Microstructures of HfOx Films Prepared via Atomic Layer Deposition Using La(NO3)3·6H2O Oxidants

Hafnium oxide (HfOx) films have a wide range of applications in solid-state devices, including metal–oxide–semiconductor field-effect transistors (MOSFETs). The growth of HfOx films from the metal precursor tetrakis(ethylmethylamino) hafnium with La(NO3)3·6H2O solution (LNS) as an oxidant was investigated. The atomic layer deposition (ALD) conditions were optimized, and the chemical state, surface morphology, and microstructure of the prepared films were characterized. Furthermore, to better understand the effects of LNS on the deposition process, HfOx films deposited using a conventional oxidant (H2O) were also prepared. The ALD process using LNS was observed to be self-limiting, with an ALD temperature window of 200–350 °C and a growth rate of 1.6 Å per cycle, two times faster than that with H2O. HfOx films deposited using the LNS oxidant had smaller crystallites than those deposited using H2O, as well as more suboxides or defects because of the higher number of grain boundaries. In addition, there was a difference in the preferred orientations of the HfOx films deposited using LNS and H2O, and consequently, a difference in surface energy. Finally, a film growth model based on the surface energy difference was proposed to explain the observed growth rate and crystallite size trends.


Introduction
Recently, hafnium oxide (HfO x ) thin films have been studied as promising electronic materials for a wide range of solid-state device applications. The excellent insulating and dielectric properties of HfO x enable its application in semiconductor devices. Thin films based on HfO x have substituted SiO 2 as the material of choice for the gate dielectric layer in metal-oxide-semiconductor field-effect transistors (MOSFETs) because of their high dielectric constant, wide band gap, large band offset, and good thermodynamic stability on Si wafers [1]. More recently, HfO x has been widely studied as a candidate insulating layer in resistors with metal-insulator-metal structures, which are used in non-volatile resistive switching memory [2]. Furthermore, HfO x doped with La and Zr has attracted attention for use in CMOS-compatible ferroelectric devices [3,4]. The dopants distort the structure of HfO x , generating a ferroelectric polar orthorhombic structure.
With the continued reduction in size and increase in complexity of semiconductor devices, a need for the fabrication of ultrathin films with precisely controlled thickness on three-dimensional device structures is becoming apparent. To meet this requirement, atomic layer deposition (ALD) is one possible thin film fabrication method [5]. To fabricate metal oxide films, a typical ALD cycle consists of four steps: pulsing the metal precursor, purging the remnant with inert gas, pulsing the oxidant, and purging the remnant with inert gas. ALD via the above basic process has the advantage of offering precisely controlled Materials 2021, 14, 7478 2 of 10 of ultrathin layers with good uniformity, as well as excellent conformal coating of surfaces with intricate structures [6].
However, because of the extremely slow growth rate in ALD, low productivity is a serious disadvantage. To enhance the throughput of the ALD method, many studies have been focused on developing batch-type ALD and spatial ALD [7,8]. In particular, both metal precursors and oxidants can modulate the characteristics of metal oxide films; that is, the choice of these materials influences the growth rate, ALD temperature window, crystalline structure, contamination, and dielectric and electrical properties. Various oxidants have been used to prepare ALD oxide films, such as H 2 O, H 2 O 2 , O 3 , and plasma-based radical oxygen [1,9]. In addition to the precursor and oxidant, the catalyst can strongly affect the deposition properties and material characteristics of films grown by ALD. To fabricate ZrO 2 films, Oh et al. used La(NO 3 ) 3 ·6H 2 O solution instead of H 2 O as an oxidant and compared the crystalline phase, grain size, and surface roughness of the resulting ZrO 2 films [10]. Interestingly, use of the La(NO 3 ) 3 ·6H 2 O solution increased the ZrO 2 film growth rate because of a catalytic effect of the La-based oxidant. In addition, HfO 2 films deposited with La(NO 3 ) 3 ·6H 2 O solution instead of H 2 O exhibited modified resistive switching characteristics [11]. However, in those studies, characterization of the specific ALD processes involved when a solution oxidant is used, in terms of self-saturation, ALD temperature window, and growth linearity, was lacking. In addition, the suggested mechanism did not adequately explain the origins of the microstructural differences observed. Accordingly, in this study, a La(NO 3 ) 3 ·6H 2 O solution was used as an oxidant for ALD, with the aim of optimizing the ALD process.
In this work, we focused on the use of La(NO 3 ) 3 ·6H 2 O as a catalytic oxidant in the ALD of HfO x films. The properties of these films were compared with those of HfO x films fabricated via ALD using H 2 O as an oxidant; the film thickness was monitored as a function of precursor and oxidant pulse time, deposition temperature, and the number of ALD cycles. The chemical, surface morphological, and structural properties of the deposited HfO x films were analyzed by X-ray photoelectron spectroscopy (XPS), atomic force microscopy (AFM), grazing-incidence X-ray diffraction (GI-XRD), and transmission electron microscopy (TEM). Furthermore, a deposition mechanism was proposed to explain the difference between the growth properties and microstructures of HfO 2 films fabricated using the La(NO 3 ) 3 ·6H 2 O solution and H 2 O.

