铁电存储器的技术背景通信类外文翻译、中英文翻译、外文文献翻译

1 铁电存储器的技术背景  概述  目前的存储器技术可以分为两种。第一种是非易失性存储器。传统上来说，他们被应用于只读存储器因为他们都有不易写入的特点。这些存储器均源于只读存储器 (ROM)技术 , 包括 EPROM, EEPROM, and Flash EPROM。  第二种是易失性存储器。易失性存储器包括 SRAM（静态存储器） 和 DRAM（动态存储器）。由于 RAM 类型的存储器易于 写入，因此它所保存的数据需要定时刷新。但由于用户易于写入这种 RAM 存储器 ，所以它是易失性。 可是它们同样会在掉电的情况下会失去所保存的数据。  铁 电存储器或是 FRAM 是一种比较完善的非易失性存储器。它是一种真正的非易失性存储器。 FRAM 存储器有 易于 写入和非易失性的优点，因此它能在断电情况下保存数据。 FRAM 产品可以保存数据达几千年。这种存储技术已经成为存储器的主流。这种存储技术可以简单的解释为对现在存储技术的概述。  什么是铁电存储器  相对于其它类型的半导体技术而言，铁电存储器具有一些独一无二的特性。传统的主流半导体存储器可以分为两类 -易失性 存储器 和非易失性 存储器 。易失性存储器包括静态存储器 SRAM（ static random access memory)和动态存储器 DRAM （ dynamic random access memory)。  SRAM 和 DRAM 在掉电时均会失去保存的数据。  RAM 类型的存储器易于使用、性能好，可是它们同样会在掉电的情况下会失去所保存的数据。   非易失性存储器在掉电的情况下并不会丢失所存储的数据。然而所有的主流的非易失性存储器均源自于只读存储器（ ROM）技术。  正如你所猜想的一样，被称为只读存储器的东西肯定不容易进行写入操作，而事实上是根本不能写入。所有由ROM技术研发出的存储器则都具有写入信息困难的特点。这些技术包括有 EPROM (几乎已经 停用 ）、 EEPROM 和 Flash。  这些存储器不仅写入速度慢，而且只能有限次的擦写，写入时功耗大。   2 铁电存储器能兼容 RAM 的一切功能，并且和 ROM 技术一样，是一种非易失性的存储器。铁电存储器在这两类存储类型间搭起了一座跨越沟壑的桥梁 -一种非易失性的 RAM。   基于 RAM 随机存储器的 FRAM 是利用铁电晶体的铁电效应实现数据存储。这是与其他非易失性存储器完全不同的机制，它是漂浮的门技术。铁电效应是铁电晶体所固有的一种偏振极化特性，与电磁作用无关。  当一个电场被加到铁电晶体材料时，铁电存储器 中的原子产生于电容器的两个电极板之间。这种电容器的构成与动态的随机存储器非常相似。不同的是存储数据不需向动态的随机存储器那样需要进行数据刷新，它是利用晶体机制进行数据存储的。这种晶体中心原子包含两种稳定状态：“ 0”状态和“ 1”状态。  由于它的基于随机存取储存器而设计的，因此它的读操作和写操作都很容易。但它和动态的随机存储器又有所不同，数据的存储状态是稳定的。因此，铁电存储器不需周期性刷新，即使在掉电的条件下， FRAM 仍能保存数据。  许多人都误解铁电这个名字 , 一个名字使用前缀 " ferro" 似乎暗示铁或 磁性 。铁电这个词也容易让人联想到铁磁。事实上，铁电存储器并没有用到 铁或磁性 的原理。他并没有受到外部磁场的影响，因为它同传统的动态随机存储器一样，操作使用的是电场。  铁电存储器的技术原理  当一个电场被加到铁电晶体时，中心原子顺着电场的方向在晶体里移动。  当原子移动时，它通过一个能量壁垒，从而引起电荷击穿。内部电路感应到电荷击穿并设置存储器。移去电场后，中心原子保持不动，存储器的状态也得以保存。铁电存储器不需要定时更新，掉电后数据能够继续保存，速度快而且不容易写坏。  铁电存储器技术和标准的 CMOS 制造工艺相兼容 。铁电薄膜被放置于 CMOS 基层之上，并置于两电极之间，使用金属互连并钝化后完成铁电制造过程。   3 Ramtron 的铁电存储器技术到现在已经相当的成熟。最初的铁电存储器采用两晶体管 /两电容器（ 2T/2C）的结构，导致元件体积相对过大。最近随着铁电材料和制造工艺的发展，在铁电存储器的每一单元内都不再需要配置标准电容器。  Ramtron 新的单晶体管 /单电容器结构可以像 DRAM 一样，使用单电容器为存储器阵列的每一列提供参考。与现有的 2T/2C 结构相比，它有效的把内存单元所需要的面积减少一半。新的设计极大的提高了 铁电存储器的效率，降低了铁电存储器产品的生产成本。   4 Ramtron 同样也通过转向更小的技术节点来提高铁电存储器各单元的成本效率。最近采用的 0.35 微米的制造工艺相对于前一代 0.5 微米的制造工艺，极大的降低了芯片的功耗，提高了单个晶元的利用率 。  所有这些令人振奋的发展都使得铁电存储器在人们日常生活的各个领域被广泛应用。从办公室复印机、高档服务器到汽车安全气囊和娱乐设施，铁电存储器不断改进性能在世界范围内得到广泛的应用。  铁电存储器的操作  一个简单的铁电晶体模型如图 1铁电存储器晶体的中心原子结构所示。在铁电晶 体中心有一个活动原子。在电场的作用下， 晶阵中的中心原子会沿着电场方向运动 到另一边 ， 反方方向的电场会使原子向着相反的方向运动。在晶体顶层和底层的原子保持稳定状态。当电场从晶体移走或是掉电的情况下，中心原子会保持在原来的位置。作为存储器件，铁电存储器是一种比较完善的存储器件。它包含了两种稳定的状态：一种是在无时间和能量的情况下不发生改变，另一种是在多变的外部环境下保持稳定。  读操作  虽然电容作为存储器件，但他不想线性电荷一样进行数据存储。要进行读操作，就要对存储单元电容中铁电晶体的中心原子位置进行记录。直接对中 心原子的位置进行检测是不能实现的。实际的读操作过程如下。在存储单元电容上施加一已知电 5 场（即对电容充电），如果原来晶体中心原子的位置与所施加的电场方向使中心原子要达到的位置相同，中心原子不会移动；若相反，则中心原子将越过晶体中间层的高能阶到达另一位置。在高能阶的作用下，充电波形上就会出现一个尖峰，把这个充电波形同参考位的充电波形进行比较，产生原子移动的比没有产生移动的多了一个尖峰，非开关电容产生普通的动态随机存储器的电荷，而开关电容则产生动态随机存储器和铁电存储器的混合电荷。存储电路决定了电容的切换。这种开关 电荷允许由电路决定存储电荷的状态。晶体原子状态的切换时间小于 1ns，完整的读操作的时间小于 70ns。  因为读操作导致存储单元状态的改变，需要电路自动恢复其内容，所以每个读操作后面还伴随一个“预充”过程来对存储器的状态进行恢复。虽然读操作被破坏，但存储无效的时间要低于 50ns。  写操作  写操作和读操作十分类似 。与其他的非易失性存储技术不同，写操作非常简单无需系统延时。数据被写到铁电的电容中。如果需要的话，新的数据很容易改变铁电晶体的状态。对于读操作， 晶体原子状态的切换时间小于 1ns，读操作的时间小于 70ns。 对于读操作，  “预充”操作伴随在  写操作之后。  