betaFeedback

WorksPhysics and Astronomy

Active Loss Engineering in Vanadium Dioxide Based BIC Metasurfaces

Abstract

Metasurfaces have unlocked significant advancements across photonics, yet their efficient active control remains challenging. The active materials required often lack continuous tunability, exhibit inadequate refractive index (RI) changes, or suffer from high losses. These aspects pose an inherent limitation for resonance-shifting based switching: when RI changes are small, the resulting shift is also minor. Conversely, high RI changes typically come with high intrinsic losses necessitating broad modes because narrow ones cannot tolerate such losses. Therefore, larger spectral shifts are required to effectively detune the modes. This paper introduces a novel active metasurface approach that converts the constraint of high intrinsic losses into a beneficial feature. This is achieved by controlling the losses in a hybrid vanadium dioxide (VO2_{2}) - silicon metasurface, supporting symmetry-protected bound states in the continuum (BICs) within the infrared spectrum. By leveraging the temperature-controlled losses in VO2_{2} and combining them with the inherent far-field-coupling tunability of BICs, we gain unprecedented precision in independently controlling both the radiative and nonradiative losses of the resonant system. Our dual-control mechanism allows us to optimize our metasurfaces and we experimentally demonstrate quality factors above 200, a maximum reflectance amplitude of 90%, a relative switching contrast of 78%, and continuous tuning from under- to over-coupling within the infrared spectral range. This study provides a foundation for experimentally and technologically simple, fine-tunable, active metasurfaces for applications ranging from molecular sensors to filters and optical modulators.

manuscript.pdf

INTRODUCTION

Metasurfaces[1, 2] have achieved remarkable advancements in photonics, including flat lenses[3, 4], holograms5,6, molecular sensors7,8, lasers9,10, and photocatalytic reactors11,12. However, their predominantly passive nature severely limits their practical use. Consequently, the focus has shifted towards developing active solutions13,14,15 that manipulate the refractive index (RI) of materials through external stimuli like electrical tuning16,17, optical illumination18,19, or temperature20,21. Apart from metasurfaces designed for quick, high contrast modulation for switches17 and lenses22, a new class of tunable metasurfaces is emerging. These prioritize the precise adjustment of amplitude, shape, and spectral position of resonances. Yet, they often fall short due to the challenges associated with the materials used. Materials with minor RI changes often lack the desired tunability23, while those with substantial RI changes exhibit high intrinsic losses like germanium-antimony-telluride24,25, indium-antimony-telluride18,26, conductive polymers27,28, and vanadium-dioxide (VO2\mathrm{VO}_2)29,30. This limitation complicates the commonly used detuning-based switching method, which relies on shifting a resonance towards larger or smaller wavelengths. Spectrally sharp resonances cannot sustain high losses. due to their long resonance lifetime. While broad resonances are capable of handling high losses and optically or electrically switched metasurfaces show great promise31,32, their spectral width and high loss limits their applicability in many applications demanding high-transmission like optical filtering. Additionally, the need for spectral selectivity in multiplexing applications is not met by these broad resonances.

Recently, symmetry-protected bound states in the continuum (BICs)33,34 have gained attention in the metasurface community. Their appeal lies in their spectrally narrow resonances quantified by the quality (QQ) factor, which is the ratio of resonance frequency ω0\omega_0 to resonance bandwidth δω\delta\omega, and in the straightforward tunability of far-field coupling through simple geometric parameters. This led to demonstrations of numerous exciting physical phenomena with applications for lasers35, chiral resonators36, molecular sensors37, and lenses38. Applying the principle of BICs to tunable metasurfaces is, however, experimentally challenging. Popular tunable materials like GST, conductive polymers, or VO2_2 have not yet been achieved experimentally due to the susceptibility of high Q-factor resonances to intrinsic losses39. Therefore, active BICs have only been realized in the THz range40 or through low RI changing materials19.

