**3. Results and Discussion**

Figure 1 shows the XRD patterns of VO2 film measured at 30 and 90 ◦C, which demonstrate the structural transition of VO2 film from monoclinic (30 ◦C) to tetragonal rutile (90 ◦C) phase. The VO2 film measured at 30 ◦C (Figure 1a) displays that two XRD peaks located at 2θ of 27.9◦ and 55.4◦ can be indexed to the (011) and (220) planes of monoclinic VO2(M) (JCPDS no.: 82–0661), respectively. When the temperature is raised to 90 ◦C (Figure 1b), two XRD peaks located at 2θ of 27.6◦ and 55.4◦ are detected, which can be assigned to the (110) and (211) planes of tetragonal rutile VO2(R) (JCPDS no.: 79–1655), respectively. In addition, because the VO2 film is grown on Si substrate, an obvious XRD peak of Si (113) at 2θ of 51.9◦ has also been detected (Supplementary materials: Part 1). Figure 1c is a comparison of Figure 1a,b, which clearly shows an XRD peak shift of VO2(M) (011) to VO2(R) (110) peak in 26◦ ≤ 2θ ≤ 30◦ and the XRD peak of Si substrate does not shift. The XRD peak-shifting behavior is a diagnostic feature for the phase transition of VO2 film from monoclinic to tetragonal rutile structure. Previously, the similar XRD peak-shifting phenomenon had also been reported by Wu et al. [56] for confirming the phase transformation of monoclinic VO2 to tetragonal rutile VO2.

**Figure 1.** XRD patterns of VO2 film measured at (**a**) 30 ◦C and (**b**) 90 ◦C. (**c**) A comparison of (**a**) and (**b**).

Figure 2 shows the SEM and TEM analyzed results for surface and cross-sectional microstructures of the VO2 film, respectively. The SEM image of surface morphology of the VO2 film (Figure 2a) displays a conformal VO2 film with bigger grains surrounded by small grains; the grain sizes of big and small grains are about 78 ± 14 and 40 ± 6 nm, respectively. According to the cross-sectional TEM bright-field and dark-field images (Figure 2b,c), it can be clearly observed that the VO2 film is grown on a native oxide layer (SiO*x*) of Si substrate and constructed from columnar grains. The thickness of the VO2 film is about 30 nm, and displays its growth rate at about 0.03 nm/cycle. Moreover, the VO2

grains are directly crystallized and grown on the top surface of the native oxide layer, which is verified by the high-resolution TEM (HR-TEM) image of the VO2/SiO*x*/Si interface (Figure 2d). The HR-TEM image also reveals a clear lattice fringe of about 0.32 nm, corresponding to the (011) plane of VO2(M). The selected-area electron-beam diffraction (SA-EBD) pattern obtained by focusing the electron beam on an individual VO2 grain is shown in Figure 2e; the SA-EBD pattern can be indexed to monoclinic VO2(M) in agreement with the XRD results.

**Figure 2.** Microstructure analyses of VO2 film. (**a**) An SEM top-view image. TEM cross-sectional (**b**) bright-field and (**c**) dark-field images. (**d**) A high-resolution TEM (HR-TEM) image of the VO2/SiO*x*/Si interface. (**e**) The selected-area electron-beam diffraction (SA-EBD) pattern obtained by focusing the electron beam on an individual VO2 grain.

Figure 3 shows the V2p, O1s, and Cl2p XPS spectra for original (before etching) and after argon-ion etching surface of VO2 film. In addition, the chemical composition of VO2 film calculated from XPS spectra are shown in Table 1. The V and O concentrations of VO2 film are about 25.7 at.% and 74.3 at.%, respectively, for the film before surface etching and are about 33.1 at.% and 66.9 at.%, respectively, for the film after surface etching. The results clearly indicate that the original surface of the VO2 film has a higher oxygen concentration because the VO2 film was exposed to air (oxygen-rich) environment, resulting in absorption of oxygen and a native oxide layer (overoxidation layer) forming on the surface of the VO2 film [26,33,34,39,41,43,57,58]. After argon-ion etching, surface contamination and the native oxide layer of the VO2 film had been removed, and the atomic proportion of V:O atom was about 1:2 in agreement with the stoichiometry of VO2. Besides, no Cl impurity had been detected in the VO2 film, demonstrating the Cl concentration in the VO2 film was lower than the detection limit of XPS (approximately 0.1 at.%). It is noteworthy that this work successfully achieved VO2 film with high purity (Cl impurity <0.1 at.%) by using a low growth temperature of 350 ◦C, which can be attributed to the additional pump-down steps in the ALD reaction cycles effectively evacuating excess precursors and byproducts [54]. In a previous study, Cheng et al. reported that implementation of pump-down steps into the gaseous-pulse cycle of ALD can effectively reduce the Cl residues. They used TiCl4 as ALD precursor to grow TiN films by using conventional four-step ALD and modified six-step ALD (adding two pump-down steps). Their results showed that the Cl residues of TiN films can be decreased from about 7.7 at.% to 2.3 at.% at the growth temperature of 300 ◦C [54].

