PaperlayerView exact revision on arXiv

Source-linked AI summary

Tunable nanophotonics enabled by chalcogenide phase-change materials

Sajjad Abdollahramezani, Omid Hemmatyar, Hossein Taghinejad, Alex Krasnok, Yashar Kiarashinejad, Mohammadreza Zandehshahvar, Andrea Alu, Ali Adibi

arXiv:2001.06335v1physics.opticsphysics.app-ph

TL;DR

Photonics offers high-speed, high-bandwidth information transport but needs greater integrability, reprogrammability, and adjustable functionality. This review surveys chalcogenide PCM-based metasurfaces and photonic integrated circuits, including material behavior, tuning mechanisms, deep-learning design, demonstrated performance, and remaining challenges.

  • Problem

    Photonic platforms provide fast, high-bandwidth information transport but remain limited in integrability and reprogrammability, while high-dimensional nanophotonic design also requires improved optimization and physical insight.

  • Method

    The review synthesizes hybrid chalcogenide PCM nanophotonics for tunable metasurfaces and photonic integrated circuits, covering material properties, switching mechanisms, applications, and deep-learning-based analysis and design.

  • Results

    The reviewed approaches demonstrate tunable optical responses, including a 1.3 µm mid-IR resonance shift with 4-fold contrast and a 350-fold resistance increase between PCM phases.

  • Takeaways & Limitations

    Chalcogenide PCMs provide a platform for reconfigurable metasurfaces and photonic integrated circuits, while reduced-dimensionality learning methods support faster design exploration and optimization.

  • Takeaways & Limitations

    Practical deployment remains constrained by switching energy, integration challenges, and material trade-offs such as slower crystallization when Se content increases in GSST.

Abstract

from arXiv · show

Nanophotonics has garnered intensive attention due to its unique capabilities in molding the flow of light in the subwavelength regime. Metasurfaces (MSs) and photonic integrated circuits (PICs) enable the realization of mass-producible, cost-effective, and highly efficient flat optical components for imaging, sensing, and communications. In order to enable nanophotonics with multi-purpose functionalities, chalcogenide phase-change materials (PCMs) have been introduced as a promising platform for tunable and reconfigurable nanophotonic frameworks. Integration of non-volatile chalcogenide PCMs with unique properties such as drastic optical contrasts, fast switching speeds, and long-term stability grants substantial reconfiguration to the more conventional static nanophotonic platforms. In this review, we discuss state-of-the-art developments as well as emerging trends in tunable MSs and PICs using chalcogenide PCMs. We outline the unique material properties, structural transformation, electro-optic, and thermo-optic effects of well-established classes of chalcogenide PCMs. The emerging deep learning-based approaches for the optimization of reconfigurable MSs and the analysis of light-matter interactions are also discussed. The review is concluded by discussing existing challenges in the realization of adjustable nanophotonics and a perspective on the possible developments in this promising area.

I. INTRODUCTION

Chalcogenide PCMs provide a route toward reconfigurable nanophotonics, addressing the fixed functionality of static metasurfaces and application-specific PICs. Their structural and compositional properties create large optical contrasts, while material selection involves trade-offs among figure of merit, switching speed, stability, and predictive design requirements.

