Gaede's molecular pump consists of a rotating cylinder with several circumferential grooves of the type just considered, into each of which projects a tongue connected with the housing. The openings on either side of the tongues are connected from groove to groove so that the high pressure side of the one is joined to the low pressure side of the next, or, in other words, the separate pumps which the grooves virtually constitute are con nected in series. The high pressure side of the outermost grooves go to the fore vacuum ; the low pressure side of the central groove is the point of lowest pressure, and goes to the vessel to be evacuated. Special devices prevent the lubricating oil from penetrating into the grooves. A modified form of molecular pump, due to Holweck, is illustrated in fig. 7. In this pump the grooves are cut in the casing, the rotating part being a simple cylinder. There are two grooves of spiral form, leading from the central opening to the outer ends of the pump, where they com municate with the space leading by the tube to the fore-vacuum. The grooves are deeper at the middle, where the pressure is low est, than at the ends. The speed of the pump is high, being as great as 7.000 to 8,000 cc. per second at a pressure of .00i mm. for one size in which it is made, and the pressure produced is less than mm. of mercury.
A great advantage of the molecular type of pump is that it deals with condensible vapours, in particular with water vapour, as readily as with gases : a disadvantage is the expense necessitated by the mechanical difficulties of construction. The cheapness and simplicity of operation of the mercury vapour pump now to be described has rendered the use of the molecular pump a restricted one.
The action of the mercury vapour pumps depends upon the diffusion of the gas into a stream of vapour, and the condensation of the vapour to prevent it passing into the receiver: according as more stress is laid upon the diffusion or the condensation these pumps have been called diffusion pumps or condensation pumps.
A better name might be vapour-stream pumps, which is here suggested. Imagine a tube through which a stream of easily con densible vapour passes, for definiteness say mercury vapour, and let a side tube from a vessel containing gas lead into this tube at right angles. A certain amount of gas will diffuse into the mercury vapour stream, and a certain amount of mercury vapour will diffuse into the side tubes. If, however, the pressure of the gases is comparatively high, and the mean free path small compared to the size of the opening into the vapour tube, impacts will be very frequent, and the diffusion very small. If the mean free path is large, the interchange of molecules takes place freely, and gas and vapour each behave as if the other was not there : the flow depends upon the partial pressure of each molecular species. The speed
at which the gas molecules dif fuse into the vapour stream can be calculated from the kinetic theory of gases, and it can be shown that it attains a maximum when the mean free path of the mercury vapour is about the di ameter of the opening through which the diffusion takes place. Gaede's original pump was based upon such considerations. Com munication between gas and va pour stream took place through a slit of carefully calculated size. The most favourable mean free path of the mercury vapour was obtained by adjusting the tem perature. The mercury vapour passing into the gas space through the slit was removed by Shortly of ter Gaede's first ac count of his pump Langmuir de vised a form of vapour-stream pump in which he insisted partic ularly upon the condensation, holding that his pump was essen tially different in principle from Gaede's. In the form illustrated in fig. 8, a stream of vapour from heated mercury M passes ver tically upwards through a tube B and is directed against a hood F, hung on thin rods R. The mercury vapour then rushes down wards and strikes the water-cooled walls of the cylindrical space, where it condenses. There is thus only a very small movement, by diffusion against the prevailing velocity of stream, of the mercury vapour upwards into the circular space A, through which the gas, entering by the wide tube T, diffuses into the mercury stream. This method of condensation has obvious advantages, and has been widely adopted, but the pump is none the less a dif fusion pump. A similar construction was afterwards adopted by Gaede, whose present one-stage diffusion pump is shown in fig. 9. The mode of operation is clear from what has just been said.
Two stage and three stage mercury vapour pumps, in which the stages are arranged in series, so that the high pressure side of one is the low pressure side of another, are made for rapid work. While, owing to the fact that at very low pressure it loses its jet form, the stream of mercury vapour in a diffusion pump exerts no such action as the steam jet does in an ordinary injector, at higher pressures the Bernoulli effect comes into operation, and is, in fact, utilised in the preliminary stage or stages of a multiple stage pump, where, owing to the higher pressure, the diffusion principle is less effective. An interesting two-stage pump devised by Dunoyer, which works with a fore-vacuum pressure no lower than 25 mm. of mercury or so, is illustrated in fig. Ia. The first stage operates on the principle of the steam injector, the rush of mercury vapour through the jet A taking the place of the rush of steam in the injector, and creating a low pressure in B. Mercury vapour also passes through the holes T into the annular space C, and creates a high vacuum in D by diffusion and condensation in the same way as it does in the Langmuir pump.