Introduction

HfO2-based ferroelectrics have attracted great attention in terms of both ferroelectric (FE) material science and non-volatile memory applications since their discovery in 20111,2,3,4,5,6,7,8,9,10,11,12,13,14,15. Optimizing properties, including remanent polarization (Pr), retention, endurance, switching speed, etc., have been significant issues for ferroelectric memories. It has been well agreed that the ferroelectricity of HfO2-based thin films originates from a metastable Pca21 orthorhombic (O) phase, which is formed during the tetragonal (T) to monoclinic (M) phase transition in thermal annealing5,16,17,18. Therefore, FE characteristics are typically observed with the coexistence of T-, O-, and M-phases and polycrystalline structure in the films, though epitaxial FE O-phase and FE rhombohedral phase films also have been demonstrated19,20,21,22,23,24. Thus, the key to controlling the FE properties mainly lies in controlling the phase composition of the films and the content of the O-phase. In the past decade, intensive studies have focused on maximizing the quantity of O-phase through optimization of material and process parameters, such as dopants and doping concentration1,8,25,26,27, deposition methods and parameters28,29,30,31,32, annealing conditions33,34, electrode materials35,36, oxygen vacancy (VO) concentration26,37,38,39, film thickness40,41, etc. And, a linear relationship between the content of O-phase and Pr has been established26,42.

Domains, as another critical fundament for FE materials in addition to the crystal phase, also play a crucial role in controlling FE properties as well known. A FE, or a polar domain, is conventionally defined as a spatial region within a single crystal or grain where the direction of spontaneous polarization remains consistent. While factors such as domain size, shape, and domain walls can affect various ferroelectric properties, the Pr of an FE film is specifically influenced by the out-of-plane component of spontaneous polarization, which is determined by the domain orientation in the film43. For epitaxial ferroelectrics usually with a single-crystalline structure, the texture of the crystal is fixed, and its symmetry is broken in different directions, resulting in multiple domains separated by domain walls44,45. As for HfO2-based FE films typically with polycrystalline structure and multiple phases, however, not only are multiple domains formed within a crystal grain, but the texture of crystal grains is also relatively flexible, both influencing the contribution of spontaneous polarization to Pr, as shown in Supplementary Fig. S12. To enhance Pr value, the goal is to achieve as many FE domains as possible which could contribute to Pr as large as possible in O-phase occupied areas, such as (001)O-oriented domains. Meanwhile, it has been reported that the (010)O-oriented domains, whose polarization direction initially lies in-plane of the film, could transform into (001)O-oriented domains under an electric field46,47. So, the goal to engineer domain orientations to get larger Pr becomes to obtain a dominance of (001)O- and/or (010)O-oriented FE domains. And, achieving the goal requires engineering both the macroscopic texture of grains and specific domain orientations inside the grains.

The engineering of the texture of HfO2-based films has been greatly done by employing single-crystalline substrates, such as yttrium-stabilized zirconia (YSZ)21,48,49,50,51,52,53,54,55, SrTiO3 (STO)20,23,56,57,58,59, and even TiN55,60,61, while the control of domain orientations inside the grains remains a challenge. Notably, the polarized O-phase (Pca21) generally transforms from the T-phase in HfO2-based films, so the understanding and controlling of the specific T-O phase transition pathway is crucial for controlling domain orientations within the crystal grains. For easy comparison of these two phases, a double-sized T-phase unit cell (a = b = \({\sqrt{2}}\)aT, c = cT) has been used33,62. It has been declared that the longest c-axis of the T-phase should transform into the longest aO-axis of the O-phase, while the equivalent a- and b-axes of the T-phase should shift to the similar bO- and cO-axes of the O-phase in T-O phase transition based on lattice parameter differences between T- and O-phases63. And, through such a T-O transition pathway, a study has successfully achieved (001)O/(010)O-oriented domains in 7%-YO1.5-substituted HfO2 (YHO-7) films by precisely controlling the epitaxial (100)-oriented T-phase on single-crystalline YSZ (001)62. Such a transition way, however, requires strict control of the single-crystalline structure of the T-phase, which is quite challenging for HfO2-based films and limits their applications greatly.

