Phase, diffraction, and etendue explain why most products sold as holograms are not, and why the real ones stay hard to build.
Walk any trade show floor and you will find a spinning LED fan painting a logo in mid-air under a sign reading "3D hologram." It is a good effect. It is also not a hologram.
The label gets attached to a lot of unrelated hardware: reflective stage illusions, rotating LED arrays, transparent display cases, headset overlays. None of it reconstructs a wavefront, which is what holography means. If you work near graphics, XR, or display pipelines, the distinction matters: these technologies differ in data model, compute budget, and failure mode.
What is a 3D hologram?
A hologram is an interference pattern that reconstructs both the amplitude and the phase of light from a scene. A normal image stores only intensity per pixel.
A hologram encodes the shape of the light wave itself, so illuminating it correctly recreates the original wavefront in free space, and your eye responds as it would to a real object.
1. The defining feature is phase, not depth
A framebuffer holds intensity: three numbers per pixel. A hologram holds a complex-valued field, amplitude and phase. Phase encodes where light came from and how far it travelled.
Dennis Gabor worked this out in 1948 while trying to beat the electron microscope's resolution limit, naming it from the Greek holos, meaning whole, because the recording keeps the whole wave rather than its brightness alone.
It earned him the 1971 Nobel Prize in Physics. An image is a projection; a hologram keeps what the projection discards.
2. Reconstruction is diffraction, not rasterization
There is no fragment shader here. Coherent light hits the hologram, its fringe structure diffracts that light, and the diffracted orders sum into the original wavefront.
The physics does the compositing. This is why a hologram fragment still shows the whole scene from a narrower angle, and why the recording looks like noise: fringe spacing carries the geometry.
3. Is a hologram fan a real hologram?
No. It is a persistence-of-vision display: LEDs on spinning blades switch at timed rotational positions, and your visual system fuses the samples into an apparent picture.
The image lives in the plane of rotation, so it is a flat image swept through a disc rather than a volume. Walk ninety degrees around one and it thins to an edge.
4. The "hologram" on stage is usually Pepper's Ghost
The 2012 Coachella appearance of a deceased rapper was billed as a hologram. It was a bright 2D projection bounced off angled foil, a Victorian theatre illusion predating holography by a century. The optics are pure reflection, so the image carries no depth information and the effect collapses from the wrong seat.
5. Volumetric displays are a real and separate category
Unlike the previous two, these put light where the image is. A published photophoretic trap display traps a cellulose particle in a shaped light field, drags it through a volume, and lights it with red, green, and blue lasers, giving full-colour images with roughly ten-micrometre points.
Because the scattering points sit inside the image volume, it is viewable from nearly any direction and avoids the clipping that affects anything modulating light at a bounded 2D surface. The catch: only real self-luminous points, never virtual ones.
6. Light-field and holographic displays sample light differently
A light-field display reproduces directional rays so different viewpoints receive different rays. It is a ray-optics approximation and inherits a spatial-versus-angular tradeoff: more views cost pixels.
Holography works in wave optics and reconstructs the field continuously instead of sampling it into discrete views. The two are not interchangeable terms, however often vendor copy treats them that way.
7. Why are holographic displays so hard to build?
One number explains most of it. A spatial light modulator's diffraction angle is bounded by pixel pitch, roughly 2 arcsin(lambda / 2p), so today's pitches of several micrometres yield a field of view in the low single digits.
The governing quantity is etendue, the product of modulator area and the solid angle it can diffract into, and it forces holographic displays to trade field of view against eyebox or display size.
Immersive headsets want roughly 120 degrees with an eyebox over 10 by 10 mm. Shrinking the pitch is the obvious fix and runs into hard fabrication limits.
8. Computing a hologram is not the same as rendering a frame
Computer-generated holography numerically simulates diffraction and interference rather than projecting geometry. Fresnel diffraction simulation is costly enough that quality and runtime trade directly against each other, which kept dynamic holography impractical for decades.
Learned approaches changed that. MIT's tensor holography trains a convolutional network to synthesise a colour 3D hologram from one RGB-D image; the published network is under 620 KB and hits 60 Hz at 1920 by 1080 on a single consumer GPU. A familiar shape: an expensive forward simulation replaced by a learned approximation that fits in cache.
9. Occlusion and per-pixel focus are still open problems
In a rasteriser, occlusion is a depth test. In holography it is a wave-propagation question, because light from a hidden surface must be blocked correctly in the field computation rather than discarded in a buffer.
The tensor holography authors are explicit that earlier physically based methods could not deliver per-pixel focal control and accurate occlusion together.
10. The real payoff is fixing vergence-accommodation conflict
Every stereoscopic headset presents two images at one fixed focal distance, so your eyes converge on a virtual object two metres away while your lenses stay focused on a panel centimetres from your face. That mismatch drives eye strain, fatigue, and degraded depth judgement. A holographic display reconstructs the wavefront an object would actually emit, so the eye focuses on the object rather than the panel. That, far more than the science-fiction imagery, is why display research keeps returning to holography.
What does this mean for developers?
Mostly it means precision when a brief lands on your desk. If it says "3D hologram," ask which technology is meant, because the pipeline changes completely. A Pepper's Ghost rig wants rendered video on a black background. A light-field display wants multi-view rendering. A holographic display wants RGB-D or a wave-optics representation, with the compute in hologram synthesis rather than your shading pass. There is no WebGPU path to a physical holographic display today and no stable cross-vendor API, but RGB-D capture and depth-aware rendering feed CGH pipelines directly.
Where the field actually stands
Real-time hologram computation is largely solved on consumer hardware. The modulators are not. Etendue is the binding constraint, and it is a fabrication problem rather than an algorithmic one, so it will not fall to a better loss function. Meanwhile most hardware sold as holographic is doing something else entirely, usually something much older.
Top comments (0)