FRAM存储单元结构  目前的 FRAM产品使用 2个场效应管和 2个电容（ 2T2C），每个存储单元包括数据位和各自的参考位，自 1993年起这种基本的单元已经被应用于产品中。 2T2C存储单          元提高了数据的可信度，特别是对于早期的 非易失性存储器是非常重要的。 2T2C存储单元结构如图 2所示。   6 图 2  2T2C存储单元结构  2T2C存储单元为每个数据位提供了一个相近的参考位，依照数据状态进行编程，读操作时一个电容会发生改变，而其它不发生变化，在设计存储器时 选择“ 0”或“ 1”任意状态。涉及到相应的存储器时， 存储电路 能非常精确地测量那个 变化和非变 化 电容器之间 不同。  存储队列中电容的变化被藉由从每一点点有差别的信号中除去。  2001年“单管单容”（ 1T1C）技术被投放市场，它使得铁电存储器产品的价格被提高。简化的 1T1C存储单元结构框图如图 3所示。                  图 3  1T1C存储单元结构框图  FRAM 的发展  正如前文所提到的，自从  1993年起基于铁电存储器 FRAM产品已经被广泛的应用于商业生产。在工业生产中，铁电技术已经趋于成熟。一些现 象已经预示着下一种主流存储技术的出现。  一方面， 很多的半导体供应商正在发展铁电。一些人关注近期产品的发展，而另一些人则关注已成熟的存储器和产品的发展。  另一方面，目前生产的低密度的铁电存储器的产品有广泛的市场。一些用户已经注意到铁电存储器的发展，铁电存储器的密度和结构，以便于对铁电产品的应用。每个新的密度的一代使得产生一系列的用户和厂家。   7 直到目前为止， Ramtron 公司是唯一的一家生产  FRAM 产品公司。由于公司的许可和授权，一些新的厂商也制造产品。在全球的范围内， FRAM发展适用于  FRAM 发展的 总资源正在剧烈的增长。这正在引起  FRAM 技术进步的里程碑。  下表是 Ramtron 公司和它的合伙人为 FRAM 技术的发展选择了历史的里程碑和近期的发展。  1984 Ramtron公司发现  FRAM 的发展技术  1989  FRAM第一次发展的过程  1993 首次制造容量为 4Kbit FRAM 存储器的商业产品  1996 容量为 16Kbit FRAM 存储器的制造  1998 厂家大量生产  0.1u的 FRAM 存储器用于飞行产品   在 64Kb FRAM中首次加入 MCU w/ 1999 在工厂中大 量生产  0.5 u的 FRAM 存储器   生产 64Kb, 256Kb FRAM 存储器  2000 3V FRAM 产品的操作示范  2001 生产 256K 1T1C 的 FRAM 存储器   在 FRAM生产过程中首次使用双层金属  生产 3V、 0.35u的产品  2002 256K 1T1C FRAM w/每周期  铁电存储器的应用   仪表  电表、水表、仪器仪表、流量表、邮资表。   8 汽车   安全气囊、车身控制系统、车载收音机、匀速控制、车载  DVD 、引擎、娱乐设备、仪器簇、  传动系、保险装置、遥感勘测 /导航系 统、自动收费系统   通讯  移动通讯发射站、  数据记录仪  、电话、收音机、电信、可携式 GPS 消费性电子产品  家电、机顶盒、等离子液晶屏电视  计算机  办公设备、雷达系统、  网络附属存储  、电子式电脑切换器。  工业、科技、医疗  工业自动控制、电梯、酒店门锁、掌上操作仪器、医疗仪器、发动机控制。  其他  自动提款机、  照相机、游戏机、 POS 功能机（可以用来以电子方式购买商品和服务）、  自动售货机。   铁电存储器在应用中所起的作用  数据收集存储   铁电存储器能够允许系统设计师更快、更频繁的写入数据，断电不易丢失。对于使用 EEPROM 的用户而言，这些是不能享受到的优良性能。   数据收集包括数据获取和存储数据，而这些数据必须在掉电的情况下仍能保留（不是暂时性的或中间结果暂存）。这些就是具有基本收集数据功能的系统或者子系统，并会随着时间而不断的发展出新的功能。在绝大多数的情况下，这个改变的过程纪录是很重要的。  配置信息存储    9 铁电存储器能够灵活实时的，并非在断电的瞬间，存储配置信息，从而帮助系统设计师克服由于突然掉电而造成的数据丢失。  配置信息的存储能够随着时间来追踪系统变化。其目标是在接通电源后恢复信息在以前的状态和位置 ，识别错误发生的起因。总的来说，数据收集通常是一个系统或者子系统的功能，然而配置信息存储则是一个低级别的工程功能，与系统的类别无关。  非易失性缓冲器   铁电存储器能够在数据发送或存储到其它非易失性媒介前，很快地存储正在运行中的数据。在这种情况下，数据信息由一个子系统传输到另一个子系统。这个信息是十分重要的并且不允许在断电的情况下丢失。在有些情况下 , 目标系统是一个更大的存储器。  而铁电存储器的快速、无限次的读写特点使得数据在被发送到另一个系统前就能及时保存。  SRAM的替代和扩展存储器   铁电存储器的快速写 入和非易失性的特点可以通过系统设计师把 SRAM 和EEPROM 的特点合而为一或者能单纯的扩展 SRAM 的功能而实现。  在很多情况下，一个系统会用到各种不同类型的存储器。铁电存储器同时具有ROM、  RAM 以及  EEPROM 的功能 ,并能节约系统内存和功耗。最常见的例子就是一个外部串行 EEPROM 的嵌入式的微控制器。铁电存储器能够取代 EEPROM,同样也能提供 SRAM 的微功能。    1 FRAM Technology Backgrounder Overview Established memory technologies are divided into two categories. First are nonvolatile memories. Traditionally, systems use them in read-only or read mostly applications since they are difficult to write. These memories are derivatives of ROM technology that include EPROM, EEPROM, and Flash EPROM. Second are volatile memories. These are RAM devices including SRAM and DRAM. Since they are easy to write, RAMs often store data that must change often. While users can write RAMs easily, they are volatile； therefore storing quantities of data in the absence of power continues to be an engineering challenge. Ferroelectric Random Access Memory or FRAM has attributes that make it the ideal