JP2004040038A - How to clean the mask - Google Patents

Abstract

A method for easily and reliably detecting contamination of a mask and cleaning the mask accordingly.
In A method, in a step S _1, the experiment of foreign matter defect transferability, or simulation by determining the critical size. In step S _2, the small section area outside the exposure area set on the mask surface and then dividing the subregions into m monitor minute area. In step S _3, to stay mask in the column between an exposure apparatus for a predetermined time t, then observing the presence or absence of a foreign object within a micro region for the m monitoring, monitor minute domains critical size or foreign matter is present Is obtained. In Step S _4, obtaining the defect occurrence probability p of the mask than the number x of the monitor minute domains foreign matter is present. In step _{S 5,} determine the _{T c} for the threshold _{p c} of the probability that the product yield below the predetermined value _p, to set a washing cycle T as _T c = T. In step S _6, to clean the mask cleaning period T.
[Selection] Fig. 2

Description

【数２】

【数３】

従って、式（４）が成り立つ。
【数４】

式（４）の解は、式（５）となる。
【数５】

【００２２】
つまり、特定領域が生き残る確率Ｐ（ｔ）は、時間と共に指数関数的に減少する。よって、マスクの露光領域全体（面積Ｓとする）が生き残る確率Ｐ_ｔｏｔ（ｔ）は、式（６）で表示される。
【数６】

【００２３】
Ｐ_ｔｏｔ（ｔ）が基準値（閾値）Ｐｃを下回る時間をＴとすると、式（６）にＰｃ、及びＴを代入して、式（６′）となるので、式（７）によりＴが求まる。
Ｐｃ＝ｅｘｐ〔−ＳτＴ〕　　　　　　　　　　　　式（６′）
【数７】

