408196 A7 B7 五、發明説明（1 ) (發明背景） ⑴發明範禱 本發明係關於一種半導體的摻雜技術，特別是一種關 於中子質化摻雜（簡稱NTD)的技術。 5 (請先聞讀背面之注$項再填寫本頁) (2)先前技術 在半導體業舆，最廣為採用的生產砂單晶法，莫過於 柴式（Czochralski *簡稱0乙）法°在0艺法中，先將超純的 多晶砂置於二氧化砂的相渦内搭融，接著*將一單晶砂種 放入此熔液中，然後再慢慢地拉出。一部份的液態砂因為 表面張力的關條而會喊著該砂種被拉出*然後，在該溶液 上方，繪微較低溫的環境下遂漸结晶。依其拉出的速度及 其他因素，直徑可達約12英寸，長約達40英寸之高琴質單 晶砂，照例I的被拉出。 經濟部標隼局負工消費合作2#印製 一般而言* CZ單晶的電阻值高於半導體製造所需之電 阻值，因之，摻雜物*如鱗原子，可被加入該熔液中，而 降低其電阻值。此種單晶砂之電阻填控制方法*有著諸多 &勺問題*尤其是電阻值的不均旬分佈° 一般來說*電阻值 由單晶的一頭到另一頭之變化相當大，在横截面之徑1¾上 也是如此。對於上述的電阻值分佈間題已在磁化柴式（MCZ )法中有了相當的改善，其中，砂單晶成長係在一沿著軸 向的磁場下（典型值約在500和4000高斯之間）完成。ώ於 溶解的砂是一電性導體，所Μ其熱對流方间將傾向於沿著 該磁力線進行。然而，kcz法雖然提供了座許的改善，但 是不均句的摻雜分佈問題依然存在。 -3 -本紙張尺度適用中國國家標準（CNS )八4規格（2丨0X297公釐） 1 5 經濟部標準局員工消費合作製 408196 A7 B7五、發明説明（2 ) 砂的原子序是14，而其原子量是28。然而，自然界存 在的砂並不全然由矽-28同位素所組成。其中，並有砂-29 的存在，比例約占4.7%的原子濃度，而矽-30佔3.1%的原 子濃度。此外，當砂-30被熱中子碰撞後，將會轉變成碟-31 (其庫子序為15)。因為所需要的麟磨子摻雑程度遠低於 矽晶中原已存在之砂—30的3.又％原子德度，所以，=^顯然 地，从一限量之中子藤射一具自然同位素成份之砂單晶， 便可將鱗原子摻雜於該砂單晶。Μ此方式摻雜之鱗原字將 均句地分佈，且將在晶格中取代部份矽原子位置成為電子 予體（donors) 〇 此CZ及MCZ摻雜砂晶的方铸，曾被先前技術揭述過， 如Tskasu等人所提出的技術（即1990年3月之美國第4,910, 156號專利）。在核武工業外能獲得的中子源，一般皆能提 供慢（熱）中子（理想地具有網量約0.025電子伏特）。其來 源包括有，直接產生中子的水池式反應爐，質子迴旋加速 器，Μ及質子直線加速铸。質子產生器一般較容易使用， 因其產生的貧子流哥从磁性操控。這些產生器所產生的快 速_子，藉由通過如鈹，自虜或鈾的物霄後，即產生快中子 。這些'决中字，接著經過周圍的減緩層（moderator)，女口 重水（D2〇)，減速後形成熱中子。 儘管中子質化摻雜（NTD)技術*應是一作均句碟原子 摻雜的理想方法*然而採用此NTD法摻雜的柴式砂單晶一 般卻無法達到所期望的結果。晶格間氧原子常出Ϊ烏於CZ及 MCZ砂單晶中，而成為雜質，究其來源，實為二氧化砂谢 一 4 - 本紙張尺度適用中國國家標準（CNS ) A4規格（2!0Χ297公釐） > (諸先閱讀背面之：¾事項再填寫本頁) 408196 A7 B7 五、發明説明（3 ) 1 5 (請先閱讀背面之注意事if再填寫本頁j 堝本身。由於其傑出的性質*晶格間氧是為微電子元件製 造過程之所需，例如，對於超大積體電路製程中所產生的 雑質進行内部去班，Μ及強化晶圓之抗熱壓特性。然而》 過多的氧原子沈殺（Oxygen Preicipitation，即锻成二氧 化矽）卻常經由NTD法中的中子撞擊而造成，而此更進而造 成砂單晶之難以預矢0的電阻值。在一些例子中，甚至發生 所有氧原子都沈澱的現象。在砂晶圓中，過量的氧沈灘導 致不欲見之少數載子壽命縮減，Μ及在逆向偏電壓下微電 子裝置出現之無可接受的大漏電流。 經濟部標率局員工消費合作_^印裝 為了使NTD法能夠成功地蓮用於CZ矽單晶，擁有本發 明的成功配方是必需的。其涵餐的參數有*最初CZ砂單晶 中的晶格間氧原子濃度，中子能譜（Μ鎘比表示），以及中 子照射量（每平方公分的總累積之中子照射量）。若不清楚 上述配方，將很可能無法預知所產出NTD CZ矽單晶的品質 。吾人認為，由於缺乏上述的知I識，Takasu等人（在1990 年3月美國專利第4,910,156號）方才建議使用許多非傳統 的紅外線探測技術，Μ作為NTD CZ晶圓品質控制T及品質保 證的方法。亦即’具有低紅夕h線穿透率的NTD CZ晶圓被認 為無法成為生產用晶圓（prime wafers)。如此*以生產後 之品質控制來面對生產過程中之不確定性，自然也無法保 證高生產良率。然而，若採用本發明之成功配方*則可保 證高品質的NTD CZ生產用晶圓之生產，因而也不需要該非 傳統式紅夕卜線探測的方法。又在Takasu等人所提的專利 申講範圍中，被允許的最大快中子照射量，是由建立在所 -5 -本紙張尺度適用中國國家標準（CNS ) A4規格（210X297公釐） 408196 A7 _B7_ 五、發明説明（4 ) 1 5 ' (請先閲讀背面之注意事項再填寫本頁) 產生之NTD晶圓上之電子元件的逆向偏壓漏電流所決定。 但是，根據本發明之成功配方所作的實驗顯示，Takasu等 所得之快中子與漏電流（其並未搮明實際數值）的關係*是 不正確的。此外，在Takasu等所敎述之快中子與慢中子的 分際是相當模糊的，但在本發明中；這卻是一個非常童要 而予嚴格定義的參數。 發明人進一步欲揭明者是，NTD法也被用於單片砂晶 圓（wafer*)的摻雜，並非只是用於一整根的矽單昂（ingot) 。例如Groves等之先前技術所述（於1993年3月所獲得乏美 國第5,212,100號專利）*然而，同樣的氧原子沈濺問題依 然存在。Μ下將詳細敘逑的本發明不僅能夠有效坤採行 NTD法來達到均句的碟原子摻雜，也能成功地避免過量氧 原子沈殺問題。 (本發明之精要） 本發昉其中之一镡目的，是提供一種CZ砂單晶之中子 質化摻雜的改善方法。 經濟部標準局負工消費合作Ί裝 本發明之次一個目的，是提供一種改善方法來避免發 生於單晶矽之傳統中子質化摻雜的過程中發生的過多氧原 子沈濺問題。 本發明之再一目的，是提供一種改善方法，Μ使得在 一個單晶砂中的氧原子沈源造成之電子效應減至最小。 本發明之又一目的，是提供一種同時结合中子質化摻 雜及氧原子控制的改善的一個方法。 本發明的進一目的，是提供—種完成前述方法的裝置。 -6 -本紙浪尺度適用中國國家標率（CNS ) Α4規格（210X297公釐） 1 5 A7 --40 只 π?__ 五、發明説明（5 ) 本發明之一個可行的目的，是允許在CZ及MCZ砂上採 用更容易獲得的低熱/快中子比（即低鎘比）的放射設施。 上述這璧目的’皆已由本發明所提供之有效配方，成 功的運用HTD法在CZ砂單晶上而達到。本發明之配方被稱 為中子增強S曲線或n-S曲線的技術，顯示了在一給定的通 常大的中子照射量下（超過每平方公分l〇lS個中子），如何 將一具有特定初始晶格間氧濃度的C2單晶砂，配合使用一 個已知鎘比（CR)(熱/快中子比）的特定中子源。雖然f宍中 子與熱（慢）中子的分際在本案中是極為重要的，但是從一 輝典型的中子源之連練能量光譜中來狹智地分別定義快中 子與熱中子是不切實際的。反之*於本發明中，此項分際 是由嚴格定義鎘比（CR)來達成。 值得注意的是，根據本發明之η-S曲線（當一個低CR之 中子源的快中子成份造成了大量的晶格空位時）顯示對於 一個具有初始晶格間氧含量9到14ppma的CZ單晶砂而言， 低CR中子源（實際值一般在5到20之間）的採用卽已可產生 令人滿意的NTD CZ砂單晶。因此，僅藉由使用容易獲得的 低CR (換句話說，高快中子成分）中子源，即可以輕易地避 免過多氧原子沈澱問題。而且，由實驗證實如Takasu等所 敏述之漏電流問題並未發生。通常*高CR (約100到5,000) 中子源是較難獲得的。傳統上對於NTD CZ砂一直認為應採 用高CR中子源的原因是由於快中子成份傾向於損害單晶矽 晶格，會因此造成助長過多氧原子沈殺的種子區（nuclea-tion centers) 〇 一 Ί 一 本紙張尺度適用中國國家標準（CNS ) Α4規格（210 X 297公釐） ^1 —^1 l^i ^^^1 nn —^n ^^^1 t^t ,(請先閲讀背面之注意事項再填寫本ί 經濟部_標準局員工消費合作$製 408136 B7 五·、發明説明（6 ) 1 (圖式說明） 圖1表示一個已經歷一般熱循環之非NTD 〇2單晶之一 個典型的氧原子減損額與初始氧濃度（含暈）關係之”S曲線 ” Ο 5 (請先閲讀背面之注意事項再填寫本頁) 圖2表示f使用於本發明的中子源之一些主要部份。 圖3表示，一個用於照射砂單晶之裝置。 圖4，是根據本發明所得•採用NTD CZ於砂單晶的氧 原子減損額與初始氧繪度（含量）之關係曲線（n-S曲線）。 圖5顯示*圖3中之矽單晶被分成許多小柱，並將其截 面面對中子流的剖面裝置圖。 (較佳實施例之詳細說明） 在無中子照射情形下，砂中氧原子沈澱的主要機制（ 在前述熱處理下），為下式所表示： 2Si + 20i Si〇2 + Six....... ♦ 〇> 其中下標I表示”晶格間”。 經濟部標準局員工消費合作製 圖1是一典型的”S曲線”，其表示了非NTD CZ之生産用 晶圓，經幾種典型熱循環後，氧原子減損額（M每百萬原 子分之一（PPma)表示）與初始氧濃度的關係；圖1中之曲線 1，是先經過8小時8000熱處理後*再經過1小時1，050°C 處理；曲線2是經過2小時800¾，再經過15小時1,〇50°C ; Μ及曲線3是經由4小時750T： *再經由16小時l，〇5〇t：。由 此可知*在微電子製造過程之典型熱循環中’ 一般的C乙晶 圓中的氧原子並不會全部沈殺，尤其是當其所含初始氧原 子濃度相當低時。 -8 -本紙浪尺度適用中國國家標準（CNS ) Α4規格（210X297公釐） 408196 at Β7 五、發明説明（7 ) 1 當有中子照射時，另外一個反應機制便會造成可觀的 效應。