nonvolatile memory. It is a true nonvolatile RAM. FRAM memory write advantages and nonvolatility make it quite suitable for storing data in the absence of power. FRAM based products have been available for several years in limited quantities. The technology is now moving rapidly toward its emergence as a mainstream memory selection. This technology note provides a brief explanation of its operation as well as an overview of the technology development status. What is FRAM? FRAM offers a unique set of features relative to other semiconductor technologies. Traditional mainstream semiconductor memories can be divided into two primary categories - volatile and nonvolatile. Volatile memories include SRAM (static random access memory) and DRAM (dynamic random access memory). SRAMs and DRAMs lose their contents after power is removed from the electronic system. RAM type devices are very easy to use, and are high performing, but they share the annoying quirk of losing their mind when the lights go out.  Nonvolatile memories do not lose their contents when power is removed. However all of the mainstream nonvolatile memories share a common ancestry that derives from ROM (read only memory) technology. As you might guess, something called read only memory is not easy to write, in fact it's impossible. All of its descendants make it very difficult to write new information into them. They include technologies called EPROM (almost obsolete now), EEPROM, and Flash. ROM based technologies are very slow to write, wear out after being written a small number of times, and use a large amount of power to write. FRAM offers features consistent with a RAM technology, but is nonvolatile like a ROM technology. FRAM bridges the gap between the two categories and creates something completely new - a nonvolatile RAM. FRAM is a RAM-based device that uses the ferroelectric effect for a storage mechanism. This is a completely different mechanism than the one used by other nonvolatile memories,  2 which use floating gate technology. The ferroelectric effect is the ability of a material to store an electric polarization in the absence of an applied electric field. Depositing a film of ferroelectric material in crystalline form between two electrode plates to form a capacitor creates a FRAM memory cell. This capacitor construction is very similar to that of a DRAM capacitor. Rather than storing data as charge on a capacitor like a DRAM, a ferroelectric memory stores data within a crystalline structure. These  Perovskite crystals maintain two stable states a  1 and a 0. Figure 1. Perovskite Ferroelectric Crystal Due to its basic RAM design, the circuit reads and writes simply and easily. However unlike a DRAM,the data state is stable. Therefore the FRAM memory needs no periodic refresh and when power fails, the FRAM retains its data. People commonly misunderstand the name ferroelectric. To many, a name using the prefix  “ferro” seems to imply iron or magnetism. The word ferroelectric also is confused with ferromagnetic. In reality, ferroelectric memories use no iron or magnetic principles. They are not susceptible to external magnetic fields as they operate entirely using electric fields just as conventional DRAMs. FRAM Technology Basics When an electric field is applied to a ferroelectric crystal, the central atom moves in the direction of the field.  