Polymorphic (HfxZr1−x)O2 (HZO) thin films exhibit ferroelectric, dielectric, and antiferroelectric properties across a wide compositional range due to the existence of orthorhombic, monoclinic, and tetragonal phases. To better understand the phase stability across the HfO2–ZrO2 compositional range, we investigate the structural evolution of HZO thin films in situ via high-temperature x-ray diffraction (HTXRD) for five different compositions [ZrO2, (Hf0.23Zr0.77)O2, (Hf0.43Zr0.57)O2, (Hf0.67Zr0.33)O2, and HfO2]. The real-time monitoring of HZO crystallization reveals a competing driving force between the tetragonal and monoclinic phase stabilities for HfO2-rich vs ZrO2-rich compositions. Additionally, we confirm an XRD peak shift toward lower 2θ with increasing temperature in ZrO2, (Hf0.23Zr0.77)O2, and (Hf0.43Zr0.57)O2 films, which we ascribe to the appearance of a metastable orthorhombic phase during heating. A monotonic trend for the onset crystallization temperature is reported for five compositions of HZO and reveals an increase in onset crystallization temperature for HfO2-rich compositions. Relative intensity fraction calculations suggest a higher fraction of monoclinic phase with increasing annealing temperature for (Hf0.67Zr0.33)O2. This study of phase stability and onset crystallization temperatures offers insight for managing the thermal budget for HZO thin films, especially for temperature-constrained processing.
As a promising lead-free alternative to traditional ferroelectric (FE) materials, HfO2-based ferroelectrics display high polarization densities in ultrathin films (>15 C/cm2), large bandgap energies (>5 eV), 10-year memory retention, and compatibility with complementary metal–oxide–semiconductor (CMOS) processing and devices.1–3 HfO2-based materials have been applied for nonvolatile memory,4 ferroelectric field effect transistors,5 pyroelectric energy harvesters,6 electrocaloric coolers,7 and supercapacitor energy harvesters.8
The origin of ferroelectricity in HfO2 is attributed to the formation of a polar orthorhombic phase (Pca21).9,10 While monoclinic (P21/c) is the most stable phase in bulk HfO2 at ambient conditions, films grown via atomic layer deposition (ALD) at thicknesses below 30 nm show the presence of the tetragonal (P42/nmc) and polar orthorhombic phases after annealing treatment.3,11 Some driving factors for orthorhombic phase stabilization have been suggested, including anisotropic stress from capping layer confinement,12 thermal expansion mismatch,13 surface energy and grain size effects,14,15 dopants, and oxygen vacancies.16,17
(HfxZr1−x)O2 (herein called HZO) has been recently highlighted due to its wide landscape of functionalities such as of being dielectric, ferroelectric, and anti-ferroelectric, which can be fine-tuned by controlling the chemical composition.13,14,18 It is known that ZrO2-rich compositions in HZO thin films can crystallize and form the polar orthorhombic phase at lower temperatures than other HfO2 dopant systems.17–19 Yu et al. successfully demonstrated ferroelectric HZO thin films on flexible polyimide substrates, where the processing temperature was held at ≤ 430 °C to avoid polymer degradation.20 In a subsequent study, HZO films grown on polyimide were also shown to have an enhanced piezoelectric response compared to conventional rigid substrates.21 Not only do low crystallization temperatures of HZO films enable new applications such as flexible devices, but also lower processing temperatures can lead to an increase in the ferroelectric properties due to suppression of the high-temperature monoclinic phase.11,18,20
Recognizing the lack of a systematic study of HZO crystallization and phase evolution across the entire solid solution, we present a study of the crystallization behavior and phase stabilization in ALD-grown HZO films across the entire composition range of the HfO2–ZrO2 solid solution. The goal of this work is to report the in situ crystallization behavior across the HZO compositional landscape and aid in the design of temperature-constrained applications. In turn, we develop a more fundamental understanding of the processing-structure relationships in HZO thin films for ferroelectric and related applications. We report a monotonic compositional dependence of the HfO2 concentration on crystallization temperature in the HfO2–ZrO2 solid solution system. Further, for ZrO2-rich compositions, we confirm XRD peak shifts toward lower 2θ during heating. The non-linear shift in 2θ occurs between 700 and 800 °C and is ascribed to a phase transformation from the tetragonal phase to the orthorhombic phase. We reveal the phase evolution of the HfO2–ZrO2 system as a function of composition using an in situ high-temperature x-ray diffraction (HTXRD) technique.
