Mist-CVD-derived Hf_0.55Zr_0.45O₂ ferroelectric thin films post-annealed by rapid thermal annealing

In 2011, Muller et al. confirmed that HfO₂ behaved as a ferroelectric material in thin films.¹ HfO₂ has long been used as an efficient high-k MOS gate insulator; thus, its ferroelectricity is expected to be applied to metal–ferroelectric–semiconductor (MFS) and/or metal–ferroelectric–insulator–semiconductor (MFIS) gate structures for ferroelectric gate FET (FeFET) non-volatile memories because of its excellent compatibility with the Si large-scale integration (LSI) process. Although several crystalline phases are known to exist in HfO₂, monoclinic, tetragonal, and cubic phases are particularly stable at ambient pressure.² These have been studied as gate oxides to scale down the gate length in MOSFETs due to their relatively high permittivity and have been put into practical use. They are paraelectric with no remanent polarization due to the presence of an inversion center in their crystalline structures. However, to exhibit ferroelectricity, the structure must be in a phase without central symmetry; among the metastable phases reported so far, one of the rectangular phases (Pca2₁) is a potential candidate,³ and it has been confirmed that the orthorhombic phase occurs.⁴ The o-phase occurs under a wide variety of conditions. So far, it has been revealed that the ferroelectricity in the HfO₂ family is induced by various dopant elements and annealing conditions, especially post annealing with and without a top metal electrode [post-metallization annealing (PMA) or post-deposition annealing (PDA)],^5–8 among other effective methods. Moreover, attempts have been made to stabilize the ferroelectric phase by epitaxial growth.^9,10

As a new candidate for future film growth, the mist-chemical-vapor-deposition (mist-CVD) technique has been proposed and developed. In general, mist CVD has been applied for the deposition of various oxide films with excellent properties, such as Ga₂O₃. Mist CVD exhibits several advantages for the deposition of oxide materials; it does not require active precursors in the atmosphere often used in conventional CVD; the mist, including precursors, is stable under atmospheric pressure, thereby eliminating the need for a vacuum. Therefore, mist CVD is generally a high-throughput and cost-effective process.^11,12 Furthermore, impurity doping and alloying can be performed by dissolving these precursors in a solution.^13–15 These are merits for low-cost preparation of thin films possessing CVD features such as 3-D and nm-order depositions on underlying patterns. Therefore, mist CVD is also a potential candidate for the back-end-of-line (BEOL) technology.

Few studies have been reported on mist-CVD-derived ferroelectric oxide films, although a number of excellent oxide films were prepared via mist CVD, as stated above. We have reported, for the first time, the preparation procedure and measurement results for as-deposited undoped HfO₂ and Hf_1−xZr_xO₂ (HZO) films via mist CVD.^16,17 However, the basic properties and appropriate preparation process of mist-CVD-derived ferroelectric HZO films have not been yet understood regardless of the above-reported pristine or unannealed ones. Mist CVD has recently been put into practical use,¹¹ although there are still several questions unexplained, such as why metastable phases are predominant^14,18 in mist CVD. Therefore, it is interesting and important to investigate the materials, including ferroelectrics, for which there are a few mist-CVD-deposited examples, in order to investigate unknown and/or new properties of mist CVD.

Atomic layer deposition (ALD) is the leading and most advanced thin film preparation technology, which has been applied for ferroelectric HZO thin film deposition. ALD-derived HfO₂ family films have been extensively researched, understood, and developed.^19,20 However, even for the improved ALD HZO or any other HfO₂-family thin film, one of the most essential issues is the endurance performance; the reported endurance pulse counts of at most an order of 10^11 21,22 are less than the practically required value of 1 × 10¹⁵.²³ One of the reasons is the intrinsically high coercive field of the HfO₂ family, leading to an earlier hard dielectric breakdown.²³ To realize high endurance performance, several studies have been conducted.^24–27 In recent years, the effects of oxygen vacancy (V_o) and its chemical effect on ferroelectric HfO₂-based thin films have been investigated.^28–30 Compared to other deposition techniques, mist-CVD deposition under atmospheric pressure and with a significantly high deposition rate (10 nm/min) is likely to produce thin films with chemical properties different from those obtained by vacuum processes such as ALD, sputter deposition, and PLD. Therefore, it becomes useful to investigate the existence of some different but effective properties and results from mist-CVD-derived HZO films.

In this study, we applied post-rapid thermal annealing (post-RTA) to mist-CVD-derived HZO thin films to explore additional physical and electrical properties and clarify basic behavior and appropriate preparation process.