As the atom moves within the crystal, it passes through an energy barrier, causing a charge spike. Internal circuits sense the charge spike and set the memory. If the electric field is  3 removed from the crystal, the central atom stays in position, preserving the state of the memory. Therefore, the FRAM memory needs no periodic refresh and when power fails, FRAM memory retains its data. It's fast, and doesn't wear out! FRAM memory technology is compatible with industry standard CMOS manufacturing processes. The ferroelectric thin film is placed over CMOS base layers and sandwiched between two electrodes. Metal interconnect and passivation complete the process. Ramtron's FRAM memory technology has matured significantly since its inception. Initial FRAM memory architectures required a two-transistor/two-capacitor (2T/2C) memory architecture, which resulted in relatively large cell sizes. Recent advances in ferroelectric materials and processing have eliminated the need for an internal reference capacitor within every cell in the ferroelectric memory array. Ramtron's new one-transistor/one-capacitor cell architecture operates like a DRAM using a single capacitor as a common reference for each column in the memory array,  4 effectively cutting the required cell area in half compared to existing 2T/2C architectures. The new architecture significantly improves the die leverage and reduces manufacturing costs for resulting FRAM memory products. Ramtron has also migrated to smaller technology nodes to increase the cost effectiveness of FRAM memory cells. A recent move to a 0.35-micron manufacturing process reduces the operating power and increases the die leverage per wafer compared to earlier generations of Ramtrons FRAM products built on the companys existing 0.5-micron manufacturing line. All of these exciting developments in FRAM memory technology are finding their way into a host of applications that people use everyday. From office copiers and high-end servers to automotive airbags and entertainment systems, FRAM memory is improving an array of products and applications worldwide. FRAM Operation A simplified model of a ferroelectric crystal is shown in Figure 1. A ferroelectric crystal has a mobile atom in the center of the crystal. Applying an electric field across a face of the crystal causes this atom to move in the direction of the field. Reversing the field causes the atom to move in the opposite direction. Atom positions at the top and bottom of the crystal are stable. Therefore removing the electric field leaves the atom in a stable position, even in the absence of power. As a memory element, the ferroelectric crystal creates an ideal digital memory. It