Reconfigurable Flat Optics with Programmable Reflection Amplitude Using Lithography‐Free Phase‐Change Material Ultra‐Thin Films

A very large dynamic optical reflection modulation from a simple unpatterned layered stack of phase‐change material ultra‐thin films is experimentally demonstrated. Specifically, this work demonstrates that properly designed systems comprising deeply subwavelength GeSbTe (GST) films, a dielectric spacer, and a metallic mirror produce a dynamic modulation of light in the near‐infrared from very strong reflection (up to R≈80% ) to perfect absorption ( A>99.995% ) by simply controlling the crystalline state of the phase‐change material. While the amplitude of modulation experimentally reaches an optical contrast higher than 104, intermediate levels of reflection in between extreme values can also be actively encoded, corresponding to partial crystallization of the GST layer. Several layered system designs are further explored and guidelines are provided to tailor the efficient wavelength range, the angle of operation, and the degree of crystallization leading to perfect absorption.


Introduction
Adjusting the absorption, reflection, and transmission properties of systems is the basis of most photonic devices engineering, from mirrors to dispersion gratings as well as photodetectors and solar cells. The recent progress in metamaterials and metasurfaces has provided new methods to precisely control these features to an unprecedented degree. For example, through careful spatial arrangement of dielectric meta-atoms with multipolar resonances, we can design metasurfaces tailored for specific optical functionalities. This concept has been exploited to demon strate flat optics such as lenses, polarizers, retroreflectors, holograms, perfect absorbers [1][2][3][4] that hold promise to surpass the performances of conventional diffractive optics components.
However, both the nanoscale of meta-atoms and the standard materials used for metasurfaces can represent a roadblock for modulation and reconfiguration purposes. Indeed, nanopatterning thin-film materials through etching processes definitively set their geo metries and limit their functionalities to a designated purpose. To take a simple example, a TiO 2 -based metasurface hologram once fabricated will only display one holographic image, what severely restricts the potential of this technology. In that context, recent works demonstrated the potential of phase-change materials (PCM) for tunable nanophotonics. [5][6][7] Indeed, this class of materials enables a very large optical modulation at the nanoscale via a fast change of phase in their crystalline structure. This large optical modulation of PCMs has been used in specifically designed nanostructures to enable active beam-steerers, [8,9] bi-directional all-optical switching, [10] dynamic modulation of light emission, [11] light absorption [12,13] or light transmission. [14] Most of these studies leverage optical frequency meta-atoms which require in-plane nanostructuration of materials that make their large-scale fabrication and industrial development difficult. Alternatively, other less technologically constraining methods exist to engineer the optical properties of devices using lithography-free planar thin-films. The most well-known example is the anti-reflection coating, for which the thickness of a transparent thin-film is set at λ/4n to minimize the reflection through A very large dynamic optical reflection modulation from a simple unpatterned layered stack of phase-change material ultra-thin films is experimentally demonstrated. Specifically, this work demonstrates that properly designed systems comprising deeply subwavelength GeSbTe (GST) films, a dielectric spacer, and a metallic mirror produce a dynamic modulation of light in the near-infrared from very strong reflection (up to 80% R ≈ ≈ ) to perfect absorption (A > > 99.995%) by simply controlling the crystalline state of the phase-change material. While the amplitude of modulation experimentally reaches an optical contrast higher than 10 4 , intermediate levels of reflection in between extreme values can also be actively encoded, corresponding to partial crystallization of the GST layer. Several layered system designs are further explored and guidelines are provided to tailor the efficient wavelength range, the angle of operation, and the degree of crystallization leading to perfect absorption.
destructive interferences while simultaneously maximizing transmission. By introducing PCMs in similar engineered thin-films, recent works demonstrated tunable structural coloration, [15,16] optically reconfigurable metasurfaces, [17] or tunable near-perfect absorption. [18][19][20][21][22][23][24]25] So far, these experimental works demonstrated actively switchable optics that are either binary and/or volatile. However, the complex features of chalcogenide PCMs, and most notably the possibility to actively set them into a state of controlled partial crystallization, may be used to a much deeper extent. Indeed, their complex refractive index can be encoded into arbitrary intermediate values between those of amorphous and those of fully crystalline. This multilevel crystallization produces stable, non-volatile states that can be driven optically or electrically via pulsed inputs. [26,27] Such reconfigurable multilevel optical properties may be exploited as additional degrees of freedom for the design space of multifunctional flat optics with a large number of intermediate states and could be the basis for numerous exciting opportunities in applications such as actively controlled reflectivity modulation, continuous optical power limiting, tunable displays, active spectral filtering, and dynamic wavefront shaping.
In this work, we design and demonstrate actively reconfigurable lithography-free flat optics whose optical properties can be continuously tuned from a strong reflection (R ≈80%) to a perfect absorption (A = (1-R)≈99.995%), that is, a contrast larger than 10 000 that is actively controlled by simply adjusting the crystalline fraction of a standard GST thin-film (see Figure 1a for an illustration of the concept). Such a modulation depth surpass most of the free-space optical modulators reported so far, [14,[28][29][30] with the additional advantage of not requiring any complex nanopatterning processes. By precisely exploiting the non-volatile multilevel states enabled by partial crystallization of GST, we provide comprehensive design rules to simultaneously tailor the wavelength range of efficiency, the angle of operation, and the degree of crystallization leading to perfect absorption. Figure 1. a) Principle of the reflectivity modulation when the phase of GST is varied from amorphous to crystalline. The layered system comprises a top GST layer, a thin dielectric layer as a spacer (in blue), and an Au layer (in yellow). As displayed in the sketches, the amplitude of reflection is controlled by the state of GST, leading to a dynamic modulation between strong reflection (left panel) and perfect absorption (center panel), as well as any intermediate values of reflection, controlled by the crystallization fraction of GST (right panel). b-e) Experimental dispersion curves of GST as a function of temperature: b) refractive index n and c) absorption coefficient k; d) evolution of refractive index and e) absorption coefficient versus temperature at 1550 nm and 1700 nm wavelength.

