TWI273600B - Integrated circuit and manufacturing method thereof, memory cell and manufacturing method thereof, method for programming memory cell and method for programming memory array multiple times - Google Patents
Integrated circuit and manufacturing method thereof, memory cell and manufacturing method thereof, method for programming memory cell and method for programming memory array multiple times Download PDFInfo
- Publication number
- TWI273600B TWI273600B TW93121700A TW93121700A TWI273600B TW I273600 B TWI273600 B TW I273600B TW 93121700 A TW93121700 A TW 93121700A TW 93121700 A TW93121700 A TW 93121700A TW I273600 B TWI273600 B TW I273600B
- Authority
- TW
- Taiwan
- Prior art keywords
- memory cell
- electrode
- memory
- material layer
- array
- Prior art date
Links
Landscapes
- Semiconductor Memories (AREA)
- Non-Volatile Memory (AREA)
Abstract
Description
1273 6iM)twf.d〇c/〇〇6 九、發明說明: 【發明所屬之技術領域】 本發明是有關於-種可電程式非揮制生記憶體與含有 此記憶體之積體電路,且特別是有關於一種記憶胞結構、 根據由超薄介電層及相關結構的累進崩潰(progressive breakdown)引發可程式電阻的操作方法。 【先前技術】 4 可電程式非揮發性記憶體技術已應用在許多用途上。 這些種種的技術會隨著記憶胞可被程式化的次數、達到程 式化所需要的電壓與儲存在每一記憶胞的資料之位元數而 改變。而且,一個重要的考量是所提供的特定記憶體技術 是否在製造步驟中能夠符合記憶胞與辅助電路的需求。 、《己隐體技術建立在浮置閘極上(如:標準的可電除且可 程式唯讀記憶體(EEPROM))或在電荷陷入層上(如:氧化矽_ 氮化矽-氧化矽)之記憶胞通常可以程式化數次。然而,這 些技術需要複雜的程式化與抹除電路,並使用複雜的充電 泵(charge pump)技術以達到程式化與抹除所要求的電壓。 而且,當於母一記憶胞中儲存多於一位元之資料時,需要 複雜的程式化與感測技術。於是,對於這些類型的快閃記 憶體而言,製作快閃記憶體所需要的製帛步驟通常包括對 於在同一積體電路上形成標準邏輯電路(如互補式金氧半 導體(CMOS)電路)不是—般要求的步驟,而這些步驟 成成本提高。 de· Graaf 等人於 a Novel High-Density Low-Cost 1273 60^twf.doc/0061273 6iM) twf.d〇c/〇〇6 IX. Description of the Invention: [Technical Field of the Invention] The present invention relates to an electrically programmable non-volatile memory and an integrated circuit including the same. In particular, it relates to a memory cell structure, an operation method for inducing a programmable resistance according to a progressive breakdown of an ultra-thin dielectric layer and related structures. [Prior Art] 4 The electrically programmable non-volatile memory technology has been used in many applications. These various techniques vary with the number of times the memory cells can be programmed, the voltage required to achieve the programming, and the number of bits of data stored in each memory cell. Moreover, an important consideration is whether the particular memory technology provided can meet the needs of the memory cell and the auxiliary circuit during the manufacturing steps. "The hidden technology is built on a floating gate (such as a standard electrically erasable and programmable read-only memory (EEPROM)) or on a charge trapping layer (eg, yttrium oxide yttrium nitride - yttrium oxide) The memory cells can usually be programmed several times. However, these techniques require complex stylization and erasing circuitry and use complex charge pump techniques to achieve the voltages required for stylization and erasing. Moreover, when storing more than one bit of data in a mother cell, complex stylization and sensing techniques are required. Thus, for these types of flash memory, the steps required to fabricate the flash memory typically include the formation of standard logic circuits (such as complementary metal oxide semiconductor (CMOS) circuits) on the same integrated circuit. The generally required steps, and these steps become cost increases. De·Graaf et al. at a Novel High-Density Low-Cost 1273 60^twf.doc/006
Diode Programmble Read-only Memory” IEDM 1996, page 7·6·1〜7·6·4中揭露一種簡單的可電程式記憶胞。根據如. Graaf等人,一種一次可程式高密度記憶體可藉由使用二 極體-抗溶絲(Diode-antifuse)結構來達成,此種二極體·抗 熔絲(Diode-antifuse)是由第一 n型多晶矽電極、第二p型 擴散電極以及在兩電極之間的一層介電層,此介電層厚度 為60埃左右’且材質為利用熱成長法形成之二氧化石夕。 在此結構中’記憶胞被程式化是藉由施加13伏特左右之 高壓以引發介電層之崩潰,由此在經程式化記憶胞的電極 之間形成物理性連接。雖然,de· Graaf等人所揭露之結 構較為緊密且容易製造,但是此種記憶胞只能程式化一次 且需要高電壓操作。 因此,目前需要提供一種可電程式非揮發·記憶體技 術’可以在低電壓操作,且其製程更可以與標準CM〇s 邏輯電路製造技術相容。同時,也需要提供一種非揮發性 兄憶胞技術,可對記憶胞進行多次程式化操作及/或在單 一記憶胞中儲存多於一位元之資料。 【發明内容】 本發明的目的就是在提供一種可電程式非揮發性記憶 胞及其製造方法,此記憶胞包括第一電極、第二電極以及 在兩^極間之一材料層,其特徵在於材料層具有基本上累 進改變的可測量特性,例如材料層的電阻、電抗、磁性、 極性、元素排列等,藉由應力的累進量表示儲存的資料來 建立可程式特性。此材料層包括一層超薄的材料,在此所 Ι2736ββ twf.doc/006 條件中a本義為材料層之厚度薄到使在正常的操作 改鐵二ί穩疋的較厚材料層的性f會隨著控制的量而 入雷2本發明之—實财此材料層包括具有蚊厚度的 J ;,此§己憶胞之特徵為利用低電壓跨過介電声一段 時間’而由應力累進量造成電阻累進改變,以建立^式 電阻,來表示儲存的資料。在—個選定記憶胞巾的可重複 程式化並不需要「抹除」,而可以提供—個多次可程式記 憶體。當然’藉由錢更多準的電㈣在單—域記憶胞 中建立夕程式化狀態,此多程式化狀態是符合多位元資料 或對應多程式化循環。因此,此種記憶胞可稱為可程式電 阻不需抹除記憶胞(programmable resistance eraseless memory) ο 在本發明之實施例中,此種記憶胞可在單一記憶胞中 儲存多位元資料;可進行多於一次的程式化操作而不需抹 除操作;可同時在單一記憶胞中儲存多位元資料且可進行 多於一次的程式化操作而不需抹除操作。而且,此記憶胞 也可儲存類比資料。 ^ 在文獻中已揭露超薄氧化層的累進崩潰是與用於電晶 體閘極的介電層之可量測性(scalability)的極限有關。如Diode Programmble Read-only Memory IEDM 1996, page 7·6·1~7·6·4 discloses a simple readable memory cell. According to Graaf et al., a one-time programmable high-density memory can be borrowed. This is achieved by using a diode-anti-dissolved wire (Diode-antifuse) which is composed of a first n-type polysilicon electrode, a second p-type diffusion electrode, and A dielectric layer between the electrodes, the dielectric layer having a thickness of about 60 angstroms, and the material being a dioxide dioxide formed by a thermal growth method. In this structure, the memory cell is programmed by applying about 13 volts. The high voltage initiates the collapse of the dielectric layer, thereby forming a physical connection between the electrodes of the stylized memory cell. Although the structure disclosed by de Graaf et al. is relatively compact and easy to manufacture, such memory cells are only It can be programmed once and requires high voltage operation. Therefore, there is a need to provide an electrically programmable non-volatile memory technology that can operate at low voltages and is compatible with standard CM〇s logic circuit fabrication techniques. ,and also To provide a non-volatile sibling cell technology, a plurality of stylized operations can be performed on a memory cell and/or more than one bit of data can be stored in a single memory cell. SUMMARY OF THE INVENTION The object of the present invention is to provide a An electrically programmable non-volatile memory cell and a method of fabricating the same, the memory cell comprising a first electrode, a second electrode, and a material layer between the two electrodes, characterized in that the material layer has a measurable characteristic of substantially progressive change, For example, the resistance, reactance, magnetism, polarity, element arrangement, etc. of the material layer, by means of the progressive amount of stress, represent the stored data to create a programmable property. This material layer comprises an ultra-thin material, here 2736ββ twf.doc/ 006 The condition a is that the thickness of the material layer is so thin that the thickness f of the thicker material layer that is stable in normal operation will be added to the mine with the amount of control. Including J with a mosquito thickness; this § recalled cell is characterized by a low voltage across the dielectric sound for a period of time 'and the resistance progressive change caused by the stress progression to establish a resistance, to represent Data stored in - a selected memory cell can be repeated stylized towel does not need to "erase", and can be provided - a multiple programmable memorized body. Of course, by using more money (4) to establish a stylized state in a single-domain memory cell, the multi-programmed state is in accordance with multi-bit data or a corresponding multi-programming cycle. Therefore, the memory cell can be referred to as a programmable resistance erase memory. In the embodiment of the present invention, the memory cell can store multi-bit data in a single memory cell; More than one stylization operation is performed without erasing the operation; multi-bit data can be stored in a single memory cell at the same time and more than one stylized operation can be performed without erasing the operation. Moreover, this memory cell can also store analog data. ^ It has been revealed in the literature that the progressive collapse of ultra-thin oxide layers is related to the limits of the scalability of the dielectric layer used for the gate of the electro-crystal. Such as
Hosoi 等人 “A New Model of Time Evolution of GateHosoi et al. "A New Model of Time Evolution of Gate
Leakage Current after Soft Breakdown in Ultra-Thin JakeLeakage Current after Soft Breakdown in Ultra-Thin Jake
Oxides,” IEDM,2002; Wang 等人 “Negative Substrate BiasOxides,” IEDM, 2002; Wang et al. “Negative Substrate Bias
Enhanced Breakdown Hardness in Ultra-Thin Oxide pMOSFETs,’’ 41st Annual International Reliability Physics 1273Enhanced Breakdown Hardness in Ultra-Thin Oxide pMOSFETs,’’ 41st Annual International Reliability Physics 1273
Symposium,Dallas,Texas 2003;以及 Under 等人,“Grower and Scaling of Oxide Conduction after Breakdown,,,41stSymposium, Dallas, Texas 2003; and Under et al., "Grower and Scaling of Oxide Conduction after Breakdown,,, 41st
Annual International Reliability Physics Symposium, Dallas Texas 2003。 累進崩潰現象在Hosoi等人之文獻中表示為r軟崩潰 (Soft breakdown)」,在Wang等人之文獻中表示為崩潰發 展「以累進的方式」。Linder等人之文獻中揭露超薄氧化 層的累進崩潰特性的特徵在於「退化速率(degradati〇n rate)」,其是由施加電壓、氧化層厚度、基底摻雜區、通 道長度決定的。在本發财,累進崩潰現象是制於在單 -記憶胞結構建立可减電阻值。結果,織胞結構可較 為緊密’可利用標準CM0S製程簡單的製选,且 低電壓操作。 ^ 於疋’本發明的實施例提供一種記憶胞,包括第一電 極二第:電極以及在兩電極間之一材料層。 責是由應力引發,例如使電壓跨過該材: 層:本發明之實_對舰加輕跨過 特 #庵^丄, 伏特。在一些實施例中,程 i if疋“加一正電壓至第-電極並施加-負電壓至 ^電極所引發’其令正電顯負電壓的絕對值小於2伏 步成m一種記憶胞的製造方法,包括於-基底上 electrode)材料層,此雷油^/上形成一電極間(inter_ .層此1簡材制之舰在於具有對應 Ι2736α〇 twf.doc/006 應力而累進改變的特性。然後,於電極間材料層上形成一 第二2°第—電極是利用提供—基底,然後植入n型或 =型,質而在基底中形成-導電區而形成的。可選擇的二 第-電極也可以在基底上_成長或沈積—或多層的導體 層而形成的。的,電極間材料層可以在第-電極上利 用成長或沈積材料層而形成的。在—實施例中,電極 料層之材質為在由摻雜絲底所形成的電極上利用熱成長 製程形成之二氧切或氮氧切。第二電極在本發明的 實把例中疋在電極間材料層上利用成長或沈積一或多声 的導體層而形成的。 曰 驟 •本發明提供—種記的製造方法,包括下列步 於-基底上形成多個第—導線,這些第—轉在第一 方向平行延伸; 於這些第一導線上形成多個第 ,-方向垂直之一第二方向平行延伸,而;以 父的陣列; 於第導線與第二導線之間_交區域形成電極間材 枓層,此電_材料層之特徵在於對應應力具有累進改變 的特性,而於相交區域形成記憶胞;以及 在基底上形成電路,以供給應力和感測記憶胞之特 性。 在本發明之實施例中,利關如淺溝渠隔離製程或 cos隔離製程在多個第—導線之間形成填滿有介電材 1273604 :twf.doc/〇〇6 $之溝渠。如此即在記憶朗列之間形成隔離結構。多個 f渠係在形成多個第—導線之前形成的。然後,多個第- f線例如藉由摻雜半導體基絲形成在溝渠之_區域。 ^實施财,乡個難係在沈積絲細於第一導線 層之後形成的。在此情況下,於形成多個溝渠的步 驟中將上述材料層分成多個導線。 在本發明之一實施例中,記憶胞是以下述步驟形成的: 撼:—⑦基紐人摻f以形成具有第―導電型態之導電 擴敢區; 於導電擴散區上形成氧化石夕層,此氧化石夕層之厚度小 於15埃;以及 在氧化石夕層上形成具有第二導電型態的楼雜多晶石夕。 合隨第二導線與電極間材料層的製程 a者所選擇的材料與所使用之材料的厚度而有所 ^發明所使用之超薄層包括氧切、氮氧化梦 及未摻雜二氧化梦’其在第—導線和第二 ^ 埃的厚度;在另一實施例中,其在第一導姊g 、、’之間具有小於15埃的厚度。_氧切或其 杜^度在下崎具有特性累較變的特徵而驗記憶體-太=據程式化及感測特性累進改變量的能力。