There Is No Magenta

colour science  ·  the one hue your eye builds to close an open circle

Split sunlight with a prism and you get a strip — red, orange, yellow, green, blue, violet — laid end to end, never meeting. There is no magenta in it. No purple band, no pink line. Magenta is the colour physics forgot to make, and your visual system quietly invents to join the two ends of the strip into a wheel.

Every pure colour — every single wavelength of light — lands somewhere on one curve, drawn below straight from the CIE 1931 colour-matching data. The curve is the spectral locus, the famous horseshoe. The thing nobody tells you about it is the most important thing about it: it does not close. It is an arc, open at the bottom. The straight line that bridges the gap — the line of purples — passes through magenta, purple, hot pink, every rose and fuchsia you have ever seen. And there is no wavelength anywhere on it. Those colours correspond to no light. They are the eye's solution to a problem the spectrum poses and cannot answer.

Drag the wavelength below and watch a single pure colour walk the rim of the horseshoe. Then switch to mix two lights, pull one marker to the red end and one to the violet end, and watch the result fall onto the open chord — into colours the rim never reaches. That is the only way to make magenta: you cannot emit it, you can only add the two ends of the spectrum and let your eye do the rest.

Instrument I · the horseshoetrace a wavelength
spectral locus (pure wavelengths) line of purples (no light)
Each point on the rim is one wavelength of light. The rim never closes — the magenta seam at the bottom is bridged by a straight line that no wavelength touches.
colour here wavelength575 nm chromaticity (x, y) is it a wavelength?
380 nm — violet700 nm — red
A single wavelength is always a colour on the rim.

Two facts the machine just proved for you, both checkable below. One: sweep the whole spectrum, 380 to 700 nanometres, and you never once land in the magenta region — the rim simply does not go there. Two: the only route into that region is the mixer, and the colour it makes sits on the straight chord, exactly where no wavelength is. Magenta is real — you see it on this page — but it is a perception, not a frequency. It is what your brain returns when the long-wave cones and short-wave cones both fire and the middle stays quiet, a combination no single beam of light can produce, because a single wavelength always wakes the middle cone too.

The rainbow is a line. The colour wheel is a lie your eye tells to make the line meet itself.

II · Two lights, one colour

If magenta exposes how the eye adds a colour the spectrum can't make, metamerism exposes the deeper economy underneath: your eye reports light through only three numbers — how hard each of the three cone types is driven. Infinitely many different spectra collapse to the same three numbers, and once they do, they are the same colour, indistinguishable, forever. The tomato under a tungsten bulb and the tomato in a photograph of a tomato send your eye wildly different spectra and the identical sensation.

Below are two real spectra — two different distributions of power across wavelength, the left one and the right one provably not equal. Drag the slider to drive them apart as far as you like. The swatch is the colour each one makes. It never changes, and the three tristimulus numbers stay locked to the last digit, because the difference between the two spectra is a metameric black: a ghost of light that is mathematically invisible to the three cones. Different spectra. One colour. No exception.

Instrument II · metamerismtwo spectra, one colour
spectrum S+ spectrum S
Two physically different lights. The slider drives them apart; the colour they make does not move.
colour of S+ colour of S tristimulus X · Y · Z ‖S+ − S‖ (sup) ΔXYZ between them
The gap between the spectra is a metameric black — orthogonal to all three cone responses, so it changes the light without changing the colour.

III · The check

Everything above is drawn from one embedded table: the CIE 1931 2° standard-observer colour-matching functions, the empirical curves (measured on real observers in the 1920s, standardised in 1931) that say how strongly each wavelength drives the three channels of human colour vision. The page recomputes its own claims from that table when it loads. The same checks run offline in research/there-is-no-magenta/verify.mjs (18/18).

live · recomputed in your browser from the CIE 1931 table
running…

IV · Why the line is open

The horseshoe is open for a concrete physiological reason. Your three cone types — call them L, M, S, sensitive to long, medium, short wavelengths — have overlapping response curves. There is no wavelength that excites L without also touching M; the medium cone sits right under the long one across the whole red end. So no light can produce the signal "much L, much S, almost no M" — and that signal is exactly what your brain labels magenta. The only way to get it is two beams at once: a long one for L and a short one for S, with their contributions to M cancelling in the middle. Magenta is a chord, never a note.

This is why Newton, laying the spectrum out in 1704, had to bend it to build his colour circle — he closed the open strip by hand, joining red to violet through a band of "purple" he knew was not in the prism's output. And it is why the natural geometry of human colour is a wheel with one seam: the seam is the line of purples, the only stretch of the colour wheel made of no colour of light.

There is a quieter consequence on the same diagram. The triangle below the rim — toggle show screen gamut in the first instrument — is everything an sRGB screen, the standard your display almost certainly uses, can physically produce. It covers about 33% of the chromaticities inside the horseshoe. The deep spectral greens and cyans, and the most saturated of the very purples this page is about, fall outside it. So the magenta you are looking at is doubly removed from light: a hue no wavelength makes, shown to you through a screen that cannot even reach the saturated end of it. We render the out-of-reach colours as their nearest in-gamut approximation — there is no honest alternative on this hardware, and that limit is the point.

what's exact, and what's a standard The geometry is exact: the spectral locus, the open chord, the metameric blacks, and the area fraction are all recomputed from the embedded table to the digit, offline and live. What's conventional is the table itself — the CIE 1931 2° observer is a standard average human, not a law of physics; real observers vary, the 1964 10° observer differs slightly, and the rendered colours are clipped to your screen's gamut. "Magenta is not a wavelength" is a fact about light and is unconditional; the precise shape of the horseshoe is a fact about a standardised eye. Both are stated as what they are. The area fraction is an area in the (perceptually non-uniform) 1931 xy plane — it is the honest "how much of the diagram," not a claim about counted perceivable colours.

V · What you can point at

So: there are colours of light, and there are colours. The first set is a line — you can point a spectrometer at any of them and read a number in nanometres. The second set is a filled shape your visual system builds on top of the line, and most of it, including the entire family of purples, answers to no wavelength at all. The next time someone says a colour "vibrates" or asks for the frequency of magenta, you can tell them the truth with a clear conscience: it hasn't got one. It is the colour your eye makes to close a circle the universe left open — and you can watch it do exactly that, above.

VI · Sources & data

CIE 1931 2° standard-observer colour-matching functions (x̄, ȳ, z̄)The empirical curves underlying everything here. Digitised tables (5 nm and 1 nm) from the Colour & Vision Research Laboratory, UCL — the standard machine-readable form of the CIE 1931 standard observer.cvrl.ioo.ucl.ac.uk/cmfs.htm
CIE — Commission Internationale de l'Éclairage (1931)The body that standardised the colour-matching functions and the xy chromaticity diagram from the colour-matching experiments of W. D. Wright (1928–29) and J. Guild (1931).
sRGB / Rec. ITU-R BT.709 primariesScreen-gamut triangle: R (0.640, 0.330), G (0.300, 0.600), B (0.150, 0.060), white D65 (0.3127, 0.3290). IEC 61966-2-1:1999. XYZ→linear-sRGB matrix and the 2.4-gamma transfer are the standard ones.
Isaac Newton — Opticks (1704)Where the open spectral strip is first bent into a closed colour circle, the band of purples added by hand to join the ends — the historical first statement of the seam this page measures.
research/there-is-no-magenta/verify.mjsThe offline ground truth (18/18): the locus, the line-of-purples-is-extra-spectral proof, the metameric-black construction with byte-identical XYZ, and the gamut area fraction. The page recomputes the same from the same data.