Northern lights colors: every auroral emission measured, and the reason your eyes cannot collect them

Every page about aurora colours gives you the same three facts. Green is oxygen. Red is oxygen higher up. Blue and purple are nitrogen lower down. Pink is a mix. NASA says it, the Canadian Space Agency says it, the National Park Service says it, and every tour operator in Lapland says it with better photographs.

All of it is true and none of it is a colour. Nobody publishes a hex code. One of the things people actually type into Google next to this question is northern lights color code, and the honest answer is that the wavelengths have been nailed down for a century and turning them into colours is an afternoon of arithmetic that nobody has bothered to write up.

So I did it. Every visible auroral emission, rendered from its measured wavelength through the CIE colour matching functions. What came out was a coincidence I was not expecting, followed by the discovery that the coincidence is wasted, and then the realisation that two of the standard explanations are wrong about which gas does what.

Every auroral emission, as a colour

The aurora is not a continuous spectrum. It is a short list of discrete emissions, which is why it is the rare natural light source where a hex code per feature is a meaningful thing to compute. The wavelengths below are the air wavelengths from the NIST Atomic Spectra Database for the atomic lines, with the emission list and the band systems taken from Mathieu, Robert and Lamy's 2025 synthetic auroral spectra paper.

One caveat belongs up front. A single wavelength is the most saturated thing a colour can be, and a screen cannot show any of them. Every emission here is outside the sRGB gamut, so each swatch is the closest your display can get, and the percentage says how far past the edge the real colour sits. This is the same wall that the visible light spectrum runs into, where not one of 321 wavelengths fits inside a monitor.

The first surprise is in the top row. The famous auroral green is #7CFF00, and that is a chartreuse. It sits at CIE chromaticity (0.3569, 0.6401), which is on the yellow side of the green primary of your screen at (0.3000, 0.6000). The 557.7 nm line is a yellow-green, not the deep emerald that every aurora photograph shows. Hold that thought, because the reason for the discrepancy turns out to be a camera setting rather than a fact about the sky.

The green line lands 2.7 nm from the peak of human vision

The CIE 1924 luminous efficiency function, the curve describing how efficiently the eye turns light into brightness in daylight, peaks at 555.0 nm. The auroral green line is at 557.734 nm. That is a gap of 2.734 nm, which puts the line at 0.99859 of maximum sensitivity.

The brightest emission in the aurora is, to four decimal places, exactly where the human eye is most sensitive. There is no causal story there, just two unrelated facts about atomic oxygen and primate retinas landing on top of each other. It is the sort of coincidence that makes the usual explanation feel satisfying. Every source tells you green dominates partly because the eye is most sensitive to green, and the numbers look like they agree.

Then you work out how bright an aurora actually is, and the coincidence falls apart.

The aurora is never bright enough to see in colour

Auroral brightness has a standard scale. The International Brightness Coefficient classes an aurora by the strength of the green line in kilorayleighs, and each class comes with a familiar comparison: class I is about as bright as the Milky Way, class IV casts shadows and is described as being as bright as the full moon.

A rayleigh is a column emission rate, 10¹⁰ photons per square metre per second into the full sphere. Convert that to a surface radiance, weight it by the luminous efficiency function and multiply by 683 lumens per watt, and each class becomes a luminance in candelas per square metre. Then compare against the CIE mesopic range, which runs from 0.005 to 5 cd/m². Below the bottom of that range you are on rods alone, and rods do not produce colour at all. Above the top you have full cone vision. In between you have some of both.

Classes I and II, which is nearly every aurora anybody ever stands under, come out at 0.00019 and 0.0019 cd/m². Both are below the 0.005 cd/m² floor of the mesopic range. At those levels the cones have nothing to work with and the retina is running on rods, which are colourblind. The official description of class I says “no colour apparent”, and the photometry says that is not a limitation of the observer, it is arithmetic.

Classes III and IV get into mesopic territory, where colour starts to come back, and this is where the real finding is. Class IV is the one people describe once in a lifetime, the kind that casts shadows. It lands at 0.19 cd/m², which is still 26 times below the level where human colour vision is fully operating.

