wunder beta💡 Fiber Optics: Information at the Speed of Light
The physics of the glass, not the map of the cables. Light in a thread wants to escape, to fade, and to spread — and every advance in fibre optics is the defeat of one of those three. Why an invisible
What you’ll learn
- Three Enemies in a Thread of GlassFrame the physics of optical fibre around three distinct enemies — escape, fade and spread — and grasp that the first two are power problems while the third is a timing problem, which is why it is the one still setting the ceiling.Light in a glass thread wants to escape through the side, to fade as it travels, and to spread out in time. Escape and fade are problems of power and both were comprehensively defeated. Spread is different in kind: a pulse can arrive at full brightness and still carry no information, because smeared ticks overlap and the receiver cannot tell one bit from the next. This course is about the glass rather than the network — the boundary, the windows, the amplifier — and it follows those three enemies to the end.
- Escape: the Step Is the Whole TrickExplain total internal reflection as the mechanism that traps light in a fibre, and understand that the core/cladding refractive index step — not the glass — is what actually guides.Light passing from slower glass into faster air bends away from the perpendicular; past a critical angle it cannot bend far enough to leave and reflects entirely back inside, with no loss mechanism at all. Colladon and Tyndall demonstrated this in 1841 with light following a jet of water. But bare glass only reflects totally where it touches air, and a real fibre is buried, taped and handled — so it must carry its own low-index outside with it. That is the cladding: not a protective coating but half the waveguide, with an index difference from the core often well under one percent, achieved by doping. Remove the step and the fibre stops guiding entirely.
- Spread I: Why the Fat Core DiedUnderstand modal dispersion as the reason wide cores failed, and see why shrinking the core to 8–10 µm abolishes rather than merely reduces it.In a wide (multimode) core, rays entering at different angles take paths of different lengths and therefore arrive at different times, smearing a sharp pulse. Nothing is lost — but the blur grows with distance, giving multimode fibre a bandwidth–distance product rather than a bandwidth. Graded-index cores partially fix this by letting outer rays run through lower-index, faster glass so they catch up. The radical fix is a wave-mechanical one: a waveguide supports only certain modes, and shrinking the core to about 8–10 µm leaves exactly one, so no second path exists to arrive late. The price — laser sources and micron-precision fusion splicing — was paid everywhere it mattered.
- Fade: the Twenty Decibel BetUnderstand why fibre loss looked like an insurmountable physics limit, and why Kao and Hockham's 1966 reframing of it as an impurity problem created the industry.Early-1960s optical glass lost roughly 99% of light in 20 metres — around 1,000 dB/km — and the consensus was that this was intrinsic to glass, which would have closed the field. In 1966 Charles Kao and George Hockham argued the loss was mostly transition-metal impurities at parts per million and that fused silica could reach below 20 dB/km: a physics limit ends a field, a purity problem is a to-do list. In 1970 Maurer, Keck and Schultz at Corning demonstrated 17 dB/km by doping silica with titanium, later reaching 4 dB/km — notably by starting from ultra-pure silica and adding dopants back rather than cleaning a dirty glass. Standard fibre today runs around 0.2–0.3 dB/km at 1550 nm, with Corning's low-loss product at 0.148 dB/km. Kao received the Nobel Prize in Physics in 2009.
- The Windows: Why 1550 WonExplain the attenuation windows as the outcome of two competing physical effects, and see why the loss minimum near 1550 nm is a floor set by the material.Rayleigh scattering from density fluctuations frozen into the glass falls as 1/λ⁴, pushing designers to longer wavelengths; infrared absorption by silicon–oxygen bonds rises steeply beyond about 1600 nm, pushing back. The minimum where they cross lands near 1550 nm, and traces of OH⁻ produce a water peak near 1383 nm that historically chopped the usable spectrum into separate windows. Typical figures: about 3 dB/km at 850 nm and 1 dB/km at 1300 nm in multimode, about 0.35 dB/km at 1310 nm and 0.25 dB/km or better at 1550 nm in single-mode. 1310 offers zero chromatic dispersion but higher loss; 1550 offers minimum loss but real dispersion — a genuinely contested trade until the EDFA settled it.