The mist-CVD apparatus and a typical process for Hf_1−xZr_xO₂ (HZO) film preparation were detailed elsewhere.^16,31 First, metal–ferroelectric–metal (MFM) stacked capacitors were fabricated by using mist-CVD HZO thin films (about 20 nm in thickness) with top Pt and bottom TiN electrodes with thicknesses 150 and 50 nm, respectively, where Pt was DC-sputtered at room temperature by Ar ions at a pressure of 2 Pa and TiN was RF-sputtered at room temperature by Ar ions at a pressure of 0.65 Pa on a Si(100) substrate. For comparison, metal–ferroelectric–n⁺ Si (resistivity: 1.7 × 10⁻³ Ω cm) (MFS) stacked capacitors were also fabricated with the same HZO film and Pt top electrode.

In preparation for mist-CVD HZO thin film, hafnium acetylacetonate [Hf(C₅H₇O₂)₄] and zirconium acetylacetonate [Zr(C₅H₇O₂)₄] were dissolved in methanol (CH₃OH) to be used as precursors during the process, where the precursor concentration in the solution was 0.02 M. Then, a mist of the precursor source solution was atomized by using an ultrasonic transducer at 2.4 MHz and then transported into the mixing part of the solution-mist gas area of the apparatus by a carrier nitrogen gas at a flow rate of 7.5 l/min for a growth time of 2 min. The precursors were chemically reacted at 400 °C on the TiN or n⁺ Si(100) substrate in the reaction area of the apparatus. Zr/(Hf+Zr) ratio, x, was set at 0.45 by adjusting the quantities of both precursors.

Different from our previous studies, the grown HZO samples were post-deposition annealed (PDA) or post-metallization annealed (PMA) by rapid thermal annealing (RTA: ULVAC-RIKO: MILA-5000, MILA-TER-P, GP-1000G). The RTA conditions were investigated with respect to the atmosphere gas, its pressure, annealing temperature, and its time.

Second, to investigate the physical properties of the HZO thin films, the crystalline structure was analyzed by 2θ−ω scanning using a glancing-angle incidence x-ray diffraction (GIXRD: Bruker D8 Discover). The surface morphology was evaluated via atomic force microscopy (AFM: SII Nano Technology Nanonavi/E-sweep) and scanning electron microscopy (SEM: JEOL JSM-7001F). Moreover, the cross-sectional image and film thickness were obtained via SEM; the thickness was also monitored by using a film thickness monitor (DekTak: XT-3). The elemental composition throughout the film was measured and analyzed via x-ray photoelectron spectroscopy (XPS: JEOL JPS-9010MX).

Third, regarding the electrical properties, the polarization–electric field (P–E) and corresponding current density–electric field (J–E) curves were measured at room temperature; then, double pulse measurements were performed to extract the real ferroelectric spontaneous polarization. Finally, endurance properties for the polarization were measured with a rectangular fatigue pulse by using a ferroelectric tester (FCE-3, Toyo Technica).

Figure 1 shows the GIXRD 2θ−ω scanning profiles of 20 nm-thick Hf_0.55Zr_0.45O₂ films deposited on both TiN and n⁺ Si(100) substrates at a mist-CVD growth temperature of 400 °C, followed by RTA at annealing temperatures from 500 to 650 °C for 20 s in air atmosphere. In the top panel of Fig. 1, 2θ (diffraction angle) values for different crystalline phases in bulk HfO₂ are included as a reference.

The profiles for the n⁺ Si substrates reveal orthorhombic, tetragonal, and cubic (o/t/c) phases for 2θ of ∼30° for every RTA temperature from 500 to 650 °C compared to the as-deposited sample. It is observed that the monoclinic (m) phase decreases monotonously with decreasing temperature. It has been reported that V_o contributes to the formation of the o-phase.³² We consider that the supply of oxygen from the atmosphere during annealing at the higher temperature may reduce V_o and promote the formation of the m-phase. In addition, AFM surface images of HZO/n⁺ Si show that the grain size and root-mean-square (rms) surface roughness of HZO increased typically from 30 to 50 and 0.25 to 0.50 nm, respectively, by increasing the RTA temperature from 600 to 700 °C for 20 s (not shown); therefore, the increase in size and roughness may be related to the increase in the ratio of m-phase.

