Calculate the Critical Angle for Total Internal Reflection
When light travels from a denser medium into a rarer one (glass into air, say), there is an angle of incidence beyond which no ray gets out at all: that is the critical angle θc, and the effect is called total internal reflection. Work out θc from two refractive indices, or the numerical aperture of an optical fibre — a quantity that follows directly from it.
This page complements Snell's law (refraction), which also computes the critical angle but within the broader context of refraction. Here the focus is on total internal reflection itself and its most direct application: guiding light inside an optical fibre.
Inputs
Result
- Critical angle θc 43.32 °
| Medium 1 (index n₁, denser) | 1.4580 |
| Medium 2 (index n₂, rarer) | 1.0003 |
| Ratio n₂/n₁ | 0.6861 |
| Critical angle θc = arcsin(n₂/n₁) | 43.32 ° |
| Beyond θc | Total internal reflection: all the light is reflected, none escapes. |
θc = arcsin(n₂ / n₁) (with n₁ > n₂)
Why θc only exists in one direction
Shine a torch from air into a swimming pool at any angle you like. A transmitted beam always appears. The surface never turns into a mirror on the way in.
Snell's law is n₁ sin θ₁ = n₂ sin θ₂. Head toward a denser medium (n₂ > n₁) and the factor n₁/n₂ is less than one, so sin θ₂ cannot reach 1. There is always a real θ₂. Reverse the trip — water into air, glass into air — and n₁/n₂ is greater than one. Past a certain incidence, the formula would need sin θ₂ > 1. No such angle exists. The refracted ray is gone. All of the light reflects. That threshold is the critical angle, and the one-way nature of the effect is not a footnote: it is the whole reason a fibre can trap a beam and a window, seen from the garden, never can.
The calculator simply refuses n₁ ≤ n₂. It is telling you the geometry has no θc in that direction, not that the sine function is broken.
Deriving θc = arcsin(n₂/n₁)
On the diagram, the transmitted ray is drawn grazing the interface. That is θ₂ = 90°, and sin 90° = 1. Drop that into Snell's law and the incident angle that remains is θc.
n₁ sin θc = n₂ · 1 → θc = arcsin(n₂ / n₁)
Arcsine is only defined for arguments up to 1, which is another way of saying n₁ must be the larger index. Equal indices give θc = 90°: you would have to skim the boundary to stay inside, which is not a useful trap. Water into air sits near 48.6°. Crown glass into air is about 41°. Diamond (n ≈ 2.42) into air is roughly 24°. A bigger step in n means a smaller θc, and a smaller θc means it is easier to keep light bouncing.
Below θc you still get a mix: some transmitted, some reflected (Fresnel). Above it the transmission channel is shut. “Total” is not a figure of speech.
Optical fibres: core, cladding, and a gap that must stay small
A fibre is two glasses, one nested in the other. The core has the higher index; the cladding sits a shade lower. A ray launched close to the axis hits that wall at a large angle of incidence — measured from the normal, so almost grazing — and if that angle is above θc the ray never leaks. Bounce, bounce, far end.
- The gap has to be real. Equal indices, no interface, no TIR: you have a glass rod, not a waveguide.
- It also has to stay small. The silica pair in the form (1.458 / 1.444) differs by 0.014. That is deliberate. A fat jump would accept a wide fan of rays, spawn many modes, and smear a pulse over kilometres. A tiny jump only keeps near-axial rays: few modes, sometimes one, and a bit that is still readable after a long haul.
Telecom silica lives in that narrow-gap world. Plastic lighting fibre, the kind that just has to look pretty on a ceiling, can afford a much larger Δn. Same physics, different job.
Numerical aperture: the light-gathering cone
Core/cladding θc is an inside story. Outside, at the cut end, the question is different: how steep a ray, arriving from air, will still be guided once it is inside? That half-angle is the acceptance angle. Its sine, in air, is the numerical aperture.
NA = √(n_core² − n_cladding²) · acceptance angle = arcsin(NA)
Plug in the defaults and NA comes out near 0.20 — a cone of about 12° half-angle. Miss that cone with a clumsily aligned laser, or flood it with a wide LED, and power is in the cladding before the fibre has started. A larger NA is easier to couple into and worse for modal dispersion. A smaller one is fussy at the launch and cleaner on the link. You do not “maximise” NA. You pick it for the regime you actually want.
The arcsin(NA) form assumes air at the input face. Index-matching gel changes the useful cone; NA itself is still a property of the two glasses.