Here, we conceive that if a T-O phase transition pathway from other orientations of T-phase to (001)O- and/or (010)O-oriented O-phase could be achieved together with the reported one as mentioned above, it would be promising to engineer the dominance of (001)O- and/or (010)O-oriented domains even in a polycrystalline structure in HfO2-based films, as schematically shown in Fig. 1. Therefore, in this work, we propose a potential way to approach this goal through controlling the interface dislocations between Hf0.5Zr0.5O2 (HZO) films and single-crystalline TiN substrates. We perform a systematic investigation of crystalline texture, domain orientation, and interface structure as well as polarization-electric field (P-E) characteristics in HZO on TiN substrates with three different orientations. Our results show that (001)O- and/or (010)O-oriented domains can be controlled in polycrystalline and multiple phases of HZO on TiN (001) and (111) with enhanced Pr values in contrast to HZO on TiN (110). This is achieved through a hidden T-O transition pathway from the longest c-axis of the T-phase towards the relatively shorter bO/cO-axis of the O-phase, in addition to the reported one towards the longest aO-axis of the O-phase. Such a hidden phase transition pathway is assisted by the periodic dislocations at the bottom HZO/TiN interface and occurs during the cooling down of the annealing process and possibly in polarization switching cycling as well. Our research highlights the potential for engineering domain orientations and further FE properties in polycrystalline HZO films.

Fig. 1: Schematics of T-O transition pathways in Hf0.5Zr0.5O2 films.
Fig. 1: Schematics of T-O transition pathways in Hf0.5Zr0.5O2 films.
Full size image

a The unit cells of T- and O-phases. b Schematics of the O-phase formation pathway. O-phase forms during the T-to-M-phase transition process. c The reported and possibly hidden T-O transition pathways. In the reported case, the (010)O- and (001)O-oriented O-phases should transform from the (100)-oriented T-phase (using a double-sized T-phase unit cell). The direction of polarization (P) should align with the c-axis of the O-phase. If hidden T-O transition pathways toward (010)O- and (001)O-oriented O-phases can be realized, it would provide greater flexibility, enabling the engineering of domain orientations and further Pr in polycrystalline HfO2-based films.

Results and Discussion

Crystalline texture of HZO on different orientations of TiN substrate

The crystalline textures of 10 nm HZO films on single-crystalline TiN (001), (110), and (111) substrates were investigated using the synchrotron-based grazing incidence X-ray diffraction (GIXRD) technique. The setup and the device structure are schematically shown in Fig. 2a. The excellent single crystallinity of TiN substrates is demonstrated in Supplementary Fig. S1. Although the quality of TiN (111) is not as good as other orientations, it would not fundamentally change the properties of HZO films grown on top (see section (A) of the Supplementary Note). Figure 2b–d shows the 2D GIXRD images of HZO on TiN (001), (111) and (110), respectively, where (111)O/(011)T and {002}O/(110)T/(002)T diffraction rings at around 30.5° and 35.5° are identified and marked. Here {002}O includes information of (200)O, (020)O, and (002)O due to their similar lattice distance. Unfortunately, the information of {110}O cannot be obtained directly because the diffraction ring is obscured by the light spot in the center. Firstly, the content of the O-phase among the samples was determined by analyzing the XRD results and performing investigations with Raman spectroscopy (see section (B) of the Supplementary Note). Interestingly, no obvious difference is observed among HZO films on TiN substrates with different crystalline orientations. This suggests that the orientation of the bottom TiN has a limited effect on adjusting the total content of the O-phase.