The profiles for the TiN films reveal o/t/c phases for every RTA temperature from 500 to 650 °C. Note that even for the highest temperature, 650 °C, the m-phase is insignificantly observed in HZO film on TiN, in contrast to that on Si. Therefore, it would be suitable from the viewpoint of HZO crystallinity to use TiN as a growth substrate for obtaining ferroelectric HZO film rather than Si. It should be noted that the peak for TiN(111) around 36.6° is obvious for the as-deposited films; however, the peak intensity decreases and eventually almost disappears after annealing above 600 °C. This suggests that the crystal structure of TiN changed due to oxidation during RTA (see Secs. III A 3 and III A 4). In addition, the AFM surface images of HZO/TiN showed that the grain size and rms surface roughness of HZO were typically 50–60 and 0.69 nm after RTA at 650 °C, respectively. This suggests that grain growth would be slightly faster on TiN than on Si. From cross-sectional SEM images (not shown), columnar grain structures were also noticeable for annealed HZO/TiN but not for the as-deposited.

The distribution of the elemental components in HZO thin film was investigated via x-ray photoelectron spectroscopy (XPS) with Ar-ion etching. Figures 2(a) and 2(b) show the XPS depth profiles for each elemental component in HZO thin films on n⁺ Si and TiN, respectively. Both HZO films were RTA-annealed at 650 °C for 20 s in air atmosphere. In these experiments, Hf, Zr, O, C, Si, Ti, and N distributions were measured; oxygen and/or V_o are always important in oxide, and carbon often becomes the origin for the leakage phenomenon even in minute amounts. The others are constituent elements of growth film and substrate.

It is observed from Fig. 2(a) that, compared to the other elements, carbon exists in small amounts throughout the HZO film, near the HZO/Si interface region, and in the Si substrate, although the top surface of the HZO film was carbon-contaminated. Unintentional carbon contamination from liquid sources used in the mist-CVD process was one of our main concerns due to an increase in the leakage current in the HZO film. However, the carbon concentration might be little enough from the result, compared with V_o discussed later. Therefore, it is reasonable to consider that the carbon deriving from liquid sources vaporized during the mist-CVD growth process at 400 °C and/or RTA at 650 °C. The distributions of Hf, Zr, and O are almost constant throughout the HZO film, away from the film surface and interface with the Si substrate. The estimated thickness of the transition layer at the HZO/Si interface is 5–8 nm, where the hafnium zirconium silicate (HfZrSiO) layer naturally exists. Hf and Zr are observed to diffuse to some extent into the Si substrate and eventually disappear. However, O is found to exist in the Si substrate within a significant depth (∼15 nm), possibly due to the Ar-ion bombardment effect on light elements such as O during the etching, and the difference in the diffusion length might be due to that in mass or chemical activity between O and both Hf and Zr.

Compared to the above results for Si, there are some differences in the behavior of each element throughout the HZO/TiN structure, as shown in Fig. 2(b). XPS intensities for Hf, Zr, and O decrease monotonously with increasing depth in the HZO film. It is also observed that they are diffused into the TiN layer, especially O, which is significantly absorbed. We consider that TiN scavenged oxygen from HZO during the annealing or deposition.³³ Similarly, N also diffuses into the HZO film and is distributed throughout. In addition, compared with the n⁺-Si case (discussed in Sec. III A 2 a), C is observed near the HZO/TiN interface and in TiN. This leads to an estimation that the TiN surface would have included C just before the HZO film growth. Moreover, the thickness of the HZO/TiN interface transition layer seems small compared to that in HZO/Si. The difference would be due to a more enhanced growth reaction of the HfZrSiO or SiO_x layer in HZO/Si than in HZO/TiN.

Since the diffusion and/or scavenging of oxygen into the TiN layer are important to generate V_o in the HZO film, inferred from the above results of Sec. III A 2 b, the XPS spectra of the Ti oxide and oxynitride near the HZO/TiN interface and in the TiN layer were measured. Figures 3(a)–3(c) show the XPS spectra of Ti 2p_1/2 and Ti 2p_3/2 in the HZO/TiN structure for the number of Ar-ion etching steps (N) of 7 (in the transition layer), 9 (shallower region in TiN layer), and 12 (deeper region in TiN layer), respectively, where the spectrum in panel (a) was separated into three spectra corresponding to TiN, TiO, and TiON referred to from literature^34,35 by calculated fitting from the measured results and then plotted. It is clear from Fig. 3 that both Ti–O and Ti–N bonds exist regardless of the depth position. For N = 12 (deeper in TiN layer), Ti–O related peaks seem to decrease compared to those for N = 7, 9. For N = 7 (in the transition layer), the TiON spectrum is clear; this indicates that the exchange between N and O occurs near the interface,²⁹ and then, N spreads into the HZO-film side. Note that the TiO peak at ∼458 eV seems to be weaker than the TiN peaks at ∼455 and 460.5 eV. This indicates that the TiO_x layer, as a chemical buffer layer suppressing the diffusion or scavenging,^36,37 was not successfully grown either due to a typically high-deposition rate of mist CVD (about 10 nm/min) or due to the reduction of oxygen by methanol.