contains two stable data states, it requires very little time and energy to change states, and is very stable over a variety of environmental conditions. Read Operation Although the memory element is a capacitor, it does not store data as linear charge. In order to read a FRAM memory cell, it is necessary to detect the position of the atoms within the Perovskite crystals. Unfortunately, they cannot be directly sensed. The FRAM read process works as follows. An electric field is applied across the capacitor. The mobile atoms will move across the crystals in the direction of the field if they are not already in the appropriate positions. In the middle of the crystal, a high-energy state holds the atoms in place when no field is present. As the atoms move through this high-energy state, a charge spike is emitted. The circuit dumps charge resulting from the applied field from the capacitor and compares it to the charge from a reference. A capacitor with atoms that switch states will emit a larger charge than a capacitor with atoms that do not switch. The no switching capacitor will emit the ordinary DRAM charge while the switching capacitor will emit the combination of the DRAM and ferroelectric charges. The memory circuit must determine which capacitor switched. This switched charge allows the circuit to determine the state of a memory cell. The state switch occurs in under 1 ns, with the complete circuit access taking less than 70 ns. Since a memory read operation involves a change of state, the circuit automatically restores the memory state. Therefore each read access is accompanied by a precharge  5 operation that restores the memory state. Although the read is destructive, the time during which the memory cell is invalid is under 50 ns. Write Operation A write-operation is very similar to a read operation. Unlike other nonvolatile memory technologies, a write-operation is very simple and requires no system overhead. The circuit applies write data to the ferroelectric capacitors. If necessary, the new data simply switches the state of the ferroelectric crystals. As with a read, the change of state occurs in under 1 ns with a full access taking under 70 ns. As with a read, a precharge operation follows a write access. FRAM Memory Architectures Current FRAM products use a two-transistor, two capacitor memory (2T2C) cell. This cell, which provides each data bit with its own reference, is a well-proven scheme. The fundamental cell design has been in field use in products since 1993. The 2T2C memory cell provides robust data retention reliability, which is especially important during the early proving stages for a new nonvolatile memory. An example of the 2T2C cell is sho wn in Figure 2.   The 2T2C memory cell provides an individual reference in close proximity for each data bit. Depending on the programmed data state, one capacitor will switch when read while the other will not switch. The assignment of 1 and 0 states is arbitrary during the memory design. Given the close proximity, the memory circuit can measure the charge difference between the switching and non-switching capacitors very precisely.  