Reconfigurable phase-change meta-absorbers with on-demand quality factor control

Perfect absorber type devices are well-suited to many applications, such as solar cells, spatial light modulators, bio-sensors, and highly-sensitive photo-detectors. In such applications, a method for the design and fabrication of devices in a simple and efficient way, while at the same time maintaining design control over the key performance characteristics of resonant frequency, reflection coefficient at resonance and quality factor, would be particularly advantageous. In this work we develop such a method, based on eigenmode analysis and critical coupling theory, and apply it to the design of reconfigurable phase-change metasurface absorber devices. To validate the method, the design and fabrication of a family of absorbers was carried out with a range of ‘on-demand’ quality factors, all operating at the same resonant frequency and able to be fabricated simply and simultaneously on the same chip. Furthermore, by switching the phase-change layer between its amorphous and crystalline states, we show that our devices can provide an active or reconfigurable functionality. Published by The Optical Society under the terms of the Creative Commons Attribution 4.0 License. Further distribution of this work must maintain attribution to the author(s) and the published article’s title, journal citation, and DOI.


Introduction
The ability to design devices capable of controlling the propagative characteristics of light in the infrared part of the electromagnetic spectrum is essential to a number of technologically important fields, such as telecommunications, sensing, imaging, spectroscopy, security and biosensing.In recent decades, the progress in these fields has been substantially influenced by the rise of plasmonic metamaterials.Plasmonic metamaterials [1,2] are artificial structures composed of metallic nanoresonators arranged in periodic sub-wavelength arrays.The resonators can be designed to have effective permittivities and permeabilities not achievable using conventional materials [3,4].Plasmonic metasurfaces [5][6][7], two-dimensional counterparts of plasmonic metamaterials, are particularly interesting from the practical point of view, as their fabrication is fairly straightforward and they can offer relatively small insertion losses.Different kinds of plasmonic metasurfaces have been proposed in the literature to serve a wide range of applications, such as beam steering [8], polarization control [9] and perfect absorption [10,11].Furthermore, to achieve dynamic tunability and/or reconfigurability, so creating active devices, plasmonic metasurfaces can also be combined with liquid crystals, gated semiconductors, or, as in this work, phase-change materials [12,13].
The phase-change material (PCM) based approach stands out from the other means of creating dynamic and reconfigurable plasmonic metasurfaces [14].PCMs typically comprise thin-films of chalcogenide alloys, such as the well-known ternary material Ge2Sb2Te5.Under electrical, optical or thermal stimulus, PCMs undergo a phase transition, i.e. switch between amorphous and crystalline states, with a resulting dramatic change in their electrical and optical properties [15].Amorphization requires temperatures in the phase-change material to reach melting point (~900 K for Ge 2 Sb 2 Te 5 ) and be followed by a very fast cooling (tens of degrees per nanosecond for Ge 2 Sb 2 Te 5 ) [15].Crystallization requires lower temperatures, lying between the glass-transition and melting temperatures [15].PCMs possess a high optical contrast (refractive index difference) between phase states in the infrared region and nearly-zero optical losses when in the amorphous state [16][17][18].They also benefit from fast switching (subnanosecond electrically [19] and tens of picoseconds optically [14]), high endurance (potentially up to 10 15 switching cycles [20]) and, importantly, they exhibit non-volatility in both phases (and even in mixed, i.e. fractionally-crystallized, phases).This allows for the utilization of chalcogenide PCMs in a variety of novel and reconfigurable infra-red devices such as non-volatile solid-state reflective displays [21], optical beam steerers [22,23], spatial light modulators [12], reconfigurable lenses [13,24], integrated photonic memories and computing devices [25][26][27] and, the topic of this paper, PCM-based reconfigurable perfect absorbers [28][29][30][31][32].
