Why a Flat Camera Lens Sees in Only One Color

Rand Simulation — Applications Engineering AI
Photonics · Ansys Lumerical FDTD · 8 min read

AI disclosure: RandSim Labs is an experimental AI-driven engineering simulation platform. Content on this site, including simulations, analyses, figures, and written materials, may be generated or assisted by AI using licensed Ansys tools. AI-generated content may contain errors and is provided for educational, informational, and demonstration purposes only. Users should independently verify all results before relying on them for engineering, design, manufacturing, safety, or other production decisions.

A spoke target imaged through a perfect lens versus a flat metalens: sharp green with red and blue color halos — The same resolution target seen two ways. Through a perfect lens (left) every spoke is crisp. Through a flat 1.0 mm metalens focused for green (right), the green channel is sharp while red and blue smear into colored halos — the whole image, solved in Ansys Lumerical, is sharp in exactly one color. This is the central limit of a single flat metalens as a camera lens.

The result: a metalens is a flat sheet patterned with millions of sub-wavelength pillars, each sized to add the phase a curved lens would — a lens made of data. We built one for a visible camera in Ansys Lumerical: a 1.0 mm, f/2 TiO₂ lens (8,724,901 pillars on a 1.1 µm stack, about 64x thinner than a hair) designed for green. It focuses beautifully — a near-diffraction-limited 1.34 µm spot at 550 nm. But the focus walks with color: red lands at 1678 µm, green at 1999 µm, blue at 2452 µm — a 774 µm spread. Put a sensor at the green focus and green is 5385x sharper than red and 9532x sharper than blue. The flatness that makes it thin makes it a one-color camera.

A camera lens has one job: take every ray leaving a point in the world and bring it back to a single point on the sensor, for every color at once. A glass lens does it with curvature and thickness. A metalens does it with a flat surface: a dense field of pillars, each a tiny antenna whose width sets how much it delays the light passing through. Choose each pillar's width to reproduce the phase a curved lens would have imposed, and a sheet a micron thick bends light like a lens millimeters thick. The question that decides whether it can be a camera lens is not whether it focuses — it is whether it focuses all colors to the same plane. That is a solve, not a formula, and it is the question here.

This study is the visible-light companion to our thermal-camera metalens, Can a Flat Lens Made of Data Focus Heat? That one showed a silicon metalens focusing long-wave infrared and the focus sliding with color; this one asks the same question for a visible camera — and answers it with a picture, not just a focal curve.

Building the lens from measured pillars

The TiO2 pillar-width map across the aperture: concentric Fresnel-zone rings — The optic itself — the width of every TiO₂ pillar across the 1.0 mm aperture, colored by size. The concentric rings are Fresnel zones: each time the required phase wraps through a full turn, the widths repeat. This map is the lens; there is no curve anywhere in it.

You cannot guess how a pillar delays light — you have to measure it. Ansys Lumerical FDTD solves Maxwell's equations for one TiO₂ pillar in its lattice, returning the phase and transmission it imposes at each color; sweeping the width builds a library of “atoms” to choose from. That library is the honest core of the study: the built atoms span a full 2π of phase at 550 nm with near-unity transmission and a clean energy balance, gated before a single pillar was placed. With the library in hand, the green-design phase at every point of the aperture picks the pillar that fits, and the lens falls out: about 8,724,901 pillars, each width a number, on a stack 1.1 µm thick.

It focuses — then the color walks off

Best-focus distance versus wavelength, red green and blue marked against the ideal 1/lambda curve — Where each color actually comes to a focus, from propagating the measured pupil. Green lands on the 1999 µm design plane (the star); red focuses short at 1678 µm and blue long at 2452 µm, tracking the tell-tale 1/λ law of a diffractive lens — the opposite sign to glass, and much stronger.

Handed the measured pupil — each cell carrying the transmission and phase its pillar was solved to have — an exact angular-spectrum propagation shows where each color comes to a point. Green lands on the 1999 µm design plane; red and blue do not. They focus at 1678 and 2452 µm, a 774 µm spread across the visible band, because the pillars deliver the right phase at exactly one wavelength and the wrong one everywhere else. This is chromatic aberration, and on a metalens it is severe and backwards from glass: the focal length falls as 1/λ, so blue (short wavelength) focuses farthest, red nearest. A single flat surface simply cannot hold all three colors to one plane.

What the sensor sees

The red, green and blue point-spread functions at the sensor plane: green tight, red and blue blurred — The point-spread function — the image of a single point of light — at the sensor (green-focus) plane. Green is a tight core; red and blue, focusing elsewhere, arrive as broad blurred discs. A camera sensor here records a green image riding on colored halos.

Fix the sensor at the green focus, where a camera would put it, and propagate a resolution target through the lens color by color. Green comes through crisp — a point images to a spot 5385x tighter than red and 9532x tighter than blue at that same plane. Red and blue, focusing hundreds of microns away, land on the sensor as broad discs, so every bright feature wears a magenta-and-blue halo. That is the hero image above: the lens is genuinely, sharply imaging — in green — and fringing everything else. Stop the target down to green light alone and a flat metalens is a superb, hair-thin camera lens; hand it white light and it is a one-color camera.

That is not a modeling artifact — it is the honest state of the art, and the reason single-layer metalenses are not yet in your phone's main camera. A flat metalens buys astonishing thinness and a fabrication that is lithography rather than grinding, at the cost of a chromatic spread a curved lens does not have. Beating it — stacking dispersion-engineered atoms, or splitting the color channels — is the open research problem the field is built around. Seeing both halves in one solved image, the crisp green and the colored halos, is exactly what tells a camera designer whether a metalens belongs in a given system.

What this model does and does not cover

Honest scope. Ansys Lumerical FDTD 2026 R1 for the TiO₂ unit-cell atom library (n = 2.4, 0.30 µm square lattice, 0.60 µm pillars, swept across the visible band), assembled into a 1.0 mm, f/2 (NA 0.25) phase mask and propagated to focus and to the sensor plane by the angular spectrum of plane waves — an exact scalar solution of the Helmholtz equation. The focus positions track the analytic 1/λ anchor to within a few percent. It is a scalar propagation of a measured pupil, not a full 3D FDTD of the whole millimeter-wide optic (which is intractable), and it treats each color band at a single index. The physics that matters — the green-design focus, the chromatic focal walk, and the one-color imaging — is robust; absolute focal distances are representative of this design, not a specific fabricated part.

Designing a metalens, a diffractive optic, or any structured surface where the pattern is the component? The design wavelength is the easy part; imaging across a band is where the physics lives, and it is a solve — unit-cell library plus propagation — not a formula. We do metasurface and diffractive-optics work in Ansys Lumerical. Rand Simulation — innovation through insight.

Why a Flat Camera Lens Sees in Only One Color

Building the lens from measured pillars

It focuses — then the color walks off

What the sensor sees

What this model does and does not cover

Recent Posts