HZO films were deposited via ALD (Ultratech Fiji G2) with both top and bottom sputtered TiN layers on undoped Si substrates. Capacitor stacks were grown in the conventional Metal-Insulator-Metal (MIM) architecture: TiN (30 nm)/HZO (30 nm)/TiN (30 nm)/Si substrate. The bottom TiN layer was sputtered onto an undoped Si wafer with a native oxide layer purchased from VA Semiconductors. Si wafers were first cleaned in an isopropyl ultrasonic bath for 15 min before RF sputtering. Afterwards, TiN sputter deposition (Kurt J. Lesker, PVD) was conducted for 25 minutes at room temperature to obtain a 30 nm amorphous TiN film, which was confirmed via ellipsometry and x-ray diffraction. After HZO deposition, a top TiN layer was also RF sputter-deposited to suppress the formation of the non-ferroelectric phases in HZO films through confinement.2,22
Immediately following the bottom TiN layer deposition, the wafer was placed into an Ultratech Cambridge Fiji G2 ALD chamber stabilized at 270 °C for HZO film deposition. Tetrakis(dimethylamino) zirconium (TDMAZ) and tetrakis(dimethylamido)hafnium (TDMAH) were used as precursors for ZrO2 and HfO2, respectively, with H2O as an oxygen source. TDMAZ and TDMAH precursors were heated to 75 °C for vapor delivery and delivered to the reactor using Ar carrier gas. For HfO2 and ZrO2, the film thickness was analyzed by ellipsometry. After 150 ALD cycles, the HfO2 and ZrO2 thicknesses were 13.5 nm and 15 nm, corresponding to 0.9 Å/cycle and 1.0 Å/cycle, respectively, consistent with the expected thickness-per-cycle values.
While typical FE HZO films are produced at thicknesses between 10 and 20 nm, a thicker film of 30 nm was chosen in order to obtain a higher signal-to-noise ratio from the diffractometer in Bragg-Brentano XRD geometry. Solid solutions of (HfxZr1-x)O2 were obtained by alternating super cycles with ratios of 0:1, 1:4, 1:1, 4:1, and 1:0 (TDMAH:TDMAZ) to target four nominal compositions of 0, 25, 50, 75, and 100% HfO2, respectively. To analyze the composition, HZO films (without top TiN) were crystallized by annealing at 500 °C for 30 s and then analyzed by x-ray Photoelectron Spectroscopy (XPS). The relative intensities of the high-resolution Zr 3d and Hf 4f peaks in XPS spectra indicated that the films contained 0%, 23%, 43%, 67%, and 100% HfO2, reasonably close to the solid solution ratios of 0%, 25%, 50%, 75%, and 100% expected from the ALD cycle ratios.
The structural evolution of the as-deposited (amorphous) HfO2–ZrO2 MIM films was monitored as a function of temperature using an x-ray diffractometer with a linear detector (Pixel1D, Empyrean, PANalytical) and a high temperature chamber (HTK1200, Anton Paar) in an N2 environment. The sample was heated at a rate of 2 °C min−1 from 25 °C to 1000 °C and cooled down at a rate of 2 °C min−1. Diffraction patterns were continuously taken during the entire heating and cooling treatment using a Cu K x-ray radiation source with 1.5418 Å and through a 2θ range of 24–45°. The temperature stage was also continuously heated, while diffraction patterns were collected for 2 minutes using a 2θ step size of 0.0263° and an integration time of 33 s per step. The total number of steps for the 2θ range of 24–45° was then approximately 800 steps. The temperature reading in each pattern is the average temperature of an individual scan (2 min per pattern and 2 °C min−1 heating rate), which yields a temperature resolution of 4 °C.
The equilibrium phase diagram of the HfO2–ZrO2 system suggests that both bulk HfO2 and ZrO2 should exist in the monoclinic phase at equilibrium.23 However, in the case of the HfO2–ZrO2 thin film system, the surface energy contribution is thought to promote a polar orthorhombic phase (Pca21) during the transformation from tetragonal (P42/nmc) to monoclinic (P21/m).14,15 In fact, the bulk-to-thin film HfO2-ZrO2 phase diagram proposed by Shibayama et al. demonstrates a compositional dependence on phase stability, indicating that higher concentrations of HfO2 will favor the monoclinic phase.24 Thus, we expect to see a similar dependence of phase stability on composition investigated in situ via HTXRD in the present study. Figures 1(a)–1(e) display the sequentially measured HTXRD patterns acquired during continuous heating of the sample for compositions of 30 nm HfO2-ZrO2 solid solution thin films including ZrO2, (Hf0.23Zr0.77)O2, (Hf0.43Zr0.57)O2, (Hf0.67Zr0.33)O2, and HfO2 from 25 °C to 1000 °C during heating.