Perfect absorbers can be utilized in a number of infrared device applications, for instance, radiation detectors [28,33,34], solar cells [35], spatial light modulators [29,36,37] and biosensing [30,[38][39][40].Various approaches have been used in the literature to design and explain the operation of perfect metamaterial absorbers, including analogies with plasmon slot waveguides [41,42], optical impedance matching [43,44], application of the first Kerker condition [45], critical coupling mechanisms [46,47] and considerations of the absorption cross section of the resonators that comprise the metasurfaces [48].All of these methods succeed in giving an explanation of the operation of perfect absorbers, however they differ significantly in terms of complexity and, importantly, their ability for quantitative design of the device response.Indeed, since the performance of metamaterial perfect absorbers can be broadly described by a few key parameters, namely the resonant frequency, bandwidth, quality factor and reflection coefficient (at resonance), a quantitative design methodology that predicts performance in these terms, in a simple and economical manner, is vital when trying to engineer device specifications that a certain application demands.For instance, in the field of infrared spectroscopy, perfect absorbers have been used to enhance the interaction of infrared radiation with biological analytes [30,38,39].Here, by controlling the quality factor of the device, one can achieve a better analyte selectivity.Moreover, for increased sensitivity of detection, accurate control of the absorber's coupling to the incoming radiation is crucial [40].To give another example, in the field of plasmonic color printing, figures of merit such as the color saturation are directly related to the quality factor of the resonant absorption spectra [48].Finally, as a last example of the importance of a quantitative methodology for designing the quality factor in optical devices, we note that in metal-insulator-metal structures used to create (optical) phase gradient metasurfaces with beam steering capabilities [23], control of the quality factor is important since it determines how sharply the optical phase varies around resonance.(We also note that the requirement of a full 2π phase coverage is present only in overcoupled resonators) [47,49].
In this paper, therefore, we derive, based on eigenmode calculations together with the concept of critical coupling [46], a design methodology that is not only relatively simple but which is also inherently well-suited to the quantitative design/selection of key absorber parameters including the resonant frequency, bandwidth and degree of coupling (minimum reflection at resonance).Moreover, we demonstrate the validity of our proposed methodology via the experimental fabrication and characterization of PCM-based reconfigurable absorbers working in the near infrared waveband (λ = 1550 nm) and with a purposely engineered quality factor.Our devices are made with a nanostrip Al metasurface placed on top of a layered structure comprising a thin ITO film, a chalcogenide phase-change material Ge2Sb2Te5 (GST) and a bottom Al mirror (see Fig. 1) [31].The ITO is used to prevent environmental oxidation of the GST layer, while it can be also utilized as a top electrode for electrical switching of the GST [21,50].By transforming the GST between the amorphous and crystalline states our device can be operated in two distinctly different regimes.When the GST is in the amorphous state, the resonant gap plasmon [51,52] of the metal-dielectric-metal structure is excited and the incoming radiation is absorbed (with the amount of absorption dependent on the degree of coupling).However, when GST layer is switched into the crystalline state the resonance condition is lifted due to the change of the refractive index, resulting in specular reflection of the incoming radiation.
We show how the absorber resonant frequency can be accurately calculated for a structure composed of materials that presents optical losses, using the result to obtain external and internal decay rates of the device [53].These decay rates define two most important parameters of a practical absorber -i.e. the quality factor and the reflection coefficient at resonance [49].It is shown that the quality factor of our devices can be controlled, for a fixed operating frequency, by changing only one design parameter -the periodicity of the nanostrips comprising the optical metasurface -while leaving other design parameters intact.This allows us to produce simple guidelines for engineering absorbers with desired characteristics, without the need to run time-consuming finite element (FEM) simulations.We test the practicability of our modeling results by fabricating several representative devices and measuring their amorphous and crystalline responses.

