Ferroelectric and field-induced ferroelectric phase formation in atomic-layer-deposited ZrO₂ thin films with TiN electrodes

Abstract

This study investigates the origin of ferroelectric (FE) phase formation in 10-nm-thick atomic-layer-deposited ZrO₂ thin films by examining their structure and electrical properties as a function of deposition temperature (T_dep, 280–310 °C) and post-metallization annealing (PMA) steps. Polarization measurements revealed only field-induced ferroelectric (FFE) properties in the films deposited at 280 °C. However, the FE-like properties were observed in the films deposited at T_dep = 285–310 °C, with double remnant polarization (2P_r) values of ∼15 μC cm⁻² at 310 °C, where the films exhibited improved crystallinity, larger grain size, and higher oxygen density. However, PMA led to increased oxygen deficiency, causing FE characteristics to vanish, leaving only FFE in the films. In addition, the wake-up process does not correspond to the permanent phase transition from the non-polar tetragonal phase to the polar orthorhombic phase. Rather than that, the internal field, induced by the asymmetric oxygen vacancy distribution across the film thickness, undergoes disparate evolution with the cycling in the FE and FFE regions in the film. The overall FE-like property of the woken-up state corresponds to the merged FE and FFE properties, which are also influenced by the applied electric field (E_app) strength during the cycling and property measurements.

---

1. Introduction

The ZrO₂ thin film is a material that exhibits various phases, including monoclinic (M), tetragonal (T), polar orthorhombic (PO), and cubic (C).¹ It exhibits the highest dielectric constant (k) of ∼40 in the T-phase with a band gap of ∼5 eV, making it suitable for use as a capacitor material in dynamic random access memory (DRAM) for an extended period.^2,3 The matured atomic layer deposition (ALD) technology of the ZrO₂ film has also contributed to its commercial application in DRAM. Moreover, it is feasible to suppress the leakage current even at extremely thin thicknesses (<5 nm) by adopting Al₂O₃ and Y₂O₃ interlayers or doping ZrO₂ with Al and Y, thereby appropriately addressing technological challenges for future DRAM fabrication.⁴

The ZrO₂ thin film at the nanoscale stabilizes the T-phase due to the influence of surface energy effects.^5–7 It has been elucidated that it possesses FFE characteristics, transitioning from a non-polar T-phase to a PO-phase at a critical field (E_crit) of approximately 1–2 MV cm⁻¹, as determined through various theoretical and experimental studies.^6,8–10 The PO-phase is characterized by the remnant polarization (P_r), which can be switched by an E_app higher than the coercive field (E_c). FFE properties can enhance the reversible charge storage density in the DRAM technology. However, decreasing E_crit and suppressing the hysteresis properties induced by the lower E_crit when decreasing the field, rather than when increasing the field, are still required.

Besides, the stable PO-phase formation in the ZrO₂ film is undesirable for DRAM applications because P_r cannot be used for DRAM operation. Nonetheless, stable PO-phase ZrO₂ might be useful for FE device applications that adopt the P_r, such as ferroelectric random access memory, ferroelectric field effect transistors, and ferroelectric tunnel junctions, due to its simple composition compared to the Hf_0.5Zr_0.5O₂ (HZO) thin film.^9,10 However, it has been extensively reported that pure ZrO₂ mostly has FFE properties due to the dominant T-phase formation, especially in the case of the ALD films, from the early stage of the related study.

Nonetheless, it has also been reported that the pure ZrO₂ film can possess stable FE properties. Starschich et al. showed that a 100-nm-thick pure ZrO₂ film deposited by chemical solution deposition has a maximum P_r of ∼12 μC cm⁻².¹¹ Chae et al. reported that a 10-nm-thick ZrO₂ film deposited by a thermal ALD on a SiO/Si substrate had a larger grain size when compared to a similar ZrO₂ film deposited on TiN substrates, and larger grain size manifested FE characteristics.¹² Xu et al. suggested that the increase in O₃ dose during O₃-based ALD decreased the in-plane tensile stress and V_O content, facilitating the transition from the T-phase to the PO-phase through field cycling.¹³ These results indicate that grain size, stress/strain, and V_O concentration influence PO-phase formation in pure ZrO₂ films. It has been extensively reported that the smaller grain size (larger surface or grain boundary area), higher tensile strain, and higher V_O concentration favor the T-phase in the HfO₂–ZrO₂ fluorite thin film system. Therefore, studying how these factors change by process conditions is essential.

In the ALD process, the substrate temperature during thin film deposition provides the energy needed to remove the ligand from the precursor and for the migration of adatoms, which can influence the presence of impurities, crystallinity,^14,15 grain size,^15–17 and stress/strain.^18,19 Shibayama et al. showed FFE characteristics in ZrO₂ thin films deposited on heavily doped p-type Ge(001) substrates by sputtering at a T_dep of room temperature but showed FE characteristics at a T_dep of 300–500 °C. Interestingly, the crystal structure analysis using X-ray diffraction showed that the M-phase fraction began to appear at T_dep ∼ 300 °C and gradually increased with increasing T_dep. These results suggest that if M-phase nucleation is promoted with increasing T_dep, the PO-phase may appear during the transition from the T-phase to the M-phase.¹⁶ Meanwhile, Xu et al. deposited ZrO₂ films on TiN substrates using the ALD method. They demonstrated that with an increase in T_dep from 250 °C to 350 °C, the interfacial layer thickens, and the in-plane tensile strain decreases, which can significantly enhance the ferroelectric properties of the ZrO₂ thin film at T_dep = 350 °C after field cycling.¹⁸ These previous results indicate that the T_dep may play a crucial role in determining the preferred phase formation and accompanying FE or FFE properties. In this study, therefore, the TiN top electrode (TE)/ZrO₂/TiN bottom electrode (BE) capacitor stack was prepared by a thermal ALD using cyclopentadienyl tris(dimethylamino) zirconium (Cp–Zr) as the Zr precursor and O₃ as the oxygen source, which guarantees stable ALD at T_dep from 280 to 310 °C from the ALD temperature window as shown in Fig. S1 in the ESI,† which was unfeasible with the commonly adopted amine-based precursors, such as tetra(ethylmethylamino)zirconium (TEMAZ), which thermally decomposes at these temperatures. Shin et al. recently reported that the HZO film can exhibit feasible FE performance without the need for a wake-up process, even at a film thickness of 5 nm, when the HZO film is grown at 310 °C. This improvement was ascribed to the appropriately controlled in-plane tensile stress and larger grain size with higher film density.²⁰ Therefore, it can be expected that the ALD ZrO₂ film grown at T_dep = 310 °C may show a practical FE performance even without any cation doping. This work also examines the influence of the PMA treatment to confirm the effect of the thermal process experienced in the back-end-of-line (BEOL) after the capacitor process. In this study the as-deposited samples were named C280, C285, C290, and C310 when grown at 280 °C, 285 °C, 290 °C, and 310 °C, respectively. The C310 sample after the PMA treatment was named C310_PMA.

2. Experimental section

2.1. Sample preparation

A TiN BE with a thickness of 50 nm was deposited on a Si (100) substrate using radio frequency (RF) sputtering (SRN120, Sorona INC.). 10-nm-thick ZrO₂ films were deposited using a 12-inch-diameter thermal ALD system (Guidance Series, Jusung Engineering). The liquid Cp–Zr precursor was vaporized directly from the vaporizer through a liquid delivery system and injected into the ALD chamber with an N₂ carrier gas flow of 1000 standard cubic centimeters per minute (SCCM). O₃ with a concentration of 260 g m⁻³ was used as an oxygen source. The optimized Zr-source flow–purge–O₃ flow–O₃ purge times were 4–4–4–6 s. The ZrO₂ thin film in this experiment had a growth per cycle (GPC) of ∼0.10 nm per cycle at deposition temperatures ranging from 280 to 310 °C.

A TiN TE was deposited on a 10-nm-thick ZrO₂ thin film using RF sputtering (SRN120, Sorona INC.) with a metal shadow mask (hole diameter: 360 μm). PMA was performed through a rapid thermal annealing (RTA) process at 400 °C for 3 minutes under an N₂ ambient condition.

2.2. Electrical, structural, and chemical characterization

Polarization–electric field (P–E) and current–electric field (I–E) characteristics were measured with a ferroelectric tester (TF Analyzer 2000, aixACCT Systems). The dielectric constant was obtained from the measured capacitance–voltage characteristics using an impedance analyzer (4194A, HP) at 10 kHz AC frequency. The voltage was applied to the TE, while the BE was grounded.

The thickness of the ZrO₂ film was measured by energy dispersive X-ray fluorescence spectroscopy (EDXRF) (Quant’X, Thermo Scientific). For crystal structural analysis, grazing incidence X-ray diffraction (XRD) analysis was conducted using an X-ray diffractometer (X'Pert Pro, PANalytical) with an incident angle of 0.5°. The microstructure of the films was analyzed using field emission scanning electron microscopy (SEM) (SUPRA 55VP, Carl Zeiss). A grain size distribution analysis was conducted using the watershed method implemented in the Gwyddion software. X-ray photoelectron spectroscopy (XPS) analysis (Axis Supra+, Kratos) was performed to investigate the chemical states of the ZrO₂ and TiN TE films. Secondary ion mass spectrometry (SIMS) (ION-TOF, TOFSIMS-5) and cross-section transmission electron microscopy (TEM) (JEOL, JEM-ARM200F) were performed to analyze the interface status and film microstructure. The density of the ZrO₂ films on the TiN BE was analyzed by X-ray reflectometry (XRR) (X’Pert Pro, PANalytical). The areal oxygen density was analyzed via wavelength dispersive X-ray fluorescence spectroscopy (WDXRF) (Bruker, S8 TIGER).