其 =明之超薄層的材質包括氮切、如氧切·氮乍, =⑽Q堆4層之多層堆疊結構、氧化料 材二 如 Al2〇3、YTa2〇5、Hf〇2、ΥΑ、Ce〇2、Ti〇 電,Annual International Reliability Physics Symposium, Dallas Texas 2003. The phenomenon of progressive collapse is expressed in the literature of Hosoi et al. as "Soft breakdown", which is expressed in the literature of Wang et al. as a "degressive way". The literature of Linder et al. discloses that the progressive collapse characteristics of ultra-thin oxide layers are characterized by "degradatiation rate" which is determined by the applied voltage, the thickness of the oxide layer, the doped region of the substrate, and the length of the channel. In this fortune, the progressive collapse phenomenon is based on the establishment of a deductible resistance value in the single-memory cell structure. As a result, the cell structure can be made compacter and can be easily fabricated using standard CMOS processes with low voltage operation. ^ 疋 疋 'An embodiment of the present invention provides a memory cell comprising a first electrode two: an electrode and a material layer between the two electrodes. Responsibility is caused by stress, for example, the voltage is crossed across the material: Layer: The reality of the invention _ the ship is lightly crossed over # 特^丄, volts. In some embodiments, the process i if "adds a positive voltage to the first electrode and applies a -negative voltage to the ^ electrode" which causes the absolute value of the positive negative voltage to be less than 2 volts to form a memory cell. The manufacturing method comprises a layer of material on the substrate, and the oil is formed on the electrode (the inter_electrode). The ship of the simple material has the characteristic of progressively changing corresponding to the stress of Ι2736α〇twf.doc/006. Then, forming a second 2° first electrode on the inter-electrode material layer is formed by providing a substrate, then implanting an n-type or a-type, and forming a conductive region in the substrate. The first electrode may also be formed on the substrate by a growth or deposition layer or a plurality of layers of conductor layers. The interelectrode material layer may be formed on the first electrode by using a layer of grown or deposited material. In the embodiment, The material of the electrode layer is a dioxotomy or oxynitride formed by a thermal growth process on an electrode formed by doping the wire. The second electrode is used in the inter-electrode material layer in the practical example of the present invention. Growing or depositing one or more layers of conductor The present invention provides a method for manufacturing a seed, comprising the steps of: forming a plurality of first wires on a substrate, the first turns extending in parallel in a first direction; forming a plurality of first wires on the first wires And - the direction perpendicular to one of the second directions extending in parallel; and; the parent array; forming an inter-electrode layer between the first wire and the second wire - the electrical layer is characterized by progressive stress Changing characteristics, forming memory cells in the intersecting regions; and forming circuits on the substrate to supply stress and sensing characteristics of the memory cells. In an embodiment of the invention, such as shallow trench isolation processes or cos isolation processes are A plurality of first-wires are formed with trenches filled with dielectric material 1273604: twf.doc/〇〇6 $. Thus, an isolation structure is formed between the memory columns. Formed before the wire. Then, a plurality of the -f lines are formed in the region of the trench, for example, by doping the semiconductor base wire. ^Implementation is made after the deposited wire is finer than the first wire layer. In this case In the step of forming a plurality of trenches, the material layer is divided into a plurality of wires. In one embodiment of the invention, the memory cells are formed by the following steps: 撼: -7 kinsin doped f to form the first conductive a conductive diffusion region of the type; forming a oxidized stone layer on the conductive diffusion region, the thickness of the oxidized stone layer being less than 15 angstroms; and forming a polycrystalline polycrystalline stone having a second conductivity type on the oxidized stone layer In the case of the material selected by the process of the second wire and the inter-electrode material layer and the thickness of the material used, the ultra-thin layer used in the invention includes oxygen cutting, nitrogen oxidizing dream and undoped Oxidation dreams 'the thickness of the first and second conductors; in another embodiment, it has a thickness of less than 15 angstroms between the first guides g , . _ Oxygen cleavage or its degree of characterization in the lower saki has a characteristic change and the memory - too = the ability to progressively change the amount according to stylization and sensing characteristics. The material of the ultra-thin layer of the bright layer includes nitrogen cut, such as oxygen cut nitrogen nitride, multi-layer stack structure of 4 layers of =(10)Q stack, oxidized material such as Al2〇3, YTa2〇5, Hf〇2, ΥΑ, Ce〇 2, Ti〇 electricity,
HfS_、HfA10x、TaOxNy、Zr〇2、ZrSi ο 2 1χ y、 2 必ix〇y、La2〇3 等也 丨 twf.doc/006 可以作為記憶胞之上述超薄層的材質。 第一電極與第二電極之組成會因本發明應用之環境而 改變。在本發明之一實施例中,第一電極包括一多晶矽層, 第二電極包括在半導體基底之導電擴散區。在另一實施例 中,多晶矽層與導電擴散區具有相反的導電型態,而形成 一類二極體(Di〇de-like)可程式電阻記憶胞。在另'實施例 中,第一電極與第二電極可包括導體組合,含有金屬如銅、 鋁、鎢、鈦、合金及其組合、p型與n型多晶矽、p型與 η型擴散區、金屬矽化物、半金屬、及其類似物。在一些 實施例中,電極之材質包括含一元素的材料,在電極間的 ,極間材料層的材質包括同一元素之化合物。舉例來說, 第與第二電極之材質包括石夕,如非晶石夕、單晶石夕、多晶 矽、金屬矽化物及其類似物,則在第一電極與策二電極之 間的電極間材料層包括一含矽的化合物,如氧化矽或氮化 梦等。 本發明也提供一種積體電路,包括如上述的記憶胞陣 列’利用麵魏藉㈣起介電層之帛補潰以程式化記 憶胞,並利用感測電路感測陣列中記憶胞的累進崩潰量。 在本,明實施例巾m責量是由記憶胞巾電阻的改變 來表示。藉由程式化與感測特性的多階變化而在單一記憶 胞建立夕位元。舉例來說,在一實施例中操作程式化邏輯 =路而對-選&記憶胞施加_偏壓,然後確認是否該偏壓 ,成預期的累進改變量。如果確認此操作失敗,然、後重試 知加電壓與確認操作,並重複此餘直到記憶胞成功的程 11 I2736Qft twf.doc/006 式化或到達重試的極限。 本發明實施例之感測電路包括一參考電流源與一電 路,此電路用以比較來自記憶胞的電流與來自參考電流源 的電流。在用於感測單一記憶胞多位元或多階特性累進改 變的實施例中,感測電路可包括多個參考電流源,且電路 用以比較來自記憶胞的電流與來自多個參考電流源之一或 多個參考電流。 ^ 本發明更提供一種記憶胞的程式化方法,包括在進行 程式化時,供應一應力至電極間材料層以引發該層的特性 累進改變。因為特性的累進改變,所以可以達到多階程式 化此夕隖程式化可應用於多次程式化一個記憶胞而不需 要抹除操作,而可於單一記憶胞中程式化多個位元,且可 以結合多位元與多次程式化。 . 根據本發明之實施例,在程式化時供應一應力至記憶 胞的流程包括使用連續的電流脈衝與確認步驟,其敘述如 下: ^ 供應一第一程式化脈衝至具有第一脈衝高度與第一脈 衝寬度之記憶胞; 測量是否該記憶胞對應該第一程式化脈衝而被程式 化;以及,如果不是 供應一程式化重試脈衝至記憶胞; 測量是否該記憶胞對應該程式化重試脈衝而被程式 化;以及,如果不是 重複供應另一程式化重試脈衝至該記憶胞與測量該記 12 >twf.doc/006 憶胞是否被程式化,制測量出該峨胞已齡式化或重 試的次數到達最大值; 其中,程式化重試脈衝各自具有脈衝寬度與脈衝高 度,其會根據一模式而改變,在此模式中至少一個程式化 重試脈衝具有與其他在模式巾的程式化重試脈衝不同的脈 衝寬度或脈衝高度。 在本發明實施例中,程式化的方法包括一確認步驟。 此確認步驟包括產生一信號’如參考電流,其可以表示出 在選定記憶胞中的特性值。然後,將此信號與參考信號作 比較,以確認預期資料的程式化。 在單一 S己憶胞被程式化多次的實施例中,此程式化的 方法包括使用例如狀態機、其他資料儲存器或邏輯電路址 構,維持施與記憶胞陣列之程式化循環數的紀錄…在確= 步驟或感測記憶胞資料時所使用的參考信號是根棣已執行 的程式化循環數而分別從對應多個程式化循環之多個參考 信號來源中選擇出的。 本發明藉由簡單的改變參考位準而允許重置儲存在記 憶胞列中的資料,其中儲存在記憶胞列中的資料是藉由設 定陣列中之記憶胞的特性高於或低於參考位準來表示資料 值。上述的術語「重置」是指設定所有記憶胞至標準值, 通常對單一位元記憶胞為「0」或對二位元記憶胞為「〇〇」 等等。此重置的方法使對陣列進行多次程式化操作以於單 一記憶胞儲存一或多個位元之資料成為可能。根據本發明 實施例之程序包括首先藉由對單位元記憶胞改變參考位準 13 7 3 6^^Qtwf.doc/006 或對多位元記_改變參考位 因此陣列中的所有記憶胞具有處於胞陣列’ 此位準是高於錢騎的參考鱗 測特性, 變參考位準岭行重置之後,使 在藉由改 準使陣列可以如上述-樣藉由施加應力;選: =朗化。於是,本發明執行重置二= 其中抹除操作係設計成藉由施加應力至記 隱胞改k已被感測之記憶胞的特性。在此種理論下,本 發明之程式財法的特徵在於「*需抹除」。 在-些實施例中,在單一記憶胞中儲存有多位元,程 式化的方法包括對被程式化至記憶朗乡位元資料提供一 數值。在確認步驟或感測在記憶胞中的多位元資料時所使 用的參考信號是從用於多位元資料之多個數值所對應的多 個參考信號中選出。 本發明也提供一種積體電路,包括一邏輯電路如通用 處理器或專用邏輯電路、一高速記憶體如靜態隨機存取記 憶體與如上述的根據由介電層的累進崩潰引發的可程式電 阻之PREM記憶胞陣列。在一些實施例中,用於程式化 記憶胞陣列的邏輯電路包括由晶載通用處理器所執行的指 令0 為讓本發明之上述和其他目的、特徵和優點能更明顯 易懂’下文特舉一較佳實施例,並配合所附圖式,作詳細 說明如下。 【實施方式】 Ι2736〇α__ 請參照第1圖至第36圖’其用以說明本發明之較佳 實施例。 第1圖至第3圖所繪示為本發明之基本記憶胞結構 圖。如第1圖所示,記憶胞包括導體10、累進崩潰介電 層11與導體12。導體10係作為第一電極。導體12係作 為第二電極。介電層11具有一厚度或結構特性,使其會 依照應力而累進改變。典型的介電材料都具有累進崩潰特 性而造成電阻的累進改變,其包括超薄氧化層,例如氮氧 化發之厚度小於20埃左右’較佳是小於15埃左右。 氮氧化矽的形成方法包括使用二氧化矽熱成長製程, 在製程期間或之後,暴露在NO或Ν20下進行氮化反應, 此步驟可以與記憶胞陣列外的周邊電路區的熱氧化製程一 起進行。介電層11也可以使用其他經過氮化製程或未經 過氮化製程的二氧化矽。介電層11也可以包括利用化學 氣相沈積法(CVD)、電漿增強化學氣相沈積法(pecvd)、 TEOSCVD、高密度電漿化學氣相沈積法(HPCVD)或其他 製程所形成的氧化矽或其他材料。介電層11所使用的材 質也可以包括利用濺鍍、脈衝氣相沈積(PVD)、喷射氣相 沈積(JVD)、原子層沈積(ALD)所形成之氧化物。對於各 種可應用之沈積技術可以參照R0ssnagel,S· Μ.等人提出 之’·Ριόπι PVD to CVD to ALD for interconnects and related applications,丨’ Interconnect Technology Conference,2001· Proceedings of the IEEE 2001 International, 4-6 June 2001HfS_, HfA10x, TaOxNy, Zr〇2, ZrSi ο 2 1χ y, 2 must be ix〇y, La2〇3, etc. tw twf.doc/006 can be used as the material of the above ultrathin layer of memory cells. The composition of the first electrode and the second electrode may vary depending on the environment in which the present invention is applied. In one embodiment of the invention, the first electrode includes a polysilicon layer and the second electrode includes a conductive diffusion region of the semiconductor substrate. In another embodiment, the polysilicon layer and the conductive diffusion region have opposite conductivity patterns to form a type of diode-like resistive memory cell. In another embodiment, the first electrode and the second electrode may comprise a conductor combination comprising a metal such as copper, aluminum, tungsten, titanium, an alloy, and combinations thereof, p-type and n-type polysilicon, p-type and n-type diffusion regions, Metal halides, semi-metals, and the like. In some embodiments, the material of the electrode comprises a material containing an element, and the material of the inter-electrode material layer between the electrodes comprises a compound of the same element. For example, the materials of the first and second electrodes include Shi Xi, such as amorphous stone, single crystal, polycrystalline germanium, metal telluride and the like, between the electrodes between the first electrode and the second electrode. The material layer includes a ruthenium-containing compound such as ruthenium oxide or nitriding. The present invention also provides an integrated circuit comprising the memory cell array as described above, which utilizes the surface of the dielectric layer to compensate for the memory cells, and uses the sensing circuit to sense the progressive collapse of the memory cells in the array. the amount. In the present embodiment, the amount of the mask is indicated by the change in the resistance of the memory cell. The octave is established in a single memory cell by multi-step changes in stylization and sensing characteristics. For example, in one embodiment, the stylized logic is operated and the _ bias is applied to the -selected & memory cells, and then the bias is determined to be the expected amount of progressive change. If it is confirmed that the operation has failed, then retry the voltage and confirm the operation and repeat the remainder until the memory cell is successful or the limit of the retry is reached. The sensing circuit of the embodiment of the invention includes a reference current source and a circuit for comparing the current from the memory cell with the current from the reference current source. In an embodiment for sensing a single memory cell multi-bit or multi-order characteristic progressive change, the sensing circuit can include a plurality of reference current sources, and the circuit is configured to compare the current from the memory cell with the plurality of reference current sources One or more reference currents. The present invention further provides a stylized method of memory cells, including supplying a stress to the inter-electrode material layer to cause a progressive change in the characteristics of the layer during