How Fiber-Optic Cables Carry Light Across Oceans

The Invisible Glass Backbone of Civilization

We routinely talk about "the cloud" and "wireless networks" as if global data floats weightlessly through the sky. In reality, wireless signals travel only the final few hundred meters between your device and the nearest cell tower or Wi-Fi router (how Wi-Fi works). Once your request reaches that local base station, it immediately drops underground into glass fibers.

When a message, financial transaction, or video stream crosses an ocean, roughly 99% of all intercontinental internet traffic travels not through satellites in orbit, but through approximately 550 active submarine fiber-optic cables totaling over 1.4 million kilometers laid directly across the seabed (how the internet works).

Why haven't satellites replaced undersea cables? The answer comes down to waveguide capacity, electromagnetic spectrum limits, and the speed of light:

• Bandwidth Density: Free-space radio and microwave satellite links share a single open atmospheric medium subject to rain fade, beam divergence, and strict frequency licensing (how electromagnetism unifies nature). Inside a fiber-optic cable, every strand of glass is a sealed, interference-free electromagnetic waveguide. A single modern transoceanic cable containing 16 to 24 pairs of glass fibers—each no thicker than a human hair—can carry 200 to 500 terabits per second (Tbps), roughly equivalent to the combined capacity of several thousand communications satellites.
• Latency and Orbital Geometry: A geostationary satellite orbits at 35,786 km above the equator, imposing a minimum round-trip propagation delay of ∼ 477 ms just to go up to space and back twice. Low-Earth Orbit (LEO) constellations orbit much closer (550 km), yet a single laser pulse traveling through an undersea cable between New York and London (5,500 km) completes the crossing in roughly 27.5 milliseconds with deterministic reliability and zero weather jitter.

Yet sending light across 6,000 km of the Atlantic or 10,000 km of the Pacific Ocean poses a staggering physics problem. Ordinary window glass absorbs and scatters more than 90% of light passing through a pane just one meter thick. How can engineers trap light inside a flexible glass thread so thin it bends around a pencil, guide it over underwater mountain ranges and abyssal plains 5,000 meters deep, and amplify the signal every 70 kilometers without ever converting the optical data back into slow electronic bits?

1. Snell's Law and Total Internal Reflection

Light normally travels in straight lines. To make light follow a curving cable over thousands of kilometers, an optical fiber exploits a fundamental boundary phenomenon of electromagnetism: refraction and Total Internal Reflection (TIR).

In a vacuum, light travels at c = 299,792,458 m/s. When an electromagnetic wave enters a transparent dielectric material such as silica glass (SiO₂), its oscillating electric field polarizes the electron clouds of the silicon and oxygen atoms. The re-radiated secondary waves interfere with the primary wave, retarding the phase velocity of the light by a factor known as the refractive index (n):

In ultra-pure telecommunications silica glass at infrared wavelengths, the refractive index is approximately n ≈ 1.444 to 1.449. Consequently, light inside a fiber-optic cable does not travel at 300,000 km/s—it travels at roughly 204,000 km/s (about 68% of its vacuum speed, or 4.9 microseconds per kilometer).

The Two-Layer Glass Waveguide

An optical fiber is not a hollow pipe with mirrors on the walls; metallic mirrors absorb a small percentage of light on every bounce, which would extinguish a beam after just a few meters. Instead, an optical fiber consists of two concentric cylinders of solid silica glass fused together with slightly different refractive indices:

1. The Core (n_core ≈ 1.449): The innermost central cylinder through which the light travels. During manufacturing, the silica glass of the core is doped with a tiny concentration of germanium dioxide (GeO₂), which slightly raises its refractive index.
2. The Cladding (n_clad ≈ 1.444): A surrounding jacket of pure silica glass (or fluorine-doped silica) with an outer diameter of exactly 125 micrometers (μm). Because n_clad < n_core by roughly 0.35%, light travels slightly faster in the outer cladding than in the inner core.