Quality factor engineering
Eigenmode calculations were used to optimize a number of device geometrical parameters so that phase-change absorbers, of the type shown in Fig. 1, could be realized with various quality factors at the resonant wavelength of λ = 1550 nm.More specifically, the following device parameters -nanostrip resonator width, w p , thickness of the GST layer, t gst , and metasurface period, w uc , -were swept using an optimization routine to achieve the required optical performance.We employed a numerical technique developed by Bai et.al. to excite the structure with a complex eigenfrequency [53].In essence, the method allows us to design a structure with a desired quality factor using the following steps: (i) Excite the device structure with a complex frequency, ω + i/τ, where 1/τ is the decay rate associated with the amplitude of the oscillation for the loaded resonator (called total decay rate in the rest of the manuscript).The real part of the excitation frequency ɷ is fixed while the imaginary part 1/τ is varied in each simulation to achieve devices with desired quality factors.(ii) Search for a device geometry (i.e., in our case, search for values of layer thicknesses, metasurface strip widths and periodicity) that makes the modulus of the electric field evaluated close to the resonator to diverge when excited with the chosen complex (resonance) frequency.This search is done using a global optimization algorithm [54,55].(iii) Once the optimal structure is found, we can relate the value of τ to the value of the quality factor as Q = ω0τ/2 [49].This means that in Step 1 we have chosen the desired quality factor of the structure, provided that we can find an appropriate device geometry in Step 2 (see Methods for further details).This results in significant gains in computational efficiency in the design of devices (as compared to conventional FEM simulations) since we essentially 'tell' the model at which frequency the resonance is required using the real part of the desired eigenfrequency, and we will introduce information about the bandwidth (quality factor) in the imaginary part of the eigenfrequency.So, we are able to directly target the resonant position and bandwidth of the device with far fewer calculations than in the case of a conventional design approach.
Ten different absorber device designs (each design being a vector in the parametric space wuc, w p and t gst ) were obtained for values of the quality factor Q = 4, 4.5, 5, 5.5, 6.In Fig. 2(a), the resulting width of the nanostrip resonators w p is plotted versus the thickness of the GST layer t gst for all cases.It can be seen that all curves cross in a very small region of the graph.This implies that for fixed values of w p and t gst (the crossing point in the plot) one can obtain any desired value of Q (within the range examined here) by varying only the width of the unit cell w uc .To verify this, another set of simulations was carried out to again find structures with Q = 4, 4.5, 5, 5.5, 6, but this time we restricted the values of w p and t gst to be within the narrow region where all the curves in Fig. 2(a) cross (more specifically, for convergence purposes, we set this region to be 2 nm wide in the w p direction and 1 nm wide in the t gst direction).The results are shown in Fig. 2(b) and confirm that we can link the quality factor of our devices to only one geometrical parameter, namely the periodicity of the top plasmonic metasurface, w uc .This is a key finding of the current work, showing how a "family" of absorbers having the same resonant frequency but different Q factors can be readily designed, and subsequently fabricated, using essentially the same device structure (apart from the value of w uc ).

Reflectance Calculations
Having designed, using the eigenmode approach described above, phase-change absorbers resonant at 1550 nm and with a range of quality factors, we next, using 'conventional' FEM techniques [31] (see Methods section), simulated their reflectance spectra over the wavelength range from 1000 to 1600 nm (note that 1600 nm is the longest wavelength examined since this corresponds to the maximum operating wavelength of our spectrophotometer -see Methods section).This enables us to observe the change in the bandwidth and in the reflection coefficient at resonance for different quality factors.Thus, in Fig. 3(a)-(c) we plot simulated reflectance spectra for devices with Q = 4.5 (wuc = 715 nm), Q = 5 (w uc = 643 nm), and Q = 5.5 (w uc = 588 nm).
We also fabricated devices having the same (nominal) nanostrip periodicities, w uc , as those in Fig. 3(a)-(c), and measured their experimental reflectance spectra.The results are shown in Fig. 3(d)-(f) (insets show SEM images of the fabricated devices).There is very good agreement between the modeling and experimental results (note that the width of the nanostrip resonators, w p , hardly deviates from the target value of 451 nm for any of the devices, as can be seen from the Appendix at the end of this manuscript).
Finally, we show experimentally the reconfigurability of our phase-change absorbers by switching the GST layer from the amorphous to crystalline state, here using controllable thermal heating on the hot plate.As discussed in the introduction, after switching, the resonant condition of the device is lifted, resulting in mirror-like specular reflection with efficiency here of above 60% (see Figs. 3(d)-3(f)).Therefore, our devices are indeed reconfigurable and could be used, for example as light modulator here working in reflection and at 1550 nm wavelength (to the best of our knowledge, this is the first experimental demonstration of GST-based reconfigurable perfect absorbers operating in the near-IR band) (See Appendix at the end of this manuscript for details on practical approaches to the switching of devices).