  • I. INTRODUCTION: Static metasurfaces retain fixed functionality after fabrication, while PICs are often application-specific, motivating highly reconfigurable and general-purpose photonic platforms.These limitations are especially relevant to applications requiring real-time tuning or FPGA-like programmability.
  • I. INTRODUCTION: Chalcogenide PCMs support reliable, repeatable switching of optical and electrical attributes, but practical systems require locally controlled, reprogrammable phase conversion.Existing approaches commonly rely on bulky heaters or focused continuous-wave laser illumination and may provide only one-way crystallization.
  • II. CHALCOGENIDE-BASED PCMS: MATERIAL PROPERTIES: GST crystallization can shift from nucleation-dominated to growth-dominated behavior after pre-conditioning, reducing GeTe crystallization time from 1 µsec to 30 nsec.A 100 nsec pre-conditioning laser pulse produces this reported reduction.
  • II. CHALCOGENIDE-BASED PCMS: MATERIAL PROPERTIES: Se doping increases GSST crystallization temperature and amorphous-phase stability but slows crystallization substantially, with tc increasing from ∼20 ns in GST to ∼3 µsec at ∼25% Se.The reported increase corresponds to more than a 100-fold reduction in crystallization speed.
  • II. CHALCOGENIDE-BASED PCMS: MATERIAL PROPERTIES: GST optical contrast is associated with Ge coordination switching between octahedral crystalline bonds and tetrahedral amorphous bonds, accompanied by ∼7% volumetric expansion.The local structural change alters optical-transition matrix elements; the precise composition-dependent contrast mechanism remains incompletely understood.
  • II. CHALCOGENIDE-BASED PCMS: MATERIAL PROPERTIES: PCM selection depends on the application because the best FOM materials differ by spectral range, while Ge2Sb2Te5 can switch faster than Ge2Sb2Se4Te1 and GeTe.Sb2S3, GeTe, and Ge2Sb2Se4Te1 provide the best FOMs in visible, NIR, and telecommunication wavelengths, respectively.

A. Hybrid plasmonic/PCM metasurfaces for global amplitude control

Hybrid plasmonic/PCM metasurfaces use phase transitions to globally tune amplitude-related optical responses across near-, mid-, and visible-to-near-IR regimes. Demonstrations include resonance shifts, modulation contrast, thermal emission, multispectral imaging, and switchable chirality.

  • Global tuning: Global PCM switching tunes plasmonic resonances and amplitude responses through changes in the effective permittivity surrounding metallic meta-atoms.The resonance shift follows the overlap between the PCM-induced permittivity change and the meta-atom electric field.
  • Global tuning: 150 nm blueshift was observed in a near-IR Fano resonance using electrically switched GLS-covered Au resonators.Electrical pulses lasted approximately 10 ms and exceeded 45 V in amplitude.
  • Global tuning: 2.5-fold transmission modulation depth and a 1.3 µm mid-IR resonance shift with 4-fold contrast were demonstrated using GST phase switching.The near-IR and mid-IR responses were obtained from global laser-driven switching of GST films.
  • Broadband and multifunctional control: A 1 µm Fano-resonance tuning range and up to 385 nm spectral shift or 60% modulation depth extended PCM-plasmonic tuning across mid-IR and vis-near-IR bands.The broadband platform used a thin Ge2Sb2Te5 layer over a perforated Au film and supported thermal or current-driven modulation.
  • Broadband and multifunctional control: PCM-plasmonic devices also enabled dynamic thermal emission, multispectral emission within 300 ns, perfect absorption with over 90% absorption, and switchable chirality.Reported examples include approximately 40% crystallization for a 9.55 µm emission peak, a 500 nm reflectance shift, and reversible circular-dichroism sign switching.

B. Hybrid dielectric/PCM metasurfaces for global amplitude control

All-dielectric and structured PCM metasurfaces provide global control of transmission, reflection, absorption, and multipolar resonances while addressing losses associated with plasmonic platforms. Phase conversion supports bistable filtering, absorption switching, anapole control, and reconfigurable display functions.

  • Platform rationale: Dynamic dielectric PCM metasurfaces address plasmonic dissipation and limited scattering efficiency through high-index, CMOS-compatible resonant nanostructures.These platforms use high-contrast PCM nanostructures for adaptive optical functionality.
  • All-dielectric and guided-mode resonances: 10% resonance shifting with reflection contrast up to 5:1 and transmission contrast of 1:3 was achieved in Ge2Sb2Te5 nanogratings under TE illumination.Raster-scanned laser crystallization produced bistable transmission and reflection resonances.
  • All-dielectric and guided-mode resonances: A guided-mode-resonance metasurface switched from a highly transmittive filter to a highly absorptive device with a 7:1 transmission contrast ratio.The transmission resonance is present in amorphous GST, while increased crystalline loss suppresses the field intensity.
  • Multipolar and cavity control: Intermediate GST phase states shifted electric-dipole modes toward anapole modes, enabling broadband switching between bright and dark scattering states.The transition was demonstrated by inducing 50% phase change in 220-nm-thick GST nanodisks.
  • Multipolar and cavity control: PCM metasurfaces further demonstrated tunable EIT responses, polarization-insensitive broadband absorption from 350 to 1500 nm, and full-color cavity display switching.These functions arise from hybridized dipole modes, patterned GST resonances, cavity modes, and phase-switched Fabry–Perot conditions.

