№ 178 · astronomy

Why stars have colours, and none of them are green

Look closely at the night sky and the stars are not all the same colour. A few are orange, many look white, and some have a cold blue cast. The colour is a thermometer: it is set mostly by how hot the star's surface is. The surprise is what the thermometer never reads. There are orange stars, white stars and blue stars, but no green ones, even though some stars give off more green light than any other colour.

Predict first

Measured per nanometre of wavelength, the Sun's light is strongest at about 500 nm, in the blue-green. What colour would the Sun look from space, with no air in the way?

Check your answer

White, with at most a faint warm tint. Its light is spread almost evenly across the colours the eye can see, so the eye's three kinds of colour receptor are all stimulated by similar amounts, and that mixture reads as white. From the ground it looks yellow mainly when it is low in the sky, where the air has scattered away more of its blue.

Hotter glows bluer

A star's surface behaves roughly like what physicists call a black body: something that gives off light only because it is hot, with a spread of wavelengths fixed by its temperature alone. The curve for that spread is broad. It rises steeply on the short-wavelength side, peaks, and falls away slowly toward the infrared. As the temperature goes up, the whole curve grows and slides toward shorter wavelengths.

What the eye sees is only a narrow window of that curve, from roughly 400 to 700 nm. For a cool star the window sits on the rising side of the curve, so the red end of the window gets far more light than the blue end; in the figure, a 3,060 K surface gives off about four times as much light at 650 nm as at 450 nm, and it looks orange. For a very hot star the window sits on the falling tail, the blue end wins, and the star looks pale blue. Jonathan Marr and Francis Wilkin, in a 2012 paper on how this is taught, put it the right way round: hot and cool stars are coloured because their spectra slope across the visible band, not because their peak sits on a particular colour.

Why no star is green

Now put the peak in the middle of the window, in the green. Because the curve is so broad, it is nearly flat across the few hundred nanometres the eye can see. A surface at 5,772 K, the Sun's, puts about a third of its visible light into each of the blue, green and red thirds of the band. Green is the largest share, but only just. Our colour vision does not pick out the brightest wavelength; it compares how strongly three kinds of receptor are stimulated, and a nearly flat spectrum stimulates all three about equally. The result is white. Try the figure's button that puts the peak exactly on green, at 530 nm: that needs 5,467 K, and the disc comes out warm white.

So as a star gets hotter, its colour does not walk around the rainbow. It runs along a single track: orange, then pale orange, white, blue-white and pale blue. Jan-Vincent Harre and René Heller worked out the colours of main-sequence stars from detailed model spectra in 2021 and plotted them on a colour wheel. The track passes through no yellow, green, cyan or purple. Sun-like stars came out slightly yellowish, and stars of type F9.5, a little hotter than the Sun, sat at the white point.

Interactive Set the surface temperature of a perfect glowing body (a black body), or pick a kind of star. The top panel shows how much light it gives off at each wavelength, with the band the eye sees shaded in its colours; switch between counting the energy per nanometre and per unit of frequency to see the peak move while the colour stays the same. The disc is the colour that light makes for the eye, worked out from the standard colour-matching curves. The bar underneath holds every colour the slider can reach. Every control works with Tab, Enter and the arrow keys.

surface temperature5,772 K

5,772 K looks white, with a faint warm tint

what the eye sees100 nm200 nm500 nm1,000 nm2,000 nm4,000 nmmost0peak 502 nm

Brightest point of the curve: 502 nm if you count energy per nanometre, 883 nm if you count it per unit of frequency.

Of the light between 400 and 700 nm: 34% is blue (400–500 nm), 35% green (500–600 nm), 31% red (600–700 nm). At 650 nm (red) it gives off 0.89 times what it gives off at 450 nm (blue).

On screen: red 1.00, green 0.88, blue 0.82 (linear, largest set to 1). Green is 88% of red.

Closest row in Mamajek’s table of main-sequence stars: G2V, 5,770 K.

