Radially Gradient-Tilted Meta-Atoms

• Radially gradient-tilted meta-atoms are metasurface elements with tilt angles that vary radially to extend the angular operating range for wide-field imaging.
• They employ analytical tilt laws and inverse-design techniques to optimize local phase matching and improve coupling efficiency under oblique incidence.
• These designs balance throughput and precision by mitigating aberrations and enabling a monolithic lens architecture with an expanded field-of-view.

Radially gradient-tilted meta-atoms are metasurface building blocks whose tilt angle is varied as a function of radial position across an aperture in order to extend the angular operating range of a metalens while preserving a monolithic, single-layer architecture. In the reported wide-field-of-view metalens framework, the tilt angle is introduced as an independent design variable in addition to the usual phase-matching requirement, so that each meta-atom is specified not only by its in-plane geometry but also by a tilt angle $\alpha$ relative to the optical axis (Zhang et al., 22 Sep 2025). In parallel, recent adjoint-based inverse-design work has formulated metasurface optimization in terms of per-meta-atom geometric parameters under general affine transformations—rotation, anisotropic scaling, and translation—providing an analytical route to optimize spatially varying orientations and related geometric degrees of freedom with one forward and one adjoint solve per iteration (Mottola et al., 16 Mar 2026). Together, these developments situate radially gradient-tilted meta-atoms at the intersection of wide-angle flat-optics design and geometry-aware metasurface inverse design.

1. Concept and distinguishing features

The defining feature of a radially gradient-tilted metasurface is that the tilt angle varies with radial coordinate rather than remaining uniform across the aperture. In the metalens framework reported in "Monolithic Expandable-FOV Metalens Enabled by Radially Gradient-Tilted Meta-Atoms" (Zhang et al., 22 Sep 2025), this radial tilt map supplements the conventional hyperbolic focusing phase profile. The result is a design in which the lens remains monolithic—one metasurface and one device—but the local meta-atom orientation is spatially tailored to better match local incidence conditions.

This differs from several established wide-field-of-view strategies. Multilayer metalenses increase degrees of freedom but at the cost of more fabrication complexity and less flatness. Quadratic-phase metalenses compensate off-axis illumination only in a paraxial sense and suffer from spherical aberration and background noise under strongly non-paraxial incidence. Uniformly tilted meta-atoms improve performance for one particular large incident angle, but typically require stitching multiple sub-lenses for different angles, which introduces registration errors and breaks the monolithic form factor (Zhang et al., 22 Sep 2025).

A common misconception is that radially gradient-tilted meta-atoms are simply another form of phase-only engineering. The cited design does not make that reduction: it explicitly treats physical tilt as a new degree of freedom. The phase is still assigned through propagation phase, but the tilt angle adds an additional mechanism for improving wide-angle response and local phase matching under oblique illumination (Zhang et al., 22 Sep 2025).

2. Physical basis and metalens phase prescription

The physical motivation for tilted meta-atoms is that upright meta-atoms are described as fundamentally disadvantaged for wide-angle imaging because their optical response and coupling efficiency degrade as the incident angle becomes large. A physically tilted meta-atom can interact more favorably with obliquely incident light, supplying an extra phase contribution and better local phase matching (Zhang et al., 22 Sep 2025). This is the basis for the reported field-of-view expansion.

The metalens uses elliptical TiO$_2$ nanopillars on silica with wavelength $\lambda = 633$ nm, height $H = 0.5~\mu\text{m}$, period $P = 0.4~\mu\text{m}$, and radius sampled from $0.02~\mu\text{m}$ to $0.2~\mu\text{m}$ (Zhang et al., 22 Sep 2025). The phase is produced via propagation phase rather than geometric phase, which makes the design polarization-insensitive. A precomputed library maps nanopillar radius to phase and transmittance (Zhang et al., 22 Sep 2025).