即是 vacancy + Sii + 201 -> SI〇2......(2) 其中之(空位）即是由‘陕中子成份造成。 5 (請先閱讀背面之注意事項再填寫本頁) 假如不依循適當的配方，而將NTD法應用於CZ5夕單晶 ，將很容葛導致如方程式②的反應，而產生不欲見的過量 氧原子沈灘。 本發明提供了一個生產NTD CZS夕晶圓的技術配方。右 教授：一”中子增強S曲線”可藉由留意一個經由在傅統氣 原子沈殿時因體積轉變而射出的晶格間砂原子（Si ϊ)(參照 方程式⑴），可Μ與一個中子造成的空位配合，而更助長 經由氧原子沈澱的氧原子減損（參照方程式⑵）而如繪成 。如本發明所述，氧原子濃度的總減損如下： [Ox] = Max([0i]o » [0r]s + min ¢2 ( CSI x] s + [S i xD 0) * [vacancy]}) 0 經濟部標準局員工消費合作4製 其中下標o與s分別表示初始狀態及S曲線預測值； [Sii]s等於0.5^[0i]s *而[vacancy]可从ί艮直接的由執 行例如MCNP (Monte Carlo N Partiel兮）電腦模擬程式（可 由在TN，〇ak Ridge之Oak Ridge美國國家實驗室取得）或美 國MS I公司之CASTEP程式ί辱至U。[vacancy]是一( CR)，矽樣本厚度，以及總中子照射量的函數。最嚴重的 氧原子減損發生在使用較低CR的中子源，使得[vacancy] 大於[Si i]s+〇時。在即此情況下，其最終的晶格間氧原子 濃度是：[〇i]f = msixt[0i]o - Ο [Οι]，0)，且圖示βδ 一 9 一 本紙張尺度適用中國國家標準（CNS ) Α4规格（210X297公釐） 408196 A7 ___B7_ 五、發明説明（8 ) η-S曲線（參考圖0。 , (請先聞讀背面之注意事項再填寫本頁) 由n-s曲線透露一個重要的訊息，就是在最普通的[Οι]ο =9〜14 ppma條件下，氧原子的減損將非常的少（參照圖4) 。針對此情形所做的實驗顯示，過多的氧原子沈澱的確並 未發生。再者，由建立在這搜生產早NTD CZ：晶圓上的 霄容及PN介面之特性結果也顯示，其所得之少數載子壽命 (參照圖1> *界面能態密度（interface state density)和 逆向偏壓下的漏電流（參照表二）與使用一般空白（即未作 中子質化）生產用晶圓的結果相同。因此本發明確實不僅 能夠應用NTD法達到均句的鱗原子摻雜*且能避兔眾所顧 慮的過度氧原子沈灘的瑪象。 餅究用的核子反應爐並不是中子的唯一來源。現在繪 參照圖2，其顯示了另一種瑕式的慢中子源的主要部份。 甶此可Μ了解，一瘦類似的轉變快中子使成慢中子源的作 法寧木脫離本案之發明精神。在_旋加速器（21)中*質子 沿著螺旋軌道（如22)蓮行，並在每次通過兩個0^電極間 的分隔間隙（25)進行加速。依此方式，高能量質子（如24) 抵達並通過導管（23)。磁力線（磁極並未標出）駕駿著_堅 質子*使其沿著導管（23)射出，並辛出口端（26)進入空氣408196 A7 B7 V. Description of the Invention (1) (Background of the Invention) ⑴Invention of the Invention The present invention relates to a semiconductor doping technology, especially a technology related to neutronization doping (referred to as NTD). 5 (Please read the note on the back before filling in this page) (2) The most widely used method for producing single crystals of sand in the semiconductor industry is the Czochralski * 0 B method In the 0 method, the ultra-pure polycrystalline sand is first placed in the phase vortex of the sand dioxide, and then a single crystal sand seed is put into the melt, and then slowly pulled out. A part of the liquid sand will cry out because of the surface tension, and then the sand will be pulled out *. Then, above the solution, it will gradually crystallize in a slightly lower temperature environment. Depending on its pulling speed and other factors, the high-quality monocrystalline sand with a diameter of about 12 inches and a length of about 40 inches can be pulled out as usual. Ministry of Economic Affairs, Standards Bureau, Consumer and Industrial Cooperation 2 # Printing Generally speaking, the resistance value of CZ single crystal is higher than the resistance value required for semiconductor manufacturing. Therefore, dopants * such as scale atoms can be added to the melt While reducing its resistance value. This kind of single crystal sand's resistance filling control method * has many & spoon problems * especially the uneven distribution of resistance value ° Generally, the resistance value varies from one end of the single crystal to the other, which is quite large in cross section. The same goes for the path 1¾. The above-mentioned problem of resistance value distribution has been improved in the MCZ method, in which the single crystal of sand grows in a magnetic field along the axial direction (typical values are about 500 and 4000 Gauss). Between) completed. Dissolved sand is an electrical conductor, so its thermal convection will tend to proceed along this line of magnetic force. However, although the kcz method provides considerable improvement, the problem of doping distribution in uneven sentences still exists. -3-This paper size is in accordance with China National Standard (CNS) 8-4 specifications (2 丨 0X297 mm) 1 5 Cooperative consumption system for employees of the Bureau of Standards of the Ministry of Economic Affairs 408196 A7 B7 V. Description of the invention (2) The atomic sequence of sand is 14, Its atomic weight is 28. However, the sand that exists in nature does not consist entirely of silicon-28 isotopes. Among them, there is the presence of sand-29, which accounts for about 4.7% of the atomic concentration, while silicon-30 accounts for 3.1% of the atomic concentration. In addition, when sand-30 is hit by thermal neutrons, it will turn into dish-31 (its library sequence is 15). Because the degree of erbium doping of Linlinzi is much lower than the 3.30% atomic degree of the existing sand in silicon