2.1.

Optimization of MPL and Chemical Etch Processing Conditions for the Creation of Microstructures in As₂S₃ Films

A femtosecond Mai Tai Ti:sapphire laser centered at 800 nm with a repetition frequency of 80 MHz is used for MPL in $4\text{-}\mu \mathrm{m}$-thick ${\mathrm{As}}_2{\mathrm{S}}_3$ films thermally deposited on a ZnSe substrate. The beam is passed through a variable attenuator to control the average power and a shutter, both of which are controlled externally though a computer. The beam is focused by either a $50\times /0.55\mathrm{NA}$ or $100\times /0.75\mathrm{NA}$ objective onto the ${\mathrm{As}}_2{\mathrm{S}}_3$ film. The substrate is secured onto a Physik Instrumente H-840 6-Axis Hexapod (programmable stage), and a camera is used to image the sample plane during irradiation. The entire system is controlled by a LabVIEW program that automates the laser writing of predetermined patterns.

A series of experiments were initially conducted in order to dial in on the optimal power, write velocity, and focus depth of laser into films. Dose arrays of raster scans, grids, and concentric circles are laser written into the film at varying average powers, write velocities, and focus depths of laser into film to test for the best etch selectivity as a function of these parameters. Examples of these laser written patterns are shown in Fig. 1. A range of powers is defined as a percentage of maximum possible average power at the sample plane from the variable attenuator. For the $50\times /0.55\mathrm{NA}$ objective, 100% power corresponds to 40 mW and for the $100\times /0.75\mathrm{NA}$ objective, 100% power corresponds to 35 mW. The etch results from these experiments indicate that a write velocity of $0.25\mathrm{mm}/\mathrm{s}$ at an average power in the 40% to 60% range induces the desired molecular change to the film without ablation. Higher powers and lower write velocities were usually found to ablate the film, whereas lower powers and higher write velocities did not result in good contrast between exposed and unexposed regions after etching. Further experimentation is in progress to determine the depth of focus into film for optimal etch selectivity, as indicated by the $+3\text{-}\mu \mathrm{m}$ legend in Fig. 1. In this scenario, the laser is focused $3\text{-}\mu \mathrm{m}$ deep into the film.

Fig. 1

(a) Raster scan dose array with varying power, indicated by the percentages. $10\text{-}\mu \mathrm{m}$ spacing, $0.25\mathrm{mm}/\mathrm{s}$ write velocity, and $50\times$ objective. (b) Grid dose array with varying power and depth of focus. “$+3\mu \mathrm{m}$” indicates the laser was focused $3\mu \mathrm{m}$ into the film. $5\text{-}\mu \mathrm{m}$ spacing, $0.25\mathrm{mm}/\mathrm{s}$ write velocity, and $100\times$ objective. (c) Same as (b) except the pattern is now a group of concentric circles. (d) Raman spectra graph for the region shown in the embedded image. The microscope image ($20\times$ magnification) is of a $10\text{-}\mu \mathrm{m}$ spaced raster scan irradiated at 60% power and $0.25\mathrm{mm}/\mathrm{s}$ write velocity with a $50\times$ objective. Each red spectrum is measured from the as-deposited film, whereas each blue spectrum is measured from a laser exposed region.

The postirradiation molecular structure of the film was investigated using micro-Raman spectroscopy. Figure 1(d) shows the Raman spectra collected at $5\text{-}\mu \mathrm{m}$ intervals over a raster scan (60% power, $0.25\mathrm{mm}/\mathrm{s}$ write velocity). A Horiba Jobin Yvon confocal micro-Raman system with an excitation wavelength of 785 nm was used. The beam was focused by a $100\times /0.75\mathrm{NA}$ objective and the Raman signal was collected for 60 s at a low optical power (reduced by optical density filter), to prevent any undesired photoinduced change to the film. The laser-induced film cross-linking causes the exposed regions of the film to exhibit a different Raman signature from the surrounding pristine film.^11,13 Most notably, the ${\mathrm{As}}_4{\mathrm{S}}_4$ molecular clusters that are present in the as-deposited/unirradiated film, characterized by the Raman peaks at 222 and $363{\mathrm{cm}}^{-1}$ ^11,13 [as illustrated by the red spectra of Fig. 1(d)] emerge during the laser exposure process (blue spectra), leading to good etch selectivity between these two regions.