Here, we propose a new approach for achieving actively fine-tunable metasurfaces based on loss-engineered hybrid VO2_2-Si resonators41,31,42 exhibiting BICs. Rather than fighting the high intrinsic losses in VO2_2's high-temperature phase30,29, we utilize it to generate a loss-based active tuning effect. Because the temperature induced phase change in VO2_2 starts locally in sub-wavelength areas43, the overall phase transition is gradual, with intermediate states readily accessible within a temperature window of roughly 30°C. When combined with the BIC's asymmetry-based tunability of radiative losses, this approach provides us with two tuning mechanisms: a passive one via the asymmetry of the structure and an active one via the temperature (Figure 1a). This dual control allows precise tailoring for various potential applications which prioritize high Q-factors, large resonance amplitudes, strong field enhancements, high modulation, or a combination of these key parameters. Our numerical and experimental demonstrations reveal that for a conventional, ellipse-pair based BIC metasurface geometry, a thin VO2_2 layer of about 30 nm is sufficient to quench the resonance. The remainder of the 750 nm high resonator can comprise any low-loss dielectric; in our case, it includes 600 nm of silicon below and 120 nm of SiO2\mathrm{SiO}_2 on top of the VO2\mathrm{VO}_2 (see a sketch of the structure in Figure 1b and the scanning electron microscope (SEM) image of the fabricated structure in Figure 1c). With κ\kappa in the low-temperature phase, we achieve Q-factors of up to 200 and a maximum reflectance amplitude of 90%. By heating the metasurface to temperatures above the phase transition, we achieve a relative switching amplitude of up to 78% between the “cold” and the “hot” phase (Figure 1d,Figure 1e). We experimentally mapped the parameter space defined by the asymmetry parameter α\alpha and temperature TT, which govern radiative and intrinsic losses, respectively. Using temporal coupled mode theory (TCMT) modelling, we identified the impact of our dual-control mechanism on both loss rates. Additionally, our results confirm the structure's versatility in transitioning between all coupling states, ranging from under-coupling through critical coupling to over-coupling.

Tuning principle and resonator geometry

Figure 1. Tuning principle and resonator geometry. (a) Schematics of the active VO2\mathrm{VO}_2-Si BIC metasurface. The resonator is coupled to incoming and outgoing radiation via the asymmetry-dependent radiative loss γrad\gamma_{\mathrm{rad}}, as depicted for the tilted ellipse-pair geometry where γrad∝sin⁡2(α)\gamma_{\mathrm{rad}} \propto\sin^2(\alpha) with the tilting angle

aa. Additionally, the mode is coupled to intrinsic losses which can be fine-tuned via the gradual phase transition of VO2\mathrm{VO}_{2} and its increasing imaginary part kk of the complex RI. This change in kk can gradually supress a resonant mode. (b) Sketch of the tilted ellipse-pair unit cell defined by the ellipses’ long axis AA, the short axis BB, the tilting angle α\alpha, the pitch in x-direction PxP_{x} and in y-direction PyP_{y}. The resonator, on top of a SiO2\mathrm{SiO}_{2} substrate, consists of a Si layer, a VO2\mathrm{VO}_{2} layer, and a capping SiO2\mathrm{SiO}_{2} layer. (c) SEM image of the fabricated metasurface with α=20∘\alpha= 20^\circ. Sketch of (d) numerical and (e) experimental reflectance spectra for the VO2\mathrm{VO}_{2}-Si metasurface in its hot and cold phases.

Results

We chose a tilted ellipse-pair geometry for our metasurface design due to its remarkable resonance robustness44. Furthermore, compared to other designs, this geometry offers a notably low spectral reflectance baseline of 2-3% in simulations. This feature, along with an exceptionally clear off-resonant spectrum, makes it particularly suitable for switching applications that value spectral cleanliness, high switching contrast, and multiplexing capability. While we focus on the tilted ellipse pair, it is worth noting that our methodology is not constrained to this specific BIC metasurface geometry and can be applied to other geometries.

To implement our loss-based switching concept, we need an active material that satisfies two main requirements. In its "off" state, the losses of the material should be minimal to ensure that resonant modes are not damped. Conversely, in its "on" state, the material should have a very high absorbance to achieve optimal resonance quenching. VO2\mathrm{VO}_{2}, especially in the mid-IR, stands out as an extraordinary choice, meeting both criteria29,30. Our atomic layer deposited VO2\mathrm{VO}_{2} (see Methods section) exhibits at 6.5 μ\mum an extinction coefficient kk of 0.011 at 25°C, and a kk of 2.42 at 80°C, as validated via ellipsometry (see Figure S1).

VO$_2$ ellipsometry plots

Figure S1. VO2_2 ellipsometry. Ellipsometry data between 4 and 8 μm for the used 30 nm VO2_2 film measured at 25°C for the cold phase and at 80°C for the hot phase. Blue solid (red dashed) lines represent the real (imaginary) part ε1(ε2)\varepsilon_1(\varepsilon_2) of the VO2_2 dielectric function.

Using simulations (see Methods section for details), we tuned the geometric parameters of our structures to resonate between 6-7 μ\mum. This leads to a 750 nm high resonator on top of a SiO2\mathrm{SiO}_{2} substrate, which is composed of a Si layer at the bottom, a VO2\mathrm{VO}_{2} layer in the middle, and a 120 nm SiO2\mathrm{SiO}_{2} capping layer which acts as a mask during the VO2\mathrm{VO}_{2}-Si etching.