crystals—30, obviously, ^ shoots a natural isotope from a limited amount of neutrons A single crystal of sand can be doped with scale atoms. The original characters of scales doped in this way will be evenly distributed, and will replace some silicon atoms in the crystal lattice as electron donors. This CZ and MCZ-doped sand crystal square casting was previously made. Technology has been disclosed, such as the technology proposed by Tskasu et al. (Ie, US Patent No. 4,910,156 of March 1990). Neutron sources available outside the nuclear weapons industry generally provide slow (thermal) neutrons (ideally with a net volume of about 0.025 electron volts). Its sources include a pool reactor that directly generates neutrons, a proton cyclotron, and M and proton linear acceleration casting. Proton generators are generally easier to use because the lean protons they generate are magnetically manipulated. The fast neutrons produced by these generators produce fast neutrons by passing through objects such as beryllium, autologous or uranium. These 'determination words' are then passed through the surrounding moderator, female mouth heavy water (D20), and decelerate to form thermal neutrons. Although the neutronization doping (NTD) technique * should be an ideal method for homogeneous atom doping *, the single-crystal firewood sand doped with this NTD method generally cannot achieve the desired results. Inter-lattice oxygen atoms often come out of the CZ and MCZ sand single crystals and become impurities. The source is actually sand dioxide. Xie Yi 4-This paper size applies Chinese National Standard (CNS) A4 specifications (2! 0 × 297mm) > (Please read the back: ¾ Matters before filling in this page) 408196 A7 B7 V. Description of the invention (3) 1 5 (Please read the notes on the back if and then fill in the j-cube itself. Since Its outstanding properties * inter-lattice oxygen is required for the manufacturing process of microelectronic components, for example, internal shifting of the mass produced in the ultra-large integrated circuit manufacturing process, and strengthening the hot-pressing resistance of the wafer. However, too many oxygen atom precipitation (forged into silicon dioxide) is often caused by neutron impact in the NTD method, which in turn causes the single crystal of sand to have a difficult to predict the resistance value of 0. In In some cases, even the precipitation of all oxygen atoms has occurred. In sand wafers, excessive oxygen sinks have led to an undesired reduction in the lifetime of minority carriers, and the inevitable appearance of microelectronic devices under reverse bias voltage Accepted large leakage current. In order to enable the NTD method to be successfully used in CZ silicon single crystals, it is necessary to have a successful formulation of the present invention. The parameters of its meal include * the lattice in the original CZ sand single crystals. Interstitial oxygen atom concentration, neutron spectrum (expressed as M cadmium ratio), and neutron exposure (total cumulative neutron exposure per square centimeter). If the above formula is not clear, the NTD produced will probably not be predicted The quality of CZ silicon single crystals. I believe that due to the lack of the above knowledge, Takasu et al. (US Patent No. 4,910,156 in March 1990) suggested the use of many non-traditional infrared detection technologies, and M was used as the NTD CZ wafer. Method of quality control T and quality assurance. That is, 'NTD CZ wafers with low red line h-line transmittance are not considered to be prime wafers. So * faced with quality control after production Uncertainty in the production process naturally cannot guarantee a high production yield. However, if the successful formulation of the present invention * is adopted, the production of high-quality NTD CZ production wafers can be guaranteed, so the non-traditional red Xi The method of line detection. In the scope of the patent application proposed by Takasu et al., The maximum allowed fast neutron exposure is established by -5-This paper standard applies Chinese National Standard (CNS) A4 specification ( 210X297 mm) 408196 A7 _B7_ V. Description of the invention (4) 1 5 '(Please read the precautions on the back before filling this page) The reverse bias leakage current of the