After the patterns are written using the laser, the sample is submerged into a solvent mixture of DIPA and DMSO for the etching process. After the sample is initially immersed for an hour, the resulting structure is analyzed using a ZYGO 6300 white-light interferometer (Fig. 2). Subsequent etching is done in 30-min increments until the profile of the structure is fully developed. Etch times vary depending on laser power and target resolution of laser-written structures, but it typically takes between 1 to 4 h of immersion. Good etch selectivity has been shown repeatedly down to $7.5\text{-}\mu \mathrm{m}$ feature spacing, as shown in Fig. 2(a). The pictured $7.5\text{-}\mu \mathrm{m}$ grid structure in Fig. 2(a) (irradiated at 60% power or 21 mW, $0.25\mathrm{mm}/\mathrm{s}$, and $100\times$ objective) was achieved with 1 h of etching. At $5\text{-}\mu \mathrm{m}$ feature spacing, etch selectivity takes longer to develop. This can be attributed to undesired partial exposure of the film in the regions between the laser write paths.

Fig. 2

(a) $7.5\text{-}\mu \mathrm{m}$ spacing grid after 1 h of etching with its respective height map. (b) $5\text{-}\mu \mathrm{m}$ spacing grid after 1 h of etching, with its height map that indicates it is not fully resolved. (c) Same grid in (b) after 240 min of etching, showing that is has become fully resolved.

The height map shown in Fig. 2(b) indicates that the structure did not fully resolve after 1 h of etching. Subsequent etching on this sample resulted in the desired structure, as evident by the height map in Fig. 2(c). This $5\text{-}\mu \mathrm{m}$-spaced structure (irradiated at 60% power or 21 mW, $0.25\mathrm{mm}/\mathrm{s}$, and $100\times$ objective) took 240 min of etching to fully resolve.

In summary, MPL is shown to be effective technique for photopatterning $4\text{-}\mu \mathrm{m}$-thick thermally deposited ${\mathrm{As}}_2{\mathrm{S}}_3$ films. Experimentation through repeated dose arrays indicate that an average irradiation power between 40% and 60% at a write speed of $0.25\mathrm{mm}/\mathrm{s}$ induces the desired molecule change in the films. Good etch selectivity for these conditions were shown repeatedly for $7.5\text{-}\mu \mathrm{m}$ feature spacing after 1 h of etching and for $5\text{-}\mu \mathrm{m}$ feature spacing after 240 min of etching.

2.2.

Multiphoton Lithography of Functional GRIN Lenses

Single-layer ${\mathrm{As}}_2{\mathrm{S}}_3$ films are inadequate for nanostructure MPL, where the incident wavelength is on the order of, or smaller than, the thickness of the film, because of the standing wave caused by reflection at the substrate interface.¹⁵ This effect leads to the production of unintended nanobead structures with large sizes and poor substrate adhesion. In this section, we investigated the properties and processing of MPL in single- and multilayered films of ${\mathrm{As}}_2{\mathrm{S}}_3$ thermally deposited onto an underlying thin film of arsenic selenide (${\mathrm{As}}_2{\mathrm{Se}}_3$) and the creation and optical characterization of a GRIN effective medium lens.

Thermal deposition was used to create a multilayered film of ${\mathrm{As}}_2{\mathrm{S}}_3$ on an anti-reflective (AR) layer of ${\mathrm{As}}_2{\mathrm{Se}}_3$. A layer of ${\mathrm{As}}_2{\mathrm{Se}}_3$ was deposited first on polished silicon wafers followed by thermal evaporation of a layer of ${\mathrm{As}}_2{\mathrm{S}}_3$. The thermal deposition process has been described previously.^8,9,15,16 The ${\mathrm{As}}_2{\mathrm{S}}_3$ layer was deposited at a rate of $30Å{\mathrm{s}}^{-1}$, and the thickness of the resulting film was $896\pm 13\mathrm{nm}$. The ${\mathrm{As}}_2{\mathrm{Se}}_3$ layer was deposited at a rate of $4Å{\mathrm{s}}^{-1}$ and the thickness of the resulting film was $203\mathrm{nm}\pm 3\mathrm{nm}$. The ${\mathrm{As}}_2{\mathrm{Se}}_3$ serves as an AR layer, which reduces reflection of the incident laser beam back into the focal volume during laser patterning.