While SiO2_2 has some absorption in the mid-IR range, it was chosen as substrate material since its thermal expansion coefficient aligns well with Si, ensuring a stable film during the VO2_2 annealing process. We sourced the RIs for VO2_2 through ellipsometry, shown in Figure S1, and took the values for Si and SiO2_2 from Shkondin et al.45 and Kischkat et al.46, respectively. To model the intermediate states of the VO2_2 layer, where it is partially in the cold and partially in the hot phase, we employed an effective medium approximation47 (Supplementary Note 1) with a hot-phase volume fraction VF=VhotVtotalVF = \frac{V_{\mathrm{hot}}}{V_{\mathrm{total}}}, defined as the ratio of volume in the hot phase VhotV_{\mathrm{hot}} divided by the total volume VtotalV_{\mathrm{total}}.

Resonance detuning-based switching typically requires a resonator predominantly composed of an active medium, because it changes the effective cavity size through refractive index shifts. Conversely, our loss-based approach only needs a minimal portion of the medium to be active. With this method, the effective cavity size hardly changes, instead, we merely introduce intrinsic losses into the system to attenuate the mode. By examining the relationship between the ellipses’ tilting angle α\alpha, the VO2_2 layer height, and the resulting reflectance modulation, we observed absolute reflectance modulation amplitudes max⁡(Rcold−Rhot)\max(R_{\mathrm{cold}} - R_{\mathrm{hot}}) of up to 80% (see Figure 2a). Note that this corresponds to a maximum relative modulation max⁡(Rcold−RhotRcold)\max\left(\frac{R_{\mathrm{cold}}-R_{\mathrm{hot}}}{R_{\mathrm{cold}}}\right) of 97% (Figure S2). Notably, as the asymmetry increases, the reflectance modulation peaks between α=15−30∘\alpha= 15 - 30^\circ, after which it declines. We ascribe this trend to two counteracting effects: the resonance lifetime and the absorption of VO2_2 in its cold state. At smaller asymmetries with high Q-factors QQ, the longer resonance lifetime (given by τ=Qω0\tau= \frac{Q}{\omega_0}) allows light to be in extended contact with the VO2_2 layer and thereby the absorption is enhanced. However, VO2_2 also exhibits absorption when in its cold state (k=0.011k = 0.011 at 6.5 µm), which results in a lower overall reflectance amplitude for small asymmetries in the cold phase. The same effects also lead to the observed behaviour for different VO2_2 layer heights: thin VO2_2 layers achieve optimal modulation with small asymmetries, and thicker layers favour larger asymmetries. Due to the dampening influence of a thicker VO2_2 layer on both the Q-factor and the reflectance amplitude in the cold phase, we proceed with a 30 nm VO2_2 layer for the remainder of the manuscript, which allows for both high Q-factors and strong resonance modulation. In Figure 2b, the absolute reflectance modulation across various asymmetries is depicted. It underlines the spectral cleanliness of the switching behaviour since no modes other than the BIC are present, and thanks to the thin VO2\mathrm{VO}_{2} layer, we can switch the metasurface with no noticeable spectral detuning and no off-resonance modulation (Figure 2b).

Numerical studies on the switching behaviour

Figure 2. Numerical studies on the switching behaviour. (a) Absolute reflectance modulation max⁡(Rcold−Rhot)\max(R_{\mathrm{cold}} - R_{\mathrm{hot}}) around 6-6.5 μ\mum as a function of the ellipse tilting angle α\alpha and the VO2\mathrm{VO}_{2} layer height hVO2h_{\mathrm{VO}_{2}}. A dashed gray line indicates hVO2=30h_{\mathrm{VO}_{2}} = 30 nm, presented in (b) where the spectral difference between cold and hot phase reflectance is visualized for various asymmetries, with a vertical offset of 0.33 between the spectra for clarity. (c) Spectra of different switching states (different VFVF) for hVO2=30h_{\mathrm{VO}_{2}} = 30 nm and α=20∘\alpha= 20^{\circ}. Spectra are vertically shifted by 0.5 for clarity. (d) Field enhancements in the VO2\mathrm{VO}_{2} layer in the cold (VF=0VF = 0), intermediate (VF=0.5VF = 0.5), and hot (VF=1VF = 1) phase for the same geometric parameters as in (c). (e) Maximal field enhancement for different VFVF. (f) Loss density mapping analog to (d). (g) Average loss density for different VFVF.

Numerical relative reflectance modulation

Figure S2. Numerical Relative Reflectance Modulation. VO2_2 layer height and asymmetry α\alpha sweep with the relative reflectance modulation max⁡(Rcold−RhotRcold)\max\left(\frac{R_{\mathrm{cold}}-R_{\mathrm{hot}}}{R_{\mathrm{cold}}}\right) plotted in color. The maximal value is 97%.

