Although 10 years have passed since the initial report of ferroelectricity in hafnia (HfO2), researchers are still intensely fascinated by this material system and the promise it holds for future applications. A wide variety of deposition methods have been deployed to create ferroelectric HfO2 thin films such as atomic layer deposition, chemical solution deposition, and physical vapor deposition methods such as sputtering and pulsed laser deposition. Process and design parameters such as deposition temperature, precursor choice, target source, vacuum level, reactive gases, substrate strain, and many others are often integral in stabilizing the polar orthorhombic phase and ferroelectricity. We examine processing parameters across four main different deposition methods and their effect on film microstructure, phase evolution, defect concentration, and resultant electrical properties. The goal of this review is to integrate the process knowledge collected over the past 10 years in the field of ferroelectric HfO2 into a single comprehensive guide for the design of future HfO2-based ferroelectric materials and devices.

HfO2 rose to prominence in the semiconductor industry when Intel announced the deployment of HfO2 high-k dielectrics in 32 nm technologies in 2008.1 While also investigating HfO2 for high-k dielectric applications for use in dynamic random access memories (DRAMs) in 2011, Böscke et al. reported the discovery of ferroelectricity in HfO2 thin films when doped with silicon and capped with TiN electrodes.2 Ferroelectricity, the property of a material with a spontaneous dielectric polarization that can be reversed with the application of an electric field, has many useful low-power memory, sensing, and energy-related applications. In the movement beyond Moore’s Law to develop ultrahigh memory- and energy-density technologies, ferroelectric materials will need to be eventually integrated into three-dimensional (3D) architectures.3 Current ferroelectric materials, however, are limited in their adaptability to 3D nanoscale capacitors due to (1) the suppression or loss of ferroelectric properties below certain critical thicknesses or sizes and (2) the lack of wide-area compatible deposition processes. Therefore, it is necessary to explore novel material candidates such as HfO2, which exhibit ferroelectricity at small thickness and can be fabricated via commercial deposition methods such as atomic layer deposition (ALD), physical vapor deposition (PVD), or other high-throughput and cost-effective methods. Methods such as pulsed laser deposition (PLD) are also useful for creating epitaxial films, which can help elucidate the origins of nanoscale ferroelectricity in HfO2.

Indeed, a unique feature of HfO2 thin films is that they can exhibit ferroelectricity in thin films (<30 nm), whereas conventional ferroelectric materials such as lead zirconate titanate [Pb(Zr,Ti)O3 or PZT], barium titanate (BaTiO3), and strontium-bismuth tantalate (SrBi2Ta2O9) typically require thicker films (>50 nm) to fully realize ferroelectricity.4 In fact, remanent ferroelectric polarization in HfO2 tends to increase as the thickness decreases down to a thickness in the range of 8–10 nm.5 Despite the relatively short history of ferroelectric HfO2, numerous applications have been proposed including ferroelectric nonvolatile memory6 like ferroelectric random access memory,7 ferroelectric field effect transistors,8 ferroelectric tunneling junctions,9 pyroelectric energy harvesters,10 electrocaloric cooling,11 supercapacitor energy harvesters,12 and neuromorphic computing.13 

Not only are HfO2 thin films suitable for ferroelectric applications, but HfxZr1−xO2 (HZO) forms a complete solid solution (Fig. 1) and has also shown useful dielectric, ferroelectric, and antiferroelectric-like properties.14 Note that, in the ferroelectric HfO2 research community, proof of true antiferroelectric origins of the pinched hysteresis is not yet widely accepted and, therefore, films showing properties reminiscent of antiferroelectrics are often referred to as antiferroelectric-like. This terminology allows a description of the effect without a claim to the origin. The most frequently studied HfxZr1−xO2 composition is x ∼ 0.5, which demonstrates robust ferroelectricity.

FIG. 1.

HfO2-ZrO2 standard pressure phase diagram. The monoclinic phase is thermodynamically most stable at room temperature for bulk materials whereas in HfxZr1−xO2 thin films, surface energy effects stabilize the tetragonal phase. Nonequilibrium factors such as quench cooling can result in a metastable orthorhombic phase not predicted by the standard phase diagram. Reprint permission obtained from Phase Equilibria Diagrams Online Database (NIST Standard Reference Database 31), Figure number Zr-186. Copyright 2021, The American Ceramic Society and the National Institute of Standards and Technology.

FIG. 1.

HfO2-ZrO2 standard pressure phase diagram. The monoclinic phase is thermodynamically most stable at room temperature for bulk materials whereas in HfxZr1−xO2 thin films, surface energy effects stabilize the tetragonal phase. Nonequilibrium factors such as quench cooling can result in a metastable orthorhombic phase not predicted by the standard phase diagram. Reprint permission obtained from Phase Equilibria Diagrams Online Database (NIST Standard Reference Database 31), Figure number Zr-186. Copyright 2021, The American Ceramic Society and the National Institute of Standards and Technology.

Close modal

In addition to their thickness scaling advantage, HfO2 and its isomorph, ZrO2, are also compatible with complementary metal-oxide-semiconductor processes, have high polarization density (Pr > 24 μC/cm2) and have a relatively large bandgap energy across the HfO2-ZrO2 composition (>5 eV).15–18 By tuning the chemical composition, engineering the defect concentration, and controlling the crystalline structure, HZO-based devices can be fabricated with a wide range of desired properties.

The appearance of ferroelectricity and resultant performance of HfO2 films heavily depends on the underlying chemical and structural characteristics, which, in turn, are often dictated by the details of the deposition process. Factors such as precursor/target material selection, deposition temperature, reactive gas flow, deposition pressure, and composition can significantly alter the functional properties of the resultant HfO2 films. Thus, a well-defined deposition process is required to synthesize films with desired properties. A fundamental understanding of the chosen deposition technique is also necessary to take advantage of its full potential and understand its limitations. Often the desired application of the HfO2 thin film will determine the most suitable deposition method and parameters. For example, to better understand the effects of strain engineering in HfO2-based materials, it may also be necessary to deposit single-phase or epitaxially strained HfO2 using methods such as pulsed laser deposition. For other applications, such as high area density or 3D substrates, a conformal technique such as atomic layer deposition may be preferred.

As the research field of HfO2 ferroelectrics continues to grow, a comprehensive review of ferroelectric HfO2 deposition methods is necessary to integrate and synthesize knowledge from different techniques and reports. This review introduces the major film deposition techniques currently deployed in the ferroelectric HfO2 community and discusses the processing-structure-property relationships of the resulting ferroelectric HfO2 and HZO films. Since device applications and limitations of ferroelectric HfO2 are reviewed elsewhere,9,19,20 the present work provides only a brief glimpse of emerging applications for ferroelectric HfO2 thin films. And while defects will be discussed in relation to the selected fabrication method, a more extensive review on defects in fluorite structures, and especially oxygen defects, can be found elsewhere.9,21–23 The present contribution, instead, provides a single and comprehensive reference to aid both academic and applied researchers in selecting and refining a HfO2 thin film deposition process. The paper is organized as follows: first, an overview of ferroelectricity in HfO2 is given, where both thermodynamic and kinetic considerations are discussed. Next, the conventional fabrication procedures for HfO2-based metal-ferroelectric-metal (MFM) devices are described, and four main deposition methods are discussed in detail. Each deposition method section gives an overview of the mechanism of the deposition technique and is then further divided into relevant process parameters that have been studied for HfO2-based ferroelectric materials.

Böscke et al. reported the first ferroelectric HfO2 doped with Si and ascribed the appearance of ferroelectricity to a noncentrosymmetric and polar orthorhombic phase (Pca21).2 This space group was not predicted in the HfO2 phase diagram. Bulk HfO2 is instead known to exist in three different crystal structures at ambient pressures: monoclinic (P21/c), tetragonal (P42/nmc), and cubic (Fm3¯m). And while the tetragonal phase in HfO2 can be stabilized at lower temperatures in thin films due to surface energy effects, the Pca21 phase in HfO2 only emerged when Böscke et al. capped the HfO2 thin film with a TiN top layer. Later, Huan et al. used first-principle calculations to describe a thermodynamic pathway for the emergence of the polar orthorhombic phase, which was said to originate from the tetragonal phase.24 Batra et al. further reported that surface energy effects could aid in stabilizing the polar orthorhombic phase, which also occur in nanoparticles of ZrO2, the isomorph of HfO2.25 In view of these first-principle studies and in conjunction with microscopy evidence from Sang et al., it is generally accepted that ferroelectricity in HfO2 is due to the Pca21 orthorhombic phase.26 Furthermore, the emergence of the orthorhombic phase requires highly strained conditions in which the shearing of the tetragonal lattice stabilizes the polar orthorhombic phase; in this sense, the Pca21 is commonly thought of as a strained tetragonal phase. Wei et al. identified another possible ferroelectric phase of rhombohedral (R3m) in epitaxially strained Hf0.5Zr0.5O2 thin films formed on an La0.7Sr0.3MnO3 (LSMO)/SrTiO3 (STO) substrate, which was suggested to emerge from a strained cubic phase.27 

A vital commonality among these reports is the role of strain as a driving factor for inducing the polar phase in HfO2 thin films. Mechanisms by which strain can promote the ferroelectric phase include anisotropic stress from capping layer confinement,2 thermal expansion mismatch,28 surface energy and grain size effects,29 dopants,30,31 and defects.21,22 These driving factors influence the stress state within the film and affect the resultant phase evolution. Deposition parameters including the material/dopant source, deposition temperature, oxygen content, layering, deposition pressure, epitaxial interfaces, electrode material, and annealing can all play a significant role in engineering desired properties of an HfO2 device. Common driving factors that are ascribed to induce ferroelectricity are summarized in Fig. 2(a). Benchmark values that are commonly reported include the remanent polarization (±Pr), which is the polarization value at zero applied field, i.e., the y-intercept of the P-E loop. Another characteristic of a ferroelectric device is also the coercive field (±Ec), which is taken at the x-intercept of the P-E loop. Ec is a measure of the strength of the reversing polarizing field E, which is required to erase the remanent polarization of the material and is related to the “memory window.” To better understand the relationship between processing parameters and resultant ferroelectric properties in HfO2 materials such as Pr and Ec, it is important to first consider the thermodynamic and kinetic driving factors that influence the emergence of the ferroelectric phase, some of which are schematically summarized in Fig. 2.

FIG. 2.

(a) MFM capacitor and the possible driving factors that contribute to ferroelectric polarization and (b) representative polarization-electric field (P-E) loop from a 10 nm thick Hf0.5Zr0.5O2 film.

FIG. 2.

(a) MFM capacitor and the possible driving factors that contribute to ferroelectric polarization and (b) representative polarization-electric field (P-E) loop from a 10 nm thick Hf0.5Zr0.5O2 film.

Close modal

Ferroelectricity in HfO2 has been achieved using numerous dopants including Si4+,32 Y3+,33 Sr2+,34 la3+,35 Gd3+,36 Al3+,37 N3−,38 and Zr4+.39 The effect of dopants on the free energy of various phases in HfO2 has been studied even prior to the initial reporting of ferroelectricity. In 1991, Lowther et al. reported that the phases of HfO2 lie within a very close energy range (∼10 meV) and that remanent lattice stresses could affect which phase ultimately dominates.39 Kisi and Howard also demonstrated the likely existence of the Pca21 phase in ZrO2, although the presence of ferroelectricity was not investigated at that time.40 After the initial report of ferroelectricity in HfO2, Materlik et al. showed that in pure HfO2 the total energy difference of the polar orthorhombic is higher than that of the stable monoclinic phase by ∼62 meV per formula unit (f.u.).29 The polar orthorhombic phase, thus, could only be thermodynamically stabilized if the total energy of the orthorhombic phase was lowered. Materlik et al. also showed that the free energy of the orthorhombic phase could be lowered by increasing ZrO2 alloying concentration.

More recently in an investigation of nearly 40 different dopants, Batra et al. showed via first-principles calculations that numerous dopants could induce a distortion of the lattice cell and promote Pca21 to the lowest free energy.41 They confirm the existence of a small energy difference between the equilibrium monoclinic phase and the polar orthorhombic which could easily be perturbed by external factors, e.g., dopants, strain, and oxygen vacancies. Given the similar parabolic relationship between dopant concentration and ferroelectric polarization among a variety of dopants, Xu et al. suggested that there may be the same underlying driving force responsible for stabilizing the orthorhombic phase.42 However, doping HfO2 films is not required to stabilize the Pca21 phase as indicated by numerous reports citing ferroelectricity in undoped HfO2.43,44

Materlik et al. suggested that the polar orthorhombic phase could also be stabilized via surface energy effects such as critical film thickness or grain size.29 It was suggested that a limited film thickness regime was necessary to limit the grain size where in smaller grains, the tetragonal phase would be preferred over the monoclinic. Furthermore, an earlier report by Batra et al. also suggested that, at small length scales, surface energy could help stabilize the orthorhombic phase.25 Experimental work conducted by Park et al. confirmed the computational findings, showing that the polar orthorhombic phase decreased monotonically with an increase in film thickness above 10 nm.45 However, while Materlik et al. proposed that only a grain size of <5 nm could thermodynamically stabilize the orthorhombic phase, experimental evidence by Park et al. showed that ferroelectricity in HZO films could be achieved in films with an average grain size over 5 nm.46 Given that ferroelectricity could be achieved at larger grain sizes, surface energy effects from grain size alone could not explain the stabilization of the orthorhombic phase. Therefore, the effect of defects cannot be ignored in considering the driving factors for thermodynamic stabilization of the polar orthorhombic phase.

Oxygen vacancies favorably increasing the orthorhombic phase in HfO2 have been seen in several studies.30,47 Hoffmann et al. considered the effect of TiN versus TaN electrodes on the ferroelectric properties of Gd:HfO2, suggesting that a higher remanent polarization in films with TaN electrodes could be due to higher oxygen vacancy concentration in HfO2.30 Hoffmann et al. reported that oxygen vacancies could decrease the free energy of the orthorhombic and tetragonal phases by 15 and 24 meV per f.u., respectively. In addition to the nonequilibrium effects of surface energy, dopants, and defects, in Sec. I C, we also discuss the role of kinetics in helping to promote ferroelectricity in HfO2.

Thermodynamics predicts that the monoclinic phase is the most stable at room temperature in bulk HfO2, but experimental evidence indicates a different reality in thin films. Park et al. considered the effects of phase transition kinetics on stabilizing the ferroelectric phase of HfO2 asserting that the tetragonal to orthorhombic phase transformation occurs during cooling in a martensitic-like process.48–50 Park et al. calculated the free energy barriers of the tetragonal to monoclinic phase transformation to be 223–263 meV per f.u. whereas the free energy barrier for the tetragonal to orthorhombic phase transformation was 22–31 meV per f.u.49 Therefore, it is likely that the tetragonal to monoclinic phase transformation is kinetically suppressed until a critical temperature. At even higher temperatures, the kinetic barrier for the tetragonal to orthorhombic phase is reached, and the monoclinic phase appears. This kinetic model explains the experimental results of diffraction studies which show a peak shift of the orthorhombic 111 (o111) and/or tetragonal 011 (t011) peak ∼30.3° peak toward lower 2θ in the work of Park et al. during in situ heating experiments for 30 nm Al-, Gd-, and Sr-doped HfO2 films.50 The peak shift toward lower 2θ was related to a change in lattice parameters, aspect ratio [c/a for tetragonal and 2a/(b + c) for orthorhombic], and unit cell volumes.50 Thus, a peak shift could provide direct evidence for a phase transformation. Shiraishi et al. estimate that the Curie temperature (Tc) of HfO2 films is found to be around 400–500 °C, although the Tc has been shown to be highly dependent on factors such as thickness and composition.51,52 In a high-temperature x-ray diffraction investigation of the HfO2-ZrO2 composition by Hsain et al., a 2θ peak shift of the o111/t011 as a function of annealing temperature was observed starting at ∼800 °C, suggesting a much higher Tc than was reported previously.53 The peak shift observed for compositions x > 0.5 in Hfx−1ZrxO2 could not be well-fitted to a linear trend, which further suggested that the origin of the peak shift was not solely thermal expansion. Hsain et al. ascribed the peak shift during in situ heating to a tetragonal-to-orthorhombic phase transformation. This finding posited that the orthorhombic phase could also be achieved during in situ slow heating as opposed to only during cooling in a martensitic process.

Lastly, we will consider the kinetic effects of rapidly cooling (quenching) in HfO2 films after a high-temperature annealing step. In a study by Toriumi et al., the remanent polarization as a function of quenching rate and mol. % Y was investigated.54 It was found that remanent polarization in undoped HfO2 was the most sensitive to quench rate, dropping by 50% when the quench rate decreased from 10 to 4 °C/s. As the doping concentration increased, however, the sensitivity to Pr on the quench rate decreased. A doping concentration of 1.7 mol. % Y was found to be relatively insensitive to the quench rate. At high quench rates, the remanent polarization is the same regardless of the doping concentration. However, Pr changes at different rates for different doping concentration when the quench rate decreases. The change in Pr “sensitivity” with different dopant concentration suggests that dopants may help to stabilize oxygen defects resulting in the stabilization of the ferroelectric phase. Without the addition of dopants, the longer time under thermal treatment relaxes the tetragonal phase into the monoclinic phase and degrades the ferroelectric properties. Thus, a high quench rate is necessary to stabilize the ferroelectric phase in undoped or lightly doped HfO2 films to suppress the thermodynamically stable monoclinic phase.

Now that we have highlighted the multiple thermodynamic (dopant, strain, grain size, defects) and kinetic (annealing temperature and cooling/heating rate) effects that help to stabilize ferroelectricity in HfO2, four major deposition methods, and their effect on ferroelectricity in HfO2 are reviewed.

HfO2-based thin films can be fabricated into an MFM capacitor stack, which allows for straightforward electrical measurements, e.g., polarization-electric field (P-E) hysteresis, current-voltage (I-V), and endurance testing (i.e., change in Pr as a function of cycling). MFM devices are typically fabricated on Si and follow the device architecture from bottom to top: Si/SiO2/metal/HfO2/metal/Ti/Pt. A schematic of a typical MFM is shown in Fig. 2. While other device configurations of HfO2 films are possible, this review focuses primarily on the MFM device structure. The conventional fabrication route follows the process flow shown in Fig. 3: (1) A substrate such as Si is cleaned and prepared for film deposition and bottom electrode, e.g., TiN, is deposited, followed by (2) HfO2 with or without a dopant, then (3) the top electrode/capping layer, typically of the same material as the first electrode layer, followed by (4) annealing treatment to crystallize the film at 400–1000 °C for 1–60 s, and then (5) deposition of a metal patterned electrode via sputtering or evaporation, and finally, (6) pattern transfer to the top TiN layer via etching in SC1. In Subsections II AII E, we focus our attention on step 3, that is, the deposition method utilized for the HfO2-based MFMs. Selection of electrode/substrate material and annealing processes are discussed as they relate to the HfO2 deposition process and selection. Due to the multitude of process parameters for each deposition technique, we highlight only those that are most relevant and of direct interest to the ferroelectric HfO2 research community.

FIG. 3.

Conventional fabrication procedure includes (1) deposition of bottom electrode via sputtering, (2) ferroelectric layer deposition using one of the surveyed methods (ALD, sputter deposition, PLD, or CSD), (3) top electrode deposition via sputtering, (4) annealing, (5) patterning of electrodes using photolithography or hard mask, and (6) etching excess TiN.

FIG. 3.

Conventional fabrication procedure includes (1) deposition of bottom electrode via sputtering, (2) ferroelectric layer deposition using one of the surveyed methods (ALD, sputter deposition, PLD, or CSD), (3) top electrode deposition via sputtering, (4) annealing, (5) patterning of electrodes using photolithography or hard mask, and (6) etching excess TiN.

Close modal

ALD is a vapor deposition technique known for its high degree of conformality and precise control over layer thickness.54–56 The conformality and reproducibility of the ALD process stems from the controlled surface deposition reactions which proceed in a sequential step-wise set of self-limiting half-reactions. An example ALD surface reaction for HfO2 from tetrakis(dimethylamino)hafnium, Hf(N(CH3)2)4 (TDMA-Hf), and water is illustrated in Fig. 4. A starting surface with available –OH groups is first exposed to a metalorganic precursor reactant, i.e., Hf(N(CH3)2)4, where the surface temperature and net reactant exposure are sufficient for the surface reaction to achieve completion (i.e., the reaction is self-limiting) without significant self-decomposition of the precursor. The reactor is then purged of unreacted precursor and product volatiles and is then exposed to a second reactant, e.g., water, to complete the second half-reaction, nominally producing a single monolayer of film. The reactor is again purged, and the cycle begins again. The reactive ligands on the metal precursor can effectively block reactions on neighboring available –OH sites, so experimentally, the total amount of material deposited per cycle is typically less than one full material monolayer. The film growth is highly conformal because the gas phase precursors are able to undergo self-limiting reactions at reactive sites on all exposed surfaces. Furthermore, the surface reaction temperature during ALD is typically less than that used during other CVD processes. Reactions during CVD commonly involve significant reactant decomposition in the gas phase near the deposition surface, whereas ALD occurs primarily via thermodynamically driven surface reactions without significant gas phase decomposition.57 

FIG. 4.

ALD process for HfO2 using Hf(N(CH3)2)4 and water as precursors.

FIG. 4.

ALD process for HfO2 using Hf(N(CH3)2)4 and water as precursors.

Close modal

The primary advantages of ALD include its precise control over film thickness, its conformality that allows deposition onto high-aspect ratio substrates, and lower process temperatures. To achieve ALD, the surface temperature must be high enough to enable reaction, but it must not exceed that needed for reactant decomposition. For ALD processes, the temperature range that fits this criterion is referred to as the “ALD temperature window” or simply “ALD window.” Typically, low temperature leads to slow growth due to reaction rate limitations and high temperature leads to decomposition and high growth rates under CVD-like conditions. For some ALD reaction systems, precursors can also physisorb at low temperatures leading to excess growth. Likewise, some reactants may desorb at high temperature without reacting, leading to decreased growth at high temperature. Key advantages such as subnanometer thickness control, high degree of conformality, and low deposition temperature have made ALD a widely used technique in the ferroelectric HfO2 community. Here, we review the choice of precursor, oxygen reactant, reaction temperature, and layering/mixing sequence in fabricating HfO2-based thin films via ALD. Process variables for ALD-based films and their reported remanent polarization, coercive field, and cycling endurance are found in Table I.

TABLE I.

Process parameters of HfO2 thin films fabricated via atomic layer deposition.

