TW200528739A

TW200528739A - Flux guides for magnetic field sensors and memories

Info

Publication number: TW200528739A
Application number: TW093139703A
Authority: TW
Inventors: Lee Kim Phan
Original assignee: Koninkl Philips Electronics Nv
Priority date: 2003-12-23
Filing date: 2004-12-20
Publication date: 2005-09-01
Also published as: WO2005064357A3; WO2005064357A2

Abstract

A magnetic sensor has a magneto resistive sensing element 210 having layers of magnetic material and one or more flux guides 120 for concentrating the field onto the sensing element, the flux guide comprising a part of one or more of the same layers used for the sensing element. By using the same layer or layers for the flux guide the guide layer can be formed in the same step as the corresponding layer of the sensing element. Such sensors can be integrated in MRAM chips. Flux guides can be used to rotate the field differently for each of 2 parallel sensing elements, to enable 2D field sensors. As there is now no need for orthogonal exchange bias directions for the two sensors, they can be integrated more easily. Flux guides can also be used to concentrate the field for writing MRAM cells, and hence reduce write current.

Description

200528739 九、發明說明：【發明所屬之技術領域】本發明相關於偵測磁場的感測器，磁阻電流感測器，具此類感測器的積體電路，製造此類感測器（或此類積體電路）的方法’使用此類感測器的方法，尤其相關於MRAM(磁性隨機存取記憶體）。【先前技術】磁場感測器己廣泛使用於電子羅盤、讀取磁頭、磁場測量、金屬偵測、汽車引擎管理及更多應用中。現有許多不同類型的磁場感測器，其中磁阻（MR)感測器在任何需要小型的應用（諸如讀取磁頭或積體電路等）中較受歡迎。根據不同效應有數種類型MR感測器，諸如各向異性磁阻（amr)、巨型磁阻（GMR)、隧穿磁阻（TMR)等。根據AMR效應的感測器已用於讀取磁頭數年之久。由於它們結構簡單及價格低廉，目前仍使用在多種應用的較小型磁性感測器中。amr 感測器具有一層各向異性磁性材料，此層的電阻並受到一外界磁場的影響，其可藉由傳送一測。GMR(巨型磁阻）感測器具有一送一電流經過此層而加以感200528739 IX. Description of the invention: [Technical field to which the invention belongs] The present invention relates to a sensor for detecting a magnetic field, a magnetoresistive current sensor, an integrated circuit having such a sensor, and manufacturing such a sensor ( Or such integrated circuits) The method of using such sensors is particularly relevant to MRAM (magnetic random access memory). [Previous technology] Magnetic field sensors have been widely used in electronic compasses, reading heads, magnetic field measurement, metal detection, automotive engine management, and more applications. There are many different types of magnetic field sensors, and magnetoresistive (MR) sensors are popular in any application that requires a small size, such as reading heads or integrated circuits. There are several types of MR sensors based on different effects, such as anisotropic magnetoresistance (amr), giant magnetoresistance (GMR), tunneling magnetoresistance (TMR), and so on. Sensors based on the AMR effect have been used to read heads for years. Due to their simple structure and low cost, they are still used in smaller magnetic sensors for many applications. The amr sensor has a layer of anisotropic magnetic material. The resistance of this layer is also affected by an external magnetic field, which can be measured by transmission. GMR (Giant Magnetoresistive) sensors have

98455.doc 200528739 隧穿障壁所取代。藉由傳送一電流經過該堆疊（在垂直於感測器水平面的方向中）而感測MR效應。目前可達成TMR效應咼達50%以上。TMR感測器係硬碟機中未來高密度讀取磁頭的最具潛力候選者。依照類型及構造，MR感測器在該感測器水平面中的一方向較靈敏，及在另一方向較不靈敏。根據Biot-Savart定律，在電流與磁場之間有一關係。在一直線導體中流經的電流在其本身周圍造成一圓形磁場。因此放置在該電流附近的磁性感測器可作為一電流感測器使用。電流感測器可應用在許多領域中，諸如IC測試等。要求必須測試及監控一 1C晶片内多種不同位置的電流。習用的1C内建式電流感測器有多種不同類型。美國專利號 5,963,038中揭示一些類型，該文獻說明藉由位於積體電路中一導體附近的磁場感測器而測量流經該導體的電流，以偵測該積體電路中的故障。該感測器可以多種不同方式建構，以便測量流經該導體的電流所產生的磁場。所揭示的範例包括一撿拾線圈感測器、霍爾（Hall)感測器、MR(磁阻）感測器及一GMR(巨型磁阻）感測器。這使得無法由外部測試設備輕易存取的導體的測試成為可能，及用於備測數個並聯路徑中個別路徑中的故障，該等並聯路徑即若僅傳導一路徑亦會通過一電阻測試。 MR感測器具有一電阻，其依經過該感測器水平面的外界磁場而定。流經該導體的電流在該導體周圍造成一圓形磁場，經過該感測器的水平面，並在該感測器水平面中垂直於該導體。該MR感測杰在此方向是靈敏的，因此沿著=感 98455.doc 200528739 、ιί印的X平面（平行於该導體）測量其電阻，以測量流經該導月旦的私*所產生的磁場強度。可使用傳統的難感測器或 GMR(巨型磁阻）感測器。若使用相同的偏塵，g難感測器較MR感測器靈敏，並能較可靠地以較少耗電測量流經該; 體的電流。接著可在具伯測電路的積體電路内部或具有合適測里配置的積體電路外部，測量該驗感測器的電阻。此類感測器用於感測高電流非常有用，但用於靜止mD =流（IDDQ)測試等應用卻不夠靈敏。此測試技術對傳統故障模型未極妥善設計(或習用邏輯測試無法測試)的閘氧化物層短路/字動閘及橋接故障等物理缺㉟已顯示極佳涵蓋範圍。對高品質及成本效益的需求，已使IDD_試廣泛I 用為電壓測試的補充測試。IDDQ測試與其他測試技術併用時，具有免除燒入（burn-in)測試需求的潛力。惟，m〇sfet 漏電正在各科技節點迅速增加，縮短故障與無故障電路的 IDDQ位準間的差異。涉及關閉電力供應及在一監控點取樣電壓中衰減的一 IDDQ測試方法如歐洲專利申請號Ep 〇 84〇 227所揭示。如世界專利號％〇 97/18481所揭示的另一技術涉及提供一主動輸出負載，其可加以開啟以測試該裝置。最終形成的電流可離開該晶片而加以偵測。98455.doc 200528739 Replaced by a tunnel barrier. The MR effect is sensed by transmitting a current through the stack (in a direction perpendicular to the horizontal plane of the sensor). At present, the TMR effect can reach more than 50%. TMR sensors are the most potential candidates for future high-density read heads in hard drives. Depending on the type and configuration, the MR sensor is more sensitive in one direction in the sensor's horizontal plane and less sensitive in the other direction. According to Biot-Savart's law, there is a relationship between current and magnetic fields. The current flowing in a linear conductor creates a circular magnetic field around itself. Therefore, a magnetic sensor placed near the current can be used as a current sensor. Current sensors can be used in many fields, such as IC testing. It is required to test and monitor the currents at many different locations in a 1C chip. There are many different types of conventional 1C built-in current sensors. Some types are disclosed in U.S. Patent No. 5,963,038, which describes detecting a fault in an integrated circuit by measuring a current flowing through the conductor with a magnetic field sensor located near the conductor in the integrated circuit. The sensor can be constructed in a number of different ways to measure the magnetic field generated by the current flowing through the conductor. The disclosed examples include a pick-up coil sensor, a Hall sensor, an MR (magnetoresistive) sensor, and a GMR (giant magnetoresistive) sensor. This makes it possible to test conductors that cannot be easily accessed by external test equipment, and to prepare for faults in individual paths among several parallel paths that will pass a resistance test if only one path is conducted. The MR sensor has a resistance, which depends on the external magnetic field passing through the horizontal plane of the sensor. The current flowing through the conductor creates a circular magnetic field around the conductor, passes through the horizontal plane of the sensor, and is perpendicular to the conductor in the horizontal plane of the sensor. The MR sensor is sensitive in this direction, so its resistance is measured along the X-plane (parallel to the conductor) of Sense 98455.doc 200528739, to measure the current generated by the private * Magnetic field strength. Either a traditional difficult sensor or a GMR (giant magnetoresistive) sensor can be used. If the same partial dust is used, the g-hard sensor is more sensitive than the MR sensor and can more reliably measure the current flowing through the body with less power consumption. The resistance of the test sensor can then be measured inside the integrated circuit with the primary test circuit or outside the integrated circuit with a suitable test configuration. This type of sensor is very useful for sensing high currents, but not sensitive enough for applications such as stationary mD = current (IDDQ) testing. This test technology has shown excellent coverage for physical defects such as gate oxide layer short-circuits / word-operated gates and bridge faults that are not very well designed with traditional fault models (or cannot be tested with conventional logic tests). The demand for high quality and cost-effectiveness has made IDD_test widely used as a supplementary test for voltage testing. IDDQ testing has the potential to eliminate the need for burn-in testing when used in conjunction with other testing technologies. However, mOsfet leakage is rapidly increasing in various technology nodes, reducing the difference between IDDQ levels of faulty and non-faulty circuits. An IDDQ test method involving shutting down the power supply and attenuating the voltage sampled at a monitoring point is disclosed in European Patent Application No. Ep 0 84 227. Another technique, as disclosed in World Patent No. 97/18481, involves providing an active output load that can be turned on to test the device. The resulting current can be detected from the wafer.

Walker 等人的 ”A Practical Built_in Current Sens〇r f0I· IDDQ Testmg(IDDQ測試的實用内建式電流感測為）（ITC20(H ’報告14.3)揭示IDDQ測試相關的一些最近工作。該文獻總結BICS的需求如下： -解析度小於2 μΑ ; 98455.doc 200528739 -尺寸小於1000個電晶體； -測试時間少於1微秒/向量，· -未造成效能劣化； -測量電流位準及方向； -經由掃描鏈控制及數位讀出； -平行運算； -自行校準；使用中為低功率耗散，閒置時無功率耗散。為旨试符合此等需求，Walker使用根據霍爾效應的 MagFET裝置。為克服低靈敏度及高雜訊位準等難題，因此使用濾波、放大及重複測試。惟，此等技術的缺點係佔空間及耗電，對於許多應用（諸如行動裝置等）空間及電力是稀罕常覺不足的資源。用於磁性感測器應用（諸如硬碟的讀取磁頭），習知係使用通量導引（或通量集中器）將多磁性通量集中到該感測器的感測層中，藉此使感測器更靈敏。圖2中顯示具有數個通量導引120的一感測器範例。由於該等通量導引，磁性通量更集中在感測元件210的感測層上。該通量導引通常由一高導磁率材料的薄膜所製成，諸如（定圖案成合適形狀的）鎳鐵合金（permalloy)。該通量導引層的厚度通常有5至數十奈米’與該感測器約以數十奈米的距離加以隔離。另一磁阻材料的應用為MRAM。最近幾年間，已熱烈深入磁性RAM(MRAM)的研究。磁性材料與CMOS科技的整合已較無問題。MRAM商品計晝在2004至2005年推出。K.-M.Η 98455.doc 200528739Walker et al. "A Practical Built_in Current Sensrf0I · IDDQ Testmg" (ITC20 (H 'Report 14.3) reveals some recent work related to IDDQ testing. This document summarizes BICS The requirements are as follows:-the resolution is less than 2 μA; 98455.doc 200528739-the size is less than 1000 transistors;-the test time is less than 1 microsecond / vector, ·-no performance degradation is caused;-the measurement current level and direction; -Controlled by scanning chain and digital readout;-Parallel operation;-Self-calibration; Low power dissipation during use, no power dissipation when idle. To meet these requirements, Walker uses MagFET devices based on Hall effect In order to overcome the problems of low sensitivity and high noise level, filtering, amplification and repeated testing are used. However, the disadvantages of these technologies are space and power consumption. For many applications (such as mobile devices), space and power are Rare and often inadequate resources. For magnetic sensor applications (such as read heads on hard disks), the conventional system uses flux guidance (or flux concentrators) to collect multiple magnetic fluxes. Into the sensing layer of the sensor, thereby making the sensor more sensitive. An example of a sensor with several flux guides 120 is shown in Figure 2. Due to the flux guidance, magnetic flux The flux is more concentrated on the sensing layer of the sensing element 210. The flux guide is usually made of a thin film of a high permeability material, such as (permalloy) (patterned into a suitable shape). The flux The thickness of the guide layer is usually 5 to several tens of nanometers' and the sensor is isolated at a distance of about several tens of nanometers. Another application of magnetoresistive materials is MRAM. In recent years, magnetic RAM has been warmly penetrated ( MRAM) research. The integration of magnetic materials and CMOS technology has been relatively easy. MRAM products were launched from 2004 to 2005. K.-M.Η 98455.doc 200528739