==========================================================================
TOTAL INTERNAL REFLECTION AT THE CORE–CLADDING BOUNDARY
==========================================================================

   Outer Silica Cladding (n_clad = 1.444, diameter = 125 µm)
  ------------------------------------------------------------------------
   Inner GeO2-SiO2 Core  (n_core = 1.449, diameter = 9 µm in single-mode)
          \                     / \                     /
           \  θ_1 > 85.2°      /   \  100% Reflection  /
  Normal    \     |           /     \     |           /
    |        \    |          /       \    |          /
  ------------+---+---------+---------+---+---------+---------------------
   Outer Silica Cladding (n_clad = 1.444)
==========================================================================

When a ray of light traveling inside the higher-index core strikes the boundary with the lower-index cladding at an angle θ₁ (measured from the perpendicular normal to the boundary), Snell's Law governs the angle θ₂ of the refracted ray in the cladding:

Because n_core > n_clad, the refracted angle θ₂ is always larger than the incident angle θ₁—the ray bends away from the normal toward the boundary surface. As the incident angle θ₁ increases, θ₂ eventually reaches 90^∘ (sin 90^∘ = 1), skimming right along the interface. This threshold angle is the critical angle (θ_c):

Whenever a pulse of light travels nearly parallel to the fiber axis and strikes the core-cladding wall at an incident angle θ₁ > 85.2^∘ (a grazing angle shallower than 4.8^∘ relative to the wall), sinθ₂ would have to exceed 1.0, which is mathematically impossible for a propagating wave in the cladding. No energy can escape into the cladding: the boundary acts as a perfect, 100% lossless mirror via Total Internal Reflection. Even if the glass fiber curves gently around the contours of the seabed, the trapped light continuously ricochets down the core.

The range of angles over which the fiber end-face can accept incoming laser light and trap it by total internal reflection is quantified by its Numerical Aperture (NA):

2. Single-Mode vs. Multi-Mode: Eliminating Modal Dispersion

If trapping light inside glass is as simple as making n_core > n_clad, why not make the glass core thick—say, 50 or 100 micrometers wide—so it is easy to align a laser with it?

Early fiber-optic systems did exactly that. Fibers with a large core diameter (50 μm or 62.5 μm) are called Multi-Mode Fibers (MMF). While they work well inside a single data center building over 100 to 300 meters, they fail catastrophically across an ocean due to modal dispersion.

In a wide 50 μm core, a single sharp pulse of light launched into the fiber splits into hundreds of distinct electromagnetic paths (or transverse waveguide modes):

• A ray traveling dead-center down the axis of the core takes the shortest physical path (L).
• A ray bouncing back and forth at the steepest allowed grazing angle near θ_c = 85.2^∘ travels a longer zigzag path (L / sinθ_c ≈ 1.0035 L).

Over a short 100 m patch cord, a 0.35% path-length difference is negligible. Over a 6,000 km transatlantic crossing, the zigzag rays travel 21 kilometers farther than the axial ray! They arrive more than 100 microseconds later, smearing crisp 10-picosecond digital pulses into an unrecognizable blur where ones and zeros overlap completely.

The 9-Micrometer Single-Mode Solution

To eliminate modal dispersion altogether, long-haul terrestrial and submarine cables exclusively use Single-Mode Fiber (SMF) (standardized under ITU-T G.652 and G.654).

In single-mode fiber, engineers shrink the germanium-doped core diameter down to just 8 to 10 micrometers (μm)—only about six times wider than the 1.55 μm wavelength of the infrared light itself. At this microscopic scale, geometric ray optics gives way to Maxwell's wave equations inside a cylindrical dielectric waveguide.

==========================================================================
MULTI-MODE vs. SINGLE-MODE FIBER WAVEGUIDE GEOMETRY
==========================================================================

1. MULTI-MODE FIBER (50 µm Core / 125 µm Cladding)
   +------------------------------------------------------------------+
   | Cladding (125 µm)                                                |
   |   ============================================================   |
   |   Core (50 µm):  \    / \    / \    / \    /  (High-order mode)  |
   |   --------------> \--/---\/---\/---\/--/----> (Axial mode)       |
   |   ============================================================   |
   +------------------------------------------------------------------+
   Result: Different path lengths -> Severe pulse spreading over >500 m.

