Light Passes Through

What really happens when light enters glass, water, a raindrop or a person's ear: transparency, transmission, refraction, rainbows, the index of refraction, chromatic aberration, caustics and translucency, explained for people who light, shoot and render pictures.

Optics · Light · 13 min

Every picture we make is made of light that has passed through something. It went through the air of the set, through the glass of a lens, often through a window, a glass of water, a raindrop or a diffusion frame, and sometimes a short way into a person's skin before it came back out. Each of those materials changes the light on its way through: it bends it, splits it into colours, focuses it into bright patterns or spreads it into a soft glow. This post walks through those changes one at a time, from **transparency** and **transmission** to **refraction** and why it happens, to **rainbows**, the **index of refraction**, **chromatic aberration**, **caustics** and **translucency**. Every figure is calculated live, so the bending you see is the bending the physics predicts.

Four things can happen to light

When light reaches a material, only four things can happen to it, and every material is just a different mix of the four. Some light **reflects** straight off the surface. Some is **absorbed** and turned into a tiny bit of heat. Some is **scattered**, knocked into a new direction by particles, fibres or cells inside. And some is **transmitted**, it goes in one side and out the other. Watch the photons below. In clear glass almost all of them pass straight through. In milk they bounce around until they find a way out. In wood they barely make it in before they are absorbed.

That gives a simple way to describe any material, with just two questions. How much does it **scatter** light inside? How much does it **absorb**? A material that does neither is **transparent**: air, water, clear glass, diamond. One that absorbs but does not scatter is transparent but tinted, like sunglasses or red wine. One that scatters a lot but absorbs little is **translucent**: frosted glass, candle wax, milk, skin, a cloud. And one that does a lot of both is **opaque**. A third number, the **index of refraction**, sets how much light bends and reflects at the surface, and we will get to it in Part 8.

What is transparency?

A material is **transparent** when light passes through it without being scattered, so the light keeps its direction and you can see a clear picture through it. That last part is the real test. Frosted glass lets plenty of light through, but it scrambles the directions, so you see brightness without a picture. The amount of light that gets through without a single scattering event falls off exponentially with the thickness and the scattering strength, and that is the light that carries the picture. Notice another detail in the figure: through frosted glass, things touching the glass stay fairly sharp and things far behind it melt away. The scattered directions spread further the further the light has to travel to the object, which is why a shower door hides the room but not a hand pressed against it.

Transparent does not mean colourless. A transparent material can still **absorb**, and when it absorbs some wavelengths more than others it becomes tinted. The rule is the Beer–Lambert law: every centimetre of material removes the same *fraction* of the light that is left, so the transmitted light falls exponentially with depth, separately for each colour. That is why a thin glass of tea is golden and a full pot is nearly black, and why water itself, which absorbs red a little more than blue, turns everything blue-green below a few metres. Underwater camera operators lose the reds first, and that is exactly why red filters and lights are part of their kit.

Light passing through an object

**Transmission** is the share of light that makes it through an object. Even perfectly clear glass never transmits everything, because every surface reflects a little. How much depends on the angle. Look straight down into a lake and you see the stones on the bottom; look across it toward the far shore and it becomes a mirror of the sky. The French physicist Augustin-Jean Fresnel worked out the formula in the 1820s, and the Fresnel equations are still the core of every physically based renderer. For glass, about 4% of the light reflects when you look straight on, and it rises steeply toward 100% as the angle becomes grazing.

This is why every 3D package has a **transmission** control. In Blender's Principled BSDF it is called Transmission Weight, in Autodesk Arnold's Standard Surface it is Transmission, and Redshift, V-Ray and Unreal Engine all have their own version. Turn it up and the surface stops scattering light diffusely and lets it through instead. The settings that come with it are always the same handful: the **IOR**, which sets how much it bends and reflects; a **roughness** that turns clear glass into frosted; a **colour and depth** that set Beer–Lambert absorption; and in some renderers a **dispersion** value that splits colours. The panel below is a mock-up of those common controls driving a small ray tracer, so you can feel what each one does.

What is refraction?

**Refraction** is the bending of light as it crosses from one material into another. It is why a straw looks broken in a glass of water, why a pool looks shallower than it is, and why a lens can focus at all. The amount of bending follows **Snell's law**: n₁ · sin θ₁ = n₂ · sin θ₂, where θ is the angle from the surface normal and n is each material's index of refraction. Going into a denser material such as water or glass, light bends toward the normal. Coming out, it bends away. Light that hits the boundary straight on does not bend at all.

That stop at the end of the curve is **total internal reflection**. When light inside water or glass meets the surface at more than the **critical angle** (about 48.6° for water and 41° for glass), Snell's law has no answer, and all of the light reflects back in. It is how optical fibre carries light around corners and why a diamond sparkles. It also gives divers a strange view: looking up from underwater, the whole sky above the surface is squeezed into a bright circle about 97° wide, called **Snell's window**, and outside that circle the surface becomes a mirror of the water below.