3. Results and discussion

3.1. Electrical properties of ZrO₂ films

To prevent potential confusion in the subsequent discussion, the terminology used to describe the double hysteresis behavior observed in the P–E curves is clarified. In particular, a distinction is made between FFE behavior and intrinsic antiferroelectric (AFE) characteristics.

According to Kittel, an AFE state is defined as one in which ions within a crystal exhibit spontaneous polarization, but neighboring ions are aligned with opposing dipole moments, resulting in a net macroscopic polarization of zero in the absence of an external electric field.^10,21 In HZO and ZrO₂ systems, AFE behavior is generally attributed to the formation of an anti-polar orthorhombic (APO) phase. In contrast, FFE behavior in these systems is typically understood as arising from a phase transition from a non-polar tetragonal (T) phase to a polar orthorhombic (PO) phase when the applied electric field exceeds a critical threshold.

Although the physical origins of AFE and FFE are distinct, FFE behavior often appears as a double hysteresis loop, resembling that of intrinsic AFE materials. As such, it can be not easy to distinguish between them based solely on P–E or I–E characteristics. However, prior studies^22–27 have reported that the dielectric permittivity of tetragonal ZrO₂ is typically in the range of 40–47, which is notably higher than that of the PO (∼26) and APO (∼20–27) phases. Therefore, dielectric permittivity may serve as an additional parameter to differentiate between AFE and FFE behavior in ZrO₂-based films. Supporting this, Fig. S2(a) in the ESI† shows the dielectric permittivity of the C280, C285, C290, C310 and C310_PMA samples. The C310 and C310_PMA films exhibit permittivity values of approximately 37 and 44, respectively—both considerably higher than those generally associated with orthorhombic phases. Accordingly, the term “AFE” is excluded in this study, and all related phenomena are described using the term “FFE” for consistency.

Fig. 1(a) shows the schematic diagram of the dynamic hysteresis measurement (DHM) used to estimate the FFE and FE properties in this study. DHM consists of four bipolar triangular pulses, and in this study, the measurement frequency was set to 1 kHz. The first and third pulses define positive and negative polarization states, respectively. The second and fourth pulses start 1 second after the relaxation period of the previous pulse, respectively, resulting in a loss of P_r. Therefore, only the data between the +_up and +_down sections of the second pulse and the −_down and −_up sections of the fourth pulse are collected. Fig. 1(b) and (c) show the I–E and P–E curves of a typical FFE material, respectively. Fig. 1(d) and (e) illustrate the same for typical FE materials. In the I–E and P–E curves, forward and reverse indicate the switching direction. As described in Section 3.2, the amorphous ratio increases as T_dep decreases in the as-deposited films, so a larger field must be applied to confirm the switching characteristics.

Fig. 1 (a) Bipolar triangular voltage pulse for DHM. I–E (b) and P–E (c) curves of FFE characteristics obtained from each pulse of DHM. I–E (d) and P–E (e) curves of FE characteristics obtained from each pulse of DHM. In the I–E and P–E curves, forward and reverse indicate the direction of switching.

On the other hand, when T_dep is high or PMA is performed, a high leakage current issue may occur. Therefore, the pulse conditions (height and duration) for the DHM must be carefully selected to prevent breakdown during field cycling while maintaining well-distinguished electrical characteristics for each film. The amplitude of the triangular pulse at field cycling to obtain the woken-up state and the amplitude of the field at P–E curve measurements remained identical.

In this section, I–E and P–E curves in pristine states measured by DHM were first analyzed according to T_dep and PMA treatment. Then, simulation results for cases where each curve could appear were analyzed to predict the phase of the ZrO₂ thin film, depending on the T_dep in both pristine and PMA-treated states. Next, the I–E and P–E curves in the woken-up state were analyzed according to T_dep and PMA treatment. Finally, the evolution of I–E curves during field cycling was discussed to determine how the T_dep and PMA treatments affect the phase formation of each ZrO₂ film.

3.1.1. I–E and P–E curves of ZrO₂ films in their pristine state. Fig. 2(a)–(e) show I–E and P–E curves of pristine state C280, C285, C290, C310 and C310_PMA films. Table 1 shows the fields (E(+_up), E(+_down), E(−_down), and E(−_up)) where the switching peak appears in each pulse area (+_up, +_down, −_down, and −_up) for these films. It also highlights the difference between the fields in which forward and reverse switching peaks occur, depending on the field direction, denoted as |E(+_up) − E(+_down)| and |E(−_down) − E(−_up)|.

Fig. 2 Pristine state I–E and P–E curves of 10-nm-thick (a) C280, (b) C285, (c) C290, (d) C310, and (e) C310_PMA thin films. (DHM condition: 3.2 MV cm⁻¹, 1 kHz for the C280 film, 2.6 MV cm⁻¹, 1 kHz for C285, C290, and C310 films, and 3.0 MV cm⁻¹, 1 kHz for C310_PMA.) The switching peaks are marked with a blue arrow, along with the field values.

Table 1 Electrical properties of C280, C285, C290, C310 and C310_PMA films in the pristine state

Samples	E(+up) [MV cm^−1]	E(+down) [MV cm^−1]	|E(+up) − E(+down)| [MV cm^−1]	E(−down) [MV cm^−1]	E(−up) [MV cm^−1]	|E(−down) − E(−up)| [MV cm^−1]
C280	∼2.91	∼0.48	∼2.43	—	—	—
C285	∼2.47	∼0.34	∼2.13	∼−1.53	∼−0.34	∼1.19
C290	∼2.26	∼0.18	∼2.08	∼−1.71	∼−0.65	∼1.06
C310	∼2.26	∼0.10	∼2.16	∼−1.71	∼−0.65	∼1.06
C310PMA	∼2.82	∼0.36	∼2.46	—	—	—

---

Fig. 2(a) shows the I–E and P–E curves of the C280 film measured with an amplitude of 3.2 MV cm⁻¹ and a frequency of 1 kHz, and Fig. 2(b)–(d) show the similar curves of the C285 C290 and C310 films measured with an amplitude of 2.6 MV cm⁻¹ and the same frequency. In C280, C285 and C290 films, the switching performance in the positive field region gradually increased as T_dep increased. In the C310 film, the switching characteristics appeared similar to those of the C290 film. This tendency was consistent with the E(+_up) values of C280, C285, C290, and C310 films, which appeared at ∼2.91, 2.47, 2.26, and 2.26 MV cm⁻¹, respectively. On the other hand, starting with the C285 film, the switching characteristics of the negative field region became apparent, gradually increasing with the increase in T_dep. In the C310 film, the switching characteristics in the negative field region were more prominent than in the positive field region. In the C285, C290, and C310 films, the switching peaks appeared at E(−_down) values of ∼−1.53, ∼−1.71, and ∼−1.71 MV cm⁻¹, respectively, but they were significantly smaller than E(+_up).

For the ideal FFE, a pair of forward and reverse switching peaks appear in the positive and negative field regions of the I–E curve, respectively, with pinched hysteresis shapes. Thus, the I–E and P–E curves of the C285, C290, and C310 films resemble those of FFE. Additionally, for the ideal FFE, the difference between fields in which forward and reverse switching occur in the positive and negative field regions, i.e., |E(+_up) − E(+_down)| and |E(−_down) − E(−_up)|, should be similar. However, as shown in Table 2, in C285, C290, and C310 films, |E(+_up) − E(+_down)| is ∼2.08–2.16 MV cm⁻¹, whereas |E(−_down) − E(−_up)| is ∼1.06–1.19 MV cm⁻¹, showing a significant difference of about twice. Therefore, the I–E and P–E curves of the C285, C290, and C310 films in their pristine state are challenging to explain in terms of FFE behavior. This work proposes a model that combines the imprinted FE and FFE behaviors, as discussed below. Fig. 2(e) shows the I–E and P–E curves of the C310_PMA film measured by DHM at 3.0 MV cm⁻¹ (1 kHz). The switching current peak appeared at a E(+_up) and E(+_down) of ∼2.82 MV cm⁻¹ and ∼0.36 MV cm⁻¹, respectively. However, as in the C280 film, the switching characteristics of the negative field voltage region, seen in the C310 film, disappeared after the PMA treatment.

Table 2 Notation that defines the characteristics of FE and FFE

Change in polarization	Polarization				Switching field
	Down	Down → 0	Up	Up → 0	Down	Down → 0	Up	Up → 0
FE	P PO↓	—	P PO↑	—	E PO↓	—	E PO↑	—
FFE	P T→PO↓	P PO↓→T	P T→PO↑	P PO↑→ T	E T→PO↓	E PO↓→T	E T→PO↑	E PO↑→T

---

The I–E and P–E curves of the ZrO₂ films in their pristine state, shown in Fig. 2, exhibited two distinct characteristics. First, the switching characteristics are more prominent in one direction of the field, and second, the difference between the fields in which forward and reverse switching occur is greater in the positive field area than in the negative field area (i.e., |E(+_up) − E(+_down)| ≫ |E(−_down) − E(−_up)|). The following section discusses the characteristics of the pristine state of each ZrO₂ film, assuming the coexistence of the (imprinted) FE and FFE phases.