C. Hybrid plasmonic/PCM metasurfaces for local amplitude control

Local PCM programming extends metasurface reconfiguration from global switching to pixel- or meta-atom-level control. Focused optical addressing enables local resonance tuning, radiation-pattern switching, and reconfigurable color generation.

  • Local programming: Local tuning is needed for reversible pixel-by-pixel programming of individual metasurface meta-atoms.The reviewed approaches use focused optical beams or localized electrical currents rather than uniform phase conversion.
  • Radiation-pattern control: Switching a GST spacer transformed a directive amorphous-state radiation pattern into omnidirectional crystalline-state scattering.The amorphous state satisfies Kerker’s condition through electric- and magnetic-dipole interference, whereas the crystalline state supports only an electric-dipole resonance.
  • Local resonance tuning: Localized laser spots tuned an individual hybrid meta-atom resonance by up to one FWHM through control of crystallized spot size, depth, and position.The device used Al nanorods covered by a 75-nm Ge3Sb2Te6 layer and sub-microsecond laser pulses.
  • Color generation: A GeTe-based MIM metasurface generated selectable cyan, magenta, and yellow pixels by switching PCM-controlled gap-plasmon absorption.The crystalline state supports resonant absorption of red, green, and blue bands, while phase switching changes the displayed color.

D. Hybrid dielectric/PCM metasurfaces for local amplitude control

Hybrid dielectric/PCM metasurfaces enable local amplitude control through electrically addressed phase switching and multilevel laser-induced crystallization, supporting displays, data storage, and grayscale holography.

  • Electrical addressing: Electrical current pulses locally switch GST pixels between amorphous and crystalline states in ITO/GST/ITO devices.The approach supports reflective and semi-transparent displays on rigid and flexible substrates.
  • Electrical addressing: 350 times higher resistance separates amorphous and crystalline phases in 300 nm×300 nm ITO/GST/ITO crossbar devices at a 2.2 V threshold.A 100 ns, 5 V RESET pulse returns the device to its high-resistance amorphous state.
  • Multilevel crystallization: Ultrashort laser pulses produce multilevel GST crystallization, mapping pulse number and energy to optical reflectivity for grayscale optical storage.The experiment writes grayscale images into a 50-nm-thick GST layer using scanned ultrafast laser pulses.
  • Multilevel crystallization: Eight crystallization levels are represented in a 34 µm-diameter spiral pattern, with pulse counts increasing from 5 to 85.Each level is characterized by its average reflectivity change and standard deviation.
  • Optical applications: Intermediate GST states enable rewritable full-color computer-generated holograms through color-selective diffractive optical components.Recorded binary patterns and thickness-tailored substrates produce tunable color-selective diffraction.

IV. ACTIVE PHASE CONTROL WITH TUNABLE PHASE-CHANGE METASURFACES

Dynamic phase-gradient metasurfaces extend PCM-based amplitude modulation toward on-demand optical components that reshape the incident wavefront for tunable beam control and focusing.

  • Motivation: Dynamic phase-gradient metasurfaces are needed to tailor incident-light phasefronts for real-world, on-demand applications.Reported functionalities include reconfigurable beam steering and tunable focusing.

A. Hybrid plasmonic/PCM metasurfaces for global phase control

Hybrid plasmonic/PCM metasurfaces use phase switching and geometric-phase designs to reconfigure global phase profiles for beam steering, focusing, and spin-orbit optical functions.