electronic components on the NTD wafer is determined. However, Experiments based on the successful formulation of the present invention show that the relationship between the fast neutrons and leakage current (which does not indicate the actual value) obtained by Takasu et al. Is incorrect. In addition, the division of fast neutrons and slow neutrons described by Takasu et al. Is quite vague, but in the present invention; this is a very strictly defined parameter. The inventors further want to reveal that the NTD method is also used for doping single wafer *, not just for a single silicon ingot. For example, as described in the prior art by Groves et al. (Patented in U.S. Patent No. 5,212,100, acquired in March 1993) * However, the same problem of oxygen atom deposition still exists. The present invention, which will be described in detail below, can not only effectively implement the NTD method to achieve uniform disc doping, but also successfully avoid the problem of excessive oxygen atom sinking. (Summary of the present invention) One of the objectives of the present invention is to provide an improved method for doping neutrons of CZ sand single crystals. The second objective of the present invention is to provide an improved method to avoid the problem of excessive oxygen atom splashing that occurs during the traditional neutronization doping of single crystal silicon. Another object of the present invention is to provide an improved method for minimizing the electronic effect caused by the oxygen atom sink source in a single crystal sand. Yet another object of the present invention is to provide a method that combines neutron protonation doping and improvement of oxygen atom control simultaneously. A further object of the present invention is to provide a device for performing the aforementioned method. -6-The paper scale is applicable to China National Standards (CNS) A4 specification (210X297 mm) 1 5 A7 --40 π? __ V. Description of the invention (5) One of the feasible purposes of the present invention is to allow And MCZ sand uses more easily available low heat / fast neutron ratio (ie low cadmium ratio) radiation facilities. The above mentioned objectives' have all been achieved by the effective formulation provided by the present invention on the CZ sand single crystal using the HTD method successfully. The formulation of the present invention is called the neutron-enhanced S-curve or nS-curve technique, and shows how, given a generally large amount of neutron exposure (more than 101 neutrons per square centimeter), C2 single crystal sand with a specific initial lattice oxygen concentration, combined with a specific neutron source with a known cadmium ratio (CR) (thermal / fast neutron ratio). Although the division between f 宍 neutrons and thermal (slow) neutrons is extremely important in this case, the fast neutrons and thermal neutrons are narrowly defined separately from the continuous energy spectrum of a typical neutron source. unrealistic. Conversely * in the present invention, this division is achieved by strictly defining the cadmium ratio (CR). It is worth noting that the η-S curve (when the fast neutron component of a low CR neutron source causes a large number of lattice vacancies) according to the present invention shows that for a compound with an initial inter-lattice oxygen content of 9 to 14 ppma For CZ single crystal sand, the use of low CR neutron sources (the actual value is generally between 5 and 20) has produced satisfactory NTD CZ sand single crystal. Therefore, the problem of excessive oxygen atom precipitation can be easily avoided simply by using an easily available low CR (in other words, high fast neutron component) neutron source. Moreover, it was confirmed by experiments that the leakage current problem as described by Takasu et al. Did not occur. Usually * high CR (about 100 to 5,000) neutron sources are more difficult to obtain. Traditionally for NTD CZ sand, the reason why a high CR neutron source should be used is that the fast neutron component tends to damage the single crystal silicon lattice, which will lead to nuclea-tion centers that promote the sinking of excessive oxygen atoms. 