An Emerging Memory Material with Outstanding Potential! Antiferroelectric Hafnium Zirconium Oxide

Ferroelectric materials are a very special class of polar compound. Based on the spontaneous electric polarization effect, they can exhibit excellent ferroelectric, piezoelectric, thermoelectric and non-linear optical properties, etc.. Therefore, they have many important applications in the fields of data storage, infrared light sensing, ultrasonic waves and photoelectric microwave communications, etc.. Ferroelectric memory has the potential to break through the von Neumann bottleneck to become one of the next generation mainstream memory technologies to realize “storage-class memory” and “in-memory computing”. With the significant progress that has been made in materials research in recent years, it has attracted more and more attention to become a hot topic in the semiconductor industry.

In the past, the fact that the Ferroelectricity of perovskite materials deteriorates rapidly when the film thickness is below a certain critical value made it difficult to continue to reduce the size of its memory cells. This meant that the memory density could not be effectively increased. Therefore, the product applications of early ferroelectric memory was limited to niche markets. More recently, however, the semiconductor-compatible material hafnium dioxide (HfO₂) was found to have ferroelectric properties. Moreover, the difficulty of the material integration process is low, and the costs are also advantageous. This has finally kicked off a new wave of industrial opportunities for ferroelectric memory. HfO₂ is already a fairly familiar material in the semiconductor field. It has long been widely used as a high-k dielectric layer in CMOS below 28 nm and as the dielectric material in DRAM capacitors. Note that HfO₂ must be doped with specific elements to produce ferroelectric properties.

Antiferroelectricity is a type of physical material property that is closely related to ferroelectricity. Whether a material is ferroelectric or antiferroelectric, it belongs to the ferroelectric materials category of applications. Recent studies have found that thin films with antiferroelectric properties can be fabricated by adjusting the proportion of Si added to HfO₂. Doping HfO₂ with zirconium (Zr) can also change its ferroelectric properties. As the doping ratio of Zr increases, the dielectric effect also increases. When the ratio of Zr doping exceeds that of hafnium (Hf), the hysteresis loop of the hafnium zirconium oxide (HfZrO₂; HZO) compound will change from ferroelectric to antiferroelectric.

Research findings have revealed that memory made using antiferroelectric HfO₂ materials have better endurance, faster switching speeds, smaller operating bias, and lighter wake-up effects, etc. than original HfO₂ memory. Once a bias is applied to an HZO material, it will produce remnant polarization, making it suitable for making volatile memory. By adjusting the processes or structural treatments, it is possible to adjust the electrodes’ work-function difference, the oxygen vacancy inside the antiferroelectric layer, or the fixed charge layer in the interface dipole to create a built-in electric field (Built-in E-field). The P-V characteristics, which were originally symmetrical under bipolar operation, can be shifted in this way, resulting in asymmetry. This allows for excellent remnant polarization even when the bias voltage is 0 V. This means it can be used to make nonvolatile memory. In addition, antiferroelectric tunnel junction memory using 3D NAND architecture can also realize stable Multilevel Cell (MLC) high density memory storage. With its diverse range of possible applications, antiferroelectric HZO has become the top choice for the next generation of ferroelectric memory materials.

Chris Chen, Director of the MA-tek Technology R&D Center 12/22/2022

An Emerging Memory Material with Outstanding Potential! Antiferroelectric Hafnium Zirconium Oxide

Institute of Optoelectronic Engineering, National Taiwan Normal University

Professor Min-Hung Lee

Postgraduate Students: Guoyu Xiang, Zhaofeng Luo, Hanzhen Zeng, Fusheng zhang, Zhixian Li, Weicheng Rui, Yitai Zhang

  (This article was provided by Professor Min-Hung Lee and edited by MA-tek)

-

Ferroelectric materials have been studied extensively [1]-[5], and many of those studies examine their applications in memory. Being similar, antiferroelectric materials are considered a type of ferroelectric material, but it also has its own unique characteristics. In recent years, with the vigorous development of ferroelectric hafnium-based oxide (Hf-Based Oxide), progress has also been made on the research of the antiferroelectric properties of this material. This article will examine materials with antiferroelectric properties from the perspectives of materials science, ferroelectric engineering and operating techniques and the development of suitable, corresponding applications.

2020 marked the 100th anniversary of the discovery of ferroelectric materials. In 1920, Valasek discovered ferroelectricity in Rochelle Salt [6]. During World War II, scientists in the United States, Soviet Russia and Japan all searched for dielectric materials that could replace mica in capacitors. The ferroelectric effect was discovered in perovskite BaTiO₃. Literature on antiferroelectric properties appeared as early as 1950 , when Professors G. Shirane and E. Sawagachi of the Tokyo Institute of Technology published about the antiferroelectric perovskite PbZrO₃ [7]-[9]. In 1951, C. Kittel of Bell Laboratories published a study proving the existence of antiferroelectricity [10]. Figure 1 is a timeline showing the development history of solid-state antiferroelectric materials [16].

Figure 1.　The Development History of Solid-State Antiferroelectric Materials [1].

In 1953, the antiferroelectric phase of perovskite PbHfO₃ was identified [11]. At the same time, L. E. Cross and B. J. Nicholson of the University of Leeds in the UK were studying NaNbO₃ crystals and demonstrated a double hysteresis loop [12]. Coupling the Devonshire model using BaTiO₃ ferroelectric materials with the theory proposed in Kittel‘s thesis explained the double hysteresis loop of NaNbO₃ [13]. In 1959, W. Cochran illustrated the theory of displacement-based transitions [14]. In terms of basic research, people’s interest in the antiferroelectric and ferroelectric boundary was ignited by the PbZrO₃-PbTiO₃ (PZT) developed by B. Jaffe, W. R. Cook and H. Jaffe in the United States in the 1960s. They used the PZT phase diagram to measure the parameters of antiferroelectric phase transitions [15]. Then, in 2007, Qimonda AG discovered ferroelectric properties in silicon-doped hafnium oxide (Si:HfO₂) and published the doping of silicon (Si) and zirconium (Zr) in hafnium oxide (HfO₂) in 2011 [17] and 2012 [18] respectively. Depending on the amount of doping, the material can have ferroelectric or antiferroelectric properties. This started the ferroelectric and antiferroelectric research of ferroelectric Hf-Based Oxide, which is compatible with current semiconductor manufacturing processes.