Lenssen等人在2000年非揮發性記憶體科技研討會（2000年 11月11至15曰於美國維吉尼亞州阿靈頓市）中提出MRAM科技的調查，名稱為"Expectation of MRAM in comparison(在比較中對MRAM的期許广。此文獻說明第一代磁性隨機存取記憶體（MRAM)係根據AMR。在1988年之後，稱為巨型磁阻（GMR)的較大磁阻效應的發現能使具較高電阻及較大 MR效應（5至15%)的較小元件實現，並因此得到較高的輸出信號。原則上，此能使MRAM用於一般的應用中。發現GMR 效應的十年後，GMR效應已應用在如HDD讀取頭及磁性感測器等商業產品中。當大的隧穿磁阻（TMR)效應展示在室溫時，在1995年左右磁性隧穿接面領域的突破甚至進一步改善MRAM的榮景。由於當時已顯示TMR效應具有高達50%以上的振幅，但由於強烈的偏壓依存性，在實際應用中可用的電阻變動目前在35%左右。 MTJ記憶單元：圖1中以一範例說明一習知MRAM的一部分，該MRAM顯示使用此類TMR科技的一積體記憶單元陣列。此結構及如何製造它為眾所熟知，在此不需再贅述。總言之，此一 TMR 型MRAM包含數個單元，其係磁性隧穿接面（MTJ)。MTJ基本上包含一自由磁層100、一絕緣層（隧穿障壁102)、一扣住磁層104，及一反鐵磁AF層106，其用以將該扣住層的磁化 ”釘在’’ 一固定方向（其通稱為交換偏壓方向）。此圖所示範例中，亦有一底層108。為求簡化，圖1所示磁性隧穿接面向 98455.doc 200528739 (MTJ)堆豐中僅顯示4個主動層。實際上可有更多層，其與該操作原理並不相關。 a μ 该寻MRAM單元在該自由磁層的磁化方向中儲存資訊 (1/0)，該磁層可在兩相反方向之間較自由旋轉。若該自由層的磁化方向與該扣住層者平行，則該㈣的電阻^，當該兩磁化方向逆平行時，則該MTJ的電阻大。用於在一: 定早兀上讀取資訊，傳送一小電流(垂直地)經過該選取單元的MTJ堆$。在5亥MTJ上測得的電壓降（與電阻成正比）係表示該單it的資訊。藉由傳送寫人電流經過數位線Du_3及^ 元線BL1-3(該等線定圖案在言亥等記憶單元的&部及上方）’可在一寫入運算期間變動一單元上的資m。電流將增加該記憶單元中的磁場（易磁化軸場及難磁化軸場）。將該^ 、昜耘式化，俾使其夠大能將該選取單元的自由層的磁化切換至-新方向。該等位元線與該等單元的難磁化轴平行 (/、在易兹化軸中產生一磁場），而該等數位線（有時稱為字線）反而在難磁化軸中產生_磁場。在—些設計中，該等關係可反過來，即位元線產生難磁化軸場，及數位線產生易磁化軸場。顯不位元線與數位線為正交，磁性隧穿接面放置在交 ***的圖中顯示難磁化軸場（由數位線產生）及易磁化軸 (位元線產生）。隶終形成的磁場與易磁化轴成4 5。，其月匕紋轉该選取單元的自由層的磁化，而不影響所有未選取的單凡。該等單元的底電極連接至具通孔的選取電晶體，其在讀取時會使用到。 9S455.doc -10- 200528739 由於最終形成的磁場與該單元的自由層的易磁化軸成 45。，因此該自由層的切換磁場最小，藉此可以最小電流完成寫入。最終形成磁場在該交點的大小為（|Hha| + |Hea|)/2，其中Hha及HEA分別為難磁化軸及易磁化軸所產生的磁場。它們通常必須具有相同大小。更多有關此類MRAM的資訊，讀者例如可參考 Ρ· K· Naji、M. Durlam、S Tehrani、J. Calder 及 M. F. DeHerrera 的 ”256 kb 3.0 V 1 TIMTJ nonvolatile magnetoresistive RAM (256 kb 3·0 V 1 TIMTJ非揮發性磁阻RAM)’’（2000年IEEE國際固態電路會議，第7·6 節），及 R· Scheuerlein、W· Gallagher、S· Parkin、A· Lee、 S· Ray、R· Robertazzi、W. Reohr的 ’’A 1 Ons Read and Write nonvolatile memory array using a magnetic tunnel junction and FET switch in each cell”（使用各單元中磁性隧穿接面及薄膜電晶體切換開關以一次啟動讀取及寫入非揮發性記憶體陣列）’’（2000年IEEE國際固態電路會議，第TA 7.2節）° 通常，若多層中的磁化方向平行，則GMR及TMR皆造成低電阻，當磁化的定向逆平行時則造成高電阻。在TMR裝置中，因電子必須隧穿通過該障壁，必須垂直於該等層水平面而施加感測電流（CPP，垂直於水平面的電流）。在gmr 裝置中，感測電流通常在該等層水平面中流動（CIP，水平面中的電流）。然而，藉由快速持續小型化的支援，市售使用CPP TMR的MRAM的發展前途似乎更有可能。亦咸知在一磁性記憶單元陣列中提供偽單元。美國專利號6466475中揭示一範例，其中在該陣列邊緣周圍提供數個 98455.doc 200528739 偽單元^避免邊緣效應。*國專利中請號/綱_ 亦揭不偽單元’用以增加MRAM讀取速度。 MRAM相關的一議題為需要相當大的電流脈衝以導通一位-，因此在MRAM上寫入係大耗電活動。知義約為數十〇e(數kA/m)。為增加此—磁場，必須傳送約5至1〇心的電流至m。MRAM成功的重要關鍵之—係減低 MRAMt的耗電（尤其是用於行動應用）。另—議題為目前在市售GMR感測器產品中，增加一組兩正交感測器用於二維磁場偵測的唯一方式係將兩個分開的正交晶片合併在一電路板上，或將兩個分開的正交晶粒合併在同一晶片封裝中。另一議題為目前在諸如在讀取磁頭等GMR4Tmr感測器中，通量導引係與感測元件分開產i，因此增加額外處理步驟的成本。【發明内容】根據一第一概念，本發明提供一磁性感測器，用以感測磁場，並具有一磁阻感測元件，其具有數層磁性材料及至少一通量導引，用以將磁場集在至該感測元件，該通量導引包括用於該感測元件的至少一相同數層的一部分。該通量導引使用同一層或相同數層的優勢為可使處理步驟的數目保持最小。導引層可與該感測元件的對應層在相同步驟中形成。即若該導引與該元件在分開的步驟中形成，兩者使用相同層仍具優勢。另一優勢為此類磁場感測器可整合在MRAM晶片中，確貫使用MRAM的相同科技。此表示它增加一點成本或未增 98455.doc 12 200528739 加成本。一些實施例使用兩個偽裝置，其如同該感測器的通里導引I又地置於—感測之旁。製造期間無需任何額外處理步驟即可將該等通量導引包含在内。模擬顯示若為外 β磁#感測$ ’具通$導引的感測器在其感測層内部的通里雄度可相當程度地增加，例如增加53%。若為電流感測器，已發現改良至約10%。使用與感測元件具有同一層或相同數層的通量導引的另一優勢為，可減低該扣住層的去磁化場（即一形狀各向異性）。作為-些實施例的額外特徵，該感測元件括—隨穿磁性接面3 Jit類特徵為該導引具有大於該感測元件的面積。另-此類特徵為在該感測元件的數個對面側具有數個導引。另—此類特徵為該等導引的大小在待測磁場的範圍中未全飽和。另—此類特徵為該感測器具有—延長形狀。此能利用各向異性以追蹤磁特性。可形成附屬請求項的額外特徵包括該感測元件係平面的，及垂直於該平面元件的水平面而伯測該隧穿電流。另 :此類特徵為該元件配置成在電阻與磁場間具有一關係， ^體上顯示無磁滞現象。另一此類特徵為該感測元件具自由兹層其具有一易磁化軸’大體上Lenssen et al. Presented a survey of MRAM technology at the 2000 Nonvolatile Memory Technology Symposium (11-15 November 2000 in Arlington, Virginia, USA), entitled "Expectation of MRAM in comparison (Expectations for MRAM in comparison. This document states that the first generation of magnetic random access memory (MRAM) was based on AMR. After 1988, the larger magnetoresistive effect of the giant magnetoresistive (GMR) It is found that smaller components with higher resistance and larger MR effect (5 to 15%) can be realized, and thus get higher output signal. In principle, this can make MRAM used in general applications. Discover GMR effect Ten years later, the GMR effect has been applied to commercial products such as HDD read heads and magnetic sensors. When the large tunneling magnetoresistance (TMR) effect was demonstrated at room temperature, magnetic tunneling was connected around 1995. Breakthroughs in the field have even further improved the prosperity of MRAM. As the TMR effect has been shown to have amplitudes of more than 50% at that time, but due to strong bias dependence, the resistance change available in practical applications is currently around 35%. MTJ Memory unit: The example illustrates a part of a conventional MRAM, which shows an integrated memory cell array using such TMR technology. This structure and how to make it are well known and need not be repeated here. In short, this TMR Type MRAM includes several cells, which are magnetic tunnel junctions (MTJ). MTJ basically includes a free magnetic layer 100, an insulating layer (tunnel barrier 102), a clasping magnetic layer 104, and an antiferromagnetic The AF layer 106 is used to "pin" the magnetization of the buckling layer in a fixed direction (which is commonly referred to as the exchange bias direction). In the example shown in this figure, there is also a bottom layer 108. For simplicity, Figure 1 The magnetic tunnel junction shown shows only 4 active layers in the 98455.doc 200528739 (MTJ) stack. In fact, there can be more layers, which are not relevant to this principle of operation. A μ The MRAM cell is in this freedom Information (1/0) is stored in the magnetization direction of the magnetic layer, and the magnetic layer can rotate relatively freely between two opposite directions. If the magnetization direction of the free layer is parallel to the person holding the layer, the resistance of the ㈣, When the two magnetization directions are antiparallel, the resistance of the MTJ is large. Read the information on the frame and send a small current (vertically) through the MTJ stack of the selection unit. The voltage drop (in proportion to the resistance) measured on the MTJ 5J represents the information of the single it. By transmitting The writer current passes through the digital line Du_3 and the ^ element line BL1-3 (the lines are fixed on the & section and above of the memory cells such as Yanhai). 'The data m on a unit can be changed during a write operation. Current The magnetic field (easy-magnetized axial field and hard-to-magnetized axial field) in the memory unit will be increased. The ^ and 昜 are hardened, so that they are large enough to switch the magnetization of the free layer of the selected unit to a new direction. The bit lines are parallel to the difficult magnetization axis of the cells (/, a magnetic field is generated in the easy-zization axis), and the digital bit lines (sometimes called word lines) instead generate a magnetic field in the difficult magnetization axis . In some designs, these relationships can be reversed, that is, bit lines generate a hardly magnetizable axial field, and digital lines generate a easily magnetizable axial field. The explicit bit line is orthogonal to the digital line, and the magnetic tunneling interface is placed in the intersecting diagram to show the hard-to-magnetize axis field (generated by the digital line) and the easy-magnetized axis (generated by the bit line). The magnetic field finally formed is 4 5 with the axis of easy magnetization. , Its moon dagger pattern transfers the magnetization of the free layer of the selected unit, without affecting all unselected single fans. The bottom electrode of these units is connected to a selection transistor with a through hole, which will be used when reading. 9S455.doc -10- 200528739 Because the magnetic field finally formed is 45 with the easy magnetization axis of the free layer of the unit. Therefore, the switching layer of the free layer has the smallest switching magnetic field, and thus writing can be performed with a minimum current. The magnitude of the resulting magnetic field at this intersection is (| Hha | + | Hea |) / 2, where Hha and HEA are the magnetic fields generated by the hard-to-magnetize axis and the easy-to-magnetize axis, respectively. They must usually be the same size. For more information about this type of MRAM, readers can refer to, for example, 256 KB 3.0 V 1 TIMTJ nonvolatile magnetoresistive RAM (256 KB 3.0 V 1 TIMTJ Nonvolatile Magnetoresistive RAM) "(2000 IEEE International Solid-State Circuits Conference, Section 7.6), and R. Scheuerlein, W. Gallagher, S. Parkin, A. Lee, S. Ray, R. Robertazzi, W. Reohr's `` A 1 Ons Read and Write nonvolatile memory array using a magnetic tunnel junction and FET switch in each cell '' (using the magnetic tunnel junction surface and thin-film transistor switch in each cell to start reading at one time And write non-volatile memory array) "(2000 IEEE International Solid-State Circuits Conference, Section TA 7.2) ° Generally, if the magnetization directions in multiple layers are parallel, both GMR and TMR cause low resistance. When the magnetization is oriented Antiparallel results in high resistance. In TMR devices, because electrons must tunnel through the barrier, a sensing current (CPP, current perpendicular to the horizontal plane) must be applied perpendicular to the horizontal planes of the layers. In a gmr device, the sensing current usually flows in the horizontal planes of these layers (CIP, current in the horizontal plane). However, with the support of rapid and continuous miniaturization, the development prospect of commercially available MRAM using CPP TMR seems more likely. It is also known to provide dummy cells in a magnetic memory cell array. An example is disclosed in U.S. Patent No. 6,646,475, in which several 98455.doc 200528739 pseudocells are provided around the edge of the array to avoid edge effects. * National patent application number / program _ also uncover the fake unit 'is used to increase the MRAM read speed. An issue related to MRAM is that a relatively large current pulse is required to turn on a bit-so writing on MRAM is a large power-consuming activity. The meaning is about several tens of e (several kA / m). To increase this—the magnetic field, a current of about 5 to 10 cores must be delivered to m. The key to MRAM's success is to reduce the power consumption of MRAMt (especially for mobile applications). Another issue is that the only way to add a set of two orthogonal sensors to two-dimensional magnetic field detection in currently available GMR sensor products is to combine two separate orthogonal chips on a circuit board, or Combine two separate orthogonal dies in the same chip package. Another issue is that in GMR4Tmr sensors such as in read heads, the flux guidance system is produced separately from the sensing element, thus adding the cost of additional processing steps. SUMMARY OF THE INVENTION According to a first concept, the present invention provides a magnetic sensor for sensing a magnetic field, and has a magnetoresistive sensing element having several layers of magnetic material and at least one flux guide for Focusing a magnetic field on the sensing element, the flux guidance includes a portion of at least one same number of layers for the sensing element. The advantage of this flux steering using the same layer or layers is that the number of processing steps can be kept to a minimum. The guide layer may be formed in the same step as the corresponding layer of the sensing element. That is, if the guide and the element are formed in separate steps, it is still advantageous to use the same layer for both. Another advantage is that such magnetic field sensors can be integrated in MRAM chips, consistently using the same technology of MRAM. This means that it adds a little or no additional cost to 98455.doc 12 200528739. Some embodiments use two pseudo-devices, which are placed side-by-side with the sensor, like the channel guide I of the sensor. This flux guidance can be included without any additional processing steps during manufacturing. The simulation shows that if the external β magnetic #sensing sensor has a high degree of acuity within the sensing layer, for example, it can increase by 53%. In the case of current sensors, improvements have been found to about 10%. Another advantage of using flux guidance with the same layer or layers as the sensing element is that the demagnetization field (ie, anisotropic shape) of the pinned layer can be reduced. As an additional feature of some embodiments, the sensing element includes a magnetic-through interface 3 Jit-like feature in that the guide has a larger area than the sensing element. Another feature of this type is that there are several guides on several opposite sides of the sensing element. Another—This type of feature is that the magnitude of these guides is not fully saturated in the range of the magnetic field to be measured. Another—such feature is that the sensor has—an extended shape. This can use anisotropy to track magnetic properties. Additional features that can form subsidiary claims include that the sensing element is planar and that the tunneling current is measured perpendicular to the horizontal plane of the planar element. Another feature of this type is that the element is configured to have a relationship between resistance and magnetic field, and no hysteresis is displayed on the body. Another such feature is that the sensing element has a free layer and it has an easy magnetization axis.

