ABSTRACT
The advent of smart thin-film technology in the evolution of Complementary Metal Oxide Semiconductor (CMOS) technology has enabled us to consider a ferroelectric crystal a useful application. Thinning a ferroelectric film with high purity means that there could be an opportunity to use ferroelectrics as a memory element. On the other hand, integrated ferroelectrics are a subject of considerable research efforts because of their potential applications as an ultimate memory device due to three reasons: First, the capability of ferroelectric materials to sustain an electrical polarization in the absence of an applied field means that integrated ferroelectric capacitors are nonvolatile. They can retain information over a long period of time without a power supply. Second, the similar architectural configuration of memory cell-array to conventional ones means that they are highly capable of processing massive amounts of data. Finally, the nanosecond speed of domain switching implies that they are applicable to a high-speed memory device.Since ferroelectric capacitors was explored for use in memory applications by Kinney et al. [10], Evans and Womack [11], and Eaton et al. [12], it has been attempted to epitomize ferroelectrics to applicable memory solutions in many aspects. In the beginning of the 1990s, silicon institutes began to exploit ferroelectrics as an application for high-density dynamic random access memories (DRAMs) [13,14]. This is because permittivity of ferroelectrics is
so high as to achieve DRAM’s capacitance extremely high and thus appropriate for high-density DRAM application. An early version of nonvolatile ferroelectric RAM (random-access-memory) used to be several kilo bits in packing density. The main materials for these nonvolatile memory applications are lead zirconate titanate (PZT) and strontium bismuth tantalate (SBT). Next steps for high-density nonvolatile memory have been forwarded [15,16]. In similar to DRAM, an attempt to build smaller unit cell in size was in the late 1990s that one transistor and one capacitor (1T1C) per unit memory was developed [17]. Since then, many efforts to build high-density FRAM have been pursued, leading to several ten mega bits in density during the 2000s [18-21].Among integrated ferroelectrics, one of the most important parameters in FRAM is sensing signal margin. The sensing signal of FRAM is proportional to remanent polarization (Pr) of a ferroelectric capacitor as follows:0rBL BL BL22 ,= = AP AV C C de eD (13.1)where A is capacitor’s area, d is capacitor’s thickness. As shown in Eq. (13.1), in principle, we have to compensate the area reduction when technology scales down. However, in practice, when the thickness of PZT ferroelectric thin film decreases, degradation of polarization tends to appear in the ferroelectric capacitor due to a dead-layer between the ferroelectric and electrodes.Key integration technologies for ferroelectric memory to become highly mass-productive, highly reliable, and highly scalable we will be presented. This covers etching technology to provide a fine-patterned cell with less damage from plasma treatments, stack technology to build a robust ferroelectric cell capacitor, encapsulation technology to protect the ferroelectric cell capacitors from process integration afterward, and vertical conjunction technology onto ferroelectric cell capacitors for multilevel metalliza-tion processes [22]. 13.2 Ferroelectricity
Ferroelectricity was first discovered by Valasek [1] in 1921. He found that the polarization of sodium potassium tartrate tetrahydrate
(NaKC4H4O6.4H2O), better known as Rochelle salt, could be reversed by the application of an external electric field. A hysteresis effect, which is characteristic for ferroelectricty, was seen and switching between the two different polarized states was possible. However, for two decades this discovery and that of the same phenomena in the KDP series [23] was considered mere artifact. Only in 1949, the ferroelectric research field emerged, due to the discovery of ferroelectricity in ceramics (BaTiO3). It belonged to what was soon to become the biggest class of ferroelectrics: perovskite oxides [24], with general formula ABO3. This falsified the assumption that hydrogen bonds were essential for ferroelectricity. Now over more than 700 ferroelectrics have been found, most of them not hydrogen bounded or oxide. Another assumption that needed to be falsified was the critical size. Up until the 1990s, ferroelectricity was believed to be destroyed below 10 nm. However, measurements on PZT films, showed switchable polarization down to thicknesses of a few nanometer [25]. Also, theoretical predictions showed that PbTiO3 will maintain a switchable spontaneous polarization down to three unit cells as long as the depolarization