programming. Because of the progressive changes in characteristics, multi-level programming can be achieved. This stylization can be applied to multiple programming a memory cell without the need for an erase operation, but can program multiple bits in a single memory cell, and Can be combined with multiple bits and multiple stylization. According to an embodiment of the invention, the process of supplying a stress to the memory cell during programming includes using a continuous current pulse and a confirmation step, which is described as follows: ^ Supplying a first programmed pulse to have a first pulse height and a pulse width of the memory cell; measuring whether the memory cell is programmed against the first stylized pulse; and, if not supplying a stylized retry pulse to the memory cell; measuring whether the memory cell corresponds to a programmatic retry a pulse is programmed; and, if it is not repeatedly supplied another stylized retry pulse to the memory cell and measuring whether the memory is programmed, the system measures the cell age The number of times of refinement or retries reaches a maximum value; wherein, the stylized retry pulses each have a pulse width and a pulse height, which are changed according to a mode in which at least one stylized retry pulse has the same mode as the other The stylized retry pulse of the towel has a different pulse width or pulse height. In an embodiment of the invention, the stylized method includes a confirmation step. This confirmation step includes generating a signal ', such as a reference current, which can represent the characteristic value in the selected memory cell. This signal is then compared to the reference signal to confirm the stylization of the expected data. In embodiments where a single S memory cell is programmed multiple times, the stylized method includes maintaining a record of the number of programmed cycles applied to the memory cell array using, for example, a state machine, other data store, or logic circuit configuration. The reference signal used in determining the step or sensing the memory data is selected from a plurality of reference signal sources corresponding to the plurality of stylized cycles, respectively, based on the number of programmed cycles executed. The invention allows the data stored in the memory cell column to be reset by simply changing the reference level, wherein the data stored in the memory cell column is set higher or lower than the reference bit by setting the characteristics of the memory cell in the array. The standard value is indicated. The above term "reset" refers to setting all memory cells to standard values, usually "0" for a single bit memory cell or "〇〇" for a binary bit memory cell, and so on. This method of resetting makes it possible to program the array multiple times to store one or more bits of data in a single memory cell. The program according to an embodiment of the present invention includes first changing the reference level by changing the reference level to the unit cell memory or changing the reference bit to the multi-bit _ so that all the memory cells in the array have Cell array 'This level is higher than the reference scale characteristic of Qian Qi, after the reset reference level is reset, so that the array can be stressed by the above-mentioned method by changing; select: = . Thus, the present invention performs a reset two = where the erase operation is designed to characterize the memory cells that have been sensed by applying stress to the cryptographic cell. Under this theory, the program finance method of the present invention is characterized by "* need to be erased". In some embodiments, a multi-bit is stored in a single memory cell, and the method of programming includes providing a value for the stylized to memory location data. The reference signal used in the confirmation step or sensing the multi-bit data in the memory cell is selected from a plurality of reference signals corresponding to a plurality of values for the multi-bit data. The present invention also provides an integrated circuit including a logic circuit such as a general purpose processor or a dedicated logic circuit, a high speed memory such as a static random access memory and a programmable resistor according to a progressive collapse caused by a dielectric layer as described above. PREM memory cell array. In some embodiments, the logic circuitry for the stylized memory cell array includes instructions executed by the on-board general purpose processor to make the above and other objects, features and advantages of the present invention more apparent. A preferred embodiment, in conjunction with the drawings, is described in detail below. [Embodiment] Ι 2736 〇 α__ Please refer to Figs. 1 to 36 for explaining a preferred embodiment of the present invention. Fig. 1 to Fig. 3 are diagrams showing the basic memory cell structure of the present invention. As shown in Fig. 1, the memory cell includes a conductor 10, a progressive collapse dielectric layer 11 and a conductor 12. The conductor 10 serves as a first electrode. The conductor 12 is used as a second electrode. The dielectric layer 11 has a thickness or structural characteristic such that it undergoes a progressive change in accordance with stress. Typical dielectric materials have progressive collapse characteristics that cause progressive changes in electrical resistance, including ultra-thin oxide layers, such as oxynitride having a thickness of less than about 20 angstroms, preferably less than about 15 angstroms. The method for forming bismuth oxynitride includes using a cerium oxide thermal growth process, and performing a nitridation reaction under exposure to NO or krypton 20 during or after the process, which may be performed together with a thermal oxidation process of a peripheral circuit region outside the memory cell array. . The dielectric layer 11 may also use other cerium oxide which has been subjected to a nitridation process or a non-nitridation process. The dielectric layer 11 may also include oxidation formed by chemical vapor deposition (CVD), plasma enhanced chemical vapor deposition (pecvd), TEOSCVD, high density plasma chemical vapor deposition (HPCVD), or other processes.矽 or other materials. The material used for the dielectric layer 11 may also include oxides formed by sputtering, pulse vapor deposition (PVD), jet vapor deposition (JVD), atomic layer deposition (ALD). For various applicable deposition techniques, reference is made to R0ssnagel, S. Μ. et al., '·Ριόπι PVD to CVD to ALD for interconnects and related applications,丨' Interconnect Technology Conference, 2001· Proceedings of the IEEE 2001 International, 4- 6 June 2001
Page(s): 3- 5 ; Jelinek,M·等人提出之"Hybrid PLD technique 15 I2736ftQ twf.doc/006 for nitrogen rich CN,layers/1 Lasers and Electro-Optics Europe,2000, Conference Digest 2000, Conference on 10-15 Sept 2000, Page(s): 1 ρρ· ; Wang,X,W·等人提出之’’Ultra-thin silicon nitride films on Si by jet vapor deposition,,, VLSI Technology,Systems,and Applications,1995· Proceedings of Technical Papers” 1995 International Symposium pn,3 1 May-2 June 1995, Page(s): 49 -52·等相關文獻。 此外,本發明也可以使用其他具有累進崩潰特性的介 電材料,包括如氧化矽-氮化矽_氧化矽QNO之多層堆疊 介電層或氧化鋁等。介電材料如Al2〇3、YTa205、Hf02、 Y2〇3、Ce02、Ti02、HfSixOy、HfSiON、HfA10x、TaOxNy、Page(s): 3- 5 ; Jelinek, M. et al., "Hybrid PLD technique 15 I2736ftQ twf.doc/006 for nitrogen rich CN, layers/1 Lasers and Electro-Optics Europe, 2000, Conference Digest 2000, Conference on 10-15 Sept 2000, Page(s): 1 ρρ· ; Wang, X, W· et al. ''Ultra-thin silicon nitride films on Si by jet vapor deposition,,, VLSI Technology, Systems, and Applications, 1995· Proceedings of Technical Papers” 1995 International Symposium pn, 3 1 May-2 June 1995, Page(s): 49-52· et al. In addition, other dielectrics with progressive collapse characteristics can be used in the present invention. Materials, including multilayer stacked dielectric layers such as yttria-tantalum nitride-yttria QNO or alumina, etc. Dielectric materials such as Al2〇3, YTa205, Hf02, Y2〇3, Ce02, Ti02, HfSixOy, HfSiON, HfA10x , TaOxNy,
Zr〇2、ZrSixOy ' La2〇3等也可以作為記憶胞之上述介電層 11的材質。 . 導體10與導體12包括導體材料,但是並不限定為金 屬、半金屬或導電摻雜半導體。導體1〇與導體12並不需 要由相同的材質所構成,但是其較佳是使用在製造過程中 已經準備好之材料。於是,導體與導體12之材質可以 使用摻雜半導體,如p型與n型多晶矽、摻雜砷化鎵等; 金屬,如紹、銅、鶴、鈦、组等;半金屬,如Tiw、TiN 等;金屬石夕化物,如WSix、TiSix。 、第2圖為繪示-較佳實施例,包括由?型多晶石夕所構 ^之第-電極I3、由厚度I5埃之氮氧化㈣構成之累進 朋潰介電層Μ與由在半導體基底巾之n型擴散區所構成 之第二電極15。第3圖為繪示另一較佳實施例,包括由^ 16 itwf.doc/006 型多晶料構成之第—電極16、累補潰介電们7與由 在半導體基底中之p型擴散區所構成之第二電極18了用 於形成第一電極與第二電極之導體(p型或n型)中所使用 之摻質的導電錢可為相同也可為不同,其選擇可取決於 製程方便性或不同的元件設計。 第4圖顯示上述由de Graaf等人所提出習知記憶胞 所使用25埃厚氮氧切層的嚴㈣潰特性。如第4圖所 不’在應力^時’間35秒後,程式化電流突然的從〇左 右升到3〇毫安培左右,顯示^介電層的嚴重崩潰。 請參照第5圖所示,本發明應用具有累進崩潰特性之 介電層’程式化電流會隨著程式化_而逐漸增加。而且, 對,^有p ^多晶;^上電極與n型埋入式擴氣下電極的15 埃氮氧化㈣言,當施加1.8伏特之電壓至上電極與-υ 伏特之電壓至下電極以進行程式化時,程式化時間,從$秒 左右至14=#左右’程式化電流會從丨毫安培近似線性的 上:至I笔女$。在堃子此種已程式化之類二極體記憶胞進 行續取操作時’其_取電壓例如是施加13伏特至上電極 與0伏特至下電極。 如第6圖所示,在讀取電壓為施加13伏特至上電極 與〇伏特至下電極的情況下,讀取電流會隨著#施加18 伏特之^壓至上電極與·15伏特之電壓至下電極以進行程 =化之程,化時間增加而增加。由第6圖可看出,讀取電 流基本亡是累進的,當程式化時間從15秒左右至15〇秒 左右’項取電流會從小於1微安培近似線性的上升至 0.25 17 itwf.doc/006 毫安培。 如第6A圖所示,在讀取電壓為施加i 極與〇伏特至下電極的情況下,·特至上電 電壓量值,棘電齡隨著程式㈣種不同程式化施加 條標記線表示施加2.5伏特增加而增加。第1 ..,._ _ 包至主予元線(上電極)盘-2 3 伙特之電驗子讀(上電極)與^ 條標_祕加2.5伏=壓= 2電極)與_1.7伏特之電壓至位元線(下電極)。第4條 線表不施加2.5伏特之電壓至字元線(上電極)與_14 之電壓至位元線(下電極)。由第6Α圖可看出,對於不同 的程式化f壓量值,讀取電流基本上是累祕,且對應程 式化時間而近赠性的上升。喊化驗越高「達到相同 的電阻累進改變量所需要的時間區間越短。舉來 程式化電壓量值為4·8伏特(2·5伏特減如2·3伏特)程式化 100毫秒(0·1秒)後,得到的讀取電流為約95微安培左右。 對於程式化電壓量值為4.5伏特(2.5伏特減去-2·〇伏特)的 情況下,程式化時間需要1秒鐘才能夠得到約95微安培 左右的讀取電流。 σ 第7圖是第6圖之讀取電流對程式化時間關係圖加上 資料位準啟始值與參考電流值。如此,藉由設定參考電流 Ref_l.l在8微安培,將代表1位元之資料程式化至記憶 胞中。在進行記憶胞程式化操作時,為了儲存等於「L 的資料值而需要施加低電壓程式化應力1〇〇秒左右。為了 18 12736題— 儲存等於「〇」的資料值而不需要施加任何程式化應力。 第8圖是第6圖之讀取電流對程式化時間關係圖加上 用於在單一記憶胞中儲存二位元資料的資料位準啟始值與 參考電流值。為了儲存等於「〇〇」的資料值而不需要施加 任何程式化應力。為了儲存等於「01」的資料值而需要施 加低電壓程式化應力75秒左右。為了儲存等於「1〇」的 資料值而需要施加低電壓程式化應力11〇秒左右。為了儲 存等於广11」的資料值而需要施加低電壓程式化應力15〇 秒左右。參考電流源是為了感測資料值的目的而設置的。 參考電流Ref-1.1在本實施例中設定為4微安培。參考電 流Ref-1.2在本實施例中設定為12微安培。參考電流 Ref-U在本實施例中設定為21微安培。藉由將讀取電流 與參考電流作比較,而可以檢測出資料值。.· 第二圖是第6圖之讀取電流對程式化時間關係圖加上 2在單-記憶胞中儲存四位对料的資料位準啟始值與 ^考電流值。如第9圖所示,為了達到單—記胞多位元儲 =而將程式化時間與參考電流位準設定的更為緊密。當 :時Γ與參考電流的差可以使用習知的感測技 術,例如已鋪用在乡階_記麵 之感測技術,以達到單一己愔己隱體等讀 多位元儲存。*錢胞贿讀存或單-記憶胞 ,胞本Γ之可程式電阻不需抹除記憶體的記 列結構圖。此結構包括隔離溝渠30、3卜32、33, -在與頁面垂直的線上延伸。於隔離溝渠⑽、Μ、3 ►twf.doc/006 之間設置有在平行線延伸的下電極導體34、35、允。下 電極導體34、35、36係形成於-絕緣基底上制用摻質 擴散而形成在一半導體基底中。超薄氧化層37、38、39 係形成於下電極導體34、35、36上。在一實施例中,超 薄氧化層37、38、39係利用單一沈積步驟,而於晶片的 整個陣列區上形成-整層薄膜。在另—實施例中,超薄氧 化層係經圖案化以符合記憶胞的佈局。上電極導體,包括 導體40覆蓋超薄氧化層37、38、%,包括與下電極導體 垂直的多辦行導線,而在兩者相交之處形成記憶胞。下 電極導體34、35、36係作鱗觸位元線與字元線之其 中之一。同樣的,上電極導體(例如導體40)係作為陣列的 字元線與位元線之其中之一。 第11圖所述為類似第1〇圖所示的記憶胞孛列結構剖 面圖’其巾上電極導體45是由p型多晶㈣構成',下電 極導體46是由n型埋人式擴散井區所構成。在—些實施 例中11型埋入式擴散井區是形成於未圖示之深η型井區 ^的:型擴散隔離井區巾,用以供給貞電壓至下電極導 11圖所不之結構中,其他與第10圖相同的構件, 在此不再贅述。 丁 第12圖所述為類似帛1〇圖所示的記憶胞陣列結構剖 :其中上電極導體47是由η型多晶石夕所構成,下電 是由ρ型埋人式擴散井區所構成。第12圖所示 ^中’構件與第1g _ _構件,在此不再贅述。 I3圖7C本發明之可程式電阻不需抹除記憶胞的基 20 12736· twf.doc/006 程圖。