www.advopticalmat.de
We further propose practical implementations of this concept in multilayered configurations designed for electrical modulation that conserves all achieved properties.

Optical Properties of GST
Phase change materials such as GST present unique chemical bonding properties (sometimes referred to as resonant bonding or "metavalent bonding") with strong electronic polarizabilities that produce a very large refractive index in a broad wavelength range, typically around n ≈ 6-7 in the infrared region. [31][32][33][34] This metavalent bonding requires a long range order between atoms and is therefore lost when the material is in an amorphous state. In that latter state, the refractive index falls back to values typical of a semiconductor, that is, n ≈ 4. Changing the phase of GST from amorphous to crystalline therefore produces an exceptionally large modulation of the refractive index. Furthermore, such a structural reorganization can be driven thermally, electrically or optically. Figure 1b,c illustrates the drastic change of refractive index n and absorption coefficient k of a GST layer upon crystallization. The dispersion curves n(λ) + ik(λ) have been measured by ellipsometry (more information in the Experimental Section) as a function of temperature. Between 120 and 160 °C, the refractive index n values almost double in the 1200-2000 nm wavelength range upon crystallization, and a transition temperature is found around 140 °C, as shown in Figure 1d. Simultaneously, the absorption coefficient k also undergoes a drastic change (Figure 1e). The co-evolution of the real part n and the imaginary part k of the complex refractive index versus temperature follows a quasi-linear variation at all wavelengths. To simplify the numerical calculations, the evolution of n and k as a function of the wavelength in the phase transition range can reasonably be approximated using a linear regression. Using this linear variation n(k), we introduce the quantity τ, that represents the crystallization fraction where 0 and 100% correspond to the amorphous and crystalline state, respectively.