field was fully compensated [26]. In 1948, the first device was made [27], but it took some years before the field of ferroelectric memory devices flourished, especially because it required techniques that could make thin films. Bulk ferroelectrics need coercive voltage of 5 kV are needed for reversing their polarization while below the submicronmeter scale the silicon level of 5 V can be reached. So when deposition techniques enabled the making of thin films this lead to the integration of ferroelectric material in silicon chips. Current research on these integrated ferroelectrics involves finite size effects, interfaces, and strain. The term ferroelectricty is a reference to the earlier discovered ferromagnetism. Analogy can be found between the electric properties of ferroelectrics and the magnetic properties of ferromagnets. However, where mechanical coupling can be neglected in ferromagnets, this is not the case for ferroelectrics. Ferroelectricty arises because of strain and displacement of charge, whereas ferromagnetism is a reordering of the spin states of the electrons.Properties of materials are closely linked to their defect structure. Numerous studies have proved that the existence of a small amount of microstructural defects can dramatically change the way of materials behaving in response to external fields. Based on these,
various kinds of functional devices have been developed, which have changed the daily life of human beings. Currently, the most important application of defects in industry is probably semiconductor devices intentionally doped with foreign atoms to realize desirable band structures to tune the behaviors of electrons. Defects are also intentionally introduced into metals and insulators to achieve better performances. Similarly, defects in ferroelectric materials are also extremely important. As a subject that has been investigated for decades, it has been proved that defects and the associated stress and electrical fields could change ferroelectric behaviors such as polarization reversal, domain kinetics, phase transition temperatures, and ferroelectric fatigue. Up to date, numerous studies have been devoted to understanding oxygen vacancies, dislocations, domain walls, voids, and microcracks in ferroelectrics. Actually, almost all aspects of ferroelectric properties are defect-sensitive. For example, doped PZTs could be either “soft” or “hard” with variable coercive fields. Oxygen vacancies play a determinant role on the fatigue process of ferroelectric oxides. Dislocations may hinder the motion of ferroelectric domain walls. Recent interests on the design and fabrication of nanodevices stem from the distinct and fascinating properties of nanostructured materials. Among those, ferroelectric nanostructures are of particular interests due to their high sensitivity and coupled and ultrafast responses to external inputs [28]. With the decrease of the size of ferroelectric component down to nanoscale, a major topic in modern ferroelectrics is to understand the effects of defects and their evolution [29]. Defects will change optical, mechanical, electrical, and electromechanical behaviors of ferroelectrics [30,31]. However, current understanding is limited to bulk and thin-film ferroelectrics and is still not sufficiently enough to describe their behaviors at nanoscale. In view of the urgent requirement to integrate ferroelectric components into microdevices and enhanced size-dependent piezoelectricity for nanosized ferroelectric heterostructure, [32] it becomes essential to explore the role of defects in nanoscale ferroelectrics. 13.2.2 CharacteristicsA material is ferroelectric when it has two distinct polarization states, which can be maintained in the absence of an electric field
and between which one can switch by applying an electric field. Key in defining whether a material is ferroelectric is the experimental setup, since one must be able to determine the switching even when a small difference is observed. For most known ferroelectrics, the onset of ferroelectricity occurs as a function of decreasing temperature. It is a structural phase transition and the transition can be either displacive or order-disorder in nature. Displacive transitions involve a small distortion of the bonds between the atoms, while in order-disorder transition the atoms rearrange from a random occupation of the lattice sites to specific define ones for each atom type. The order parameter for ferroelectrics is the spontaneous polarization that is caused by the atomic arrangement of ions in the crystal structure. This can depend either on the position of the ions (conventional ferroelectrics) or on the charge ordering of multiple valences (electronic ferroelectrics). A polar displacement of the atoms in an unit cell results in