以上のように、Ｐ_ｔｏｔ（ｔ）が基準値Ｐｃなった時点でマスクを洗浄するということを条件とするならば、式（６）から求められる式（７）を満たす周期Ｔでマスクを洗浄すれば良いことが分かる。
【００２４】
本発明では、マスクの露光領域でなくモニター用の小領域を観察するので、モニター用の小領域の面積ｓに対応した時定数τ（＝τ’　）を求め、モニター用の小領域を観察した結果とτ’　より式（７）を用いてＴを求めることが本発明の基礎となる基本的な原理である。
【００２５】
上記目的を達成するために、本発明に係るマスクの洗浄方法は、リソグラフィ工程で用いるマスクを洗浄してマスクに付着した異物を除去する際に、
マスクの露光領域以外の領域に小区域の監視領域を設定する第１のステップと、
リソグラフィ工程で使用する描画装置のカラム内にマスクを所定時間保持して監視領域への異物の付着の有無を観察し、異物の付着確率を実験的に求める第２のステップと、
実験的に求めた異物の付着確率に基づいて、マスクに付着した異物に起因してリソグラフィ工程で転写パターンに欠陥が発生する欠陥発生確率を求め、欠陥発生確率に基づいてマスクの洗浄周期を決定する第３のステップと
を有し、第３のステップで決定した洗浄周期に従ってマスクを洗浄することを特徴としている。
【００２６】
第２のステップで監視領域への異物の付着の有無はＳＥＭ等で観察する。ＳＥＭ等で微小パーティクル状の異物を検査するので、マスク検査の信頼性が非常に高くなる。
本発明方法で、描画装置のカラムとは、露光ビームを形成する機構を収め、かつマスクを保持するカラム状の室を言う。通常、真空下に維持されている。
【００２７】
本発明方法の好適な実施態様では、第１のステップでは、小区域の監視領域を更に複数個のモニター用微小領域に区画し、
第２のステップでは、マスクに付着した異物が転写パターンに欠陥を生じさせる異物の最小寸法として定義される臨界サイズを決定し、次いで臨界サイズ以上の異物が付着している微小領域の付着領域数を計数して実験的に付着確率を求め、
第３のステップでは、実験的に求めた付着確率から統計学的処理により欠陥発生確率を求める。
【００２８】
本発明方法は、描画方法の方式、描画装置の種類に制約無くそれらで使用されるマスクの洗浄に適用でき、特にＬＥＥＰＬ等で用いるステンシル等倍マスクや、Ｘ線リソグラフィで用いる等倍マスクの洗浄に好適に適用できる。
本発明方法では、洗浄周期Ｔを設定した後は、都度、洗浄周期Ｔを設定する必要はなく、同じ洗浄周期Ｔでマスクを洗浄することにより、転写パターンに生じる欠陥発生確率を所望の値以下にすることができる。
【００２９】
ここで、洗浄周期Ｔを求め、洗浄周期Ｔが到達する毎に、定期的にマスクを洗浄する本発明方法の利点について、図５を参照して説明する。
最初に洗浄周期Ｔを求めておき、洗浄周期Ｔ毎に定期的にマスクを洗浄する本発明方法では、マスクをパターン形成に使用することのできる実質的な時間は、図５（ａ）に示すように、左下がり斜線部分を除いた時間である。図５（ａ）の左下がり斜線部分は洗浄周期Ｔを求めるために要する時間を示す。
一方、マスク露光の所定の経過時間毎にマスクを洗浄する従来のやり方では、マスクをパターン形成に使用することのできる実質的な時間は、図５（ｂ）に示すように、右下がり斜線部分を除いた時間である。図５（ｂ）の右下がり斜線部分はマスクの汚染検査に要する時間を示す。
従って、本発明方法を適用することにより、図５（ｂ）の検査時間（右下がり斜線部分）−図５（ａ）図のＴを求めるための時間（左下がり斜線部分）だけ、マスク露光に使える時間が増大する。
【００３０】
本発明方法は、微小領域をモニターして臨界サイズ以上の異物が付着しているかどうか判断しているので、本発明方法の検査時間が、マスク全面を検査する従来の方法の検査時間に比べて短くなる。
図５では、マスクの全面を汚染検査する従来のやり方の検査時間（図５（ｂ）の右下がり斜線部分）の一回分と図５（ａ）のＴを求めるための時間（左下がり斜線部分）の幅の長短で、本発明方法の検査時間が短いことを図示している。