2,300 K5,000 K10,000 K20,000 K40,000 K

The peak depends on how you count

There is a second reason not to read a star's colour from its peak: the peak is not one fixed place. You can count a star's energy per nanometre of wavelength, or per unit of frequency. Both describe the same light, but the steps are spaced differently, so the curves have different shapes. Marr and Wilkin point out that for the Sun the first peaks near 500 nm and the second near 880 nm, in the infrared. Switch the figure between the two: the peak jumps, and the colour of the disc does not change at all, because the light is the same.

Real stars, and how their temperature is read

Harre and Heller also found where the black-body picture holds best. Model stars of types K0 to K1, around 5,200 to 5,300 K, came out almost exactly the colour of a black body at the same temperature. The coolest stars, around 2,300 K, look orange, not red; the black body is redder, which the authors suggest may be part of why such stars are called red dwarfs. Above about 12,000 K the colour hardly changes to the eye, and stars hotter than about type B5 all converge on the same blue. They also note that at low light the eye weighs colours differently, making reds look darker and blues brighter, so a star seen at night need not match these colours exactly.

Astronomers do not find temperatures from the peak either. Marr and Wilkin note that they use ratios of spectral lines or of the light measured through two different filters. A classic version measures a star's brightness through two standard filters, B for blue and V for visual. The difference B−V is the star's colour index: the bigger the number, the redder the star. In Eric Mamajek's table of average main-sequence stars, a star of the Sun's type, G2V, has a B−V of 0.65 and a temperature of 5,770 K, while an M5V star has 1.83 and 3,060 K. This is still how temperatures are estimated in bulk: Zenghua Zhou and colleagues noted in 2026 that most stars in modern sky surveys have no spectrum, and published new colour–temperature relations with an internal precision of about 30 to 50 K for the best colour combinations.

The figure uses 5,772 K for the Sun, the nominal value the International Astronomical Union fixed in 2015 as a standard for converting between solar units. Set the slider there and the disc is white: the same light that, drawn per nanometre, peaks in the blue-green.

Where this comes from

  1. Digital color codes of stars linked only, not reproduced
    Jan-Vincent Harre, René Heller · arXiv:2101.06254 · 2021
    arxiv.org/abs/2101.06254
  2. A better presentation of Planck's radiation law linked only, not reproduced
    Jonathan M. Marr, Francis P. Wilkin · arXiv:1109.3822 · 2011
    arxiv.org/abs/1109.3822
  3. Intrinsic Colors, Temperatures, and Bolometric Corrections of Pre-Main Sequence Stars linked only, not reproduced
    Mark J. Pecaut, Eric E. Mamajek · arXiv:1307.2657 · 2013
    arxiv.org/abs/1307.2657
  4. Empirical colour--effective temperature relations in the SDSS system from IRFM temperatures of GALAH and APOGEE stars reuse permitted with attribution
    Zenghua Zhou, Luca Casagrande, Xiaobin Zhang, Jianping Xiong, Jiao Li, Yanjun Guo, Zhanwen Han and Xuefei Chen · arXiv:2606.11594 · 2026
    arxiv.org/abs/2606.11594
  5. Nominal values for selected solar and planetary quantities: IAU 2015 Resolution B3 linked only, not reproduced
    Andrej Prsa, Petr Harmanec, Guillermo Torres, Eric Mamajek, Martin Asplund, Nicole Capitaine, Joergen Christensen-Dalsgaard, Eric Depagne, Margit Haberreiter, Saskia Hekker, James Hilton, Greg Kopp, Veselin Kostov, Donald W. Kurtz, Jacques Laskar, Brian D. Mason, Eugene F. Milone, Michele Montgomery, Mercedes Richards, Werner Schmutz, Jesper Schou and Susan G. Stewart · arXiv:1605.09788 · 2016
    arxiv.org/abs/1605.09788
  6. A Modern Mean Dwarf Stellar Color and Effective Temperature Sequence (version 2022.04.16) linked only, not reproduced
    Eric Mamajek · 2022
    www.pas.rochester.edu/~emamajek/EEM_dwarf_UBVIJHK_colors_Teff.txt
  7. Simple Analytic Approximations to the CIE XYZ Color Matching Functions (Journal of Computer Graphics Techniques 2(2)) linked only, not reproduced
    Chris Wyman, Peter-Pike Sloan & Peter Shirley · 2013
    jcgt.org/published/0002/02/01/