2. SINGLE-MODE FIBER (9 µm Core / 125 µm Cladding)
   +------------------------------------------------------------------+
   | Cladding (125 µm)                                                |
   |   ------------------------------------------------------------   |
   |   Core (9 µm):   ===================>  (Single LP_01 Mode Only)  |
   |   ------------------------------------------------------------   |
   +------------------------------------------------------------------+
   Result: Zero intermodal dispersion -> Transoceanic 10,000+ km reach.
==========================================================================

A cylindrical waveguide's ability to support multiple modes is governed by its normalized frequency parameter (the V-number):

where a is the core radius (4.5 μm) and λ is the vacuum wavelength (1.55 μm). Whenever V < 2.405 (the first root of the Bessel function J₀), all higher-order zigzag modes are physically cut off and radiate out of the core. Only a single fundamental electromagnetic mode—the LP₀₁ (HE₁₁) mode, featuring a smooth bell-shaped Gaussian intensity profile centered on the core—can propagate. Because every photon in a given wavelength channel travels in the exact same spatial mode at the same group velocity, intermodal dispersion drops to zero.

3. Why 1550 Nanometers? The Physics of Ultra-Pure Silica

Even with zero modal dispersion and 100% lossless Total Internal Reflection at the core boundary, light traveling through the bulk silica glass still loses intensity over distance. This loss is measured in decibels per kilometer (dB/km):

Because the decibel scale is logarithmic, every 3 dB of attenuation cuts the optical power in half; 10 dB reduces it to 10%; 20 dB reduces it to 1%; and 30 dB reduces it to 0.1%.

In 1966, standard optical glass had an attenuation of roughly 1,000 dB/km—meaning that after just 100 meters, the light power dropped by a factor of 10⁻¹⁰, leaving not a single photon detectable. That year, physicist Charles K. Kao and George Hockham published a landmark paper proving that this extreme opacity was not an intrinsic property of silica, but the result of trace metallic impurities (just a few parts per million of iron, copper, and chromium ions) and dissolved water (OH⁻ hydroxyl ions). By 1970, Robert Maurer, Donald Keck, and Peter Schultz at Corning Glass Works invented chemical vapor deposition techniques to synthesize fused silica purity exceeding 99.9999999%, shattering the 20 dB/km barrier and earning Kao the 2009 Nobel Prize in Physics.

Once transition-metal impurities and OH⁻ water peaks are eliminated, pure silica glass is bounded by two fundamental, inescapable physical walls that pinch together to form a narrow window of supreme transparency:

1. At Shorter Wavelengths — Rayleigh Scattering (∝ λ⁻⁴): Because glass is an amorphous solid frozen as it cooled from a liquid, it contains sub-microscopic density and compositional fluctuations smaller than the wavelength of light. Just as atmospheric molecules scatter short blue wavelengths of sunlight to make the sky blue, these frozen density fluctuations scatter light inside the glass with an intensity inversely proportional to the fourth power of wavelength (α_Rayleigh ∝ λ⁻⁴). Moving from visible red light (650 nm) into the near-infrared at 1550 nm increases λ by a factor of 2.38, slashing Rayleigh scattering by (2.38)⁴ ≈ 32-fold.
2. At Longer Wavelengths — Infrared Multiphonon Absorption (λ > 1650 nm): Beyond 1600 nm, the photon energy approaches the natural vibrational resonance frequencies of the covalent silicon–oxygen (Si–O) chemical bonds in the glass lattice. The glass begins absorbing infrared photons directly as heat, causing attenuation to skyrocket above 1650 nm.

Between the downward slope of Rayleigh scattering (λ⁻⁴) and the upward wall of Si–O infrared bond absorption sits the global minimum of silica attenuation: the Telecommunications C-Band (Conventional Band) centered around λ = 1550 nanometers (1530–1565 nm).