HfO x Film Fabrication
HfO x thin films were deposited on Si(100) substrates by ALD using tetrakis (ethylmethylamino) hafnium (TEMAH) as the metal precursor and 40 wt% La(NO 3 ) 3 ·6H 2 O solution (LNS) as an oxidant. To better understand the effects of LNS, HfO x thin films were also prepared using H 2 O as the oxidant, and the deposition and material properties of these were compared with those of the thin films prepared using LNS. Prior to the deposition of the HfO x thin films, p-type Si(100) substrates were cleaned in a dilute HF solution to remove the native oxide film. The metal precursor, TEMAH, was volatilized at 60 • C and delivered into a vacuum chamber filled with a pure N 2 carrier gas (>99.999%). The liquid oxidant (LNS or H 2 O) was vaporized at room temperature, and the vaporized oxidant was introduced into the chamber without any carrier gas. A cycle of ALD consisted of the following four steps: (1) pulsing TEMAH with N 2 carrier gas, (2) purging for 60 s with N 2 gas (>99.999%), (3) pulsing the oxidant (LNS or H 2 O), and (4) purging for 60 s with N 2 gas (>99.999%). To optimize the HfO x ALD process involving LNS, the substrate temperature (200-400 • C), TEMAH pulse time, and LNS pulse time were varied.

Analyses of HfO x Thin Films Properties
The thickness of the film deposited on the Si substrate was measured using spectroscopic ellipsometry (UVISEL, Horiba, Kyoto, Japan). XPS (Theta Probe, Thermo Fisher Scientific Co., Waltham, MA, USA) was used to analyze the chemical bonding of the HfO x Materials 2021, 14, 7478 3 of 10 films and to ascertain the presence or absence of La in the HfO x films. The morphological properties of the deposited films were characterized by AFM (XE-100, Park Systems, Suwon, Korea). Furthermore, the crystalline phase and crystallite size and orientation were determined by GI-XRD (SmartLab, Rigaku, Tokyo, Japan) and TEM (Tecnai F20 G 2 , FEI, Hillsboro, OR, USA). For the cross-sectional TEM image, a sample was prepared using the focused ion beam system, and for the plan view TEM image, samples were prepared using the ion milling system.

ALD Process for HfO x Film Growth Using LNS
The characteristics of HfO x thin films deposited using LNS as an oxidant were investigated by varying several process parameters (Figure 1a-d). Figure 1a, b show film growth saturation curves as a function of Hf-precursor (TEMAH) and LNS pulse times, respectively, with a substrate temperature of 300°C over 50 cycles. For LNS pulse times of ≥0.2 s, the self-limiting characteristic of the reaction was apparent, as shown in Figure 1a; for these experiments, TEMAH was injected into the process chamber for more than 1 s. Figure 1b shows the film growth saturation curve as a function of LNS pulse times greater than 0.2 s. After examining Figure 1a and b, the optimized HfO x film deposition process conditions were determined to be a substrate temperature of 300 • C, Hf-precursor pulse time of 1 s, and LNS pulse time of 0.2 s. Note that when H 2 O was used as an oxidant instead of LNS, the oxidant pulse time was the same (0.2 s).