Where TIR actually shows up
Once a ray is stuck, you can send it a long way, snake it around a corner, or bounce it until the stone looks expensive. Four jobs, one geometry.
- Fibre-optic telecoms — kilometres of silica, a tiny index step, losses of a fraction of a decibel per kilometre. Broadband in the ground is TIR in series, not a metaphor.
- Endoscopy — a bundle, each fibre carrying its own sliver of light. Some fibres light the scene, some bring the image back. The tube can bend: provided incidence stays above θc, the ray follows the bend.
- Diamond fire — n ≈ 2.42, so θc ≈ 24°. Light that enters the table rattles around the pavilion until it finds a face steep enough to escape. Brilliant-cut angles are not jewellery fashion; they are TIR engineering. Crown glass at 1.52 lets go too soon, which is why it looks dull beside a stone.
- TIR prisms — binoculars, periscopes, camera viewfinders. A 45° glass prism turns a beam without a silver coat, and with less loss than a metal mirror. More on the Porro layout below.
Frustrated total internal reflection
Past θc, no energy leaves — if medium 2 is thick. Right against the interface, though, there is a field that dies away in a fraction of a wavelength: the evanescent wave. It does not carry power off to infinity. It just sits there, stuck to the surface.
Bring a second dense medium within a few hundred nanometres and that tail has somewhere to go. Some light tunnels across the gap, even though geometric Snell forbids a refracted ray. That is frustrated TIR. The original total reflection did not spring a leak. You changed the boundary conditions.
ATR crystals in infrared spectroscopy, waveguide couplers, some fingerprint sensors, near-field microscopy: all of them farm this tail. “Total” means total for a half-space that goes on forever, not for a three-layer sandwich built on the scale of λ.
If you came here hoping the calculator would model the tunnelled fraction: it will not. That needs the gap thickness, the wavelength, and a bit of wave optics. Qualitatively, closer and thinner means more light sneaks through.
Critical angle vs Brewster's angle
Two “special” angles at a surface. Two formulae that look cousinly on a blackboard. They are not the same effect, and mixing them up is the classic lab howler.
θc = arcsin(n₂ / n₁) · θB = arctan(n₂ / n₁)
Brewster's angle is where the p-polarised part of the reflection vanishes. The bounce that remains is s-polarised. It exists in both directions, with two different values: glass into air around 33°, air into glass around 57°. Nothing disappears from the transmitted beam. You still have a refracted ray; you have merely cleaned the polarisation of the reflected one.
θc exists only for n₁ > n₂, and it shuts the door on all the light, not one polarisation. Arcsine and arctangent of the same ratio do not land on the same number. If a protocol says “the angle where the reflection goes away”, that is Brewster, and only for p-light. If it says “the angle where nothing gets out”, that is θc.
The Porro prism in binoculars
Take a 45-45-90 block of crown glass. Light comes in through the hypotenuse and hits a leg at 45°. Glass-to-air θc is about 41°. Forty-five is safely above that, so the face acts as a mirror with no silver on it. A metal coating would have eaten a few percent at each bounce. The prism does not.
A pair of Porro prisms, crossed, erects the image of a terrestrial telescope and folds the path — which is why classic binoculars have that offset “Z” shape. Compact roof-prism designs (Schmidt–Pechan, Abbe–König) play the same game in a tighter box. Some of their faces no longer sit above θc, so those get silvered, or a phase coating. The physics is unchanged: Snell's law decides whether the glass can do the job alone.
Drop the same Porro into water and θc climbs to about 61° (arcsin(1.333/1.52)). Forty-five degrees is no longer enough. The free mirror switches off. TIR is not a property of glass. It is a property of the pair.
Looking up from underwater
From the floor of a pool, the world above the surface does not fill the whole sky. It is packed into a bright disk — Snell's window — whose angular radius is just θc, about 48.6° from the vertical. Outside that cone the underside of the water is a mirror. You see tiles, the far wall, sometimes your own feet.
The disk is therefore roughly 97° across. Everything happening in air (the board, the sky, the lifeguard) is squeezed into it, with a strong fisheye crush toward the rim. You have not lost the outside world. You have the whole of it, shrunken, sitting at the zenith, ringed with reflections.
That is the same θc the calculator prints for water into air. Off the diving board, none of this happens: air into water has no critical angle, and the surface stays a window. One pool, two views, depending on which side of the interface you are standing on.