We think first of two detached bands of stars, one, the white dwarf group, formed by stars in which all the electrons are torn off the atoms; and the other, the main-sequence, formed of stars in which the atoms are still surrounded by their _K_-rings of electrons, while the exterior rings have been torn off. Starting from about the middle of the main-sequence is the spur branch leading up to the red giants, as shewn in fig. 22. As we pass along this, the internal temperatures of the stars decrease, so that the stellar atoms are less broken up than in main-sequence stars. In the red giants at the extreme end, even _M_-ring electrons may still remain. RUSSELL’S HYPOTHESIS Two entirely different explanations of this distribution of stars have been suggested. In 1925 Russell put forward a theory which centred primarily around the fact that the temperatures at the centres of the main-sequence stars are all very nearly equal. Let us simplify the situation for a moment by imagining it to be an ascertained fact that the temperatures at the centres of _all_ stars are precisely the same, say 32,000,000 degrees. If this were a sure fact, it would be natural to conjecture that the stars had some sort of controlling mechanism by which they continually adjusted their central temperatures to this exact figure, so that if ever the temperature fell below 32,000,000 degrees the mechanism would come into play and raise the temperature to precisely this amount, while if it increased to above this figure, the mechanism would come into play and depress it. Such controlling mechanisms are of course common in engineering practice; there are for instance the safety-valves which keep the pressure in a boiler always uniform, the Watts-governor which keeps an engine going always at the same rate of speed, and the thermostat which keeps the temperature of a room constant. A mechanism is already known for raising the temperature at a star’s centre. If a star is not generating any energy at all in its interior, either by the annihilation of matter or otherwise, its emission of radiation causes it to shrink, and this, as we have seen (p. 285), causes its temperature to rise. Thus it is easy to keep a star’s central temperature up to 32,000,000 degrees by arranging that no energy shall be generated so long as the temperature at the centre is below 32,000,000 degrees, and this is the main hypothesis on which Russell’s theory is based. He supposes that no energy at all is generated by matter at temperatures below 32,000,000 degrees, but that, as soon as this temperature is reached, matter begins to annihilate itself in sufficient quantity to provide for the radiation of a star.