We first discuss the onset of crystallization for each composition. In Fig. 1, all ZrO2-rich compositions start as XRD amorphous and then first crystallize into the tetragonal or orthorhombic phase (t-phase: P42/nmc, o-phase: Pca21, Pmn21, Pcba, or Pnma) with 2θ of ∼30.3° corresponding to closely overlapping reflections of the orthorhombic 111 reflection (here, referred to as o111) and the tetragonal 011 reflection (here, referred to as t011). Since these two closely overlapping reflections are difficult to separately identify, here, they are referred to collectively as o111/t011. Note that in Fig. 1, we choose to list the Pca21 phase as reference peaks to indicate the orthorhombic phase (in contrast to the other three non-polar orthorhombic phases) because it is difficult to discern the non-polar from the polar phase due to closely overlapping peaks. In addition to the standard peaks for HZO, the capping layer TiN is commonly known to oxidize and form a thin layer of TiO2 on the surface, which leads to the additional observed peaks in the diffraction pattern.25 Therefore, peaks at approximately 25.3° and 27.5° appear throughout the HfO2–ZrO2 compositions and may belong to TiO2 brookite and rutile phases, respectively, since they are not a signature of any phases corresponding to the HZO films.
The onsets of crystallization for the HZO films, obtained from evaluating the onset of diffraction intensity from the largest-intensity diffraction peaks, were determined to be 300 °C ± 5 °C (ZrO2), 328 °C ± 5 °C [(Hf0.23Zr0.77)O2], 363 °C ± 5 °C [(Hf0.43Zr0.57)O2], 375 °C ± 5 °C [(Hf0.67Zr0.33)O2], and 386 °C ± 5 °C (HfO2). As shown in Fig. 2, these results reveal a monotonic compositional dependence of the onset crystallization temperature with the increasing HfO2 concentration. Although heating is expected to impart microstructural changes, there were no significant changes observed in the XRD intensities after the onset of crystallization, indicating no change in the film's preferred crystallographic orientation after becoming visible in XRD. Fitting the crystallization temperature data to a least squares regression line, shown in Fig. 2, reveals a linear dependence in Zr-rich HZO compositions. However, at the highest HfO2 concentrations, the observed crystallization temperature is lower than that suggested by the linear trend, indicating a more constant crystallization temperature at higher HfO2 concentrations. Since HfO2-ZrO2 is an isomorphous solid solution, it is instructive to compare the crystallization temperature trends with other solid solution experiments where crystallization data were also collected starting from a solid amorphous phase. A similar trend was reported in the SrHfO3–SrTiO3 solid solution where an increase in crystallization temperature was observed with increasing Hf fraction. In the SrHfO3–SrTiO3 solution, the physical origins of the crystallization temperature behavior with the composition were not fully understood.26 However, the increase in crystallization temperature with increasing Hf was ascribed to the heavier atomic mass of Hf (relative to Ti). This is consistent with data in Fig. 2, which shows higher crystallization temperatures for films containing more Hf (178.49 amu) relative to Zr (91.224 amu). It is also reasonable considering that heavier atoms will require more thermal energy to diffuse and crystallize. The generally observed increase in crystallization temperature in HfO2-rich compositions may also be explained by the fact that HfO2 is known to exhibit a higher equilibrium crystallization temperature than ZrO2. Thus, according to rules of solid-solution mixing, mixing of ZrO2 with HfO2 may lower the crystallization temperature of intermediate solid solutions.27,28