Modulation Principle
When a dielectric layer is deposited on a reflective surface, for example, a thick gold layer, most of the incident light is reflected back. However, depending on the thickness of this dielectric layer, the reflected light fields from the first and second interfaces may either be in phase or in opposite phase. This layer can therefore be used as a means to adjust the reflectivity of the system via constructive or destructive interferences. The reflection properties of a generic system comprising a semi-infinite reflective substrate and two thin layers on top, can be expressed as follows: r 01 is the Fresnel coefficient at the first interface and r 123 the effective reflection coefficient of the combined second and third interface. β 1 = π λ 2 n  1 d 1 and β 2 = π λ 2 n  2 d 2 , with n  1 , d 1 and n  2 , d 2 the complex refractive indices and thicknesses of layers 1 and 2, respectively.
From this equation, we can see that there are different ways of tailoring the reflection of such a multilayer stack. The simplest one is to adjust the thickness of the thin-film until obtaining a desired reflection value at a given wavelength. A similar effect can be obtained by choosing a material with appropriate complex refractive index so as to adjust the optical path (n.d). Interestingly, such system can be engineered to reach a perfect optical absorption via a mechanism called critical coupling. In Equation (1), this regime can be reached when the numerator equals zero and gives us the following conditions: where k 1 is the extinction coefficient of layer 1, λ the wavelength, and Φ 01 and Φ 123 are the phases of r 01 and r 123 , respectively. Simply put, Equations (4) and (5) set conditions in terms of amplitude and phase to reach the perfect absorption. Rearranging Equation (4), we find a more intuitive relation: where α 1 is the absorption in layer 1. Equation (6) implies that, if the phase conditions of Equation (5) are fulfilled, only two parameters suffice to attain the perfect absorption: the thickness and the absorption of the top dielectric layer. More intuitively, this critical coupling condition corresponds to a regime where the overall radiative losses and absorption losses are equals. This principle has been used to obtain perfect absorption using ultra-thin layers of lossy materials. [35] By introducing a transparent layer between the reflective surface and the top lossy layer we add a supplementary degree of freedom to adjust the reflected phase and reach destructive interferences. This approach has been used by Long et al. [36] . In their work, by tuning R 123 and Φ 123 , they provide guidelines to reach the critical coupling conditions in a tri-layer system using various combinations of materials. Interestingly, in PCMs one can actively modify the complex refractive index and may therefore dynamically modulate the optical reflection of a stack of thin-films. Using PCM as a top layer, we not only considerably expand the design space to reach the perfect absorption regime, but we can also optimize such a system to maximize the amplitude of reflectivity modulation. Due to the number of free parameters, in our work we do not use directly Equation (1) but numerical computation of r. The theoretical optical simulations are done with a home-made software based on the transfer matrix method [37] and applied in the following work for normal incidence θ = 0°.

www.advopticalmat.de
To illustrate this concept, in the following, we give an example of a simple planar device that maximizes the reflection modulation. We define a layered system composed, from bottom to top, of Au/TiO 2 /GST (respectively, 200, 67.5, and 15 nm). Each thicknesses were calculated to maximize the reflectivity contrast between the amorphous and crystalline states of GST at a given wavelength (here 1550 nm). The calculated amplitude modulation as a function of crystallization is shown in Figure 2. In the amorphous state (crystallization fraction τ = 0%), we start with a very strong reflectivity R = 83%, and this value is progressively reduced during the crystallization. At the fully crystallized state, we obtain a perfect absorption regime (A ≈ 99.9999%), since the reflectivity reaches a value of 0.000016%. In order to give a figure of merit for the amplitude of the reflection modulation, here we define the contrast ratio C between R max and R min , the maximum and minimum of reflectivity, respectively, at a given wavelength during the phase change of GST: In this case, C is equal to 5.18 × 10 6 . There could be different ways of quantifying the amplitude of modulation, but this simple figure of merit provides a proper and straightforward comparison between our different designs. The contrast C is also related to another figure of merit usually used: the extinction ratio ER defined as −10 × Log(C), which would reach a value of -68 dB here. This system is very simple but yet very powerful given the large amplitude of modulation we can get by simply changing the phase of GST. Furthermore, properly controlling the crystallization fraction should enable an active access to a desired level of reflection in between the extrema. In other words, we can dynamically prepare this sample to have any arbitrary level of reflectivity between 83% and 0%, and this value will be kept over time owing to the nonvolatile properties of phase-change materials.
In this example and throughout this work, our method to optimize the system is as follows: For Au/Spacer/GST, we compute the reflectvity at a given angle of incidence and wavelength while varying the thicknesses of the two layers above Au. For each couple of thicknesses of Spacer and GST, we compute the reflectivity modulation by varying the crystallization rate from 0 to 100 % and we extract the two extrema of reflectivty, the contrast ratio C, and also the crystallization fraction at which occurs the perfect absorption. By this method, the optimal GST/spacer layer thicknesses can be selected providing best absorption and maximal reflectivity, that is, highest amplitude of modulation.
In the following, we show how we can fully exploit the design space offered by this platform to tailor many features of the modulation, such as the wavelength of operation and the crystallization fraction at which the perfect absorption occurs.

