A spectral atlasPlates I–V380–730 nm1.7–3.3 eV

Nothing
is coloured

Light has a wavelength. Matter has energy gaps. Neither of those things is a colour. Colour is what your nervous system makes of the gap between them — and it makes it out ofthree numbers.

K 393 Ca⁺G 431 FeF 486 Hb 518 MgE 527 FeD 589 NaC 656 HB 687 O₂
Sunlight, spread out. The dark bars are missing light — wavelengths swallowed on the way here by atoms whose electrons happened to have a gap of exactly that size. Fraunhofer counted 574 of them in 1814 without knowing what they were. Each one is a fingerprint of an element sitting between you and the sun.
I
In the atom
10−10 morbital radius

An electron can only owe round numbers

An electron bound to an atom cannot hold any amount of energy it likes. It can occupy this level, or that one, and nothing in between — the way you can stand on a stair but not halfway up one.

So when light passes through, most of it passes. A photon is only absorbed if its energy matches a gap exactly. Hit the match and the electron leaps a rung and the photon is gone. Miss by a little and nothing happens at all.

Photon energy and wavelength are the same fact stated twice:E = hc/λ, or in the units that make it legible,λ(nm) = 1240 / E(eV). Which means the entire visible world — every colour you have ever seen — lives inside asingle octave of energy, roughly 1.8 to3.1 electronvolts. Below it, infrared. Above, ultraviolet.

Hydrogen · the Balmer jumpschoose a landing
Transition
n = 3 → 2
Energy gap
1.89 eV
Wavelength
656 nm
Emitted light

Hydrogen's levels sit at En = −13.6/n² eV. Only four of its jumps land inside the visible octave — all of them ending onn = 2. That is the whole reason a hydrogen discharge tube glows pink: red at 656, cyan at 486, violet at 434, mixed. Two of these four are lines C and F in the solar plate above. Same atom, seen dark instead of bright.

Absorption and emission are one process run in two directions. An atom lit from behind subtracts its lines from the background — you see a dark bar. The same atom, excited and left alone in the dark, drops back down and hands the energy back as a photon — you see a bright line at exactly that wavelength. Sodium eats yellow from the sun at 589 nm and is why a street lamp is yellow.
II
In the material
10−9 mmolecule, lattice, particle

What you see is the leftovers

A lone atom has sharp lines. Pack atoms into a molecule or a crystal and the levels smear into broad bands — so a real material doesn't nibble a hairline out of the spectrum, it takes a bite.

The colour of a leaf, a carrot, a sapphire, a gold ring: in every case you are looking at the light that was not good enough to move an electron. Colour by subtraction. Pick a material and watch it eat.

Absorption chamberwhite light in · remainder out
Absorbs at
662 nm
Gap
1.87 eV
Mechanism
π → π* in the porphyrin ring
You see

Chlorophyll a takes two bites — deep blue and deep red — and hands back the middle. Green is not what the leaf is. Green is what the leaf refused.

Gold is the odd one. In a bulk metal the electrons form a continuous sea and reflect nearly everything, which is why most metals are grey. Gold'sinterband transition sits at about 2.4 eV, so it absorbs blue and violet and reflects the rest — a metal with a colour. Grind it to 20-nanometre particles and the electrons start sloshing collectively instead: theplasmon resonance lands near 520 nm, it eats green, and the gold turns ruby red. Medieval glaziers were doing quantum optics with a crucible and no idea.
III
In the eye
2 × 10−6 mcone diameter

The spectrum arrives, and is immediately thrown away

The leftover light reaches your retina still carrying its full spectral detail — an intensity for every wavelength, hundreds of independent numbers. Your eye replies with three.

There are three kinds of cone. Each holds a pigment whose own electrons have their own preferred gap, so each responds across a broad, overlapping hump. A cone cannot reportwhich wavelength arrived — only how hard it was hit. One photon of green and two dimmer photons of yellow-green produce the same shrug.

Cone response · drag the wavelength545 nm
L · long
0.84
M · medium
1.00
S · short
0.01
Reported as

Peaks near 448, 547 and 576 nm. Note how little separates M and L — about 29 nm — and that the so-called "red" cone actually peaks in yellow-green. Red is not a receptor. Red is a comparison, and it happens later. (Curves derived from the CIE 1931 standard observer.)

IV
In the nerve
10−2 mretina to cortex

Three numbers, subtracted from each other

Before the signal has even left the retina, the three cone values are recombined into three opponent channels — not sent onward as they are.

L + M becomes brightness. L − M becomes the red–green axis.S − (L + M) becomes the blue–yellow axis. It is a compression scheme, and a good one: neighbouring cone responses are highly correlated, so differencing them throws away redundancy and keeps the news.

It also explains the shape of colour experience. Why you can imagine a reddish-blue and a yellowish-green, but not a reddish-green — that would be one channel holding two opposite signs at once, which is not a thing the wiring can express. And why staring at red for thirty seconds leaves a green ghost: you fatigued one end of a see-saw.

Magenta · the colour with no wavelengthproof by absence

630 nm alone
L high · M low · S none

450 nm alone
S high · M low · L none

both at once
L high · S high · M low

No wavelength produces that third swatch. Search the whole spectrum and you will not find magenta in it — the two ends never meet.

Magenta is an invention. The pattern "L and S loud, M quiet" is one the spectrum can never generate on its own, so the brain has no wavelength to file it under. It issues a new label instead, and bends the line of the spectrum into a circle to close the gap. Magenta is the seam. It is as vivid as any other colour and it corresponds to nothing out there at all.
V
The proof
two spectraone colour

If colour were in the light, these would look different

Here are two beams. One is a broad, smooth hill of light. The other is three narrow spikes with nothing between them — physically almost nothing in common.

Their cone triples are identical, so your eye has no mechanism by which to tell them apart. They are the same colour. Not similar: the same, in the only sense the word has.

Metamerscomputed, not asserted
Beam A · L M S
0.436 0.439 0.125
Beam B · L M S
0.436 0.439 0.124
Perceived

Beam B's three spike heights were solved for, not chosen by eye: the weights that make its integral against the colour-matching functions equal beam A's. This is the whole basis of your screen. It cannot produce a smooth spectrum and never tries to — it produces metamers of one, three lights at a time.

Which also means colour is a lossy hash. Two objects can match perfectly under one light and split apart under another, because a different illuminant reweights the spectrum before your cones ever see it. Paint matched in the shop and wrong in daylight is not a manufacturing defect. It is the compression showing its seams.

So: an electron with a gap of 1.87 eV, a photon that fits it, a molecule that keeps the rest, three pigments that disagree about how much they were hit, two subtractions, and a name. Somewhere in that chain the light stopped and the colour started, and there is no line in the physics where it happened.The green is not in the leaf. It is not in the light either.