Variations in the capacitors across the memory array are eliminated from consideration by having a differential signal for each bit.  The 1T1C technology entered the market in 2001, it significantly improves the cost-per-bit ratio of resulting FRAM memory products. resulting FRAM memory products. A simplified  6 diagram of the 1T1C cell is shown below in Figure 3. FRAM Development As mentioned earlier, FRAM-based products have been commercially available since 1993. The considerable feature advantages of FRAM technology have stirred interest within the industry. Several signposts point to its emergence as the next mainstream memory technology. On the supply side, numerous semiconductor suppliers are developing ferroelectric processes. A few concentrate on near-term production while others are eyeing the longer-term opportunity for more sophisticated memories and embedded products. On the demand side, a broad market has developed for low-density FRAM products that are currently in production. Many potential users are watching the FRAM roadmap, looking for FRAM densities and configurations that will be suitable for their applications. Each new density generation enables a range of new potential users and applications. Until recently, Ramtron was the only company producing FRAM products. As a result of its successful licensing program, several new vendors are in the process of establishing production capability. The total resources being applied to FRAM development on a global basis are increasing dramatically. This is causing acceleration in the advancement of FRAM technology and its process milestones. The following table shows selected historical milestones and the near-term roadmap for FRAM technology development by Ramtron and its partners.   1984 Ramtron founded to develop FRAM technology 1989  First FRAM fab installed for process development 1993 First FRAM commercial product introduced 4Kbit FRAM memory in volume production 1996 16Kbit FRAM memory in volume production 1998 foundries open pilot-production lines First MCU w/ embedded 64Kb FRAM prototype 1999   64Kb, 256Kb FRAM memories in production 2000 3V operation FRAM products demonstrated 2001 256K 1T1C FRAM in production First embedded product using two-layer metal  7 FRAM process in production 3V operation products enter production  2002 256K 1T1C FRAM w/Real Time Clock FRAM Product Applications Metering electric power water gas flow tax postage Automotive airbag body control car radio cruise control DVD engine entertainment instrument clusters power train safety telematics/navigation toll tag  Communications cell base stations data logger phones radio telecom portable GPS Consumer Electronics home automation set top plasma and LCD TV Computing office equipment RAID network attached storage KVM switch Industrial, Scientific and Medical industrial automation elevator hotel lock handheld instrument medical motor cont