MaterialaDoping %t (nm)Dopant precursorHfO2 precursorOxygen sourceTd (°C)Annealing conditionsTop/bottom electrodesPrEcEnduranceRef.
HfO2 None 6.9 None TDMA-Hf Ozone 260 650 TiN/TiN 13.5 N/A 78  
HfO2 None None TEMA-Hf Ozone 280 650/30s/N2 TiN/TiN 10.4 108 using 3.2 MV/cm 44  
Al:HfO2 4.8 mol. % 16 TMA TEMA-Hf Ozone N/A 800,1000/20s/N2 TiN/TiN N/A 37  
Al:HfO2 N/A 10 TMA HfCl4 O2 plasma 300 800/20s/N2 TiN/TiN 15 N/A N/A 50  
Al:HfO2 3 at. % 8.5 TMA TEMA-Hf Ozone 250 700/10s/N2 TiN/TiN 5.1 N/A 77  
Al:HfO2 2.2 cat. % 10 TMA TEMA-Hf Ozone 280 650/20s TiN/TiN 16.5 N/A 95  
Gd:HfO2 3.4 cat. % 100 Gd(iPrCp)3 HfCl4 H2300 650/20s TaN/TaN 32 1.5 106 using 4 MV/cm 30  
Gd:HfO2 2 mol. % 10 Gd(iPrCp)3 HfCl4 H2300 1000/1s TiN/TiN 12 1.75 N/A 36  
HZO/Al2O3 50 at. % 12 TDMA-Zr TDMA-Hf N/A N/A N/A Ti/Au /p-type Si 21 7.5 N/A 13  
La:HfO2 10 cat. % 12 La-(iPrCp)3 TEMA-Hf H2O/O2 280 800C/20s/ N2 TiN/TiN 28 105 using 4.2 MV/cm 35  
La:HfO2 6.0 cat. % 10 La-(iPrCp)3 TEMA-Hf Ozone 280 650/20 TiN/TiN 23.6 N/A 95  
Si:HfO2 4 mol. % 10 4DMA-S TEMA-Hf Ozone N/A 1000C/20s/N2 TiN/TiN 10 N/A 2  
Si:HfO2 3.8 mol. % 10 TDMA-Si TEMA-HF N/A 350 650C/N2 TiN/TiN 15 106 using 3 MV/cm 6  
Si:HfO2 2-8.5 mol. % 10 TDMAS TDMA-Hf O2 200 700–900/5–60s/ N2 TiN/TiN 16 N/A 32  
Si:HfO2 4 mol. % 10 TDMA-S TDMA-Hf O2 plasma 200 900/40s TiN/TiN 15 1.5 N/A 32  
Si:HfO2 2.7 cat. % 10 3DMAS TEMA-Hf Ozone 280 650/20s TiN/TiN 18.8 N/A 95  
Si:HfO2 in 3D trenches 13:1 aspect ratio N/A 10 N/A N/A N/A N/A N/A TiN/TiN 14 1010 using 2 MV/cm 9  
Sr:HfO2 7.9 at. % 10 Sr(tBu3Cp)2 HfCl4 H2300 800/20s/N2 TiN/TiN 23 106 using
3 MV/cm 
34  
Sr:HfO2 N/A 10 N/A N/A N/A 300 800/20s/ N2 TiN/TiN 17.5 N/A N/A 47  
Y:HfO2 5.2 mol. % 10 Y(MeCp)3 TEMA-Hf Ozone N/A 600/20s/N2 TiN/TiN 24 1.2 N/A 33  
Zr:HfO2 50 at. % 20 TDMA-Zr TDMA-Hf Ozone 250 400C/60s/N2 TiN/TiN 25.45 N/A 5  
Zr:HfO2 50 at. % 2.5 TEMA-Zr TEMA-Hf H2240 400 TiN/Si N/A 16  
Zr:HfO2 50 at. % 10 N/A N/A N/A 500 500 TiN/TiN 17 NA 14  
Zr:HfO2 50 at. % 10 TEMA-Zr TEMA-Hf Ozone 280 600 TiN/TiN 15 1.2 NA 45  
Zr:HfO2 50–70 at. % TEMA-Zr TEMA-Hf Ozone 280 800/20s/N2 TiN/TiN 15 1.5 N/A 49  
Zr:HfO2 50 at. % 17 TEMA-Zr Hf(N(CH3)2) (C8H17N2N/A 350 700/10s/O2 [Pt]/[Pt/TiO2/SiO29.3 N/A 110  
Zr:HfO2 50 at. % N/A TEMA-Zr TEMA-Hf Ozone, H2∼270–300 600/20s/N2 TiN/TiN 13 N/A N/A 60  
Zr:HfO2 50 at. % N/A CpZr[N(CH3)2]3 TEMA-Hf Ozone, H2∼270–300 600/20s/N2 TiN/TiN 17.5 N/A N/A 60  
Zr:HfO2 50 at. % 10 TEMA-Zr TEMA-Hf ozone 260 450/N2 TiN/TiN 15 109 using
2.5 MV/cm 
67  
Zr:HfO2 50 at. % 10 TDMA-Zr TDMA-Hf ozone 260 450/N2 TiN/TiN 20 109 using 2.5 MV/cm 67  
Zr:HfO2 50 at. % 10 TDMA-Zr TDMA-Hf O3 250 400/60s/N2 TiN/TiN 24 1.1 N/A 65  
Zr:HfO2 50 at. % 10 TDMA-Zr TDMA-Hf H2250 400/60s/N2 TiN/TiN 19 1.2 N/A 67  
Zr:HfO2 50 at. % 10 CpZr[N(CH3)2]3 Cp-Hf[N(CH3)2]3 Ozone 300 450/20s/N2 TiN/TiN 17.5 105 using
4 MV/cm 
73  
Zr:HfO2 50 at. % 10 CpZr[N(CH3)2]3 Cp-Hf[N(CH3)2]3 O2 plasma 300 450/20s/N2 TiN/TiN 17.5 1.5 104 at
4 MV/cm 
73  
Zr:HfO2 50 at. % 10 CpZr[N(CH3)2]3 TEMA-Hf Ozone 280 600/20s/N2 TiN/TiN 19 N/A 79  
Zr:HfO2 50 at. % 8.4 TEMA-Zr TEMA-Hf Ozone 280 500/30s/N2 TiN/TiN 12 1.2 N/A 96  
Zr:HfO2 50 at. % 8.4 TEMA-Zr TEMA-Hf Ozone 280 500/30s/N2 Pt/TiN 1.5 N/A 80  
Zr:HfO2 50 at. % 10 CpZr[N(CH3)2TEMA-Hf Ozone 230–300 600/20s/N2 TiN/TiN 22 N/A N/A 80  
Zr:HfO2
(6 nm HfO2 /
6 nm ZrO2
50 at. % 12 TDMA-Zr TDMA-Hf H2150 600/20/N2 [Au/Ni]/[Pt/Ti/SiO2/Si] 12 N/A 89  
Zr:HfO2 nanolaminate (1 nm HfO2/1 nm ZrO2) × 4 50 at. % ZyALD TDMA-Hf Ozone 285 500/10 min/N2 TiN/TiN 25 N/A 87  
Zr:HfO2 nanolaminate (3 nm HfO2/
3 nm ZrO2
50 at. % ZyALD TDMA-Hf Ozone 260 500/10 min/N2 TiN/TiN 12 1.1 N/A 88  
Zr:HfO2 nanolaminate H15Z50H15 60 at. % TEMA-Zr TEMA-Hf Ozone 280 N/A TiN/TiN 12 1010 at 2.5 MV/cm 86  
Zr:HfO2 with 1.6 nm Al2O3 50 at. % 21.6 TEMA-Zr, TMA TEMA-Hf H2O 250 (HfO2), 150 (Al2O3450/30s/N2 TiN/TiN 30.19 105 at 2.7 MV/cm 145  
Zr:HfO2 with 1 nm Al2O3 middle layer 50 at. % 40 TEMA-Zr, TMA TEMA-Hf Ozone 280 500/30s/N2 TiN/TiN 11.4 N/A 92  
Zr:HfO2 with 2 nm Al2O3 50 at. % 20 TDMA-Zr, TMA TDMA-Hf Ozone 280 (HfO2), 250 (Al2O3550/30s/N2 TiN/TiN 16 108 at 2.5 MV/cm 93  
Zr:HfO2 with HfO2 seed layer 50 at. % 10 TEMA-Zr TEMA-Hf Ozone 300 600 TiN/TiN 22.1 1.11 N/A 90  
MaterialaDoping %t (nm)Dopant precursorHfO2 precursorOxygen sourceTd (°C)Annealing conditionsTop/bottom electrodesPrEcEnduranceRef.
HfO2 None 6.9 None TDMA-Hf Ozone 260 650 TiN/TiN 13.5 N/A 78  
HfO2 None None TEMA-Hf Ozone 280 650/30s/N2 TiN/TiN 10.4 108 using 3.2 MV/cm 44  
Al:HfO2 4.8 mol. % 16 TMA TEMA-Hf Ozone N/A 800,1000/20s/N2 TiN/TiN N/A 37  
Al:HfO2 N/A 10 TMA HfCl4 O2 plasma 300 800/20s/N2 TiN/TiN 15 N/A N/A 50  
Al:HfO2 3 at. % 8.5 TMA TEMA-Hf Ozone 250 700/10s/N2 TiN/TiN 5.1 N/A 77  
Al:HfO2 2.2 cat. % 10 TMA TEMA-Hf Ozone 280 650/20s TiN/TiN 16.5 N/A 95  
Gd:HfO2 3.4 cat. % 100 Gd(iPrCp)3 HfCl4 H2300 650/20s TaN/TaN 32 1.5 106 using 4 MV/cm 30  
Gd:HfO2 2 mol. % 10 Gd(iPrCp)3 HfCl4 H2300 1000/1s TiN/TiN 12 1.75 N/A 36  
HZO/Al2O3 50 at. % 12 TDMA-Zr TDMA-Hf N/A N/A N/A Ti/Au /p-type Si 21 7.5 N/A 13  
La:HfO2 10 cat. % 12 La-(iPrCp)3 TEMA-Hf H2O/O2 280 800C/20s/ N2 TiN/TiN 28 105 using 4.2 MV/cm 35  
La:HfO2 6.0 cat. % 10 La-(iPrCp)3 TEMA-Hf Ozone 280 650/20 TiN/TiN 23.6 N/A 95  
Si:HfO2 4 mol. % 10 4DMA-S TEMA-Hf Ozone N/A 1000C/20s/N2 TiN/TiN 10 N/A 2  
Si:HfO2 3.8 mol. % 10 TDMA-Si TEMA-HF N/A 350 650C/N2 TiN/TiN 15 106 using 3 MV/cm 6  
Si:HfO2 2-8.5 mol. % 10 TDMAS TDMA-Hf O2 200 700–900/5–60s/ N2 TiN/TiN 16 N/A 32  
Si:HfO2 4 mol. % 10 TDMA-S TDMA-Hf O2 plasma 200 900/40s TiN/TiN 15 1.5 N/A 32  
Si:HfO2 2.7 cat. % 10 3DMAS TEMA-Hf Ozone 280 650/20s TiN/TiN 18.8 N/A 95  
Si:HfO2 in 3D trenches 13:1 aspect ratio N/A 10 N/A N/A N/A N/A N/A TiN/TiN 14 1010 using 2 MV/cm 9  
Sr:HfO2 7.9 at. % 10 Sr(tBu3Cp)2 HfCl4 H2300 800/20s/N2 TiN/TiN 23 106 using
3 MV/cm 
34  
Sr:HfO2 N/A 10 N/A N/A N/A 300 800/20s/ N2 TiN/TiN 17.5 N/A N/A 47  
Y:HfO2 5.2 mol. % 10 Y(MeCp)3 TEMA-Hf Ozone N/A 600/20s/N2 TiN/TiN 24 1.2 N/A 33  
Zr:HfO2 50 at. % 20 TDMA-Zr TDMA-Hf Ozone 250 400C/60s/N2 TiN/TiN 25.45 N/A 5  
Zr:HfO2 50 at. % 2.5 TEMA-Zr TEMA-Hf H2240 400 TiN/Si N/A 16  
Zr:HfO2 50 at. % 10 N/A N/A N/A 500 500 TiN/TiN 17 NA 14  
Zr:HfO2 50 at. % 10 TEMA-Zr TEMA-Hf Ozone 280 600 TiN/TiN 15 1.2 NA 45  
Zr:HfO2 50–70 at. % TEMA-Zr TEMA-Hf Ozone 280 800/20s/N2 TiN/TiN 15 1.5 N/A 49  
Zr:HfO2 50 at. % 17 TEMA-Zr Hf(N(CH3)2) (C8H17N2N/A 350 700/10s/O2 [Pt]/[Pt/TiO2/SiO29.3 N/A 110  
Zr:HfO2 50 at. % N/A TEMA-Zr TEMA-Hf Ozone, H2∼270–300 600/20s/N2 TiN/TiN 13 N/A N/A 60  
Zr:HfO2 50 at. % N/A CpZr[N(CH3)2]3 TEMA-Hf Ozone, H2∼270–300 600/20s/N2 TiN/TiN 17.5 N/A N/A 60  
Zr:HfO2 50 at. % 10 TEMA-Zr TEMA-Hf ozone 260 450/N2 TiN/TiN 15 109 using
2.5 MV/cm 
67  
Zr:HfO2 50 at. % 10 TDMA-Zr TDMA-Hf ozone 260 450/N2 TiN/TiN 20 109 using 2.5 MV/cm 67  
Zr:HfO2 50 at. % 10 TDMA-Zr TDMA-Hf O3 250 400/60s/N2 TiN/TiN 24 1.1 N/A 65  
Zr:HfO2 50 at. % 10 TDMA-Zr TDMA-Hf H2250 400/60s/N2 TiN/TiN 19 1.2 N/A 67  
Zr:HfO2 50 at. % 10 CpZr[N(CH3)2]3 Cp-Hf[N(CH3)2]3 Ozone 300 450/20s/N2 TiN/TiN 17.5 105 using
4 MV/cm 
73  
Zr:HfO2 50 at. % 10 CpZr[N(CH3)2]3 Cp-Hf[N(CH3)2]3 O2 plasma 300 450/20s/N2 TiN/TiN 17.5 1.5 104 at
4 MV/cm 
73  
Zr:HfO2 50 at. % 10 CpZr[N(CH3)2]3 TEMA-Hf Ozone 280 600/20s/N2 TiN/TiN 19 N/A 79  
Zr:HfO2 50 at. % 8.4 TEMA-Zr TEMA-Hf Ozone 280 500/30s/N2 TiN/TiN 12 1.2 N/A 96  
Zr:HfO2 50 at. % 8.4 TEMA-Zr TEMA-Hf Ozone 280 500/30s/N2 Pt/TiN 1.5 N/A 80  
Zr:HfO2 50 at. % 10 CpZr[N(CH3)2TEMA-Hf Ozone 230–300 600/20s/N2 TiN/TiN 22 N/A N/A 80  
Zr:HfO2
(6 nm HfO2 /
6 nm ZrO2
50 at. % 12 TDMA-Zr TDMA-Hf H2150 600/20/N2 [Au/Ni]/[Pt/Ti/SiO2/Si] 12 N/A 89  
Zr:HfO2 nanolaminate (1 nm HfO2/1 nm ZrO2) × 4 50 at. % ZyALD TDMA-Hf Ozone 285 500/10 min/N2 TiN/TiN 25 N/A 87  
Zr:HfO2 nanolaminate (3 nm HfO2/
3 nm ZrO2
50 at. % ZyALD TDMA-Hf Ozone 260 500/10 min/N2 TiN/TiN 12 1.1 N/A 88  
Zr:HfO2 nanolaminate H15Z50H15 60 at. % TEMA-Zr TEMA-Hf Ozone 280 N/A TiN/TiN 12 1010 at 2.5 MV/cm 86  
Zr:HfO2 with 1.6 nm Al2O3 50 at. % 21.6 TEMA-Zr, TMA TEMA-Hf H2O 250 (HfO2), 150 (Al2O3450/30s/N2 TiN/TiN 30.19 105 at 2.7 MV/cm 145  
Zr:HfO2 with 1 nm Al2O3 middle layer 50 at. % 40 TEMA-Zr, TMA TEMA-Hf Ozone 280 500/30s/N2 TiN/TiN 11.4 N/A 92  
Zr:HfO2 with 2 nm Al2O3 50 at. % 20 TDMA-Zr, TMA TDMA-Hf Ozone 280 (HfO2), 250 (Al2O3550/30s/N2 TiN/TiN 16 108 at 2.5 MV/cm 93  
Zr:HfO2 with HfO2 seed layer 50 at. % 10 TEMA-Zr TEMA-Hf Ozone 300 600 TiN/TiN 22.1 1.11 N/A 90  
a

All films reported as polycrystalline.

1. Precursor choice

In this section, we discuss the merits and drawbacks of different types of Hf and Zr precursors used in the deposition of ferroelectric HfO2 and HZO films. HfO2 can be deposited from a wide array of precursors such as halide-based compounds (e.g., HfCl4) or alkylamides {e.g., Hf[N(CH3)2]4 and Hf[N(CH3)(C2H5)]4} the latter commonly referred to as TDMA-Hf and TEMA-Hf, respectively.58–60 Commercial ALD processes utilizing Hf-precursors have typically employed HfCl4; however, there has been intensive research in processes utilizing alternative precursors due to the difficulty in handling powdery HfCl4 precursors which require a high vaporization temperature (200 °C) and suffer from additional concerns related to the corrosive reaction by-product (HCl).61 Another issue plaguing HfCl4 precursors is their poor adsorption probability, which limits their applicability to three-dimensional structures where conformal coating into high-aspect features would be necessary.62 A review of the major precursors used in processing studies of HfO2 (Table I) indicates that TDMA-Hf and TEMA-Hf are the most prevalent in ferroelectric HfO2 fabrication methods. Recently, heteroleptic alkylamide/cyclopentadienyl precursors, e.g., CpHf[N(CH3)2]3 have been utilized to deposit ferroelectric HfO2 at higher deposition temperatures.63 

The ALD window is heavily dependent on the selected precursor. As described above, the ALD window is the optimal temperature range for a reaction to occur between the precursor and the surface and typically falls in the range of 150–400 °C for Hf and Zr precursors.60,64,65 The growth windows reported for the most ubiquitous precursors in the deposition of ferroelectric HfO2 are 250–350 and 200–300 °C for TEMA-Hf and TDMA-Hf, respectively. If the process occurs at temperatures below the ALD window, the precursor may not completely react or may condense in the ALD chamber.61 The resultant films are likely to have high levels of impurity due to unreacted compounds and will suffer from unfavorable electrical properties such as high leakage current. At reaction temperatures above the ALD window, the precursors may undergo decomposition or desorption, which may result in a higher growth per cycle (GPC) due to increased ligand reactions that can result in unwanted impurities in the film.66 

In many cases, it is desirable to deposit an amorphous film for ferroelectric HfO2 applications. This is because partially crystallized films may lower the energy barrier for transforming nuclei into the unfavorable monoclinic phase.49 On the other hand, as-deposited amorphous films enable greater control over phase transformations during postdeposition heat treatment. Therefore, the chamber temperature during the ALD reaction is an important consideration in fabricating functional HfO2 films. It may be desirable to deposit at low temperatures to minimize in situ crystallization during deposition, while ensuring that temperatures are high enough to volatilize impurities and unreacted compounds.

Several comparative precursor studies have revealed that the composition of the precursor ligands strongly affects the residual carbon concentration, grain size, and resultant electrical properties.60,67 For example, Materano et al. conducted a comparative study of HfO2 and ZrO2 precursors with the goal of depositing amorphous HZO films to control the subsequent crystallization and phase evolution during thermal treatment.60 While the remanent polarization in HfO2 films fabricated with various precursor combinations was found to be within the same range of 15–20 μC/cm2, the density and dielectric permittivity were found to differ slightly depending on the precursor used. CpZr[N(CH3)2]3-based films were found to have a higher density value of 8.5–12.8 vs 6.5–9.3 g/cm3 for Hf[N(CH3)(C2H5)]4-based films. However, films fabricated with the latter precursor achieved a maximum permittivity of 32.5 compared to 30 in films produced from the cyclopentadienyl precursor.

Materano et al. suggested that ALD deposition temperatures should be limited to 300 °C to deposit a near-amorphous film and avoid the growth of nuclei.22 Materano et al. also demonstrated optimized film growth with TEMA-Hf at low deposition temperature application (250–270 °C), whereas cyclopentadienyls or halide-based precursors required higher deposition temperatures (300 °C) for film growth. While the latter group of precursors showed the lowest impurity levels and highest density films, TEMA-Hf-based films were found to exhibit the highest k-values. Although the cyclopentadienyl (CpZr[N(CH3)2]3) precursor exhibits an extended ALD window up to 500 °C, other studies have shown that the bulkiness of their ligand structure results in lower GPC (<0.05 nm per cycle) compared to other precursors.68,69 A comparison of different types of impurities, i.e., hydrogen and carbon and the GPC for Hf-precursors is presented in Fig. 5. While the hydrogen and carbon impurities are relatively low for cyclopentadienyl precursors, the GPC shown in [Fig. 5(c)] is also relatively diminished. Thus, a consideration in selecting an adequate precursor is the priority associated with maximizing growth rate while minimizing impurity concentration.

FIG. 5.

Impurity concentration as a function of temperature for (a) carbon, (b) hydrogen, (c) chlorine, and nitrogen in atomic percent in various Hf precursors and (d) growth per cycle for common Hf-based precursors as a function of deposition temperature. Reprinted with permission from M. Materano, C. Richter, T. Mikolajick, and U. Schroeder, J. Vac. Sci. Technol. A 38, 022402 (2020). Copyright 2020, American Vacuum Society.

FIG. 5.

Impurity concentration as a function of temperature for (a) carbon, (b) hydrogen, (c) chlorine, and nitrogen in atomic percent in various Hf precursors and (d) growth per cycle for common Hf-based precursors as a function of deposition temperature. Reprinted with permission from M. Materano, C. Richter, T. Mikolajick, and U. Schroeder, J. Vac. Sci. Technol. A 38, 022402 (2020). Copyright 2020, American Vacuum Society.

Close modal

Kim et al. also reviewed the ferroelectric performance of HZO films fabricated with the two most common precursors, TEMA(Hf,Zr) and TDMA(Hf,Zr).67 Similar to Materano et al., Kim et al. found that TDMA-based precursors resulted in lower C impurity concentration and slightly larger grain sizes which favored the orthorhombic phase and mitigated the wake-up effect.67 The endurance and leakage properties of both TEMA- and TDMA-based films are shown in Fig. 6. TEMA-based films were more prone to thermal decomposition and resulted in higher C concentration (3.9 vs 2.3 at. %).

FIG. 6.

Comparative endurance properties of HZO films fabricated with precursors (a) tetrakis(ethylmethylamino) (TEMA) and (b) tetrakis(dimethylamino) (TDMA) and their (c) current density. Reprinted with permission from B. S. Kim, S. D. Hyun, T. Moon, K. Do Kim, Y. H. Lee, H. W. Park, Y. B. Lee, J. Roh, B. Y. Kim, H. H. Kim, M. H. Park, and C. S. Hwang, Nanoscale Res. Lett. 15, 72 (2020). Copyright 2020 under a Creative Commons License; and republished with permission from M. H. Park, H. J. Kim, Y. J. Kim, T. Moon, K. D. Kim, Y. H. Lee, S. D. Hyun, C. S. Hwang, J. Mater. Chem. C 3 6291 (2015). Copyright 2015, Royal Society of Chemistry; permission conveyed through Copyright Clearance Center, Inc.

FIG. 6.

Comparative endurance properties of HZO films fabricated with precursors (a) tetrakis(ethylmethylamino) (TEMA) and (b) tetrakis(dimethylamino) (TDMA) and their (c) current density. Reprinted with permission from B. S. Kim, S. D. Hyun, T. Moon, K. Do Kim, Y. H. Lee, H. W. Park, Y. B. Lee, J. Roh, B. Y. Kim, H. H. Kim, M. H. Park, and C. S. Hwang, Nanoscale Res. Lett. 15, 72 (2020). Copyright 2020 under a Creative Commons License; and republished with permission from M. H. Park, H. J. Kim, Y. J. Kim, T. Moon, K. D. Kim, Y. H. Lee, S. D. Hyun, C. S. Hwang, J. Mater. Chem. C 3 6291 (2015). Copyright 2015, Royal Society of Chemistry; permission conveyed through Copyright Clearance Center, Inc.

Close modal

As illustrated by the two aforementioned comparative studies, each precursor has a specific ALD window in which the optimum deposition temperature lies. While deposition temperatures of 300 °C have been shown to demonstrate robust ferroelectricity in these studies, low temperature deposition may also be used to suppress the lateral grain growth during thermal annealing which may stabilize the orthorhombic phase. Lower growth temperature (200 °C) can induce the incorporation of residual carbon into the HfO2 film and may help to generate ferroelectricity in nominally undoped films. Kim et al. reported the growth of 9-nm thick HfO2 films deposited at 220 °C using TEMA, which exhibited remanent polarization values of 10.4 μC/cm2 and could endure up to 108 switching cycles.44 However, it is imperative to balance the benefits of a low deposition process with a tolerance for defects; as demonstrated in Fig. 6, using a TEMA precursor at a low deposition temperature resulted in a higher wake-up effect and higher leakage current density.

In summary, a variety of precursor choices are commercially available and can be readily applied to the fabrication of ferroelectric HfO2. The two most ubiquitous precursor choices are TDMA-Hf and TEMA-Hf, while new cyclopentadienyls are also gaining attention. The choice of the precursor is closely linked to the choice of deposition temperature, which can be tuned to alter the impurity concentration of resultant films, as well as the as-deposited amorphous state. It is important to consider the ALD window of specific precursors to ensure stable growth per cycle and minimize unreactive or decomposed ligands which lead to inferior quality films.