We expand our numerical analysis beyond the search for maximum switching in order to also study the intermediate switching states. Figure 2c captures the progressive degradation of the resonance with increasing VFVF for a fixed asymmetry α=20∘\alpha= 20^\circ. It reveals a gradual decrease in both QQ-factor and amplitude as VFVF increases, until the resonance almost vanishes. To study the gradual quenching of the BIC in more detail, Figure 2d presents simulated maps of the electric field enhancement (FE) within the VO2\mathrm{VO}_{2} layer for varying VFVF. For VF=0VF = 0 (cold phase), there are pronounced FE hotspots at the tips of the ellipses (FE=1019FE = 1019), which reveal the electric dipole in each of the ellipses that form the BIC. While VFVF increases, the field configuration remains consistent, but the FE eventually drops by almost an order of magnitude at VF=1VF = 1 (hot phase, FE=126FE = 126). Figure 2e depicts the maximal FE extracted from the simulations across all VFVF in increments of 0.1, demonstrating an almost linear decline. To offer further insight into the attenuation of the mode, we present simulated loss density maps (power absorbed per unit volume48) for the varying VFVF in Figure 2f. At VF=0VF = 0, losses manifest in regions with high FE (the tips of the ellipses, consistent with Figure 2d), a typical behaviour of resonant structures. In contrast, for VF=1VF = 1, the losses intensify and disperse, peaking in the centers of the ellipses, which indicates non-resonant absorption. The intermediate state with VF=0.5VF = 0.5 is a combination of cold and hot phase characteristics with losses both in the center and at the tips. As shown in Figure 2g, the average loss density within the VO2\mathrm{VO}_{2} layer increases most rapidly for VFVF between 0.2 and 0.5. A higher VFVF lead to a saturated loss level, until it declines for VF=0.9−1VF = 0.9 - 1, possibly due to the further decreasing FE.

Having verified through simulations that a thin VO2\mathrm{VO}_{2} layer within a BIC resonator can dampen the resonance efficiently and in a controlled manner, we next focus on experimental verification of this effect. We fabricated the BIC metasurface as detailed in the Methods section. All optical measurements were conducted using a Spero microscope (Daylight Solutions), which operates in the wavelength range starting from 5.6 μm\mu\mathrm{m}, as described in the Methods section. The upper wavelength limit was set to 7.5 μm\mu\mathrm{m} because, at longer wavelengths, the SiO2\mathrm{SiO_2} phonon band of the substrate and the capping layer dominates the optical response. Aiming to demonstrate the independent tuning of both intrinsic and radiative losses, we fabricated metasurfaces with varying asymmetries, ranging from α=0∘\alpha= 0^\circ to α=40∘\alpha= 40^\circ in the increments of 5°. Figure 3a presents their reflectance spectra measured at 30°C (RcoldR_{\mathrm{cold}}) and 100°C (RhotR_{\mathrm{hot}}), from which we extracted their absolute reflectance modulation characteristics (Rcold−RhotR_{\mathrm{cold}} - R_{\mathrm{hot}}) in Figure 3b. Notably, the maximum modulation varies with the asymmetry, reaching almost 50% at α=20∘\alpha= 20^\circ. Generally, the switching process does not compromise the clean low-temperature spectra and produces only minimal off-resonance wavelength modulation. The broadening of the mode at high temperatures is relatively small, while the Q-factor of the reflectance modulation is on par with the resonance in its cold phase. The off-resonance modulation observed around 5.8 μm\mu\mathrm{m} for α=40∘\alpha= 40^\circ arises from the BIC interacting with the onset of the first diffraction order at 5.7 μm\mu\mathrm{m}.

Experimental switching verification

Figure 3. Experimental switching verification. (a) Measured reflectance spectra at 30°C (blue) and 100°C (red) for increasing α\alpha. Insets in the top right corners show SEM images of the respective unit cell with the dimensions of 4400 x 2860 nm2^2. (b) Reflectance modulation (R30∘C−R100∘CR_{30^\circ\mathrm{C}} - R_{100^\circ\mathrm{C}}) extracted from the spectra in (a). (c) Temperature-dependent variation of the absolute reflectance modulation max⁡(R30∘C−R100∘C)\max(R_{30^\circ\mathrm{C}} - R_{100^\circ\mathrm{C}}) for the same asymmetries as in (a) and (b). (d) Full hysteresis modulation pattern for α=20∘\alpha= 20^\circ. (e) Absolute max⁡(R30∘C−R100∘C)\max(R_{30^\circ\mathrm{C}} - R_{100^\circ\mathrm{C}}) and relative reflectance modulation max⁡(R30∘C−R100∘CR30∘C)\max\left(\frac{R_{30^\circ\mathrm{C}}-R_{100^\circ\mathrm{C}}}{R_{30^\circ\mathrm{C}}}\right) across all asymmetries.

