas pressure, substrate temperature, and time of each step in each step are considered. Given the length, the composition and thickness of the film grown in each step (in this case, a simple product of the time for performing the step and the growth rate) can be obtained from the relational expression of the functions Cm and gr. The profile can be obtained.
[0068]
Actually, even when epitaxially growing a film having a profile whose composition changes approximately continuously, such as an HBT structure having a SiGe graded composition layer, the partial pressure of the source gas is actually changed approximately continuously. Since it is difficult to do this, a method of changing the gas partial pressure every short step is almost adopted. In that case, since the process conditions can be considered to be normally constant in each step, the method of calculating the composition and the growth rate / film thickness for each step is more practical.
[0069]
As described above, the sample structure can be estimated if the process conditions (gas partial pressure, substrate temperature) from the start of growth of the crystal layer to the end of growth are known. As described above, a program for estimating a sample structure from process conditions is generally called a process simulator. Therefore, in this specification, the process simulator is used in the meaning of this program.
[0070]
-Basic procedure of manufacturing process-
FIG. 1 is a flowchart showing a management procedure in a sample evaluation and thin film manufacturing process according to an embodiment of the present invention.
[0071]
First, assuming that a manufacturing process (process P) of 50 wafers per lot is performed in step ST101, an evaluation method M (for example, a spectroscopic ellipsometry method) is performed on the first wafer (evaluation wafer). ) To obtain actual measured values of physical quantities (for example, Δ, Ψ spectra).
[0072]
On the other hand, in step ST102, an initial estimated value of the process condition is set. In step ST103, the sample structure is estimated by the process simulator. In step ST104, a theoretically predicted value of the physical quantity measurement value is calculated.
[0073]
In step ST105, the actual measured value and the theoretical calculated value are compared with each other. At this time, an evaluation value for evaluating the degree of difference between the actual measurement value and the theoretical prediction value is calculated.
[0074]
Next, in step ST106, it is determined whether or not the evaluation value is a minimum value. If the evaluation value is not the minimum value, a new estimated value is set in step ST107, and then the process returns to step ST103, and steps ST104 to ST106 are performed. Repeat the process.
[0075]
If it is determined in step ST106 that the evaluation value is minimal, the process proceeds to step ST108 to determine an estimated value of the sample structure, and then in step ST109, it is determined whether or not the sample structure is in an appropriate range. If the sample structure is within the proper range as a result of this determination, the process proceeds to step ST110, and the next process is performed with the set process conditions.
[0076]
On the other hand, if the result of determination in step ST109 is that the sample structure is not within the appropriate range, the process proceeds to step ST111, and the process conditions estimated in step ST108 are corrected out of the appropriate range, Process wafers other than the evaluation wafer. However, the process may be performed again on the evaluation wafer.
[0077]
-Description of each step in the flowchart-
Hereinafter, the specific contents of the above processing will be described by taking as an example the case of evaluating a gradient composition SiGe-HBT structure.
[0078]
The measured value of the physical quantity in step ST101 is assumed to use an optical evaluation method, and in this embodiment, a spectroscopic ellipsometry method is used. The present invention can also be implemented using other optical evaluation methods such as a spectral reflectance measurement method. In particular, as the physical quantity, it is preferable that the variable can take a plurality of real numbers when the wavelength of light is a variable. This is because if a plurality of actually measured values cannot be obtained, the fitting process between the actually measured values and the estimated values using the least square method or the like cannot be performed with high accuracy.
[0079]
In this embodiment, a sample having an epitaxially grown SiGe film is measured by a spectroscopic ellipsometer method to obtain Ψ and Δ spectra at a plurality of wavelengths. The dependence of the optical constant of the SiGe film on the Ge composition ratio is remarkable in the short wavelength region, but light with a short wavelength is easily absorbed by Si. It is preferable to carry out measurement over the entire range.
[0080]
FIG. 2 is a diagram showing an example of Δ and Ψ spectra obtained as a result of measurement by the spectroscopic ellipsometry method. As described above, when the spectral sample is irradiated with linearly polarized light, the obtained reflected light has a p-polarized component when the electric field vector is parallel to the incident surface, and s When the polarization component is used and the complex reflectances are Rp and Rs, ρ≡Rp / Rs is still a complex number. Therefore, ρ can be expressed as ρ≡tan ΨeiΔ using two real numbers ψ and Δ. FIG. 2 is a spectrum obtained by measuring these two physical quantities of Ψ and Δ for light of each wavelength.