3.1.2. Phase prediction of the pristine state through simulation. The shifted I–E and P–E curves along the E direction in the pristine state film, shown in the previous section, could be related to the internal bias field (E_int) formed inside the thin film.^28–34 In addition, for the ZrO₂ of the T-phase grain, switching occurs through field-induced transition, so a larger field of >∼1–2 MV cm⁻¹ may be required for transition.^6,8 Therefore, under the influence of E_int, (1) FE, (2) FFE, and (3) mixed states (FE + FFE) were assumed to understand the complex I–E curves in the pristine state. In addition, the I–E curves in Fig. 3(a)–(f) for these cases were obtained using Miller's analytic equations.^35,36

Fig. 3 Schematic I–E curves according to E_int for the case of (a) FE, (b) FFE, and (c) FE and FFE at a DHM pulse amplitude of 4.0 MV cm⁻¹. For FE, E_PO↓ and E_PO↑ were applied at 0.5 MV cm⁻¹ and −0.5 MV cm⁻¹, respectively. For FFE, E_T→PO↓, E_PO↓→T, E_T→PO↑, and E_PO↑→T were applied at 3.0 MV cm⁻¹, 1.0 MV cm⁻¹, −3.0 MV cm⁻¹, and −1.0 MV cm⁻¹, respectively. I–E curves under the influence of E_int = +1.0 MV cm⁻¹ for the cases of (d) FE, (e) FFE, and (f) FE and FFE mixing at a pulse amplitude of 2.6 MV cm⁻¹ in DHM. In each figure, the forward switching peak of FE is shown as P_PO↓, the reverse switching peak is shown as P_PO↑, the forward switching peaks of FFE are shown as P_T→PO↓, P_T→PO↑, and the reverse switching peaks are shown as P_PO↓→T and P_PO↑→T.

The notation is defined in Table 2, taking into account the phase and polarization states to prevent confusion regarding the switching characteristics in FE and FFE. For the PO-phase of FE, when the polarization state is switched in the down (up) direction, the state of polarization is defined as P_PO↓ (P_PO↑), and the field is defined as E_PO↓ (E_PO↑). In the case of FFE, the field-induced transition from the T-phase to the PO-phase results in switching characteristics, so the state of polarization in the down (up) direction is P_T→PO↓ (P_T→PO↑), and the field is defined as E_T→PO↓ (E_T→PO↑). Conversely, the polarization state when the transition from the down (up) state of the PO-phase to the T-phase occurs is defined as P_PO↓→T (P_PO↑→T), and the field at this moment is defined as E_PO↓→T (E_PO↑→T).

Fig. 3(a)–(c) show the expected I–E curves for FE, FFE, and FE + FFE under E_int. E_PO↓, E_PO↑, E_T→PO↓, E_PO↓→T, E_T→PO↑, and E_PO↑→T were assumed to be 0.5, −0.5, 3.0, 1.0, −3.0, and −1.0 MV cm⁻¹, respectively. In the I–E curves of Fig. 3(a)–(c), a sufficiently large amplitude of DHM of 4.0 MV cm⁻¹ was assumed to illustrate how each switching peak shifts according to E_int. Fig. 3(a) shows the shift of I–E curves for E_int = 0, +0.5, and +1.0 MV cm⁻¹ for the case where the thin film has only FE properties. E_int in the + direction indicates the occurrence of E_int from TE to BE, shifting the I–E curve in the negative bias direction. For the black curve under E_int = 0 MV cm⁻¹, switching for P_PO↓ was observed at E_PO↓ = 0.5 MV cm⁻¹, and switching for P_PO↑ was observed at E_PO↑ = −0.5 MV cm⁻¹, indicating a symmetrical shape for the zero field axis. However, the curve shifts in the negative bias direction according to the magnitude of E_int (red and yellow curves) while maintaining the overall switching peak shapes unchanged.

Fig. 3(b) shows the shift of I–E curves for E_int = 0, +0.5, and +1.0 MV cm⁻¹ when the thin film has only FFE properties. The I–E curve shifts in the negative field direction as E_int increases, as in the case of FE in Fig. 3(a). However, because the measurement pulse amplitude was assumed to be 4 MV cm⁻¹, for E_int = +0.5 MV cm⁻¹ and +1.0 MV cm⁻¹, not all of the T-phase was transitioned to PO_↑ during the −_down pulse, and some portion remained in the T-phase. The partial peak of the red and yellow curves near −4 MV cm⁻¹ indicates this limited transition, and, thus, the P_PO↑→T peak from PO_↑ back to the T-phase (at −1.5 and −2.0 MV cm⁻¹ for the red and yellow curves) was smaller than that in the case of E_int = 0 MV cm⁻¹. Fig. 3(c) shows the I–E curve of the FE + FFE state, which is the combination of the curves in Fig. 3(a) and (b).

Fig. 3(d)–(f) show the corresponding I–E curves for the cases illustrated in Fig. 3(a)–(c), respectively, under an E_int of +1.0 MV cm⁻¹ and a DHM pulse amplitude limited to 2.6 MV cm⁻¹. Since forward switching and reverse switching of FE thin films occurred at E_PO↓ = −0.5 MV cm⁻¹ and E_PO↑ = −1.5 MV cm⁻¹, respectively, the forward switching and reverse switching peaks corresponding to P_PO↓ and P_PO↑ could be observed, as shown in Fig. 3(d). On the other hand, since switching in the positive and negative field areas of the FFE thin film with E_int = +1.0 MV cm⁻¹ occurred at E_T→PO↓ = 2.0 MV cm⁻¹ and E_T→PO↑ = −4.0 MV cm⁻¹, respectively, the switching peak corresponding to P_T→PO↑ could not be observed. Because the P_T→PO↑ switching has not occurred during the sweep down to −2.6 MV cm⁻¹, the P_PO↑→T peak, which is supposed to appear at −2.0 MV cm⁻¹ during the sweeping back to the zero field, also does not appear. In this case, even though the sample exhibits the FFE property, it can appear as an FE curve with a negative E_int (or shifted in the positive field direction), as shown in Fig. 3(e). Fig. 3(f) shows the I–E curve for thin films with mixed FE and FFE with E_int = +1.0 MV cm⁻¹. The curve is a combination of the curves in Fig. 3(d) and (e), exhibiting an asymmetric FFE-like shape. Therefore, it can be noted that the switching I–E curves can have diverse shapes depending on the E_app region and the composition of the involved phases. Next, the experimental I–E curves in Fig. 2 are compared with the schematic I–E curves in Fig. 3(d)–(f).

The I–E curves in Fig. 2(a) C280 and Fig. 2(e) C310_PMA are similar to those in Fig. 3(e), indicating that they exhibit FFE properties rather than the positively shifted FE property. The I–E curves in Fig. 2(b) C285, Fig. 2(c) C290, and Fig. 2(d) C310 are similar to those in Fig. 3(f), indicating that they mostly have FE + FFE phases. The evident peak positions in the I–E curve of C310 render the E_int (E_int↓) identification feasible, estimated from the |E(−_down) − E(−_up)| value in Table 1. It was ∼1.06 MV cm⁻¹, and a similar value is estimated for C290.

Despite the apparent similarity to conventional switching behavior, the FFE characteristics were further examined by applying differential high-magnitude (DHM) pulses in the negative field region, incrementally increasing the field amplitude until dielectric breakdown occurred. This approach was employed to better resolve the coexistence of FE and FFE states. Fig. S3 (ESI†) presents the I–E curves of the 10-nm-thick C310 film as a function of the applied electric field. At an applied field of 3.6 MV cm⁻¹, the forward P_T→PO↑ switching peak was partially obscured by leakage current; however, a subtle reverse P_PO↑→T peak emerged near 1.4 MV cm⁻¹. As the field increased to 4.2 MV cm⁻¹, this reverse peak became more pronounced. The difference in switching fields associated with the FFE component was approximately 2.2 MV cm⁻¹ in the negative direction and 2.3 MV cm⁻¹ in the positive direction, indicating symmetry in the field response. The resulting I–E profile closely resembles that shown in Fig. 3(c), providing strong evidence that samples C285 through C310 exhibit mixed FE and FFE switching characteristics.

Table 3 shows the expected phase, electrical properties, and E_int↓ for ZrO₂ films in their pristine state. C280 and C310_PMA films show FFE characteristics in the T-phase, while C285, C290, and C310 films are expected to have mixed FE and FFE characteristics in a mixture of the T- and PO-phases simultaneously. E_int in the positive direction was formed inside these thin films, and the expected values from the I–E curves of the C285, C290, and C310 films were ∼1.19, ∼1.06, and ∼1.06 MV cm⁻¹, respectively.