No aurora on the standard scale reaches full colour vision. The 2.7 nm coincidence between the green line and the peak of the photopic curve is real, and it is useless, because the photopic curve is not the one in service when you are looking at an aurora.

Checking the model against the moon

That calculation is doing a lot of work, so it is worth testing against the one external benchmark the brightness scale hands over. Class IV is defined as being roughly as bright as full moonlight. Take the 0.19 cd/m² surface luminance, integrate it over a full hemisphere of sky, and you get a ground illuminance of 0.61 lux. Full moonlight is about 0.25 lux.

That is a factor of 2.4, in the direction you would expect, because real auroral curtains cover part of the sky rather than all of it. A class IV display over half the sky lands within a few percent of full moonlight. The radiometry reproduces a comparison that was written down by people looking at the sky rather than calculating, which is about as much confirmation as this kind of model gets.

What rods do to the colour balance

Since the aurora is a rod-dominated object, the interesting question is not how the emissions rank in daylight vision but how they rank under the scotopic curve, which peaks at 506 nm rather than 555 nm. The reordering is severe. Per photon, relative to the green line:

• The 427.8 nm violet band is 50.2 times more effective at stimulating rods than it is at stimulating daylight vision.
• The 470.9 nm blue band gains 20.1 times.
• The 520.0 nm nitrogen line gains 3.7 times.
• The red doublet loses a factor of 30, dropping from 0.221 of the green line photopically to 0.0074 scotopically.

The green line itself is 51.7 nm away from the rod peak and sits at 0.3608, so even the brightest thing in the aurora is 2.77 times down from the best the dark-adapted eye can do. The 557.7 nm line is almost perfectly placed for the visual system that is switched off and mediocrely placed for the one that is running.

Rods cannot report colour, so none of this becomes hue. It becomes brightness, which is why a faint aurora reads as a moving grey or colourless glow with maybe a hint of green at the bright edges, and why the red that a camera finds easily is the first thing to vanish from naked-eye viewing. This is the same limit on colour discrimination showing up at the bottom of the brightness range instead of the bottom of the difference range.

Green dominates mostly because of chemistry, not eyes

The standard claim deserves a number rather than a mention. Using the measured intensity ratios from the auroral statistics literature, an active aurora runs the green line at about 19 kR against 1.2 kR in the red doublet, so there are 16 green photons per red photon before anything biological happens. The eye then adds its own preference, which is worth 4.52 times per photon once you account for both the efficiency curve and the energy per photon.

Multiply them and green outweighs red by 72.3 to 1 in luminance. Split that into its two causes:

The eye is responsible for a 4.5x slice of a 72x effect. The aurora is green because excited atomic oxygen at 100 km mostly decays down the 557.7 nm channel, and the eye bias is a modest amplifier on top of a decision the atmosphere already made. Writing it the other way round, which is how most explanations are phrased, gets the proportions wrong by a factor of about four.

Oxygen cannot make pink

Here is where the standard account is not merely imprecise but wrong. Every version of it says pink and magenta are mixtures of the other auroral colours. Space.com puts pink and yellow down to red, green or blue “blending together”. If you take that literally and mix the two oxygen emissions, it does not work.

I swept the ratio of the red doublet against the green line across four orders of magnitude and tracked the hue angle of the result. The mixture travels from green at 113 degrees through yellow and orange to red at 43 degrees, and stops. Magenta is at about 330 degrees. Oxygen's two emissions sit on the same side of the colour circle, so every mixture of them lands on the arc between them. There is no combination of the green line and the red doublet that produces pink, at any ratio.

What does produce it is nitrogen, using both of its band systems at once. The N2+ first negative group is violet at hue 319 degrees. The N2 first positive group is red at hue 40 degrees. Those two are on opposite sides, and mixing them crosses the purple line:

The ratio needed to hit magenta at 330 degrees is about 0.12 of first positive against first negative, and I checked that the answer does not depend on my assumption about the relative strengths of the N2+ bands. Across four different band ratio assumptions the magenta appears at ratios from 0.07 to 0.27 and always at the same hue, 329.8 to 330.1 degrees. The result is robust to the one number in it I had to assume.