Having considered the effect of the chemical composition on initial crystallization, now, we will discuss the effect of the chemical composition on phase evolution. In Fig. 1, the monoclinic phase (m-phase: P21/c) would appear as m111 and m reflections at angles of 28.3° and 31.6°, respectively. However, in the present data shown in Figs. 1(a)–1(c), these reflections are suppressed. ZrO2-rich compositions, (a) ZrO2 and (b) (Hf0.23Zr0.77)O2 and (c) (Hf0.43Zr0.57)O2, show minimal intensity at monoclinic reflections m111 and m . A weaker intensity peak at 31.6° emerges after 800 °C for the (Hf0.43Zr0.57)O2 composition, which may indicate some small fraction of m-phase appearing. Here, monoclinic peaks in Bragg-Brentano geometry are weaker due to low signal-to-noise at higher elevated temperatures; therefore, Grazing Incidence X-ray Diffraction (GIXRD) measurements on the post-annealed sample (Fig. S2) indicate the presence of two low-intensity peaks at 28.3° and 31.6° for (Hf0.43Zr0.57)O2. This confirms the emergence of the m-phase at this composition albeit at a smaller phase fraction compared to the o/t-phase. With higher HfO2 concentrations such as (d) (Hf0.67Zr0.33)O2 and (e) pure HfO2, more pronounced peaks at 28.3° and 31.6° become visible with increasing HfO2, indicating a larger m-phase fraction than that observed in ZrO2-rich compositions. Additionally, visible peaks at 24.2° and 24.6° emerge for composition (e) and further support the presence of the m-phase. The m-phase for (Hf0.67Zr0.33)O2 emerges at approximately 375 °C, which is the same temperature as the onset of the tetragonal or orthorhombic phase in this composition. This finding indicates that ZrO2-rich compositions or 50–50% (Hf:Zr) may be the preferred composition to suppress the m-phase, which is not the desired phase for ferroelectric HfO2. In Fig. 1(e), the pure HfO2 composition displays characteristic m111, m, and m200 peaks of the pure m-phase. Additionally, it should be noted that the transformation to the m-phase is irreversible. Our cooling measurements (Fig. S1) confirmed that the m-phase remained stable upon cooling for HfO2-rich compositions.25
We next discuss the possible tetragonal-to-orthorhombic phase transformation during heating for ZrO2-rich compositions. An inflection point is present in Figs. 1(a)–1(c) where the o111/t011 peak shifts toward lower 2θ beginning at approximately 750 °C for the given compositions. The degree of 2θ shift was measured to be 0.31°, 0.50°, and 0.37° for ZrO2, (Hf0.23Zr0.77)O2, and (Hf0.43Zr0.57)O2, respectively. A magnified 2θ region of 29.5° to 31.5° is replotted in Fig. 3(a) for (Hf0.43Zr0.57)O2 since this composition is of the most interest for ferroelectric applications and is representative of the behaviors observed in ZrO2 and (Hf0.23Zr0.77)O2. Other HZO Bragg peaks in the XRD pattern do not shift considerably. Representing the 2θ position as a function of temperature, Fig. 3(b) reveals an inflection point between 700 and 800 °C, suggesting that a phase transformation is occurring, an observation consistent with previous work.25 Because thermal expansion is expected to produce a linear change in the 2θ position vs temperature, the observed change indicates that other processes occur. Specifically, the non-linear o111/t011 peak shift in Fig. 3(b) from 30.3° toward 29.9° is ascribed to a phase transformation from the tetragonal phase to the orthorhombic phase at elevated temperatures.3,18,24,25,29
In order to assess phase evolution during cooling, o111/t011 2θ peak positions for three compositions, ZrO2, (Hf0.23Zr0.77)O2, and (Hf0.43Zr0.57)O2, were extracted from the cooling data (Fig. S1) and plotted as a function of temperature (Fig. S3). The linearity was assessed using a least squares regression line. Interestingly, the 2θ peak shift as a function of temperature for ZrO2 and (Hf0.23Zr0.77)O2 during cooling is close to linear, suggesting that further phase transformations do not occur upon cooling and that phases evolved during heating are retained upon cooling. Instead, a strong linear fit (R2 > 0.96) for these compositions suggests linear thermal contraction during cooling. The (Hf0.43Zr0.57)O2 composition, however, showed a weaker linear fit (R2 = 0.86), especially deviating at lower temperatures. The slightly non-linear o111/t011 inflection toward lower 2θ may suggest an additional phase transformation during the cooling process in (Hf0.43Zr0.57)O2, which would be consistent with a previous report that suggested that the orthorhombic phase occurs while cooling during rapid thermal annealing (RTA) processing.3 In the current study, however, data also suggest that a phase transformation may be occurring during in situ heating at slower heating rates than in RTA (the heating rate in the present experiment is 2 °C min−1, whereas in common RTA treatment, it is approximately 600 °C min−1). Therefore, kinetic driving factors may play additional roles in the tetragonal-to-orthorhombic phase transformation.