尚、図５では、洗浄時間は、本発明方法でも従来の方法でも同じ時間を要するので、図示していない。
以上のことから、同じ時間当たりのパターン形成の処理量、つまりスループットは、本発明方法を用いるほうが大きくなる。
【００３１】
【発明の実施の形態】
以下に、添付図面を参照して、実施形態例に基づいて本発明をより詳細に説明する。
実施形態例
本実施形態例は、本発明に係るマスクの洗浄方法の実施形態の一例であって、ＬＥＥＰＬに使用するマスクに適用したものである。図１は本実施形態例の方法を適用して洗浄するマスクを使用するＬＥＥＰＬの構成を示す概念図である。
ＬＥＥＰＬシステム１０は、等倍近接露光装置であって、図１に示すように、電子ビームＢを出射する電子銃１２、電子ビームＢを平行なビームにするコンデンサレンズ１４、電子ビームＢの広がりを制限するアパーチャー１６Ａ、Ｂ、対として構成されているメインデフレクターのセット１８、２０、及び電子ビームの微調整を行う対の微調整用デフレクター２２、２４を有する。
【００３２】
メインデフレクターのセット１８、２０は、電子ビームＢが平行なままでラスターまたはベクトル走査モードの何れかで、且つステンシルマスク２６に垂直に入射するように偏向させる機能を有する。
ステンシルマスク２６は、ウエハＷから間隔Ｄだけ（Ｄ＝約５０μｍ）離隔してウエハＷ上に配置される。ステンシルマスク２６の厚みＴは約５００ｎｍである。ここで、ステンシルマスク２６の厚みＡとは、所謂メンブレン部の厚みをいい、メンブレン部を支える梁がある場合は、梁の厚みは例えばマスクの基板となるウエハ厚と等しいと考えられる。
【００３３】
次に、式（７）が本発明方法の基本的原理であるということを前提として、図２を参照し、実際にＬＥＥＰＬで使用するステンシルマスク２６（以下、マスクと言う）の欠陥発生確率からＴを求める具体例を説明する。図２は本実施形態例の方法を実施する際の手順を示すフローチャートである。
第１段階として、露光装置、つまりＬＥＥＰＬ内で生じる異物（パーティクル）がマスク上に付着し、致命的な欠陥になる確率（欠陥発生確率）を求める。ここで、致命的な欠陥とは、それがウエハ上に転写され、製造された半導体装置の性能、動作を阻害してしまう大きさの欠陥を言う。
第２段階として、欠陥発生確率に基づいて、マスクの洗浄周期Ｔを求める。
【００３４】
第１段階では、先ず、マスク上に落下する異物数はマスクの使用時間に比例するとし、マスク上の異物数がゼロの初期状態から、時間ｔの間にマスクの露光領域（面積Ｓ）上に落下する異物数をｎｔとする。
描画装置内からマスク上に落下する異物数がマスクの使用時間に比例するという仮定は、装置トラブルによる突発的な異物の増加などの特殊な場合を除いては、極く自然な仮定である。
【００３５】
ステップＳ_１で、異物欠陥転写性の実験、又はシミュレーションにより臨界サイズを決定する。臨界サイズとは、マスクに付着した異物が転写パターンに欠陥を生じさせる異物の最小寸法である。
致命的な欠陥になる可能性のある臨界サイズ以上の異物の異物全体に対する割合をαとすると、面積Ｓのマスクの露光領域上に落下した異物のうち致命的な異物数Ｎは、Ｎ＝αｎｔである。
マスクの露光領域Ｓを仮想的にＭ個の小領域に分割して考える。任意に取り出したある小領域に致命的な異物が存在しない、即ちその小領域が生き残っている、つまりマスクとして小領域が機能する確率Ｐ_０は、ウエハに関する歩留理論で議論されている様々なモデルに基づいた公式で表現できる（例えば、Ｓ．　Ｍ．　Ｓｚｅ，
ＶＬＳＩ　Ｔｅｃｈｎｏｌｏｇｙ，　２ｎｄ　Ｅｄｉｔｉｏｎ，　Ｃｈａｐ　ｔｅｒ　１４　）。
【００３６】
最も単純なポアソンモデルでは、Ｐ_０は式（８）で表示される。
【数８】

式（８）は、式（６）とは指数の肩の定数が変わっているが、原理的には時間に対して指数関数的に歩留まりが下がるという意味で同じである。
【００３７】
また、シーズモデルでは、式（９）となる。確率Ｐ_０は時間と共に漸近的に０に近づいていく。
【数９】