Critically coupled absorbers
The eigenmode and reflectance calculations presented in the previous section allow for accurate extraction of the external and internal decay rates -important characteristics of a practical perfect absorber.The quality factor of a resonating structure, such as perfect absorber, can be expressed through its total decay rate as Q = ω 0 τ/2, where ω 0 is the resonance frequency and 1/τ is the total decay rate.In turn, the total decay rate 1/τ is defined by the internal and external decay rates of the structure, according to the relation: Here 1/τ 0 is the internal decay rate, i.e. the time measure of the internal losses of the resonator through electron scattering, and 1/τ e is the external decay rate, i.e. the time measure of the external losses of the resonator through the emission of radiation.The reflection coefficient Γ(ω) can be expressed in terms of the decay rates as [49]: Thus, once the resonant frequency has been identified via eigenmode calculations, and the reflectance coefficient at resonance evaluated (via FEM simulations of the type used to generate Fig. 3(a)-(c), but carried out only at ω 0 ), then Eq. ( 1) and ( 2) can be solved simultaneously to obtain the values of 1/τ 0 and 1/τ e at ω 0 .
The decay rates thus calculated for a set of perfect absorbers with fixed values of t gst and w p but with differing values of Q are plotted in Fig. 4(a).The interplay between internal and external decay rates determines the onset of critical coupling of the absorber to the incident radiation [47].As is well known, the device is critically coupled when 1/τ 0 = 1/τ e and in such a case all the incoming radiation is perfectly absorbed at resonance.This condition implies matching of the free space impedance to the impedance of the resonating structure.It can also be interpreted as the when the absorption cross section is equal to the unit cell area [48].According to the results shown in Fig. 4(a), critical coupling should occur in our devices when Q = 5 (when w uc = 643 nm).This is confirmed in Fig. 4(b), where the reflectance at resonance is shown for all Q values examined here; the reflectance is zero (perfect absorption) only for the case when Q = 5.Thus, although we can easily obtain (design) absorbers with a range of Q factors, this comes at a price of higher reflection at resonance (but the reflectance is still below 10% at resonance for all the devices studied here).
It is also convenient to introduce another measure of the performance of the perfect absorber, namely the coupling coefficient CC = τ 0 /τ e , which is plotted in Fig. 4(c).Absorbers with CC > 1 (τ 0 > τ e ) are overcoupled, those with CC > 1 (τ 0 < τ e ) are undercoupled, while those with CC = 1 are critically coupled.The results of Fig. 4(c) also demonstrate a clear correspondence between the degree of coupling and the quality factor, reiterating the point that it is not possible to design-in an arbitrary value of Q for any given degree of coupling (for fixed t gst and w p values, as used here).
It is also convenient to introduce another measure of the performance of the perfect absorber, namely the coupling coefficient CC = τ 0 /τ e , which is plotted in Fig. 4(c).Absorbers with CC > 1 (τ 0 > τ e ) are overcoupled, those with CC > 1 (τ 0 < τ e ) are undercoupled, while those with CC = 1 are critically coupled.The results of Fig. 4(c) also demonstrate a clear correspondence between the degree of coupling and the quality factor, reiterating the point that it is not possible to design-in an arbitrary value of Q for any given degree of coupling (for fixed t gst and w p values, as used here).
To explore in a little more detail the occurrence of critical coupling in our phase-change absorbers, we compare in Fig. 5 FEM simulated and experimentally measured reflectance spectra for five sets of devices, having nanostrip periodicities ranging from 544 to 816 nm.There is generally good agreement between the simulated (Fig. 5  In particular, we note that, according to the simulations, the absorber with a nanostrip periodicity of 643 nm (w uc = 643 nm) should be the one which is critically coupled, while in the experimental results it is the device with w uc = 715 nm which is closest to being critical coupled.This discrepancy can be explained by an internal decay rate that is larger in the fabricated devices as compared to that in the simulations.This most likely arises due to fabrication imperfections, in particular a non-negligible roughness of the surface of the Al film used to make the periodic nanostrips.The roughness of Al films is known to increase electronic scattering, so affecting their effective optical properties (n and k) and increasing losses and decreasing the quality factor [56].Indeed, from Fig. 4(a) it can be seen that if the internal losses were increased slightly (increased internal decay rate) then the absorber with Q = 4.5 (wuc = 715 nm) would be (closest to being) at the critical coupling condition, while devices with higher Q values (w uc < 715 nm) would possess a monotonically increasing reflectance at resonance.This is precisely the behavior seen in the experimental results of Fig. 5(b).Thus, although the underestimation of the internal losses of our devices resulted in the mismatch between simulated and measured values of the minimum absorber reflectance, this does not affect general findings of the current work.