Figure 4 shows a series of XPS spectra of Hf 4f_5/2 and Hf 4f_7/2 for Hf oxides, featuring the HfO_2−x suboxide existing on the TiN layer as a parameter of its depth position with the number of Ar etching steps (N) as same as the above. At binding energy of 15–16 eV, the peak of the HfO_2−x suboxide³⁸ appears at all the depths. In particular, for N = 3–8, their peak intensities seem to be almost equal. Furthermore, the integrated areas of the XPS spectra originating from Hf 4f_7/2 and Hf 4f_5/2 for HfO₂ and from HfO_2−x with the reported peak energy were calculated. Then, the area ratio of the integrated XPS signal for the suboxide to that for stoichiometric HfO₂ vs the depth was replotted in Fig. 5(a). It is seen that the HfO_2−x suboxide ratio is increased from ∼15% to ∼30% with increasing depth, except for the HZO surface region. The ratio is estimated as ∼30% at the HZO/TiN interface. Therefore, it is understood from the result of XPS spectra originating from Hf suboxide that the distribution of V_o in the HZO/TiN structure corresponds to the depth profile of oxygen in Fig. 2(b), where the oxygen decreased with increasing depth. Here, it should be stressed that the V_o concentration is much larger for mist CVD (>∼15%) than for ALD (∼1%).³⁹ This would induce a difference in electrical properties between mist CVD and ALD.

On the other hand, by using the same procedure, the ratio of V_o for the n⁺ Si sample is plotted in Fig. 5(b). Unlike the result for TiN, there is almost little V_o, suggesting that the Si substrate would scavenge little oxygen due to the existence of a SiO_x buffer layer grown on Si. Accordingly, the result leads to a belief that oxygen diffusion to the bottom electrode is one to two orders weaker in n⁺ Si than in TiN.

We measured the current density corresponding to the polarization in 20 nm-thick HZO thin films on both n⁺ Si and TiN after post-metallization annealing (PMA) under different RTA conditions of ambient gas and gas pressure. Figure 6 shows current densities under 2 MV/cm at 1 kHz vs gas pressure after RTA at 600 °C for 30 s as a parameter of the kind of substrate or gas. Note that this current includes leakage and displacement currents in addition to a polarization reversal current. The HZO film on TiN annealed in pure N₂ showed a current density larger than 300 mA/cm² for all pressures (not shown), suggesting that a large amount of V_o contributed to the current conduction. In agreement with this, it is found from Fig. 6 that for all RTA conditions and substrates, the current decreases monotonically with increasing pressure, meaning that the leakage is the lowest at 10⁵ Pa, i.e., atmospheric pressure, which indicates the suppression of V_o. The current in n⁺ Si is lower than in TiN. First, this is due to the grown HfZrSiO or the SiO_x layer between HZO and n⁺ Si behaving as an effective insulating buffer layer. Second, it is due to the fact that there are few V_o in HZO on n⁺ Si, while there are many on TiN, as discussed in Sec. III A 4. Furthermore, the leakage tendencies on both n⁺ Si and TiN are qualitatively similar after annealing in O₂ and N₂ gases. This might also reflect the role of the above-mentioned interface layer, such as HfZrSiO or TiO_x and TiON, in suppressing the leakage. We observed that the leakage currents after annealing in pure O₂ and air were comparable for all conditions (not shown here). This suggests that a certain amount of oxygen is sufficient to suppress the leakage on both n⁺ Si and TiN. Compared to most ALD-derived samples,^5,6 where annealing in a nitrogen atmosphere does not increase but decrease the leakage current, mist-CVD-derived samples, which need to be annealed in the oxygen atmosphere to reduce the leakage current, would show an increase in V_o during nitrogen annealing as previously mentioned. The reason why the passivation by nitrogen annealing does not work is unknown at present. However, oxygen vacancies would be significant because, in the results, the leakage was reduced in the oxygen atmosphere.

