Material Insights · Understanding the Material World

Light, semiconductors and information: from the transistor to the network: Light, lenses and optical fibre

All courses / Understanding Engineering / Module 9: Light, semiconductors and information: from the transistor to the network Part 2 of 5

Light, lenses and optical fibre

Light is the part of the spectrum our eyes use, and it is also the part that glass handles best. This part shows how bending at a surface makes a lens, why a solar concentrator's lens cannot squeeze sunlight to a point, and how total internal reflection traps light in a hair-thin glass thread that carries far more data than copper can.

2.1 Light bends when it changes speed, and a curved surface turns that bending into a lens

Light slows inside matter; the refractive index n is its speed in a vacuum divided by its speed in the material. Water has 1.33, glass and acrylic about 1.5, diamond 2.4. When a beam crosses a surface at a slant, the side that enters first slows first and the beam swings towards the perpendicular, like a cart whose one wheel runs into mud before the other. Snell's law fixes the new direction: n times the sine of the angle from the perpendicular has the same value on both sides.

A lens is shaped so that this bending sends every ray from one point of an object to one point behind it. For a thin lens, 1/f = 1/u + 1/v, with u the object distance, v the image distance and f the focal length; its power in dioptres is 1/f, with f in metres.

Take the flat, grooved Fresnel lens used on some solar concentrators and solar cookers. It is a lens cut into rings and pressed flat into a thin acrylic sheet, here 300 mm across. With f = 0.5 m its power is +2 dioptres. The Sun is so far away that 1/u is effectively zero, and the image sits at v = f. The Sun's disc spans 0.53 degrees, so its image is 0.5 m x 0.0093 = 4.6 mm across, and the sunlight on the whole lens lands in that spot, about 4,000 times ordinary sunshine.

Lenses have two flaws. Glass and plastic bend blue more than red, so one lens brings colours to focus at slightly different distances; in this acrylic lens blue focuses about 9 mm before red. For collecting heat that hardly matters, so one cheap moulded lens will do; a sharp picture needs several elements of different glasses, at a cost in weight and price. The second is diffraction: a wave passing an opening spreads, so no lens makes a spot smaller than about 1.22 x wavelength x f-number. At f/4 and 1,550 nm that is 7.6 micrometres, close to the width of the light-carrying region of the fibre in the next section, so launching a laser into fibre needs micrometre alignment.

Data Estimate 2: a Fresnel lens on a solar concentrator
CaseWorkingResult
Lens power, f = 0.5 m1 / 0.5 m+2 dioptres
Image of the Sun, 1.5 x 10^11 m away1/v = 1/0.5 - 1/(1.5 x 10^11), per metrev = 0.500 m
Image of a lamp 2 m away1/v = 2 - 0.5 = 1.5 per metrev = 0.667 m, 167 mm further back
Width of the Sun's image0.5 m x 0.0093 radianabout 4.6 mm
Sunlight collected, 300 mm lens, 90 per cent passed1,000 W/m2 x 0.0707 m2 x 0.9about 64 W
Concentration(300 mm / 4.6 mm)^2about 4,000 suns
Colour spread, acrylic500 mm x 0.0084 / 0.49blue focuses about 9 mm before red
Smallest spot by diffraction, f/4 at 1,550 nm1.22 x 1.55 micrometres x 4about 7.6 micrometres

All values calculated. The Sun's disc spans 0.53 degrees, or 0.0093 radian. Acrylic's index is 1.4976 for blue and 1.4892 for red light. Real concentrators reach less than the ideal because of reflection, groove shading and tracking error. Site pages: 'Refraction and lenses', 'Concentrated solar power'.

Source: Bloomfield, How Things Work (Wiley, 2015), ch. 14, 'Optics and Electronics' (lens equation, f-number, diffraction); example and numbers computed by the author
  • Bloomfield, How Things Work (Wiley, 2015) · ch. 13, 'Light' (refraction, index of refraction, dispersion); ch. 14, 'Optics and Electronics' (lens equation, diffraction)

2.2 Past a critical angle no light escapes from glass, and a fibre uses that to bend light round corners

Reverse the direction, from a denser material into a less dense one, and the ray bends away from the perpendicular. Tilt it far enough, past the critical angle, and no light leaves at all: the surface sends all of it back. The critical angle obeys sin(angle) = n_outside / n_inside: 1.0 / 1.5 gives 42 degrees for glass against air, and 1.0 / 1.33 gives 49 degrees for water against air.

The fibre that a broadband installer brings into a flat uses this over tens of kilometres. Its core has a slightly higher index than the cladding around it. At 1,550 nm a typical pair is about 1.449 and 1.444, a difference of a third of one per cent. Then sin(angle) = 1.444 / 1.449 = 0.9965, the critical angle is 85.2 degrees, and any ray that stays within about 4.8 degrees of the fibre's axis is held in. The cladding is 125 micrometres across, the width of a hair.

Held in is not the same as arriving in order. In a wide core the steepest ray allowed would zigzag 1 / 0.9965 times, or 0.35 per cent, further than one running straight down the middle. Per kilometre that is 3.5 m of extra path. Light in the core moves at about 207,000 km a second, so the extra path costs 17 nanoseconds. At 10 Gb/s a bit lasts 0.1 ns, so after one kilometre each bit would be smeared over some 170 of its neighbours. Long links avoid this by shrinking the core to about 9 micrometres, only a few wavelengths, which leaves room for a single path: single-mode fibre. Wide-core multimode fibre survives across a building or a data-centre hall, a few hundred metres, where cheaper sources and connectors are worth the shorter reach.