MPL was performed using a system and process described previously.¹⁷ MPL was conducted using the output from a mode-locked Ti:sapphire laser ($\lambda =800\mathrm{nm}$, 76 MHz repetition rate, and 120 fs pulse duration). An acousto-optic modulator was used for the electronic control of the intensity of the laser beam. The laser pulses were focused into the film with a $100\times /1.4\mathrm{NA}$ oil-immersion objective (Nikon). GRIN lenses were photopatterned by translating the film along programmed paths in the $x$, $y$, and $z$ directions at a speed of $50\mu \mathrm{m}{\mathrm{s}}^{-1}$ while keeping the laser beam fixed, and then opening and closing a shutter in synchrony with the movements to control local exposure. The average focused laser power was varied from 0.05 to 0.50 mW. The detailed chemical etch processing has been described elsewhere.^8,16

Incorporating the etch-resistant ${\mathrm{As}}_2{\mathrm{Se}}_3$ AR-layer between the ChG fabrication layer and the substrate reduces the amplitude of a standing wave caused by reflection at the substrate interface.^15,18 This reduction resulted in the generation of pillars with uniformly cylindrical shapes with increased material–substrate adhesion that were more likely to obtain a smaller size before delamination than the bead-shaped features created in the single layered film [Fig. 3(a)]. The uniform shapes also allowed for more predictable fill fractions when compared to stacked bead shaped nanostructures.^8,15 Predictable fill fractions are important for GRIN lens design when determining effective index and optical function. Large pillar arrays of uniformly cylindrical shapes were demonstrated in the previous works.^8,15

Fig. 3

(a) Comparison of pillars fabricated in single-layered film (top) and multi-layered film with an AR coating (bottom). (b) Diagram of the source beam interaction with sample. (c) GRIN effective-medium lens ($\text{diameter}=20\mu \mathrm{m}$) designed to have a spherical refractive index profile and fabricated by MPL in thermally deposited ${\mathrm{As}}_2{\mathrm{S}}_3$. This device consists of cylindrical pillars, which increased in fill fraction (effective refractive index) from the outside diameter to the inner. (d) Transverse intensity scans within focal plane of the GRIN effective medium lens ($\text{diameter}=100\mu \mathrm{m}$) designed to have a spherical refractive index profile.

GRIN effective-medium lenses with diameters of 20 and $100\mu \mathrm{m}$ were successfully fabricated in the multilayered film. Figure 2(c) shows scanning electron microscope images of the top-down view of the GRIN lens having a diameter of $20\mu \mathrm{m}$. The inset to Fig. 2(c) shows the top-down view of the GRIN lens having a diameter of $100\mu \mathrm{m}$.

The GRIN lenses consisted of cylindrical pillar features separated by a pitch of 500 nm. The lenses were designed to have a spherical refractive index profile that ranged from $n=1.06$ to 2.45, the latter representing a region completely filled in with ${\mathrm{As}}_2{\mathrm{S}}_3$. The corresponding total optical phase delay from center to edge of the lens would be $1.26\pi$. The index profile was created by fabricating pillars having diameters that decreased with radial distance from the center of the lens, so that the corresponding decrease in local fill-fraction of the high-index ChG creates a local refractive index that follows that targeted spherical refractive index profile. Across the lens, the pillars had consistent height, and within a small designed region, the diameters of the pillars were similar. Widths of the pillars ranged from $120\mathrm{nm}\pm 14\mathrm{nm}$ fabricated at a power of 0.16 mW to pillars large enough to overfill the entire area. The pillar heights were measured to be a $907\mathrm{nm}\pm 10\mathrm{nm}$ across the structure and were fixed based by the original film height.