此製程包括形成下電極導體5〇、在下電 =成介電層51,此介電層51在低電壓具有 ;一1二”性、以及在介電層51上形成上電極導體52。 適用;开在介電層形成之前’需使下電極導體50 在、、尤積具1乳化層或其他介電層。舉例來說,導體層 ’下電極導體5〇包括形成有助S -阳終I '的成核層或結晶層。下電極導體也可以包括 i行^避免材料擴散至介電層,以保護將發生的累進崩 摻質製程’其中利用植入n型 然後,在下—步驟區而形成下電鱗體55。 此介電層58在低電軸介.電層58, 電Μι 進朋潰的特性。之後,在介 電層:1上形成由ρ型多晶矽 第15圖表示第13圖所示的製程,其Τ導體二 :6丄^半ί體基底62的擴散區而形成下電極導體 3電/:=,在下電極導趙6°上形成介二, =電層63在低電a具有累進崩潰的特性 電層^ ^形成曰由n型多晶石夕所構成之上電極導體64 第16圖疋本發明之可程式電 泰 列結構的製造流程圖。如第“:記=記 +導體基底膽,例如是具极供一 _中形成隔離溝渠101_10 :=;:。於基底 长二實施例中,其形成 21 ►twf.doc/006 方法例如是淺溝渠隔離製程(STI)。在另一實施例中,也 可以使用LOCOS氧化製程而形成隔離結構。然後,植入 N型摻質106,以於隔離溝渠10M05之間形成埋入式擴 散區107-110。在一些實施例中,係形成有深η型井區, 接著於深η型井區内形成ρ型隔離井區,然後,再於ρ型 隔離井區内形成η型埋入式擴散區。深η型井區與ρ型隔 離井區是在形成隔離溝渠101-105之前或之後形成的。 在η型埋入式擴散區107-110形成之後,於η型埋入 式擴散區107-110表面形成超薄介電層111-114。在本實 施例中,單晶矽中的η型埋入式擴散區107-110表面是一 個良好的表面,此表面可形成適用於記憶胞的氮氧化矽。 在其他實施例中,此表面也可經過處理,以預備形成介電 層。 * 之後,於基底100上沈積一層ρ型多晶矽115,然後 圖案化ρ型多晶矽115以形成與由η型埋入式擴散區 107-110所構成之位元線垂直相交的字元線。在字元線與 位元線的相交區域形成類二極體可程式電阻不需抹除記憶 胞,其可以利用習知的字元線與位元線編碼結構來進行存 取。在本發明之實施例中,上下電極分別包括位元線與字 元線的一部份’而且上下電極分別和位元線與字元線接 觸。在另一實施例中,上下電極在相交區域更包括形成一 層額外的材料層,此材料層分別和位元線與字元線接觸。 第17圖是本發明另一實施例陣列結構的製造流程 圖。其構件與第16圖給予標號的構件類似。如第17圖所 22 f.doc/006 不,先提供一半導體基底100。於基底1〇〇表面利用微影 技術定義出陣列區域,或者在基底表面植型掺質12〇, 以形成η型擴散區12卜於基底1〇〇中形成隔離溝渠122_ 126以截斷η型擴散區121並填入介電材料,而形成埋入 式擴散位元線127_130。從此製程之後,製造流程與上述 第16圖的製造流程相同。 根據第16圖與第17圖的製造流程而可以製作出一陣 列結構,此陣列結構之上視圖如第18圖所示,此陣列結 構具有埋入式擴散位元線200-202與多晶矽字元線2〇弘 205,且兩者彼此垂直相交。記憶胞形成在兩者相交的區 域,如相交區域206。 第19圖與第20圖是本發明之不同的記憶胞陣列的感 測電路結構之示意圖。在第19圖與第2〇圖中,·陣列25〇 代表記憶胞陣列,其是由在圖式中水平排列的多條字元線 與垂直排列的多條位元線所構成。記憶胞是以二極體符號 來表示由第16圖至第18圖的製程所製造出的類二極體可 私式電阻不需抹除記憶胞。使用標準的編碼技術,位元線 之一連接至資料輸出線251。資料輸出線251連接至電流 模式感測放大器252-254,此3個感測放大器係用於測量 如第8圖所示的單一記憶胞二位元儲存的資料。每一個電 流模式感測放大器係連接至一參考電流源。於是,電流模 式感測放大器252係連接至供應參考電流Ref-L3之參考 電流源。電流模式感測放大器253係連接至供應參考電流 Ref_1.2之參考電流源。電流模式感測放大器254係連接 23 I2736i)4Q twf.doc/006 至供應參考電流Ref_l.l之參考電流源。感測放大器的輸 出在導線255-257上然後被譯碼以計算出儲存在選定記憶 胞中的二位元的值。 第20圖是本發明的另一種記憶胞陣列的感測電路結 構之示意圖。來自陣列250的資料輸出線251連接至單一 電流模式感測放大器260。在導線265上供給至感測放大 器之參考電流是由開關261-263來選擇,這些開關261-263 係分別連接至供應參考電流Ref-1.3、Ref-1.2、Ref-l.l之 參考電流源。在其他實施例中,並不使用數位感測放大器, 而且儲存在記憶胞中的資料係以類比值輸出而被送出。 第21圖是應用可程式電阻不需抹除記憶胞pREM陣 列270之記憶體元件的簡單組成方框圖。記憶體元件包括 行譯碼器271與列譯碼器272,行譯碼器271與列譯碼器 272連接至位址匯流排273。來自電壓源275之用於讀取 及程式化操作的供應電壓經由列譯碼器272與行譯碼器 271供給至陣列中的選定記憶胞。感測放大器與輸入資料 結構276連接至列譯碼器272之輸出端、輸入資料匯流排 280與輸出資料匯流排281。讀取與程式化狀態裝置277 連接e己憶體元件的各構件。狀態裝置可由專用邏輯、可程 式邏輯陣列結構、通用處理器執行指令或這些執行的組合 來執行。 如上所述,可程式電阻不需抹除記憶胞陣列也可用於 在單一 §己憶胞中儲存多位元之資料。在其他實施例中,此 種s己憶胞陣列適用於多次程式化循環。請參照第22圖至 24 ltwf.doc/006 第25圖,執行第一程式化循環以設定用於感測在位準 Ref_l之單一位元的參考電流(如第22圖所示)。執行第二 程式化循環以設定用於感測在位準Ref_2之單一位元的參 考電流(如第23圖所示)。執行第三程式化循環以設定用 於感測在位準Ref_3之單一位元的參考電流(如第24圖所 示)。執行第四程式化循環以設定用於感測在位準Ref_4 之單一位元的參考電流(如第25圖所示)。記憶胞程式化 之次數是適用在特定的執行,此次數是由電阻累進改變的 可靠程式化數目的能力,以及在存取記憶體時可清楚區別 所產生的電流位準來決定。讀取與程式化狀態裝置(如第 21圖之標號277所示)係設定成尋找出經執行的程式化循 環數目,以致於適當的參考電流可供應至感測電路。 第26圖是本發明的可程式電阻不需抹除記憶胞之基 本程式化規則。對於第一程式化循環3〇〇,設定用於感測 與驗證的第一電流值(方塊301)。然後,執行應力/程式化 操作以引發一累進崩潰量,使得經程式化記憶胞在讀取時 可以產生大於第一參考電流值的一輸出電流(方塊3〇2)。 在另一實施例中,在程式化操作時施加一應力,使用連續 的短脈衝逐漸的引起電阻的累進改變,短脈衝例如是具有 會改變的波長及/或電壓位準、或固定的波長及電壓位準, 使得記憶胞特性具有多種的累進崩潰控制總量。然後,執 行確説、插作以確定成功的程式化(方塊303)。如果確認失 敗,接著以下一應力脈衝重試應力程式化操作。如果確認 成功,就完成第一程式化循環(方塊304)。如第26圖所示, 25 ltwf.doc/006 只要設定了累進參考電流,第二程式化循環31〇、第三程 式化循環320、第四程式化循環320等等也可以使用相同 的基本程序來執行。 三個典型的程式化操作是有用的,在這些操作中連續 的施加脈衝以在記憶胞特性上建立選擇的累進改變量,^ 據程式規則執行每個脈衝,其中重試規則包括·· 、(1)一確認步驟,以測量是否達到選擇的位準,如果 衝則在每一循環中供應具有等脈衝電壓與等脈衝長度 (2)—確認步驟,以測量是否達到選擇的位準,如果 為否則在母一接縯循環中供應具有增加的脈衝電壓盘蓉 脈衝長度之脈衝。 〃 (3)—確認步驟,以測量是否達到選擇的位嗶,如果 為否,則在母一接續循環中供應具有增加的脈衝長产 脈衝電壓之脈衝。 又、 、(4)一確認步驟,以測量是否達到選擇的位準,如果 ^否,則在接續的循環中供應一脈衝,此脈衝在連續的接 續循環的一或多個步驟之期間,其脈衝寬度與脈衝長度之 其中之一或兩者會改變。 在根據第2圖所示結構的較佳實例中,程式化程序包 括固定η型擴散電極在-2伏特左右之電壓、對p型多= 矽電極逐步的施加〇·5伏特左右至2伏特左右之電壓,以 固定的脈衝寬度(例如lms或i〇ms)在每一步驟増加〇1 伏特之電壓,在每一脈衝之間使用一確認步驟,並且在當 26 rf.doc/006 記憶胞通過確認步驟後停止。 在根據第2圖所不結構的另一定電壓操作中,程式化 程序包括固定η型擴散電極在-2伏特左右之電壓,並對p 型多晶矽電極施加2伏特左右之電壓,以供應等脈衝高 度,並供應具有等脈衝寬度(例如lms或l〇ms)之脈衝, 在每一脈衝之間使用一確認步驟,並且在當記憶胞通過確 認步驟後停止。當然,脈衝高度與脈衝寬度可視特定系統 之需要而改變。 本發明的可程式電阻不需抹除記憶胞顯示出卓越的程 式化干擾與讀取性能。程式化干擾之效能可參照第27a 圖至第27D圖來說明之。第27A圖表示陣列部分的示意 圖’緣示出字元線400-403與位元線410-413。在字元線 401與位元線411之交錯區域的記憶胞A,藉由施加18 伏特之電壓至字元線401上的電極與施加·^伏特之電壓 至位元線411上之電極而被程式化。在字元線4〇1與相鄰 位元線412之交錯區域的記憶胞B ,從字元線接收到18 伏特之電壓,但是位元線接地。在位元線411與相鄰字元 線402之交錯區域的記憶胞c,從位元線接收到_15伏特 之電壓,但是字元線接地。 第27B圖表示選定記憶胞a,其中上電極42〇是由p 型多晶矽所構成,下電極421是由n型埋入式擴散(或井) 區所構成,電極間介電層422是由15埃之氮氧化矽所構 成。3·3伏特左右的程式化電位跨過兩電極促成對於元件 的Ρ_η接合的偏壓模式,供給電壓的絕對值小於2伏特, 27 12736 QAtwf.doc/006 造成跨過兩電極之電阻累進減少,導致讀取電流的累進增 加’如第6圖所示。 第27C圖表示未選定記憶胞C,其中上電極423是 由P型多晶矽所構成,下電極424是由η型埋入式擴散(或 井)區所構成,電極間介電層425是由15埃之氮氧化矽所 構成。從位元線接收到-L5伏特之電壓,但是字元線接地。 對於元件的ρ-η接合的是處於反向偏壓模式。 第27D圖表示未選定記憶胞β,其中上電極426是 由Ρ型多晶矽所構成,下電極427是由η型埋入式擴散(或 井)區所構成,電極間介電層428是由15埃之氮氧化矽所 構成。從字元線接收到1·8伏特之電壓,但是位元線接地。 第28圖顯示出施加如記憶胞β那樣的偏壓使記憶胞 在程式化應力下經過10000秒後的程式化電流仍然很低。 第29圖顯示出施加如記憶胞c那樣的偏壓,使記憶胞在 程式化應力下經過10000秒後的程式化電流仍然很低。而 且’從圖式的刻度並無法看出程式化干擾。 弟30Α圖至第30D圖是本發明之記憶胞陣列讀取與 讀取干擾狀態圖。 本發明的可程式電阻不需抹除記憶胞的讀取性能可參 照第30Α圖至第30D圖是來說明之。第30Α圖表示陣列 部分的示意圖’繪示出字元線5〇〇-5〇3與位元線5〇4-5〇7。 ,字兀線502與位元線505之交錯區域的記憶胞a藉由 轭加1·3伏特之電壓至字元線5〇2上的電極,並使位元線 507上之電極接地而進行讀取。在字元線5〇2與相鄰位元 28 12736· twf.doc/006 線506之交錯區域的記憶胞Bi,從字元線接收到ι·3伏 特之電壓,但是位元線設定在一抑制電壓為13伏特左 右。在字元線503與相鄰位元線5〇5之交錯區域的記憶胞 Β2 ,從字元線接收到〇伏特之電壓,從位元線接收到接 地電位。在位元線506與相鄰字元線5〇3之交錯區域的記 憶胞C,從字元線接收到接地電位,從位元線接收到13 伏特左右之抑制電壓。 第30Β圖表示選定記憶胞Α ,其中上電極52〇是由ρ 型多晶石夕所構成,下電極521是由η型埋入式擴散(或井) 區所構成,電極間介電層522是由15埃之氮氧化石夕所構 成。在如第30Β圖所示的記憶胞中,介電層經程式化至 低電阻狀態。1·3伏特左右的讀取電位跨過兩電極促成對 於元件的ρ-η接合的偏壓模式,而在元件中引發可被感測 出的電流。 第30C圖表示選定記憶胞Α,其中上電極523是由ρ 型多晶矽所構成,下電極524是由η型埋入式擴散(或井) 區所構成’電極間介電層525是由15埃之氮氧化石夕所構 成。在如第30C圖所示的記憶胞中,介電層並未經程式 化至低電阻狀態。L3伏特左右的讀取電位跨過兩電極^ 成對於元件的ρ-η接合的偏壓模式,但是在元件 引發電流。 、’ ^ 第30D圖表示處於程式化狀態之未選定記憶胞, 其中上電極527是由p型多晶石夕所構成,下電極似是由 η型埋入式擴散(或井)區所構成,電極間介電層529是由 29 Ι27360β^_ 15埃之氮氧化矽所構成。下電極接收丨3伏特左右的抑 制偏壓,且上電極接地◊對於元件的ρ-η接合的是處於反 向偏壓模式,而且基本上沒有電流 。同樣的’如第30Α ,所不的記憶胞Β2與記憶胞C都不會產生電流,無論其 是處,經程式化低電阻狀態或未程式化高電阻狀態。 第31圖為讀取電流對閘極電壓關係圖,其中標記線 550表不已程式化記憶胞,標記線551表示未程式化記憶 胞(“未使用”)。如圖所示,當記憶胞被反向偏壓(Vg<〇), 基本上沒有電流。在偏壓狀態之前,Vg小於十分之一伏 特,在經私式化與未程式化記憶胞中都有少量的電流。然 而,在本實施例中Vg於13伏特(線552)左右,經程式化 s己憶胞顯示出具有較未程式化記憶胞為高的電流。 第32圖表示本發明之記憶胞的耐久性。對於處於不 同程式化位準的已程式化記憶胞(標記線56〇與561)與未 程式化記憶胞(標記線562),在長讀取時間的情況下,讀 取電流仍保持接近固定值。 而且,如第33圖所示,資料保持特性也很好。在長 時間鬲溫烘烤之情況下,對於不同程式化位準的已程式化 記憶胞(標記線565與566)與未程式化記憶胞(標記線 567),讀取電流仍保持接近固定值。 因為本發明的記憶胞具有良好的穩定性、耐久性與保 持性,此記憶胞可以程式化多次而適用於單一記憶胞^位 元儲存的記憶胞陣列。第34圖是第6圖的讀取電流對程 式化時間關係圖,加上用於單一記憶胞二位元儲存的資料 I2736IAQ itwf.doc/006 ^準啟始值與參考電流,^也可以支❹次程式化循 本發明之f次程式化、單-記憶胞多位元之實施例的參考 電流值不思圖。在第1循環’為了儲存等於「00」的資^ 值而不需要施加任何程式化電壓(應力)。為了儲存等於 οι」的貝料值而需要施加低電壓程式化電壓(應力)25秒 左右。,了儲存等於「1G」㈣料值而需要施加低電壓程 式化電壓(應力)35秒左右。為了儲存等於「u」的資料值 而需要施_低電壓程式化電壓(應力⑷秒左右。參考電流 源是為了感測資料值的目的而設置的。參考電流 在本實%例中没定$ 1〇微安培。參考電、流Ref_12在本 實施例中蚊為22微安培。參考電流RefL1.3在本實施 例中設定為35微安培。藉由將讀取錢與參f電流作比 較,而可以檢測出資料值。對於第2程式化循環,藉由將 讀取電流與參考電流Ref_2j、Ref-2;2、Ref-2 3作比較, 而可以檢測出資料值。對於第3程式化循環,藉由將讀取 電流與參考電流Ref-3·:!、Ref-3.2、Ref_3.3作比較,而可 以檢測出資料值。 第35圖與第36圖是本發明之具有可程式電阻不需抹 除s己憶胞應用在「系統晶片(SyStein On a Chip,SOC)」的 實施例。可程式電阻不需抹除記憶胞(PREM)的製程完全 與標準互補式金氧半導體(CM〇s)製程相容,在實施例中 只需要增加一道光罩,所以相當的適用於製作SOC產品 中緊密的非揮發性記憶體。對於具有p+型多晶矽上電極 31 Ι2736βα twf.doc/006 與η+型埋入式擴散下電極 道額外的H 構’只需要增加一 、貝卜㈣罩’以疋義出用於η+型埋 步驟,而其他STI製程、介電声之擴㈣的植入 石㈣占制於0 u ▲ 電廣之化成製程與P+型多晶 夕少成製耘疋共早CMOS結構的形成製程。 的低電壓操作使得此種記議於低;^卜== 一種良好的選擇。 一低功丰%楗至 第35圖繪製的積體電路包括可程式 需 =〇>,)陣列601、邏輯電路6〇3例如專用 或可知式閘極陣列邏輯電路、與靜態隨機存取記憶體 ^M) 602 4REM陣列可用於儲存較多的不&資 料例如可㈣閘極_麟_财㈣程式規格。、 可用於儲存在邏輯電路’操作細所使用的 第36圖是本發明的另-系統晶片的實施例。第36圖 繪製的積體電路包括可程式餘不輯除記,隨(pREM) 陣列7(U、邏輯電路7G3例如專用邏輯電路或可程式閘極 陣列邏輯電路、靜態隨機存取記憶體(SRAM)搬與通用 處理器704。PREM陣列701可用於儲存通用處理器7〇4 所執行的扣令程式。而且,由外加控制器儲存在SRAM7〇2 或從PREM陣列701傳送至SRAM 7〇2的指令可供應至 處理器,並由處理器執行以控制pREM陣列7〇1的程式 化0 總而言之,本發明提供一種稱為pREM(可程式電阻 不需抹除記憶體(Programmable Resistor With Erase-less 32 1273· twf.doc/006Zr〇2, ZrSixOy 'La2〇3, etc. may also be used as the material of the dielectric layer 11 of the memory cell. The conductor 10 and the conductor 12 comprise a conductor material, but are not limited to metal, semi-metal or conductive doped semiconductors. The conductor 1 and the conductor 12 do not need to be composed of the same material, but it is preferable to use a material which has been prepared in the manufacturing process. Therefore, the material of the conductor and the conductor 12 can be doped semiconductor, such as p-type and n-type polysilicon, doped gallium arsenide, etc.; metal, such as Shao, copper, crane, titanium, group, etc.; semi-metal, such as Tiw, TiN Etc.; metal lithology, such as WSix, TiSix. Figure 2 is a pictorial representation - a preferred embodiment, including The first electrode I3 of the type polycrystalline stone, the progressive electrode layer composed of the nitrogen oxide (4) having a thickness of I5 Å, and the second electrode 15 composed of the n-type diffusion region of the semiconductor substrate. FIG. 3 is a view showing another preferred embodiment, including a first electrode 16 composed of a ^16 itwf.doc/006 type polycrystalline material, a bulk dielectric layer 7 and a p-type diffusion in a semiconductor substrate. The conductive electrode of the second electrode 18 formed by the region for forming the dopant used in the conductor (p-type or n-type) of the first electrode and the second electrode may be the same or different, and the selection may depend on Process convenience or different component design. Fig. 4 is a view showing the strict (four) collapse characteristics of the above-mentioned 25 angstrom thick oxynitride layer used by the conventional memory cells proposed by de Graaf et al. As shown in Fig. 4, after 35 seconds between the stresses, the stylized current suddenly rises from about 〇 to about 3 mA, indicating a serious breakdown of the dielectric layer. Referring to Fig. 5, the application of the dielectric layer of the present invention with progressive collapse characteristics will gradually increase with the stylization. Moreover, there are p ^ polycrystals; ^ 15 electrodes of the upper electrode and the n-type buried gas diffusion electrode are oxidized (4), when a voltage of 1.8 volts is applied to the upper electrode and the voltage of -υ volts to the lower electrode When stylized, the stylized time, from around $second to around 14=', the stylized current will be approximately linear from 丨 milliamperes: to I female. When the dice memory cell such as the dice is renewed, the voltage is applied, for example, to apply 13 volts to the upper electrode and 0 volt to the lower electrode. As shown in Figure 6, when the read voltage is 13 volts applied to the upper electrode and the volts to the lower electrode, the read current will be applied to the upper electrode with a voltage of 15 volts to the next 15 volts. The electrode is increased in the course of the process, and the time is increased. It can be seen from Fig. 6 that the read current is basically progressive, and when the stylized time is from about 15 seconds to about 15 sec, the current will rise from less than 1 microamperes to approximately 0.25 17 itwf.doc /006 milliamps. As shown in Fig. 6A, in the case where the reading voltage is the application of the i-pole and the volt-volt to the lower electrode, the magnitude of the voltage is applied to the voltage, and the age of the spine is applied with the different stylized application bar marking lines of the program (four). 