【００３８】
少なくとも一つの小領域が死んでしまう確率、つまり一つの小領域がマスクとして使用できなくなる確率ｐは、式（１０）で表される。
【数１０】

確率ｐに対する閾値ｐ_ｃを例えばｐ_ｃを０．１とすると、確率ｐがｐ_ｃ以下であるうちに、又はｐ_ｃに到達した時点でマスクの再洗浄を行うことにすれば、異物による歩留低下を所定の値以下に小さくすることができる。ｐ_ｃは、以下の式（１０′）で表される。
ｐ_ｃ＝１−｛Ｐ_０（Ｔ_ｃ）｝^Ｍ　　・・・・・　　式（１０′）
ここで、Ｔ_ｃを洗浄の特性時間と呼ぶ。特性時間Ｔ_ｃが経過する前にマスクを洗浄すれば、製品歩留りを所定の値にできる時間、即ち洗浄周期Ｔである。つまり、洗浄周期Ｔの最大値は特性時間Ｔ_ｃである。
【００３９】
例えば、４０ｍｍ角の露光領域をもつＬＥＥＰＬ用のマスクの場合、これを１００μｍ角の小領域に分割すると、Ｍ＝１６００００となる。
シーズモデルを用いると、式（９）で（αｎＴ_ｃ）／Ｍ＜１であることと、式（１０′）でｐ_ｃ＝０．１であることより、
αｎＴ_ｃ≒０．１　　　　　　　　・・・・・　　式（１１）
となる。
【００４０】
真空下で露光する露光装置、例えばＬＥＥＰＬの場合、マスク搬送時はともかく、露光中の異物落下はきわめて少ないと考えられる。仮に１日に１個の異物が面積Ｓのマスクの露光領域上に落下し、臨界サイズ以上のものの割合を１０％とすると、式（１１）よりＴ_ｃはほぼ１日（２４時間）となる。
つまり、１日１回定期洗浄を行うことにより、異物起因する製品の半導体装置の歩留低下を１０％以下に保つことができる。
洗浄の特性時間Ｔ_ｃを決定するためには、露光領域上で、後述するモニター用微小領域のうち少なくとも一つのモニター用微小領域が死んでしまう確率ｐを求めることが必要である。この確率ｐの値を、露光領域外の小区画領域の異物観察結果から推定する。
【００４１】
そのために、ステップＳ_２で、図３に示すように、露光領域外の小領域をマスク面上に設定し、次いで小領域をｍ個のモニター用微小領域に分割する。
例えば、図３に示すように、露光領域外の小領域をマスク上に配置し、露光領域と同様に、この小領域のメンブレンもｍ個の面積ｓのモニター用微小領域に分割する。例えば、１ｍｍ角の小区画メンブレンを１００μｍ角のモニター用微小領域に分割すると、ｍ＝１００になる。
【００４２】
ステップＳ_３で、マスクを所定時間ｔの間露光装置のカラム内に滞在させ、その後、ｍ個のモニター用微小領域内の異物の有無を観察し、臨界サイズ以上の異物が存在するモニター用微小領域の個数ｘを求める。
次いで、ステップＳ_４で、異物が存在するモニター用微小領域の個数ｘより少なくとも一つのモニター用微小領域が死んでしまう確率ｐを求める。確率ｐを求めるには、先ず、確率ｐの推定値ｐ′を求める。
【００４３】
ｘ個のモニター用微小領域に致命的な異物が存在している確率ｐの推定値ｐ′は、ｐ′＝ｘ／ｍである。ｐ′は確率分布をもち、ｍが充分大きい場合、平均値μ＝ｐ、分散ｏ^２　＝ｐ（１−ｐ）／ｍの正規分布で近似できる（宮川公男、基本統計学）。
例えば、ｍ＝１００、ｘ＝２の場合、区間推定を行うことにより、９５％の確からしさで、０．００６＜ｐ＜０．０７であることを推定することができる。
【００４４】
次いで、ステップＳ_５で、製品歩留りを所定の値以下にする、確率ｐの閾値ｐ_ｃに対するＴ_ｃを求め、Ｔ_ｃ＝Ｔとして洗浄周期Ｔを設定する。
ステップＳ_５では、先ず、歩留モデル式（８）又は（９）のパラメーターＭを決定する。本実施形態例では、安全を見るために、上限の値、例えば０．０７を時間ｔでのｐ（ｔ）として式（１２）を解くことにより、歩留モデルのパラメーターＭを決定する。
【数１１】

例えば、シーズモデルの場合、前述の様にαｎｔを決定することができる。次に、決定したＭをパラメーターとして用いて式（１３）を解けば、特性時間Ｔ_ｃを決定することができる。
【数１２】