The GRIN lenses were characterized using the output from an Er:YAG laser ($\lambda =2.94\mu \mathrm{m}$, $170\text{-}\mu \mathrm{s}$ pulse duration) with the system shown schematically in Fig. 3(b). The laser beam over-filled the sample area so that the intensity distribution across the microlens was uniform. An optical fiber coupled to an IR photodetector (ThorLabs PDA 20H) was mounted on a three-axis translation stage and positioned behind the sample. Transmitted light collected by the fiber provided a relative measure of the local intensity as a function of position behind the lens. Measurements of transmitted light far from the lens provided a baseline measurement that can be used to correct for loss due to reflection from the supporting silicon substrate.

Figure 3(c) shows data from the optical characterization of the $100\text{-}\mu \mathrm{m}$ diameter GRIN lens shown in the inset. The black trace is a line scan of light intensity transmitted through the silicon substrate far from the lens. The transmission is roughly 50%, as expected due to reflection losses at both the optical interfaces of the front- and backsides of the silicon substrate. The transmission versus position is uniform within the shot-to-shot variation of the pulsed laser, which confirms that the lens will be uniformly illuminated when positioned in the probe beam. Figure 3(d) shows the transverse intensity scans within the focal plane of the GRIN lens. Irradiance profiles generated by scanning in the $x$ or $z$ direction of the lens resulted in peaks of $\sim 250\%$ of $I_0$.

A homogenous phase shifting aperture “block” with a $100\text{-}\mu \mathrm{m}$ diameter was created and characterized in the same manner as the lens. This was done in order to benchmark the transmission of the lens and account for increased throughput on the front-interface of the substrate that may result from refractive-index matching caused by the nanostructures. At the focal point of the lens, the intensity was found to be two times more intense than that transmitted through the uniform block, and four times as intense as light transmitted through the unpatterned substrate away from the lens. The transmitted light behind the lens had a clear focal structure with an FWHM of $32.1\mu \mathrm{m}\pm 1.9\mu \mathrm{m}$ for the $x$ scan and $28.6\mu \mathrm{m}\pm 1.5\mu \mathrm{m}$ for the $z$ scan. The effective width of the incident beam may be taken to be equivalent to the effective aperture set by the width of the GRIN lens, which was $\sim 140\times 250{\mu \mathrm{m}}^2$ at 90% intensity.

The addition of an AR layer enabled the fabrication of robust reproducible pillars with better material substrate adhesion and smaller dimensions via MPL in ${\mathrm{As}}_2{\mathrm{S}}_3$ film. GRIN lenses with spherical index profiles were created and IR characterization indicated optical functionality. This work opens new routes to ChG-based electronics and optical devices such as detectors, sensors, photonics waveguides, and acousto-optics.

3.1.

Nanopatterning of Chalcogenide Phase Change Material via Thermal Scanning Probe Lithography

Stylus-based writing based on AFM techniques offers significantly improved resolution not constrained by optical diffraction. So far, AFM-based patterning of PCMs has been limited to external laser-assisted heating or using conductive atomic force microscopy (c-AFM).^44,45 However, the c-AFM method is limited to substrates that can form a conductive loop with the AFM tip to facilitate Joule heating. An alternative AFM-related technique, TSPL, has recently demonstrated its promise for high-resolution lithography processes in nanoelectronics,^46–49 thermochemistry,^50,51 and biomolecular nanodevices.⁵² The approach enables simultaneous in situ patterning and AFM imaging [Fig. 4(a)], and unlike c-AFM, it does not require a closed circuit, making it compatible with a diverse range of substrates. Using this technique, we demonstrated nanoscale patterning of an amorphous GSST film with 30-nm thickness. Micro-Raman measurements confirmed that the thermal AFM writing process leads to local crystallization of GSST, evidenced by a shift in the Raman signal peak from near 160 to $120{\mathrm{cm}}^{-1}$, associated with the crystalline phase of GSST [Fig. 4(b)]. In Fig. 4(c), a TSPL-written 100-nm-wide serpentine line pattern is imaged by tracing the changes of GSST volume upon switching.⁵³ The result indicates the potential of the thermal AFM technique for nanoscale patterning of PCMs.