Tailoring the System for a Desired Wavelength of Operation
As explained in the previous section, for a given 3-layer system (Au/dielectric spacer/GST), the amplitude of reflectivity modulation at a given wavelength can be maximized by choosing an optimized couple of thicknesses for the Spacer/GST system. As shown in Figure 3b, the progressive change of phase of GST (the crystallization fraction is symbolized by different colors in the graph) produces an ≈ -50 dB extinction ratio at λ = 1500 nm. By simply adjusting the GST thickness, we can tailor the wavelength at which this optimized contrast occur. This is illustrated in Figure 3c, where a large contrast C is obtained at different wavelengths for three Au/ITO/GST layered systems. In the following, we define an "8-bit bandwidth" corresponding to the range of wavelength where the contrast C is higher than 256 (i.e., the minimum amplitude modulation needed to encode data via an 8-bit grayscale level). In Figure 3c, we show that the respective 8-bit bandwidths for three GST thicknesses 17, 20, and 23 nm span a wavelength range larger than 150 nm.
Note that the range of validity of this study is limited below 1300 nm due to the increasing absorption of GST but remains valid beyond 1700 nm, throughout the mid-infrared domain. This implies that this system can be designed to operate at arbitrary wavelengths in the IR range, with a contrast ratio higher than 256.
Importantly, one may wonder how it is possible to shift the wavelength of operation by 150 nm by simply adding 3 nm of GST to the system. The reason behind this effect is that the crystallization fractions τ at which occur the minimum and maximum reflectivity are different for each of the three cases considered in Figure 3c. In the following section, we describe how the system can be designed to operate at specific crystallization fractions.

Reaching the Perfect Absorption at Designed Crystallization Fractions of GST
We have seen that it is possible to set PCM layers into states of partial crystallization, hence enabling multilevel intermediate values of complex refractive index. However, most of the reported PCM-based photonic devices are designed to operate as binary devices that simply exploit the fully amorphous and crystalline states. We provided an example of such functionality in Figure 2 where the device presents a maximal reflectivity when GST is amorphous and a perfect absorption when GST is crystalline, but this was just an illustrative example to highlight the full amplitude modulation of reflectivity one can obtain. With this system, we can indeed go well beyond the sole binary modulation, exploit all intermediate states, and select at which crystallization fraction will occur the perfect absorption. For each layered system shown in Figure 3c, the perfect absorption is reached at a given value of crystallization fraction τ pa . Using our generic method, we can tailor τ pa for a given wavelength (here 1550 nm) by adjusting the respective thicknesses of the couple spacer layer/GST layer. This principle is illustrated in Figure 3d-f in which the crystallization fraction for perfect absorption is designed to occur at τ pa = 98%, 75%, 37%, 15%, and 0% for ITO/GST thicknesses of (85 nm/14.5 nm), (65 nm/21 nm), (0 nm/63,5 nm), (250 nm/179 nm), and (380 nm/157 nm), respectively. It is worth noting that we are able to switch the amplitude from maximal to minimal reflectivity by either starting from the amorphous or the crystalline phase (green and blue curves). The layered system without spacer, that is, with GST directly on Au layer is also a solution since it modulates light with high contrast ratio (yellow curve in Figure 3d-f). We should notice also that for every case the maximum of reflectivity is always higher than 50% and can reach nearly 80% for optimized ITO spacer. Moreover the calculated perfect absorption is always better than 99.98 % (see Figure 3d).
Note that the reflectivity curves R(τ) shown in Figure 3d,e also present an abrupt change of optical phase at the minimum of reflectivity, as displayed in Figure 3f and in agreement with recent works on tunable perfect absorbers. [23,38]  for different thicknesses of GST (17, 23, and 23 nm for the blue, red, and yellow curves, respectively). The black dashed line is the limit for a contrast of 256 (8-bit grayscale level). d-f) Tailoring the crystallization fraction for the perfect absorption by adjusting the respective thicknesses of the Au/ITO/GST system. The reflectivity at λ = 1550 nm, versus the crystallization fraction is displayed in linear scale in (d) and in logarithmic scale in (e). The phase of the reflectivity is shown in (f).