Fig. 2: Synchrotron-based GIXRD results of pristine Hf0.5Zr0.5O2 films on single-crystalline TiN substrates with different orientations.
Fig. 2: Synchrotron-based GIXRD results of pristine Hf0.5Zr0.5O2 films on single-crystalline TiN substrates with different orientations.
Full size image

a Schematic of the device structure and the synchrotron-based GIXRD measurement setup. The 2D GIXRD image of HZO b on TiN (001), c on TiN (111), and d on TiN (110). (111)O/(011)T and {002}O/(110)T/(002)T diffraction rings are identified and marked, respectively. The intensity-azimuth angle (\(\gamma\)) curves of {002}O/(110)T/(002)T and (111)O/(011)T peaks of HZO e on TiN (001), f on TiN (111), and g on TiN (110). The bottom row illustrates the preferred out-of-plane orientations of specific lattice planes of T- and O-phases in HZO films on TiN (001), (111), and (110), concluded from the GIXRD results.

Then the in-plane and out-of-plane textures are taken into account. From the marked rings in Fig. 2b–d, the peak intensities of {002}O/(110)T/(002)T and (111)O/(011)T for HZO on TiN (001), (111), and (110) were extracted as a function of γ (the azimuth angle) in Fig. 2e–g, respectively. For HZO on TiN (001) and (111), the intensity of the {002}O/(110)T/(002)T peak is higher at around γ = 90° (out-of-plane) than around γ = 30° and 150° (in-plane), while the (111)O/(011)T peak intensity shows peaks near γ = 35° and 145°. Both indicate a highly out-of-plane orientation of {002}O and (110)T/(002)T. In contrast, for HZO on TiN (110), two round peaks in the intensity of (111)O/(011)T are observed at γ = 55° and 125°, indirectly indicating the preferential orientation of {110}O and (112)T/(020)T at γ = 90°. A detailed analysis of XRD results is presented in section (C) of the Supplementary Note. These observations suggest that the texture of the T/O grains in HZO films can be engineered by controlling the orientation of the TiN substrate. And, {002}O/(110)T/(002)T grains can be controlled out-of-plane by using TiN (001) and TiN (111) substrates, which is the base to further obtain (001)O- and/or (010)O-oriented domains. Note that the above diffraction information of (110)T and (002)T is obtained based on the original unit cell of the T-phase. With a double-sized unit cell, it should be written as (200)- and (002)-textured. This notation is also used in the following text.

Domain orientations tuning with different orientations of the substrate

To obtain microscopic information on the domains inside the O-phase grains, high-resolution scanning/transmission electron microscopy (S/HRTEM) investigations on these samples in the pristine state were further carried out, as shown in Fig. 3. The out-of-plane growth direction of the film in different areas can be determined from their corresponding Fast Fourier Transform (FFT) patterns. Consistent with GIXRD results, most of the areas exhibit T-phase in all films (see Supplementary Fig. S14), while O-phase crystals/domains are also observed clearly. In principle, three orientations of domains are predictable from each kind of O-phase texture, for example, (100)O-, (010)O- and (001)O-oriented domains are expected from {002}O texture. Very interestingly, however, (010)O-oriented domains are dominantly present in O-phase grains on TiN (001), while (001)O-oriented ones appear in some areas from multiple image statistics (Fig. 3a). Both types of domains can contribute to Pr completely. Namely, the (100)O-oriented domain, which has no contribution to Pr, is avoided. For HZO films on TiN (111), as shown in Fig. 3b, (010)O-oriented domains are also predominantly present, while (001)O-oriented domains are not found along with (100)O-oriented ones in the observed areas. Additionally, a minor fraction of (111)O-oriented domains is also observed in this case (see Supplementary Fig. S15). In contrast, all three kinds of (101)O-, (110)O-, and (011)O-oriented domains predicted from the out-of-plane texture of {110}O in XRD results are observed in HZO grains on TiN (110). Since the cO-axes (aligned with the spontaneous polarization direction) of (101)O- and (011)O-oriented domains form a 45° angle with the out-of-plane direction, they can only contribute partially to Pr, while the (110)O-oriented domains have no contribution to Pr due to the in-plane cO-axis orientation. Thus, these domains are limited to getting high Pr theoretically. Nevertheless, these results clearly demonstrate the tuning effect of TiN substrate orientation on domain orientations in HZO. And, the goal to get the dominance of (001)O- and/or (010)O-oriented domains in polycrystalline HZO is achievable by employing a TiN (001) substrate and possibly by using TiN (111) as well.

