Interesting, thanks! 
When speaking of pigments, there’s no problem defining white, or producing a substance whose color is an excellent approximation of white: A white substance is one which reflects all of the light incident upon it, regardless of wavelength. It’s only for defining white light that you have a problem.
As to my comment about Vega being the definition of “white” for astronomers, the way astronomers define color is generally “color indices”: You look at a star (or other object) through one filter, and measure its magnitude, and then through another filter, and subtract the two magnitudes. Zero magnitude through any given filter is defined to be how Vega appears through that filter, so for any pair of filters, Vega’s color index is always zero, meaning colorless.
I’m not sure it’s quite that simple actually.
First of all, when you say light obviously that needs to be taken as visible light to a human: no material can reflect all em radiation, and who knows what color combinations might look like to animals with different color gamuts.
So secondly, since we’re talking about human color perception, i would guess (it seems quite hard to Google) that different proportions of reflection would be perceived as the same white. For example, one material that reflects 95% of RGB and another that reflects 98% of RG and 92% of B might look the same to the human eye.
I’m not being critical, i just find it interesting to dig into.
You’re quite right, of course. I’ve been fascinated by the way different spectra can look like the same color to the human eye, and the consequences this has. It’s an interesting topic for a science fiction story – alien races, even if they’re sensitive to the same visible wavelengths that humans are, might have very different perceptions if they use a significantly different tristimulus color gamut, or if they were, say, bistimulus or quadrostimulus. A display on a standard human color monitor or television, or the printing in a three-color page might look grotesquely incorrect to them. Our color photographs and color movies and even color holograms wouldn’t look right, either. To accommodate a wide range of alien races with different color sensors, we might have to use a completely different technology. Some kind of Lippmann color system, maybe. (And that’s ignoring the possibilities of races that have gamuts that extend into the infrared or ultraviolet, a la one of the characters in Alfred Bester’s The Stars my Destination.)
But Chronos is pretty much correct for the case of a white surface, if illuminated by a source that is agreed to be “white” Your counterexample of a surface reflecting 0.98 R 0.98 G and 0.92B isn’t that far off the ideal – our eyes and the software in our brain that processes all this is pretty tolerant of small variations.
It’s that fact, I think, which makes our tristimulus color technology as relatively simple as it is. Imagine how difficult things would be if you had to precisely match color gamuts at every stage. Through the years, we’ve been incredibly sloppy about how we define “Red” Green" and “Blue” (or “Cyan” “Magenta”, “Yellow”), yet, even though the color filters, film responses, and light sources don’t precisely match, we still get a good approximation of the original color.
We could define white pigment as that pigment that does not change the visible spectrum of light it reflects. If you illuminate it with red light, it looks red. Ours eyes don’t detect the reflective properties of a material, they detect light.
I guess light is the main question, since the question was about the colour of the sun.
Vega is an interesting question. I assume Vega is white, by definition, when viewed at sea level and at the top of Mauna Kea. Or even from orbit? So is it defined as a local spectral calibration standard, and implicitly includes local spectral changes due to the atmosphere, rather than an absolute standard? Which would make sense. But that doesn’t make it a standard for human perception of colour. In the extreme, low on the horizon, Vega would be a clearly red colour to the eye. But by definition, it is the white stars next to it in the sky are judged by.
No black body has a flat spectrum, so even Vega’s spectrum is just a particular temperature we have decided to use as a white reference, and doesn’t meet a quantitative criterion such as equal energy at each wavelength in the visible spectrum.
A lot of people cite “equal output at each wavelength” as a definition for white light. I don’t think any of these are people in optics. As I say, the definition of white light is from the sun, or from a blackbody. The Standard Illuminants A, B, C, and D used by the CIE committee were light from a tungsten filament (pretty close to a blackbody when hot, and also widely used at the time for interior lighting) or various types or ordinary daylight. See here for details:
To tell the truth, I can’t even think of any kind of source that has a flat output. You could construct one (using a bank of independently-controlled LEDs that peaked at different wavelengths), but I don’t know what advantage it would have. I haven’t run the calculation, but I strongly suspect that the CIE coordinates of such a “flat white” source would be pretty close to the loci of the standard white illuminants.
I don’t disagree, which is why I said “a quantitative” criterion. That just seems to be the one with the least baggage. Otherwise there isn’t any single definition. It is all down to human perception, and that gets you to the loci of standard illuminants, all of which are in some manner human relative, even if only relative to human technology of the time. We have a set of different black body temperatures, some with modification, and choose depending upon circumstances, which one is white. Which does tend to underline that there really isn’t any physical definition of a single white colour. At least not one that can be used to describe the perceived colour of a star. Astronomers don’t use illuminants - the things they are interested in mostly emit their own light. So they choose yet another black body as their reference when defining colour.
The equal energy at each wavelength is a bit of a backwards definition. We define a white noise spectrum this way, but take the idea for the name from a naive ideal of light. Porting it backwards is really unjustified. But it is at least simple and universally understood, and useful away from human perception.
Yeah, I’m sure that this is the reason that people think that white light ought to be defined the same way. But, as a practical matter, it really isn’t. Blackbody radiation is easy to generate, but a flat spectral output across a long wavelength range isn’t.
Greg Egan explored this topic (with modified, heptachromat humans) in Seventh Sight in the Instantiation collection.