Table 3 Expected phase, electrical properties, and E_int of 10-nm-thick ZrO₂ in the pristine state

Samples	Expected phase and properties		E int↓ [MV cm^−1]
C280	T-phase	FFE	—
C285	T-phase + PO-phase	FE + FFE	∼1.19
C290	T-phase + PO-phase	FE + FFE	∼1.06
C310	T-phase + PO-phase	FE + FFE	∼1.06
C310PMA	T-phase	FFE	—

---

3.1.3. I–E and P–E curves of the ZrO₂ films in the woken-up state. Fig. 4(a)–(e) show the I–E and P–E curves of 10-nm-thick C280, C285, C290, C310 and C310_PMA films, respectively, in woken-up states. Due to the different switching properties of the samples, the wake-up field cycling conditions were optimized for each case, as shown in the figure caption. Table 4 shows the switching fields (E(+_up), E(+_down), E(−_down), and E(−_up)) where forward or reverse switching peaks appear in each pulse area (+_up, +_down, −_down, and −_up) for C280 and C310_PMA films. It also illustrates the difference between the fields in which forward and reverse switching peaks occur, depending on the field direction, through |E(+_up) − E(+_down)| and |E(−_down) − E(−_up)|. Table 5 displays the fields (E(+_up) and E(−_down)) at which forward or reverse switching peaks occur in each pulse area (+_up and −_down) for C285, C290, and C310 films. It also shows the 2P_r values.

Fig. 4 Woken-up state I–E and P–E curves of 10-nm-thick (a) C280, (b) C285, (c) C290, (d) C310, and (e) C310_PMA thin films (field cycling conditions: 3.2 MV cm⁻¹, 100 kHz, 4.6 × 10⁶ cycling for the C280 film, 2.6 MV cm⁻¹, 100 kHz, 10⁷ cycling for C285, C290, and C310 films, and 3.0 MV cm⁻¹, 100 kHz, 10⁷ cycling for C310_PMA). The switching peaks on each curve are marked with a blue arrow, along with the field values.

Table 4 Electrical properties of C280 and C310_PMA films in the woken-up state

Samples	E(+up) [MV cm^−1]	E(+down) [MV cm^−1]	|E(+up) − E(+down)| [MV cm^−1]	E(−down) [MV cm^−1]	E(−up) [MV cm^−1]	|E(−down) − E(−up)| [MV cm^−1]
	Forward switching	Reverse switching		Forward switching	Reverse switching
C280	∼2.24	∼0.35	∼1.89	∼−2.46	∼−0.67	∼1.79
C310PMA	∼2.04	∼0.15	∼1.89	∼−2.24	∼−0.48	∼1.76

---

Table 5 Electrical properties of C285, C290, and C310 films in the woken-up state

Samples	E(+up) [MV cm^−1]	E(−down) [MV cm^−1]	2Pr [μC cm^−2]
	Forward switching	Reverse switching
C285	∼1.38	∼−0.29	∼3.5
C290	∼1.22	∼−0.31	∼9.6
C310	∼0.80	∼−0.57	∼15.1

---

Fig. 4(a) and (e) show the I–E and P–E curves of the C280 and C310_PMA films after field cycling under the conditions of 4.64 × 10⁶ cycles at 3.2 MV cm⁻¹ and 100 kHz and 10⁷ cycles at 2.8 MV cm⁻¹ and 100 kHz, respectively. For the C280 film, the data after 4.64 × 10⁶ cycles were used due to increased leakage current observed at 10⁷ cycles, as shown in Fig. S4(a) and (b) (ESI†). In contrast to the switching characteristics observed only in +_up and +_down pulses in the pristine state, the I–E and P–E curves in the woken-up state exhibited switching characteristics in all pulse areas, including +_up, +_down, −_down, and −_up. The C280 and C310_PMA films show E(+_up) of ∼2.05 and ∼2.04 MV cm⁻¹ and E(−_down) of ∼−2.40 and ∼−2.24 MV cm⁻¹, respectively. The E(+_down) and E(−_up) values of C280 and C310_PMA are ∼0.16 and ∼0.15 MV cm⁻¹ and ∼−0.54 and ∼−0.48 MV cm⁻¹, respectively. Therefore, the field differences between forward and reverse switching in the positive and negative field regions were similar; i.e., |E(+_up) − E(+_down)| is ∼1.89 MV cm⁻¹ for C280 and ∼1.89 MV cm⁻¹ for C310_PMA, and |E(−_down) − E(−_up)| is ∼1.86 MV cm⁻¹ for C280 and ∼1.76 MV cm⁻¹ for C310_PMA. This finding suggests that the C280 and C310_PMA films have the T-phase even after the wake-up step, as expected in the previous section. However, it was noted that the woken-up state also shows the switching peak in the negative field region, and E(+_up) in the positive field area is lower than that in the pristine state. The former can be explained by mitigating the E_int↓ intensity through field cycling.^37,38 However, when E_int↓ is mitigated, the switching peak of the positive field region must appear at higher fields. In contrast, the experiments show an opposite trend, requiring a different interpretation.

Previous studies have reported that field cycling lowers the E_crit of FFE films with a fluorite structure due to the involvement of defect redistribution, de-pinning, or the occurrence of reversible phase changes when a sufficiently high electric field is applied during cycling.^9,10,39–41 In addition, Shin et al. reported that the switching peak decreased from 3.9 to 2.2 MV cm⁻¹ after field cycling in the 9.4-nm-thick ZrO₂ film deposited at 250 °C and treated with PMA.¹⁰ Similar effects may have decreased E(+_up) in C280 and C310_PMA due to a decrease in E_crit resulting from the field cycling.

Fig. 4(b)–(d) show the I–E and P–E curves of C285, C290, and C310 films in the woken-up state at 2.6 MV cm⁻¹, 100 kHz, and 10⁷ cycles, respectively. In the I–E curves of the pristine state films, switching peaks were observed in all four areas: +_up, +_down, −_down, and −_up. However, they appeared only in +_up and −_down after the wake-up. In addition, P–E curves of the pristine state exhibited pinched hysteresis, whereas those in the woken-up state displayed FE-like hysteresis. The FE-like curves of C285, C290, and C310films show E(+_up) and E(−_up) values of ∼1.38, ∼1.22, and ∼0.80 MV cm⁻¹, and ∼−0.29, ∼−0.31, and ∼−0.57 MV cm⁻¹, respectively, indicating a more symmetrical shape with increasing T_dep. In addition, the 2P_r values of the C285, C290, and C310 films were ∼3.5, ∼9.6, and ∼15.1 μC cm⁻², respectively. However, such FE-like I–E and P–E curves do not guarantee that the films are composed of only the PO-phase during the field cycling, as discussed below.

As in the C280 and C310_PMA cases, the C285, C290, and C310 films may experience changes in E_int, E_crit, and E_c during field cycling. However, unlike C280 and C310_PMA, which only show FFE characteristics, C285, C290, and C310 films have mixed PO- and T-phase, so the effect of field cycling on each phase may differ. In particular, since charge injection occurs mainly during polarization switching,^42–44 it is necessary to analyze how these effects are reflected in the characteristics near the E_c of FE and the E_crit of FFE under the influence of E_int↓. Therefore, in the next section, the origin of the characteristics of the woken-up state of the C285, C290, and C310 films with mixed FE and FFE is discussed in detail by observing the evolution of I–E through field cycling.

3.1.4. Evolution of I–E curves of the mixed phase under the influence of E_int↓. In this section, offset field cycling analysis was conducted on the C310 film, which exhibits the most pronounced FE + FFE characteristics, to investigate how field cycling affects FE and FFE characteristics in the ZrO₂ film under the E_int↓ effect.

Fig. 5(a) shows the bipolar triangular pulse when field cycling was carried out at −1 ± 1.6 MV cm⁻¹ (or between −2.6 and 0.6 MV cm⁻¹) by offsetting the field by −1 MV cm⁻¹. The uppermost panel of Fig. 5(b) shows the I–E curve of the pristine state sample, with the sweeping field ranging from −2.6 to 2.6 MV cm⁻¹. It shows that the FFE switching, which requires a positive field as high as ∼2.3 MV cm⁻¹, cannot occur under this field cycling condition (with a maximum positive field strength of only 0.6 MV cm⁻¹). In contrast, the FE switching in the negative field region requires only ∼−1.7 MV cm⁻¹, so the applied maximum negative field (−2.6 MV cm⁻¹) during field cycling can induce the FE switching. Therefore, the field cycling can affect only the switching properties of the FE portion of the film. Fig. 5(b) illustrates the evolutionary process of I–E curves for the pristine and10⁵, 10⁶, and 10⁷ cycled states, respectively. As the cycling proceeds, the P_PO↑ and P_PO↓ (blue arrows) peaks of FE shift in the positive field direction, indicating that E_int↓ decreases. In contrast, the switching peaks for P_T→PO↓ and P_PO↓→T (red arrows) of FFE remained almost unchanged, corroborating the discussions above. Fig. 5(c) shows the pulse configuration when cycling at +1 ± 1.6 MV cm⁻¹ (or between −0.6 and 2.6 MV cm⁻¹) by applying an offset of +1 MV cm⁻¹. Following the discussions above, this cycling field condition may impact the FFE portion without influencing the FE portion of the film. Fig. 5(d) shows a similar evolutionary process of I–E curves to that in Fig. 5(b) with a +1 ± 1.6 MV cm⁻¹ cycling field. Unlike Fig. 5(b), the P_T→PO↓ and P_PO↓→T (red arrows) transitions of FFE were shifted in the negative field direction, indicating an increase in E_int↓ with increasing cycling. In contrast, the peaks for P_PO↑ and P_PO↓ (blue arrows) of FE show little change, indicating that the field cycling at +1 ± 1.6 MV cm⁻¹ influences only the FFE portion. This finding also suggests that the film has FE and FFE portions in parallel rather than stacked. Therefore, they are affected by the field independently.