- The Amplifier That Refuses to Touch ElectricityUnderstand the erbium-doped fibre amplifier as the invention that made long-haul fibre possible, and grasp that its transparency to format — not merely its gain — is what changed the economics.Before the EDFA, long links needed O-E-O regenerators every ~50 km: high-speed electronics on the seabed, each engineered for one specific bit rate, so upgrading the transmitters obsoleted the whole chain. In 1987 David Payne's group at Southampton, with Emmanuel Desurvire's group at Bell Labs independently, demonstrated a low-noise high-gain erbium-doped fibre amplifier: a pump laser excites erbium ions in the core, and passing signal photons stimulate identical photons out of them. The signal never stops being light, so the amplifier cannot know the bit rate and boosts every colour in its band at once. Erbium's gain band, fixed by atomic physics at roughly 1530–1565 nm, happens to coincide with silica's loss minimum — a coincidence that decided that the internet would run at 1550 nm.
- Spread II: Dispersion Is the CeilingUnderstand chromatic dispersion as the return of the spreading enemy in single-mode fibre, why the loss minimum and dispersion zero sit at different wavelengths, and why the nonlinear limit is the real modern ceiling.Single-mode fibre abolishes modal dispersion but not spreading, because no modulated source is truly monochromatic — information is spectral width — and silica's index varies with wavelength, so a pulse's colours travel at different speeds. Material and waveguide dispersion cancel near 1310 nm, but the loss minimum is at 1550 nm, so the EDFA forced the world to accept real dispersion. Engineers answered with dispersion-shifted fibre (which proved bad for multi-colour systems, since equal speeds worsen interaction), dispersion-compensating fibre, and finally coherent detection with DSP that inverts the distortion arithmetically. What resists is polarization-mode dispersion, which drifts randomly with temperature and vibration, and the Kerr nonlinearity, where high power makes the glass's index depend on intensity — producing a nonlinear Shannon limit that the industry is now close to.
- Why Fibre Beat Copper (It Isn't Speed)Dismantle the belief that fibre beat copper on speed, and identify the real advantage: silica's loss is essentially independent of how fast you signal, while copper's rises steeply with frequency.Light in silica travels at about two-thirds of c — a refractive index of ~1.47 slows it — and a signal on good copper propagates at roughly 60–70% of c, so fibre is not faster, which is why intercontinental latency is a distance problem no technology fixes. Copper's real handicap is that the skin effect crowds high-frequency current into an ever-thinner surface layer, so attenuation rises steeply with frequency and bandwidth trades directly against reach. Silica's ~0.2 dB/km at 1550 nm is set by the carrier wavelength and stays essentially flat across a band tens of nanometres wide: the glass is equally transparent to a slow whisper and to a hundred gigabits a second. That indifference is the entire advantage.
Questions this course answers
A fibre link delivers a strong, bright signal to the receiver — plenty of power, well above the noise — but carries no usable data. Which enemy is at work?
This is the distinction the course turns on. Escape and fade are POWER problems — they make the signal dim. Spread is a TIME problem: no energy is lost at all, but each sharp tick arrives blurred, overlapping its neighbours, until the receiver cannot say where one bit ends and the next begins. That's inter-symbol interference, and it is entirely compatible with a beautifully strong signal. It's also why spread is the enemy still setting the ceiling today: you cannot fix a timing problem by shouting louder.
Why is a real optical fibre built as a core surrounded by cladding, rather than as a bare glass thread?
Total internal reflection at a glass–air boundary is genuinely perfect, which is why Tyndall's water jet works. But a working fibre is never surrounded by air: it's buried, taped, bundled and handled, and anywhere something touches a bare core the index step vanishes at that point and light escapes there. Cladding makes the fibre carry its own low-index outside, fused on permanently, so the reflecting boundary is deep inside where the world cannot touch it. The cladding isn't a coating — it's half the waveguide, and without the index step between the two glasses the fibre stops guiding entirely.
Why does making the core smaller — down to about 8–10 microns — eliminate modal dispersion rather than merely reduce it?