For normal incidence, the stated phase profile is the hyperbolic focusing phase

$$\Phi(x,y)=\frac{2\pi}{\lambda}\left(\sqrt{x^2+y^2+f^2}-f\right).$$

For oblique incidence, the phase is written as

$$\Phi(x,y)=\frac{2\pi}{\lambda}\left[\sqrt{(x-x_0)^2+(y-y_0)^2+f^2}-\sqrt{x^2+y_0^2+f^2}+(x\sin\theta_x+y\sin\theta_y)\right],$$

where $x_0,y_0$ specify the focal point in the image plane and $_2$0 are the incidence-angle components (Zhang et al., 22 Sep 2025).

Within this framework, the radially varying tilt is introduced because inner zones see smaller effective obliquity, whereas outer zones can be tilted more to compensate stronger phase-gradient demands. This suggests that the radial tilt law functions as a local obliquity-matching mechanism layered onto the usual focusing-phase prescription.

3. Analytical tilt laws and design workflow

The reported workflow combines phase matching and a prescribed tilt map. First, a target focusing phase is derived from the oblique-incidence hyperbolic phase function. Second, a nearest-neighbor matching algorithm selects the meta-atom radius whose precomputed phase best approximates the target phase. Third, the tilt angle is assigned according to an analytical radial rule (Zhang et al., 22 Sep 2025).

Two explicit tilt distributions are reported.

The first is the PDTM, or Position-Dependent Tilted-Meta-atom metalens, intended for high throughput and efficiency. Its tilt varies linearly with radius: $_2$1 with $_2$2 as half the maximum FOV angle and $_2$3 as the metalens radius (Zhang et al., 22 Sep 2025). The tilt is smallest near the center and largest at the edge.

The second is the ADTM, or Aperture-Dependent Tilted-Meta-atom metalens, intended for higher imaging precision. Its tilt is defined geometrically as

$_2$4

where $_2$5 is the distance between an aperture and the metalens (Zhang et al., 22 Sep 2025). In this configuration, the aperture acts as a spatial filter that preferentially routes different field angles to meta-atoms whose tilts are better matched to those rays. The reported parameters are metalens diameter $_2$6, focal length $_2$7, NA $_2$8, aperture radius $_2$9, and aperture distance $\lambda = 633$0 (Zhang et al., 22 Sep 2025).

These two laws illustrate two distinct operating philosophies: direct radial tilt grading for efficiency-centric operation, and tilt grading combined with aperture-assisted angular selectivity for precision-centric operation. A plausible implication is that radial tilt design is not a single prescription but a family of spatial orientation rules constrained by the intended balance among throughput, aberration control, and angular selectivity.

4. Relation to affine geometric transformations and adjoint inverse design

A separate but closely related methodological development appears in "Adjoint-Based Gradient Evaluation for Metasurface Inverse Design via Affine Geometric Transformations" (Mottola et al., 16 Mar 2026). There, the metasurface is modeled as a collection of $\lambda = 633$1 2D scatterers or meta-atoms in the $\lambda = 633$2-plane in a TE$\lambda = 633$3 setting, and each meta-atom is described by a compact parameter vector rather than a pointwise permittivity distribution. The transformed coordinates are

$\lambda = 633$4

with rotation angle $\lambda = 633$5, anisotropic scaling factors $\lambda = 633$6, and rigid translation vector $\lambda = 633$7 (Mottola et al., 16 Mar 2026). The corresponding parameter vector is

$\lambda = 633$8

The central adjoint result is

$\lambda = 633$9

where $H = 0.5~\mu\text{m}$0 is the adjoint field, $H = 0.5~\mu\text{m}$1 the forward field, $H = 0.5~\mu\text{m}$2 a purely geometric weighting term, and $H = 0.5~\mu\text{m}$3 depends on the chosen cost function (Mottola et al., 16 Mar 2026). The paper emphasizes a decoupling: the adjoint source is determined by the cost function, whereas the weighting term is determined by the geometric transformation. Different affine transformations therefore do not require new adjoint PDE solves; only the boundary weight $H = 0.5~\mu\text{m}$4 changes (Mottola et al., 16 Mar 2026).

