Graham's _tube atmolyser_ consisted of a long tobacco-pipe stem placed inside a rather shorter and considerably wider tube of glass; the pipe stem was fixed by passing through two corks, one at each end of the glass tube; through one of these corks there also passed a short piece of glass tubing. When the instrument was employed, the piece of short glass tubing was connected with an air-pump, and one end of the pipe stem with the gaseous mixture--say ordinary air. The air-pump being set in motion, the gaseous mixture was allowed to flow slowly through the pipe stem; the lighter ingredient of the mixture passed outwards through the pipe stem into the wide glass tube more rapidly than the heavier ingredient, and was swept away to the air-pump; the heavier ingredient could be collected, mixed with only a small quantity of the lighter, at the other end of the pipe stem. As Graham most graphically expressed it, "The stream of gas diminishes as it proceeds, like a river flowing over a pervious bed."
Graham then contrived a very simple experiment whereby he was able to measure the rate of motion of the molecules of carbonic acid. He introduced a little carbonic acid into the lower part of a tall cylindrical jar, and at the close of certain fixed periods of time he determined the amount of carbonic acid which had diffused upwards through the air into the uppermost layer of the jar. Knowing the height of the jar, he now knew the distance through which a small portion of carbonic acid passed in a stated time, and regarding this small portion as consisting of a great many molecules, all moving at about equal rates, he had determined the average velocity of the molecules of carbonic acid. A similar experiment was performed with hydrogen. The general results were that the molecules of carbonic acid move about in still air with a velocity equal to seventy-three millimetres per minute, and that under the same conditions the molecules of hydrogen move with a velocity equal to about one-third of a metre per minute.[12]
The Bakerian Lecture for 1849, read by Graham before the Royal Society, was entitled "On the Diffusion of Liquids." In this paper he describes a very large number of experiments made with a view to determine the rate at which a salt in aqueous solution diffuses, or passes upwards into a layer of pure water above it, the salt solution and the water not being separated by any intervening medium. Graham's method of procedure consisted in completely filling a small bottle with a salt solution of known strength, placing this bottle in a larger graduated vessel, and carefully filling the latter with water. Measured portions of the water in the larger vessel were withdrawn at stated intervals, and the quantity of salt in each portion was determined. Graham found that under these conditions salts diffused with very varying velocities. Groups of salts showed equal rates of diffusion. There appeared to be no definite connection between the molecular weights of the salts and their diffusion-rates; but as Graham constantly regarded diffusion, whether of gases or liquids, as essentially due to the movements of minute particles, he thought that the particles which moved about as wholes during diffusion probably consisted of groups of what might be called chemical molecules--in other words, Graham recognized various orders of small particles. As the atom was supposed to have a simpler structure than the molecule (if indeed it had a structure at all), so there probably existed groups of molecules which, under certain conditions, behaved as individual particles with definite properties.