[0081]
Next, in step ST102, an initial value of the process condition is estimated. Here, when setting the process conditions, if three values of the substrate temperature and the flow rate of disilane and germane are specified, the growth rate and the Ge composition ratio can be obtained. Therefore, like a laminated film having a SiGe film having a gradient composition The substrate temperature and the gas flow method are determined so that a trapezoidal profile having a desired Ge composition ratio can be obtained. Therefore, first, it is necessary to set a profile structure having a desired Ge composition ratio and to determine process conditions for obtaining this profile structure.
[0082]
FIGS. 3A to 3D show, in order, the designed Ge composition ratio profile in the graded composition SiGe-HBT laminated film, the profile when the conditions deviate, and the condition estimation using the process simulator. It is a figure which shows the profile of the sample structure after a profile and correction | amendment.
[0083]
As shown in FIG. 3A, in the case of the gradient composition SiGe-HBT, as described above, a trapezoidal Ge composition ratio profile is designed to generate an internal electric field for accelerating carriers in the base layer. To do. In this embodiment, the thickness of the SiGe buffer layer is 40 nm, the thickness of the SiGe graded composition layer is 40 nm, the thickness of the Si cap layer is 50 nm, and a total of 130 nm of laminated films is epitaxially grown on the Si substrate. The Ge composition ratio of the SiGe buffer layer is made uniform 15%. The Ge composition ratio of the SiGe graded composition layer is set to 15% in the part adjacent to the SiGe buffer layer and 0% in the part adjacent to the Si cap layer, and the Ge composition ratio is approximately linear in the SiGe graded composition layer. Change (in stages).
[0084]
FIG. 4 is a diagram showing process conditions for obtaining the trapezoidal profile structure shown in FIG. Usually, for easy control, the substrate temperature T is kept constant during growth and disilane (Si ₂ H ₆ ) Flow rate is constant, and germane (GeH _Four The flow rate ratio is controlled by modulating only the flow rate. That is, the Ge composition ratio of the SiGe film is changed stepwise by decreasing the germane flow rate stepwise. At this time, if the germane flow rate is reduced, the Ge composition ratio can be lowered. However, since the growth rate is lowered at the same time, the germane flow rate is modulated so as to obtain a desired profile.
[0085]
FIG. 5 is a flowchart showing a sample structure estimation procedure by the process simulator in step ST103.
[0086]
First, in step ST150, the process from the start to the end of the epitaxial growth process is represented as a flow of basic steps in which the process conditions can be considered constant in each step. In step ST151, the gas pressure (or flow rate), substrate temperature, and step in each step are represented. Give the length of time. Specifically, the time t0, t1, t2, t3,... At which the parameters (in this example, only the germane flow rate) change along the flow of the process conditions as shown in FIG. A plurality of basic steps are defined.
[0087]
Next, in step ST152, the composition (Ge composition ratio) of the crystal layer grown in that step is calculated from the pressure ratio (or flow ratio) of the source gases. When the composition has substrate temperature dependency, the calculation is taken into consideration. That is, it corresponds to obtaining the Ge composition ratio d (t) in each step based on the above-described equation (2).
[0088]
In step ST152, the thickness (product of growth rate and time) of the crystal layer grown in the step is calculated from the pressure of the source gas and the substrate temperature. In step ST154, the composition and thickness of the layer in all steps are calculated. The final structure is calculated from the calculation result.
[0089]
FIG. 6 is a diagram showing a profile in the epitaxial layer thickness direction of the Ge composition ratio estimated by the process simulator. It can be seen that the profile shown in the figure has a slightly convex shape when the outline is viewed.
[0090]
FIG. 7 is a graph showing the germane flow rate dependency of the Ge composition ratio in the SiGe layer. The Ge flow rate ratio is disilane (Si ₂ H ₆ ) And germane (GeH) _Four ) And the flow rate ratio of germane to the total flow rate. However, the growth rate of the SiGe layer depends on the substrate temperature and the gas flow rate. In the case of UHV-CVD, since the substrate temperature is relatively low, even if a certain amount of source gas is flowed, the reaction on the substrate surface cannot catch up and the growth rate does not depend on the gas flow rate. Easy to rush into.
The normal process is performed only in this reaction-limited region.