Fig. 3: HRTEM images of Hf0.5Zr0.5O2 films before wake-up on different orientations of TiN substrate.
Fig. 3: HRTEM images of Hf0.5Zr0.5O2 films before wake-up on different orientations of TiN substrate.
Full size image

a For HZO on TiN (001), in O-phase occupied areas, the (010)O-oriented domains are dominant, and the (001)O-oriented domains appear in a small part of the observed areas. b For HZO on TiN (111), the (010)O-oriented domains are dominant. c For HZO on TiN (110), the (110)O-, (101)O-, and (011)O-oriented domains are dominant.

P r engineering beyond O-phase content controlling

The effect of domain orientation on the Pr value of HZO is further considered. As mentioned in the introduction part, the Pr value of HZO is conventionally controlled by the content of the O-phase. Our HZO films on different orientations of TiN substrates show a similar content of O-phase but different domain orientations. And, in principle, the predominant (001)O- and/or (010)O-oriented domains in HZO on TiN (001) and (111) should enable enhanced Pr performance in contrast to that on TiN (110) even though O-phase content is similar. To confirm this, we experimentally investigated the P-E characterizations of the above samples. Figure 4a–c shows P-E curves of HZO films on three orientations of TiN substrates in the pristine state and after wake-up. Pristinely, all devices displayed an antiferroelectric (AFE)-like characterization, consistent with a large T-phase content in this state. After wake-up, typical FE P-E hysteresis curves were observed for all three samples. The comparison of Pr values after wake-up in three cases is shown in Fig. 4d. Clearly, the 2Pr of HZO on TiN (001) is slightly larger than that of HZO on TiN (111) but notably higher than that on TiN (110). The slightly smaller Pr of HZO on TiN (111) than that on TiN (001) should be due to the effect of a small part of (111)O-oriented domains whose spontaneous polarization contributes to Pr partly. Therefore, these results are exactly as expected from the predominant orientation of domains in HZO on each substrate. In addition, Pr evolutions in wake-up were also conducted as shown in Fig. 4e. During the polarization switching cycling, the difference in the Pr values of HZO films on TiN (001)/(111) from HZO on TiN (110) becomes larger. Thus, HZO films on TiN (001)/(111) exhibit an enhanced wake-up effect compared with those on TiN (110), macroscopically. Furthermore, Fig. 4f compares the relationship between Pr and O-phase content in the reported literature26,42 and this work. Despite the reported linear relationship, our results directly and clearly demonstrated that the Pr value can be controlled by domain orientations beyond the O-phase content.

Fig. 4: Electrical characterizations of Hf0.5Zr0.5O2 capacitors with different orientations of TiN substrate.
Fig. 4: Electrical characterizations of Hf0.5Zr0.5O2 capacitors with different orientations of TiN substrate.
Full size image

P-E hysteresis curves of HZO capacitors a on TiN (001), b on TiN (111), and c on TiN (110) in the pristine state and after wake-up. d The comparison of after-wake-up P-E hysteresis curves of HZO capacitors on different orientations of TiN substrate. e The wake-up evolution of HZO capacitors on different orientations of TiN substrate. f The linear relationship between the value of Pr and the O-phase content of HZO film reported in literature26,42 and in our work.