To assess the fine-tuning capabilities of our metasurfaces, we varied the temperature from 25°C to 110°C in 5°C increments. The absolute reflectance modulation amplitude max⁡(Rcold−Rhot)\max(R_{\mathrm{cold}} - R_{\mathrm{hot}}) across different asymmetries is plotted in Figure 3c. The phase transition occurs between 45°C and 75°C, with the most significant changes around 60°C. This offers a sizable 30°C temperature window to fine tune the optical response. In line with the spectra in Figure 3a,Figure 3b, the highest maximal modulation is achieved with α=20∘\alpha= 20^\circ, while 10° and 30° show similar results. An asymmetry of 40° yields much lower modulation, and 0° shows negligible change. Driving the metasurface through the whole modulation cycle by heating and cooling reveals a typical hysteresis, as expected for VO2\mathrm{VO_2}-modulated systems29. Our dual-tuning approach lets users prioritize either the absolute reflectance modulation discussed so far or the relative reflectance modulation max⁡(Rcold−RhotRcold)\max\left(\frac{R_{\mathrm{cold}}-R_{\mathrm{hot}}}{R_{\mathrm{cold}}}\right). Based on this preference, the optimal structural asymmetry can be chosen, with the peaks for relative and absolute reflectance modulation in our presented structure occurring at α=10∘\alpha= 10^\circ and α=20∘\alpha= 20^\circ, respectively, see Figure 3e.

Having demonstrated the experimental switching and fine-tuning capabilities of our metasurfaces, we now take a closer look into the effects the switching has on the resonance characteristics. Our investigation involves spectra from eight distinct asymmetries (α=5−40∘\alpha= 5 - 40^\circ in 5∘5^\circ increments) and a temperature series comprising 18 steps (T=25−110∘CT = 25 - 110^\circ\mathrm{C} in 5∘C5^\circ\mathrm{C} increments). This results in a 2D parameter matrix of 144 unique resonance spectra. To model these resonances, we used temporal coupled mode theory (TCMT), an established framework for interpreting complex light-metasurface interactions49,50. Central to TCMT are intrinsic γint\gamma_{\mathrm{int}} and radiative γrad\gamma_{\mathrm{rad}} loss rates, signifying energy loss within the resonator and energy loss to external channels, respectively. Controlling their ratio is crucial since they dictate the Q-factor, control the FE, and determine the resonance behaviour. In our modelling, we conceptualize our system as a singular resonator coupled to two ports, one symbolizing incoming/reflected light and the other indicating transmitted light, as detailed in Supplementary Note 2. Next, we fit the TCMT model to our experimental data by using unconstrained parameters. Only the intrinsic loss γint\gamma_{\mathrm{int}} is constrained for all α\alpha at a given temperature since it predominantly depends on the intrinsic loss of VO2_2. In Figure 4a, we replot the experimental reflectance data from Figure 3a (T=25T = 25°C, varying α\alpha, solid gray lines) and overlay them with the results of the TCMT fit (dashed black lines).

Insight into experimental data through TCMT analysis

Figure 4. Insight into experimental data through TCMT analysis: (a) Measured spectra (gray solid line and shading) and TCMT-derived fits (dashed black line) for metasurfaces at 25°C with α\alpha values from 5° to 40°. (b) Radiative loss γrad\gamma_{\mathrm{rad}} and (c) intrinsic loss γint\gamma_{\mathrm{int}} extracted from TCMT fits of experimental data across various asymmetries α\alpha and temperatures. While γrad\gamma_{\mathrm{rad}} exhibits an increase with α\alpha and is constant over temperature, γint\gamma_{\mathrm{int}} behaves oppositely: it increases with temperature and is unaffected by α\alpha. (d) Overlay of data from (b) and (c) displaying the difference between the two loss rates γrad−γint\gamma_{\mathrm{rad}} - \gamma_{\mathrm{int}}. Brown shading marks undercoupled resonator regions (γrad<γint\gamma_{\mathrm{rad} } < \gamma_{\mathrm{int}}), green denotes overcoupled regions (γrad>γint\gamma_{\mathrm{rad}} > \gamma_{\mathrm{int}}), and white represents critically coupled resonators, indicated by the s-shaped dashed black curve. (e) γrad\gamma_{\mathrm{rad}} and γint\gamma_{\mathrm{int}} plotted as functions of the angle α\alpha at a temperature of 60°C, based on the data slice highlighted by the dashed horizontal line in (d). (f) γrad\gamma_{\mathrm{rad}} and γint\gamma_{\mathrm{int}} plotted against temperature for α=10\alpha= 10°, corresponding to the data slice marked by the dashed vertical line in (d).