These materials have attracted a great deal of attention because of their rich scientific potential and practical high energy field applications. Antiferroelectricity occurs when the polarity of two adjacent single crystals inside a material is parallel but arranged with the electric dipoles in opposite directions. The two reconstitute a new lattice. Given a strong enough electric field, the phase in the opposite direction of the electric field can be reversed into a ferroelectric field of the same polarity, at which point the double hysteresis loop can be observed. When the applied electric field is zero, the two opposite and parallel polarities cancel each other out. Conversely, if the applied electric field cannot cause the reversal, it means that the antiferroelectric material’s resistance is greater than that of the applied electric field.

Figure 2(a) is a schematic diagram of paraelectric, ferroelectric and antiferroelectric phase transition states. Figure 2(b) shows the typical hysteresis loop of antiferroelectric properties measured for antiferroelectric materials [5,6]. In the low polarization state, when the bias voltage increases to the coercive field EF, the amount of polarization rises sharply. However, when the field direction is reversed, the P-E hysteresis loop decreases sharply along another path, that of the coercive field EA (< EF), forming another loop before reaching the linear region.

Figure 2.　(a) Schematic Diagram of Paraelectric, Ferroelectric and Antiferroelectric Properties in PbZrO3 ; (b) Typical Double Hysteresis loop of Pb0.98La0.02(Zr0.66Ti0.10Sn0.24) 0.995O3  [8]。

Before hafnium-based oxide, most well-known ferroelectric materials were perovskite materials. In a ABO₃ structure perovskite, A ions occupy positions at the eight corners of the cubic lattice shown in Figure 3. Oxygen atoms occupy the central positions of the six faces of the cubic lattice. Together, they form a face-centered cubic lattice. Finally, B ions occupy the central position of the cubic lattice, thus forming the BO₆ octahedral structure. The A cation is a rare earth ion, the B cation is Al³⁺, and the O^2- is an anion. Figure 4 is the PbZrO₃ perovskite structure [19], where the dark green atoms are Pb, the light green atoms are Zr, and the gray atoms are O. The arrow is the displacement of Pb in the antiferroelectric crystal structure. The asymmetrical phase is an important factor affecting antiferroelectricity. The early research of ferroelectric applications mostly used perovskite materials, but some of the elements involved are toxic. Therefore, in view of environmental and thermal budgeting considerations, its current use in CMOS processes is limited.

Figure 3.　Structural Diagram of ABO3 Perovskite Ferroelectric Material.

Figure 4.　Crystal Structure of PbZrO3 (Pb: Dark Green, Zr: Light Green, O: Gray); See the Basic Antiferroelectric Structure of PbZrO3, where the Arrow Indicates the Phase Transition Caused by Ion Displacement [19].

In recent years, the ferroelectric and antiferroelectric phases of the two-dimensional material CuBiP₂Se₆ have been widely tested. As shown in Figure 5(a), in AgBiP₂Se₆, the sulfide framework has octahedral voids filled by Ag(Bi) and P-P pairs. Bulk crystal Ag⁺ and Bi³⁺ sites exhibit an antiferroelectric-like ordering [21]. Adjacent layers are bonded together via van der Waals interactions, and the weakness of the van der Waals bonds allow the material to be exfoliated into two-dimensional layers. Detailed density functional theory (DFT) studies indicate that balancing the ionic and van der Waals interactions can help control the ferroelectric and antiferroelectric orderings in these materials [20]. Figure 5(b) is the image of dipoles in β-In₂Se₃ taken via ADF-STEM (Annular Dark Field Scanning Transmission Electron Microscope). Atom displacements are measured and boundaries observed.

Figure 5.　(a) The Ag and Bi Positions in the CuBiP2Se6 Crystal Structure, Where an Antiferroelectric-Like Ordering is Exhibited; (b) Atomic Displacements in β-In2Se3 [20].

Early studies of ferroelectric applications mostly used perovskite materials such as: lead zirconate titanate (PbZrTiO₃, PZT), barium titanate (BaTiO₃, BTO), strontium bismuth tantalate (SrBi₂Ta₂O₉, SBT), and strontium titanate (SrTiO₃, STO), etc.. However, due to the toxicity of some of the elements in perovskite ferroelectric materials, it faces miniaturization problems, as shown in Figure 6 [22]. Therefore, a solution needed to be found. Qimonda discovered in 2007 that hafnium oxide doped with silicon (Si:HfO₂) exhibits ferroelectric properties. Subsequently, many other research circles have published that doping hafnium-based oxides with silicon (Si), aluminum (Al), gadolinium (Gd), strontium (Sr), lanthanum (La), and zirconium (Zr) can all result in ferroelectric properties, as shown in Figure 7 [23]. In 2011, T. S. Böscke proposed that the ferroelectric phase of the hafnium oxide system is an Orthorhombic Phase. The formation of this crystal phase is a transformation from the Tetragonal Phase through a cooling process. Figure 8 is its schematic diagram [17].

Thin films with antiferroelectric properties can be obtained by adjusting the ratio of silicon (Si) in hafnium oxide [17]. In Figure 9, it can be seen that the minimally doped HfO2 film exhibits ferroelectricity, and, as the SiO2 doping gradually increases, the hysteresis loop (black) and capacitance (thick red) will change accordingly. Finally, the hysteresis loop obtained when the film is doped with 5.6 mol.% shows the transformation from ferroelectric to antiferroelectric.

Figure 9. Capacitance and Hysteresis loop of Different SiO2 Doping Concentrations [17].