次いで、ステップＳ_６で、洗浄周期Ｔでマスクを洗浄する。
【００４５】
以上の本実施形態例の方法により、高精度マスクの半導体装置製造への適応が容易になり、特にステンシル等倍マスクを用いるＬＥＥＰＬ技術や、等倍マスクなどを用いるＸ線リソグラフィなど次世代のリソグラフィ技術を有効に活用することが可能になり、例えば１００ｎｍルール以降の半導体回路形成の量産化に役立ち、半導体産業に貢献することが出来る。
また、等倍近接露光のＬＥＥＰＬ用マスク等の洗浄に好適に適応できる。ＬＥＥＰＬ用マスクは、マスクの作製メーカーだけでなく、半導体メーカーも自前で、しかも短期間でマスクを作成することが可能であり、半導体メーカーが必ずと言っていいほど持っているＳＥＭ等の検査装置を用いて、マスク検査を行うことも可能であり、検査装置も含めたマスク製造装置トータルのコストを抑えることができる。
【００４６】
【発明の効果】
本発明方法によれば、小画の監視領域を設け、監視領域への異物の付着を検査することにより、マスク上のパターン形成の欠陥発生率を統計的に推定し、それによりマスクの洗浄周期を決定している。本発明方法を適用することにより、ＴＡＴの短い簡便で確実なやり方でマスクの洗浄周期を設定し、その洗浄周期に従って、マスクを洗浄することができる。
また、ＳＥＭ等で微小パーティクル状の異物の付着を検査するので、マスクの信頼性が高くなる。更には、小画の監視領域のみＳＥＭ等により検査するので、マスク検査のスループットが大きく、かつ一度、洗浄周期を決定すれば、以降のマスク検査を行う必要がないので、低コストでマスクを良好な状態に維持することができる。
【図面の簡単な説明】
【図１】実施形態例の方法を適用して洗浄周期を決定するマスクを使用するＬＥＥＰＬの構成を示す概念図である。
【図２】実施形態例の方法を実施する際の手順を示すフローチャートである。
【図３】マスク面にモニタ用微量領域を設定することを説明する模式図である。
【図４】マスク汚染の原因を説明する模式図である。
【図５】図５（ａ）及び（ｂ）は、本発明方法及び従来の方法について、それぞれ、洗浄と製品歩留りとの関係を示すグラフである。
【符号の説明】
１０……ＬＥＥＰＬシステム、１２……電子銃、１４……コンデンサレンズ、１６……アパーチャー、１８、２０……メインデフレクターのセット、２２、２４……微調整用デフレクター、２６……ステンシルマスク。[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a method of cleaning a mask, and more particularly, to a method of cleaning a mask based on the inspection result after inspecting the mask for contamination, and replacing the conventional method of cleaning a mask with a short, simple and reliable TAT. The present invention relates to a mask cleaning method in which a mask cleaning cycle is set in a simple manner, and the mask is cleaned according to the cleaning cycle.
[0002]
[Prior art]
One of the evaluations of a semiconductor device is the degree of integration of how many circuits are integrated in the semiconductor device, which largely depends on the size of a circuit pattern formed on a substrate.
In recent years, in order to meet demands for higher integration of semiconductor devices, finer circuit patterns, and the like, semiconductor integrated circuit manufacturing technology has been remarkably developed in recent years.
In connection with a semiconductor integrated circuit manufacturing technique, one of the manufacturing processes of a semiconductor device such as an IC or an LSI includes a lithography process of forming a fine circuit pattern on a semiconductor substrate. In the lithography process, conventionally, a photolithography method using ultraviolet light has been generally applied.
[0003]
By the way, as the circuit pattern is further miniaturized, the resolution limit of light starts to be concerned, and lithography technology with higher resolution using charged particle beams such as electron beams and ion beams and X-rays is attracting attention. .
For example, an exposure technique using a charged particle beam has a great feature in that a fine pattern of 100 nm or less can be easily formed by reducing the beam diameter to the order of nm. Among them, the electron beam lithography technology has been practically used for a long time.
[0004]
However, in the direct drawing method using a so-called electron beam or the like, which draws while scanning such an extremely narrow electron beam, it takes an enormous amount of time to draw on a large area or to form a large pattern. In other words, there is a problem that the throughput decreases, that is, the processing amount per unit time decreases.
For this reason, in the manufacture of semiconductor integrated circuits, photolithography using ultraviolet light as a light source is still used as a mainstream lithography method. It has been used only in limited areas, such as next-generation trial device prototypes that cannot be accommodated by lithography.
[0005]
In order to solve such a low throughput, instead of directly drawing with a Gaussian-shaped electron beam as in the past, a predetermined pattern is formed on a substrate using an electron beam variably shaped via an electron optical system. In the 1980's, a method of drawing directly on a computer appeared.
In the 1990s, a lithography technique for reducing a partial collective pattern and drawing on a wafer substrate by a method called block exposure or a cell projection method appeared in the 1990s. P177 and FIG. 5).
With these technological advances, the throughput of electron beam direct writing has been dramatically improved.
[0006]
Furthermore, SCALPEL (see www.lucent.co.jp/press/99#2#5.html, etc.) developed by Lucent Technology, etc., and PREVAIL developed jointly by IBM and Nikon Corporation (See "Project \ Exposure \ with \ Variable \ Axis \ Immersion \ Lenses \: \ A \ High-Throughput \ Electron \ Beam \ Approach \ E.P.P.E.P.E.P.E.P.E.P.E.P.E.P.E.P.E.P.E.P.E.P.E.P.E.P.E.P.E.P.E.P. It is believed that line lithography can further increase throughput.
[0007]
However, in these electron beam reduction drawing methods, it is necessary to increase the energy of the electron beam in order to converge the electron beam as desired and create a sharp image. For this reason, the energy of the electron beam in the partial block exposure and the cell projection method described above is generally 50 keV, but the energy of the electron beam is as high as 100 keV in the electron beam reduced writing method.
At such a high energy, the mechanism for controlling the electron beam optical system also becomes large-sized, and there is a disadvantage that the cost of the drawing apparatus is particularly increased. In addition, in the case of a high energy electron beam, electrons pass through the resist film without emitting much energy in the resist film on the wafer, so that the resist sensitivity per number of electrons decreases.
[0008]
Therefore, as long as the resist has the same sensitivity, the higher the energy of the electron beam, the larger the required amount of electron beam current and the higher the electron density in the electron beam. As the electron density in the beam increases, the dilemma of defocusing the electron beam and causing degradation in pattern resolution occurs.
Further, as the amount of electron beam current increases, the influence of the proximity effect, which causes distortion in a formed pattern as a result of backscattering from the wafer under the resist film to the resist film, increases. Furthermore, as the amount of electron beam current increases, the mask, the resist film, and even the wafer are heated, and the distortion of the formed pattern is increased.
Therefore, in the above-mentioned electron beam reduced writing method, while it is necessary to limit the amount of electron beam current in order to maintain required pattern accuracy, there is a trade-off relationship that when the amount of electron beam current is limited, the throughput decreases. there were.
[0009]
In order to avoid these problems, an exposure method for forming a pattern with a low energy electron beam has been developed. The development of an exposure method using a low-energy electron beam has been reported by Utsumi, who reports that the proximity effect is substantially reduced with a low-energy electron beam: "Low-voltage-alternative-for-electron-beam-lithography" (J Vac. Sci {TechB {10} (6), November / December {3094-3098}).
Based on the theory of this paper, the technology disclosed in U.S. Pat. No. 2,951,947 by Sony Corporation, LEEPL and Tokyo Seimitsu Co., Ltd. LEEPL (see Low Energy E-beam Proximity Projection Lithography: www.leepl.com and Nikkan Kogyo Shimbun / December 4, 2000), which is a new lithography technique using an electron beam, is being developed.
