■ INTRODUCTION

Layered van der Waals (vdW) materials have revolutionized the photonic devices with their ease of integration in forming heterostructures 1 and their intriguing characteristics that have triggered novel applications. Specifically, the naturally occurring hyperbolicity of vdW materials such as hexagonal boron nitride (hBN) and α-MoO 3 has led to the prediction and demonstration of canalization, 2 sub-wavelength imaging, 3 negative refraction, 4 photonic magic angle, [5][6][7] and high infrared absorption. 8 The inherent layered nature of these vdW materials is quite valuable in integrating with other materials to create structures whose optical and electrical responses can be tuned dynamically.

Phase-change materials have proven to be an indispensable tool in the area of optical modulation, offering repeatable and active modulation. Since the first revelation of its dynamic phase-change properties in 1959, 9 the interest in the active modulation properties of VO 2 has been thriving. When the temperature applied to VO 2 increases beyond a critical point (T c ), insulator-to-metal transition (IMT) takes place, which renders its monoclinic (insulator) phase rutile (metallic). 10 While the microscopic dynamics of this intriguing phase transition for VO 2 remains elusive, 11 the experimental implementation of it in devices has been successful in the recent decade. The lower T c required for the IMT of VO 2 than that of other phase-change materials, 9,12 along with a variety of possible routes to achieve IMT (thermally, 13 electrically, 14 mechanically, 15 and optically 16,17 ), has provided ample opportunities to tailor the IMT of VO 2 to numerous applications. 18,19 The essence of optical modulation in the mentioned photonic applications is the IMT of VO 2 , which marks a sizeable change in the refractive index of this material. Some of the notable applications include optical diodes, 20 tunable metamaterials, [21][22][23] thermal emitters, 10,24 and infrared absorbers. 25 Marrying the tunability of VO 2 and the rich optical response of vdW materials through intricate designs, researchers have successfully demonstrated devices with tunable optical responses. 26,27 In near-field optics, α-MoO 3 and hBN are known to support hyperbolic phonon polaritons that yield increased photonic density of states and are highly sensitive to the optical characteristics of the immediate environment. This very fact has led to the realization of tunable hyperbolic phonon polaritons in both intensity and frequency through scattering-type scanning near-field optical microscopy characterization. 28,29 However, reports on the far-field tunability of the optical response of hyperbolic materials with VO 2 do not exist to the best of our knowledge. Unlike phonon polaritons which require momentum-matching techniques, the lattice vibration modes known as optical phonons (OPhs) can directly be coupled to photons. 30 α-MoO 3 is an outstanding example of a vdW material that has three Reststrahlen (RS) bands in the infrared region and demonstrates in-plane anisotropy from ultraviolet to infrared regions with three orthogonal OPh modes in the infrared region near respective transverse optical (TO) frequencies. [31][32][33][34][35][36] Benefitting the inplane anisotropy of α-MoO 3 through OPhs can be consequential for polarization-sensitive photonic devices in the second infrared atmospheric window (8-14 μm, equivalent to 700-1400 cm -1 ). 30,37 Another untapped potential of α-MoO 3 is its high reflectance in this atmospheric window, which can be consequential for radiative cooling and thermal camouflage applications, 38,39 which have been revitalized thanks to thin films 40 and metamaterials 39 research. 41,42 Camouflage requires high reflectance within the transmissive window. 39,43 The signature of the RS band is its high reflectivity; the RS bands in [100] and [010] directions of α-MoO 3 lie within this atmospheric window, which can make this material highly suitable for camouflage applications. The combination of α-MoO 3 and VO 2 can also pave the way to scalable tunable thermal camouflage applications. 43,44 Here, we experimentally demonstrate the tunability of OPhs and cavity modes in α-MoO 3 with the IMT of VO 2 through far-field Fourier transform infrared (FTIR) measurements in a multilayer structure. From the experimental results, we find that the frequency and intensity of the reflected signal can be modulated. Our simulated results shed light on the origin of this spectral behavior. We finally conclude by demonstrating the inherent tunability in the OPh frequency and absorption intensity in α-MoO 3 that is possible through this type of heterostructure, which can pave the way for tunable, large-scale optical components for infrared applications.

■ RESULTS AND DISCUSSION

Design and Experiments. The structure investigated in this study is a multilayer system composed of a 90 nm VO 2 film on a SiO 2 substrate (Figure 1a) and an α-MoO 3 flake which is transferred onto VO 2 -SiO 2 (Figure 1b). The temperature-dependent reflectance measurements are carried out using the FTIR system where the sample is placed onto a heating ceramic, as illustrated in Figure 1a,b. The measured spectral reflectance curves for different temperature values are depicted in Figure 1c,d for samples without (Figure 1a) and with the α-MoO 3 flake (Figure 1b), respectively.

As the temperature is increased in steps of 10 °C (Figure 1c,d), the IMT takes place, and the reflectance increases due to the metallic nature of VO 2 after the IMT. This dramatic change in spectral reflectance occurs around 70 °C, which corresponds to the VO 2 phase transformation being triggered. 11 From the reflectance curves for the sample with α-MoO 3 (Figure 1d), we observe a shift in the reflectance intensity for OPhs in the [100] direction (OPh x ) at 812 cm -1 . 30 Also evident from Figure 1d is a larger dip in reflectance for the room-temperature (RT) curve at 789 cm -1 , which corresponds to the Fabry-Perot (FP) mode due to the thick α-MoO 3 layer (air-MoO 3 -VO 2 stack) which dissipates in the VO 2 layer. This mode blue-shifts (moves to a higher frequency) as the temperature increases due to the modified phase shift imparted on the reflected light off of the MoO 3 -VO 2 interface. There is a small dip at 1008 cm -1 , evident in Figure 1d, which is due to the OPhs in the [010] direction of α-MoO 3 (OPh z ). For ω < 1000 cm -1 , the reflectance dips are attributed to the OPhs of SiO 2 . The modes are modeled and discussed in further detail in the following section.

■ DISCUSSION

In order to investigate the impact of the phase change on OPhs of α-MoO 3 , we first studied VO 2 in isolation. The spectral reflectance for the glass-VO 2 system (Figure 1a) is represented in Figure 1c. The IMT process for VO 2 starts with metallic islands that expand as temperature is increased, which can be modeled with the effective medium theory (EMT) 45,46

where ε eff,j is the effective dielectric function in direction j. ε I and ε M stand for, respectively, the insulator and metal phases of VO 2 taken from Wan et al. 47 (Supporting Information). The filling factor (FF) is the in-plane filling ratio of the metallic islands to the insulator host within the VO 2 film, which approaches 1 as the temperature is increased. L j is the depolarization factor in direction j, which represents the shape of the metallic inclusion and satisfies L x + L y + L z = 1. 48 Given the in-plane (transverse) isotropy of VO 2 , the EMT yields inplane and out-of-plane permittivities (ε x = ε y ≠ ε z ). Since the heating is applied in steps of 10 °C with long wait times (slow heating steps), the inclusions are disk-like in shape, with the disk axis parallel to the optical axis (z direction). This translates to L x = L y ≈ 0 and L z ≈ 1. 46 The Lorentz formulation is used for the SiO 2 model (Supporting Information), and a value of FF = 0.5 is used for the curves corresponding to 60 °C. α-MoO 3 is modeled with the Lorentz equation 34 (Supporting Information). The mentioned dielectric models are employed to a 4 × 4 transfer matrix method (TMM) 49 to simulate the reflectance curves for three distinct temperature values, that is, RT, 60 °C, and 90 °C. The simulation results for the sample without the α-MoO 3 layer are depicted in Figure 2a (dashed curves), which agree well with their corresponding measurement results (solid curves). The incident polarization angle is 30°for all measurements and simulations. The weighted-average incidence angle of the FTIR spectrum is 15°, which is also used for simulations. The reflectance simulations for the structure with α-MoO 3 are compared to the measurements in Figure 2b, and Figure 2c illustrates a close-up view of the FP and OPh x modes. As temperature rises from RT to 90 °C (hot), the measured FP mode demonstrates a frequency shift of 4 cm -1 from 789 to 794 cm -1 and an intensity change of 0.15 from 0.27 to 0.42. The simulated values suggest a similar shift of 7 cm -1 in frequency and 0.26 in intensity. The measured OPh x mode at 812 cm -1 demonstrates an intensity shift of 0.11 and a frequency shift of 2 cm -1 . Although the simulations show a frequency shift of 0.94 cm -1 for the OPh x mode, such a small shift is too close to the highest resolution of our FTIR system, which is 2 cm -1 , which can explain the higher shift observed in measurements. Small frequency shifts are also expected in the literature for thin layers. 50 The OPh z mode (997 cm -1 ) in Figure 2d demonstrates a frequency shift of 2 cm -1 and an intensity shift of 0.28 when the temperature increases from RT to hot. The simulations suggest a frequency shift of 1.2 cm -1 .

The reflectance value for the RT phase remains high for the transmissive window (700-1250 cm -1 ) and is low for the nontransmissive window (1250-2000 cm -1 ), demonstrating camouflage characteristics. Once VO 2 changes the phase to hot, aside from the increased reflectance (better camouflage capability), the absorption is also increased (Supporting Information) within the nontransmissive window (1250-2000 cm -1 ). Hence, the emissivity increases within this band, and radiative cooling is achieved in the hot phase.

The highest observed frequency shift is for the FP (or the Etalon) mode trapped inside α-MoO 3 and dissipated to VO 2 . The average thickness of the measurement area for the α-MoO 3 layer is 1.27 μm (Supporting Information). Figure 3 demonstrates the spectral reflectance for different α-MoO 3 thicknesses (t MO ) for RT and hot phases of VO 2 . The blue traces in Figure 3a (RT) starting near 800 cm -1 to smaller frequency values are the FP modes in α-MoO 3 . As the thickness increases, more FP modes are realizable in α-MoO 3 . A similar scenario is observed in the hot simulation results of Figure 3b. The sharp vertical blue lines near 1000 cm -1 in Figure 3a,b are OPh z modes. The red regions between 800 and 1000 cm -1 in Figure 3a,b are high-reflectance bands representing the RS bands in the [100] direction of α-MoO 3 , a signature of the natural hyperbolicity of this material. As temperature is increased, the reflectance increases in general due to the IMT of VO 2 . The behavior of reflectance and transmittance can be explained by investigating absorption. The lossy behavior of α-MoO 3 is overshadowed by VO 2 for ω > 1000 cm -1 and ω < 800 cm -1 (Supporting Information). As temperature increases, the VO 2 layer becomes more lossy. 15 For 800 cm -1 < ω < 1000 cm -1 , the RS band in the [100] direction of α-MoO 3 reflects the incident light, and the light does not reach VO 2 underneath to be dissipated. An insightful and simple definition to understand the dominance of the absorption behavior of VO 2 is the electric loss tangent. 51,52 The absolute inverse electric loss tangent is defined as tan -1 δ e = |ε 1 |/ε 2 , where ε 1 and ε 2 are, respectively, the real and imaginary parts of the permittivity. A lower inverse tangent signifies higher absorption, which is the case for VO 2 . For ω < 800 cm -1 , for example, VO 2 is the most lossy material in the multilayer system (Supporting Information). To further verify the observed modes and shifts, another sample is fabricated with layers consisting of Pt (100 nm), VO 2 (90 nm), and α-MoO 3 from the bottom to the top, as illustrated in Figure 4a. This structure includes a 220 nm α-MoO 3 layer (Supporting Information). Using Pt as a bottom reflector layer isolates the observed reflectance from the effects of glass phonon modes near 997 cm -1 . Figure 4b demonstrates the simulated spectral reflectance versus FF (which is temperature-dependent) plot that clearly outlines the observed OPh x and OPh z modes, respectively, near 810 and 1006 cm -1 . Due to the thinner α-MoO 3 layer in this sample, FP modes do not exist. Figure 4c,d demonstrates the measured (solid) and simulated (dotted) spectral reflectance at three different temperature values. The experimental OPh x mode (dip) in Figure 4c demonstrates a blue shift of 2 cm -1 and an intensity shift of 0.1 as temperature is increased to hot from RT (inset of Figure 4c). Simulated shifts are 0.53 and 0.08 cm -1 . Figure 4d demonstrates the measured and simulated spectral reflectance curves for the same sample without α-MoO 3 , which agree well. The difference between the simulation and measurement mainly arises for the intensity of the resonances, which is expected due to the sharpness of these resonances and the limiting resolution of FTIR spectra. Throughout this study, the chosen thickness values for α-MoO 3 and VO 2 were not the optimum case and were chosen as the reported values due to the fabrication and transfer challenges that are concomitant to current VO 2 and α-MoO 3 growth techniques. The frequency and intensity tuning can be further enhanced by modifying the thicknesses of the layers (Supporting Information).

■ CONCLUSIONS

We demonstrated experimentally a frequency tuning of 2 cm -1 for both OPh x and OPh z modes in [100] and [010] crystal directions of α-MoO 3 . We reported the reflectance intensity tuning of 11% and 0.28 for these phonon modes. Using the EMT and dielectric models of each layer along with the TMM and finite-difference time-domain simulations, the measurements were verified with simulations. The origin of observed and simulated spectral characteristics was also traced back using loss tangent analysis. The effect of the glass substrate was eliminated using a Pt metal layer instead, which further verified the shifts and modes. Our findings can be tailored to the design of advanced tunable photonic devices in the second atmospheric window in the infrared region.

■ METHODS VO 2 Deposition. The VO 2 thin films were deposited in a 5 mTorr O 2 environment at 550 °C substrate temperature, and the PLD laser energy was set to be 321 mJ with 5 Hz pulse frequency. A postdeposition annealing at 550 °C for 30 min was performed in the same 5 mTorr O 2 environment. This temperature is safe for the thick Pt layer and does not allow for the dewetting of the underlying Pt layer. 53 The follow-up step does not introduce thermal shock to the sample since the flakes are grown and transferred.

α-MoO 3 Deposition and Transfer. α-MoO 3 flakes were grown using low-pressure physical vapor deposition. For this process, 50 mg of MoO 3 (Sigma-Aldrich) powder was spread evenly within an alumina boat. This boat was placed within a 1 inch diameter quartz tube and at the center of a small Lindberg tube furnace. A 41 in. 2 rectangular piece of a SiO 2 /Si wafer (300 nm oxide thickness) was placed face-up downstream in a colder zone of the furnace. These pieces were suspended on the top of the alumina boat and were located roughly 4 cm from the center region. The pressure was maintained at 2.8 Torr with a carrier gas of O 2 at a flow rate of 25 sccm. The center of the furnace was then heated to 675 °C over a period of 25 min and then to 700 °C over a period of 5 min. Upon reaching 700 °C, the furnace was immediately opened, thereby quenching the deposition. A simple tape transfer was adopted to deposit the flakes onto VO 2 films.

FTIR Characterization. Mid-infrared reflectance measurements were obtained with a Hyperion 2000 IR microscope coupled to a Bruker Vertex 70 FTIR spectrometer. For the FP structures, a mercury cadmium telluride detector was used. The Cassegrain objective was 15×, and the aperture dimensions were 30 × 30 μm 2 .

program (NSF DMR-1720319) at the Materials Research Center; the IIN; the Keck Foundation; and the State of Illinois, through the IIN.

■ ABBREVIATIONS α-MoO 3 , α-phase molybdenum trioxide h-BN, hexagonal boron nitride FP, Fabry-Perot IMT, insulator-to-metal transition EMT, effective medium theory FF, filling factor OPh, optical phonon TO, transverse optical LO, longitudinal optical RS band, Reststrahlen band vdW, van der Waals s-SNOM, scattering-type scanning near-field optical microscopy FDTD, finite-difference time-domain FTIR, Fourier transform infrared