Analyses of HfOx Thin Films Properties
The thickness of the film deposited on the Si substrate was measured using spectroscopic ellipsometry (UVISEL, Horiba, Kyoto, Japan). XPS (Theta Probe, Thermo Fisher Scientific Co., Waltham, MA, USA) was used to analyze the chemical bonding of the HfOx films and to ascertain the presence or absence of La in the HfOx films. The morphological properties of the deposited films were characterized by AFM (XE-100, Park Systems, Suwon, Korea). Furthermore, the crystalline phase and crystallite size and orientation were determined by GI-XRD (SmartLab, Rigaku, Tokyo, Japan) and TEM (Tecnai F20 G 2 , FEI, Hillsboro, OR, USA). For the cross-sectional TEM image, a sample was prepared using the focused ion beam system, and for the plan view TEM image, samples were prepared using the ion milling system.

ALD Process for HfOx Film Growth Using LNS
The characteristics of HfOx thin films deposited using LNS as an oxidant were investigated by varying several process parameters (Figure 1a-d). Figure 1a, b show film growth saturation curves as a function of Hf-precursor (TEMAH) and LNS pulse times, respectively, with a substrate temperature of 300 ℃ over 50 cycles. For LNS pulse times of ≥0.2 s, the self-limiting characteristic of the reaction was apparent, as shown in Figure  1a; for these experiments, TEMAH was injected into the process chamber for more than 1 s. Figure 1b shows the film growth saturation curve as a function of LNS pulse times greater than 0.2 s. After examining Figure 1a and b, the optimized HfOx film deposition process conditions were determined to be a substrate temperature of 300 °C, Hf-precursor pulse time of 1 s, and LNS pulse time of 0.2 s. Note that when H2O was used as an oxidant instead of LNS, the oxidant pulse time was the same (0.2 s).  The temperature window for ALD with the TEMAH precursor and LNS oxidant was characterized by measuring the thickness of deposited HfO x films after 50 ALD cycles as a function of the substrate temperature from 200 • C to 400 • C (Figure 1c). The ALD process temperature window is defined as the temperature range over which a constant thickness is deposited, which was determined to be below 350 • C in this study. The temperature window for the process involving LNS was found to differ from that of the process with H 2 O, which was in the range of 200 • C to 400 • C [12]. HfO x films were deposited under the optimized ALD conditions-a TEMAH pulse time of 1 s, LNS pulse time of 0.2 s, and temperature of 300 • C-using various numbers of ALD cycles. As can be seen in Figure 1d, the thickness of the HfO x thin films increased linearly with the number of cycles. A HfO x growth rate of 1.6 Å per cycle was obtained with the use of the LNS oxidant. Interestingly, this rate was two times faster than that measured when H 2 O was used with the same pulse time (0.8 Å per cycle). Figure 2a shows a cross-sectional TEM image of HfO x deposited on an Si substrate after 125 cycles of the optimized ALD process using LNS (TEMAH pulse time, 1 s; LNS pulse time, 0.2 s; 300 • C). The TEM image, as expected, clearly depicts an interface region consisting of native oxide (SiO x ). The thickness of the HfO x film was uniform, and the average thickness was 20 nm. The surface morphology of the HfO x film is apparent in the 2 × 2 µm 2 AFM image in Figure 2b. The root mean square (RMS) roughness value was determined to be 1.74 nm. Finally, it should be noted that the TEM and AFM results verify that HfO x films without pinholes or cracks were successfully deposited using LNS.