Without Spacer: The Au/GST System
To experimentally demonstrate these concepts, we have fabricated Au/GST systems with GST thicknesses of 53 and 58 nm, respectively (more details in the Experimental Section). The modulation of reflectivity is observed for both samples upon thermally driven crystallization (the reflectivity in dB, i.e, 10Log(R), is represented in Figure 4a,b). With the dispersion curves shown in Figure 1, theoretical reflectivity spectra R theo (λ) for each crystallization fraction τ have been computed and fitted to the experimental R exp (λ) spectra for each temperature T. The fitted thicknesses for GST are found to be 54.75 and 59 nm, in line with the targeted 53 and 58 nm deposited thicknesses. These fittings (videos shown in Supporting Information) allow to retrieve the relations between the tem-perature T and the crystallization fraction τ for each sample (Supporting Information). Consequently, the theoretical reflectivity R theo (λ, τ) can be converted and represented with temperature scale as R theo (λ, T). These retrieved theoretical curves versus the wavelength and temperature are shown in Figure 4g,h and are in full agreement with the experimental ones (in Figure 4d,e). A deep absorption is observed around 110 and 125 °C before the sharp phase transition for samples 1 and 2 with 53 and 58 nm GST thickness, respectively. As shown in Figure 4c, the reflectivity modulation for the two GST layer systems is maximal at 1342 and 1457 nm, respectively, and reach a minimum of reflectivity of 0.026%. This translates into an experimental absorption of A ≈ 99.97%. As expected from our calculations, the thinner GST layer (53 nm) requires a smaller crystallization fraction to reach the perfect absorption and therefore enters that total absorption regime at a lower temperature. At the complete crystallization www.advopticalmat.de of GST, the reflectivities are maximal and are 43% and 52%, respectively. The contrast ratio C(λ) versus the wavelength is reconstructed and shown in Figure 4f, and the maximal contrast value are 1783 and 1910 for samples with 53 and 58 nm GST thickness (i.e., extinction ratios of -32 and -33 dB, respectively). Moreover, in both cases, the contrast is higher than 256 in the [1318-1367] nm and [1430-1480] nm range, i.e., in a bandwidth of 50 nm (see Figure 4f). Inside the bandwidth, the maximal reflectivity is always higher than 43%.

