Tensor-driven geometric phase in nonlinear AlGaAs metasurfaces

Abstract: Dielectric metasurfaces provide a unique platform for efficient harmonic generation and optical wavefront manipulation at the nanoscale. While several approaches are available for performing wavefront shaping, the one exploiting geometric phase streamlines significantly the design and fabrication process. It has been recently shown that, in III-V semiconductor alloys, the rotation of the crystal axes affects the phase and amplitude of second-harmonic generation (SHG) induced by circularly polarized light [1]. Based on this notion, we fabricated and characterized two aluminum gallium arsenide metasurfaces displaying the versatility of the geometric phase design approach through nonlinear beam steering and structured-light generation on the harmonic field.

• The paper demonstrates nonlinear wavefront shaping in AlGaAs metasurfaces, using a tensor-driven geometric phase that varies with crystal orientation, achieving spin-dependent beam steering and vortex generation.
• The authors fabricate metasurfaces with rotated elliptical pillars, validating that the tensor-induced phase can control the second harmonic (SH) phase and amplitude without sacrificing conversion efficiency.
• The study showcases two demonstrators: a beam-steering device deflecting right- and left-circularly polarized SH beams by ±12°, and a structured-light device generating vortices with topological charge ℓ = 1 and encoding a helical phase.

Overview

This paper reports the experimental demonstration of nonlinear wavefront shaping in AlGaAs metasurfaces using a geometric phase mechanism that is unique to materials with a rotation-sensitive second-order susceptibility tensor. Building on a prior theoretical observation that rotating the crystal axes of (100) AlGaAs modifies both the phase and amplitude of circularly polarized second-harmonic generation (SHG) (2601.18246), the authors design, fabricate, and characterize two metasurfaces — one for beam steering and one for structured-light generation — validating that the tensor-driven phase can be exploited as a practical wavefront-engineering tool without sacrificing conversion efficiency.

Physical mechanism

The key distinction from conventional Pancharatnam–Berry (PB) metasurfaces lies in the behavior of the $\chi^{(2)}$ tensor under rotation. In most previously studied nonlinear PB platforms, the nonlinear response is rotationally invariant, so rotating a meta-atom produces only the canonical geometric phase terms. For zincblende crystals such as AlGaAs, however, the nonlinear susceptibility transforms under rotation about the $z$-axis as $\sin(2\phi)$, so an in-plane rotation by an angle $\beta$ generates an effective tensor

$$\chi^{(2)\prime} = \cos(2\beta)\chi^{(2)}_1 + \sin(2\beta)\chi^{(2)}_2,$$

where $\chi^{(2)}_2$ is $\chi^{(2)}_1$ rotated by $\pi/4$. The resulting SH field is a coherent superposition of contributions from the two tensor orientations, with complex coefficients $a$ and $b$ set by the resonant response of the nanostructure.

The authors derive the total SH phase as the sum of two contributions:

$z$0

where $z$1 is the tensor-induced term and the remaining terms follow from total angular momentum (TAM) conservation. For rotation-invariant media this reduces to the familiar $z$2 or $z$3 laws; for AlGaAs, an additional $z$4-dependent phase appears for both LCP and RCP SH components. This constitutes a phase function not previously reported in metasurface design, arising purely from the interplay between resonator orientation and crystallographic anisotropy — notably, while the material is linearly isotropic, the effect manifests exclusively in the SH field.

Design and numerical validation

The resonator geometry is a half-cylinder-like elliptical pillar (height 400 nm, semi-axes 325 nm and 375 nm) chosen for its broken in-plane symmetry, which permits vertical SH emission. With $z$5 rotational symmetry, the TAM selection rule $z$6 admits all integer $z$7, including the $z$8 components radiating along the vertical direction. Fully vectorial FEM simulations (COMSOL) confirm that single-resonator SH intensity varies with orientation angle due to the changing alignment of internal fields relative to the crystal axes, while the SH phase follows the predicted analytic model closely, indicating the phase is governed by geometry rather than resonance effects.

