The former formulæ of silica, SiO (Si = 14) and SiO_{3} (Si = 42), were first changed into the present one, SiO_{2} (Si = 28), on the basis of the following arguments:--An excess of silica occurs in nature, and in siliceous rocks free silica is generally found side by side with the silicates, and one is therefore led to the conclusion that it has formed acid salts. It would therefore be incorrect to consider the trisilicates as normal salts of silica, for they contain the largest proportion of silica; it is much better to admit another formula with a smaller proportion of oxygen for silica, and it then appears that the majority of minerals are normal or slightly basic salts, whilst some of the minerals predominating in nature contain an excess of silica--that is, belong to the order of acid salts. At the present time, when there is a general method (Chapter VII.) for the determination of atomic weights, the volumes of the volatile compounds of silica show that its atomic weight Si = 28, and therefore silica is SiO_{2}. Thus, for example, the vapour density of silicon chloride with respect to air is, as Dumas showed (1862), 5·94, and hence with respect to hydrogen it is 85·5, and consequently its molecular weight will be 171 (instead of 170 as indicated by theory). This weight contains 28 parts of silicon and 142 parts of chlorine, and as an atom of the latter is equal to 35·5, the molecule of silicon chloride contains SiCl_{4}. As two atoms of chlorine are equivalent to one of oxygen, the composition of silica will be SiO_{2}--that is, the same as stannic oxide, SnO_{2}, or titanic oxide, TiO_{2}, and the like, and also as carbonic and sulphurous anhydrides, CO_{2} and SO_{2}. But silica bears but little physical resemblance to the latter compounds, whilst stannic and titanic oxides resemble silica both physically and chemically. They are non-volatile, crystalline insoluble, are colloids, also form feeble acids like silica, &c., and they might therefore be expected to form analogous compounds, and be isomorphous with silica, as Marignac (1859) found actually to be the case. He obtained stannofluorides, for example an easily soluble strontium salt, SrSnF_{6},2H_{2}O, corresponding with the already long known silicofluorides, such as SrSiF_{6},2H_{2}O. These two salts are almost identical in crystalline form (monoclinic; angle of the prism, 83° for the former and 84° for the latter; inclination of the axes, 103° 46´ for the latter and 103° 30´ for the former), that is, they are isomorphous. We may here add that the specific volume of silica in a solid form is 22·6, and of stannic oxide 21·5.