  • Design rationale: Resonant dispersive meta-atoms limit bandwidth, motivating Pancharatnam-Berry metasurfaces for dispersionless phase control and amplitude-phase decoupling.Spatially varying identical meta-atom orientations also alleviate fabrication tolerance.
  • Dynamic phasefront switching: GST switching changes a cylindrical metalens focal line from z = 0.5 mm in the amorphous state to z = 1 mm in the crystalline state.The metasurface uses two meta-atom types that redirect light in opposite directions across states.
  • Spin-orbit control: GST-state switching activates or deactivates geometric-phase functions including spin Hall effect, vortex-beam generation, and holography.In the amorphous state, the polarization conversion ratio exceeds 80% across 8.5–10.5 µm; crystalline GST reduces it to 10%.
  • Wideband control: Wideband wavefront switching uses rotating U-shaped Ge2Sb2Te5 nanoantennas with complementary cross-polarized transmittance responses between material states.The design multiplexes two nanoantenna types with different orientations under circularly polarized illumination.
  • Beam steering: A GST-based plasmonic metasurface enables anomalous reflection at a predesigned angle in the amorphous state and a different response after crystallization.Pixelation allows different regions to be designed for distinct steering angles.
  • Local phase manipulation: Plasmonic GST slit metalenses demonstrate local phasefront engineering with on-axis and off-axis focusing from binarized phase distributions.The figure compares the planar metalens structure, field distribution, and measured focusing patterns with simulations.

B. Hybrid dielectric/PCM metasurfaces for global phase control

Hybrid dielectric/PCM metasurfaces provide bistable global phase control by redesigning dielectric meta-atoms for distinct phase profiles before and after crystallization.

  • Wideband wavefront switching: High-index U-shaped Ge2Sb2Te5 nanoantennas provide state-dependent cross-polarized transmittance for wideband near-IR wavefront switching.Two antenna types are optimized with opposite amorphous/crystalline transmittance behavior under circularly polarized illumination.
  • Varifocal metalens design: Full-wave simulations sweep “I”, “H”, and “+” Huygens meta-atom geometries to identify the 16 optimal phase combinations.The resulting library supports experimentally demonstrated bistable focusing.
  • Varifocal metalens design: A GSST Huygens-meta-atom library discretizes two continuous [0, 2π] phase profiles into four levels, requiring 16 meta-atoms for all state-transition phase combinations.The profiles target bifocal focusing with NA = 0.45 in the amorphous state and NA = 0.35 in the crystalline state.
  • Varifocal metalens performance: The bistable varifocal metalens achieves diffraction-limited performance above 20% in both states with a 29.5 dB switching contrast ratio near λ = 5.2 µm.It resolves USAF 6.2 in the amorphous state and USAF 5.6 in the crystalline state.

C. Hybrid plasmonic/PCM metasurfaces for local phase control

Hybrid plasmonic/PCM metasurfaces use GST phase changes to locally tune optical phase, amplitude, and polarization. Demonstrated devices provide broadband modulation and reconfigurable focusing, while the metalens example is limited by access to two PCM states.

  • C. Hybrid plasmonic/PCM metasurfaces for local phase control: A GST-filled Au-nan slit metalens spectrally tunes its Fabry–Pérot resonance through crystallization fraction, controlling phase over 0.56π at 1.55 µm.Experiments accessed only amorphous and fully crystalline states, restricting feasible optical phase patterns to two.
  • C. Hybrid plasmonic/PCM metasurfaces for local phase control: Addressable MIM structures exploit multistate GST transitions to tailor reflected amplitude, phase, and polarization with a broadband phase shift up to 315° and amplitude modulation up to 60%.Electro-thermal simulations support these near-IR responses.
  • C. Hybrid plasmonic/PCM metasurfaces for local phase control: The resulting varifocal metalens changes its focal point from ∼5λ to ∼20λ while preserving a resolution comparable to the Airy disk.Each meta-atom is addressed through its crystallization fraction.