〇 一 Ί A paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 mm) ^ 1 — ^ 1 l ^ i ^^^ 1 nn — ^ n ^^^ 1 t ^ t, (please first Read the notes on the back and fill out this. Ί Ministry of Economic Affairs _ Standards Bureau employee consumption cooperation $ system 408136 B7 V. Description of the invention (6) 1 (Schematic illustration) Figure 1 shows a non-NTD 〇2 order that has undergone a general thermal cycle A typical "S-curve" of the relationship between the oxygen atom depletion and the initial oxygen concentration (including halo). 5 (Please read the precautions on the back before filling this page). Figure 2 shows the neutron source used in the present invention. Some of the main parts. Figure 3 shows a device for irradiating sand single crystals. Figure 4 is obtained according to the present invention. • Using NTD CZ on the amount of oxygen atom depletion and initial oxygen drawing (content) of sand single crystals. Relationship curve (nS curve). Figure 5 shows * The silicon single crystal in Figure 3 is divided into many (Detailed description of the preferred embodiment) The main mechanism of the precipitation of oxygen atoms in the sand under the condition of no neutron irradiation (under the aforementioned heat treatment) is as follows: The formula shows: 2Si + 20i Si〇2 + Six ....... ♦ 〇> where the subscript I represents "inter-lattice". Figure 1 is a typical "S-curve" of the employee consumption cooperation chart of the Bureau of Standards of the Ministry of Economic Affairs ", Which shows the relationship between the oxygen atom depletion amount (M per million atomic parts (PPma)) and the initial oxygen concentration after several typical thermal cycles for non-NTD CZ production wafers; Figure 1 Curve 1 is after 8000 heat treatment for 8 hours * and then at 1,050 ° C for 1 hour; curve 2 is after 800 hours for 2 hours and then 1,050 ° C for 15 hours; Μ and curve 3 are after 4 Hour 750T: * After 16 hours 1,050t: From this we can see that * in the typical thermal cycle of the microelectronics manufacturing process, the oxygen atoms in the general C and B wafers will not all sink, especially When the initial oxygen atom concentration is quite low. -8-This paper applies the Chinese National Standard (CNS) Α4 Grid (210X297 mm) 408196 at Β7 V. Description of the invention (7) 1 When neutrons are irradiated, another reaction mechanism will cause considerable effects. That is vacancy + Sii + 201-> SI〇2 .. .... (2) One of them (vacancies) is caused by the 'Shaan neutron component'. 5 (Please read the precautions on the back before filling this page) If you do not follow the appropriate formula, apply the NTD method to CZ5 Evening single crystals will lead to reactions such as Equation (2), which will generate undesired excess oxygen sinking beaches. The invention provides a technical formula for producing NTD CZS wafers. Professor You: A "neutron-enhanced S-curve" can be noticed by an inter-lattice sand atom (Si ϊ) (refer to equation 射) emitted by the volume transition of the Futong gas atom Shen Dian (see equation ⑴). The vacancies caused by the ions cause the loss of oxygen atoms precipitated by the oxygen atoms (see equation ⑵) and are drawn as shown. According to the present invention, the total impairment of the oxygen atom concentration is as follows: [Ox] = Max ([0i] o »[0r] s + min ¢ 2 (CSI x] s + [S i xD 0) * [vacancy]} ) 0 The system of employee consumption cooperation 4 of the Bureau of Standards of the Ministry of Economics, where the subscripts o and s represent the initial state and the predicted value of the S curve, respectively; [Sii] s is equal to 0.5 ^ [0i] s *, and [vacancy] can be directly determined from Run, for example, MCNP (Monte Carlo N Partiel) computer simulation program (available from Oak Ridge US National Laboratory at Oak Ridge, TN) or the CASPE program from MS I of the United States to U. [vacancy] is a function of CR (CR), silicon sample thickness, and total neutron exposure. The most severe oxygen atom impairment occurs when a neutron source with a lower CR is used, so that [vacancy] is greater than [Si i] s + 0. In this case, the final inter-lattice oxygen atom concentration is: [〇i] f = msixt [0i] o-Ο [Οι], 0), and the graph βδ-9 is a paper size applicable to Chinese national standards (CNS) A4 specification (210X297 mm) 408196 A7 ___B7_ V. Description of the invention (8) η-S curve (refer to Figure 0. (Please read the notes on the