Fig. 5 (a) Bipolar triangular pulse when field cycling was carried out at ±1.6 MV cm⁻¹ with an offset of −1 MV cm⁻¹. (b) Evolution of I–E curves obtained by offset field cycling carried out on a 10-nm-thick C310 film. (c) Bipolar triangular pulse when field cycling was carried out at ±1.6 MV cm⁻¹ with an offset of +1 MV cm⁻¹. (d) Evolution of I–E curves by the offset field cycling carried out on the film. (e) Bipolar triangular pulse when field cycling was carried out at ±2.6 MV cm⁻¹. (f) Evolution of I–E curves by the field cycling without offset carried out on the film. Switching peaks P_PO↓, P_PO↑ of FE (red arrows) and P_T→PO↓, and P_PO↓→T of FFE (blue arrows) are shown in Fig. 5(b), (d) and (f).

Fig. 5(e) shows the pulse when field cycling was carried out at ±2.6 MV cm⁻¹ on the C310 film, which is the condition that leads to the woken-up state, as shown in Fig. 4(d). Therefore, the P_PO↑ and P_PO↓ of the FE grain, as well as the P_T→PO↓ and P_PO↓→T of the FFE grain, were repeatedly switched. Fig. 5(f) shows the evolutionary process of I–E curves for the pristine and 10⁵, 10⁶, and 10⁷ cycled states, respectively. In this case, the FE and FFE switching peaks shifted in the positive and negative field directions, respectively, indicating that the imprint effect of the FE and FFE portions became weaker and stronger with the cycling. As a result, the switching current of FE and FFE overlapped after 10⁷ cycles, and the overall I–E curve shape became similar to that of an imprinted (in the positive field direction) FE material. Therefore, it can be inferred from the above discussions that the FE-like I–E curve is an outcome of combining the less imprinted FE phase with the severely imprinted FFE phase. Furthermore, as shown in Fig. S5 (ESI†), this behavior persisted under field cycling conditions of 2.6 MV cm⁻¹ up to 10⁹ cycles and 3.0 MV cm⁻¹ up to 10⁷ cycles, confirming that the endurance performance is comparable to that reported in previous studies^45–50 on HfO₂- and ZrO₂-based FE thin films. Fig. S6(a) and (b) in the ESI† show the evolution of I–E curves with a field cycling of ±2.6 MV cm⁻¹ for the C285 and C290 films, yielding results similar to those observed for the C310 film. In other words, FE-like curves of the C285, C290, and C310 films in the woken-up state resulted from combining imprint-relaxed FE and imprint-enhanced FFE phases. These findings indicate that the more positively shifted FE-like curve of the sample deposited at a lower T_dep in the woken-up state (Fig. 4) can be attributed to the higher FFE compared to the FE portion. Conversely, when the T_dep increased, the I–E curve became more symmetrical (less imprinted), indicating that the FE portion increased.

The microscopic origin of such intriguing findings can be understood from the following model. The structural analysis in the following section reveals that the ZrO₂ film contained a higher V_O concentration near the TE interface than the BE interface, resulting in the positive, i.e., from TE to BE direction, E_int (E_int↓ in the previous section) across the entire film area, which contains that FE and FFE portions. Fig. 6(a) schematically illustrates this configuration in its pristine state. When the cycling field in Fig. 5(a) was applied, the maximum negative field (−2.6 MV cm⁻¹) was sufficiently large to switch the polarization of the FE portion, P_PO↑, despite the E_int↓. Although the maximum positive field (0.6 MV cm⁻¹) was similar to the nominal coercive field (∼0.5 MV cm⁻¹), the previous P_PO↑ could be switched entirely to P_PO↓ with the aid of E_int↓. Therefore, even with such an asymmetric electric field configuration, the FE portion could experience facile FE switching during the wake-up cycling. In this case, the usual imprint field relaxation occurs, and the E_int↓ is weakened, corroborating the shift of the FE switching peak shown in Fig. 5(b). Fig. 6(b) illustrates these circumstances. However, the FFE portion could hardly undergo such switching. As shown in Fig. 3(b), the FFE portion requires a negative field as high as ∼4 MV cm−1 for the T → PO↑ transition. However, the maximum negative field could be only ∼−1.6 MV cm−1 (= −2.6 MV cm−1 + 1 MV cm−1 (E_int↓)), so the T → PO↑ transition could not occur. Also, the T → PO↓ transition under positive bias could not occur even with the E_int↓ because the maximum positive field was limited to ∼1.6 MV cm−1 (= 0.6 MV cm−1 + 1 MV cm−1 (E_int↓)). The T → PO↓ transition required a positive E_crit as high as ∼2.3 MV cm−1. Therefore, no changes in the FFE characteristics are expected, which was indeed the case.

Fig. 6 (a) Schematics of E_int↓ in the ZrO₂ film stack. (b) Schematics showing the origin of E_int↓ changes caused by switching in FE and FFE regions respectively in maximum negative (E_app = −2.6 MV cm⁻¹) and maximum positive (E_app = 0.6 MV cm⁻¹) fields during offset field cycling with 1 ± 1.6 MV cm⁻¹. (c) Schematics showing the origin of E_int↓ changes caused by switching in FE and FFE regions respectively in maximum positive (E_app = 2.6 MV cm⁻¹) and maximum negative (E_app = −0.6 MV cm⁻¹) fields during offset field cycling with −1 ± 1.6 MV cm⁻¹.

When the cycling field shown in Fig. 5(c) is applied, the circumstances for the FE and FFE portions become distinct. In this case, the maximum negative field became slightly positive (−0.6 MV cm⁻¹ + 1 MV cm⁻¹ = 0.4 MV cm⁻¹), so no P_PO↑ could be acquired in the FE portion. Since no P_PO↑ switching could occur, the FE polarization always remained at PO↓, and no imprint relaxation was expected, which was indeed the case in Fig. 5(d). In contrast, in this case, the maximum positive field (2.6 MV cm⁻¹ + 1 MV cm⁻¹ = 3.6 MV cm⁻¹) exceeded the E_crit of ∼2.3 MV cm⁻¹ of the FFE, resulting in the T → PO↓ transition. During the field sweep down to −0.6 MV cm⁻¹, the PO↓ → T transition occurred, resulting in the two FFE switching peaks in the positive field region. However, the T → PO↑ and PO↑ → T transitions could not occur as these required a much larger negative cycling field. Therefore, the FFE switching could occur only in the positive field region, which may not relax the E_int↓ in the FFE portion. Instead, the exceptionally high positive field may induce the injection of positive (negative) charge from the TE (BE) to the ZrO₂ film, further strengthening the E_int↓. The FFE switching peak shifts in the negative field direction during cycling, as shown in Fig. 5(d), which can be understood in terms of unipolar charge injection. Fig. 6(c) illustrates the schematic diagram that demonstrates this behavior.

When the cycling field amplitude increased in both directions, as shown in Fig. 5(e), changes in the FE and FFE regions, i.e., relaxation and strengthening of E_int↓, respectively, occurred simultaneously. The resulting changes are illustrated in Fig. 5(f). The upper schematic diagrams of Fig. 6(b) and (c) show this behavior.

According to Shin et al.,¹⁰ the minimum switching field for the E_T→PO↓ transition induced by field cycling in ZrO₂ thin films was approximately 2.1–2.2 MV cm⁻¹. However, in the case of the C310 film, the minimum switching field for the FFE component was reduced significantly to ∼0.8 MV cm⁻¹. Furthermore, as shown in Fig. S7 (ESI†), the dielectric constants of the C310 film in the pristine and woken-up states (after field cycling at 2.6 MV cm⁻¹, 100 kHz) were approximately 36 and 37, respectively. These findings suggest that the evolution of the pinched hysteresis loop—reflecting a mixed FE and FFE state in the pristine film—into a more FE-like hysteresis after field cycling cannot be explained by domain depinning or phase transition mechanisms alone. Therefore, the emergence of FE-like properties with increased P_r in the woken-up state of an imprinted FE and FFE-mixed ZrO₂ thin film is more appropriately explained by the superposition of E_int↓-relaxed FE and E_int↓-enhanced FFE portions.

3.2. Structural and chemical properties of ZrO₂ films

In this section, the structural and chemical properties were analyzed to determine the origin of changes in the electrical properties of the ZrO₂ thin film concerning the T_dep and PMA.

Fig. 7(a) shows the XRD patterns of C280, C285, C290, C310, and C310_PMA with 10-nm thickness. It shows the peaks of the TiN TE and BE around 36.8°, and the peaks around 30.5° indicate o(111)/t(101) of the ZrO₂ film. Fig. 7(b) shows the results of the Gaussian fitting of the main peaks o(111)/t(101) of ZrO₂ films.^51–54 The as-deposited ZrO₂ film exhibited an increase in peak intensity with increasing T_dep. Notably, the peak intensity of C310_PMA was slightly higher. Fig. S2(b) in the ESI† shows the relative crystallinity and dielectric permittivity of each film, calculated from the XRD main peak area in Fig. 7(b). Considering the reported dielectric constants^22–27 of ZrO₂ phases in previous studies—approximately ∼22 for the amorphous, ∼26 for the PO-phase, ∼20–27 for the APO-phase, and ∼40–47 for the T-phase—the dielectric permittivity of C280, C285, C290, and C310 (ranging from ∼27 to ∼37) increases with crystallinity. These XRD and dielectric permittivity results indicate that the crystallinity of ZrO₂ films increases in the order of C280 < C285 < C290 < C310 < C310_PMA.