"On-Demand" quality factors
The above findings clearly show that by fixing two main design parameters of the proposed perfect absorbers -the GST layer thickness, t gst , and width of the plasmonic strip resonators, w p , -it is possible to control the quality factor and coupling coefficient of the device by changing only the resonator periodicity, w uc .Thus, once optimal values of t gst and w p are found, the need to run conventional FEM simulations (that can be both time and computer power consuming) can be avoided.In fact, as we show below, we can go further and produce simple guidelines that allow the design of perfect absorbers with specific 'on-demand' Q factors.
These simple guidelines are summarized and tested in Fig. 6.The internal decay rate 1/τ 0 is observed to vary approximately linearly with the inverse of the width of the unit cell, w uc , as shown in Fig. 6(a).In contrast, the external decay rate, 1/τ e , increases linearly with w uc , as can be seen from Fig. 6(b).Therefore, using Eq. ( 1) and ( 2), we can obtain a simple and quick way to calculate the Q value, for a given value of w uc , via the relationship: where k 0 ( = 1.761x10 16 nm −1 s −1 ) and C 0 ( = 3.409x10 13 s −1 ) are the regression coefficients of the linear fitting in Fig. 6(a), and where k e ( = 2.261x10 11 nm −1 s −1 ) and C e ( = −8.612x10 13s −1 ) are the regression coefficients of the linear fitting shown in Fig. 6(b).The results of applying Eq. ( 3) to find Q for the cases with w uc = 560, 600, 675, 775 nm are shown in Fig. 6(c), where they are compared to Q values calculated via eigenmode simulations.It is clear that that there is excellent agreement between both techniques.Thus, using eigenmode calculations and critical coupling concepts, we have shown that it is possible to design perfect absorbers having a particular desired quality factor (at least over the range of Q values studied here, see the appendix for a discussion on what affects the range of Q values achievable), simply by changing the periodicity of the top plasmonic metasurface.

Conclusions
In conclusion, we have presented a design method to control the main aspects that define the performance of perfect absorbers, i.e., the resonant frequency, bandwidth and reflection coefficient at resonance.To validate the method, we designed and fabricated a family of reconfigurable phase-change perfect absorbers having desired quality factors, obtaining very good agreement between theory and experiment.Moreover, analysis based on a critical coupling approach allowed us to establish simple relationships between device geometry and performance.In particular, we showed how the width of the unit cell is (for fixed values of other device geometric parameters) linked directly to the value of Q.As a result, we could greatly simplify the design and fabrication process of the absorber devices.We demonstrated this by producing a family of devices with a range of 'on-demand' quality factors, all operating at the same resonant frequency and able to be fabricated simultaneously on the same chip (i.e., all with the same layer thicknesses and all with the same width of the top optical metasurface nanostrips).Furthermore, by switching the phase-change (GST) layer in the absorber structure between its amorphous and crystalline states, we showed that our devices can be switched on and off, thus allowing for the design of active and reconfigurable components.For example, we might imagine an application where, by monitoring the interaction of incident radiation with a series of selectively enabled phase-change absorber devices having various quality factors, one can estimate in a quick and simple way the bandwidth of the signal under detection.More generally, our design approach is likely to find use wherever precise control over the resonant frequency and quality factor of plasmonic metasurface resonant absorbers type structures is desired.
The rate of energy dissipation in the resonator is proportional to the amount of the stored energy.Solving the above equation we obtain and by combining both these equations together we get Q ω τ = .It thus follows that The time constant for the amplitude decay is related with the time constant for the energy decay as τ = 2τ eng [49].This means that the amplitude of the oscillations in the system decays twice as fast as the dissipation of energy.We can relate the quality factor Q to the complex eigenfrequency of the system ω 0 + i¤τ via Q = ω 0 τ¤2.The time constant for the decay rate is equal to the time necessary for the amplitude to decay by a factor of e −1 .The bigger the time constant of the amplitude decay, the lower is the rate at which the energy is dissipated in the system and the higher is the quality factor.