In addition to the ZrO2-rich compositions, it is interesting to note that the non-linear peak shift also occurs for pure ZrO2 films in Fig. 1. The presence of the orthorhombic phase and peak shifting toward lower 2θ in ZrO2 films have also been previously reported in the literature during nonequilibrium processes.19,24 Due to the nonequilibrium nature of capping layer confinement and free energy effects, it is reasonable that an orthorhombic phase may also appear in pure ZrO2 thin film systems, as we infer in the present experimental work.30 Figure 3 shows that, for all three ZrO2-rich compositions, the transformation point toward lower 2θ is completed near the same temperature of approximately 800 °C. This suggests a common thermodynamic driving force for the potential tetragonal-to-orthorhombic phase transformation in ZrO2-rich compositions.
The temperature stability of the o111/t011 phase was further investigated by inspecting the relative peak intensity fractions. While cooling data (Fig. S1) and GIXRD results (Fig. S2) reveal that the m-phase also emerges in (Hf0.43Zr0.57)O2, the monoclinic peak intensities were much weaker and impossible to fit in the in situ heating experiment. Therefore, the composition, which clearly showed the co-existence of both o/t- and m-phases during heating, (Hf0.67Zr0.33)O2, was selected to study the temperature dependence on the ortho/tetra phase stability. HTXRD in situ heating data were used to calculate the relative intensity fraction of ortho/tetra phases collectively (since they are not readily discriminated) relative to the m-phase at temperatures of 500–1000 °C using PDXL peak-fitting software31 for the (Hf0.67Zr0.33)O2 composition. We emphasize that the relative intensity fraction is not equivalent to the phase fraction but provides an indication of how phase fractions trend during the in situ HTXRD experiment. Fitting diffraction profiles to a pseudo-Voigt model, the integrated intensity fraction is ortho/tetra phase and m-phase . The overlapping o111 and t011 reflections were used to calculate the combined ortho/tetra relative intensity fraction (Io/t), while the m111 and m 11 peaks were used for the monoclinic phase (Im). The resulting relative intensity fraction for the HfO2-rich composition (Hf0.67Zr0.33)O2 is plotted in Fig. 4, which reveals that increasing annealing temperatures promote an increase in the m-phase fraction with temperature. Data in Fig. 4 suggest that low temperature processing is favorable for maintaining the tetra/ortho phase. Note that there is little measurable m-phase relative intensity in compositions below 50% HfO2, indicating a favorable tetra/ortho phase stability for ZrO2-rich compositions at the measured conditions.
In summary, in situ HTXRD is performed on 30 nm thin films for five compositions of the HfO2–ZrO2 system and the crystallization pathway is reported. It is observed that the monoclinic phase is suppressed in ZrO2-rich compositions, confirming that HfO2-rich compositions are likely to favor the monoclinic phase, while the tetragonal phase dominates ZrO2-rich compositions. A 2θ peak shift from 30.3° toward 29.9° is observed for ZrO2, (Hf0.23Zr0.67)O2, and (Hf0.43Zr0.57)O2, which is ascribed to a tetragonal-to-orthorhombic phase transformation. A monotonic compositional dependence for the onset crystallization temperature is found for the HfO2–ZrO2 system with higher onset temperature corresponding to increased concentration of HfO2. Relative intensity fraction calculations indicate that lower temperature annealing is favored for maintaining the tetra/ortho phase and suppressing the m-phase. Knowledge of the structural evolution of the HfO2–ZrO2 system and critical temperatures, such as the onset of crystallization and the tetragonal-to-orthorhombic phase transformation temperature, is an important requirement to exercise control over the HZO structure for ferroelectric and related properties. In addition, the demonstrated technique of in situ HTXRD can be applied to future structural investigation of fluorite and other crystal systems, e.g., La/Gd-alloyed HfO2, ferroelectric ZrO2, and La-alloyed ZrO2.
See the supplementary material for in situ HTXRD cooling data; GIXRD post-in situ for (Hf0.43Zr0.57)O2; and fitted 2θ position and full width at half maximum vs temperature for ZrO2, (Hf0.23Zr0.77)O2, and (Hf0.43Zr0.57)O2 compositions.
This work is based upon the work supported by the National Science Foundation, as part of the Center for Dielectrics and Piezoelectrics under Grant Nos. IIP-1841453 and IIP-1841466. HAH is supported by the National Science Foundation Graduate Research Fellowship Program (No. DGE-1746939). This work was performed at the Analytical Instrumentation Facility (AIF) at North Carolina State University and at the NC State Nanofabrication Facility (NNF), both of which are supported by the State of North Carolina and the National Science Foundation (Award No. ECCS-1542015). AIF and NNF are members of the North Carolina Research Triangle Nanotechnology Network (RTNN), a site in the National Nanotechnology Coordinated Infrastructure (NNCI).