To further illustrate the difficulties in defining a “white spectrum”: Multiple folks have already mentioned a hypothetical spectrum with the same amount of energy in every wavelength band. But surely, it would be just as natural to call “white” a spectrum with the same energy in every frequency band… and yet, those are two completely different spectra. Nor is either actually physically possible, without some boundaries, because both would yield infinite total energy (the frequency one via an ultraviolet catastrophe, and the wavelength one via infrared catastrophe).
But don’t colors have absolute values in terms of wavelength, so that an emitter or reflector of blue light always emits or reflects light waves of a higher energy than an emitter or reflector of red light
OTOH the mapping of names to colors varies widely across different cultures, languages, and times. We know that “orange” was named for the fruit, so we must have called that color something else before we knew what oranges were. (I think it was “yellow”, which I find very hard now to wrap my mind around now.)
Sort of. A pure spectral colour - that is a single wavelength maps to a single colour in our perception. But we can see colours that are not single wavelengths, and they map to different colours. Worse we can see colours that do not exist as single wavelengths. Purple for instance. (Violet is not purple.)
Our eye perceives colour as a mixture of three stimuli. These stimuli are processed to yield a colour. But for almost any colour that isn’t a single wavelength there are an infinite number of combinations of the three stimuli values that can yield the same perceived colour.
If you look at a CIE colour space diagram you can observe a number of things.
- The colours along the bottom edge of the diagram have no spectral existence. There is no one wavelength of light that makes them up, and they are made of a mix of far red and far blue.
- The curved edge of the diagram corresponds to pure spectral colours. The only way to see these is with a pure single wavelength light source. However, where the edges are reasonably straight, you can get very close by mixing a set of nearby colours, both of shorter and longer wavelengths.
- Every other colour in the diagram is a tristimulus mix, and is created by a mix of the three stimulus values. As a good approximation you can choose any set of wavelengths and come up with a mix that yields those colours. None of the colours inside the diagram exist as pure spectral colours.
And if you want to get really weird:
Take note of the spectral sensitivity curves. There is no point on them where only one of the receptor types responds. Which means that even for pure spectral colors, our receptors don’t experience a pure response.
If you could somehow manually stimulate just one receptor type, say with a focused laser, you could see hyperspectral colors that can’t be seen in real life.
Not sure what you mean by this. The three receptors overlap over the visible spectrum, so that for most pure spectral colors you’ll get some response from each of them. In some places you’ll effectively only get two (above about 550 you have very little “blue” response, but lots of red and green). Only at the extreme ends – far violet or extreme red – will you get mostly one color sensor type firing. Out that far you’re at the extreme limits of human vision, and the response is so small that it’s difficult to see the colors there. But it is true that if you look at a spot of light from a very far red laser you can see it, because the intensity of the laser is great enough to make the light visible. And in that case you would virtually see light from one sensor. But it’s not because it’s “focused” – just bright. Nor is the color “hyperspectral”. It’s still part of the spectrum, and even part of the visible spectrum; just a part you rarely see because light there is usually dim. But because any of those long red wavelengths effectively only stimulate the one color cone, they’re going to look the same. Without stimulation of the other color cones you’re not going to have a means for the brain to distinguish between them.
http://hyperphysics.phy-astr.gsu.edu/hbase/vision/colcon.html
I suppose the same thong ought to happen at the blue end of the spectrum, but I haven’t heard about it.
I mean, their flag represents the sun as a red circle, so that makes sense
The very fact that people perceive “impossible” colours seems to prove colours are artefacts. Same for different spectra that look the same.
Physical impossibility isn’t necessarily a reason to not use something as a standard. It’s the precision of the definition that matters, not its physical existence. For example, empty, flat space does not exist, yet that’s what we measure many quantities against. Spectra could be compared to a theoretical spectrum.
It’s a very poor excuse for a practical standard that’s used experimentally and empirically. A standard you can’t make or use is only of interest as a theoretical possibility. Like empty flat space.
That’s why standards that are in real use are physically realizable items, even if they might not be trivial. That’s why the standard kilogram was for many years an actual physical weight, and why, from 1960 until 1983, the standard meter was defined as 1650763.73 times the length of the orange-red line of Krypton 86 in vacuum.
My point is that if you could stimulate just the M (“green”) receptor explicitly, without stimulating the S or L receptors, then you could see a “greener than green” color. Because even a pure spectral green in real life stimulates the L receptor. Well, at least it seems like it would be greener-than-green, but we don’t actually know since no one has seen it yet. I brought up the laser for its ability to focus to a small point, not for its pure spectrum. One can imagine other means of stimulating just one receptor type; say, electrical or chemical (I’m not claiming any of these are actually practical).
As you mention, you can get a fairly pure violet or red by going far out past the usual ends of the spectrum, but that’s not particularly practical (the short-wavelength, high-intensity blue would be damaging if nothing else), and in any case that approach wouldn’t work for green.
It’s possible to get a taste of these “impossible colors” by inducing receptor fatigue (i.e., staring at a strong color and then perceiving the negative of that), but it’s not quite the same thing.
Perhaps this has already been pointed out in one of the posts but, if the sun was yellow then white objects would appear yellow. It would, in fact, be impossible for any object to appear any other color than yellow or a reduced shade unless illuminated by a non-Sol light source.