Gallium Phosphide Optical Metasurfaces for Visible Light Applications

Optical metasurfaces promise compact, lightweight, and cost-effective alternatives to refractive optics for visible-light applications by controlling amplitude, phase, and polarization at subwavelength scales. However, material choices remain limited: plasmonic metals suffer high losses; silicon has high index but significant visible absorption; titanium dioxide is low-loss but lower index, requiring high-aspect-ratio features. The study investigates gallium phosphide (GaP), which combines high refractive index (>3.1 across visible, ~3.6 at 500 nm) with transparency above ~540 nm due to its 2.24 eV indirect bandgap. The research question is whether sputtered amorphous GaP thin films on transparent substrates can enable efficient visible metasurfaces, overcoming historical challenges in growing/depositing GaP on glass or sapphire and nanopatterning GaP films. The purpose is to design, fabricate, and characterize GaP-based metasurfaces (two blazed gratings and a metalens) to assess feasibility and performance in the visible range.

Prior work established visible metasurfaces using titanium dioxide and plasmonic gold structures, with plasmonic losses limiting efficiency for imaging. Silicon and other dielectrics have demonstrated NIR/visible metasurfaces, but silicon absorption and TiO2’s lower index impose trade-offs. GaP has been used in LEDs and in photonic crystal nanocavities with high Q, and GaP nanoantennas demonstrated low-loss visible responses. Simulations proposed GaP-based metalenses and Huygens metasurfaces. Epitaxial GaP growth (MBE/MOCVD) is restricted to lattice-matched substrates (e.g., Si), with recent epitaxial lift-off enabling GaP on glass but with process complexity and limited area. RF sputtering has been reported for amorphous GaP thin films with promising optical properties, suggesting a scalable path to high-index, low-loss films suited for metasurfaces.

Design: Three metasurface devices targeting visible wavelengths were designed. (1) An asymmetric transmissive blazed grating implemented with infinitely long GaP nanobeams in a subwavelength pitch p=380 nm, optimized for λ=520 nm with up to three phase levels per period; film thickness 110 nm; nanobeam widths 25 nm and 75 nm with 55 nm edge-to-edge spacing. (2) A Pancharatnam–Berry Optical Element (PBOE) blazed grating using unit cells (190 nm × 190 nm) of two unequal nanobeam widths; geometric phase realized via discrete rotations; p=380 nm allowed two orientations per period to favor a single diffraction order by breaking mirror symmetry. (3) A metalens using single GaP nanobeams as half-waveplates; geometric phase φ(x,y)=2θ(x,y) for RCP incidence, targeting high polarization conversion to LCP. Film deposition: GaP films were deposited via RF sputtering (Kurt J. Lesker Lab18). Target: 3-inch GaP (99.9%) bonded to copper. Conditions: 100 W RF power; 7.5 mTorr chamber pressure; deposition rate ~0.5 Å/s. Substrates: silica and sapphire, RCA cleaned. Deposition temperatures: 450–550 °C. Post-deposition rapid thermal anneal (RTA): 750 °C for 5 min (substrates must tolerate high temperature). Ellipsometry measured n≈3.8→3.0 for λ=450–800 nm; extinction coefficient k≈0 for λ>560 nm. AFM showed RMS roughness <1 nm; SEM cross-sections confirmed film morphology. Patterning and etch: Wafers were spin-coated with HSQ e-beam resist and a conductive aquaSAVE layer to mitigate charging. E-beam lithography: Vistec VB300, 100 kV, 2 pA; exposed areas 0.25–1 mm². Development: salty developer (1 wt% NaOH + 4 wt% NaCl, 4 min) to achieve sub-10 nm resolution. Pattern transfer: ICP-RIE (per Schneider et al. 2018) with GaP etch rate ~25 nm/min. Residual HSQ (5–10 nm) was left in place to avoid damage. SEM verified pattern fidelity; minimum features down to ~20 nm achieved for dense PBOE patterns. Simulation and measurement: Finite element simulations (COMSOL 5.4) computed unit-cell responses (2D or 3D per symmetry). Asymmetric grating geometry optimized for relatively flat high efficiency T1 diffraction over incident angles −20° to +20° for TM polarization. Diffraction efficiency measured using a custom goniometer; incident polarization and angle scanned. For metalens, focal spot characterized at λ=450 nm; spot FWHM measured and compared to diffraction limit.