Why is light redirected during refraction?

Light bends because it **slows down**. In a vacuum light travels at about 300,000 km per second, in water at about 225,000 and in glass at about 197,000. The frequency of the wave cannot change at the boundary (each crest that arrives has to leave), so when the speed drops the crests bunch up and the wavelength gets shorter. Now picture a wavefront arriving at an angle. One end of it enters the glass first and slows down while the other end is still racing through the air. The front has to pivot to stay connected, and the direction of travel, which is always at right angles to the front, pivots with it.

A classic way to feel this is a pair of wheels on an axle, rolling off a smooth road onto sand at an angle. The first wheel to touch the sand slows down while the other keeps going at road speed, so the axle swings around, and by the time both wheels are in the sand the whole thing is travelling in a new, steeper direction. Roll it straight in and both wheels slow together, so there is no turn. Roll it back out of the sand and it turns the other way. That is refraction in one picture: a change of speed across a boundary that is crossed at an angle.

Light wave "turning" or "bending"

There is a second, beautiful way to see the same thing, from the 17th-century mathematician Pierre de Fermat: light always takes the path that takes the **least time**. Think of a lifeguard on a beach who has to reach a swimmer in the sea. Running is faster than swimming, so the straight line is not the fastest route. It is better to run a bit further along the sand and swim a shorter distance. Drag the entry point and watch the time. The fastest route bends exactly at the shoreline, and its angles obey Snell's law with the running and swimming speeds in place of the speeds of light.

Refraction does not need a sharp boundary. When the index of refraction changes gradually, light does not snap to a new direction, it **curves**. Over a hot road on a summer day the air near the asphalt is hotter and thinner, with a slightly lower index, so rays heading down toward the road are bent smoothly back up. The camera then sees the sky where the road should be, a shimmering pool of "water" that is really a mirage. Any cinematographer who has shot down a long road on a long lens has filmed this, together with its cousin, heat haze: the same bending, made unsteady by moving pockets of warm air.

Refraction and reflection in rainbows

A **rainbow** is refraction, reflection and refraction again, inside millions of falling drops. Sunlight enters a drop and bends, reflects once off the back of the drop, and bends again on the way out. Because water's index is a little higher for violet than for red, each colour comes out at a slightly different angle. The trick is that rays entering at many different heights all leave at nearly the same angle, about 42° from the direction opposite the sun for red and about 40.6° for violet. Light piles up at that angle, so that is where we see a bright band.

That one rule explains everything you see in the sky. The bow is a circle centred on the point exactly opposite the sun, which is why it sits lower when the sun is high and vanishes when the sun is above 42°. Red is on the outside. Inside the bow the sky is brighter, because every drop sends some light to angles smaller than 42°, and just outside it there is a darker strip called **Alexander's band**. Light that reflects twice inside the drop makes a fainter **secondary bow** at about 51°, with its colours reversed. For filmmakers that has practical uses: put your back to the sun, and know that a rainbow you can see at golden hour will be gone by midday.

One number for how much light slows down

The **index of refraction**, or IOR, is simply how many times slower light travels in a material than in a vacuum: n = c / v. Air is 1.0003, so close to 1 that we usually ignore it. Water is 1.333, ordinary window and lens glass is around 1.5, and diamond is 2.42, which means light crawls through diamond at less than half its vacuum speed. The higher the number, the more light bends when it enters. Click a material and the glass ball is re-rendered with that IOR. Watch how the upside-down picture inside it shrinks and how much more of the surface turns into reflection.

The IOR does a second job that is easy to miss: it also sets how **reflective** the surface is. At a straight-on angle the reflectance is ((n − 1) / (n + 1))², so water reflects about 2%, glass about 4% and diamond about 17%. That is why many renderers use an IOR of 1.5, about 4% reflectance, as the default for almost every non-metal, from plastic to painted wood. It is also why a diamond looks brighter than glass even before it sparkles, and why skin, at about 1.4, has a softer sheen than a glass of water would at the same angle.

When colours focus in different places

The index of refraction is not one fixed number. It changes slightly with wavelength, and in almost every clear material it is higher for blue light than for red. This is called **dispersion**, and it is what split the beam in Fig. 01. Glass makers describe it with the **Abbe number**: a high number like 64 for crown glass means little dispersion, a low number like 36 for flint glass means a lot. The curves below are calculated from the Sellmeier formulas glass makers publish, and the Abbe numbers in the readout are computed from those same curves.