back before filling this page) An important point revealed by the ns curve The message is that under the most common conditions of [Οι] ο = 9 ~ 14 ppma, the oxygen atom loss will be very small (see Figure 4). Experiments conducted in this case show that excessive oxygen atom precipitation does indeed It did not happen. Furthermore, the characteristics of the early capacity of the NTD CZ: wafer and the PN interface built on this search and production also show the minority carrier lifetime (see Figure 1 > * Interface Energy State Density (interface) State density) and reverse current leakage current (refer to Table 2) are the same as the result of using a normal blank (that is, not neutronized) to produce wafers. Therefore, the present invention does not only apply the NTD method to achieve uniform sentence Scaling atoms doped * and can avoid the excessive oxygen that rabbits worry about The mammoth of the sunken beach. The nuclear reactor used for cake research is not the only source of neutrons. Now referring to Figure 2, it shows the main part of another flawed slow neutron source. A thin and similar transformation of fast neutrons makes the slow neutron source Ningmu out of the inventive spirit of the case. In the cyclotron (21), * protons travel in a spiral orbit (such as 22), and every time Accelerate through the separation gap (25) between the two 0 ^ electrodes. In this way, high-energy protons (such as 24) arrive and pass through the conduit (23). The magnetic field lines (the magnetic poles are not marked) drive _ firm protons * Make it exit along the duct (23) and enter the air at the outlet end (26)
Cfj 〇 在出口端（26)射出的高能量質子，隨後通過一個如鈹 ，線或鈾的物質曆（27)，(一般厚度在0.1到1公分之間）， 後者轉變質子流成為快中子流（如10)。瑄些快中子然後遇 過環繞在出D端（26)及物質層（27)的減綴;#(如水、重水（ -10 -本紙張尺度逍用中國國家標準（CNS > A4規格（21〇X297公釐） 1 5 經濟部標準局員工消費合作#製 2 2 408196 A7 B7五、發明説明（9 ) D2O)、石墨等等）成為慢中子。減緩層的厚度是，達到理想 鎘比（CR)之越碼厚度。 一個特定的中子源之鎬比（意指熱/快中子數比）在本 發明中嚴格的定義為：假如在一給定的時段褢，Μ—典型 的蓋闲計數器（Geiger* counter*)測童一'被中子照射適一'段 時間的金箱，得到的累積伽碼（gamma)射線總數被記為A * 而且在相同的情況下，一中子照射相同大小但包覆在0.5 毫米厚的縮金屬内的金箱*其伽碼射線累積總數記為B， 貝(J觸比（CR)即定義為（A—1B)/B 〇在A中，金箱的中子活彳b( activation)是由於熱中子與快中子共同的作用，然而在B 中，熱中子太部份被錦金屬層所吸收，因而金箱的洁化性 大部份由於快中子的作用。 在圖3中，顯示了除圖2中的原子爐以夕卜的其他形式慢 中子源如何首自被使用於NTD辱射CZ砂單晶〇5tD31)上。而後 者，因其圍繞在中子減緩層或核心（28)的夕卜圍，因此確定 其會接受至tl·隱中子（如29)的照射。且為了要確保中子照射 是均句的且相等地供給於全部的矽單晶（如31)，有作法（ 圖未示）能讓砂單晶（31)繞著中子源（28)旋轉（軌迨32)， 並使得全部的砂單晶从自己的軸心、旋轉〇«!軌道33) , Μ及 相對於中子源（28)將所有砂單晶作上下移動（方向34)。 本發明不論是藉由圖2與圖3的方式，或是由一核子反 應爐的放射方式實施，可Μ應用於CZ或MCZ型的單晶砂。 依本發明生產出的砂單晶，將允許有效的控制氧原子濃度 及麟原子摻雜。 -11 - ---.-------0^-- (請先閱讀背面之注$項再填寫本頁) 訂 本紙張尺度適用中國國家標準（CNS ) Μ規格（210Χ29?公釐） ί 5 10 15 經 濟 部21 標 準 局 貝 工 消 費 合 作2% % 五、發明説明（10 408196 A7 B7 上述兩種不同的實胞例，總中子照射童控制在每平方 公分5x1017垄中子之間，而對一中子通量約每平方 公分每秒ΙΟ14個中子的中子源而言，該矽單晶受照射的時 間，約在5小時到10天之間。此方法對於直徑約在2到30英吋 且長度在1到150公分之間的砂單晶是有效的。對於直徑大 於S英吋的矽單晶，最好能夠將該砂單晶切成每段長約20 公分的小段，並且將各小段橫切面面對著中子流（參照_5 )以避免中子通董衰減。 圖4係顯示：根據本發明之傳授*經過中子質化接雜（ NTD)的單晶矽之氧原子減損與初‘始氧原子濃度的關係曲線 。ifcb—曲線相對圖1之曲線1的曲濂41 ,是先經過1小時 8〇〇1*再經過8小時1,〇50°〇的熱處理，而相對於圖1之曲 線2的曲線42，是先經2小時800°C，再經16小時1,050¾熱 處理*而相對於圖1中之曲線S的曲線43，是先經4小時750 ’再經16小時1,050*0的熱處理。由Μ上可Μ 了解，假 如本發明之配方未被使用而盲目的對CZ或MCZ矽單晶作中 子質化摻雜*將很可能落於風4的n-s曲線中晶格間氧大量 減撗的部份。此比之圖1(無NTD，S曲線）預測氧沈澱情形 要巌重得多〇 當矽簞晶的直徑超過8英吋，假使使用如圖3所示的中 子照射架構，則很可能會發生一些中子衰減。因此，對於 這些較大的矽單晶，應該切成約20公分厚的小塊，並且令 其橫切面面對著中子源，而環繞於中子源的四周，如圖5 所示。為了確保這瘦中子能均句地照射且相等地供給所有 —12 — -(請先閲讀背面之注項再填寫本頁) _；__ 本紙張尺度適用中國國家標準（CNS ) A4規格（210X297公釐） 5 408196 A7 B7 五、發明説明（n) 的小塊（如51)，則必須使用一種裝置使得小塊（51)旋繞於 中子源（28)(依軌道32)，並讓所有的矽單晶小塊自Μ他們 的軸心旋轉（軌道如53)，並且可相辦於中子源（28)將所有 的單晶（51)作上下移動（方向依34)。 在此另外提供表一及表二的數據供參考Μ證明本發明 之有效性。表一顯示少數載子的壽命（其相關於在砂晶圓 中的電子與電洞復合區（recombination 的數目）及 對於數種不同砂晶圓樣本的載子表面產生速率（其相對於 在石夕晶圓表面的載子產生逢率（surface gener*ia1;ic>ri velo city))，而其中所有的砂晶圓樣求，除了該空白晶圓夕卜* 皆經使用本發明的酡方，MNTD法撥雜了磷原子，並且隨 後接受了前述所示的熱處理。可Μ由表一知，其中NTD CZ 晶圓和空自晶圓在少數載子壽命及表面載子產生摔率上並 沒有真正的差異。 表二乃比較相似於表一中的諸晶圓樣本上之ΡΝ介面在 5伏特逆向偏壓下之漏電流值。其中<Jra是指面向漏電流（ area eurM^ent;)，夕卜HS-漏電流(per*iphepa 1 ciirr ent)。在實驗的精確度内*本案之NTD CZ晶圓之Jrp和Jra 值*基本上與一空白晶闯所測得的值相同。 (請先閲讀背面之注意事項再填寫本頁) -9Cfj 〇 High-energy protons emitted at the outlet (26), and then passed through a material calendar (27), such as beryllium, wire, or uranium (typically between 0.1 and 1 cm thick), which converts the proton stream into fast neutrons Stream (such as 10). Some fast neutrons then encountered the minus affixes surrounding the D end (26) and the material layer (27); # (such as water, heavy water (-10-this paper size uses Chinese national standards (CNS > A4 specifications ( 21〇X297 mm) 1 5 Consumption Cooperation of Employees, Bureau of Standards, Ministry of Economic Affairs # 2 2 408196 A7 B7 V. Description of Invention (9) D2O), graphite, etc.) become slow neutrons. The thickness of the retardation layer is to reach the ideal cadmium The over-thickness thickness of the ratio (CR). The pick ratio of a particular neutron source (meaning thermal / fast neutron number ratio) is strictly defined in the present invention: if a given period of time, M—typical The Geiger * counter * measures a child's golden box that has been irradiated with neutrons for a certain period of time, and the total cumulative gamma rays obtained are recorded as A *, and in the same case, one The neutron is irradiated with a gold box of the same size but wrapped in a 0.5 mm thick shrink metal. The cumulative total number of gamma rays is recorded as B, and the J contact ratio (CR) is defined as (A-1B) / B. In A, the activation of the neutron b (g) in the golden box is due to the joint action of thermal neutrons and fast neutrons. However, in B, the thermal neutrons are partly covered by gold. It is absorbed by the layer, so the cleanliness of the gold box is largely due to the role of fast neutrons. In Figure 3, it shows how slow neutron sources other than the atomic furnace shown in Figure 2 are used first. It is on NTD insult CZ sand single crystal 05tD31). The latter, because it surrounds the neutron moderation layer or core (28), it is determined that it will accept tl · Hidden neutrons (such as 29) In order to ensure that the neutron irradiation is uniform and equally supplied to all silicon single crystals (such as 31), there are methods (not shown) that allow the sand single crystal (31) to surround the neutron source (28 ) Rotate (orbit 迨 32), and make all the sand single crystals rotate from their own axis, «! Orbit 33), and move all the sand single crystals up and down (direction 34) relative to the neutron source (28). ). The present invention can be applied to CZ or MCZ-type single crystal sand regardless of whether it is implemented by the method of Figs. 2 and 3 or the radiation method of a nuclear reactor. The single crystal of sand produced according to the present invention will allow effective control of the oxygen atom concentration and the lin atom doping. -11----.------- 0 ^-(Please read the note on the back before filling in this page) The size of the paper is applicable to the Chinese National Standard (CNS) M specifications (210 × 29? Mm) ） 5 10 15 Ministry of Economic Affairs 21 Standard Bureau of Shellfish Consumer Cooperation 2%% 5. Description of the invention (10 408196 A7 B7 The above two different cases of real cells, the total neutron exposure is controlled at 5x1017 ridges per square centimeter. For a neutron source with a neutron flux of about 1014 neutrons per square centimeter per second, the exposure time of the silicon single crystal is between about 5 hours and 10 days. This method has a diameter of about Sand single crystals with a length of 2 to 30 inches and a length of 1 to 150 cm are effective. For silicon single crystals with a diameter larger than S inches, it is best to cut the sand single crystals into segments with a length of about 20 cm. And to face the neutron flow (see _5) to avoid the attenuation of the neutrons. Figure 4 shows the teachings according to the present invention * after neutron quality doping (NTD). The relationship between the oxygen atom depletion of single crystal silicon and the initial oxygen atom concentration. Ifcb—The curve is compared with the curve 41 of curve 1 in FIG. At 800001 *, 8 hours of heat treatment at 1,050 ° C, and the curve 42 with respect to curve 2 of Fig. 1 is first subjected to 800 ° C for 2 hours, and then 1,050¾ heat treatment for 16 hours. The curve 43 of the curve S in FIG. 1 is subjected to a heat treatment of 750 'for 4 hours and then 1,050 * 0 for 16 hours. From the above, we can understand that if the formula of the present invention is not used, the CZ or MCZ will be blindly Silicon single crystals doped with neutrons dopant * will most likely fall in the part of the interstitial oxygen reduction in the ns curve of wind 4. This is more important than the predicted oxygen precipitation