Fig. 4

Thermal scanning probe patterning of the phase-change material GSST. (a) Optical micrograph of a thermal probe (NanoFrazor, Swisslitho AG) that offers simultaneous in situ imaging and patterning capabilities. (b) Raman spectra of as-deposited GSST film and the same film in regions patterned with TSPL. (c) Depth map displaying a 100-nm-wide crystalized line resulting from volume contraction upon GSST crystallization. (d) A snapshot cross-sectional temperature profile in a 30-nm-thick GSST film on Si after $300\mu \mathrm{s}$ thermal AFM tip dwell time, simulated using the finite-element method. This profile corresponds to the maximum crystallization depth for GSST, before going over the melting temperature ($\sim 890\mathrm{K}$) on the surface.

To further elucidate the PCM patterning mechanism, we performed finite-element simulation of the transient temperature profile in GSST during thermal nanopatterning. The model was implemented in the commercial COMSOL heat module. The thermal probe is approximated as a point heat source with $4\times {10}^{-6}\mathrm{W}$ power following our experimental parameters. Figure 4(d) shows the temperature distribution inside a 30-nm-thick GSST film after a tip-dwelling time of $300\mu \mathrm{s}$. The highly localized temperature profile suggests that nanoscale patterning with resolution better than 20 nm is viable with our TSLP technique. The simulation also indicates that only the top 15 nm of the film is crystallized, largely because the thermally conductive Si substrate serves as an efficient heat sink. By modulating the power on the tip, we envision a modulation on the crystallization depth by reaching different amorphous/crystalline areal distributions, which enables applications such as color depth modulation for grayscale patterning⁴⁵ and continuously phase tuning for holographic⁵⁴ and metasurface applications.⁴² The result validates the potential of TSPL for nanopatterning of PCM ultrathin films, providing a pathway toward realizing topologically optimized, reconfigurable photonic devices.

3.2.

Solution-Deposited Chalcogenide Ge-Sb-Se-Te Films as a Route Toward Low-Loss Phase-Change Optical Materials on Large-Area Nonplanar Platforms

Although GSST PCMs have been fabricated using either a standard melt-quench (bulk) or thermal evaporation (film/waveguides), an ability to conformally deposit PCM layers on large-area nonplanar platforms such as lenses or substrates with complex surface topography has been increasingly demanded. Employing previously demonstrated methods of dispersing chalcogen-containing alloys into solution, one can form films of these alloys following composition-specific protocols. Solution deposition (SD)⁵⁵ is a cost-effective and scalable method where substrates with virtually any shape and size are amenable to forming conformally coated layers, as illustrated in Fig. 5(a). Here, we discuss the ongoing progress made on processing high-performance ${\mathrm{Ge}}_2{\mathrm{Sb}}_2{\mathrm{Se}}_{5-x}{\mathrm{Te}}_x$ PCM alloys, using SD.

Fig. 5

(a) Contrasts in refractive index and optical loss upon ${\mathrm{Ge}}_2{\mathrm{Sb}}_2{\mathrm{Se}}_4{\mathrm{Te}}_1$ switching and an illustration highlighting that SD enables large-area conformal coating; (b) SD process flow; (c) atomic percentage of constituent elements in resulting SD ${\mathrm{Ge}}_2{\mathrm{Sb}}_2{\mathrm{Se}}_4{\mathrm{Te}}_1$ films as a function of thiol-EDA mixing ratio and their comparison with a starting parent alloy composition; and (d) optimum chemical process window dictated by intersolvent solubility and solute dissolution.

Figure 5(b) illustrates the step-wise SD process flow. The process begins with the preparation of bulk ${\mathrm{Ge}}_2{\mathrm{Sb}}_2{\mathrm{Se}}_{5-x}{\mathrm{Te}}_x$ ($x=0.5$ to 1.0) formed using a standard melt-quench method. The bulk alloys are ground and filtered through a $120\text{-}\mu \mathrm{m}$ sieve to realize fine powder and then dissolved into a homogeneous solvent mixture of ethanethiol (thiol) and ethylenediamine (EDA).⁵⁶ Polished and plasma cleaned ZnSe substrates are dipped into the solution using a speed-controlled robotic arm to create films of uniform thickness and surface quality. Finally, the as-coated films undergo soft- and hard-heat treatment to remove residual solvents. To realize optimal performance, SD films must maintain parent alloy stoichiometry, exhibit spatially uniform surface quality, be optically thick ($t\ge 1\mu \mathrm{m}$), be chemically compatible with the underlying IR transparent substrates, exhibit low optical loss in the mid-wave IR, and convert to desired crystalline phases upon thermal or laser treatment. These key attributes assure minimum degradation to the film’s phase change response.