In a lens, dispersion means each colour has its own focal length. Blue focuses a little closer to the lens than red, so no single plane is sharp for every colour. This is **longitudinal (axial) chromatic aberration**, and you see it as purple and green fringes in front of and behind the point of focus, especially on fast lenses wide open. The classic fix, from the 1700s, is the **achromatic doublet**: a converging lens of low-dispersion crown glass cemented to a weaker diverging lens of high-dispersion flint. The flint undoes the colour spread of the crown without undoing all of its focusing power, so red and blue meet again.

The other kind, **lateral (transverse) chromatic aberration**, changes the *size* of the image with colour, so the red, green and blue pictures are slightly different sizes. The centre of the frame is clean and the fringes grow toward the edges, which is where you see red and cyan edges on dark branches against a bright sky. Because it is just a scaling of colour channels, software can undo it well: Adobe Lightroom and Camera Raw have a Remove Chromatic Aberration checkbox, and many cameras correct it in the lens profile before you ever see the file. Compositors do the opposite, adding a touch of it to CG elements so they sit in plates shot through real lenses, using a lens distortion node with a dispersion setting such as the one in Blender's compositor.

Light focused into bright patterns

**Caustics** are the bright, sharp patterns made when a curved transparent surface focuses light: the dancing net of light on the floor of a swimming pool, the bright spot inside the shadow of a wine glass, the ripples of light reflected from a river onto the underside of a bridge. Every little wave on the pool's surface acts like a weak lens. Where the surface curves one way it gathers light into a bright line; where it curves the other way it spreads light out into a dim patch. The deeper the water, the further the light travels after bending, and the sharper the lines become, until at large depths they cross and blur again.

Caustics are famously hard to render, and the reason is worth knowing if you work in CG. Most modern renderers are **path tracers**: they start at the camera, bounce rays around the scene and hope to hit a light. A caustic on the floor can only be lit by a path that goes through the glass and lands exactly on the small sun, and random bounces rarely do that, so the result is noisy fireflies. Tracing from the light instead, the idea behind **photon mapping**, finds caustics easily. This is why some renderers have dedicated caustic modes, why Blender added "shadow caustics" to Cycles in version 3.2, and why large soft lights make caustics far easier to render than a small hard sun.

What is translucency?

A material is **translucent** when light passes through it but is scattered along the way, so it glows rather than shows a picture. Paper, frosted glass, leaves, wax, marble, jade, milk and skin are all translucent. The light enters, takes a random walk through fibres, cells or particles, and leaves somewhere else, often in a different direction and a different place from where it entered. That is why a candle seems to glow from inside and why a backlit leaf lights up green. Change the scattering below and you can push the slab from transparent through translucent to opaque. The curve under it shows where the light comes out: the wider it is, the softer and blurrier the glow.

In CG this is called **subsurface scattering**, or SSS, and it is what makes a digital face look like skin instead of plastic. Inside skin, blood and tissue absorb green and blue much more than red, so red light travels furthest before it comes back out. That is why the edge of a shadow on skin has a warm red fringe and why ears glow red when backlit. Renderers expose this as a **subsurface radius** per colour channel. Blender's Principled BSDF, for example, defaults to a radius of 1.0, 0.2 and 0.1 for red, green and blue, multiplied by a scale. The figure below takes the same light and shadow and spreads it under the surface with three different settings, so you can see exactly where that warmth comes from.

Sources and further reading

  • Eugene Hecht, *Optics* (5th ed., Pearson, 2017): Fresnel equations, Snell's law, Huygens' principle, dispersion, rainbows
  • Schott AG, optical glass data sheets for N-BK7, F2 and SF11 (Sellmeier coefficients), via refractiveindex.info
  • Richard Feynman, *QED: The Strange Theory of Light and Matter* (Princeton, 1985): least time and the lifeguard picture
  • M. G. J. Minnaert, *Light and Color in the Outdoors* (Springer, 1993): rainbows, Alexander's band, mirages, Snell's window
  • Blender Manual, "Principled BSDF" and "Lens Distortion Node": docs.blender.org
  • Blender 3.2 release notes, Cycles "Shadow Caustics": developer.blender.org
  • Autodesk Arnold User Guide, "Standard Surface": help.autodesk.com
  • Henrik Wann Jensen, *Realistic Image Synthesis Using Photon Mapping* (A K Peters, 2001)
  • H. W. Jensen, S. Marschner, M. Levoy, P. Hanrahan, "A Practical Model for Subsurface Light Transport", SIGGRAPH 2001

*All figures are illustrations computed in your browser, not photographs or screenshots of any software. The physics in them (Snell's law, Fresnel reflectance, Sellmeier dispersion, Beer–Lambert absorption, ray statistics in raindrops, photon random walks) is calculated for real; material values marked as illustrative are rounded to show behaviour, not measured. Software features are described from the makers' own documentation as of October 2026.*

Back to the blog