A hidden T-O phase transition pathway resulting in domain orientation tuning

Finally, we discuss how the domain orientations and Pr values are affected by the orientation of the TiN substrate. It is particularly important to understand how the (100)O-oriented domain, whose polarization has no contribution to Pr, is avoided in HZO films on TiN (001) and (111), which is critical for enhancing the Pr value. As introduced, the T-O phase transition pathway is the key to understanding. By employing synchrotron-based in-situ GIXRD investigation for HZO on TiN (001) in annealing (see section (D) of the Supplementary Note), we confirmed that the {002}O-oriented O-phase (mixture of (200)O, (020)O, and (002)O) is transformed from (110)T/(002)T-oriented T-phase, which aligns with the GIXRD and S/HRTEM results discussed earlier. To get more intuitive insights, in-situ TEM investigations in thermal annealing were further performed. As demonstrated in Fig. 5a, b, (002)T-oriented T-phase nucleated first in the heating process, observed under the microscope at 550 °C. After cooling to room temperature (RT), it transited to (020)O-oriented O-phase partly, with remaining T-phase part intergrown with transited O-phase in the same grain. Namely, a T-O transition pathway, where the longest c-axis of the T-phase transforms to the shorter bO-axis of the O-phase, is observed. Interestingly, this transition pathway is fundamentally different from the previously reported one where the longest c-axis of the T-phase transforms to the longest aO-axis of the O-phase, so we consider it a hidden T-O phase transition pathway. Moreover, the reported T-O transition from (110)T-oriented T-phase to (020)O-oriented O-phase is also observed (see Supplementary Fig. S17). The (002)O-oriented one observed in some areas in the HRTEM images may arise from this transition as well, considering the similar lattice constants of bO and cO. Therefore, through both the hidden and reported T-O phase transition pathways, the dominance of (020)O- and (002)O-oriented O-phase (i.e., (010)O- and (001)O-oriented domains) whose spontaneous polarization could completely contribute to Pr under the electric field, is achieved in O-phase areas of polycrystalline HZO on TiN (001) after annealing.

Fig. 5: Hidden T-O transition pathway in Hf0.5Zr0.5O2 on TiN (001).
Fig. 5: Hidden T-O transition pathway in Hf0.5Zr0.5O2 on TiN (001).
Full size image

a, b The in-situ HRTEM images in annealing showing the hidden T-O transition pathway from the c-axis of T-phase to the bO-axis of O-phase. Solid red lines mark grain boundaries in the film, and dashed lines highlight the T-O phase boundaries within a single grain. c The HRTEM image showing the intergrowth of T- and O-phases within a grain of HZO on TiN (001) after annealing. The amplified atom arrangement images on the right reveal periodic dislocations at the interface between HZO and TiN (001), occurring at regular intervals of every 5 HZO cells with 6 TiN cells. d The strain map calculated from diffraction spots marked by the circle. Periodic dislocations are generated at the interface with the release of strain, which contributes to local epitaxial growth of HZO on TiN (001). e Schematics of the hidden T-O transition pathway observed from the HRTEM images. The c-axis of the T-phase transforms into the bO-axis of the O-phase. Simultaneously, the a- and b-axes of the T-phase transform into cO- and aO-axes of the O-phase, respectively.

Next, we discuss how the hidden transition happens. Notably, despite the large interface lattice mismatch between TiN and HZO, epitaxial growth of T/O grains is realizable locally through periodic dislocations. The period of the interface dislocations is determined by the lattice mismatch between HZO and TiN. More specifically, as shown in Fig. 5c and the amplified atom arrangement images derived from it, the a-axis of the T-phase aligns with that of TiN at the interface, with periodic dislocations occurring every 5 double-sized T-phase unit cells with 6 TiN unit cells, forming a (002)-oriented T-phase. This is understandable because the lattice constant along the a-axis of the T-phase is about 6/5 of that in TiN. The intergrown (020)O-oriented O-phase also follows similar interface dislocations. Additionally, as shown in Fig. 5d, the strain map derived from Fig. 5c indicates that the lattice-mismatch-induced stress is primarily concentrated at the interface dislocations but released within the film. Therefore, it is considered that the interface dislocations provide tolerance, enabling the hidden phase transition inside HZO grains with the release of strain. In this way, the a-axis of the T-phase transforms to the cO-axis of the O-phase, while the longest c-axis of the T-phase transforms to the shorter bO- rather than the longest aO-axis of the O-phase, resulting in (010)O-oriented domains while avoiding (100)O-oriented ones. This is also understandable because the (100)O plane exhibits a buckling structure, with Hf/Zr atoms shifting along the out-of-plane (aO-axis) direction, which disturbs the flatness of the (100)O plane, introducing strain and raising its formation energy (see Supplementary Fig. S18). Moreover, the lattice constant along the a-axis of the T-phase is similar to that along the cO-axis of the O-phase, making this transition easier to happen.