With Spacer: The Au/ITO/GST System
To experimentally demonstrate the role of the spacer, we have fabricated an Au/ITO/GST system with respective thicknesses of 200/66/21 nm. Figure 5a shows the measured reflectivity spectra upon thermal crystallization of GST, and Figure 5b displays a 2D map of the R(T,λ) values in dB. The largest light modulation resulting from the crystallization of GST occurs at a wavelength of 1485.9 nm, with a reflectivity that is gradually tuned from 78% to 0.0052%, as illustrated in Figure 5d. Strikingly, this measured minimum of reflectivity translates into an absorption A ≈ 99.995%, that is, a nearly complete absorption of light. At this wavelength, we therefore measure a contrast C higher than 15 000 and an extinction ratio ER = −42 dB. The measured contrast is higher than 256 in the wavelength range 1423-1515 nm, that is a bandwidth of 92 nm (see Figure 5c). For proper comparison with other works, we also give here the modulation depth (MD) of the system, defined as the difference between reflectivity extrema: MD = R max − R min . For this sample, the maximal modulation depth is MD = 0.78.
The usefulness of the spacer is clearly highlighted in this experiment: By adding a properly designed spacer layer, we experimentally show we can both strongly enhance the modulation amplitude and tailor the crystallization fraction of GST for which perfect absorption occurs. It is worth noting (see Figure 5d) that we have recorded 158 distinct optical reflectivity levels during the thermal crystallization. Encoding in 256 grayscale levels appears therefore experimentally feasible.
These results are experimental demonstrations of the previously listed advantages of our system: i) a large dynamical modulation from strong reflection to perfect absorption with contrasts higher than 10 4 ; ii) a designer wavelength of operation, tailored via the GST thickness; iii) an adjustable crystalline fraction at which occurs the perfect absorption.

System with Electrodes
We have also computed the reflectivity for a more practical system in which GST is sandwiched between two ITO layers at 1550 nm wavelength. ITO is a transparent conductive oxide allowing the application of an electric field between the two Figure 5. Experimental results on the Au/ITO/GST layered system. a) Evolution of reflectivity versus thermal crystallization. Each colored curve represent a different temperature, from 27 °C (blue) to 190 °C (red). b) Experimental reflectivity 10Log(R exp ) value versus wavelength and temperature. c) Experimental contrast C versus the wavelength, reaching a maximum of C = 15 000. The dashed black line represents the level above which an 8-bit encoding is possible. d) Measured reflectivity curve during phase transition at λ = 1485.9 nm, leading to a dynamic modulation from strong reflection to perfect absorption.

www.advopticalmat.de
conductive layers or between the top ITO layer and Au layer. This layered system can be a good candidate to modulate light electrically and, in addition, the top ITO layer can serve as a passivation layer to protect GST. For a given thickness of bottom ITO layer (60 nm), we found that there is always a couple of GST/top ITO thickness that provides a large modulation of reflectivity during GST phase variation. It is illustrated in Figure 6a at λ = 1550 nm. The region inside the black contour corresponds to GST and top ITO thickness values, where both the maximal reflectivity and contrast are higher than 50% and 256, respectively. We can see that absorptions greater than 99.999% can theoretically be reached. For solutions with thin top ITO thickness, the maximum of reflectivity is higher than 70%. The bottom ITO layer thickness can be arbitrary chosen. However, the reflectivity can be maximized with an optimal bottom ITO thickness which is around 70 nm. Similar results are also observed for the other wavelengths in the 1300-1700 nm range. Concerning the efficiency bandwidth, we found that the Au/ITO/GST layered system can encode an 8-bit contrast within a 150 nm bandwidth. For the AU/ITO/GST/ITO layered system of Figure 6c,d, the bandwidth is 220 nm. Furthermore, by choosing appropriate Au/GST/ITO thicknesses, the bandwidth could be as wide as 300 nm (e.g., with 200/59/302 nm, results shown in Figure S6, Supporting Information).