2. Oxygen source and dose time

The effect of oxygen sources on a variety of different HfxZr1−xO2 compositions has been studied. Typical oxygen source reactants employed in the ALD of HfO2 and HZO include oxygen sources such as deionized water (H2O), ozone (O3), oxygen gas (O2) (both gaseous and plasma), and hydrogen peroxide (H2O2). While H2O is commonly available in most ALD systems, O2 plasma and ozone may be preferred due to their enhanced reactivity. However, during the coating of high-aspect ratio 3D structures, wall collisions can lead to species loss (i.e., oxygen recombination and ozone decomposition), making it difficult to achieve saturation and uniform coverage.70 Here, we review several studies that have compared the properties of HfO2 and HZO thin films using various oxygen sources.65,71–77

In 2003, Park et al. evaluated the properties of dielectric HfO2 films grown using either H2O or O3 as the oxygen source.71 As expected, the stronger oxidation power of O3 increased the oxygen concentration in HfO2 films and resulted in lower leakage current. In fact, many have adopted O3-based processes, especially for ALD at higher deposition temperatures. At higher temperatures, precursor decomposition can lead to carbon contamination. O3 can remove the unwanted carbon via oxidation to form volatile CO and CO2.64,72

Kim et al. compared TDMA-based films deposited with either O3 or H2O and found using x-ray diffraction that both reactants enabled the formation of the orthorhombic phase.65 O3-based HZO films demonstrated a Pr value of 23.5 μC/cm2 compared to 38.7 μC/cm2 for films produced from H2O. It was suggested that the diminished electrical properties of the H2O-based films were due to the incorporation of atomic hydrogen. Interestingly, cycling endurance of both films was reported to be ∼107 cycles at a constant cycling field. Kim et al. claimed that the similarity of the endurance behavior suggests that H2O does not generate more oxygen vacancies which would lead to earlier breakdown. If atomic hydrogen originating from H2O acted as a reducing agent, then hydrogen might be expected to generate oxygen vacancies that help to stabilize the ferroelectric orthorhombic phase. Since the ferroelectric properties were not improved in films processed with H2O, Kim et al. surmised that the hydrogen derived from H2O most likely acts as an impurity due to its negative effects on remanent polarization and lack of effect on endurance limit.

As discussed previously, O2 or O3 may be favored in some cases due to its stronger oxidizing power compared to H2O. The differences between HfO2-based films fabricated via O2 versus O3 oxygen sources on properties of HfO2-based films were also examined by Alcala et al. in HZO films fabricated from cyclopentadienyl-based precursors.73 Alcala et al. similarly found that O3 led to reduced impurity concentration, which expanded the HfxZr1−xO2 composition range in which ferroelectricity could still be achieved. For example, Hf0.25Zr0.75O2 films fabricated via O3 demonstrated pristine (non-pinched) ferroelectric hysteresis, while Hf0.25Zr0.75O2 films fabricated via O2 plasma showed a pinched hysteresis. Figure 7 shows how O2 plasma films demonstrated better GPC stability across a broader range of deposition temperatures (250–350 °C), whereas comparable Pr values for O3 films were restricted to a narrower range of deposition temperatures (270–300 °C).

FIG. 7.

Remanent polarization as a function of ALD deposition temperature for HZO films fabricated with TEMA-Hf and CpZr[N(CH3)2]3 precursors and two different oxygen source, O2 plasma and O3. Reprinted with permission from R. Alcala, C. Richter, M. Materano, P. D. Lomenzo, C. Zhou, J. L. Jones, T. Mikolajick, and U. Schroeder, J. Phys. D: Appl. Phys. 54, 035102 (2021). Copyright 2021, IOP Science.

FIG. 7.

Remanent polarization as a function of ALD deposition temperature for HZO films fabricated with TEMA-Hf and CpZr[N(CH3)2]3 precursors and two different oxygen source, O2 plasma and O3. Reprinted with permission from R. Alcala, C. Richter, M. Materano, P. D. Lomenzo, C. Zhou, J. L. Jones, T. Mikolajick, and U. Schroeder, J. Phys. D: Appl. Phys. 54, 035102 (2021). Copyright 2021, IOP Science.

Close modal

The aforementioned studies demonstrate that the precursor’s ALD window often constrains the targeted deposition temperature range which, in turn, influences the choice of an appropriate oxygen source. Therefore, all three parameters, e.g., precursor, oxygen source, and deposition temperature should be considered simultaneously. It is also interesting to note that the effect of oxygen source on electrical properties of HZO films may differ with differently paired precursors for HZO and HfO2 films. While Kim et al. used TDMA-Hf at 250 °C to study the effects of O3 or H2O reactants, Lee et al. used TEMA-Hf at various deposition temperatures (160–320 °C) to investigate the effect of O3 and H2O on properties of HfO2 gate dielectrics.74 Lee et al. demonstrated that an ALD process could be achieved at a temperature as low as 160 °C using H2O combined with the TEMA-Hf precursor, whereas a temperature of at least 240 °C was required to achieve ALD with O3. They ascribed this effect to the proton-induced ligand exchange reaction that occurs in the case of ALD with H2O, whereas the O3 process required a higher temperature to overcome the activation barrier to dissociate O3 and promote the O-radical-mediated ALD reaction.

Alternative oxygen sources such as H2O2, while not as common, have also been investigated as a potential oxygen source for ALD. Films with lower leakage properties can be produced with H2O2 due to the higher oxidizing power of H2O2 compared to that of H2O, due to the oxidation of impurities as noted above for O3.75 In addition, H2O2-based processes were found to be compatible with trench structures with an aspect ratio of ∼15, indicating high conformality in comparison with O2 plasma. H2O2 may be a promising choice that balances the need for strong oxidizing power with the need for high-aspect ratio conformality.

While we have seen that O3 also demonstrates one of the strongest oxidizing powers, a commonly reported issue is that high concentrations of O3 in ALD processes can result in excess oxygen and promote the growth of interfacial oxide layers such as SiO2 or TiO2.69,76 One way to mitigate this issue is to compromise between oxidizing power and oxidant pulse time (i.e., extent of exposure during the ALD cycle). The polarization switching kinetics of Al-doped HfO2 fabricated with TEMA precursors was studied by varying the ozone dose times where it was found that an increased dose time from 3 to 10 s lowered the remanent polarization from 10.2 to 5 μC/cm2. It was suggested that a dose time of 3 s generated the largest remanent polarization (Pr = 5.1 μC/cm2) due to the higher concentration of oxygen vacancies in the film.77 Similar trends have been reported in undoped HfO2 which demonstrate that lower dose times significantly improve remanent polarization without the introduction of carbon or other dopants.78 

Schroeder et al. investigated the impact of O3 dose time on HfxZr1−xO2 thin films of varying composition fabricated with TEMA-Hf and CpZr[N(CH3)2]3 precursors.79 Measuring the remanent polarization as a function of O3 dose time, they found that the magnitude of the ferroelectric polarization is strongly affected by dose time for every composition (Fig. 8). A critical dose time was required to achieve maximum polarization for each composition. Since the remanent polarization values may vary ±10 μC/cm2 for dose times between 0.1 and 10 s, dose time should be optimized for a selected composition.

FIG. 8.

Effect of dose time on remanent polarization for different stoichiometries of HfO2-ZrO2 films. Reprinted with permission from U. Schroeder, M. Materano, T. Mittmann, P. D. Lomenzo, T. Mikolajick, and A. Toriumi, Jpn. J. Appl. Phys. 58, SL0801 (2019). Copyright 2019, The Japan Society of Applied Physics.

FIG. 8.

Effect of dose time on remanent polarization for different stoichiometries of HfO2-ZrO2 films. Reprinted with permission from U. Schroeder, M. Materano, T. Mittmann, P. D. Lomenzo, T. Mikolajick, and A. Toriumi, Jpn. J. Appl. Phys. 58, SL0801 (2019). Copyright 2019, The Japan Society of Applied Physics.

Close modal

The effect of dose time on the endurance and wake-up properties of HZO films deposited from TEMA-Hf and CpZr[N(CH3)2]3 precursors was also investigated by Materano et al.80 It has been previously reported that the wake-up phenomenon, that is, the change in remanent polarization as a function of electric field cycling, may be caused by the redistribution of oxygen and charged defects which elicit a phase transformation to the orthorhombic phase. In Materano et al., low oxygen content during processing yielded a higher wake-up effect most likely due to the increased oxygen vacancies within low-oxygen content films. These defect sites can act as domain pinning centers, which could be relaxed upon electric field cycling and lead to the noticed increase in remanent polarization as a function of cycling. Materano et al. also reported a lower cycling endurance, i.e., number of cycles until breakdown, for films fabricated with higher oxygen dose time. The reduced endurance in samples fabricated with a high dose time can be explained by a higher fraction of the non-ferroelectric monoclinic phase.80 

Mittmann et al. studied the effects of ozone dose time on the leakage and retention properties in HZO films.81 For dose times greater than 10 s, the leakage current increases several orders of magnitude, likely due to oxygen interstitial defects creating compressive stress. Shorter ozone doses, on the other hand, lead to incomplete reactions of the metalorganic precursor and result in higher leakage current. Thus, Mittmann et al. suggested an optimal dose time of 1–5 s to obtain an oxygen vacancy concentration of ∼0.6%. Retention was also measured as a function of ozone dose time at 85 and 125 °C. Retention improved with increasing dose time up to 1 s which was correlated with an increase in Pr, and imprint was found to be the lowest for films fabricated with the 1 s dose time. Thus, Materano et al.80 and Mittmann et al.81 show that oxygen dose time can influence not only remanent polarization, but also other electrical properties of interest such as cycling endurance, leakage, and retention.

Another factor that can alter the oxygen content in ALD-based films is the type of electrode material employed in the MFM device. While most studies utilize TiN electrodes, other groups have investigated the potential use of IrO2 or TaN, which have different oxygen-scavenging behavior.82,83 Using HZO films from TEMA-Hf and CpZr[N(CH3)2]3 precursors, Mittmann et al. found that the use of oxygen-rich electrodes, i.e., IrO2, led to a reduction in oxygen vacancy concentration within films fabricated from where the excess oxygen stabilizes the deleterious monoclinic phase and degrades the ferroelectric properties.82 Therefore, oxygen-deficient or oxygen-scavenging electrodes are better suited for improving the properties of ferroelectric HfO2. The effect of TaN electrodes was investigated in detail by Fields et al. using TDMA-Hf and TDMA-Zr precursors, where the N concentration of the electrodes was varied and structural and electrical properties of HZO films were characterized.83 Leakage current measurements correlated the change in TaN stoichiometry with oxygen point defect concentrations where electrodes with lower N concentration had higher leakage current density due to increased oxygen defects. The importance of oxygen vacancies in promoting ferroelectricity was supported by the higher remanent polarization values of N-deficient electrodes in which Ta acts as a strong oxygen scavenger and leads to the formation of an increased amount of oxygen vacancies in HZO films with Ta-rich electrodes. Such electrode studies conducted by Mittmann et al. and Fields et al. point to the importance of considering the effects of electrode oxygen-scavenging potential film ferroelectric properties.

We can understand the effect of oxygen supply on ferroelectric properties by considering the effect of oxygen vacancies and interstitials on the thermodynamic energy barrier for various HfO2 phases. Materano et al. used density functional theory calculations to assess the effect of oxygen defects on the crystallization pathways for various phases in HfO2.80 Materano et al. calculated the mean energy paths as a function of film stoichiometry using nucleation theory. Considering the phase dependence on formation energy h, a difference Δh exists between crystallographic phases as a function of defect concentration. The phase transition rate is affected because R ∼ exp(−G∗/kT) is dependent on the thermodynamic energy barrier G∗ = αγ3/|Δh|2 and is dependent on defect concentration, where G* is the minimum energy needed to form a stable nucleus of the lower energy phase, k is Boltzmann’s constant, α is the nuclei geometry factor, and γ is the interfacial energy. Oxygen vacancies lower the phase energy Δh for all phases relative to the monoclinic phase. While the orthorhombic phase is favored with the formation of oxygen vacancies, the tetragonal phase is more favored with increasing oxygen vacancy concentration. From a kinetics point of view, oxygen vacancies do not strongly affect the barrier height of the phase transformation states; however, oxygen interstitials are suggested to lower or completely remove the barrier for the monoclinic phase which would effectively suppress ferroelectricity.

Materano et al. also provided insight into the convoluted effects of doping and oxygen defects.22 The model developed by Materano et al. suggests that oxygen vacancies slow down the rate of transformation from the tetragonal to the monoclinic phase during grain growth according to Ostwald’s model. Without a dopant, it was suggested that oxygen vacancies would redistribute within the HfO2 layer and lead to the transformation to the monoclinic phase. Therefore, the role of the dopant is to interact with oxygen vacancies and effectively stabilize them, which results in the stabilization of the orthorhombic phase.

In summary, the choice of precursor, oxygen source, and reactant dose time play a crucial role in tuning the ferroelectric properties and performance of HfO2-based films and must be considered concurrently. The deposition temperature is also an important parameter where it may be necessary to compromise between low impurity concentration (higher deposition temperature) and enhanced control of phase evolution through amorphous as-deposited states (lower deposition temperature). Strong oxidants produce generally denser films with favorable leakage current properties, but have issues such as decreased conformality, growth of interfacial oxides, and higher reactivity. While weaker oxidants such as H2O result in films with increased carbon impurities, they are generally more conformal and have moderate growth rates. Alternative oxygen sources such as H2O2 may be considered as a compromise between H2O and O3, and the oxygen dose time can be varied to improve ferroelectric remanent polarization in HfO2 films. Differences in optimal oxygen dose time may vary with differently paired precursors, oxygen source, and dopants; generally, however, a low to moderate oxygen dose time is required to induce ferroelectricity in HfO2-based films and suppress the monoclinic phase. The resultant oxygen distribution within an ALD film may also be influenced by the electrode material, where ferroelectric properties are enhanced with oxygen-scavenging electrodes.

3. Layering and mixing

Up to this point, we have discussed HfO2 and HfxZr1−xO2 films as homogeneous mixtures in which pulses of Hf and Zr precursors are alternated each ALD cycle. In recent years, new superstructures have been realized via intentional layering. Studies have investigated HfO2-ZrO2 nanolaminates,84–87 bilayers,88,89 seed layers,90–93 and dopant layering.94,95 In this section, we make the distinction between solid solution, homogeneous HfxZr1−xO2 films and HfO2-ZrO2, with the latter being deposited intentionally thicker layers to form layered structures.

Weeks et al. demonstrated one of the first ferroelectric HfO2-ZrO2 nanolaminates in 2017 and found that ALD deposition temperature strongly influences the orthorhombic phase stability during subsequent annealing.87 At lower ALD deposition temperatures (265 °C), antiferroelectric-like behavior characterized by a pinched hysteresis was observed, whereas at higher deposition temperatures (280 °C) the films exhibited a strong ferroelectric response before cycling. Weeks et al. ascribed the noticeable difference in wake-up and remanent polarization to the changes in the lattice distortion during heating. Lattice parameter distortion may occur when crystallites coalesce in as-deposited films, and the effect may be increased at higher deposition temperatures due to the larger crystallite growth. This lattice distortion appears to favor the formation of the ferroelectric phase.96 In Weeks et al., the highest remanent polarization was reported for the (1 nm HfO2/1 nm ZrO2) × 4 nanolaminate at a Pr (25 μC/cm2), which was higher than that of the homogeneous HZO solid-solution thin film (19.5 μC/cm2).

A similar investigation was conducted by Park et al. in 2019 of HfO2-ZrO2 nanolaminates and superlattice structures, in which it was found that parameters such as the starting material, thickness ratio between HfO2 and ZrO2, and the single-layer thickness influence ferroelectric properties.86Figure 9 shows P-E loops for several bilayers and trilayers from the work of Park et al. When a nanolaminate was deposited with a ZrO2 layer thicker than 1.1 nm at the bottom, the monoclinic phase fraction was found to be higher than those of the solid solution or HfO2-starting films. The driving factors for the monoclinic phase formation were suggested to be the increased as-deposited crystallization of ZrO2 and the epitaxial tensile strain between the ZrO2-HfO2 layers as was previously shown by Weeks et al. Despite the 50:50 composition for Hf:Zr, which thermodynamically favors the orthorhombic phase in thin films, the kinetics of the monoclinic phase formation could play a larger role in the resulting phase composition after annealing. The effect in the ZrO2-starting films may arise due to the larger lattice parameters of ZrO2, compared to those of HfO2, which may induce a state of in-plane tensile strain on the top HfO2 layer. Resultant strain could significantly reduce the energy barrier for the tetragonal to monoclinic phase transformation. Park et al. showed that single-layer thickness had a measurable effect on the crystalline phases present only in the case for ZrO2-starting samples, where the monoclinic peak intensity grew for larger layer thicknesses. HfO2-ZrO2-HfO2 stacks had the most favorable properties including endurance of 1010 cycles with a Pr of 12 μC/cm2.

FIG. 9.

Polarization-electric field loops for nanolaminate samples with (a) 40:40 and (b) 20:60 HfO2:ZrO2 ALD ratios for bilayers and trilayers, (c) P-E loops for HfO2:ZrO2:HfO2 trilayers with different ALD cycle ratios, and (d) 2Pr as a function of ZrO2 fraction for trilayers, bilayers, and solid solution (SS), (e) cross-sectional TEM image, (f) high angular dark field image, (g) EDS image, [(h) and (i)] high-resolution TEM image of namolaminate capacitor at two different regions, and (j) grain size distribution of bilayers and trilayers analyzed via the watershed method. Reprinted with permission from M. H. Park, H. J. Kim, G. Lee, J. Park, Y. H. Lee, Y. J. Kim, T. Moon, K. D. Kim, S. D. Hyun, H. W. Park, H. J. Chang, J.-H. Choi, and C. S. Hwang, Appl. Phys. Rev. 6, 041403 (2019). Copyright 2019, AIP Publishing LLC.

FIG. 9.

Polarization-electric field loops for nanolaminate samples with (a) 40:40 and (b) 20:60 HfO2:ZrO2 ALD ratios for bilayers and trilayers, (c) P-E loops for HfO2:ZrO2:HfO2 trilayers with different ALD cycle ratios, and (d) 2Pr as a function of ZrO2 fraction for trilayers, bilayers, and solid solution (SS), (e) cross-sectional TEM image, (f) high angular dark field image, (g) EDS image, [(h) and (i)] high-resolution TEM image of namolaminate capacitor at two different regions, and (j) grain size distribution of bilayers and trilayers analyzed via the watershed method. Reprinted with permission from M. H. Park, H. J. Kim, G. Lee, J. Park, Y. H. Lee, Y. J. Kim, T. Moon, K. D. Kim, S. D. Hyun, H. W. Park, H. J. Chang, J.-H. Choi, and C. S. Hwang, Appl. Phys. Rev. 6, 041403 (2019). Copyright 2019, AIP Publishing LLC.

Close modal

McBriarty et al. also showed that a larger orthorhombic phase fraction emerged for HfO2-ZrO2 bilayers in which the ZrO2 layer was unconstrained during annealing (occupying the top layer), whereas samples that had a ZrO2-starting layer resulted in a higher tetragonal phase fraction and reduced remanent polarization by more than a half.88 Thus, multiple studies reviewed here suggest that nanolaminates starting with an HfO2 layer may be favored for maximizing the ferroelectric polarization.

While the aforementioned studies considered the effects of a starting layer on nanolaminate properties, others have investigated the role of a starting material, also called “seed layer” deposited below homogenously mixed films, on ferroelectric properties.90,91,93 A seed layer refers to a thin film that is deposited before the functional layer, e.g., ferroelectric, dielectric, and antiferroelectric of interest. Seed layers have been shown to influence the microstructure, crystallographic texture, and electrical properties of thin films in other material systems such as PMN-PT, BMT-PT, and PZT.97–99 In the HfO2 literature, there is still ongoing debate about which seed layers are preferred for improved ferroelectric response. For example, Gaddam et al. found that a maximum remanent polarization could be achieved in HZO films formed by ALD when an HfO2 seed layer was used as a starting surface for deposition90 similar to the Park et al. investigation of nanolaminates.86 However, this effect was only limited to seed layer thicknesses of 10 Å, after which the Pr decreased. Interestingly, a different study reported that ZrO2 could be used as either a top or bottom (seed) layer at the HZO/TiN interfaces to influence structural or electrical properties.90 Surprisingly, in this study, Onaya et al. found that MFM stacks with a bottom ZrO2 layer exhibited a higher remanent polarization value compared to HZO films with no seed layers, albeit the improvement was small (12 vs 15 μ/cm2).91 The reported increased remanent polarization due to a ZrO2-starting layer directly contrasts the observations made by Park et al., which suggested that ZrO2 as a starting material degrades the ferroelectric performance.86 Differences in which starting layer improves ferroelectric performance may be ascribed to the overall structure of the films (nanolaminates versus mixed films) as well as film thicknesses and annealing treatment. However, this discrepancy highlights the need for further research on the effects of starting layers in nanolaminates and mixed films.

Prior reports have not only focused on seed layers but also inserting layers into the HZO ferroelectric layer to control the grain size of the resultant film, also known as the insertion of an interfacial layer. In 2014, Kim et al. were able to control the grain size of a 40 nm film by depositing the HZO film in an HZO/Al2O3/HZO sequence; that is, an HZO solid solution film with a middle 1-nm Al2O3 layer.91 By insertion of Al2O3, it was possible to decrease the grain size and mitigate the evolution of the nonferroelectric monoclinic phase. The interfacial-layered stack resulted in a remanent polarization value (10 μC/cm2) that was 11 times larger than that of a 40 nm Hf0.5Zr0.5O2 film. Other work on Al2O3 has shown that it can be used as an effective capping layer instead of the conventional TiN.93 

In some cases, the opposite of discrete layering is desired. ALD studies have reported improved solid solution mixing with dopant layering,94in situ precursor mixing,92 and atomic partial layer deposition (APLD).85 In 2014, Lomenzo et al. deposited Si-doped HfO2 with various concentrations and distributions, finding that inhomogeneous layering effectively lowered the remanent polarization.94 More recently, Mart et al. showed that in situ precursor mixing could improve ferroelectric switching.95In situ precursor mixing was realized by a modified ALD process in which the TEMA-Hf and dopant precursor are pulsed back-to-back before the oxidant step. As shown in Fig. 10, in situ precursor mixing allowed Mart et al. to reach lower dopant concentrations that were previously inaccessible and improve compositional uniformity. A similar method referred to as APLD was employed by Hernández-Arriaga et al. in which the precursor doses and exposure times were varied to obtain fractional coverage in monolayers of HfO2 and TiO2.85 As a result, HfO2-TiO2 layers showed improved chemical homogeneity. A schematic of APLD is shown in Fig. 11 in which a nonsaturating pulse of Hf is cycled with a saturating pulse of Ti, followed by a complete oxygen pulse. The “partial” nature of this technique comes from the nonsaturating Hf pulse, which allows better and more homogenous mixing with the subsequent Ti pulse as illustrated in Fig. 11(b). While Hernández-Arriaga et al. did not report ferroelectric properties, it is likely that APLD could be readily applied for planar ferroelectric HfO2-based thin films. It is important to note that partial ALD recipes may not be well-suited for high-aspect areas where the limited availability of reaction sites may prohibit conformal growth.

FIG. 10.

Dopant concentration measured via XPS vs dopant ALD pulse fraction for 10 nm thick (a) Al-, (b) Si-, and (c) La-doped HfO2. Linear scaling of the XPS measurement result vs pulse fraction is observed for all three material systems. Reprinted (adapted) with permission from C. Mart, K. Kühnel, T. Kämpfe, M. Czernohorsky, M. Wiatr, S. Kolodinski, and W. Weinreich, ACS Appl. Electron. Mater. 1, 2612 (2019). Copyright 2019, American Chemical Society.

FIG. 10.

Dopant concentration measured via XPS vs dopant ALD pulse fraction for 10 nm thick (a) Al-, (b) Si-, and (c) La-doped HfO2. Linear scaling of the XPS measurement result vs pulse fraction is observed for all three material systems. Reprinted (adapted) with permission from C. Mart, K. Kühnel, T. Kämpfe, M. Czernohorsky, M. Wiatr, S. Kolodinski, and W. Weinreich, ACS Appl. Electron. Mater. 1, 2612 (2019). Copyright 2019, American Chemical Society.

Close modal
FIG. 11.

APLD results in improved film uniformity and may allow smaller dopant concentrations to be reached. (a) Pulse sequence used to grow via APLD with nonsaturating Hafnium pulse, (b) schematic of HfO2-TiO2 nanolaminates, (c) schematic of HfO2-TiO2 grown via APLD. Reprinted with permission from H. Hernández-Arriaga, E. López-Luna, E. Martínez-Guerra, M. M. Turrubiartes, A. G. Rodríguez, and M. A. Vidal, J. Appl. Phys. 121, 064302 (2017). Copyright 2017, AIP Publishing LLC.

FIG. 11.

APLD results in improved film uniformity and may allow smaller dopant concentrations to be reached. (a) Pulse sequence used to grow via APLD with nonsaturating Hafnium pulse, (b) schematic of HfO2-TiO2 nanolaminates, (c) schematic of HfO2-TiO2 grown via APLD. Reprinted with permission from H. Hernández-Arriaga, E. López-Luna, E. Martínez-Guerra, M. M. Turrubiartes, A. G. Rodríguez, and M. A. Vidal, J. Appl. Phys. 121, 064302 (2017). Copyright 2017, AIP Publishing LLC.

