Why don't metals glow from red to yellow to green to blue etc.? Why only red, then yellow and then white? Shouldn't all wavelengths be emitted one by one as the temperature of the metal increases?
If some metals do glow at with different colours, could you give me examples of such metals and the reason why this happens in specific cases?
The physics of why the heated metal glows like a black body has already been thoroughly covered in the previous answers. However, in order to completely bridge the gap with the physiology of color perception (which has been alluded to in some answers), it is worth showing a picture of the Planckian locus:
Plot by PAR, from Wikimedia Commons
This is the set of all the colors a black body can have, plotted in a chromaticity diagram. It is computed by combining the black body emission formula with the color matching functions, which are a mathematical model of our color vision.
This graph clearly shows the path of a black body going hotter: red → orange → yellow → white → blue. Now, one may wonder by which coincidence it hits right into the white, rather than going slightly above (through the greens) or below (through the purples). That question, however, is backwards. The good question would be “Why have we chosen to name ‘white’ a color from the Planckian locus”. This is the question of the definition of white, and it is not straightforward.
In the physicists jargon, the name white is often used to mean a “flat spectrum”, i.e. one in which the power per unit frequency does not depend on the frequency. When talking about actual visible colors, however, it has a completely different meaning:
This leaves the notion of while light ill defined: the light coming from a white surface has the same spectrum as whatever illuminant (meaning: light source) was shined to it. Then the light from any typical illuminant could be considered, in some sense, to be “white light”.
In practice, in the realm of color science, there are some so called “standard illuminants” which are deemed white. Most notably D65 and D55. These are meant to model natural daylight. The choice of daylight as a reference light source is obvious given that our species has evolved in a world where daylight has always been the standard light source, and thus our natural white reference.
The spectrum of daylight varies with the weather and with the height of the sun above the horizon, but it is never too far from a black body spectrum. Which is probably not very surprising given that the Sun itself is a pretty good black body.
When metals (or any materials) get very hot, they emit "black body radiation". It's a funny name, because even "non black" bodies emit this thermal radiation. There may be some (small) deviations from the general law due to surface emissivity, but not so much that you would notice.
Black body radiation is not a single wavelength, it is a broad spectrum given by Planck's law. I gave a number of examples of this in this earlier answer. I will just reproduce one of the curves from there:
This shows that the general shape of the emission curve is always the same - the only thing that changes is the position of the peak (which follows a 1/T law known as Wien's Displacement Law), and the area under the curve, which increases with the fourth power of the temperature (Stefan-Boltzmann law).
As I mentioned, in a "real" material there may be some "bumps" in the spectrum due to changes in emissivity with wavelength - but to the best of my knowledge this is not enough to stop the appearance of red-yellow-white as things get hotter. However, when things get very, very hot they might appear blue. However, at that temperature the intensity of light is so great (and with so much UV and shorter wavelength light) that I would highly recommend not trying to look at the source. But you can see this in the T=10000 curve above, which has much more blue than red light in it.
When I refer to a colour, eg green, I an referring to a range of wavelength which on entering the eye are perceived to be a range of colours depending on the intensity of each of the wavelengths within that range eg yellowish green, green, bluish green.
The black body emission spectrum at a number of temperature is shown below.
The colours shown on the graph illustrate the colours our eye would perceive if illuminate by the range of wavelengths spanned by a particular colour.
The brain interprets the information from the colour receptors (cones) and so you perceive a colour which sometimes can be "unexpected".
For example a combination of red and green light the eye perceives as yellow.
At a temperature of $3000 \, \rm K$ the vast majority of the visible light entering the eye is red and so the object is seen to be red.
As the temperature increases there is more red light but also orange light so the colour of the object changes and from a dull red it becomes a much brighter red.
At higher temperature there is a greater range of wavelength each stimulating the cones and because of this range the colour of the object is moving towards white not the individual colours which are emitted.
Raising the temperature even more which makes the proportions of the colours roughly equal the object makes the object look white.
An even higher temperature when it is very dangerous to look at the hot object directly because of the uv light which is also emitted, the white has a bluish ting.
A metal rod heated in an induction furnace shows the changing colours extremely well.
If you wanted to see a green hot object what you need to do is filter out the red and blue ends of the light from the object.
All wavelengths are included in what you call "white", thus all wavelengths are there.
The evolution you are describing is the evolution of the black body radiation as heat is increased.
You can see in the link that as temperature increases the curve moves to the left, to smaller wavelengths.
When the temperature of a blackbody radiator increases, the overall radiated energy increases and the peak of the radiation curve moves to shorter wavelengths.
Here is the left side of a black body radiator, which a hot metal is.
Backbody radiation in the visible spectrum at 800, 850, 900, 950, 1000, 1050, 1100, 1150, 1200 and 1250 K.
As the temperature of the sample rises , the spectrum moves to the left, and more and more higher energy photons are released.
As heat increases , to the left, more and more photons at visible energies appear, first the red, then the yellow, then by the time temperature reaches green in the spectrum you already see "white" because that is the way color perception works in our eyes. From then on it is white for our perceptions.
The curves are almost flat at this level of detection and at temperature s where metal is incandescent , still the number of photons emitted is biased towards the infrared, there will always be more photons at a lower frequency than the next band of the spectrum thus blue cannot dominate over the "white" soup our eyes see.
Shouldn't all wavelengths be emitted one by one as the temperature of the metal increases?
Blackbody emission can be grossly oversimplified as "we start with the lowest and then we keep adding more energetic wavelengths as the temperature increases". That's why when we get to the point that everything visible was added to the mix, humans perceive "white". White is the full spectrum.
So it goes like:
If you wanted to see blues and greens, it would have to stop emitting reds first. And it's not how it works, quantum mechanic explains why.
Of course, this is gross oversimplification. Other answers show that as the temperature increases, the emission curve not only gets wider (what I've described here) but also gets slightly shifted up (what OP expected). But the first effect is large while the other is small in comparison, so the widening into white dominates. Only with the hottest of the hot (like arc lamps) we get to the point where we begin to perceive the light as "slightly bluish".
Also, there is no need to limit to metals. The only thing about metals is that you can't get them really hot. Which, accidentally, makes the experiment stop about right where the white is prettiest.
I made a figure to underscore what I think are the most crucial points to understand why black body radiation is never perceived as green:
The result of this is that it is impossible to stimulate the M (green) cone with black body radiation without also stimulating either the L (red) or the S (blue) cone to a similar or larger degree.
The black body radiations that we perceive as most red and blue do not actually peak where the L and S cones peak in sensitivity. The L cone sensitivity peak around $570$ nm while the S cone sensitivity peak around $442$ nm. Using Wien's displacement law we find that the black body radiations for $\sim 5100$ K and $\sim 6550$ K would peak at these wavelengths. But in this figure (from Wikipedia), showing the perceived color of black body radiations for different temperatures, we can see that both of those temperatures would be perceived as more or less white.
The most red and blue black body radiations instead peak at much longer and shorter wavelengths. These are instead the Planck curves for which the slope is at its steepest between the peaks of the spectral sensitivity curves, producing the maximal contrast in stimulation between the different cone types.
One important thing to note, as galinette did in a comment, is that metals do vaporize soon after they get warm enough to emit a significant amount of blackbody radiation in the visible portion of the electromagnetic spectrum. As graphs posted by others indicate, a metal would emit light perceived as white around 5000 K. It would have to be warmer than this for the graph to be situated such that the emitted radiation in the visible portion of the spectrum appears blue. Iron, for example, has a melting point of 1811 K and a boiling point of 3134 K. If iron got warm enough to have a blue tint, you wouldn't be able to see it for very long.