Here, we discuss key process conditions that have deterministic impacts on the resulting quality of coated films. Specifically, the atomic percentages of constituent elements in resulting films are significantly associated with the thiol-EDA mixing ratio. Although Ge and Sb elements have been shown to be easily soluble using amine solvents such as propylamine and EDA,^57,58 Se and Te are inherently more challenging to dissolve due to their preferential dissolution kinetics, necessitating the introduction of an additional solvent such as thiol, as shown in prior efforts.^56,59 Specifically, Te has very low solubility in amine solvents, and thus is partially transferred to a substrate from the solution, disrupting the stoichiometry of resulting films from that of the target parent alloy. This variation, as a function of solvent chemistry, is illustrated in Fig. 5(c). Meanwhile, the composition data extracted using energy dispersive spectroscopy (EDS) on resulting films in Fig. 1(c) indicate that an increase (decrease) in the volume percentage of thiol (EDA) leads to an increase (decrease) in the atomic percentages of Se and Te (Ge and Sb) in a resulting film. Their atomic percentages match those of a parent alloy when the thiol-EDA mixing ratio becomes 50% to 50%. This indicates that the dissolutions of Se/Te and Ge/Sb are likely facilitated using thiol and EDA, respectively. To ensure that resulting films maintain the parent alloy stoichiometry in a spatially uniform manner, thiol and EDA are required to be homogeneously mixed. The top $x$ axis of Fig. 5(d) shows intersolvent solubility as a function of thiol-EDA mixing ratio where H and S correspond to homogeneous and separated, respectively, indicating the presence of an immiscibility zone across 20% to 40% volume percentages of thiol to avoid. The $y$ axis of Fig. 5(d) shows the maximum solute loading level within solutions with varying thiol-EDA mixing ratios where $Y$ and $N$ correspond to yes (complete dissolution) and no (partial dissolution), indicating the maximum loading level achieved at a thiol-EDA mixing ratio of 50% to 50%, which facilitates material transfer throughput from a parent alloy to a coated film. Therefore, the 50% to 50% ratio, represented by the green window in Fig. 5(d), is an optimum condition, which would induce the stoichiometry of resulting films to be spatially homogeneous and consistent with that of a parent alloy while maximizing a film deposition rate.