Microstructure of HfO x Films Prepared Using LNS
in purple for comparison in (a,b,d). The growth rates were 1.6 Å per cycle for the process involving LNS and 0.8 Å per cycle for that involving H2O.
The temperature window for ALD with the TEMAH precursor and LNS oxidant was characterized by measuring the thickness of deposited HfOx films after 50 ALD cycles as a function of the substrate temperature from 200 °C to 400 °C ( Figure 1c). The ALD process temperature window is defined as the temperature range over which a constant thickness is deposited, which was determined to be below 350 °C in this study. The temperature window for the process involving LNS was found to differ from that of the process with H2O, which was in the range of 200 °C to 400 °C [12].
HfOx films were deposited under the optimized ALD conditions-a TEMAH pulse time of 1 s, LNS pulse time of 0.2 s, and temperature of 300 °C-using various numbers of ALD cycles. As can be seen in Figure 1d, the thickness of the HfOx thin films increased linearly with the number of cycles. A HfOx growth rate of 1.6 Å per cycle was obtained with the use of the LNS oxidant. Interestingly, this rate was two times faster than that measured when H2O was used with the same pulse time (0.8 Å per cycle). Figure 2a shows a cross-sectional TEM image of HfOx deposited on an Si substrate after 125 cycles of the optimized ALD process using LNS (TEMAH pulse time, 1 s; LNS pulse time, 0.2 s; 300 °C). The TEM image, as expected, clearly depicts an interface region consisting of native oxide (SiOx). The thickness of the HfOx film was uniform, and the average thickness was 20 nm. The surface morphology of the HfOx film is apparent in the 2 × 2 μm 2 AFM image in Figure 2b. The root mean square (RMS) roughness value was determined to be 1.74 nm. Finally, it should be noted that the TEM and AFM results verify that HfOx films without pinholes or cracks were successfully deposited using LNS. The crystalline phase of the deposited HfOx films on the Si substrate with optimized process parameters (TEMAH pulse time of 1 s, oxidant pulse time of 0.2 s, and temperature of 300 °C) was identified via GI-XRD. Most of the diffraction peaks of the HfOx films prepared using LNS and H2O can be assigned to the monoclinic phase (JCPDS 06-0318), in agreement with the ALD results reported in the literature (Figure 3) [13]. The result was different from that in the case of ZrO2, the phase structure of which changed from a tetragonal phase to a monoclinic phase when LNS was used [10]. For the HfOx films prepared with H2O, the normal direction of the (1 ̅ 11) plane at 28.9° was the preferred orientation, as can be clearly seen in the XRD pattern in Figure 3. In addition, the peak assigned The crystalline phase of the deposited HfO x films on the Si substrate with optimized process parameters (TEMAH pulse time of 1 s, oxidant pulse time of 0.2 s, and temperature of 300 • C) was identified via GI-XRD. Most of the diffraction peaks of the HfO x films prepared using LNS and H 2 O can be assigned to the monoclinic phase (JCPDS 06-0318), in agreement with the ALD results reported in the literature (Figure 3) [13]. The result was different from that in the case of ZrO 2 , the phase structure of which changed from a tetragonal phase to a monoclinic phase when LNS was used [10]. For the HfO x films prepared with H 2 O, the normal direction of the (111) plane at 28.9 • was the preferred orientation, as can be clearly seen in the XRD pattern in Figure 3. In addition, the peak assigned to the (111) planes at 31.6 • was broad. When LNS was used as the oxidant, a broad peak at around 32.1 • appeared, which was similar to the (111) plane observed for the film prepared using H 2 O.

Microstructure of HfOx Films Prepared Using LNS
to the (111) planes at 31.6° was broad. When LNS was used as the oxidant, a broad peak at around 32.1° appeared, which was similar to the (111) plane observed for the film prepared using H2O. The interplanar distance and crystallite size were calculated from the angular positions of the respective preferred orientations in the XRD patterns; these values are listed in Table 1. The crystallite size of the HfOx films was determined using the Scherrer equation, D = kλ/Bcosθ, where λ is the XRD wavelength (1.5418 Å ), k is the shape factor, B is the full width at half maximum of the measured XRD peak in radians, and θ is the Bragg angle. A relatively small crystallite size of 1.7 nm was obtained for the HfOx films deposited using LNS.

Oxidant Preferred Orientation Diffraction Angle of Preferred Orientation
Crystallite Size from XRD (nm) H2O (1 ̅ 11) 28.9 7.9 LNS (111) 32.1 1.7 For more details on the difference in the crystallite size, the size of the crystallite was directly observed via dark-field TEM plan views (Figure 4). For the HfO2 films fabricated using H2O, as shown in Figure 4a, large crystallites were observed, with diameters of more than 20 nm. In the case of the films prepared using LNS (Figure 4b), small crystallites were found with diameters of less than 10 nm. The difference between the XRD and TEM crys- The interplanar distance and crystallite size were calculated from the angular positions of the respective preferred orientations in the XRD patterns; these values are listed in Table 1. The crystallite size of the HfO x films was determined using the Scherrer equation, D = kλ/Bcosθ, where λ is the XRD wavelength (1.5418 Å), k is the shape factor, B is the full width at half maximum of the measured XRD peak in radians, and θ is the Bragg angle. A relatively small crystallite size of 1.7 nm was obtained for the HfO x films deposited using LNS. For more details on the difference in the crystallite size, the size of the crystallite was directly observed via dark-field TEM plan views (Figure 4). For the HfO 2 films fabricated using H 2 O, as shown in Figure 4a, large crystallites were observed, with diameters of more than 20 nm. In the case of the films prepared using LNS (Figure 4b), small crystallites were found with diameters of less than 10 nm. The difference between the XRD and TEM crystallite size results is likely to be related to fact that the shape factor was applied collectively. Nonetheless, the trend of smaller crystallite sizes for the films made using LNS is consistent for the results obtained via the two characterization methods.
tallite size results is likely to be related to fact that the shape factor was applied collectively. Nonetheless, the trend of smaller crystallite sizes for the films made using LNS is consistent for the results obtained via the two characterization methods.