SEM images of the fabricated devices
Dimensions of the fabricated devices were checked using scanning electron microscopy (SEM) with the results presented in Fig. 7.One can see that the achieved periodicity and width of nanostrip resonators hardly deviates from the target values of w uc and w p .

Optical Constants
The optical constants used in the numerical simulations presented in the article are referenced and plotted below, see Fig. 8.The optical constants for aluminum (Al) are extracted from the work of Rakic et al. [57], where the experimental data is fitted using the Brendel-Bormann model.The optical constants for indium tin oxide (ITO) were obtained by ellipsometry measurements.Finally, optical constants of GST are the same as in the work by Ruiz de Galarreta et al. [23].

Maximum and minimum achievable Q in the structure
The maximum and minimum values for the quality factor fabricated for the structure in the paper (with the restriction of fixed width of the nanostrip and thickness of GST) were found to be Q = 3.55 and Q = 6.26.For values of the quality factor smaller than 3.55 the period in the structures starts to be large enough (>1000 nm) to have higher diffractive orders within the range of measurement.For values of the quality factor bigger than 6.26, the width of the unit cell is close to the width of the nanostrip, making the fabrication of these structures rather difficult.Other kinds of restrictions not arising from the device geometry but due to the optical constants of the materials used may also affect the limits of the range for the achievable Q factors.In particular, the internal losses will be determined by the interaction of the excited resonant fields and the materials that comprise the device.In that case we may expect higher quality factors (lower internal losses) for all-dielectric and hybrid dielectric/plasmonic structures as compared to the purely plasmonic approach presented here.

Phase switching of devices
As discussed in the main text, the switching of phase-change devices requires, for amorphization, temperatures in the phase-change material to reach melting point (~ 900 K for Ge2Sb2Te5) and be followed by a very fast cooling (tens of degrees per nanosecond for Ge 2 Sb 2 Te 5 ), while crystallization requires heating to temperatures lying between the glasstransition and melting temperatures.Such heating can be delivered electrically by Joule heating (as is routinely carried out in electrical phase-change memory devices, see.e.g [15], Raoux et al.), optically (as in re-writable optical disks, see also [15], Raoux et al.) or by using some form of embedded micro-heater (as demonstrated e.g. as in [50], Au et al.).In the results below, we demonstrate switching of our absorber type devices using both optical and micro-heater approaches.
Optically induced switching: A 405 nm wavelength (blue) diode laser was scanned across the device to induce reversible switching of absorber devices.The duration and power of the laser pulses was appropriately selected in order of crystallize or re-amorphize the phase-change layer (specifically crystallization pulses were of 3.6 mW power, at the sample, and 200 ns duration; re-amorphization pulses were of 13 mW power and 15 ns duration).The reflectance spectrum of the device was measured after each crystallization/amorphization (or SET/RESET) process, and the device was also observed in an optical microscope -see Fig. 9.In Fig. 9(a), it can be seen that repeated cycling of the device can be achieved using this laser scanning approach, but that the re-amorphization stage is probably incomplete (since the spectrum for the amorphous phase of the device shifts in wavelength and decreases in absorption with each cycle).Incomplete re-amorphization is not un-expected however when using a scanning laser, since as the laser scans, the periphery of the laser spot can heat previously amorphized regions of the phase-change layer up to a temperature conducive to crystallization (but note that such effects can be reduced/ameliorated by using very large spot sizes, as for example in [16]).In-situ switching using micro-heater approach: Although optical switching of the phasechange absorber device is possible, it is of course an ex-situ approach (requiring an external laser), and a more attractive approach for practicable devices is some form of in-situ excitation.An already successfully demonstrated in-situ approach is that using a form of micro-heater, as e.g. in [50].Indeed, for the absorber-type devices, it may well be possible to use the top patterned metal layer as a form of micro-heater, i.e. the metal nanostrips play a dual role, providing both plasmonic effects that drive the optical response and providing heater elements.The basic structure and operation of such an absorber device is shown in Fig. 10