For rotation, the paper derives

$H = 0.5~\mu\text{m}$5

so that the geometric weight is effectively

$H = 0.5~\mu\text{m}$6

For anisotropic expansion along direction $H = 0.5~\mu\text{m}$7,

$H = 0.5~\mu\text{m}$8

with

$H = 0.5~\mu\text{m}$9

For translation along direction $P = 0.4~\mu\text{m}$0,

$P = 0.4~\mu\text{m}$1

with

$P = 0.4~\mu\text{m}$2

(Mottola et al., 16 Mar 2026).

This framework is not presented as a direct implementation of the wide-field metalens in (Zhang et al., 22 Sep 2025). However, it explicitly allows each meta-atom to have its own independent orientation and position-related variables, and the supplied discussion states that spatially varying orientation patterns, including radial ones, can be imposed or optimized by choosing $P = 0.4~\mu\text{m}$3 as a function of position. The cited example

$P = 0.4~\mu\text{m}$4

is given as a way to enforce a radially oriented pattern (Mottola et al., 16 Mar 2026). This suggests a methodological bridge between prescribed radial tilt laws and gradient-based optimization of spatially varying orientation fields.

5. Demonstrated devices and performance characteristics

The wide-field metalens study reports two demonstrated designs, both operating over a total FOV of $P = 0.4~\mu\text{m}$5, from $P = 0.4~\mu\text{m}$6 to $P = 0.4~\mu\text{m}$7 (Zhang et al., 22 Sep 2025). The two designs emphasize different trade-offs.

Design	Key geometry	Reported performance
PDTM	diameter $P = 0.4~\mu\text{m}$8; focal length $P = 0.4~\mu\text{m}$9; NA $0.02~\mu\text{m}$0	average transmittance $0.02~\mu\text{m}$1; average diffraction efficiency $0.02~\mu\text{m}$2
ADTM	diameter $0.02~\mu\text{m}$3; focal length $0.02~\mu\text{m}$4; NA $0.02~\mu\text{m}$5; aperture radius $0.02~\mu\text{m}$6; $0.02~\mu\text{m}$7	average diffraction efficiency about $0.02~\mu\text{m}$8 for $0.02~\mu\text{m}$9 to $0.2~\mu\text{m}$0; transmittance about $0.2~\mu\text{m}$1 at $0.2~\mu\text{m}$2 to $0.2~\mu\text{m}$3 at $0.2~\mu\text{m}$4

For the PDTM, the reported behavior includes clean focusing across $0.2~\mu\text{m}$5 to $0.2~\mu\text{m}$6. At $0.2~\mu\text{m}$7, the peak intensity drops to about $0.2~\mu\text{m}$8 of the on-axis value. The Strehl ratio decreases linearly with angle, reaching $0.2~\mu\text{m}$9 at $$\Phi(x,y)=\frac{2\pi}{\lambda}\left(\sqrt{x^2+y^2+f^2}-f\right).$$0, and the MTF remains strong on-axis but falls off significantly beyond about $$\Phi(x,y)=\frac{2\pi}{\lambda}\left(\sqrt{x^2+y^2+f^2}-f\right).$$1 (Zhang et al., 22 Sep 2025). The interpretation given is that PDTM favors throughput and uniform energy coupling across angle, while precision degrades more quickly off-axis.