磁場的方向。用於_MTTA“丄~ # J 、勺自由層等磁性元件，總有效各 :異性有兩個來源一為材料因晶體結構造成的固有各向向異性。前去p 為延長形狀造成的形狀各、吊Μ於後者。若此兩各向異性對齊，總各向異性則為兩者的和。一右兩者正父，則有效各向異性將跟 98455.doc 13 200528739 =強者’強度僅為兩者的差。即若該兩各向異性正交作 :狀各向異性較強，有效各向異性(易磁化軸)將仍在延長的，中。此係用於感測器必須整合在臟鹰中的例子。整個晶片的結晶各向異性相同，但藉由在該等mram單元^ =各向異性方向，將與該等難规單元形狀正交的數個 ^器形狀定圓案，並可得到該等感測器。因此，利用一二材料的各向異性特f可提供改良的電阻特性，諸如不用引入太多磁滞現象即可使靈敏度增至最大等。另一The direction of the magnetic field. It is used for magnetic components such as _MTTA "丄 ~ # J, free layer of spoon, etc., there are always two effective sources: anisotropy has two sources. One is the inherent anisotropy of the material due to the crystal structure. If the two anisotropies are aligned, the total anisotropy is the sum of the two. If there is a right father, the effective anisotropy will follow 98455.doc 13 200528739 = stronger's strength is only The difference between the two. That is, if the two anisotropies are orthogonal: the shape anisotropy is strong, the effective anisotropy (easy magnetization axis) will still be extended, and this is used for the sensor must be integrated in An example in the dirty eagle. The crystal anisotropy of the entire wafer is the same, but by using the mram unit ^ = anisotropic direction, the shape of several squares orthogonal to the shape of the difficult unit is rounded. These sensors can be obtained. Therefore, using the anisotropic characteristics of one or two materials can provide improved resistance characteristics, such as maximizing sensitivity without introducing too much hysteresis. Another

:為該接面包括-扣住磁層，其在朝垂直於該自由磁層的〜兹化轴方向具有—磁化。明顯地該輸出可為任何事物， =表不疋否偵測到磁場的邏輯信號到表示對一已知精確位準測里的類比或數位信號。例如可對偵測到的輸出實施合適的後處理，以適應該應用的精確度或雜訊免除。: The interface includes a -clamping magnetic layer which has -magnetization in a direction perpendicular to the axis of the free magnetic layer. Obviously, the output can be anything, which means whether the logic signal of the magnetic field is detected to indicate the analog or digital signal in a known accurate level measurement. For example, the detected output can be appropriately post-processed to suit the accuracy or noise immunity of the application.

雕本發明的另一概念為提供一電流感測器，用以感測一導體中的電流，《流感測器包括上述磁場感測器、，其設置以谓測電流造成的磁場。此尤其為磁場感測器的有用應务月的另概念為提供用以感測磁場的感測器，其 ^有-感測元件及-通量導引1以將磁場集中在該感測凡件上，该感測7G件係一磁阻感測元件，俾便該感測元件、之長軸大體上與感測的磁場正交。此可提供一特別靈敏的感測器。古本發明的另一概念提供一積體電路，其具有一内建式電流感測器，該電流感測器包括上述電流感測器。作為一額外特徵，該電流感測器設置成感測靜止電流（IDDQ)。作為 98455.doc -14- 200528739 另一額外特徵，該積體電路具有多個電流感測器，其鏈結在一掃描鏈中。本务明的第二概念提供一磁性記憶體，其具有至少一記憶單元，該等單元中之至少一者具有至少一通量導引，用以集中用以寫入該單元的磁場。使用通量導引可有助於減低寫入该冗憶單元所需的寫入電流。此對於電池供電他行動裝置尤其重要。 …、在°己隐元件之旁設置的偽MTJ的自由層可當作通量導引使用。由於此類通量導引，在一些實施例中，寫入電流了減低、、、勺10/。’總寫入耗電並可減低約至游。(依實施及 :否僅有：磁場或兩磁場(易磁化軸及難磁化軸)配備有通里導引而定）。_些實施例的特徵為該等通量導引與該單元 =由磁層共面。此可提供該磁場的特別有效集中。一些 ““列的特徵為該等通量導引係以偽MTJ的形 <，其具二相關聯冗憶單元的所有層。此可確保該等導引不用任施。人驟即可共面，因此可以極少或無額外成本地實用些實施例的額外特徵’該導引包括該記憶單元使 =同-層(或數層)的一部分。此能使該二外處理步驟即可坌等Μ…、而額 π成，其可有助於減低成本及維持可靠度。 ==徵’該單元係-隧穿接面。作為另一此配置^ 導引重疊數疊間的一通孔。此有助於保持 _ y在，使可整合的單元數增至最大。 …、、作為另—此類特徵，通量導引在該單元的四側上形成。 98455.doc -15 - 200528739 作為另一此類特徵，該通量導引包括_偽單元。本發明的另一概念提供一磁性感測器，其具有：一第一感測元件，用以測量一第一方向中的磁場；一第二感測元件，用以測量一第二方向中的磁場；及一通量導引，用於，等感測元件中之至少一者，該通量導引配置成旋轉該磁場方向經過該感測元件。綠万疋轉有助於使 .....“ μ 〜里从砀砸場的方向。若其可避免該等感測器具有不同定向的需要，則可使 &較簡單。其可使一 MRAM晶片内的積體二維磁場感測器能達成。在一些實施例中’該感測器具有不同幾何圖形的不同通量導引。該等通量導引當作轉向器，在兩個不冋方向中旋轉該等感測元件的感測方向。纟自該等感測器 ^合併信號可提供測量磁場向量的相關資訊。此方式的優、：用於兩感測元件不再需要正交交換偏壓方向。若該等 ::導引係偽裝置，此方式可避免將該等通量導引定圖案 :的分開步驟’藉此可較容易整合該等感測器。尤其； :小或無額外成本地整合在—MRAM晶片内。該等導引少狀可為平行四邊形。作為另-特徵’纟第-及第二感測元件上皆設置通量導。作為另一此類額外特測磁場的正交分量。此1感測器配置成偵里此方式可使判定磁場方向較容易。為另-此類特徵’將該等感測元件相同交換偏壓方向。茈w 、午讥一有合在— n較容易製造，並有助於使其能整已。作為另一此類特徵，該等感測元件係整合在同 98455.doc 200528739 一曰曰片上。此有助於使空間及製造成本能減低。作為另一此類額外特徵，在該等感測元件的兩側上設置通量導引， =方；忒等感測元件任一者的通量導引係平行。此能使該通置的旋轉較有效率。作為另一此類額外特徵為判定（來自該等第一及第二感測元件的）磁場方向及大小的電路結構。作為另一此類特徵，該電路結構配置成考量未由該等通量導引旋轉或放大的背景磁場量。此可使計算能較準確。作為另一額外此類特徵，該等通量導引配置成優先由磁靜電交互作用而非交換父互作用以耦合該等感測元件。此可增加該磁場旋轉的效率。先前即已開發磁性隧穿接面用於記憶體應用，本發明人了解其可調適以作為感測器使用（先不管記憶單元及電流感測器一定具有不同特性的事實）。在一記憶單元中，一自由層的磁阻迴路（MR迴路）應與一較大抗磁力（約數十〇e)成直角，並具有兩個不同的剩餘磁化強度狀態。（剩餘磁化強度指在施加磁場在樣本已達到飽和後減至〇之後再恢復磁化）。此外，該迴路的中心必須在零磁場。相對地，一電流感測器一方面必須儘可能對磁場有大敏感度（用於高靈敏度），另一方面必須具有小（或無）磁滞現象。另一概念提供用以感測磁場或感測電流（或寫入一 MRAM)的對應方法。另一概念提供製造感測器的方法，該方法包括在同一步驟中形成一層感測元件及一層用於該感測元件的通里導引的步驟。另一概念提供製造具上述特徵 98455.doc 200528739 的MRAM的方法。该寻附屬請求項中任一者的特徵可互相結合，獨立請求項中任一者結合。熟諳此藝者將明白進優勢，尤其在本發明未知曉的其他技藝上。兹將參昭至^ 說明可如何實施本發明。明顯地，不用背離本發明的二主: 項可作出許多變化及修改。因此，應可清楚了解本的形式僅用於說明，不希望用以侷限本發明的範圍。【實施方式】本發明將相關讀㈣殊實_並參照至料附圖加以說明，但本發明並非侷限於該等實施例及附圖，而是附請求項加以界；t。所示附圖僅為示意，並非用以:定: 附圖中，用於說明目的，會誇大該等元件的大小= 例繪製。本說明及後附請求項中使用"包括詞之處，並Another concept of the present invention is to provide a current sensor for sensing a current in a conductor. The "flu tester includes the magnetic field sensor described above, and it is arranged to measure a magnetic field caused by a current. Another useful concept of the magnetic field sensor, in particular, is to provide a sensor for sensing a magnetic field, which includes a sensing element and a flux guide 1 to focus the magnetic field on the sensing element. On the part, the sensing 7G part is a magnetoresistive sensing element, and the long axis of the sensing element is substantially orthogonal to the magnetic field sensed. This provides a particularly sensitive sensor. Another concept of the present invention is to provide an integrated circuit having a built-in electric influenza detector, and the current sensor includes the above-mentioned current sensor. As an additional feature, the current sensor is configured to sense a standstill current (IDDQ). As another additional feature of 98455.doc -14-200528739, the integrated circuit has multiple current sensors, which are linked in a scan chain. The second concept of the present invention provides a magnetic memory having at least one memory unit, and at least one of the units has at least one flux guide for concentrating a magnetic field for writing into the unit. Using flux steering can help reduce the write current required to write to the memory cell. This is especially important for battery-powered mobile devices. …, The free layer of the pseudo MTJ set beside the hidden element can be used as a flux guide. Due to such flux guidance, in some embodiments, the write current is reduced by 10%. ‘Total write power consumption can be reduced to about. (Depending on the implementation and whether: only the magnetic field or two magnetic fields (easy magnetization axis and difficult magnetization axis) are equipped with a directional guide). A feature of some embodiments is that the flux guidance is coplanar with the unit by the magnetic layer. This can provide a particularly effective concentration of the magnetic field. Some "" columns are characterized by such flux guidance in the form of a pseudo-MTJ < which has all the levels of the associated memory cells. This ensures that such guidance is not required. Humans can be coplanar, so the extra features of some embodiments can be implemented with little or no extra cost. The guide includes the memory unit to be part of the same-layer (or several layers). This enables the two external processing steps to wait for M ... and π, which can help reduce costs and maintain reliability. == 征 'This unit is a tunnel junction. As another such arrangement, ^ guides a through hole between the overlapping stacks. This helps keep _ y on and maximizes the number of units that can be integrated. ... As another feature of this type, flux guidance is formed on the four sides of the unit. 98455.doc -15-200528739 As another such feature, the flux guidance includes a pseudo-unit. Another concept of the present invention provides a magnetic sensor including: a first sensing element for measuring a magnetic field in a first direction; and a second sensing element for measuring a magnetic field in a second direction. A magnetic field; and a flux guide for waiting for at least one of the sensing elements, the flux guide being configured to rotate the direction of the magnetic field past the sensing element. Lvwanzhuanzhuan helps make ..... "μ ~ li from the direction of the field. If it can avoid the need for these sensors to have different orientations, it can make & simpler. It can make An integrated two-dimensional magnetic field sensor in an MRAM chip can be achieved. In some embodiments, 'the sensor has different flux guides with different geometries. These flux guides are used as redirectors. The sensing directions of the sensing elements are rotated in different directions. The combined signals from the sensors ^ can provide relevant information for measuring the magnetic field vector. The advantage of this method is that it is no longer needed for the two sensing elements. Orient the bias direction orthogonally. If these :: guides are pseudo-devices, this method can avoid the separate steps of guiding the flux to the pattern: 'This makes it easier to integrate these sensors. Especially; : Small or no additional cost integrated in the -MRAM chip. These guides can be parallelograms. As another feature, the first and second sensing elements are provided with flux guides. As another The quadrature component of the extra-special magnetic field. This 1 sensor is configured to detect The direction of the magnetic field is easier. For another-such feature, the same biasing direction is exchanged for these sensing elements. 茈 w and afternoon are combined-n is easier to manufacture and helps to make it straight. Another such feature is that the sensing elements are integrated on the same sheet as 98455.doc 200528739. This helps to reduce space and manufacturing costs. As another such additional feature, the sensing elements A flux guide is set on both sides of the, = square; the flux guide system of any one of the sensing elements is parallel. This can make the rotation of the placement more efficient. As another such additional feature is the judgment Circuit structure of magnetic field direction and magnitude (from the first and second sensing elements). As another such feature, the circuit structure is configured to take into account the amount of background magnetic field that is not rotated or amplified by the flux guidance This allows the calculation to be more accurate. As another additional such feature, the flux guidance is configured to preferentially couple the sensing elements by magnetic-electrostatic interactions rather than exchange parent interactions. This can increase the magnetic field The efficiency of rotation. The magnetic tunneling interface is used in memory applications. The inventors understand that it can be adapted to be used as a sensor (regardless of the fact that the memory unit and the current sensor must have different characteristics). In a memory unit, a free The layer's magnetoresistive circuit (MR circuit) should be at right angles to a large coercive force (about tens of e) and have two different states of residual magnetization. (Residual magnetization refers to the saturation of the sample when the magnetic field is applied. The magnetization is restored after reducing to 0). In addition, the center of the loop must be at zero magnetic field. On the other hand, a current sensor must be as sensitive as possible to magnetic fields (for high sensitivity) on the one hand, and on the other hand Must have small (or no) hysteresis. Another concept provides a corresponding method to sense a magnetic field or a current (or write to a MRAM). Another concept provides a method of manufacturing a sensor, the method includes In the same step, a layer of a sensing element and a layer for guiding the sensing element are formed. Another concept provides a method of manufacturing an MRAM with the above characteristics 98455.doc 200528739. The features of any one of the subordinate request items can be combined with each other, and any one of the independent request items can be combined. Those skilled in the art will appreciate the advantages, especially in other techniques not known to the present invention. Reference will be made here to explain how the present invention can be implemented. Obviously, many variations and modifications can be made without departing from the two main points of the present invention. Therefore, it should be clearly understood that the form of the present invention is only for illustration, and is not intended to limit the scope of the present invention. [Embodiment] The present invention will be described in detail with reference to the attached drawings, but the present invention is not limited to these embodiments and drawings, but is bounded by the appended claims; t. The drawings shown are for illustration only, and are not intended to: determine: In the drawings, for illustration purposes, the size of these elements will be exaggerated = drawing for example. Where "quote" is used in this note and the appended claims, and