2.5 volts increase and increase. 1st..,._ _ package to the main elementary line (upper electrode) disk-2 3 mate special electrical test read (upper electrode) and ^ bar mark _ secret plus 2.5 volts = pressure = 2 electrodes) and _ A voltage of 1.7 volts to the bit line (lower electrode). Article 4 The line meter does not apply a voltage of 2.5 volts to the word line (upper electrode) and the voltage of _14 to the bit line (lower electrode). It can be seen from Figure 6 that for different stylized f-pressure values, the read current is basically tiring and corresponds to the increase in the program time. The higher the test, the shorter the time interval required to reach the same resistance progressive change. The programmed voltage value is 4·8 volts (2.5 volts minus 2·3 volts) stylized 100 milliseconds (0 After 1 second, the resulting read current is about 95 microamps. For a stylized voltage value of 4.5 volts (2.5 volts minus -2 volts), the programming time takes 1 second. A read current of about 95 microamperes can be obtained. σ Fig. 7 is a diagram of the read current vs. stylized time relationship of Fig. 6 plus the data level start value and the reference current value. Thus, by setting the reference current Ref_l.l at 8 microamperes, stylizes data representing 1 bit into memory cells. In the memory cell programming operation, in order to store the data value equal to "L, it is necessary to apply low voltage stylized stress." About seconds. For the 18 12736 problem - store the data value equal to "〇" without applying any stylized stress. Figure 8 is a diagram of the read current vs. stylized time relationship of Figure 6 plus the data level start value and the reference current value for storing the binary data in a single memory cell. In order to store data values equal to "〇〇", no stylized stress is required. In order to store the data value equal to "01", it is necessary to apply a low voltage stylized stress for about 75 seconds. In order to store data values equal to "1", it is necessary to apply a low voltage stylized stress of about 11 sec. In order to store a data value equal to a wide 11", it is necessary to apply a low voltage stylized stress of about 15 sec. The reference current source is set for the purpose of sensing the data value. The reference current Ref-1.1 is set to 4 microamperes in this embodiment. The reference current Ref-1.2 is set to 12 microamperes in this embodiment. The reference current Ref-U is set to 21 microamperes in this embodiment. The data value can be detected by comparing the read current with the reference current. The second figure is the read current vs. stylized time relationship diagram of Figure 6 plus 2 data level starting values and test current values for storing four pairs of materials in a single-memory cell. As shown in Figure 9, the stylized time is set more closely with the reference current level in order to achieve single-cell multi-bit storage. When the difference between the time and the reference current can be used, a conventional sensing technique can be used, for example, a sensing technique that has been applied to the town level _ face to achieve a multi-bit storage such as a single self-concealed body. * Money bribery reading or single-memory cell, the programmable resistance of the cell is not required to erase the memory structure chart. This structure includes isolation trenches 30, 3b, 32, 33, - extending on a line perpendicular to the page. The lower electrode conductors 34, 35 extending in parallel lines are disposed between the isolation trenches (10), Μ, 3 ► twf.doc/006. The lower electrode conductors 34, 35, 36 are formed on the insulating substrate to be doped and diffused to form a semiconductor substrate. Ultrathin oxide layers 37, 38, 39 are formed on the lower electrode conductors 34, 35, 36. In one embodiment, the ultra-thin oxide layers 37, 38, 39 utilize a single deposition step to form a full film over the entire array area of the wafer. In another embodiment, the ultra-thin oxide layer is patterned to conform to the layout of the memory cells. The upper electrode conductor, including the conductor 40, covers the ultra-thin oxide layer 37, 38, %, including a plurality of wires that are perpendicular to the lower electrode conductor, and forms a memory cell where the two intersect. The lower electrode conductors 34, 35, 36 are one of the scale contact line and the word line. Similarly, the upper electrode conductor (e.g., conductor 40) is one of the word and bit lines of the array. Figure 11 is a cross-sectional view of the memory cell array structure similar to that shown in Fig. 1 'the upper electrode conductor 45 is composed of p-type polycrystal (4)', and the lower electrode conductor 46 is diffused by n-type buried structure. The formation of the well area. In some embodiments, the 11-type buried diffusion well region is formed in a deep η-type well region (not shown): a type of diffusion isolation well region for supplying a 贞 voltage to the lower electrode guide 11 In the structure, other components that are the same as those in FIG. 10 will not be described again. Figure 12 is a cross-sectional view of a memory cell array structure similar to that shown in Figure 12, in which the upper electrode conductor 47 is composed of n-type polycrystalline glaze, and the power-off is performed by a p-type buried diffusion well region. Composition. The "Medium" and the 1g___members shown in Fig. 12 are not described here. I3 Figure 7C The programmable resistor of the present invention does not need to erase the base of the memory cell 20 12736· twf.doc/006. The process includes forming a lower electrode conductor 5A, a lowering of the dielectric layer 51, the dielectric layer 51 having a low voltage, and a top electrode conductor 52 formed on the dielectric layer 51. Before the dielectric layer is formed, it is necessary to make the lower electrode conductor 50, especially the emulsion layer or other dielectric layer. For example, the conductor layer 'lower electrode conductor 5' includes the formation of the auxiliary S-positive I a nucleation layer or a crystalline layer. The lower electrode conductor may also include a row of electrons to prevent diffusion of the material to the dielectric layer to protect the progressive collapse dopant process that will occur, wherein the implanted n-type is then used in the lower-step region. The lower scale body 55 is formed. The dielectric layer 58 is characterized by a low electrical axis dielectric layer 58 and an electric layer. After that, a dielectric layer: 1 is formed on the dielectric layer: 1 by the p-type polysilicon. The process shown in the figure has a germanium conductor 2: a diffusion region of the semiconductor substrate 62 to form a lower electrode conductor 3 electric /:=, forming a second layer on the lower electrode guide 6°, = the electric layer 63 is low Electrical a has a characteristic electrical layer of progressive collapse ^ ^ Formation 之上 is composed of n-type polycrystalline slabs of upper electrode conductor 64 Figure 16 . Thai flowchart for producing electrically programmable structure, such as a column of ": note = + conductor base note gall, for example with a pole forming isolation trenches 101_10 _: =;:. In the second embodiment of the substrate, the formation of the 21 ►twf.doc/006 method is, for example, a shallow trench isolation process (STI). In another embodiment, the LOCOS oxidation process can also be used to form the isolation structure. Then, an N-type dopant 106 is implanted to form a buried diffusion region 107-110 between the isolation trenches 10M05. In some embodiments, a deep η-type well region is formed, followed by a p-type isolation well region in the deep η-type well region, and then an n-type buried diffusion region is formed in the p-type isolation well region. The deep η well zone and the ρ type isolation well zone are formed before or after the formation of the isolation trenches 101-105. After the formation of the n-type buried diffusion regions 107-110, ultra-thin dielectric layers 111-114 are formed on the surface of the n-type buried diffusion regions 107-110. In the present embodiment, the surface of the n-type buried diffusion region 107-110 in the single crystal germanium is a good surface which can form niobium oxynitride suitable for the memory cell. In other embodiments, the surface can also be treated to prepare a dielectric layer. * Thereafter, a p-type polysilicon 115 is deposited on the substrate 100, and then the p-type polysilicon 115 is patterned to form a word line perpendicularly intersecting the bit line formed by the n-type buried diffusion regions 107-110. Forming a diode-like programmable resistor at the intersection of the word line and the bit line does not require erasing the memory cell, which can be accessed using conventional word line and bit line coding structures. In an embodiment of the invention, the upper and lower electrodes respectively comprise a bit line and a portion of the word line' and the upper and lower electrodes are in contact with the word line and the bit line, respectively. In another embodiment, the upper and lower electrodes further comprise an additional layer of material in the intersection region, the material layers being in contact with the word lines and the bit lines, respectively. Figure 17 is a flow chart showing the manufacturing process of an array structure according to another embodiment of the present invention. The components are similar to those given in Figure 16. As shown in Fig. 17, 22 f.doc/006 No, a semiconductor substrate 100 is provided first. The array region is defined by lithography on the surface of the substrate 1 or the dopant is implanted on the surface of the substrate to form an n-type diffusion region 12 to form an isolation trench 122_126 in the substrate 1 to intercept the n-type diffusion. The region 121 is filled with a dielectric material to form a buried diffusion bit line 127_130. After this process, the manufacturing process is the same as the manufacturing process of Fig. 16 above. According to the manufacturing flow of FIG. 16 and FIG. 17, an array structure can be fabricated. The top view of the array structure is as shown in FIG. 18, and the array structure has buried diffusion bit lines 200-202 and polycrystalline characters. Line 2 〇Hong 205, and the two intersect perpendicularly to each other. The memory cells are formed in areas where the two intersect, such as the intersection area 206. Fig. 19 and Fig. 20 are schematic views showing the structure of the sensing circuit of the memory cell array of the present invention. In Fig. 19 and Fig. 2, the array 25 代表 represents a memory cell array which is composed of a plurality of