For the ADTM, effective focusing is reported from $$\Phi(x,y)=\frac{2\pi}{\lambda}\left(\sqrt{x^2+y^2+f^2}-f\right).$$2 to $$\Phi(x,y)=\frac{2\pi}{\lambda}\left(\sqrt{x^2+y^2+f^2}-f\right).$$3. At $$\Phi(x,y)=\frac{2\pi}{\lambda}\left(\sqrt{x^2+y^2+f^2}-f\right).$$4, normalized focal intensity remains about $$\Phi(x,y)=\frac{2\pi}{\lambda}\left(\sqrt{x^2+y^2+f^2}-f\right).$$5 of the on-axis peak. The Strehl ratio drops from $$\Phi(x,y)=\frac{2\pi}{\lambda}\left(\sqrt{x^2+y^2+f^2}-f\right).$$6 at $$\Phi(x,y)=\frac{2\pi}{\lambda}\left(\sqrt{x^2+y^2+f^2}-f\right).$$7 to $$\Phi(x,y)=\frac{2\pi}{\lambda}\left(\sqrt{x^2+y^2+f^2}-f\right).$$8 at $$\Phi(x,y)=\frac{2\pi}{\lambda}\left(\sqrt{x^2+y^2+f^2}-f\right).$$9, and the MTF shows diffraction-limited focusing from $$\Phi(x,y)=\frac{2\pi}{\lambda}\left[\sqrt{(x-x_0)^2+(y-y_0)^2+f^2}-\sqrt{x^2+y_0^2+f^2}+(x\sin\theta_x+y\sin\theta_y)\right],$$0 to $$\Phi(x,y)=\frac{2\pi}{\lambda}\left[\sqrt{(x-x_0)^2+(y-y_0)^2+f^2}-\sqrt{x^2+y_0^2+f^2}+(x\sin\theta_x+y\sin\theta_y)\right],$$1, with sub-diffraction performance at larger angles. Imaging simulations of a USAF target and a complex pattern show readable detail across the FOV, with smallest resolvable detail around $$\Phi(x,y)=\frac{2\pi}{\lambda}\left[\sqrt{(x-x_0)^2+(y-y_0)^2+f^2}-\sqrt{x^2+y_0^2+f^2}+(x\sin\theta_x+y\sin\theta_y)\right],$$2 (Zhang et al., 22 Sep 2025). The paper characterizes ADTM as the higher-precision design because aperture filtering suppresses aberration-prone rays while the tilt pattern improves wavefront matching.

The reported performance metrics are total transmittance, diffraction efficiency, Strehl ratio, and MTF, respectively defined in the supplied description as integrated focal-plane energy over incident energy, energy within a radius of $$\Phi(x,y)=\frac{2\pi}{\lambda}\left[\sqrt{(x-x_0)^2+(y-y_0)^2+f^2}-\sqrt{x^2+y_0^2+f^2}+(x\sin\theta_x+y\sin\theta_y)\right],$$3 FWHM around the focus over total focal-plane energy, ideal FWHM over actual FWHM, and Fourier transform of the PSF (Zhang et al., 22 Sep 2025).

6. Validation, scope, and technical interpretation

The affine-transformation inverse-design study validates adjoint gradients against finite-difference gradients for rotation angle $$\Phi(x,y)=\frac{2\pi}{\lambda}\left[\sqrt{(x-x_0)^2+(y-y_0)^2+f^2}-\sqrt{x^2+y_0^2+f^2}+(x\sin\theta_x+y\sin\theta_y)\right],$$4, width or anisotropic deformation, and centroid translation, reporting excellent agreement in all cases (Mottola et al., 16 Mar 2026). Its numerical examples use rectangular meta-atoms with rounded corners made of TiO$$\Phi(x,y)=\frac{2\pi}{\lambda}\left[\sqrt{(x-x_0)^2+(y-y_0)^2+f^2}-\sqrt{x^2+y_0^2+f^2}+(x\sin\theta_x+y\sin\theta_y)\right],$$5 at $$\Phi(x,y)=\frac{2\pi}{\lambda}\left[\sqrt{(x-x_0)^2+(y-y_0)^2+f^2}-\sqrt{x^2+y_0^2+f^2}+(x\sin\theta_x+y\sin\theta_y)\right],$$6 nm. One inverse-design example optimizes a 128-meta-atom metasurface where the only design variables are the rotation angles. The optimized structure focuses light at a prescribed point, with a clear focal spot at approximately $$\Phi(x,y)=\frac{2\pi}{\lambda}\left[\sqrt{(x-x_0)^2+(y-y_0)^2+f^2}-\sqrt{x^2+y_0^2+f^2}+(x\sin\theta_x+y\sin\theta_y)\right],$$7, and the optimized rotation angles vary sharply and non-symmetrically across the array. Additional examples target two and three focal points, again using rotation angles as the design variables, and again yielding highly nonuniform orientation patterns (Mottola et al., 16 Mar 2026).