Figure 7 shows the polarization–electric field (P–E) hysteresis curve [in panel (a)] and the corresponding current density–electric field (J–E) hysteresis [in panel (b)] measured at 0.5 kHz after an initial poling of about ten cycles with a maximum imposed electric field of 2 MV/cm for the Pt/HZO/TiN stacked capacitor under RTA-annealing in air atmosphere at 550 °C for 0 s (spike anneal). A ferroelectric curve is obtained with concave regions, as shown in Fig. 7(a), although it is significantly fat and shows an apparent large polarization density at E = 0, P(E = 0), compared with typical values of around 10–30 µC/cm² for the remanent polarization density, P_r, in HZO thin films.^1,5–8 According to Sec. III B 1, this is due to the inclusion of the non-reversal polarization due to the leakage originating from several sources such as the V_o in the film. From Fig. 7(b), a noticeable polarization reversal current appears at ∼1.3–1.4 MV/cm, which is within the typical range for the reported HZO films.^1,5–8 It has been recently suggested that a polarization reversal current effect similar to ferroelectricity may occur due to electron trapping by V_o.^40,41 This ferroelectric-like phenomenon was detected at low frequencies, i.e., 0.001–10 Hz, and depends strongly on the frequency.⁴² This sample showed hysteresis loops for frequencies higher than 0.5 kHz; therefore, the obtained hysteresis is properly considered as originating from ferroelectricity.

Similar to the models of P–E hysteresis discussed previously,⁴³ the background leakage curve in this sample showed an exponential increase with imposing electric field, which corresponds to several conduction mechanisms such as Poole–Frenkel one.

On the other hand, the annealed HZO/n⁺ Si samples showed polarization reversal but mainly included paraelectricity. These samples showed a polarization reversal current but lacked reproducibility among them. This would be due to HZO/n⁺ Si exhibiting few V_o, as shown in Fig. 5(b), which contribute to ferroelectricity³² because oxygen was not scavenged sufficiently from the HZO film on n⁺ Si, unlike the HZO/TiN system.

To remove the non-reversal polarization component from the total polarization, a double-pulse method was applied to the P–E measurement. The measurement procedure was described elsewhere.^16,17,44 Briefly, by applying a double-pulse triangular wave, a hysteresis curve comprising only the ferroelectric component is extracted by subtracting the charge (without the ferroelectric component) obtained from the second imposed non-inverting triangular pulse from the total charge obtained from the first imposed inverting triangular pulse. Figure 7(c) shows the P–E curve for the positive bias region after the double-pulse method is applied. As a result, it is revealed that the film has definitely a component of ferroelectric spontaneous polarization. Moreover, P(E=0) originating from the ferroelectric component, P_r, became 20 µC/cm², which was ∼45% of that from the normally measured P–E curve. Nonetheless, P_r was significantly improved from those of the previous ones without annealing¹⁷ because the contribution of the ferroelectric polarization was much increased by RTA in this work. It was revealed that the obtained value of ferroelectric polarization is close to the typically reported one.^1,5–8

Next, to evaluate the endurance characteristics for the Pt/HZO/TiN ferroelectric capacitor, we measured the same sample as in Sec. III B 2. Figures 8(a) and 8(b) show the polarization current density–electric field (J–E) and polarization–electric field (P–E) hysteresis curves measured at 0.5 kHz, respectively, with the number of endurance pulse cycles as a parameter of the pulse being rectangular with the frequency of 100 kHz and pulse height of 2.5 MV/cm.

As shown in Fig. 8(a), the current increases for endurance pulse cycles from 10⁵ to 10⁸, before decreasing for cycle numbers up to 2 × 10⁹. In response to this behavior, it is observed from Fig. 8(b) that P(E = 0) shows a similar behavior—it increases with increasing endurance pulse cycle number from 10⁵ to 10⁸, before decreasing with an increase in the cycle number up to 2 × 10⁹. The squareness of the P–E hysteresis is improved for the largest number of cycles of 2 × 10⁹, which directly relates to the decrease in the current, as shown in Fig. 8(a).

Basically, from the XPS results explained in Sec. III A 3, it is reasonable to consider that during the imposed pulses, charged oxygen, its vacancy, and charged nitrogen can move across the HZO/TiN interface (because there seems to be little TiO_x at the interface, which may become a diffusion barrier and prevent the diffusion). We consider that the increase in the leakage current up to 10⁸ cycles is caused by the formation of V_o due to the scavenging of oxygen into TiN. It is reported that the leakage current in the HfO₂ film decreases generally by introducing nitrogen near the sites of V_o.⁴⁵ This suggests that during the fatigue process, V_o is generated at the side of the TiN bottom electrode (growth substrate) and is located near the nitrogen sites, which is supported by the XPS results in Sec. III A 3. As reported that nitrogen doping decreases the leakage current density,⁴⁶ the results further suggest that after nitrogen is supplied to HZO from TiN, the bottom electrode contributes to the reduction of the leakage current during 10⁸ to 2 × 10⁹ cycles.