To assess the phase change behavior of SD films, ${\mathrm{Ge}}_2{\mathrm{Sb}}_2{\mathrm{Se}}_4{\mathrm{Te}}_1$ films underwent in situ grazing-incidence $x$-ray diffraction (GIXRD) at elevated temperatures ranging from room temperature to 325°C in an Ar environment, as shown in Fig. 6(a). The GIXRD data collected at room temperature show weak yet noticeable peaks corresponding to Te crystals. Those collected at elevated temperatures up to 325°C show the evolution of peaks corresponding to hexagonal ${\mathrm{Sb}}_2{\mathrm{Te}}_3$ crystals, which give rise to an increase in refractive index of the film,¹⁹ whereas the Te peaks are carried over to high-temperature data. Given the fact that in situ GIXRD data collected from thermally deposited ${\mathrm{Ge}}_2{\mathrm{Sb}}_2{\mathrm{Se}}_4{\mathrm{Te}}_1$ films show ${\mathrm{Sb}}_2{\mathrm{Te}}_3$ peaks with a relatively higher intensity,¹⁹ Te crystals existing in SD films are estimated to partially suppress the formation of the desirable ${\mathrm{Sb}}_2{\mathrm{Te}}_3$ crystals upon heat treatment. This necessitates a method to reduce inherent Te crystals in starting films, thereby promoting the formation of desirable ${\mathrm{Sb}}_2{\mathrm{Te}}_3$ crystals. To achieve this, Te content was reduced from ${\mathrm{Ge}}_2{\mathrm{Sb}}_2{\mathrm{Se}}_4{\mathrm{Te}}_1$ to ${\mathrm{Ge}}_2{\mathrm{Sb}}_2{\mathrm{Se}}_{4.5}{\mathrm{Te}}_{0.5}$ in the parent alloys. Figure 6(b) shows the EDS-extracted atomic percentages of constituent elements on various locations of parent alloy, a coated film with the adjusted composition of ${\mathrm{Ge}}_2{\mathrm{Sb}}_2{\mathrm{Se}}_{4.5}{\mathrm{Te}}_{0.5}$, and a picture of the sample in the inset. Here, the uniform film coverage, spatial composition homogeneity, and compositional match from a parent alloy to a coated film are apparent, which are further cross confirmed by homogeneous elemental maps at a micron scale, as shown in Fig. 6(c). Figure 6(d) shows that the GIXRD data collected from the SD ${\mathrm{Ge}}_2{\mathrm{Sb}}_2{\mathrm{Se}}_{4.5}{\mathrm{Te}}_{0.5}$ films clearly indicate no presence of Te crystals in their as-coated form, while the intensity of ${\mathrm{Sb}}_2{\mathrm{Te}}_3$ peaks is observed to be stronger and sharper, indicating a higher volume fraction and uniform size distribution of the crystals as compared to those from ${\mathrm{Ge}}_2{\mathrm{Sb}}_2{\mathrm{Se}}_4{\mathrm{Te}}_1$ films. The IR transmission spectra (not corrected for Fresnel loss) of the SD films are shown in Fig. 6(e) where both as-coated and heat-treated films exhibit broadband transparency from mid-wave to long-wave IR. However, the short-wave IR transparency noticeably decreases upon heat treatment likely due to ${\mathrm{Sb}}_2{\mathrm{Te}}_3$ crystals-induced optical scattering. Preliminary laser writing to access a tendency toward laser-induced crystallization was carried out on the as-coated film, and visible arrays of patterns on film surfaces illustrate the films’ photo sensitivity and phase change, as shown in Fig. 6(f).

Fig. 6

(a) In situ GIXRD data of SD ${\mathrm{Ge}}_2{\mathrm{Sb}}_2{\mathrm{Se}}_4{\mathrm{Te}}_1$ films; (b) EDS-extracted composition of ${\mathrm{Ge}}_2{\mathrm{Sb}}_2{\mathrm{Se}}_{4.5}{\mathrm{Te}}_{0.5}$ parent alloys and SD films; (c) EDS-mapped spatial distributions of constituent elements in SD ${\mathrm{Ge}}_2{\mathrm{Sb}}_2{\mathrm{Se}}_{4.5}{\mathrm{Te}}_{0.5}$ films; (d) in situ GIXRD data of SD ${\mathrm{Ge}}_2{\mathrm{Sb}}_2{\mathrm{Se}}_{4.5}{\mathrm{Te}}_{0.5}$ films; (e) transmission spectra of SD ${\mathrm{Ge}}_2{\mathrm{Sb}}_2{\mathrm{Se}}_{4.5}{\mathrm{Te}}_{0.5}$ films pre- and postheat treatment; and (f) MPL induced modifications on SD ${\mathrm{Ge}}_2{\mathrm{Sb}}_2{\mathrm{Se}}_{4.5}{\mathrm{Te}}_{0.5}$ films.

In summary, the approach which demonstrates SD ${\mathrm{Ge}}_2{\mathrm{Sb}}_2{\mathrm{Se}}_{4.5}{\mathrm{Te}}_{0.5}$ films as a candidate PCM is presented as a route toward conformal PCM media. Such strategies are expected to greatly increase design flexibility providing new opportunities for optical applications such as reconfigurable metalenses and other advanced devices. As shown, the process’ scalability and applicability to diverse composition and substrate material systems is also advantageous. Finally, further insight into the switching behavior of the SD films to understand the limitations in refractive index and electrical conductivity modifications upon either heat treatment or laser exposure will extend integration potential in advanced devices.