Fig. 1 .
Fig. 1.(a) 2D cross section of the unit cell of the reconfigurable phase-change absorber device.Dimensions that define the optical response of the device are shown, as well as the different material layers used.(b) 3D schematic of (part of) the device.

Fig. 2 .
Fig. 2. (a) Perfect phase-change absorbers of various geometries optimized to possess desired quality factors (each color indicates devices with equal values of Q).(b) Width of the unit cell w uc against the corresponding value of Q (the data in this plot corresponds to the optimized devices in the region where curves cross in (a)).Dashed lines are guides for the eye only.

Fig. 3 .
Fig. 3. (a), (b), (c) Simulated reflectance spectra for designs with fixed values of w p and t gst and with Q = 4.5, 5, 5.5 respectively and with the GST layer in the amorphous (colored lines) and crystalline (black lines) states.(d), (e), (f) Experimental reflectance spectra for fabricated devices with the same designs as in top panels (and again with the GST in amorphous and crystalline phases).Insets in (d) to (f) show SEM images of the fabricated devices.

Fig. 4 .
Fig. 4. (a) Total, internal, and external decay rates as a function of Q.(b) Minimum reflectance and (c) coupling coefficient as a function of Q. Dashed lines in (b) and (c) are guides for the eye only.
(a)) and experimental (Fig. 5(b)) results, apart from a mismatch in the minimum reflectance values.

Fig. 5 .
Fig. 5. (a) Modelled reflectance spectra for a family of phase-change absorbers (GST layer in the amorphous state) obtained by varying only the periodicity of the metasurface (the Al nanostrips).(b) Experimental reflectance spectra for fabricated devices having the same geometries as in (a).

Fig. 6 .
Fig. 6.(a) Internal decay rate as a function of the inverse of the plasmonic metasurface periodicity w uc (dashed line is a linear best-fit).(b) External decay rate as a function of the inverse of the plasmonic metasurface periodicity w uc (dashed line is a linear best-fit).c) Calculated values of Q using the data in (a) and (b) against the values for Q obtained from FEM modeling for the same structures (the dashed "45° line" is shown for visual guidance only).
τ being the time constant for the energy decay.The quality factor is defined as 0 eng

Fig. 8 .
Fig. 8. Optical constants (permittivity) used in numerical simulations.(a) Real and imaginary parts of aluminium permittivity extracted from the work of Rakic et al. [57] (b) Real and imaginary part for the GST permittivity in the amorphous and crystalline state from the work by Ruiz de Galarreta et al. [23] (c) Real and imaginary part of ITO permittivity obtained by ellipsometry measurements.

Fig. 9 .
Fig. 9. (a) Reflectance spectra of the absorber device in the pristine (as-deposited) state as well as after several swithing (SET/RESET) cycles.(b) Optical microscope images of the device after a sequence of switching cycles.

Figure 9 (
b) shows optical microscope images of the device after successive SET/RESET cycles; changes in optical contrast of the phase-change layer are readily seen.
: Fig.10(a)shows the basic concept, Fig.10(b) an SEM image of an as-fabricated device, and Fig.10(c)shows the reflectance of the device as it is repeatedly switched between amorphous (low reflectance) and crystalline (high reflectance) states.

Fig. 10 .
Fig. 10.(a) Schematic of an absorber device in which the top patterned metal layer also acts as a micro-heater to switch the phase-change layer beneath.(b) SEM image of the active area of a fabricated device of the type shown in (a).(c) Measured reflectance of device of the type shown in (b) as it is subjected to repeated cycling (by applying pulsed voltages to the micro-heater).