We extracted the loss rates γrad\gamma_{\mathrm{rad}} (Figure 4b) and γint\gamma_{\mathrm{int}} (Figure 4c) as well as the corresponding total Q-factor (Figure S3) from all spectra and mapped them across the entire α\alpha-TT-parameter space. As γrad\gamma_{\mathrm{rad}} represents the energy radiated out of the resonator, its expected rise with increasing asymmetry is evident, with no dependence on temperature. Conversely, γint\gamma_{\mathrm{int}}, which represents scattering and material losses, exhibits a substantial change with temperature around the phase transition of VO2_2. Analyzing these loss rates individually underlines a central observation: the two ‘tuning knobs’ (asymmetry & temperature) modulate the two loss channels (γrad\gamma_{\mathrm{rad}} & γint\gamma_{\mathrm{int}}) independently, allowing selective tailoring of the resonance characteristics. Of special importance is the regime of critical coupling, where γrad\gamma_{\mathrm{rad}} equals γint\gamma_{\mathrm{int}}, marking a sweet spot for a plethora of photonic devices11,51,52. It allows optimal energy transfer, heightens sensitivity, and maximizes light absorption. Figure 4d illustrates the interplay of the two loss rates within our parameter space by plotting their difference γrad−γint\gamma_{\mathrm{rad}} - \gamma_{\mathrm{int}}. The brown zone corresponds to undercoupling (γrad<γint\gamma_{\mathrm{rad}} < \gamma_{\mathrm{int}}), mainly observed in regions with high γint\gamma_{\mathrm{int}} (elevated temperatures) and low γrad\gamma_{\mathrm{rad}} (minimal asymmetry). Overcoupling (γrad>γint\gamma_{\mathrm{rad}} > \gamma_{\mathrm{int}}), on the other hand, is prevalent at lower temperatures and high asymmetries. The s-shaped dashed black curve marks the domain where resonators are critically coupled.

Q-factor mapping

Figure S3. Q-factor mapping. Total fitted Q-factor according to our TCMT model, calculated from the two loss rates γrad\gamma_{\mathrm{rad}} and γint\gamma_{\mathrm{int}} by Q=ω0(γrad+γint)−1Q = \omega_0(\gamma_{\mathrm{rad}} + \gamma_{\mathrm{int}})^{-1}.

Our method of integrating a γrad\gamma_{\mathrm{rad}}-adjustable BIC with the dynamically tunable γint\gamma_{\mathrm{int}} of VO2_2 allows our metasurfaces to modulate between under, over, and critical coupling, dependent on temperature (shown in Figure 4e for T=60T = 60°C) or asymmetry (depicted in Figure 4f for α=10\alpha= 10°). A single metasurface can thus be effortlessly tuned through all coupling regimes, with the steepest changes within the phase transition temperature range of 45°C to 75°C. The path of critical coupling within our 2D parameter space can be further adjusted by simply adjusting the height of the VO2_2 layer (see Figure 2a).

Discussion

Our study introduces an innovative method for active control of metasurfaces by harnessing loss-based tuning within a phase-change material in a hybrid BIC metasurface. We utilized the passive Q-factor control of BICs and combined it with the active temperature-sensitive tunability of losses within VO2\mathrm{VO}_2. This approach enabled us to quench the resonance-based reflectance peak of our metasurfaces, achieving a numerical absolute reflectance modulation of 80% and a relative modulation of 97%. In experiments, these values were realized as 47% and 78%, respectively.

A standout benefit of our approach is that it eliminates the need for thick layers of active materials. Instead, thin layers within the resonator, amounting only to 1/25 of the resonator's volume, are sufficient to suppress the mode. This suppression proves particularly effective at high Q-factors, resulting in a pristine reflectance spectrum free from off-resonant modulations and spectral clutter. Our method provides separate and individual control over the system's key loss channels: the radiative loss, passively controlled by the asymmetry factor of our geometry, and the intrinsic loss, tunable both actively via temperature and passively through the layer thickness of VO2\mathrm{VO}_2. We further experimentally verified the fine-tuning capability of our metasurfaces, largely attributed to the broad hysteresis of VO2\mathrm{VO}_2. This property enables robust intermediate switching states, leading to an adjustable modulation of the reflectance. To gain a comprehensive understanding of our system, we mapped the parameter space spanned by the asymmetry parameter α\alpha and temperature – the pivotal factors controlling radiative and intrinsic losses. TCMT modelling within this space underline our robust dual control over the loss. Moreover, our findings validate the ability of our structures to transition across all coupling states, from under- to critical-, and then to over-coupling.

This research, while providing substantial insights into loss-based metasurface tuning, also traces a path for applications where precision and controllability of resonances is vital. As the nanophotonics community dives deeper into active metasurfaces, the results presented in this study will likely serve as foundation for more application-based developments. Our dual-control mechanism allows new approaches for advanced sensing of biological samples or gases. Further possible directions include optical switches, fine-tunable filters, thermal radiators, and programmable metasurfaces.