[0010]
In LEEPL, the energy of the electron beam is about 1-4 keV, preferably about 2 keV.
Also, in LEEPL, since the pattern of the mask uses so-called equal-size proximity exposure that is the same as that of the pattern on the wafer, the mask is arranged at a position very close to the wafer, about 50 μm above the wafer covered with the resist film. Cage. Further, in order to form an ultrafine pattern of, for example, 100 nm or less, it is necessary to form an ultrafine pattern of 100 nm or less on the mask.
[0011]
[Problems to be solved by the invention]
By the way, in the close-to-one-size exposure method represented by LEEPL, as described above, since the mask is arranged close to a position of about 50 μm from the wafer, particles generated from the resist film on the wafer during pattern writing are formed. Is more likely to adhere to the mask as compared to a mask arranged at a distance from the wafer.
Then, when particles adhere to these masks serving as a master for manufacturing a semiconductor device to form a mask defect, the mask defect is transferred as a defect to all semiconductor devices patterned using this mask. If a defect transferred to a semiconductor device exists as a defect that affects the performance of the semiconductor device, the semiconductor device becomes a defective product.
For this reason, it is usually necessary to inspect all the masks used as the original masters for foreign substances such as carbon-based contaminants as mask defects.
[0012]
Further, as the miniaturization of the semiconductor device progresses, the size of the problematic defect also becomes finer. In particular, in the electron beam equal-size exposure and X-ray equal-size exposure represented by the above-described LEEPL and the like, the pattern size of the mask is the same as the pattern size of the semiconductor device. become.
Since it is technically difficult to inspect defects caused by such minute foreign matter by an optical method, it is necessary to perform a defect inspection of a mask by a method using a charged particle beam such as an SEM or FIB. It is becoming.
[0013]
Inspection using a charged particle beam requires a longer inspection time than inspection using an optical method. The reason is that the inspection using a charged particle beam requires inspection of a mask to be inspected in a vacuum in order to inspect using charged particles (electrons in the case of SEM) as a probe. It takes a long time to evacuate the inspection room of the inspection device that accommodates the laser beam, and if a scanning method such as SEM is used, in order to inspect a minute defect, a probe beam such as an electron beam is narrowed and scanned. It is necessary.
[0014]
In particular, when the magnification is increased to increase the resolution, a single scan area (hereinafter, referred to as a shot) becomes smaller, so that the number of shots increases in inverse proportion to the square of the magnification for the same observation area, and the throughput increases. Becomes lower. As a result, the manufacturing cost of the semiconductor device increases due to an increase in the mask inspection TAT.
In this case, it is economically problematic to form a pattern by applying an electron beam equal-magnification exposure method in the manufacture of a semiconductor device, and it is difficult to put it into practical use.
[0015]
Therefore, in order to make pattern formation by the electron beam equal-magnification exposure method practical, it is important to simply and reliably inspect the mask for contamination and then clean and remove contaminants from the mask. Cleaning the mask with a short cleaning cycle without fear of fear is a factor that hinders the productivity of the exposure operation.
Accordingly, an object of the present invention is to provide a method for easily and reliably detecting contamination of a mask and cleaning the mask accordingly.
[0016]
[Means for Solving the Problems]
In order to achieve the above object, the present inventor has studied the mechanism of mask contamination and aimed to establish a mask cleaning cycle corresponding thereto, as described below.
Here, the mechanism by which particulate carbon-based contaminants are generated from the resist film on the wafer during drawing and adhere to the mask will be briefly described.
One type of vacuum evaporation method is an electron beam evaporation method. The electron beam evaporation method is a method in which a sample target is irradiated with a high-energy electron beam of, for example, about 10 keV in a vacuum, the sample target is heated, molecules of the sample target are vaporized, and the target is vapor-deposited.
The preferred acceleration energy of LEEPL is 2 keV, but also in LEEPL, similarly to the electron beam evaporation method, at the time of pattern drawing, the constituent molecules of the resist film on the wafer are heated and vaporized to become a low molecular weight carbon compound, Particles or layered contaminants adhere to the wafer side surface of the mask.
[0017]
Therefore, it is important to detect the presence or absence of the contaminant on the mask surface and the amount of the contaminant attached.
Incidentally, there are two types of contaminants attached to the mask as shown in FIG. The first contaminant is a contaminant generated from a resist film or the like, and the second contaminant is a contaminant due to the exposure apparatus (drawing apparatus) itself.
[0018]
Since it can be concluded that the deposition of the first contaminant is approximately proportional to the writing time of the mask, empirically and experimentally, by actually performing the writing test of the mask for a long time using the writing apparatus, The time required for a deadly amount of contaminants to adhere to the mask surface can be easily grasped.
On the other hand, the second contaminants include vacuum holding grease used in a column connection part of a drawing apparatus (exposure apparatus), foreign matter remaining in a column that has not been removed even by washing at the time of starting the apparatus, For example, contaminants such as organic matter and fine metal dust adhere to the mask, in other words, are contaminants that adhere to the mask in a falling image.
[0019]
The inventor of the present invention has conceived of determining the cleaning cycle of the mask by estimating the probability of occurrence of a defect due to so-called falling foreign matter shown in FIG.
The inventor of the present invention determines the cleaning cycle T of the mask from the result of foreign matter observation in a small area on the mask. The cleaning cycle T is generally a characteristic time related to the mask defect occurrence probability, and it has been found that the cleaning cycle T can be obtained according to the following equations (6) and (7).
Hereinafter, a process of sequentially obtaining the equation (6) will be described.
[0020]
First, let τdt be the probability that a foreign substance having a fatal size that causes a defect in a semiconductor device will adhere to a unit area (hereinafter, referred to as a specific region) on a mask at time dt.
Assuming that the probability that the specific area survives as usable after the elapse of the time t, that is, the probability that a fatal foreign object does not adhere is P (t), the following equation (1) is used to formula (3).
[0021]
(Equation 1)

