Walk the independent claim. Snap's grant US12204112B2, "Waveguide and diffraction grating for augmented reality or virtual reality display" (issued January 21, 2025; inventors including Alexandra Crai), is a granted patent. Its CPC mix — G02B 27/0172 head-mounted-display optics, the diffraction-grating class G02B 5/1819, and G06T 19/006 augmented-reality rendering — places it squarely in diffractive-waveguide art rather than the reflective camp.

The structure claim 1 recites is more specific than "a grating that couples light out." It defines an input region that couples light into the waveguide so it propagates under total internal reflection (TIR), and an output region that provides three distinct diffractive interactions: a first interaction that out-couples light propagating along a first axis toward the viewer, a second that out-couples light propagating along a perpendicular second axis toward the viewer, and a third that turns light so it continues to propagate under TIR along an axis perpendicular to its prior direction. In other words, the same output region both spreads the image across the eyebox (pupil expansion in two directions) and releases it to the eye — and the claim ties all three jobs to one engineered grating region.

The load-bearing limitation, though, is what the claim does next: it divides that output region into a plurality of zones, each with different non-zero diffraction efficiencies for those three interactions. Claim 1 recites a first zone and a second zone offset along the second axis, and requires that in the second zone the efficiency of one interaction be both greater than the other interaction in that same zone and greater than the same interaction back in the first zone — "so as to reduce rainbow artefacts." That closing phrase is the point of the invention. Uniform gratings leak light unevenly and scatter stray orders that a viewer perceives as colored rainbow ghosting; spatially grading the diffraction efficiency zone-by-zone is the disclosed mechanism for suppressing it while keeping luminance even across the field.

“A waveguide for use in a virtual reality, VR, or augmented reality, AR, device, is disclosed.”— U.S. Patent No. 12,204,112 source

The dependent claims show how the efficiency grading is physically realized, which is useful for gauging how tightly the structure is pinned. Claim 4 builds each zone from two overlaid rectangular periodic arrays of optical structures — two 2D lattices with rectangular symmetry, spatially offset from one another by a factor that is deliberately not half the lattice period. Shifting one sub-lattice against the other changes the local diffraction efficiency without changing the grating's period, so the designer tunes efficiency by offset rather than by re-pitching the grating. Claims 5 and 6 then specify the offsets that make the structures "continuous" along one axis versus the other, and claim 7 adds a third zone that receives light directly from the input region and is biased toward the turning (third) interaction — the fan-out region of the pupil expander. Claim 11 adds that the second zone's structures have a larger cross-sectional area than the first zone's, another lever on local efficiency. Claim 12 notes the input region can be a simple 1D linear grating.

What it reads on is the display optics of AR smart glasses, and the assignee is the notable part. Snap is known for its camera-and-AR software, yet it builds Spectacles hardware and holds genuine waveguide IP. This grant places a software-first company among the diffractive-waveguide holders alongside Magic Leap and Meta — and it does so with a claim that addresses one of the practical failure modes (rainbow artefacts) that separates a demo-grade waveguide from a shippable one.

Scope discipline matters here because the claim is narrower than "diffractive AR optics." It is anchored to a two-dimensional pupil-expanding output region whose zones carry different non-zero efficiencies for three named interactions, arranged so a specified efficiency ordering holds between zones. A reflective-waveguide headset — Lumus-style geometric optics — operates in a different regime and would not read on it. Equally, a single-grating out-coupler without the zoned, graded-efficiency structure, or one that spreads the pupil in only one axis, falls outside the recited limitations. The defensible element is the zoned efficiency grading for artefact reduction, not diffraction generally.

Granted status makes US12204112B2 a live obstacle on the diffractive path rather than an aspiration. Claims 15 through 20 extend the same structure to "an augmented reality or virtual reality display comprising the waveguide," so the protection runs from the optical component up to the finished display. As with the diffractive-versus-reflective and flat-versus-curved splits, entering AR optics means choosing which thicket to navigate — and Snap has staked a diffractive position, at the level of grating micro-structure, that a competitor on that path must account for.

It is worth dwelling on why the two-axis pupil expansion in claim 1 raises the stakes. A diffractive waveguide for glasses has to take a small projector pupil and replicate it across a much larger eyebox so the image stays visible as the eye moves and across a range of face shapes. Claim 1's three interactions — out-couple along the first axis, out-couple along the perpendicular second axis, and turn light to propagate along a perpendicular axis under TIR — are precisely the operations a two-dimensional expander performs, with claim 13's fourth zone added to "selectively turn the light" toward the second zone and claim 14's fifth zone mirroring the second on the opposite side of the first along the second axis for symmetry. Pupil replication of this kind is exactly where rainbow artefacts and luminance roll-off appear, so the zoned-efficiency grading is not a cosmetic add-on; it is the mechanism that makes a two-axis expander deliver a uniform image. That is the disclosed engineering, and it is the part a competitor cannot simply paraphrase away.

For a landscape analyst the takeaway is twofold. First, the assignee list for AR optics is wider than the headlines suggest: a social-camera company holds granted, structurally detailed diffractive-waveguide claims. Second, the claim's center of gravity is image quality, not merely image delivery — the zoned-efficiency, rainbow-suppression limitation is where the engineering value and the enforceable scope coincide. Report the structure as the patent recites it, and the strategic read follows: this is a manufacturing-grade optics claim from a company most observers file under software.