The J. Müller group published in 2012 that, if HfO2 is doped with Zr [23], the Zr to HfO2 doping ratio will affect the ferroelectric properties. The literature uses Atomic Layer Deposition (ALD) to more accurately control the ratio of Hf to Zr so that it can be optimized, thus facilitating the study of ferroelectric properties. Figure 10(a) shows the hysteresis loop with the Zr doping ratio from low to high and the relationship between the dielectric value and the electric field. The higher the Zr doping ratio, the greater the dielectric value. When the Zr ratio exceeds that of the Hf, the hysteresis loop changes from ferroelectric to antiferroelectric.

Figure 10.　The Relationship Between Polarization and Dielectric Value vs. Electric Field for Different Zr Ratios [23].

Figure 13. Structure Simulation Constructed According to the NLS Model [29].

By adding a Back-Switching Field (EBS) to the ferroelectric model to form an antiferroelectric model then making adjustments based on experiment data, the ferroelectric and antiferroelectric time constants were calculated to be 1203 ns and 223 ns respectively, as shown in Figure 14(a) and (b). This constant directly affects the speed with which the polarization direction changes. So, as frequency increases, antiferroelectric components respond faster than ferroelectric ones [28].

Figure 14. P-V Characteristics of Simulated Ferroelectric Capacitors at Different Frequencies: (a) Ferroelectric Capacitor and (b) Antiferroelectric Capacitor [28].

Since the Polarization-Voltage (P-V) characteristics of antiferroelectric capacitors is very different from those of ferroelectric capacitors. Figure 15(a) is diagrams the P-V characteristics of typical ferroelectric capacitors and antiferroelectric capacitors under bipolar operation [30]. Where the bias voltage is 0 V, ferroelectric capacitors have good, clear remnant polarization, making them suitable for use in Non-Volatile-Memory. Antiferroelectric capacitors, on the other hand, show weak remnant polarization and are not suitable for non-volatile memory. Compared to ferroelectric materials, however, they have better endurance, smaller operating biases and slighter Wake-Up effects. Next, we will examine the application of antiferroelectricity in random-access-memory.

The P-V characteristics of antiferroelectric capacitors are like two symmetrical ferroelectric polarization loops in the positive and negative polarity regions respectively. Figure 15(b) applies a bias voltage for unipolar operation in the positive polarity region (0 V to 3 V). Here, the ferroelectric capacitor loses the ferroelectric properties it originally had under bipolar operation, whereas the antiferroelectric capacitor maintains good ferroelectric properties under the unipolar bias. This unipolar operation method reduces the operating voltage. However, an additional fixed bias voltage needs to be applied to the antiferroelectric capacitor. As such, the antiferroelectric capacitor is more suitable for use in volatile memory.

Figure 15.　The P-V Characteristics of Ferroelectric Capacitors and Antiferroelectric Capacitors Under (a) Bipolar and (b) Unipolar Operation

By adjusting the Work-Function Difference of the electrodes [31], the oxygen vacancyinside the antiferroelectric layer [32], the fixed charge layer and the Interface Dipole [33] create a Built-in E-Field that offsets the P-V characteristics that were symmetrical under bipolar operation, resulting in asymmetry. This asymmetrical antiferroelectric P-V has excellent remnant polarization when the bias voltage is 0 V, making it suitable for non-volatile memory, as shown in Figure 16(a). The non-volatile polarization characteristics of the antiferroelectric capacitors in Figure 16(b) and (c) relate to the Work-Function Difference of RuOx and TiN electrodes respectively. The number of oxygen vacancy inside the antiferroelectric layer is about 1018 cm-3 plus the fixed charge of the alumina layer and the dipoles at the interface between the alumina layer and the antiferroelectric layer.

Figure 16.　(a) Antiferroelectric P-V Shift Caused by the Built-in E-Field Results in Asymmetry; (b) The Non-Volatile Polarization Characteristics of Antiferroelectric Capacitors with the Work Function Difference of RuOx and TiN Electrodes Respectively; The number of oxygen vacancy inside the antiferroelectric layer is about 1018 cm-3 and (c) the fixed charge of the alumina layer and the dipoles at the interface between the alumina layer and the antiferroelectric layer [31][33].

In the current literature, Intel has published an approach for using the Work-Function Difference between RuOx and TiN electrodes in the 3D array structure to realize Embedded Dynamic Random-Access Memory (eDRAM) technology, as shown in Figure 17(a). Figure 17(b) shows how the internal electric field and dipole defects of the material affect the polarization. Figure 17(c) shows that different internal electric fields will have different Work-Functions, resulting in differences in polarization that lead to differences in Memory Windows (MW).

Figure 17.　(a) TEM Cross Section of eDRAM 3D Array Structure; (b) Influence of Internal Electric Fields and Dipole Defects on Polarization; (c) Relationship Between the Work-Function Difference and MW [32].

Ferroelectric Tunnel Junction (FTJ) memory is basically a resistive, two-terminal component, as shown in Figure 18. There are two types: the single-layer ferroelectric structure and the double-layer ferroelectric and dielectric structure. The single-layer ferroelectric structure consists of two layers of metal sandwiching a ferroelectric film to form a Metal/Ferroelectric Layer/Metal (MFM) structure. The Screening Length caused by the use of electrode materials at both ends is different, producing a difference in Barrier Height that then modulates the On and Off Tunneling Current. The double-layer ferroelectric and dielectric structure is a ferroelectric thin film sandwiched between two metal layers and a dielectric layer, forming a Metal/Ferroelectric Layer/Dielectric Layer/Metal (MIFM) double-layer ferroelectric tunnel junction component. The principle is that, when passing through different polarization directions, the energy barrier height of the ferroelectric layer is different, thus resulting in tunneling with different Tunnel Electro-Resistances (TER).