未排除其他元件或步驟。當參照至 W 双石3使用的不定冠阔或疋艰詞（例如”一”、” 一個”、” 除非有特別說明，否則亦包括該名詞的複數。說明中及請求項中第一、第二、第三等用詞係用 A刀類似元件’不必然是說明-排列或時間先後順序。 $了解如此使用的詞在適當環境下可互相替換，本發明在 =中所述實施例能以本文中說明或描繪以外的其他順序呆作。此外’說明中及請求項中上、下、上方、下方等用詞係用於說明目的，並不必然用以铐明如 ^ ffl 以說明相關位置。應了解如此吏用的詞在適當環境下可互相替換’本發明在本文中所述 98455.doc 18 200528739 灵施例月b以本文中說明或描繪以外的其他定向操作。圖3及4，本發明的第一實施例本發明人了解MTJ裝置可調適成當作磁場感測器使用。例如在電流（如功率電流或1〇〇電流）的無接觸測量中，此等裝置有許多應用。此等裝置可設置在任何種類的CM0S晶片或其他晶片中。當然亦可實施在MRAM晶片中。用於MRAM 單元的相同製造技術可稍加改變，用以建立積體電流感測為。由於此等感測器的實施不需要花費任何額外的遮罩或額外的處理步驟，因此特別適合MRAM晶片（或含嵌入式 MRAM的晶片）中的功率接腳測試及mDx測試。若該等感測态將位於MRAM晶片内部，則它們必須確實與MRAM使用相同的技術，不用任何額外的處理步驟。將沈積MRAM/感測器堆疊，接著並馬上在該過程後端中的兩個金屬化位準之間將該堆疊定圖案。此可使感測器設計較嚴格及較無彈性，雖然將常因低成本而偏好此設計。例如，以習用方式製造通量導引會花費如1¾離、沈積、微影及钱刻等數個額外處理步驟，因此並非想要的方式，在許多例子中亦不容許此方式。本發明實施例可提供一簡單方式以建立該等磁場感測器的通量導引’該等磁場感測器整合在MRAM晶片中，並可確實分享相同的MRAM製程。如圖3及4所示，設置在真實感測元件附近的兩個偽MTJ元件的自由層可作為通量導引使用，將較多通量集中至該感測器中，藉此使偵測限制降低。簡言之，藉由模擬已發現，使用該等偽裝置作為通量導引，若 98455.doc -19- 200528739 為磁場感測器’通量密度可增加53%。用於電流感測器（其使用如MRAM中的相同方式以產生磁場），可改良約1〇%。圖3以剖面圖說明感測元件21〇，及作為通量導引12〇使用的偽裝置。該等層對應至圖丨所示記憶單元的分層。有一感測層（自由磁層200)、一絕緣層（隧穿障壁2〇2)、一扣住磁層 204及反鐵磁人^層206，其用以將該扣住層的磁化，，釘住，，在一固疋方向（交換偏壓方向卜此圖所示範例中，亦有一底層208，當作下接觸層。為求簡化，圖3所示磁性隧穿接面 (MTJ)堆疊中僅顯示4個主動層。實際上，可有更多層（其與操作原理並不相關）。該兩毗鄰偽MTJ裝置的感測層可當作位於該兩裝置之間的作用（真實）感測器的通量導引。此設計容許該磁場感測器的通量導引無需任何額外處理步驟便可產生。該兩偽裝置的自由層因其磁化可自由旋轉，因此現在當作通量導引使用。該等偽裝置的扣住層因其磁化固定，因此在測量期間將不會影響通量變動。圖4是上視圖。為在該兩偽裝置之間產生較佳效率及同質磁場，最好該兩偽裝置（當作通量導引使用）大於該感測器。該等通量導引的導磁率可藉由變化其幾何圖形而加以調整。應選擇該等偽裴置的縱橫比及大小，俾便其在待測磁場範圍中未全飽和，但仍具備夠大導磁率以集中該通量。將该等通量導引定圖案應儘可能靠近該感測器，但仍與該感測器隔離。它們越靠近，該等通量導引越有效率。當然隶小間隔係依微影解析度及姓刻技術而定。儘管該等通旦 98455.doc -20- 200528739 導引與該感測器之間有小間隔，一上接觸層在該感測器上方的位置並不重要。事實上，其可有一些容忍度。即若該感測器的上接觸層與該等通量導引中之一接觸，因該等通量導引在底部並不相連，因此通常不會造成任何問題。若作為使用MRAM技術的内建式或積體電流感測器使用，可在如用於MRAM的字線或位元線的相同下金屬化層中將該電流導體定圖案。該導體係流電地與該MTj裝置隔離。當傳送待測電流通過該線時，電流在該線本身周圍產生一磁場，該MTJ裝置將感測該磁場。該MTJ裝置的電阻變化即表示該磁場’因而表示該電流。圖5至1 〇，模擬結果：在此等圖中說明使用有限元件技術的模擬結果。執行該等模擬用於兩例子：磁場感測器及電流感測器，兩例子皆计异其具備及不具備該等通量導引的結果。圖5說明用於兩例子中橫越一磁性感測器的通量密度的χ分量（磁應力”具通量導引及不具通量導引。在此χ方向表示與感測器水平面平行的方向。可明顯見到，具有該等通量導引，感測器中的通量密度戲劇性地增加53%。在模擬中，為求簡化，僅考量感測器的自由層及該等通量導引。假設感測器的感測層（自由層）及該等通量導引的導磁率分別約為2〇〇及1〇〇〇。此為實際數值，並依該等通量導引及該感測器的幾何圖形而定。感測器寬度為1㈣，通量導引寬度為5㈣，它們之間的間隔為1〇〇_。自由層厚度為5nm(其通常用於mram 堆：g：中）。在該自由層水平面的方向中施加5〇e的定量及同 98455.doc -21 - 200528739 質磁場。圖5(修改版）：圖6說明若為未具通量導引的感測元件2丨〇的先前技藝所算出的磁場線影像。圖7說明若為根據本發明具有通量導引 120的實施例的一對應影像。用於比較，兩影像中使用相同密度縮放比例。若為具通量導引的感測器，明顯見到該等磁％線由於該等通量導引而較集中在該感測器的感測層上。在圖7中，為求簡明僅顯示該等通量導引的一部分，該寺通里導引貫際上在兩方向延伸更遠。圖8及9說明一電流感測器的磁場線的對應影像，分別具有及未具通量導引。該電流感測器例子應用相同幾何圖形及參數，其中，不施加一同質磁場至該感測器，而在該感測器下方距離150 nm處放置一導體13〇，其剖面為3〇〇 nm (南）xl5〇0 nm(橫向）。傳送一 15 mA的電流經過該導體（在垂直於該圖的方向中），在導體本身周圍產生一圓形磁場。在該感測器位置，產生的磁場約5 〇e，其與圖6及圖7中的例子相容。同樣，在圖9中，為求簡明僅顯示該等通量導引的一部分，該等通量導引實際上在兩方向延伸更遠。 < 像中了看出改善並不如圖6及7所示的明顯。原因是該等通量㈠丨的-部分僅有較小部分接近該制器，其中較罪近水平方向的通量線實際上對通量導引效應有貢獻。詳細模擬資料顯示，相較於未具通量導引的例子，在具有通里導引的感測器内部的通量密度增加約丨，如圖1 〇在圖表中次明與感測器中心相距不同距離橫越該感測器寬度 98455.doc 200528739 的通量密度。此外，該等通量導引可合併使用電流線的碾鍍（如世界專利號02/41367 A2所揭示），其常用於MRAM位凡/數位線。此例中該等通量導引的效應約增加丨〇%，超越具碾錢層的導體線所產生的改良式磁場。其他效應：在一未具通量導引的感測器中，該感測器的延長外形引起一形狀各向異性：Kd=l/2 μ0 Μ?(Νχ·Νγ)，其中Ms係該層的飽和磁化，NX&NY分別為橫向及縱向的去磁化因子。此形狀各向異性傾向使該感測器的任何磁層的磁化對齊至縱向。該70件的縱橫比越大，形狀各向異性變得越強。用於具約1微米寬度的感測器，期望具縱橫比（AR)5比7的延長形狀，用以穩定該自由層，及藉此減低磁滯現象。惟，用於扣住層（其磁化有目的地固定在橫向），該形狀各向異性將傾向強迫磁化離開該扣住方向，其可造成該感測器的不良行為。藉由#近該感測器的兩邊放置兩個較大的偽裝置，減低該感測H的扣住層的去磁化場，並藉此減低形狀各向異性。原因是該兩偽裝置的扣住層亦固定在與該❹U器的扣住層相同的方向中’藉此使其固定方向較不會造成不利影響。類似方；則即中所述效應，該自由層（感測層）在形狀各向異性中會見到相同的減低”隹，該效應較小，原因是該自由層（通常由鎳鐵合金製成）的飽和磁化小於該扣住層（通常一、戴化始（CoFe)製成）的餘和磁化。此例中的自由層表現如同其較大的有效寬度，因此有較小的从。然而，該自由層 98455.doc -23- 200528739 的一些形狀各向異性對壓制磁滯現象有用，但另一方面該形狀各向異性傾向劣化該感測器的靈敏度。因此在各特殊應用中應找到一妥協值。在具通量導引的感測器的幾何形狀設計期間，必須考量減低該自由層的有效形狀各向異性，以得到想要的形狀各向異性值。雖然所述實施例參照至使用TMR效應的磁性感測器，惟該構想並不限於此，但例如亦可應用至GMR感測器。圖11至15，應用通量導引以增強MRAM單元的寫入：亦可應用上述相同通量導引原則以藉由增加在單元上由寫入電流產生的通量密度而增加MRAM。此在寫入期間可省下20%的耗電。該等通量導引的實施不需要花費任何額外的處理步驟，僅需改變該遮罩的設計。此外，若該等通量導引可剛好放入該等MTJ記憶單元11 〇間的未使用空間中，則不需要加大該單元的尺寸。圖11及12以上視圖及剖面圖說明未具通量導引的先前技藝MRAM(對應至圖1的設計）。在字（或數位）線方向中，只要各單元夠大足以在該石夕基板上藏置該隔離電晶體，及該等單元間仍無重大磁性麵合，該等單元可儘可能靠近地設置。沿著此方向設置通量導引通常會加大單元尺寸。惟，沿著位元線方向，由此需要一些空間用於底電極通孔1 50，因此該等單元間的間隔明顯較大。此未使用空間可用來放置偽MTJ作為通量導引。圖13及14根據本發明以上視圖及剖面圖說明一 mram的第一實施例。該設計對應至圖1的設計，並使用相似參考數字以適當說明。每一 MTJ記憶元件（圖中以MTJ表示）夾在兩 98455.doc -24- 200528739 個偽MTJ(其作為通量導引（FG)12〇)之間。該等偽MTJ設置在位元線方向中。該等偽MTJ與該等記憶體元件可同時準確地以相同方式定圖案。部分位於相同通孔上方的兩通量導引可分享同一底電極作為記憶體元件。惟該等底電極中的連接在该荨單元之間總必須打斷。該位元線係經由頂電極通孔160(其準確地位於該記憶體元件上方）而連接至該記憶體元件的頂電極。因此該等偽MTJ係開放電路，不會以電影響該記憶體元件。該等偽MTJ與該記憶體元件間的間隔應儘可能地小，例如丨〇〇 nm左右或更小（依技術能力而定）。該等偽MTJ應僅佔用該等元件間未用到的空間。使用該等偽MTJ作為通量導引，傳送至一數位線的寫入電流（其在此範例中產生難磁化軸場）可減低約1〇%，以得到如未具通量導引的例子中的相同磁場。換言之，該難磁化軸場的耗電（與電流平方成正比）幾乎可減低2〇%。若考慮用於兩磁場（難磁化軸場及易磁化軸場）的總耗電，則總寫入耗電可減少約10%(假設位元線電流等於數位線電流）。該等通量導引（FG)(以偽MTJ的形式）在位元線方向中的數個記憶體元件之間定圖案。若未知何故該數位線方向巾的數㈣憶體元件之間需要較大空間（例如因為需要較大空間用於該隔離電晶體)，則可藉由在兩方向中放置通量導引而使用此未使用空間。其中在兩方向（即難磁化在圖1 5中，說明一第二實施例，軸場及易磁化軸場方向）中放置該等偽單元。在豆他方面該設計對應至圖丨3及U的設計，並使用彳目似參考數字以適當 98455.doc -25- 200528739 說明。在此例中，在位元線及數位線的電流皆可減少1 〇〇/0，用於兩磁場的寫入耗電可減少2 0 %。圖1 6至21，用於二維感測器的通量導引·· 藉由介紹此類感測器，將簡短地說明TMR感測器的研發。將MRAM整合至CMOS科技中已產生將磁場感測器包含在MRAM晶片内的新機會，原因是MRAM及磁性感測器係根據相同的效應’ I1遂穿磁阻（TMR)效應。該等磁性感測器將與MRAM使用完全相同的科技，僅有幾何形狀設計不同；因此將磁性感測器整合至MrAM晶片内可無額外成本地增加MRAM晶片的額外價值。積體磁性感測器的應用可包括··用於電力接腳測試或IDDx測試的高度靈敏電流感測态，用於MRAM寫入控制的外部磁場感測器，用kMRAM 的主動保護，積體羅盤等等。如上述，磁阻感測器可根據該等磁阻效應之一，諸如巨型磁阻（GMR)或隧穿磁阻（TMR)等。一 TMR感測器包含一感測兀件，其為一磁性隧穿接面（MTJ，如上述圖4及5所示）。春圖16至19以示意圖說明一根據TMR效應的磁場感測器。使用相似參考數字以適當說明。自由層200的磁化較自由依施加磁場而旋轉，反而該扣住層的磁化總指向一固定方向（在此範例為X方向）。該自由層在y方〜向中有-易磁化軸（通常係藉由將在y方向中延長的一疋圖案）。若無施加磁場，自由層的磁化（M胃* )會停在、：軸的方向。在乂方向中出現一磁場時，Μ“偏離丫方 °而旋轉，朝向該磁場的方向（X方向）。磁場越強，角度_ 98455.doc -26- 200528739 Μ14 x方向間的角度）越小。該MTJ的電阻（藉由垂直於從該 MTJ的頂私極至底電極的堆疊傳送—感測電流而測量）根據以下交換函數而變動·· R(0)/R(〇)= 1 +MR*(1 - cos^)/2 其中R⑺及R(0)係該MTJ在Θ分別為非零及為零時的電阻； MR係磁阻比（通常約3〇%至4〇%)。一 TMR感測器通常設計成在待測磁場方向垂直於該自由層的易磁化軸（即在该扣住磁化的方向中）時適當地操作。此方向稱為該感測器的感測方向1 了電流感測器之外，其中夕應用而要—維（2D)磁場债測能力，即感測器系統必須提供該磁場向量投射在該晶片水平面的相關資訊。此磁場投射（以下稱平面磁場H)可分解Hx&Hy，各由一分開的感測器加以偵測。用以伯測_2D磁場向量，需要兩個正交的感測器。各感測器僅對該磁場的—分量（Ηχ或Η力敏感，來自該兩分量的合併信號並提供該平面磁場的方向及大小的相關資訊。在GMR及TMR感測器中，該感測方向係由該扣住層的磁化方向作決定性的判定。如上述，此層由下方的反鐵磁層釘住。此釘住效應稱為交換㈣。藉由在將該堆疊退火至攝氏數百度，接著在出現一磁場時冷卻下來的處理期間設 :該交換偏f方向。在冷卻及移除磁場後，該交換偏塵方二：二：先：施加的磁场方向中。此步驟可在該等感測元田^圖案之W或之後完成。因該等感測器係從相同多層堆豐定圖案’因此通常具有相同的交換偏壓方向。在目前市 9S455.doc -27- 200528739 售的GMR感測器產品中，產生2D磁場偵測的一組兩個正交感測器的唯一方式係在一 PCB上結合兩個分開的正交晶片’或在相同晶片封裝中結合兩個分開的正交晶粒。為將兩個正交感測器整合至相同基板，需要將分開的感 ’貝J卯局σ卩退火。貫際上此方式極困難或不可能。尤其該等感測器若與MRAM整合，目前則無法做到此處理步驟，原因是用於MRAM僅可能有一交換偏壓方向。用於GMR感測器，局部改變交換偏壓方向的一方式，係在出現朝一期望方向的磁場時，藉由傳送高電流脈衝通過一些選取的感測荔長條’苓考K-M.Lenssen的"Magnetoresistive sensors and meiT10ry(磁阻感測器及記憶體），’會NATO ASI文章，2001年 1 2月）电流產生的熱將會使選取的感測器的交換偏壓方向重設至新方向，而其餘感測器仍保持原狀（保留在先前設定白）°亥方法初始用以在惠氏（Wheatstone)橋接器中設疋不同感測器的相對交換偏壓方向。此方法僅可用於非積月且感’到态’及僅用於無障壁的GMR感測器。用於TMR感測杰，傳送此一高電流（因此具高電壓）並不適當，原因是該裝置谷易在曝露至大於約丨v電壓時損壞。、炫將。兒明製造一對（或數對）積體TMR感測器的一簡單方弋用灰2D磁場测量（諸如2D外部磁場測量等）或用於一電子羅盤。使用偽裝置的通量導引，在不同角度下定圖案， :用以將磁場向量分解成2個分量，而該等感測器的交換偏 ^ ^向仍為平行。意即該裝置不需要局部不同的交換偏壓向因此在用以製造一 MRAM的步驟之外不需要額外的 98455.doc 200528739 處理步驟。因此該方法極適用於一 MRAM晶片内的感測器整合。使用與上述圖1至1 5相關的相同通量導引原則。惟在此例中’該等通量導引不僅用以增加通量密度，亦用以將平面磁場分解成兩個旋轉分量，各分量由一分開的感測器加以偵測。此等感測|§具有相同的交換偏壓方向，卻搞合至具不同定向的通量導引。圖19中說明該方法的原則，圖19以示意上視圖說明2D磁性感測器A及B及相關聯的通量導引。感測器A(310)及感測器b(320)在X方向具有相同的交換偏壓方向，並在y方向具有易磁化軸，藉此它們設計成在X 方向感測最佳。各感測器在其左側及右側具有2個平行四邊形的通量導引300。該兩組感測器_通量導引以足夠的一些㈣互相分開’以避免互相影響。該等通量導引係偽肋衣置’因此其疋圖案方式可相同於該等工作感測器。感測杰八的各通量導引有2側與該感測器的縱向平行，及另2側在 Υ’方向，其與X形成一角度α。在此範例中，α為45。（但亦可為一不同值）。感測器Β的各通量導引有2側與該感測器的縱向平仃及3 2側在X方向，其與父形成一角度卢。通常卜〇(即該兩感測器的通量導引在χ方向上為鏡像。為從該等感測器得到同等及對稱信號，該等通量導引的鏡像是必要的。為防止D亥等通里導引在零磁場飽和（其中該磁性力矩較佳朝向-方向）’該等平行四邊形的通量導引必須具有小的層中的各向異性的一重要來源為形狀各〜、！·生（由4層的延長形狀所造成）。因此該等通量導引的形 98455.doc -29- 200528739 =不可在任何方向（明顯）延長。此外，該等力矩在零磁場應分割成數個領域。此可藉由使該等導引夠大（至少在各側為數微米）而完成。 ’' 當出現一施加平面磁場H時，該等通量導引沿著其縱向集中通里，並在該感測器兩測產生磁性電荷（見圖19的感測器 A乾例）。該等電荷因此在χ方向（其與該感測器的感測方向致）中產生通里線。因該等通量線寧可沿著該等通量導引勺傾斜側私動，因此該通量在磁場平行於此方向（在此範例中，此為感測器Α的y，方向）時最強。明顯地，該等通量導引作為一通量轉向器以旋轉該感測器的感測方向，從感測的X方向至y，方向。同樣地，感測對朝向χ，方向的磁場最敏感。當兩感測器皆在使用中時，一平面磁場Η似乎被 :解成兩個正交方向(即X，及y，)。來自該等感測器的信號不含糊地由一電子電路加以記錄，其接著合併並提供該磁場向量的相關資訊（諸如方向及大小等）。該等感測器的行為會看似有兩個正交的感測器。現象為理想例子，其中已假設該等通量導引產生的磁场退比背景磁場強。實際上，情況有些不同。事實上見No other elements or steps are excluded. When referring to the indefinite or broad words used in W Double Stone 3 (such as "a", "a", "" unless otherwise specified, the plural of the noun is also included. First and second in the description and in the request The third and third words are similar to the use of A-knife similar elements. 'It is not necessarily a description-arrangement or chronological order. $ Understand that the words used in this way can be replaced with each other under appropriate circumstances. In addition to the description or depiction in the order. In addition, the words "up, down, up, down, etc." in the description and in the request are used for illustration purposes, and are not necessarily used to indicate relevant positions such as ^ ffl. It should be understood that the words used by such officials can be replaced with each other under appropriate circumstances. The present invention is described in this document 98455.doc 18 200528739 灵例月 b b is operated in other orientations than described or depicted in this article. Figures 3 and 4, this First Embodiment of the Invention The inventor understands that MTJ devices can be adapted to be used as magnetic field sensors. For example, in contactless measurement of currents (such as power current or 100 current), these devices have many applications These devices can be set in any kind of CMOS chip or other chips. Of course, they can also be implemented in MRAM chips. The same manufacturing technology for MRAM cells can be slightly changed to establish integrated current sensing behavior. Because of this The implementation of such sensors does not require any additional masking or additional processing steps, so it is particularly suitable for power pin testing and mDx testing in MRAM chips (or chips with embedded MRAM). If these sensing states Will be located inside the MRAM wafer, then they must indeed use the same technology as the MRAM without any additional processing steps. The deposited MRAM / sensors are stacked and then immediately at the two metallization levels in the back end of the process This stacks the pattern at a time. This can make the sensor design more rigid and inflexible, although this design will often be preferred due to low cost. For example, custom-made flux guidance can cost There are several additional processing steps such as lithography and money engraving, so this is not the desired way, and it is not allowed in many examples. Embodiments of the present invention can provide a simple way to establish Flux guidance of equal magnetic field sensors' These magnetic field sensors are integrated in the MRAM chip and can indeed share the same MRAM process. As shown in Figures 3 and 4, two The free layer of the pseudo MTJ element can be used as a flux guide to concentrate more flux into the sensor, thereby reducing the detection limit. In short, it has been found through simulation that these pseudo devices are used As a flux guide, if 98455.doc -19- 200528739 is a magnetic field sensor, the flux density can be increased by 53%. For a current sensor (which uses the same method as in MRAM to generate a magnetic field), it can be improved Approximately 10%. Figure 3 illustrates a cross-sectional view of the sensing element 21o and the dummy device used as the flux guide 120. These layers correspond to the layers of the memory cells shown in FIG. There is a sensing layer (free magnetic layer 200), an insulating layer (tunnel barrier 200), a clasping magnetic layer 204 and an antiferromagnetic layer 206, which are used to magnetize the clasping layer, Pegging, in a fixed direction (exchange bias direction) In the example shown in this figure, there is also a bottom layer 208 as the lower contact layer. For simplicity, the magnetic tunnel junction interface (MTJ) stack shown in Figure 3 Only 4 