At 1550 nm, modern submarine single-mode fiber (ITU-T G.654.E with a pure-silica core) achieves an extraordinary attenuation of just 0.15 to 0.18 dB/km. After traveling through 18 kilometers of solid glass, half of the original laser photons are still going strong!

4. Erbium-Doped Fiber Amplifiers (EDFAs): Boosting Light with Light

Even at 0.16 dB/km, crossing the 10,000 km Pacific Ocean without amplification would subject a laser beam to 1,600 dB of total loss—an attenuation factor of 10⁻¹⁶⁰. Since the observable universe contains only about 10⁸⁰ atoms, not a single photon would ever reach the other shore. The signal must be amplified every 60 to 80 kilometers along the ocean floor.

In the first generation of submarine fiber cables laid in the late 1980s (such as TAT-8 across the Atlantic), every undersea repeater had to perform an Optical-Electrical-Optical (O-E-O) conversion: a photodiode converted the incoming faint light pulses into an electrical current, electronic transistors amplified and retimed the bits, and a new semiconductor laser fired fresh light into the next fiber segment. O-E-O repeaters had a fatal bottleneck: they were locked to a single bit rate and a single laser wavelength. Upgrading the speed of a cable meant hauling 6,000 km of cable off the seabed to replace every electronic chip.

In 1987, David Payne at the University of Southampton and Emmanuel Desurvire at Bell Labs invented a device that revolutionized global communications: the Erbium-Doped Fiber Amplifier (EDFA).

How Quantum Stimulated Emission Amplifies Data Purely as Light

An EDFA contains zero high-speed data electronics. Instead, inside each seabed repeater housing, engineers splice a short 10-to-30-meter coil of silica fiber whose core has been doped with trace amounts of a rare-earth lanthanide element: trivalent Erbium ions (Er³⁺).

By a remarkable coincidence of atomic physics, the quantum energy gap between the first excited metastable state (⁴I_13/2) and the ground state (⁴I_15/2) of an Er³⁺ ion embedded in silica glass corresponds precisely to photons with wavelengths between 1530 nm and 1565 nm—the exact C-band transparency window of silica fiber!

==========================================================================
ERBIUM-DOPED FIBER AMPLIFIER (EDFA) QUANTUM MECHANISM
==========================================================================

1. PUMPING (Population Inversion):
   Continuous 980 nm Pump Laser ---> Excites Er3+ ions in a 20 m coil
                                     into the metastable 4I_13/2 level
                                     (lifetime ~ 10 milliseconds).

2. STIMULATED EMISSION (Optical Amplification):
   Weak 1550 nm Data Photon arrives ---> Triggers excited Er3+ ion to decay,
                                         emitting a SECOND 1550 nm photon
                                         with IDENTICAL wavelength, phase,
                                         polarization, and direction!
==========================================================================

Here is how an EDFA amplifies a faded transoceanic signal in flight:

1. Optical Pumping (980 nm): Inside the repeater, a continuous-wave semiconductor pump laser operating at 980 nm (or 1480 nm) shines steady light into the Erbium-doped fiber coil via a wavelength-selective coupler. The Er³⁺ ions absorb these 980 nm pump photons and jump into an excited energy state (⁴I_13/2) that has an unusually long quantum lifetime of ∼ 10 milliseconds, creating a population inversion where most Erbium ions sit cocked like loaded springs.
2. Stimulated Emission (1550 nm): When the faint, attenuated 1550 nm data pulses arrive from the previous 70 km of ocean cable and pass through the Erbium coil, their oscillating electromagnetic fields trigger stimulated emission—the exact quantum mechanism Albert Einstein discovered for lasers. Each incoming 1550 nm signal photon stimulates excited Er³⁺ ions to drop back to their ground state and release cloned 1550 nm photons with the exact same wavelength, phase, polarization, and travel direction.
3. All-Optical Gain (+15 to +20 dB): Within twenty meters of Erbium-doped glass, the signal power is boosted by a factor of 30× to 100×, completely erasing the loss of the preceding 70 km span without ever slowing the light down or converting it into electricity.