D. Hybrid dielectric/PCM metasurfaces for local phase control

Hybrid dielectric/PCM metasurfaces locally reconfigure transmitted or reflected phase profiles through PCM-state control of resonant meta-atoms. Reported platforms combine multilevel optical switching, high phase shifts, and adaptive beam-control functionalities.

  • D. Hybrid dielectric/PCM metasurfaces for local phase control: Optical writing with 85-fs, 730-nm femtosecond pulse trains enables multilevel GST switching using 0.59-µm diffraction-limited patterning.A spatial light modulator writes patterns, while an imaging system reads the exposed zones.
  • D. Hybrid dielectric/PCM metasurfaces for local phase control: All-dielectric GST nanobars locally control incident-light phase profiles by combining amorphous, partial, and fully crystalline meta-atom states in dynamic gradient metasurfaces.The structures vary the period of constituent supercells to realize the phase gradients.
  • D. Hybrid dielectric/PCM metasurfaces for local phase control: GST nanoposts enclosed by Si nanorings provide a 325° phase shift and transmittance above 0.6 at 1340 nm.The design uses strong electric- and magnetic-dipole interference to enhance the field inside the PCM core.
  • D. Hybrid dielectric/PCM metasurfaces for local phase control: Ge2Sb2Se4Te1 nanoribbons achieve ∼270° phase agility and performance above 45% around 1.55 µm by operating off resonance to suppress dissipative loss.Locally transforming each meta-atom with a focused optical beam enables adaptive beam deflection.

V. PHASE-CHANGE PHOTONIC INTEGRATED CIRCUITS

PCM-based photonic integrated circuits are presented as a route toward reconfigurable, high-speed photonic systems. Their potential is linked to the small footprints, low power consumption, and high bandwidth of programmable silicon photonic devices.

  • V. PHASE-CHANGE PHOTONIC INTEGRATED CIRCUITS: PCM-based PICs are envisioned for next-generation high-speed CMOS-compatible computation, communications, computers, and data-storage systems.The review situates these circuits within recent progress in large-scale programmable silicon photonic devices.

A. Integrated phase-change photonic switches and modulators

Integrated phase-change switches and modulators use thermal, optical, electrical, and mixed-mode stimuli to achieve reversible, broadband, and tunable photonic operation. Device architectures improve contrast, loss, switching speed, and integration, while optical pumping remains difficult to scale.

  • Switching mechanisms: Phase transitions in integrated photonic switches are driven thermally, photothermally, or electrothermally, with the mechanism determining switching performance and applications.The review organizes integrated switches and modulators according to how the PCM phase is converted.
  • Broadband architectures: Directional couplers provide broadband switching, including more than 30 nm bandwidth with -10 dB cross talk and approximately 1 dB loss in the telecommunication band.The reported 1×2 and 2×2 switches used asymmetric directional couplers incorporating Ge2Sb2Te5.
  • Broadband architectures: GSST directional-coupler switches achieved 0.01 to 0.4 dB insertion loss in the C-band and cross talk over 15 dB.GSST’s lower extinction coefficient enabled low-loss broadband 1×2 and 2×2 switching.
  • Reversible modulation: Resonant and hybrid waveguide designs exploit PCM refractive-index or absorption contrast for reversible modulation, including ER up to 33 dB and a 42 dB contrast ratio with GSST.The hybrid modulator used 440 ps pulses and multiple intermediate states, while the GSST resonator combined less than 0.5 dB insertion loss with high contrast.
  • Interaction enhancement: Photonic-crystal integration enhanced PCM-light interaction, producing higher switching contrast with approximately 14% lower power consumption than bare-waveguide switches.High-quality-factor cavities increased absorption in the crystalline state and enabled resonance shifts from GST’s refractive-index contrast.
  • Scalable integration: Patterned and encapsulated GST nanodisks reduced deformation-related loss, enabling reversible switching with less than 1 dB insertion loss, while electrical and mixed-mode approaches improved scalability.Electrical switching used integrated heaters, and mixed-mode devices combined optical and electrical operations or readout.
  • Scalable integration: Free-space optical pumping is difficult to scale because beam alignment and focusing are slow and diffraction-limited, while routing pump light and switching large PCM areas are cumbersome.These constraints motivate electro-optic and mixed-mode approaches for reversible switching of larger PCM inclusions.