active layers are shown in it. In fact, there can be more layers (which are not related to the operating principle). The sensing layers of the two adjacent pseudo MTJ devices can be regarded as the role (real) between the two devices The flux guidance of the sensor. This design allows the flux guidance of the magnetic field sensor to be generated without any additional processing steps. The free layer of the two pseudo-devices can be rotated freely because of their magnetization, so it is now used as a flux The guidance layer of these pseudo devices is fixed because of its magnetization, so it will not affect the flux change during the measurement. Figure 4 is the top view. In order to produce better efficiency and homogeneity between the two pseudo devices Magnetic field, preferably the two pseudo devices (used as flux guidance) are larger than the sensor. The permeability of the equal flux guidance can be adjusted by changing its geometry. The aspect ratio and size of these pseudo-penetrations should be selected so that they are not fully saturated in the range of the magnetic field to be measured, but still large enough Permeability to concentrate the flux. The pattern of flux guidance should be as close to the sensor as possible, but still isolated from the sensor. The closer they are, the more efficient the flux guidance. Of course The small interval is determined by the lithographic resolution and the last name engraving technology. Although there is a small interval between the Tongdan 98455.doc -20- 200528739 and the sensor, an upper contact layer is above the sensor The position of is not important. In fact, it may have some tolerance. That is, if the upper contact layer of the sensor is in contact with one of the flux guides, the flux guides are not connected at the bottom. , So it usually does not cause any problems. If used as a built-in or integrated current sensor using MRAM technology, the current can be in the same lower metallization layer as the word line or bit line used for MRAM Conductor pattern. The conductive system is galvanically isolated from the MTj device. When transmitting When the current to be measured passes through the line, the current generates a magnetic field around the line itself, and the MTJ device will sense the magnetic field. The resistance change of the MTJ device indicates the magnetic field, and thus the current. Figures 5 to 10, simulation Results: The results of simulations using finite element technology are illustrated in these figures. These simulations were performed for two examples: magnetic field sensors and current sensors, both of which differed in whether they had or did not have such flux derivatives. The results are shown in Figure 5. Figure 5 illustrates the χ component (magnetic stress) with and without flux guidance for the flux density across a magnetic sensor in both examples. Here the χ direction represents and senses The horizontal plane of the sensor is parallel. It can be clearly seen that with such flux guidance, the flux density in the sensor has increased dramatically by 53%. In the simulation, for simplicity, only the free layer and Such flux guidance. It is assumed that the sensing layer (free layer) of the sensor and the magnetic permeability of the flux guidance are about 2000 and 1,000 respectively. This is an actual value and depends on the flux guidance and the geometry of the sensor. The sensor width is 1 为, the flux guide width is 5㈣, and the interval between them is 100_. The thickness of the free layer is 5 nm (which is typically used for mram stacks: g: medium). In the direction of the horizontal plane of the free layer, a quantity of 50e and a magnetic field with the same mass as 98455.doc -21-200528739 are applied. Fig. 5 (modified version): Fig. 6 illustrates the magnetic field line image calculated by the prior art if the sensing element 2 is not provided with flux guidance. FIG. 7 illustrates a corresponding image for an embodiment with flux guidance 120 according to the present invention. For comparison, the same density scaling is used in both images. For a sensor with flux guidance, it is obvious that the magnetic% lines are concentrated on the sensing layer of the sensor due to the flux guidance. In Fig. 7, for the sake of brevity, only a part of such flux guidance is shown, and the temple guidance is extended in both directions. Figures 8 and 9 illustrate corresponding images of magnetic field lines of a current sensor with and without flux guidance, respectively. The current sensor example uses the same geometry and parameters. Among them, a homogeneous magnetic field is not applied to the sensor, but a conductor 13 is placed at a distance of 150 nm below the sensor, and its cross section is 300 nm (South) xl500 nm (transverse). Passing a 15 mA current through the conductor (in a direction perpendicular to the figure), a circular magnetic field is generated around the conductor itself. At this sensor position, a magnetic field of about 50e is generated, which is compatible with the examples in Figs. Similarly, in Fig. 9, for brevity, only a part of the flux guidance is shown, and the flux guidance actually extends further in both directions. < It can be seen in the image that the improvement is not as obvious as shown in FIGS. 6 and 7. The reason is that only a small part of the-part of these fluxes is close to the device, where the near-horizontal flux line actually contributes to the flux guiding effect. The detailed simulation data shows that compared with the case without flux guidance, the flux density inside the sensor with flux guidance increases by about 丨, as shown in Figure 1 Flux density across the sensor width 98455.doc 200528739 at different distances. In addition, such flux guidance can be combined with current line milling (as disclosed in World Patent No. 02/41367 A2), which is commonly used for MRAM bit / bit lines. In this example, the effect of such flux guidance has increased by about 0%, surpassing the improved magnetic field generated by the conductor wire with a lapping layer. Other effects: In a sensor without flux guidance, the extended shape of the sensor causes a shape anisotropy: Kd = 1/2 μ0 Μ? (Νχ · Νγ), where Ms is the layer Saturation magnetization, NX & NY is the demagnetization factor of the horizontal and vertical, respectively. This shape anisotropy tends to align the magnetization of any magnetic layer of the sensor to the longitudinal direction. The larger the aspect ratio of the 70 pieces, the stronger the shape anisotropy becomes. For sensors with a width of about 1 micron, an extended shape with an aspect ratio (AR) of 5 to 7 is desired to stabilize the free layer and thereby reduce the hysteresis phenomenon. However, for the clasp layer (its magnetization is purposefully fixed in the lateral direction), the shape anisotropy will be inclined to force the magnetization away from the clasp direction, which may cause the sensor to behave badly. By placing two larger dummy devices near both sides of the sensor, the demagnetization field of the pinch layer of the sensor H is reduced, thereby reducing the shape anisotropy. The reason is that the retaining layers of the two pseudo-devices are also fixed in the same direction as the retaining layers of the device, thereby making the fixing direction less adversely affected. Similar side; that is, the effect described in the free layer (sensing layer) will see the same reduction in shape anisotropy "隹, the effect is smaller because the free layer (usually made of nickel-iron alloy) The saturation magnetization of is smaller than the coercivity and magnetization of the pinning layer (usually made of CoFe). The free layer in this example behaves like its larger effective width, so it has a smaller dependency. However, The free layer 98455.doc -23- 200528739 has some shape anisotropy that is useful for suppressing hysteresis, but on the other hand the shape anisotropy tends to degrade the sensitivity of the sensor. Therefore, one should be found in each special application Compromise value. During the design of the geometry of the sensor with flux guidance, it is necessary to consider reducing the effective shape anisotropy of the free layer to obtain the desired shape anisotropy value. Although the embodiment refers to The magnetic sensor using the TMR effect, but the concept is not limited to this, but it can also be applied to GMR sensors, for example. Figures 11 to 15 apply flux guidance to enhance the writing of MRAM cells: The above can also be applied Same flux The principle of induction is to increase MRAM by increasing the flux density generated by the write current on the cell. This can save 20% of the power consumption during writing. The implementation of such flux guidance does not require any additional cost For the processing steps, it is only necessary to change the design of the mask. In addition, if the flux guidance can be placed in the unused space between the MTJ memory units 110, it is not necessary to increase the size of the unit. The views and cross-sections above 11 and 12 illustrate the prior art MRAM (corresponding to the design of Figure 1) without flux guidance. In the word (or digital) line direction, as long as each unit is large enough on the Shixi substrate The isolation transistor is hidden, and there is still no significant magnetic contact between the units. The units can be placed as close as possible. Setting the flux guide along this direction usually increases the size of the unit. However, along the position In the direction of the element line, some space is required for the bottom electrode through hole 150, so the spacing between these cells is significantly larger. This unused space can be used to place a pseudo MTJ as a flux guide. Figures 13 and 14 are based on this Description of the invention above view and sectional view one The first embodiment of the mram. This design corresponds to the design of FIG. 1 and is appropriately described using similar reference numbers. Each MTJ memory element (represented by MTJ in the figure) is sandwiched between two 98455.doc -24-200528739 pseudo MTJs (It serves as a flux guide (FG) 12). The pseudo MTJs are set in the bit line direction. The pseudo MTJs and the memory elements can be accurately patterned in the same way at the same time. Partly located The two flux guides above the same through hole can share the same bottom electrode as a memory element. However, the connection in the bottom electrodes must always be interrupted between the net cells. The bit line is through the top electrode through hole 160 (Which is exactly above the memory element) and connected to the top electrode of the memory element. Therefore, these pseudo MTJs are open circuits and will not affect the memory elements electrically. The space between the pseudo MTJs and the memory element should be as small as possible, for example, about OO nm or less (depending on technical capabilities). These pseudo MTJs should only occupy unused space between these components. Using these pseudo MTJs as flux guides, the write current (which generates a difficult-to-magnetize axial field in this example) transmitted to a digit line can be reduced by about 10% to get an example without flux guides. In the same magnetic field. In other words, the power consumption of this difficult-to-magnetize axial field (which is proportional to the square of the current) can be reduced by