situation in Figure 1 (without NTD, S curve). It is much heavier. When the diameter of the silicon crystal is more than 8 inches, if the neutron irradiation structure shown in Figure 3 is used, some neutron attenuation is likely to occur. Therefore, for these larger silicon single crystals, It should be cut into small pieces about 20 cm thick, with its cross-section facing the neutron source and surrounding the neutron source, as shown in Figure 5. In order to ensure that the thin neutrons can be illuminated uniformly and Equally supply all —12 —-(Please read the note on the back before filling out this page) _; __ This paper size applies to Chinese national standards (CNS ) A4 size (210X297 mm) 5 408196 A7 B7 5. For the small piece (such as 51) of the invention description (n), a device must be used to make the small piece (51) spiral around the neutron source (28) (according to orbit 32) ), And let all the silicon single crystal small pieces rotate from their axes (orbits such as 53), and can be handled by the neutron source (28) to move all the single crystals (51) up and down (the direction depends on 34 The data in Tables 1 and 2 are provided here for reference to prove the effectiveness of the present invention. Table 1 shows the lifetime of minority carriers (which is related to the number of electron and hole recombination regions in the sand wafer) and the carrier surface generation rate (which is relative to The surface gener * ia1; ic > ri velo city) on the surface of the wafer, and all of the sand wafer samples are obtained, except for the blank wafer. The MNTD method is doped with phosphorus atoms, and then subjected to the heat treatment shown above. As can be seen from Table 1, NTD CZ wafers and air-free wafers do not have the minority carrier lifetime and surface carrier generation rate. The real difference. Table 2 is similar to the leakage current value of the PN interface on the wafer samples in Table 1 under a 5 volt reverse bias. Among them, < Jra refers to the leakage current (are eurM ^ ent;) , Xi Bu HS-leak current (per * iphepa 1 ciirr ent). Within the accuracy of the experiment * the Jrp and Jra values of the NTD CZ wafer in this case * are basically the same as the values measured by a blank crystal break. (Please read the notes on the back before filling out this page) -9
T 經濟部中20*標準局員工消費合|印裝Consumption of 20 * Standard Bureau employees in the Ministry of Economic Affairs | Printing
Process NTD 80〇υ/2ίι HID 900t；/0.5h HID ιοοου/ο.^ NTD 1000t：/0.5h NTO lOOO'C/O.Bh blank Jra (ηΑ/αιί2) 2.0 2.2 6.4 9.3 10.3 25.0 Jrp (πΑ/cid) 12.0 10.6 9.0 0.8 4.7 1.9 表——：金氧半導體電容⑽S Capacitor)量彻I —13 — 本紙張尺度適用中國國家標準（CNS ) A4規格（2!〇X297公釐） 408196 五、發明説明（12 ) A7 B7 1 5Process NTD 80〇υ / 2ί HID 900t; /0.5h HID ιοοου / ο. ^ NTD 1000t: /0.5h NTO lOOOO'C / O.Bh blank Jra (ηΑ / αιί2) 2.0 2.2 6.4 9.3 10.3 25.0 Jrp (πΑ / cid) 12.0 10.6 9.0 0.8 4.7 1.9 Table——: Metal Oxide Semiconductor Capacitors CapS Capacitor) Quantities I — 13 — This paper size is applicable to China National Standard (CNS) A4 specifications (2.0 × 297 mm) 408196 V. Description of the invention (12) A7 B7 1 5
Process NTD 80〇lC/2h _ 850t71h NTD 90010/0.5h NTD 1000¾/0.5h blank Minority Carrier Lifetime (ms) 0.83 2.14 3.48 1.84 1.97 Surface Gen* Velocity (oi/s) 14.1 17.7 13.9 17.2 25.1 表二：介面漏電流（Junction Leakage— Current)量測 .(請先閲讀背面之注意事項再填寫本頁) 由上述實施例及圖示之說明可Μ了解，本發明哲Μ根 據前述之說明而遂行同一領域內的技術轉換或_化實施’ 15 故任何形式上及細部的技術改變，都不脫離本發明的技術 .! 範疇。 經 2| 橾 準 工 消 費 合 作2$ 製 -14 - 本紙張尺度逋用中國國家標準（CNS ) Α4規格（21〇Χ2.97公釐）Process NTD 80〇lC / 2h _ 850t71h NTD 90010 / 0.5h NTD 1000¾ / 0.5h blank Minority Carrier Lifetime (ms) 0.83 2.14 3.48 1.84 1.97 Surface Gen * Velocity (oi / s) 14.1 17.7 13.9 17.2 25.1 Table 2: Interface leakage Current (Junction Leakage — Current) measurement. (Please read the precautions on the back before filling out this page.) From the above examples and illustrations, you can understand that the philosophy of the present invention is based on the foregoing description. Technology transformation or implementation '15 Therefore, any form and detail of technical changes will not depart from the scope of the technology of the present invention !. After 2 | 橾 quasi-industrial consumption cooperation 2 $ system -14-This paper size adopts China National Standard (CNS) A4 specification (21〇 × 2.97 mm)