[0091]
In the reaction rate limiting region, the temperature dependence of the growth rate is a so-called thermal excitation process type. When the growth rate is g, the substrate temperature is T, and the activation energy is Ea, the reaction rate between g and temperature T is as follows.
g ∝ exp (-Ea / T * k)
There is a relationship. Here, k is a Boltzmann constant.
[0092]
Here, the epitaxial growth of the SiGe layer is characterized in that the activation energy Ea is dependent on the Ge composition ratio, and the value of the activation energy Ea decreases as the Ge composition ratio increases. As a result, the growth rate of the SiGe layer increases as the Ge composition ratio increases at the same temperature, and the substrate temperature dependency of the growth rate decreases.
[0093]
FIG. 8 is a graph showing the germane flow rate dependence of the growth rate (nm / min) of the SiGe layer. FIG. 8 shows the relationship between the growth rate at the substrate temperature of 587 ° C. and the germane flow rate ratio. As shown in the figure, germane (GeH _Four ) The higher the flow rate ratio, the greater the growth rate. In the vicinity of 600 ° C., when the substrate temperature changes by 1 ° C. (1 degree), the growth rate changes by about 3% in the case of the Si layer (the point where the Ge flow ratio is 0), whereas the Ge composition It has been found that the growth rate change is only about 2% in the SiGe layer with a rate of 15%.
[0094]
As described above, setting the substrate temperature, the disilane flow rate, and the germane flow rate as process conditions, and calculating the structure that will be completed if these three quantities are specified at any time during the film forming process. Can do. However, in the actual process, when the process condition deviates from the set condition, the gas flow rate rarely deviates from the set value, and in most cases, the substrate temperature deviates from the set value. In addition, since the substrate temperature hardly changes with time, it is often sufficient to set one variable having the substrate temperature as a constant value as the structure estimation parameter. Of course, depending on the situation, the gas flow rate may shift or the substrate temperature may change with time. In such a case, these values can also be used as fitting parameters. As described above, if a method for estimating a structure from process conditions is used, all structures that can be actually created can be expressed by parameters.
[0095]
As a method for estimating the structure from the process conditions, the relationship among the constant growth temperature, the constant gas flow rate ratio, the growth rate, and the Ge composition ratio as described above is clarified, and an arbitrary You can take a method of calculating the structure under the growth conditions, but separately from this method, create a sample by dividing the process conditions in advance, and find the relationship between the resulting structure and the process condition values. In addition, it is possible to take a method of obtaining the sample structure under the conditions other than the sample creation by interpolation. This method is convenient when the number of process parameters is small.
[0096]
Next, in step ST104, Δ, Ψ, which would be obtained when optical evaluation of a laminated film having such a structure is performed by a spectroscopic ellipsometry method based on the profile structure of the Ge composition ratio shown in FIG. The spectrum (theoretical predicted value of the physical quantity) is calculated. In other words, the profile in the depth direction of the optical constant can be calculated according to the profile of the Ge composition ratio, and the theory of the measurement value of the laminated film by the spectral ellipsometry method or the spectral reflectance measurement method according to the profile of the optical constant. Expected values can be calculated.
[0097]
The specific calculation method of the theoretical prediction value is described in detail in “Ellipsometry and Polarized Light (Elsevier Science Ltd Published 1987)” by Azzam et. Al. The calculation principle is that the sample is a laminated film composed of a number of layers having constant optical constants within the layer, and light propagation in each layer and reflectance and transmittance at each interface are obtained from the optical constants of each layer. There is. That is, if the relationship between the composition of each layer and the optical constant (refractive index n, extinction coefficient k) is known, the measured value of the physical quantity when the structure of the laminated film estimated by the process simulator is optically evaluated is predicted. be able to.
[0098]
The relationship between the composition of SiGe or SiGeC and the optical constant is described in, for example, the paper “RTCarline et.al. Appl.Phys.Lett.64 No.9 p.1114-1116,1994”. The algorithm for calculating the constant is described in, for example, the paper “Snyder et.al. Appl.Phys. 68 No.11 p.5925-5926, 1990”.
[0099]
Therefore, if the process condition value is input to the process simulator, the sample structure can be calculated, and the theoretical predicted value of the measured value obtained by the spectroscopic ellipsometry obtained from the sample structure can be calculated.
[0100]
FIG. 9 shows the spectra (theoretical predicted values) of Δ and Ψ that the present inventors have simulated from the laminated film having the Ge composition ratio profile shown in FIG. Note that, in the spectrum simulation shown in FIG. 9, the apparatus constants specific to the used ellipsometry spectrometer are taken into consideration. Therefore, when it is assumed that a laminated film having the same structure is measured using another ellipsometry spectrometer, the same spectrum as shown in FIG. 9 is not always obtained.