The ray picture suggests option B — less room to zigzag — but that would only reduce the problem. The honest answer is a wave answer: a waveguide supports only certain standing patterns, its modes, and shrinking the core removes them one at a time like a narrowing organ pipe losing its low notes. Below a certain size exactly one mode remains, so modal dispersion isn't reduced, it's abolished definitionally — the physics won't permit a second path to exist. The price is a core smaller than the dot on an i, which is why single-mode fibre needs lasers and fusion splicers. The industry paid it everywhere it mattered.
What made Kao and Hockham's 1966 argument so consequential?
Glass fibres had been proposed before, and total internal reflection had been demonstrated by Colladon and Tyndall in 1841. The received wisdom in the 1960s was that fibre's enormous loss was intrinsic — that scattering and absorbing light is simply what glass does — which would have closed the field permanently. Kao and Hockham argued the loss was mostly transition-metal impurities at parts per million, and that fused silica could get below 20 dB/km. That reframing is the whole thing: a physics limit ends a field, a purity problem is a to-do list. Corning reached 17 dB/km in 1970; Kao received the Nobel Prize in 2009.
Why is there a loss minimum near 1550 nm rather than the loss simply falling forever as wavelength increases?
Two effects pull in opposite directions. Rayleigh scattering — from density fluctuations frozen into the glass as it cooled, the same physics that makes the sky blue — falls as the fourth power of wavelength, so longer is better. But go too long and the photons start exciting the silicon–oxygen bonds themselves and vanish as heat, which climbs steeply past ~1600 nm. Between a falling curve and a rising one there is a minimum, and it lands near 1550 nm. Both curves are fixed by the material, so there is no fourth window with less loss and there never will be. The OH⁻ water peak at ~1383 nm is a contaminant that chopped the spectrum into islands, but it isn't what sets the floor.
Beyond avoiding electronics on the seabed, what was the EDFA's most consequential property?
A regenerator has to DECIDE what each bit was, so it must be built for a specific bit rate — invent a faster transmitter and every unit between you and your destination is obsolete, which meant upgrading the ends of a link required re-laying the middle. An EDFA just makes light brighter via stimulated emission; it never converts anything, so it cannot know or care what the bits are, and it boosts every colour in its band at once in the same piece of glass. That's what unfixed a cable's capacity from the moment it was laid, and what made wavelength-division multiplexing physically possible. It emphatically does NOT work at any wavelength: erbium's gain band is fixed by atomic physics at roughly 1530–1565 nm — which is why 1550 won.
Grounded in trusted sources
- Wikipedia — Optical fiber — https://en.wikipedia.org/wiki/Optical_fiber
- Nobel Prize in Physics 2009 — press release (Charles K. Kao) — https://www.nobelprize.org/prizes/physics/2009/press-release/
- IEEE Spectrum — How Charles Kao Beat Bell Labs to the Fiber-Optic Revolution — https://spectrum.ieee.org/how-charles-kao-beat-bell-labs-to-the-fiberoptic-revolution
- Fibermanialink — Fiber Optic Wavelengths Explained: 850 vs 1310 vs 1550 nm — https://www.fibermanialink.com/blog/why-fiber-optic-use-850nm-1310nm-1550nm/
- MapYourTech — Common Optical Wavelengths: 850nm, 1310nm, 1550nm — https://mapyourtech.com/common-optical-wavelengths-850nm-1310nm-1550nm/
- Eduard Rhein Stiftung — Invention of the erbium-doped fibre amplifier (EDFA) — https://www.eduard-rhein-stiftung.de/en/erfindung-des-erbium-dotierten-faserverstarkers-edfa/
- FiberLabs — Erbium-Doped Fiber Amplifier (EDFA) — https://www.fiberlabs.com/glossary/erbium-doped-fiber-amplifier/
- ScienceDirect Topics — Doped Fiber Amplifier — https://www.sciencedirect.com/topics/engineering/doped-fiber-amplifier
Every Wunder lesson is built from real, reputable sources — never invented.
Related Science courses
Wunder is a personalized learn-anything platform — tell it any topic and it builds a beautiful, fact-checked course in minutes, with narration, a knowledge check, and a college-style University track.
Browse more Science courses · All topics · Home
© 2026 Wunder Learning LLC · Terms & Privacy