Chemical Bonding and Elemental Content of HfO2 Films Deposited Using LNS
XPS analyses were carried out to determine the differences between the HfOx films deposited using H2O and LNS in terms of their chemical bonding characteristics and compositions. All the analyzed HfOx films were deposited at thicknesses of up to 20 nm on Si substrates. The measured XPS results were deconvoluted using a Shirley background and Gaussian line shapes. Figure 5a,b show Hf 4f XPS spectra of the HfOx films deposited using H2O and LNS, respectively. All the spectra consist of two peaks assigned to 4f5/2 and 4f7/2 electrons and could be fitted with two sets of the doublet peaks, for which the spin-orbit splitting was 1.68 eV. For the specimens prepared using the H2O oxidant (Figure 5a), peaks at 16.60 eV and 18.28 eV forming one of the doublets were assigned to Hf 4+ 4f5/2 and Hf 4+ 4f7/2 of stoichiometric HfO2, respectively [14,15]. The doublet at lower energy, 16.00 and 17.68 eV, was assigned to Hf suboxide (HfO2−x, 0 < x < 2), that is, the individual peaks in the doublet were assigned to Hf n+ 4f5/2 and Hf n+ 4f7/2 (n < 4), respectively [16,17]. It is apparent from the deconvolution results that the fully oxidized Hf 4+ doublet is much more intense than the suboxidized Hf n+ doublet ( Figure 5a). As shown in Figure 5b, for the film prepared using LNS, the doublet assigned to stoichiometric HfO2 was located at the same binding energy as that of the film prepared using H2O (Hf 4+ 4f5/2, 16.60 eV; Hf 4+ 4f7/2, 18.28 eV). Moreover, the positions of the peaks corresponding to Hf n+ 4f5/2 (16.15 eV) and Hf n+ 4f7/2 (17.83 eV) were similar to those for the film deposited with H2O as the oxidant. However, the Hf 4+ :Hf n+ ratio was different for the film prepared using LNS. The ratio of suboxidized Hf n+ was significantly increased. Figure 5c,d show O 1 s spectra of HfOx films deposited using H2O and LNS. These were deconvoluted into three components, respectively related to Hf-O bonding in stoichiometric HfO2, oxygen vacancies (Vo), and hydroxyl groups (-OH). In the spectra of both films, three peaks located at 530.1 ± 0.1, 531.2 ± 0.1, and 532.0 ± 0.1 eV, were identified by deconvolutions and assigned to H-O bonding, oxygen vacancies, and hydroxyl groups, respectively [18]. As shown in Figure 5c, for the HfOx film deposited using H2O as an oxidant, Hf-O bonding was found to be dominant. However, when the HfOx film was deposited using LNS, the oxygen vacancy and hydroxyl group contents were significantly higher, as shown in Figure 5d. Non-lattice oxygen peaks, such as oxygen vacancy peaks and hydroxyl peaks, contributed to suboxide content in the oxide layer [11,19].