[0101]
Next, in step ST105, the actual measured value shown in FIG. 2 is compared with the theoretical expected value shown in FIG. Then, for example, the dispersion value σ shown in the expression (1) is obtained as an evaluation value, and it is determined whether or not the evaluation value is minimal in step ST106, and the process returns from step ST107 to step ST103 until the evaluation value becomes minimal. repeat. At this time, in the process of step ST107, a process condition in which only the substrate temperature T is changed from 600 ° C. shown in FIG. 4 by, for example, 1 ° C. is set as a new estimated value, and the sample structure in step ST103 is set based on this process condition. Estimate
[0102]
As the process conditions, the substrate temperature, the disilane flow rate, and the germane flow rate are set. Therefore, in step ST107, these three amounts of values at arbitrary times during the film forming process are newly set. You can also. However, as described above, in the actual process, it is almost the substrate temperature T that deviates from the setting in the process conditions. In this example, only one variable of the substrate temperature T is used as the structure estimation parameter. .
[0103]
When the rapid drop method is used, it is considered that the probability that the evaluation value reaches the minimum is most likely based on the differential coefficient of the evaluation value when the substrate temperature T is changed in the process of step ST107. In the direction, the value of the substrate temperature T is corrected, and the evaluation value converges to a minimum.
[0104]
Next, in step ST108, an estimated value of the most probable process condition is obtained based on the physical quantities obtained in step ST101 (Δ and Ψ spectra shown in FIG. 2). If it is determined in step ST109 that the process condition deviates from an appropriate range (for example, 600 ° C. ± 0.5 ° C.), the condition is corrected in step ST111. That is, it is determined whether or not the process condition is to be corrected by the process from step ST109 to ST110 or ST111.
[0105]
For example, when it is estimated that the measured value of the physical quantity shown in FIG. 2 is obtained from a laminated film epitaxially grown under the condition that the substrate temperature T is 2 ° C. lower than 600 ° C., the substrate temperature T is defined. The parameters are changed so that the substrate temperature T is increased by 2 ° C., and the subsequent wafer is processed. As a result, since an appropriate countermeasure can be taken for the cause of the abnormality, a structure having a desired profile can be formed from the next time.
[0106]
-Effect of the embodiment-
In the case of a crystal growth process for producing an actual transistor, an insulating film, a polysilicon layer, etc. are already formed in the substrate before the crystal growth. In the case of a substrate on which a pattern has already been formed in this way, due to differences in thermal emissivity, the same temperature is not obtained even if the same heating method is used for an Si substrate on which no pattern is formed. Even in the case of substrates having the same type of pattern, the growth temperature changes due to process variations in the thickness of the insulating film layer. Therefore, samples that must actually be evaluated often have a curve portion in the profile of the Ge composition ratio, which is formed with the temperature shifted. In addition, it is very difficult to make the substrate temperature uniform in the plane, and portions where the films are formed at different temperatures are formed in the substrate plane. In that case, in the conventional method, as shown in FIG. 11 (c), the theoretically predicted value of the measured value of the physical quantity by the spectroscopic ellipsometry method, etc., on the premise of the structure that the stacked film may not actually have. Is estimated. Therefore, as shown in FIG. 11D, only a physical structure deviated from the design (for example, a profile structure with a Ge composition ratio) can be obtained.
[0107]
On the other hand, according to the present embodiment, as will be described below, it is possible to predict a structure that can actually occur, grasp that the temperature is deviated from the standard condition, and perform condition correction based on that.
[0108]
For example, when the process conditions are set using the trapezoidal profile structure shown in FIG. 3A as a design value and the profile structure shown in FIG. 3B is obtained, the process returns from step ST107 to step ST103, and step ST103. The most probable profile structure that the laminated film will actually have is obtained at the stage of repeating the process of .about.ST106.
[0109]
That is, as shown in FIG. 3 (c), it occurs when the substrate temperature T is higher than the set value, or when the substrate temperature T is lower than the set value. Information on the composition and film thickness of each layer (SiGe buffer layer, SiGe gradient composition layer, and Si cap layer) in the laminated film is obtained. Since the process conditions are corrected based on the information, a laminated film having a trapezoidal profile almost as shown in FIG. 3D can be formed in the subsequent processes.