5. DWDM and Coherent Optics: Packing 100+ Laser Colors Into One Hair

Because an Erbium-Doped Fiber Amplifier amplifies the entire 1530 nm to 1565 nm C-band simultaneously regardless of modulation format, engineers realized they did not have to send just one laser beam down a fiber.

Using Dense Wavelength-Division Multiplexing (DWDM), a shore terminal combines 64 to 128 independent laser beams, each tuned to a slightly different infrared wavelength (color) spaced just 50 GHz (0.4 nm) or 37.5 GHz (0.3 nm) apart on the ITU-T frequency grid. Because light waves of different frequencies do not interfere linearly inside glass, all 100+ laser channels travel down the same 9 μm core simultaneously—like a hundred parallel freight trains sharing the same track without colliding—and every EDFA on the ocean floor boosts all 100+ channels at once.

Furthermore, modern submarine terminals no longer blink lasers simply ON and OFF (On-Off Keying). Instead, they use Coherent Optical Communications powered by Application-Specific Integrated Circuits (ASICs) (how semiconductors and transistors work):

1. Quadrature Amplitude Modulation (Dual-Polarization 16-QAM / 64-QAM): The transmitter encodes multiple bits per symbol by modulating not just the intensity of the light, but its in-phase and quadrature optical phase across two orthogonal polarization axes (horizontal and vertical electric field orientations), transmitting 400 Gbps to 800 Gbps on each individual wavelength.
2. Digital Chromatic Dispersion Compensation: Even inside a single-mode fiber, slightly different frequencies within a single laser pulse travel at microscopically different speeds (chromatic dispersion), and slight core asymmetries split the two polarizations (polarization-mode dispersion). Because a coherent receiver mixes the incoming signal with a local oscillator laser to measure the full complex electric field (amplitude, phase, and polarization), its Digital Signal Processor (DSP) passes the digitized waveform through an inverse mathematical filter—digitally undoing 10,000 km of pulse distortion in real time without needing dispersion-compensating fibers on the seafloor.
3. Space-Division Multiplexing (SDM): Modern high-capacity cables (such as Google's Dunant and Grace Hopper or Meta's 2Africa and Anjana) scale capacity further by packing 12 to 24 fiber pairs inside the cable and sharing pump lasers across multiple EDFAs, reaching total cable capacities of 250 to 500 Tbps.

6. Anatomy of a Submarine Cable and High-Voltage Power Feeding

People often imagine an undersea internet cable as a massive steel pipe meters wide. In fact, across 85% of its transoceanic route on the deep abyssal plain (3,000 to 6,000 meters underwater), a deep-sea cable is only 17 to 21 millimeters in diameter—about the thickness of a standard garden hose or a marker pen!

Why so thin? In the pitch-black abyssal ocean, the seabed is cold (1–2^∘C), calm, and free of ship anchors or fishing trawlers. Making the deep-sea cable lightweight and narrow is essential so that when a cable ship lays or retrieves it across a 5 km vertical water column, the hanging cable does not snap under its own suspended weight. Conversely, on the shallow continental shelf (depths less than 1,000 to 1,500 meters), where commercial fishing trawlers and ship anchors scrape the bottom, the cable is wrapped in single or double layers of heavy galvanized steel armor wires, swelling to 40 to 50 mm in diameter, and buried 1 to 3 meters beneath the seabed by a high-pressure sea plow.

Powering 150 Seabed Repeaters Across 10,000 Kilometers

Every 60 to 80 km along the cable sits a torpedo-shaped beryllium-copper or titanium pressure vessel about 3 meters long and weighing 500 kg: the Submarine Repeater, housing the 980 nm pump lasers and Erbium-doped fiber coils. How do you supply electrical power to 150 repeaters stretched across the Pacific Ocean?