Close modal

To summarize, we reviewed in this section several distinct practices of layering and mixing of HfO2-ZrO2 thin films via ALD. By controlling the ALD cycle number, order, and ratio, different nanolaminate, bilayer, and interfacial architectures are possible. Factors governing properties in nanolaminates include individual layer thickness, starting layers, deposition reactants, and temperature conditions. While research in this field is still nascent, we have reviewed several methods presently available for controlling the microstructure and properties of HfO2-ZrO2 films by leveraging processing options unique to ALD.

PVD is a process by which material from a condensed phase (target) is transformed into a vapor phase and then back into a thin film condensed phase onto a selected substrate. The most common PVD techniques are sputter deposition (direct current, or DC, and radio frequency, or RF, types), evaporation, and PLD. Sputter deposition has been extensively used for the fabrication of polycrystalline films while PLD is more commonly employed for epitaxial film growth. Here, we discuss sputter-deposited films while PLD is discussed in Sec. II C. From an industry-perspective, sputter deposition offers the advantage of very high deposition rates and the associated low cost.100 Sputter deposition can also be accomplished at room temperature or <200 °C. Unlike ALD, which has a specified temperature window in which an effluent reaction can be achieved, sputter deposition is typically temperature independent and can, therefore, more readily accommodate temperature-sensitive substrates such as polymers. Another key advantage of sputter deposition stems from its use of ceramic targets which results in lower carbon impurities compared to ALD which requires the use of metalorganic precursors. However, a major downside to sputter deposition also comes from its physical and nonchemical mechanism of deposition, which makes sputter deposition a “line of sight” technique and, therefore, is not ideal for coating three-dimensional structures. While sputtered films generally have a composition close to the target material, greater control over the composition in doping may require the use of dual targets. A schematic mechanism of sputtering is illustrated in Fig. 12.

FIG. 12.

Schematic of a coplanar physical vapor deposition (sputter deposition) sputtering system showing sputtering gas bombarding onto the biased target and ejecting material onto the oppositely biased substrate for film growth.

FIG. 12.

Schematic of a coplanar physical vapor deposition (sputter deposition) sputtering system showing sputtering gas bombarding onto the biased target and ejecting material onto the oppositely biased substrate for film growth.

Close modal

Magnetrons utilize strong electric and magnetic fields to confine charged plasma particles close to the surface of the sputter target. The electrons in the plasma travel in a “racetrack” pattern, which enables them to collide more frequently with argon atoms and produce more ions. Argon ions that are bombarded onto the cathodically connected target eject target particles which ballistically travel from the target to the substrate surface where they are deposited.

In contrast to conventional DC sputtering, RF sputtering is more commonly used for the deposition of insulating materials to mitigate the charge build-up on insulating targets. Modulating an AC bias across the anode-cathode mitigates charge build-up on the target albeit with the expense of additional RF power supplies. In reactive sputtering, a reactive gas such as oxygen or nitrogen is introduced into the chamber and sputtered particles undergo a chemical reaction which forms a molecular compound and is then deposited onto the substrate. The composition can thus be controlled via relative pressures of the inert and reactive gas species. In this section, we review selected parameters of interest in sputtered HfO2 thin films, which include dopant material, oxygen flow, and vacuum level which can be used to fine-tune electrical and structural properties of resultant HfO2-based films. The process control variables for sputter deposition-based films and their remanent polarization and coercive field are summarized in Table II.

TABLE II.

Chemical names and formulas for precursor abbreviations used in Table I.

Chemical nameChemical formulaAbbreviation
Cylcopentadienyl C5H5 Cp 
Tetrakis(dimethylamino) - X [N(CH3)2]4TDMA-X 
Tetrakis(ethylmethylamino) - X X[N(CH3)(C2H5)]4 TEMA-X 
Tris(dimethylamido)-silane (CH3)2N)3SiH TDMA-S 
Tetrakis(dimethylamino)-silane [(CH3)2N]4Si 4DMA-S 
Tris(isopropylcyclopentadienyl) gadolinium(III) Gd(C5H4CH(CH3)2)3 Gd(iPrCp)3 
Tris(methylcyclopentadienyl) yttrium (CH3C5H4)3Y(MeCp)3 
Bis(tri-tert-butylcyclopentadienyl) strontium [(t-C4H9)3C5H2]2Sr Sr(tBu3Cp)2 
Tris(isopropyl-cyclopentadienyl) X (C3H7C5H4)3X((iPrCp)3
Trimethylaluminium Al(CH3)3 TMA 
Tris(dimethylamino)cylcopentadienyl-zirconium Cp[N(CH3)2]3 Zr ZyALD 
Chemical nameChemical formulaAbbreviation
Cylcopentadienyl C5H5 Cp 
Tetrakis(dimethylamino) - X [N(CH3)2]4TDMA-X 
Tetrakis(ethylmethylamino) - X X[N(CH3)(C2H5)]4 TEMA-X 
Tris(dimethylamido)-silane (CH3)2N)3SiH TDMA-S 
Tetrakis(dimethylamino)-silane [(CH3)2N]4Si 4DMA-S 
Tris(isopropylcyclopentadienyl) gadolinium(III) Gd(C5H4CH(CH3)2)3 Gd(iPrCp)3 
Tris(methylcyclopentadienyl) yttrium (CH3C5H4)3Y(MeCp)3 
Bis(tri-tert-butylcyclopentadienyl) strontium [(t-C4H9)3C5H2]2Sr Sr(tBu3Cp)2 
Tris(isopropyl-cyclopentadienyl) X (C3H7C5H4)3X((iPrCp)3
Trimethylaluminium Al(CH3)3 TMA 
Tris(dimethylamino)cylcopentadienyl-zirconium Cp[N(CH3)2]3 Zr ZyALD 

1. Dopant and target selection

Like ALD-grown HfO2 thin films, a variety of dopants can be introduced into sputter deposition-grown HfO2 films to elicit a ferroelectric response. HfO2 films have been fabricated via RF sputtering using a variety of dopants [Zr4+,101–105 Y3+,106–109 Fe3+,110 Si4+,42 Sc2+,42 Y3+,42 Ge4+,42 and N3– (Refs. 38 and 42)]. Interestingly, it has been observed that much lower doping concentrations or even no doping can achieve the orthorhombic phase in sputtered HfO2 films, relative to ALD-fabricated films.111 One reason might be that co-sputtering from dual targets could potentially lead to better intermixing. In ALD, the monolayer deposition process can limit the quality of atomic intermixing, which may lead to a higher doping concentration requirement for stabilizing ferroelectricity. Y-doped HfO2 films are the most heavily studied sputter deposition-based films, and several studies focus on the optimization of Y-doped HfO2-based devices. In a dual-target system of metallic Hf and Y, Olsen et al. explored the effect of Y-concentration on the structure of HfO2 films by varying the sputter power of the Y target.106 Olsen et al. found that as Y concentration increased from 0.2 to 5.0 mol. %, the monoclinic phase became increasingly suppressed. Ferroelectric hysteresis measurements revealed the highest remanent polarization (Pr ∼ 10 μC/cm2) at Y-doping concentrations of 1.9 mol. %.99 Films fabricated via sputtering are typically reported to require a smaller amount of doping (<2 mol. %) in contrast with Y:HfO2 films deposited via pulsed laser deposition where the ideal doping concentration varies between 3.8 (Ref. 112) and 5 mol. %.113 In contrast, for chemical solution-deposited Y:HfO2 films, the concentration of dopant is typically held around 5 mol. %.114,115 While the distinction between 1.9 and 5 mol. % may be within the tolerance limit of typical methods for determining chemical composition, i.e., energy dispersive spectroscopy (EDS), the trend of requiring lower doping concentration in sputter deposition films than in ALD films is consistent across a variety of dopant systems.

One example includes HfxZr1−xO2 films, which were fabricated by Luo et al. where a Zr concentration of ∼15–16 mol. % showed the best ferroelectric properties.102 Alloying concentration was controlled by changing the sputtering power from 90 to 150 W. A concentration of 15.65 mol. % demonstrated both the highest intensity fraction of the orthorhombic phase in grazing incidence x-ray diffraction (GIXRD) and the highest remanent polarization value shown in Fig. 13. In contrast, the optimal alloying concentration for ALD-based HfxZr1−xO2 films is typically reported around 50:50.14 One hypothesis to explain the lower alloying requirement for HfO2 films fabricated via sputter deposition system may involve oxygen vacancies. Hard XPS results from Baumgarten et al. show higher Hf3+ spectral components present in sputter deposition-grown films, relative to films grown via ALD which showed negligible Hf3+ components where Hf3+ spectral components are an indicator of oxygen vacancies present in the film.116 Sputter deposition-grown films also showed a more pronounced wake-up effect, i.e., the increase in remanent polarization as a function of electric field cycling, compared to ALD-grown films which was suggested to be due to higher oxygen vacancy mobility in sputter deposition films. Higher concentration of oxygen vacancies in sputter deposition-grown films could, thus, preferentially stabilize the orthorhombic phase at lower doping concentrations. Luo et al. calculated the Helmholtz free energy difference of HfO2 phases relative to the monoclinic phase as a function of temperature and found that a higher concentration of oxygen vacancies could effectively lower the concentration of ZrO2 required to generate the largest amount of orthorhombic phase.102 The higher physical energy of bombardment in the sputter deposition system could partially damage metal-oxide bonds and be one source for additional oxygen vacancy generation.

FIG. 13.

(a) TEM micrograph of metal-insulator-metal capacitor, (b) XPS ZrO2 content of HfO2-ZrO2 films measured as a function of sputtering power, (c) GIXRD of films fabricated via various sputtering powers, and (d) corresponding Pr and orthorhombic peak intensity. Reprinted with permission from Q. Luo et al., IEEE Electron Device Lett.40, 570–573 (2019). Copyright 2019, IEEE.

FIG. 13.

(a) TEM micrograph of metal-insulator-metal capacitor, (b) XPS ZrO2 content of HfO2-ZrO2 films measured as a function of sputtering power, (c) GIXRD of films fabricated via various sputtering powers, and (d) corresponding Pr and orthorhombic peak intensity. Reprinted with permission from Q. Luo et al., IEEE Electron Device Lett.40, 570–573 (2019). Copyright 2019, IEEE.

Close modal

In a study of Y-doped HfO2 deposited via RF magnetron sputtering at room temperature, the effect on sputtering Ar ion beam energies (300–800 eV) on ferroelectric properties was studied. Remanent polarization in Y:HfO2 was found to be heavily dependent on the ion beam energy used during deposition.117 The assisted Ar ion beams were emitted from an anode-layer ion source and the beam current was maintained at 600 mA with an effective width of 120 mm. GIXRD measurements were conducted on four Y:HfO2 films deposited with different ion beam energies of 0, 300, 500, and 800 eV. No diffraction appeared in films fabricated with an ion beam energy of 0 eV. When the ion beam energy was raised to 300 eV, cubic reflections 111, 200, and 220 emerged which suggested that the 300 eV ion beam energy gave rise to the crystallization of Y:HfO2 into the cubic phase. At 500 eV, the orthorhombic and monoclinic phases appeared; the orthorhombic phase was differentiated from cubic and tetragonal with P-E measurements which indicated that films were ferroelectric. At even higher ion beam energy of 800 eV, the diffraction peaks of the orthorhombic phase weakened, and the monoclinic phase dominated. Liang et al. argued that the Gibbs free energy, surface energy, and strain energy effects alone cannot explain the ion beam energy dependence on ferroelectric properties. Thus, the kinetics of the phase transformation must be considered. In accordance with Ostwald’s step rule, a structural transformation ends with the phase whose free energy is closest to that of the parent phase instead of the most stable phase with the lowest energy. Kim et al. suggest that the bombardment of the Ar ion beam during the sputtering process could transfer the additional kinetic energy to the sputtering particles which could help overcome the energy barrier for the cubic to orthorhombic phase which leads to enhanced remanent polarization.

One advantage of sputter deposition over ALD may be the increased availability of PVD targets over ALD precursors, which allows more HfO2-dopant systems to be explored. For example, Shiraishi et al. investigated FeO1.5 doping compositions from 0 to 0.14 mol. % where a composition of 0.06 mol. % generated the highest remanent polarization value of 8.8 μC/cm2.110 While Fe dopants are not common in the ferroelectric HfO2 literature, the low doping concentration required to elicit ferroelectric properties in the study by Shiraishi et al. further supports the hypothesis of lower required doping concentrations for sputter deposition HfO2 films to achieve ferroelectricity.

Xu et al. considered not only the effect of cation (Si4+, Sc2+, Y3+, Ge4+, Zr4+) dopants but also an anion (N3−) dopant in sputtered HfO2 films.38,118 The doping window for various selections is shown in Fig. 14, which indicates a relatively narrow doping window compared to ALD-grown films. Xu et al. proposed that for the case of N-doping in HfO2 films, two factors should be considered independently: first, the introduction of oxygen vacancies when N3− replaces O2−, and second, the formation of Hf-N and N-O covalent bonds. The latter of these two contributions is thought to act differently than cation doping. A concentration of 0.34% N3− was effective in generating ferroelectric properties (Psw ∼ 20 μC/cm2) and 2.44% N completely suppressed ferroelectricity. N bonding effects, specifically the formation of covalent N-O bonds, hinder oxygen vacancy motion at higher doping concentrations, thereby generating a sensitive doping tunability for N-doped HfO2.118 The effect of electronically passivated oxygen vacancies in the form of an oxygen vacancy and nitrogen-on-oxygen-site defect (VN2) complexes was also discussed in Baumgarten et al. although it was suggested that the role of neutralized oxygen vacancies played a more critical role in ALD films.116 

FIG. 14.

Switchable polarization (Psw) of 27 nm doped HfO2 films as a function of doping concentration for different dopants H, Si, Y, Sc, Ge, and Zr. Reprinted with permission from X. Xu, F.-T. Huang, Y. Qi, S. Singh, K.M. Rabe, D. Obeysekera, J. Yang, M.-W. Chu, and S.-W. Cheong, Nat. Mater. 20, 826 (2021).Copyright 2017, AIP Publishing.

FIG. 14.

Switchable polarization (Psw) of 27 nm doped HfO2 films as a function of doping concentration for different dopants H, Si, Y, Sc, Ge, and Zr. Reprinted with permission from X. Xu, F.-T. Huang, Y. Qi, S. Singh, K.M. Rabe, D. Obeysekera, J. Yang, M.-W. Chu, and S.-W. Cheong, Nat. Mater. 20, 826 (2021).Copyright 2017, AIP Publishing.

Close modal

Oxygen content in sputter deposition-grown films can also be a function of the target quality and age. Over the lifetime of a metallic oxide target, erosion will occur in the typical “racetrack” fashion that causes the trajectory of ions to change. Severin et al. discussed the effects of target aging on structural properties of ZnO films and illustrates how a higher bombardment by oxygen ions may cause structural defects and changes in oxygen concentration at the center of the deposited films due to back-sputtering.119 The effect of deleterious back-sputtering may also explain the lower deposition rates observed with target aging. Not only is target aging a concern for repeatability in film quality, but another issue that may arise is differences in oxygen content found in targets across vendors. While vendors report the purity of the metallic component in their targets, e.g., 99.999% Hf with trace Zr, they do not report the exact stoichiometry of metal oxide targets or their oxygen concentration. Thus, using two different targets may yield slightly different oxygen concentration in sputter-deposited films. In addition, the oxygen content within a target may also vary with age with the depletion of oxygen over time. Given that differences of 0.5–1 at. % in oxygen concentration may elicit differing oxygen vacancy concentrations, and in turn, ferroelectric properties, the differences in oxygen concentration across different vendor sputter deposition targets and over a target lifetime is a major concern for film quality and repeatability in the sputter deposition process. To further understand the importance of oxygen and oxygen vacancies in sputter deposition-grown HfO2 films, Sec. II B 2 focuses on studies that consider the effect of oxygen flow during sputter deposition on the properties of HfO2 ferroelectric films.

2. Oxygen content

HfO2 films are typically made via reactive sputtering using oxygen gas to deposit nominally pure and doped films.42,94–106 This section discusses the role of oxygen concentration and resultant oxygen vacancies or interstitials on ferroelectric properties of HfO2 as they have been the focus of many recent studies.43,105,109 It may be easier to form oxygen vacancies in sputter-deposited films because nonreactive gas, i.e., argon, can be exclusively used during deposition to achieve film deposition. In contrast, ALD requires a strong oxidizing agent such as water or ozone to promote a viable surface reaction and takes place at a higher pressure, leading to higher oxidation during film growth.

Kim et al. studied the effect of Ar/O2 gas ratio on the crystallographic structure of nominal HfxZr1−xO2 films via XRD and observed an increase in the intensity of the orthorhombic peak when the Ar/O2 gas ratio was increased from 2.0/1.0 to 2.0/0.2 SCCM.43 The gas ratio during sputter deposition likely affected the oxygen stoichiometry within the HZO films; thus, a more oxygen-deficient film was correlated with a higher orthorhombic phase portion. The memory window, which is related to the coercive voltage, also showed a dependence on Ar/O2 ratio, increasing from 0.45 to 0.99 V when O2 flow was decreased. A HfxZr1−xO2 film study by Lee et al. confirmed the results observed by Kim et al. and showed that increasing the O2/(Ar + O2) ratio during deposition similarly reduced ferroelectric properties of resultant films. At a gas composition of O2/(Ar + O2) = 0, no diffraction peaks were observed for the monoclinic phase. Instead, a high intensity orthorhombic peak at ∼30° was observed. Increasing the O2/(Ar + O2) resulted in monoclinic phase formation and diminished the intensity of the ∼30° orthorhombic peak, which suggests that oxygen-deficient growth conditions may be favorable for realizing ferroelectric properties in sputter-deposited films. Both findings from the reports of Kim et al. and Lee et al. hint that, in low oxygen growth environments, the root cause of enhanced ferroelectric properties may be oxygen vacancy formation.41,102

Mittmann et al. explored the origins of the ferroelectric phase in undoped HfO2 deposited via sputtering and the influence of oxygen flow on phase evolution.120 In this study, the concentration of oxygen was found to directly correlate with the stabilization of the orthorhombic phase where decreased O2/(O2 + Ar) was found to promote the ferroelectric phase. Mittmann et al. proposed that oxygen-rich environments instead favored the nucleation of nanocrystals of the monoclinic phase and led to the suppression of ferroelectricity. High remanent polarization values of 20 μC/cm2 were achieved for films deposited at a 0:5 SCCM O2/(O2 + Ar) condition. Another interesting outcome of this study was that Mittmann et al. found that the required crystallization temperature was much higher (800–1000 °C) for sputter deposition-grown HfO2 films compared to crystallization temperatures reported for ALD-grown HfO2 films (500–600 °C).111 Mittmann et al. suggested that the higher annealing temperature may be required due to the lower thermal energy sputtering process which may create a higher activation energy for nucleation. Another factor could be the comparatively low amounts of impurities in sputter deposition films which could help facilitate heterogenous nucleation, unlike ALD films which have higher concentration of C impurities. Another factor that may explain the differences in required annealing temperatures to achieve ferroelectricity in sputter deposition versus ALD films could be the higher thermal budget in ALD-based processes which are typically conducted at around 300 °C whereas sputter deposition processes are typically accomplished at room temperature.

In a later study, Mittmann et al. examined the effects of ZrO2 alloying and oxygen flow concurrently, finding that sputter deposition with ZrO2 was more likely to create oxygen interstitials whereas HfO2 deposition favored vacancies.120 The mechanism of oxygen defects stabilization of ferroelectricity in HfO2 is thought to be similar for sputter deposition and ALD films, and the mechanism is discussed more thoroughly in Sec. II A 2. The phase fraction of orthorhombic plus tetragonal and 2Pr values as a function of power ratio and oxygen flow is reproduced in Fig. 15. Oxygen flow was also found to strongly impact the deposition rate and the resultant grain diameter. Films with larger grains were more likely to possess the monoclinic phase; thus, the flow of oxygen was found to not only influence phase evolution via oxygen defects, but also through surface energy effects.

FIG. 15.

(a) Combined orthorhombic and tetragonal phase fractions (o + t)/(o + t + m) in sputter-deposited HZO films under different oxygen flow conditions. (b) Remanent polarization 2Pr of HZO films deposited in varying ambient oxygen and at different HfO2/ZrO2-power ratios normalized to 5 SCCM oxygen flow. Reprinted with permission from T. Mittmann, M. Michailow, P. D. Lomenzo, J. Gärtner, M. Falkowski, A. Kersch, T. Mikolajick, and U. Schroeder, Nanoscale 13, 912 (2021). Copyright 2021, Royal Society of Chemistry.

FIG. 15.

(a) Combined orthorhombic and tetragonal phase fractions (o + t)/(o + t + m) in sputter-deposited HZO films under different oxygen flow conditions. (b) Remanent polarization 2Pr of HZO films deposited in varying ambient oxygen and at different HfO2/ZrO2-power ratios normalized to 5 SCCM oxygen flow. Reprinted with permission from T. Mittmann, M. Michailow, P. D. Lomenzo, J. Gärtner, M. Falkowski, A. Kersch, T. Mikolajick, and U. Schroeder, Nanoscale 13, 912 (2021). Copyright 2021, Royal Society of Chemistry.

Close modal

Oxygen flow during the deposition of HfO2 films is not the only processing step during which oxygen defects can be introduced or suppressed. The supply of oxygen during annealing may also contribute to defect formation. While annealing HfO2 thin films is commonly undertaken in the N2 atmosphere, some have investigated the effects of annealing in O2 gas.105,109 Sputtered films in Lee et al. were annealed in O2 gas with different annealing times where the strongest ferroelectric response (Pr of 16.5 μC/cm2) occurred in films with an O2 gas anneal of the shortest duration tested, 1 min.105 Similarly, Suzuki et al. showed that RF sputtered HfO2 films with no O2 process gas demonstrated the best ferroelectric performance. It was hypothesized that the oxygen-rich environment during deposition generated interstitial oxygen, which has a deleterious effect on the insulating properties of HfO2 films.109 Inversely, it could also be said that a lower O2 flow during sputter deposition aids in the emergence of oxygen vacancies which help stabilize the ferroelectric phase in HfO2.

Controlling the exposure to oxygen gas before film deposition may also influence ferroelectric properties in sputter-deposited films. Szyjka et al. investigated the effect of oxygen gas supplied prior to the sputter deposition of HfO2 films using hard XPS.121 The supply of oxygen gas before deposition was found to stabilize the TiN/HfO2 interface with a self-limiting TiO2 layer of up to 3.5 nm thickness and prevented the formation of an interfacial HfN layer. Moreover, the ratio of O2/Ar during sputter deposition of HfO2 strongly impacts the remanent polarization with Pr decreasing from 12 to 7.5 μC/cm2 when O2/Ar is set higher than 0.75/20 SCCM. Thus, oxygen supply during film deposition and annealing is an important parameter to consider in optimizing a sputter deposition film process.

In summary, we have considered the role of oxygen content in sputter deposition-based HfO2 films. We described how oxygen vacancies may be easier to form in the sputter deposition method versus ALD due to the possibility of using very low oxygen process gas (<1:10 O2/Ar ratio) as well as the mechanism of physical bombardment which may more readily dislodge atoms and form defect complexes. In general, the O2/Ar gas ratio should be kept to maximize the remanent polarization in sputter deposition HfO2 films. Oxygen processing can also be optimized before and after deposition of the film; a predeposition O2 flow is possible to create a thin TiO2 layer that can prevent the formation of an HfN layer. To achieve the highest remanent polarization, it is also recommended to use N2 process gas rather than O2 during annealing treatment.

3. Deposition pressure

In addition to reactive oxygen flow in the sputter deposition process, the pressure during sputter deposition may also play an important role in deposition and crystallization of ferroelectric HfO2 films. The pressure during deposition can affect the microstructure and chemical quality of resultant films. One example is a study by Lee et al. on HfxZr1−xO2 sputter-deposited films, in which GIXRD results revealed that monoclinic phase 111-type peaks increased in intensity as working pressure increased from 0.13 to 1.33 Pa.105 It is important to note that the base pressure was kept constant at 6.66 × 10−5 Pa for both deposition processes; thus, one would expect the impurity levels to be similar for both deposition processes. It is well known that deposition pressure may have a measurable effect on the stoichiometry and mechanical stress on resultant oxide films.122,123 In zinc oxides, increasing the sputtering pressure has been correlated with a decrease in the intrinsic compressive stress by a factor of 2.123 For example, one study of sputter-deposited cerium oxide decreasing the deposition pressure from 30 to 20 mTorr was found to decrease the oxygen content by 20 at. %.122 

Altering the background pressure during sputter deposition impacts the mean free path and scattering of deposited species. In this way, the scattering of lighter element species such as oxygen or nitrogen may increase or decrease with changes in the deposition pressure. Increasing the working pressure could decrease the mean free path of particles in the deposition chamber and may alter the stoichiometry and, in turn, influence the phase composition of films. Increasing the working pressure from 0.13 to 1.33 Pa in the study by Kim et al. altered the number of bombarded particles in the chamber and the resputtering of oxygen anions may have led to increased scattering of oxygen species and resulted in higher oxygen content films.