[0110]
In other words, in the present embodiment, the theoretical predicted value of the measured value of the physical quantity by the spectroscopic ellipsometry method or the like is estimated on the premise of the profile structure that is actually estimated to be possessed by the laminated film. As a result, a structure as designed (here, a trapezoidal profile structure with a Ge composition ratio) can be obtained.
[0111]
That is, it is possible to improve the characteristics of the semiconductor device and reduce the variation in characteristics.
[0112]
(Other embodiments)
As an algorithm for calculating the evaluation value, basically, an algorithm of a conventional method can be used, and an evaluation value other than the variance value shown in the equation (1) can also be used. Further, as an algorithm for determining a minimum value, other existing algorithms for determining a minimum value can be used instead of the rapid descent method.
[0113]
Further, the determination process in step ST109 may be replaced with a determination as to whether or not the estimated value of the device structure (Ge composition ratio profile structure) is within an allowable range.
[0114]
Furthermore, a communication function may be provided between the evaluation apparatus and the manufacturing apparatus that performs the film forming process, and the next process may be automatically corrected based on the evaluation result. This is particularly effective when the evaluation apparatus is configured to be housed in a common housing with the manufacturing apparatus.
[0115]
The structure estimation by such process simulation and the fitting procedure for comparing the theoretically predicted value of the physical quantity with the actual measurement value can be performed by a device different from the device that measures the actual physical quantity value. It is preferable to deploy a computer with a built-in hard disk storing an algorithm for performing this fitting procedure, since it can be performed consistently from physical quantity measurement to final sample structure estimation.
[0116]
Further, the processing shown in the flowchart shown in FIG. 1 can be automatically performed by recording the procedure on a recording medium. For example, by recording the procedure from step ST102 to step ST107 as a program on a computer-readable recording medium, the process condition in which the structure to be controlled is formed can be automatically estimated. Examples of the recording medium include a CD-ROM, a magnetic tape using a magnetic material, and an FD, a non-volatile memory such as an EEPROM, and a DVD, and any of them may be used.
[0117]
In implementing the present invention, it is not necessary to perform all parts of the sample structure with a process simulator. For example, in the case of a gradient composition HBT structure, the SiGe buffer layer portion and the portion that becomes a uniform composition film such as a Si cap layer use the film thickness and composition ratio as direct parameters as in the past, and the SiGe gradient composition layer. It may be easier to use only the process simulator. Even in this case, since the profile of the SiGe graded composition layer can be appropriately expressed as compared with the conventional method, the evaluation accuracy is remarkably improved as compared with the conventional method.
[0118]
In the embodiment described above, the SiGe graded composition HBT structure by UHV-CVD method is created and evaluated by the spectroscopic ellipsometry method, but the present invention is not limited to this combination. For example, the spectral reflectance measurement method and the spectral ellipsometry method have a lot in common. Therefore, in the procedure shown in the flowchart of FIG. 1, the physical quantity for actually measuring and calculating the theoretically predicted value is simply replaced with the reflectance from Ψ and Δ. It can be implemented in almost the same way. Even with other evaluation methods, the present invention can be implemented as long as the sample structure is estimated and the theoretical predicted value of the physical quantity can be calculated.
[0119]
Even if it is not the UHV-CVD method, the relationship between the process conditions and the structure that will be formed by using the process conditions has been clarified. Implementation of the invention is possible. For example, it can be easily carried out by epitaxial growth by LP-CVD or MBE. Also, in the RTA process for impurity diffusion, activation, and silicide formation, the process conditions are controlled by a very small number of parameters such as temperature and time, but the formed structure has a very complicated profile. Suitable for the implementation of the present invention.
[0120]
In addition, even for a film having a simple structure such as an oxide film or a nitride film, the process condition value to be corrected by the method of the present invention can be easily obtained. Therefore, by applying the present invention, an effect of improving the process can be obtained. .
[0121]
However, the present invention is particularly meaningful when applied to a structure that is composed of a plurality of elements and whose composition changes. In such a structure, if the composition changes, the growth state (for example, the growth rate) often changes accordingly. Therefore, it is possible to grasp the correct structure or correct the conditions with the conventional method. Is difficult. On the other hand, as in the present invention, the physical quantity (structure) that can actually occur when the process conditions are changed, and then the measured value of the physical quantity is predicted based on the physical quantity (structure). The structure can be controlled almost as designed. Examples of such a film composed of a plurality of elements include SiGe crystal film, SiGeC crystal film, AlGaAs crystal film, Si _y C _1-y There are compound semiconductor films such as a film and an InP film, and the present invention can be applied to any semiconductor film.