ferroelectricity. These displacements might be coupled to strain or other nonpolar atomic displacements. Thus, ferroelectric crystals need a polar space group. Also, their structure should have several polar variants with discrete states of polarization. The latter is ensured by the principle that a ferroelectric crystal structure is a small symmetry-breaking distortion of a higher-symmetry reference state, called the paraelectric state. At the Curie temperature (Tc), the material transits from the ferroelectric state to the nonpolar paraelectric phase upon increasing temperature. However, sometimes the ferroelectric breakdowns before its transition temperature is reached and no paraelectric state is found. Ferroelectric transitions can be characterized by phonon spectroscopy. X-ray and neutron diffraction can be used to get the crystal structure of the material and determine its polarization switching origin. Most of the ferroelectric transitions are second order and can be described by Landau theory. With help of the Lyddane-Sachs-Teller relation, the link is made with the vanishing frequency of a polar mode. This leads to the theoretical description of ferroelectricity that is given by the so-called soft-mode description [33]. Upon cooling, a material from above the transition temperature, a normal mode of vibration of the crystal decreases to zero frequency. Hence, a ferroelectric transition can usually be associated with the condensation of a soft (or low frequency) mode. This condensation distorts the crystal structure and leads to the appearance of a long-
range polar order. For this transition, the anharmonic interaction between the phonons is important. Also, polar phonons are temperature dependent, and thus ferroelectricity as well. The distinction between displacive and order-disorder transitions can be made by looking at the dynamics of the transition, whether the soft mode is propagating of diffusing in character. With aid of this soft-mode concept and the assumption that a ferroelectric state is only a small distortion of a higher-symmetry state, first-principle studies can be done. Phonons related to the paraelectric reference structure are computed and the unstable modes used as guides to identify energy-lowering distortions. These models can predict crystal structure and polarization. They are especially of interest for the behavior of ultrathin films, which are built up of a few layers of atoms. Together with experimental structural determinations, first-principle studies form a powerful way to determine atomic arrangements and electronics states in the crystal on an atomic scale [34]. 13.2.2.1 Polarization and hysteresisThe appearance of a hysteresis cycle is essential for ferroelectricity. However, not all solids with electrical hysteresis are ferroelectric. Hysteresis can have extrinsic causes due to mobile charge defects and PN-junctions. For an ideal ferroelectric, the P-E hysteresis loop is symmetric. From it, one can define the remanent polarization states and the coercive fields. This coercive field must be lower than the breakdown field of the material, to enable switching. To measure the electric properties of a ferroelectric we need to include it in a device, the simplest one being a capacitor. In this device, a ferroelectric is sandwiched between two metal electrodes. This means that in measurements the system as a whole including all its components is involved. So one needs to distinguish between ferroelectric response and electrical responses dominated by the system. For an overview of how measurements with limitless artifacts can be performed, the reader is referred to the textbook of Rabe et al. [35]. One trick is to vary the frequency since artifacts are usually highly frequency dependent.In Fig. 13.1, the P-V hysteresis loop for several PbTiO3/SrTiO3 samples with different polarizations is visualized. The correspond-ing I-V characteristic shows clear switching peaks at the coercive voltage, another characteristic of ferroelectricity. From these
switching currents, the polarization can be determined. For an infinite crystal, the polarization is defined as an integrated current though the transformation from one ferroelectric variant to anoth-er. By measuring the switched charge/current versus the voltage, one can obtain the P-V hysteresis loop and determine the remnant polarization with the aid of integration techniques. Nowadays, most of these measurements are carried out by commercial apparatus of Radiant technologies or AixAcct, but previously the Sawyer-Tower circuit was used [36]. For most devices, the polarization values and time that the material remains switched are the values of most interest. Also, current measurements to determine the leakage current and dielectric permittivity of the device are common.