Chemical Bonding and Elemental Content of HfO 2 Films Deposited Using LNS
XPS analyses were carried out to determine the differences between the HfO x films deposited using H 2 O and LNS in terms of their chemical bonding characteristics and compositions. All the analyzed HfO x films were deposited at thicknesses of up to 20 nm on Si substrates. The measured XPS results were deconvoluted using a Shirley background and Gaussian line shapes. Figure 5a,b show Hf 4f XPS spectra of the HfO x films deposited using H 2 O and LNS, respectively. All the spectra consist of two peaks assigned to 4f 5/2 and 4f 7/2 electrons and could be fitted with two sets of the doublet peaks, for which the spin-orbit splitting was 1.68 eV. For the specimens prepared using the H 2 O oxidant (Figure 5a), peaks at 16.60 eV and 18.28 eV forming one of the doublets were assigned to Hf 4+ 4f 5/2 and Hf 4+ 4f 7/2 of stoichiometric HfO 2 , respectively [14,15]. The doublet at lower energy, 16.00 and 17.68 eV, was assigned to Hf suboxide (HfO 2−x , 0 < x < 2), that is, the individual peaks in the doublet were assigned to Hf n+ 4f 5/2 and Hf n+ 4f 7/2 (n < 4), respectively [16,17]. It is apparent from the deconvolution results that the fully oxidized Hf 4+ doublet is much more intense than the suboxidized Hf n+ doublet ( Figure 5a). As shown in Figure 5b, for the film prepared using LNS, the doublet assigned to stoichiometric HfO 2 was located at the same binding energy as that of the film prepared using H 2 O (Hf 4+ 4f 5/2 , 16.60 eV; Hf 4+ 4f 7/2 , 18.28 eV). Moreover, the positions of the peaks corresponding to Hf n+ 4f 5/2 (16.15 eV) and Hf n+ 4f 7/2 (17.83 eV) were similar to those for the film deposited with H 2 O as the oxidant. However, the Hf 4+ :Hf n+ ratio was different for the film prepared using LNS. The ratio of suboxidized Hf n+ was significantly increased. Figure 5c,d show O 1 s spectra of HfO x films deposited using H 2 O and LNS. These were deconvoluted into three components, respectively related to Hf-O bonding in stoichiometric HfO 2 , oxygen vacancies (V o ), and hydroxyl groups (-OH). In the spectra of both films, three peaks located at 530.1 ± 0.1, 531.2 ± 0.1, and 532.0 ± 0.1 eV, were identified by deconvolutions and assigned to H-O bonding, oxygen vacancies, and hydroxyl groups, respectively [18]. As shown in Figure 5c, for the HfO x film deposited using H 2 O as an oxidant, Hf-O bonding was found to be dominant. However, when the HfO x film was deposited using LNS, the oxygen vacancy and hydroxyl group contents were significantly higher, as shown in Figure 5d. Non-lattice oxygen peaks, such as oxygen vacancy peaks and hydroxyl peaks, contributed to suboxide content in the oxide layer [11,19]. Thus, the O 1 s XPS analysis results are in accord with the Hf 4f XPS analysis results, confirming that the HfO x film deposited using LNS contained more suboxides.  When an HfOx film prepared using LNS was utilized as a resistive switching layer, a higher current density, compared to the HfOx film prepared using H2O, was measured in the highly resistive state, indicating that these films formed more current paths [11]. The high current was caused by the increased number of grain boundaries, because of the smaller sizes of the crystallites in the films deposited with the use of LNS. The increase in the non-lattice oxygen (Vo, -OH) content supplies defect states to the bandgap of oxide films. In addition, as grain boundaries are considered to be reservoirs of oxygen vacancies, the increase in the number of oxygen vacancies is expected with the increase in grain boundaries. Accordingly, the XPS results in Figure 5 are consistent with the previously reported resistive switching results [11], the XRD results as presented in Figure 3, and the TEM results presented Figure 4.
The widely reported crystalline phase of HfOx films is monoclinic. However, the doping of metal element into HfOx makes the film orthorhombic phase, showing ferroelectric characteristic. Therefore, to verify the presence or absence of La in the composition of the films, addition XPS data was analyzed. In general, in the XPS analysis for La, spin-orbit peaks of La 3d5/2 and La 3d3/2 appeared near 835 eV and 850 eV, respectively, and each spin-orbit component was further split via multiplet splitting [20]. However, the XPS data in Figure 6 only show background signals. It indicated that the La content was lower than the detection limit of our XPS analysis method. Thus, the results indicated that the La in the LNS oxidant hardly had very little influence on the chemical composition of the deposited HfOx films. The XRD results indicate that the HfOx existed in the monoclinic phase in the films, and the XPS results show that the La content was below the detection limit. Thus, we conclude that La in LNS is not affected by the chemical reaction with TEMAH, When an HfO x film prepared using LNS was utilized as a resistive switching layer, a higher current density, compared to the HfO x film prepared using H 2 O, was measured in the highly resistive state, indicating that these films formed more current paths [11]. The high current was caused by the increased number of grain boundaries, because of the smaller sizes of the crystallites in the films deposited with the use of LNS. The increase in the non-lattice oxygen (V o , -OH) content supplies defect states to the bandgap of oxide films. In addition, as grain boundaries are considered to be reservoirs of oxygen vacancies, the increase in the number of oxygen vacancies is expected with the increase in grain boundaries. Accordingly, the XPS results in Figure 5 are consistent with the previously reported resistive switching results [11], the XRD results as presented in Figure 3, and the TEM results presented Figure 4.
The widely reported crystalline phase of HfO x films is monoclinic. However, the doping of metal element into HfO x makes the film orthorhombic phase, showing ferroelectric characteristic. Therefore, to verify the presence or absence of La in the composition of the films, addition XPS data was analyzed. In general, in the XPS analysis for La, spin-orbit peaks of La 3d 5/2 and La 3d 3/2 appeared near 835 eV and 850 eV, respectively, and each spin-orbit component was further split via multiplet splitting [20]. However, the XPS data in Figure 6 only show background signals. It indicated that the La content was lower than the detection limit of our XPS analysis method. Thus, the results indicated that the La in the LNS oxidant hardly had very little influence on the chemical composition of the deposited HfO x films. The XRD results indicate that the HfO x existed in the monoclinic phase in the films, and the XPS results show that the La content was below the detection limit. Thus, we conclude that La in LNS is not affected by the chemical reaction with TEMAH, as observed for a ZrO 2 film prepared using LNS [10]. As a consequence, it can also be concluded that the HfO x deposited using LNS is probably not ferroelectric. as observed for a ZrO2 film prepared using LNS [10]. As a consequence, it can also be concluded that the HfOx deposited using LNS is probably not ferroelectric.