For the periodic metasurface (period 900 nm), a broad reflectance resonance centered at 1555 nm enhances pump confinement. The simulated SH phase follows the expected trend overlaid with a periodic modulation attributable to resonance shifts as a function of unit-cell orientation. The authors note this modulation could be mitigated by more rotation-tolerant designs or pulsed pumps with finite bandwidth — an acknowledged limitation of the current implementation. A lookup table built from these simulations selects four elements at rotations of $z$9, $\sin(2\phi)$0, $\sin(2\phi)$1, and $\sin(2\phi)$2 for the two demonstrators.

Fabrication and experimental results

Devices were fabricated on an MBE-grown stack (GaAs substrate, Al$\sin(2\phi)$3Ga$\sin(2\phi)$4As buffer, 400 nm Al$\sin(2\phi)$5Ga$\sin(2\phi)$6As layer) via EBL with HSQ resist and ICP-RIE etching (SiCl$\sin(2\phi)$7/Ar), followed by selective oxidation to place the metasurface on a low-index AlOx cladding. Each 100 μm-diameter device was pumped at 1550 nm with 200 fs pulses from an OPA in reflection geometry.

Beam steering: A supercell of four rotated elements implements a linear phase ramp $\sin(2\phi)$8, deflecting RCP SH to $\sin(2\phi)$9 and LCP SH to $\beta$0. Polarization-resolved Fourier-plane measurements show RCP SH predominantly in the $\beta$1 order and LCP SH in the $\beta$2 order, in agreement with vectorial simulations. Because both chiralities are generated simultaneously, the device intrinsically performs spin-dependent beam splitting at the harmonic frequency.

Structured light: A four-quadrant design with $\beta$3 rotations encodes a $\beta$4 ($\beta$5) helical phase for RCP (LCP) SH around the quadrant vertex. Interferometry in a generalized Mach–Zehnder configuration yields a fork-like pattern confirming a vortex with topological charge $\beta$6, and Fourier analysis of the interferogram recovers a helical phase profile matching angular-spectrum simulations. The far field shows four bright spots rather than a uniform ring, attributed to phase-mask discretization (angular diffraction effects) and reproduced by simulation.

Efficiency and terminology considerations

A notable quantitative result is that the SH yield of both wavefront-shaping devices is comparable to that of a uniform reference metasurface composed of identically tilted meta-atoms, measured via lock-in detection with a Si photodiode. The beam-steering device shows slightly lower collected power, attributed to radiation into higher diffraction orders outside the collection NA rather than to reduced intrinsic efficiency. Since the approach does not rely on tuning optical resonances, it is inherently broadband and compatible with both local and nonlocal resonant designs.

The authors also raise a terminological point: unlike canonical Berry or Aharonov–Bohm phases, which are true holonomies, the phase demonstrated here lacks a clear holonomic interpretation. They suggest adopting a broader definition in which "geometric" encompasses any phase control achieved through structural geometry rather than wavelength- or material-dependent mechanisms. This candid acknowledgment highlights an unresolved conceptual question in the nonlinear-metasurface literature.

Limitations and open questions

Several constraints qualify the results. The resonant modulation of SH amplitude and phase with unit-cell orientation introduces design-dependent non-idealities that are only partially mitigated in the present devices. The structured-light demonstrator exhibits discretization artifacts in the far field, inherent to the four-level phase encoding. The absence of a holonomic description of the observed phase remains an open theoretical question, as does the extension of the framework to other tensor-symmetry classes beyond zincblende, where the specific phase dependence would differ according to the rotation properties of $\beta$7.

Conclusion

The paper establishes a practical design principle combining the anisotropic $\beta$8 tensor of AlGaAs with rotated dielectric resonators, yielding a purely nonlinear geometric phase that enables spin-dependent beam steering and vortex-beam generation at the second harmonic with efficiency comparable to uniform metasurfaces. The framework streamlines fabrication through a minimal element set and extends naturally to other zincblende materials, broadening the toolkit for nonlinear flat optics.