From air, which contains a _mixture_ of oxygen and nitrogen, the nitrogen alone cannot be removed, because it has no inclination to combine directly or readily with any substance; and although it does combine with certain substances (boron, titanium), these substances combine simultaneously with the oxygen of the atmosphere.[6] However, oxygen may be separated from air by causing it to combine with substances which may be easily decomposed by the action of heat, and, in so doing, give up the oxygen absorbed--that is, by making use of reversible reactions. Thus, for instance, the oxygen of the atmosphere may be made to oxidise sulphurous anhydride, SO_{2} (by passing directly over ignited spongy platinum), and to form sulphuric anhydride, or sulphur trioxide, SO_{3}; and this substance (which is a solid and volatile, and therefore easily separated from the nitrogen and sulphurous anhydride), on further heating, gives oxygen and sulphurous anhydride. Caustic soda or lime extracts (absorbs) the sulphurous anhydride from this mixture, whilst the oxygen is not absorbed, and thus it is isolated from the air. On a large scale in works, as we shall afterwards see, sulphurous anhydride is transformed into hydrate of sulphuric trioxide, or sulphuric acid, H_{2}SO_{4}; if this is allowed to drop on to red-hot flagstones, water, sulphurous anhydride, and oxygen are obtained. The oxygen is easily isolated from this mixture by passing the gases over lime. The extraction of oxygen from oxide of mercury (Priestley, Lavoisier), which is obtained from mercury and the oxygen of the atmosphere, is also a reversible reaction by which oxygen may be obtained from the atmosphere. So also, by passing dry air through a red-hot tube containing barium oxide, it is made to combine with the oxygen of the air. In this reaction the so-called barium peroxide, BaO_{2}, is formed from the barium oxide, BaO, and at a higher temperature the former evolves the absorbed oxygen, and leaves the barium oxide originally taken.[7] [6] A difference in the physical properties of both gases cannot be here taken advantage of, because they are very similar in this respect. Thus the density of oxygen is 16 times and of nitrogen 14 times greater than the density of hydrogen, and therefore porous vessels cannot be here employed--the difference between the times of their passage through a porous surface would be too insignificant. [Illustration: FIG. 27.--Graham's apparatus for the decomposition of air by pumping it through india-rubber.]