Relationship between Surface Energy and Crystalline Properties
It is plausible that the difference between the surface energies of the film specimens prepared using LNS and H2O influences the differences in growth rate and crystallite size. Various research groups have investigated the relationship between surface energy and molecule adsorption. Michiardi et al. reported that when a NiTi alloy underwent oxidation, the total free energy of the alloy increased, and the increase in surface energy caused an increase in the protein adsorption [21]. Moreover, Hayami and Otani reported that in the vapor-liquid-solid process of nanowire growth, for a surface with a higher surface energy, the droplet binding energy was higher [22]. Thus, for droplet binding, the (001) plane, with the highest surface energy, is preferential. In addition, many researchers have reported the correlation of three characteristics of surface energy, orientation, and growth rate. Penn et al. reported that titanium oxide nanoparticles grow rapidly along the [001] direction, driven by the relatively high surface energy of the (001) plane [23].
Finally, we present an explanation for the correlation between the surface energy, growth rate, and crystalline characteristics of the HfOx films. The proposed possible growth mechanisms for the HfOx films are shown in Figure 7. It is possible that adsorption is more favorable when the surface energy is higher. In HfOx films, the surface energy of the (111) plane is 21% higher than that of the (1 ̅ 11) plane [24]. When H2O is used as an oxidant, HfOx exists mostly as the (1 ̅ 11) plane, which has the lowest surface energy. The low surface energy prevents the precursor from being absorbed on the surface, and hence vertical growth of the HfOx film was slow. Meanwhile, to reduce the total free energy of the film, the size of the crystallites increases inside the

Relationship between Surface Energy and Crystalline Properties
It is plausible that the difference between the surface energies of the film specimens prepared using LNS and H 2 O influences the differences in growth rate and crystallite size. Various research groups have investigated the relationship between surface energy and molecule adsorption. Michiardi et al. reported that when a NiTi alloy underwent oxidation, the total free energy of the alloy increased, and the increase in surface energy caused an increase in the protein adsorption [21]. Moreover, Hayami and Otani reported that in the vapor-liquid-solid process of nanowire growth, for a surface with a higher surface energy, the droplet binding energy was higher [22]. Thus, for droplet binding, the (001) plane, with the highest surface energy, is preferential. In addition, many researchers have reported the correlation of three characteristics of surface energy, orientation, and growth rate. Penn et al. reported that titanium oxide nanoparticles grow rapidly along the [001] direction, driven by the relatively high surface energy of the (001) plane [23].
Finally, we present an explanation for the correlation between the surface energy, growth rate, and crystalline characteristics of the HfO x films. The proposed possible growth mechanisms for the HfO x films are shown in Figure 7. as observed for a ZrO2 film prepared using LNS [10]. As a consequence, it can also be concluded that the HfOx deposited using LNS is probably not ferroelectric.