The FTJ has a Low-Resistance State (LRS) and a High-Resistance State (HRS). This means that the FTJ is capable of distinct 0 and 1 storage states. The difference between single-layer and double-layer structure FTJs is that they have significantly different electron tunneling lengths. The electrons of a single-layer FTJ only need to tunnel through the ferroelectric layer. In a double-layer FTJ, however, the electrons tunnel through the dielectric layer to form the On current, while they need to tunnel through the ferroelectric layer and the dielectric layer for the Off current, as shown in Figure 19(a) [35]. In 2019, the B. Max research team at Germany’s Technical University of Dresden published a 12 nm double-layer FTJ structure using HZO and 1 nm Al2O3 [35]. The correlation between the Set Voltage and the switch current is shown in Figure 19(b). The polarization of the ferroelectric layer is adjusted by changing the writing and erasing voltage. This adjusts the tunneling energy barrier, which changes the value of the tunneling current.

Figure 18.　Energy Band Diagram of (a) Single-Layer and (b) Double-Layer Ferroelectric Tunnel Junction Components [34].

Figure 19.　(a) Switching State Energy Band Diagram of Double-Layer FTJ; (b) Dependence of Set Voltage on the On Current and Off Current; Current Values Read at 2 V Under Different Set Voltages [35].

In antiferroelectric components, the Built-in Electric Field can be created using the method discussed above so that the antiferroelectric component will have remnant polarization when the bias voltage is 0 V, as shown in Figure 14(a). To improve the memory density, there are the 3D NAND and Minimal incremental Layer Cost (MiLC) architectures shown in Figure 20(a). It is one of the strategies that can surpass the memory unit area of 4F2 embedded Non-Volatile Memory (eNVM). The Transmission Electron Microscope (TEM) image of the 3D NAND antiferroelectric tunnel junction memory cross-section is shown in Figure 20(b) and (c). Figure 20(d) and (e) prove through Fast-Fourier Transformation (FFT) and Energy-Dispersive X-ray Spectroscopy (EDS) analysis that this antiferroelectric hafnium zirconium oxide has excellent crystal qualities and shows the correctness of the positioning of the related elements. The I-V characteristics of the 3D vertical antiferroelectric tunnel junction in particular displayed an excellent current ratio (Ion/Ioff ratio) > 100x at 1 V, as shown in Figure 20(f) [36].

Figure 21(a) and (b) show the current ratio characteristics of the double-layer ferroelectric and antiferroelectric tunnel junctions respectively with Al2O3 thicknesses of 0 nm to 4 nm. The an antiferroelectric tunnel junction with an Al2O3 thickness of 2 nm to 4 nm has a current of ~100x. It can be seen that modulating antiferroelectricity through engineering can enable it to have better performance than ferroelectricity in tunnel junction memory components. In addition, an antiferroelectric tunnel junction with an Al2O3 thickness of 2 nm has the minimum operating bias. Figure 21(c) and (d) show the performance of the antiferroelectric tunnel junction with an Al2O3 thickness of 2 nm as the writing voltage is gradually increased. It has excellent Multi-Level Cell (MLC) characteristics, linear symmetry and stable deep learning characteristics [36].

Figure 20.　(a) 3D NAND and MiLC Architecture; (b) and (c) TEM Cross-Sectional View of Antiferroelectric Tunnel Junction Component for 3D NAND Architecture; (d) and (e) for Fast-Fourier Transformation and EDS Analysis; (f) The I-V Characteristics of 3D vertical FTJ Show an Excellent Current Ratio (Ion/Ioff ratio) of > 100x at 1 V [36].

Figure 21.　(a) and (b) Show the Current Ratio Characteristics of Double-Layer Ferroelectric and Antiferroelectric Tunnel Junctions with Al2O3 Thicknesses from 0 nm to 4 nm; (c) and (d) Demonstrate the Performance of the Antiferroelectric Tunnel Junction with an Al2O3 Thickness of 2 nm by Gradually Increasing the Voltage; It Has Excellent Multi-Level Cell Characteristics and Deep Learning Characteristics [36].

In 2018, the team led by Tokyo University’s Professor Masaharu Kobayashi [37] successfully verified the applicability of antiferroelectric materials in field-effect-transistors by adding an antiferroelectric capacitor to the gate of the field-effect-transistor and adjusting the area ratio. The ratio of the capacitor to the gate area is 1:32. This approach successfully produced a field-effect-transistor with a memory window, as shown in Figure 22. In 2019, Professor Kai Ni’s team at the Rochester Institute of Technology [38] proposed using multi-layer ferroelectric or single-layer antiferroelectric materials to realize the Multi-Peak Ec concept to achieve stable multi-level memory. They demonstrated the concept using simulations, as shown in Figure 23, achieving a Multi-Peak EC P-V loop and memory probability distribution.

A single-layer antiferroelectric component achieved multi-level operation in 2022 [39], verifying its feasibility. The ferroelectric-antiferroelectric field-effect-transistor used a single-layer quasi-antiferroelectric hafnium zirconium oxide (AFE-FE-FET), as shown in Figure 24(a) and (b). Using quasi-antiferroelectric hafnium zirconium oxide has the advantage of having remnant polarization at 0 V and Multi-Peak EC characteristics. AFE-FE-FET achieves ultra-low write/erase voltages (|VP/E| = ±4 V). The Multi-Peak EC concept is used to produce stable multi-level operations, as shown in Figure 25. Its endurance is >105, and it has excellent Data Retention as well as a time of >104 seconds at 65 °C.

From the detailed discussion above on these components from the perspective of application, we can see that the application of antiferroelectric materials in memory has promise. Though it was originally difficult for it to achieve non-volatile storage, its high endurance, slight wake-up effect and quick responses, etc. can be adjusted via engineering and improved for specific uses. As such, the related materials science is very important. Developing materials analysis methods suitable for use on materials with antiferroelectric properties, such as crystal phase analysis, polarization domain judgment, and oxygen vacancy distribution analysis, etc., is an area that can be further developed in the future.