Discussion
Recent works exploited PCMs as a means to modulate the free space optical reflectivity or absorption but with either a limited amplitude of modulation [20,24,39] and/or with complicated geometries (metal-insulator-metal architectures or devices requiring several nanopatterning technological steps [14,22,[40][41][42] ). Here, we reveal that there is no need to design and fabricate complex nanostructured devices to reach a very large modulation of reflection/absorption and that a properly designed flat stack of thin-films could produce a dynamical reflectivity modulation with contrast and extinction ratios as high as 10 6 and -67 dB, respectively. These figures of merits actually surpass state-of-the-art modulators such as commercially available lithium niobate or liquid crystals free-space modulators that usually present extinction ratio >10 dB. Our experimental demonstration of a -42 dB extinction ratio also compares favorably to recent experimental works on nanopatterned devices with functional materials such as graphene or VO 2 , which reported extinction ratios in the range -10 to -25 dB. [14,30] Furthermore, designer modulation properties, tailored for specific requirements, can be obtained by adjusting the thicknesses of both the PCM and the dielectric spacer, providing two degrees of freedom to design the system. The use of a spacer has been reported elsewhere for binary switch The black contour delimits the region inside which the contrast C is higher than 256 and the maximum of reflection is higher than 50%. c,d) Light modulation and angle of incidence: c) simulation of the reflectivity of an Au(200nm)/ITO(60nm)/GST(20nm)/ITO(320nm) layered system versus crystallization rate τ and angle of incidence θ at 1550 nm wavelength; d) contrast C (blue curves) and maximal reflectivity R max (red curves) at two angles of incidence (0°, full line and at 30°, dashed line).

www.advopticalmat.de
configuration, [15,22] but its role remained elusive and had not been explicitly pointed out. A simplified view would describe the PCM and spacer thicknesses as two separate control sticks to adjust the amplitude and phase of reflected fields, respectively. By adjusting the thickness of the spacer, we alter the interference via the introduction of a supplementary phase. On the other hand, by selecting the good thickness of GST layer, it becomes easy to tune the wavelength range of efficient light modulation. Furthermore, in this work, we restricted ourselves to the use of GST, but many other PCMs with various merits are readily available. For example, the emerging large bandgap PCM Sb 2 S 3 could be exploited to expand these designs to the visible range. [43] The different methods presented here can directly be applied to any other tunable material and should therefore enable demonstrating large reflectivity modulation devices ranging from the visible to the mid-infrared ranges. Such tailorable properties are much needed, as each given application will require specific crystallization fraction, wavelength of operation, and optical contrast. Additionally, it is possible to maximize the modulation using both ultra-thin layers of PCMs and low crystallization fraction, as shown in Figure 3d-f: These two combined features should lead to lower energy consumptions and faster switching times.
Another strong advantage of having flat, deeply-subwavelength layers, compared to patterned or metal-insulator-metal architectures, is the relaxed angle-dependence. So far, all theoretical and experimental results have been presented for normal incidence, but the results remain valid up to an angle of ±30°. Modulation and perfect absorption are very good in this range of angles, as shown in Figure 6c,d. Actually, we can see that even at 30° the wavelength bandwidth of large modulation is nearly 200 nm. Very importantly, all the methods described in this work can be directly applied to optimize the working conditions around a particular angle simply by adjusting the thicknesses of GST, spacer and potential ITO electrodes. As an example, we have designed an Au/ITO/GST/ITO system that presents a modulation contrast higher than 10 5 at an incident angle of 45° (see Figure S7, Supporting Information).
In view of these different features such as design flexibility, large reflectivity modulation, and ease of fabrication, many straightforward applications could be envisioned: flat integrated free space modulator, optical shutter, optical power filter, or optical limiter. Using an array of electrodes, this platform could find applications in faster adaptive optics platforms which are used in increasing fields of applications. [44,45] The unpatterned and low-thickness nature of the platform makes its fabrication readily accessible to large-scale fabrication for potential integration as dynamically tunable optics in large-dimension optical components that are typically used in very large instruments such as telescopes or membrane optics in spacecrafts [46][47][48] Finally, we want to emphasize that the unique physical properties of PCMs enable to explore more complex modulation schemes, and in particular by exploiting the multilevel states of partial crystallization. Indeed, in our experimentally shown -42 dB dynamic range modulation of reflectivity, we have been able to set more than 150 distinct states of reflectivity and to reach perfect absorption at an intermediate crystallization state of GST (see Figure 4). Each of these measured reflectivity level corresponds to distinct non-volatile states of partial crystallization. Provided one can precisely and spatially control the local crystalline fraction of PCM, we envision this multilevel non-volatile encoding of reflectivity to be used for a wealth of applications, including grayscale metasurfaces for complex wavefront shaping, reflectivity control devices for attitude control in solar sails [49] or new kinds of SLMs producing drastically improved holograms' definition compared to the one generated by binary SLM. [50,51]