almost 20%. If the total power consumption for two magnetic fields (difficulty magnetized axial field and easy magnetized axial field) is considered, the total write power consumption can be reduced by about 10% (assuming that the bit line current is equal to the digital line current). The flux guidance (FG) (in the form of a pseudo MTJ) patterns a number of memory elements in the bit line direction. If it is unknown why a large space is needed between the memory elements of the digital line direction towel (for example, because a larger space is needed for the isolated transistor), it can be used by placing a flux guide in both directions This unused space. Wherein, the pseudo-units are placed in two directions (ie, difficult to magnetize in FIG. 15, a second embodiment, the axial field and the easy-to-magnetize axial field direction). In terms of other aspects, the design corresponds to the design of Figures 3 and U, and it is explained with appropriate reference numbers 98455.doc -25- 200528739. In this example, the currents on both the bit line and the digital line can be reduced by 1000/0, and the power consumption for writing in two magnetic fields can be reduced by 20%. Figures 16 to 21, flux guidance for two-dimensional sensors ... By introducing such sensors, the development of TMR sensors will be briefly explained. The integration of MRAM into CMOS technology has created a new opportunity to include magnetic field sensors in MRAM chips because MRAM and magnetic sensors are based on the same effect, the I1 tunneling magnetoresistance (TMR) effect. These magnetic sensors will use exactly the same technology as MRAM, with only different geometric designs; therefore, integrating magnetic sensors into MrAM chips will add additional value to MRAM chips without additional cost. Applications of the integrated magnetic sensor can include highly sensitive current sensing states for power pin testing or IDDx testing, external magnetic field sensors for MRAM write control, active protection with kMRAM, integrated circuits Compass and more. As mentioned above, magnetoresistive sensors can be based on one of these magnetoresistive effects, such as giant magnetoresistance (GMR) or tunneling magnetoresistance (TMR). A TMR sensor includes a sensing element, which is a magnetic tunnel junction (MTJ, as shown in Figures 4 and 5 above). Spring Figures 16 to 19 illustrate diagrammatically a magnetic field sensor based on the TMR effect. Similar reference numbers are used for proper explanation. The magnetization of the free layer 200 is relatively free to rotate according to the applied magnetic field. Instead, the magnetization of the pinned layer always points in a fixed direction (X direction in this example). The free layer has an easy magnetization axis in the y-direction to the direction (usually by a pattern of lines that will extend in the y-direction). If no magnetic field is applied, the magnetization of the free layer (M stomach *) will stop in the direction of the :: axis. When a magnetic field appears in the 乂 direction, Μ "rotates away from the Y ° and rotates toward the direction of the magnetic field (X direction). The stronger the magnetic field, the smaller the angle _ 98455.doc -26- 200528739 Angular angle between the M14 x directions) The resistance of the MTJ (measured by sensing the current perpendicular to the stack from the top private electrode to the bottom electrode of the MTJ—sense current) varies according to the following exchange function. R (0) / R (〇) = 1 + MR * (1-cos ^) / 2 where R⑺ and R (0) are the resistances of the MTJ when Θ is non-zero and zero respectively; MR is the magnetoresistance ratio (usually about 30% to 40%). A TMR sensor is usually designed to operate properly when the direction of the magnetic field to be measured is perpendicular to the easy magnetization axis of the free layer (that is, in the direction in which the magnetization is held). This direction is called the sensing direction of the sensor 1 In addition to the current sensor, its application is essential-the ability to measure 2D magnetic field debt, that is, the sensor system must provide information about the projection of the magnetic field vector on the horizontal plane of the chip. This magnetic field projection (hereinafter referred to as the plane Magnetic field H) can be decomposed Hx & Hy, each detected by a separate sensor. To measure the _2D magnetic field vector, two Orthogonal sensors. Each sensor is only sensitive to the-component (Ηχ or Η force) of the magnetic field, and the combined signal from the two components provides information about the direction and magnitude of the plane magnetic field. In the measurement device, the sensing direction is determined by the magnetization direction of the pinning layer. As mentioned above, this layer is pinned by the antiferromagnetic layer below. This pinning effect is called exchange ㈣. The stack is annealed to several hundred degrees Celsius, and then the cooling period is set in the direction of the exchange bias f during a magnetic field. After cooling and removing the magnetic field, the exchange bias is two: two: first: the direction of the applied magnetic field Medium. This step can be completed after the W of the sensing Yuan Tian ^ pattern. Because the sensors are from the same multi-layer stack Fengding pattern 'so they usually have the same exchange bias direction. In the current market 9S455. doc -27- 200528739 GMR sensor products sold, the only way to generate a set of two orthogonal sensors for 2D magnetic field detection is to combine two separate orthogonal chips on a PCB 'or on the same chip Combining two separate orthogonals in a package In order to integrate two orthogonal sensors to the same substrate, it is necessary to anneal the separate sensors. This method is extremely difficult or impossible in general. Especially if these sensors and MRAM Integration, this processing step cannot be achieved at present, because it is only possible to have a bias bias direction for MRAM. For GMR sensors, a way to locally change the bias bias direction is when a magnetic field in a desired direction appears By transmitting high current pulses through some selected sensing strips of 'Ling Kao KM. Lenssen's " Magnetoresistive sensors and meiT10ry (magnetoresistive sensors and memory),' NATO ASI Article, 2001 1 2 The heat generated by the current will reset the exchange bias direction of the selected sensor to a new direction, while the remaining sensors remain intact (retained in the previously set white). The method is initially used in Wheatstone (Wheatstone ) The relative exchange bias direction of different sensors is set in the bridge. This method can only be used for non-monthly GMR sensors that sense 'to state' and for barrier-free only. For TMR sensing, it is not appropriate to transmit this high current (and therefore high voltage), because the device valley is susceptible to damage when exposed to a voltage greater than about vv. Xuan Jiang. A simple way to make a pair (or several pairs) of integrated TMR sensors is to use a gray 2D magnetic field measurement (such as a 2D external magnetic field measurement) or an electronic compass. The pseudo-device flux guidance is used to set the pattern at different angles: to decompose the magnetic field vector into 2 components, and the exchange bias of these sensors is still parallel. This means that the device does not require locally different exchange bias directions and therefore does not require additional 98455.doc 200528739 processing steps beyond the steps used to make an MRAM. Therefore, this method is very suitable for sensor integration in an MRAM chip. The same flux guiding principles used in relation to Figures 1 to 15 above are used. However, in this example, the flux guidance is used not only to increase the flux density, but also to decompose the planar magnetic field into two rotational components, each of which is detected by a separate sensor. These senses | § have the same exchange bias direction, but they fit into flux guidance with different orientations. The principle of the method is illustrated in Fig. 19, which is a schematic top view illustrating the 2D magnetic sensors A and B and the associated flux guidance. Sensor A (310) and sensor b (320) have the same exchange bias direction in the X direction and have an easy magnetization axis in the y direction, whereby they are designed to sense optimally in the X direction. Each sensor has two parallelogram-shaped flux guides 300 on its left and right sides. The two sets of sensors_flux guidance are separated from each other 'enough to avoid mutual influence. These flux guides are pseudo-ribbed garments' so that their 疋 pattern can be the same as these working sensors. Sensing each side of the flux of Jieba has two sides parallel to the longitudinal direction of the sensor, and the other two sides are in the Υ ′ direction, which forms an angle α with X. In this example, α is 45. (But it can also be a different value). Each flux of the sensor B is guided with 2 sides in the X direction with the longitudinal plane of the sensor and 3 2 sides, which form an angle with the parent. Generally, the flux guidance of the two sensors is a mirror image in the χ direction. In order to obtain equal and symmetrical signals from the sensors, the mirror image of the flux guidance is necessary. To prevent D Hai et al. ’S guidance at zero magnetic field saturation (where the magnetic moment is preferably oriented in the-direction). The flux guidance of these parallelograms must have a small layer of anisotropy. An important source is the shape. ! · (Caused by the extended shape of 4 layers). Therefore the flux-guided shape 98455.doc -29- 200528739 = cannot be extended in any direction (obviously). In addition, such moments should be divided at zero magnetic field There are several fields. This can be done by making the guides large enough (at least a few microns on each side). '' When an applied planar magnetic field H appears, the flux guides are concentrated in their longitudinal direction. And generate magnetic charges in the two measurements of the sensor (see the dry example of Sensor A in Figure 19). These charges therefore generate a tonally line in the χ direction (which is the same as the sensing direction of the sensor). .Because these flux lines would rather move privately along the inclined side of these flux guiding spoons, This flux is strongest when the magnetic field is parallel to this direction (in this example, this is the y, direction of sensor A). Obviously, the flux guidance acts as a flux redirector to rotate the sensing The sensing direction of the sensor is from the sensed X direction to the y, direction. Similarly, the sensing is most sensitive to the magnetic field facing the χ, direction. When both sensors are in use, a flat magnetic field Η seems to be: Resolve into two orthogonal directions (ie, X, and y,). The signals from these sensors are unambiguously recorded by an electronic circuit, which then merges and provides relevant information (such as direction and magnitude) of the magnetic field vector Etc.) The behavior of these sensors would appear to have two orthogonal sensors. Phenomenon is an ideal example where it has been assumed that the magnetic field generated by the flux guidance is stronger than the background magnetic field. In fact, the situation A little different. See you in fact