[0122]
【The invention's effect】
According to the present invention, an actual measurement value by an optical evaluation method is predicted through a structure that can actually occur as a structure of an element of a semiconductor device, and the structure of the element of the semiconductor device is evaluated by comparing the predicted value with the actual measurement value. Therefore, it is possible to provide an accurate structure evaluation reflecting a structure that can actually occur, a semiconductor device using the structure, and a recording medium for automatically performing the structure.
[Brief description of the drawings]
BRIEF DESCRIPTION OF DRAWINGS FIG. 1 is a flowchart showing a management procedure in a sample evaluation and thin film manufacturing process according to an embodiment of the present invention.
FIG. 2 is a diagram showing an example of Δ and Ψ spectra obtained as a result of measurement by a spectroscopic ellipsometry method.
FIGS. 3A to 3D show, in order, the profile of the designed Ge composition ratio in the laminated film of the gradient composition SiGe-HBT, the profile when the conditions deviate, and the conditions using the process simulator. It is a figure which shows the profile for estimation, and the profile of the sample structure after correction | amendment.
4 is a diagram showing process conditions for obtaining the trapezoidal profile structure shown in FIG.
FIG. 5 is a flowchart showing a sample structure estimation procedure by a process simulator;
FIG. 6 is a view showing a profile in the epitaxial layer thickness direction of a Ge composition ratio estimated by a process simulator.
FIG. 7 is a graph showing the germane flow rate dependence of the Ge composition ratio in a SiGe layer.
FIG. 8 is a graph showing the germane flow ratio dependence of the growth rate of the SiGe layer.
9 is a spectrum of Δ and Ψ simulated from the laminated film having the Ge composition ratio profile shown in FIG. 6;
FIG. 10 is a flowchart showing a management procedure in a conventional sample evaluation and thin film manufacturing process.
FIGS. 11A to 11D show, in order, a designed Ge composition ratio profile in a graded composition SiGe-HBT laminated film, a profile when the conditions deviate, and a conventional structure assuming a uniform structure; It is a figure which shows the profile of an estimated profile and the sample structure after correction | amendment.

Claims (2)

Step (a) of obtaining a plurality of actual measurement values of physical quantities of the above elements by an optical evaluation method for one evaluation wafer among a plurality of wafers including the elements of the semiconductor device;
Assuming a process condition for forming the element of the evaluation wafer, and calculating a structure of the element formed through a process using the assumed process condition by calculation (b);
Calculating an expected value of a plurality of measured values of the physical quantity obtained when the structure of the element obtained in the step (b) is evaluated by the optical evaluation method;
Estimating the structure of the element based on the plurality of actually measured values of the physical quantity of the element of the evaluation wafer and the predicted value of the plurality of measured values;
When the difference between the estimated structure of the element of the evaluation wafer and the design structure of the plurality of wafers is out of an allowable range, at least a wafer other than the evaluation wafer among the plurality of wafers. Correcting the process conditions of the process (e) and
Including
The above process, Ri epitaxial growth process der crystal film on the substrate on which a pattern is formed,
The crystal film is a crystal film that includes Si and Ge and includes a structure in which the band gap changes in an inclined manner,
The method of manufacturing a semiconductor device, wherein the process condition is a temperature of the substrate . A recording medium embeddable in a computer used for characterization of elements of a semiconductor device by an optical evaluation method,
A procedure (a) for capturing a plurality of actual measured values of physical quantities of the elements of the semiconductor device;
Assuming a process condition for forming the element, and calculating a structure of the element formed through a process using the assumed process condition by calculation (b);
A procedure (c) for calculating an expected value of a plurality of measured values of the physical quantity obtained when the structure of the element obtained in the procedure (b) is evaluated by the optical evaluation method;
Recording a program for causing a computer to execute the step (d) of estimating the structure of the element based on the plurality of actual measurement values of the physical quantity of the element and the predicted value of the plurality of measurement values ;
The above process is an epitaxial growth process of a crystal film on a substrate on which a pattern is formed,
The crystal film is a crystal film that includes Si and Ge and includes a structure in which the band gap changes in an inclined manner,
The process condition is a computer-readable recording medium which is the temperature of the substrate .

Images