Stress-Induced In Situ Modification of Transition Temperature in VO2 Films Capped by Chalcogenide

We attempted to modify the monoclinic–rutile structural phase transition temperature (Ttr) of a VO2 thin film in situ through stress caused by amorphous–crystalline phase change of a chalcogenide layer on it. VO2 films on C- or R-plane Al2O3 substrates were capped by Ge2Sb2Te5 (GST) films by means of rf magnetron sputtering. Ttr of the VO2 layer was evaluated through temperature-controlled measurements of optical reflection intensity and electrical resistance. Crystallization of the GST capping layer was accompanied by a significant drop in Ttr of the VO2 layer underneath, either with or without a SiNx diffusion barrier layer between the two. The shift of Ttr was by ~30 °C for a GST/VO2 bilayered sample with thicknesses of 200/30 nm, and was by ~6 °C for a GST/SiNx/VO2 trilayered sample of 200/10/6 nm. The lowering of Ttr was most probably caused by the volume reduction in GST during the amorphous–crystalline phase change. The stress-induced in in situ modification of Ttr in VO2 films could pave the way for the application of nonvolatile changes of optical properties in optoelectronic devices.


Introduction
In optoelectronic components such as switches, waveguides, transistors, and memories, the operation principle requires the control of electronic signals by light irradiation or the control of photonic signals by an electric field. To realize such devices that are based on photon-electron interaction, materials that show phase transition accompanied by significant change in both electric properties (conductivity, etc.) and optical properties (reflectance, etc.) are strong candidates.
Vanadium dioxide (VO 2 ) undergoes structural phase transition near room temperature (68 • C in a bulk under the atmospheric pressure) between a high-temperature phase with a rutile-type structure (R phase) and a low-temperature phase with a monoclinic structure (M phase) [1,2]. The electrical, optical, and thermal properties of VO 2 abruptly change at the transition temperature (T tr ). In the high-temperature R phase, reflectance in the infrared region and electrical conductivity significantly increase compared to the M phase. The M-R phase transition of VO 2 can be induced not only by heat, but also by electric field [3,4], light [5,6], and mechanical strain [7], suggesting the possibility of realizing VO 2 -based electrical/optical switching devices operated with these stimuli.
When VO 2 is in the M (or R) phase at a given temperature, its T tr is supposed to be higher (lower) than that temperature. Therefore, if the T tr can be tuned for a certain temperature range, it means that the phase can be switched reversibly between R and M in this temperature range. This stimulus-induced phase switching might lead one to expect that VO 2 would be applied not only in optical switches, but also in electrical resistance change memory devices. However, the R phase induced by an external stimulus is volatile, i.e., it recovers to the M phase once the stimulus is removed. A certain continuous energy is generally required to maintain the R phase [8][9][10].
The use of strain could be a realistic approach to maintain the R phase of VO 2 , or to maintain the low T tr , without supplying continuous extrinsic energy. It is known that shortening of c R axis is accompanied by lowering of T tr in VO 2 . In what follows, the subscripts M and R indicate the phases of VO 2 . Muraoka and Hiroi reported the lowering of T tr in VO 2 films with short c R axes grown on rutile TiO 2 (001) substrates [11]. Cao et al. demonstrated a phase transition in a VO 2 microbeam by applying mechanical stress, and revealed a wide-ranging relationship between T tr and the strain along the c R axis [7]. If the strain were nonvolatile and reversible, then so would be the modification of T tr , and bistability would be realized in a temperature range in which the T tr could be modified by strain. Sources of strain in previous studies on the modulation of T tr in VO 2 films include lattice mismatches with the substrate or the underlayer [11][12][13] or doping using different elements [14,15]. However, the factor of modulation in these experiments was induced at the deposition stage, and hence, its T tr was no longer controllable after the formation of the film.
