the first infinitely thin layer of the liquid; but this layer, while it is becoming hot, undergoes a certain dilatation, becomes lighter than the rest of the fluid mass, and ascends immediately to the upper part of the vessel, whence it can have no longer any influence on the pile. It is replaced by a second layer, which undergoes a similar process, and this again by others; so that by these partial renovations of the liquid screen, the hinder part of the glass applied to the aperture of the tube is not in contact with heated molecules, and retains the same temperature for a long time.
It was extremely difficult to make flat glass vessels with very regular surfaces of the same thickness throughout, and with the opposite sides exactly parallel. Metallic frames and glasses joined with gum could not be employed because of the corrosive action of the several liquids. After many a fruitless effort to surmount this difficulty, I thought at last that the process by which the index of refraction of liquids is measured in optics might be available in this case also. With this view I had quadrangular pieces of two centimetres in breadth and nine centimetres in length cut out of several pieces of the same mirror unsilvered and sufficiently thick. I laid close to the two faces of each of the pieces from which the excision had been made two flakes made out of another and a much thinner glass. It is known that the mere adhesion of two plates of polished glass is sufficient to prevent the passage of liquids. However, in order to be more secure, I introduced each recipient between two metallic frames, which held the thin glasses in their places by means of four screws placed at the angles. The liquid was poured into these vessels at a small aperture made at the top, and furnished with a glass stopper. In such a system there could be no doubt of the parallelism of the faces and the equal thickness of the layers.
The results furnished by the several bodies, both solid and liquid, I have disposed in several tables, each of them exhibiting at the top the common thickness of the screens employed and, beside the substance, the indications of the thermomultiplier and the quantity of rays transmitted as compared with the whole radiation. This distribution, while it allows the use of plates of different thicknesses, has the additional advantage of presenting distinct groups of each class of bodies. The free radiation in each case was 30°. In order to link the results of these tables together, I have commenced the second and the third with the numbers given by a flake of glass placed in the same circumstances as the plates which constitute each group: thus the glass set down in the table of liquids was contained between the two thin plates of the recipients, and made of the thick looking-glass employed in their construction. It was therefore exactly of the same thickness as the liquid layers, and, like them, came into contact with the thin plates which formed the faces of the recipients. But as those faces themselves intercepted a portion of