The topic of working pressure on ferroelectric films was also investigated by Bouaziz et al. who found lower working pressure is favorable for eliciting ferroelectricity in HfO2 films.103 Bouaziz et al. evaluated the effect of the working pressure on the structure and properties of magnetron sputter-deposited HZO films from a single Hf/Zr target. HZO films deposited under lower working pressure (5 × 10−3 mbar) conditions showed considerably higher orthorhombic peak intensity upon annealing and minimum monoclinic peak intensity. In comparison, HZO films fabricated under higher working pressure (5 × 10−2 mbar) conditions showed strong monoclinic 113¯ peak intensity in as-deposited films and no appearance of the orthorhombic peak at any of the annealing conditions (400, 500, and 600 °C). Corroborating the GIXRD measurements, P-E measurements indicated a ferroelectric response (Pr value ∼ 5 μC/cm2) for samples deposited at low working pressure while no ferroelectric hysteresis was observed for samples processed at high working pressure. These results agree with the observations made by Lee et al. and suggest that lower working pressure during sputter deposition may be beneficial to enhanced properties in sputter deposition-based ferroelectric HfO2 films. Bouaziz et al. attributed the differences in structure and properties of films fabricated under low and high working pressure due to the propensity of lower working pressure to generate amorphous films, which could increase the likelihood of the appearance of the orthorhombic phase upon annealing.

In summary, Sec. II B introduced the effect of dopant, oxygen content, and deposition pressure in the processing of ferroelectric HfO2-based films via sputter deposition. The sputter deposition studies discussed here, their process parameters, and reported ferroelectric properties are listed in Table III. While there are similarities between HfO2-based films fabricated via ALD and sputter deposition, such as the effect of oxygen defects, there are also many differences that emerge from the variation in energetics and chemistry of ALD versus sputter deposition. Sputter deposition is a physical deposition method where the ballistic nature of the deposition may be more likely to create oxygen vacancies and related defects, which in turn may decrease the required doping concentration to induce ferroelectricity. The oxygen content, mechanical stress, and resultant phase composition may also be controlled via deposition pressure where decreased deposition pressures have resulted in the decrease in monoclinic phase formation in HfO2 films. Sputter-deposited films may also exhibit lower carbon contamination due to the absence of carbon-based precursors depending on the chosen deposition pressure. Since many sputter deposition processes are conducted at room temperature, sputtering may also be beneficial in suppressing the formation of nuclei in the chamber and thus result in better control of the phase formation during annealing. However, there are also drawbacks associated with sputter-deposited HfO2 thin films. For example, sputter deposition is not conformal like ALD, making applications for three-dimensional structures difficult. Sputter deposition is also not self-limiting, and the deposition rate may be a function of the target chemical composition and age, where targets near the end of their lifetime typically show changes in deposition rates. Moreover, the oxygen concentration in metal oxide targets may vary across vendors and careful attention should be dedicated to ensuring the repeatability of film quality.

TABLE III.

Process parameters of HfO2 thin films fabricated via sputter deposition.

MaterialDoping cat. %t (nm)Target material (s)Dopant target power (W), sizeHfO2 target power (W), sizeBase pressure (mbar)Deposition pressure (mbar)Reactive gas flows (SCCM)Td (°C)Annealing conditionsTop/bottom electrode (°C)PrEcRef.
N:HfO2 0.34 at. % 28 HfO2 N/A 80 6.00E – 08 1.20E – 03 Ar:N2 15–10:5–10; 20 total 25 600/30s Ge(111)/TiN 10 38  
Polycrystalline X:HfO2 Sc, Y, Nb, Al, Si, Ge, or Zr 27 HfO2 5–100 80 N/A N/A 20 25 500–700 Ge(111)/TiN 10 42  
Polycrystalline HfO2 none 24 HfO2 N/A 100 N/A 3.50E – 03 Ar:O2 2:0.2–1.0 25 600/30s/N2 Al/Al N/A 0.3 43  
Polycrystalline Zr:HfO2 50% 15 HfO2, ZrO2 200 200 N/A 1.30E – 02 N/A 500 None TiN/Au 101  
Polycrystalline Zr:HfO2 15.64 mol. % 10 HfO2, ZrO2 110, 3in 60, 3 in. N/A N/A N/A 25 500/30s/N2 TiN/TiN 10 0.75 102  
Polycrystalline Zr:HfO2 16.2 mol. % 10 HfO2, ZrO2 110, 3in 60, 3 in. N/A N/A N/A 25 500/30s/N2 TiN/TiN 0.5 102  
Polycrystalline Zr:HfO2 17.54 mol. % 10 HfO2, ZrO2 110, 3in 60, 3 in. N/A N/A N/A 25 500/30s/N2 TiN/TiN 2.5 0.25 102  
Polycrystalline Zr:HfO2 50% 10 HZO 100 100, 4 in. 5.00E – 07 5.00E – 03 Ar:O2
50:10 
25 600/30s/N2 TiN/TiN N/A 1.1 103  
Polycrystalline Zr:HfO2 50% 10 HZO 100 100, 4 in. 5.00E – 07 5.00E – 02 Ar:O2
50:10 
25 600/30s/N2 TiN/TiN 1.1 103  
Polycrystalline Zr:HfO2 14.60% 30 HfO2, ZrO2 60 150 N/A 8.00E – 03 Ar:O2 3:2 25 N/A Al/Al N/A N/A 104  
Polycrystalline Zr:HfO2 50% 11 HfO2, ZrO2 190, 4in 120, 4 in. 6.66E – 07 1.30E – 03 Ar:O2 1:0 25 600/60s/N2 TiN/TiN 10 105  
Polycrystalline Zr:HfO2 50% 11 HfO2, ZrO2 190, 4in 120, 4 in. 6.66E – 07 1.30E – 03 Ar:O2 0.975:0.025 25 600/60s/N2 TiN/TiN 105  
Polycrystalline Y:HfO2 0.9 mol. % 12 HfO2, Y2O3 20 150 N/A 1.10E – 03 N/A 25 1000/1s/N2 TiN/TiN 12.5 106  
Polycrystalline HfO2 None 12 HfO2, Y2O3 None 150 N/A 1.10E – 03 N/A 25 1000/1s/N2 TiN/TiN 106  
Polycrystalline Y:HfO2 1.21 mol. % 37 Hf, Y 15, 60 × 5 mm 103.5, 150 × 50 × 5 mm 1.50E – 05 1.20E – 02 Ar:O2 20:20 400 750/40s Si/TiN 10 2.5 107  
Polycrystalline Y:HfO2 with 1 nm HfO2 1.5 mol. % 10 Hf, Y 30 100 N/A 7.00E – 04 Ar:O2 20:20 25 850/40s/N2 Si/TiN 14 2.2 108  
Epitaxial Y:HfO2 7 mol. % 24 YHO-7 50 50 N/A 2.70E – 01 Ar:O2 1:0 25 1000/10s/N2 ITO/YSZ 10 109  
Epitaxial Y:HfO2 7 mol. % 24 YHO-7 50 50 N/A 2.70E – 01 Ar:O2 1:0 25 1000/10s/N2 ITO/YSZ N/A N/A 109  
Epitaxial Fe:HfO2 0.06 mol. % 20 HfO2, FeO1.5 N/A N/A N/A 5.10E – 05 N/A 25 900/10min/N2 ITO/YSZ 8.8 110  
Polycrystalline HfO2 None 20 HfO2 None 100, 3 in N/A 1.20E – 03 Ar 20 25 800/20s/N2 TiN/TiN 10 111  
Polycrystalline Zr:HfO2 35% 10 HfO2, ZrO2 200, 3 in 200, 3 in 1.00E – 09 1.20E – 03 Ar:O2 20:1.25 25 800/20s/N2 TiN/TiN 16 N/A 120  
Polycrystalline Zr:HfO2 50% 10 HfO2, ZrO2 100 100 5.00E – 07 5.00E – 02 Ar:O2 50:10 25 600/30s/N2 TiN/TiN 149  
MaterialDoping cat. %t (nm)Target material (s)Dopant target power (W), sizeHfO2 target power (W), sizeBase pressure (mbar)Deposition pressure (mbar)Reactive gas flows (SCCM)Td (°C)Annealing conditionsTop/bottom electrode (°C)PrEcRef.
N:HfO2 0.34 at. % 28 HfO2 N/A 80 6.00E – 08 1.20E – 03 Ar:N2 15–10:5–10; 20 total 25 600/30s Ge(111)/TiN 10 38  
Polycrystalline X:HfO2 Sc, Y, Nb, Al, Si, Ge, or Zr 27 HfO2 5–100 80 N/A N/A 20 25 500–700 Ge(111)/TiN 10 42  
Polycrystalline HfO2 none 24 HfO2 N/A 100 N/A 3.50E – 03 Ar:O2 2:0.2–1.0 25 600/30s/N2 Al/Al N/A 0.3 43  
Polycrystalline Zr:HfO2 50% 15 HfO2, ZrO2 200 200 N/A 1.30E – 02 N/A 500 None TiN/Au 101  
Polycrystalline Zr:HfO2 15.64 mol. % 10 HfO2, ZrO2 110, 3in 60, 3 in. N/A N/A N/A 25 500/30s/N2 TiN/TiN 10 0.75 102  
Polycrystalline Zr:HfO2 16.2 mol. % 10 HfO2, ZrO2 110, 3in 60, 3 in. N/A N/A N/A 25 500/30s/N2 TiN/TiN 0.5 102  
Polycrystalline Zr:HfO2 17.54 mol. % 10 HfO2, ZrO2 110, 3in 60, 3 in. N/A N/A N/A 25 500/30s/N2 TiN/TiN 2.5 0.25 102  
Polycrystalline Zr:HfO2 50% 10 HZO 100 100, 4 in. 5.00E – 07 5.00E – 03 Ar:O2
50:10 
25 600/30s/N2 TiN/TiN N/A 1.1 103  
Polycrystalline Zr:HfO2 50% 10 HZO 100 100, 4 in. 5.00E – 07 5.00E – 02 Ar:O2
50:10 
25 600/30s/N2 TiN/TiN 1.1 103  
Polycrystalline Zr:HfO2 14.60% 30 HfO2, ZrO2 60 150 N/A 8.00E – 03 Ar:O2 3:2 25 N/A Al/Al N/A N/A 104  
Polycrystalline Zr:HfO2 50% 11 HfO2, ZrO2 190, 4in 120, 4 in. 6.66E – 07 1.30E – 03 Ar:O2 1:0 25 600/60s/N2 TiN/TiN 10 105  
Polycrystalline Zr:HfO2 50% 11 HfO2, ZrO2 190, 4in 120, 4 in. 6.66E – 07 1.30E – 03 Ar:O2 0.975:0.025 25 600/60s/N2 TiN/TiN 105  
Polycrystalline Y:HfO2 0.9 mol. % 12 HfO2, Y2O3 20 150 N/A 1.10E – 03 N/A 25 1000/1s/N2 TiN/TiN 12.5 106  
Polycrystalline HfO2 None 12 HfO2, Y2O3 None 150 N/A 1.10E – 03 N/A 25 1000/1s/N2 TiN/TiN 106  
Polycrystalline Y:HfO2 1.21 mol. % 37 Hf, Y 15, 60 × 5 mm 103.5, 150 × 50 × 5 mm 1.50E – 05 1.20E – 02 Ar:O2 20:20 400 750/40s Si/TiN 10 2.5 107  
Polycrystalline Y:HfO2 with 1 nm HfO2 1.5 mol. % 10 Hf, Y 30 100 N/A 7.00E – 04 Ar:O2 20:20 25 850/40s/N2 Si/TiN 14 2.2 108  
Epitaxial Y:HfO2 7 mol. % 24 YHO-7 50 50 N/A 2.70E – 01 Ar:O2 1:0 25 1000/10s/N2 ITO/YSZ 10 109  
Epitaxial Y:HfO2 7 mol. % 24 YHO-7 50 50 N/A 2.70E – 01 Ar:O2 1:0 25 1000/10s/N2 ITO/YSZ N/A N/A 109  
Epitaxial Fe:HfO2 0.06 mol. % 20 HfO2, FeO1.5 N/A N/A N/A 5.10E – 05 N/A 25 900/10min/N2 ITO/YSZ 8.8 110  
Polycrystalline HfO2 None 20 HfO2 None 100, 3 in N/A 1.20E – 03 Ar 20 25 800/20s/N2 TiN/TiN 10 111  
Polycrystalline Zr:HfO2 35% 10 HfO2, ZrO2 200, 3 in 200, 3 in 1.00E – 09 1.20E – 03 Ar:O2 20:1.25 25 800/20s/N2 TiN/TiN 16 N/A 120  
Polycrystalline Zr:HfO2 50% 10 HfO2, ZrO2 100 100 5.00E – 07 5.00E – 02 Ar:O2 50:10 25 600/30s/N2 TiN/TiN 149  

Sections II A and II B introduced and reviewed methods for depositing primarily polycrystalline ferroelectric HfO2-based thin films. In some cases, the growth of epitaxial HfO2 films may be desired. Epitaxial films are films with well-defined single crystalline orientations in both out-of-plane and in-plane directions with respect to their substrate. The synthesis of perfect ferroelectric films on suitable substrates has shown that appropriately strained ferroelectric thin films can exhibit properties superior to their bulk counterparts.119 Many studies concerning epitaxial ferroelectric films have shed light onto the nature of ferroelectric switching, domain structure, and interface effects.124–126 Common epitaxial substrates/electrodes that have been employed include LSMO, yttria-stabilized zirconia (YSZ), and indium tin oxide (ITO). A common impetus for epitaxial studies is to understand the origin of ferroelectricity in HfO2 thin films by removing convoluting features such as grain boundaries, defects, and phase impurity that plague many ALD and sputter deposition studies.

The deposition of HfO2 epitaxial films has been frequently accomplished via PLD. Photon energy of high fluence (typically 1–2 J/cm2) is used to vaporize a target material via pulsed laser beam.123 The PLD technique can be summarized in three main steps: (1) the ablation of material from the target, (2) formation of a highly energetic laser plume, and (3) growth of the film on the substrate. A simplified schematic of a PLD chamber is represented in Fig. 16. Many process parameters, e.g., laser fluence, background pressure, and substrate temperature, are modified to control film morphology and stoichiometry. PLD offers significant benefits over other film deposition methods including the relative ease of controlling film stoichiometry, a high deposition rate (∼100 s Å/min), nanometer thickness control, cleanliness of the process (due to the use of the laser, a filament is not required), and the ability to use numerous target materials which allows the growth of complex oxides. The opportunity to deposit films under nonthermal conditions allows researchers to access chemically nonequilibrium or metastable compositions. PLD may be coupled with an ultrahigh vacuum (1.0 × 10−9 mbar) system to reduce impurities within the chamber during film growth as can similarly be implemented for sputter deposition. Since the deposition pressure for PLD is typically maintained in O2 background gas at a 10−4 −10−1 mbar range, deposition may also be achieved with only moderate vacuum requiring a roughing pump.

FIG. 16.

Simplified schematic of a PLD chamber. Pulsed laser deposition process occurs in three main steps: (1) Highly energetic laser is focused onto a target, (2) ablation plume forms, and (3) matter condenses on substrate holder and leads to film growth.

FIG. 16.

Simplified schematic of a PLD chamber. Pulsed laser deposition process occurs in three main steps: (1) Highly energetic laser is focused onto a target, (2) ablation plume forms, and (3) matter condenses on substrate holder and leads to film growth.

Close modal

The laser beam-solid interaction involved in PLD is quite complex and depends on the laser pulse parameters as well as the parameters of the target material. To briefly describe the mechanism of PLD, the method is divided into three main stages. In the first stage of the process, a laser is focused onto the surface of the target with a sufficiently high energy density and short pulse duration leading to the ablation of the target material. The ablation mechanism involves many phenomena such as electronic excitation, exfoliation, and hydrodynamics. In the second stage, the emitted plume of material migrates toward the substrate where the uniformity is governed by the target-to-substrate distance, laser spot size, and plasma temperature. The third and final stage occurs when high-energy ablated material condenses onto the surface of the substrate.

The crystallinity and quality of the film growth is largely dependent on the mobility and diffusion of atoms on the surface. An adsorbed atom or ion may diffuse across the surface and become bonded as an adatom. In this case, the diffusion of the adatom across the surface is governed by Ds = Doexp{–εD/kT}, where εD is the activation energy for diffusion, k is Boltzmann’s constant, and Do is a diffusion coefficient. Sufficient surface diffusion is required for the adatoms to attain thermodynamically stable positions; therefore, the temperature may be increased to promote rapid and defect-free growth. One potential issue with enhancing the surface diffusion via substrate temperature is the concurrent increase in surface-to-bulk and bulk interdiffusion, which causes a blurring of the boundary line. Thus, the surface mobility of atoms may be optimized not only with the substrate temperature but also via the energy transfer mechanism from impinging species. For a more in-depth description of the physics of the PLD technique, the reader is referred to Willmott and Huber.127 

The ability to create highly stoichiometric, metastable, and epitaxial films via PLD has spurred researchers to investigate the strain effects in HfO2 films. In 2015, the first report of epitaxial orthorhombic HfO2 was published, although the remanent polarization was not reported.128 Fan et al. investigated substrate-induced strain on the effects of ferroelectricity in ZrO2 thin films in 2016.101 Later in 2016, Katayama showed that textured YO1.5-HfO2 films could be grown onto (111)-oriented yttria-stabilized zirconia substrates via PLD.129 In Sec. II C, we describe the effect of substrate choice, deposition temperature, and oxygen flow on PLD-grown film properties. The influence of these process variables on remanent polarization and coercive field is illustrated in Table IV.

TABLE IV.

Process parameters of HfO2 thin films fabricated via PLD.

MaterialDoping %t (nm)Deposition pressure (mbar O2)Base pressure (mbar)Fluence (mJ/cm2)Repetition (Hz)Target materialTd (°C)Annealing conditionsSubstrateTop/bottom electrodePrEcEnduranceRef.
Epitaxial Y:HfO2 7 mol. % 20 0.0133 N/A N/A HfO2, Y2O3 700 None (100)YSZ N/A N/A N/A N/A 128  
Epitaxial Y:HfO2 7 mol. % 14 0.0133 N/A N/A HfO2, Y2O3 700 None ITO/(111) YSZ N/A 10 N/A 129  
Epitaxial Y:HfO2 7 mol. % 15 0.0133 N/A N/A HfO2, Y2O3 25 1000/10s/N2 (111) ITO-coated (111) YSZ Pt 18 N/A 137  
Polycrystalline Y:HfO2 3.8 mol. % 15 0.01 2.00E – 07 5 HfO2,
2 Y2O3 
HfO2, Y2O3 200 600/60s/N2 SiO2/Si TiN/TiN 15 0.9 107 at 0.8 MV/cm 112  
Polycrystalline Y:HfO2 7 mol. % 930 0.0133 N/A HfO2, Y2O3 25 1000/10s/N2 (111)Pt/TiOx/SiO2/(001)Si Pt 17 1.6 N/A 134  
Polycrystalline Y:HfO2 5 mol. % 12 0.00067 N/A N/A N/A HYO 25 760–870/1 min/N2 Si(001)/SiO2 Pt/ITO 13 N/A 113  
Epitaxial Zr:HfO2 50 at. % 0.1 N/A HZO 800 N/A (001)-oriented LSMO/STO LSMO 34 5.1 N/A 27  
Epitaxial Zr:HfO2 50 at. % 7.7 0.1 N/A N/A HZO 800 None SrTiO3/LSMO Pt 34 109 cycles 130  
Epitaxial Zr:HfO2 50 at. % 9.5 0.1 N/A N/A N/A HZO 800 None TbScO3 LSMO 23 N/A 132  
Epitaxial Zr:HfO2 50 at. % 9.5 0.1 N/A N/A N/A HZO 800 None GdScO3 LSMO 22 N/A 132  
Epitaxial Zr:HfO2 50 at. % 9.5 0.1 N/A N/A N/A HZO 800 None NdScO3 LSMO 20 N/A 132  
Epitaxial Zr:HfO2 50 at. % 9.5 0.1 N/A N/A N/A HZO 800 None DyScO3 LSMO 19 N/A 132  
Epitaxial Zr:HfO2 50 at. % 9.5 0.1 N/A N/A N/A HZO 800 None SrTiO3 LSMO 11.5 N/A 132  
Epitaxial Zr:HfO2 50 at. % 9.5 0.1 N/A N/A N/A HZO 800 None MgO LSMO 7.5 N/A 132  
Epitaxial Zr:HfO2 50 at. % 9.5 0.1 N/A N/A N/A HZO 800 None NdGaO3 LSMO N/A 132  
Epitaxial Zr:HfO2 50 at. % 9.5 0.1 N/A N/A N/A HZO 800 None YAlO3 LSMO N/A 132  
Epitaxial Zr:HfO2 50 at. % 9.5 0.1 N/A N/A N/A HZO 800 None LAST LSMO N/A 132  
Epitaxial Zr:HfO2 50 at. % 9.5 0.1 N/A N/A N/A HZO 800 None LaAlO3 LSMO 3.5 N/A 132  
Epitaxial Zr:HfO2 50 at. % 15 6.70E – 04 N/A 6.50E – 01 HZO 700 500/30 s/air (001)-YSO TiN 15 N/A 133  
Epitaxial Zr:HfO2 50 at. % 9.2 0.02 NA N/A N/A N/A 800 None LSMO/STO(001) Pt 10 2.6 N/A 135  
Epitaxial Zr:HfO2 50 at. % 9.2 0.1 NA N/A N/A N/A 800 None LSMO/STO(001) Pt 24 2.6 N/A 135  
Epitaxial Zr:HfO2 50 at. % 9.2 0.2 NA N/A N/A N/A 800 None LSMO/STO(001) Pt 17 2.6 N/A 135  
MaterialDoping %t (nm)Deposition pressure (mbar O2)Base pressure (mbar)Fluence (mJ/cm2)Repetition (Hz)Target materialTd (°C)Annealing conditionsSubstrateTop/bottom electrodePrEcEnduranceRef.
Epitaxial Y:HfO2 7 mol. % 20 0.0133 N/A N/A HfO2, Y2O3 700 None (100)YSZ N/A N/A N/A N/A 128  
Epitaxial Y:HfO2 7 mol. % 14 0.0133 N/A N/A HfO2, Y2O3 700 None ITO/(111) YSZ N/A 10 N/A 129  
Epitaxial Y:HfO2 7 mol. % 15 0.0133 N/A N/A HfO2, Y2O3 25 1000/10s/N2 (111) ITO-coated (111) YSZ Pt 18 N/A 137  
Polycrystalline Y:HfO2 3.8 mol. % 15 0.01 2.00E – 07 5 HfO2,
2 Y2O3 
HfO2, Y2O3 200 600/60s/N2 SiO2/Si TiN/TiN 15 0.9 107 at 0.8 MV/cm 112  
Polycrystalline Y:HfO2 7 mol. % 930 0.0133 N/A HfO2, Y2O3 25 1000/10s/N2 (111)Pt/TiOx/SiO2/(001)Si Pt 17 1.6 N/A 134  
Polycrystalline Y:HfO2 5 mol. % 12 0.00067 N/A N/A N/A HYO 25 760–870/1 min/N2 Si(001)/SiO2 Pt/ITO 13 N/A 113  
Epitaxial Zr:HfO2 50 at. % 0.1 N/A HZO 800 N/A (001)-oriented LSMO/STO LSMO 34 5.1 N/A 27  
Epitaxial Zr:HfO2 50 at. % 7.7 0.1 N/A N/A HZO 800 None SrTiO3/LSMO Pt 34 109 cycles 130  
Epitaxial Zr:HfO2 50 at. % 9.5 0.1 N/A N/A N/A HZO 800 None TbScO3 LSMO 23 N/A 132  
Epitaxial Zr:HfO2 50 at. % 9.5 0.1 N/A N/A N/A HZO 800 None GdScO3 LSMO 22 N/A 132  
Epitaxial Zr:HfO2 50 at. % 9.5 0.1 N/A N/A N/A HZO 800 None NdScO3 LSMO 20 N/A 132  
Epitaxial Zr:HfO2 50 at. % 9.5 0.1 N/A N/A N/A HZO 800 None DyScO3 LSMO 19 N/A 132  
Epitaxial Zr:HfO2 50 at. % 9.5 0.1 N/A N/A N/A HZO 800 None SrTiO3 LSMO 11.5 N/A 132  
Epitaxial Zr:HfO2 50 at. % 9.5 0.1 N/A N/A N/A HZO 800 None MgO LSMO 7.5 N/A 132  
Epitaxial Zr:HfO2 50 at. % 9.5 0.1 N/A N/A N/A HZO 800 None NdGaO3 LSMO N/A 132  
Epitaxial Zr:HfO2 50 at. % 9.5 0.1 N/A N/A N/A HZO 800 None YAlO3 LSMO N/A 132  
Epitaxial Zr:HfO2 50 at. % 9.5 0.1 N/A N/A N/A HZO 800 None LAST LSMO N/A 132  
Epitaxial Zr:HfO2 50 at. % 9.5 0.1 N/A N/A N/A HZO 800 None LaAlO3 LSMO 3.5 N/A 132  
Epitaxial Zr:HfO2 50 at. % 15 6.70E – 04 N/A 6.50E – 01 HZO 700 500/30 s/air (001)-YSO TiN 15 N/A 133  
Epitaxial Zr:HfO2 50 at. % 9.2 0.02 NA N/A N/A N/A 800 None LSMO/STO(001) Pt 10 2.6 N/A 135  
Epitaxial Zr:HfO2 50 at. % 9.2 0.1 NA N/A N/A N/A 800 None LSMO/STO(001) Pt 24 2.6 N/A 135  
Epitaxial Zr:HfO2 50 at. % 9.2 0.2 NA N/A N/A N/A 800 None LSMO/STO(001) Pt 17 2.6 N/A 135  

In epitaxial studies, a substrate with a similar lattice parameter to that of the film of interest, e.g., orthorhombic HfO2 (a = 5.01 Å, b = 5.05 Å, c = 5.24 Å)14 is used for lattice-matching. In 2019, Lyu et al. revealed enhanced ferroelectricity in HZO films using STO template layers and LSMO electrodes.130 Interestingly, this study reported an out-of-plane d111 corresponding to the orthorhombic phase of 2.940 Å, which is smaller than the equivalent (111) interplanar spacing found in other studies of HZO films. Lyu et al. claimed that the smaller spacing may be due to differences in strain of the LSMO electrodes and thermal expansion mismatch between oxide layers and Si and STO substrate. Smaller than expected d-spacings in HfO2-based ultrathin films have been reported elsewhere and are linked to the discovery of the rhombohedral phase in HfO2, which were also fabricated with an epitaxial characteristic. Wei et al. also utilized PLD to fabricate ultrathin HZO films on LSMO-buffered (001)-SrTiO3 substrates.27 Using φ scans and θ−2θ scans around asymmetric {111} reflections via synchrotron x-ray diffraction, Wei et al. observed that the in-plane d-spacing for (d111¯=d11¯1=d111=2.94Å ) was smaller than the out-of-plane 111 reflection (d111 = 2.98 Å). A combination of texture mapping and selected area diffraction patterns elucidated differences in in-plane and out-of-plane d-spacings, which revealed a multiplicity that was well matched to that of a rhombohedral cell.