In Figure 3a, the V2p3/2 peak of the original VO2 film (before surface etching) can be fitted with two peaks at binding energy of about 517.2 and 515.6 eV, which can be assigned to V5+ and V4+, respectively [26,33,34,39,41,43,57,58]. Musschoot et al. [43] and Sliversmit et al. [57] reported that the V5+ signal is mainly contributed from the native oxide layer (overoxidation layer) of VO2 film. After surface etching (to remove the native oxide layer), the V2p3/2 peak has a maximum at 515.6 eV (assigned to V4+), which primarily confirms VO2 stoichiometry. In Figure 3b, the O1s XPS peaks are located at binding energy of about 529.8 and 530.5 eV for the original and after-surface-etching VO2 film, showing a peak-shifting phenomenon. The similar peak-shifting phenomenon of O1s XPS peak for VO2 film after surface etching by argon ion sputtering had also been observed by Musschoot et al. [43].

**Figure 3.** (**a**) V2p, (**b**) O1s, and (**c**) Cl2p XPS spectra for original (before etching) and after argon-ion etching surface of VO2 film.



Figure 4a,b show the selected temperature-dependent Raman spectra of the VO2 film for heating and cooling cycles, respectively. It is noticed that the full temperature-dependent Raman spectra of the VO2 films for temperatures between 30 and 80 ◦C in heating and cooling cycles are shown in Figures S1 and S2, respectively. As shown in Figure 4, four Raman peaks at 194, 224, 305, and 616 cm−<sup>1</sup> are associated with the monoclinic phase VO2 [25,26,59–61]. The peaks of 194, 305, and 616 cm−<sup>1</sup> are assigned to Ag phonon vibration modes [25,26,59,60] and the peak of 224 cm−<sup>1</sup> can be assigned to Ag + Bg mode [61]. The low-frequency phonons at 194 and 224 cm−<sup>1</sup> relate to lattice motion involving V–V bonds, while the other peaks are attributed to V–O bonds [26,59–61]. Peaks located at 301, 520, and 935–990 cm−<sup>1</sup> are contributed from the silicon substrate that compared with the Raman spectrum of the silicon substrate (Figure 5). Moreover, the phonon intensities of 194, 224, and 616 cm−<sup>1</sup> gradually disappear as the temperature increases and display the reversibility during the cooling cycle. However, the peak intensity of 305 cm−<sup>1</sup> does not show an evident change due to an overlap signal between 305 and 301 cm−<sup>1</sup> for VO2 and silicon substrate, respectively.

**Figure 4.** Temperature-dependent Raman spectra of the VO2 film: (**a**) heating cycle, (**b**) cooling cycle, and relative Raman intensity of the Ag phonon mode at (**c**) 194, (**d**) 224, and (**e**) 616 cm<sup>−</sup>1.

**Figure 5.** Raman spectrum of the silicon substrate.

Furthermore, the plots of normalized Raman intensity variations for the Ag phonon vibration mode at 194, 224, and 616 cm−<sup>1</sup> are shown in Figure 4c,d,e, respectively. The normalized Raman intensity of VO2 film was calculated from the equation below:

$$I\_{\rm NRI} = \frac{I\_T - I\_{\rm 80}}{I\_{\rm 30} - I\_{\rm 80}} \tag{1}$$

where *I*NRI is the normalized Raman intensity, *IT* is the Raman intensity measured at indicated temperature (*T*), *I*<sup>30</sup> and *I*<sup>80</sup> are the Raman intensities measured at 30 and 80 ◦C, respectively. It can be seen clearly that the plots of Raman intensity vibrations show a hysteresis feature for Raman shift at 194, 224, and 616 cm−1. The phase transition temperatures of VO2 film estimated by the differential curves (as inserts) are about 65, 63.9, and 64.5 ◦C for 194, 224, and 616 cm−<sup>1</sup> in the heating process, respectively. In the cooling process, the phase transition temperatures of VO2 film are about 57.6, 56.6, and 58.7 ◦C for 194, 224, and 616 cm−1, respectively. Therefore, the overall SMT temperatures

estimated from the middle of the hysteresis curves are about 61.3, 60.25, and 61.6 ◦C for 194, 224, and 616 cm<sup>−</sup>1, respectively.

The temperature-dependent sheet-resistance (SR) variation of VO2 film is shown in Figure 6, displaying a thermal hysteresis variation. Besides, the SR variation has approached two orders of magnitude across the semiconductor-to-metal transition (SMT) of the VO2 film (SR changed from 2.2 × 104 to 2.7 × <sup>10</sup><sup>2</sup> <sup>Ω</sup>/ for the temperature raised from 40 to 80 ◦C) so that the value of the resistance ratio agrees with the typical VO2 film thickness less than 50 nm (typically, the resistance ratios of most VO2 films across the SMT are in the range of 102–103 for thickness <50 nm) [29]. Furthermore, a sharp drop of SR can be clearly observed in the heating cycle, determining a phase transition temperature of about 63 ◦C, and a sharp rise of SR in the cooling cycle with a phase transition temperature of about 56 ◦C can be also seen in Figure 6. Therefore, the SMT temperature estimated from the middle of thermal hysteresis SR variation is about 60.0 ◦C.

**Figure 6.** Temperature-dependent sheet-resistance variation of the VO2 film.

There are several parameters that may affect the temperature-dependent electrical properties of VO2, such as changes in impurity content, stoichiometry, strain, oxygen vacancies, and the presence of grain boundaries [8–12]. In this work, the SMT temperature of VO2 film evaluated from the temperature-dependent Raman spectra and sheet-resistance variation is about 61 ± 1 ◦C, slightly different from the well-known 340K (~67 ◦C), which can be reasonably attributed to the influence of grain boundary density because the VO2 film has a polycrystalline structure with considerable grain boundaries.