B. Integrated phase-change photonic binary and multilevel memories

Integrated chalcogenide PCM memories use optical pulses and evanescent waveguide coupling to encode binary and multiple nonvolatile transmission states. Architectures extend this control to wavelength-selective access and improved multilevel programming efficiency.

  • GST’s dramatic optical contrast, sub-nanosecond switching, and reproducibility over 10^12 cycles support integrated nonvolatile photonic memory.These properties make GST suitable for optical memory operation across visible and near-infrared wavelengths.
  • An eight-level GST memory achieved approximately 1 GHz switching, 13.4 pJ switching energy, and arbitrary switching among intermediate states.The device controls the crystalline fraction of an evanescently coupled GST film.
  • WDM access used three ring resonators with 1 nm-separated resonances around 1550 nm and 1 × 1 µm^2 GST cells.Different wavelengths address individual memory elements through wavelength-selective resonator filtering.
  • PWM switching controls GST crystallization and amorphization through pulse width, enabling distinct memory levels with fixed pulse amplitude.Changing pulse width controls the resulting GST phase and transmission state.
  • A dual-pulse programming scheme achieved 34 nonvolatile levels, corresponding to more than 5 bits, using a fixed 250 ns double-step erase pulse.The approach uses stepwise crystallization for erasing and a rectangular writing pulse for amorphization.
  • Evanescent GST coupling avoids placing highly absorptive crystalline GST directly in the waveguide, reducing attenuation and associated power consumption.The GST cell instead interacts with the guided field from above the waveguide.

C. Integrated phase-change photonic arithmetic processors

Integrated PCM photonics combines storage and computation in the same optical hardware. Demonstrations implement abacus-like arithmetic, matrix-vector multiplication, and other logic operations using programmable GST states.

  • A GST waveguide-crossing array performs addition, multiplication, subtraction, and division through progressive crystallization of individually addressable PCM cells.The all-photonic abacus demonstrates the base-ten operation 6 + 6 = 12.
  • The shared storage-and-processing architecture places computational operations and data storage in the same location rather than continuously transferring data between CPU and external memory.This is presented as an all-photonic alternative to the von Neumann arrangement.
  • Integrated phase-change photonics experimentally implemented multiplication of a (1 × 2) matrix by a (2 × 1) vector.The device used a Si3N4/SiO2 waveguide with a 10 nm GST layer and a 10 nm ITO capping layer.

D. Integrated phase-change photonic synapses and neuromorphic processors

Chalcogenide PCM photonics implements tunable synaptic weights and neuron-like switching in integrated waveguides and resonators. These elements support all-photonic spiking networks with supervised and unsupervised learning for pattern recognition.

  • A tapered SiN-waveguide synapse with evanescently coupled GST islands represents synaptic weight through optical transmission.Low-power probe pulses and higher-power pump pulses use different wavelengths to read and modify the synaptic state.
  • The synaptic transmission change varies exponentially and monotonically with the number of optical pulses, resembling the STDP learning rule.The relationship is fitted as ∆T = A e^(r×N) and compared with ∆w = A e^(-∆t/τ).
  • An all-photonic spiking neural network integrated four neurons, sixty synapses, and 140 optical elements.The platform supported both supervised and unsupervised learning.
  • PCM neuron switching changes probe-light coupling from a resonator to an output waveguide, producing integrate-and-fire-like output spikes.Crystalline PCM keeps the probe resonant and suppresses output, whereas amorphization makes the probe off-resonance and transmits it.
  • The neurosynaptic platform performed pattern recognition using both supervised and unsupervised learning configurations.Supervised learning keeps the spiking feedback open, while unsupervised learning closes it.