As for HZO on TiN (111), the transition from the longest c-axis of (110)T-oriented T-phase to the shorter cO-axis of the intergrown O-phase is observed, leading to (010)O-oriented domains and avoiding (100)O-oriented ones like the case on TiN (001) (Supplementary Fig. S19b). (001)O-oriented domains are not observed, possibly due to the different space tolerance at the interface or the limited areas observed. As for HZO on TiN (110), both reported and hidden T-O transitions occur possibly, forming three kinds of (101)O-, (011)O-, and (110)O-oriented domains due to its different interface structure (Supplementary Fig. S19c).

In addition, considering the wake-up difference on three orientations of TiN substrates, our recent study demonstrated that the T-O phase transition continues in polarization switching cycling, resulting in a wake-up effect from AFE-like characteristics to FE properties64. S/HRTEM images of HZO films on TiN (001) after wake-up reveal a significant increase in (020)O/(002)O-oriented O-phase while a reduction in (110)T/(002)T-oriented T-phase, confirming a substantial T-O transition during wake-up (Supplementary Fig. S21), consistent with the P-E loop changes in Fig. 4. T-O phase transition under polarization switching cycling is like a continuation of T-O phase transition in thermal annealing. Both should happen through the pathway with the smallest energy barrier. So, we consider that a similar T-O phase transition pathway also appears in the wake-up of HZO, which is indicated by the intergrowth grain in TEM images. Since the T-phase content is very high and the T-O phase transition is not sufficient before wake-up, the Pr difference is not obvious in the pristine state. While the T-O phase transition becomes much more sufficient after wake-up in polarization switching, the Pr difference becomes much more obvious.

In conclusion, the control of (001)O- and/or (010)O-oriented domains in O-phase grains of polycrystalline HZO, whose spontaneous polarization can contribute to Pr completely, has been achieved along with enhanced Pr by employing TiN (001) and TiN (111) single-crystalline substrates. This achievement benefits from a hidden transition from the longest c-axis of the T-phase to the shorter bO- or cO-axis of the O-phase in intergrown T/O grains assisted by the periodic interface dislocations in addition to reported one from the longest c-axis of the T-phase to the longest aO-axis of O-phase. These findings shed light on governing the domain orientations and FE properties of HZO by controlling the interface dislocations.

Methods

Device fabrication

(1) Synthesis of TiN bottom electrodes: epitaxial TiN thin films were deposited on (001)-, (110)-, and (111)-oriented MgO substrates by a homemade magnetron sputtering epitaxy system using a 2-inch Ti (purity of 99.995%) target and the N2 (purity of 99.999%) reactive gas. The base vacuum pressure of the growth chamber is 3 × 10−8 Torr. The TiN films were deposited at a temperature of 1100 °C under a pure nitrogen pressure of 0.02 Torr. During the deposition, all three MgO substrates were mounted on a rotatable SiC absorber heated by a laser. The temperature of MgO substrates was monitored at the back side of the SiC absorber using an infrared pyrometer. To ensure uniformity, the sample holder was rotated at a speed of 5 rpm throughout the deposition. The radio frequency (RF) generator power was kept at 100 W throughout the entire deposition process.

(2) Growth of HZO films: HZO films were deposited on TiN bottom electrodes using thermal ALD (TFS200 Beneq Thermal/Plasma Enhanced Atomic Layer Deposition) at a substrate temperature of 250 °C. Hf(N(CH3)2)4, Zr(N(CH3)2)4, and H2O were used as the Hf precursor, Zr precursor, and oxygen source, respectively. The growth rates per cycle of HfO2 and ZrO2 were nearly identical (0.09 and 0.07 nm/cycle, respectively). Therefore, HZO films with a 0.5:0.5 Hf: Zr ratio could be deposited by using a 1:1 HfO2:ZrO2 ALD cycle ratio, with a 0.5 s H2O dose time in each cycle.