Fig. 7 (a) XRD patterns of 10-nm-thick C280 (blue), C285 (green), C290 (orange), C310 (red), and C310_PMA (black) thin films. (b) XRD main peaks of the films using a Gaussian curve-fitting. (c) Deconvolution of the XRD peak of the C310 film (red) for comparison with the C310_PMA film (black). The peak at 2θ = 30.49° of the C310 film was deconvolved into two parts indicating the T-phase (2θ = 30.43°) and the PO-phase (2θ = 30.54°).

Comparing the 2θ positions of the films, there was no significant difference in the other films. However, the 2θ for C310_PMA tended to increase. Fig. 7(c) shows the XRD main peaks of C310 and C310_PMA, with the peak of C310 deconvolved for phase transition analysis. The peak position for C310 was 30.49°, whereas for C310_PMA, it was 30.54°, representing an increase of 0.05°. The XRD raw peak (light red) of the C310 film was broad and appeared not to converge into a single peak, whereas the raw peak (light gray) of the C310_PMA film seemed to sharpen and converge into a single peak. Since the T-phase peak position of ZrO₂ increases by approximately 0.1° compared to the PO-phase peak, the two deconvolved peaks of 30.43° and 30.54° in C310 may imply PO-phase and T-phase fractions, respectively.^51–54 This result is consistent with the electrical characteristics, which show only the FFE characteristics in C310_PMA, unlike the FE and FFE characteristics observed in C310.

Fig. 8 shows the volumetric fractions of grain sizes for 10-nm-thick C280, C290, C310, and annealed C310 films. The annealed C310 film was annealed in an N₂ atmosphere at 400 °C for 3 minutes to emulate the PMA condition without TiN TE deposition. During TiN TE removal for grain size analysis after PMA, the top surface morphology may be damaged, making it difficult to obtain proper SEM images. Nevertheless, Park et al. compared the annealed HZO film with and without a TiN TE to examine the effect of the TiN capping layer on grain size and its distribution. The results conclude that the effect of the TiN TE on grain size distribution is negligible.¹⁷ Therefore, the grain size distribution of the C310_PMA film was confirmed using an annealed film without a TiN TE. Fig. S8 in the ESI† shows the top-view SEM images used to measure the grain size of each film. The grain sizes of C280, C290, C310, and annealed C310 films, measured using the software Gwyddion on the SEM images, were 10.2 nm, 11.7 nm, 11.7 nm, and 12.6 nm, respectively. These findings suggest that the increase in T_dep or PMA treatment can contribute to the enlargement of grain size in 10-nm-thick ZrO₂ thin films. The tendency of the measured grain size is consistent with the degree of crystallization observed in each film by XRD.

Fig. 8 Distribution of grain size of 10-nm-thick C280, C290, C310, and annealed C310 thin films measured using software Gwyddion based on the SEM top-view image.

Materlik et al. calculated the free energy for each phase of ZrO₂ according to the grain size. They found that the grain size region where the PO-phase is the most stable compared to the T- and M-phases did not appear, but as the grain size increased, the free energy tended to decrease rapidly in the order of the T → PO → M-phase.⁶ Ye et al. reported through density functional theory (DFT) calculations that the fraction of the T-phase decreases as the grain size increases in the ZrO₂ film below 18-nm-thick, while the fraction of the PO-phase increases.⁷ Chae et al. also reported that the formation of the T-phase is favored when the grain size is small, and the formation of the PO-phase is favored when the grain size is larger, based on calculations of the free energy of the T- and PO-phases in ZrO₂.¹² Therefore, it is possible that the fractions of the PO-phase gradually increased in this study with the gradual increase in grain size with increasing T_dep.

However, the grain size of the annealed C310 film, which exhibited only FFE properties, was 12.6 nm and appeared to be 0.9 nm greater than that of the 11.7 nm C310 film, which showed the largest 2P_r value. This may suggest that factors other than grain size contribute more to PO-phase stabilization in ZrO₂ thin films.

Fig. 9(a)–(d) show SIMS depth profiles of C280, C290, C310, and C310_PMA films. Films for depth profile analysis were made of a 10-nm-thick TiN TE/10-nm-thick ZrO₂/50-nm-thick TiN BE stack, and the analysis was conducted until the TiN BE intensity was obtained. In each figure, the depth direction of negatively charged ions O⁻ (blue), TiO⁻ (green), TiN⁻ (black), and ZrO²⁻ (red) are shown and matched well with the stack of samples.

Fig. 9 SIMS depth profiling results of 10-nm-thick (a) C280, (b) C290, (c) C310, and (d) C310_PMA thin films. The negatively charged O⁻ (blue), TiO⁻ (green), TiN⁻ (black), and ZrO²⁻ (red) species are depicted in each figure. (e) Cross-sectional TEM image of a TiN TE/6-nm-thick C310/TiN BE film stack.

The TiO⁻ and O⁻ signals in the middle of the TiN TE and within the ZrO₂ layer decrease with increasing T_dep. Also, the ZrO₂ signal tail into the TiN BE becomes longer as T_dep decreases. These findings indicate that denser ZrO₂ films are formed at higher T_dep, which is corroborated by the estimated densities of the ZrO₂ thin films, analyzed through XRR (5.931 g cm⁻³ (C280), 6.215 g cm⁻³ (C290), and 6.291 g cm⁻³ (C310)), as shown in Fig. S9(a)–(c) of the ESI.†

Fig. 9(e) shows the TEM image of the TiN TE/6-nm-thick C310/TiN BE stack. It was found that the interfacial oxide film is thicker near the TiN BE than near the TiN TE, corroborating the SIMS results. The formation of a thick oxide film on the TiN BE may be attributed to the continuous exposure of the TiN BE to O₃ during the deposition of ZrO₂ thin films. In addition, since the TiO_x layer formed by oxidation at the beginning of the ZrO₂ deposition acts as a physical barrier between the TiN BE and ZrO₂, the formation of V_O inside ZrO₂ may be suppressed during the subsequent ALD process.^55–58 On the other hand, since TiO_x in the TiN TE was created by combining with oxygen at the upper interface of ZrO₂ after ZrO₂ deposition was completed, the upper part of ZrO₂ may have a relatively high concentration of V_O compared to the lower part. Therefore, the difference in V_O distribution within the ZrO₂ thin film may have resulted in the formation of E_int↓ from TiN TE to TiN BE, as shown in the aforementioned Fig. 6(a). As a result, the shift of the I–E and P–E curves in the pristine and the woken-up states, as described in Section 3.1, may occur in the positive direction.

Fig. 10(a) shows the total areal density of oxygen measured by WDXRF for C280 and C310 films deposited on TiN BE/Si substrates (left columns), as well as the areal density of oxygen in TiN BE/Si substrates exposed under identical conditions without the Zr precursor (right columns). Fig. 10(b) shows the areal density of elements Zr and O of C280 and C310, measured by EDXRF and WDXRF, respectively. To estimate the areal density of O only in the ZrO₂ thin film, the areal density of oxygen in the substrate was subtracted from the total areal density of oxygen in Fig. 10(a). The areal densities of Zr and O of C280 were 4.88 μg cm⁻² and 1.17 μg cm⁻², respectively, whereas for C310 they were 4.75 μg cm⁻² and 1.33 μg cm⁻². The Zr : O composition ratios derived from these values were 1 : 1.4 for C280 and 1 : 1.6 for C310, indicating a higher O content in the C310 film. (Since the areal density of O in TiN BE/Si exposed to the oxidizing atmosphere is expected to be higher than that of O in the TiN BE oxidized during ZrO₂ deposition, the estimated O content in the film may be slightly underestimated.)

Fig. 10 (a) The areal density of oxygen measured by WDXRF for 10-nm-thick ZrO₂ films deposited on C280/TiN BE/Si and C310/TiN BE/Si stacks (left columns), along with a reference TiN BE/Si sample deposited under identical conditions but without the Zr precursor (right column). (b) Areal density of Zr in C280 and C310 films measured by EDXRF (left). The areal density of oxygen for the ZrO₂ films alone, obtained by subtracting the TiN baseline oxygen density from the total areal density in (a), is shown on the right. Despite a slightly higher Zr areal density in C280 (4.88 μg cm⁻²) compared to that in C310 (4.75 μg cm⁻²), the oxygen areal density was lower in C280 (1.17 μg cm⁻²vs. 1.33 μg cm⁻²), indicating a lower oxygen-to-zirconium ratio at the lower T_dep.

Fig. 11(a) shows the XPS Zr 3d spectra of the C280, C310, and C310_PMA samples with a 6 nm-thick TiN TE. The binding energies of the Zr 3d_5/2 states for C280, C310, and C310_PMA were 182.5 eV, 182.7 eV, and 182.4 eV, respectively. This finding suggests that ZrO₂ films deposited at 280 °C have a lower oxidation state compared to those deposited at 310 °C, which is consistent with the XRF results shown in Fig. 10. When the ZrO₂ film was annealed with the TiN TE present, the binding energy of the Zr 3d_5/2 peak decreased compared to the as-deposited case, indicating that part of the oxygen ions had migrated to the TiN TE and increased the V_O concentration near the TE interface.