  • Bloomfield, How Things Work (Wiley, 2015) · ch. 14, 'Optics and Electronics' (optical fibres, total internal reflection, single-mode fibre)
  • International Telecommunication Union, Recommendation ITU-T G.652, Characteristics of a single-mode optical fibre and cable (08/2024) · G.652.D cabled attenuation of at most 0.3 dB per km at 1,550 nm and 0.4 dB per km from 1,310 to 1,625 nm; https://www.itu.int/dms_pubrec/itu-t/rec/g/T-REC-G.652-202408-I!!TOC-HTM-E.htm
  • Prysmian, 'Single-mode optical fibre G.652.D' data sheet (read November 2026) · cabled attenuation at most 0.22 dB per km at 1,550 nm; mode field diameter 10.4 micrometres at 1,550 nm; cladding 125 micrometres; group index 1.468; https://ec.prysmian.com/sites/ec.prysmian.com/files/media/documents/G652D.pdf

2.3 A fibre carries millions of times more data over distance than copper because its carrier is so high and its loss so low

Compare the Category 6A twisted-pair cable behind an office wall with a single-mode fibre.

The first is the carrier. A signal can hold information only in proportion to the frequency range it occupies. Copper cable is used up to about 500 MHz. Light at 1,550 nm has a frequency of 193 THz, which is 390,000 times higher, and the usable window, 1,530 to 1,565 nm, covers about 4.4 THz of spectrum on its own. The second is loss. Current in copper meets resistance, and at high frequency it also crowds to the surface (skin effect) and leaks into the insulation, so loss climbs steeply with frequency: the TIA-568 limit for Category 6A cable at 500 MHz is 45.3 dB per 100 m. Fibre loses 0.2 dB per km. On a loss budget of 20 dB, copper reaches about 44 m and fibre 100 km, over two thousand times further. Third, light in one fibre does not leak into the next, whereas copper pairs pick up each other's signals.

Real long-haul systems send many wavelengths down one fibre, each from its own laser, and carry tens of terabits per second per fibre pair, the same order as the table's ceiling. A kilometre of four-pair Category 6A cable holds about 18 kg of copper; the glass in a kilometre of fibre weighs 27 grams.

Fibre gives up things too. It cannot carry power, a cut needs fusion splicing with trained crews, tight bends leak light, and each end needs a laser and a photodiode. That is why copper still wins for a metre or two inside a rack and for powering a phone, and fibre wins for everything beyond.

Data Estimate 3: copper twisted pair against single-mode fibre
QuantityCat 6A copperSingle-mode fibreRatio
Frequency range usedup to 500 MHz4.4 THz in one bandabout 9,000 times
Lossabout 450 dB per km0.2 dB per kmabout 2,300 times
Distance for 20 dB of lossabout 44 m100 kmabout 2,300 times
Rate in common use10 Gb/stens of Tb/s per fibre pairabout 2,000 or more
Rate x distance (bit-metres per second)10 Gb/s x 100 m = 10^1220 Tb/s x 100 km = 2 x 10^18about 2 million times
Upper bound from Shannon's formula (Part 5)not needed4.4 THz x 6.7 bits/s/Hz x 2 polarisations = 58 Tb/s-
Conductor or glass per kmabout 18 kg copper (8 wires of 0.57 mm, 23 AWG)about 27 g silica (125 micrometre thread)about 700 times

All ratios calculated. Copper loss is the TIA-568 limit for Category 6A cable, 45.3 dB per 100 m at 500 MHz, as printed on a maker's data sheet; real cable usually does somewhat better. Copper mass: 8 x 0.258 mm2 x 1 km x 8,960 kg/m3. Glass mass: 0.0123 mm2 x 1 km x 2,200 kg/m3. The Shannon bound assumes 20 dB signal-to-noise, an assumption; fibre non-linearity pulls real systems below it.

Source: International Telecommunication Union, Recommendation ITU-T G.652 (08/2024); trueCABLE, 'Cat6A Augmented CMP 4PR U/UTP 23AWG' data sheet (read November 2026), fibre attenuation near 1,550 nm; Category 6A insertion-loss limit at 500 MHz and 23 AWG conductors, https://cdn.shopify.com/s/files/1/0014/6404/1539/files/Cat6A-Unshielded-Plenum.pdf; other numbers computed by the author

See it in the table

H₂O WaterCu Copper

Check your understanding

A Fresnel lens with a focal length of 0.5 m faces the Sun, whose disc spans 0.0093 radian. About how wide is the Sun's image?

Image width = focal length x angle = 0.5 m x 0.0093 = 4.6 mm. A source of finite size cannot be focused to a point, however good the lens.

In a fibre the core index is 1.449 and the cladding index 1.444. What happens to a ray that meets the boundary at 88 degrees to the perpendicular?

The critical angle is about 85.2 degrees. A ray meeting the boundary at a larger angle to the perpendicular, closer to grazing, cannot leave the core.

A fibre loses 0.2 dB per km. How much of the light remains after 50 km?

50 km x 0.2 dB = 10 dB, and 10 dB is a factor of ten, so 10 per cent remains. After 100 km, 20 dB, only 1 per cent remains.
← Part 1: Electromagnetic waves, the spectrum and radioMark read · Part 3: Photons and semiconductors: diodes, solar cells and LEDs →