You cannot run a parallel AC power cord: over 10,000 km, alternating current would suffer massive capacitive charging currents across the polyethylene insulation (how electrical transformers step voltage up and down). Instead, every repeater on the ocean floor is wired in series along the cable's single tubular copper conductor, powered by Power Feed Equipment (PFE) inside the shore Cable Landing Stations (CLS):

• Constant Direct Current (∼ 1.0 A): The PFE regulates a strictly constant DC current of roughly 0.8 to 1.2 Amperes flowing through the copper tube. Inside each repeater, the 980 nm pump lasers are wired in series with Zener-diode bypass protection, dropping roughly 35 to 50 V per repeater (P = IV ≈ 40–50 W per repeater).
• Bipolar ± 10 kV to ± 15 kV Feed with Ocean-Earth Return: Over 10,000 km, the copper tube has a total resistance of roughly 0.7 Ω/km × 10,000 km = 7,000 Ω, plus 150 × 45 V = 6,750 V across the repeaters—requiring nearly 14,000 Volts of total driving potential (V = IR + V_repeaters). To minimize voltage stress on the polyethylene insulation, the western shore station feeds +7,000 V DC and the eastern shore station feeds −7,000 V DC, placing a virtual 0 V ground point in the middle of the ocean. The return path for this current is the conductive saltwater of the ocean itself, injected via huge titanium or copper ocean-ground beds buried at each beach! If one shore station loses power during a disaster, the opposite shore station can ramp up to the full 14 kV to keep the entire cable alive single-ended.

7. When a Cable Breaks: OTDR and Deep-Sea Repair Ships

Roughly 150 to 200 submarine cable faults occur globally every year. Contrary to popular belief, shark bites account for almost zero modern outages (modern armored and HDPE-jacketed cables easily resist bites). Instead, over 70% of all cable cuts are caused by human maritime activity—commercial fishing vessels dragging bottom-trawling nets or cargo ships accidentally dropping multi-ton anchors in shallow coastal zones—while the remainder come from undersea earthquakes, volcanic landslides, and abrasive abyssal turbidity currents (underwater avalanches of mud and sand racing down continental canyons).

When a cable is severed, global traffic rarely goes dark because Border Gateway Protocol (BGP) routers automatically reroute packets onto alternate cables within milliseconds (how packets travel across the internet). Meanwhile, a specialized cable repair ship is dispatched to splice the glass back together:

1. Pinpointing the Break from Shore via OTDR: Engineers do not have to guess where in a 10,000 km ocean the cable snapped. If the copper tube is exposed to seawater, shore technicians measure the DC electrical resistance to the seawater short (R = ρ L / A). For optical breaks, engineers use Coherent Optical Time-Domain Reflectometry (C-OTDR): the shore terminal fires a test laser pulse down the fiber and measures the tiny fraction of Rayleigh-backscattered light returning from each repeater span, timing the exact microsecond the reflection stops (Δ d = v · Δ t2). This pinpoints the break to within a few hundred meters across an entire ocean.
2. Grapnel Recovery on the Seabed: The repair ship sails to the GPS coordinates and lowers a specialized flatfish grapnel on a multi-kilometer steel tow wire. In shallow water, a Remotely Operated Vehicle (ROV) uses electromagnetic tone detectors to locate the buried cable, jetting away the mud. In deep water (4,000+ m), the ship drags a cutting-and-holding grapnel perpendicular to the cable route across the seabed until tension gauges spike, hooking the severed end and winching it slowly to the ship's deck.
3. Microscopic Fusion Splicing at Sea: On board the ship's cleanroom laboratory, technicians strip back the steel armor, copper tube, and stainless steel vault to expose the hair-thin 125 μm glass fibers. Using a robotic electric-arc fusion splicer with microscopic camera alignment, they align the 9 μm cores of the old cable and a new replacement spool to sub-micrometer precision and fuse them with a 2,000^∘C electric arc, achieving a splice loss under 0.03 dB. After X-raying the molded polyethylene insulation joint to verify zero air bubbles that could short out at 15 kV under 500 atmospheres of pressure, the ship winches up the second severed end, splices it onto the other side of the spare cable loop, and lowers the repaired omega-loop back to the ocean floor.