到該感測元件有兩磁場：-個由該等通量導引產生（其在X =向中）另一個為該平面磁場H的背景磁場（其未由該等通導T放大。期待後者明顯小於前者。因此，該感測器内側取終形成的通量未朝向X方向，但依所施平面磁場的方向而（稍）傾斜。惟，稍後的模擬將說明仍可從該等感測器得到不含糊的信號。 98455.doc •30- 200528739 在其他應用中，當僅使用該等通量導引以增加通量密度 ^ a等通里導引與該感測器間的間隙應儘可能地小。惟在本毛明亥等通量導引係作為通量轉向器，而非僅作為通星本中时因此，較佳該感測器由磁靜電交互作用而非交換交互作用㈣合至該等通量導引。原因是習知交換福合係用以利用極小變動而強力地維持該等磁性力矩的方向。若該感測器與該等通量導引之間有一些交換耦合，則该等通ϊ線在進入該感測器時會寧可在該等通量導引内側移動時繼、續維持在相同方向(即，分“口在感測器A的，方向）。當然，此將減低該感測器的有效性。模擬結果（未顯示）已證實此爭議。該感測器與該等通量導引間的間隔應依此調整，俾打斷該交換耦合，但該磁靜電交互作用仍強。實際上，最小間隔將約為1〇至2〇 nm。在此節中，使用有限元件方法以模擬一對使用平行四邊形通量導引的感測器範例。在該模擬中，為求簡化，僅考 ΐ遠自由層。該感測器大小為6X1 μιη2。該等通量導引具有兩個45。的傾斜側，及側邊具有6/xm及8 μχη的尺寸。該等通量導引與該感測器間的間隔為〇·2 μηι。一 2 〇e的同質磁場施至該糸統中。圖20a及20b說明當磁場朝〇。及45。方向時感測器a的模擬磁性通量線影像範例。該等影像顯示該等通量導引的確將該等通量線轉向靠近在該感測器位置的X方向。感測器B的通量影像類似，但在X軸上成鏡像。已計算在不同施加磁場方向在感測器A及b中心的x方向 98455.doc -31 - 200528739 中的通1松、度（感應B)。因該等感測器對χ方向中的磁場敏感，因此若該感測器在此磁場範圍内的特性係線性（此假設係真貝的）’則该感測器的信號與此方向中的磁場成正比。口此，u玄等感測态信號的曲線（磁阻性）會類似於圖2 1所示的 Β與角度的相對曲線。圖21顯示該等如曲線（類似於信號曲線）像正弦波，並互相為移相。此確實為該設計目的。在理想例子中，相位差應為90。，意即該平面磁場可由一正交座標系統及〆）加以判疋，或將該磁場分解成2個正交分量。惟，如上述，並非總是如此。實際上，該等通量線橫越該感測器移動的方向並非總與X方向平行，俾在該等信號之間造成較小的相位差 (小於9〇°)。在提供的範例中，實際的相位差約65。。意即該座標系統屬非正交。然而，來自兩感測器的信號仍不含糊，此仍容許在0至360。的全角度範圍上判定施加磁場的方向可使用習用合併電路結構（或使用如微處理器或檢視表 T習用數位電路結構）將圖21所示的兩信號轉換成一角度總言之，已說明藉由包含不同幾何形狀的通量導引用於兩個分開的感測器以建構2D磁場感測器的方式“亥等通旦導引作為通量轉向器、1以將該等感測器的感财向= 蝴不同方向。來自該等感測器的數個合併信號提供測置磁場向量的相關資訊。此設計的優點為兩感測器並不要正交交換偏壓方向’藉此該等感測器不用額外的處理：驟而可整合在該MRAM晶片中，或換言之， " 成寺感測器可 98455.doc 200528739 元全與MRAM製程相容。該等通量導引事實上可為偽壯置，意即它們可用與該等感測器及MRAM元件相同的= 式，在相同時間定圖案。因此，將該等通量導引定圖案亦不需花費額外步驟。有限元件模擬已證明依此構成的兩感測器確實可提供不含㈣㈣，其可用以判定㈣磁場的角度及大小。明顯不像其他的通量導引設計，應依此構成該等通量導引’俾壓制在該等通量導引與該感測器間的交換耦合。在該等通量導引與該感測方向(χ方向)之間，該等導引的角度不需侷限於45。（雖⑽。可能是最適角度）。兩感測器的通量導引不需成鏡像（雖然鏡像是得到料感測器同等及對稱行為的最適配置)。具有如上述特殊形狀的通量導引亦可由軟磁材料的分開層定圖帛，而非使用該等偏裝置的自由層。在此例中，可使用涉及一些額外步驟的習用技術以製造該等通量導引。在此例中’此方式仍有不需改變交換偏壓方向以形成一組正交感測器的優點。實際上’該科行四邊料量導引的线轉角可造成問題（即其可作為扣住中外此意即接近此等轉角的磁性力矩會被釘住’難以在磁性反轉過程期間旋轉。結果是其可增加該等通量導引的抗磁性（係不良結果）。輕易解決此問題的方式為將該等尖銳轉角弄圓。微磁模擬(未顯示)已證明圓轉角會壓制好該釘住效應。圓形將不會明顯影響該等通量導引的導引效應。可使用任何合適習用電路結構作為測㈣感測元件電阻 98455.doc -33- 200528739 的偵測電路結構，以適應該應用。用於MRAM的讀出電路為眾所熟知，亦可用於感測器。通常使用一運算放大器 (op-amp)經由一偏壓電晶體以放大橫越一負載電阻所見的電壓，该電阻與該感測器串聯。在該感測器上的偏壓係定位在一較固定值（約200 mV)，及該感測器電阻中的變動造成電流中的變動，其造成該負載電阻上的電壓變動。接著放大此電壓變動。該電路的缺點為其在改變該感測器電阻日可造成該定位電壓的一些變化。美國專利號6,2〇5,〇73 B1 在MTJ記憶體讀出的内文中揭示一改良式電路。在此設計中’亦將一偏壓控制運算放大器輸出饋至偏壓電晶體的輸出。由橫越該感測器的電壓饋入該偏壓控制運算放大器的負輸入。在此設計中，該偏壓控制運算放大器的負回饋容許在該感測器上主動將該電壓定位，其可提供較穩定信號及較快速讀出時間。結論：上述電流感測器可實施於許多種類的積體電路中，尤其是CMOS電路及MRAM電路。此類感測器的輸出可跟著已建立的實施而耦合在掃描鏈中，用以將許多感測器輸出多工至該積體電路的至少一輸出上。此類積體電路可用於習用客戶設備中，尤其是膝上型電腦、行動電話等行動裝置。如上述’用以债測磁場強度的感測器具有使用磁性隨穿接面的一感測元件，以及一偵測電路結構，該感測元件具有一隨磁場變化的電阻，該感測元件包括一隧穿接面，及該情測電路結構配置成偵測流經該隧穿接面的隧穿電流。如 98455.doc -34- 200528739 l長等形狀各向異性正交於該磁場。該感測器可具有一磁阻感測7L件，其具數層磁性材料及至少一通量導弓丨，用以將4磁场集中至該感測元件上，該通量導引包括用於該感測元件的相同層的一部分。藉由使用該通量導引的同或數）層，該導引層可與該感測元件的對應層在同一步驟中形成。此類感測器可整合在MRAM晶片中。用於兩並聯感測兀件中之各一者，可使用通量導引不同地旋轉該磁場，以使2D感測器致能。由於該兩感測器茲不需正交交換偏壓方向’可較輕易將兩感測器整合。通量導引亦可用以集中寫入MRAM單元的磁場，並因此減低寫入電流。可預見在後附請求項的範圍内會有其他變化。【圖式簡單說明】藉由參照至附圖將較佳了解本發明特點，附圖說明本發明數個較佳實施例。其中：圖1說明一習知MRAM設計；圖2說明具通量導引的一習知感測器；圖3及4以圖說明本發明的第一實施例；圖5說明一實施例的通量密度剖析資料；圖6及7說明感測器周圍磁場線的影像；圖8及9說明導體及感測器周圍的磁場線；圖10說明另一實施例的通量密度圖；圖11及12以圖說明一習知MRAM ; 圖13及14根據本發明另一實施例以圖說明一MRAM ; 圖15以圖說明本發明另一實施例； 98455.doc -35- 200528739 圖16至18以圖說明一TMJ感測器；圖19、20a及20b根據本發明另一實施例以圖說明感測器，以判定一磁場方向；及圖2 1根據本發明另一實施例說明來自第一及第二感測元件的通量密度。【主要元件符號說明】 100, 200 自由磁層 102, 202 絕緣層（隧穿障壁） 104, 204 扣住磁層 106, 206 反鐵磁（AF)層 108, 208 底層 110 MTJ記憶單元 120, 300 通量導引 130 導體 160 通孔 210 磁阻感測元件 310, 320 感測器There are two magnetic fields to the sensing element: one is generated by the flux guidance (which is in the X = direction) and the other is the background magnetic field of the plane magnetic field H (which is not amplified by the conduction conductance T. Expect the latter It is significantly smaller than the former. Therefore, the flux formed on the inside of the sensor does not face the X direction, but is (slightly) tilted according to the direction of the applied planar magnetic field. However, later simulations will show that these sensors can still The sensor gets an unambiguous signal. 98455.doc • 30- 200528739 In other applications, when only such flux guidance is used to increase the flux density ^ a gap between the isotropic guidance and the sensor should be As small as possible. However, when the flux guidance system such as Ben Maominghai is used as a flux redirector, rather than only in the star book, therefore, it is better that the sensor be interacted by magnetostatic rather than exchange. Coupled to such flux guidance. The reason is that it is known that the exchange system is used to maintain the direction of the magnetic torque with a small change. If there is some between the sensor and the flux guidance, Exchange coupling, the communication lines would rather The inner side of the guide is continuously and continuously maintained in the same direction (that is, the direction of the sensor A is the same). Of course, this will reduce the effectiveness of the sensor. Simulation results (not shown) have confirmed this Controversy. The interval between the sensor and the flux guide should be adjusted accordingly, interrupting the exchange coupling, but the magnetic-electrostatic interaction is still strong. In fact, the minimum interval will be about 10 to 2 In this section, a finite element method is used to simulate a pair of sensors using parallelogram flux guidance. In this simulation, for simplicity, only the far free layer is considered. The size of the sensor is 6X1 μιη 2. The flux guide has two inclined sides of 45 °, and the sides have dimensions of 6 / xm and 8 μχη. The interval between the flux guide and the sensor is 0.2 μm A 20e homogeneous magnetic field is applied to the system. Figures 20a and 20b illustrate examples of simulated magnetic flux line images of sensor a when the magnetic fields are oriented at 0 ° and 45 °. These images show the communication The volume guidance does turn the flux line closer to the X direction at the sensor position. The flux image of the sensor B is similar, but it is mirrored on the X axis. The x directions of the sensors A and b in the direction of the applied magnetic field have been calculated. 9855.doc -31-200528739 B). Because these sensors are sensitive to the magnetic field in the χ direction, if the characteristics of the sensor in this magnetic field range are linear (this assumption is true), then the signal of the sensor is in this direction. The magnetic field is proportional to the magnetic field. In this case, the curve (magnetoresistance) of the sensing state signal such as uxuan will be similar to the relative curve of B and the angle shown in Figure 21. Figure 21 shows the contour curve (similar to the signal (Curves) are like sine waves and are phase-shifted to each other. This is indeed the design purpose. In an ideal example, the phase difference should be 90. That means the plane magnetic field can be judged by a system of orthogonal coordinates and 〆), or This magnetic field is decomposed into two orthogonal components. However, as mentioned above, this is not always the case. In fact, the direction in which the flux lines traverse the sensor is not always parallel to the X direction, causing a small phase difference (less than 90 °) between the signals. In the example provided, the actual phase difference is about 65. . This means that the coordinate system is non-orthogonal. However, the signals from the two sensors are still unambiguous, which is still tolerated between 0 and 360. To determine the direction of the applied magnetic field over the full range of angles, a conventional combined circuit structure (or a conventional digital circuit structure such as a microprocessor or look-up table T) can be used to convert the two signals shown in FIG. 21 into an angle. In summary, it has been explained that A way of constructing a 2D magnetic field sensor by using flux guidance with different geometries for two separate sensors "Hai et al. Guided as a flux redirector, 1 to sense the sensitivity of these sensors Financial direction = different directions. Several combined signals from these sensors provide information about the measured magnetic field vector. The advantage of this design is that the two sensors do not exchange the bias directions orthogonally. The sensor does not require additional processing: it can be integrated into the MRAM chip, or in other words, " Chengsi sensor can be 98455.doc 200528739 yuan compatible with the MRAM process. Such flux guidance can actually be Pseudo-stabilization means that they can be patterned at the same time using the same = pattern as these sensors and MRAM elements. Therefore, no additional steps are required to pattern the flux guidance pattern. Finite element simulations have been Prove this structure The formed two sensors can indeed provide no ㈣㈣, which can be used to determine the angle and size of the ㈣ magnetic field. Obviously unlike other flux guidance designs, such flux guidance should be constituted accordingly. The exchange coupling between isoflux guidance and the sensor. Between the flux guidance and the sensing direction (χ direction), the angle of the guidance need not be limited to 45. (Although ⑽. Probably the most suitable angle.) The flux guidance of the two sensors does not need to be mirrored (although mirroring is the most suitable configuration to obtain the same and symmetrical behavior of the material sensor). The flux guidance with the special shape as described above can also be soft magnetic Separate layers of material are used instead of free layers using such biasing devices. In this example, conventional techniques involving some additional steps can be used to make such flux guidance. In this example, 'this approach is still There is an advantage of not needing to change the direction of the exchange bias to form a set of orthogonal sensors. In fact, the line corners of the bank's four-sided material volume guide can cause problems (that is, it can be used to hold the Chinese and foreign means that it is close to this The magnetic moment of the equal angle will be pinned 'difficult in magnetic Rotation during the sexual reversal process. The result is that it can increase the flux-guided diamagnetism (a bad result). The easy way to solve this problem is to round these sharp corners. Micromagnetic simulation (not shown) has been Prove that the round corners will suppress the pinning effect. The round shape will not significantly affect the guiding effect of such flux guidance. Any suitable conventional circuit structure can be used as the resistance of the radon sensing element 98455.doc -33- 200528739 The detection circuit structure is suitable for this application. The readout circuit for MRAM is well known and can also be used for the sensor. Usually an op-amp is used to amplify the crossover through a bias transistor. A voltage seen by a load resistor, which is connected in series with the sensor. The bias voltage on the sensor is positioned at a relatively fixed value (about 200 mV), and the change in the sensor resistance causes the current to , Which causes the voltage on the load resistance to change. Then zoom in on this voltage change. The disadvantage of this circuit is that it can cause some changes in the positioning voltage when the resistance of the sensor is changed. U.S. Patent No. 6,205,073 B1 discloses an improved circuit in the content read from the MTJ memory. In this design, the output of a bias-controlled op amp is also fed to the output of the bias transistor. A negative input to the bias-controlled operational amplifier is fed by a voltage across the sensor. In this design, the negative feedback of the bias-controlled operational amplifier allows the voltage to be actively positioned on the sensor, which can provide a more stable signal and faster readout time. Conclusion: The above current sensors can be implemented in many types of integrated circuits, especially CMOS circuits and MRAM circuits. The output of such sensors can be coupled in a scan chain following established implementations to multiplex the output of many sensors to at least one output of the integrated circuit. Such integrated circuits can be used in conventional customer equipment, especially mobile devices such as laptops and mobile phones. As described above, the sensor for measuring the strength of a magnetic field with a debt has a sensing element using a magnetically-connected interface and a detection circuit structure. The sensing element has a resistance that changes with the magnetic field. The sensing element includes A tunnel junction and the circuit structure are configured to detect a tunneling current flowing through the tunnel junction. Such as 98455.doc -34- 200528739 l anisotropy of length and other shapes orthogonal to the magnetic field. The sensor may have a 7L piece of magnetoresistive sensing, which has several layers of magnetic material and at least one flux guide bow for concentrating 4 magnetic fields on the sensing element. The flux guidance includes Part of the same layer of the sensing element. By using the XOR layer of the flux guide, the guide layer can be formed in the same step as the corresponding layer of the sensing element. Such sensors can be integrated in MRAM chips. For each of the two parallel sensing elements, the magnetic field can be rotated differently using flux guidance to enable the 2D sensor. Since the two sensors do not need to exchange the bias direction orthogonally, the two sensors can be integrated more easily. Flux steering can also be used to focus the magnetic field written into the MRAM cell and thus reduce the write current. Other changes are foreseen within the scope of the appended claims. [Brief description of the drawings] The characteristics of the present invention will be better understood by referring to the accompanying drawings, which illustrate several preferred embodiments of the present invention. Among them: Fig. 1 illustrates a conventional MRAM design; Fig. 2 illustrates a conventional sensor with flux guidance; Figs. 3 and 4 illustrate a first embodiment of the present invention; Fig. 5 illustrates a conventional embodiment of the present invention; Volume density analysis data; Figures 6 and 7 illustrate images of magnetic field lines around the sensor; Figures 8 and 9 illustrate conductors and magnetic field lines around the sensor; Figure 10 illustrates a flux density diagram of another embodiment; Figures 11 and Figure 12 illustrates a conventional MRAM; Figures 13 and 14 illustrate a MRAM according to another embodiment of the present invention; Figure 15 illustrates another embodiment of the present invention; 98455.doc -35- 200528739 Figures 16 to 18 Figures illustrate a TMJ sensor; Figures 19, 20a, and 20b illustrate the sensor according to another embodiment of the present invention to determine a magnetic field direction; and Figure 21 illustrates from the first and Flux density of the second sensing element. [Description of main component symbols] 100, 200 Free magnetic layer 102, 202 Insulating layer (tunneling barrier) 104, 204 Detach magnetic layer 106, 206 Antiferromagnetic (AF) layer 108, 208 Bottom 110 MTJ memory unit 120, 300 Flux guidance 130 conductor 160 through hole 210 magnetoresistive sensing element 310, 320 sensor