It has been more than a hundred years since American scholar Joseph Valasek first published his research in 1920 showing that Rochelle salt (KNaC4H4O6·4H2O) had a spontaneous polarization effect. Due to the low mechanical strength and deliquescence of Rochelle salt, the discovery of the ferroelectric phenomenon at the time was little more than symbolic in terms of scientific significance. However, the uniqueness of ferroelectric physical properties and the broad application prospects still drew many researchers to invest in it and search actively for new ferroelectric materials with better application potential. Therefore, for half a century, the whole world was focused on the research of perovskite ferroelectric materials. BaTiO₃ (BTO) and Pb[ZrxTi1-x]O₃ (PZT), two classic materials with strong, stable ferroelectric properties, were developed, making it possible for ferroelectric materials to be used in military and commercial applications. BaTiO₃ in particular had an especially simple structural composition. As such, it provided a convenient model of reference for studying crystal structure changes in ferroelectric phase transitions. In 1949, it also led scholar A. F. Devonshire to establish a theory for predicting ferroelectric phenomena (it described the free energy of ferroelectric materials as a double well with an energy barrier between the two polarized states). This theory still plays an extremely important role in understanding ferroelectric properties and accelerating the research of ferroelectric materials.

Generally speaking, common ferroelectric material applications mainly use their piezoelectricity, thermoelectricity, electro-optic effect, and high dielectric properties. Basically, all ferroelectric materials possess both ferroelectricity and piezoelectricity.

Ferroelectricity refers to the spontaneous polarization of a material within a certain temperature range. Since the positive and negative charge centers in the ferroelectric lattice do not coincide, it can generate an electric dipole even when there is no external electric field. This spontaneous polarization can change direction under the influence of an external electric field. When the temperature is higher than a certain critical value, however, the lattice structure of the ferroelectric material will change, and the positive and negative charge centers will overlap so that the spontaneous polarization phenomenon disappears. This critical temperature value is called the Curie Temperature (Tc).

Piezoelectricity is a property that realizes the conversion between mechanical energy and electrical energy. If an external force is applied to the material in a certain direction to make it deform, the material will become polarized inside, and charge will be generated on the surface. This is called the piezoelectric effect. Conversely, if an electric field is applied to the material, it will deform and generate mechanical force. This is called the inverse piezoelectric effect. In addition, there are some special multiferroic ferroelectric materials, such as BiFeO3, in which non-equilibrium carriers will become excited under light irradiation, causing an asymmetrical change inside the electron cloud and thereby inducing microscopic polarization and producing many special physical phenomena, such as the anomalous photovoltaic effect and the photorefractive effect, etc..

In recent years, the rise of emerging technologies such as artificial intelligence, the Internet of Things, 5G communications and smart vehicles have created a need for real-time analysis of a huge amount of information. Existing high-capacity storage devices like DRAM and NAND Flash cannot meet the needs of future technological applications in terms of either power consumption or data access speed. Moreover, as the semiconductor process line width shrinks below 14 nm and mainstream transistor technology migrates to advanced structures such as FinFET and GAA, the embedded NOR Flash memory unit that has long been used in CMOS chips has also been unable to keep up with SoC development needs. We must have new embedded, non-volatile memory technology in order to match the ASIC and MCU manufactured by the next generation of advanced processes.

Ferroelectric memory is a capacitive component based on the principle of spontaneous electric polarization that has not only reliable non-volatility but also extremely fast reading and writing speeds, high endurance and ultra-low power consumption. It also has great advantages in terms of process complexity and cost. As such, it is the most likely to become the emerging storage solution of the post Moore’s Law era.

Scholars were using the perovskite ferroelectric material BTO as early as the late 1950s to develop the first FeFET memory. The manufacturing process for the device was very simple. All that needed to be done was to replace the gate dielectric layer of the MOSFET transistor with a ferroelectric material. However, predictions made according to first-principles calculations say that, after the film thickness is about 6 single lattices below the critical value, the ferroelectricity of perovskite materials will deteriorate rapidly. This limits how much the memory density can be improved. Therefore, early product applications were limited to certain niche markets. Since German scholars discovered in 2011 that doped hafnium dioxide has good ferroelectric properties, however this material began attracting attention and was quickly introduced into the application development of ferroelectric memory.

HfO₂ is a ceramic material with a wide energy gap and high dielectric constant. In recent years, it has been widely used in advanced semiconductor manufacturing processes. It has been used to replace SiO₂ as the gate insulator in MOSFETs to solve the size limit problem faced by component miniaturization. The main advantage of HfO₂ over traditional perovskite ferroelectric materials is not only that it is fully compatible with the semiconductor process but also, more importantly, that HfO₂ thin films can retain their ferroelectricity even at a thickness of only 10 nm. Furthermore, it has a Curie Temperature of up to 470 °K, allowing it to operate at room temperature without issue.

HfO₂ only exhibits ferroelectric properties when it is doped with other elements and forms a special crystal phase. Doping elements commonly used in academic research include Si, Y, Sr, La, Ge and N, etc.. It has three main crystal phases: Monoclinic, Tetragonal and Orthorhombic. Among them, the Monoclinic phase has the lowest energy, but only the Orthorhombic has the required ferroelectric properties. How to select the appropriate type and ratio of doping elements, interface substances and annealing conditions, etc. to get HfO₂ to form a stable ferroelectric material is a subject still being researched. However, HfO₂ has already begun to show great market application potential as a ferroelectric memory material.

Whether it is ferroelectric or antiferroelectric, it belongs to the ferroelectric materials field of application. The formation of ferroelectricity is due mainly to the fact that, in some dielectric crystals, the structure of the unit cell results in positive and negative charge centers that do not coincide. This causes an electric dipole to appear, resulting in an electric polarization intensity not equal to zero, meaning that the crystal has spontaneous polarization. Usually, the directions of spontaneous polarization in a ferroelectric material are not the same except within small areas. Those small regions where the polarization of each unit cell is the same are called Ferroelectric Domains. The area between such domains are referred to as domain walls.