While the mechanism by which the rhombohedral phase emerges in epitaxial HZO films is still unclear, Wei et al. suggested that both epitaxial strain and nanoparticle pressure play a role in the stabilization of the R3m phase.27 Crystallites grown in the energetically favorable (111) orientation are subjected to an epitaxial compressive strain, which could elongate the pesudocubic cell in the out-of-plane direction and form the rhombohedral cell. This effect is diminished at larger film thicknesses due to the relief of internal pressure from the formation of the monoclinic phase; thus, the rhombohedral phase is most likely to be present in ultrathin films <9 nm. While first-principles simulations by Wei et al. revealed the possibility of an epitaxial compression corresponding to an out-of-plane d111 ∼ 3.25 Å for a ferroelectric R3m phase with a theoretical Pr ∼ 15 μC/cm2, the computational models were unable to definitively link with the observed experimental report (which demonstrated out-of-plane d111 ∼ 2.98 Å and Pr ∼ 34 μC/cm2) due to the multiphase nature of films grown with larger thicknesses. Recent work by Zhang et al. confirmed that polarization could in theory be achieved in the R3m phase under large compressive strain (∼5%); however, the rhombohedral phase remains more stable than the orthorhombic phase only in the case of ultrathin films (<5 monolayers).131 

Despite the differences in phase composition reported by Wei et al. and Lyu et al., both authors emphasized the importance of the strain relationship between the HZO layer and the underlying LSMO/STO substrate to achieve ferroelectricity. In 2019, Estandía et al. further studied the role of epitaxial strain and compression on ferroelectric properties by using substrates of various lattice parameters.132 Bilayers of HZO/LSMO were grown via pulsed laser deposition on substrates with lattice parameters ranging from 3.71 to 4.21 Å on different substrate materials shown in Fig. 17. Estandía et al. described how the epitaxial strain experienced by the electrode, LSMO in the present case, can further translate into a strain state experienced by the HZO and LSMO interface as was previously shown in the works of Wei et al. and Lyu et al. The range of Pr values reported by Estandía et al. was 5 μC/cm2 for films on LSAT [lattice parameter (as) = 3.868 Å] and substrates with a smaller lattice parameter, to around 25 μC/cm2 for films on TbScO3 (as = 3.96 Å) and GdScO3 (as = 3.97 Å). Electrical properties were analyzed in conjunction with GIXRD and 2θ-χ measurements where a higher intensity of the orthorhombic phase was observed for films deposited onto substrates which possessed a larger lattice parameter. As demonstrated by the increasing intensity of the orthorhombic 111 peak of HZO as a function of substrate lattice parameter, the amount of orthorhombic phase in HZO can be enhanced with substrates of pseudocubic lattice parameter larger than 3.87 Å. Such an increase in lattice parameter gives rise to a tensile strain state between the LSMO bottom electrode and HZO film.

FIG. 17.

(a) Lattice mismatch between LSMO electrode and substrate onto which HZO layers were deposited, (b) XRD θ -2θ symmetric scans of HZO/LSMO bilayers on various substrates. Reprinted with permission from S. Estandía, N. Dix, J. Gazquez, I. Fina, J. Lyu, M. F. Chisholm, J. Fontcuberta, and F. Sánchez, ACS Appl. Electron. Mater. 1, 1449 (2019). Copyright 2019, American Chemical Society.

FIG. 17.

(a) Lattice mismatch between LSMO electrode and substrate onto which HZO layers were deposited, (b) XRD θ -2θ symmetric scans of HZO/LSMO bilayers on various substrates. Reprinted with permission from S. Estandía, N. Dix, J. Gazquez, I. Fina, J. Lyu, M. F. Chisholm, J. Fontcuberta, and F. Sánchez, ACS Appl. Electron. Mater. 1, 1449 (2019). Copyright 2019, American Chemical Society.

Close modal

At this point, it is valuable to discuss the discrepancies between the works of Estandía et al. and Lyu et al. with that of Wei et al., the latter of which reported a rhombohedral phase in HfO2 thin films. Both Estandía et al. and Lyu et al. reported that HZO films grown on STO with a lattice parameter of ∼3.905 Å were composed of predominately orthorhombic phase; moreover, the lattice mismatch between that of HZO and STO was ascribed as a key enabler of orthorhombic phase stabilization. In the case of Wei et al., HZO grown on STO/LSMO achieved the rhombohedral phase under similar growth conditions to the aforementioned studies. A major factor that may have uniquely promoted stabilization of the rhombohedral phase in Wei et al. may include the use of ultrathin film thicknesses (1.5–9 nm) which experience significant in-plane compressive strain (up to ∼8% at the HZO-LSMO interface). In reconciling the similar ferroelectric properties reported across these epitaxial studies, it is also important to consider that the orthorhombic phase may still be the primary driver of ferroelectricity in epitaxial thin films. Wei et al. reported from DFT calculation that the R3m phase is weakly polar with a Pr ∼ 1 μC/cm2. Incorporating epitaxial compression into the DFT simulation resulted in a Pr ∼ 15 μC/cm2, which still does not explain the reported Pr of 34 μC/cm2 for 5 nm thick films. While it is difficult to discriminate the phase composition due to the closely overlapping diffraction signatures, the polar orthorhombic phase likely remains the leading driver of ferroelectricity in most epitaxial thin films. Lastly, while the elusive rhombohedral phase was only experimentally reported by Wei et al., it is possible that epitaxial studies using similar substrates may have overlooked the detection of R3m due to the similarity of their d111 and the in-depth texture mapping required to discern the differences in the symmetry, i.e., d111 ≠ d11–1. Regardless of the seeming discrepancies among these studies, the strain imposed between the substrate and the HZO ferroelectric film remains to be a major driver of ferroelectricity in epitaxial HfO2 films.

Tensile and compressive stresses are well-known drivers for the orthorhombic phase transition in polycrystalline films. In 1998, Kisi and Howard reported that the transition from tetragonal to orthorhombic phase ZrO2 was facilitated via compressive stress along the a-b axis and tensile stress along the c-axis.40 The necessary stress was ∼760 MPa for tensile and ∼50 MPa for compressive stress. Other studies such as Park et al. have also proposed coalescence stress as the driving force for stabilizing the orthorhombic phase.45 In the Volmer–Weber-type growth mode, stress is formed during the island coalescence stage. In Park et al., HZO films were deposited onto TiN/Si substrates and ferroelectricity only present with tensile strains in excess of a critical value of 1.5% (corresponding to 4.7 GPa tensile strain). Strain, thus, plays an important role in the emergence of ferroelectricity in both polycrystalline and epitaxial films.

A strain effect analogous to that of the capping layer effect in polycrystalline films may be induced by substrates in epitaxial films. The function of the capping layer in polycrystalline films is to provide mechanical confinement in such a way that prevents the volume expansion and shearing of the HfO2 unit cell and inhibiting the monoclinic phase formation.2,6 Similarly, the HZO/LSMO bilayer may also experience mechanical confinement from the underlying substrate. To better understand the mechanism of “transferring” the epitaxial relationship from the substrate to the electrode and then to the ferroelectric film, Li et al. grew epitaxial HZO films on differently oriented (001), (011), and (111) YSZ substrates with TiN bottom electrodes.133 Cross-sectional, high-resolution transmission electron microscopy (TEM) reveals lamellar ferroelectric nano-domains. Selected-area electron diffraction patterns obtained at the interface of the HZO/TiN interface reveal epitaxial growth between HZO and TiN and that the underlying substrate is accessed through an interfacial oxide reaction, which causes the formation of a TiO2 bridge layer. EDS also revealed interatomic interdiffusion at the TiN/HZO interface with an out-of-plane lattice constant consistent with that of (011)-TiO2. Li et al. asserted that the interface strain was favorable to stabilize the ferroelectric orthorhombic phase. Taken collectively, the aforementioned studies reveal the importance of the HfO2/TiN interface and the effects of strain engineering on ferroelectric properties.

The growth window, or the range of process parameters and film characteristics in which ferroelectricity is achieved in PLD films, is also an important parameter to consider in PLD and may be wider than the growth window found in ALD and sputter deposition films. Mimura et al. studied 7% Y-doped polycrystalline HfO2 (YHO7) films with a thickness range from 10 to 930 nm on platinized (001) Si substrates.134 According to x-ray diffraction analysis, the authors found that films demonstrated a majority of orthorhombic or tetragonal phase even when films reached a thickness of 930 nm and demonstrated a remanent polarization of 14–17 μC/cm2. However, XRD patterns also showed that the lower-symmetry monoclinic phase appeared in films with thicknesses greater than 210 nm. Despite the presence of the deleterious monoclinic phase, clear hysteresis loops of comparable remanent polarization were observed for all films ranging from 27 to 930 nm. Mimura et al. suggest that dopant selection may be a more important parameter in stabilizing the orthorhombic phase for a higher film thickness range.

In the HfO2-ZrO2 system, while the two oxides form a complete solid solution with a stable monoclinic phase at room temperature (Fig. 1), the solubility of YO1.5 in HfO2 is much lower and thus the HfO2-YO1.5 chemical phase which promotes the orthorhombic phase may still be present at higher thicknesses. In HfO2-ZrO2, on the other hand, an increase in the film thickness yields the monoclinic phase as surface energy effects diminish in thicker films. Y-doping of HfO2 is more likely to impact the stabilization of the orthorhombic phase via alternative mechanisms such as the formation of dopant-oxygen bonding and the stabilization of oxygen defects which has been identified as key enablers of the orthorhombic phase.22,41 These drivers of ferroelectricity may be able to persist in thicker films in the case of Y-doped HfO2 and suggests that dopant selection and concentration, rather than thickness, may be a more important factor for stabilizing the orthorhombic phase in thicker, PLD-grown HfO2-based films.

Lyu et al. similarly studied the effect of deposition parameters on the structural and electrical properties of PLD-grown HZO films.135 The growth window of HZO was investigated by varying film thickness, deposition temperature, and oxygen pressure in the chamber. Increasing film thickness showed decreased remanent polarization and orthorhombic phase fraction, similar to the behavior of HfO2 films deposited via ALD and sputtering methods. The dependence of film thickness on ferroelectric polarization in PLD-based HZO contrasts with the trends observed in Y-doped HfO2 by Mimura et al., which found little correlation between film thickness and remanent polarization. This observation suggests that dopant choice can dramatically affect the phase stability in HfO2 at different thickness ranges. Thus, rare earth dopants such as yttrium may be useful in stabilizing the ferroelectric orthorhombic phase in thicker films than the solid-solution HfO2-ZrO2.

Surprisingly, and in contrast to ALD and sputter deposition films, PLD films grown by Lyu et al. showed increased remanent polarization with increasing oxygen pressure during deposition. Additionally, XRD analysis revealed a concurrent increase in the monoclinic phase and a decrease in the orthorhombic phase during low oxygen pressure deposition (0.02 mbar). Oxygen pressure was correlated with the phase evolution through the lattice strain experienced by the HZO layer where the 111 peak shifted toward lower 2θ with lower oxygen pressure processing. One possible explanation for this observation may be due to the differences in oxygen concentration of the PLD target versus that of ALD or sputter targets, which could affect the overall optimized oxygen concentration.

Another key difference between PLD and other deposition methods is the phase transformation pathway experienced by the deposited films. ALD or sputter deposition is typically accomplished at < 300 °C temperatures and requires HfO2 films to be subsequently annealed to achieve ferroelectricity. The phase transformation pathway proceeds according to the following sequence: the amorphous or cubic phase is transformed to tetragonal during annealing, and then strain and surface energy effects promote the metastable orthorhombic phase (strained tetragonal) during cooling. In PLD films (which are typically epitaxial), on the other hand, the starting phase is typically the thermodynamically favored monoclinic phase which transforms to the orthorhombic phase during heating either during deposition or in a postdeposition annealing process.27,128–130,132–134,136 Whereas PLD-based films are more likely to follow the classical thermodynamic pathway predicted in the HfO2 phase diagram (Fig. 1), nonequilibrium effects dominate for ALD- and sputter-deposited films, resulting in a phase transformation pathway not predicted by the thermodynamic phase diagram.

In situ XRD heating experiments on Y-doped HfO2 by Mimura et al. showed that the phase evolution proceeded from monoclinic to tetragonal to orthorhombic and is plotted as a function of temperature in Fig. 18.137In situ XRD results of Mimura et al. directly contrast the phase formation studies in ALD-based HfO2 films, which typically describe that the transformation to the monoclinic phase is irreversible and not parent to the orthorhombic phase.45,49,50 The phase evolution observed in ALD-based HfO2 films follows (1) as-deposited amorphous with small nuclei of tetragonal phase, (2) upon heating a mixed phase of tetragonal and orthorhombic forms, and (3) further heating leads to an irreversible phase transformation to the monoclinic phase. However, in the case of epitaxial 7 mol. % Y-doped HfO2 on (111) ITO-coated, (111)YSZ substrates, the phase transformation proceeds from monoclinic to tetragonal to orthorhombic, i.e., low symmetry to higher symmetry. This phase transformation resembles that of the thermodynamically predicted pathway and is most likely due to the slow heating process of the in situ measurement and the thermodynamically favorable epitaxial nature of the films. However, the report by Mimura et al. also differs from other epitaxial studies which claim an as-deposited orthorhombic or rhombohedral phase in PLD-based films; such discrepancy between Mimura et al. and other reports may be a function of the chosen doping concentration, film thickness, or other PLD processing conditions. While unique opportunities exist for PLD-based HfO2 films such as achieving ferroelectricity in thicker films, perfect phase purity, and novel phase evolution pathways, the characteristics of a PLD-grown HfO2 film may be more strongly dependent on the specific experimental setup present in the individual research groups due to the complex nature of the deposition method compared to the more generalizable ALD and sputter deposition.

FIG. 18.

(a) In situ heating XRD for 7% Y-doped HfO2 films measured from 30 to 1000 °C by ψ scanning at 2θ = 15.7° (λ = 0.0827 nm). (b) Changes in peak intensity as a function of heating and cooling cycle, (c) XRD θ-2θ profiles for YHOY at 1000 °C. Reprinted with permission from T. Mimura, T. Shimizu, T. Kiguchi, A. Akama, T. J. Konno, Y. Katsuya, O. Sakata, and H. Funakubo, Jpn. J. Appl. Phys. 58, SBBB09 (2019). Copyright 2019, The Japan Society of Applied Physics.

FIG. 18.

(a) In situ heating XRD for 7% Y-doped HfO2 films measured from 30 to 1000 °C by ψ scanning at 2θ = 15.7° (λ = 0.0827 nm). (b) Changes in peak intensity as a function of heating and cooling cycle, (c) XRD θ-2θ profiles for YHOY at 1000 °C. Reprinted with permission from T. Mimura, T. Shimizu, T. Kiguchi, A. Akama, T. J. Konno, Y. Katsuya, O. Sakata, and H. Funakubo, Jpn. J. Appl. Phys. 58, SBBB09 (2019). Copyright 2019, The Japan Society of Applied Physics.

Close modal

In this section, we reviewed the role of multiple PLD parameters such as the choice of the substrate (influencing strain), film thickness, dopant selection, and oxygen pressure on the structural and electrical characteristics of HfO2 films, which are summarized in Table IV. Much focus has been dedicated to strain engineering in epitaxial ferroelectric HfO2 films, which has resulted in the understanding that a tensile stress greater than 0.5% may be the most favorable to the formation of the orthorhombic phase. Investigations on the effect of thickness scaling and phase evolution have uncovered the fact that our generalized understanding of HfO2 films in the ALD or sputter deposition systems may not translate to an understanding of HfO2 grown via PLD. For example, Mimura et al. showed that ferroelectricity could be achieved in PLD films grown up to 930 nm134 and that PLD films could transform from the monoclinic to the favorable orthorhombic phase upon heating.137 Moreover, Lyu et al. also revealed that unlike what is observed in ALD and sputter-deposited films, increasing the oxygen pressure during deposition leads to an increased phase fraction of the orthorhombic phase.130 Thus, while PLD-grown epitaxial HfO2 films can provide a new understanding of the structure of HfO2 and the origin of its ferroelectricity, the effect of PLD parameters on structure and properties may not be so easily generalizable to other deposition methods. Notwithstanding, PLD has been useful in elucidating the nature of film-interface strain relationships in HfO2-based films and has also allowed researchers to expand our understanding of the limitations, or lack thereof, of ferroelectricity in HfO2.

While the techniques reviewed thus far including ALD, sputter deposition, and PLD are the most prevalent deposition techniques used in the growth of ferroelectric HfO2 thin films, others have also employed chemical solution deposition (CSD) in the deposition of ferroelectric HfO2-based films.138–143 CSD is a highly versatile technique with deposition options ranging from spin-coating, dip-coating, to even inkjet printing or aerosol jet.144 CSD is already being employed in the field of electronic thin films due to its cost-effectiveness, ease of use, and high yield. Other advantages of the CSD method include the wide tunability with various dopant systems and the possibility of generating ferroelectricity in thicker films. The synthesis methods and process variables and their effect on ferroelectric properties are summarized and reported in Table V. In this section, we review the structure and properties of CSD-based ferroelectric HfO2 films achieved using a variety of different dopants and review current opportunities and challenges of CSD-grown ferroelectric HfO2.

TABLE V.

Process parameters of HfO2 thin films fabricated via CSD.

MaterialDoping cat. %t (nm)SCPaPrecursor chemicalsBaking temp (°C)Annealing conditionsTop/bottom electrodePrEcEnduranceRef.
Polycrystalline HfO2 None 136 3000 rpm for 30s Hafnium 2,4-pentandionate (Alfa Aesar, 99.5%) 120 for 2 h 700/60s/O2 [Au/TiN]/[TiN] 22.56 N/A 107 cycles at 0.7 MV/cm 145  
Polycrystalline CeO2-HfO2 15 mol. % 78 500 rpm for 16s, 3000 rpm for 30s hafnium 2, 4-pentanedionate (C20H28HfO8, Kindo, 97%), cerium nitrate hexahydrate (CeN3O9⋅6H2O, Aladdin, 99.95%) and 2,4-pentanedione 180 for 180 s, 300 for 180 s 800 Pt/n-type Si 20 1.5 109 at 2.9 MV/cm 142  
Polycrystalline La:HfO2 5.00% 45 3000 rpm for 30 s Hf(IV)-acetylacetonate (Alfa Aesar, 97% purity) and La(III)-acetate hydrate (Sigma-Aldrich, 99.9% purity) 215 for 5 min 800/90 s/Ar:O2 1:1 [Pt]/[Pt/Si] 7.7 N/A 140  
Polycrystalline (La, Nd, Sm, Er, or Yb):HfO2 5.2 mol. % 42 N/A Hafnium 2,4-pentandionate (Alfa Aesar) and the 2,4-pentandionate of the corresponding doping metal (Alfa Aesar) N/A 800/90 s/Ar:O2 1:1 [Pt]/[Pt/Si] 12.5 1.6 N/A 139  
Polycrystalline Y:HfO2 5.2 mol. % 42 N/A Hafnium 2,4-pentandionate (Alfa Aesar) and the 2,4-pentandionate of the corresponding doping metal (Alfa Aesar) N/A 800/90 s/Ar:O2 1:1 [Pt]/[Pt/Si] 19 1.6 N/A 139  
Polycrystalline Y:HfO2 5 mol. % 40 N/A Hafnium (IV) acetylacetonate [Hf(acac)4; Wako Pure Chemicals], zirconium (IV) acetylacetonate [Zr(acac)4; Wako Pure Chemicals] and yttrium (III) acetylacetonate, [Y(acac)3; Wako Pure Chemicals] 120 for 1 h 800/3 min/vacuum [Pt/Ti]/ [SiO2/Si] 17 1.5 N/A 114  
Polycrystalline Y:HfO2 5.2 mol. % 3000 rpm for 60 s Hafnium chloride (Sigma Aldrich, 98%) and yttrium chloride 60 for 1 h 800/90s [Pt(111)]/ [TiO2/SiO2/Si] N/A N/A N/A 115  
Polycrystalline Y:HfO2 5.2 mol. % 35 N/A Solgel-educt (hafnium ethoxide, Alfa Aesar, 99.9%), 2,4-pentanedionate (Sigma-Aldrich, GC-grade) per hafnium ion; yttrium 2,4-pentanedionate (Sigma Aldrich, 99.95%) 60 for 30 min 700/5 min/O2 [Pt]/[Pt/Si] 13 2.1 N/A 138  
Polycrystalline Y:HfO2 5 at% 55 N/A Hafnium 2,4-pentandionate (Alfa Aesar) and the Yttrium 2,4-pentandionate (Alfa Aesar) 7 layers × 4 times bake at 200 700/20 min/O2 Pt/n-type Si 15 5.5 N/A 141  
Polycrystalline ZrO2 None 50 3000 rpm for 30 s Zirconium dichloride oxide (ZrOCl2⋅8H2O; 99.99%) 120, 260, 380 for 5 min each sequentially 450/10 min then 700/30s TiN/Si(100) 8.5 1.2 N/A 143  
Polycrystalline HZO  40 3000 rpm for 50 s Hf(O-i-C3H7)4, Zr(O-i-C3H7)4, and 2-methoxyethanol 300/5 min/air 600-800/10 min/N2 Pt/ [(111)Pt/TiO2/(100)Si] 7.5 N/A 146  
Polycrystalline HZO and ZrO2 50 at. % ZrO2 100 N/A Hafnium 2,4-pentandionate and zirconium 2,4-pentadionate 160 for 6 h 800/90s Pt/111 Pt 4.3 × 104 cycles at 2 MV/cm 31  
MaterialDoping cat. %t (nm)SCPaPrecursor chemicalsBaking temp (°C)Annealing conditionsTop/bottom electrodePrEcEnduranceRef.
Polycrystalline HfO2 None 136 3000 rpm for 30s Hafnium 2,4-pentandionate (Alfa Aesar, 99.5%) 120 for 2 h 700/60s/O2 [Au/TiN]/[TiN] 22.56 N/A 107 cycles at 0.7 MV/cm 145  
Polycrystalline CeO2-HfO2 15 mol. % 78 500 rpm for 16s, 3000 rpm for 30s hafnium 2, 4-pentanedionate (C20H28HfO8, Kindo, 97%), cerium nitrate hexahydrate (CeN3O9⋅6H2O, Aladdin, 99.95%) and 2,4-pentanedione 180 for 180 s, 300 for 180 s 800 Pt/n-type Si 20 1.5 109 at 2.9 MV/cm 142  
Polycrystalline La:HfO2 5.00% 45 3000 rpm for 30 s Hf(IV)-acetylacetonate (Alfa Aesar, 97% purity) and La(III)-acetate hydrate (Sigma-Aldrich, 99.9% purity) 215 for 5 min 800/90 s/Ar:O2 1:1 [Pt]/[Pt/Si] 7.7 N/A 140  
Polycrystalline (La, Nd, Sm, Er, or Yb):HfO2 5.2 mol. % 42 N/A Hafnium 2,4-pentandionate (Alfa Aesar) and the 2,4-pentandionate of the corresponding doping metal (Alfa Aesar) N/A 800/90 s/Ar:O2 1:1 [Pt]/[Pt/Si] 12.5 1.6 N/A 139  
Polycrystalline Y:HfO2 5.2 mol. % 42 N/A Hafnium 2,4-pentandionate (Alfa Aesar) and the 2,4-pentandionate of the corresponding doping metal (Alfa Aesar) N/A 800/90 s/Ar:O2 1:1 [Pt]/[Pt/Si] 19 1.6 N/A 139  
Polycrystalline Y:HfO2 5 mol. % 40 N/A Hafnium (IV) acetylacetonate [Hf(acac)4; Wako Pure Chemicals], zirconium (IV) acetylacetonate [Zr(acac)4; Wako Pure Chemicals] and yttrium (III) acetylacetonate, [Y(acac)3; Wako Pure Chemicals] 120 for 1 h 800/3 min/vacuum [Pt/Ti]/ [SiO2/Si] 17 1.5 N/A 114  
Polycrystalline Y:HfO2 5.2 mol. % 3000 rpm for 60 s Hafnium chloride (Sigma Aldrich, 98%) and yttrium chloride 60 for 1 h 800/90s [Pt(111)]/ [TiO2/SiO2/Si] N/A N/A N/A 115  
Polycrystalline Y:HfO2 5.2 mol. % 35 N/A Solgel-educt (hafnium ethoxide, Alfa Aesar, 99.9%), 2,4-pentanedionate (Sigma-Aldrich, GC-grade) per hafnium ion; yttrium 2,4-pentanedionate (Sigma Aldrich, 99.95%) 60 for 30 min 700/5 min/O2 [Pt]/[Pt/Si] 13 2.1 N/A 138  
Polycrystalline Y:HfO2 5 at% 55 N/A Hafnium 2,4-pentandionate (Alfa Aesar) and the Yttrium 2,4-pentandionate (Alfa Aesar) 7 layers × 4 times bake at 200 700/20 min/O2 Pt/n-type Si 15 5.5 N/A 141  
Polycrystalline ZrO2 None 50 3000 rpm for 30 s Zirconium dichloride oxide (ZrOCl2⋅8H2O; 99.99%) 120, 260, 380 for 5 min each sequentially 450/10 min then 700/30s TiN/Si(100) 8.5 1.2 N/A 143  
Polycrystalline HZO  40 3000 rpm for 50 s Hf(O-i-C3H7)4, Zr(O-i-C3H7)4, and 2-methoxyethanol 300/5 min/air 600-800/10 min/N2 Pt/ [(111)Pt/TiO2/(100)Si] 7.5 N/A 146  
Polycrystalline HZO and ZrO2 50 at. % ZrO2 100 N/A Hafnium 2,4-pentandionate and zirconium 2,4-pentadionate 160 for 6 h 800/90s Pt/111 Pt 4.3 × 104 cycles at 2 MV/cm 31  
a