VI. EMERGENCE OF DEEP LEARNING IN ANALYSIS, DESIGN, AND OPTIMIZATION OF PHASE-CHANGE NANOPHOTONICS

Deep learning addresses the complexity and limited physical insight of conventional phase-change metasurface design by reducing response and design dimensionality. The resulting tools accelerate optimization while supporting knowledge discovery about parameter roles and device performance.

  • Motivation and challenges: Conventional metasurface optimization becomes computationally demanding in high-dimensional design spaces and may not reveal internal wave-propagation dynamics.Iterative methods are also not guaranteed to converge to a global optimum.
  • Deep-learning design models: A DNN forward-and-inverse model relates metasurface design parameters to the real and imaginary parts of the complex optical response.Separating response components smooths abrupt phase or amplitude changes during training and accelerates design relative to conventional optimization.
  • Dimensionality reduction: An autoencoder reduces optical-response dimensionality, while a feed-forward network maps design parameters into the reduced response space.The pseudo-encoder supports inverse design and extracts reduced design parameters.
  • Dimensionality reduction: The dimensionality-reduction method is not problem-specific after training for a metasurface class and can incorporate prior information such as fabrication-insensitivity preferences.It provides a trade-off between accepted error and simulation complexity or time.
  • Knowledge discovery: Neural-network weights were used to assess the roles of design parameters in phase-change metasurface responses and inform improved designs.The approach provided knowledge about light-matter interaction alongside optimization.
  • Demonstration: A 10×200 metasurface problem was reduced to 5×10 with less than 0.01% error using 5000 random-structure simulations.The example used a GST/SiO2/Au reconfigurable metasurface and produced a reflective band-stop filter.

VII. PERSPECTIVE AND OUTLOOK

The outlook identifies chalcogenide PCMs as a route toward increasingly tunable and reconfigurable metasurfaces and photonic integrated circuits, while emphasizing energy and speed constraints in integrated devices. It also highlights metagratings, BICs, exceptional points, and deep-learning-assisted design as promising directions for expanding functionality.

  • VII. PERSPECTIVE AND OUTLOOK: Chalcogenide PCMs are expected to remain important for actively tunable metasurfaces spanning microwave through optical applications.The review connects PCM tunability with lenses, imaging, optical information processing, nonlinear optics, analog computation, and cloaking.
  • VII. PERSPECTIVE AND OUTLOOK: Reversible PCMs could enable erasable and rewritable metagrating patterns for efficient rerouting of incident waves without conventional subwavelength structures.Metagratings can redirect incident waves with 100% efficiency, while their wavelength-scale periodicity alleviates fabrication challenges.
  • VII. PERSPECTIVE AND OUTLOOK: PCM-based dielectric metasurfaces could make bound states in the continuum tunable and reconfigurable, extending their applications beyond predominantly static devices.BICs provide very large Q-factors and potentially infinite lifetimes, with demonstrated relevance to lasers, nonlinear optics, and sensors.
  • VII. PERSPECTIVE AND OUTLOOK: Exceptional points in non-Hermitian photonic structures are identified as another unusual scattering phenomenon that could benefit from PCM integration.Exceptional points occur when pairs of eigenstates and corresponding eigenvalues coalesce; in parity-time-symmetric structures, they can arise at real frequencies.
  • VII. PERSPECTIVE AND OUTLOOK: Deep-learning-based optimization is presented as a way to address the complexity created by large numbers of metasurface design parameters.Increasing design freedom expands the optimization landscape but also makes conventional electromagnetic design more difficult.
  • VII. PERSPECTIVE AND OUTLOOK: Integrated PCM photonics remains constrained by switching energy and crystallization time, motivating architectures that enhance electromagnetic fields around PCM elements.Reported switching energies range from 50 pJ to a few nJ, whereas plasmonic nanogaps, photonic crystal cavities, and crossing waveguides are proposed to reduce transformation energy.
Loading 2001.06335v1…