word lines arranged horizontally in the drawing and a plurality of bit lines arranged vertically. The memory cell is a diode symbol to indicate that the diode-like private resistor manufactured by the processes of Figs. 16 to 18 does not need to erase the memory cell. One of the bit lines is connected to the data output line 251 using standard encoding techniques. The data output line 251 is coupled to current mode sense amplifiers 252-254, which are used to measure data stored in a single memory cell as shown in FIG. Each current mode sense amplifier is connected to a reference current source. Thus, current mode sense amplifier 252 is coupled to a reference current source that supplies reference current Ref-L3. The current mode sense amplifier 253 is connected to a reference current source that supplies a reference current Ref_1.2. The current mode sense amplifier 254 is connected to 23 I2736i) 4Q twf.doc/006 to the reference current source supplying the reference current Ref_1. The output of the sense amplifier is then decoded on conductors 255-257 to calculate the value of the two bits stored in the selected memory cell. Figure 20 is a diagram showing the structure of a sensing circuit of another memory cell array of the present invention. Data output line 251 from array 250 is coupled to a single current mode sense amplifier 260. The reference current supplied to the sense amplifier on conductor 265 is selected by switches 261-263 which are respectively coupled to reference current sources supplying reference currents Ref-1.3, Ref-1.2, Ref-l.l. In other embodiments, digital sense amplifiers are not used, and data stored in the memory cells are sent out as analog output. Figure 21 is a block diagram showing the simple composition of a memory element that uses a programmable resistor without erasing the memory cell pREM array 270. The memory elements include a row decoder 271 and a column decoder 272, and a row decoder 271 and a column decoder 272 are connected to the address bus 273. The supply voltage from voltage source 275 for reading and programming operations is supplied to selected memory cells in the array via column decoder 272 and row decoder 271. The sense amplifier and input data structure 276 is coupled to the output of column decoder 272, input data bus 280 and output data bus 281. Each component that is connected to the stylized state device 277 is connected to the e-memory component. The state device can be executed by dedicated logic, a programmable logic array structure, a general purpose processor execution instruction, or a combination of these. As described above, the programmable resistor does not need to erase the memory cell array and can also be used to store multi-bit data in a single memory cell. In other embodiments, such an array of cells is suitable for multiple stylized cycles. Referring to Figures 22 through 24 ltwf.doc/006, Figure 25, a first stylized loop is executed to set a reference current for sensing a single bit at level Ref_1 (as shown in Figure 22). A second stylized loop is executed to set a reference current for sensing a single bit at level Ref_2 (as shown in Figure 23). A third stylized loop is executed to set a reference current for sensing a single bit at level Ref_3 (as shown in Figure 24). A fourth stylized loop is executed to set a reference current for sensing a single bit at level Ref_4 (as shown in Figure 25). The number of memory cells is programmed for a particular execution. This number is determined by the ability to reliably change the number of resistors that are progressively changed, and the level of current that can be clearly distinguished when accessing the memory. The read and stylized status device (shown as reference numeral 277 in Figure 21) is set to find the number of programmed cycles to be executed so that the appropriate reference current can be supplied to the sensing circuit. Figure 26 is a basic stylization rule for the programmable resistor of the present invention without erasing the memory cell. For the first stylized cycle 3, a first current value for sensing and verification is set (block 301). Then, a stress/stylization operation is performed to induce a progressive collapse amount so that the programmed memory cell can generate an output current greater than the first reference current value when reading (block 3 〇 2). In another embodiment, a stress is applied during the staging operation, and a continuous short pulse is used to gradually cause a progressive change in the resistance, such as a wavelength and/or voltage level that may change, or a fixed wavelength and The voltage level makes the memory cell characteristics have a variety of progressive collapse control totals. Then, the assertion and insertion are performed to determine the successful stylization (block 303). If the confirmation fails, then the following stress pulse retry the stress stylization operation. If the validation is successful, the first stylized loop is completed (block 304). As shown in Fig. 26, 25 ltwf.doc/006 can also use the same basic program as long as the progressive reference current is set, the second stylized loop 31〇, the third stylized loop 320, the fourth stylized loop 320, and the like. To execute. Three typical stylized operations are useful in which successive pulses are applied to establish a selected progressive change in memory cell characteristics, and each pulse is executed by a program rule, wherein the retry rule includes ··, ( 1) A confirmation step to measure whether the selected level is reached, if the rush is supplied with equal pulse voltage and equal pulse length (2) in each cycle - a confirmation step to measure whether the selected level is reached, if Otherwise, a pulse having an increased pulse voltage of the pulse length of the pulse is supplied in the mother-in-one cycle. 〃 (3)—Confirmation step to measure whether the selected position is reached. If no, the pulse with the increased pulse length of the pulse voltage is supplied in the parent-continuous cycle. And (4) a confirmation step to measure whether the selected level is reached, and if no, a pulse is supplied in the successive cycle, the pulse being during one or more steps of the continuous successive cycle One or both of the pulse width and the pulse length will change. In a preferred embodiment according to the structure shown in Fig. 2, the stylization program includes a fixed n-type diffusion electrode at a voltage of about -2 volts, and a p-type multi= 矽 electrode stepwise application of about 5 volts to about 2 volts. The voltage is applied at a fixed pulse width (eg, lms or i〇ms) at a voltage of 1 volt at each step, using a confirmation step between each pulse, and when 26 rf.doc/006 memory cells pass Stop after confirming the steps. In another constant voltage operation that is not structured according to Fig. 2, the stylization procedure includes fixing the voltage of the n-type diffusion electrode at about -2 volts, and applying a voltage of about 2 volts to the p-type polysilicon electrode to supply an equal pulse height. And supplying a pulse having an equal pulse width (for example, lms or l〇ms), using a confirmation step between each pulse, and stopping when the memory cell passes the confirmation step. Of course, the pulse height and pulse width can vary depending on the needs of the particular system. The programmable resistor of the present invention does not need to erase the memory cells to exhibit excellent program disturb and read performance. The performance of stylized interference can be explained by referring to Figures 27a to 27D. Figure 27A shows a schematic view of the array portion showing the word lines 400-403 and the bit lines 410-413. The memory cell A in the interleaved region between the word line 401 and the bit line 411 is applied by applying a voltage of 18 volts to the electrode on the word line 401 and applying a voltage of volts to the electrode on the bit line 411. Stylized. The memory cell B in the interleaved region of the word line 4〇1 and the adjacent bit line 412 receives a voltage of 18 volts from the word line, but the bit line is grounded. The memory cell c in the interleaved region between the bit line 411 and the adjacent word line 402 receives a voltage of _15 volts from the bit line, but the word line is grounded. Figure 27B shows the selected memory cell a, wherein the upper electrode 42 is composed of p-type polysilicon, the lower electrode 421 is composed of an n-type buried diffusion (or well) region, and the inter-electrode dielectric layer 422 is composed of 15 It is composed of arsenic oxynitride. A stylized potential of around 3. 3 volts across the two electrodes contributes to a bias mode for the Ρ_η junction of the component, the absolute value of the supply voltage is less than 2 volts, and 27 12736 QAtwf.doc/006 causes a progressive reduction in resistance across the two electrodes, This results in a progressive increase in read current as shown in Figure 6. Figure 27C shows the unselected memory cell C, wherein the upper electrode 423 is composed of P-type polysilicon, the lower electrode 424 is composed of an n-type buried diffusion (or well) region, and the inter-electrode dielectric layer 425 is composed of 15 It is composed of arsenic oxynitride. A voltage of -L5 volts is received from the bit line, but the word line is grounded. The ρ-η junction of the component is in a reverse bias mode. Figure 27D shows the unselected memory cell β, wherein the upper electrode 426 is composed of erbium-type polysilicon, the lower electrode 427 is composed of an n-type buried diffusion (or well) region, and the inter-electrode dielectric layer 428 is composed of 15 It is composed of arsenic oxynitride. A voltage of 1·8 volts is received from the word line, but the bit line is grounded. Figure 28 shows that the bias voltage applied to the memory cell β causes the memory cell to still have a low stylized current after 10,000 seconds of stylized stress. Figure 29 shows the application of a bias voltage such as memory cell c, so that the programmed current of the memory cell after 10,000 seconds of stylized stress is still very low. And 'the stylized interference is not seen from the scale of the graph. Diagrams 30 to 30D are diagrams of the memory cell array read and read interference states of the present invention. The readability of the programmable resistor of the present invention without erasing the memory cell can be explained with reference to Figs. 30 to 30D. Figure 30 shows a schematic diagram of the array portion Illustrating the word line 5〇〇-5〇3 and the bit line 5〇4-5〇7. The memory cell a in the interleaved region between the word line 502 and the bit line 505 is applied by applying a voltage of 1·3 volt to the electrode on the word line 