These results do not establish that in-plane affine rotation and out-of-plane physical tilt are interchangeable. They do, however, show that metasurface performance can depend strongly on spatially varying orientation variables and that those variables can be optimized efficiently within an adjoint framework (Mottola et al., 16 Mar 2026). This is significant for radially gradient-tilted meta-atoms because the metalens paper likewise treats local orientation as a first-class design variable rather than a fixed byproduct of phase encoding (Zhang et al., 22 Sep 2025).

Another misconception is that radial tilt merely redistributes phase without affecting wide-angle coupling. The supplied account explicitly argues that local tilt can improve coupling efficiency, reduce aberration, and better match the incident wavefront under oblique illumination (Zhang et al., 22 Sep 2025). By contrast, the inverse-design paper emphasizes that optimal orientation fields may be rapidly varying and non-symmetric, which indicates that radial regularity is a design choice rather than a universal optimum (Mottola et al., 16 Mar 2026). This suggests that radial tilt laws are one structured subset of a broader orientation-design space.

7. Trade-offs, scalability, and research context

The principal trade-off reported for radially gradient-tilted metalenses is efficiency versus precision. PDTM is described as efficiency-centric, whereas ADTM is precision-centric (Zhang et al., 22 Sep 2025). Improved angular robustness can therefore be associated with lower transmittance or reduced sharpness depending on the tilt strategy. Even in the ADTM case, transmittance decreases with angle, and the method is not presented as removing aberrations entirely at large incidence angles (Zhang et al., 22 Sep 2025).

Fabrication complexity is also identified as a limitation: tilted meta-atoms are more irregular than upright pillars, although distributed lithography or nanoimprint lithography are mentioned as candidate realization routes (Zhang et al., 22 Sep 2025). In addition, the wide-angle performance is presented in simulation. This suggests that the practical competitiveness of radially gradient-tilted meta-atoms will depend not only on optical design but also on process control for non-upright nanoscale geometries.

The FOV is reported as expandable by tuning the tilt-angle configuration and the metalens diameter (Zhang et al., 22 Sep 2025). A reconfigured design is stated to extend effective operation to about $$\Phi(x,y)=\frac{2\pi}{\lambda}\left[\sqrt{(x-x_0)^2+(y-y_0)^2+f^2}-\sqrt{x^2+y_0^2+f^2}+(x\sin\theta_x+y\sin\theta_y)\right],$$8–$$\Phi(x,y)=\frac{2\pi}{\lambda}\left[\sqrt{(x-x_0)^2+(y-y_0)^2+f^2}-\sqrt{x^2+y_0^2+f^2}+(x\sin\theta_x+y\sin\theta_y)\right],$$9 in the showcased example. Within the inverse-design perspective, the ability to optimize per-meta-atom geometric variables under affine transformations with one forward and one adjoint solve per iteration provides a computational argument for exploring such enlarged orientation-design spaces at scale (Mottola et al., 16 Mar 2026).

In current research context, radially gradient-tilted meta-atoms therefore denote more than a specific metalens layout. They identify a design paradigm in which spatially varying local orientation is elevated to the same level of importance as local phase prescription. One branch of this paradigm prescribes analytical radial tilt laws for monolithic wide-field metalenses (Zhang et al., 22 Sep 2025); another develops adjoint machinery for efficient gradient evaluation over orientation and other affine degrees of freedom (Mottola et al., 16 Mar 2026). Taken together, these works frame orientation-engineered metasurfaces as a technically distinct route beyond upright meta-atoms, phase-only encoding, and stitched wide-angle architectures.