Fig. 11 (a) XPS spectra of the Zr 3d of 10-nm-thick C280 (blue), C310 (red), and C310_PMA (black) films. (b) XPS spectra of the Ti 2p of 6-nm-thick TiN TE deposited on Au (gold), C280 (blue), and C310 (red) films, respectively. (c) XPS spectra of the Ti 2p of PMA-treated 6-nm-thick TiN TE deposited on Au (gold) and C310 (red) films, respectively.

Fig. 11(b) shows the XPS Ti 2p spectra of the 6-nm-thick TiN TE when deposited on 10-nm-thick C280, 10-nm-thick C310, and 50-nm-thick Au, respectively, to compare the oxygen scavenging effect of the TiN TE. In the Ti 2p spectra, Ti(0) and Ti(IV) represent the binding energies for TiN and TiO₂, respectively. The Ti 2p spectra of C280 and C310 films were similar, but the intensity of the Ti(0) peak was relatively high in the TiN TE deposited on Au. Fig. S10(a)–(c) in the ESI† show the deconvoluted Ti 2p spectra of the TiN TE deposited on C280, C310, and Au films, respectively, assuming the presence of metallic Ti, Ti²⁺, Ti³⁺, and Ti⁴⁺ ions. The proportion of Ti(0) peaks was 20.7% on C280, 20.5% on C310, and 21.8% on Au, respectively, indicating that V_O can be formed within ZrO₂ during TiN TE deposition. Fig. 10(c) shows the similar XPS spectra of Ti 2p for the 6-nm-thick TiN TE deposited on 10-nm-thick C310 (red) and 50-nm-thick Au (gold), respectively, after PMA treatment. Fig. S10(d) and (e) (ESI†) show the deconvolution results. The proportion of Ti(0) in the C310 sample decreases to 11.5%, indicating that the TiN TE scavenges oxygen from the underlying ZrO₂ during PMA. A similar decrease to 15.4% in the TiN/Au sample indicates that the ∼7% decrease should be attributed to oxidation by the residual oxygen in the RTA chamber. Still, it is evident that PMA induced the oxygen scavenging effect from the ZrO₂ layer, resulting in the formation of V_O near the TiN TE interface.

As previously discussed, the interfacial oxide layer between the TiN bottom electrode and the ZrO₂ film can act as a physical barrier to suppress oxygen scavenging from the TiN substrate, helping to preserve oxygen stoichiometry favorable for PO-phase stabilization. The thickness of this oxide layer is expected to increase with T_dep, which may further promote PO-phase formation by modulating strain at the interface. Also, the strain state originating from the substrate can influence grain growth during crystallization.^12,18,56 According to the SIMS analysis, the oxygen scavenging effect from the top TiN electrode is evident from the relative intensities of O⁻ and TiO⁻ ions, and is more pronounced in films deposited at lower temperatures. This suggests that films deposited at lower temperatures—with a higher amorphous fraction compared to the more densely crystallized films grown at higher temperatures—are likely to exhibit a higher concentration of V_O, which may contribute to the stabilization of the T-phase.

As T_dep increases, both the oxygen content and the crystallinity of the ZrO₂ thin films, reflected by grain size, also increase. In the C280 sample, the low T_dep fails to supply sufficient thermal energy for effective crystallization, resulting in smaller grains (∼10.2 nm) with lower crystallinity. Previous studies^{6,7,12,15–17} have reported that such small grain sizes tend to stabilize the tetragonal T-phase over the PO-phase. In addition, incomplete ligand desorption during ALD at low temperatures may hinder the formation of stable Zr–O bonds, thereby reducing oxygen incorporation and increasing the V_O concentration. This is supported by WDXRF and XPS analyses, which reveal lower oxygen content and reduced Zr oxidation states in low-temperature-deposited films, indicating a higher V_O concentration in the C280 sample.

Although the C310 sample exhibits improved crystallinity with larger grains (∼12.6 nm), PMA enhances oxygen scavenging from the TiN electrode, resulting in a further decrease in oxygen content and oxidation state. Literature reports^{7,13,55,59–62} have shown that increased V_O concentration lowers the total energy of the T-phase more significantly than that of the PO-phase, energetically favoring T-phase formation. These findings support our observation that FFE behavior becomes dominant in low-temperature-deposited ZrO₂ films, and that FFE characteristics are exclusively observed in PMA-treated films due to enhanced V_O formation.

Therefore, minimizing V_O is essential for stabilizing the PO-phase and achieving reliable ferroelectric behavior.

4. Conclusions

This study investigated the evolution of FE and FFE behavior in 10-nm-thick ZrO₂ films with TiN electrodes as a function of T_dep and PMA. At low T_dep (C280), only FFE behavior was observed, whereas higher T_dep (C285, C290, and C310) led to a gradual emergence of FE-like behavior, as evidenced by increased 2P_r values (∼3.5, 9.6, and 15.1 μC cm⁻²) after field cycling.

The FE and FFE switching characteristics, analyzed via I–E and P–E measurements, revealed that E_int↓ dynamics are governed by the available switching pathways. In FE regions, symmetrical P_PO↑ and P_PO↓ switching leads to E_int↓ relaxation, while in FFE regions, unidirectional switching—limited to P_T→PO↓ and P_PO↓→T transitions, with no occurrence of P_T→PO↑ and P_PO↑→T—enhances E_int↓. This explains the merged switching peaks and apparent FE-like response observed in C285, C290, and C310 during wake-up, which does not necessarily indicate a full transition from FFE to FE.

Structural and chemical analysis confirmed that higher T_dep increases grain size and reduces V_O concentration, favoring PO-phase formation. However, in the C310 sample, PMA-induced oxygen scavenging increased V_O concentration and restored FFE-like behavior despite improved crystallinity and grain size. These findings indicate that V_O concentration plays a more dominant role than grain size in stabilizing the PO-phase, emphasizing the importance of V_O control for achieving robust ferroelectricity in ZrO₂ thin films.

Author contributions

J. Song designed and performed the experiments and wrote the manuscript. J. H. Shin, S. Y. Byun, I. S. Lee, and H. W. Nam assisted with the device fabrication. Y. J. Kim assisted in the maintenance of the ALD system and electrical measurement. K. D. Kim, S. H. Lee, H. S. Park, and S. J. Shin assisted with data interpretation, helped review the manuscript, and provided comments. C. S. Hwang supervised the whole research and manuscript preparation.

Conflicts of interest

The authors declare they have no conflict of interest.

Data availability

The data supporting this article have been included as part of the ESI.† The data that support the findings of this study are available from the corresponding author upon reasonable request.

Acknowledgements

This work was supported by the National Research Foundation of Korea (Grant No. 2020R1A3B2079882) and the Technology Innovation Program (20017224, Development of a New ALD System for Next-Generation Mass-Production Memory Devices) funded by the Ministry of Trade, Industry, and Energy (MOTIE, Korea).