SCP, spin coat parameters.

In 2014, Starschich et al. reported the first ferroelectric Y-doped HfO2 films fabricated via CSD using precursors hafnium 2,4-pentandionate and yttrium 2,4-pentandionate via spin-coating.138 The chemicals were mixed into a solution of propionic acid and propionic acid anhydrite and then spin cast onto a Pt bottom electrode. In 2015 Starschich et al. published a study detailing the optimization of their CSD process flow in which the authors controlled parameters such as intermediary temperature, annealing temperature, and dopant type.139 Starschich et al. highlighted the importance of the intermediary heating step which took place after each layer of spin coating before final annealing. After the solution was spin cast onto a Pt substrate, pre-annealing heat treatment was conducted from 215 to 295 °C to evaporate the solvent and solidify/densify the film. The film densified through a chemical reaction in which hafnium clusters containing bridging oxo-groups and carboxylate form in the presence of propionic acid. The full process flow is depicted in Fig. 19. Starschich et al. found that increasing the pre-annealing heat treatment from 215 to 295 °C resulted in a 40% reduction in remanent polarization. One explanation for this was that the lower heat treatment maintained more solution-like homogeneity and led to better film quality. An annealing temperature of 800 °C generated the largest remanent polarization and a peak shift of the orthorhombic/tetragonal peak at around 2θ ∼ 30° was observed in XRD measurements. The highest remanent polarization of 20 μC/cm2 was found at a doping concentration of 5.2 mol. % Y. The most pronounced issue noted by Starschich et al., however, was that all films suffered from a wake-up effect in which the application of 1000 cycles was required to attain the maximum polarization. Moreover, endurance of 100 nm ferroelectric ZrO2 from CSD was found to be limited to 43k cycles.31 

FIG. 19.

Schematic of process flow of CSD and pyrolysis are repeated until the desired thickness is achieved. The final layered film is annealed and patterned with electrodes for electrical measurement. Reprinted with permission from S. Starschich, D. Griesche, T. Schneller, and U. Böttger, ECS J. Solid State Sci. Technol. 4, P419 (2015). Copyright 2015, The Electrochemical Society, IOP Publishing.

FIG. 19.

Schematic of process flow of CSD and pyrolysis are repeated until the desired thickness is achieved. The final layered film is annealed and patterned with electrodes for electrical measurement. Reprinted with permission from S. Starschich, D. Griesche, T. Schneller, and U. Böttger, ECS J. Solid State Sci. Technol. 4, P419 (2015). Copyright 2015, The Electrochemical Society, IOP Publishing.

Close modal

Other dopants were also investigated such as La, Nd, Sm, and Er; interestingly, all dopants showed nearly identical P-E hysteresis curves and remanent polarization. Starschich et al. also plotted the remanent polarization as a function of yttrium doping concentration for three different deposition techniques, CSD, ALD, and sputter deposition shown in Fig. 20. The effect of yttrium doping concentration on remanent polarization for CSD films was similar to that of the ALD films33 but sputter deposition films required a doping concentration of less than 2 mol. % to achieve the highest remanent polarization. Starschich et al. suggested that sputter deposition may have a lower doping window due to differences in film microstructure or differences in stoichiometry resulting from the deposition at higher vacuum.

FIG. 20.

Comparison of the effect of yttrium concentration on remanent polarization of ALD-, CSD-, and PVD-based films. Reprinted with permission from S. Starschich, D. Griesche, T. Schneller, and U. Böttger, ECS J. Solid State Sci. Technol. 4, P419 (2015). Copyright 2015, The Electrochemical Society, IOP Publishing.

FIG. 20.

Comparison of the effect of yttrium concentration on remanent polarization of ALD-, CSD-, and PVD-based films. Reprinted with permission from S. Starschich, D. Griesche, T. Schneller, and U. Böttger, ECS J. Solid State Sci. Technol. 4, P419 (2015). Copyright 2015, The Electrochemical Society, IOP Publishing.

Close modal

In 2020, Chen et al. confirmed ferroelectricity in nominally pure HfO2 thin films deposited via CSD on Si/TiN electrodes.145 They found that HfO2 thin films exhibited ferroelectricity in thickness ranges from 34 to 136 nm, where 136 nm films showed a Pr of 22.56 μC/cm2 and could endure 107 switching cycles. Chen et al. employed a layer-by-layer annealing technique to achieve control of the grain size similar to Starshich et al. The observation of ferroelectricity at larger thicknesses was attributed to both the annealing technique and the residual carbon from the incomplete decomposition. Chen et al. argued that interstitial carbon resulted in disruption to the grain growth and led to the favorability of the orthorhombic phase in thicker films. However, the incomplete decomposition of organic precursors consisting of acetylacetone and propionic acid also increased the leakage current as was shown by the P-E hysteresis loops where warping of the loop tip, a characteristic feature of leakage contribution within a film, was observed. Despite the leaky P-E loops, switching current measurements could unambiguously prove the existence of ferroelectric switching in undoped, CSD-grown HfO2 films.

A major advantage of CSD growth is the ability to produce thicker films which may pave a way toward novel applications for functional HfO2 materials. Schenk et al. prepared 1-μm thick ferroelectric 20 at. % La-doped HfO2 films on platinized Si substrates. Strain versus electric field loops obtained via double-beam laser interferometer revealed a piezoelectric coefficient of 20 pm/V for the pristine state and 5 pm/V for the fatigued case after 460 cycles; moreover, the Pr value was reported to be 9 μC/cm2.

The existence of 1-μm thick ferroelectric HfO2 seemingly contrasts with many of the previous studies referenced in this article; however, it is important to emphasize the differences in growth window that can be achieved for ferroelectric HfO2 via different deposition methods. The stabilization of thick ferroelectric HfO2 can be understood via grain size and grain boundary effects. In ALD-based films, grains typically span the entire film thickness for thin films <30 nm. With increasing film thickness, the grain boundary density decreases relative to the volume of the film, and the more thermodynamically stable monoclinic phase becomes preferred. Unlike ALD-based films, grain size and grain boundary density in CSD-based films are insensitive to film thickness. Starschich et al. showed that CSD-made 390-nm ZrO2 films exhibit a fine-grain microstructure with a grain size radius of ∼10 nm.31 In addition, it is also possible that nonsurface energy effects may dominate in the stabilization of the orthorhombic phase as was seen in the case of Y:HfO2 films by Mimura et al. where films with grains up to ∼200 nm still exhibited ferroelectricity.134 In the case of Schenk et al., La:HfO2 films showed a fine microstructure with an average grain size of ∼10 nm. While the fine-grained microstructure allowed the establishment of the orthorhombic phase in thicker films (1 μm) than possible from ALD, such fine grains also resulted in low density films which led to films with higher leakage current and worse endurance characteristics compared to ALD films.

Schenk et al. highlighted the challenge of pronounced wake-up effect in CSD-derived HfO2 films. To mitigate the wake-up effect, Samanta et al. proposed a multistep chemical solution deposition/annealing process for Y:HfO2 films by which a furnace anneal was utilized to mitigate oxygen deficiency in films during annealing.141 Endurance measurements were conducted up to 1100 cycles under triangular waves and the highest remanent polarization Pr ∼ 17 μC/cm2 was achieved in the first cycle with subsequent cycles leading to a gradual fatigue and diminished remanent polarization. The multiannealing step process was directly correlated with the reduction of the wake-up effect by evaluating the endurance characteristics of films exposed to different annealing repetitions in between deposition cycles, i.e., 1, 2, 3, 4 intervals of 20 min O2 furnace annealing evenly spaced for a seven-layer deposition. The film that was only annealed once during seven layers of deposition showed a + Pr wake-up of 54.13% whereas a film annealed twice in the oxygen atmosphere with the same number of layers only showed a + Pr wake up of 13.79%.

In addition to minimizing the wake-up effect in Y-doped HfO2 films, others have explored ways to improve the maximum remanent polarization. Zheng et al. used HfO2-CeO2 solid solution thin films fabricated in n-type Si (100) substrates which yielded films with an endurance of 109 cycles, a remanent polarization of ∼20 μC/cm2, and virtually no observed wake-up effect.142 Ferroelectricity in pure ZrO2 thin films was also established via CSD using all-inorganic aqueous salt precursors, which yielded a remanent polarization of 8.5 μC/cm2.143 Interestingly, Wang et al. observed via GIXRD a thickness dependence on the monoclinic phase formation in ZrO2 films where the thickest films studied (50 nm) displayed 111 and 1¯11 peaks characteristic of the monoclinic phase. While Wang et al. did not report the average grain size of their ZrO2 thin films, based on the thickness dependence of the phase composition, it is likely that grain size increased as a function of thickness and yielded the stabilization of the monoclinic phase with increasing thickness unlike in other CSD studies where the thickness was inconsequential to ferroelectricity.31,140

While Zr:HfO2 films are the most commonly employed composition in ALD-based ferroelectric HfO2 studies, Nakayama et al. investigated the structure and properties of CSD-derived 40 nm HfO2-ZrO2 thin films on platinized Si substrates.146 Nakayami et al. reported that a 50:50 composition, similar to compositions used in ALD studies, generated the highest ferroelectric response. The relative intensity of the orthorhombic/tetragonal to monoclinic phase peaks was found to decrease as a function of annealing temperature, a trend that is also commonly observed in ALD-derived HfO2 films. The largest remanent polarization was reported as 8 μC/cm2 for films annealed at 700 °C. The phase transformation pathway of CSD-based HZO was found to follow the tetragonal to orthorhombic to monoclinic pathway commonly observed in ALD films. A similar phase transformation pathway between CSD and ALD-based HZO films suggests a common underlying mechanism for the establishment, and disappearance, of ferroelectricity in Zr:HfO2 thin films.

This section reviewed the use of chemical solution deposition in the creation of ferroelectric HfO2 films with dopants. Several trends are observed including, (1) CSD enables the achievement of thicker ferroelectric HfO2 and ZrO2 films due to the fine grain microstructure of CSD-based films, (2) several similarities exist between ALD and CSD-derived HfO2 films; notably, optimal concentrations of Y-doping and a similar phase transformation pathway as a function of annealing temperature, (3) an important process parameter for CSD optimization is the intermediary annealing step in between layer depositions where the duration, temperature, and atmosphere need to be optimized for each fabrication process, and finally (4) CSD-based HfO2 films are generally plagued by poor endurance characteristics and deleterious wake-up effect albeit recent studies have made strides in improving these two issues. CSD remains a powerful deposition technique due to the benefits of low-cost and scalable production in which organic or inorganic precursors and spin coating are employed during fabrication. In the future, challenges that must be resolved include improving film density in order to decrease leakage current and minimizing defects such as oxygen vacancies to mitigate the wake-up effect and low endurance limit.

While ferroelectricity in HfO2 was previously thought to be limited to the thin film regime, recent reports point to the possibility of growing bulk ferroelectric HfO2. Two orthorhombic phases, Pbca and Pca21, were reported in 5 at. % Gd-doped HfO2 synthesized from a co-precipitation method.147 Banerjee et al. used perturbed angular correlation spectroscopy and compared experimental results to density functional theory to assign crystal structures to the phases present in bulk HfO2. Despite a phase constitution of ∼17% Pca21, no ferroelectric properties could be measured.

Ferroelectric hysteresis via PUND measurement was only recently reported by Xu et al. who used a laser-diode-heated floating zone (LDFZ) to grow a 12 mol. % Y-doped HfO2 single crystal.148 The LDFZ furnace was able to reach a maximum temperature of 3000 °C and achieve a high quenching rate which was claimed as key to obtaining the metastable Pca21 phase in bulk Y-doped HfO2. Xu et al. crystallized the single crystal from a molten liquid above 2700 °C and then supercooled the material which resulted in a supersaturated state. The fast quench rate was integral to constraining yttrium diffusion and chemical pressure of a supersaturated state drove the formation of the orthorhombic phases. Xu et al. conducted density functional theory calculations and showed that the energy landscape and pathways to the ferroelectric orthorhombic phase in bulk HfO2. The energy barrier from the tetragonal to the ferroelectric orthorhombic phase was calculated to be the lowest, especially with the lattice constants were fixed, further suggesting the importance of a fast quench rate. While remanent polarization values only reached 3 μC/cm2, ferroelectricity in bulk HfO2 indicated no wake-up effect and abundant 90°/180° ferroelectric domains which shows promise for future device applications.

This review describes the working principles of four main deposition methods that have been deployed in the past 10 years to successfully synthesize ferroelectric HfO2 films. Our analysis reveals that process windows yielding maximum ferroelectric polarization can be significantly different across different deposition techniques. In ALD- and sputter deposition-based films, for example, a low oxygen concentration is typically preferred for ferroelectric films, while for PLD films, higher oxygen pressures during deposition may yield better ferroelectric properties. In sputter-deposited films, lower doping concentrations are required to achieve the same level of robust ferroelectricity compared to ALD and CSD films. The dimensional limitations of ferroelectricity for HfO2 may also vary on the deposition technique, where PLD films can achieve ferroelectricity in both ultrathin and >900 nm regimes, whereas ferroelectricity is typically limited to film thicknesses in the tens of nanometers for ALD- and PLD-grown films.

The motivation for using any given deposition technique is ultimately an act of balancing its strengths and weaknesses. We detail the major advantages and limitations of each technique in this report and summarize these findings in Fig. 21. It is noted that a key advantage of ALD is the ability to exhibit monolayer control over film growth, leading to novel film architectures such as nanolaminates, bilayers, trilayers, and seed layers. However, a significant drawback of ALD is the residual carbon contamination stemming from the use of metalorganic precursors. Sputter-deposited films, on the other hand, are derived from ceramic targets and have very little carbon contamination and can be deposited at room temperature. Sputter deposition is a line-of-sight technique, however, which makes adaptation to three-dimensional substrates less practical. Another major concern in sputter-deposited films is repeatability due to target aging and oxygen content variety across differently sourced target materials. PLD, in contrast, offers the alluring advantage of phase-pure epitaxial films without the convoluting effects of grain boundaries, defects, or mixed phases, and can thus be leveraged to draw key insight into the mechanisms of ferroelectricity in HfO2. The nonthermal nature of PLD offers researchers access to metastable phases and multicomponent compositions, which can result in the discovery of new material systems. However, the complexity of PLD mechanisms and data rendered from near-perfect films may make it difficult to generalize findings across different research groups. CSD, in contrast, is one of the most cost-effective techniques due to its high yield and use of inexpensive equipment. CSD has been used to create thicker ferroelectric films with a wide range of dopants, although the control of film thickness may be challenging. Finally, we briefly discussed recent progress in growing bulk HfO2 ceramics using laser heated float zone crystallization and co-precipitation techniques. In summary, we reiterate that no deposition method is superior to the other, but the selection of a deposition method and its process parameters must be determined by practical considerations and the desired end-use application.

FIG. 21.

Benefits and drawbacks of each deposition method surveyed in this report for ferroelectric HfO2.

FIG. 21.

Benefits and drawbacks of each deposition method surveyed in this report for ferroelectric HfO2.

Close modal

This contribution is based upon the work supported by the National Science Foundation (NSF), as part of the Center for Dielectrics and Piezoelectrics under Grant Nos. IIP-1841453 and IIP-1841466. H.A.H. was supported by the National Science Foundation Graduate Research Fellowship Program (No. DGE-1746939) INTERN supplemental grant and the IIE-GIRE Program funded by the National Science Foundation (No. 1829436). A.P. acknowledges support from a fellowship at the US CCDC Army Research Laboratory, administered by Oak Ridge Associated Universities (ORAU) through the U.S. Department of Energy Oak Ridge Institute for Science and Education and the National Science Foundation under Award No. CMMI-1634955.

The data that support the findings of this study are available from the corresponding author upon reasonable request.