Still, there are several ways to modulate the strain of a thin film in situ. Attempts have been made to control the strain of thin films using piezoelectric materials [16][17][18][19][20][21][22][23][24][25][26]. They include some studies on the modification of the magnetic properties of (La, Sr)MnO 3 [16][17][18], CoFeB [19], and Ni [20] films, as well as some on the modification of the transition properties of VO x films [22][23][24][25][26], all grown on piezo layers or piezo substrates. It has been reported that the T tr of VO 2 films can be controlled through the strain of (1−x)Pb(Mg 1/3 Nb 2/3 )-xPbTiO 3 (PMN-PT) crystalline substrates by 1.35 • C [23] or by 6 • C [24]. Applying uniaxial pressure to a film with the tip of a scanning probe microscope also functions to induce local strain [27,28]. In the present report, we propose a method of capping the target thin film with a material in which amorphous-crystalline phase transition easily occurs. Generally, the density of a solid material in a crystalline phase is higher than that of the same composition material in an amorphous phase. Therefore, in a bilayered sample consisting of a strain-generator layer and VO 2 , one could expect that amorphous-crystalline phase changes in the strain-generator layer would cause in-plane compressive strain in the VO 2 layer, resulting in the modulation of T tr in VO 2 .
Ge 2 Sb 2 Te 5 (GST) is a typical material that undergoes reversible and nonvolatile switching between amorphous and crystalline phases. A number of researchers have worked on the application of this material in the field of optical and electrical memory devices. Previously, we studied the optical and thermal properties of GST [29][30][31][32] and developed several optical devices using it [33][34][35][36][37]. The amorphous-crystalline phase change of GST is accompanied by a volume contraction of 6.8% [38]. Supposing an isotropic volume change, this value can be converted to a linear compressive strain by 1 − (1 − 0.068) 1/3 = 2.3%. Assume that this strain is fully transferred to the VO 2 layer in touch with the GST layer, and that the c R axis of VO 2 lies in-plane. According to the relationship between T tr and the strain along c R axis in VO 2 microbeams, compressive strain for 2.3% along c R could lower the T tr by about 30 • C [7]. Moreover, previous reports on VO 2 thin films grown on TiO 2 (001) substrates implied that the strain effect on T tr was more pronounced in thin films than in microbeams [11,39]. These reports showed T tr values lower than those in the bulk by more than 50 • C in films with c R compression of only −0.6%. The controllability of T tr within several tens of • C implies switchability between the M and R phases in a range of several tens of • C, possibly satisfying the requirement of the operation temperature range of commercial devices.
To confirm the above concept, we herein report the in situ reduction of T tr in a VO 2 layer by amorphous-crystalline phase change in a GST layer that caps the VO 2 . During the preparation of this manuscript, we learned of a recent paper by Meng et al. on reflectivity measurements of GST/VO 2 bilayered samples [40]. We would like to note that our study focuses on the modulation of T tr in VO 2 , whereas they targeted the function as a four-value memory device by combining the phase changes of GST and VO 2 .

Sample Preparation
Bilayered films consisting of VO 2 and amorphous GST layers were prepared on sapphire substrates of either R-plane (1−102) or C-plane (0001). Figure 1 schematically shows the sample preparation processes. In order to realize the amorphous phase, the GST layer should be deposited after VO 2 , since the growth of crystalline VO 2 films requires a temperature higher than the crystallization temperature of GST (161 • C) [41].  2 was grown by means of either pulsed laser deposition (PLD) or rf-biased reactive magnetron sputtering on substrates heated at 500 and 400 • C, respectively. A bias power of 5 W was applied to the sample stage during the sputtering process. Other deposition conditions can be found in previous reports [42,43].
The GST film was deposited by a rf magnetron sputtering method in an argon (Ar) atmosphere of 0.5 Pa with an output power of 100 W. Intentional substrate heating was not carried out. We prepared two types of samples with and without a SiN x buffer layer between GST and VO 2 (see Section 3.2). The SiN x layer was grown by rf-sputtering a Si 3 N 4 target with a power of 200 W under 0.5 Pa of Ar. Crystallization of the GST layer was achieved by postannealing the sample at 200 • C (with a heating speed of 20 • C min −1 ) for 2 min in Ar atmosphere or in air. In what follows, the step with the pristine VO 2 layer before GST deposition is referred to as "VO", the step with the as-deposited GST layer on VO 2 as "AD", and the step after the postannealing of the bilayered sample as "PA". Figure 2 shows the X-ray diffraction (XRD) profiles of a GST/VO 2 bilayered sample in steps AD and PA at RT in a geometry aligned with respect to VO 2 40-2 M or 002 R diffraction peak. The weak intensity of the Al 2 O 3 substrate peaks is because the VO 2 100 M plane was slightly misoriented against the (1-102) plane of the substrate. The absence of GST peaks before annealing proved the amorphous nature of the as-deposited GST layer, while the diffraction peaks corresponding to NaCl-type GST that appeared after annealing indicated the success of the crystallization process. The c R axis is always supposed to lie in plane in VO 2 films grown on C-plane Al 2 O 3 , whereas one of three geometries allows the c R axis to lie in plane on R-plane Al 2 O 3 . Table 1 shows the thickness of each layer in the four samples (A-D) reported in the present article.

Characterization
To investigate T tr of VO 2 at each step, we performed temperature-controlled measurements of optical reflection intensity and electrical resistance.
Optical reflection intensity was measured in an Ar atmosphere of 1 atm using a heating/cooling stage (HFS-91, Linkam, Tadworth, UK) installed in an optical microscope. White light from a halogen lamp was shone on the sample surface through a half mirror, and the reflected light was detected by a laser power meter (Vega, Ophir, Jerusalem, Israel). The sample temperature was swept at a rate of either 5 • C min −1 or 3 • C min −1 . In order to perform the measurement to temperatures below RT, liquid nitrogen-cooled air was supplied in this stage. It is known that the temperature dependence of the optical properties of GST is negligible in the whole temperature range of the measurements (−20 • C minimum and 120 • C maximum) [29].
It is possible to evaluate the resistance of the VO 2 layer only when the insulating buffer layer is inserted between GST and VO 2 layers, since the resistivity of the crystalline GST is comparable with that of VO 2 . A 10 nm thick SiN x layer was employed as the insulating layer. The resistance as a function of temperature (R-T) of the VO 2 layer in GST/SiN x /VO 2 multilayered samples was measured with a two-probe scheme using tungsten-carbide (WC) probes. Both edges of the VO 2 layer were covered during deposition of SiN x and GST layers to make these areas accessible by the probes. The resistance was measured by a multimeter (2000, Keithley, Solon, OH, USA), while the sample temperature was swept with a homemade temperature control stage, in which the power for a Peltier device was controlled by a computer.

Optical Reflection
For Sample A with the ultrathin GST layer of 5 nm thick, we observed the reflection from the film side. Figure 3a-c show the temperature dependence of the optical reflection intensity of Sample A in steps VO, AD, and PA, respectively. The intensity is normalized with the values at 100 • C. Significant evolution of~10% in reflectance was detected even through the GST layer (Figure 3b,c). The sharp change of the reflectance was attributed to the phase transition of VO 2 . Figure 3d shows the temperature differential profiles of the reflected light intensity (dI/dT) in the heating runs in the three steps. A slight lowering of T tr , i.e., by about 4 • C, was revealed in step PA with respect to step AD. It is striking that the stress-induced modulation of T tr in VO 2 was realized even when the VO 2 layer was 10 times thicker (50 nm) than the GST layer (5 nm). For optical switch applications, combinations of a thin GST layer and a thick VO 2 layer may be of use.  For this sample, the reflection measurements were carried out in a substrate-side incident configuration, since the opaque nature of 200 nm thick GST prevented taking measurements on the film-side. The reflection intensity on the low-temperature side was stronger than that on the high-temperature side in step VO (Figure 3e), whereas the temperature dependence was inverted in steps AD and PA (Figure 3f,g). To confirm if such a difference in temperature dependence was reasonable, we calculated the reflectivity at a wavelength of 700 nm by using analytical solutions of electromagnetic waves [44] that propagate in multilayered structures, based on the dielectric functions of SiO 2 and VO 2 taken from the literature [45,46], and that of GST measured by us. For a sample with a pristine VO 2 layer, modeled with a SiO 2 (semi-infinite thickness)/VO 2 (30 nm thick)/vacuum (semi-infinite thickness) multilayered structure, the electromagnetic calculation reproduced higher reflectivity when the VO 2 was in the M phase compared to the R phase. With a GST layer [SiO 2 (semi-infinite)/VO 2 (30 nm)/GST (200 nm)/vacuum (semi-infinite) structure], on the other hand, the calculations revealed lower reflectivity in case of the M phase VO 2 compared to the R phase, regardless of phases in GST. These simulations were consistent with the experimental results, supporting the hypothesis that both types of abrupt changes in the reflection intensity observed in Samples A and B were caused by the temperature-induced M-R or R-M phase transition. Figure 3h shows the dI/dT profiles in the heating runs in the three steps. Either the valley or the peak in each curve is supposed to correspond to the phase transition. The center temperature of transition during heating runs existed at 68 and 72 • C in steps VO and AD, respectively, suggesting that deposition of the GST layer did not drastically affect T tr . In contrast, the postannealing process obviously lowered the T tr . The transition in step PA took place in a range of 25-50 • C with the center temperature being~41 • C, suggesting a lowering of T tr by approximately 30 • C caused by annealing. The T tr changed more drastically in Sample B compared with Sample A; this can be understood by supposing that the smaller the thickness ratio between VO 2 and GST layers, the more stress the VO 2 layer will suffer from the shrinkage of the GST layer.

Electrical Resistance
Here, one may wonder if the lowering of T tr reported above was an interdiffusion effect, which may have occurred during the postannealing at 200 • C. The possibility of the Ge-doping effect was rejected, since it is known to increase the T tr [47]. Still, a Sb-or Te-doping effect cannot be excluded at this moment. The migration of oxygen ions from VO 2 towards GST could cause the reduction of VO 2 , which could be another factor to decrease its T tr [48,49]. Proof of the strain effect on the decrease of T tr requires a way to prevent interdiffusion. In addition, the GST and VO 2 layers should be insulated from each other when one evaluates the electrical conductivity of the VO 2 layer in step PA, since the conductivity of the crystallized GST is comparable with that of R-phase VO 2 . To distinguish the GST crystallization effect from the interdiffusion effect, and to achieve R-T measurements of solely the VO 2 layer under GST, we prepared samples with a SiN x buffer layer, which was intended to play two roles, i.e., as an interdiffusion barrier and an insulator. Figure 4a shows the R-T curves of Sample C with a thick VO 2 layer (100 nm) in steps AD and PA. In both cases, the phase transition was clearly observed through significant resistance change, i.e., four orders of magnitude. The comparable resistance values in the M phase in both steps implied that the SiN x layer functioned as the insulator, preventing the current flow in the crystalline GST layer. Figure 4b shows temperature differential profiles of logarithm of the resistance of Sample C in the heating runs. One can see a shift of the valley between steps AD and PA, which suggests the decrease of the T tr caused by the postannealing, even with a configuration where the interdiffusion effect was excluded. It was assumed that this behavior was the result of shrinkage of the c R axis caused by in-plane compressive strain due to the crystallization of GST. The fitting to Pseudo-Voigt functions of the differential curves indicated a T tr of~74 and~72 • C in steps AD and PA, respectively. We performed a similar experiment for another sample with a thinner VO 2 layer (6 nm, Sample D). The R-T curves of the sample showed a resistance change of more than one order of magnitude, suggesting the existence of a weak phase transition (Figure 4c). The fitting of the temperature differential profiles revealed T tr of~58 and~52 • C in steps AD and PA, respectively, suggesting a lowering of T tr for about 6 • C during the crystallization of GST (Figure 4d). Thinner than the VO 2 layer in sample C, the VO 2 layer in this sample was probably more severely affected by the stress from GST. To qualitatively observe the difference of the degree of stress, we performed XRD 2θ − ω scans of Samples C and D in steps VO and PA ( Figure 5). In Sample D, a shift of the VO 2 020 M peak to a lower angle occurred when the ultrathin VO 2 layer was capped with crystalline GST. The elongation of the out-of-plane lattice suggested the shrinkage of the in-plane lattice and c R axis, which is in agreement with our understanding. In contrast, no significant shift was observed in Sample C with a thick VO 2 layer, which was consistent with the less drastic shift of its T tr .

Discussion
The transition temperatures obtained from all the experiments in this study are summarized in Table 1. It is easy to predict that the lower the thickness ratio between VO 2 and GST layers, the larger the stress that VO 2 suffers from GST when the GST is crystallized. The tendency found in the results, i.e., a more significant drop of T tr in the sample with smaller VO 2 /GST thickness ratio, supports the hypothesis that the modification of T tr was induced by the stress which occurred during the crystallization of GST. In the present study, mechanical strain was introduced into the VO 2 films in situ, unlike the previous films on TiO 2 (001) with static strain. On the other hand, studies on piezo-induced in situ T tr modification of VO 2 resulted in reductions of only 1.35 • C [23] or 6 • C [24]. The strain that would be induced through the volume change at an amorphous-crystalline phase change (2.3%) could be significantly superior to that induced through the piezoelectric effect (± 0.2%) [50].
Nevertheless, the shift of T tr for~6 • C, observed in R-T measurements (Sample D), was not as huge as that observed in reflectance measurements,~30 • C (Sample B). One reason for this may be that the optical reflectance is sensitive in detecting the property change at the GST/VO 2 interface, whereas the electrical resistance contains the property of the VO 2 film for the whole thickness. The VO 2 lattice at the film/substrate interface was probably pinned by the substrate lattice, even when that at the GST/VO 2 interface was shrunk due to stress from GST. The pinning effect could be pronounced when the VO 2 layer is thin. Optimization of the thicknesses of the GST, VO 2 , and buffer layers would be required. Another reason for this, in particular regarding Samples C and D, could be that the SiN x layer may have absorbed a large part of the stress from the GST layer and weakened the deformation of the VO 2 . ZnS-SiO 2 , an insulating material commonly used in optical disks, is supposed to be softer than SiN x , and could be useful for improving the transport efficiency of the stress. The change of T tr for~30 • C in the present results may not be large enough for device applications, and therefore, transportation of the 2.3% strain from GST to VO 2 with a higher efficiency would be sought after.
More importantly, the application to optical devices such as memories and switches would require an amorphization process of the crystalline GST layer. Amorphization is currently performed in commercial optical disks by rapid local heating of the material, which should be in the order of µm 2 in area and tens of ns in duration [51]. Such a process will be examined in our future research.

Conclusions
In GST/VO 2 bilayered films prepared on Al 2 O 3 substrates, the phase transition properties of the VO 2 layers were compared between the steps before and after crystallization of the GST layer. It was shown that the crystallization of the GST layer lowers the phase transition temperature of the VO 2 layer, either with or without a SiN x buffer layer. The nonvolatile modification of T tr was probably induced through strain in the VO 2 layer, which originated in the volume shrinkage of the GST layer at its amorphous-crystalline phase change. The shift of T tr caused by the crystallization of the GST layer was by approximately 30 • C for a GST/VO 2 bilayered sample with thicknesses of 200/30 nm. Once an amorphization process has been established, the presently-described devices will possibly be proved to function as reversible, nonvolatile resistance change memory or optical switching devices with operation temperature ranges of several tens of degrees Celsius. The operation mechanisms of the present samples would represent new guidelines in the strain engineering field, and would greatly broaden the possibilities of strain-driven devices.