Materials and Methods

Simulations for our study were performed using the commercial finite element solver CST Studio Suite (Simulia). The software was configured with adaptive mesh refinement, periodic boundary conditions, and operated in the frequency domain.

For sample fabrication, first, 600 nm of Si were deposited onto a fused silica substrate by an ion beam sputtering system with Kaufmann ion sources (Kaufman & Robinson, Inc.) at room temperature by using argon ions at 600 eV. Then, 30 nm of VO2\mathrm{VO}_{2} were deposited by 250 cycles atomic layer deposition (Cambridge NanoTech Fiji 200) running at 150°C. Tetrakis(dimethylamino)vanadium (TDMAV) pre-heated to 87°C and water were used as precursors, that were let into the chamber within 0.6 s pulses with the consecutive waiting times of 8 s and 3 s, respectively. Immediately after the deposition, the sample was annealed in a vacuum tube furnace at 500°C for 5 minutes under 15 sccm flow of oxygen. Finally, 120 nm of SiO2\mathrm{SiO}_{2} for the etch mask were deposited onto the sample from a stoichiometric powder in an electron beam evaporator (Bestec, 8 kV, 1 Å/s) at room temperature.

Nanostructuring of the Si – VO2\mathrm{VO}_{2} – SiO2\mathrm{SiO}_{2} multilayer film starts with a deposition of a 400 nm layer of positive electron beam resist, ZEP520A (Zeon Corporation), followed by a coating of a conductive polymer, using ESPACER (Showa Denko K.K). Electron beam lithography was carried out at 20 kV using an eLINE Plus system (Raith). Development was achieved with a subsequent bath in amyl acetate and MIBK:IPA in a 1:9 ratio. The polymer mask was used to dry-etch the SiO2\mathrm{SiO}_{2} layer and then removed by the Microposit Remover 1165 (Microresist). A second, selective dry-etching process based on SF6\mathrm{SF}_{6} and Argon was applied to etch the VO2\mathrm{VO}_{2} and Si layers. The ellipse's long axis AA measures 2460 nm, the short axis BB is 980 nm, the x-direction pitch PxP_{x} is 4400 nm, and in the y-direction pitch PyP_{y} is 2860 nm.

Optical measurements were conducted using a spectral imaging MIR microscope, Spero (Daylight Solutions). The microscope featured a 4×4 \times magnification objective (NA=0.15NA = 0.15) and provided a 2 mm22\ \mathrm{mm}^2 field of view. This system features three tunable quantum cascade lasers that consistently cover the 5.65.6-10.5 μm10.5\ \mu\mathrm{m} wavelength range, offering a spectral resolution of 2 cm−12\ \mathrm{cm}^{-1}. The lasers emit linearly polarized light. Since the SiO2\mathrm{SiO}_2 substrate restricts transmission, all measurements were carried out in reflection mode. To regulate the sample's temperature during experiments, a home-made sample holder was used. This holder consists of a thermally insulated copper contact surface, a temperature sensor, and four heating resistors. Temperature adjustments were controlled by a feedback loop to ensure stability.

ACKNOWLEDGMENTS

This work was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under grant numbers EXC 2089/1 – 390776260 (Germany’s Excellence Strategy) and TI 1063/1 (Emmy Noether Program), the Bavarian program Solar Energies Go Hybrid (SolTech), and the Center for NanoScience (CeNS), Faculty of Physics/LMU Munich. BUT team was supported by the Grant Agency of the Czech Republic (21-29468S) and the sample fabrication was done in CzechNanoLab Research Infrastructure supported by MEYS CR (LM2023051). S.A.M. acknowledges the Lee-Lucas Chair in Physics and the EPSRC (EP/W017075/1).

CONFLICT OF INTEREST

The authors declare no competing financial interest.

supplementary_material

SUPPLEMENTARY NOTES

Effective Medium Approximation

To model the inhomogeneous nature of the intermediate state of VO2\mathrm{VO}_{2} in our metasurfaces, we utilize the effective medium approximation which is widely adopted in literature[1, 2, 3]. It calculates the average permittivity of the composite material consisting of domains of VO2\mathrm{VO}_{2} in its hot and cold phases and treats it as a homogeneous medium. This approximation is justified due to the highly subwavelength size of the domains in different phases within the polycrystalline film. The effective permittivity εeff\varepsilon_{\mathrm{eff}} of the VO2\mathrm{VO}_{2} layer can be expressed as a weighted sum of the permittivities of its individual constituents, given by

εeff=VF⋅εVO2,hot+(1−VF)⋅εVO2,cold\varepsilon_{\mathrm{eff}} = VF \cdot\varepsilon_{\mathrm{VO}_{2},\mathrm{hot}} + (1 - VF) \cdot\varepsilon_{\mathrm{VO}_{2},\mathrm{cold}}

As the VO2\mathrm{VO}_{2} undergoes the phase transition, the volume fraction VFVF, defined as VF=VhotVtotalVF = \frac{V_{\mathrm{hot}}}{V_{\mathrm{total}}} varies, leading to changes in the effective permittivity εeff\varepsilon_{\mathrm{eff}}, see Figure S4. By integrating this effective medium approximation-based permittivity into our models, we could simulate the metasurfaces behaviour in intermediate states more accurately.