5〇2, and grounding the electrode on the bit line 507. Read. The memory cell Bi in the interleaved region between the word line 5〇2 and the adjacent bit 28 12736· twf.doc/006 line 506 receives a voltage of ι·3 volt from the word line, but the bit line is set in a The suppression voltage is around 13 volts. The memory cell 2 in the interleaved area between the word line 503 and the adjacent bit line 5〇5 receives the voltage of the volts from the word line and receives the ground potential from the bit line. The memory cell C in the interleaved region between the bit line 506 and the adjacent word line 5〇3 receives the ground potential from the word line, and receives a suppression voltage of about 13 volts from the bit line. The 30th panel shows the selected memory cell, wherein the upper electrode 52 is composed of p-type polycrystalline asmatte, and the lower electrode 521 is composed of an n-type buried diffusion (or well) region, and the interelectrode dielectric layer 522 It is composed of 15 angstroms of arsenic oxide. In the memory cell as shown in Fig. 30, the dielectric layer is programmed to a low resistance state. A read potential of about 1/3 volts across the two electrodes contributes to a bias mode for the pn-n junction of the component, while inducing a current that can be sensed in the component. Figure 30C shows the selected memory cell, wherein the upper electrode 523 is composed of p-type polysilicon and the lower electrode 524 is composed of an n-type buried diffusion (or well) region. The interelectrode dielectric layer 525 is 15 angstroms. Nitrous oxide is formed by the evening. In the memory cell as shown in Fig. 30C, the dielectric layer is not programmed to a low resistance state. A read potential of around L3 volts across the two electrodes into a bias mode for the pn-n junction of the component, but induces a current in the component. , ' ^ Figure 30D shows the unselected memory cells in a stylized state, where the upper electrode 527 is composed of p-type polycrystalline shi, and the lower electrode seems to be composed of an η-type buried diffusion (or well) region. The inter-electrode dielectric layer 529 is composed of 29 Ι27360β^15 Å of arsenic oxynitride. The lower electrode receives a suppression bias of about 3 volts, and the upper electrode ground ◊ is in the reverse bias mode for the ρ-η junction of the element, and there is substantially no current. Similarly, as in the 30th, neither the memory cell 2 nor the memory cell C generate current, whether it is, a programmed low resistance state or an unprogrammed high resistance state. Figure 31 is a plot of read current vs. gate voltage, where mark line 550 represents a stylized memory cell and line 551 represents an unprogrammed memory cell ("unused"). As shown, when the memory cell is reverse biased (Vg < 〇), there is substantially no current. Before the bias state, Vg is less than one tenth of a volt, with a small amount of current in both the private and unprogrammed memory cells. However, in this embodiment, Vg is about 13 volts (line 552), and the programmed singular cells show a higher current than the unprogrammed memory cells. Fig. 32 shows the durability of the memory cell of the present invention. For stylized memory cells (markers 56〇 and 561) and unprogrammed memory cells (marker line 562) at different stylized levels, the read current remains close to a fixed value for long read times. . Moreover, as shown in Fig. 33, the data retention characteristics are also good. In the case of long-term warm baking, the read current remains close to a fixed value for the stylized memory cells (markers 565 and 566) and unprogrammed memory cells (marker 567) with different stylized levels. . Because the memory cell of the present invention has good stability, durability and retention, the memory cell can be programmed multiple times to be applied to a memory cell array of a single memory cell. Figure 34 is the diagram of the read current vs. stylized time in Figure 6, plus the data for the single memory cell binary storage I2736IAQ itwf.doc/006 ^ quasi-start value and reference current, ^ can also support The reference current values of the embodiment of the f-stylized, single-memory multi-bits of the present invention are not considered. In the first cycle, it is not necessary to apply any stylized voltage (stress) in order to store the value equal to "00". In order to store the value of the material equal to οι", it is necessary to apply a low voltage stylized voltage (stress) for about 25 seconds. The storage is equal to the "1G" (four) material value and the low voltage programming voltage (stress) needs to be applied for about 35 seconds. In order to store the data value equal to "u", it is necessary to apply a low voltage stylized voltage (stress (4) seconds or so. The reference current source is set for the purpose of sensing the data value. The reference current is not determined in this example. 1 〇 microamperes. The reference electric current, flow Ref_12 is 22 microamperes in the present embodiment. The reference current RefL1.3 is set to 35 microamperes in this embodiment. By comparing the read money with the reference f current, The data value can be detected. For the second stylized cycle, the data value can be detected by comparing the read current with the reference currents Ref_2j, Ref-2; 2, Ref-2 3. For the third stylization The loop can detect the data value by comparing the read current with the reference currents Ref-3·:, Ref-3.2, and Ref_3.3. Figures 35 and 36 show the programmable resistor of the present invention. There is no need to erase the embodiment of the "SyStein On a Chip (SOC)". The process of programmable resistors without the need to erase the memory cell (PREM) is completely compatible with the standard complementary metal oxide semiconductor (CM). 〇s) process compatibility, in the embodiment only need to add a mask, so It is suitable for the production of tight non-volatile memory in SOC products. For the p-type polycrystalline upper electrode 31 Ι 2736βα twf.doc/006 and the η+ type buried diffusion lower electrode, the additional H-structure only needs to be increased by one. , Beb (four) cover 'for the η + type buried step, while other STI process, dielectric sound expansion (four) of the implant stone (four) accounted for 0 u ▲ Dianguang chemical formation process and P + type polycrystalline The formation process of the CMOS structure is achieved by the Xiacheng system. The low voltage operation makes this kind of discussion low; ^ Bu == A good choice. A low power %楗 to the integrated circuit drawn in Figure 35 Array 601, logic circuit 6〇3, such as dedicated or known gate array logic circuit, and static random access memory ^M) 602 4REM array can be used to store more & For example, the information can be (4) gate _ _ _ _ (four) program specifications. Figure 36, which can be used to store logic circuitry's operations, is an embodiment of another system wafer of the present invention. The integrated circuit drawn in Fig. 36 includes programmable residuals, with (pREM) array 7 (U, logic circuit 7G3 such as dedicated logic circuit or programmable gate array logic circuit, static random access memory (SRAM) And the general purpose processor 704. The PREM array 701 can be used to store the deduction program executed by the general purpose processor 7〇4, and is stored by the external controller in the SRAM7〇2 or from the PREM array 701 to the SRAM 7〇2. The instructions can be supplied to the processor and executed by the processor to control the stylization of the pREM array 7.1. In summary, the present invention provides a program called pREM (Programmable Resistor With Erase-less 32) 1273· twf.doc/006
Memory))的非揮發性記憶體,其電極間介電層隨著應力 而改變的特性是電阻。在一實施例中,PREM記憶胞具有 P+型多晶石夕閘極、超薄氧化層與n+型擴散區,其中使用 超薄氧化層的累進崩潰作為資料儲存的特徵。pREM元件 可以達到與CMOS製程完全相容、多次程式化(multi-time programming)、多階記憶胞(multi-level 11}盥 壓 ,等特點。邏輯電路、S議與非揮== =易的整合在SOC的應用上。記憶胞資料保存效能也非 ㊉良好。,而且’記憶胞並沒有檢_程式化干擾現象 取干擾現象。 雖然本發明已以較佳實施例揭露如上,然其 二任何熟習此技藝者’在不脫離本發明之精神 ,备可作些許之更動與潤飾,因此本發^月之保護 乾圍§視後附之申請專利範圍所界定者 【圖式簡單說明】 記憶本發明之含有累進崩潰介電層的可程式電限 imr二極體可程式電-記==含 極的 4 層的類二極艘可程式電阻記憶胞有累進崩潰介電 第4圖是deGraaf等人所提岣知介電層的災難性 33 >c/006 崩潰行為示意圖。 第5圖是第2圖的累進崩潰記憶胞之程式化電流對程 式化時間關係圖。 第6圖疋第2圖的累進崩潰記憶胞之讀取電流對程式 化時間關係圖。 第6A圖是第2圖的累進崩潰記憶胞具有四種不同程 式化,加電壓量值之讀取電流對程式化時間關係圖。 第7圖是本發明之適用於測量程式化在記憶胞内之資 料的參考電流值Ref-U關係圖。 一 =8圖疋本發明之適用於測量程式化在記憶胞内之二 位元,料的參考電流值關係圖。 、,第9圖是本發明之適用於測量程式化在記憶胞内之資 料的15參考電流值關係圖。 · 第圖是本發明之記憶胞陣列的基本陣列結構圖。 第11圖是本發明之第2圖所示記憶胞的陣列之基本 陣列結構圖。 第12圖是本發明之第3圖所示記憶胞的陣列之基本 陣列結構圖。 第13圖是本發明之記憶胞的基本製造流程圖。 第14圖是本發明之第2圖所示記憶胞的基本製造流 程圖。 第丨5圖是本發明之第3圖所示記憶胞的基本製造流 程圖。 第16圖是本發明之第11圖所示陣列結構的基本製造 34 ltwf.doc/006 流程圖。 一種基 第17圖是本發明之第u胃所示陣列結 本製造流程圖。 第17圖所示製造 流程所製造出來之陣列結構之上視圖。 i 第19圖是本發明的陵万|恭μMemory)) A non-volatile memory whose electrical properties change between stresses due to stress. In one embodiment, the PREM memory cell has a P+ type polycrystalline slab gate, an ultrathin oxide layer, and an n+ type diffusion region, wherein progressive collapse using an ultrathin oxide layer is characteristic of data storage. pREM components can be fully compatible with CMOS processes, multi-time programming, multi-level memory (multi-level 11), etc. Logic circuits, S-discussion and non-swing == = The integration of the SOC is not good. The memory cell data storage performance is not good. And the memory cell is not detected. The stylized interference phenomenon takes the interference phenomenon. Although the present invention has been disclosed in the preferred embodiment, the second Anyone skilled in the art will be able to make some changes and refinements without departing from the spirit of the present invention. Therefore, the protection of the hair of the month is defined by the scope of the patent application attached [simplified description] Memory The programmable voltage limit imr diode of the progressive collapse dielectric layer of the present invention can be programmed to be electrically-remembered == the 4-layered diode of the pole-like resistor has a progressive crash dielectric. Figure 4 is deGraaf The catastrophic 33 >c/006 crash behavior diagram of the dielectric layer is shown in Figure 5. Figure 5 is a stylized current vs. stylized time diagram of the progressive crash memory cell in Figure 2. Figure 6 2 picture of progressive crash memory cell reading The flow-to-stylized time relationship diagram. Figure 6A is a diagram showing the four types of progressive crash memory cells in Fig. 2 with four different stylized, voltage-valued read current versus stylized time relationships. Figure 7 is a diagram of the application of the present invention. A reference current value Ref-U relationship diagram for measuring data stored in a memory cell. One = 8 Figure 疋 The present invention is applicable to measuring a reference current value relationship diagram of a two-bit material in a memory cell. Figure 9 is a diagram showing the relationship between the reference current values of the present invention for measuring the data stored in the memory cells. The first diagram is a basic array structure diagram of the memory cell array of the present invention. Fig. 12 is a basic array structure diagram of an array of memory cells shown in Fig. 2. Fig. 13 is a basic array structure of an array of memory cells of the present invention. Fig. 14 is a basic manufacturing flow chart of the memory cell shown in Fig. 2 of the present invention. Fig. 5 is a basic manufacturing flow chart of the memory cell shown in Fig. 3 of the present invention. Fig. 16 is a flowchart of the present invention. The basic structure of the array structure shown in Figure 11 34 ltwf.doc/006 Flowchart. A base Figure 17 is a manufacturing flow chart of the array shown in the first embodiment of the present invention. Figure 17 is a top view of the array structure manufactured by the manufacturing process. 19 is the Ling Wan of the present invention|Cong μ
圖疋本發明的驅動記憶體之積體電路 第18圖是根據本發明之第16圖、第17圖 第20圖 意圖。 的組成結 第21圖 第22圖是本發明的多次程式化實施例在第一次程式 化後用於感測一位元資料的參考電流位準關係圖。> 第23、圖是本發明的多次程式化實施例在第二次程式 化後用於感測一位元資料的參考電流位準關係圖。 第24圖是本發明的多次程式化實施例在第三次程式 化後用於感測一位元資料的參考電流位準關係圖。 第25圖是本發明的多次程式化實施例在第四次程式 化後用於感測一位元資料的參考電流位準關係圖。 第26圖是本發明的多次程式化模式圖。 第27A圖至第27D圖是本發明之記憶胞陣列程式化 與程式化干擾狀態圖。 第28圖是第27A圖至第27D圖中之第27C圖所示 之未選定記憶胞之程式化干擾狀態圖。 第29圖是第27A圖至第27D圖中之第27D圖所示 35 itwf.doc/006 之未選定記憶胞之程式化干擾狀態圖。 第30A圖至第30D圖是本發明之記憶胞陣列讀取與 讀取干擾狀態圖。BRIEF DESCRIPTION OF THE DRAWINGS Fig. 18 is a view showing the sixteenth, seventeenth and twentyth aspects of the present invention in accordance with the present invention. Composition Figure 21 Figure 22 is a diagram of a reference current level relationship for sensing a bit of metadata after the first stylization of the multi-stylized embodiment of the present invention. > Fig. 23 is a diagram showing a reference current level relationship for sensing a bit of metadata after the second stylization of the multiple stylized embodiment of the present invention. Figure 24 is a diagram of a reference current level relationship for sensing a bit of metadata after a third stylization of the multiple stylized embodiment of the present invention. Figure 25 is a diagram showing the reference current level relationship for sensing a bit of metadata after the fourth stylization of the multiple stylized embodiment of the present invention. Figure 26 is a diagram of a plurality of stylized patterns of the present invention. Figures 27A through 27D are diagrams showing the stylized and stylized interference states of the memory cell array of the present invention. Fig. 28 is a diagram showing the stylized interference state of unselected memory cells shown in Fig. 27C of Figs. 27A to 27D. Fig. 29 is a diagram showing the stylized interference state of unselected memory cells of 35 itwf.doc/006 shown in Fig. 27D to Fig. 27D of Fig. 27A to Fig. 27D. Fig. 30A to Fig. 30D are diagrams showing the state of reading and reading interference of the memory cell array of the present invention.
第31圖是選擇已程式化記憶胞與第30B圖、第30C 圖所示之選擇未程式化記憶胞之讀取電流對閘極電壓關係 圖。 >Figure 31 is a graph showing the relationship between the read current and the gate voltage of the selected unprogrammed memory cell in the selected memory cell and the 30B and 30C. >
第32圖是選擇已程式化記憶胞與第30B圖、第30C 圖所示乏選擇未程式化記憶胞之讀取電流對讀取時間關係 圖。Figure 32 is a graph showing the read current versus read time for the selected memory cells and the unselected unprogrammed memory cells shown in Figure 30B and Figure 30C.
第33圖是選擇已程式化記憶胞與第30B圖、第3〇C 圖所示之選擇未程式化記憶胞之讀取電流對保持時間關係 圖。 第34圖是本發明之多次程式化、單一記憶胞多位元 之實施例的參考電流值示意圖。 第35圖是本發明之具有可程式電阻不需抹除記憶胞 陣列、專用邏輯電路、靜態隨機存取記憶體的記體電路之 方框示意圖。 第36圖是本發明之具有可程式電阻不需抹除記憶胞 陣列、通用處理器、專用邏輯電路、靜態隨機存取記^體 的記體電路之方框示意圖。 ^ 【圖式標示說明】 10、12、40 ·•導體 1 卜 14、17、51、58、63 :介電層 13、16:第一電極 36 Ι2736〇α twf.doc/006 15、18 :第二電極 30、31、32、33、100、1(U、102、103、104、105、 122、123、124、125、126 :隔離溝渠 34、35、36、46、48、50、55、60 :下電極導體 37、38、39 :超薄氧化層 45、47、52、59、64 :上電極導體 > 56、 61、106、120 :摻質 57、 62、100 :基底 107、108、119、110 :埋入式擴散區 111、112、113、114 :超薄介電層 115 :多晶矽 121 :擴散區 、 127、128、129、130、200、201、202 :埋入式擴散 位元線 203、204、205 :多晶矽字元線 206 :相交區域 250、270、601、701 : PREM 陣列 251 :資料輸出線 252、253、254、260 :感測放大器 255、256、257、265 :導線 261、263、263 :開關 271 :行譯碼器 272 :列譯碼器 273 :位址匯流排 37 12736顯 twf.doc/006 275 :電壓源 276 :感測放大器與輸入資料結構 277 :讀取與程式化狀態裝置 280 :輸入資料匯流排 281 :輸出資料匯流排 300、301、302、303、304、310、320、330 方塊 400、401、402、403、500、5(U、502、503 :字元線 410、411、412、413、504、505、506、507 :位元線 420、 423、426、520、523、527 ··上電極 421、 424、427、521、524、528 :下電極 422、 425、428、522、525、529 :電極間介電層 550、551、560、56卜 562、565、566、567 :標記線 552 :線 ·.Figure 33 is a graph showing the read current versus hold time for selecting a non-stylized memory cell as shown in Figure 30B and Figure 3C. Figure 34 is a schematic diagram of reference current values for an embodiment of the multi-programmed, single memory cell multi-bit of the present invention. Fig. 35 is a block diagram showing the recording circuit of the present invention having a programmable resistor without erasing a memory cell array, a dedicated logic circuit, and a static random access memory. Figure 36 is a block diagram showing the recording circuit of the present invention having a programmable resistor without erasing the memory cell array, a general purpose processor, a dedicated logic circuit, and a static random access memory. ^ [Illustration description] 10, 12, 40 ·• Conductor 1 Bu 14, 17, 51, 58, 63: Dielectric layer 13, 16: First electrode 36 Ι 2736 〇 α twf. doc / 006 15, 18: Second electrode 30, 31, 32, 33, 100, 1 (U, 102, 103, 104, 105, 122, 123, 124, 125, 126: isolation trenches 34, 35, 36, 46, 48, 50, 55 60: lower electrode conductors 37, 38, 39: ultra-thin oxide layers 45, 47, 52, 59, 64: upper electrode conductors > 56, 61, 106, 120: dopants 57, 62, 100: substrate 107, 108, 119, 110: buried diffusion regions 111, 112, 113, 114: ultra-thin dielectric layer 115: polysilicon 121: diffusion region, 127, 128, 129, 130, 200, 201, 202: buried diffusion Bit lines 203, 204, 205: polycrystalline word line 206: intersecting areas 250, 270, 601, 701: PREM array 251: data output lines 252, 253, 254, 260: sense amplifiers 255, 256, 257, 265 : Wires 261, 263, 263: Switch 271: Row Decoder 272: Column Decoder 273: Address Bus 37 37736 twf.doc/006 275: Voltage Source 276: Sense Amplifier and Input Data Structure 277: Read and stylized state loading 280: input data bus 281: output data bus 300, 301, 302, 303, 304, 310, 320, 330 blocks 400, 401, 402, 403, 500, 5 (U, 502, 503: word line 410 411, 412, 413, 504, 505, 506, 507: bit lines 420, 423, 426, 520, 523, 527 · upper electrodes 421, 424, 427, 521, 524, 528: lower electrodes 422, 425 428, 522, 525, 529: inter-electrode dielectric layers 550, 551, 560, 56 562, 565, 566, 567: marking line 552: line ·.
602、 702 : SRAM 603、 703 :邏輯電路 704:處理器 A、B、B卜B2、C :記憶胞 38602, 702: SRAM 603, 703: logic circuit 704: processor A, B, B, B2, C: memory cell 38
Claims (1)
Applications Claiming Priority (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US56070703P | 2003-07-21 | 2003-07-21 | |
US10/641,897 US7132350B2 (en) | 2003-07-21 | 2003-08-15 | Method for manufacturing a programmable eraseless memory |
US10/642,249 US7180123B2 (en) | 2003-07-21 | 2003-08-15 | Method for programming programmable eraseless memory |
US10/641,846 US20050035429A1 (en) | 2003-08-15 | 2003-08-15 | Programmable eraseless memory |
Publications (2)
Publication Number | Publication Date |
---|---|
TW200516591A TW200516591A (en) | 2005-05-16 |
TWI273600B true TWI273600B (en) | 2007-02-11 |
Family
ID=38621591
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
TW93121700A TWI273600B (en) | 2003-07-21 | 2004-07-21 | Integrated circuit and manufacturing method thereof, memory cell and manufacturing method thereof, method for programming memory cell and method for programming memory array multiple times |
Country Status (1)
Country | Link |
---|---|
TW (1) | TWI273600B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10991433B2 (en) * | 2019-09-03 | 2021-04-27 | Silicon Storage Technology, Inc. | Method of improving read current stability in analog non-volatile memory by limiting time gap between erase and program |
-
2004
- 2004-07-21 TW TW93121700A patent/TWI273600B/en active
Also Published As
Publication number | Publication date |
---|---|
TW200516591A (en) | 2005-05-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US7132350B2 (en) | Method for manufacturing a programmable eraseless memory | |
JP5042233B2 (en) | Deposited semiconductor structure and fabrication method for minimizing n-type dopant diffusion | |
TW475267B (en) | Semiconductor memory | |
US9343673B2 (en) | Method for forming metal oxides and silicides in a memory device | |
KR101196052B1 (en) | Memory devices, memory device constructions, constructions, memory device forming methods, current conducting devices, and memory cell programming methods | |
KR101128246B1 (en) | Non-volatile programmable memory | |
TWI309083B (en) | Memory cell comprising switchable semiconductor memory element with trimmable resistance | |
US6836423B2 (en) | Single level metal memory cell using chalcogenide cladding | |
US6919578B2 (en) | Utilizing atomic layer deposition for programmable device | |
TW465026B (en) | Semiconductor device, method of manufacturing semiconductor device and communication method | |
US6984548B2 (en) | Method of making a nonvolatile memory programmable by a heat induced chemical reaction | |
JP2013505581A (en) | Three-dimensional polysilicon diode having low contact resistance and method for forming the same | |
US20030206481A1 (en) | Trench isolated cross-point memory array and method of fabrication | |
US20050190601A1 (en) | Programmable resistor eraseless memory | |
JP2003303941A (en) | Self-aligned programmable phase change memory | |
US8320159B2 (en) | Resistance variable nonvolatile memory device | |
TW200931413A (en) | Method of programming cross-point diode memory array | |
US7777268B2 (en) | Dual-gate device | |
KR101735187B1 (en) | Variable resistor, non-volatile memory device using the same, and method of fabricating thereof | |
CN113257850A (en) | Resistive memory cell and related array structure | |
US7180123B2 (en) | Method for programming programmable eraseless memory | |
TW201216455A (en) | Nonvolatile variable resistance element and method of manufacturing the nonvolatile variable resistance element | |
TWI273600B (en) | Integrated circuit and manufacturing method thereof, memory cell and manufacturing method thereof, method for programming memory cell and method for programming memory array multiple times | |
JP5015420B2 (en) | Programming method for programmable erasure-free memory | |
TW200805637A (en) | Nonvolatile memory array having modified channel region interface |