The polarization directions and strengths of ferroelectric domains differ, and they are randomly distributed throughout the material, canceling each other out so that the material as a whole exhibits no polarization. When an electric field is applied to the ferroelectric material, however, the polarization direction of the ferroelectric domains become consistent and reach Saturation Polarization. When the applied electric field exceeds the Positive Coercive Field or is lower than the Negative Coercive Field, the direction of the electric dipole of the ferroelectric material can be changed. When the applied electric field is removed, there will still be remnant polarization in the ferroelectric material. Therefore, in essence, ferroelectric memory is very suitable for use in non-volatile memory components.

Antiferroelectricity refers to when two adjacent single crystals in a material have parallel electric dipoles with opposite polarities, and the two reconstitute a new unit cell. When the applied electric field is zero, the two opposite and parallel polarities in the lattice cancel each other out so that the spontaneous polarization is zero macroscopically. When a sufficiently strong external electric field is applied, the polarity phase in the direction opposite that of the electric field can be reversed, forming a ferroelectric phase of the same polarity, at which point you can observe a double hysteresis loop similar to that of ferroelectric materials. However, when the electric field decreases and returns to zero, the curve will form another closed loop before reaching the linear region, so there is no remnant polarization.

Antiferroelectricity is a material characteristic that may be enhanced or weakened by changes in parameters such as temperature, pressure, external electric field and growth method. In particular, at a sufficiently high temperature, the ferroelectricity will disappear. Most antiferroelectric materials currently being used are made of HfO₂ doped with Zr, generally referred to as hafnium zirconium oxide. As the Zr doping ratio increases, the dielectric value of the HZO also increases. When the doping ratio of the Zr exceeds that of the Hf, the hysteresis loop of the HZO will change from ferroelectric to antiferroelectric.

According to recent research findings, antiferroelectric materials usually have better endurance, faster switching speeds, smaller operating biases and smaller wake-up effects than ferroelectric materials. In addition, with the appropriate processing and structural improvements, antiferroelectric materials can be used in volatile memory and non-volatile memory as well as to realize multi-level, high density memory functions. In 2022, the University of Tokyo in Japan successfully developed a 3D vertical field-effect-transistor that used antiferroelectric materials in place of the original ferroelectric materials in the gate insulators, thus verifying the concept of 3D stacked storage. The research was presented at the 2022 IEEE symposium on silicon nanoelectronics. There is a good chance that these materials will be chosen for use in the production of ultra-high density memory components with smaller sizes and lower power consumption in the future.

Incidentally, in addition to its applications in memory, antiferroelectric 2D materials can also be used as new optoelectronic materials which exhibit the diverse properties of superconductors, metals, semiconductors and insulators. It can also be used in traditional MOSFETs to improve the dielectric properties of the gate oxide through the negative capacitance effect. Using a negative capacitance ferroelectric material to make the gate layer will cause the current to increase faster relative to the original gate voltage, thereby reducing the transistor’s Subthreshold Swing. The subthreshold swing is a performance index that measures the transition rate between the On and Off states of the transistor. It represents the amount of change in the gate voltage required to change the source-drain current 10 times, also known as the S factor. The smaller the S value, the faster the ON/OFF rate.

At the end of 2019, the International Electron Device Meeting (IEDM) listed ferroelectric memory as a separate agenda topic, “Sec 15: Memory Technology-Ferroelectric”, for the first time. This move made clear the importance that both industry and academic circles place on this new trend of ferroelectric memory technology research and development. The most exciting recent achievement in ferroelectric memory research is the great leap forward it appears to have taken in terms of storage endurance. In particular, the discovery that doping HfO₂ with La increases the number of times that the memory can store over 1011 cycles. This is already approaching the endurance of DRAM. The Belgium research institute IMEC also presented their antiferroelectric HZO capacitor at the 2022 IEDM conference. Not only could it cycle up to 1011 times, but it also has a better hysteresis loop and a lighter wake-up effect. Antiferroelectric capacitor technology combines high performance, small sizes, and semiconductor process compatibility. They have the potential to become the key to enabling a new generation of embedded and stand-alone ferroelectric random-access memory (FeRAM) technology. In addition, in 2019, Purdue University used the semiconductor-compatible ferroelectric material α-In2Se3 and successfully combine transistors with “Ferroelectric RAM” to realize a new type of ferroelectric semiconductor field-effect-transistor structure with both information processing and storage functions and published it in the renowned Nature Electronics journal.

As more and more research teams around the world invest in its development, the future architecture of ferroelectric memory has begun to take shape. In the past, the properties of the ferroelectric materials used limited their application to niche markets. As research on more and more novel ferroelectric materials such as HfO₂ and HZO mature, however, it is believed that ferroelectric memory will soon see a new wave of industrial development opportunities. This article provided a comprehensive introduction to the development of “anti” ferroelectric materials and technology applications to help readers quickly learn about and understand this advanced technology and its excellent market potential. Professor Min-Hung Lee’s research expertise centers mainly on forward-looking transistors, high power components and solar cells. After obtaining a Ph.D. from National Taiwan University, he served in the Electronics Research and Service Organization (ERSO) and Display Image Technology Center (DTC) of the Industrial Technology Research Institute for several years, accumulating a wealth of industry-university research and development experience. Since 2007, Professor Lee has been teaching at Taiwan Normal University and devoting himself to academic research. He has published more than 150 journal papers and acquired numerous important invention patents. He has been awarded the National Science Council/Ministry of Science and Technology Outstanding Young Scholar Award twice and, in 2019, was elected as an IEEE Senior Member.

Professor Lee and his research team have been active participants in the “Ministry of Science and Technology’s Semiconductor Moon Shot Project – Key Technologies for Thermal Simulation of Materials, Processes, Components and Circuits for Next Generation Technology Nodes” since 2018. Their outstanding research also won the “2022 Future Technology Award”. He has made great contributions to enhancing Taiwan’s core technological advantages in the field of advanced semiconductors. MA-tek is very honored to cooperate with Professor Lee this year to carry out industry-university cooperation and provide all the analysis services that his team needs in the research of ferroelectric RAM processes. MA-tek has the complete testing equipment and professional technical experience to meet the various analysis and testing needs of advanced semiconductor components’ manufacturing processes, packaging and failure analysis.