Conclusion
We have experimentally demonstrated a large optical modulation from a strong reflection R ≈ 80% to a perfect absorption (A ≈ 99.995%) with a contrast ratio of ≈15000 using simple unpatterned thin-layers of PCM. We further show that we can tailor by design: i) the efficient working wavelength conditions throughout the infrared domain, ii) the angle of illumination over all angles, and iii) the crystalline fraction of the GST layer at which the tunable perfect absorption occurs. For a fixed layered system thickness, we can obtain a 150 to 300 nm wavelength window enabling modulation with high contrast (higher than 256). Tuning the crystallization fraction where occurs perfect absorption enables a simple binary switch that could theoretically lead to a very deep extinction ratio (up to -68 dB) between the amorphous and crystalline phase. On the other hand, tailoring the crystallization fraction where occurs perfect absorption could be used to modulate light electrically in an ultrathin, low-energy consumption platform via grayscale level schemes.

Experimental Section
Sample Fabrication: The fabrication starts by depositing a 200 nm Au layer (e-beam evaporation) on a silicon substrate, followed by depositing a GST layer. GST layers were obtained by using a Bühler SYRUSpro 710 machine associated with an OMS 500 optical monitoring system. Granules made with stoichiometric Ge 2 Sb 2 Te 5 were placed into a Mo liner. A focused electron beam was then used to heat-up the material with typical current of a few tens of mAmp. Specific e-beam pattern was developed in order to secure uniform evaporation of the material. Samples were placed onto a rotating calotte situated at a distance of about 600 mm from the crucible to allow achieving layers with good uniformity over the substrate aperture. Deposition were carried out at room temperature, and GST was evaporated at a rate of 0.25 nm s -1 that was controlled with a quartz crystal microbalance. A relative precision of the thickness better than 1 nm and an absolute precision within 2% was achieved using this technique. Finally, previous expertise [52] has shown that by properly adapting the deposition parameters, it is possible to keep a composition close to that of a raw material, securing that the properties of the initial material will be maintained. Thin-films of ITO were deposited by magnetron sputtering in a partial Ar/O 2 atmosphere and the substrate was kept at 250 °C during deposition.
Measurements of GST Optical Properties: The GST samples were optically characterized between 260 and 2100 nm using a Horiba Jobin-Yvon spectroscopic ellipsometer. To study the crystallization of the GST layers, samples were placed on a heating stage for 5 min before being allowed to cool down to room temperature. The ellipsometry spectrum is then measured at room temperature, to avoid crystallization during the measurement. This process is repeated using the same sample and increasing the temperature of the stage by 10 °C steps, starting from 60 up to 180 °C. The different spectra are then individually fitted using www.advopticalmat.de a model comprising a GST layer following a Tauc-Lorentz dispersion formula. This enables the extraction of the refractive index and extinction coefficient of the GST for each intermediate crystallization fraction from amorphous to fully crystalline.
Reflection Measurements: For reflectivity measurement, the light of a broadband superluminescent LED (SLED) was sent on the GST side with a 5× M plan apo NIR Mitutoyo objective. The focused spot diameter at the sample was around 30 μm and the cone angle of illumination was around 3°. The reflected light was sent to an optical spectrum analyzer (Anritsu MS9740 A), and the spectrum was recorded in the 1250-1650 nm range. The theoretical dynamic range of this optical spectrum analyzer was 90 dB. In the sample measurements, due to the incoming light power and the dynamic range of the optical spectrum analyzer, it was possible to measure up to -57 dB extinction ratio. The samples were heated with a 1 °C min -1 temperature ramp. A single reflectivity spectrum was recorded every 2 °C step for samples without ITO spacer and 1 and 0.5 °C for sample with ITO spacer. The reflected spectra were all normalized to the reflection spectrum of a silver mirror in order to accurately measure the absolute reflectivity of the sample R exp (λ, T).

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.