Effective permittivity

Figure S4. Effective Permittivity. Effective permittivity of VO2\mathrm{VO}_2 for different volume fractions VFVF, using kk and nn values at a wavelength of 6.5 μm6.5\,\mu\mathrm{m}.

TCMT Model

TCMT provides a straightforward method of modelling resonant structures. Fan et al. presents a general description[4], which we adapt for our single resonance system, coupled to two ports corresponding to reflectance and transmittance. The temporal dynamics of the resonance amplitude aa can be written as

dadt=(−γrad+γint2+iω0)a+γradsin(1)\frac{da}{dt}=\left(-\frac{\gamma_{\mathrm{rad}}+\gamma_{\mathrm{int}}}{2}+i\omega_0\right)a+\sqrt{\gamma_{\mathrm{rad}}}s_{\mathrm{in}} \tag*{(1)}

with ω0\omega_0 as the resonance frequency of the BIC. sins_{\mathrm{in}} represents the time dependent input field. Next, we define the relationship between the scattered field souts_{\mathrm{out}} with the resonance mode and the incoming field by

sout=[sreflectedstransmitted]=S[asin]s_{\mathrm{out}}=\begin{bmatrix}s_{\mathrm{reflected}}\\s_{\mathrm{transmitted}}\end{bmatrix}=S\begin{bmatrix}a\\s_{\mathrm{in}}\end{bmatrix}

where SS represents the scattering matrix given by

S=[−iγrad00iγrad].S=\begin{bmatrix}-i\sqrt{\gamma_{\mathrm{rad}}}&0\\0&i\sqrt{\gamma_{\mathrm{rad}}}\end{bmatrix}.

By using equation (1) under steady state conditions, we are now able to calculate the reflectance RR and the transmittance TT via their relation to sreflecteds_{\mathrm{reflected}} and stransmitteds_{\mathrm{transmitted}}:

R=∣sreflected∣2=γrad∣a∣2=γrad∣κiω+γrad+γint2∣2∣sin∣2R=\lvert s_{\mathrm{reflected}}\rvert^2=\gamma_{\mathrm{rad}}\lvert a\rvert^2=\gamma_{\mathrm{rad}}\left\lvert\frac{\kappa}{i\omega+\frac{\gamma_{\mathrm{rad}}+\gamma_{\mathrm{int}}}{2}}\right\rvert^2\lvert s_{\mathrm{in}}\rvert^2
T=∣stransmitted∣2=γrad∣sin∣2T=\lvert s_{\mathrm{transmitted}}\rvert^2=\gamma_{\mathrm{rad}}\lvert s_{\mathrm{in}}\rvert^2

κ\kappa denotes the coupling coefficient between the resonant mode and the external ports. We used this theoretical framework to fit our experimental reflectance spectra to extract parameters like γrad\gamma_{\mathrm{rad}} and γint\gamma_{\mathrm{int}}. For fitting purposes we normalized sins_{\mathrm{in}} to 1.

Supplementary Figures

References

  1. [1]Voloshenko, I. et al. Microscopic nature of the asymmetric hysteresis in the insulator-metal transition of VO2 revealed by spectroscopic ellipsometry. Appl. Phys. Lett. 113, (2018).DOI
  2. [2]Liu, B., Gong, M., Zhang, J., Tian, S. & Zhao, X. Comparative study of the metal insulator transition of a VO2film with simultaneous infrared thermography and electric measurements. AIP Adv. 11, (2021).DOI
  3. [3]Leahu, G., Li Voti, R., Sibilia, C. & Bertolotti, M. Anomalous optical switching and thermal hysteresis during semiconductor-metal phase transition of VO2 films on Si substrate. Appl. Phys. Lett. 103, (2013).DOI
  4. [4]Fan, S., Suh, W. & Joannopoulos, J. D. Temporal coupled-mode theory for the Fano resonance in optical resonators. J. Opt. Soc. Am. A 20, 569 (2003).pubmed.ncbi.nlm.nih.gov/12630843

Paper details

vlnc:W_z3j5udy7
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents