* 112. For suppose a firm partition, full of little doors (themselves without mass) to be placed so as to divide the vessel into two, and set a demon at each door, with instructions to open it for an instant whenever he sees he can thereby let a quick-moving particle escape from the first compartment to the second, or a slow-moving particle from the second into the first. Then, _because the tendency is not to a uniform distribution of velocity_ among the particles, but to a distribution which involves quicker and slower in certain proportions, we may imagine this process to be carried on long enough to make a considerable difference in the average velocities of the particles in the two compartments, though the numbers of particles in each compartment may remain almost unchanged. The consequence will of course be a greater pressure per square inch on the walls of the second compartment than of the first; and thus, if the partition wall were moveable, a certain amount of work might be obtained by allowing it to move. Thus a group of particles originally incapable, without external assistance, of doing work, may be rendered capable of doing work by mere _guidance_ applied by finite intelligence. * 113. Now let us refer for a moment to the molecular theory of gases, and we see that what the demons (without any expenditure of work, each being, so far as he is required, virtually a combination of two intelligent perfect engines, one working direct, the other reversed) have guided the gas to do, is to transfer heat from a colder to a hotter portion of the gas. The only reason why this does not occur without the assistance of demons (at least to an extent, or for a length of time, sufficient to produce a sensible effect) lies in the enormous number of particles per cubic inch in even the most rarefied gas. Hence, _solely because of the excessive numbers and minuteness of the particles of matter_, the one chance of escape from Carnot’s proposition is denied us, and therefore we must allow that, so far as the physical universe is concerned, a reversible heat-engine is the best possible. But if a reversible heat-engine be the best possible, then the principle which we have italicised in Art. 107 must hold good, and from this it follows that only a portion of the heat passing through a perfect engine can be transformed into useful work unless the condenser of the engine be at the absolute zero of temperature—a condition which can never be attained. 114. It thus appears that at each transformation of heat-energy into work a large portion is degraded, while only a small portion is transformed into work. So that while it is very easy to change all of our mechanical or useful energy into heat, it is only possible to transform a portion of this heat-energy back again into work. After each change too the heat becomes more and more dissipated or degraded, that is, less and less available for any future transformation.