1.
S.
Natarajan
 et al,
International Electron Devices Meeting, 2008, San Francisco, CA
, (
IEEE International
,
New York
,
2009
), p.
3
.
2.
T. S.
Böscke
,
J.
Müller
,
D.
Bräuhaus
,
U.
Schröder
, and
U.
Böttger
,
Appl. Phys. Lett.
99
,
102903
(
2011
).
3.
C. H.
Yu
,
2006 International Microsystems, Package, Assembly Conference Taiwan
,
Taipei
, 18–20 October 2006 (
IEEE
,
New York
,
2006
), pp.
1
6
.
4.
J. F.
Ihlefeld
,
D. T.
Harris
,
R.
Keech
,
J. L.
Jones
,
J.-P.
Maria
, and
S.
Trolier-McKinstry
,
J. Am. Ceram. Soc.
99
,
2537
(
2016
).
5.
S. J.
Kim
 et al,
Appl. Phys. Lett.
112
,
172902
(
2018
).
6.
S.
Mueller
,
S. R.
Summerfelt
,
J.
Muller
,
U.
Schroeder
, and
T.
Mikolajick
,
IEEE Electron Device Lett.
33
,
1300
(
2012
).
7.
J.
Okuno
 et al,
2020 IEEE Symposium on VLSI Technology
,
Honolulu, HI, USA
, 16–19 June 2020 (
IEEE
,
New York
,
2020
), pp.
1
2
.
8.
H.
Mulaosmanovic
,
E. T.
Breyer
,
S.
Dünkel
,
S.
Beyer
,
T.
Mikolajick
, and
S.
Slesazeck
,
Nanotechnology
32
,
502002
(
2021
).
9.
J.
Muller
 et al,
2013 IEEE International Electron Devices Meeting
,
Washington, DC, USA
, 9–11 December 2013 (
IEEE
,
New York
,
2013
), pp.
10.8.1
-
10.8.4
.
10.
M. H.
Park
,
H. J.
Kim
,
Y. J.
Kim
,
T.
Moon
,
K. D.
Kim
, and
C. S.
Hwang
,
Nano Energy
12
,
131
(
2015
).
11.
M. H.
Park
,
H. J.
Kim
,
Y. J.
Kim
,
T.
Moon
,
K. D.
Kim
,
Y. H.
Lee
,
S. D.
Hyun
, and
C. S.
Hwang
,
Adv. Mater.
28
,
7956
(
2016
).
12.
M. H.
Park
,
H. J.
Kim
,
Y. J.
Kim
,
T.
Moon
,
K. D.
Kim
, and
C. S.
Hwang
,
Adv. Energy Mater.
4
,
1400610
(
2014
).
13.
H.
Ryu
,
H.
Wu
,
F.
Rao
, and
W.
Zhu
,
Sci. Rep.
9
,
020383
(
2019
).
14.
J.
Müller
,
T. S.
Böscke
,
U.
Schröder
,
S.
Mueller
,
D.
Bräuhaus
,
U.
Böttger
,
L.
Frey
, and
T.
Mikolajick
,
Nano Lett.
12
,
4318
(
2012
).
15.
J. F.
Ihlefeld
 et al,
J. Appl. Phys.
128
,
034101
(
2020
).
16.
A.
Chernikova
 et al,
ACS Appl. Mater. Interfaces
8
,
7232
(
2016
).
17.
M. C.
Cheynet
,
S.
Pokrant
,
F. D.
Tichelaar
, and
J.-L.
Rouvière
,
J. Appl. Phys.
101
,
054101
(
2007
).
18.
C.
Gaumer
 et al,
Microelectron. Eng.
88
,
72
(
2011
).
19.
M. H.
Park
,
Y. H.
Lee
,
T.
Mikolajick
,
U.
Schroeder
, and
C. S.
Hwang
,
MRS Commun.
8
,
795
(
2018
).
20.
S. J.
Kim
,
J.
Mohan
,
S. R.
Summerfelt
, and
J.
Kim
,
JOM
71
,
246
(
2019
).
21.
M. H.
Park
 et al,
J. Mater. Chem. C
8
,
010526
(
2020
).
22.
M.
Materano
,
P. D.
Lomenzo
,
A.
Kersch
,
M. H.
Park
,
T.
Mikolajick
, and
U.
Schroeder
,
Inorg. Chem. Front.
8
,
2650
(
2021
).
23.
A.
Chouprik
,
D.
Negrov
,
E. Y.
Tsymbal
, and
A.
Zenkevich
,
Nanoscale
13
,
011635
(
2021
).
24.
T. D.
Huan
,
V.
Sharma
,
G. A.
Rossetti
, and
R.
Ramprasad
,
Phys. Rev. B
90
,
064111
(
2014
).
25.
R.
Batra
,
H. D.
Tran
, and
R.
Ramprasad
,
Appl. Phys. Lett.
108
,
172902
(
2016
).
26.
X.
Sang
,
E. D.
Grimley
,
T.
Schenk
,
U.
Schroeder
, and
J. M.
LeBeau
,
Appl. Phys. Lett.
106
,
162905
(
2015
).
27.
Y.
Wei
 et al,
Nat. Mater
17
,
1095
(
2018
).
28.
M. H.
Park
 et al,
Adv. Mater.
27
,
1811
(
2015
).
29.
R.
Materlik
,
C.
Künneth
, and
A.
Kersch
,
J. Appl. Phys.
117
,
134109
(
2015
).
30.
M.
Hoffmann
 et al,
J. Appl. Phys.
118
,
072006
(
2015
).
31.
S.
Starschich
,
T.
Schenk
,
U.
Schroeder
, and
U.
Boettger
,
Appl. Phys. Lett.
110
,
182905
(
2017
).
32.
P. D.
Lomenzo
,
Q.
Takmeel
,
S.
Moghaddam
, and
T.
Nishida
,
Thin Solid Films
615
,
139
(
2016
).
33.
J.
Müller
 et al,
J. Appl. Phys.
110
,
114113
(
2011
).
34.
T.
Schenk
,
S.
Mueller
,
U.
Schroeder
,
R.
Materlik
,
A.
Kersch
,
M.
Popovici
,
C.
Adelmann
,
S.
Van Elshocht
, and
T.
Mikolajick
,
2013 Proceedings of the European Solid-State Device Research Conference (ESSDERC)
,
Bucharest
, 16–20 September 2013 (
IEEE
,
New York
,
2013
), pp.
260
263
.
35.
U.
Schroeder
 et al,
Inorg. Chem.
57
,
2752
(
2018
).
36.
S.
Mueller
,
C.
Adelmann
,
A.
Singh
,
S.
Van Elshocht
,
U.
Schroeder
, and
T.
Mikolajick
,
ECS J. Solid State Sci. Technol.
1
,
N123
(
2012
).
37.
S.
Mueller
,
J.
Mueller
,
A.
Singh
,
S.
Riedel
,
J.
Sundqvist
,
U.
Schroeder
, and
T.
Mikolajick
,
Adv. Funct. Mater.
22
,
2412
(
2012
).
38.
L.
Xu
,
T.
Nishimura
,
S.
Shibayama
,
T.
Yajima
,
S.
Migita
, and
A.
Toriumi
,
Appl. Phys. Express
9
,
091501
(
2016
).
39.
J. E.
Lowther
,
J. K.
Dewhurst
,
J. M.
Leger
, and
J.
Haines
,
Phys. Rev. B
60
,
014485
(
1999
).
40.
E. H.
Kisi
and
C. J.
Howard
,
Zirconia Engineering Ceramics
(
Trans Tech Publications Ltd
, Kapellweg 8, 8806 Bäch SZ, Switzerland,
1998
), pp.
1
36
.
41.
R.
Batra
,
T. D.
Huan
,
G. A.
Rossetti
, and
R.
Ramprasad
,
Chem. Mater.
29
,
9102
(
2017
).
42.
L.
Xu
,
T.
Nishimura
,
S.
Shibayama
,
T.
Yajima
,
S.
Migita
, and
A.
Toriumi
,
J. Appl. Phys.
122
,
124104
(
2017
).
43.
M. G.
Kim
and
S.
Ohmi
,
Jpn. J. Appl. Phys.
57
,
11UF09
(
2018
).
44.
K. D.
Kim
,
M. H.
Park
,
H. J.
Kim
,
Y. J.
Kim
,
T.
Moon
,
Y. H.
Lee
,
S. D.
Hyun
,
T.
Gwon
, and
C. S.
Hwang
,
J. Mater. Chem. C
4
,
6864
(
2016
).
45.
M.
Hyuk Park
,
H.
Joon Kim
,
Y.
Jin Kim
,
W.
Lee
,
T.
Moon
, and
C.
Seong Hwang
,
Appl. Phys. Lett.
102
,
242905
(
2013
).
46.
M. H.
Park
 et al,
Nanoscale
9
,
9973
(
2017
).
47.
M.
Pešić
 et al,
Adv. Funct. Mater.
26
,
4601
(
2016
).
48.
M. H.
Park
,
Y. H.
Lee
, and
C. S.
Hwang
,
Nanoscale
11
,
019477
(
2019
).
49.
M. H.
Park
 et al,
Nanoscale
10
,
716
(
2018
).
50.
M. H.
Park
 et al,
Adv. Electron. Mater.
4
,
1800091
(
2018
).
51.
T.
Shiraishi
 et al,
Appl. Phys. Lett.
108
,
262904
(
2016
).
52.
Y.
Tashiro
,
T.
Shimizu
,
T.
Mimura
, and
H.
Funakubo
,
ACS Appl. Electron. Mater.
3
,
3123
(
2021
).
53.
H. A.
Hsain
,
Y.
Lee
,
G.
Parsons
, and
J. L.
Jones
,
Appl. Phys. Lett.
116
,
192901
(
2020
).
54.
A.
Toriumi
,
L.
Xu
,
Y.
Mori
,
X.
Tian
,
P. D.
Lomenzo
,
H.
Mulaosmanovic
,
M.
Materano
,
T.
Mikolajick
, and
U.
Schroeder
,
2019 IEEE International Electron Devices Meeting (IEDM)
,
San Francisco, CA, USA
, 7–11 December 2019 (
IEEE
,
New York
,
2019
), pp.
15.1.1
-
15.1.4
.
55.
S. M.
George
,
Chem. Rev.
110
,
111
(
2010
).
56.
G. N.
Parsons
,
S. M.
George
, and
M.
Knez
,
MRS Bull.
36
,
865
(
2011
).
57.
R. W.
Johnson
,
A.
Hultqvist
, and
S. F.
Bent
,
Mater. Today
17
,
236
(
2014
).
58.
K.
Kukli
,
M.
Ritala
,
M.
Leskelä
,
T.
Sajavaara
,
J.
Keinonen
,
A. C.
Jones
, and
J. L.
Roberts
,
Chem. Mater.
15
,
1722
(
2003
).
59.
K.
Kukli
,
T.
Pilvi
,
M.
Ritala
,
T.
Sajavaara
,
J.
Lu
, and
M.
Leskelä
,
Thin Solid Films
491
,
328
(
2005
).
60.
M.
Materano
,
C.
Richter
,
T.
Mikolajick
, and
U.
Schroeder
,
J. Vac. Sci. Technol. A
38
,
022402
(
2020
).
61.
C. S.
Hwang
,
Atomic Layer Deposition for Semiconductors
, 1st ed. (
Springer
,
New York
,
2014
).
62.
J.
Aarik
,
A.
Aidla
,
A.
Kikas
,
T.
Käämbre
,
R.
Rammula
,
P.
Ritslaid
,
T.
Uustare
, and
V.
Sammelselg
,
Appl. Surf. Sci.
230
,
292
(
2004
).
63.
H.
Machida
,
T.
Kada
,
M.
Ishikawa
,
A.
Ogura
, and
Y.
Ohshita
,
Jpn. J. Appl. Phys.
43
,
966
(
2004
).
64.
S. W.
Smith
,
A. R.
Kitahara
,
M. A.
Rodriguez
,
M. D.
Henry
,
M. T.
Brumbach
, and
J. F.
Ihlefeld
,
Appl. Phys. Lett.
110
,
072901
(
2017
).
65.
S. J.
Kim
 et al,
Appl. Phys. Lett.
115
,
182901
(
2019
).
66.
T.
Suntola
,
Appl. Surf. Sci.
100–101
,
391
(
1996
).
67.
B. S.
Kim
 et al,
Nanoscale Res. Lett.
15
,
72
(
2020
).
68.
J.
Niinistö
,
M.
Putkonen
,
L.
Niinistö
,
F.
Song
,
P.
Williams
,
P. N.
Heys
, and
R.
Odedra
,
Chem. Mater.
19
,
3319
(
2007
).
69.
J.
Swerts
,
N.
Peys
,
L.
Nyns
,
A.
Delabie
,
A.
Franquet
,
J. W.
Maes
,
S. V.
Elshocht
, and
S. D.
Gendt
,
J. Electrochem. Soc.
6
, G26 (2010).
70.
R. S.
Pessoa
,
M. A.
Fraga
,
W.
Chiappim
, and
H. S.
Maciel
,
Emerging Materials for Energy Conversion and Storage
(
Elsevier
,
New York
,
2018
), pp.
83
114
.
71.
H. B.
Park
 et al,
J. Appl. Phys.
94
,
3641
(
2003
).
72.
X.
Liu
,
S.
Ramanathan
,
A.
Longdergan
,
A.
Srivastava
,
E.
Lee
,
T. E.
Seidel
,
J. T.
Barton
,
D.
Pang
, and
R. G.
Gordon
,
J. Electrochem. Soc.
152
,
G213
(
2005
).
73.
R.
Alcala
,
C.
Richter
,
M.
Materano
,
P. D.
Lomenzo
,
C.
Zhou
,
J. L.
Jones
,
T.
Mikolajick
, and
U.
Schroeder
,
J. Phys. D: Appl. Phys.
54
,
035102
(
2021
).
74.
S. Y.
Lee
,
H. K.
Kim
,
J. H.
Lee
,
I.-H.
Yu
,
J.-H.
Lee
, and
C. S.
Hwang
,
J. Mater. Chem. C
2
,
2558
(
2014
).
75.
M.-J.
Choi
,
H.-H.
Park
,
D. S.
Jeong
,
J. H.
Kim
,
J.-S.
Kim
, and
S. K.
Kim
,
Appl. Surf. Sci.
301
,
451
(
2014
).
76.
J.
Park
,
M.
Cho
,
S. K.
Kim
,
T. J.
Park
,
S. W.
Lee
,
S. H.
Hong
, and
C. S.
Hwang
,
Appl. Phys. Lett.
86
,
112907
(
2005
).
77.
S.-J.
Yoon
,
S.-Y.
Na
,
S.-E.
Moon
, and
S.-M.
Yoon
,
J. Vac. Sci. Technol. B
37
,
050601
(
2019
).
78.
A.
Pal
,
V. K.
Narasimhan
,
S.
Weeks
,
K.
Littau
,
D.
Pramanik
, and
T.
Chiang
,
Appl. Phys. Lett.
110
,
022903
(
2017
).
79.
U.
Schroeder
,
M.
Materano
,
T.
Mittmann
,
P. D.
Lomenzo
,
T.
Mikolajick
, and
A.
Toriumi
,
Jpn. J. Appl. Phys.
58
,
SL0801
(
2019
).
80.
M.
Materano
,
T.
Mittmann
,
P. D.
Lomenzo
,
C.
Zhou
,
J. L.
Jones
,
M.
Falkowski
,
A.
Kersch
,
T.
Mikolajick
, and
U.
Schroeder
,
ACS Appl. Electron. Mater.
2
,
3618
(
2020
).
81.
T.
Mittmann
,
M.
Materano
,
S.-C.
Chang
,
I.
Karpov
,
T.
Mikolajick
, and
U.
Schroeder
,
2020 IEEE International Electron Devices Meeting (IEDM)
,
San Francisco, CA, USA
, 12–18 December 2020 (
IEEE
,
New York
,
2020
), pp.
18.4.1
18.4.4
.
82.
T.
Mittmann
 et al,
Phys. Status Solidi RRL
15
,
2100012
(
2021
).
83.
S. S.
Fields
 et al,
Adv. Mater. Interfaces
8
,
2100018
(
2021
).
84.
D.
Cao
,
X.
Cheng
,
L.
Zheng
,
D.
Xu
,
Z.
Wang
,
C.
Xia
,
L.
Shen
,
Y.
Yu
, and
D.
Shen
,
J. Vac. Sci. Technol. B
33
,
01A101
(
2015
).
85.
H.
Hernández-Arriaga
,
E.
López-Luna
,
E.
Martínez-Guerra
,
M. M.
Turrubiartes
,
A. G.
Rodríguez
, and
M. A.
Vidal
,
J. Appl. Phys.
121
,
064302
(
2017
).
86.
M. H.
Park
 et al,
Appl. Phys. Rev.
6
,
041403
(
2019
).
87.
S. L.
Weeks
,
A.
Pal
,
V. K.
Narasimhan
,
K. A.
Littau
, and
T.
Chiang
,
ACS Appl. Mater. Interfaces
9
,
013440
(
2017
).
88.
M. E.
McBriarty
 et al,
Phys. Status Solidi B
257
,
1900285
(
2020
).
89.
Y. W.
Lu
,
J.
Shieh
, and
F. Y.
Tsai
,
Acta Mater.
115
,
68
(
2016
).
90.
V.
Gaddam
,
D.
Das
, and
S.
Jeon
,
IEEE Trans. Electron Devices
67
,
745
(
2020
).
91.
T.
Onaya
,
T.
Nabatame
,
N.
Sawamoto
,
A.
Ohi
,
N.
Ikeda
,
T.
Nagata
, and
A.
Ogura
,
APL Mater.
7
,
061107
(
2019
).
92.
H. J.
Kim
,
M. H.
Park
,
Y. J.
Kim
,
Y. H.
Lee
,
W.
Jeon
,
T.
Gwon
,
T.
Moon
,
K. D.
Kim
, and
C. S.
Hwang
,
Appl. Phys. Lett.
105
,
192903
(
2014
).
93.
J.
Wang
 et al,
IEEE Electron Device Lett.
40
,
1937
(
2019
).
94.
P. D.
Lomenzo
,
Q.
Takmeel
,
C.
Zhou
,
Y.
Liu
,
C. M.
Fancher
,
J. L.
Jones
,
S.
Moghaddam
, and
T.
Nishida
,
Appl. Phys. Lett.
105
,
072906
(
2014
).
95.
C.
Mart
,
K.
Kühnel
,
T.
Kämpfe
,
M.
Czernohorsky
,
M.
Wiatr
,
S.
Kolodinski
, and
W.
Weinreich
,
ACS Appl. Electron. Mater.
1
,
2612
(
2019
).
96.
M.
Hyuk Park
,
H.
Joon Kim
,
Y.
Jin Kim
,
T.
Moon
, and
C.
Seong Hwang
,
Appl. Phys. Lett.
104
,
072901
(
2014
).
97.
T. M.
Correia
and
Q.
Zhang
,
Ferroelectrics
405
,
275
(
2010
).
98.
L.
Liu
,
R.
Zuo
, and
Q.
Liang
,
Ceram. Int.
39
,
3865
(
2013
).
99.
R.
Laishram
,
J. Phys. Chem. Solids
104
,
8
(
2017
).
100.
S.
Bonomi
and
L.
Malavasi
,
J. Vac. Sci. Technol. A
38
,
060803
(
2020
).
101.
Z.
Fan
,
J.
Xiao
,
J.
Wang
,
L.
Zhang
,
J.
Deng
,
Z.
Liu
,
Z.
Dong
,
J.
Wang
, and
J.
Chen
,
Appl. Phys. Lett.
108
,
232905
(
2016
).
102.
Q.
Luo
 et al,
IEEE Electron Device Lett.
40
,
570
(
2019
).
103.
J.
Bouaziz
,
P.
Rojo Romeo
,
N.
Baboux
, and
B.
Vilquin
,
J. Vac. Sci. Technol. B
37
,
021203
(
2019
).
104.
K. C.
Das
,
N.
Tripathy
,
S. P.
Ghosh
,
P.
Sharma
,
R.
Singhal
, and
J. P.
Kar
,
Vacuum
143
,
288
(
2017
).
105.
Y. H.
Lee
,
H. J.
Kim
,
T.
Moon
,
K. D.
Kim
,
S. D.
Hyun
,
H. W.
Park
,
Y. B.
Lee
,
M. H.
Park
, and
C. S.
Hwang
,
Nanotechnology
28
,
305703
(
2017
).
106.
T.
Olsen
,
U.
Schröder
,
S.
Müller
,
A.
Krause
,
D.
Martin
,
A.
Singh
,
J.
Müller
,
M.
Geidel
, and
T.
Mikolajick
,
Appl. Phys. Lett.
101
,
082905
(
2012
).
107.
P.
Zhao
,
D.
Zhou
,
F.
Liu
,
N.
Sun
, and
S.
Li
,
Mater. Res. Express
6
,
086325
(
2019
).
108.
Y.
Zhang
,
J.
Xu
,
D.-Y.
Zhou
,
H.-H.
Wang
,
W.-Q.
Lu
, and
C.-K.
Choi
,
Ceram. Int.
44
,
012841
(
2018
).
109.
T.
Suzuki
,
T.
Shimizu
,
T.
Mimura
,
H.
Uchida
, and
H.
Funakubo
,
Jpn. J. Appl. Phys.
57
,
11UF15
(
2018
).
110.
T.
Shiraishi
,
S.
Choi
,
T.
Kiguchi
,
T.
Shimizu
,
H.
Uchida
,
H.
Funakubo
, and
T. J.
Konno
,
Jpn. J. Appl. Phys.
57
,
11UF02
(
2018
).
111.
T.
Mittmann
 et al,
Adv. Mater. Interfaces
6
,
1900042
(
2019
).
112.
F.
Huang
 et al,
Phys. Chem. Chem. Phys.
19
,
3486
(
2017
).
113.
H.
Yamada
,
Y.
Toyosaki
, and
A.
Sawa
,
J. Appl. Phys.
124
,
105305
(
2018
).
114.
T.
Murakami
,
K. I.
Haga
, and
E.
Tokumitsu
,
Jpn. J. Appl. Phys.
59
,
SPPB03
(
2020
).
115.
S.
Kim
and
J.
Hong
,
Ceram. Int.
43
,
S158
(
2017
).
116.
L.
Baumgarten
,
T.
Szyjka
,
T.
Mittmann
,
M.
Materano
,
Y.
Matveyev
,
C.
Schlueter
,
T.
Mikolajick
,
U.
Schroeder
, and
M.
Müller
,
Appl. Phys. Lett.
118
,
032903
(
2021
).
117.
H.
Liang
,
B.
Zhang
,
Y.
Guo
,
X.
Guo
,
S.
Ren
,
Y.
Li
,
Y.
Lu
, and
R.
Lang
,
Ceram. Int.
47
, 27843 (
2021
).
118.
L.
Xu
,
S.
Shibayama
,
K.
Izukashi
,
T.
Nishimura
,
T.
Yajima
,
S.
Migita
, and
A.
Toriumi
,
2016 IEEE International Electron Devices Meeting (IEDM)
,
San Francisco, CA, USA
, 3–7 December 2016 (
IEEE
,
New York
,
2016
), pp.
25.2.1
-
25.2.4
.
119.
D.
Severin
,
O.
Kappertz
,
T.
Nyberg
,
S.
Berg
,
A.
Pflug
, and
M.
Wuttig
,
J. Appl. Phys.
105
,
093302
(
2009
).
120.
T.
Mittmann
,
M.
Michailow
,
P. D.
Lomenzo
,
J.
Gärtner
,
M.
Falkowski
,
A.
Kersch
,
T.
Mikolajick
, and
U.
Schroeder
,
Nanoscale
13
,
912
(
2021
).
121.
T.
Szyjka
,
L.
Baumgarten
,
T.
Mittmann
,
Y.
Matveyev
,
C.
Schlueter
,
T.
Mikolajick
,
U.
Schroeder
, and
M.
Müller
,
ACS Appl. Electron. Mater.
2
,
3152
(
2020
).
122.
C. C.
Chin
,
R. J.
Lin
,
Y. C.
Yu
,
C. W.
Wang
,
E. K.
Lin
,
W. C.
Tsai
, and
T. Y.
Tseng
,
Physica C
260
,
86
(
1996
).
123.
J.
Hinze
and
K.
Ellmer
,
J. Appl. Phys.
88
,
2443
(
2000
).
124.
V. G.
Koukhar
,
N. A.
Pertsev
, and
R.
Waser
,
Phys. Rev. B
64
,
214103
(
2001
).
125.
Y.
Lin
and
C. L.
Chen
,
J. Mater. Sci.
44
,
5274
(
2009
).
126.
C. H.
Ahn
,
K. M.
Rabe
, and
J. M.
Triscone
,
Science
303
,
488
(
2004
).
127.
P. R.
Willmott
and
J. R.
Huber
,
Rev. Mod. Phys.
72
,
315
(
2000
).
128.
T.
Shimizu
,
K.
Katayama
,
T.
Kiguchi
,
A.
Akama
,
T. J.
Konno
, and
H.
Funakubo
,
Appl. Phys. Lett.
107
,
032910
(
2015
).
129.
K.
Katayama
,
Appl. Phys. Lett.
6
, 112901 (
2016
).
130.
J.
Lyu
,
I.
Fina
,
R.
Bachelet
,
G.
Saint-Girons
,
S.
Estandía
,
J.
Gázquez
,
J.
Fontcuberta
, and
F.
Sánchez
,
Appl. Phys. Lett.
114
,
222901
(
2019
).
131.
Y.
Zhang
,
Q.
Yang
,
L.
Tao
,
E.
Tsymbal
, and
V.
Alexandrov
,
Phys. Rev. Appl.
14
,
014068
(
2020
).
132.
S.
Estandía
,
N.
Dix
,
J.
Gazquez
,
I.
Fina
,
J.
Lyu
,
M. F.
Chisholm
,
J.
Fontcuberta
, and
F.
Sánchez
,
ACS Appl. Electron. Mater.
1
,
1449
(
2019
).
133.
T.
Li
 et al,
J. Mater. Chem. C
6
,
9224
(
2018
).
134.
T.
Mimura
,
T.
Shimizu
, and
H.
Funakubo
,
Appl. Phys. Lett.
115
,
032901
(
2019
).
135.
J.
Lyu
,
I.
Fina
,
R.
Solanas
,
J.
Fontcuberta
, and
F.
Sánchez
,
CS Appl. Electron. Mater.
1
,
27
(2019).
136.
Z.
Fan
,
J.
Chen
, and
J.
Wang
,
J. Adv. Dielect.
06
,
1630003
(
2016
).
137.
T.
Mimura
,
T.
Shimizu
,
T.
Kiguchi
,
A.
Akama
,
T. J.
Konno
,
Y.
Katsuya
,
O.
Sakata
, and
H.
Funakubo
,
Jpn. J. Appl. Phys.
58
,
SBBB09
(
2019
).
138.
S.
Starschich
,
D.
Griesche
,
T.
Schneller
,
R.
Waser
, and
U.
Böttger
,
Appl. Phys. Lett.
104
,
202903
(
2014
).
139.
S.
Starschich
,
D.
Griesche
,
T.
Schneller
, and
U.
Böttger
,
ECS J. Solid State Sci. Technol.
4
,
P419
(
2015
).
140.
T.
Schenk
,
N.
Godard
,
A.
Mahjoub
,
S.
Girod
,
A.
Matavz
,
V.
Bobnar
,
E.
Defay
, and
S.
Glinsek
,
Phys. Status Solidi RRL
14
,
1900626
(
2020
).
141.
S.
Samanta
 et al,
Adv. Mater. Interfaces
8
,
2100907
(
2021
).
142.
S.
Zheng
,
Z.
Zhao
,
Z.
Liu
,
B.
Zeng
,
L.
Yin
,
Q.
Peng
,
M.
Liao
, and
Y.
Zhou
,
Appl. Phys. Lett.
117
,
212904
(
2020
).
143.
J.
Wang
,
D.
Zhou
,
W.
Dong
,
X.
Hou
,
F.
Liu
,
N.
Sun
,
F.
Ali
, and
Z.
Li
,
Ceram. Int.
47
,
016845
(
2021
).
144.
R. M.
Pasquarelli
,
D. S.
Ginley
, and
R.
O’Hayre
,
ChemInform
43
, 5406 (
2012
).
145.
H.
Chen
,
Y.
Chen
,
L.
Tang
,
H.
Luo
,
K.
Zhou
,
X.
Yuan
, and
D.
Zhang
,
J. Mater. Chem. C
8
,
2820
(
2020
).
146.
S.
Nakayama
,
H.
Funakubo
, and
H.
Uchida
,
Jpn. J. Appl. Phys.
57
,
11UF06
(
2018
).
147.
D.
Banerjee
,
R.
Sewak
,
C. C.
Dey
,
D.
Toprek
, and
P. K.
Pujari
,
Mater. Today Commun.
26
,
101827
(
2021
).
148.
X.
Xu
,
F.-T.
Huang
,
Y.
Qi
,
S.
Singh
,
K. M.
Rabe
,
D.
Obeysekera
,
J.
Yang
,
M.-W.
Chu
, and
S.-W.
Cheong
,
Nat. Mater.
20
,
826
(
2021
).
149.
J.
Bouaziz
,
P.
Rojo Romeo
,
N.
Baboux
,
R.
Negrea
,
L.
Pintilie
, and
B.
Vilquin
,
APL Mater.
7
,
081109
(
2019
).