Please feel free to reach out to us if you have any questions : marketing@matek.com

References：
[1]V. A. Isupov, “Ferroelectric and antiferroelectric perovskites Pb (B′0.5B′′0.5) O3.” Ferroelectrics, vol. 289, no. 1, pp. 131-195., 2003.
[2]G. A. Smolenskii, V. A. Bokov, V. A Isupov, N. N. Kraini, P. E Pasynkov, and A. I. Sokolov, “Ferroelectrics and related materials.” New York: Gordon and Breach, 1984.
[3]K. M. Rabe, “Antiferroelectricity in oxides: A reexamination” Functional metal oxides: new science and novel applications, pp. 221-244., 2013.
[4]X. Tan, C. Ma, J. Frederick, S. Beckman, and K. G. Webber “The Antiferroelectric ↔ Ferroelectric Phase Transition in Lead-Containing and Lead-Free Perovskite Ceramics” Journal of the American Ceramic Society, vol. 94, no. 12 pp. 4091-4107.,2011.
[5]T. Mitsui, “Ferroelectrics and Antiferroelectric” Springer Handbook of Materials Data, pp. 901-934., 2018.
[6]J. Valasek, “Piezo-Electric and Allied Phenomena in Rochelle Salt” vol. 17, no. 4, pp. 475, 1921.
[7]G. Shirane, E. Sawaguchi, and Y. Takagi, “Dielectric Properties of Lead Zirconate” Physical Review, vol. 84, no. 3, pp. 476, 1951.
[8]G. Shirane, “Ferroelectricity and Antiferroelectricity in Ceramic PbZrO3 Containing Ba or Sr.” Physical Review, vol. 86, no. 2, pp. 219, 1952.
[9]E. Sawaguchi, H. Maniwa, and S. Hoshino “Antiferroelectric Structure of Lead Zirconate” Physical Review, Vol. 83, no. 5, pp. 1078, 1951.
[10]C. Kittel “Theory of Antiferroelectric Crystals” Physical Review, vol. 82, no. 5, pp. 729., 1951.
[11]G.Shirane, and R. Pepinsky“Phase Transitions in Antiferroelectric PbHfO₃” Physical Review, vol. 91, no. 4, pp. 812., 1953.
[12]L.E. Cross, and B.J. Nicholson “LV. The optical and electrical properties of single crystals of sodium niobate” The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science , vol. 846, no. 376, pp. 453-466, 2009.
[13]L. E. Cross “VII. A thermodynamic treatment of ferroelectricity and antiferroelectricity in pseudo-cubic dielectrics” The Philosophical Magazine: A Journal of Theoretical Experimental and Applied Physics, vol. 1, no. 1, pp. 76-92., 1995.
[14]W. Cochran “Crystal stability and the theory of ferroelectricity” Advances in Physics, vol. 9, no. 36, pp. 387-423, 1960.
[15]H. Jaffe, “Piezoelectric Ceramics” Journal of the American Ceramic Society, vol. 41, no. 11, pp. 494-498, 1958.
[16]C. A. Randall, Z. Fan, I. Reaney, L. Q. Chen, and S. Trolier‐McKinstry , “Antiferroelectric: History, fundamentals, crystal chemistry, crystal structures, size effects, and applications” Journal of the American Ceramic Society,. vol. 104, no. 8, pp. 3775-3810, 2021.
[17]T. S. Böscke, J. Müller, D. Bräuhaus, U. Schröder, and U. Böttger “Ferroelectricity in hafnium oxide thin films” Applied Physics Letters, vol. 99, no. 10, pp. 102903, 1958.
[18]J. Müller, T. S. Böscke, U. Schröder, S. Mueller, D. Bräuhaus, U. Böttger, L. Frey, and T. Mikolajick “Ferroelectricity in Simple Binary ZrO2 and HfO₂” Nano letters, vol. 12, no. 8, pp. 4318-4323, 2012.
[19]D. L. Corker, A. M. Glazer, J. Dec, K. Roleder, and R. W. Whatmore“A Re-investigation of the Crystal Structure of the Perovskite PbZrO3 by X-ray and Neutron Diffraction” Acta Crystallographica Section B: Structural Science, vol. 53, no. 1, pp. 135-142, 1997.
[20]J. R. Reimers, S. A. Tawfik and M. J. Ford “van der Waals forces control ferroelectric–antiferroelectric ordering in CuInP2S6 and CuBiP2Se6 laminar materials” Chemical science, vol. 9, no. 39, pp. 7620-7627, 2018.
[21]C. Xu, Y. Chen, X. Cai, A. Meingast, X. Guo, F. Wang, Z. Lin, T. W. Lo, C. Maunders, S. Lazar, N. Wang, D. Lei, Y. Chai, T. Zhai, X. Luo, and Y. Zhu “Two-Dimensional Antiferroelectricity in Nanostripe-Ordered In2Se3” Physical Review Letters, vol. 125, no. 4, pp. 047601, 2020.
[22]M. H. Park et al., "Review and perspective on ferroelectric HfO₂-based thin films for memory applications," MRS Communications, vol. 8, no. 3, pp. 795-808, 2018.
[23]J. Muller et al., "Ferroelectricity in simple binary ZrO2 and HfO₂," Nano letters, vol. 12, no. 8, pp. 4318-4323, 2012.
[24]Milan Pešic et al., "Comparative Study of Reliability of Ferroelectric and Anti-Ferroelectric Memories" IEEE Trans Device Mater Reliab, VOL. 18, NO. 2, JUNE 2018