Notes and references

1. U. Schroeder, M. H. Park, T. Mikolajick and C. S. Hwang, Nat. Rev. Mater., 2022, 7, 653–669.
2. D. Panda and T.-Y. Tseng, Thin Solid Films, 2013, 531, 1–20.
3. J. Robertson, Eur. Phys. J.: Appl. Phys., 2004, 28, 265–291.
4. H. Seo, I. W. Yeu, D. S. Kwon, D. G. Kim, J. Lim, T. K. Kim, H. Paik, J. H. Choi and C. S. Hwang, Adv. Electron. Mater., 2022, 8, 2200099.
5. S. Shukla and S. Seal, Int. Mater. Rev., 2005, 50, 45–64.
6. R. Materlik, C. Künneth and A. Kersch, J. Appl. Phys., 2015, 117.
7. K. H. Ye, I. W. Yeu, G. Han, T. Jeong, S. Yoon, D. Kim, C. S. Hwang and J.-H. Choi, Appl. Phys. Rev., 2023, 10, 031419.
8. S. E. Reyes-Lillo, K. F. Garrity and K. M. Rabe, Phys. Rev. B: Condens. Matter Mater. Phys., 2014, 90, 140103.
9. S. D. Hyun, H. W. Park, M. H. Park, Y. H. Lee, Y. B. Lee, B. Y. Kim, H. H. Kim, B. S. Kim and C. S. Hwang, Adv. Electron. Mater., 2020, 6, 2000631.
10. J. Shin, D. H. Shin, K. Do Kim, H. Seo, K. H. Ye, J. W. Jeon, T. K. Kim, H. Paik, H. Song and S. H. Lee, J. Mater. Chem. C, 2024, 12, 15423–15434.
11. S. Starschich, T. Schenk, U. Schroeder and U. Boettger, Appl. Phys. Lett., 2017, 110, 032903.
12. K. Chae, S. F. Lombardo, N. Tasneem, M. Tian, H. Kumarasubramanian, J. Hur, W. Chern, S. Yu, C. Richter and P. D. Lomenzo, ACS Appl. Mater. Interfaces, 2022, 14, 36771–36780.
13. B. Xu, L. Collins, K. M. Holsgrove, T. Mikolajick, U. Schroeder and P. D. Lomenzo, ACS Appl. Electron. Mater., 2023, 5, 2288–2295.
14. T. T. Ngoc Van, D. Jang, E. Jung, H. Noh, J. Moon, D.-S. Kil, S.-W. Chung and B. Shong, J. Phys. Chem. C, 2022, 126, 18090–18099.
15. K. Kim, M. Park, H. Kim, Y. Kim, T. Moon, Y. Lee, S. Hyun, T. Gwon and C. Hwang, J. Mater. Chem. C, 2016, 4, 6864–6872.
16. S. Shibayama, T. Nishimura, S. Migita and A. Toriumi, J. Appl. Phys., 2018, 124, 184101.
17. M. H. Park, Y. H. Lee, H. J. Kim, T. Schenk, W. Lee, K. Do Kim, F. P. Fengler, T. Mikolajick, U. Schroeder and C. S. Hwang, Nanoscale, 2017, 9, 9973–9986.
18. B. Xu, R. Ganser, K. M. Holsgrove, X. Wang, P. Vishnumurthy, T. Mikolajick, U. Schroeder, A. Kersch and P. D. Lomenzo, ACS Appl. Mater. Interfaces, 2024, 16, 32533–32542.
19. Y. Lee, H. W. Jeong, S. H. Kim, K. Yang and M. H. Park, Mater. Sci. Semicond. Process., 2023, 160, 107411.
20. S. J. Shin, H. Kim, S. Byun, J. Shin, J. Choi, S. H. Lee, K. D. Kim, J. H. Song, D. H. Shin, S. H. Lee, I. S. Lee and H. Nam, unpublished work.
21. C. Kittel, Phys. Rev., 1951, 82, 729.
22. S. K. Kim and C. S. Hwang, Electrochem. Solid-State Lett., 2007, 11, G9.
23. Y. Zhang, H.-X. Chen, L. Duan, J.-B. Fan, L. Ni and V. Ji, Solid State Sci., 2018, 81, 58–65.
24. X. Zhao and D. Vanderbilt, MRS Online Proc. Libr., 2002, 747, T5.2/N7.2.
25. X. Zhao and D. Vanderbilt, Phys. Rev. B: Condens. Matter Mater. Phys., 2002, 65, 075105.
26. J.-H. Kim, V. Ignatova, P. Kücher, J. Heitmann, L. Oberbeck and U. Schröder, Thin Solid Films, 2008, 516, 8333–8336.
27. X. Zhao, D. Ceresoli and D. Vanderbilt, Phys. Rev. B: Condens. Matter Mater. Phys., 2005, 71, 085107.
28. T. Schenk, M. Hoffmann, J. Ocker, M. Pesic, T. Mikolajick and U. Schroeder, ACS Appl. Mater. Interfaces, 2015, 7, 20224–20233.
29. F. Fengler, M. Hoffmann, S. Slesazeck, T. Mikolajick and U. Schroeder, J. Appl. Phys., 2018, 123, 204101.
30. P. Buragohain, A. Erickson, P. Kariuki, T. Mittmann, C. Richter, P. D. Lomenzo, H. Lu, T. Schenk, T. Mikolajick and U. Schroeder, ACS Appl. Mater. Interfaces, 2019, 11, 35115–35121.
31. K. Takada, S. Takarae, K. Shimamoto, N. Fujimura and T. Yoshimura, Adv. Electron. Mater., 2021, 7, 2100151.
32. M. Park, H. Kim, Y. Kim, T. Moon, K. Kim, Y. Lee, S. Hyun and C. Hwang, J. Mater. Chem. C, 2015, 3, 6291–6300.
33. S. Hong, Y. Lee, D. Ahn and S.-E. Ahn, Appl. Phys. Lett., 2021, 118, 013504.
34. P. D. Lomenzo, M. Materano, T. Mittmann, P. Buragohain, A. Gruverman, T. Kiguchi, T. Mikolajick and U. Schroeder, Adv. Electron. Mater., 2022, 8, 2100556.
35. A. K. Saha and S. K. Gupta, IEEE Device Research Conference (DRC), 2018, pp. 1–2.
36. S. Miller, J. Schwank, R. Nasby and M. Rodgers, J. Appl. Phys., 1991, 70, 2849–2860.
37. L. Chen, Z. Liang, S. Shao, Q. Huang, K. Tang and R. Huang, Nanoscale, 2023, 15, 7014–7022.
38. M. Pešić, F. P. G. Fengler, L. Larcher, A. Padovani, T. Schenk, E. D. Grimley, X. Sang, J. M. LeBeau, S. Slesazeck and U. Schroeder, Adv. Funct. Mater., 2016, 26, 4601–4612.
39. K. Do Kim, Y. H. Lee, T. Gwon, Y. J. Kim, H. J. Kim, T. Moon, S. D. Hyun, H. W. Park, M. H. Park and C. S. Hwang, Nano Energy, 2017, 39, 390–399.
40. P. D. Lomenzo, C.-C. Chung, C. Zhou, J. L. Jones and T. Nishida, Appl. Phys. Lett., 2017, 110, 232904.
41. M. H. Park, H. J. Kim, Y. J. Kim, Y. H. Lee, T. Moon, K. D. Kim, S. D. Hyun, F. Fengler, U. Schroeder and C. S. Hwang, ACS Appl. Mater. Interfaces, 2016, 8, 15466–15475.
42. H. W. Park, S. D. Hyun, I. S. Lee, S. H. Lee, Y. B. Lee, M. Oh, B. Y. Kim, S. G. Ryoo and C. S. Hwang, Nanoscale, 2021, 13, 2556–2572.
43. Q. Jiang, H. J. Lee, G. H. Kim and C. S. Hwang, Adv. Mater., 2009, 21, 2870–2875.
44. A. Tagantsev and G. Gerra, J. Appl. Phys., 2006, 100, 051607.
45. Y. Choi, C. Han, J. Shin, S. Moon, J. Min, H. Park, D. Eom, J. Lee and C. Shin, Sensors, 2022, 22, 4087.
46. M. G. Kozodaev, A. G. Chernikova, E. V. Korostylev, M. H. Park, R. R. Khakimov, C. S. Hwang and A. M. Markeev, J. Appl. Phys., 2019, 125, 034101.
47. Y. Peng, W. Xiao, Y. Liu, C. Jin, X. Deng, Y. Zhang, F. Liu, Y. Zheng, Y. Cheng and B. Chen, IEEE Electron Device Lett., 2021, 43, 216–219.
48. K. Tahara, K. Toprasertpong, Y. Hikosaka, K. Nakamura, H. Saito, M. Takenaka and S. Takagi, Proc. Symposium on VLSI Technology, 2021, pp. T7-3.
49. B. Y. Kim, I. S. Lee, H. W. Park, Y. B. Lee, S. H. Lee, M. Oh, S. K. Ryoo, S. R. Byun, K. D. Kim and J. H. Lee, Adv. Mater. Technol., 2023, 8, 2300146.
50. M. H. Park, H. J. Kim, Y. J. Kim, W. Jeon, T. Moon and C. S. Hwang, Phys. Status Solidi RRL, 2014, 8, 532–535.
51. A. Kersch, R. Ganser and M. Trien, Front. Nanotechnol., 2022, 4, 1026286.
52. J. Muller, T. S. Boscke, U. Schroder, S. Mueller, D. Brauhaus, U. Bottger, L. Frey and T. Mikolajick, Nano Lett., 2012, 12, 4318–4323.
53. H. A. Hsain, Y. Lee, G. Parsons and J. L. Jones, Appl. Phys. Lett., 2020, 116.
54. Y. H. Lee, S. D. Hyun, H. J. Kim, J. S. Kim, C. Yoo, T. Moon, K. D. Kim, H. W. Park, Y. B. Lee and B. S. Kim, Adv. Electron. Mater., 2019, 5, 1800436.
55. J. Lee, K. Yang, J. Y. Kwon, J. E. Kim, D. I. Han, D. H. Lee, J. H. Yoon and M. H. Park, Nano Convergence, 2023, 10, 55.
56. A. A. Koroleva, A. G. Chernikova, S. S. Zarubin, E. Korostylev, R. R. Khakimov, M. Y. Zhuk and A. M. Markeev, ACS Omega, 2022, 7, 47084–47095.
57. E. I. Suvorova, O. V. Uvarov, K. V. Chizh, A. A. Klimenko and P. A. Buffat, Nanomaterials, 2022, 12, 3608.
58. J. S. Kwak, Y. H. Do, Y. C. Bae, H. S. Im, J. H. Yoo, M. G. Sung, Y. T. Hwang and J. P. Hong, Appl. Phys. Lett., 2010, 96, 223502.
59. M. H. Park, Y. H. Lee, H. J. Kim, Y. J. Kim, T. Moon, K. Do Kim, S. D. Hyun, T. Mikolajick, U. Schroeder and C. S. Hwang, Nanoscale, 2018, 10, 716–725.
60. Z. Zhao, Y.-R. Chen, J.-F. Wang, Y.-W. Chen, J.-R. Zou, Y. Lin, Y. Xing, C. Liu and C. Hu, IEEE Electron Device Lett., 2022, 43, 553–556.
61. 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., 2020, 2, 3618–3626.
62. S. Shibayama, J. Nagano, M. Sakashita and O. Nakatsuka, Jpn. J. Appl. Phys., 2020, 59, SMMA04.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d5tc01449b

---