This explains something the mixture story cannot. The pink in an aurora appears along the lower border of the curtain, and that is not a coincidence of viewing geometry. Both of the band systems that make it belong to molecular nitrogen, and both need the dense air low in the atmosphere, below about 100 km, where N2 is still abundant and the incoming electrons are energetic enough to reach. Oxygen's green line peaks above that and its red doublet far above it. The pink sits at the bottom of the curtain because the chemistry that makes it only happens at the bottom. A colour mixture argument predicts pink wherever red and blue overlap, which is not what the sky does.

The teal in every photograph is a white balance setting

Back to the chartreuse. The physics says the green line is #7CFF00 at hue 116 degrees. Aurora photographs are reliably a deeper, bluer, emerald green. Something is moving it, and the candidate is obvious once you think about how these pictures get made: aurora photographers set a cool white balance, usually somewhere between 3200 K and 4000 K, because it renders the night sky as a dark blue rather than a muddy brown.

Chromatically adapting the green line from a 3800 K assumed illuminant to a D65 display, which is what that camera setting does to the file, gives this:

Nearly 28 degrees of hue, which is the whole distance between a yellow-green and the saturated green people recognise. The iconic aurora colour is a real emission seen through a deliberate and sensible photographic choice. At 5500 K the shift is 23 degrees, at 2800 K it is 33, so the exact emerald varies with whoever processed the image, which is also why no two aurora photographs quite agree on the colour.

None of which makes the photographs dishonest. They are showing you a colour your eyes genuinely cannot collect, for the reason in the brightness table above, and they have to pick some white balance to do it. It is worth knowing that the number is a choice. The same thing happens in reverse when the dress splits a room, where the illuminant your visual system assumes decides what colour you report.

What to actually expect

Putting the photometry together with the emission colours gives a more useful answer than a colour chart does.

• A faint aurora will look grey or colourless. At class I and II you are below the threshold for colour vision by a factor of thousands. Seeing it as a pale moving glow is the correct response of a working visual system, not a failure of yours.
• Green arrives first. It has 16 times the photons and the best placement on both efficiency curves, so it is the only colour that shows up while you are still near the bottom of the brightness range.
• Red is a camera colour. The doublet loses a factor of 30 under rod vision. A sensor has no scotopic mode and will record it happily while you see nothing.
• Pink at the bottom edge means a strong event. It needs electrons energetic enough to drive nitrogen below 100 km, which is why it shows up in the displays people remember.
• Give it 30 minutes. Full dark adaptation is slow, and since you are operating at the very bottom of the mesopic range, the difference between a partly and a fully adapted eye is the difference between colour and no colour.

A note on what this does and does not establish

The emission wavelengths are measured and the colour science is standard, so the hex codes are as solid as the gamut problem allows. The intensity ratios come from published auroral statistics, and real auroras vary across a wide range, which is why I quoted the brightness findings by IBC class rather than as single numbers. The one quantity I assumed rather than sourced is the relative strength of the N2+ bands, and I tested the pink result against four alternatives to check it did not matter.

The mesopic boundary at 0.005 cd/m² is a standard adaptation threshold rather than a hard switch, and colour perception for a large dim source can begin a little below it. That nuance moves classes I and II around slightly. It does not touch the headline, which is that class IV sits 26 times below full photopic vision, because that number depends only on the top of the mesopic range.

I have also treated the aurora as a flat extended source of uniform brightness, which real curtains are not. A bright rayed structure can be locally much brighter than its class average, and that is probably where most reports of genuinely vivid naked-eye colour come from.

Try the hard version

An aurora is colour judgement on the worst possible terms: dim, moving, and with nothing to compare against. If you want to find out how good you are under easier conditions, that is what this site is for. The colour memory game takes away the reference and makes you hold a target in your head. Hue sort is the direct test of the thing this article keeps measuring, which is whether you can order colours by hue angle, and spot the difference shrinks a gap until you lose it. There is more in the full set.

For the neighbouring questions, what colour lightning is uses the same line-spectrum method on a much hotter source, why the sky is blue covers the daytime version of a glowing atmosphere, star colours measures the emission of everything further away than this, and CIEDE2000 explains the distance measure behind the hue angles here.