You'd assume lens flare comes from dirty glass. It doesn't. It comes from physics that no amount of cleaning fixes — specifically, from how light bounces between glass elements inside the barrel, in gaps a coating on the outer surface was never built to reach.
In 1935, Carl Zeiss introduced T coating: a single layer of magnesium fluoride applied to a lens's exterior surfaces. Uncoated glass reflects roughly 4% of incident light at each air-glass interface. A single MgF₂ layer, tuned for the middle of the visible spectrum, cuts that down to around 1.5%. A real improvement — and not the whole story, because it only fixes the surfaces it touches.
Where the Ghost Actually Comes From
Coating handles one kind of reflection: light entering a coated element, bouncing off that element's own back surface, and some of it escaping back out the front. The coating suppresses this by design.
There's a second kind coating doesn't reach. Light passes through the first element, crosses the air gap, and reflects off the back surface of the second element — then travels backward through that gap and the first element again. That's an internal Fresnel reflection, happening at an air-glass interface deep in the lens stack that the outer coating was never applied to, or that a coating there is tuned for a different angle and geometry than a stray internal bounce actually presents.
This internal reflection is what produces the ghost: a dim, often hard-edged inverted image of the light source, distinct from the soft haze of general veiling flare. Because the light takes multiple bounces through different glass types, path length varies slightly by wavelength — a dispersion effect — so the ghost usually isn't neutral. One edge reads blue, the other red.
Why the Ghost Has a Color at All
At an air-glass interface, reflectance follows the Fresnel relation: for light hitting straight-on, R = ((n₁ − n₂) / (n₁ + n₂))². Air sits at n≈1.0, typical optical glass around n≈1.5, which gives R≈0.04 — the 4% every uncoated surface reflects. An MgF₂ layer (n≈1.38) works by interference: at a quarter-wave optical thickness, the wave reflecting off the coating's front surface and the wave reflecting off its back surface arrive out of phase and cancel each other, dropping reflectance toward that 1.5% figure.
That cancellation is tuned to one wavelength and imperfect at the spectrum's edges — blue and red don't cancel as cleanly as the green the coating was optimized for. The residual reflection that survives is what shows up as colored fringing on the ghost. It isn't a manufacturing defect. It's the coating running into the limit of what a single-layer interference trick can do across a whole visible spectrum.
The color also depends on the glass itself. Low-index crown glass disperses light less than high-index flint glass, so a lens built mostly from crown elements throws faint, close-to-neutral ghosts, while one with a flint element in the middle throws stronger, more saturated color.
Why It Became a Look Instead of a Flaw
Modern lenses use multi-layer coatings — ten or more layers of alternating materials, engineered to suppress reflection across the full visible range and a wide span of incidence angles. They knock down both the surface reflections T coating targeted and the internal ones it couldn't touch. The image gets optically cleaner: less veiling glare, fainter ghosts.
For a lot of shooters, that cleanliness reads as a loss rather than a fix. A colored ghost is a signature of a specific optical design — the glass types, the element count, the coating generation of its era. A 1950s Zeiss and an uncoated older design don't throw the same ghost, and the difference is a fingerprint of when and how the lens was built, not a variable anyone can dial in after the fact. Digital sensors' higher sensitivity to low-level light makes these artifacts more visible than they were on film, which is part of why the vintage-lens-flare look has stayed in demand even as coating technology moved past the T-coating era. For more on how optical artifacts differ from the digital effects they get compared to, see Bloom Is Not Halation — a related case of an effect people mix up with something it isn't. On the color-separation side of what a coating layer does and doesn't touch, see our piece on why spectral curves overlap.
The ghost moves with the lens, distorts with the aperture, and shifts with focal length — it's a geometric relationship between light source and glass stack, not a fixed shape. Chasing that specific signature still means shooting the actual lens against a strong point source, because the physics that produces it is the same physics a coating was designed to fight and never fully could.