Relationship between Surface Energy and Crystalline Properties
It is plausible that the difference between the surface energies of the film specimens prepared using LNS and H2O influences the differences in growth rate and crystallite size. Various research groups have investigated the relationship between surface energy and molecule adsorption. Michiardi et al. reported that when a NiTi alloy underwent oxidation, the total free energy of the alloy increased, and the increase in surface energy caused an increase in the protein adsorption [21]. Moreover, Hayami and Otani reported that in the vapor-liquid-solid process of nanowire growth, for a surface with a higher surface energy, the droplet binding energy was higher [22]. Thus, for droplet binding, the (001) plane, with the highest surface energy, is preferential. In addition, many researchers have reported the correlation of three characteristics of surface energy, orientation, and growth rate. Penn et al. reported that titanium oxide nanoparticles grow rapidly along the [001] direction, driven by the relatively high surface energy of the (001) plane [23].
Finally, we present an explanation for the correlation between the surface energy, growth rate, and crystalline characteristics of the HfOx films. The proposed possible growth mechanisms for the HfOx films are shown in Figure 7. It is possible that adsorption is more favorable when the surface energy is higher. In HfOx films, the surface energy of the (111) plane is 21% higher than that of the (1 ̅ 11) plane [24]. When H2O is used as an oxidant, HfOx exists mostly as the (1 ̅ 11) plane, which has the lowest surface energy. The low surface energy prevents the precursor from being absorbed on the surface, and hence vertical growth of the HfOx film was slow. Meanwhile, to reduce the total free energy of the film, the size of the crystallites increases inside the It is possible that adsorption is more favorable when the surface energy is higher.
In HfO x films, the surface energy of the (111) plane is 21% higher than that of the (111) plane [24]. When H 2 O is used as an oxidant, HfO x exists mostly as the (111) plane, which has the lowest surface energy. The low surface energy prevents the precursor from being absorbed on the surface, and hence vertical growth of the HfO x film was slow. Meanwhile, to reduce the total free energy of the film, the size of the crystallites increases inside the film, eliminating the grain boundaries of with relatively high energies due to lattice mismatch. However, in the case of LNS, a higher surface energy favors the adsorption of the precursor to lower the total energy of the film. It can be hypothesized that differences in precursor adsorption affect the vertical growth rate of HfO x films. In addition, high surface energies affect the crystallization of deposited films. Owing to the high surface energy, many additional nucleation sites existed on the surface of the film. The growth of crystallites at the nucleation sites occurred via the addition of atoms from the precursor. The nuclei grow into crystallites, and the crystallites also offer a surface with high surface energy. Therefore, crystallite growth and the generation of a high-energy surface occurs cyclically. For this reason, a faster HfO x film deposition rate and smaller crystallites were observed when LNS was used in the ALD process.

Conclusions
Nanocrystalline HfO x films were successfully synthesized by ALD using LNS. The ALD process and film characteristics were compared with those obtained when a conventional oxidant, H 2 O, was used. Typical ALD characteristics such as a self-limiting process, temperature window, and linear dependence of film growth on ALD cycle number were apparent. Interestingly, the growth rate when LNS was used was twice as high as that when H 2 O was used. The XRD results demonstrated that the HfO x films deposited with either LNS or H 2 O both consisted of the monoclinic phase, but there was a difference in the orientation preference. Using LNS, the preferred orientation was (111), which has a higher surface energy than the (111) orientation, the preferred orientation of HfO x prepared using H 2 O. The TEM results revealed that the crystallite size of the HfO x film grown using LNS was smaller than that of the film grown using H 2 O. The XPS results showed that the HfO x films prepared using LNS had more suboxides or defects. This was consistent with the fact that the TEM results revealed a higher number of grain boundaries because of the smaller size of the crystallites. Moreover, since it was established via XPS that La dopant atoms were not present in HfO x , it is not likely that this material is ferroelectric. Finally, we suggested a growth mechanism model based on the XRD, TEM, and XPS results. It was found that the high surface energy of HfO x films grown using LNS accelerates the adsorption of the precursor and offers more nucleation sites, resulting in small crystallites and a fast growth rate.