• Material properties: RF-sputtered amorphous GaP films on glass and sapphire exhibited high index and low loss suitable for visible metasurfaces: n≈3.8 at 450 nm decreasing to ≈3.0 at 800 nm; k≈0 for λ>560 nm; surface roughness <1 nm RMS.
• Asymmetric blazed grating (TM mode at 520 nm): Simulations predicted 40–50% diffraction efficiency into first transmitted order (T1) over −20° to +20° incidence. Measurements showed lower efficiency than the ideal model; including a ~30 nm residual GaP layer in gaps (attributed to microloading-limited etch clearing) reconciled simulation with experiment. Despite this, performance improved over prior silicon implementations (previously ~30% transmission mode flat diffraction), due to GaP’s lower visible absorption. Device size: 1 mm × 1 mm. Measurement noted a −15° cutoff due to index mismatch between sapphire substrate (n≈1.8) and index-matching oil (n≈1.56).
• PBOE blazed grating: Achieved polarization-sensitive diffraction with high polarization extinction ratio (PER) between orthogonal polarizations ranging from ~60:1 to ~15:1, while maintaining up to ~30% diffraction efficiency into T1 for TM polarization at 520 nm. Patterns demonstrated dense features down to ~20 nm.
• Metalens: At λ=450 nm, measured focal spot FWHM ≈ 1.1 µm, matching the diffraction-limited FWHM (≈1.1 µm), indicating near-diffraction-limited focusing and effective polarization conversion.
• Process scalability: Although demonstrated with e-beam over 0.25–1 mm², authors indicate compatibility with higher-throughput lithographies (immersion or nanoimprint) for scaling.

The study demonstrates that sputtered amorphous GaP thin films can serve as a practical high-index, low-loss platform for visible metasurfaces on transparent substrates (glass, sapphire), addressing material limitations of silicon (absorption) and TiO2 (lower index and high aspect ratio requirements). The asymmetric grating’s broad-angle, moderate-to-high T1 efficiency and improved performance over silicon validate GaP’s optical advantage. The PBOE grating shows that geometric-phase control is feasible with dense GaP nanostructures, enabling high polarization selectivity without excessive loss. The metalens achieving diffraction-limited focusing at 450 nm confirms that GaP nanobeams can provide the requisite half-wave retardation and efficient polarization conversion in the blue-visible regime. Remaining discrepancies between ideal simulations and measurements (e.g., residual GaP in gaps due to microloading, index-matching constraints) highlight fabrication and testing considerations rather than fundamental material limitations. Overall, the results support GaP as a viable alternative material for high-performance visible metasurfaces, with potential for further gains through process optimization.

This work introduces a scalable RF-sputtered amorphous GaP platform for visible-light metasurfaces on transparent substrates and experimentally validates three device classes: an asymmetric blazed grating with broad-angle efficiency, a high-PER PBOE blazed grating with up to ~30% T1 efficiency, and a diffraction-limited metalens at 450 nm. The films exhibit high refractive index and low loss across the visible, enabling compact designs versus lower-index dielectrics and outperforming silicon in the visible due to reduced absorption. Future research should focus on: (i) optimizing deposition and RTA to further reduce loss and fine-tune refractive index; (ii) improving etch uniformity/selectivity to eliminate residual layers and mitigate microloading; (iii) scaling to larger areas via immersion or nanoimprint lithography; (iv) extending device designs to broader bandwidth, higher efficiency, and additional functionalities; and (v) refining measurement setups (e.g., index matching) for accurate angular performance characterization.

• Etch microloading led to incomplete GaP removal in narrow gaps, necessitating modeling with a ~30 nm residual layer to match measured efficiencies; this reduced device performance versus ideal designs.
• Measurement limitations included an angular cutoff (~−15°) caused by index mismatch between sapphire (n≈1.8) and index-matching oil (n≈1.56).
• High-temperature RTA (750 °C) restricts substrate choices to those compatible with such thermal budgets.
• Fabrication used e-beam lithography over small areas (0.25–1 mm²); broader scalability requires alternate lithography.
• High spatial frequency gratings allowed only limited phase levels/orientations per period, constraining ideal blaze approximations.
• Achieving dense ~20 nm features increases patterning and etch challenges; residual HSQ (5–10 nm) was left in place, which could influence optics marginally though deemed low impact.