(Equation 2)

(Equation 3)

Therefore, equation (4) holds.
(Equation 4)

The solution of equation (4) becomes equation (5).
(Equation 5)

[0022]
That is, the probability P (t) that the specific area survives decreases exponentially with time. Therefore, the probability P that the entire exposure region (area S) of the mask will survive_tot(T) is represented by equation (6).
(Equation 6)

[0023]
P_totAssuming that the time during which (t) is less than the reference value (threshold) Pc is T, Pc and T are substituted into Expression (6), and Expression (6 ′) is obtained, so T is obtained from Expression (7).
Pc = exp [−SτT] Equation (6 ′)
(Equation 7)

As described above, P_totIf the condition is that the mask is to be cleaned when (t) reaches the reference value Pc, it can be seen that the mask should be cleaned at a period T that satisfies Expression (7) obtained from Expression (6).
[0024]
In the present invention, since the small area for monitoring is observed instead of the exposed area of the mask, the time constant τ (= τ ′) corresponding to the area s of the small area for monitoring is obtained, and the small area for monitoring is observed. Obtaining T from the result and τ ′ using equation (7) is a basic principle underlying the present invention.
[0025]
In order to achieve the above object, the mask cleaning method according to the present invention, when cleaning the mask used in the lithography process to remove foreign substances attached to the mask,
A first step of setting a small area monitoring area in an area other than the exposure area of the mask;
A second step of holding a mask in a column of a writing apparatus used in a lithography process for a predetermined time, observing whether or not foreign matter has adhered to a monitoring area, and experimentally calculating the probability of foreign matter adhesion;
Based on the experimentally determined foreign matter adhesion probability, the defect occurrence probability of causing a defect in the transfer pattern in the lithography process due to the foreign matter adhering to the mask is determined, and the mask cleaning cycle is determined based on the defect occurrence probability The third step to
And cleaning the mask in accordance with the cleaning cycle determined in the third step.
[0026]
In the second step, the presence or absence of foreign matter adhering to the monitoring area is observed with an SEM or the like. Since a minute particle-like foreign substance is inspected by an SEM or the like, the reliability of the mask inspection becomes very high.
In the method of the present invention, the column of the drawing apparatus refers to a column-shaped chamber that holds a mechanism for forming an exposure beam and holds a mask. Usually, it is kept under vacuum.
[0027]
In a preferred embodiment of the method of the present invention, in the first step, the monitoring area of the small area is further divided into a plurality of monitoring small areas,
In the second step, a critical size is defined, which is defined as the minimum size of foreign matter that causes foreign matter adhering to the mask to cause a defect in a transfer pattern. Is calculated experimentally to determine the adhesion probability,
In the third step, a defect occurrence probability is determined by statistical processing from the adhesion probability experimentally determined.
[0028]
The method of the present invention can be applied to cleaning of a mask used in a drawing method without limitation on a method of the drawing method and a type of a drawing apparatus. It can be suitably applied to
In the method of the present invention, after setting the cleaning cycle T, it is not necessary to set the cleaning cycle T each time. By cleaning the mask in the same cleaning cycle T, the probability of occurrence of defects occurring in the transfer pattern can be reduced to a desired value or less. Can be
[0029]
Here, the advantage of the method of the present invention in which the cleaning cycle T is obtained and the mask is periodically cleaned each time the cleaning cycle T reaches will be described with reference to FIG.
In the method of the present invention in which the cleaning cycle T is first determined and the mask is periodically cleaned at each cleaning cycle T, the substantial time during which the mask can be used for pattern formation is shown in FIG. Thus, it is the time excluding the diagonally shaded area. The hatched portion in the lower left part of FIG. 5A indicates the time required for obtaining the cleaning cycle T.
On the other hand, in the conventional method of cleaning the mask every predetermined elapsed time of the mask exposure, the substantial time during which the mask can be used for pattern formation is, as shown in FIG. It is the time excluding. The shaded area on the lower right in FIG. 5B indicates the time required for the contamination inspection of the mask.
Therefore, by applying the method of the present invention, the mask exposure is performed only for the inspection time (hatched portion falling rightward) in FIG. 5 (b) -the time for obtaining T in FIG. 5 (a) (hatched portion falling leftward). The usable time increases.
[0030]
Since the method of the present invention monitors a micro area to determine whether or not foreign matter having a size larger than the critical size is attached, the inspection time of the method of the present invention is shorter than the inspection time of the conventional method of inspecting the entire mask. Be shorter.
In FIG. 5, the inspection time of the conventional method for inspecting the entire surface of the mask for contamination (the downward-sloping hatched portion in FIG. 5B) and the time for obtaining T in FIG. The width of ()) indicates that the inspection time of the method of the present invention is short.
In FIG. 5, the cleaning time is not shown because the same time is required for the method of the present invention and the conventional method.
From the above, the processing amount of pattern formation per time, that is, the throughput, is larger when the method of the present invention is used.
[0031]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, the present invention will be described in more detail based on exemplary embodiments with reference to the accompanying drawings.
Embodiment example
The present embodiment is an example of an embodiment of a mask cleaning method according to the present invention, and is applied to a mask used for LEEPL. FIG. 1 is a conceptual diagram showing the configuration of LEEPL using a mask to be cleaned by applying the method of the present embodiment.
The LEEPL system 10 is a 1: 1 proximity exposure apparatus, and as shown in FIG. 1, an electron gun 12 for emitting an electron beam B, a condenser lens 14 for converting the electron beam B into a parallel beam, and a spread of the electron beam B. It has a limiting aperture 16A, B, a set of main deflectors 18, 20 configured as a pair, and a pair of fine deflectors 22, 24 for fine adjustment of the electron beam.
[0032]
The sets of main deflectors 18, 20 have the function of deflecting the electron beam B so that it remains parallel and in either raster or vector scanning mode and is incident perpendicular to the stencil mask 26.
The stencil mask 26 is arranged on the wafer W at a distance D (D = about 50 μm) from the wafer W. The thickness T of the stencil mask 26 is about 500 nm. Here, the thickness A of the stencil mask 26 refers to the so-called thickness of the membrane portion, and when there is a beam supporting the membrane portion, the thickness of the beam is considered to be equal to, for example, the thickness of a wafer serving as a substrate of the mask.
[0033]
Next, assuming that the equation (7) is the basic principle of the method of the present invention, referring to FIG. 2, the defect occurrence probability of a stencil mask 26 (hereinafter referred to as a mask) actually used in LEEPL will be described. A specific example of obtaining T will be described. FIG. 2 is a flowchart illustrating a procedure when the method according to the present embodiment is performed.
As a first step, the probability that foreign matter (particles) generated in the exposure apparatus, that is, the LEEPL, adheres to the mask and becomes a fatal defect (defect occurrence probability) is determined. Here, a fatal defect refers to a defect having a size that is transferred onto a wafer and impairs the performance and operation of a manufactured semiconductor device.
As a second stage, a mask cleaning cycle T is obtained based on the defect occurrence probability.
[0034]
In the first stage, first, it is assumed that the number of foreign particles falling on the mask is proportional to the use time of the mask. From the initial state in which the number of foreign particles on the mask is zero, the exposure area (area S) of the mask is changed during time t. Let nt be the number of foreign objects falling on the surface.
The assumption that the number of foreign substances falling on the mask from the inside of the drawing apparatus is proportional to the mask usage time is a very natural assumption, except for a special case such as a sudden increase in foreign substances due to an apparatus trouble.
[0035]
Step S₁Then, the critical size is determined by experiment or simulation of foreign matter defect transferability. The critical size is the minimum size of the foreign matter that causes the foreign matter attached to the mask to cause a defect in the transfer pattern.
Assuming that the ratio of foreign matter having a critical size or more that can become a fatal defect to the total foreign matter is α, the number N of fatal foreign matter out of foreign matter that has fallen on the exposure region of the mask having the area S is N = αnt It is.
It is assumed that the exposure region S of the mask is virtually divided into M small regions. There is no fatal foreign matter in a small area arbitrarily taken out, that is, the small area survives, that is, the probability P that the small area functions as a mask₀Can be expressed by formulas based on various models discussed in the yield theory for wafers (for example, S. @ M. @ Sze,
VLSI {Technology, {2nd Edition, {Chapter}}}.
[0036]
In the simplest Poisson model, P₀Is represented by equation (8).
(Equation 8)