(3) Growth of top electrodes: first, 30-nm-thick poly-TiN top electrodes were deposited via reactive magnetron sputtering at room temperature through a hard shadow mask with three different hole sizes (~6800, 11,800, 38,000 μm2). The base pressure was reduced to 4 × 10−4 Pa, the sputter power was set to 250 W, and the gas atmosphere consisted of Ar: N2 in a ratio of 50:4 sccm. Then, 30-nm-thick W electrodes were deposited without removing the hard mask, the sputter power for W is 80 W and the atmosphere is Ar = 30 sccm. After the top electrode deposition, post-metallization annealing (PMA) was performed for 30 s at 550 °C in an N2 atmosphere to promote the crystallization of the HZO films.

Electrical characterizations

The electrical measurements were carried out using a Cascade Microtech Summit 11000 B-S Semi-automated Probe Station, which is a commonly used apparatus for such purposes in the field. During all electrical measurements, a bias voltage was applied to the top electrode, and the bottom electrode was grounded. Polarization-electric field curves were measured utilizing a semiconductor parameter analyzer (Keithley 4200A-SCS) operating at a frequency of 1 kHz. Meanwhile, endurance tests were conducted at a frequency of 250 kHz with an applied electric field of 3.5 MV/cm.

Structural characterizations

(1) GIXRD characterizations: the 2θ-ω scans were conducted using a high-resolution X-ray diffractometer (Bruker D8 Discovery). Monochromatic Cu Kα radiation with a wavelength of 1.5406 Å was utilized for this purpose. The GIXRD measurements were carried out at the 02U2 beamline situated within the Shanghai Synchrotron Radiation Facility (SSRF), where a fixed energy of 10 keV (λ = 1.24 Å) and an incidence angle of 3° were employed. 2θ of high energy X-ray (λ1 = 1.24 Å) is converted into Cu-Kα form (λ2 = 1.54 Å) based on λ1/sinθ1 = λ2/sinθ2 for a convenience. For in-situ 2D-GIXRD investigation during the thermal annealing process, crystalline information was captured by a MarCCD 225 detector every 3 s during annealing in N2. The thermal annealing procedure consisted of three distinct stages: heating at a controlled rate of 20 °C/min, holding the temperature at 550 °C for 5 min, and cooling at a 30 °C/min rate. The 2D diffraction images obtained were processed and converted into one-dimensional XRD data using Fit2D and Dioptas software, enabling detailed analysis of the structural evolution during the annealing process.

(2) Raman spectroscopy: Raman measurements were carried out at room temperature with a Raman microscope (Renishaw inVia Qontor), using a 532 nm wavelength laser, 100× objective lens, and 100 s total measurement time per spectrum.

(3) TEM characterizations: STEM samples were prepared using a focused ion beam with Ga+ ions to a thickness of ~30 nm. Scanning transmission electron microscopy (STEM) and High-Resolution Transmission Electron Microscopy (HRTEM) imaging were performed on a Thermo Scientific Themis Z microscope operating at 300 keV and equipped with a spherical aberration corrector for the condenser lens. The atomic-resolution high-angle annular dark-field (HAADF) STEM image was collected with a 25-mrad convergent angle and a collection angle of 41–200 mrad. The analysis of the FFT patterns was employed to determine the phase of HZO. The in-plane strain map (εxx) was calculated using the Geometric Phase Analysis (GPA) method. A PicoFemto in-situ MEMS heating TEM holder was employed to investigate the temperature-dependent structural evolution of the HZO film using high-resolution imaging. The sample was heated from RT to 550 °C at a rate of 20 °C per minute, held at 550 °C for 1.5 h, and then cooled down to RT at a rate of 30 °C per minute.