Figure 8(c) shows P(E = 0) vs imposed endurance pulse counts, revealing the endurance properties of the HZO thin film under imposed rectangular pulses for up to 3 × 10⁹ counts with an amplitude of 2.5 MV/cm at 100 kHz. In response to the results of Figs. 8(a) and 8(b), the so-called “wake-up” phenomena are observed for pulse counts from 10⁵ to 10⁸. It is generally considered from the above-obtained results that the “wake-up” corresponds to the period of increase in the leakage due to the generation and movement of V_o near the HZO/TiN interface. After 10⁸ pulse counts, P(E = 0) starts to decrease. This reflects the behavior explained in Fig. 3 that nitrogen would move and locate near the V_o. Subsequently, the non-polarization reversal component of P(E = 0) was reduced due to the decrease in leakage. The Pt/HZO/TiN capacitor was confirmed to endure up to 2 × 10⁹ counts, while it reached a hard breakdown before 3 × 10⁹ ones. In the fatigue process after the wake-up, as reported in Ref. 47, the P–E hysteresis improved as shown in Fig. 8(a). These phenomena are quite similar to the previous ones, indicating that the mist-CVD-derived HZO film possesses essentially similar properties to the HZO films grown by other deposition methods such as ALD. The polarization in Fig. 8(c) is composed largely of the parasitic non-linear leakage current. Therefore, to show a more precise polarization endurance plot, the component due to the parasitic current should be eliminated from the value of polarization in Fig. 8(c).

In view of the above results, the leakage current and endurance characteristics were markedly influenced by the diffusion and movement of oxygen, its vacancy, and nitrogen during annealing and pulse cycling. The intensity and/or frequency of these dynamics in the HZO film should be related to the significantly fast reaction process inherent in mist CVD between supplied mist liquid molecules and heated growth substrate. The next challenge for improving the electrical properties, especially P–E curves, of mist-CVD-derived HZO film is to further reduce the background leakage current, including the large-concentrated V_o as its major origin. To solve this, it would be proper to perform a milder process at the film growth, then reduce the non-thermal equilibrium conditions in the growth reaction, slowing down the deposition rate, and optimizing the increase in oxygen supply as carrier gas. Afterward, post-RTA processes should be also reconsidered.

We have newly applied a series of post-annealing processes through RTA on mist-CVD-derived HZO thin films. A ferroelectric polarization was confirmed with its concave P–E shape and noticeable polarization reversal currents. Those ferroelectric properties of the HZO thin film provided quantitative values for P_r and E_c of about 20 µC/cm² and 1–1.5 MV/cm, respectively, similar to the reported ones from other growth methods. It is also revealed that the background leakage should be further reduced in the mist-CVD HZO film compared to that recently reported by ALD. The origin of leakage was strongly related to V_o generated in the film and near the HZO/bottom electrode interface. Nonetheless, to reduce the leakage, it was found effective to use atmospheric pressure in air and oxygen in the post-RTA process. Endurance behaviors for the mist-CVD HZO film revealed results qualitatively similar to those prepared by the other methods for both “wake-up” and “fatigue” phenomena, showing that the mist-CVD HZO film endured up to 2 × 10⁹ counts.

We consider that the next challenge is to find milder process conditions in mist CVD, then reduce the non-thermal equilibrium conditions in the growth reaction, slowing down the deposition rate, and optimizing the oxygen supply to reduce V_o.

The authors would like to thank Professor Kazuo Takahashi, Electronics, Kyoto Institute of Technology, for assisting with XPS measurements and their analyses.

The authors have no conflicts to disclose.

Sho Tanaka: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Validation (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). Yuki Fujiwara: Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Visualization (equal); Writing – original draft (equal). Hiroyuki Nishinaka: Methodology (supporting); Resources (supporting); Validation (supporting); Writing – review & editing (supporting). Masahiro Yoshimoto: Resources (supporting); Writing – review & editing (supporting). Minoru Noda: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Funding acquisition (equal); Investigation (equal); Methodology (equal); Project administration (equal); Resources (equal); Supervision (equal); Validation (equal); Writing – original draft (equal); Writing – review & editing (equal).

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