Equation (8) is different from Equation (6) in that the constant of the shoulder of the exponent is changed, but is the same in principle in that the yield decreases exponentially with time.
[0037]
In the seeds model, the equation (9) is used. Probability P₀Approaches 0 asymptotically over time.
(Equation 9)

[0038]
The probability p that at least one small region dies, that is, the probability p that one small region cannot be used as a mask is expressed by Expression (10).
(Equation 10)

Threshold p for probability p_cTo p_cIs 0.1, the probability p is p_cOr p_cIf the mask is re-cleaned at the time when the temperature reaches the threshold value, it is possible to reduce the yield reduction due to the foreign matter to a predetermined value or less. p_cIs represented by the following equation (10 ′).
p_c= 1- ｛P₀(T_c)｝^M・ ・ ・ ・ ・ ・ ・ Formula (10 ')
Where T_cIs called the characteristic time of cleaning. Characteristic time T_cIf the mask is cleaned before the time elapses, the time required for the product yield to be a predetermined value, that is, the cleaning cycle T. That is, the maximum value of the cleaning cycle T is the characteristic time T_cIt is.
[0039]
For example, in the case of a LEEPL mask having an exposure area of 40 mm square, if this is divided into small areas of 100 μm square, M = 160000.
Using the seeds model, (αnT_c) / M <1 and p in equation (10 ′)._c= 0.1,
αnT_c{0.1} ... Formula (11)
It becomes.
[0040]
In the case of an exposure apparatus that performs exposure under a vacuum, for example, LEEPL, it is considered that foreign matter drops during exposure are extremely small, regardless of whether the mask is carried. Assuming that one foreign substance falls on the exposure area of the mask having the area S per day and the ratio of those having a critical size or more is 10%, the expression_cIs almost one day (24 hours).
In other words, by performing regular cleaning once a day, it is possible to keep the yield reduction of the semiconductor device of the product due to the foreign matter at 10% or less.
Cleaning characteristic time T_cIt is necessary to determine the probability p that at least one of the monitoring minute regions to be described later will die on the exposure region. The value of the probability p is estimated from the foreign substance observation result of the small section area outside the exposure area.
[0041]
Therefore, step S₂Then, as shown in FIG. 3, a small area outside the exposure area is set on the mask surface, and then the small area is divided into m monitor small areas.
For example, as shown in FIG. 3, a small area outside the exposure area is arranged on the mask, and the membrane of this small area is divided into m small areas for monitoring having an area s, similarly to the exposure area. For example, if a 1 mm square small section membrane is divided into 100 μm square monitoring minute regions, m = 100.
[0042]
Step S₃Then, the mask is allowed to stay in the column of the exposure apparatus for a predetermined time t, and thereafter, the presence or absence of foreign matter in the m monitoring minute areas is observed. Find x.
Then, step S₄Then, the probability p that at least one monitoring minute region dies is obtained from the number x of the monitoring minute regions where the foreign matter exists. To determine the probability p, first, an estimated value p 'of the probability p is determined.
[0043]
The estimated value p 'of the probability p that a fatal foreign object exists in the x monitoring minute regions is p' = x / m. p ′ has a probability distribution, and if m is sufficiently large, the average value μ = p and the variance o²It can be approximated by a normal distribution of = p (1-p) / m (Kimio Miyagawa, basic statistics).
For example, when m = 100 and x = 2, by performing section estimation, it is possible to estimate that 0.006 <p <0.07 with 95% certainty.
[0044]
Then, step S₅, The threshold p of the probability p to make the product yield below a predetermined value_cT for_cAnd T_c= T is set as the cleaning cycle T.
Step S₅First, the parameter M of the yield model equation (8) or (9) is determined. In this embodiment, the parameter M of the yield model is determined by solving equation (12) with p (t) at time t as the upper limit value, for example, 0.07, in order to see safety.
[Equation 11]

For example, in the case of the seed model, αnt can be determined as described above. Next, by solving equation (13) using the determined M as a parameter, the characteristic time T_cCan be determined.
(Equation 12)

Then, step S₆Then, the mask is cleaned in the cleaning cycle T.
[0045]
The method of the present embodiment facilitates adaptation of a high-precision mask to the manufacture of a semiconductor device. In particular, next-generation lithography such as LEEPL technology using a stencil equal-magnification mask and X-ray lithography using an equal-magnification mask, etc. The technology can be used effectively, which contributes to mass production of semiconductor circuit formation of, for example, the 100 nm rule or later, and can contribute to the semiconductor industry.
In addition, the present invention can be suitably applied to cleaning of a LEEPL mask or the like of 1: 1 proximity exposure. LEEPL masks can be produced not only by the mask maker but also by the semiconductor maker on their own and in a short period of time, and the inspection equipment such as SEM that the semiconductor maker almost always has Can be used for mask inspection, and the total cost of the mask manufacturing apparatus including the inspection apparatus can be reduced.
[0046]
【The invention's effect】
According to the method of the present invention, a monitoring area of a small image is provided, and the occurrence of a defect in pattern formation on the mask is statistically estimated by inspecting the adherence of a foreign substance to the monitoring area. Is determined. By applying the method of the present invention, the mask cleaning cycle can be set in a simple and reliable manner with a short TAT, and the mask can be cleaned according to the cleaning cycle.
In addition, since the adhesion of the minute particle-like foreign matter is inspected by the SEM or the like, the reliability of the mask is improved. Furthermore, since only the small monitor area is inspected by SEM or the like, the mask inspection throughput is large, and once the cleaning cycle is determined, there is no need to perform the subsequent mask inspection, so that the mask can be manufactured at low cost. State can be maintained.
[Brief description of the drawings]
FIG. 1 is a conceptual diagram showing a configuration of a LEEPL using a mask for determining a cleaning cycle by applying a method of an embodiment.
FIG. 2 is a flowchart illustrating a procedure when the method according to the embodiment is performed.
FIG. 3 is a schematic diagram for explaining setting a small amount area for monitoring on a mask surface.
FIG. 4 is a schematic diagram illustrating a cause of mask contamination.
5 (a) and 5 (b) are graphs showing the relationship between cleaning and product yield for the method of the present invention and the conventional method, respectively.
[Explanation of symbols]
10 LEEPL system, 12 Electron gun, 14 Condenser lens, 16 Aperture, 18, 20 Main deflector set, 22, 24 Fine deflector, 26 Stencil mask

Claims (2)

When cleaning the mask used in the lithography process to remove foreign matter attached to the mask,
A first step of setting a small area monitoring area in an area other than the exposure area of the mask;
A second step of holding the mask in a column of a writing apparatus used in a lithography process for a predetermined time, observing whether or not foreign matter has adhered to the monitoring region, and experimentally calculating the probability of foreign matter adhesion;
Based on the adhesion probability of the foreign matter obtained experimentally, a defect occurrence probability that a defect occurs in a transfer pattern in a lithography process due to the foreign matter attached to the mask is determined, and the mask is cleaned based on the defect occurrence probability. A third step of determining a cycle, wherein the mask is cleaned according to the cleaning cycle determined in the third step. In the first step, the monitoring area of the small area is further divided into a plurality of monitoring small areas,
In the second step, a critical size is defined, which is defined as the minimum size of the foreign matter that causes the transfer pattern to have a defect in the foreign matter attached to the mask. The number of areas is counted and the adhesion probability is obtained experimentally,
2. The method according to claim 1, wherein in the third step, the defect occurrence probability is obtained by a statistical process from the experimentally obtained adhesion probability.

Images