98455.doc 36-98455.doc 36-

Claims

200528739 十、申請專利範圍： !，一種磁性感測器，用以感測磁場並具有一磁阻感測元件°亥凡件具有數層磁性材料及至少一通量導引，用以將磁場隼巾5 、一木中至该感測元件上，該通《導引包括用於該感測兀件之至少—相同層之一部分。 ★二求項1之感測器，該感測元件包括一隧穿磁性接面。士明求項1或2之感測器，該通量導引具有大於該感測元件之面積。 4·如μ求項1或2之感測器，在該感測元件之數個對面側上具有數個通量導引。长貝1之感心，該等通量導引大小在待測磁場範中未完全飽和。 6. 8. 9. 10· 11. 12. 如2求項1之感測器，該感測元件具有一延長形狀。二明求項2之感測器，該感測元件係平面的，及該隧穿電流係垂直於該平面元件之水平面而加以偵測。如請求項2之感測器’該感測元件配置間具有-關係，其大體上顯示無磁滞現象。山之如請求項1之感測器’該感測元件具有一自向大體上垂直於待測磁場之方向具有-易磁化轴。如凊求項!之感測器’言亥感測元件包括一扣有與該自㈣層U磁化㈣_角度之磁化。…、如晴未項1〇之感測器’其中該角度大體上為直角。Γ種磁性感測器，具有：—第—感測元件，用以測量- 弟一方向中之磁場；及一筮— 弟一感測元件，用以測量一不 98455.doc 200528739 同方向中之磁場，及—通量導引，用於該等感夠元件之至少—者，該通量導引配置成經由該感測元件旋轉該磁場之方向。 ~ 13 ·如明求項i 2之感測器，該感測器具有不同之通量導引，其用於各感測元件具有不同幾何形狀。 14·如明求項13之感測器，該通量導引配置成在各感測元件之不同方向中旋轉該磁場。 15·如請求項12之感測器，其具有電路結構，配置成合併來自該等感測元件之輸出，以提供該磁場向量之相關資訊。 16. 如請求項12之感測器，該等第一及第二感測元件及其通量導引配置成偵測該磁場之正交分量。 17. 如請求項12之感測器，該等通量導引各包括至少一側，其疋向於一不同於該感測元件之感測方向之角度。 18·如請求項17之感測器，其中該等通量導引係平行四邊形。 19·如請求項12之感測器，該等感測元件係整合在相同晶片上。 20·如請求項12之感測器，用於該等感測元件二者之一之數個通量導引係平行地定向。 21 _如請求項15之感測器，該電路結構配置成考量未由該等通量導引旋轉或放大之背景磁場總量。 22· —種電流感測器，用以感測一導體中之電流，該電流感測器包括如先前請求項中任一項之感測器，其設置成偵測該電流造成之磁場。 23·如請求項22之電流感測器，在平行於該導體寬度之方向 98455.doc 200528739 中，該感測元件寬度係小於該導體之寬产。 24· —種積體電路，具有—内涂：兩、、六武鬥逯式迅抓感测器，其具有如社求項22或23之電流感測器。 μ 25·如請求項24之積體電路，嗜雷泣片 Θ电抓感測為配置成感測靜止電流（IDDQ)。 26. 如請求項24之積體電路，Α呈右力具具有在一知描鏈中鏈結之多個電流感測器。 27. -種客戶設備，具有如請求項24至26中任—項之積路。 28. -種使用感測元件感測磁場之方法，該感測元件且有一磁性随穿接面’該接面具有一隨該磁場變化之電阻，，磁場由-通量導引加以集中，該方法特徵為偵測流過： Ρ遂穿接面之隧穿電流之步驟。 29. -種製造如請求項^中任—項之感測器之方法，該方法包括在相同步驟中形成一層感測元件及一層通量^ 之步驟。曰里导引 30. 記憶單元，該等單元中至用以集中用於寫入該單元一種磁性記憶體，具有至少一少一者具有至少一通量導引，之磁場。 31·如請求項30之磁性記憶體，該導引包括該記憶單元使之相同層或數層之一部分。％用 32·如請求項30或31之磁性記憶體，該等通量導引包括數個偽磁性記憶單元，其位於一對應記憶體元件之旁。 33.如請求項30之磁性記憶冑，該等通^導引與該對應記憶 98455.doc 200528739 單元之自由磁層共面。 34. 如請求項30之磁性記憶體，該單元包括一隨穿接面。 35. 如請求項3〇之磁性記憶體，數個通量導引係在該單兩側形成。 3 6 · —種製造磁性隨拉— 存取記憶體（MRAM)之方法，該方法包括在相同步驟中元之通詈道^ >成一層記憶單元及一層用於該記憶單導弓I之步驟。 98455.doc200528739 10. Scope of patent application:!, A magnetic sensor for sensing a magnetic field and having a magnetoresistive sensor element. Haifan has several layers of magnetic material and at least one flux guide to align the magnetic field. Towel 5, a piece of wood to the sensing element, the guide includes at least one part of the same layer for the sensing element. ★ The sensor of claim 2, wherein the sensing element includes a tunneling magnetic interface. The sensor of Shi Ming seeking item 1 or 2, the flux guide has an area larger than that of the sensing element. 4. A sensor such as μ seeking term 1 or 2, having several flux guides on several opposite sides of the sensing element. Changbei 1 feels that these flux guidance magnitudes are not fully saturated in the magnetic field range to be measured. 6. 8. 9. 10 · 11. 12. The sensor according to item 2 of item 1, the sensing element has an extended shape. The sensor of Erming Q2, the sensing element is planar, and the tunneling current is detected perpendicular to the horizontal plane of the planar element. For example, in the sensor of claim 2, there is a relationship between the sensing element configurations, which generally shows no hysteresis. The sensor of Shanzhi as claimed in claim 1, the sensing element has an orientation that is substantially perpendicular to the direction of the magnetic field to be measured and has an easy magnetization axis. The sensor of the above item! The sensing element includes a magnetization that is angled to the U magnetization angle of the self-propelled layer. …, Such as the sensor of Qingwei item 10 ′, wherein the angle is substantially a right angle. Γ magnetic sensor with:-the first-sensing element to measure-the magnetic field in the direction of a brother; and the first-a sensing element to measure a magnetic field in the same direction 98455.doc 200528739 A magnetic field and a flux guide for at least one of the sensing elements, the flux guidance is configured to rotate the direction of the magnetic field via the sensing element. ~ 13 · If the sensor of item i 2 is found, the sensor has different flux guidance, which is used for each sensing element to have a different geometry. 14. The sensor of claim 13, wherein the flux guidance is configured to rotate the magnetic field in different directions of each sensing element. 15. The sensor of claim 12, having a circuit structure configured to combine outputs from the sensing elements to provide information about the magnetic field vector. 16. As in the sensor of claim 12, the first and second sensing elements and their flux guidance are configured to detect orthogonal components of the magnetic field. 17. As in the sensor of claim 12, the flux guides each include at least one side which is oriented at an angle different from the sensing direction of the sensing element. 18. The sensor of claim 17, wherein the flux guidance is a parallelogram. 19. The sensor of claim 12, the sensing elements are integrated on the same chip. 20. The sensor as claimed in claim 12, wherein a plurality of flux guides for one of the sensing elements are oriented in parallel. 21 _As in the sensor of claim 15, the circuit structure is configured to consider the total background magnetic field that is not rotated or amplified by these flux guides. 22 · A current sensor for sensing a current in a conductor, the current sensor comprising a sensor as in any of the previous claims, which is arranged to detect a magnetic field caused by the current. 23. The current sensor of claim 22, in a direction parallel to the width of the conductor 98455.doc 200528739, the width of the sensing element is smaller than the width of the conductor. 24 · —A kind of integrated circuit, which has—inner coating: two, six, and six-bucket fast grab sensors, which have current sensors such as social item 22 or 23. μ 25. As in the integrated circuit of claim 24, the thundercry film Θ electric grasping sensing is configured to sense a static current (IDDQ). 26. As in the integrated circuit of claim 24, A is a right-handed tool having a plurality of current sensors linked in a chain. 27. A client device having a product path as claimed in any of claims 24 to 26. 28. A method of sensing a magnetic field using a sensing element, the sensing element having a magnetically-connected surface, the mask has a resistance that changes with the magnetic field, and the magnetic field is concentrated by a flux guide, the The method is characterized by the step of detecting the tunneling current flowing through: 29. A method of manufacturing a sensor as claimed in any of the items ^, the method comprising the steps of forming a layer of sensing elements and a layer of flux ^ in the same step. 30. Memory unit. These units are used to focus on writing into the unit. A magnetic memory with at least one magnetic field with at least one flux guide. 31. The magnetic memory of claim 30, wherein the guide includes the memory unit as part of the same layer or several layers. % Use 32. Magnetic memory as claimed in item 30 or 31. Such flux guidance includes several pseudo magnetic memory cells, which are located next to a corresponding memory element. 33. As in the magnetic memory of claim 30, the general guidance is coplanar with the free magnetic layer of the corresponding memory 98455.doc 200528739 unit. 34. The magnetic memory of claim 30, the unit comprising a threaded interface. 35. As in the magnetic memory of claim 30, several flux guides are formed on both sides of the single sheet. 3 6 · —A method for manufacturing magnetic pull-and-access memory (MRAM), which includes the same steps in the same steps ^ > forming a layer of memory unit and a layer for the memory single guide bow I step. 98455.doc