[25]M. Pešic´ et al., “Physical mechanisms behind the field-cycling behavior of HfO₂-based ferroelectric capacitors,” Adv. Funct. Mater., vol. 26 no. 25, pp. 4601–4612, Jul. 2016, doi: 10.1002/adfm.201600590.
[26]M. Pesic et al., “Root cause of degradation in novel HfO₂- based ferroelectric memories,” in Proc. IEEE Int. Rel. Phys. Symp. (IRPS) , Pasadena, CA, USA, 2016, pp. MY-3-1-MY-3-5, doi: 10.1109/IRPS.2016.7574619.
[27]K.-Y. Hsiang, Y.-C. Chen, F.-S. Chang, C.-Y. Lin, C.-Y. Liao, Z.-F. Lou, J.-Y. Lee, W.-C. Ray, Z.-X. Li, C.-C. Wang, H.-C. Tseng, P.-H. Chen, J.-H. Tsai, M. H. Liao, T.-H. Hou, C. W. Liu, P.-T. Huang, P. Su, and M. H. Lee, “Novel Opposite Polarity Cycling Recovery (OPCR) of HfZrO2 Antiferroelectric-RAM with an Access Scheme Toward Unlimited Endurance, ” accepted by Technical Digest, International Electron Device Meeting (IEDM) , San Francisco, Dec. 3-7, 2022.
[28]X. Lyu, M. Si, X. Sun, M. A. Capano, H. Wang and P. D. Ye, "Ferroelectric and Anti-Ferroelectric Hafnium Zirconium Oxide: Scaling Limit, Switching Speed and Record High Polarization Density," 2019 Symposium on VLSI Technology, 2019, pp. T44-T4
[29]Y. Chen, K. Hsiang, Y. Tang, M. -H. Lee and P. Su, "NLS based Modeling and Characterization of Switching Dynamics for Antiferroelectric/Ferroelectric Hafnium Zirconium Oxides," 2021 IEEE International Electron Devices Meeting (IEDM) , 2021, pp. 15.4.1-15.4.4,
[30]K.-Y. Hsiang, C.-Y. Liao, Y.-Y. Lin, Z.-F. Lou, C.-Y. Lin, J.-Y. Lee, F.-S. Chang, Z.-X. Li, H.-C. Tseng, C.-C. Wang, W.-C. Ray, T.-H. Hou, T.-C. Chen, C.-S. Chang, and M. H. Lee, “Correlation between Access Polarization and High Endurance (˜ 1012 cycling) of Ferroelectric and Anti-Ferroelectric HfZrO2,” IEEE International Reliability Physics Symposium (IRPS) , 2022, pp. pp. P9-1-P9-4.
[31]M. Pešić, U. Schroeder, S. Slesazeck, and T. Mikolajick, “Comparative Study of Reliability of Ferroelectric and Anti-Ferroelectric Memories, ” IEEE Transactions on Device and Materials Reliability, vol. 18, no. 2, pp. 154-162, 2018.
[32]S.-C. Chang, N. Haratipour, S. Shivaraman, T. L. Brown-Heft, J. Peck, C.-C. Lin, I.-C. Tung, D. R. Merrill, H. Liu, C.-Y. Lin, F. Hamzaoglu, M. V Metz, I. A Young, J. Kavalieros, and U. E. Avci, "Anti-ferroelectric HfxZr1-xO2 Capacitors for High-density 3-D Embedded-DRAM." IEEE International Electron Devices Meeting (IEDM) , 2020, pp. 28.1.1-28.1.4.
[33]K.-Y. Hsiang, C.-Y. Liao, J.-H. Liu, J.-F. Wang, S.-H. Chiang, S.-H. Chang, F.-C. Hsieh, H. Liang, C.-Y. Lin, Z.-F. Lou, T.-H. Hou, C. W. Liu, and M. H. Lee, “Bilayer-based Antiferroelectric HfZrO2 Tunneling Junction with High Tunneling Electroresistance and Multilevel Nonvolatile Memory,” IEEE Electron Device Letters, vol. 42, no. 10, pp. 1464-1467, 2021.
[34]K. . -Y. Hsiang et al., "Bilayer-Based Antiferroelectric HfZrO2 Tunneling Junction With High Tunneling Electroresistance and Multilevel Nonvolatile Memory," in IEEE Electron Device Letters, vol. 42, no. 10, pp. 1464-1467, Oct. 2021, doi: 10.1109/LED.2021.3107940.
[35]B. Max, M. Hoffmann, S. Slesazeck and T. Mikolajick, "Direct Correlation of Ferroelectric Properties and Memory Characteristics in Ferroelectric Tunnel Junctions," in IEEE Journal of the Electron Devices Society, vol. 7, pp. 1175-1181, 2019, doi: 10.1109/JEDS.2019.2932138.
[36]K. . -Y. Hsiang et al., "Dielectric Layer Design of Bilayer Ferroelectric and Antiferroelectric Tunneling Junctions Toward 3D NAND-Compatible Architecture," in IEEE Electron Device Letters, vol. 43, no. 11, pp. 1850-1853, Nov. 2022, doi: 10.1109/LED.2022.3204445.
[37]Y. Chen et al., "NLS based Modeling and Characterization of Switching Dynamics for Antiferroelectric/Ferroelectric Hafnium Zirconium Oxides," 2021 IEEE International Electron Devices Meeting (IEDM) , 2021, pp. 15.4.1-15.4.4, doi: 10.1109/IEDM19574.2021.9720645
[38]K. Ni et al., "A Novel Ferroelectric Superlattice Based Multi-Level Cell Non-Volatile Memory," 2019 IEEE International Electron Devices Meeting (IEDM) , 2019, pp. 28.8.1-28.8.4, doi: 10.1109/IEDM19573.2019.8993670
[39]C.-Y. Liao et al., “Multipeak coercive electric-field-based multilevel cell nonvolatile memory with antiferroelectric-ferroelectric field-effect transistors (FETs) ,” IEEE Trans. Ultrason., Ferroelectric., Freq. Control, vol. 69, no. 6, pp. 2214‒2221, Jun. 2022.